Butrint 3: Excavations at the Triconch Palace (Butrint Archaeological Monographs) 1842179802, 9781842179802

116 40 389MB

English Pages 374 [401] Year 2011

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Butrint 3: Excavations at the Triconch Palace (Butrint Archaeological Monographs)
 1842179802, 9781842179802

Table of contents :
Cover
Preface
Butrint: Gërmimet në Pallatin e Trikonkës
1. Introduction
2. The domus and the Triconch Palace
3. The 5th- to mid 7th-century occupation of the triconch area
4. The medieval occupation of the triconch area
5. The Merchant’s House and the city wall (5th to 7th centuries)
6. The medieval occupation of the Merchant’s House
7. The mosaic pavements and painted walls of the domus
8. The domus and Triconch Palace as aristocratic residences
9. Urban change and the Triconch Palace site in the 5th to 7th centuries
10. From Roman insula to medieval quarter?
11. Conclusion
Bibliography
Index
Plates

Citation preview

Butrint Archaeological Monographs

BUTRINT 3

EXCAVATIONS AT THE TRICONCH PALACE

William Bowden and Richard Hodges

with contributions by Ylli Cerova, Andrew Crowson, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako, Matthew Logue, John Mitchell, Nevila Molla, Jonathan Price, Alessandro Sebastiani and Emanuele Vaccaro

OXBOW BOOKS FOR THE BUTRINT FOUNDATION Oxbow Books

Oxford and Oakville

Published by Oxbow Books, Oxford, UK © Oxbow Books, William Bowden, Richard Hodges and the individual authors 2011 ISBN 978-1-84217-980-2

This book is available direct from Oxbow Books, Oxford, UK (Phone: 01865-241249; Fax: 01865-794449) and The David Brown Book Company PO Box 511, Oakville, CT 06779, USA (Phone: 860-945-9329; Fax: 860-945-9468) or from our website www.oxbowbooks.com A CIP record is available for this book from the British Library Library of Congress Cataloging-in-Publication Data Bowden, William. Butrint 3 : excavations at the Triconch Palace / William Bowden and Richard Hodges ; with contributions by Ylli Cerova ... [et al.]. p. cm. -- (Butrint archaeological monographs ; v. 3) Includes bibliographical references and index. ISBN 978-1-84217-980-2 (hardback) 1. Triconch Palace Site (Butrint, Albania) 2. Butrint (Albania)--Antiquities, Roman. 3. Butrint (Albania)--Antiquities, Byzantine. 4. Excavations (Archaeology)--Albania--Butrint. I. Hodges, Richard. II. Title. III. Title: Butrint three. IV. Series. DR998.B88B68 2011 939’.865--dc23 2011028471

Front cover: The Triconch Palace (c. AD 420) (Studio Inklink). Back cover: Comic masks from the mosaics of the domus (Martin Smith). Printed in Great Britain by Short Run Press, Exeter

For Pippa Pearce

Butrint Archaeological Monograph Series: 1. 2. 3. 4.

Byzantine Butrint: Excavations and Surveys 1994–1999 Richard Hodges, William Bowden and Kosta Lako Roman Butrint: An Assessment Inge Lyse Hansen and Richard Hodges Butrint 3: Excavations at the Triconch Palace William Bowden and Richard Hodges Butrint 4: The Archaeology and Histories of an Ionian Town Inge Lyse Hansen, Richard Hodges and Sarah Leppard

Contents

Preface Richard Hodges

vii

Butrint: Gërmimet në Pallatin e Trikonkës Nevila Molla

xi

1. Introduction William Bowden and Richard Hodges

1

THE EXCAVATIONS 2. The domus and the Triconch Palace William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako 3. The 5th- to mid 7th-century occupation of the triconch area William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

11 56

4. The medieval occupation of the triconch area William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

118

5. The Merchant’s House and the city wall (5th to 7th centuries) William Bowden, Ylli Cerova, Andrew Crowson, and Emanuele Vaccaro

152

6. The medieval occupation of the Merchant’s House William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

203

INTERPRETATION AND DISCUSSION 7. The mosaic pavements and painted walls of the domus John Mitchell

231

8. The domus and Triconch Palace as aristocratic residences William Bowden

277

9. Urban change and the Triconch Palace site in the 5th to 7th centuries William Bowden

303

10. From Roman insula to medieval quarter? Richard Hodges

319

11. Conclusion Richard Hodges

327

Bibliography Index Plates

339 346 351

Preface Richard Hodges

‘Excavations have a clear voice and speak, to those who have the ear to listen, with a tone so bewitching and persuasive that it sustains the excavator and helps him overcome the moments of despondency that for one reason or another are never absent. Thankfully, there are also moments of intense and pure joy, and these compensate for the pain of past sufferings, and instil courage to endure those of the future.’ (Ugolini 1937, 47) The Triconch Palace excavations at Butrint seemed a wonderfully simple endeavour when in 1993 we first agreed with our Albanian colleagues to work here. Situated in a level glade beside the Vivari Channel, a little way off the well-worn tourist path through Butrint, it has a certain charm as a secret spot blessed by early morning sunshine. Here mussel-fishermen, squatting on their haunches, cooked up their catches before we began our excavations, and villagers from neighbouring Vrina cast their simple lines from the reed bank. Only with time did we comprehend the difficulties of excavating in thick woodland that was regularly waterlogged by the high water-table, and suffered from a stifling humidity because the trees screened off the afternoon breezes from the Ionian Sea. Butrint, ancient Buthrotum, situated three kilometres from the Straits of Corfu on the Vivari Channel at the south end of Lake Butrint, is a typical illustration of an ancient city that declined in late antiquity before experiencing a Mid Byzantine revival that endured until the later Middle Ages. With its fertile coastal niche extending ten kilometres into a mountainous valley and with access to legendary amounts of fish in Lake Butrint, it appears to fit the stereotype of a Mediterranean coastal location. Previous excavations by the Italian Archaeological Mission led by Luigi Maria Ugolini and by members of the Albanian Institute of Archaeology had concentrated upon understanding key monuments, mostly associated with the civic centre of the ancient city. There had been no attempt, therefore, to chart the history of an insula as it might have been transformed, through time, into some kind of medieval quarter. The challenge for the Butrint Foundation team, as it

appeared to us in 1993, was to confront the question of the changing nature of a Mediterranean port not through the study of isolated monuments situated within an established historical narrative, but through establishing and explaining generational changes in the urban fabric of the city. At that time, few such areas within ancient cities had been treated in this way to stratigraphic excavations on a large scale, seeking evidence of the changing rhythms of occupation. The sheer volume of information appeared to make such an approach virtually impossible. Nevertheless, mindful of the significant results from stratigraphic excavations in Mediterranean ports such as Carthage, Tunisia (Hurst and Roskams 1984), Cherchel, Algeria (Potter 1995) and the Crypta Balbi in Rome (Manacorda 2001), as well as from on-going excavations in the Agora at Athens and at Corinth, the Triconch Palace appeared to be a promising and intriguing challenge that might shed much new light on urban processes in the Adriatic region. Our point of departure for the excavation methodology was Martin Carver’s Arguments in Stone (1993) which readily assumed (as we did) that a north European methodology (rooted in north European historiographic traditions) might be easily translated to a Mediterranean context. As in many similar projects, such an assumption was soon to be dispelled. First, our Albanian collaborators, as we have recorded elsewhere, had their own historical paradigm rooted in sustaining a national myth (Bowden and Hodges 2004) which took no account of contemporary historiography. Second, although our approach involved sampling on a major scale, identifying stratigraphic deposits as predicated by Carver’s method, this was complicated by tree cover, by the changing and high water table, and most of all by the realization that only open-area excavation with an immense commitment to labour and post-excavation analysis offers a suitable instrument for interpreting interperiod and intra-settlement differentiations. Carver’s method did not work in the context of late antique and medieval Butrint. Problems of residuality, the repeated remodelling and re-use of structures throughout the Roman

viii

Preface

period and the large-scale secondary movement of deposits in antiquity (during construction work and terracing) meant that the results of trenching and small-scale interventions were inconclusive at best and totally misleading at worst. As a result, our initial investigations at the Triconch Palace from 1994–99, summarized in Byzantine Butrint (2004), although they dramatically increased our understanding of Butrint, provided only an imprecise overview of the town and its changing topography. Support by the Packard Humanities Institute, first for the large-scale excavations between 2000–2003, then for the concomitant post-excavation process, enabled us to achieve our original aims of charting the urban history of Butrint over almost 1500 years. As will be clear from this volume, due to the complexities of the excavation itself, major episodes and important features of Butrint’s urban history only became clear as we wrestled with the voluminous documentation and huge body of finds. Added to this, as will be apparent, our new interpretation has greatly benefited from the experience gained in making other stratigraphic excavations at Butrint (Bowden and Hodges forthcoming). We must conclude that, although long-term large-scale excavation is by far the most costly and unfashionable of archaeological techniques, the results of the Butrint Project suggests that it remains an essential element of the study of abandoned Roman towns. For their patience and tenacious support, first and foremost, we are particularly grateful to Lord Rothschild and Lord Sainsbury of Preston Candover, who established the Butrint Foundation. We also owe a special debt to Dr. David Packard, President of the Packard Humanities Institute, for his singular support in a partnership with the Butrint Foundation. Our thanks, too, to other supporters of these excavations and the associated archival research: in particular, Dame Drue Heinz of the Drue Heinz Trust, the David and Lucille Packard Foundation, and the World Monuments Fund. We also wish to acknowledge a grant from the British Academy for the 1994 season. The conservation and presentation of the excavated site in 2005 was made possible by support from the Howard and Nancy Marks Fund. We are especially indebted to Sir Patrick Fairweather, formerly British ambassador to Albania (1992–96), and from 1997–2004, the Director of the Butrint Foundation. Thanks, too, to Daniel Renton, Director of the Foundation between 2004–6 who oversaw the conservation and presentation of the Triconch Palace; to Rupert Smith, Director of the Foundation during 2007–8; and to the current Director, Brian Ayers. Our thanks to Iris Pojani and Diana Ndrenika, successive Directors of the International Centre for Albanian Archaeology in Tirana. Thanks also to our Albanian colleagues: Lorenc Bejko, Neritan Ceka, Ylli Cerova, Dhimitër Çondi, Reshad Gega, Ilir Gjipali, Shpresa Gjongecaj, Gjerak Karaiskaj, Kosta Lako, Telemark Llakhana, Etleva Nallbani, Guri Pani, Luan Përzhita and Artan Shkrelli. Finally, we owe a special debt to the

Director of the Institute of Archaeology during the course of these excavations, Professor Muzafer Korkuti. The project began with the support of the British School at Rome, where Maria Pia Malvezzi and Tommaso Astolfi played an instrumental role in the early years. Since 1996 it has formed part of the research programme of the Institute of World Archaeology in the University of East Anglia, Norwich. Richard Hodges and Gjergj Saraçi served as co-directors of the Butrint Foundation project from 1994–96; and Richard Hodges and Kosta Lako with Ilir Gjipali were codirectors from 1998–99. Richard Hodges and Ilir Gjipali were co-directors from 2000–2009. The field seasons were: September 1994; April 1995; September-October 1995; April 1996; September 1996; and September 1998. The September-October 1999 season was devoted to studying the finds. There then followed four major seasons of excavations in April-May 2000; June-July 2001; June-July 2002; and June-July 2003. Sally Martin acted as project manager of the 1994–2003 seasons. Kosta Lako and Oliver Gilkes supervised the excavations at the Triconch Palace except for the final seasons in 2002–3 when Ylli Cerova, Andrew Crowson and Karen Francis assisted by William Bowden oversaw the completion of the excavations. The comprehensive study of the walls of Butrint was undertaken by Richard Andrews between 1994–99, and has been updated by a team making a photogrammetric survey from the University of Siena, directed by Giovanna Bianchi and led by Nevila Molla during 2007. Site photography was by James Barclay-Brown, William Bowden, Andrew Crowson, Steve Diehl, Oliver Gilkes, Martin Smith and Massimo Zanfini. Site conservation and presentation of the excavated remains in 2005 was masterminded by Daniel Renton assisted by Jerry O’Dwyer and Albana Hakani, with on-site conservation of the mosaics by Jacques Neguer and Elda Omari. Additional mosaic conservation was undertaken by Agron Islami in 2010. The illustrative panels at the site were made by Studio Inklink of Florence. A small army of excavators took part in this project. Special thanks should go to Dave Boschi, Jayne Bown, Emily Glass, Simon Greenslade, Charlotte Hodges, Valbona Hysa, Rovena Kurti, Sarah Leppard, Nevila Molla, John Percival, Erjona Qilla, Emmy Rothschild and Louise Schofield. The processing of the finds, which will be published separately, was managed by Dave Boschi and Inge Lyse Hansen, assisted by Ilir Papa, Liri Shametaj and Sabina Veseli. Finds photography was undertaken by James Barclay-Brown, Michael Grayley and Martin Smith. Paul Reynolds has studied the Roman pottery and Joanita Vroom has studied the medieval ceramics. The late Sarah Jennings was responsible for glass, and John Mitchell for the small finds. The Roman coins have been studied by Richard Abdy and Sam Moorhead, and the Byzantine and medieval coins by Pagona Papadopoulou. Adrienne Powell

Preface managed all the faunal remains. John Giorgi and Alexandra Livarda reviewed the palaeobotanical evidence, Petrus De Rijk studied the industrial waste, and David Bescoby provided assistance with the geomorphology and issues relating to seismic episodes. Management of the team’s accommodation and daily routine was overseen by Gjoni Marko from 2001–5. Mention should also be made of the tireless Muzafer Lazë, our driver and general factotum. In writing this report we should like to single out the support and help of Andrew Crowson, Amy Culwick, Karen Francis, Inge Lyse Hansen, Oliver Gilkes, Matthew Logue, John Mitchell, Nevila Molla, Jonny Price, Paul Reynolds, Alessandro Sebastiani and Joanita Vroom, as well as the helpful advice of Paul Arthur, Kim Bowes, Sandro De Maria, the late Riccardo Francovich, Shelley Hales, John Haldon, Nicholas Purcell, Jean Pierre Sodini, Giuliano Volpe and Konstantinos Zachos. The completion of the volume owes much to the efforts of Simon Greenslade and

ix

Sarah Leppard who prepared most of the illustrations for publication. The index was prepared by Andrew Crowson. We are very grateful to Tara Evans at Oxbow for her hard work and patience. Few excavations at Butrint, as it turned out, have proved as challenging to manage and as puzzling to interpret. For these reasons we are deeply conscious of all those who supported us over the 18 years since we first visited the Triconch Palace. In particular, we would like to single out the contribution made by Pippa Pearce, who has overseen our finds conservation since the first season in 1994. Pippa’s contribution to the Butrint Project, however, has been much more than conservation. Her humour and resilience in the face of the challenges of the project and its working environment have been a fundamental part of the excavations and an inspiration to the team. It is therefore with thanks and great affection that we dedicate this volume to her.

Butrint: Gërmimet në Pallatin e Trikonkës Nevila Molla

1. Hyrje William Bowden dhe Richard Hodges Programi i Fondacionit Butrinti është përqendruar mbi studimin e Butrintit, qyteti i një porti antik dhe mesjetar që ndodhet në Shqipërinë e jugut, 3 km nga ngushtica e Korfuzit, i cili deri pak kohë më parë ishte i panjohur në raport me kontekstin mesdhetar. Pozicioni i tij në kryqëzimin e rrugëve të Mesdheut i ofronte qytetit-port, kontroll të plotë të rrugëve detare që përshkonin Mesdhuen në drejtim të perëndimit, në veri drejt detit Ardiatik dhe në jug drejt detit Jon. Ashtu si dhe qyteti i lashtë, Dyrrhachium (Durrësi modern) në veri të tij, Butrinti kontrollonte edhe rrugët tokësore në drejtim të zonave të brendshme malore të Ballkanit. Realizimi i një projekti me synim sërvejimin dhe gërmimin jo vetëm të pikave kyçe brenda qytetit por edhe të mikro-zonës së tij lagunore që nga viti 1994, na mundësoi krijimin e një paradigme të re mbi karakterin e qyteteve-port në kontekstin e detit Adriatik, që nga periudha helenistike e deri në atë otomane. Gërmimet na kanë mundësuar t’u përgjigjemi pyetjeve që kanë lindur gjatë 15 vitesh pune kërkimore në Butrint, por në të njëjtën kohë ato duhen kuptuar brenda një konteksti gjithnjë e në ndryshim që karakterizon debatin më të gjerë akademik lidhur me urbanizmin në periudhën romake dhe atë mesjetare. Ky volum paraqet gërmimin e një sektori të gjerë urban në Butrint pozicionuar buzë kanalit të Vivarit, të zhvilluar në dy faza (1994–1999 dhe 2000–2003) si rezultat i të cilit, u zbuluan të gjitha ambientet e të ashtuquajturit Pallati i Trikonkës dhe të strukturave ngjitur me të, të emërtuara Shtëpia e Tregëtarit. Synimi i ketij gërmimi ishte të kuptonim çfarë zhvillimi përjetoi një rezidencë e përmasave të mëdha gjatë periudhës romake dhe çfarë ndodhi me të në periudhën post-romake.

Gërmimet 2. Domus-i dhe Pallati i Trikonkës W. Bowden, K. Francis, O. Gilkes dhe K. Lako Ky kapitull paraqet në detaj gërmimin e një ndërtese të periudhës romake dhe të periudhës së vonë romake të ndodhur në zonën qendrore dhe atë lindore të sitit të Pallatit të Trikonkës (zonat 2 dhe 3). Gërmimi regjistroi një seri fazash të artikulimit struktural që përjetoi zhvillimi i një rezidence elitare, duke filluar me shfaqjen e strukturave më të hershme në këtë zonë të sitit e duke vazhduar me transformimin e tyre në rezidencë elitare të përfaqësuar nga vetë pallati madhështor i trikonkës, braktisja e të cilit si i tillë ndodh pak kohë pas ndërtimit të tij, në fillim të shek. V. Duket se në fillim të shek. II e ndoshta edhe më herët, një strukturë e konsiderueshme ndërtohet në pjesën qendrore të zonës së trikonkës, si një pronë e vetme me një shtrirje hapësinore të ngjashme me atë të domus-it me peristil që ndërtohet në fund të shek. IV- fillim të shek. V. Edhe pse vetëm pak ambiente të pjesës perëndimore të saj datojnë në shek. II është e mundur që, që në këtë fazë kompleksi të ketë qenë ndërtuar rreth një kopshti apo oborri qendror, të cilit më vonë i mbivendoset peristili i domus-it të sapo përmendur. Evidenca bindëse nga gërmimi indikojnë ekzistencën e dy rrugëve të qytetit të kësaj periudhe përgjatë anës lindore dhe perëndimore të kompleksit, të njëjtat anë në të cilat gjenden edhe dy hyrjet kryesore për në brendësi të ndërtesës. Një nga kanalet e sistemit të drenazhimit kalon gjithashtu përgjatë anës perëndimore të saj. Gjatë shek. III dhe të IV ky kompleks merr formën e një ndërtese qytetare tipike për periudhën në fjalë, hyrja kryesore dhe publike e të cilit ndodhej në perëndim ku një derë e vogël hapej në një vestibul, dyshemeja e të cilës ishte dekoruar me mozaikun më cilësor në të gjithë kompleksin, ndërsa muret, me kolona iluzionare (të pikturuara në afresk). Pas saj, vizitori hynte në një galeri (korridor të gjatë) gjithashtu të dekoruar me mozaik dhe kolonadë në

xii

Nevila Molla

afresk që në fund të çonte në dhomën e pritjes, fillimish me plan katror e që më pas pajiset me apsidë. Edhe dyshemetë e këtij ambienti dhe të apsidës janë dekoruar me mozaikë dhe një shatërvan me plan oktagonal ndodhej në mes të dhomës së pritjes. Dekorimi luksoz dhe karakteristikat e arkitekturës publike që mbart kjo pjesë e ndërtesës indikojnë qartë se ajo është zona publike e banesës dhe flasin për një rol të caktuar që duhet të ketë pasur pronari i saj në jetën qytetare në Butrint. Kjo zonë ndahej qartë nga pjesa private e banesës e pozicionuar në veri, lidhja me të cilën realizohej nëpërmjet një dere në murin verior të galerisë. Një seri dhomash në jug krijonte një zonë të ndërmjetme mes zonës publike dhe zonës private të pozicionuar në lindje të banesës, ndërkohë që pak dihet për zonën perëndimore të saj. Edhe pse evidenca datuese është e kufizuar, duket se në gjysmën e dytë të shek. IV e ndoshta rreth viteve 400, domus-i përjetoi një fazë të rëndësishme rindërtimi. Fokusi i banesës zhvendoset tashmë nga galeria dhe dhoma apsidale e pritjes me pamjen mbi kanalin e Vivarit, në oborrin-peristil të ri të vendosur në qendër të kompleksit. Po kështu, hyrja kryesore për në banesë zhvendoset nga ana lindore në atë perëndimore ku një vestibul me kolona të çonte drejt peristilit, pragu i të cilit dekorohej me mozaik. Një mbishkrim me emrin dhe rangun e pronarit të banesës indikon qartë që kjo ishte hyrja kryesore në ndërtesë. Dyshemetë e shtëpisë së rindërtuar rreth vitit 400 janë të shtruara në një nivel prej rreth 0.60m më lartë se ato të mëparshmet, çka sugjeron një moment të përkohshëm të daljes së një pjese të saj jashtë përdorimit. Kjo situatë e re mund të shpjegohet me ngritjen e nivelit të ujit në kanalin e Vivarit që e detyron pronarin të zhvendosë hapësirat e banueshme në pjesën veriore të kompleksit. Gjithashtu, një fenomen sizmik mund të ketë qenë katalizatori i ndryshimeve që pësoi kompleksi. Gërmimet në forumin e qytetit kanë zbuluar të dhëna bindëse për ndodhinë e një lëvizjeje të ndjeshme tektonike në gjysmën e dytë të shek. IV. Faza finale e rindërtimit konsiston në vënien në jetë të një programi të rëndësishëm zgjerimi, si pasojë e të cilit kompleksi shtrihet drejt lindjes ku ndërtohet struktura tri-konkëshe. Gjithashtu, hyrjet e kompleksit zhvendosen përsëri duke u pozicionuar në jug dhe në veri me qëllim që vizitori që hynte në banesë nga portiku të mund të shkonte direkt në triclinium. Megjithëse kompleksi i zgjeruar përbënte një investim të rëndësishëm në burime financiare dhe krahun e punës që është dashur të impenjohej, duket se ai u braktis përpara se projekti të përfundohej, siç indikojnë mungesa e shtrimit të mozaikëve në dyshemetë e portikut dhe e çdo shenje suvaje apo dekori në muret dhe dyshemenë e triclinium-it. Arsyeja për braktisjen e projektit nuk mund të dihet me qartësi. Sidoqoftë, nga çereku i dytë i shek. V, në mos më herët, në disa zona të Pallatit të Trikonkës fillon të shfaqet një mënyrë shumë e ndryshme të jetese.

3. Përdorimi i Zonës së Trikonkës gjatë shek. V – mesi i shek. VII W. Bowden, K. Francis, O. Gilkes dhe K. Lako Pas braktisjes së projektit rindërtues rreth vitit 420, zonat jugore dhe perëndimore të domus-it shndërrohen në ambiente për nevoja utilitare shtëpiake, e ndoshta disa prej dhomave shërbejnë edhe për zhvillim të një aktiviteti industrial në shkallë të vogël, siç tregon zbulimi i një vatre në fundin perëndimor të ish-galerisë dhe i dy furrave në dhomat e zonës perëndimore të kompleksit. Një numër gropash për trungje dhe shkopinj në dyshemenë e galerisë indikojnë se kjo hapësirë mund të ketë qenë ndarë në një seri ambientesh. Megjithate, gërmimi tregon se dhoma e pritjes me apside dhe portiku perëndimor vazhdojnë të ruajnë karakterin e tyre fillestar deri në fillim të shek. VI. Kjo na bën të jemi të kujdesshëm në interpretimin e shfaqjes së ketyre elementëve, duke pasur parasysh shembullin e vilës në S. Giovanni di Ruoti në Basilicata, ku përkrah ambienteve kryesorë të saj, qe rreth viteve 400 shfaqet edhe depozitimi i mbeturinave. Është e mundur që në këtë rast kemi të bëjmë jo me braktisjen e plotë të konceptit të rezidencës elitare nga banorët e Pallatit të Trikonkës, por thjesht me ri-përkufizimin e këtij koncepti. Gjatë gjysmës së dytë të shek. V zona e Pallatit të Trikonkës vazhdon të përdoret për një seri të ndryshme aktivitetesh shtëpiake, vëçanërisht ambientet menjëherë pas hyrjes me kolona, në perëndim, ku vazhdojnë të shfaqen furra dhe struktura rrethore për qëllime magazinimi e mure ndarës që ndoshta krijojnë dhoma fjetjeje, duke e kthyer këtë zonë të pallatit në një kompleks relativisht të madh shtëpiak. Zona e jugut, e ndoshta dhe ajo e veriut fillojnë të përdoren gjithnjë e më shumë për aktivitete të tilla si përgatitja e llaçit. Për fat të keq, zhvendoja e shtresave në triclinium nuk na lejon të sugjerojmë se çfare roli mund të ketë luajtur ai në situaten e re në të cilën ndodhet kompleksi. Sugjerimi se ky kompleks vazhdon të jetë një pronë e vetme e zotëruar nga një individ përforcohet nga fakti që muri rrethues i qytetit ndërtuar në çerekun e fundit të shek. V devijon në këtë pikë për të respektuar zonën e Pallatit të Trikonkës dhe të Shtëpisë së Tregëtarit ngjitur me të. Fillimi i shek. VI e gjen Pallatin e Trikonkës pothuajse të transformuar. Çatitë e shumë prej dhomave ishin shkatërruar dhe bashkë me to edhe sensi i unitetit dhe koherencës së vetë kompleksit, të cilët i kishin mbijetuar shek. V. Struktura të reja ndërtohen mbi depozita të trasha që mbulojnë ambientet në zonën jugore ku në veçanti vlen të përmendet prania e një shkallareje prej guri në fundin lindor të galerisë së dikurshme të domus-it, që indikon ekzistencën e një kati të dytë, i cili ndoshta zhvillohej mbi hyrjen-vestibul të dikurshme. Ndërkohë, zona perëndimore e banuar gjatë shek. V braktiset dhe mbushet me hedhurina dhe tjegulla të thyera që formojnë një shtresë që shtrihet edhe në portikun perëndimor. Oborri qendror nuk mbulohet nga shtresa dheu çka flet për vazhdimin e përdorimit të tij në një farë mase. Megjithatë, në çerekun e dytë të shek.

Butrint: Gërmimet në Pallatin e Trikonkës VI është e qartë që kompleksi i Pallatit të Trikonkës ishte pothuajse i braktisur dhe jeta zhvillohej në zonat përreth tij si p.sh., në Shtëpinë e Tregëtarit. Shenja të një përdorimi sezonal për qëllime peshkimi por edhe për aktivitet të kufizuar industrial vërehen në strukturat pranë kanalit të Vivarit, ndërkohë qe gjysma e parë e shek. VI shënon edhe shfaqjen për herë të parë të varreve, disa prej të cilëve në amfora, në brendësi të kompleksit të trikonkës. Sidoqoftë, duhet pasur parasysh se panorama që rezulton në këtë zonë të qytetit, sigurisht që nuk reflekton të gjithë realitetin në Butrintin e shek. VI kur ndërtohen monumente si Baptisteri dhe Bazilika e Madhe. Përdorimi i një zone të trikonkës si varrezë vazhdon në gjysmën e dytë të shek. VI dhe në fillimin e shek. VII. Niveli i kufizuar i aktivitetit që rezulton këtu kontraston ndjeshëm me atë në Shtëpinë e Tregëtarit, banorët e së cilës duket se përdorin strukturat e trikonkës si burim materiali ndërtimi deri sa në mesin e shek. VII, ky kompleks braktiset përfundimisht dhe shenjat e para të ripërdorimit shfaqen vetëm në shek IX.

4. Zona e Pallatit të Trikonkës në periudhën mesjetare W. Bowden, A. Culwick, K. Francis, O. Gilkes, K. Lako dhe J. Price Është e vështirë të përcaktohet niveli i aktivitetit në zonën e trikonkës për periudhën nga mesi i shek. VII deri në shek. IX. Vlen të përmendet se sasia e kufizuar e materialit qeramik që daton në këtë periudhë u gjet e gjitha në brendësi të oborrit qendror të ish-domus-it, çka sugjeron se edhe në këtë periudhë oborri qëndroi i hapur. Rishfaqja e aktivitetit në shek. X dëshmohet nga një strukturë e mbivendosur në një prej ambienteve të zonës jugore e cila në shek. XI zëvendësohet nga një strukturë e ndërtuar me gurë. Të tjera struktura të tilla, të sinjalizuara nga prania e gropave për fiksimin e trungjeve mbi muret e antikitetit të vonë u gjetën në zonën perëndimore. Dimensionet e konsiderueshme të gropave në mure flasin edhe për dimensione të konsiderueshme të ndërtesave që ato formonin. Megjithatë, mungesa e të dhënave datuese për shkak të shtresës së hollë të dheut që mbulonte kreshtat e mureve, na lejon të sugjerojmë vetëm një datim të përgjithshëm, shek X–XIII, bazuar mbi faktin që sasia e konsiderueshme e materialit të rikuperuar nga kjo periudhë tregon se kjo është periudha më intensive e përdorimit të zonës së Trikonkës gjatë gjithë mesjetës. Oborri qendor i dikurshëm duket se vazhdon të qendrojë i hapur siç sugjeron pjesa e madhe e monedhave që u gjet në të, një fakt ndoshta i lidhur me përdorimin e oborrit si treg i hapur sezonal. Në fillim të shek. XIII, ashtu si dhe në Shtëpinë e Tregëtarit, aktiviteti jetësor në trikonkë duket se merr fund dhe konstatohet vetëm akumulim mbeturinash e ndoshta përdorim i zonës si varrezë. Nuk dihet se kur kjo zonë transformohet në pyll por në fotot e Ugolinit të viteve 1920–të dendësia e tij duket më e vogël se sa ajo e sotmja.

xiii

Sidoqoftë, stratigrafia e sipërme e zonës së Trikonkës u gjet shumë e trazuar, e ndoshta përgjegjës për këtë mund të jenë rrënjët e pemëve dhe kultivimi agrar i terrenit në periudhën post-mesjetare.

5. Shtëpia e Tregëtarit dhe muri rrethues i qytetit në shek. V–VII W. Bowden, Y. Cerova, A. Crowson dhe Emanuele Vaccaro Shtëpia e Tregëtarit e gërmuar në vitet 2001 dhe 2002 ndodhet ngjitur me krahun perëndimor të kompleksit të trikonkës dhe formon qartë një kompleks më vete që nga periudha më e hershme e saj, si dhe gjatë periudhës romake e deri ndoshta edhe në atë mesjetare. Sekuenca kronologjike e hershme është vetëm relativisht e përcaktuar si e para shek. V, çka nuk mundëson krijim e një ideje të qartë lidhur me natyrën dhe funksionin e strukturave më të hershme të kompleksit, të cilat ndoshta janë të lidhura me ndonjë ndërtesë tjetër në veri të Shtëpisë së Tregtarit. Një ndryshim i ndjeshëm ndodh në momentin e zgjerimit të kompleksit në drejtim të kanalit Vivari me ndërtimin e tre ambienteve, 38, 43, dhe 47. Duket se kompleksi ndante në krahun e tij lindor të njëtën rrugë me domus-in e viteve 400. Ndërtimi i një muri bllokues të kësaj rruge mes sallës apsidale të domus-it dhe murit lindor të Shtëpisë së Tregtarit nga pronari i domus-it, sugjeron se ky i fundit duhet të ketë qenë pronari i të dy komplekseve. Fokusi i ndërtesës në këtë fazë të hershme duket se ishte kanali i Vivarit siç indikon dhe dhoma 39 me pamje të hapur drejt tij. Megjithatë nuk mund të thuhet me siguri nëse kompleksi shërben për shfrytëzimin e peshkimit apo si pikë ankorimi ku zhvillohej aktivitet tregëtar. Periudha nga fillimi i shek. të V deri në mesin e tij përfaqësohet nga sekuenca e dyshemeve në dhomën 37, mbivendosja e të cilave tregon përdorim intensiv të këtij ambienti. Kjo sekuencë u korrespondon pjesërisht dy momenteve të rëndësishme kronologjike në domus-in fqinj: ndërtimi i peristilit dhe zgjerimi i kompleksit rreth vitit 400. Gjatë gjysmës së dytë të shek. V Shtëpia e Tregëtarit fillon të shfaqë gjithnjë e më shumë karakteristika të ngjashme me domus-in e trikonkës, ndër të cilat më e rëndësishme është prania e një kati të dytë në dhomën 37, ndërtuar mbi pilastra prej guri, të cilat konstatohen për herë të parë në këtë zonë të sitit. Ky moment korrespondon me shtresën sterile prej llumi të vënë re në galerinë e domusit, çka përforcon idenë e përballjes së banorëve të zonës buzë kanalit me problemin e ngritjes së nivelit të ujit. Megjithatë, sasia e konsiderueshme e fragmenteve prej mermeri gjetur në dhomën 37, fragmenti i një tasi prej xhami me dekorim në gdhendje, një objekt prej fildishi i ngjashëm me gurët e shahut, etj, kontrastojnë për nga natyra me gjetjet në trikonkën e kësaj faze e indikojnë një standard më të lartë jetese të banorëve të Shtëpisë së Tregëtarit. Sigurisht që ndërtimi i murit rrethues të qytetit shkaktoi riorganizimin e kompleksit në fjalë, disa prej dhomave

xiv

Nevila Molla

të të cilit duket se dalin jashtë përdorimit. Sidoqoftë, duket se muri respekton praninë e këtij kompleksi duke e zhvendosur linjën e tij për të përfshirë kompleksin në brendësi të qytetit. Edhe pse kjo zgjedhje mund të jetë bërë vetëm për të shfrytëzuar muret ekzistuese (një fenomen i vënë re edhe në pika të tjera të murit të antikitetit të vonë) është domethënës fakti që në zonën e oborrit të Shtëpisë së Tregëtarit ndodhet një nga të paktat porta të murit të qytetit që janë identifikuar nga kjo periudhë. Datimi i tij në fundin e shek. V u mundësua nga gjetja e materialit qeramik në themelin e hapur për ndërtimin e tij. Ndërtimi i murit rrethues të qyetit mund të shoqërohet me fushatën fortifikuese të ndërmarrë nga perandori Anastas (491–518) dhe pse fortifikimi i qyteteve në Iliricum duket se ishte pjesë e një procesi të vazhdueshëm të nisur nga Theodosi II. Çereku i fundit të shek. V, ashtu si dhe në trikonkë, shënon fillimin e procesit të shkatërrimit dhe të përdorimit të një pjese të kompleksit si burim materiali ndërtimi. Ky proces duket se ndodh paralel me atë të procesimit të midhjeve e guaskave të tjera siç indikon shfaqja e shtresave me mbetje të tilla në ambientin 38 dhe oborrin e shtëpisë (39). Rreth mesit të shek. VI një strukturë dy katëshe me çati tjegullash ndërtohet në pozicion këndi me murin e qytetit. Jetesa ne katin e sipërm të një banese bëhet një tipar gjithnjë e më i zakonshëm në antikitetin e vonë. Gjetje si objekte xhami dhe metali, llampa dhe qeramikë fine të zbulura nën shtresën e çatisë së rënë sugjerojnë se dhe pse pronar i një strukture me cilësi ndërtimi modeste, ky pronar kishte akses ndaj objektesh dhe mallrash importi cilësore. Ndërtimi i banesës ngjitur me murin mund të indikoje përdorimin e saj nga një funksionar zyrtar apo ushtarak por nga ana tjetër, mund të flasë edhe për mundësinë e dobësimit të autoritetit civil në mesin e shek. VI deri në atë masë sa individë privatë ishin në gjendje të zotëronin pika kyçe të qytetit për nevoja të tyre. Braktisja e kompleksit ndodh në fundin e shek. VI, datuar nga një monedhë e Justinianit II, e viteve 574/575 dhe një pjatë e tipit ARS e datuar në vitet 575–600. Ideja për ta shoqëruar këtë braktisje me destabilitetin e shkaktuar nga invazioni i dokumentuar sllav i viteve 586/7 është tunduese por duhet pasur parasysh se përcaktimi i shkallës së efektit të këtij sulmi është shumë i vështirë. Pas braktisjes së kompleksit duket se zona përdoret për procesimin dhe gatimin e midhjeve, aktivitet që shkaktoi akumulimin e një sasie të madhe mbetjesh të kësaj natyre në zonën brenda portës së qytetit, e cila qendron e hapur deri në shek. XIII. Vërehet se këtu mungon prania e varreve që karakterizon momentin final të antikitetit të vonë në Pallatin e Trikonkës, diferencë e cila mund të sugjerojë zotërues të ndryshëm të zonave në fjalë. Ashtu si dhe në Trikonkë megjithatë, periudha mes shek. VII – IX karakterizohet nga një sasi shume e vogël materiali, çka e bën njohjen e historisë së kompleksit për këta shekuj të pamundur.

6. Shtëpia e Tregtarit në periudhën mesjetare W. Bowden, A. Crowson, M. Logue dhe A. Sebastiani Niveli i aktivitetit për shek. VII-XI në zonën e Shtëpisë së Tregtarit është i vështirë të përcaktohet, por funksionimi i portës në murin rrethues tregon se kjo zonë mund të ketë qenë e përshtatshme për zbarkim dhe procesim të midhjeve në mënyrë sezonale, çka përforcohet nga prania e vatrave të përkohshme të zjarrit që mundësonin gatimin e tyre. Nga shek. IX e në vazhdim vërehet një rritje në praninë e monedhave dhe qeramikës por nga fundi i shek. IX e gjatë shek. X vërehet gjithashtu procesi intensiv i depozitimit të mbeturinave, gjë qe tregon se ky kompleks mbetet një zonë periferike e Butrintit në këtë periudhë. Shenja të një përdorimi më të formalizuar dhe koherent shfaqen në periudhën mes vitit 1025 e shek. XII, kur vërehet prania e një rrugice të shtruar me kalldrëm nga porta në murin rrethues në brendësi të kompleksit. Ndoshta në po këtë periudhë mbi themelet e një prej dhomave të mesit të shek. VI (40) ndërtohet një strukturë e re, funksioni i të cilës nuk është i qartë. Nga shek. XII vërehet një rënie në intensitetin e aktivitetit në zonën në fjalë dhe regjistrohet vetëm prania e vatrave të rrethuara me struktura murore mbrojtëse dhe e gropave të trungjeve që dëshmojnë praninë e një ose më shumë strukturave prej druri. Një ndryshim i ndjeshëm vërehet gjatë shek. XIII dhe XIV, mesa duket i nisur me bllokimin e portës në murin e qytetit, e cila kishte funksionuar për rreth 700–800 vjet. Ky moment ndiqet nga një proces akumulimi i një dheu të zi dhe me përmbajtje të theksuar humusi, i cili përmban një sasi të madhe qeramike, kocka kafshësh, etj., që duket se vijnë si pasojë e depozitimit të mbeturinave shtëpiake nga ndonjë banesë në afërsi, ndoshta në veri të ish-Shtëpisë së Tregtarit dhe flasin qartë për vazhdimin e banimit të sitit në pjesë të tjera të tij. U vunë re gjithashtu, shenja të kultivimit të kësaj zone por në të njëjtën kohë identifikimi i një numër varresh të përqendruar sidomos në dhomën 40 tregon përdorimin e një pjese të saj edhe si varrezë. Rreth fundit të shek. XIV muri rrethues i qytetit pajiset me një kullë të re të ndërtuar mbi themelet e një prej ambienteve të antikitetit të vonë, e cila projektonte jashtë linjës së murit rrethues. Ky moment mund të shoqërohet me një sërë riparimesh të mundshme që duhet të ketë realizuar Castellanus i Butentro pas zotërimit të qytetit nga venecianët në vitin 1386. Në shek. XV dhe XVI vazhdon kultivimi agrar i zonës por dhe depozitimi i mbeturinave, në nivelet e sipërme të të cilave u gjet një sasi e madhe fragmentesh qeramike të shek. XV–XVI, prania e të cilës mund të lidhet me përdorimin e kullës së vonë mesjetare. Kjo e fundit duket se i nënshtrohet procesit të rrënimit në shek. XVII, kur regjistrohen edhe momentet e fundit të depozitimit të materialit arkeologjik ne zonën e Shtëpisë së Tregëtarit.

Butrint: Gërmimet në Pallatin e Trikonkës

Interpretim dhe Diskutim 7. Mozaikët e Domus-it J. Mitchell Në periudhën e vonë romake shtëpitë e elitës sociale përfaqësonin një arenë të vërtetë të paraqitjes së identitetit dhe të statusit social të individëve që i zotëronin ato. Të projektuara për të akomoduar një numër gjithnjë e më të madh vizitorësh, me fokus struktural rreth një oborri të madh peristil (ekuivalent i atriumit të rezidencave të periudhës së hershme romake) nga i cili mund të hyhej në të gjitha ambientet, këto shtëpi mbartin një ansambël shumë të pasur dekorativ të përbërë nga elementë skulpturorë, afreksë, mozaikë, etj. Në domus-in e trikonkës elementi dekorativ më i ruajtur përfaqësohet nga mozaiket e dyshemeve të cilët vijnë nga faza 2 dhe 3 e datojnë nga shek. III deri në fund të shek. IV. Nga faza e shek. III ruhet mozaiku i vestibulës dhe ai i dhomës së pritjes. Mozaiku në dyshemenë e vestibulës spikat për ngjyrat e larmishme dhe të forta, ndër të cilat dominojnë e bardha dhe e zeza të shoqëruara nga roza, e kuqa dhe dy nuanaca të blu-grisë, si dhe për cilësinë e lartë dhe preçizionin në punimin e tij. Ai përbëhej nga një panel qendror i dekoruar me një dredhëz komplekse me motiv svastike, brenda të cilit paraqiten maska nga repertori i Komedisë së Re, përfshirë edhe dy protagonistët kryesorë: një vajzë e re e shoqëruar nga babai i saj i vrenjtur. Këto dy maska bashkë me dy të tjera duhet të kenë rrethuar një temë qendrore në mes të dyshemesë, që ndoshta mund të ketë qenë një figurë e plotë e Dionisit, Zotit të teatrit. Motivi i maskave tragjike dhe komike shfaqet shpesh në repertorin dekorativ të shtëpive romake nga shek. II deri në atë të IV, i ekzekutuar në forma të ndryshme (mozaikë, gdhendje, në reliev, etj.). Simbolika e tyre mbart nuanca komike që e bëjnë vizitorin të ndjehet i mirëpritur, ndërkohë që grimasa dhe goja e hapur i japin imazhit një forcë trembëse ndaj vizitorit keqdashës, duke u dhënë këtyre motiveve funksion apotropaik (Maguire and Duncan-Flowers 1989, 16–18; Prentice 1906). Ngjyrat e mozaikut të vestibulës karakterizojnë edhe mozaikun e galerisë 30 metërshe në perëndim të saj, por motivet dekorative këtu ndryshojnë. Një drejtkëndësh i gjatë që shtrihet në të gjithë dyshemenë mbart një dekor kompleks prej oktagonësh në ngjyrat bardhë e zi, të mbushur me trekëndësha dhe katrorë me elementë dekorativë si svastika dhe kryqe që theksojnë pjesën qendrore të secilit prej tyre. Korniza rreth panelit qendror përbëhet nga dy fasha; e para trengjyrëshe (e kuqe, rozë, blue e lehtë ose gri) me motive floreale dhe ajo në kufi me murin gjithashtu me motive floreale mjaft të ngjashme me atë në mozaikun e vestibulës. Shtrati mbi të cilin janë vendosur të dy mozaikët e këtyre ambienteve ka përbërje të njëjtë: cocciopesto, rërë, fragmente tjegullash dhe gëlqere, sipër të cilit është shtruar një shtresë më fine me përzierje pudre prej tjegulle, guri dhe gëlqereje. Mozaiku i dhomës së pritjes u përket dy fazave që korrespondojnë me fazën e ndërtimit të kësaj dhome dhe

xv

me atë të modifikimit të saj nëpërmjet shtimit të apsidës. Në fazën e parë të tij, mozaiku përbëhet nga një seri komplekse motivesh abstrakte të vendosur në panele të rradhitur nga qendra e dhomës në drejtim të mureve. Një motiv rrethor i vendosur në qendër të saj nuk ruhet mjaftueshëm. Katër këndet e dyshemesë dekoroheshin me vazo lulesh shumëngjyrëshe, prej të cilave ruhen vetëm dy. Në fazën e dytë një fashë e gjerë dhe e çrregullt prej tre kuadratësh shtrohet në zonën veriore përpara apsidës së sapo ndërtuar, duke shënuar një lloj pragu për në apsidë. Dekori konsiston në motive gjeometrike, rozeta katër-petalëshe dhe kryqe, disa prej teserave të të cilëve janë prej xhami. Mozaiku shumëngjyrësh i apsidës nuk ruhet mirë, por në konturet e saj vërehet një fashë në formë harku, e cila është dekoruar me motive në formë rruazash dhe me një kryq në kulmin e tij. Në brendësi të harkut ndodhet një maskë e madhe femërore, prej të cilës ruhet vetëm pjesa e sipërme. Pjesa qendrore e apsidës ishte e mbushur me gjethe shumëngjyrëshe të mbivendosura ndaj njëra-tjetrës. Ky stil dekorimi është tipik për apsidat duke filluar nga shek. II dhe III e vazhdon të përdoret gjatë periudhës së vonë romake (shembuj: Germain 1969, 79, no. 98, pl. XXV–Timgad). Një fragment amfore nga Gaza, gjetur në shtratin e mozaikut të apsidës dhe datuar në shek. IV ofron një ide për datën e shtrimit të këtij mozaiku. Mozaiku i dyshemesë së vestibulës perëndimore përmban një dekor më të thjeshtë prej një motivi të vetem, peltae, egzekutuar në rozë dhe blu mbi një fushë të bardhë që i jep atij karakterin e një tapeti. Spikat pranija e një mbishkrimi të ekzekutuar me tesera në ngjyrë të kuqe në buzën perëndimore të dyshemesë, pranë pragut për në hyrjen e peristilit. Dhe pse i fragmentuar, përmbajtja e mbishkrimit është deshifruar si emri dhe rangu i pronarit të shtëpisë. Emri ose titulli mbaron me arius, kurse rangu i tij, lamprotato, ekuivalent me clarissimus, indikon rang senatorial. Mbishkrimi është vendosur në një zonë ku mund të vërehet qartë nga vizitorët e shtëpisë. Mozaikë polikromë dekoronin dyshemetë e katër portikëve të peristilit të domus-it në vitet e fundit të shek. IV, por duket se vetëm ai në portikun perëndimor ka arritur deri në ditët e sotme. Nga ansambli i larmishëm dekorativ prej motivesh gjeometrike, floreale dhe zoomorfe, kryqe, etj., vlen të vihet veçanërisht në dukje pranija e dy syve të mëdhenj të mbivendosur ndaj njëri-tjetrit në panelin qendror të pjesës veriore të këtij portiku. Ky është një tjetër motiv apotropaik, funksioni i të cilit është mbrojtja e shtëpisë ndaj ‘syrit të keq’. Ai nuk ndeshet shpesh në repertorin dekorativ të mozaikëve (Engemann 1975: tab. 10a – Jekmejeh afer Antioch, 11a–b – Rome, 14c – Sousse). Sondazhi në një prej zonave të dëmtuara të mozaikut prodhoi qeramikë të tipit ARS, e cila daton prej viteve 360 deri në fillim të shek. V, si dhe një fragment ene kuzhine të fundit të shek. IV – fillim i shek. V, një fragment mermeri të punuar në reliev që ndoshta i përket një stibadium-i dhe mund të datojë në mesin e shek. IV, si dhe fragmente xhami dhe suvaje. Ky material ofron një datim të përgjithshëm për shtrimin e mozaikut në fjalë rreth fundit të shek. IV.

xvi

Nevila Molla

8. Arkitektura elitare dhe hapësira në Butrintin e antikitetit të vonë W. Bowden Studimet e dekadave të fundit i kanë zgjeruar mjaft njohuritë tona mbi natyrën e rezidencës nga periudha romake në ate antike të vonë, si rezultat i të cilave shtëpia romake shihet si një element që mbart dinamiken dhe fleksibilitetin e lartë të marrdhënieve shoqërore në botën romake nëpërmjet përdorimit të modeleve arkitektonike komplekse dhe një fjalori dekorativ të pasur. Këto modele mbeshteten fort në huazimin nga tradita e arkitekturës dhe dekorit të ndërtesave publike. Është e rëndësishme të vihet re se luksi dhe madhështia e komplekseve private të kësaj kohe nuk mund të gëzonin status nëse ato nuk ishin në raport me rolin e individit në jetën urbane publike si pronar tokash ose prodhues (Purcell 1991; Wallace-Hadrill 1994, 4–6). Pjesa më e madhe e të dhënave vjen nga qytetet vezuviane, Pompei dhe Herkulanum, por një pjesë e tyre njihet edhe nëpërmjet përshkrimeve të Vitruvius dhe autorëve të tjerë të kohës, të cilët ofrojnë terminologjinë me të cilën emërtojmë ambientet dhe hapësirat e gjetura nga gërmimi (tablinum, triclinium, cubiculum, etj.). Si rezultat i një procesi të gjerë urbanizimi qytetet romake u dominuan nga ndërtesa të tilla luksoze të cilat shpesh mund të mbushnin insulae të tëra. Jo të gjitha rezidencat e kohës së vonë romake ishin me dimensione kaq madhështore e megjithatë, edhe shtëpitë më të vogla reflektojnë të njëjtin zhvillim arkitektonik, çka ka bërë disa studiues të sugjerojnë dallimin e një ‘shtrese të mesme’ në shoqërinë e kohës (Ellis 2006). Rasti i Pallatit të Trikonkës e pasuron studimin e kësaj teme, mbi të gjitha sepse na kujton të jemi të kujdesshëm në intepretimin e këtij tipi rezidencash bazuar vetëm në formën e tyre në planimetri. Gërmimi i tij tregon se megjithë natyrën madhështore që karakterizon triclinuimin, projektimi dhe ndërtimi i pjesës tjetër të kompleksit ishin të kushtëzuar nga prania e strukturave të mërparshme, çka e detyroi pronarin e Trikonkës të dështojë në arritjen e një projekti të artikuluar qartë. Për pasojë, pjesët përbërëse të tij janë të pozicionuara në mënyrë jo të rregullt ndaj njëra-tjetrës dhe dimensionet e strukturave më të hershme minimizohen në ekstrem nga madhështija e peristilit asimetrik dhe triclinium-it. Ashtu si dhe në rastin e vilave më të hershme romake ndërtimi i Pallatit te Trikonkës ishte një proces gradual i realizuar nga disa breza, një faktor ky që pati një ndikim të ndjeshëm në arritjen e koherencës arkitektonike të kompleksit. Pallati i Trikonkës që shihet sot në Butrint nëpërmjet mbetjeve të tij në planimetri, është në fakt një grup dhomash dhe fasadash madhështore të ardhura nga periudha të ndryshme, të cilat kurrë nuk mundën të krijojnë formën e plotë të asaj për të cilën ishin ideuar. Lehtësisht lind pyetja nëse një faktor i tillë ndikoi në braktisjen e programit të ndërtimit të fazës së fundit për shkak të ndoshta, humbjes së interesit dhe shpresës se ndërtesa do te arrinte ndonjëherë modelin arkitektonik elitar që kishte aspiruar. Në të njëtën

kohë, megjithatë, duhet pasur parasysh se me zhvillimin që konstatohet në Pallatin e Trikonkës mund te jetë e ngjashme panorama e shumë siteve-domus që shërbejnë si shembuj përfaqësues, por të cilët i njohim vetëm nëpërmjet planimetrive të ndërtesave që publikohen gjithnjë e në vazhdim, e që për pasojë marrin një formë kanonike.

9. Ndryshimi urban dhe siti i Pallatit të Trikonkës W. Bowden Gërmimi arkeologjik në Pallatin e Trikonkës dhe Shtëpinë e Tregëtarit paraqet për shek. V–VII një panoramë që po bëhet relativisht e zakontë për qytetet e antikitetit të vonë. Që prej viteve 420, vendin e kompleksit madhështor, e zënë banesa të vogla të vendosura brenda skeletit të shtëpisë së mëparshme. Megjithatë, edhe pse qartësisht të një tjetër shkalle në krahasim me domus-in nuk duhet mbitheksuar ideja e varfërimit të këtyre banesave, të cilat formonin një grup në vetvete dhe në rastin e Shtëpisë së Tregëtarit kishin madje, ambicje për të praktikuar aspirata dhe elemente të një arkitekture elitare. Këto ndërtesa shfaqen në zonën e Trikonkës ndoshta rreth 75 vjet më herët se sa koha e ndërtimit të strukturave madhështore religjoze të Butrintit (Bowden and Mitchell 2004) kështuqë, ato nuk mund të konsiderohen si indikuese të një procesi të përgjithshëm varfërimi të qytetit (ashtu si edhe shfaqja e ndërtesave religjoze nuk mund të konsiderohet si indikues i një prosperiteti të përgjithshëm) (Bowden 2003, 105–59). Një argument i ngjashëm paraqitet nga Lewit (2005, 255) në lidhje me shfaqen e këtij fenomeni në site urbane të Evropës perëndimore. Në Barcelonë për shëmbull, nënndarja e banesave, përdorimi i dyshemeve prej balte dhe i spoilias, dhe shpërdorimi i zonave publike koinçidon me periudhën kur qyteti ishte një qendër e rëndësishme peshkopale dhe rezidenca e mbretit të vizigotëve. Po kështu, në Mérida, nën-ndarja e banesave të mëdha ndodh në të njëjtën kohë me periudhën kur madhështia e arkitekturës religjoze arrin pikun, ndërsa në Toulouse ky proces i shek. V ndodh njëkohësisht me ndërtimin e kishave në shkallë të gjerë. Këto konstatime shërbejnë për të na kujtuar natyrën shumë-faqëshe të procesit të ndryshimit në ambientin urban të antikitetit të vonë. Lewit e interpreton këtë lloj të ri banese si reflektues të traditave në ndryshim të shoqërisë së antikitetit të vonë sesa si varfërim i saj. Edhe pse arkeologjia e Pallatit të Trikonkës shtyn drejt idesë së një lloj thjeshtëzimi të arkitekturës shtëpiake të shek. V dhe VI, është e qartë se Butrinti ishte një komunitet që mbeti në kontakt me rrjetin e botës mesdhetare të shkëmbimit të mallrave gjatë gjithë shek. VI (Reynolds 2004). Megjithatë, me fillimin e shek. VII këto lidhje tranformohen në gati asgjë dhe qyteti nuk merr më monedha ose të paktën, nuk vërehet t’i përdorë ato në shkëmbimin e përditshëm. Ky reduktim i lidhjeve të Butrintit me botën mesdhetare dhe fundi i ekonomisë monetare duhet të këtë ndodhur brenda kohëzgjatjes së jetës njerëzore. Prindërit

Butrint: Gërmimet në Pallatin e Trikonkës e këtij brezi të banorëve të Butrintit për më tepër, duhet të kenë parë dhe marrë pjesë në ndërtimin e Bazilikës së Madhe dhe Baptisterit. Ndërkohë që gërmimet në Pallatin e Trikonkës paraqesin një panoramë të zakontë për Mesdheun e antikitetit të vonë, ato sjellin qartë në vëmendjen tonë faktin se procesi i ‘tranzicionit urban’ solli ndryshime të ndjeshme, të cilat ishin të dukshme për banorët e qytetit të antikitetit të vonë.

10. Nga insula romake në lagje mesjetare R. Hodges Në mijëvjeçarin pas shek. VI – fillimit të shek. VII zona e Trikonkës dhe Shtëpisë së Tregëtarit përdoren vetëm në mënyrë episodike. Pjesë e arsyes për këtë ishte përmbytja sezonale e zonës buzë kanalit të Vivarit por gjithashtu, shfaqja e fenomenit sizmik gjatë shek. IV i bëri banorët e Butrintit të kuptojnë që nga shek. V porblematikën e përdorimit të kësaj zone të qytetit për banim. Megjithatë, gërmimi arkeologjik tregon se zona nuk u braktis plotësisht. Përkundrazi, duke qenë në arterien që lidhte liqenin e pasur në peshk të Butrintit me ngushticën e Korfuzit, kjo zonë ofronte një pozicion optimal për përdorim në periudha të caktuara të vitit. Gjatë periudhës së shek VII–IX, për të cilën të dhënat mbi vendbanimin janë të kufizuara, duket se kastron i Butrintit ishte përqendruar në zonën perëndimore të fortifikimeve të qytetit, për t’u transferuar më pas në fushën e Vrinës (Hodges 2008). Procesimi i midhjeve duhet të ketë filluar që në shek. VIII–IX dhe pse evidenca për këtë nuk është definitive. Në fillimin e shek. X duket se zona buzë kanalit ishte fokusi i një aktiviteti intensiv të procesimit të midhjeve, një aktivitet i përqendruar në zonën brenda portës së qytetit, ndërkohë që peristili i vjetër vazhdonte të mirëmbahej si një shesh qendror i hapur, i cili duhet të ketë ofruar një locus të përshtatshëm për zhvillimin e aktivitetit tregëtar. Evidenca strukturore dhe materiale sugjeron se Butrinti përjetoi një periudhë zhvillimi domethënës në shek. X–XII. Përdorimi i zonës së Trikonkës bëhet shumë më intensiv se më parë dhe pjesë e këtij procesi mund të ketë qenë sjellja e një sasie të madhe materiali për nivelim të disa zonave. Është e mundur që në këtë periudhë të jetë ndërmarrë gjithashtu një program riparimi dhe fortifikimi. Por si një sektor rezidencial, trikonka, në ndryshim me insulën që përfaqësonte në periudhën romake, nuk u zhvillua kurrë në dimensionin e një lagjeje të krahasueshme me lagjet e qyteteve të periudhës së mesme bizantine siç ndodh p.sh., në Amorium (Lightfoot dhe Lightfoot 2007) ose Pergamum (Buchwald 2007). Kjo periudhë përfundoi me humbjen e Butrintit ndaj kualicionit të kryqëzatës së katërt, në mos edhe më herët. Tashmë nukleusi i Butintit mesjetar, ndoshta për shkak të ngritjes së nivelit të ujit, ishte zhvendosur në shpatullat e akropolit dhe rrethohej nga një unazë kishash që fillonte nga zona e gjimnazit dhe e kishës se Baptisterit dhe kalonte

xvii

pranë portës perëndimore, në perëndim dhe në Bazilikën e Madhe në lindje, duke shënuar kufijtë jugorë të zonës së banuar të qyetit. Me bllokimin e portës 1 dhe 2 dhe në pamundësi për akses ndaj kanalit Vivari, zona brenda mureve kthehet tashmë në pronë bujqësore. Faza e fundit që i takon periudhës veneciane ka në vëmendje vetëm mbrojtjen e burimeve të çmuara peshkore. Materiali i pasur qeramik tregon se megjithë praninë e varfër të evidencës strukturore, ata që merreshin me peshkim gëzonin standartet e jetesës urbane, të paktën përsa u përket objekteve të tyre të tavolinës. Ishte pa dyshim vlera e burimeve peshkore qe i shtyu venecianet të mbronin Butrintin me çdo kusht deri në vitin 1797, kur ai dhe Korfuzi përfundimisht ranë në duart e francezëve me nënshkrimin e traktatit të Campo Formio.

11. Përfundime R. Hodges Kur filluam këtë projekt në vitin 1994 me kolegët tanë shqiptarë, ishim të tundurar ta interpretonim Butrintin sipas modelit të ndërtuar nga Ugolini, bazuar mbi gërmimet e tij të viteve 1928–1936. Sipas tij Butrinti konfigurohej si një nukleus i vogël i vendosur në kreshtën e kodrës dhe i rrethuar nga një zonë e ngushtë urbane në këmbet e saj, pa pësuar ndryshime të ndjeshme për gjithë periudhën nga epoka helenistike deri në atë veneciane. Ky interpretim nuk merrte në konsideratë ndryshimet e ndjeshme ambientale e sidomos, ndryshimin në nivelin e ujit me të cilin Butrintit i është dashur të përballet vazhidimisht. Gërmimet në Pallatin e Trikonkës tregojnë se me braktisjen e qendrës së qytetit romak pronat pranë kanalit te Vivarit u zgjeruan në shkallë dhe ambicje, një situatë që ndryshoi me ndërtimin e murit të qytetit në fund të shek. V. Fortifikimet ishin një element domethënës në lidhje me identitetin urban dhe përdoreshin shpesh si një simbol për të përfaqësuar qytetet në periudhën e vonë romake (La Rocca 2000). Megjithatë, ambicja e Buthrotianëve u eklipsua nga Saranda ku një mur rrethues më cilësor, me kulla mbrojtëse, u ndërtua në të njëtën kohë me kompleksin religjoz të Dyzet Shenjtorëve në kodrën mbi port (Bowden and Mitchell 2003; Hodges 2006, 218–20). Me ndërtimin e ketij sanktuari të ri me proporcione metropolitane, Saranda duket se është tashmë qendra dominuese e rrethinave të saj dhe përfaqëson gjithashtu, një qendër religjoze të krahasueshme me kompleksin e Shën Mihalit në malin Gargano, në Pulian veriore, të dy këta, rajone që përjetojnë prosperitet në fillim të shek. VI. Por Butrinti me gjithë aktivitetin e tij të gjatë tregtar nuk qendron në të njëjtin rang me Dyzet Shenjtorët. Kjo e fundit në një mënyrë apo një tjetër, mundi të përvetsonte famën e Butrintit të periudhës helenistike si një vend pelegrinazhi. Kështu, edhe pse Butrinti mbeti një vend qendror në nivel krahinor gjatë periudhës se mesme bizantine, ai ishte nën hijen e Dyzet Shenjtorëve deri në shek. XIII kur, Korfuzi zhvillohet si qendër administrative dhe tregtare.

xviii

Nevila Molla

Banorët pranë kanalit në Butrint duket se ishin pronarë tokash me të drejtë peshkimi, pasuria e të cilëve varej nga kërkesa në rajon për produktet lokale të Butrintit. Një paralel i përshtatshëm për këtë është shembulli i venecianëve rreth një mijëvjeçar më vonë, në shek. XVII, XVIII, kur sasi shumë të mëdha peshku furnizonin Korfuzin. Në këto kushte mund të supozojmë se retorika e shtëpive qytetare romake gjatë shek. II–V bazohej mbi kombinimin e vlerave rezidenciale klasike me tregtinë private të peshkut dhe produkteve agrare. Në kontrast me këtë situatë në shek. XI, rrugët me kalldrëm dhe fortifikimet e riparuara sugjerojnë praninë e mundësive ekonomike për të rindërtuar lagjen buzë kanalit. Dhe pse jo me dendësi të madhe popullsie duket se ekziston vullneti për ringjalljen e topografise urbane të Butrintit. Në fillim të shek. XIII, ngritja e nivelit të ujit e bëri këtë të pamundur. Parë në këtë prizëm, mbetjet arkitektonike duken të çuditshme. Rrënojat nga periudha romake janë madhështore dhe në fakt, mbishkrimi në mozaikun e domus-it të shek. IV

flet për praninë e një njeriu të lartë. Ndoshta prosperiteti i Butrintit në periudhën romake, i ilusturar mirë nga shtëpitë qytetare në bregun e Vivarit, ishte më i dukshëm sepse në pesë shekujt nga periudha e Augustit deri ne atë të Anastasit, Butrinti jo vetëm rivalizoi, por ndoshta e eklipsoi Korfuzin, ku ende nuk janë gjetur prona të tilla. Gërmimet në Pallatin e Trikonkës kanë zbuluar pa dyshim evolucionin tentativ të një lagjeje urbane edhe pse historia e saj nuk është ajo që kishim imagjinuar kur gërmimet filluan në vitin 1994. Sigurisht që këto gërmime kanë hedhur dritë mbi historinë urbane bizantine por dhe në këtë rast panorama që kemi zbuluar është episodike dhe në një linjë me ritmet e ndryshueshme të lidhjeve të Butrintit me detin Adriatik. Kolonia romake e ri-themeluar nga Augusti, ashtu si dhe porti bizantin i shek. XI luajtën një rol të rëndësishëm si vende qendrore të vlerësuara për burimet në peshk, por si një qytet Butrinti ri-përjetoi vetëm shkurtimisht ambicjet e arkitektit të tij augustian.

1  Introduction William Bowden and Richard Hodges



[A]mong the fundamental problems of Byzantine history it would be hard to name one that has been studied less than has that of the cities. (Ostrogovsky 1959, 47)

When Ostrogovsky wrote these words, the cities of the post-Roman Mediterranean were little understood. Over the following decades, however, the late antique city has become the subject of an ongoing academic debate and the study of Roman and post-Roman urbanism has effectively become an academic sub-discipline, generating a huge body of literature within the wider field of late antique studies (e.g. Brogiolo, Gauthier and Christie 2000; Brogiolo and Ward Perkins 1999; Carver 1993; Christie and Loseby 1996; Lavan 2001; Liebeschuetz 2001; Popovič 1984; Rich 1992). The Butrint Foundation’s programme has focused on Butrint in southern Albania, hitherto a little-known ancient and medieval port in a Mediterranean context. Given the opportunity to survey and excavate not only in several key areas within the deserted town, but also within its lagoonal micro-region, since 1994 we have been developing a new paradigm for the nature of an Adriatic Sea port between the Hellenistic and Ottoman ages (cf. Hodges 2006). As outlined in the introduction to Byzantine Butrint (Hodges et al. 2004), the first volume dealing with the results of the Butrint Project, the ongoing debate on urban transformation formed the academic background to the Butrint Project, which is in its fifteenth year at the time of writing. Unsurprisingly over this long duration the debate has changed and fluctuated, while at the same time the project itself has developed and responded to the changing circumstances of post-communist Albania. The Butrint Project in 2011 is responding to questions that have developed out of the research at the site over the last 17 years, yet also has to be understood in the context of the changing nature of the wider debate on Roman and medieval urbanism. This volume describes the excavations of a large urban sector of Butrint. The excavations followed small scale excavations made by earlier archaeological missions to

Butrint, but in this case involved uncovering almost all the so-called Triconch Palace and part of an adjacent building, which we called the Merchant’s House.1 This sector of Butrint lies midway along the Vivari Channel, and it was clearly a locus of activity long before, and long after, the construction of the 5th-century Triconch Palace. Indeed, the present volume describes the sequence of occupation here from the 2nd century or earlier until the 16th century and later. This sequence encompasses a succession of Roman residential buildings that last in one form or another until the desertion of the Roman town in the early 7th century. It also encompasses the reoccupation of the area, first intermittently in Middle Byzantine times, then intensively for a brief spell in the 10th to 12th centuries, before the channel-side plots were turned over to allotments and the occasional location of fishing activities prior to the making of the woodland park today. The archaeology of this area then, represents a distinctive section of the ancient city, different in character from the more familiar public centre around the theatre, different, too, from the hillslopes of the acropolis, or indeed, the acropolis itself. In many ways, the excavations reveal the rhythms of Butrint as a small but enduring port close to the meeting of the Adriatic and Ionian Seas and chart its changing maritime connections. This report does not describe the changing topography of the port in Roman and post-Roman times. This will only be possible once the full reports on other excavations in the forum, in the suburb on the Vrina Plain, and the maritime villa at Diaporit, for example, have been completed. Nevertheless, we offer a further opportunity (enlarging upon our earlier observations – see Bowden 2003; Hansen 2009; Hodges 2006; Hodges 2008; Hodges, Bowden and Lako 2004) to examine how an important sector in this town evolved, and how this history begins to throw new light on our understanding of urbanism in the Adriatic and Ionian Seas.

William Bowden and Richard Hodges



Phoenicê

Saranda

ALBANIA

CORFU Butrint GREECE 0

10km

Lake Butrint Diaporit

Butrint National Park boundary

Butrint Vrina

Xarre Mursia

Malathrea Çiflik

0

Çuka e Aitoit

5km

Figure 1.1. Butrint in relation to its surrounding region

The site of Butrint Butrint sits at the crossroads of the Mediterranean, commanding the sea-routes up the Adriatic Sea to the north, across the Mediterranean to the west, and south through the Ionian islands. Like ancient Dyrrhachium (modern Durrës) to the north it also controlled a land-route into the mountainous interior. The abandoned ancient and medieval port is located 3 km inland from the Straits of Corfu in south Albania (Fig. 1.1). For much of its long history the settlement was confined to a hill on a bend in the Vivari Channel, which connects the Straits to the large inland lagoon of Lake Butrint. A narrow plain, formerly a marsh, separates the channel from a band of hills to the south,

along which runs the present frontier between Albania and Greece. Immediately east of Lake Butrint, a range of hills and low mountains rise up to 824 m, effectively creating a basin around the ancient city and the inland lake. The walled city, inscribed as a Unesco World Heritage Site in 1992, covers an area of c. 16 hectares (Fig 1.2), but geophysical survey on the eastern side of the Vivari Channel shows that at times in antiquity Butrint covered as much as 30 hectares (see Bescoby 2007) (Fig. 1.3). The walled city comprises two parts: the acropolis and the lower city. The acropolis is a long narrow hill, approximately 200 m long and 60 m at its widest, that rises up to 42 m above sea level at its east end; its sides are accentuated

1.  Introduction



Figure 1.2. The walled city of Butrint, with the Triconch Palace in the immediate foreground

by a circuit of walls that separate it from the natural and artificial terraces gathered around the flanks of the hill. The lower city occupies the lower-lying contours down to the edge of the Vivari Channel. Remains of a cemetery are recorded on the spine of the hill running west from the acropolis (Budina 1988; Ugolini 1937, 174), but its extent is unknown. The most obvious monument outside the city walls, on the opposite side of the channel, is the Triangular Castle, which after 1572 became the nucleus of the early modern settlement (Karaiskaj 1980, 33–35; Karaiskaj 2009, 95–105; Leake 1835, 95). Beyond the fortress to the east, opposite the walled site, there are substantial remains of Roman to Byzantine date. These formed part of the Roman and late antique town. The so-called Triconch Palace at Butrint was first noted in the 1920s, when the Italian Archaeological Mission to Butrint interpreted a tri-apsidal building as “una chiesa bizantina”. The interpretation of the building as a martyrium church was followed subsequently until 1994 when it was recognised as the triclinium of a palatial late Roman domus. From 1994 until 2003, the Triconch Palace and its surroundings were the subject of extensive excavations revealing a rich late antique and medieval occupation sequence.2 The intention of these excavations was to understand how a major residence developed during the Roman period and to understand what happened to it in the post-Roman period. It was intended that these excavations would inform the ongoing debates on the nature of late Roman aristocratic housing (discussed further in Chapter 8) and provide an insight into the changing social and economic fortunes of Butrint itself between the Roman and medieval periods.

A brief overview of the excavations The Triconch Palace lies of the south side of the city close to the Vivari Channel that connects Lake Butrint to the Straits of Corfu. The excavated area encompassed the area of the late antique domus together with an area to the west that appeared to have a different, possibly commercial, function (Fig. 1.4). It became clear that the area represented three separate building plots and something of the relationship between these plots (and perhaps their owners) could be discerned through the excavation (Bowden, Hodges, and Lako 2002). Of equal importance was the relationship between the buildings and the Vivari Channel and the way in which those relationships were changed by the construction of the late antique city wall, which was dated by the excavations to the first half of the 6th century. The earliest origins of channel-side occupation in the Triconch Palace area could not be determined. However, as the Roman colony of Buthrotum prospered, expanding out from its Hellenistic nucleus at the foot of the acropolis hill, a modest town house occupied the plot beside the Vivari Channel. The earliest known structures date to the 2nd century AD. By the end of the 4th century this was enlarged into a major house – a domus – with a small peristyle. This relatively small but affluent property provided the starting point for making a much more ambitious dwelling here at the start of the 5th century. By occupying the adjacent properties, the architect was able to enlarge the peristyle, create an impressive marine doorway, and add a triconch dining room of substantial proportions. But the palace was never completed. Instead, its rooms were occupied by several small structures. Notwithstanding the enclosure of the area behind a town fortification in the first half of



Lake Gate

Lion Gate Well of Junia Rufina

Mausoleum

North Gate

Great Basilica

Acropolis Basilica

West Gate

Theatre

Bridge

Forum Gymnasium

Western Defences

William Bowden and Richard Hodges

Tower Gate

Castle

Baptistery

Monument Temple mausoleum

Triconch Palace Merchant's House Vrina Plain settlement

Venetian Tower

Triangular Castle

0

Figure 1.3 Butrint, showing all extant monuments

200m

1.  Introduction

Plot 1

Plot 2

 Plot 3

Merchant's House

Triconch Palace

0

20m

Figure 1.4. The excavated area of the Triconch Palace and the Merchant’s House with possible property boundaries marked

the 6th century, these small dwellings and their successors were used intermittently until the mid to late 6th century, with their occupants able to access a range of goods from across the central and eastern Mediterranean. Major change came by the mid 7th century when the area was abandoned, and thereafter this area was seldom occupied in any permanent form. The rising water table, we surmise, may have made parts of it uninhabitable in the winter months. From the 10th to 12th centuries, however, the walled town became the focus of renewed activity, evidenced by deposition of significant quantities of material in the former courtyard area of the Triconch Palace. This was seemingly the context for new construction – at least one post-built dwelling that later was re-fashioned in stone was discovered here. Yet within a century or so, occupation of the area once again diminished. From the 13th century onwards, this section of the medieval town was maintained, we surmise, principally for its gardens and in all probability fishing activities. The adjacent area to the west of the Triconch Palace contained a substantial but smaller later Roman building, the so-called Merchant’s House (although any commercial use remains ambiguous). It was not possible to excavate this complex fully, but one part had a lower floor decorated with fragments of marble veneer. At the end of the 5th century it was incorporated into the structure of the new city wall, which nonetheless allowed the owners of the property to maintain their access to the channel, which was facilitated by a small gate in the wall. Occupation of

the building actually ceased at about this time, but in the ground separating it from the city wall and channel, there was evidence of occupation until the later 6th century. Thereafter, this channel-side area was partially covered in crushed mussel shells, indicating fishing activity, probably throughout the Middle Byzantine period. A cobbled surface or road, and a number of buildings of varying construction, suggest a short-lived bid to reoccupy this area in the 11th century, as occurred in the adjacent Triconch Palace area. Thereafter, following the closure of the gate to the channel in the 13th century and the construction of a tower at the start of the 14th century, like the Triconch Palace area, this ground was maintained primarily as gardens or agricultural land for nearby properties whose presence is evidenced by quantities of 13th- to 15th-century ceramics.

Methodology and this report The first phase of the Triconch Palace excavations took place over four seasons between 1994 and 1999. Owing to the logistical difficulties of working in Albania during this period, this work was mainly an assessment exercise that involved sampling the archaeological assemblage in different parts of the site. Although relatively limited in scale this work succeeded in both producing major excavated assemblages and defining the basis of the occupation sequence at the site. These excavations are described in some detail in Gilkes and Lako 2004, although some elements will be repeated here in the context of this final report.



William Bowden and Richard Hodges

Figure 1.5. The Triconch Palace excavations in progress during the last season of excavation in 2003 From 2000–2003 major open area excavations were carried out, focused on the southern and eastern parts of the site (Fig. 1.5). These succeeded in clarifying and refining the sequence provided by the 1994–99 work and allowed us to understand the development of the Triconch Palace as a building and the contemporaneous use of the Merchant’s House area to the west. A complex occupation sequence dating to the period following the abandonment of the Triconch as a grandiose residence was also recovered, while in the Merchant’s House area the sequence provided reliable evidence for the late 5th-century construction of the city wall, together with evidence of a sequence of structures that extended to the end of the 6th century. Both the Triconch and Merchant’s House areas also produced complex sequences of intermittent occupation from the 10th to the 15th centuries. In consequence the eight major phases identified in the 1994–99 report have been expanded to fifteen (Tables 1.1 and 1.2). In an effort to render this final report more readily comprehensible, phases have been limited to those that denote substantive changes in layout and function. Smaller structural alterations have been accommodated within sub-phases where necessary. Although the upper levels of walls often collapsed or were demolished it is apparent that foundations or the lower parts of walls would have remained visible and were often subsequently reused. Consequently the plans show all structures that were not

clearly buried by later deposits, although in some cases they may have only stood to just above foundation level. This report attempts to present the evidence in a way that is sufficiently comprehensive to allow readers to formulate their own interpretation if they so wish. However, the excavations of the Triconch Palace and the adjacent Merchant’s House area produced more than 3000 individual context records and consequently a full description of the archaeological sequence would be both impractical and indigestible. This report will therefore take the form of a narrative description of the phases of construction and occupation on the site, with descriptions of individual contexts and relationships limited as far as is possible to those that are pivotal for the definition and understanding of these phases. This volume is limited to descriptions and discussion of the archaeological and structural sequences. The primary record is housed on the Integrated Archaeological Database system developed by the York Archaeological Trust. Full reports on the finds will appear in subsequent volumes. A common problem of archaeological reports is that interpretations and sequences that were tentative or ambiguous at the moment of excavation become more unequivocal during the process of writing the results in narrative form. At the Triconch Palace excavations, very complex depositional sequences were encountered and although we are confident that the analysis of the site

1.  Introduction



Table 1.1. 1994–1999 phasing 1994–1999 phasing

Phase 1

3rd century AD and earlier: buildings before the domus.

2

4th century: the domus.

3

Early 5th century: the Triconch Palace.

4

Mid to late 5th century: the city wall.

5

Late 5th century: later occupation.

6

Early to mid 6th century (c. 525–575): burials, demolition and rubbish tipping.

7

13th century: medieval reoccupation.

8

14th to 19th centuries: later medieval and post-medieval deposition.

Table 1.2. Revised phasing following 2000–2003 excavations

1: 3rd c. and earlier (c. AD 100–250) 2: 3rd to 4th c. 3: early 5th c. (c. AD 400) 4: early 5th c. (c. AD 420) 5: early to mid 5th c. (c. AD 420–450)

Triconch Palace Earliest occupation The early domus

Merchant’s House

9: mid to late 6th c.

The peristyle domus The Triconch Palace Post-built structures and industrial/domestic occupation. Reduced dwelling in west wing Construction of city wall; possible flooding and other occupation Reoccupation of southern wing and roof collapses in west wing Demolition and dumping; extensive use by fishermen; beginning of burials Industrial activity in some rooms

10: late 6th to 7th c.

Demolition and burial

11: mid 7th to early 10th c. 12: early 10th to late 10th c.

Minimal occupation/activity Stone and post-built structures and quite intensive activity Continued deposition in courtyard; further post- and stone-built structures

6: mid to late 5th c. 7: early 6th c. 8: mid 6th c. (c. AD 525–550)

13: early 11th to late 12th c.

14: early 13th to 14th c.

15: 15th c. and later

Renewed burial; little defined activity in Triconch area; quantities of material indicating nearby activity No discernable activity in Triconch area

is reliable and robust as a whole, it would be dishonest to suggest that no ambiguities are present within the interpretation presented here. Consequently, we have attempted where possible to provide some indication of the degree of confidence with which elements of the sequence have been identified, together with offering alternative explanations and possibilities where appropriate. The workforce was composed primarily of experienced excavators from UK and Ireland-based professional rescue archaeology units (who volunteered their services to the project), Albanian student volunteers (previously

Earliest excavated parts of Merchant’s House Continued modification

Extension of west wing and addition of upper storey Construction of city wall and abandonment and demolition of parts of complex Mussel processing Construction of two-storey building next to city wall Abandonment of two-storey building; continued mussel processing Mussel processing Mussel processing a) Construction of two-storey building and cobbled surface b) Construction of tower and new timber buildings Repair of tower and insertion of burials; blocking of gate; agricultural activity Agricultural activity and dumping of refuse

trained within the Butrint training excavation) and local workmen. The latter were generally used only for removal of vegetation and topsoil, shovelling spoil and moving wheelbarrows. This is obviously a departure from many excavations in the Mediterranean where, through legal requirements or tradition, most excavation is carried out by labourers, and the detailed nature of the sequence recovered reflects the use of professional archaeologists accustomed to working with complex stratified sequences. One final point: from the outset it was envisaged that the Triconch Palace area would be conserved and



William Bowden and Richard Hodges presented as a ‘new’ (i.e. post-Ugolini) monument within the archaeological park at Butrint (Martin 2001). Being beside the Vivari Channel fortifications, it offers a natural midway point for an alternative trail to the path through Butrint, which begins at the Venetian Tower and ends at the Water Gate close to the Great Basilica. Richard Andrews created a design for the conservation and presentation of the Triconch Palace area in 2001. These were adopted by the Butrint National Park, and the work was completed in July 2005 under the overall direction of Daniel Renton assisted by Jerry O’Dwyer and Albana Hakani (Fig. 1.6). Agron Islami (Institute of Monuments, Tirana) aided by Elda Omari (University of Padua) was responsible for consolidating all the mosaic pavements here (Fig. 1.7).

Notes 1

2

Figure 1.6. The Triconch Palace following the conservation programme

The evidence for this building having a commercial function remains ambiguous. However, the name ‘Merchant’s House’ stuck from an early stage of the excavation, and as such is present throughout the primary archive. Although we are aware of the problems caused by the persuasive power of such names, we have decided to continue using it in order to avoid having to eliminate it from the primary record. Equally the use of more neutral terms such as “Plot 1 structures” risks rendering the report more indigestible than it is already. For the 1994–2003 excavations see Bowden et al. 2000; Bowden, Hodges and Lako 2002; Bowden and Mitchell 2007; Hodges, Bowden and Lako 2004; Gilkes et al. 2002; Gilkes and Lako 2004; Hodges et al. 1997. For earlier work and interpretation see Anamali 1993: 470; Lako 1990; Meksi 1988: 207–8; Ugolini 1937: 176.

Figure 1.7. Cleaning of the mosaics in progress prior to consolidation

THE EXCAVATIONS

2  The domus and the Triconch Palace William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

Introduction This chapter details the excavations of the Roman and late Roman dwellings in the central and eastern areas of the Triconch Palace site (plots 2 and 3), from the earliest occupation phases noted on the site, until the last and most grandiose of the elite dwellings on the site: the Triconch Palace itself and its abandonment as an elite residence shortly after its construction in the early 5th century. The complex late antique levels of the 5th to 7th centuries are examined in the following chapter. In an attempt to make a large body of data more accessible, detailed descriptions of the structures and archaeology of each phase are concluded with short summary and interpretation sections.

Phase 1: 3rd century and earlier (c. AD 100–250) The earliest occupation (Fig. 2.1) The area of the triconch was occupied from at least the 2nd century AD and probably earlier, although the high ground water meant that it was only possible to reach deposits of this date on the west side of the site. Owing to the fragmentary nature of these earliest structural remains, they have been grouped together under one phase, although it is clear that not all were erected as part of one contemporaneous construction programme. Nonetheless, the evidence suggests that from around the 2nd century onwards the central area of the site was occupied by a unified complex of buildings that may well have been arranged around a central garden or courtyard. The most significant assemblages were recovered from excavations within Rooms 29 and 31, where a major 2ndcentury deposit (3394) overlay a small structure represented by walls 5662 and 3392 (Fig. 2.2). This structure abutted wall 1282 (below the level of the later stylobate 1356). Wall 1282 in turn abuts wall 1355. Fragmentary traces of a further late 2nd- to early 3rd-century building, represented by walls 5525 and 5657, were also found, with dating provided by a late 2nd-century deposit (5645) beneath

the walls, and a deposit (5523) dating to 200–250 that lay above them. Wall 1282 seems to have delineated the western extent of a major complex. Only one entrance on this side (later blocked by 1835) can now be seen (Fig. 2.3), although much of 1282 was truncated by the insertion of the later stylobate 1356. Of particular interest in this area was the discovery of a trackway (5554), which ran between the major structure represented by walls 1355 and 1282, and that which was represented by walls 5525 and 5567, before curving to avoid the building comprising walls 5662 and 3392 (Figs 2.2 and 2.4). This trackway took the form of a wide linear depression (reminiscent of a ‘hollow way’) and showed no sign of a surface surviving from its earliest phases, although a rough mortar surface (5545) may have been laid subsequently. The earlier phases of the trackway date to the first half of the 3rd century (dating provided by underlying deposit 5523 (AD 200–250) and 5546 (AD 240–250) which filled much of the depression) (Fig. 2.5). As noted above, the deposit beneath the trackway (5523) overlay the earlier structure represented by walls 5525 and 5657 and so the track clearly post-dated this structure. Traces of earlier structures were also detected below the later mosaic in the long gallery (Room 18). It is clear that, as at the nearby villa of Diaporit, buildings underwent frequent alteration during this period, with many minor changes occurring over a very short space of time. On the eastern side of the central plot, there is also evidence for earlier structures, although in this case the early date must be extrapolated from wall relationships rather than stratified materials. Wall 1058/1062/1082/1083 (hereafter referred to as 1058) is thought to be contemporary with the 2nd-century wall 1282 (mentioned above) on the basis of its position and the fact that (like 1282) it predates the walls of Rooms 19–23 (see below). Both 1282 and 1058 share a similar distinctive construction technique (reminiscent of that of the scaenae frons of the theatre) featuring corners (1976 and 1083) made from well cut limestone blocks separated by mortar courses with horizontal and vertical scoring (Fig. 2.6). They are

1314 1316

1388

1314 1350 1306

1350

1352

13521353

5614

5610 1353

Trackway

5610

5554

55835582

1355

5525

Trackway

5662

5554

1010

5580

1939

5589

5584 5581

5580 5589

1282

5581

5662 3392

5584

5582

3392

1164

1939 5583

13555614

5525 5657

1388

1010

1169

1350 1306

1350

5657

1316 1169

1164

1282

1058 1282 1058 1282

1976

1082 5000

1083

1976

1082 0 5000

Figure 2.1. Phase 1 (3rd century and earlier) showing features discussed in text 0

1353

5657

1353

5657 5657

1353

1355

5525

1355 5525

1355

5525

Trackway 5554

Trackway Trackway 55545554

5662

5662

5662

3392

1282

3392

3392

0

5m 0 0

20m

1083

1282

20m

linked by a long southern wall (5000), which seemingly underlies walls 1976 and 1083 (which form the corners of walls 1282 and 1058). Wall 5589 (the continuation of 1058) abuts a structure comprising walls 5580, 5581, 5582, 5583, 5584 and 1939 (Fig. 2.7). One element of this building (5584) is constructed using the distinctive opus mixtum noted in late 1st- and 2nd-century buildings at Diaporit in which diagonally cut bricks are used with the long diagonals forming the external face of the wall, thereby economising on the number of bricks required. Wall 1939 continued as wall 5610 to form the south wall of a possible northern range of buildings. It is noteworthy that there is only one entrance on this eastern side of the complex (see below for further discussion in relation to the phase 2 house). Equally the angle of the eastern boundary walls suggests that they respect a pre-existing land division, perhaps relating to another property to the east (on the area designated plot 3), or alternatively to a road or alley leading to the channel

1282

5m

5m

Figure 2.2. Detailed plan of western end (walls 1282 and rooms 29 and 31) showing phase 1 (3rd century and earlier) features

2.  The domus and the Triconch Palace

South

1381

13

North

1282

1282

1835 0

tile

1m

Figure 2.3. Elevation of wall 1282 showing blocking 1835

Figure 2.5. Section through 3rd-century trackway showing deposits noted in text (1 m scale)

Figure 2.4. View of 3rd-century trackway beneath Rooms 29 and 31, looking north (1 m scale) side (see Fig. 2.1). The presence of a major city drain on this alignment (see Fig. 2.49) (a drain that also appears in the so-called Gymnasium to the north) suggests that the second hypothesis is more likely, although no clear sign of a road was detected during excavation. A further series of early walls (1169, 1350, 1306) are present on the northeast side of the site, which were overlain by deposits of the mid 3rd to 4th century. These are discussed in detail by Gilkes and Lako (2004, 154–6) and are shown here on Fig. 2.1. In the earlier report it was only possible to say that these features dated to the 3rd century or earlier. However, the dating of the structures on the west

Figure 2.6. Wall 1058 (3rd century or earlier) showing distinctive construction of rectangular blocks and horizontal and vertical scoring in the pointing

of the site to the 2nd century or earlier (which the coherence of the building plan and wall relationships imply are almost certainly contemporary with those to the northeast) suggests that this dating should be revised downwards. Also of note are a series of piers (1010, 1164, 1314, 1316, 1388) that may form part of an arcaded structure, perhaps the façade of a separate complex (see Gilkes and Lako 2004, 154–6 for detailed discussion of the associated stratigraphy).

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

14

1939 1939

5583 5583 5584 5584

5582 5582

5580 5580

5589 5589

out over a garden immediately adjacent to the Vivari Channel. At the same time it is likely that a reception room (24) was added to the eastern end of this gallery. The southern range of rooms (19–22) also dates to this phase, and seems to form a part of the house to which visual access was more restricted. There is no direct stratified dating evidence relating to this phase, as water levels meant that it was not possible to reach foundation levels for any of the walls concerned. All the dates are based on the relationships between the structural elements of this phase and those of earlier and later phases that are associated with dated deposits. Some broad indication of date is also provided by stylistic analysis of mosaics and wall painting.

Room 18: the long gallery The gallery (18) is composed of the long phase 1 southern wall (5000), which abuts walls 1976 and 1083 (which 5581 5581 form the corners of walls 1282 and 1058). The insertion of a northern wall (1368), which clearly abuts the earlier structures to the west (1282) and east (1058), created a long 0 5m 0 5m tile tile corridor-like space.1 While the elements of this northern wall are more difficult to interpret, there is no doubt that Figure 2.7. Structure comprising walls 1939 and 5580–5584, the gallery postdates the 2nd-century buildings described in which predates the structure represented by 1058/5589 phase 1, and formed a unified structure that acted as a focal point within the 3rd- to 4th-century complex (Fig. 2.9). The gallery was floored with two distinctive and Summary/interpretation finely executed mosaic pavements (see Chapter 7). That It is probable that by the 2nd century, if not before, to the eastern end was a small square pavement, lying the central part of the triconch area was occupied by a immediately in front of the principal entrance to the gallery, substantial structure that may have constituted a single and was one of the finest Roman pavements known from property covering the same area as the later (phase 3) Butrint. It depicted a series of theatrical masks set in four peristyle domus of the late 4th or early 5th century. Only square panels separated by complex geometric borders. The structures on the west side of the complex have been pavement was composed of extremely small, closely set confirmed as dating to the 2nd century, however. It is tesserae. A pair of pilasters and a threshold separated this likely also that this building was arranged around a central area from the main part of the gallery, which was paved courtyard or garden in the space later occupied by the with a single unified geometric design more than 25 m phase 3 peristyle. in length. Although the tesserae used within this larger The complex was probably flanked by roads to the east pavement were slightly bigger than those of the smaller and west; that to the west is indicated by the ‘hollow way’, eastern pavement, the overall effect would have been while the presence of the eastern road is suggested by the extremely impressive. alignment of the eastern boundary walls and the presence Traces of painted decoration surviving at the eastern of a major city drain. Entrances to the complex seemingly end of the gallery indicate that its north wall and part existed from both these roads. The presence of the 3rdof the southern wall were decorated with a scheme that century ‘hollow way’ also provides an intriguing glimpse illusionistically depicted a full-scale colonnade rising above into the appearance of this area of Butrint in which roads a brightly panelled skirting stylobate-wall. It is likely that were not paved or metalled but instead must have been little this scheme (described in Chapter 7) replicated a genuine more than muddy tracks (not dissimilar to those found in colonnade or arcade that allowed a view to the south over many modern Albanian villages). a garden and to the Vivari Channel beyond. A series of flat stones that remain visible towards the western end of wall 5000 form a threshold between the gallery and the area to Phase 2: 3rd to 4th century the south, suggesting that wall 5000 acted as a stylobate The buildings before the peristyle domus (Fig. 2.8) for this arcade or colonnade (Figs 2.10 and 2.11). Thus, it Probably during the later 2nd or 3rd centuries an extensive seems likely the principal purpose of this gallery was to range of buildings was constructed on the south side of the provide a view across the channel. site. The focus of these new buildings was an opulently The eastern area with the most elaborate of the pavements decorated long gallery or corridor (18) that probably looked probably formed an entrance vestibule, with a door (later

1042

1042 Room 27

Room 27 Room 25

Room 25 ?

Room 26

1282 ?

1696 1282

Room 26 Room 23

1568

1568 3040

1599

1599 1381 1378 1381 Room 22

13731375 Room 21 Room 20 1369

1378 1981

1981

1696

Room 23

Room 24

Room 24 1429

1429

3040 5586

1373

Room 22 1368

Room 21 1369 Room 18

1368

1364 5586 Room 19

1364 5077 1375 1372 Room 19 1367 1361 Room 20 1367

1372

5077 1361

1368 1976 1368 1976

Room 18 5000

5000

1083

0

1083

Figure 2.8. Phase 2 (3rd to 4th century) showing wall numbers and room 0 numbers mentioned in text

20m 20m

Figure 2.9. View across excavations from west, with the long gallery (Room 18) on the right of the photograph and the apsidal reception room (24) in the immediate foreground

16

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

Figure 2.10. Flat stones of wall 5000 at the western end of the long gallery (Room 18), which may be the remains of a stylobate for a colonnade separating the gallery from a garden space to the south

Figure 2.12. The vestibule, looking south. The theatrical masks in the mosaic are clearly visible as is the later stairblock in the northeast corner (JBB)

Figure 2.11. The eastern entrance and vestibule (looking west). A much altered doorway can be seen in the eastern wall of the vestibule. The door in the later wall in the immediate foreground may have existed but its present form owes much to reconstruction by the Albanian Institute of Monuments (JBB)

blocked) at the eastern end of the gallery opening onto this space (Figs 2.11 and 2.12). The fact that this was the only eastern entrance in the whole complex explains the presence of such a fine pavement at one end of the gallery. The visitor arriving at the eastern entrance would have been afforded a dramatic view along this opulently decorated space. The gallery has not been precisely dated although the wall relationships at either end of Room 18 indicate that it post-dates the 2nd-century phase 1 buildings, while on stylistic grounds the mosaics and wall paintings are unlikely to date to much later than the 3rd century (see Chapter 7).

Room 24: the reception room At the western end of the long gallery a door allowed access to a large rectangular room with an octagonal fountain (Fig. 2.13). The north wall of this room (1696), although levelled when the room was later extended, abutted the 2nd-century wall 1282, while the southern wall (1429) abutted wall 1976 (the corner of 1282). A short length of wall (1975) was added to the north side of 1976 to create a central door (1.68 m wide) in the west wall of the long gallery.2 The fountain was situated in the southern half of the room and was axially aligned with a large central doorway in the south wall (3.66 mm in width). It is octagonal in plan, 0.84 m in width and originally 0. 27 m deep. In a second phase the depth was reduced to 0.15 m. The sides of the basin and the original floor are revetted with slabs of polished Proconnesian marble, still in position. The opening had a 0.26 m wide framing surround, which was also set with polished marble slabs. These elements have all been detached and lost, but fragments of pavonazzetto and africano slabs of the same thickness found in the room may derive from this surround (Figs 2.14 and 2.15). Water was probably supplied via lead piping (subsequently robbed) laid in a plaster and tile-lined duct that entered the fountain on its south east side. There is no sign of a drain and it is thus likely that the fountain was also drained via this duct. It is evident that the base of the fountain was relined on one occasion, with a thin layer of

1282 1282

Room 24 Room 24

1696 1696

Room 23 Room 23

1981 1981

1975 1975 1429 1429

Phase 2 Phase 2

Room 18 Room 18

1976 1976 0 0

1281 1281

5m 5m

1282 1282

Room 24 Room 24 Room 23 Room 23

1981 1981

1975 1975 tile tile Phase 2b Phase 2b

1998 1979 1429 1998 1979 1429

Room 18 Room 18

1976 1976

0 5m 0 5m Figure 2.13. The two structural phases of Room 24 (both dating to the 3rd to 4th century) with the apse belonging to the later phase (2b)

18

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

Figure 2.14. Overall view of Room 24 (JBB)

Figure 2.15. The fountain in Room 24 (JBB)

2.  The domus and the Triconch Palace

19

Room 26

Room 23

Room 22

Room 21

Room 20 Room 19

Room 24

Room 18

0

10m

Figure 2.16. Plan of the southern range highlighting access routes between different parts of the building

silt (1602) separating the upper surface (1601) from the lower (1603). In its earliest, rectangular, phase, the reception room was paved with a complex geometric mosaic (see Chapter 7). The innermost panel of this mosaic (which may have contained a figurative medallion or a central marble roundel) is slightly offset to the north to accommodate the fountain, creating a rather awkward visual arrangement in the rectangular room. On stylistic grounds this mosaic may date to the later 3rd or 4th century. At a subsequent date wall 1696 was demolished and the room was extended with the addition of an apse (1281) 4.7 m in width (see below phase 2b). Room 24 was almost certainly a major public reception room within the domus of the 3rd and 4th centuries and continued to fulfil this function until the principal entrance of the complex was moved at the end of the 4th century. Its progressive aggrandisement reflects the changing currents of elite housing, an aspect that will be discussed in detail in Chapter 8. Rooms 19–22 The third major element in this 3rd- to 4th-century domus was a range of five rooms that lay to the north of the long gallery described above. The only access to these rooms from the gallery was via a single large door in the centre of the gallery’s northern wall. Although there are other doorways between the gallery and the southern range, all appear to be later insertions and it is only the central doorway that was part of the original layout of the building.

This doorway led into Room 21, the central of the five rooms that linked the gallery to the courtyard or garden space beyond, and also to the rooms to the east. The southern wall of these rooms was formed by wall 1368 (the north wall of the gallery), while the north was formed by walls 1373 and 1381. A series of cross walls (bonded to the northern and southern walls; 1361, 1372, 1364, 1367, 1375, 1369, 1378) formed the divisions between the rooms. The walls were constructed using a crude opus incertum of limestone blocks bonded with a reddish pink mortar of varying shades and consistency. Doorjambs in this phase were constructed with more precision using rectangular dressed limestone blocks. The doors between these rooms appear to have delineated precise means of access between the different areas of the house (Fig. 2.16). Only the central room (21) could be accessed from the long gallery, with a large door in wall 1368 allowing access to Room 21 and a clear view through the room into the area beyond (probably an open courtyard or garden in this phase). From Room 21, as well as gaining access to the central courtyard, it was possible to turn right through a single door into Room 20 which was in turn linked by two doors to Room 19 (which appears to have been sub-divided by wall 5077, with a door allowing access between the northern and southern halves). Both Rooms 19 and 20 could only be entered via Room 21. Neither Rooms 19 nor 20 allowed access into the central courtyard and the remainder of the house beyond. They thus formed a self-contained entity within the house.

20

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

In contrast the rooms to the west (22 and 23) could seemingly only be entered from the central courtyard and from Room 26. Although a narrow door does exist between the central room (21) and that immediately adjacent (22), it lacks the dressed rectangular blocks that seem to characterise the doorjambs of this phase and is therefore likely to be a later insertion.3 Consequently, although the southern range of rooms appears to be more or less identical in terms of size and form, it seems that the rooms to the east served a very different function to those to the west. This will be further examined in Chapter 8. The original floor levels of the rooms of the south wing were not conclusively identified in any of the rooms (and in Rooms 19 and 20 the excavations did not reach these levels), although excavations in Rooms 21 and 22 both suggested that any floors had been largely robbed for building materials in antiquity (Figs 2.17 and 2.18). The lowest level reached in Room 21 was a rough mortar floor (5624) that was revealed in the northeast corner of the room. This may have formed the base of a stone or tile floor that was subsequently robbed, but its level in relation to the walls and rough construction suggests that it is more likely to relate to later phases of occupation. In Room 22, the remains of a stone floor (1950) were found above a clay bedding deposit (1964). Only one complete flagstone (0.94 × 0.47 m) survived in situ (Fig. 2.18; the fact that it was cracked in several places perhaps accounts for the fact that it was not removed in antiquity with the rest of the floor. None of the rooms produced convincing stratified deposits of earlier than the 5th century. The west wing (Rooms 25–27) At the same time that the southern wing was laid out it is probable that a further set of rooms (25–27) was also erected, although the relationships between the constituent walls have been obscured by subsequent major alterations (Fig. 2.8). However, wall 1378 (which divides Rooms 22 and 23) continues to the north (as wall 1599). An east–west return of 1599 formed the north wall of Room 26 (although this was subsequently truncated by the insertion of later stylobate 1292), while 1599 may also have continued northwards to join 1568 (the eastern wall of Room 27). It is unknown whether there was a door between Rooms 25 and 27 in this phase. It is clear that Rooms 25 and 26 were divided by a wall that was subsequently replaced by the later stylobate 1292, with a door linking Rooms 25 and 26. A further door linked Rooms 26 and 23. The west wing of the house was thus formed by a substantial tripartite suite of rooms that faced out onto the central area and which were also linked to the south wing. What is not clear, however, is whether the western external door in wall 1282 was still open in this period. The triconch area The central area of the triconch excavations, which lay between the two roads described in phase 1, was clearly intensively occupied between the 2nd and 4th centuries, as

was the area of the so-called ‘Merchant’s House’ described in Chapter 5. The nature and use of the area occupied by the later triconch triclinium and its associated rooms are less clear, and, indeed, it is possible that this area remained an open space during this period. Although excavation in this area was relatively minimal and it is thus possible that earlier buildings lie at a lower level, several factors point to an absence of buildings in the triconch area. The domus in the central area is clearly an amalgam of numerous building phases and displays a complex structural history. In contrast, the triconch has no earlier elements (with the exception of wall 1068, which forms its eastern boundary) and displays a unity of construction that is entirely absent from the remainder of the complex. It is possible that all earlier structures were simply razed to a level that allowed the builders of the triclinium and its associated rooms to construct anew, but this would be in marked contrast with the central area (and indeed all other examples of mid–late Roman building in the city) where earlier structural elements were reused wherever possible. A single demolished earlier wall (1042) was noted beneath the northern part of the triclinium suggesting that the earlier Italian and Albanian excavations had reached a level that was sufficient to reveal earlier remains had they been present. It is possible that structures in this area were built using materials that were unlikely to be detected by the earlier excavations; a putative foundation of a pisé wall (1119/1008) was noted in Room 10 (Gilkes and Lako 2004, 155–57). Nonetheless, the available evidence indicates that occupation of this area (plot 3) before the construction of the triclinium was significantly different to that in the area covered by the 3rd- and 4th-century domus.

Phase 2b: Alterations to the 3rd- to 4thcentury complex Room 24 The reception room described above was subsequently further aggrandised when wall 1696 was demolished and the room was extended with the addition of an apse (1281) 4.65 m in width (Fig. 2.13). The apse utilised a distinctive construction technique of opus mixtum in which large square bricks (0.30 × 0.30 × 0.03 m) were laid in bands of three courses separated thick mortar courses of up to 40 mm in thickness (Fig. 2.19). The bands of brick coursing acted as string courses between courses of limestone blocks bonded with a thick pink mortar. It is clear from this solid construction that the apse was intended to support a semi-dome. The mosaic was also extended into the apse, with a radiating design around a semi-circular central motif that does not survive (see Chapter 7). The apse was not a perfect semi-circle and the mosaicist was seemingly unable to accommodate a change in design, instead compensating for the extra space with additional bands that sat awkwardly in the available space. Stratigraphically this mosaic obviously post-dates the rectangular mosaic, although there is no other archaeological indication as to its date.

2.  The domus and the Triconch Palace

1950

21

5624

1950

5624

Room 21 Room 22

Room 21

Room 22 0 Figure 2.17. Plan of Room 21 and 22 showing earliest levels reached and floors 5624 and 1950

0

5m 5m

Figure 2.18. Room 22 showing a single flagstone which may be the only surviving trace of the original floor

22

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

Figure 2.19. Room 24 apse masonry (1281) showing use of distinctive brick levelling courses (JBB)

It is possible that the construction of the apse coincided with the narrowing of the wide southern doorway with walls 1979 and 1998, although it is difficult to see how this would accord with the aggrandizement of the room suggested by the apse (Figs 2.13 and 2.20). Consequently this narrowing may be later in date. The addition of the apse was a significant elaboration of the principal reception room of the house, and can be seen in the context of the increasing need for local potentiores to appear within specific architectural settings when meeting guests and clients (see Chapter 8). The west wing (Rooms 25–27) A significant intermediate stage in the west wing of house is evidenced by the insertion of an opus signinum floor (3440) in Rooms 25 and 26. This floor abuts a series of door blockings (1835 in wall 1282, 3431 in wall 1599 and possibly 1383 in wall 1384 that separated Rooms 26 and 23) (Fig. 2.20). Although the floor was only revealed in Room 26 (Fig. 2.21), it clearly extended under the later stylobate (1292), which formed the threshold to Room 25, and under the mosaic pavement of Room 25. Traces of painted wall plaster were found on the walls (including on the face of the blocking 3431) (Fig. 2.21), while later demolition deposits (3476) also contained quantities of red, pink and blue painted plaster (see below). These alterations signal significant changes in access into and within the building. The only visible western entrance into the complex was blocked, as was the access between Room 26 and Room 23. A door to the courtyard

was also blocked, although access to the courtyard was presumably maintained through Room 25. The precise date of these alterations, together with the addition of the apse in Room 24, is unknown. They may be broadly contemporary with one another (the masonry of the apse is of a very similar technique to that of the door blocking 1835), but all that can be said for certain is that they post-date the addition of the long gallery and Rooms 19–23 but predate the creation of the colonnaded western entrance and mosaic in Room 25. This last is important in that it provides a further indication that the colonnaded western entrance and the associated central peristyle reflect a major subsequent reorientation of the house (see below). Summary/interpretation It seems that during the 3rd to 4th century the central area was occupied by a reasonably substantial townhouse, the origins of which lay in the phase 1 building of the 2nd and 3rd centuries. Unlike the earlier building, however, sufficient remains of the 3rd and 4th century house to allow some judgements to be made regarding its spatial and functional arrangement. The house appears to have been laid out around a central open space, perhaps a garden or courtyard. There is no trace of a peristyle relating to this phase. The principal public entrance to the house was on the east side, where a small door allowed access to an elaborately decorated vestibule. This vestibule was paved with the finest mosaic found in the complex, while the walls were decorated with

2.  The domus and the Triconch Palace

23

Room 27 1568

Room 25

1599 1835 1282

Room 26

3431

1383

Room 23

Room 24

Room 18

1998

1979 0

5m

Figure 2.20. Plan of the west wing showing door blockings 1835, 3431 and 1383 an illusionistic colonnade. The visitor arriving at the door (or passing it) would be afforded an impressive view down the length of a long corridor, or gallery, also paved with a fine mosaic (see Hales 2003: 107 and Chapter 8 in this volume). It is likely that the painted colonnade on the north side of the gallery reflected a genuine colonnade or arcade on the southern side that allowed a view over a garden and out to the Vivari Channel beyond. A visitor could then proceed along this gallery to the principal reception room of the house. This room was originally rectangular in plan but was subsequently

embellished with an apse. Both the rectangular and apsidal phases were decorated with mainly geometric mosaic pavements. An octagonal fountain was an early feature and was respected by the earlier phase of mosaics. The view along the gallery also allowed a partial view of this fountain through the east door of the reception room. This richly decorated part of the building formed an explicitly public area, in which much of the owner’s dayto-day business would have been carried out. The most opulent decoration of the complex (at least that we know of) was concentrated in this area, with the use of painted

Elevation

24 1528

North

South

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

1599 3431, showing layers 3355 and 3476, Room 26 West-facing elevation of wall 1599 and blocking 3431 1599

1599

1292

1383

Elevation

3355

3355 North

3476 1599

1528

1599

3431

3476 Plaster

South

Plaster 1599

3440

1292

1383 3355

3355 3476

3476 Plaster

Plaster 3440

Interpretation North

later mosaic floor

early wall stylobate wall

Interpretation

South early wall

early wall southern wall

late 4th- to 5th-century demolition rubble North

later mosaic floor

blocking

early wall stylobate wall

late 4th- to 5th-century demolition rubble early wall

blocking

Plaster

South

Plaster

early wall opus signinum floor

late 4th- to 5th-century demolition rubble

0

southern wall

1m

late 4thto 5th-century demolition Plaster 3431 and later stylobate Figure 2.21. West-facing elevation showing opus signinum floorrubble 3440 in relation to blocking Plaster 1292 opus signinum floor

colonnades (and probably genuine colonnades) recalling public architecture and hence the owner’s role in civic life (Hales 2003, 127–29; also Wallace Hadrill 1994, 17–37). There was a clear separation between this gallery and the reception room and the remainder of the house to the north. While proceeding along the gallery towards the reception room, the visitor could glimpse the remainder of the house through a large door in the northern wall of the gallery. This door (which was the only link between the ‘public’ area and the remainder of the house) allowed a view through Room 21 (the central room of the southern wing of the house) through to the central garden or courtyard. The southern range of rooms offered an intermediate stage within the hierarchy of space created between the public eastern entrance and the more private central courtyard in the form of the two eastern rooms, 19 and 20. These could only be entered from Room 21 and did not allow access into the courtyard. By contrast, the two western rooms of the southern wing (22 and 23) do not appear to have been accessible from Room 21, but were instead only linked to the central courtyard and the west wing. Relatively little is known of the west wing of the building, but it seems that Rooms 25 and 26 were originally part of a range of rooms that were accessible from both the courtyard and Room 23. It is also possible that a western entrance gave access from the ‘hollow way’ road that skirted the western side of the complex. All the visible doorways in this room were subsequently blocked during phase 2b prior to the addition of an opus signinum floor.

0

1m

Phase 3: Early 5th century (c. AD 400) The peristyle domus (Fig. 2.22) Probably around AD 400, the domus of the 3rd and 4th centuries underwent a major reconstruction that moved the focus of the house to the north, and effectively transformed its orientation and the way in which it functioned as an elite dwelling. The principal entrance of the house was moved from the eastern end of the long gallery to the western side of the house, where it was axially aligned with a new peristyle courtyard that was erected within the central area of the house. This coincided with an apparent reduction in the importance accorded to the long gallery and its associated apsidal reception room. As with the preceding phase, absolute dating evidence is limited but a variety of evidence indicates a date around the start of the 5th century for this transformation. The western entrance (Rooms 25–27) The creation of the major western entrance into the domus involved substantially raising the floor level in Room 25/26 (an alteration that also seems to have been carried out in the central courtyard area (see below). A large spread of grey mortar-rich clay (3476) was found over the opus signinum floor of Room 25/26, with a greater depth apparent against the walls (Fig. 2.21). The clay contained quantities of green, pink, blue and red painted wall plaster together with fragments of marble paving or veneers, and it is possible that it represents a temporary abandonment of the building during which sections of plaster collapsed from the surrounding walls (which themselves may have

2.  The domus and the Triconch Palace

25

3222

Room 27 1293

5582 1356

Room 25

5581

1831

3040

5587 5585 1292 Room 26

1898

5213

5586

1373

Room 19 1981

Room 23 5241 Room 24

Room 21

Room 20

Room 22

Room 18

1998 1979

1083

Room 28

Room 36 1994

1795

0

20m

Figure 2.22. Phase 3 (c. AD 400) showing wall numbers and room numbers discussed in the text been partly made from clay). Similar levels were found in Room 27, where deposits 5600 and 5607 also contained quantities of painted plaster. In Room 26, 3476 was overlain by a thick layer of demolition rubble (3355) (Figs 2.21, 2.23 and 2.28), which was concentrated in the middle of the room, and may represent a deliberate attempt to raise the floor level. Part of the material for this levelling may have been derived from the partial demolition of the walls that separated Rooms 26, 25 and 27, and the partial demolition of wall 1282 that divided Room 25 from the area to the west. The removal of these walls allowed the transformation of Room 25 into a colonnaded vestibule that gave access to the house from the west (Figs 2.24, 2.25). Pairs of elegant narrow columns created tripartite entrances on all sides of Room 25 (walls 1292, 1293, 1356 and 1831), which led to Room 23 to the south, Room 27 to the north and the new peristyle to the east (see below). The columns, which were only 0.18–0.20 m in diameter, were seated on small bases, which in turn rested on well-cut limestone thresholds. The seating for the column bases survives in situ (as do two of the bases

themselves) giving intercolumnar spaces of c. 1.5 m. It is likely that there was an additional room or courtyard between the colonnaded entrance and the street to the west in order to restrict access to the building. However, no trace has been found of this additional structure. The only pavement that survived in the west wing from this phase was a geometric mosaic (1528) that was found in Room 25 (see Chapter 7). This was laid on the raft of limestone rubble 3355 (noted above) that extended beneath the stylobate 1292 into Room 26 (see above and Fig. 2.28). On its eastern side a mosaic inscription recorded a name and rank, which presumably was that of the owner of the house and was intended to inform visitors as to whose house it was that they were entering (Figs 2.26, 2.27). This was read by Charlotte Roueché as ...]ẠṂαρίῳ τῷ λανπ[ρ]οτά[τῳ|.. suggesting that the owner was of senatorial rank (pers. comm.).4 There is some doubt as to the relationship between the eastern colonnade stylobate (1831) and the mosaic, as the stylobate foundation appears to cut the mosaic, thus truncating the upper half of the inscription. This may, however, be due to the mosaic subsiding (thus leaving

Room 25 1292

3405 pit section line

Room 26 Rubble layer 3355

0

5m

Figure 2.23. Room 25/26 showing rubble 3355 in relation to pit 3405. See also Fig. 2.24

Figure 2.24. The colonnaded entrance vestibule (Room 25) seen from the west (JBB)

3222

Room 27 1293

1356

Room 25

1831 5587

1292 5213

Room 26

1898

0

5m

Figure 2.25. The west range and west portico, showing positions of mosaic pavements. The square in the west portico shows the location of the sondage described on p. 32

Figure 2.26. The mosaic inscription in the entrance vestibule (Room 25), apparently recording the name of the owner and the fact that he was of senatorial rank (JBB)

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

28

East-facing section through pit 3405 showing its relationship with stylobate 1292, Room 26 East-facing section through pit 3405 showing its relationship with stylobate 1292, Room 26

0

0.50m

Figure 2.27. The surviving letters of the mosaic inscription

Elevation South Elevation

1292

1528

3414

South

3414 3355

3406

3405

3406

3405

3355

1292

3440 3440

1528

North North

0

1m

0

1m

Interpretation South Interpretation 5th-century

occupation level South 5th-century 4th- to 5thoccupation level

century levelling 4th- to 5thcentury levelling

fill

pit cut

fill

pit cut cocciopesto floor

North

stylobate wall

mosaic floor

stylobate wall

mosaic floor

North

0

1m

cocciopesto floor

Figure 2.28. Section of pit 3405, showing its relationship with the south0stylobate of Room 25 1m (1292)

the foundation proud of the pavement) or the stylobate being repaired and thus cutting the pavement. Certainly, the mosaic could not realistically predate the creation of the colonnaded room. No other floor surfaces associated with this phase were found in the west wing, and 5th-century hearths, which apparently post-date the use of the west wing as part of an elite dwelling, were found directly over levels that predate the insertion of the colonnaded entrance (see Chapter 4, phase 5). The mosaic pavement was c. 0.50 m above the level of the opus signinum floor that preceded it (Fig. 2.21).

Dating evidence for the creation of this western entrance is extremely poor. In the west wing itself, dating is limited to a small assemblage of 5th-century pottery found in the fill (3406) of a small pit (3405) that cut rubble layer 3355 in Room 26 but which clearly underlay the stylobate of the southern colonnade (1292) (Figs 2.23 and 2.28). In Room 27 the plaster-rich deposit 5607 (which may equate with 3476 in Room 26 and thus also predate the colonnade) contained pottery probably dating to the 4th century). Other dating evidence from elsewhere in the site suggests, however, that this major alteration does date to the very first years of the 5th century (c. AD 400) (see below).

2.  The domus and the Triconch Palace

Figure 2.29. The small peristyle of the domus of phase 3 (c. AD 400) seen from the north. The south and west edges of the courtyard are obscured by the stylobates of the much larger peristyle of the Triconch Palace

29

Figure 2.30. Detail showing the masonry construction of the columns on the small peristyle (30 cm scale)

Figure 2.31. The well or cistern head of the phase 3 peristyle (looking southwest)

Figure 2.32. Wall 5585, which formed part of the rear wall of the eastern portico of the small peristyle of phase 3 (c. AD 400), butting wall 1373 (the north wall of the south wing). This clearly demonstrates that the peristyle of phase 3 post-dates the south wing of the domus

The Peristyle Courtyard (Room 16) The earliest peristyle courtyard for which evidence survives is contemporary with the western entrance described above. In its centre was a rectangular courtyard some 9.5 m long and 8 m wide, delineated by a low stylobate wall (3222/5587) built of limestone rubble that survives on its northern and eastern sides (the south and west sides being largely obliterated by the later stylobate wall of the much larger peristyle of phase 4) (Fig. 2.29). On the eastern side, the lowest courses of two masonry columns 0.47 m in diameter can be seen. These are constructed using bricks or tiles cut into a triangular form (Fig. 2.30). The two columns are separated by a distance of 2.5 m, and this intercolumnation and construction can probably be extrapolated for the rest of the colonnade that surrounded this central courtyard. The centre of the courtyard was paved with large irregularly shaped flat limestone slabs, set into a clay base. A well or cistern head was slightly offset to the south of

the centre of the courtyard (Fig. 2.31). This was solidly constructed of mortared limestone masonry, and was 1.69 m in diameter with an internal opening of 0.67 m. When found it had been demolished to pavement level, although it is likely that it had a more elaborate superstructure that had been robbed in antiquity. It is not clear whether it was a well or the head of a cistern supplied by the run-off from the building’s roof. The proximity of the site to the brackish waters of the Vivari Channel suggests that it is unlikely that a well in this position could have supplied drinking water. The courtyard was surrounded by a quadriporticus. On the north, south and west sides the roofs would have been supported by the existing structures. However, on the east side, the construction of the portico required the addition of a short length of wall (5585) to support the roof of the eastern portico and to create a coherent peristyle courtyard, with a range of buildings on the east side (see below). This wall clearly abuts the earlier structures to the north (5581/5582) and south (1373) (Fig. 2.32). Wall 1373 is the

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

30

0

0 0

Existing floors & walls

5m

Existing floors & walls

5m

Existing floors & walls

5m

0

20m

0 in relation to the pavement in 20m Figure 2.33. Cross-section showing height of portico pavement the early long gallery (Room 18) (a difference of some 0.44–0.50 m) 20m 0

north wall of the south wing of the complex. The existing central courtyard cannot have existed without wall 5585 to support its roof and consequently the courtyard must post-date the southern wing. The porticoes were paved with well-executed geometric mosaics incorporating figures of birds, repetitive motifs and magical symbols such as knots and the evil eye intended to ward off misfortune (see Chapter 7). These survive in a partial condition in the western portico (1830), while a small section extends into the northern portico. The mosaic is bedded on a layer of pinkish grey mortar, which

survives over a slightly larger area than the mosaic itself. It is notable that, like the floor in the western entrance, the portico mosaics are at a higher level than the known floors of the earlier phase, with the mosaic in the long gallery being some 0.44–0.50 m lower than that of the portico (Figs 2.33 and 2.34). Much of the dating and interpretation of this phase rests on the relationship between this mosaic and a long drain (1966/3402/3272/5213). This drain runs south from the flagstone courtyard, beneath the western portico mosaic, through the doorway to Room 22 and finally across the

2.  The domus and the Triconch Palace

31

Figure 2.34. View of the excavations in early July 2003 showing groundwater in the long gallery (Room 18), while the higher peristyle remains dry

Figure 2.35. The drain (5213) running from the paved courtyard through the door into Room 22. This door was then blocked prior to the laying of the mosaic in the portico. The drain then cuts the mosaic in the long gallery, demonstrating that the small peristyle post-dates the long gallery

Figure 2.36. Drain 5213 in the south portico of the phase 3 (c. AD 400) peristyle

32

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

Figure 2.37. Foundations of the eastern range of the phase 3 and earlier domus, levelled during the construction of the Triconch Palace of phase 4 (c. AD 420) long gallery (18) where it apparently cuts a series of mortar layers above the 3rd-century mosaic (Figs 2.35, 2.36). The drain ran beneath a blocking wall (1898) that closed the door between the portico and Room 22 (suggesting that access to the southern wing became more restricted). This blocking wall (removed during the course of the excavation) was also clearly abutted by the pinkish grey mortar bedding of the mosaic (Fig. 2.36). It is likely that the insertion of the drain (contemporary with the construction of the peristyle), the blocking of the doorway, and the laying of the portico mosaic took place as consecutive actions as part of the same building phase. The portico mosaic hence clearly post-dates both the blocking and the drain, and thus belongs to a later phase than the mosaic in the long gallery and the south wing as a whole. Other key dating evidence for this mosaic derives from a small sondage excavated through a gap in the mosaic in the west portico (see Fig. 2.25) into the lower makeup (10005) or statumen for this floor. This recovered a stamped base of an ARS dish, such as ARS 67, datable to between c. 360 and the early 5th century, a fragment of a cooking vessel of later 4th- to 5th-century type, a large fragment of marble furniture with promegranates carved in flat relief (possibly from the display front of a dining stibadium) probably work of the mid 4th century, as well as vessel glass and fragments of painted plaster.

The east and north wings Very little is known of the east and north wings of the peristyle domus as both were razed to their foundations to allow the construction of the much larger peristyle of the later triconch phase. The plan of the north wing is entirely unknown as this area remains unexcavated and the northern stylobate of the triconch phase peristyle removed all traces of doors into this northern wing. However, something of the east wing can be discerned from the foundations revealed during the excavation of the courtyard (Figs 2.22 and 2.37). The insertion of wall 5585 to support the roof of the eastern portico created a slightly awkward L-shaped room, assuming that walls 3040 and 5586 were left intact. It was entered via two doors from the eastern portico. It seems clear that the principal requirement was the creation of the peristyle and that any resulting awkward spaces were accommodated within the design, which retained the earlier rooms of the east wing as far as possible. The south wing (Rooms 18–24) The fact that the drain (5213) that led from the peristyle cuts the mosaic in the long gallery (18) indicates that this space was no longer conceived of as part of the prestige ‘public’ area of the house. A rough flagstone drain cover bisected the elegant pavement, which was seemingly covered with a layer of mortar (1771, 5158) by this period (Fig. 2.35). This

2.  The domus and the Triconch Palace

33

Figure 2.38. Changes in access to Room 22, probably associated with the phase 3 rebuilding of the domus (c. AD 400) mortar appears to be an attempt to level out undulations in the mosaic. It lipped up against the remains of the painted plaster in the eastern end of the room and was clearly never intended to be anything more than a rough floor covering. The mortar surface (5624) identified in Room 21 (see above) may also belong to this phase. Room 22 was also no longer directly accessible from the peristyle in this period (due to the insertion of blocking wall 1898), but a door was then cut through wall 1369 that allowed access to Room 22 from the central room of the south wing (Room 21) (Fig. 2.38). A door was also cut (through walls 1383/1384) between Room 23 and Room 26 reopening the previously blocked access between Room 23 and the west wing, an action made necessary by the blocking of the door between Rooms 22 and 23 with wall 5241 (Figs 2.39, 2.40).5 Access was also maintained between the peristyle area and the gallery via Room 21 where the principal doors to the north and south were left open at this stage.6 There was no direct access from the eastern wing of the domus to Rooms 19 or 20, while doorways between Rooms 19, 20 and 21 were all maintained during this period. With the exception of some mortar surfaces (1694) (Room 24) 1771, 5158, 5212 (Room 18)) laid over the mosaics in Rooms 18 and 24 (Fig. 2.41), no pavements were found in any of the rooms of the southern wing, and it is difficult to ascertain how they functioned in this period. Certainly, Room 22 is unlikely to have been intended as a public room as it was now bisected by a large drain. Although this drain may have been covered there was clearly some attempt to restrict access to the room, by blocking the entrance from the peristyle area and from Room 23. The bath-house (Rooms 28 and 36) (Figs 2.42, 2.43) The small bath-house that lies on the southwest corner of

Figure 2.39. View looking east along south wing showing Room 23 with door to Room 26 reopened and door to Room 22 blocked

the domus, adjacent to the Merchants’ House area, can also be tentatively placed in this phase. Dating evidence for this building is circumstantial but it must post-date Room 38 in the Merchants’ House area as the bath-building effectively blocks the eastern door into Room 38. The original bath was a small rectangular building

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

34

Figure 2.40. Door blocking 5241. View looking west prior to excavation of Room 23

1694

5212

Room 18

5158

1771

Room 24 0

10m

Figure 2.41. Schematic plan showing mortar floors 1694, 1754 and 1845 (Room 24) and 5158 (Room 18). These may have been intended as preparation for new floors associated with the rebuilt domus of phase 3

formed by walls 1795 and 1994 with internal measurements of 6.75 × 2.80 m. Only half the structure was excavated although the tops of all the walls were revealed and hence the entire plan is known. A door in the northern wall opened into what was probably a small tepidarium or vestibule some 2.20 m in length. A line of substantial rectangular pilae (of which four were revealed, each around 0.75 × 0.30 m) probably supported a partition between this area and

a hotter caldarium with a hot plunge pool at its southern end. The floor of the caldarium was supported by pilae made of circular tiles, while traces of more substantial and closely set pilae adjacent to the praefurnium arch in the southern wall probably indicate the position of the hot plunge pool. An irregular apse or large niche in the west wall of the bath was probably intended to house a statue or other ornament.

2.  The domus and the Triconch Palace

35

1981 1981

Room 36 1994

1795

Room 36

1795

1994

First phase First phase

0

5m

0

5m

Figure 2.43. Bath-house (Room 36). View looking north (2 m scale)

Room 28

1981

Room 28

1981

Room 36 1994

1795

Room 36

1795

1994

Second phase Second phase

0

5m

0 5mand Figure 2.42. The phase 3 domus bath-house (Rooms 28 36) showing the two successive construction phases

The walls of the bath were constructed using irregular mortared limestone rubble and brick/tile, while the pilae were built using bricks and tiles bonded with clay. The mixed nature of the construction materials used suggests that the bricks and tiles were re-used. In a secondary phase, a small octagonal room (Room 28) was added to the north, its walls abutting both wall 1795 and the southwest corner of the reception room (24) (wall 1981) (Fig. 2.44). It was irregular in plan, measuring 4.65 m across, with an entrance on its northeast side close to the main southern door of the reception room. Two opposing sides of the octagon were formed by substantial semi-circular niches, presumably intended for statues. There were clearly two phases of floor in this room. The lowest (1932) was a mortar floor (visible only in section). This was overlain by a thin layer of mortar (1978), perhaps the bedding for a more elaborate floor, followed by a layer of silt (1931) above which was a pavement composed of pieces of blue schist, marble and limestone (1978) set into a mortar base (Fig. 2.45). The interpretation of layer 1931, which separates the earlier and later floor surfaces, raises important questions

36

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

Figure 2.44. Octagonal bath-house vestibule (Room 28). View looking showing Northeast-facing section in doorway towest Room 28 east niche (2 m scale)

Southeast

Northwest

1798 1814

1816 1931

1812

1978 1932 0

1m

Interpretation Southeast

Northwest

pavement

mortar bedding

wall

wall

mortar/bedding

4th-century silt layer

mortar floor 0

1m

Figure 2.45. Section across the door of the octagonal vestibule (Room 28), showing contexts 1932, 1978 and 1931 in bath-house door, beneath the later floor

2.  The domus and the Triconch Palace for understanding and dating the bath-house. The layer contains early 3rd-century pottery, which could suggest that the bath house is relatively early in date. However, the ceramic assemblage from 1931 has elements in common with later deposits 1110/1111 and 1227 on the far north side of the site, including a possible pot join between 1931 and 1111 (Gilkes and Lako 2004, 155–7). This last factor suggests that both 1110/1111 and 1931 are partly secondary deposits derived from earlier material. The depositional date of 1110/1111 is unlikely to be earlier than AD 350, and thus the same could be suggested for 1931. Thus AD 350 could be suggested as a terminus post quem for the later floor in the octagonal room. Other factors might also indicate a late date for the bathhouse. The date of wall 3057/3273 in Room 38 (which must predate the bath-house as the bath-house effectively blocks one of the entrances into the room) in the Merchant’s House area is clearly important, although this can only be said to predate the early 5th century (as wall 3057 is abutted by mid 4th- to early 5th-century deposit 3302). The bathhouse was subsequently truncated by the late antique city wall (3230) (see Chapter 5, phase 6b). Perhaps the best indication for a late date for this bath-house is its position in relation to the long gallery (18) and the apsidal reception room (24). The bath-house could only be accessed via these rooms, and it is unlikely that this arrangement would have been adopted while the gallery and reception rooms remained the principal public spaces in the complex. In the later Roman period, partly due to Christian ambivalence towards the practice, bathing became an increasingly private activity (a situation perhaps reflected in the limited size of the Triconch Palace bath) (Baldini-Lippolis 2001, 64–5). It is thus likely that the bath suite would be situated away from the more public areas of the elite domus rather than only being accessible through them. It therefore seems most probable that the bath is contemporary with the peristyle domus phase in which the focus of the house moved away from the gallery/reception room axis.

Summary/interpretation Although the dating evidence associated with this phase is limited, it seems that during the second half of the 4th century, perhaps close to AD 400, the domus underwent a major reconstruction. This involved a substantial shift in the focus of the house away from the long gallery and apsidal reception room of the 3rd and early 4th centuries that overlooked the channel, towards a new peristyle courtyard that occupied the central area of the complex. The principal entrance into the house was also moved in this phase when the eastern entrance that gave access into the long gallery was supplanted by a major new western entrance in which a colonnaded vestibule led to the peristyle. A mosaic announcing the owner’s name and rank greeted the visitor on the threshold of the peristyle, clearly suggesting that this was now the principal entrance into the house.

37

The long gallery and apsidal reception room were no longer the principal public areas of the house and this is reflected in the fact that a large drain running from the peristyle was cut into the splendid mosaic of the gallery with no effort made to hide the damage. It is also likely that the small bath-house that lies to the southwest was also built at this time. Those using the bath would have had to walk down the gallery and pass through the apsidal reception room in order to reach the bath; consequently, it is unlikely that these remained areas that visitors were intended to frequent on a regular basis. There is some indication that there was a short period of disuse or an episode of collapse prior to this reconstruction. A thick layer of clay and wall plaster was found beneath the structures of the new western entrance, although this may possibly have resulted from deliberate demolition and levelling. There can be no doubt that the floor levels of the new buildings were significantly higher than those that preceded them. The mosaic of the western entrance was nearly 0.5 m higher than the opus signinum floor of the earlier building, while the mosaic of the peristyle was at a similarly high level compared to the mosaics of the long gallery and reception room (a difference of 0.44–0.50 m being apparent between the two). The reasons for this rise in floor levels are not clear although it is possible that rising water levels in the Vivari Channel were already having a negative effect on channel-side buildings. It may be these increasingly waterlogged conditions that persuaded the owner of the domus to move the opulent public areas of the house northwards away from the channel-side and to raise the pavement level within the building. The reconstruction of the domus around AD 400 with floors at a level 0.50 m higher than that of the earlier phases may well be connected with seismic activity. Recent excavations in the forum (Hernandez and Çondi 2008) have demonstrated that there was a major seismic event in the second half of the 4th century, which resulted in the forum pavement developing a marked incline to the south. The southern edge of the pavement was found to be 0.60 m below the level of the pavement in the north of the forum. This drop in level is obviously reminiscent of the difference between the pavement levels of the domus of AD 400 and its predecessor. It may be that a sudden tectonic slump meant that the earlier building became waterlogged as well as suffering significant structural damage, requiring its reconstruction at a higher level.

Phase 4: c. AD 420 The Triconch Palace (Fig. 2.46) Early in the 5th century (c. AD 420) an attempt was made to greatly aggrandise this already large domus, which involved extending the building onto the adjacent building plot to the east (plot 3). The peristyle courtyard was reconstructed on a larger scale and plot 3 was incorporated into the plan of the main residence with a major extension that included a huge three-apsed triclinium (the triconch after which the building is named). The principal entrances into the house

38

1126

Room 12 1144

Room 14

Room 17

Room 9

1168 1350

1169

Room 16

5421

Room 29 1421

Room 7

2002

Room 27

Room 1/2

1044

Room 1

1359

Room 31

Room 8

Room 2

Room 25

1356

1760

Room 3/4 1761

1002

Room 26

5586

Room 23 Room 24

1040

1045

Room 21

Room 20

Room 19

Room 5

1057

1056

1091

Room 6

1054

1067

1055

Room 11

1068

Room 22 1093

1064

1063

Room 18

1086

Room 35 Room 28

1088

1097

1087

1216 1218

Room 36 1994

1795

0

Figure 2.46. The Triconch Palace of phase 4 (c. AD 420), with structural elements discussed in the text

20m

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

1002

1354

1420

Room 10

1012

1195

2.  The domus and the Triconch Palace

39

Figure 2.47. Aerial view of the Triconch Palace (after conservation) showing the enlarged peristyle and the new triclinium of phase 4 (c. AD 420) (Alket Islami) were moved once more, with the western entrance blocked and major new entrances constructed to the north and west. The triconch triclinium itself was excavated during earlier Italian and Albanian campaigns, together with the doubleapsed vestibule (11); consequently, no stratified deposits were identified in these areas. However, dating evidence relating to this phase was recovered on the western side of the site and from the east portico of the peristyle. The peristyle The new peristyle courtyard of the Triconch Palace was almost three times the size of the structure it replaced, with interior dimensions of 21.5 × 11.7 m. Its construction involved the almost total demolition of the northern and eastern wings of the earlier domus as well as the earlier smaller peristyle. However, despite this the builders used elements of earlier buildings wherever possible, resulting in a rather uneven structure in which the northern and eastern porticoes were considerably larger than those to the south and west (Fig. 2.47). On its southern and western sides, the porticoes of the new structure were the same widths as those of the earlier structure, with the portico walkways merely lengthened to cater for the increased size of the new peristyle. While on the western side this only involved the extension of the portico as far as the line of the old north wing, to the south the portico was more than doubled in length, measuring 31.60 m between the east and west wings. The southern and western limits of the old flagstone courtyard were

Figure 2.48. The portico stylobate of phase 4 (c. AD 420), showing use of spolia. The earlier fixing holes are clearly visible on the block beneath the column base

40

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

Figure 2.50. The two surviving column bases at the eastern end of the south portico

Figure 2.49. The earlier city drain, which partly dictated the line of the eastern portico of the peristyle of phase 4 (c. AD 420) (1 m scale)

retained, although the stylobate wall that surrounded the earlier courtyard was replaced with a much more substantial structure built using faced limestone blocks with an upper surface of c. 0.75 × 0.75 × 0.24 m. These were spolia, perhaps derived from a monumental building within the city. In many cases they had sockets for iron ties that had fixed them in their previous locations (Fig. 2.48). The new stylobates were 0.75 m wide, and were clearly intended to support a much more monumental structure than those of the smaller courtyard, which were only 0.45 m in width. The porticoes on the north and east sides of the peristyle were much wider than those to the south and west. This was partly expedient, reflecting the existence of earlier buildings that could be used within the new structures. This is particularly evident on the north side where the portico is 5.30 m wide, much larger than the south portico which was only 3 m across. The increased width of the north portico is due to the fact that its walls reuse those of the north wing of the earlier domus. Thus wall 1169/1305 became the back wall of the new north portico, while the foundations of walls 1939/5610 were used to support the

stylobate and colonnade. The width of the portico thus reflects the width of the demolished north wing of the peristyle domus that presumably lies beneath it. On the east side, the line of the portico stylobate (2002) was dictated by the east side of the pre-existing city drain that had previously delineated the extent of the domus complex (Fig. 2.49). It followed the alignment of the drain and was therefore not at an exact right angle to the north and south porticoes. This slightly skewed angle was also followed by the triconch triclinium. Like the north portico, the east portico was c. 5.30 m wide. This dimension must have been deliberately intended to follow that of the north portico as the position of the triconch façade (1002) (which formed the back wall of the portico) was not determined by pre-existing structures. Little trace of the peristyle colonnades survives although two bases remain in situ at the eastern end of the southern portico (Fig. 2.50). These give a column width of 0.45 m, separated by an interaxial distance of 1.88 m. This indicates that there were eight columns on the north and south sides of the peristyle and four on the east and west sides. The actual corners of the peristyle were supported by four right-angled piers made of roughly coursed limestone masonry (Fig. 2.51). The east portico and south portico had no traces of floor surfaces, and the small areas of the north portico that were exposed also showed no sign of ever having been paved. The mosaic of the early peristyle partially survived in the west portico, and it is possible that the builders of the enlarged peristyle intended to retain this earlier pavement in the new structure. It was certainly envisaged that the floor level of the new peristyle would be the same as that of its predecessor. The earlier walls that were removed to make way for the new peristyle (5585, 5586, 3039, 1085) were all demolished to below this level, while a large threshold block at the east end of the north stylobate would also function with a portico pavement at this level (Fig. 2.52). The interior of the courtyard similarly showed no sign of surfacing, and it is possible that it may have been intended

2.  The domus and the Triconch Palace

41

Figure 2.51. Right-angle pier on the northwest corner of the peristyle (1 m scale)

Figure 2.52. Threshold block at east end of north stylobate

for use as a garden. No deposits contemporary with the construction of the triconch were found within the courtyard, however, and as is discussed below, it seems that the triconch was never finished or occupied as a prestige residence.

explicitly Christian nature of elite identity as expressed by the owner of the Triconch Palace, a factor also present in some of the earlier decorative elements of the domus. The Christian symbols in the lunettes were intended to serve an apotropaic function, protecting the triclinium from attack by evil spirits, although they would have been invisible to anyone approaching the triconch from its principal entrances, and visitors would only be able to appreciate them once they were inside the triclinium itself. Once inside, the triconch guests would have been presented with a lofty and architecturally impressive space (Figs 2.56, 2.57). The two lateral apses of the triclinium were 6.60 m across with a depth of 4.25 m. The central apse, however, was narrower at 6.16 m but substantially deeper at 4.80 m. The reason for this variation is not clear, although the greater depth of the central apse may have been intended to provide more room for the principal dining couch. All three apses had niches, presumably for statuary, set into their inner faces, although in the southern apse much of the wall was taken up by windows offering a view over an open space, possibly a garden, towards the Vivari Channel to the south. Consequently, it is likely that the surviving niche was the only one present in the southern apse. The central apse is likely to have had at least one window above the level of the niches, but that to the north was surrounded by other structures and consequently had no access to external natural light. This would, in fact, have meant that the interior of the triconch was far from symmetrical and would have presented a slightly uneven appearance. The floors of the northern and eastern apses were slightly raised (evidenced by the two cross walls (1044 and 1045) that run between the apse terminals) while that to the south appears to have been lower and lit by the principal windows of the complex (Fig. 2.58). The masonry of the tri-apsidal room was composed of small limestone blocks, with a mid-wall levelling of 3–4 courses of thin bricks or re-used tiles (Figs 2.59, 2.60, 2.61). The niches were lined with brick and the edge of the surviving window was also reinforced with brick (Fig. 2.62).

The triconch triclinium The focal point of the new enlarged complex was the triconch triclinium (Room 1) on the eastern side of the building, which was by far the largest and most architecturally sophisticated structure within the complex. It was situated on the central axis of the enlarged peristyle, from which it was entered via a central doorway. This meant that it did not occupy an axial position in relation to the earlier colonnaded western vestibule (25), although this was in any case superseded as an entrance in this phase. As noted above, the façade of the triclinium was not at an exact right angle to the peristyle. This meant that the builder had to compensate for this within the triconch itself, seemingly attempting to keep the triconch axially aligned with the peristyle rather than with its own façade. This can be seen from the angles of the interior walls within the triconch, none of which are quite at right angles to the façade (Fig. 2.53). The façade of the building rose above the eastern portico of the peristyle as indicated in the reconstruction (Fig. 2.54). It was articulated with four splendid stone windows that would have been placed above the level of the portico roof, allowing the evening light to illuminate the central apse and its occupants. These windows were found in fragments in the major drain that ran in front of the eastern portico (see below, phase 8). The windows were 0.427 m wide and 0.66 m in height to the base of the lunette. They took the form of single slabs of stone, pierced with a pattern of overlapping scales. In the lunettes of two of the windows, monograms in the form of a cross and a rho were flanked by small panels pierced with round holes, while the lunettes of the remaining two windows depicted a star or flower (Fig. 2.55). These windows are of particular importance in that they demonstrate the

Room 12

Room 14 Room 9 Room 10

Room 7

Room 8

Room 2

Room 1/2

Room 1

Room 16

Room 6

Room 3/4 Room 5

Room 11

0

10m

Figure 2.53. The triconch triclinium with the east portico, showing the new entrances to north and south (via Rooms 11 and 12) and the location of the section across the eastern portico (Fig. 2.75)

The masonry was bonded using a distinctive pebbly mortar that is characteristic of this phase of construction. The walls of the conches themselves were up to a metre thick, presumably reflecting the extra load placed on them by the semi-domes with which the conches were roofed. The spring of the southern vault still survives, indicating that the vaults sprang from a height of 3.8 m above pavement level. Two large rectangular piers (c. 1.75 × 1.35 m) reinforce the junctions between the central and lateral apses where the stresses from the combined weight of the semi-domes would have been most acute. The narrower width of the central apse also indicates that the central arch was higher

than those of the lateral conches. The central part of the room was probably covered with a simple pitched timber roof, as the absence of piers at the western terminals of either of the lateral apses suggests that the use of a dome or vault is unlikely. The central apse contains a curious masonry structure, c. 4 m in width, which could conceivably be the base for a stibadium, the semi-circular dining couch characteristic of late antique triclinia, although it is difficult to see how such a crude structure could have been used in such a way (Fig. 2.63). It consists of two curving earth-bonded stone structures, which may originally have formed a single

2.  The domus and the Triconch Palace

Figure 2.54. Reconstruction of the triconch façade and east portico (Daniel Voisey)

Figure 2.55. The most complete of the stone windows from the triconch façade

43

44

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

C N

A

B

B

A-A A 0

C 5

10m

B-B 0

5

10m

C-C

Figure 2.56. Reconstruction of the triconch triclinium, redrawn from Sheila Gibson’s original with the addition of the four windows

2.  The domus and the Triconch Palace

45

Figure 2.57. Reconstruction of the triclinium interior, depicted with fabric drapes to reflect the lack of evidence for decoration. It seems more likely, however, that the triclinium was never finished or used (Daniel Voisey)

Figure 2.58. Axonometric reconstruction of triclinium showing different floor levels in apses (Sheila Gibson)

Figure 2.59. The masonry of the triclinium on its southwest corner (looking east from the peristyle) (JBB)

Figure 2.60. The exterior masonry of the triclinium with the one window surviving to full height (2 m scale)

Figure 2.61. Overall view of the triclinium (with interior partially backfilled for conservation)

2.  The domus and the Triconch Palace

47

Figure 2.63. Possible foundation of stibadium in central conch (2 m scale)

Figure 2.62. The surviving niche in the west side of the south conch (1 m scale)

semi-circular feature, one third of which became displaced. At its widest (at the southern terminus) it measures 1.28 m. The masonry is composed of unshaped limestone blocks and fragments of tile, built around a core of clay and rubble. In view of the evidence that this phase of the building remained incomplete, it seems most likely that this structure relates to subsequent secondary occupation of the triconch. No trace of wall or floor decoration was found within the triconch complex. It is possible that if the building was decorated with marble veneers and an opus sectile floor, then all the decorative elements may have been removed after the building’s abandonment. However, the walls showed no sign of plaster backing or fixing holes for such veneers, and it seems unlikely that such decoration would have left no trace at all. Alternatively, as discussed by Gilkes and Lako (2004, 174–5) it is possible that other forms of decoration such as textiles may have formed the predominant elements of any decorative scheme. However, the archaeological sequence from the 2000–3 excavations broadly confirms the hypothesis outlined in the report on the 1994–99 excavations, suggesting that the final phase of construction remained unfinished (see below). A series of small rooms was created around the triconch (Rooms 5, 6, 7 and 8). All were accessible from the triconch. Two of the rooms (5 and 7) were entered through

doors in the lateral sides of the triclinium adjacent to the main door. One of these rooms (5), which was inserted rather awkwardly into the area between the southern apse of the triconch and the eastern apse of the new entrance vestibule, appears to have later functioned as the small chapel mentioned by Ugolini (see Chapter 10) (Ugolini 1937, 176; cf. Gilkes and Lako 2004, 170–1). A rectangular niche was set into the apex of the apse, presumably for a statue, and a window provided a view to the southeast. The apse also acted as a buttress for the semi-dome of the southern conch. This architectural expedient may partially explain the odd arrangement of the building. Rooms 6 and 8 meanwhile were irregularly shaped spaces created by the insertion of the triconch structure into the uneven space formed by the structure’s outer walls (1012, 1040 and 1067). They opened respectively into the northern and southern conches. Neither was accessible from outside the triclinium. The eastern outer wall that delineated the triconch complex (1040) follows an earlier alignment represented by wall 1068, which was abutted by 1040. It seems likely that these walls represent a property boundary that remained current in the early 5th century when the triconch was built, as there seems little reason for wall 1040 to otherwise follow this alignment. Thus, although the owner of the triconch was able to expand onto the property to the east of the original domus, this is more likely to represent a change of ownership rather than a total breakdown of earlier structures of property ownership. The closing of the western entrance Probably at the same time that the triconch triclinium complex was erected and the new northern and southern entrances were created, the earlier western entrance was blocked by the addition of two new interconnecting rooms on the western side of the complex. These rooms (29 and 31) were well-built and substantial constructions, comprising walls 1359, 1354, 1420, 1421, 1760 and 1761 (Fig. 2.64). All the walls, with the exception of 1359 were keyed in to one another and belonged to a unitary phase of construction.

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

48

1354

5496

1420

5421 Room 29

5422 5500

1421

1359 5575

Room 31 1760

Figure 2.65. Mortar surface 5575 in Room 31 (looking north) (2 m scale)

1761 0

5m

Figure 2.64. Rooms 29 and 31 showing walls and mortar floor 5421

The walls were built of roughly coursed limestone with tile fragments, bonded with a white pebbly mortar and varied from between 0.60 and 0.75 m in thickness. The southern room (31) was almost square, measuring 5.28 × 6.21 m, with the smaller room (29) measuring 6.25 × 4.14 m. A construction cut (5496, 5500) for these walls cut an underlying surface (5422). Room 29 contained traces of a mortar surface (5421) that was possibly a base for a more elaborate pavement of tile or stone, and which clearly post-dated walls 1354, 1420 and 1421, overlying the fill of construction cut 5496. The mortar floor contained early 5th-century pottery, and overlying deposits (5364 and 5409 contained good assemblages dating to the first half of the 5th century. Both these latter contexts also overlay a series of post-holes (see below). In Room 31 the walls themselves cut a trample surface (5575) equated with 5422 that contained 3rd- to 4th-century pottery as well as possible early 5th-century material (Fig. 2.65). It thus seems reasonable to envisage these walls and the blocking of the western entrance as being contemporary with the building of the triclinium and the expansion of the peristyle c. AD 400–420, as the focus of the building moved east to the triconch itself. This change in focus was reflected in the construction of new grandiose entrances at the northern and southern ends of the eastern portico.

The marine entrance The new southern entrance to the complex was via a large double-apsed room (11) with a length of 9.91 m from the centre of the apses and a width of 5.65 m between the two thresholds (Fig. 2.66). This was built as a unit with the triclinium: apse 1055 was clearly of a unified construction with the façade of the triclinium (1002) and the apse (1054) of the triconch’s small ancillary room (5). The western walls of the double-apsed room (1056, 1091 and 1093) all abut the earlier walls 1058/1062 (although these latter walls belong to the pre-3rd century phase 1 building). Like the triconch this building had been emptied during earlier excavations. On the northern side of the room, a colonnaded threshold (1057) opened onto the peristyle, mirroring the colonnaded entrance from Room 14 to the north and establishing a north–south axial symmetry. A pair of columns (for which the bases remain) created a tripartite entrance 4.57 m wide, the intercolumnar spaces varying from 1.20–1.25 m. The bases give a column width of 0.38 m. Like the peristyle stylobate, the white limestone blocks used for the threshold were all spolia, with clear socket marks from their previous usage (Fig. 2.67). Access to the building from the south was through a door 2.30 m wide marked by a substantial threshold for a door that opened inwards, suggesting that this was an external entrance, rather than a means of accessing a private garden or closed courtyard (Fig. 2.68). The door opened onto an area (35) that may have formed the approach from a road that ran along the bank of the Vivari Channel, or alternatively may have led to a private jetty that allowed direct access from the channel. The double-apsed plan of Room 11 is certainly suggestive of a major public entrance to the

2.  The domus and the Triconch Palace

49

Figure 2.66. The marine entrance (Room 11), looking from the south towards the largely unexcavated peristyle and east portico. Note the continuation of the large drain beneath the western apse

Figure 2.67.Column bases on the threshold between the marine entrance and the eastern portico

Figure 2.68. The marine entrance, looking west along the south range. Note threshold block indicating a door that opened inwards

50

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

Figure 2.69. Rooms 12 and 19 (seen in 1994 at a very early stage in the project with dense vegetation over most of the complex) with the threshold blocks for the northern entrance into Room 19 clearly visible on the right

complex. It is a typical late Roman architectural form, used in entrances to a variety of different building types, including churches (Krautheimer 1986: figs 34, 43, 88), houses (Ghislanzoni 1962; Guidobaldi 1986, 183, figs 18–19; Wilson 1983) and tombs (von Hesberg 1994, figs 129, 130). As with the rest of the triconch complex, no trace of wall or floor decoration was found in this vestibule. The city drain that ran across the front of the eastern portico of the peristyle can also be seen running beneath the western apse, indicating that the earlier excavators went well below the level of the original floor.

Figure 2.70. Detail of threshold for northern entrance into Room 19

The northern entrance and the north wing The new residence had major entrances on both the north and south sides, which probably superseded the axial western entrance. To the north, three limestone slabs indicate the presence of a substantial threshold (3.0 m wide) in the northern wall of Room 12, and this appears to have been the main entrance into the complex (Figs 2.69 and 2.70). From here the visitor would pass through Room 9 and hence into the peristyle, via a door indicated by a further threshold block or step (although it is possible that this block is not now in its original position). Alternatively, it may have been possible to enter Room 14 from Room 9, although there is now no indication of a doorway. This could be due to the difference in levels between the threshold of Room 12 and the floors of the peristyle and Room 14 (0.60–0.70 m), which means that there must have been a series of steps within Room 9 or down into Room 14. Traces of a stone pavement can be noted in the southeast corner of Room 9 although it is not clear if this belongs to the original building or a subsequent phase (Fig. 2.71). As noted above, the northern wing of the original domus was presumably demolished to enable the construction of the northern portico of the enlarged peristyle. The northern

2.  The domus and the Triconch Palace

51

Figure 2.71. Room 9, with traces of stone pavement in the near (southeast) corner. The trench along the wall relates to earlier excavations by Kosta Lako (2 m scale) East-facing section across Room 17 1112 South

North 1114

1222

1223

1126

1223

1263

1227 1350

1114 1227 1271

1279

1m

0

Interpretation medieval humic build-up South

North mortar surface 3rd-century or earlier wall

rubble make-up

6th-century and later tips

baulk 10th-century grave

4th-century demolition rubble 4th-century demolition rubble

6th-century and later tips rubble make-up

early 5thcentury wall

4th-century demolition 4th-century demolition

Figure 2.72. East-facing section across Room 17 showing levels that predate the construction of the north wing

boundary of the complex was extended some 5 m to the north to enable the construction of an entirely new north wing. A new wall (1126) was constructed, cutting through the 4th-century demolition layer 1227 (see Figs 2.72, 2.73). It was of a rather haphazard and undulating construction and formed the limit of a new and rather narrow range of rooms on the north side of the enlarged peristyle. Elements of the outer wall of the earlier building (1305, 1306, 1350 and 1169) were retained within the southern wall of the new northern wing. Wall 1169 was abutted by a new wall

(1195), which extended the building to the east (mirrored by the double apsed vestibule that extended the southern wing to the east). Two rooms of this new northern wing (14 and 17) have been excavated (Gilkes and Lako 2004). Room 14 seems to have acted as an entrance between the peristyle and the northern wing (Fig. 2.74). On the south side of the room, a wide threshold, 6 m across, supported two columns with a seating diameter of 0.35 m that created a colonnaded entrance onto the peristyle. Two further short walls (1144

52

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako main entrance

1126

Room 12 1144 Room 14

Room 17

Room 9

1168

1350

1305 1306

1169

1012

1195 1002 0

10m

Figure 2.73. The new north wing of the Triconch Palace, with location of the section in Fig. 2.72

Figure 2.74. Room 14 showing threshold and column seatings (looking west)

and 1168), separated by another threshold block, divided Room 14 from Room 17, although neither was keyed in to the main walls of the building. The original floors of Room 17 seem to have consisted merely of a patchy mortar surface (1222), over a make-up level of rubble (1223). No trace of floor levels was found in Room 14. The new phase is easily distinguished from the earlier structures, as the masonry consists of coursed limestone and tile fragments, mostly bonded with a distinctive pebbly mortar. It is thus clear that the threshold in Room 12 indicates that there was a major northern entrance to the complex, although it remains uncertain as to how one arrived at the peristyle from this entrance. The most likely route seems to be through Rooms 9 and 14, entering the peristyle via the colonnaded screen that mirrored that of the double apsed room to the south.

Evidence for construction dates and abandonment Construction levels relating to the triconch phases of the building were noted in four areas: the east portico of the peristyle; the additional rooms on the west side of the complex (29 and 31); the new north wing (Room 17) and the triconch complex itself (Room 10). A trench excavated between the stylobate of the east portico (3171/1962) and the façade wall of the triconch itself (1002) located a foundation cut (3262) for the triconch façade. This was cut into layers 3161 and 3042, of which the latter produced pottery dating to c. AD 400–420 (Reynolds 2002). Layer 3042 also was also apparently cut by the portico stylobate. (Fig. 2.75). On the north side of the site, wall 1126 (the outer wall of the north wing) cut deposit 1227 (see above Fig. 2.72). This deposit contained 4th-century pottery (Reynolds

2.  The domus and the Triconch Palace

53

South-facing section across the eastern portico West

1002

3170 3171/1962

3262 3042

3161

3458

5589

East

3263

3162 0

1m

Interpretation East

West construction cut stylobate wall of eastern portico phase 1 wall

construction cut

early 5th-century make-up drain

western wall of the triconch triclinium

early 5th-century make-up early 5th-century make-up

Figure 2.75. Section across eastern portico showing construction levels (see Fig. 2.53 for location)

2004: 225), as well as a coin of 364–378 (SF 522) and two further coins of 408–423 (SF 521) and 410–435 (SF 511), although the 5th-century date of the later coins by comparison with the pottery led Paul Reynolds to suggest that they may be intrusive. He alternatively suggested that the ceramic assemblages might reflect redeposited material with later intrusive material (such as the coins) representing the actual date of the final deposition. However, at present it is only possible to say that wall 1126 dates to after the mid 4th century. Similar caveats apply to the material from contexts 1110/1111 from Room 10, which are also cut by a foundation trench for a triconch phase wall (Gilkes and Lako 2004: 155). This was predominantly 3rd-century pottery, but with later, apparently intrusive, material dating to the latter part of the 4th century. A sherd of ARS 31, apparently from the same vessel as one found in this deposit, was found in the bath-house on the other side of the site (see above, context 1931) giving some indication as to the extent to which material has been moved around the site. 1110/1111 was cut by the foundation trench (1029) of wall 1012/1015, which is the southern wall of Room 10 and also forms the north retaining wall of the triconch complex itself (Fig. 2.76). As noted previously it is likely that the Triconch Palace was never completed or used as a prestige residence, and areas of the complex were rapidly put to other uses. The dates of these earliest phases of post-triconch occupation (examined in detail in the following chapter) provide a terminus ante quem for the construction of the triconch complex and as such will be noted briefly here. Of particular relevance are a series of deposits excavated on the western side of the site in Rooms 22, 23, 26 and 29, which appear to relate to secondary occupation in the second quarter of the 5th century or before. In Room 22, three possible hearths (1951, 1952, 1953) were found directly overlying a clay bedding layer (1964), suggesting that the pavement of the room (which

survived fragmentarily as 1950) had been removed before the creation of the hearths. These features were overlain, however, by a thin but extensive horizon of clay, ash, and charcoal, which contain material dating to c. AD 400. In Room 23 an important series of deposits were found associated with the room’s use as a storeroom, including the in situ remains of a dolium or pithos. This sequence will be discussed in detail in the following chapter, but it suggests that this usage for Room 23 dates from the early 5th century onwards. The lowest deposit in this sequence (5327) dated to the late 4th or early 5th century. This deposit also extended through the doorway to Room 26, and it seems likely that Rooms 23 and 26 were used in tandem for small-scale domestic and industrial purposes. In Room 26 (the room immediately to the south of the colonnaded western entrance) a complex series of deposits relating to a hearth (3441) and a kiln (3256), which were among the earliest ‘post-domus’ features identified, were found to date to the first half of the 5th century. Of particular significance is a large yellow and black silty deposit (3414) that overlies the phase 3 domus threshold (1292) between Rooms 25 and 26, and which dated to the early 5th century. A subsequent series of deposits (3361, 3339) associated with the demolition or abandonment of the hearth and kiln contained ceramics dating respectively to the late 4th to early 5th century and c. AD 425–50. However, there were ceramic joins between the latter deposit (3339) and the earliest deposit in the sequence (3414) while a series of small pits and post-holes also disturbed the sequence. Nonetheless, it is clear that small-scale industrial or domestic activity was occurring in Room 26 from as early as AD 425. Room 29 was the northernmost of the two rooms added during the triconch phase that blocked the western entrance. The walls (1354, 1420 and 1421) cut layer 5575, which containing pottery dating to AD 400–450 (as well as 3rd and 4th-century material). A series of post-holes were also found cutting the early 5th-century mortar floor of the room (5421). These were overlain by deposits 5364 and 5409,

William Bowden, Karen Francis, Oliver Gilkes and Kosta Lako

54

1023

Section

1021

Section

1021

1023

1014 1014 1015

1037 1037

1009 1009

1015

1022

1010

1029

1119

1029

1119

1022

1010 1008 1008 1027 1027

1012 1012

0

5m

0

5m

West-facing section across Room 10 1012

West-facing section across Room 10 North

1038

North

1021

1038

1037 1037

1030 1019

1009 1009

1021

1008

1019

1009

1029

1009

1029

1008

1030

1012

South South

1118 1118

0

1m

0

1m

Interpretation Interpretation construction cut for an earlier pisé building removed and infilled prior to the construction of the Triconch northern wall of Room 10 northern wall of Room 10

construction cut

mortar surface

construction cut 4th-century demolition rubble mortar surface 4th-century demolition rubble

southern retaining wall of the Triconch

southern retaining construction cut wall of the Triconch construction cut for an earlier pisé mortar surface building removed and infilled prior to the construction of the Triconch construction cut early wall 4th-century demolition rubble mortar surface 4th-century demolition rubble

early wall

Figure 2.76. Room 10 showing relationship between structures and early deposits (see also Gilkes and Lako 2004. 155–7)

2.  The domus and the Triconch Palace both containing good assemblages of ceramics dating to the first half of the 5th century. Thus, as in the case of Rooms 23 and 26, it is clear that secondary activity was occurring in buildings of the triconch construction phase from relatively early in the 5th century. Other indications of secondary occupation in the formerly opulent buildings comes from Room 18 where a series of features, including post-holes, slots and possible partitions, were found cutting an early 5th-century clay and mortar deposit (5140), and underlying a large spread of green silts and clays (1759, 1768, 5204, 5074). These latter deposits dated to the second half of the 5th century, thus indicating that this secondary occupation in Room 18 occurred prior to that date. However, it is possible that Room 18 had fallen out of use at the same time as the earlier peristyle domus was constructed (with the clay and mortar deposit 5140 possibly predating the insertion of the drain (5213) that ran from the peristyle and which cut the Room 18 mosaic). Finally it should be noted that the bath-house (Room 28) fell from use during the 5th century, indicated by a charcoal-rich deposit (1973), which covered the sub-floor of the hypocaust. This layer, which presumably post-dates the last firing, contains ceramics dating to the 5th century. Summary/interpretation In its final phase as an elite residence the Butrint domus underwent a massive programme of expansion that saw it almost doubled in size and extended onto the adjacent plot to the east. This construction programme, which seems to have been carried out around AD 400–420, involved a major enlargement of the peristyle and the addition of the triconch triclinium with associated elaborate entrances to the north and south. The north wing of the building was moved further northwards to accommodate the expanded peristyle, and the earlier western entrance was blocked by the insertion of two new rooms. The new buildings differed in some key respects from those that preceded them. In particular, the triconch and the remainder of the new east wing have a unity of construction and design that is entirely absent from the rest of the complex. Rather than adapting the remains of earlier buildings, the east wing is an entirely new construction. This suggests either that all the buildings previously on the eastern plot were demolished, or that, perhaps more likely, much of the eastern plot was not previously occupied by substantial masonry buildings. The fact that other elements of the phase 4 building did reuse elements of earlier structures where they were available (for example the north, south and west porticoes reflect earlier buildings) suggests that the latter explanation is more likely. One of the most significant changes entailed in the new arrangement was the closure of the western entrance and the creation of two new entrances to the south and north respectively. Both of the new entrances meant that visitors entered the east portico of the peristyle and from there could proceed directly to the triclinium. The entrance on the south side was particularly grandiose and may have

55

been intended for use by those who would arrive by boat. Alternatively, a road may have run along the bank of the Vivari Channel, although there is no indication that this was the case from plans of the earlier buildings. The northern entrance allowed access from the city side of the complex, probably from an earlier road running to the channel-side that was otherwise blocked by the triclinium complex. The fact that the owner of the house was in a position to apparently block a road to the channel-side gives some indication of the ways in which members of the elite increasingly felt entitled to manipulate the urban environment in an overt fashion. Although the expanded triconch represented a significant investment of labour and resources, the project was apparently abandoned before its completion. The peristyle remained incomplete, with the earlier flagstone pavement of its predecessor remaining incongruously in the southwest corner, and only the western portico paved with the mosaic of the previous phase. The remainder of the porticoes were left bare, awaiting mosaics that were never laid, while the massive triconch triclinium showed no trace of wall or floor decoration. The reasons for this abandonment remain unclear, an issue that will be explored in detail in Chapter 8 Nonetheless, from the second quarter of the 5th century, if not before, a very different mode of living was pursued in parts of the Triconch Palace. The archaeological sequences associated with this subsequent occupation form the subject of the next chapter.

Notes 1

2

3

4

5

6

1368 will be used in the context of this report to describe this long northern wall, although during excavation a series of different numbers were ascribed to denote particular elements of this wall (1079, 1374, 1376, 1377, 1871). It is almost certain, however, that all were part of the same phase of construction. It is possible that 1975 was a later addition and that the original door between the long gallery and the reception room was wider (2.56 m) and offset to the south to allow a clear view along the gallery to the fountain. The door between Rooms 21 and 22 and a door between Room 19 and the central courtyard were both marked on fig. 9.8 in Gilkes and Lako 2004, as was a gap in the wall between the long gallery (18) and Room 23. This reflected the state of knowledge derived from the 1994–99 excavations, where only the wall tops had been revealed over much of the southern range. It should be noted that senatorial status had become considerably devalued by the early 5th century, with more than 3000 men of senatorial rank in each half of the empire (Heather 1998, 190). The reopening of the door between Rooms 23 and 26 in this phase is suggested by the threshold level of the reopened door, which is the same as that of the peristyle mosaic and around 0.18 m below the level of the mosaic in Room 25, the latter discrepancy explained by the height of the colonnaded thresholds surrounding the mosaic. Although the blocking wall 1380 is undated, this access must have remained in this period to allow access to the bath-house.

3  The 5th- to mid 7th-century occupation of the triconch area William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

Introduction This chapter details the archaeological and structural sequences from phases 5 to 10, which encompass the period from the abandonment of the phase 4 building programme until the temporary cessation of activity in the area of the domus around the middle of the 7th century. As detailed in the previous chapter, the Triconch Palace was at least partly abandoned as an elite residence around AD 420, shortly after the erection of the triconch triclinium and its associated phase 4 structures. During the 5th and 6th centuries parts of the Triconch Palace saw intensive use for small dwellings and industrial activity. The ruins of the Roman domus were used both to house these activities and as a quarry with some parts of the complex demolished for building materials. This material may have been used for the construction of the city wall in the first half of the 6th century, or in the many church-building projects that were carried out in the area during the late 5th and the first half of the 6th century. From the middle of the 6th century burials were inserted in parts of the complex, a process that continued until the area was seemingly entirely abandoned around the middle of the 7th century. The division of this period into defined phases is perhaps slightly misleading; it is clear that occupation and use of the various areas of the site was varied and intermittent. While clearly defined episodes of use and abandonment can be detected in some areas, in others occupation was seemingly continuous or followed its own internal rhythm. However, the use of chronologically defined periods has been continued as a means of structuring a large body of complex data.

Phase 5: Early to mid 5th century (c. AD 420–450) Post-built structures and industrial/domestic occupation (Fig. 3.1) As noted at the close of Chapter 2, the hypothesis that the

Triconch Palace was never completed in its final phase is supported by the evidence for small-scale domestic or industrial activity in parts of the complex from the first half of the 5th century onwards. It is the fact that this date is so close to that suggested for the construction of the triclinium itself (c. AD 420), which lends weight to the idea that the phase 4 complex was never completed. The principal sequences associated with this activity come from the south and west wings of the building. This may be largely because the east wing was dug to foundation levels by earlier Italian and Albanian excavations and consequently we know nothing of the later deposits. Equally, excavation of the north wing has been minimal. However, the excavations of the peristyle revealed no sequences that are comparable to those noted in the south and west wings and consequently it seems that this area was left as an open space. The south wing (Rooms 18, 21, 22) (Fig. 3.2) By the early 5th century Room 18 had clearly lost its earlier splendour. The rough mortar surfaces (1771, 5158), that were seemingly laid during the alterations that accompanied the construction of the phase 3 peristyle domus, were covered by an intermittent clay and mortar deposit (5140, 5212), which was perhaps a by-product of the building work of phase 4 (see Fig. 3.78). Deposit 5140 contained ceramics dating to the late 4th and early 5th century and was cut by a number of features relating to a secondary use of the room during the first half of the 5th century. These included post-holes and slots relating to possible internal divisions within the room, including what may be a partition running lengthways down the centre of the room. This was noted in the eastern part of the room during the excavation, but the composite photograph of the floor (Fig. 3.3) suggests that this division may have run the entire length of the room.1 The mortar floor to the south of this central divide shows noticeably more wear than it does to the north. This suggests either that there was a central partition but that the part to the south saw more use than

3  The 5th- to mid 7th-century occupation of the triconch area

57

Room 27 Room 29

Room 25

Phase 5 areas of identified activity 20m

0

Room 26 Room 21 Room 23

Room 22

Room 18

0

20m

Figure 3.1. The triconch showing areas of known activity in phase 5 (c. AD 420–450)

Room 22

Room 21 1952

5132

1953 1951

5608

5205 1964 5249 postholes 5239

postholes 5140

5212

Room 18

1771

timber partition

5172

5258

0

Figure 3.2. The south wing in phase 5 (c. AD 420–450)

10m

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

58

Figure 3.3. Composite photograph of mosaic from Room 18. Note the possible partition running lengthways along the centre of the mosaic. See also Figs 7.2–7.6 (MZ)

5212

5239

Room 18 5217

1975

1976

0

2m

Figure 3.4. Possible hearth 5239, which lay above the mortar/clay deposit 5212

that to the north, or that the northern half of the corridor was covered by a lean-to roof that protected the northern half of the floor from weathering. At the west end of the room, a small hearth (5239), represented by an area of burnt tiles and clay and sitting in a shallow cut (5217) into clay/mortar deposit 5212, abutted wall 1975 to the south of the doorway to Room 24 (Fig. 3.4). A small group of flagstones in front of the central door into Room 21 may mark a threshold. Later deposits ran through this door and so it is clear that it remained open at this point. A number of possible post- or stake-holes were also noted cutting through the mortar 5212 at this end of the room, although no coherent structures could be discerned. Two irregular stone structures (5249, 5258) were erected against the small piers that had divided the entrance vestibule of the phase 2 domus from the remainder of the long gallery (Fig. 3.5). A small rectangular pit (5172) was found against the south wall (5000) of the room. At the east end of the room, no deposits of this period were found and it seems that the mortar floor (1771) of the earlier period remained clear and in use. As noted above, the extent to which Room 18 was roofed in this period is unknown and the function of the room as a whole remains unclear. It seems unlikely that the central part of the room served as living space in this period, but the apparent maintenance of the clean mortar surface at the eastern end suggest that the former vestibule

Figure 3.5. The east end of Room 18 and stone structure 5249

3  The 5th- to mid 7th-century occupation of the triconch area

59

Section

1381

1952

1378

5132 1951

5087

1953

1964

5241

5205

0

5m

Figure 3.6. Room 22 in phase 5 (c. AD 420–450) showing features noted in text and location of section shown in Fig. 3.52

Figure 3.7. Phase 5 (c. AD 420–450) features in Room 22 looking west (2 m scale)

60

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako shallow stone-lined ‘tank’ or basin (5132) that was cut into clay level 1964. An approximately square cut, around 1 × 1 m, was positioned in the northwest corner of the room between walls 1378 and 1381. The cut was lined with vertically placed flat limestone slabs bonded with a light cream mortar. The base was also lined with mortar. Interpretation of this feature is difficult. The way that its outer edges are built suggests that the ‘tank’ is unlikely ever to have stood to a greater height than its present remains. A further spread of mortar (5087), possibly a mortar mixer, was found on the east side of the room against wall 1378 and door blocking 5241 (see below Fig. 3.52). A small clay-bonded masonry structure (5113/5205) in the southeast corner of the room is probably contemporary with these features, although dating evidence is not conclusive. The structure, measuring 1.85 × 1.55 m, was built abutting walls 1368 and 1369, and was composed of two insubstantial walls that survive only to a height of a single course. Its function is unclear, although quantities of mussel shells within its fill (3206) might suggest that it was used to cook and prepare mussels, an activity that is dominant in later phases in the area. With the exception of the stone tank (5132) and the stone feature in the southeast corner (5113/5205), all these features were overlain by an extensive thin trample horizon (1823) composed of clay, charcoal and ash with frequent inclusions of mussel shell. Pottery from this level suggested a date of around AD 400 for its deposition, although such an early date would be surprising in view of the postulated construction dates for the triconch triclinium of c. AD 420.

Figure 3.8. Hearth 1952 (represented by a single large tile) (2 × 2 m scale) continued to function as a roofed space. The hearth at the western end could have been open to the air. In Room 21, a clay and mortar deposit 5608 was noted extending throughout much of the excavated area of the room and may be equated to the similar deposit 5140 in Room 18, which was connected to it. Like 5140, the clay deposit 5608 overlay a rough mortar surface (5624) that probably relates to the remodelling of phase 4. Room 22 was also used for small-scale domestic or industrial activity (Figs 3.6–3.7). The bases of three hearths (1951, 1952 and 1953) were found above clay level 1964 (into which the long drain was cut). The hearths were directly on the clay and thus post-date the robbing of the floor (1950) (of which a few possible traces survived). All were on the eastern side of the room. The best surviving hearth was 1952, which had a tile base of which one central tile and a number of fragments were found (Fig. 3.8). Hearth 1951 was an area of burning and ash measuring 0.5 × 0.52 m, while 1953 was a sub-rectangular area of burning surrounded by small limestone blocks that may represent the fragmentary remains of a hearth surround. These features were seemingly contemporary with a

The bath-house (Rooms 28, 36) (Fig.3.9) It is likely that the bath-house had gone out of use by the time that the triconch building programme ceased. Pottery dating to the 5th century was recovered from a charcoalrich clay layer (1973) that overlay another clay layer (1990), which in turn sealed the mortar sub floor. Layer 1973 also contained hammerscale from iron smithing, which may derive from the smithing activities in Rooms 18 and 24. Sealing 1973 was 1934, a mortar and tile layer relating to collapse or demolition in either phase 8 or 9. Evidence for continued occupation in the southern range during these phases, and activity in the Merchant’s House of the same date, suggests that there may also have been limited occupation in the former bath-house area. A single post-hole (1993; fill 1995) was recorded cutting rubble 1934 and may reflect evidence of a post-built structure in the late 6th century. Layers 1887 and 1790 sealed this post-hole; these contained pottery of a 7th-century date and possible medieval amphora sherds. These rubble layers are probably collapse of wall 1795 or associated doorway/arch and represent a process of demolition seen across much of the southern range during phase 10. The west wing (Rooms 23, 25–27, 29) (Fig. 3.10) Rooms 23 and 26 appear to have formed a unit during the early 5th century when they were apparently used

North North

2m 2m

61

0 0

mid 6th-century trample deposit mid 6th-century trample deposit late 5th-century demolition late 5th-century demolition

2018 2018

early 5thcentury early 5thbath-house century wall bath-house wall

2017 2017

1934 1934

city wall city wall

1994 1994

South South

bath-house mortar sub-floor bath-house mortar sub-floor abandonment or latest abandonment or latest firing deposit firing deposit bath-house firing deposit post hole bath-house firing deposit post hole

1991 1991 1973 1973

7th-century collapse 7th-century collapse

13th-century horizons and humic build-up 13th-century horizons and humic build-up early 5thcentury early 5thcentury bath-house bath-house wall step wall step

1797 1797 1934 1934 1790 1790 1974 1974

Figure 3.9. Section through bath-house (Rooms 28, 32, 36) showing abandonment deposits

7th-century collapse 7th-century collapse

1798 1798

0 0

marble, blue schist and limestone marble, bluefloor schist and limestone floor

early 5thearly 5thcentury bath-house century wall bath-house wall

North North 1812 1812 1772 1772 1795 1795 1114 1114 South South

for domestic habitation. In Room 23, the earliest datable deposit was a thin clay layer (5327) (Fig. 3.11) containing ceramics of the late 4th or early 5th century, which overlay a series of mortar and rubble deposits (5341, 5368) (Fig. 3.11). It is possible that this layer (5327) is levelling for a robbed floor associated with the phase 3 peristyle domus (5327 is in fact 120–200 mm below the level of the mosaic in the entrance vestibule (Room 25)). Layer 5327 extended through the door into Room 26 (which had been reopened at the same time as the creation of the phase 3 western entrance). The clay level 5327 clearly formed an occupation level in phase 5, indicated by the in situ remains of a large dolium at least 1 m in diameter that was positioned just inside the door to Room 26 (to the west of the door itself) (Figs 3.12 and 3.13). A patch of mortar (5326) may represent the remains of a mortar floor above the clay. A pit or post-hole (5393) was cut into the clay in the northwest corner of the room adjacent to the dolium. The sequence was sealed by a clay layer (5302) containing pottery dating to AD 425–450. However, four lamps found in a tile roof collapse layer (5014) above the clay may date to this phase of activity. They were found in each corner of the room and were probably placed on shelves or in niches in the walls, where they remained until the roof collapsed around the end of the 5th century (see below phase 7). In Room 26 (the room immediately to the south of the colonnaded western entrance), a complex series of deposits were found dating to the first half of the 5th century. No traces of a floor surface relating to the phase 3 peristyle domus or phase 4 triconch were found, and the early 5th-century levels were found directly above the rubble level 3355 that extended under the mosaic in Room 25 to the north (see Fig. 2.26). This suggests that Room 26 originally had a stone or marble floor that was conclusively robbed following the abandonment of the triconch building project. In the northwest corner of the room, an oval platform of clay (3362), around 2 m across and 0.15 m deep, was created. This acted as the base for a large oven or hearth made of tiles bonded with clay (3256) (Figs 3.14, 3.15, 3.16). An underlying tile platform supported a rectangular hearth, also made of tiles, surrounded by an oval border of tile fragments and clay. A box-shaped flue or chimney made of limestone blocks with a tile base, led from the back of the furnace and abutted the western wall 1282, where presumably the smoke and air was directed upwards to the roof of the building. Two areas of paving made from large, finger-impressed tiles were attached to the furnace at ground level on its western and northern sides. A substantial post-hole (3410) and nine smaller post-holes and stake-holes (3480) were located close to the furnace on its southern side. These probably represent the remains of an adjacent timber structure associated with the furnace – perhaps timber supports for bellows or tools. They were cut into a mottled black and yellow clay deposit (3414) that directly overlay the rubble layer 3355 and which overlay

2m 2m

3  The 5th- to mid 7th-century occupation of the triconch area

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

62

to assign any particular use to the structure. The deposit contained many fragments of animal bone, some worked. In particular, on the northern side of the oven close to the chimney, a delicately carved bone handle was recovered, in the form of an extended arm, with the hand holding a scallop shell (SF 2154). On the surface of layer 3361 was a small area of schist paving and an almost complete rotary quern made of granite (SF 2169). The ephemeral remains of another, smaller hearth dating to this phase (3441), were found in the opposite northeast corner of the room at ground level (Fig 3.17), The structure consisted of a square or oval platform made of tile fragments and clay, and surrounded by limestone blocks, although much of the structure was destroyed by a later pit (3433). This hearth went out of use and was buried beneath a sandy clay deposit (3339), which also covered the oven deposit 3361, although the oven itself apparently remained in use. A new hearth (3326) was built adjacent to the oven (Figs 3.17 and 3.18). This was a square structure, 0.70 m across, consisting of a flat tile platform built at ground level with a border of pieces of limestone. A large shallow rubbish pit (3433), some 0.80 m across and 0.25 m deep, containing oyster and mussel shells was cut into 3339, and appeared to be contemporary with this second hearth (Fig. 3.17). The whole sequence was sealed by a clay levelling deposit (3257) containing pottery dating to c. AD 425–450. This deposit contained a number of small copper coins (one of which was datable to 402–445 while three others could be ascribed to a broader 5th to 6th century period) and a heavy iron and bronze chain fitting, perhaps for a cauldron (SF 2133). Ceramic joins between layer 3339, which overlay the first hearth, and the earliest deposit in the sequence (3414) indicate that the sequence may have been disturbed by the post-holes that cut the lower deposit. Equally there were joins between 3257, which sealed the oven and second

the southern colonnade (1292) of Room 25. This seems to be stratigraphically the earliest deposit in the sequence, and dates to the early 5th century. The furnace itself appears to have been in operation over a relatively long period, evident from successive lenses of ash and burnt clay, with an associated charcoal-rich deposit 3361, which radiated out from the furnace, showing signs of trampling and continual deposition. This layer contained substantial quantities of pottery dating to the late 4th and early 5th century, a coin dating to 388–402,2 and many fragments of fine-walled glass oil-lamps with flared rims and narrow bases. At least one of the glass fragments was decorated and a single large lump of melted glass, possibly derived from glass working, was found close to the furnace. This provides the only evidence of industrial activity associated with the oven (no residues were discernible on the tile hearth), making it impossible

5615 5594

5421

Room 27

Room 29

Room 25

3362 3441 3326

3256 3410

5393

3433 Clay layer 3414

Room 26

Dolium 5238 5326

5341

Room 23

Clay layer 5327

5368

Room 23

0

5m

Figure 3.10. Phase 5 (c. AD 420–450) features in the west wing (Rooms 23, 2–28, 29, 31)

0

5m

Figure 3.11. Layers 5341 and 5368 in Room 23 beneath clay 5327

3  The 5th- to mid 7th-century occupation of the triconch area

63

1292 3256 3362

Figure 3.12. View looking north across Room 23 showing clay layer 5237 and the dolium immediately inside the door from Room 26 (2 m scale)

3414

flue

3480 3410

3326

0

tile

1m

Figure 3.13. In situ base of dolium in Room 23 (1 m scale)

Figure 3.14. Hearth 3256 with hearth 3326 beyond (1 m scale) East-facing section across kiln 3256, Room 26 3362

South 3414

3256

1599

3355

3476

1292

North 1292

3440

0

3256

1m

Interpretation

3362 3414

flue

clay platform

South

3480 3410

3326

kiln

paving

early-mid 5thcentury occupation late 4th-century rubble

North stylobate wall early wall

late 4th-century rubble cocciopesto floor

0

tile

1m

Figures 3.15 and 3.16. Hearth 3256 with section showing relationship to earlier features

East-facing section across kiln 3256, Room 26 South 3414 3476

3362

3256

North 1292 3355

1599

0

1m

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

64

3433

3326 3339

3441

tile

0

1m

Figure 3.17. Hearths 3441 and 3326 and pit 3433. Deposit 3339 covered hearth 3441 prior to the creation of hearth 3326 and pit 3433 but 3441 is shown here for ease of reference

hearth, and 3339. Nonetheless it is clear that small-scale industrial or domestic activity was occurring in Room 26 from relatively early in the 5th century. Traces of occupation relating to this phase were also noted in Room 27, where the tile base of a hearth (5615) and an associated ash deposit (5594) containing 5th-century pottery, were found above the plaster deposit 5600 (see Chapter 2, phase 3) (Figs 3.10, 3.19, 3.20). Like Room 26, there was no trace of a floor surface relating to the west wing of the phase 3 peristyle domus, with the hearth lying directly on deposits that are likely to predate the western entrance (the plaster deposit 5600 is around 0.40 m below the level of the mosaic in Room 25), suggesting a marble or flagstone floor that had been robbed out (see Chapter 2, phase 3). Other evidence of occupation in the first half of the 5th century was found in Room 29, the northernmost of the two rooms added during the triconch phase that blocked the western entrance (Fig. 3.21). Here a series of post-holes (5419, 5454, 5456, 5458, 5460, 5462, 5464, 5466, 5468, 5470, 5472, 5502, 5504) were found cutting the mortar floor of the room (5421) and the earlier surface beneath (5422). No clear pattern could be discerned, although the post-holes could have supported a makeshift roof. The fact that no overlying tile deposits were found could suggest that these rooms were never roofed as part of the phase 4 (triconch) construction and thus the rooms were turned into more makeshift shelters after the abandonment of construction. The post-holes were overlain by deposits 5364 and 5409 (Fig. 3.22), both containing good assemblages of ceramics

Figure 3.18. Hearth 3326, looking south (1 m scale)

dating to the first half of the 5th century. Summary/interpretation After the abandonment of the construction of the triconch and its associated structures in around AD 420, a number of the abandoned rooms of the complex were turned over to more ephemeral occupation. In particular, the southern and western wings of the domus were occupied for domestic purposes and perhaps also served small-scale industrial functions. A hearth was found at the western end of Room 18 (the long gallery), while post- and stake-holes attest to wooden partitions elsewhere within the room. It is possible that the gallery was sub-divided into living and working areas, although no clear plan could be discerned from the post- and stake-holes. Rooms 23, 26 (and perhaps 25, 27, 29 and 31) were clearly interlinked in this initial early 5th-century phase. Room 26 (which lay to the south of the former colonnaded western entrance) contained two hearths, one of which was a fairly substantial structure that may in fact have been an oven or furnace. Immediately inside the door to Room 23, a large dolium probably served to store water, while the room itself was lit with four oil lamps found within a

3  The 5th- to mid 7th-century occupation of the triconch area

65

later roof collapse deposit. This dwelling was entered from the courtyard via the colonnaded western entrance, and a trampled silt deposit appears to have developed over the mosaic pavement of the entrance. It is possible to imagine a slightly dilapidated looking but unified complex of rooms on the western side of the courtyard. They were presumably still roofed (at least partially) although the presence of the colonnaded entrance

5594 5615

5600

Room 27 0

tile

1m

Figure 3.19. Hearth 5615 and ash deposit 5594

5502

5466

5421

Figure 3.20. Hearth 5615 (partially hidden beneath the phase 6 circular structure 5336) and ash deposit 5594 (1 m scale)

5468 5470

5456 5504

5464

5454 5460 5458

5419 5462

5422

5472

0

5m

Figure 3.21. Room 29, showing post-holes dating to the first half of the 5th century cutting surfaces 5421 and 5422

66

North-facing section in doorway from Room 29 to Room 31 William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

East

West 5370 5369

1359

5364

1421

5409 5421

0

50 cm

Interpretation East

West blocking

demolition wall

wall early-mid 5th-century rubble

early-mid 5th-century silt mortar floor

0

50 cm

Figure 3.22. Section in doorway between Rooms 29 and 31 showing deposits 5364, 5409, and surface 5421

would have meant that the area with the furnace was well ventilated. The colonnaded entrance also gave access to Rooms 29 and 31, which had been added during the phase 4 (triconch) building programme. The presence of numerous post-holes could indicate that these latter rooms had never been roofed when they were originally erected, forcing the later occupants to erect temporary roofs. Although the phase 5 occupation of the triconch area appears to have much in common with the secondary occupation and sub-division observed in many late antique contexts, which has been the subject of considerable debate, we should be wary of taking this suggestion too far. The earlier doors appear to have been left open through much of the building, maintaining access between its different rooms and areas. Equally, although some areas of the building were clearly the focus of secondary occupation, which differed considerably from the earlier opulence of the building, the hearths, trampled dirt and wattle partitions were absent from some areas. In particular, the west portico (and the rest of the peristyle) appears to have remained clear of debris until the early 6th century, as does the apsidal reception room (24). In both these areas the earliest accumulations of rubbish do not appear to have occurred until 500 at the earliest. The situation may be analogous to that noted at S.

Giovanni di Ruoti in Basilicata, where midden deposits developed alongside the principal rooms of the villa from as early as AD 400 (Small and Buck 1994). Thus it is possible that concepts of elite dwelling were not entirely abandoned in the Triconch Palace area, but were instead redefined according to the mores of the time, a factor that is equally relevant when discussing the 6th-century phases of the site. This will be discussed further in Chapter 9.

Phase 6: Mid to late 5th century Possible flooding and other occupation (Fig. 3.23) Probably during the second half of the 5th century, there was a marked change in the southern part of the Triconch Palace complex. Thick deposits of relatively sterile green-grey silt were noted throughout the remains of the long gallery (Room 18), which also extended into the central room of the southern wing (Room 21). It is possible that these layers represent a period of flooding and/or abandonment. Beyond the long gallery, however, the type of smallscale activity that was characteristic of the early to mid 5th century (phase 5) appears to have continued uninterrupted into the second half of the 5th century.

3  The 5th- to mid 7th-century occupation of the triconch area The south wing (Rooms 18–22) (Fig. 3.24) In Room 18 (the long gallery that fronted the southern wing of the domus), the hearth and post-holes of phase 5 (c. AD 420–450) described above were sealed by a series of green-grey clay-silt deposits (1205, 1759, 1768 in the eastern end of Room 18; 5074 in the central part of Room 18), which lay above the late 4th- to early 5th-century deposit 5140 (Fig. 3.25). A brownish grey clay layer (5204) at the western end of Room 18 may also be equated with these layers (see also Fig. 3.78). Layer 5074, which occupied much of the central part of Room 18, contained tile and limestone fragments, flecks of crushed mussel shell and a considerable quantity of vessel and window glass (particularly towards the southern wall of the gallery), together with coins dating to the period 445–498. It also extended through the central door into Room 21 in the south wing, where it was excavated as context 1982. The dating of these layers is unclear. They contain a wide range of ceramics dating from the mid 5th century until the mid 6th century. Ceramic joins between 5074 and later layers (5005, 5013, 5155, 5175) as well as with the earlier layer beneath (5140) attest to the contamination caused by the intrusion of later features. However, the fill (5139) of a later pit (5141) that cuts the green silt layer 5074 contains a ceramic assemblage dating to c. AD 500, which could provide a terminus ante quem for the deposition of these green silts. The interpretation of these deposits is difficult, although the finely sorted and rather sterile nature of the silts could suggest an episode of flooding or inundation in the second half of the 5th century. Alternatively they could represent rain-washed or wind-blown silts that accumulated in the gallery during a period of abandonment.3 Although the chronology of this sequence is not well defined, the next major phase of activity in the long gallery certainly post-dates these deposits. As noted above, this later phase appears to commence around AD 500. Elsewhere in the southern wing this phase of disuse is seemingly absent, and there appears to be considerable activity in the second half of the 5th century. In Room 19 (Fig. 3.26), the earliest level reached was a possible mortar surface (5623) that overlay a compact clay surface or make-up level (5635). These levels were only revealed in the southern half of the room (to the south of wall 5077), but the clay make-up contained pottery dating to AD 450–475, suggesting that the room was reused in the second half of the 5th century or later. The possible floor itself was around 0.18–0.20 m higher than the earlier floor in Room 18 (the long gallery). The surface (5623) was cut by the foundation of a semi-circular masonry feature (5425), which abutted the wall that divided the two halves of the room (5077), as well as the blocking wall (5144) that filled the doorway in 5077.4 The semi-circular feature (5425) took the form of a shallow cut 2.60 × 1.57 × 0.20 m, surrounded by a clay-bonded wall built from limestone rubble (Fig. 3.27). Only two courses of this surrounding wall survived, but the presence of rubble of a similar size

67

and shape in the interior fill suggests that it originally stood to a greater height. The function of this feature is unclear, although it may have served as a storage bin. A small contemporary pit (5625) and a limestone and mortar feature were excavated in the southeast corner of the room. The fills of these features (5612, 5626) contained pottery dating to the second half of the 5th century, while a tile rubble layer (5551) that overlies the features contained pottery dating mainly to 450–475 (with one later piece that was probably intrusive). The very limited date range of the material associated with this usage of Room 19 indicates that this phase of occupation was short-lived. In the central room of the southern wing (Room 21) (Fig. 3.28) there are traces of what may be greenish silt deposits similar to those that filled much of the long gallery (Room 18), although they are nowhere near as consistent or substantial. The silty layer 5609 is confined to an area close to the door to Room 18 and may be a continuation of the green silt deposit 5074 that filled the central part of Room 18. A further grey silt deposit (1982) was noted in the northwest part of the room underlying later features, although no dating material was recovered from either deposit. Although Room 21 may have contained silt deposits similar to those that signalled a temporary hiatus in activity in the long gallery, there was clearly further activity in Room 21 during the second half of the 5th century. A large mortar mixer (1924) was found in the northwest corner of the room adjacent to the central door into the peristyle (Fig. 3.29). It took the form of a sub-circular depression measuring c. 1.35 m across with a thickness of around 0.28 m at the point where the mortar lipped up against the adjacent walls. Several thin layers of mortar remained within the confines of this depression, some of which contained noticeable quantities of crushed tile, although the majority were composed of hard white mortar with inclusions of sand, shell and chert. To the east were the remains of a mortar surface (5314) and the possible base of a tank represented by a sub-rectangular ridge of mortar (1983) in the northeast corner of the room. A further ridge of mortar (5353) was banked up against the southeast corner of the room. These features were overlain by rubble deposits dating to around the start of the 6th century (1827/5305) (Fig. 3.98). By contrast, Room 22 contains little that is clearly datable to this phase. The only exceptions are two postholes (1940, 1942) that are cut into the occupation level (1823) of phase 5 to the west of the earlier drain. These were overlain by a layer of silt (1969), which, together with a further sterile clay layer (5027) (see below Fig. 3.52), may represent abandonment and silting of the same type encountered in the long gallery. It is also of note that the adjacent Room 23 contains no levels that are datable to this period. However, in the case of both Room 22 and 23 it is also possible that they remained in use and were clear of debris. Indeed, it was probably in this phase that the door between Rooms 22 and 23 was reopened, with a new door (5242) being cut through the blocking wall (5241) that had

68

1147

1140

Room 14

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

Room 29

5336

5365

Room 27 5364 1515 1799

Room 31

1535

1840 1800

1839

Room 25 3257

Room 26 3258

5357

1942

1924

Room 23

1940

Room 21

5242

Room 22 brown clay 5204

Room 19 Room 20

1982

5609

5642 5623

grey-green clay 5074

Room 18

grey-green clay 1768/1205

5642 5425 5623

5635

5625

5642

green silt 1759

0

Figure 3.23. The triconch showing areas of known activity in phase 6 (mid to late 5th century)

20m

3  The 5th- to mid 7th-century occupation of the triconch area

1942

1924

1940

Room 19

1982

Room 23 5242

Room 22

Room 20

Room 21

5642

5609

brown clay 5204

69

Room 18 grey-green clay 5074

5623 grey-green clay 1768/1205

Section

5425

5623 5625

5642 5635

5642 green silt 1759

0

10m

Figure 3.24. Schematic plan of the south wing in phase 6 (mid to late 5th century)

been inserted during the creation of the western entrance of the peristyle domus (Fig. 3.30). The bath-house (Room 36) (Fig. 3.9) In the former caldarium (Room 36), the 5th-century deposit (1973), which overlay the subfloor was in turn overlain by 1934, a layer of mortar and tile containing late 5th-century material. The west wing (Rooms 25, 26, 27) (Fig. 3.31) Although the long gallery may have become disused in the second half of the 5th century, other parts of the complex continued to be occupied. In Room 25, the central room of the western entrance with the mosaic pavement, a small hearth (1840) made of tiles on a clay base was found above a trample deposit (1490), together with an associated burnt deposit (1839). The hearth, measuring 0.80 × 0.50 m was positioned in the centre of Room 25, and thus did not entirely block the access to the adjacent rooms. The trample deposit (1490), on which the hearth was placed, immediately overlay the mosaic in Room 25, and dated to c. 457–475 (Fig. 3.58). Pottery from 1490 spanned a range from c. 425–475, but the deposit included a coin of Leo I (457–474) suggesting a later date within this period. In Room 26, to the south of the colonnaded entrance (25) activity also continued into the second half of the 5th century. The clay levelling deposit 3257 (dating to AD 425–450, see phase 5 above) that filled the room to a depth of 120 mm and which sealed the earlier ovens and hearths, was used as the base for a large dolium (3258) of which the lower third was found in situ (Fig. 3.32). The fill of this storage vessel (3184) contained coins of indeterminate

4th to 5th century date, animal bone and pottery, although its original purpose is unknown. Possible fragments of a rough limestone pavement survived in the southwest corner of the room. A particularly important sequence relating to the second half of the 5th century was recovered from Room 27, which lay immediately to the north of the colonnaded western entrance (25) (Fig. 3.33). Here a series of tip levels and clay horizons (5474, 5475, 5484, 5494, 5514, 5516, 5595) were found above the hearth of phase 5 (Fig. 3.34). These clearly represented debris from occupation, containing a rich assemblage of animal and fish bones and pottery (5475, 5484, 5494) and mussel shells (5516). The material dated from the mid to late 5th century, and was clearly derived from nearby domestic occupation. This was overlain by a compacted layer of demolition material (5390) dating to the late 5th century. The rubble deposit (5390), which seems to have spread throughout the interior of Room 27, served as the foundation for an extraordinary circular structure built within the walls of the room (Fig. 3.35). This structure (5336) used the north and east walls of the room (1353, 1568) as its principal supports, while a large buttress (5391) provided reinforcement to the southwest. The wall was built using blocks of limestone laid in fairly even courses with some attempt at creating a smooth interior face. The masonry was bonded primarily with earth. The principal component of the buttress 5391 was a large Corinthian capital from which all detail had been removed, with only the outline of the volutes remaining (Fig. 3.36). This formed the outer part of the buttress, with limestone rubble packed between the capital and the circular structure, and thus providing support

South-facingsection sectionthrough throughRoom Room18 18 South-facing

West

5136

5175 5136 5057 = 1676 18 South-facing section through Room 5175 5259

West West

5047 5057 = 1676 5175

5136

5047

5136

5175

5209 5209

5209

5136

5047

5213

5213 5175 5158

5045 = 1740 50565213

5158

5058

5045 = 1740

1205=1768

1809 1809

drain fill drain fill early 5th-century drain cut

West

West West 3rd- to 4th-century

early 5th-century mortar surface associated mosaic bedding 3rd- to 4th-century with the domus

3rd- to 4th-century mosaic bedding

late 5th- to early 6th-century flooding/abandonment deposit late 5th- to early 6th-century flooding/abandonment deposit

3rd- to 4th-century mosaic

1771 1771

mid 6th-century mid 6th-century occupation level mid tolevel late 6th-century occupation

early 5th-century secondary mortar surface early 5th-century

post hole 3rd- to 4th-century mid 6th-century post hole

industrial refuse

mid 6th-century occupation level

East East

2m 2m

2m

mid to late 6th-century midindustrial to late 6th-century refuse industrial refuse

mid to late 6th-century mid tokiln latewaste 6th-century kiln waste

late 5th- to early 6th-century flooding/abandonment deposit

baulk baulk

late 5th- to early 6th-century late 5th- to early 6th-century flooding/abandonment deposit flooding/abandonment deposit

late 5th- to early 6th-century flooding/abandonment deposit

early 5th-century early 5th-century secondary mortar surface early 5th-century mortar associated earlysurface 5th-century secondary mortar surface with the domus mortar surface associated with the domus mid to late 6th-century mussel shell deposits mid to late 6th-century mussel shell deposits

3rd- to 4th-century mosaic

mosaic early 6th-century 3rd- to 4th-century staircase mid 6th-century mosaic East early 6th-century late 5th- to early 6th-century occupation level staircase flooding/abandonment deposit late 5th- to early 6th-century late 5th- to early 6th-century flooding/abandonment deposit early 5th-century flooding/abandonment deposit early 6th-century mortar surface associated late 5thto early 6th-century staircase 2nd-century 0 with the domus flooding/abandonment deposit wall

East

occupation level

Figure 3.25. South-facing mortar surface associated 0 with the domus early 5th-century to 4th-century the early 5th-century E–W section3rd-through secondary mortar surface mosaic mortar surface associated long gallery (Room 18) early 5th-century with the domus 3rd- to 4th-century secondary mortar surface early 5th-century showing grey-green silt 3rd- to mosaic 4th-century secondary mortar surface mosaic deposits

2m

0 0

0 mid to late 6th-century industrial refuse

1058 1058

1759 1759

East 1058

1759

2m

mid to late 6th-century mussel shell deposits

late 5th- to early 6th-century hole flooding/abandonmentpost deposit

1058

1759 1273

late 5th- to early 6th-century flooding/abandonment deposit baulk early 5th-century secondary mortar surface

1273 1273 1277=1977 1277=1977

East 1809

1277=1977

0

1771 0

post hole

3rd- to 4th-century mosaic early 5th-century secondary mortar surface

1058

early 5th-century mortar surface associated early 5th-century mosaic beddingdrain fill with the domus early 5th-century early 5th-century 3rd- to 4th-century secondary mortar surface drain fill early 5th-century 3rdto 4th-century 3rd- to 4th-century mortar surface associated mid to late 6th-century mortar associated earlysurface 5th-century 3rd-mosaic to 4th-century bedding 3rd- to 4th-century mortar surface associated with the domus3rdearly 5th-century mosaic mussel shell deposits tobedding 4th-century mosaic bedding the domus mortar surface associated mosaic bedding early 5th-century earlywith 5th-century mosaic mortar bedding with the domus secondary surface mid to late 6th-century mosaic bedding mortarwith surface cut theassociated domus early drain 5th-century post hole mussel shell deposits with the domus drain cut

early 5th-century early 5th-century 3rd- to 4th-century drain cut surface associated mortar mosaic beddingdrain fill with the domus early 5th-century drain cut

East

1277=1977

mid to late 6th-century mid to late 6th-century mid to late 6th-century mid to late 6th-century mid 6th-century industrial refuse occupation level occupation level secondary kiln waste occupation level mid to late 6th-century drain mid to late 6th-century mid to latecut 6th-century mid to late 6th-century to late 6th-century mid to late 6th-century mid tomid late 6th-century mid to late 6th-century mid 6th-century industrial refuse mid to late 6th-century occupation level occupation level industrial refuse secondary mid to late industrial refuse6th-century mid to late occupation level occupation level kiln waste occupation level6th-century secondary industrial refuse drain cut mid to late 6th-century occupation levelbaulk occupationmid level to late 6th-century West mid to late 6th-century drain cut mid 6th-century secondary mid to late 6th-century industrial refuse occupation level occupation level secondary drain cut occupation level kiln waste baulk drain cut

Interpretation

1273

1273

1771

1809 1809

1809

1277=1977

1759

Interpretation Interpretation

Interpretation

5140 5140

1809

1771 1205=1768

Interpretation 1809

1205=1768 1205=1768

1205=1768

5158 5158

1205=1768

1809

5158 5158 5058

1205=1768

5158 1205=1768

5140 5140

5140

5058 5058

5140

1205=1768

5140

5140 5140

5056 = 1676

5158

5158

West

5058 5140 baulk

5058 5058

late 5th- to early 2m 6th-century flooding/abandonment deposit

2nd-century wall

East 2m

2nd-century wall

0

2m

early 5th-century mortar associated earlysurface 5th-century with the domus mortar surface associated

with the domus

mid 6th-century occupation level mid 6th-century occupation level

3rd- to 4th-century mosaic 3rd- to 4th-century mosaic

early 6th-century early staircase 6th-century staircase late 5th- to early 6th-century flooding/abandonment deposit late 5th- to early 6th-century flooding/abandonment deposit 0

0

East East 2nd-century wall 2nd-century wall 2m

2m

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

5140

5074

1809

5045 = 1740 5074 5058

1205=1768 5140

5056

5074

5056 = 1676 5056 = 1676 5045 = 1740 5045 = 1740

baulk baulk

5071 5071

5140

5074 5074

5058

5045 = 1740

5140

5071 5158

5056 5056

5045 = 1740 5158 5045 = 1740

5074 5058

5074

5158 5158

5140 5158

5056

5045 = 1740

5071 5056 = 1676 5045 = 1740

5056 = 1676

baulk

5074

5209

5213

5259

5259 baulk 5074

5071

5057 = 1676 5074 5213 5158

5209 West

5047 5047

5259 5259

5057 = 1676 5057 = 1676

70

South-facing section through Room 18 South-facing section through Room 18 West

3  The 5th- to mid 7th-century occupation of the triconch area

5077

5144 5077 5077

5077 5144 5144

5077 5077 5642 5635

5642 5642 5635 5635

5642 5642 5425 5642

5623 5425 5425

5623 5623 5623

71

5625 5623 5623 5625 5625 5642

Figure 3.27. Semi-circular masonry feature 5425 in relation to blocking 5144. Looking west (2 m scale)

5642 5642

0

2m

2m Figure 3.26. Room 19 in phase 6 (mid to0 late 5th century) 0

2m

1924 1983

5314 1982

5608

5609

5353

0

2m

Figure 3.28. Room 21 in phase 6 (mid to late 5th century) showing mortar features 1924 and 1983 and grey-green silt layers similar to those from the long gallery (Room 18) 1378 South

5241

Figure 3.29. Mortar mixer 1924, looking north (top) and mortar feature 1983, looking north (1 m scale) (bottom)

1378 5242 1381

5190 0

Figure 3.30. Elevation of wall between Rooms 22 and 23 showing door 5242 cut through blocking wall 5241

North

1m

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

72

Room 25

5365

Room 29

1292

5336

Room 27

Room 26

5364 1515 1799 1535

3257 clay levelling

1840 1490 1800

Room 31

3258 dolium

1839

1599

Room 25 1383

Room 26

0

3257 3258

0

5m

Figure 3.32. Dolium 3258 cutting layer 3257 in Room 26

5m

5336

Figure 3.31. The west wing (Rooms 25, 26, 27) during phase 6 (mid to late 5th century) 5373

in the area that was not supported by the external walls of the room. On the interior, a thin layer of olive brown clay and tile (5416) may have formed a rough surface. The circular building had exterior dimensions of 4.20 m from east to west and 4.40 m from north to south, with respective internal dimensions of 3.25 m and 3.60 m. At its widest the wall had a thickness of 0.70 m, although at the points where it abutted the outer walls of Room 27 (1353, 1568) it measured as little as 0.20 m. The wall survived to a height of 0.30–0.45 m and it is clear that there was no entrance to the interior at ground level. It is possible that the buttress (5391) may have formed the base for a flight of external stairs to a higher entrance but there is no way of knowing whether this was the case. Neither is there any indication as to whether the building was roofed or whether the earlier roof of Room 27 was still sufficiently intact to provide the cover required. This round building remains enigmatic. The deposits within it provided no clue as to its function, and there are no obvious parallels of this date from Butrint or elsewhere.

re-used column 5391

0

2m

Figure 3.33. Room 27 in phase 6 (mid to late 5th century), showing the enigmatic round structure 5336 It is clear, however, that it was built to serve a function that could not be accommodated by the existing Room 27 with its open colonnade to the south. The lack of an entrance at ground level could suggest a suspended floor, and the substantial nature of the walls could also be indicative of a need to keep out rats or similar. It seems possible, therefore, that the round structure served for the storage of perishable foodstuffs, such as grain, cured meat or fish.

South sections across structure 5336 Room 27 area 3  and Theeast-facing 5th- to mid 7th-century occupation ofinthe triconch

Section West

East 5391

North

South 5337

5336

5336

5381 5390

5474

73

5475

5516

5484 5416

5514

5515 0

Interpretation West

East buttress

North

South late 5th-century demolition

wall

1m

wall

late 5th-century demolition

mussel shell deposit

mid to late 5th-century domestic waste deposits

floor

late 5th-century demolition rubble

mid to late 5th-century domestic waste deposits

surface

Figure 3.34. Composite section across Room 27 showing tip levels 5390, 5474, 5475, 5484, 5494, 5514, 5516, 5595 and the interior (5416) surface of structure 5336

Figure 3.35. The round structure 5336, looking north

Whatever the function of the circular building it is clear that it was a short-lived structure. Deposits from the interior of the building (5381 and 5337) indicate that it was abandoned by the end of the 5th century and used for the disposal of domestic rubbish and building material. Further to the west, Rooms 29 and 31 also contained further structural elements that probably date to the second half of the 5th century (Fig. 3.37). In Room 29 a layer of rubble (5364), containing mid 5th-century material, was spread across the room, apparently to act as a levelling

layer. This lay above a silt layer 5409 that sealed the postholes of the previous phase and which contained material of c. 400–450. Above this layer a clay-bonded wall (5365) was built, partitioning off the western end of the room (Fig. 3.38, 3.39). This was an insubstantial structure 0.40 m wide, with an opening towards its northern end allowing access between the two areas of the room. A thin spread of mortar in the northern part of the room was considered to be part of the rubble layer 5364 and probably represents the remains of a surface associated with this phase.

74

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

5365

mortar

5364

1515 1799 1535 1800

Figure 3.36. Buttress 5391 showing reused section of Corinthian capital. Traces of the volutes were just discernable at ground level tile

A similar partitioning also occurred in Room 31 to the south. A clay-bonded wall (1535) divided the room into eastern and western halves (Figs 3.40, 3.42). Only the central part of this wall survived but it is likely that there were two doors into the western part of the room. The date of this structure may be later than that of the partition wall in Room 29. The wall overlies deposits (3347, 5572) that were equated with 5409 (the early to mid 5th-century deposit in the adjoining room), although, in fact, the material from these deposits could be later, possibly containing late 5th- or early 6th-century material. A mortar bedding layer (1515) and the fragmentary remains of an associated tile floor (1800) may belong to the same phase as the wall (Fig. 3.41). The mortar layer extended throughout the room, although the tile floor was best preserved immediately to the east of wall 1535, where a section measuring 1.80 × 0.75 m survived intact. Pottery from the mortar bedding (1515) dates to the late 5th century; however, the dating evidence associated with this pavement is ambiguous and it is possible that the date range could extend into the 6th century. To the west of wall 1535 a further area of tiles (1799) was thought to be a hearth upon excavation, although it may be part of an equivalent floor to 1800. It was buried beneath a rubble layer (1536) dating to the 6th century.

0

5m

Figure 3.37. Plan of Rooms 29 and 31 in phase 6 (mid to late 5th century)

The central courtyard As with phase 5, the central courtyard is notable in this phase for being apparently free of activities that disrupted its unity and coherence as a space, although a thin silt occupation horizon (5357) was noted in the southern portico; this contained an assemblage of pottery dating to c. 450–475 or later (Fig. 3.43). It seems therefore that the portico probably remained in use in a way that allowed free circulation between the various areas of the complex. The north wing (Rooms 14 and 17) Excavation of the north wing of the building was not as extensive as those of the southern and western wings of the building but Rooms 14 and 17 contained a series of features that post-dated the phase 4 Triconch Palace (Gilkes and Lako 2004: 168–70). Two substantial rectangular mortar features (1140 and 1147) were found against the northern wall of Room 14 (which was added as part of the Triconch Palace construction programme) (Figs 3.44, 3.45). Both features were solid blocks of mortar with clearly defined vertical edges that suggested the presence of surrounding

3  The 5th- to mid 7th-century occupation of the triconch area

75

Figure 3.38. View west across Room 29, partitioned by clay-bonded wall 5365. The mortar and rubble surface (5364) is visible in the foreground (1 m scale)

Figure 3.39. Detail of clay-bonded wall 5365, looking south (1 m scale)

Figure 3.40. Room 31, showing clay-bonded wall 1535, tile floors 1799 and 1800 (Room 31) looking south (1.8 m scale)

Figure 3.41. Detail of Room 31 tile floor 1800, showing mortar bedding 1515 (1.8 m scale)

South

North

1515 5571=5364

1535

5572=5409 0 Figure 3.42. East-facing elevation of wall 1535, showing relationship with underlying layers

1m

76

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

wooden structures, which had subsequently rotted away. Feature 1147 was a solid rectangular block of gravelly mortar measuring 1.90 × 1.20 m, placed in the angle between walls 1126 and 1144. It survived to a height of up to 0.25 m and its southern edge was sharply defined, sloping inwards towards the base (Fig. 3.44), clearly indicating a supporting structure that had not survived. Feature 1140 was of a similar size (1.88 × 1.10 m), and also had sharply defined edges surviving to a height of 0.10 m. It was composed of a gravelly lime mortar laid over a rough base of tiles. The mortar was thickest towards the eastern end of the feature where there were traces of a sub-circular depression within the mortar surface. No dating evidence was recovered from these features, and we know only that they post-date the construction of Room 14 (part of the phase 4 Triconch Palace) and predate the major dumps of tile and rubbish of c. 525–550 (represented by layer 1152). Consequently they may belong to phases 5 or 6. It seems likely, however, that they are the remains of wooden sided tanks used for the mixing of mortar, although it is not known what building project they might be associated with, as the city wall is probably slightly later, but there are no other major masonry structures of the later 5th century that are known from the area. It is even possible that they are related to the construction of the triconch triclinium and were intended to be removed after work was completed. Alternatively, they may have provided mortar for building work in the unexcavated area further to the north. In Room 17, a clay occupation layer (1194) dating to 450–500 was noted above a mortar floor (1200) (associated with the original construction of the phase 4 Triconch Palace) (Fig. 3.45). This layer may be contemporary with the use of the ‘tanks’ in the adjacent Room 14 and also provides a further terminus ante quem for the abandonment of the Triconch Palace as a luxury dwelling. Summary/interpretation During the second half of the 5th century, the area of the Triconch Palace continued to be used for a series of activities probably associated with domestic occupation. In particular, the west wing saw continued activity in the rooms leading off from the colonnaded western entrance of the earlier domus. A hearth was positioned in the central colonnaded room (25), while Room 26 contained a large storage vessel similar to that which had been used in the adjoining Room 23 in the previous phase. To the north of the colonnaded area, Room 27 was used to house a remarkable circular storage structure (perhaps a granary or a store for other foodstuffs). Further to the west, meanwhile, Rooms 29 and 31 had partition walls inserted, closing off the western halves perhaps for use as private sleeping quarters. It is possible, therefore, that the western part of the Triconch Palace continued to function as part of a relatively large domestic establishment. All rooms in the west wing remained interlinked and access to the southern wing may have been reinstated with the cutting of a new door between Rooms 22 and 23.

Activity within the southern wing during the second half of the 5th century is less clearly defined, and it seems that the long gallery (18) may have been temporarily abandoned following an episode of flooding represented by extensive deposits of sterile silt. Activity in the remaining rooms of the southern wing was also less clearly defined than in those of the west wing (perhaps because excavation was less extensive and these levels were only partly reached in Room 19 and not reached at all in Room 20). However, Room 19 may have a rough clay surface dating from this period, while Room 21 contained a large mortar mixer datable to the second half of the 5th century. No deposits of this period were recovered from Rooms 22 and 23, although this may be because they continued to be used and were thus kept clear of debris. It is particularly noteworthy that the central courtyard remained largely clear although a thin silt occupation level was noted along the southern portico. It seems that the courtyard continued to function as a central open space with the porticoes remaining clear of obstruction or refuse. As noted above in relation to phase 5, it is also of note that the apsidal reception room (24) remained free of build-up or deposition in this period. The overall impression of the Triconch Palace area in the second half of the 5th century is of a complex that remained under single ownership and which essentially retained its coherence as a large dwelling, although one that falls short of our expectations of the lifestyle of late antique potentiores. The principal domestic area remained on the western side of the complex, with the southern (and perhaps the northern) wings used increasingly for other activities, such as the preparation of mortar. The unfortunate removal of the later stratigraphy in the triconch triclinium itself means that we are unable to say what role the triconch may have played within this dwelling. The suggestion that the Triconch Palace area remained the single property of an individual or family of some status or influence is reinforced by the fact that the city wall (erected in the first half of the 6th century) deviates markedly from its course in order to avoid the area of the Triconch Palace and the adjacent Merchant’s House (see Chapter 5).

Phase 7: Early 6th century Partial reoccupation of the southern wing and roof collapses in the west wing (Fig. 3.46) The early to mid 6th century was marked by significant change in the triconch area. In the long gallery to the south (Room 18), there is evidence of occupation in the form of pits and post-built structures. At the eastern end of this room a stone-built stair block indicates that at least one of these post-built structures must have had a second storey. The west wing, however, was abandoned, and the roofs of the rooms collapsed, ending the late antique occupation phases of this part of the building. The only exception was Room 31 on the far west of the complex, where a series of post-holes attest to continued occupation.

3  The 5th- to mid 7th-century occupation of the triconch area

77

5357 5357

10m 10m

00

West-facing West-facing section section across across the the Triconch Triconch peristyle peristyle (Room (Room 16) 16) North North

South South 1833 1833

1875 1875

5357 5357

1957 1957

5m 5m

00

Interpretation Interpretation

mid mid to to late late 5th-century 5th-century occupation occupation level level South South

North North early early 5th-century 5th-century northern northern stylobate stylobate wall wall of of the the Triconch Triconch peristyle peristyle

early early 5th-century 5th-century domus domus peristyle peristyle

early early 5th-century 5th-century southern southern stylobate stylobate wall wall of of the the Triconch Triconch peristyle peristyle 0 0

Figure 3.43. Thin silt horizon 5357 in south portico dating to c. 450–475

5m 5m

78

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

Figure 3.44. Mortar features 1140 and 1147 in Room 14 (1 m scale)

In the portico there are some indications of a gradual build-up of rubbish and deposition, although much of the area apparently remained relatively free of deposition. Material also began to build up over the mosaic pavement of the apsidal reception Room 24. The south wing (Rooms 18–22, 24) (Fig. 3.47) In the long gallery (Room 18) a number of features were noted immediately above the sterile grey-green silt levels of the preceding phase. The most significant of these was a substantial masonry structure (1277) built at the east end of the room above the clay layer 1205/1759 and abutting the east wall (1879) (Fig. 3.48). The structure was rectangular and measured 1.96 × 1.26 m, thus partially covering the original eastern doorway of the domus complex (long since blocked), and appeared to be the lower part of a stair block. It was of mortared construction, built around a rubble core, with the traces of certainly one step and perhaps two steps remaining. These were shallow (c. 0.20 m in height) and widely spaced, the lowest measuring 0.70 m across. It is possible that this was a low flight of steps leading to an elevated doorway into Room 19 to the north, although the relative level of the contemporary occupation horizons in Room 19 suggests that this is unlikely; if the builders had wished to create a means of access between the long gallery and Room 19 then a door could simply have been cut in the northern wall (1871). It is more likely that this stair block indicates that an upper floor had been added to the eastern end of the gallery. However, the shallow gradient of the surviving stair block suggests that an additional wooden staircase ran along the north wall (1871) at right angles to the stair block.

The upper floor probably only existed at the eastern end of the gallery, covering an area identical to that of the eastern vestibule of the 3rd century. The arch that originally separated the vestibule from the remainder of the long gallery (represented by the two plastered pilasters 5250 and 5251) probably provided support together with the surrounding walls, as no post-holes were identified with this phase. Patches of compacted mortar (1212 and 1214) were also found above the silt level 1768, and probably represent a surface in the lower floor of this two-storey structure (Fig. 3.49). In the centre of the long gallery (18), a large area of white mortar (1789) was found overlying clay layers 1768 and 5074. In the very centre of the room, close to the south wall was a fairly shallow oval pit with a U-shaped profile (5141), which had been cut into clay layer 1768/5074, down to the level of the mosaic. The fill of this pit (5139) consisted of burnt and charcoal-rich silty clay, containing late 5th-century pottery. The fill, which seemed to have been burnt in situ, also contained the complete neck of a Samos amphora (dating to c. 500 AD). The amphora neck had been deliberately sawn off at the shoulder and placed upright within the pit. The discovery of other examples of truncated amphora necks in burnt contexts across the site, suggests that the Samos neck may have been utilised within the pit as a structural pillar, on which to support a part of a hearth platform or small oven. Further examples of these amphora ‘pillars’ were found within later 6th-century deposits in the long gallery (5100, 5202; see below) and also in within Room 22 (3182). Two shallow, circular post-holes (5152/5153 and 5257) situated approximately 1 metre from pit 5141 and 0.30 m apart belong to this same

1147

1200

1140

Room 17

Room 14

3  The 5th- to mid 7th-century occupation of the triconch area 1143 1186

South-facing section

79

Room 12 1147

1200

Room 12

1140

Room 17

Room 14

1143

1186

1200

South-facing section

1147

Room 9

1140

Room 17

Room 14

1143

1186

Room 9

South-facing section

5th-century mortar surface

0

5th-century mortar surface

5th-century mortar surface

South-facing section across Roomsacross 14 and 17Rooms South-facing section West

Room 17

West 1143 1193

1143 1143 1193

1194

West Interpretation

old excavation trench

1185

1185

1186 Room 17

East

Room 14 old excavation trench

1133

Room 14

1196

1002 East

1163 1152

1196

1200

1133

1196

1152

1002

1163

1163 old excavation trench

1186

Room 14

East

recent humic deposition 14th-century+ build up

West 13th-century Interpretation Room 17 6th-century pit masonry final occupation collapse structure layer 14th-century+ build up West c.525-550 mortar floor 6th-century masonry collapse structure c.525-550

1152

1114

1200

Interpretation

1114

1185

1194

10m

1114

1186

1193

0

14 and 17

Room 17

Room 17 1194 1200

10m

Room 14

South-facing section across Rooms 14 and 17 West

0

Room 17

13th-century pit

mid 6th-century dumping Room 14 recent humic deposition earth-bonded blocking threshold block mid 6th-century dumping

earth-bonded blocking final occupation layer threshold 14th-century+ build up Figure 3.45. Plan of Roomsblock 14 and 17 (with phase mortar floor

mid 6th-century dumping

old excavation trench

East

final occupation layer mid 6th-century dumping

0

Room 14

recent humic deposition

final occupation layer

6 features labelled) with south-facing section

2m

mid 6th-cent 0 2m mid 6th-century dumping 13th-century 6th-century phase. A second, oval-shaped pit (1769) waspitsituated to interspersed with lenses of mussel shells (1740, 1788, masonry earth-bonded blocking final occupation final occupation l collapse east of the the immediate post-holes (Fig. 3.47). 5259), and containing pottery dating to around the middle structure layer threshold Atc.525-550 the western end of the long gallery a fairly shallow, block of the 6th century (see below phase 8 and Fig. 3.78). mortar floor circular mound of sandy clay (5203) was found overlying In Room 19 (Fig. 3.50), a remnant of a probable tile the sterile clay horizon 5204. It contained tile fragments, surface (5639) was found in the southern half of the room, limestone and burnt mortar and was defined on the western where it survived adjacent to wall 1871. A possible bedding side by the ephemeral remains of a narrow rubble wall. It layer (5579) for this surface lay above the layer of tile probably represents a small domestic hearth. rubble (5551) and a layer of loam (5656) that post-dated the All these features in the long gallery were overlain features of phase 6. Both 5551 and 5656 dated to the second by a series of clay deposits (5013, 5045, 5058, 5175) half of the 5th century. In the northern part of the room, a

80 William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

Room 27 5337 5526 5443 1515

Room 31

5489

5451

Room 25

5485 5487 5493

1449

1449

1422/1841

Room 26

door cut through wall

1380

5272 5014 1300

1487/3231/5026

Room 23

Room 22

Room 21

5210

5202

5271

5599

5579 5551 5639

Room 20

Room 24 Room 18

5141 amphora 5152 neck 1769

5311

1789

5074/1768 1214

1277

Room 19

1212

5257 1216 1218

1217

0

Figure 3.46. The triconch showing areas of known activity in phase 7 (early 6th century)

20m

3  The 5th- to mid 7th-century occupation of the triconch area

81

door cut through wall

1380

5272 5014

5311 5271

1487/3231/5026

5599 5579 5639

1300

5210 5202

amphora neck

5141

5152

5074/1768

1789

5551

1277

1214

1769

1212

5257 10m

0

Figure 3.47. The south wing in phase 7 (early 6th century)

Room 19 1871

5250 support for upper storey?

1277 1768

Figure 3.48. Masonry stair block 1277 at the east end of Room 18 (JBB)

1214 5251

further small area of tile paving (5311) was found on the same level although the dating evidence was not as clear as that of the pavement in the southern half of the room. The two areas of paving are not directly associated with each other, as the door between the two halves of the room had been blocked by this stage. It may, however, have been at this point that a door was cut into wall 1058 on the north side of the room, thereby creating access between Room 19 and the portico area. The area of pavement to the south (5639) was overlain by a layer of silt (5553) containing pottery dating to c. 550–575, indicating a date in the first half of the 6th century for this occupation phase. Room 20, which lies immediately to the west of Room 19, may also have been used in the first half of the 6th century. A rammed clay level (5599), which may represent a surface, was found extending right across the room (see Fig. 3.67). This was the lowest excavated deposit in Room

1212

0

2m

Figure 3.49. Plan of possible early 6th-century two-storey building at eastern end of the long gallery (Room 18) showing stair block 1277, and mortar floor 1212 and 1214

20 and contained pottery dating to c. 500–525. However, no other associated features were found at this level. In the central room of the southern wing (21) there appears to have been relatively little activity, but it seems to have been during the first half of the 6th century that the substantial blocking wall (1380) was inserted into the door that allowed access to Room 21 from the central courtyard

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

82

extensive layer of broken tiles, 0.10–0.30 m in depth, that spread through the full extent of the room (Fig. 3.52). The tile contexts (1487, 3231, 5026) contained pottery dating to the late 5th century. The tiles were fragmentary, with few complete examples, suggesting that this material was dumped within the room rather than representing the actual collapse of its roof. The deposit spread through the door into Room 23 to the west, where it did appear to represent a primary roof collapse (see below context 5014). Around AD 500 or slightly later, secondary occupation horizons began to appear in Room 24, the apsidal reception room that had seemingly been kept clear of debris and deposition since the abandonment of the phase 4 (triconch) building programme some 80 years previously. A thick horizon of grey clay (1300), up to 0.30 m deep, overlay the thin layers of mortar (1694) that covered parts of the mosaic (see Chapter 2, phase 3, Fig. 2.44) (Fig. 3.53). This contained a rich assemblage of material dating from c. 500–550, as well as numerous earlier coins. It is possible that the clay derives from flooding or inundation in the same way as the silts and clays found in the long gallery (18; see above), although the assemblage of material derived from Room 24 appears to be significantly later.

Door cut through wall 1058

5272

5311

5271

5551

5579 5639

1871

tile

0

2m

Figure 3.50. Room 19 in phase 7 (early 6th century) showing possible remnant tile surfaces

(Fig. 3.51). The blocking wall was built above a clay layer (5305) that extended over the mortar mixer of phase 6, and which was equated with 5357, a silty layer above the mosaic in the portico dating to 450–475 or later (discussed above in phase 6). Layer 5305 contained coins extending into the second half of the 5th century, although the range of pottery present in 5305 probably extends into the early 6th century. Within Room 21 the relationship between the blocking and the deposits that immediately post-dated it was truncated by a cut for a later furnace or oven (1821) dating to around the mid 6th century (discussed below), although this certainly indicates that the blocking dates to the first half of the 6th century. The deposits that abutted the blocking wall on its north side in the portico cannot be demonstrated to be any earlier than AD 580. The late 5th or early 6th centuries saw the beginning of a series of roof collapses and the deposition of associated deposits of broken tile. This process commenced in the southwestern and western rooms of the complex, progressively spreading to other areas during the first half of the 6th century. In Room 22, the features of phase 5 were covered by an

The city wall It was probably at the end of phase 7 or the start of phase 8 that the city wall was erected (see Chapter 5). Unlike the excavations of the Merchant’s House to the west, the excavations of the triconch area produced no new data on the city wall to supplement that published by Andrews et al. (2004) and Gilkes and Lako (2004), although recent work by Bianchi et al. (2007; 2009) has furnished a more complete picture of the wall circuit as a whole. The construction of the fortification involved the building of a long stretch of new wall (1217), which as elsewhere in the circuit incorporated elements of earlier structures where possible (1216 and 1218). The fact that the remainder of the new wall to the south of the triconch area seems to be of a single build clearly indicates that there were no buildings between the domus and the channel as had earlier structures existed then elements of them would have been utilised within the wall circuit. What can also be said is that the construction of the wall would have partially severed the connection between the triconch area and the Vivari Channel. Although a gate (gate 2) was included to the east of the triconch (see Fig. 10.1), it is notable that this gate lay beyond the eastern boundary wall of the triconch complex and thus served the property to the east (although how long this eastern boundary wall remained standing is unknown). The west wing (Rooms 23, 25, 26, 27) (Fig. 3.54) The tile levels recorded in Room 22 were particularly evident in the rooms of the west wing, which appear to have been largely unoccupied from the late 5th century onwards. In Room 23, which had seemingly functioned together with Room 26 during the 5th century, there was no evidence of activity in this period, and a large deposit of broken tiles

3  The 5th- to mid 7th-century occupation of the triconch area

3378

1380

83

1373

West

East 1369

1375 5305

tile

0

1m

Figure 3.51. Elevation showing the blocking (1380) of the door between Room 21 and the peristyle above the late 5th- to early 6th-century deposit 5305 East-facing section across Room 22 showing tile demolition layer 5026 = 1487

Section

South

Section

5017 North

South

5136

5024 5017 5136

1452=5025

5017 5050

5050

519

1381

5136 1930

5190

5027 5087

5026 = 1487 0

2m

Interpretation South

Interpretation late 6th-century rubble

South

North wall

posthole wall late 6th-century rubble tile

late 6th-century rubble

drain capping

mortar-mixer

late 5th-century tile dump

mid to late 5th-century clay level

late 6th-century rubble tile

Figure 3.52. Section showing phase 7 tile layers in Room 22. See Fig. 3.6 for location

(5014) containing material dating to the late 5th century was found extending across the room (Fig. 3.55). The tiles, which included both imbreces and tegulae, were lying in a distinctive horizontal manner (Fig. 3.56), and almost certainly represent the collapse of the roof of Room 23. This deposit was notable for the presence of four lamps, one from each corner of the room; these had presumably lit the room in its last phase of occupation (see above, phase 5). It also extended through the doorway in wall 1378 into Room 22 to the east where it was recorded as 1487, 3231 and 5026, although these deposits had the appearance of dumped material rather than a primary roof collapse (see Fig. 3.52). The tile levels also extended through Room 26 where they were excavated as 1422 and 1841, a thick deposit of fragmentary imbreces and tegulae that also contained fragments of floor tile (Fig. 3.57). The deposit was more than 0.25 m deep and contained pottery dating to the late 5th century, the date range perhaps extending into the early 6th century. The deposit also continued into the former

western entrance (Room 25) where it was recorded as 1449, a thick layer of tile fragments and rubble that overlay the trampled occupation deposit 1490 (dating to the third quarter of the 5th century) that covered the earlier mosaic pavement. Deposit 1449 itself contained material dating to c. 450–500 and continued across the mosaic of the western portico (Fig. 3.58). Substantial tile rubble layers (5337, 5338, 5381) were also found overlying the circular storage structure in Room 27 to the north (see above Fig. 3.34) while 1449 was also recorded as continuing into the southern part of Room 27 (Fig. 4.5). Although most of the rooms of the west wing appear to have been abandoned by the late 5th century, Room 31 (one of the westernmost rooms originally added as part of the phase 4 Triconch Palace) continued in use (Fig. 3.59). A series of post-holes (5526, 5453, 5489, 5451, 5485, 5487, and 5493) were noted cutting the late 5th-century mortar bedding (1515) for pavement 1800. The largest post-hole measured 0.35 m across and 0.20 m deep, while the smallest measured 0.18 m across and 0.17 m deep. They formed a

pos hol

wall

84

East facing section through Room 24

1431 1980

1389 1300 1391 1390 1465

Interpretation

1458 1391

1442 1392 1443 1590

1458 1391

1440 1457

1114

1458 1391 1392

1300

1446 2024

1441

1447

early 11th- to late 12th-century Medieval "black soil" occupation level demolition layer

1444 1448

1295

1299

1280 1300

North 1281

1298

1299

1694

1301 0

1320 2m

construction cut

13th-century horizons and humic build-up

South

6th-century demolition blocking wall

mosaic floor

construction cut

6th-century demolition

fountain

6th-century occupation level early 11th- to late 12th-century pit/post-hole

mid 6thcentury wall

late 6thcentury wall

6th-century demolition

mortar over mosaic floor

6th-century occupation level

apse wall

7th-century backfill

North

mid 6th-century kiln lower kiln structure kiln firing platform

Firing chamber

construction cut for kiln

6th-century demolition

Figure 3.53. Section through Room 24 showing clay deposit (1300) dating to AD 500–550

0

2m

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

1114

South

3  The 5th- to mid 7th-century occupation of the triconch area

85

5337

1449

1449

1422/1841

5014

1487/3231/5026

10m

0 Figure 3.54. The west wing, showing extent of tile rubble deposits

clear L-shape (3.40 × 1.40 m) that was on approximately the same alignment as the walls of the earlier room.5 The central courtyard As noted above in relation to phase 6, the central courtyard area remained free of obstruction and deposition until towards the end of the 5th century, although a thin silt deposit (5357) had begun to develop in the south portico c. 450–475 or later. However, around the end of the 5th century it is apparent that the west portico was also seeing a certain amount of deposition. In particular, in the western portico, the tile dump 1449 (dating to 475–500) (mentioned above in relation to Room 25) seemingly extended across the colonnade stylobate of the western entrance and across the mosaic of the western portico (see Fig. 3.58). Much of the courtyard, however, remained relatively unencumbered with debris until well into the 6th century.

Summary/interpretation By the early to mid 6th century the appearance of the Triconch Palace area was changing significantly. Many of the buildings had seemingly lost their roofs by this point and the apparent unity and coherence that had been maintained throughout much of the 5th century had disappeared. The long gallery to the south (Room 18), which had formed one of the main public areas of the earlier domus, was reoccupied during this period with pits and post-built structures cut into the silt and clay deposits that had built up during the 5th century. The most substantial of these structures was at the eastern end of the gallery where a masonry stair block indicated the presence of a second floor that probably covered the area of the original entrance vestibule. Other rooms of the southern wing (19 and 20) also saw the insertion of rudimentary floor levels during the first part of the 6th century.

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

86

In the west wing, however, the intensive domestic occupation of the 5th century came to an end and the rooms became filled with large dumps of broken roof tile. The nature of most of these deposits does not indicate that they were primary collapse levels but rather that they were dumps of broken material from elsewhere in the complex, perhaps tile that could not otherwise be salvaged for other building projects. These tile deposits also extended into Room 22 of the south wing and into the western portico of the courtyard. The floor of the apsidal reception room (24), which had seemingly been kept clear of secondary deposits throughout the 5th century, became covered with a thick deposit of clay, suggesting that any function that it had retained as a reception room was lost be this point. Perhaps the most fundamental alteration that occurred in this phase was the insertion of a wall that blocked the door

5014

5014 5014

between Room 21 and the central portico, thus severing a major access route with the complex. This door had dominated the north-south axis within all the building phases prior to the Triconch Palace of the early 5th century and its closure signalled the end of the relationship between the south wing and the courtyard (and thus between the south wing and the rest of the complex). It is interesting, however, that this blocking wall was not a rough clay-bonded construction but a well built mortared wall, suggesting that some attempt was made to preserve the appearance of the southern portico. Indeed, although there are some indications of deposition within the south and west porticoes by the start of the 6th century, the central areas of the courtyard still remained free of debris even at this late stage. As with the previous phase, it is particularly unfortunate that the later stratigraphy within the east wing and the triconch triclinium was removed by earlier clearance excavations, as the attempt to maintain the courtyard could suggest that the triclinium continued to function in some way through much of the 5th century. However, by the end of the 5th century it is clear that occupation of the rest of the Triconch Palace complex was sparse and that many of the buildings were roofless, abandoned and derelict. This was the situation in the triconch area at the moment when the city wall was erected (sometime during the first half of the 6th century, perhaps around AD 525) (see Chapter 5). The wall would have fundamentally altered the relationship between the triconch area and the Vivari Channel, although the channel remained potentially

late wall cutting late wall late cutting wall cutting tile spread the tile spread the tile the spread

of the tile area of the tileof area area the tile spread 5014 spread 5014 spread 5014

0

0

0

5m

Figure 3.55. Partial plan of tile level 5014 in Room 23, with indication of the overall spread of this level

5m 5m Figure 3.56. Part of tile level 5014 (Room 23) looking east

1384

1383 1384

East

West 1841

1422 tile

1282

0

Figure 3.57. Section/elevation showing tile deposits 1422 and 1841 in door between Rooms 23 and 26

2m

3  The 5th- to mid 7th-century occupation of the triconch area

87

South-facing section across Room 25 and the western portico showing deposit 1449 Section East

West 1545 1831

1449

1545 1449

1490

1528

1832

1830

tile

0

2m

Interpretation East

West

late 5th- to early 6th-century demolition

late 6th-century demolition

late 5th-century occupation level

collapsed wall

wall

late 6th-century demolition wall

mosaic

mosaic

late 5th- to early 6th-century demolition

Figure 3.58. Section showing tile deposit 1449 in Room 25 and in western portico

accessible via two gates (see Chapter 11). It is possible that some of the quarrying of building materials that was occurring in the triconch area in the early 6th century was associated with the city wall construction, although it is equally likely that material was being removed for use elsewhere in the town. Unfortunately the dating of the city wall can only be placed within a broad period of the first half of the 6th century, and so cannot be tied to the demolition of any specific structures.

5451 5485 1800

1799

Phase 8: Early to mid 6th century (c. AD 525–550)

5487

Fishermen, blacksmiths, burials and demolition (Fig. 3.60) During the first half of the 6th century the Triconch Palace area became increasingly dilapidated. The façade of the triconch triclinium was at least partially demolished and the porticoes of the peristyle were also probably dismantled, perhaps to supply materials for the major Christian buildings that were constructed elsewhere in the city during this period or alternatively to provide materials for the city wall which was constructed in the first half of the 6th century (perhaps around AD 525) The southern wing of the complex was used primarily by fishermen. Large spreads of mussel shells appear throughout the rooms of the southern wing, as the mussels were shelled on the bank of the Vivari Channel where they were landed. The apsidal reception room was used intensively as part of this activity, attested by the dozens of post- and stake-holes found within the room. The first half of the 6th century also saw the beginnings of the area’s use for burial, with interments datable to before AD 550 noted in the south wing. The south wing (Rooms 18–22, 24) In the long gallery (18) all of the features of phase 6 were

5453

5526 5489

1515

5493

tile

0

5m

Figure 3.59. Room 31 in phase 7 (early 6th century), showing L-shape post-hole structure cutting earlier floor make-up 1515 sealed by a series of distinctive mussel shell deposits that extended through much of the south wing (see above Fig. 3.25). In the centre of the long gallery, a distinctive lens of almost pure crushed mussel shell, c. 50 mm deep (1788/5259) was found above the clay 5074. This deposit extended north through the doorway into Room 21, where it was excavated as 1847.6 Lying above mussel lens 1788/5259 was a compacted deposit of silty clay (5175) that also contained a high percentage of crushed mussel shell. The pottery contained within this deposit suggests

88

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

a deposition date to the mid/late 6th century. Layer 5175 also covered the eastern end of the long gallery (where it was excavated as 1740/5045/5058) and extended to the north through the doorway into Room 21. At the western end of the long gallery a further clay deposit (5013) also seems to date to this phase although no mussel shells were found within it. An intermittent spread of mortar, possibly representing the remains of an occupation surface, was found above this deposit. Deposit 5013 also appeared to extend to the north, through the doorway into Room 23 (cut at an unknown earlier date). The use of Room 18 for the purpose of processing mussels probably also resulted in the dumping of mussel shell deposits (1538, 1797) over the remains of the bathhouse (Rooms 28 and 32) (see above Fig. 3.9). The use of the southern wing for iron smithing began in the apsidal former reception room (24), where around 127 small pits, post-holes, and stake-holes were noted cutting the grey clay layer (1300) that from the start of the 6th century had begun to build up over the earlier mortar surfaces (Figs 3.61, 3.62). Obviously not all of these postand stake-holes date from the same phase of occupation, and it is possible that some were cut from a higher level but were not recognised. However, it is possible that a rectangular post-built structure can be discerned in the northwest corner of the room (Fig. 3.63), while other lines and groups of post-holes may represent partitions. There is also evidence of a rough masonry wall (2024) built above layer 1300, with no visible foundation (Fig. 3.53). This wall runs east–west and approximately follows the line of the earlier wall (1696) that formed the northern end of Room 24 prior to the construction of the apse. There seems therefore to be a separation between the apse area and the rest of the room. This may be related to the use of the apse to house a hearth or furnace, which was clearly used and rebuilt on a number of occasions (Figs 3.64, 3.65). The earliest of these was a small limestone structure (1561) built above layer 1300, which was filled with burnt clay, ash and charcoal (1562). A rough surface of tiles and limestone (1564) associated with this structure was then overlain by a further small hearth made of tiles and clay (1573) that may have replaced the original oven (1561). A second working surface (1554) that overlies the earlier surface (1564) is probably related to this second phase of hearth. The fragmentary remains of what may be another rebuilding of this furnace (1524, 1529) overlie a series of thin burnt deposits associated with the earlier structures. The primary evidence for the use of this feature is hammerscale from the ash and charcoal deposit 1562, indicating blacksmithing. This sequence of furnaces covers a substantial area (c. 3.00 × 2.40 m) within the apse, and it seems that the apse, which was presumably covered with a semi-dome, formed a good sheltered location for a series of blacksmiths’ hearths. The rough wall 2024 may have acted as a means of partitioning this working area off from the makeshift (and combustible) structures that were erected in the rest of the

room. We may imagine perhaps a room that was occupied seasonally, possibly because its walls survived to a greater height than other parts of the complex, where the interior had a series of temporary shelters erected within it. These were abandoned at the end of the season and repaired or replaced at the start of the following year, resulting in the dense palimpsest of post-holes and ovens revealed by the excavations. The furnaces in the apse were overlain by a rubble deposit (1280) that contained pottery dating to the first half of the 6th century, while the post-hole sequences were beneath a layer of dumped tile and building material (1392) that may be of a slightly later date (Fig. 3.53).7 Elsewhere in the south wing, as noted above, there was also evidence of this intermittent use of the remains of the domus by fishermen. In Room 21 (the central room of the south wing) the mussel-rich deposits noted above (1847) extended through the doorway from the long gallery (Figs 3.66, 3.90) and as 5270 extending beneath the later door blocking 1379 into Room 20 (Fig. 3.98). Mussel rich silt deposits 1820 and 1826 were also noted as overlying the late 5th-century tile deposit 1487 (see above) that extended from Room 21 into Room 22.8 In Room 22 itself, spreads of mussel shells (1486, 3172) extended across much of the excavated area, where they overlay the dumped tile deposits of the late 5th or early 6th century (1487, 3231, 5026), although they did not appear in the section on the western side of the room (Fig. 3.53). In Room 19, occupation may have been slightly less ephemeral, indicated by a line of three post-holes (5528, 5530, and 5532) almost parallel to wall 1879 on the eastern edge of the north half of the room (Figs 3.67, 3.68). The post-holes cut a thin late 5th-century occupation horizon (5271) and were seemingly beneath a clay level (5403) dating to between 500 and 550. However, given the shallow depth of these post-holes (c. 0.14 m) it is very likely that they belong to one of the medieval post-built phases noted in higher levels of Room 19.9 Room 20 was the only one of the rooms of the southern wing where burials were noted for the period prior to AD 550. Two graves were found cutting into the rammed clay level 5599 (which dated to c. 500–525) (Fig. 3.69. Grave 5634 (Fig. 3.70) was that of an infant or neonate, buried within a 6th-century Gazan amphora that was placed within a cut lined with vertically placed limestone slabs and with a base of tile and limestone. The cut was 0.80 m long, 0.40 m wide and 0.25 m deep. The amphora had been sawn or broken off at the shoulder and the body placed in head first. The second grave (5636) was that of a child of c. 2–5 years buried alla cappuccina (Fig. 3.71 and 3.72). The tomb was composed of a north–south oriented cut 1.4 m long, 0.60 m wide and 0.27 m deep, the base of which was lined with a pair of large tegulae on which the body was placed before it was covered using large imbreces. Although Room 20 was used for burial it was also used for similar purposes to the other rooms. There was evidence of secondary structures above the graves in the

Room 17

Room 14

1193

1152

3  The 5th- to mid 7th-century occupation of the triconch area

5382 2000

Room 29 Room 31

1523/1536 2096

1564 1554 1301

2095

1486

1561

1573

Room 23

Room 22

Room 21

3172

1847

5634 Room 20 5599 5565 5592

Room 19

5636

Room 24 1300

5637

5528 5530 5271 5532

Room 18 5013

5259

5175

5259

1788

5074

5045/5058 1740

0

20m

Figure 3.60. The triconch showing areas of known activity in phase 8 (c. 525–550) 89

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

90

form of post-holes (5565, 5592, 5637) and the dumping of mussel shells. A large spread of shells (5522) was found in the eastern half of the room, covering one of the graves (5636) and one of the post-holes (5637). One of the other post-holes (5565) was seen to cut this deposit. At the same time, the wall dividing Rooms 20 and 21 (1375) seems to have partially collapsed or been demolished, evidenced by a large spread of compact mortar and occasional rubble (5566). Although most of the usable stone had been removed, pieces of mortared limestone masonry were still clearly visible in the deposit. The whole sequence lay beneath a large spread of dumped tile rubble (5509) in which the latest pottery dated to c. 550–575.

post-built post-built structurestructure post-built structure

2024

1300

1300 1300

0

0

5m

Figure 3.61. Pits and post-holes cutting the0early-mid 6thcentury layer 1300 in Room 24

Figure 3.62. Post-holes in Room 24. These were almost certainly mainly cut through layer 1300, but the smaller ones were only detected when the mosaic was cleaned

The west wing (Rooms 23, 29 and 31) In Room 23 large spreads of clay and mortar (5006, 5018) were deposited above the tile dump/roof collapse 5014 (see above phase 7). These deposits were found throughout the room and may derive from the demolition or robbing of the walls, or alternatively from the collapse of walls made of rammed and plastered clay, a building technique that may have been used in much of the complex. Walls built in this technique would have collapsed relatively soon after the 5m roofs had collapsed or been removed. The deposit contained 5m pottery dating to the first half of the 6th century. The only other feature noted in the west wing for the first half of the 6th century was a single amphora burial (5382) in the eastern half of Room 29 adjacent to wall

Figure 3.63. Possible post-built structure in Room 24

3  The 5th- to mid 7th-century occupation of the triconch area 1355 and cutting into the mid 5th-century levelling deposit (5364) (Fig. 3.73). The amphora (5384) (probably Cretan and dated to c. AD 500) had been broken to allow the interment of an infant, and inserted into a cut 0.54 × 0.37

91

× 0.23 m. It was positioned in the grave cut using two pieces of slate or schist for support. The broken pieces of amphora had then been laid over the body as a cover, and the cut had then been carefully filled with three layers of

1561 1573

1564

1554 1301

2024 0

4m

Figure 3.64. Furnaces in apse of Room 24

3172

Figure 3.65. Furnace 1561 and related structures in the apse of Room 24 (1 m scale)

Room 21 1486

1847

Room 22

Room 18

0

5m

Figure 3.66. Plan showing early to mid 6th-century mussel shell deposits in Rooms 21 and 22

92

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

5634

5528

Room 20

5565

5592 5522

5530

5599 5271

5637

5532

Room 19

5636

0

5m

Figure 3.67. Rooms 19 and 20 in phase 8 (c. AD 525–550)

Figure 3.68. Room 19 showing post-holes 5528, 5530, 5532, looking east (2 m scale)

3  The 5th- to mid 7th-century occupation of the triconch area

93

5641 5632 5634

5634

0

5565

1m

5599

5631 5592

5637

5636 5636

0 0

5618

1m

5m

Figure 3.69. Graves in Room 20 belonging to phase 8 (c. AD 525–550)

Figure 3.70. Grave 5634

Figure 3.71. Grave 5636 prior to excavation (1 m scale)

94

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

Figure 3.72. Grave 5636 after removal of the cover (1 m scale)

Figure 3.73. Amphora burial 5382 in Room 31 (1 m scale)

2000

2096

2095

0

10m

Figure 3.74. Overall plan of central courtyard showing location of drain fill 2000 (which contained the windows from the façade of the triconch) and the developing early to mid 6th-century deposits in the porticoes

3  The 5th- to mid 7th-century occupation of the triconch area

95

roof tile (5376). The possibly early date of the amphora could indicate that this burial dates to around the start of the 6th century (therefore belonging in phase 7). Elsewhere in the west wing there was relatively little activity, although a rather thick (0.15 m) and compact silt horizon (1523, 1536) in Room 31 may represent a final stage of late antique occupation. The deposit was found throughout the western half of the room, overlying the postholes of phase 7 and contained an assemblage of material dating to c. 525–550.10 The central courtyard (Fig. 3.74) Although the central courtyard and portico may have retained its coherence until the start of the 6th century, it is clear that between 500 and 550 major changes occurred. The drain that ran along the front of the east portico (which seems to have been a major city drain) was found to be choked with debris from the collapse or demolition of the portico and the façade wall of the triconch triclinium (Fig. 3.75). The drain fill (2000) included large pieces of tile (c. 0.40 × 0.30 m), which had perhaps fallen from the portico roof, fragments of column and capital and, most significant of all, the four stone windows that had embellished the triclinium façade (see above). Pottery from this fill dated variously to the 4th, 5th and 6th centuries, although a sherdjoin between 2000 and the major deposits over the north wing (see below for context 1152) could suggest a date c. 525–550 for the drain fill.11 At around the same time a thin trampled layer (2096) containing mussel shells developed in the eastern portico and sealed the stylobate wall of the eastern portico colonnade. This was overlain by a mid 6th-century rubble deposit (2095) (see below). The north wing In the north wing several burials were found in Rooms 14 and 17 that were originally dated to the first half of the 6th century and were previously published as such (Gilkes and Lako 2004: 170–72), given that they appeared to lie beneath a substantial deposit of broken tiles (1193) that was equated with the major tile and pottery dump (1152) in Room 14. However, radiocarbon dating on one of these burials produced a date of AD 880–1020 and as such these burials are now placed in phase 12 (see Chapter 4). This suggests either that the deposit of broken tiles (1193) should not be equated with 1152 or that the tops of the grave cuts were missed during the excavation (see Fig. 4.23).12 In Room 14 the principal event was the deposition of a thick layer of tiles and domestic rubbish (see Fig. 3.45). Although individual tips were identified within the deposit the number of sherd joins between the individual contexts suggest that the whole deposit should be treated as a unit (1152) (see Reynolds 2004 for detailed discussion of this context). The deposit extended through the colonnaded entrance of Room 14 and into the eastern portico (where joins were noted with the deposit filling the drain in front of the eastern portico (2000, see above). The deposit contained a vast assemblage of pottery dating to the first half of the

Figure 3.75. Drain in front of east portico with rubble (2000) visible beneath arch (1 m scale)

6th century, but it is most likely that it dates to 525–550 (and probably to the latter part of this period). Room 14 also contained one grave (1160) recorded as being cut into 1152 but overlain by a further deposit (1133) that was subsequently combined with 1152 due to pot joins. It is therefore unclear if the grave was inserted during this period of dumping, or if the cut through 1133 was not recognised (as was probably the case with the graves in Room 17). The grave was lined with rough limestone slabs, and is thus typologically similar to the later (10th century) burials in Room 17. As such it has been provisionally placed in phase 12 with the Room 17 burials. Summary/interpretation By the second quarter of the 6th century the structural coherence of the Triconch Palace had been lost and the complex was to all intents and purposes abandoned as an area of habitation, although it is clear that occupation was continuing nearby (for example in the Merchant’s House area). Equally, the sequences at the Triconch Palace can clearly not be extrapolated outwards into the rest of the city as this is precisely the period when the major Christian monuments such as the Baptistery and the Great Basilica were constructed.

96

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

There are few signs of permanent use of the Triconch Palace buildings and the archaeology instead points to intermittent (perhaps seasonal) use of the parts of the building that were adjacent to the Vivari Channel. 13 Fishermen and women were landing their catches of mussels on the banks of the channel and cooking and shelling them on the spot, before taking the cooked mussels to sell elsewhere. Exactly the same activity can be seen on the banks of the Channel today and the waterfront of the present site of Butrint is covered with the same sort of layers of mussel shells that are found in the 6th-century levels of the Triconch Palace. The fishermen were simply using the nearest ruined buildings as shelter and discarding the waste shells around them. These deposits of shells show that despite the construction of the city wall (see above phase 7) the waterfront of the Vivari Channel continued to be accessible from the triconch area, suggesting that the gate in the Merchant’s House area (gate 1) and the gate to the east of the triconch complex (gate 2) remained generally open and passable. The apsidal reception hall (24) was perhaps standing to a slightly greater height by dint of its more robust construction, and consequently a lengthy series of temporary structures were erected within its walls, probably to house a continuing series of blacksmiths’ hearths, evidence for which was found in the apse. The multiple hearths and postholes suggest that this activity may have been seasonal with structures and hearths abandoned during the winter months and repaired or rebuilt in the following season. Other rooms within the complex also had wooden structures erected within the confines of their walls although none appears to have been particularly substantial or long-lived. Other parts of the building were seemingly quarried for building materials during this period, particularly for columns, capitals and any marble veneers that still remained. The eastern portico and the façade of the triconch triclinium were certainly demolished in this phase. The roofs were stripped of any usable tile (it is noticeable that almost no complete tiles were recovered during the excavation) and any material that couldn’t be used elsewhere was simply dumped in the abandoned buildings. It was also in the first half of the 6th century that the first graves appeared in the complex. These graves belonged to infants and children, although we should not imagine a situation comparable to some early Roman contexts where the respect accorded to child burials was less than that given to adults. The children’s graves in the triconch were constructed with some care, and effort was made to protect the corpses through the use of tile and stone covers. It is striking, however, that the commemoration of the sites of these graves appears relatively short and that probably within a generation the areas where bodies were interred were used for the dumping of building rubble and mussel shells.

Phase 9: Mid to late 6th century (c. AD 550–575) Industrial activity in Rooms 18, 21 and 24 (Fig. 3.76) In the second half of the 6th century the remains of the southern wing were used intensively for industrial activity, with a series of furnaces inserted into the former long gallery. These were used for a variety of purposes which included blacksmithing, indicated by hammerscale found in contemporary deposits. Similar activities were also carried out within the shell of Room 21, while a large furnace was also constructed in the apse of the earlier reception Room 24. Away from the southern wing, however, there appears to have been little activity apart from the continued insertion of burials and the sporadic dumping of rubbish. The south wing (Rooms 18, 21 and 24) In the second half of the 6th century, a series of furnaces were constructed in the remains of the long gallery (Room 18). At the western end of the room were four separate but adjacent furnace structures (Fig. 3.83). A large openended rectangular structure (5181), made of rectangular limestone blocks and tile fragments, was built in the northwestern corner of Room 18. It was 0.60 m in width and survived to two courses (c. 0.50 m) in height. A layer of clay (burnt orange) c. 0.3 m thick was used to bond and line the structure. Structure 5181 partially obstructed the doorway from Room 18 into the apsidal room (24) to the west (Figs 3.77, 3.78, 3.83). At the southern end of the furnace were what appeared to be the remains of a semicircular stoke pit (5196), with a rounded base and steep, almost vertical edges. The pit contained a fairly compacted, black charcoal fill (5188) containing a little early 6th-century pottery. This fill almost certainly represents the final residue or rake-out from furnace structure 5181. The confirmed use of the adjacent furnace (5088) as a smithing hearth would support a similar interpretation for furnace 5181. The second furnace structure (5088) identified at the western end of the long gallery was located approximately 1.6 m to the east of furnace 5181 (Figs 3.79, 3.80). Two parallel walls, 0.40 m apart and up to 1.20 m long, butted the remains of the north wall of Room 18. On the south side the structure appeared to be open-ended. The furnace walls were constructed of rough limestone and travertine blocks. As with furnace 5181, clay (burnt orange in the firing process) had been used to line the structure and to bond the stones together. All the stones showed signs of intense heat and many were fractured as a result. One of the fills (5021) within the furnace contained numerous curved pieces of fired clay – possibly the remains of kiln furniture. The upper fill of the furnace (5089) (above 5021) contained ferrous hammerscale indicating its use for iron smithing. A further multi-layered feature of burnt clay and stone abutted the eastern side of furnace 5088 (Figs 3.80, 3.81,

3  The 5th- to mid 7th-century occupation of the triconch area

5362

5574

2071

2077

5356

1479

1298 1444

Room 24

1821

5181 5088 5181 5217 5013 5200

5156

Room 21 5154

5178 5162

5155

10007

Room 18

1740/5045 1796

5166

0

20m

Figure 3.76. The triconch showing areas of known activity in phase 9 (c. 550–575) 97

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

98

3.82). This was a linear mound, measuring 1.6 × 1 m and aligned east–west, alongside the north wall of Room 18. Two truncated amphora necks had been set vertically upright on the surface of the lowest fill (5157) and were supported on small tile platforms. The positioning of these pieces of pottery and their association with furnace deposits suggests that they were being utilised to support materials that were being worked. The amphora pieces appear to have been utilised in the same manner as the

other examples found within the demolished remains of contemporary furnace deposit 5202, located in the southwestern corner of the room (see below), as well as within the phase 7 pit (5139) located in the centre of the room (see above). The amphora supports placed on the surface of deposit 5157 were covered by a reddish brown clay (5100) containing mortar, tile and abundant mid 6th-century pottery fragments, although the very tops of the amphora necks protruded through the surface of the

Figure 3.77. Furnace 5181 at east end of Room 18 (2 m scale) East facing elevation of kiln 5181 at the western end of room 18 1282

5000

1976

South

1975

5407

5013 5204 1771

5010

1377

5181

North

5212 1809 1m

0

tile

Interpretation

late 6th-century door blocking associated with kiln 5181

late burial cut 3rd-century wall

South

3rd-century wall

3rd-century wall

3rd- to 4th-century door blocking

3rd-century wall

mid 6th-century levelling late 6th-century kiln

late 5th century abandonment 3rd-century mosaic

mid 5th-century rubble

early 5th-century mortar floor

wall plaster

Figure 3.78. Side elevation of furnace 5181 and deposits in doorway between Rooms 18 and 24

North

3  The 5th- to mid 7th-century occupation of the triconch area clay, which is probably a collapsed or demolished clay and tile superstructure or platform. The whole feature probably represents a subsidiary structure associated with the adjacent, contemporary furnace 5088. The remains of a fourth possible furnace structure were located in the southwestern corner of the room (Fig. 3.83). The feature consisted of a deep oval pit with vertical sides (5217) containing a mottled orange clay fill (5202). The fill was jumbled and tightly packed and contained a very large quantity of late 6th-century pottery and amphorae fragments. Within the fill was another example of a truncated amphora neck, which had been set on a tile platform in the same manner noted above. A compacted layer of limestone and tile rubble (5178) was found surrounding the furnace structures at the western end of the long gallery (18) and is thought to represent the remains of a rough floor surface associated with a small workshop or processing factory (Figs 3.81, 3.83). The stone and tile pieces within 5178 appeared to have been set closely together in a single layer to form cobbles. The cobbled surface (which contained mid 6th-century pottery) gradually disappeared to the east. On the south side of the room there was evidence for the repair of the cobbled surface with patches and spreads of mortar and gravel. A large pit (5200), probably associated with the furnaces and containing early to mid 6th-century pottery, was cut into the phase 7 clay layer 5013. Approximately 2 m to the east of 5200 was a second, shallower circular pit (5166) containing charcoal, tile and 5th- to 6th-century pottery. In the centre of Room 18, the first feature post-dating the mortar and mussel shell deposit 5175 was an oval pit (5154) that was cut against the north wall of the room. The pit was 2 × 1.6 m and orientated north–south. Its fill (5155) contained a large quantity of tile fragments, as well as early 6th-century pottery, mortar lumps, limestone, animal bones, crushed mussel shell and large pieces of iron slag and nails suggesting that it functioned as a general refuse pit within the room.. A further large furnace (5156) was constructed in the centre of the long gallery (Figs 3.84, 3.85, 3.86). This was one of the best preserved and most substantial furnaces of the whole workshop complex and consisted of an oval structure that was constructed against what must have been the remains of the north wall. The furnace cut through layers 5175 and 5074 and into the fill of pit 5154 (suggesting that it was one of the later features of this phase). Two substantial stone and clay walls (5156), 0.40 m wide, were constructed within a foundation cut (5194). Three courses of these walls survived to a height of 0.35 m. The two side walls of the furnace were 0.60 m in length and aligned north–south, perpendicular to the north wall of the room. In common with furnaces 5181 and 5088, the southern end of furnace 5156 appeared to be open-ended, although the flat base of the furnace sloped upwards to meet the contemporary ground surface at this end. The inner edges of the furnace were lined with clay, 0.4–0.5 m thick, which was burnt orange and which retained the impressions of the hands that had shaped the lining. The lining did not extend across the base of the structure.

99

Figure 3.79. Furnace 5088, looking north

Within the furnace, the lowest of three fills (5171) contained pieces of burnt limestone, mortar, 6th-century pottery and pieces of corroded iron, including nails. Like the features noted above, this fill also contained hammerscale derived from iron smithing (J. McDonnell pers. comm.). The middle fill (5159) containing a high percentage of crushed mussel shell, fragments of tile and pottery and a large, curved fragment of fired clay – possibly part of a pithos. The uppermost fill of the furnace (5126) also contained a large quantity of crushed mussel shell, as well as white flecks of mortar, limestone, tile, mid to late 6thcentury pottery and some iron slag. Contemporary with upper fill 5126 but surrounding the furnace was a dark spread of charcoal-rich silty clay (5071) that also contained fragments of limestone and mortar, as well as minute, trampled flecks of tile, mussel shell and 5th- to mid 6th-century pottery. Layer 5071 extended across the width of the room, and spread approximately 1–1.5 m on either side of furnace 5156. A notably large quantity of iron nails (1.8 kg) and iron slag (265 g) was contained in this deposit, which probably represents refuse and working residues associated with the use of furnace 5156, probably for iron smithing. To the east of the central doorway into Room 21, there was further possible evidence for industrial activity in the

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

100

5100 Amphora

5100 Amphora

5157

5178 5157

Furnace 5088

5178

5357 5357

1m

0

10m 10m

00

South-facing section through kiln 5088 in room 18 Kiln 5088 5021 West West-facing section across the Triconch peristyle (Room 16) West-facing section across the Triconch peristyle (Room 16)

North North 1875 1875

5005

Amphora

5004

5174

5100

5157 18 5005 South-facing section through kiln 5088 in room West

1m

0

5021

5013

Furnace 5088

tile

1833 1833

0

Amphora

South South

5357 5357

1957 1957 00

East

East

5178 1m

5004

5m 5m

Figure 3.80. 5174Plan and section of furnace 5088 and adjacent mound 5100of burnt clay with amphora necks Interpretation Interpretation

North tileNorth

5157

Kiln 5088

early 5th-century 5th-century northern northern early stylobate wall wall of of the the Triconch Triconch peristyle peristyle stylobate

5178

5013

mid to to late late 5th-century 5th-century mid occupation1m level South occupation level South

0 early 5th-century 5th-century early domus peristyle peristyle domus

early 5th-century 5th-century southern southern early stylobate wall wall of of the the Triconch Triconch peristyle peristyle stylobate 00

Figure 3.81. Mound with amphora necks to east of furnace 5088 and surface 5178 (1 m scale)

5m 5m

3  The 5th- to mid 7th-century occupation of the triconch area

101

Figure 3.82. Detail of amphora necks adjacent to furnace 5088

5100

5088

5157

5181 5196

5178 5013

5217 5166 5200

0 Figure 3.83. Room 18 showing furnaces and surface 5178 and associated features

2m

102

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

form a mound of burnt clay, limestone and tile rubble, which may represent the demolished remains of another furnace lying directly above mussel spread 1788. The furnaces appear to be associated with the ephemeral remains of a structure noted at the eastern end of the long gallery. This was represented by an L-shaped ridge with a 100–150 mm deep rectangular impression on one side and with a higher ground level maintained to the north and east (probably the inside) of the structure. This structure possibly represents the remains of a small shelter built within the confines of Room 18 or, more likely, an L-shaped timber partition that divided domestic and industrial activities in the room (10007) (Fig. 3.76). This division is suggested by the absence of furnaces and industrial deposits at the eastern end of the room and by a semicircular feature of tiles and slate (1796) placed within a slight hollow on the

Figure 3.84. Furnace 5156, looking north (1 m scale)

surface of the phase 7 mussel deposit 1740. This feature possibly represents the remains of a small, domestic corner hearth or oven. A small concentration of 11 small coins (of which the latest identifiable dated to AD 498–565) on the surface of silt 5045 may also be associated with this domestic usage. The remainder of the eastern end of Room 18 was covered by a deposit (1273) that contained compacted mortar and small stones, which may represent a rough cobble floor. This deposit covered the remains of the phase 7 structure represented by the masonry stair block (1277), showing that it had gone out of use by this time. In Room 21 two contemporary furnaces or ovens (1479 and 1821) were also constructed during this period (Fig. 3.87). Furnace 1479 was constructed in the northwest corner of the room and was an irregular construction of limestone blocks 1.16 m long × 0.79 m wide (Figs 3.88). It was cut into a rubble deposit dating to the early to mid 6th century (1810) that overlay the early 6th-century rubble deposit 1827/5305, and was filled with a series of ashy and burnt deposits (1477). Furnace 1821 was a more substantial construction built on the eastern side of Room 21 abutting the (now partially demolished) remains of wall 1375. A deep cut, 2.45 m long and 1.30 wide, through the underlying rubble layers (1826 and 1827) was filled with a rectangular masonry structure bonded and lined with clay (Figs 3.87, 3.89, 3.90). This structure was left open at its eastern end. A series of ash and charcoal levels are sealed by a rubble deposit (1836) that seems to mark the final collapse of the furnace. In both Room 18 and Room 21 the end of this phase was marked by distinct tips of rubble and domestic refuse and the insertion of burials. In the central and eastern areas of

Figure 3.85. Furnace 5156 sectioned and shown in relationship to pavement levels in Room 18

5156 5071

Profiles through furnace 5156

5071

1m

0

tile

5156

5194

West 5259

5194 East

5175 5074 5140

1809

5155 5154

5156 North

5155

1368

5071 South 5175 5259

5074 5140

5156 5071

tile

1809 0

1m

0

tile

posthole

Figure 3.86. Plan and section of furnace 5156 Profiles through furnace 5156

5071

5156

5194

West 5259

5194 East

5175

1809

1479

5074 1810/1826 5140 Section

5155

5154

5156 North

1368 5074 5140

tile

posthole

1821 5071 South 5175 5259

Figure 3.88. Furnace 1479, looking west (1 m scale)

1809 1m

0

0

2m

Figure 3.87. Room 21 in phase 9 with furnaces 1479 and 1821 (showing location of section Fig. 3.90

Room 18 an extensive deposit (1676/5056/5057) of dark grey-black mixed deposit of clay, silt and sand, containing tile, limestone rubble, mortar, charcoal and burnt clay, seemingly marked the end of industrial activity in the long gallery (Figs 3.25, 3.91). Sherd joins with 1676 were also noted in deposits in Room 31 (1473, 1516, 1523) although

Figure 3.89. Furnace 1821 (looking east) showing southern half upon completion of excavation in 2003. The northern wall of the furnace was (unfortunately) removed in 2000 during the excavation of the northern section of the room to create a complete section through the room and the furnace (shown in Fig. 3.90) (1 m scale)

1m

Figure 3.90. Section of furnace 1821

mussel depsoit

early 6th-century trample layer construction cut for furnace rubble layer

mussel deposit ashy fill

mid 6thcentury furnace 3rd-century wall East

Interpretation

tile

primary fill primary fill cut predating the furnace flue?

ashy fill late 6th-century furnace collapse

1854

1944 1956

clay lining

late 6th-century demolition

mid 6th-century demolition/collapse/mussel deposit

trample layer

1821 1870

1837

1848 1945

1836 1818 1375 East

North-facing section through furnace 1821 in Room 21

1828

1847

1827

3rd-century wall

mussel depositt rubble

early 6th-century tile dump

1m 0 1982 1847

1805 1826

1415

early 6th-century demolition

West

1847 1870

1487

1893

1805

West

trample layer

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

104

this may be a result of post-depositional disturbance. At the western end of Room 18 was 5004, a deposit of similar character and date to 1676 noted above. A similar deposit (1415) sealed the two latest furnaces to the north in Room 21 (see above Fig. 3.90). Both 1676 and 1415 contained a large number of 5th- to 6th-century coins and pottery dating to c. 550–575. Also in the western half of the long gallery (18) a deposit of loose, orange-yellow sand (5136) (beneath 5057) containing limestone, mortar lumps and late 5th- to late 6thcentury pottery was deposited across the centre of the room. This deposit extended across the northwestern half of the room and into the southern half of Room 22. Significantly, 5136 partially overlay the demolished remains of part of the north wall of the long gallery (5190) and therefore must represent at least the partial demolition of that wall, close to Room 22. 5136 may equate with rubble deposit (1452/5025), which overlay the mid 6th-century mussel shell deposit 1486 in Room 22 (see Fig. 3.52). Around the middle of the 6th century a further possible furnace was also built in the apse of the former reception Room 24. A substantial cut (1460), 3.4 m in length , was made in the rubble layer (1280) that sealed the previous phase of furnaces and which dated to the mid 6th century. This cut was used to construct a masonry structure (1298), built of limestone bonded with clay (Figs 3.92, 3.93). The actual furnace structure was 1.60 m in length with the remainder of the cut forming an possible rectangular firing chamber. A channel or flue was also cut through the masonry of the apse itself in order perhaps to increase the draw on the furnace. It is hard to see how this structure actually functioned although burnt materials from the fills and surrounding area suggest an industrial function, which was how it was interpreted by the excavators. At the same time a rough masonry wall (1444) was built approximately along the line of the earlier partition wall (1445). No terminus ante quem for the abandonment of this furnace is available, as no dating evidence was recovered from the fills or overlying levels. The central courtyard (Fig. 3.94) It is clear that the central courtyard saw little activity in this period beyond the deposition of waste material and the accumulation of natural silts. Pottery of the mid 6th century and after was recovered from a thick gritty clay deposit (5356) up to 0.40 m deep in the southern portico, while a major deposit of silty clay (5362) was recorded as overlaying the flagstone courtyard of the earlier peristyle and all the earlier wall foundations within the area defined by the later portico stylobates. During excavation it was thought that this represented a make-up level associated with the construction of the phase 4 Triconch Palace, but in fact much of the ceramic assemblage dates to the mid 6th century and after. Sterile silt deposits (5573 and 5574) were found overlying 5362, which may represent natural accumulation resulting from wind and rain action. These in turn were overlain by a rubble deposit (2071/1988/2077/2095) containing pottery of the mid 6th

3  The 5th- to mid 7th-century occupation of the triconch area

1452=5025

105

1415

Room 22 Room 21 5136

5004 Room 18

1676/5056/5057

0

10m

Figure 3.91. Plan showing extent of rubble deposits 5136, 1676, 5056, 5057, 1415 and 1452/5025 that apparently mark the end of the industrial activity in the south wing c. AD 575

century and later. These deposits probably resulted from periodic demolition interspersed with the casual disposal of rubbish. Layer 2095 (Fig. 3.74) contained pottery with sherd joins with the major phase 8 deposit 1152 found in the north wing (see above), giving some indication of the extent to which material was being moved around the site during this period.14

We may perhaps see the Triconch Palace in this period as a quarry and rubbish dump that was seasonally used (perhaps by those in the adjacent Merchant’s House area) owing to its position adjacent to the Vivari Channel.

Summary/interpretation It is clear that the southern part of the Triconch Palace area adjacent to the Vivari Channel continued to be used during the second half of the 6th century for a variety of industrial purposes, which certainly included blacksmithing, the working of carbon steel and complex activities that may have involved the use of copper-based flux in welding (De Rijk in prep.).15 The presence of multiple furnaces in the long gallery and in Rooms 22 and 24 is perhaps best interpreted as evidence of seasonal usage, with furnaces abandoned at the end of the season, and either reused the following year, or rebuilt if they had fallen too far into decay. The evidence tends to suggest a rather piecemeal usage over a number of years rather than intensive industrial activity. The buildings themselves were probably totally ruinous and only partly standing by this point. By around 575 some of the northern wall of the long gallery was standing at around its present height, evidenced by deposits that clearly ran over the demolished remains. The central courtyard area meanwhile, saw the accumulation of natural silts with some deposits that may result from demolition and quarrying.

During the later 6th and early 7th centuries the primary use of the triconch seems to have been for burial. The west wing and central courtyard show continued evidence of demolition and robbing of walls, the only evidence of possible habitation consisting of a mortar floor in Room 29.

Phase 10: Late 6th to early 7th century (c. AD 575–650) (Fig. 3.95)

The west wing (Rooms 25, 26, 29) The abandonment of the western range and the robbing and demolition of the walls that began around AD 550–575 continued in the late 6th century, with the exception of a short period of activity in Room 29, where a thin spread of mortar (1436), probably a surface, was laid down. This surface was supported by a make-up level (1439=5331) (see Fig. 4.42), which in turn covered the early to mid 6th-century amphora burial (5382). The make-up layer (1439=5331) contained 34 coins, the latest of which dated to AD 582–602, while the mortar spread itself contained coins of which the latest was a half follis dated to AD 583–584. Both levels also contained other domestic detritus including animal bone and glass, together with late 6th- to early 7th-century pottery (from 5331). The mortar surface also contained the disarticulated skeletal remains of a baby

William Bowden, Karen Francis, flueOliver Gilkes, and Kosta Lako

106

flue

1298 1280

1298

1280

1444 1444

0

2m

0

2m

East-facing section through furnace 1298 in Room 24 South East-facing section through furnace 1298 in Room 24 1114 South 1280

North North 1281

1114 1295

1300

1280 1694

tile 1300

1694

1299 1460 1299

1295 1301

1298

1301

0 1298

1460

Interpretation tile South Interpretation

1281 1m 1m

0

North

South

North

13th-century horizons and humic build-up 7th-century backfill 13th-century horizons and humic build-up mid 6th-century furnace

apse wall

7th-century backfillfiring chamber lower furnace structure firing mid 6th-century furnace platform firing chamber lower furnace structure construction cut for furnace mortar floor firing platform 6th-century occupation level

apse wall

6th-century demolition mortar floor 6th-century demolition 6th-century occupation level

Figures 3.92 and 3.93. Section through furnace 1298 and plan showing location in apse in Room 24 construction cut for furnace

3  The 5th- to mid 7th-century occupation of the triconch area

107

5362

5574

2071

2077

5356 5355

10m

0

West-facing section across Room 16 North

5573

5574

5362 1833

1875

5355

5356

1830 South

5357

1957 0

Interpretation natural accumulation wind/rain North

early 5th-century northern stylobate wall of the Triconch courtyard

6th-century silt early 5th-century domus courtyard

5m Medieval demolition

mid 6th-century clay deposit

early 5th-century southern stylobate wall of the Triconch courtyard 0

Figure 3.94. The central courtyard and deposits in the porticoes in phase 9

early 6th-century door blocking South mid to late 5th-century occupation level 5m

108 William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

5362

Room 25

1783 5574

Room 26

5355

1841 1496

3075 1454

Room 23

5006

Room 22

3158

3123 3123

Room 24

1435 5535 1415/5264 1379

Room 21

5511

Room 20 5509

5534

5508

Room 19 5480 5553

5073 5109

Room 18

1238 5044

1234 1758

1269

0

Figure 3.95. The triconch showing areas of activity in phase 10 (c. AD 575–650)

20m

3  The 5th- to mid 7th-century occupation of the triconch area

109

and the lower jawbone of a young child. After this period of activity, Room 29 may have been wholly or partially demolished, indicated by a succession of rubble deposits (1417, 1416/5329) (see Fig. 4.42). In the east entrance to Room 25 an infant grave (1783) was cut into (1545) (a deposit overlying the phase 9 tile dump 1449), which also extended into Rooms 26 and 27 and contained limited pottery from the mid to late 6th century (see also Fig. 4.5). The grave was oval in plan, U-shaped in profile with a flat base and was oriented northeast–southwest. Its dimensions were 0.67 × 0.14 m and it was cut to a depth of 90 mm. The skeleton was placed in the centre of the grave in a supine position (Fig. 3.96). Its shallow depth suggests that the grave may have been cut from a slightly higher level, although the presence of the east stylobate of Room 25 (which the body was laid on) clearly restricted the depth of the grave. Wall 1567 was later constructed over the top of the burial, blocking the east entrance to Room 25. Wall 1567 may have been built during this phase but is more likely to be later. The central courtyard (Room 16) Little activity is discernable within the central courtyard during the later 6th and 7th centuries, although it is notable that the central space remained free of burials in this period. The south wing (Rooms 18–24) (Fig. 3.97) During the late 6th century the south wing also saw a mixture of demolition and silting, suggesting that the buildings were left to decay, while perhaps being periodically quarried for stone. There are, however, signs of sporadic occupation. In Room 19, a silt layer (5553) containing pottery of c. AD 550–575 developed in the southwest corner of the room. This was cut by a large feature (5480), which probably belongs to this phase. This was an irregular cut of unknown purpose some 2.20 m in length placed in the corner formed by walls 1999 and 2002. Its fills (5481, 5482) were overlain by a layer (5402) containing (presumably residual) mid 6th-century pottery. In Room 20, mortar, tile and stone rubble contexts (5534, 5511), which lay above the mid to late 6th-century tile rubble (5509) may indicate partial demolition of the south range. The mortar and rubble deposits extended into Room 21 (where they were recorded as 1415/5264 and contained pottery as late as AD 575). These deposits pre-date the claybonded wall (1379) that blocked or narrowed the doorway between Rooms 20/21, as the construction trench for the wall (5535) cut 5534 (Fig. 3.98). This suggests either that the rubble layers (5534 and 5511) did not relate to the demolition of the walls of the room, or that this demolition was only partial. In Room 20, rubble layer 5534 was cut by 5 pits or postholes (5520, 5541, 5543, 5549 and 5557) (Fig. 3.99). These were oval shaped cuts around 0.35 m in depth. It is difficult to discern a structure from the plan of these features, and

Figure 3.96. Grave 1783 (looking west). The stylobate of the eastern colonnade of Room 25 is clearly visible (1 m scale)

they may represent supports for parts of the earlier buildings that are likely to have been in considerable disrepair at this juncture. A spread of mussel shells (5508) in the southeast corner of Room 20 is stratigraphically contemporary with these features, although it contains ceramics of early 6thcentury date. Also possibly contemporary with this phase is a rectangular stone-lined pit (5510) which abuts the wall separating Room 20 from the south portico (1373) (Fig. 3.100). The pit, which measured 1.25 × 0.53 m was excavated to a depth of 0.62 m although this was not the full extent. The dark organic fill (5495) contained 6th- to 7th-century pottery as well as mussel shells, although the purpose of this feature is unknown.16 As noted above, the rubble deposits extended into Room 21 (as 1415=5264) and sealed the mid to late 6th-century furnaces 1479 and 1821. This deposit (1415) contained 16 coins of 4th- to 5th-century date and pottery dating to c. AD 550–575. The rubble was cut by an east–west orientated grave (1435) (Fig. 3.101). The skeleton (1434) was centrally positioned within the grave with its head towards the western end, while a single limestone slab positioned upright on the southern side may have acted a grave marker. To the left hand side of the skeleton was found a hoard of iron and bronze finger rings (SF 0581), while a single strip of iron was positioned over the right hand side of the skeleton (SF 0580). Of the 16 small finds

1841 1496

1456 1454

3075 5006 3158

Room 23

3123

Room 24

5557

1434

1435

Room 20

Room 21

3126

5509

5535

Room 22

1415/5264

Room 19

5511

5549 5541

1379

3123

5480

5543 5508

5534

5520

5553

5075 5073 5111

Room 18

5109

1238

5043

1241

1236 5044

1266 1269

1234 1758 1756

10m

0

Figure 3.97. The south wing in phase 10 1376 North

South

5536

1379

5535 1373

1375

5268 5270

1983

1079 5353

5305 0

2m

tile

Interpretation late 6th- to 7thcentury door blocking

North

South

fill cut for door blocking

wall

wall

tile

mid - late 5th-century mortar feature

early - mid 6th-century mussel rich deposit

early 6th-century rubble layer

tile rubble layer

mid - late 5th-century mortar feature

Figure 3.98. West-facing elevation of wall 1375/1376 separating Rooms 20 and 21, showing blocking 1379 and underlying deposits 5510

5557

5510

5557 5549

5535 5549 5535 1379

5549

5510 5511

5557

1379

5541 5509

5541

1379

5541

5508

5508 5543 5534 5534

5509

5535

5511 5509

5543

5511

5543

5508

5534

5520

5520 0

5520

Figure 3.99. Late 6th-century post-holes 5m in Room 20 0 and pits and mussels 5508 0

5m

5m

3  The 5th- to mid 7th-century occupation of the triconch area

Figure 3.100. Stone-lined pit 5510 (probably late 6th century) in Room 20 (1 m scale)

111

Figure 3.101. Grave 1435 (1 m scale)

1456

1455

1454

0

1m 1456

3126

1454

3135

3158

0

3075

1m

3126

3158

3122 3123 3122

0

1m 3123

0

2m

Figure 3.102. 6th-century graves in Room 22

excavated from the grave fill (1428), four were iron nails/ nail fragments (SF 0579, SF 0582, SF 0689 and SF 0692). These nails and soil staining in the north and northwest parts of the grave suggest the burial may have had an associated coffin. It is possible that this grave dates to the

medieval period, but it seems reasonable to associate it with the remainder of the graves in the southern range. Rubble deposits were also found in Room 22, where a thick layer (c. 0.30 m deep) largely composed of fragmented pinkish yellow mortar, was found spread across the entire

112

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

Figure 3.103. 6th-century amphora burial 1454 (1 m scale)

Figure 3.104. 6th-century amphora burial 3126

Figure 3.105. Infant burial 3122 (20 cm scale)

room. This layer (3075) contained ceramics dating to AD 580–600, and may be contemporary with the rubble deposits in Rooms 20 and 21 noted above. Two burials were cut into these late 6th-century rubble deposits (Figs 3.102, 3.103, 3.104). A neonate amphora grave (1454) (cutting 1452/5025) was orientated north–south parallel to wall 1893, while another amphora burial (3158) (cutting 3075), was also found in the central area close to drain 1930. A third infant burial, without an amphora container, was found in the south west corner of Room 22 (3122) (Fig. 3.105). Both the amphorae from 1454 and 3158 dated to the first half of the 6th century. The rubble deposit extended into Room 23 (where it was recorded as 5006). Pottery in this layer indicates a deposition date of post AD 550. Directly above this layer a dividing wall (1382) was built running north south down the centre of the room (Fig. 3.106). No overlying stratigraphy survived in association with this wall. It can only be said to post-date the rubble deposits noted above and predate three stone-lined graves (1511, 1514, 1520) that are medieval (see Chapter 4, phase 12). The fact that

it subdivides the room suggests that the walls may still have been standing to some height, although physically it is built at a very high level above the late antique floor levels, suggesting that it may be medieval in date. Room 18 (the long gallery) was also used for burial during this period. At the western end of Room 18, two infant burials (5073, 5109) were noted cutting 5004 (a deposit sealing the industrial phases and dated to AD 550–575) (Figs 3.107, 3.108). The burials were both stonelined and east–west orientated with heads to the west. No coherent dating evidence was recovered from overlying deposits, although a rubble layer (5003) was equated by the excavators with a 6th-century deposit (5068) further to the east. Deposit 5068 was one of a number of deposits associated with the demolition of the phase 9 industrial levels and post-dated the major post AD 550 deposit 1676 (see phase 9 above). Equally, the fact that these are infant burials like many of the other confirmed late antique burials from the triconch is suggestive of a late antique date. A group of six burials was also noted at the eastern end of Room 18 (5044), (1211), (1234), (1238), (1269) and

3  The 5th- to mid 7th-century occupation of the triconch area

113

Figure 3.106. Room 23, showing dividing wall 1382 (looking north). The wall can only be said to post-date the mid to late 6th century

5075 5073

5075

5111 0

5109

5073

1m

5004

5111

0

5109

2m

Figure 3.107. Possible 6th-century burials at west end of Room 18

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

114

Figure 3.108. Possible 6th-century burials 5073 and 5019 during excavation (looking south) 1241 1241

1211 1211 1238 1238

1239 1239

1267 1267

1236 1236

5043 5043

1756 1756

1234 1234 5044 5044 1m 1m

0 0

1758 1758

0 0

1m 1m 1211 1211 1233 1233

5043 5043

1757 1757

1234 1234

1266 1266

1241 1241 1238 1238 1267 1267

1236 1236 5044 5044

1269 1269

1758 1758

1756 1756

0 0

1266 1266 1269 1269

2m 2m

Figure 3.109. Probable 6th-century burials at the eastern end of Room 18

3  The 5th- to mid 7th-century occupation of the triconch area

115

Figure 3.110. Grave 1758 prior to excavation (1 m scale)

Figure 3.111 Grave 1234 before and after excavation (1 m scale)

Figure 3.112. Grave 5044 (1 m scale)

(1758) (Fig. 3.109). No coherent deposits were recovered from above these burials, but two of the graves were clearly of late antique date. Grave 1269 was that of a premature foetus (c. 34 weeks) inserted into a possibly Calabrian amphora of the late 6th or 7th century. This grave in turn post-dated grave 1758 (that of an infant buried in a tilelined grave, covered with imbreces) (Fig. 3.110). Grave 1234 was also probably of late antique date in that it was a classic ‘cappuccina’ burial, in which the body (that of an adult female) was covered with a pitched tile roof (Fig. 3.111). All these graves were cut into an undated context

1233, from which the bones of a foetus (1211) were also recovered. A further infant grave (5044) (Fig. 3.112) was noted as being overlain by 1606, a silt layer of similar character and date to the mid to late 6th-century deposit 1676 that sealed the industrial levels of phase 9.17 It is likely therefore that the burials were inserted as the silt abandonment deposits began to develop following the cessation of industrial activity in Room 18.18 A single amphora burial (1496), containing a neonate, was inserted in the apse area of Room 24, above the furnace noted above in phase 9 (1298), and built against the apse wall (1281) (Figs 3.113, 3.114). The amphora (1548) is a type paralleled at the Crypta Balbi in Rome, probably dating to the mid to late 7th century (Reynolds 2002). The bath-house (Rooms 28, 32, 36) (see above Fig. 3.9) The mussel deposits (1538, 1797) (noted above (phase 8) as resulting from mussel processing in Room 18 during the

116

William Bowden, Karen Francis, Oliver Gilkes, and Kosta Lako

1496

1496

0

1m

1298

0

2m

Figure 3.113. Plan of apse area of Room 24 showing mid to late 7th-century amphora burial 1496 cut into kiln 1298 mid 6th century), within Rooms 28/32 are sealed by rubble deposits (1806=1772, 1778=1773) and may represent collapsed wall debris. Similar deposits were noted in Room 36 (1790 and 1887) of which 1790 contained ceramics of 7th-century date. These deposits overlay a single post-hole (1993) that was cut into the late 5th-century deposit 1934, indicating only that this post-hole dated to some time during the 6th century. The rubble layers within the bath-house probably result from the collapse of wall 1795 or associated doorway/arch and represent a process of demolition seen across much of the southern range during phase 10.

Figure 3.114. Mid to late 7th-century amphora burial 1496 (1 m scale)

Summary/interpretation A series of rubble dumps across most of the southern wing, the bath-house, the western wing and the central courtyard suggest that the late 6th and 7th centuries saw the continuation of demolition and robbing of masonry across much of the triconch area. Only Room 20 in the southern wing and Room 29 in the western wing produced any real evidence for occupation of the Triconch Palace area in this period (although it should be noted that the Merchant’s House area saw considerable use at the end of the 6th century). In the southern wing, post-holes, small pits and an associated deposit of mussel shells suggest that Room 20 was used after its walls had been partially demolished, although the construction of a wall narrowing or blocking the door between Rooms 20 and 21 indicates that the dividing wall must have continued to stand to some height at this point. In the western wing the mortar floor in Room 29 represents a short period of activity during the late 6th

3  The 5th- to mid 7th-century occupation of the triconch area century and possibly into the early 7th before the room was used to dump demolition material. Graves in Rooms 18, 22, 23, 25 provide evidence that the use of the area of burial, which started around the mid 6th century, continued during the late 6th and early 7th century. These ranged from single burials such as that in the apse of Room 24, to the cluster of burials at the eastern end of Room 18. The burials that can definitely be assigned to this period are mainly (although not invariably) those of children, mainly because they are often buried in amphorae that allow them to be dated even in instances where the related stratigraphical relationships are unclear. This limited activity in the area of the triconch can be contrasted with that which was occurring contemporaneously in the Merchant’s House area in the adjacent plot to the west, where new two-storey buildings were erected at the end of the 6th century. We may perhaps envisage the triconch area acting as a peripheral part of the Merchant’s House complex, whose occupants used the former domus primarily as a quarry and a graveyard. Following this period, the triconch area was seemingly abandoned entirely by the mid 7th century, with little sign of reoccupation until the later 9th century.

Notes 1

2 3 4 5

6

High levels of ground water meant that the mosaic was never entirely dry during the excavation. It was cleaned again for conservation and recording in 2007 when this central division became apparent. Throughout this volume coins are referred to solely by context to correspond with the concordance table in the forthcoming report on the coins by Sam Moorhead. With the benefit of hindsight the use of micromorphology would have potentially solved some of these ambiguities. This blocking wall (5144) was at a level below that which was reached by the excavation so the precise date of its insertion is unknown. Although these post-holes were recorded as underlying a series of late antique deposits, it is possible that they are medieval and were not recognised within the darker overlying deposits. The fact that the deepest post-hole was only 0.22 m in depth could also support this interpretation. The site records suggest that 1847 may lie beneath the late 5th-century tile dumps in Room 21, indicating that this mussel shell deposit may belong in the previous phase. Indeed, it is likely that the sporadic dumping of mussel shells may have occurred in some rooms from the late 5th

7 8

9 10

11 12

13 14

15 16

17 18

117

century onwards, interspersed with the dumping of rubbish and demolition debris. Layer 1392 contained a wide range of material of which the later pottery could indicate a date of c. 575, although much of the material belonged to the second quarter of the 6th century. The site records indicate that the mussel shell level 1847 lay beneath the 5th-century rubble layer 1487, suggesting that the scattering of mussel shells was occurring from the late 5th century onwards. The site records noted that the post-holes had been spotted only after the area had been cleaned for photographs, and considered it likely that they had been cut from a higher level. The silty horizon 1523 in fact contained a much wider range of ceramics that extended as late as the 7th century. However, the excavators noted that there was a possibility of contamination from overlying layers and therefore recorded the lower part of the deposit separately as 1536, in which the pottery assemblage has a narrower date range. Much of the pottery from 2000 dated to the later 5th century (c. 450–475) and it is possible that that the filling of the drain commenced in the 5th century. The relationship between 1152 and 1193 was destroyed by the late medieval pit (1186) but the excavator was convinced that they were the same deposit. However, 1193 does not contain the huge quantities of finds that characterise 1152. Present-day fishermen at Butrint collect mussels from approximately June to September. Although there are sherd joins with 1152 Paul Reynolds considers the ceramics from this deposit (2095) to be markedly different to those from 1152, and to have more in common with the later major deposit 1676 from Room 18 in the south wing. All these deposits also contain some (probably) intrusive medieval material, deriving from the later build-up of medieval deposits in the courtyard area. Analysis of the hammerscale showed that it contained droplets of copper alloy, which was probably used as part of a flux preparing the iron for welding. The fill also contained medieval pottery but the construction of this feature and its relationship with the wall (1373) suggests that it is most likely to be a late antique feature, although the absence of overlying stratified deposits means this is not certain. The site records are contradictory on this relationship, which could not be established beyond doubt during the excavation. An earth-bonded wall (1276) built on the south stylobate of Room 18, was originally thought to be associated with this burial phase (Gilkes and Lako 2004, 171–2). However, this wall is probably more likely to have been part of the earlier industrial phases in Room 18.

4  The medieval occupation of the triconch area William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

Introduction This chapter looks at the archaeological deposits and structures forming phases 11 to 14 of the Triconch Palace. These phases reflect the semi-abandonment of the triconch area during the mid 7th to late 9th century, followed by more consistent occupation from the 10th to 13th century and subsequent abandonment and development of woodland from the 14th to 19th century. As described in the previous chapter, some of the rooms of the former Triconch Palace were used as the basis for small dwellings and industrial areas during much of the 5th and 6th centuries. The mid 6th to mid 7th centuries also saw a process of demolition, evidenced by the removal of walls and structures. The use of the area for burial that started in the 6th century continued into the 7th century and possibly beyond, although many of the burials have very limited dating evidence associated with them. It is clear, however, that at least some of the burials date to the medieval period. The extent of activity in the triconch area between the mid 7th and mid to late 9th centuries is difficult to ascertain. It is of interest, however, that the limited quantity of pottery that can be dated to this period is almost all found within the central courtyard area of the former Triconch Palace, suggesting that the courtyard remained an open and accessible area during this period. The reoccupation of the triconch area during the 10th century is reflected by a post-built structure in the south wing (in Room 19), which was replaced by a stone structure during the 11th century. A stone-built structure and well in the west wing (in Rooms 25 and 26) are less securely dated, but may also belong to this phase. Other rooms in these areas also contain evidence of activity in the form of pits, post-holes and mussel shell layers. The post-built structures also included an extensive series of post-holes cut into the wall-tops of the late antique buildings, which had apparently been levelled to their current height in the medieval period. These post-holes clearly show that the walls were used as the foundations for quite substantial

structures that were built over parts of the south and west wings. However, no dating evidence was found in relation to these post-holes, as the wall-tops were only covered by thin topsoil horizons. Consequently, it can only be suggested that the most likely date range for these buildings is from the 10th to 13th centuries, on the basis that the quantities of material recovered indicate that this period saw the most intensive usage of the triconch area during the medieval period. Burials occurred from as early as the 10th century in the northern and southern wings, while sporadic burials were probably also inserted in the later medieval period. The burials were characterised by a more marked presence of adults than was the case in Late Antiquity, where burials were mainly those of infants and children. The former central courtyard seems to have continued as an open space, and it is of note that the majority of coin finds were concentrated in this area. This could suggest its periodic use for trading activity although this remains largely a matter of speculation. The concentration of coins in the courtyard area may simply reflect the fact that the majority of 10th- to 13th-century material was found in this area. The use of the remains of the triconch triclinium in the later medieval period is unknown, although it is possible that the building was used as a church (with associated fairs or markets occurring in the courtyard) (see Gilkes and Lako 2004, 170 and Chapter 10 in this volume). By the late 13th century, as in the adjacent Merchant’s House area, major domestic occupation had ended and the only discernable activity was rubbish dumping and possibly burial. The date when woodland developed over the triconch area is unknown, although Ugolini’s photographs of Butrint suggest that the tree cover was sparser in the 1920s-1930s than it is at present, indicating that the dense woodland that covers the site today is a relatively recent phenomenon. However, the medieval stratigraphy of the triconch (particularly in the area of the central courtyard) was found to be highly disturbed, with intrusive later medieval material appearing in the earliest post-Roman contexts. This

4  The medieval occupation of the triconch area could be due to excavators failing to recognise horizons within the dark humic soils that constituted the upper levels of the triconch stratigraphy, but the consistency with which this intrusive material occurs suggests that root disturbance or late- and post-medieval cultivation may be responsible. Partly as a result of this, phases 12 and 13 (which can be distinguished in other parts of the site) have been combined in the central courtyard area.

Phase 11: Mid 7th to early 10th century Minimal activity Little or no activity took place in the triconch area between the 7th and early 10th centuries and little pottery that can be confirmed as dating to this period was found anywhere within the triconch area, with the exception of small quantities noted in the central courtyard. The central courtyard (16) The central courtyard is the only area of the triconch that contained notable (although still very small) quantities of pottery dated to the later 7th, 8th or 9th centuries, although it should be noted that pottery dating to this period only constituted 1% of the total combined medieval ceramic assemblage from the Triconch and Merchant’s House. All this material occurred either in late antique contexts (2071, 5362), or residually in later contexts (1699, 2052, 5276, 5322, 5359, 5361/5399, 5539, 5540). With the exception of 5359, which appears to be a relatively coherent context of 10th- to 12th-century date (see below), all these contexts contain a range of material spanning the medieval period and sometimes extending into the postmedieval period.1 No discrete early medieval contexts were identified. However, the fact that early medieval pottery was seemingly confined to this area suggests that the medieval use of the courtyard space commenced during this period (or continued uninterrupted from late antiquity) although the nature of this occupation can only be speculated upon. The lack of late 6th- and early 7th-century burials in the courtyard (which remained free of graves throughout its post-Roman history) is also intriguing, adding to the impression that the courtyard was maintained as a defined space from late antiquity onwards. Summary/interpretation The archaeological evidence suggests that there was little activity in the triconch area between the mid 7th and early 10th centuries. The former rooms of the late Roman building contained no evidence relating to this period, although there are some indications that the courtyard remained a defined space that saw some limited activity (even if this activity was only the occasional disposal of small quantities of rubbish). However, even these very limited finds of early medieval material in the triconch area does confirm that activity of some sort did continue in the lower city during this period. This occupation has also been confirmed by the 8th- to 9th-century deposits discovered in the 2007 and 2008 excavations of two of

119

the towers in the Western Defences (Hodges 2008: 65–71; Hodges et al. 2009).

Phase 12: Early to late 10th century

(Fig. 4.1) Renewed occupation, building and burial During the 10th century there was renewed occupation of the triconch area. Stone- and post-built structures were erected over the south wing (Room 19) and the former west wing (Rooms 25, 27 and 31), although it should be noted that the dating of the latter is not secure. The post-built structures also partly reused the levelled walls of the Triconch Palace as foundations. Pits and mussel shell deposits attest to renewed productive activity. This reoccupation and structural activity corresponds to a similar phase in the adjacent Merchant’s House area. Considerable quantities of pottery and 10th-century coins found in the deposits from the central courtyard suggest the start of quite intense activity that continues seemingly uninterrupted until at least the 12th century. Owing to the uncertainty surrounding some of the dating and stratigraphy in this period a number of features have been treated as a single combined phase (see phase 12–13 below). However, in several areas it has been possible to date structures and features with more confidence to this earlier period; hence, these are treated separately within this section. A series of burials has also been placed within this period, as AMS radiocarbon dates suggest they date primarily to between the late 9th and early 11th centuries, although it is clear that sporadic burial continued into the 12th century and later. The west wing (Rooms 25, 27, 31) It was probably during the 10th century that a building was erected within Rooms 25 and 27, partly reusing the walls of the earlier building and partly using new foundations. Whether these relatively crude foundations provided support for a timber superstructure is not clear. The building was a rough oval shape measuring 5.5 × 2.8 m, and consisted of newly built walls of rough limestone blocks and slabs bonded together with mortar (1410, 1567, 3260). Wall 1567 was built above the earlier stylobate (1568) for the colonnaded entrance between Room 25 and the east portico of the domus, although the medieval wall and the stylobate were separated by a number of intervening deposits including a late antique infant burial (1783) (discussed in Chapter 3, phase 10). The large late antique circular structure in Room 27 (5336) (discussed in Chapter 3, phase 6) appears to have influenced the plan of this later building, which seems to partly use extant sections of the late antique walls as foundations (Figs 4.2, 4.3). A substantial well (1600) may also be associated with this building (Fig. 4.4). This was a carefully constructed circular feature, cut through a late antique deposit (1545) (also cut by the infant grave 1783 noted above) and through the buried mosaic pavement of Room 25. The well lay

120

1186

1264

Room 17

3260

Room 31

1800

5526 5489 5485

1515

5487

5453 5451 5493

1767 1410 1558 1600 1545

1198

Room 14

1160

1199

Room 27

Room 25

5322

1567

5229

5528 5530

Room 23 Room 24

5476

1514

1511

1520 1408

5237

5532

Room 19

5441 5437

5447 5429 5435 5414

5010

5443 5445

5433 5439 0

Room 28

5231

5427 5431

2m

1793 5325

0

Figure 4.1. The triconch area showing areas of activity in phase 12 (early to late 10th century)

20m

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

1191 1260

1231

4  The medieval occupation of the triconch area

121

3260

5526 5489

1410

5451

5485 1515

1767

5453

5487

1558

1600

5493

1567 1545

0

5m

Figure 4.2. Phase 12 (early to late 10th century) structures in Rooms 25, 27, 31. The L-shaped line of post-holes on the left is more likely to be of late antique date just to the south of the oval building. The well, which was 1.2 m in diameter, was carefully lined with unbonded limestone blocks. A poorly preserved rough paved surface consisting of tightly packed stone and tile (1558) abutted both the wall of the oval building (1410) and the well, and was presumably associated with both structures. Dating evidence associated with this building is extremely poor. A possible occupation deposit (1767), a thin silty layer that built up within the walls, contained no datable material (Fig. 4.5). When the walls were removed, wall 1410 was found to contain residual 6th- and 7th-century pottery. However, the fact that no later material was found within the walls or layer 1767 may be significant. Given the quantities of 10th-century and later material present on the site, it is likely that this material would have also been present within 1767 had this deposit dated to later than the 10th century.

The demolished or collapsed remains of the structure (1557), which partially covered the tile pavement, were sealed by deposits 1409 (noted in Rooms 25 and 27) and 1411 (which covered the south part of Room 25 and extended into Room 26). These contained ceramics predominantly dating to the 10th–14th centuries. The upper fill of the well (1560) contained a few sherds of 15th-century pottery, although it is likely that this is the result of later slumping in the well fill.2 The lower fills were not excavated due to the high level of the ground water in the triconch area. On a basis of the stratified dating evidence, therefore, the building can only be ascribed a date of between the late 7th and 14th centuries, but the material from contexts 1409 and 1411, together with the apparently intensive use of other parts of the triconch in this phase suggests that a construction date in the 10th century is most likely.

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

122

Figure 4.3. Phase 12 (early to late 10th century) structures in Room 25, looking north (1m horizontal scale)

Figure 4.4. Well 1600. This upper section was removed down to the level of the earlier mosaic pavement (1 m scale)

South-facing section through southern end of Room 27 1767

1410 1114 1409

West 1355

1449

East 1545

1568

0

1m

Interpretation medieval horizons and humic build-up

10th-century wall

10th-century occupation level

West

East

1st- to 3rdcentury wall

late 5th-century tile dump

10th- to 12thcentury trample

stone block mid to late 6th-century collapsed wall demolition from 1568

0

3rd- to 4thcentury wall

1m

Figure 4.5. South-facing section through southern end of Room 27, showing 10th- to 12th-century deposit 1409 covering the levelled remains of the phase 12 (early to late 10th century) building represented by wall 1410

4  The medieval occupation of the triconch area

5528 5528

123

5229 5229

5530 5530 5403 5403 5532 5532

5476 5476

5429 5447 54475429

5237 5237

5231 5231

5414 5414 5427 5427 5441 5441 5437 5435 54375435 5402 5402

5445 5445

5433 5433 5443 5443 5439 5439

5431 5431

5225 5225 Structural Structural posts posts Partition Partition posts posts 0 0

2m2m

Extension Extension posts posts Threshold Threshold deposit deposit

0 0

2m2m

Figure 4.6. Phase 12 (early to late 10th century) building in Room 19, with interpretation of posts

An ancillary structure relating to this building may also have been erected in the adjacent Room 31, indicated by an L-shape set of post-holes and stake-holes (5526, 5453, 5489, 5451, 5485, 5487, and 5493) (see Fig. 3.59). Although these features are probably of late antique date, it is possible that they were not recognised as cutting through the overlying deposits (1516 and 1523) (noted in Chapter 3, phase 7). If so, they may relate to the stone structure in Rooms 25 and 27 and represent some kind of fenced area or out-building. The south wing (Rooms 19 and 23) During the 10th century, a post-built structure, possibly a dwelling, was built within the remains of Room 19 (Figs 4.6, 4.7). A complete pot dating to the 10th century was found within what appeared to be a threshold deposit, allowing the building to be dated with some confidence. The post-holes were cut into a probable late antique deposit (5402) (see Chapter 3, phase 10), which overlay the levelled remains of wall 5077 (the Roman wall that bisected the room). This indicates that the remains of this wall were not a factor in the plan of the building (although

the alignment appears similar). Thirteen post-holes were identified within the southern half of Room 19. Of these post-holes, 10 are recorded as cutting layer 5402 while the remaining three are noted as cutting 5403 (Fig. 4.8).3 The south side of the structure was supported by four upright posts, represented by post-holes 5427, 5435, 5437 and 5441. These varied in width from 0.32 m – 0.54 m, with a maximum depth of 0.34 m (although the recorded depths may not represent the original depths, see note 3 above). Post-holes 5435 and 5437 were larger than the other post-holes in the building and may have formed key supports for the timber structure. Alternatively 5435 may represent two smaller circular post-holes, with the second possibly relating to a rebuild. A re-cut of the phase 10 feature 5480 may have formed the southwest corner. Four further post-holes to the south (5445, 5443, 5439 and 5431) may also have formed a small extension or leanto. Post-hole 5439 may have been used as a central post to this subsidiary structure, with 5443 as a possible re-cut. It is also possible that these peripheral post-holes relate to an external staircase, perhaps reusing the late antique stairblock 1277, located in the northeast corner of Room 18.

124

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

Figure 4.7. Phase 12 (early to late 10th century) post-built structure in Room 19, looking east (2 m scale)

S

N

5528 5530 S

N E

W

5403 N

5532

S

5447

5476

SW

NE

5429

E

W

S

N

5402 5441

N

5427

S

S

N

E

5414

5445

W

5435 S

N

S

N

S

N

E

W

5433

E

W

0 5443

5439

5437 5431

0

2m Profiles

1m

Figure 4.8. Profiles of post-holes of phase 12 (early to late 10th century) building in Room 19

4  The medieval occupation of the triconch area In Chapter 3 (phase 7), it is noted that the gradient of the surviving stair block suggests that an additional wooden staircase ran along the north wall 1871 at right angles to the stair block. It is possible that 1277 may have still been visible during phase 12 and used as such for access into Room 19. One post-hole (5225) is visible cut into the top of wall 1896 which could have formed support for such a structure, although this feature is undated. The eastern wall of the building was seemingly formed by four further post-holes (5429, 5528, 5530 and 5532). Three of these post-holes were in the northern half of Room 19 running parallel to wall 1879. These may be of late antique date (see Chapter 3, phase 8), but the depths of these post-holes (0.14 m, 0.15 m and 0.17 m respectively) suggest that they were cut from a higher level than that at which they were noticed by the excavators, and relate to the other medieval post-holes recorded as cutting layer 5402.4 Three further post-holes (5229, 5231 and 5237) were also recorded cutting into the top of the levelled remains of wall 1879. They share the same alignment as post-holes 5528, 5530 and 5532 suggesting that all six are contemporary. The presence of internal supports (represented by postholes 5476 and 5447) suggests that the structure may have had an upper storey. It is possible that the post-holes cut into wall 1879 formed an external staircase accessed from the central courtyard, while, as noted above, the southernmost post-holes in Room 19 may also have supported an external staircase. Within the structure, a layer of clay, tile and limestone pieces (5129) was noted above layer (5402) and beneath the base for the phase 13 house 5002, but otherwise no evidence for a contemporary occupation layer or floor surface was recovered. The dating evidence associated with this building is limited. Four of the post-holes (5427, 5435, 5437 and 5441) that ran in a line east west across the central southern part of Room 19 contained residual pottery dating to the 5th to 6th century. However, a further post-hole (5414) within this line contained an almost complete pot dated by Joanita Vroom to the 10th century (Vroom in prep.) (Fig. 4.9). This pot would have left no space for a post, and its upright position within the hole suggests that it was placed there deliberately (rather than being a chance inclusion after the post had been removed). It is likely therefore that 5451 was not a post-hole but a threshold deposit, created by the owner to protect and bring luck to the building. Consequently the construction of the building can reasonably be placed in the 10th century. Starting from the 12th century, a number of burials were inserted into the south wing in a sequence which may extend into the 12th century. In Room 23, three graves were found underlying an undated deposit (5022/5023), and cut into a deposit dating to c. AD 550 or later (5006=5018) (Fig. 4.10). Grave 1511 (skeleton 1509) was a stone-lined inhumation, defined by limestone slabs, containing the remains of an infant skeleton c. 9 months old, with its head placed to the southwest (Fig. 4.11). Grave 1514 (skeleton

125

Figure 4.9. Complete 10th-century pot in post-hole 5414 (which may in fact be a foundation or threshold deposit) (30 cm scale)

1512) was a sub-rectangular tile and stone-lined inhumation containing the remains of an infant of c. 6 months–1 year old, with the head placed to the southwest (Fig. 4.12). AMS radiocarbon dating of the skeleton gave a date of AD 1010–1170. Grave 1520 meanwhile was a stone-lined inhumation cut into the top of wall 1382 that subdivided Room 23. The skeleton (5018) was that of a male of 40 years old or more with the head placed to the southwest (Fig. 4.13). AMS radiocarbon dating produced a calibrated date of AD 780–980. In the doorway between Rooms 18 and 23 a stone-lined burial (5012) containing the skeleton (5010) of a woman between 30 and 50 years of age was found, apparently partly cut into the wall (Fig. 4.14). AMS radiocarbon dating produced a calibrated date of AD 890–1030 for the burial. Skeleton 5010 was also maternally related to one of the skeletons from the north wing (1224) (dated by AMS radiocarbon to AD 880–1020) as was skeleton 1512 described above, and it is clear that the interment of all three skeletons occurred in broadly the same period (see below).5 In Room 24 a stone-lined grave (1408) containing a child of 3–4 years was cut through the levelled top of wall 1282 (one of the earliest Roman walls in the complex) (Figs 4.15). AMS dating produced a range of results from AD 980–1140. In the remains of the small octagonal vestibule of the domus bath building (Room 28), a stone-lined burial (1793) containing the body of a woman of around 25 years old was encountered just below the topsoil (1114) (Figs 4.16, 4.17). No reliable dating evidence was recovered although the burial cut a rubble layer (1772/1773) that contained a single sherd of medieval amphora (Fig. 4.16). A further stone-lined burial (5325) was found in a trench dug between the long gallery (Room 18) and the city wall (an area designated as Room 35). The grave, which had a stone capping, contained the body of a child of around 4–5 years old (Figs 4.18, 4.19). This grave was sealed by a layer of clay (5273) that contained pottery ranging from

126

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

1512

1509

1514

5006=5018

1511

Room 23 0

1m

0

1m

1514 1511 1520

1518

1520

5012

5010 5010

0

5011 0

1m

1m

0

5m

Figure 4.10. Possible 10th-century burials in Room 23

Figure 4.11. Grave 1511 in Room 23 (1 m scale)

Figure 4.12. Grave 1514 in Room 23 (1 m scale)

4  The medieval occupation of the triconch area

Figure 4.13. Grave 1520 in Room 23 (1.6 m scale)

127

Figure 4.14. Grave 5010 in Room 23 (20 cm scale)

1282

1408

1406

0

0

1408

1m

5m

Figure 4.15. Grave 1408, cut into the wall between Rooms 18 and 24

1791

1793 1794

0

1m

Room 28

1772/1773

1793

0

2m

Figure 4.16. Grave 1793 in Room 28

Figure 4.17. Burial 1793 in Room 28 (1 m scale)

Figure 4.19. Grave 5325, adjacent to city wall

4  The medieval occupation of the triconch area

5293

5324

0

129

5325

1m

5325

City wall

0

2m

Figure 4.18. Grave 5325, adjacent to city wall

the 10th–14th centuries, although given its proximity to the surface it may well date to the latter part of this period. The north wing (Rooms 14 and 17) In the north wing (Rooms 14 and 17) several burials were found that were originally dated to the first half of the 6th century (Gilkes and Lako 2004: 170–72) (see Figs 2.72 4.20, 4.21). However, AMS radiocarbon dating of several of the skeletons subsequently indicated that these burials dated from the late 9th–11th centuries. In Room 17, two graves (1198/1220 and 1199) were found cutting an occupation level (1194) dating to the second half of the 5th century. Grave 1198/1220 was a vertical sided cut 1.20 × 0.47 m lined with limestone and tile fragments, which contained the bodies of two children of around 5–7 years old (1224) and 4–6 years old (1225) respectively (Figs 4.22, 4.23). Grave 1199 was that of an adult of around 45–60 years (1226) buried within a vertical sided cut lined with large limestone slabs set on their edges. A cover of limestone slabs was also present although the grave had been truncated by a medieval pit (1186) (Fig. 4.23). Both these graves apparently lay beneath a substantial deposit of broken tiles (1193) that

was thought to be part of the major deposit 1152 noted in Room 14.6 AMS radiocarbon dating of skeleton 1224 subsequently gave a calibrated date of AD 880–1020 and it appears likely that either 1193 was a later medieval deposit, or that the excavators missed the grave cuts. As noted above skeleton 1224 is also related via the maternal line to skeletons 1518 (grave 1520) and 5010 (grave 5012) from Room 23 (Fenton et al. in prep. for mitochondrial DNA analysis and the human skeletal remains). On the north side of the medieval pit (1186) traces were found of a further grave (1260) (Fig. 4.21). Unlike the other graves in Room 17 there was no sign of a tomb structure and the skeleton (1259) had been badly disturbed by the cutting of the pit. An AMS radiocarbon sample from the skeleton gave a calibrated date of AD 970–1040. A further stone-lined grave (1191) contained an adult male (1189) between the ages of 30 and 45 and the semi-disarticulated skeleton of a child (1188) suggesting that the child was buried first and then the grave was opened up again for the adult (Fig. 4.24). AMS dating of skeleton 1189 produced a calibrated date of AD 890–1030. A further infant grave (1231) in Room 17 also may have belonged to this sequence (Fig. 4.21). The grave was defined by upright limestone

130

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

1194

1191 1260

1264

Room 17

1231

Room 14

1186

1160 1198

1152

1199

10th-century graves

10m

0

Figure 4.20. Graves in the north wing

blocks, but was very badly disturbed. Finally, a grave (1265) containing a child of 6–8 years was noted running under the western baulk of the excavation, cutting the 4th-century demolition deposit 1227 and lying beneath the 6th-century and later deposit 1114 (Fig. 4.25) (see Gilkes and Lako 2004, 155).7

Room 14 also contained one grave (1160), recorded as being cut into the major mid 6th-century deposit 1152 but overlain by a further deposit (1133) that was subsequently combined with 1152 due to pot joins. It is possible that the grave was inserted during this period of dumping, but it is also likely that the cut through 1133 was not recognised (as

4  The medieval occupation of the triconch area

1191

1231

1260 1231

0

131

50cm

1259

1260

1186

Room 17 1198

0

1m

1199

0

5m

Figure 4.21. Graves in eastern part of Room 17 (1191, 1198, 1199, 1231, 1260). For details of 1191, 1198 and 1199, see Figs 4.22, 4.23, and 4.24 was apparently the case with the graves in Room 17). The grave was lined with rough limestone slabs and contained a neonate in a supine position (Fig. 4.20). Summary/interpretation During the 10th century there was renewed activity over the area of the former Triconch Palace, with post-built structures erected within the remains of at least one room of the south wing (Room 19) and probably within Rooms 25, 27 and 31 in the west wing. It is not clear whether these buildings were part of a complex under single ownership or whether they represent two separate establishments. However, the buildings in the west wing are small and relatively insubstantial and seem unlikely to have formed any kind of dwelling. In contrast, that found within Room 19 appears to have been quite strongly built, and possibly included an upper storey (Fig. 4.26). This building type, with its external staircase to a second storey, appears reminiscent of the late 6th-century building discovered adjacent to the city wall in the Merchant’s House area and it is tempting to see it as part of a late antique building style that continued into the Middle Ages. How these buildings related to the contemporary activity in the central courtyard (described below) is unknown.

Figure 4.22. Grave 1198/1220 looking west (1 m scale)

132

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

Figure 4.24. Grave 1191, looking west, with skeleton 1188 at feet of skeleton 1189 (1.4 m scale)

Phase 12–13: Early 10th to 12th century (Fig. 4.27) Deposition in courtyard

Figure 4.23. Sequence showing grave 1199 and tile deposit 1193 cut by the medieval pit 1186 (looking west)

At the same time parts of the former triconch complex were being used for burial. Stratigraphically the medieval burials from the triconch are extremely problematic and difficult to date, but the radiocarbon dating of a number of the skeletons shows that a 10th- to 11th-century date for most of the skeletons is likely. The link between three skeletons through mitochondrial DNA could suggest different generations of the same family buried in the same area.8

Much of the triconch area saw apparently continuous activity between the 10th and 12th centuries, represented in particular by a series of extensive deposits in the central courtyard area (16). Although a depositional sequence of layers was identified in this area, the finds within these deposits spanned the entire period (probably as a result of post-depositional disturbance). Consequently, dividing them into the two phases (12 and 13) that can be distinguished elsewhere in the triconch and Merchant’s House areas would be an arbitrary exercise; therefore it has been decided to combine them. The central courtyard (16) As noted above, although no coherent early medieval deposits have been identified within the courtyard area, a limited number of finds indicate that such activity as occurred between the late 7th and early 10th centuries was concentrated in this area. The earliest identifiable medieval deposit in the courtyard was a rubble deposit (5355) that extended throughout the portico and lipped over the

1191

Room 17 1265

1231 1265

1186

1198

1264 0

1260

1199

1m

0 Figure 4.25. Partially excavated grave 1265, showing location in Room 17

Figure 4.26. Hypothetical reconstruction of the Room 19 building (SL)

5m

134

5321

Courtyard (16) 20m

0

5032

5148 5221

5022/5023 5260

5028

1498 1492

3097 3095

Room 22

3093

Room 23 5262

5034 5036

5666

3079

3111

5219 3119

5134 5150 5050

5015

3109

3181

5114

5063 5146

5122

5227

Room 18

5183

5123 5121 5118 5116 5225 5233

5066 5059

5061 5127

0

Figure 4.27. Features that can only be broadly dated to phases 12–13 (early 10th to 12th century) in the triconch

10m

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

5322

4  The medieval occupation of the triconch area

135

0

10m

0

10m

West-facing section across Room 16 5276

West-facing section across Room 530116

5276

5301

North

5322

5265

5322

5265

North 1875 1875

5359 5359

5361 5361

5362 5362

North

10th- to 11th-century 10th- to 12th-century humic build-up humic build-up 10th- to 11th-century 10th- to 12th-century humic build-up humic build-up

10th- to 12th-century humic build-up 10th- to 12th-century humic build-up

6th-century silt 6th-century silt

5573

1833

5574

5573

1957

5356

5357

5356

5357

1957

South 5355 5355

0 mixed 13th- to 14th-century humic/rubble build up 10th- to 12th-centurymixed 13th- to 14th-century 10th- to 11th-century humic/rubble build up humic build-up demolition 10th- to 12th-century 10th- to 11th-century humic build-up demolition

1830 1830 5m 5m

South

North early 5th-century northern stylobate wall of the Triconch early 5th-century northern courtyard stylobate wall of the Triconch courtyard

5574

0

Interpretation Interpretation

1833

South

early 5th-century domus courtyard early 5th-century domus courtyard

South early 6th-century mid 6th-century door blocking clay deposit natural accumulation early 6th-century mid 6th-century wind/rain door blocking clay deposit mid to late 5th-century natural accumulation wind/rainearly 5th-century southern occupation level mid to late 5th-century stylobate wall of the Triconch early 5th-century southern occupation level courtyard stylobate wall of the Triconch courtyard 0 5m 0

Figure 4.28. West-facing section through courtyard (Room 16) deposits

5m

136

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

stylobate wall 5597 into the courtyard itself (Fig. 4.28). The deposit may represent the demolition of the northern walls of the south wing. Deposit 5355 contains pottery dating from AD 450 to 550 and AD 580 to 585 but also 37 sherds of medieval pottery dating to the 10th to 12th centuries.9 Overlying the rubble deposit 5355 was a clay deposit (5361/5399/5539), which contained comparatively little in the way of coarse components or finds.10 Pottery dated from between the 4th to 12th centuries (including a single sherd of 8th- to 9th-century pottery) although the majority dated to the 10th and 11th centuries, suggesting that the deposit dates primarily to this period. This deposit was overlain by a thick layer of dark greybrown humic soil (5359), which varied from between 0.20– 0.45 m in depth. A considerable quantity of material was recovered from this level including 429 sherds of pottery dating from between the 7th and 16th centuries. 87% of this pottery dated to the 10th to 12th century, however, suggesting that this was the principal period of deposition. There were also three coins with dates of AD 945–950, AD 970–976 and AD 1075–1080 (as well as four residual 5th-century nummi).11 This was overlain by a similar deposit (5322), which also covered the extent of the courtyard. Although this was a shallower deposit than 5359 (varying in depth from 0.10–0.17 m) it contained substantially more material: 1335 sherds of pottery, of which 91.5% dated to the 10th to 12th century, together with two coins of AD 963–969 and three of AD 976–1030/5. Other finds included a bone pipe (SF 3034), a copper alloy link chain (SF 3037), pins (SF 3049, 3050), strip (SF 3055), strap end (SF 3091), fish hook (SF 3096), finger ring (SF 1144), a ceramic loom weight (SF 2022), a whetstone (SF 3092) and an iron knife fragment (SF 1141). Subsequent to the deposition of 5322, a small D-shaped structure (5321) made of rough irregular limestone rubble bonded with earth was built in the southeast corner of the courtyard (Figs 4.29, 4.30). The structure, measuring 2.95 × 2.12 m, stood alone within the courtyard and was not aligned with any of the earlier triconch walls. It only survived to a height of one course and no obvious explanation exists as to its purpose, although it may have served as a rough foundation for a wooden agricultural feature (such as a water trough for livestock). It is likely that it was a relatively short-lived structure, as the sequence of deposition within the courtyard area seemingly continued throughout the 10th–12th centuries. The D-shaped structure and the preceding sequence of humic deposits was overlain by further deposits (5301 and equivalents) with a similar soil matrix but a much greater number of coarse components, including significant quantities of limestone rubble, suggesting perhaps another phase of demolition or collapse of surviving elements of the triconch buildings.12 The pottery and finds from this deposit, however, are similar to those from the underlying layers comprising 2509 sherds of pottery dating from the late 7th to 16th century, but with 97% dating to the 10th to 12th centuries. The deposit included five coins

dating respectively to AD 945–950, 970–976, 977–989, 1078–1081 and 1508–1516 (together with three residual late antique coins). Small finds included a copper alloy hoop (SF 3006), ring (SF 3003), a complete hair pin (SF 0891), a fragment of decorated sheet with a punch design (SF 1066), and a stone pestle (SF 0988). The south wing (Rooms 18, 22 and 23) As well as the post-built structure in Room 19, which can be placed with some confidence in the 10th century (and which is consequently discussed above in phase 12), other rooms within the south wing show signs of considerable occupation during phases 12 and 13. Room 18, the former long gallery, was certainly used during the medieval period, evidenced by a number of post-holes suggesting a substantial post-built structure partly built using the levelled walls as foundations (as was also the case elsewhere in the triconch complex). Dating evidence associated with these post-holes is very limited; that which does exist indicate that they could date to any period within the 10th to 14th centuries. Phasing of these post-holes is equally problematic and it is not clear which of these post-holes are contemporary. They are tentatively placed in this phase on the basis that the 10th to 12th centuries saw the most concentrated activity in the triconch in the medieval period (based on the quantities of material recovered from the central courtyard). In the eastern half of Room 18, a series of six post-holes (5116, 5118, 5120, 5122, 5123, 5227) were noted cutting the northern wall of the room (1079) while two further postholes (5225 and 5233) respectively cut the late antique walls 1896 and 1879 (Fig. 4.31). These varied in depth between 0.22 and 0.32 m (with the exception of 5225, which was only 0.10 m deep) and were between 0.20 m and 0.34 m wide. In the western half of Room 18, ten further postholes (5015, 5059, 5061, 5063, 5066, 5114, 5127, 5146, 5183) were noted as cutting though late antique rubble demolition layers (5003 and 5068) (see Chapter 3, phase 10). These varied in depth from 0.12 m to 0.36 m, although were generally much wider (between 0.31 m and 0.48 m) than those cut into the wall tops. Four post-holes (5028, 5050, 5134 and 5150) also used wall 1079 as a foundation at the western end of Room 18. Two of these (5050 and 5150) were very substantial (0.50 m and 0.55 m deep respectively), while 5028 was much shallower (0.23 m). Rooms 22 and 23 contain further evidence of medieval occupation in the south wing of the triconch (Figs 4.32, 4.38). In Room 22 a series of 13 post-holes were recorded cutting the late antique levels 1452/5025. These post-holes (1492, 1498, 3079, 3093, 3095, 3097, 3109, 3111, 3119, 3181) varied between 0.20 m and 0.38 m in depth. However, in several of the shallower post-holes, the excavators noted that it was possible that they had not been recognised at the level from which they were cut and consequently may well have originally been deeper. They varied in breadth from 0.23 m to 0.69 m, although the identification of some of the larger features as post-holes is perhaps open to question.

4  The medieval occupation of the triconch area

137

5321

0

2m

5321

0

Figure 4.29. D-shaped structure 5321 in courtyard (16) (c. 12th century)

Figure 4.30. D-shaped structure 5321 in courtyard (16) (c. 12th century), looking southeast (1 m scale)

10m

138

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

5219

5028 5134

5015

5150

5050 5114

5183

5146

5122

5227

5123

5120

5118 5116

5225 5233

5066 5059

5061

0

5m

Figure 4.31. Post-built structures in Room 18. The dates of those cut into the wall tops are unknown

5032

5148

Room 23 5221 5260

5022/5023

5262

5015

5034 5036

1492

3097 3095

1376

Room 22 1452/5025

3093 3079

5666

1498

5219

3111

3119

3109 3181

5134 5150 5050

5114

5227

0

5m

Figure 4.32. Post-holes in Rooms 22 and 23, dating to 10th to 12th centuries or later

Four post-holes (3079, 3093, 3095/3097, 3119) formed a clear line, which was parallel with a line of six post-holes (5032, 5034, 5036, 5148, 5219, 5221) cut into the top of wall 1378 (which divided Rooms 22 and 23). The post-holes cut late 6th-century deposits (1452/5025, 3075), but overlying deposits were identified in only one case, that of 5150, which was overlain by deposit 5007 that contained 13th to 16th century pottery. In the remainder of cases the post-holes were recognised after the removal

of the topsoil (1114). Consequently, it can only be argued that they probably date to this 10th- to 12th-century period when medieval activity in the triconch seems to have been at its highest, and when the levelled walls of the late antique structures were used as foundations. The plan of any individual building is difficult to determine, although the use of the earlier buildings as a foundation implies a quite substantial structure, probably involving also the post-holes noted in Room 18 (described above).

4  The medieval occupation of the triconch area Room 23 contained only three post-holes (5260, 5262 and 5666) (Fig. 4.32), cut into an undated deposit (5022/5023) apparently overlying three graves described above (1511, 1514, 1520) (Fig. 4.18).13 These were very shallow post-holes (70–100 mm in depth) and were either cut from higher up and not recognised in the overlying late and post-medieval topsoil horizons, or were part of a very lightweight structure. They are likely to be associated with other post-holes cut into wall 1376. As with Room 22 these post-holes cannot be dated precisely and may belong to phases 12, 13, or later. No later stratified deposits were identified and the post-holes were overlain directly by the dark humic topsoil that was typical of the upper levels of the triconch. Summary/interpretation Between the late 9th and 16th centuries, the central courtyard saw around 1 m of deposition, as layers of dark humic soils and rubble built up within this former open space. Although the finds span a wide date range, the vast majority date to the 10th–12th centuries, suggesting that this was the primary period of deposition. Equally, 13 of the 17 medieval coins recovered from these deposits dated to the 10th and 11th centuries. Interpretation of these deposits is extremely difficult. It is possible that they represent garden soils or midden deposits derived from the rubbish of those who lived in the buildings that had been erected over the remains of the south and west wings. Indeed, the quantities of intrusive later pottery found within the lower deposits within the sequence could be indicative of disturbance through cultivation. Alternatively these deposits may represent deliberate levelling activity with large quantities of material brought in from elsewhere in Butrint to reclaim the area adjacent to the channel that was affected by seasonal flooding. Although the evidence is inconclusive, this interpretation is perhaps supported by evidence from elsewhere in the town (see Chapter 10). Throughout this period the courtyard remained an open area, the only structure built within its walls being the ephemeral D-shaped structure (5321) of which the function is unknown. The courtyard (Room 16) in this period contained the majority of the medieval ceramics from the triconch area, with huge quantities of amphorae, primarily from southern Italy, represented among the ceramic finds. Table wares, by contrast were very limited within this 10th- to 12thcentury group, suggesting an intriguing picture in terms of the occupants’ access to imported goods. The quantities of pottery and coins found within the courtyard have led to the suggestion that it was the site of trade activity or a periodic fair (Hodges and Vroom 2007). Indeed, the total of 53 late 9th- to 11th-century coins from the triconch is relatively high by comparison with other sites in the region. However, a high quantity of coins for this period is typical of Butrint as a whole (see Papadopoulou in prep.), and consequently the high number of coins from the triconch cannot be taken as an indicator

139

of specific mercantile activity at the site. Indeed, even when all coins from the triconch are taken into account for this period, it only represents an average rate of one coin lost every three to four years, which seems unlikely to imply significant trading activity. The south wing of the former domus saw quite intensive occupation that can probably be associated with this 10th- to 12th-century period. The post-built structures in Rooms 18, 22 and 23 can be reasonably associated with the more securely dated 10th-century (phase 12) structures in Room 19 and also those in the west wing. There is also the (admittedly negative) evidence of the limited number of finds of this period from the south wing in contrast with the substantial quantities of material recovered from the courtyard. This suggests that the area of the former south wing was actually occupied during this period and thus remained clear of deposition. Had the whole area been abandoned, we might expect a more even spread of material across the levelled remains of the buildings. The nature of this occupation is very difficult to ascertain, although the ceramic assemblage from the courtyard is intriguing in this regard. Those who were living in the triconch area clearly had access to significant quantities of imported produce shipped in southern Italian amphorae, but little table ware; the latter becomes evident as supply became more focused on the Eastern Mediterranean in the late 11th century. It is tempting to suggest a military presence, supplied with imported produce but with little access to more luxurious items like fineware. This could also perhaps be associated with the first reconstruction of the city wall (probably dating to the 11th century onwards). Alternatively, the occupation of the triconch may relate to a civilian community primarily accessing imports through military supply. However, these must remain very tentative suggestions in the absence of comparable assemblages from elsewhere.

Phase 13: Early 11th to late 12th century (Fig. 4.33) New structures During the 11th and 12th centuries, the activity in the triconch that started in the 10th century seems to have continued without interruption. Although, as noted above, in some areas phases 12 and 13 cannot be separated, in the west and south wings clear structural phases can be identified. Consequently, the separation between these phases has been maintained even though the dating evidence associated with these building is very limited. In Rooms 25 and 26 in the west wing, the stone structure of phase 12 was replaced by a post-built structure, while conversely in the south wing the post-built structure in Room 19 was replaced by a stone structure. There was also some further activity in the remains of the apsidal reception room (Room 24).

140

Room 31 1480

1473

1516

1549 5665

1542 1540

1855 1859

Room 26

1451

3178

1394 1391

1401 1396 1398

Room 23

10009

5260

10010

5262

5129

10008

5036

Room 24

20m

0

1857

5402

5142

5666 5002

1441

10011

5134 1443 1390

Room 18

Room 19

5007

coin hoard

0

Figure 4.33. Phase 13 (early 11th to late 12th century) features in the triconch

10m

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

1503

4  The medieval occupation of the triconch area

tree

1480

141

1409

Room 25 1473

1503 1516

1411

1516

Room 31

tree

1549 5665

1542 1540

1855 1859

Room 26

1857 1451

3178

0

5m

Figure 4.34. Phase 13 (early 11th to late 12th century) features in Rooms 25, 26, and 31

The west wing (Rooms 25, 26, 31) (Fig. 4.34) As noted above (see phase 12) the remains of the ephemeral stone structure in Rooms 25 and 26 were sealed by deposits 1409 and 1411 following its abandonment. Deposit 1411 was subsequently cut by three post-holes (1540, 1542, 1549) in Room 25 and a further two post-holes (1451, 3178) in Room 26. A further four post-holes (1855, 1857, 1859 and 5665) were cut into the Roman wall 1282. The post-holes were of varied size, measuring 0.15–0.48 m in depth and 0.20–0.50 m in width. A rectangular building of around 6 × 4.5 m can be extrapolated from these remains, although it is likely that other post-holes were missed during the excavation. This building clearly post-dates deposits 1411 and 1409, although the dates of these layers are not certain. Both contain single coins of the 10th century and 10th- to

12th-century pottery (as well as small quantities of material from earlier and later periods). Deposit 1411 was equated by the excavators with other deposits in the courtyard (1500, 2052, 5276), although the date range of the material within these deposits covered the 10th–16th centuries.14 Other activity in the west wing occurred in Room 31, where a crude wall of limestone blocks, mortar lumps and tile fragments (1503) was built on top of a probable late antique deposit (1516), dating to around AD 550–575 (Figs 4.34, 4.35). The wall’s construction was very rough and the stones irregular and uneven. It curved round in a rough arc from the south to the west wall of the room. It was post-dated by thin layers found inside and outside the wall, which may be related occupation deposits. On the exterior of the wall to the northeast was a layer of clay (1480), which appeared to be the base of a rough

142

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

pavement made of limestone slabs up to 0.40 m in length. On the interior of wall 1503, and within the confines of the south and west walls of Room 31, a further thin clay layer was found (1473), which may be an occupation deposit associated with the wall. Finds within this layer include residual late 6th-century pottery, a single sherd of 10th- to 12th-century pottery, and a coin of 1150. This meagre dating evidence could suggest a 12thcentury occupation date for this building, although the extent to which material is mixed within the later stratigraphy of the triconch means that this date can be tentative at best. The most likely interpretation of this building is as an animal pen associated with the post-built structure in Rooms 25 and 26, with animals kept in the southwest corner of the room. The south wing (Rooms 19, 23, 24) The most substantial structure of this period was built in Room 19, where a stone building was erected, apparently as a replacement for the 10th-century post-built structure

5079

10008 10008

5079

10009 10009 10010 10010

5142

5129 5142

5129 5080 5080

10011

10011

0

0

Figure 4.35. Wall 1503, probably dating to the 12th century. looking west (2 m scale)

Figure 4.36. Circular stone building in Room 19, probably dating to the 11th century

Figure 4.37. Circular stone building in Room 19, probably dating to the 11th century, looking east (2 × 2 m scale)

2m

2m

4  The medieval occupation of the triconch area

143

1394 1391

1391 1391

1401 1396 1394 1398

1401 1394 1441 1401 1396 1398 13961443 1441 1398 1441 1390 1443 1443 1390 1390

0

5m

Figure 4.38. Phase 13 (early 11th to late 12th century) features in Room 24

0 of phase 12 (Figs 4.34, 4.37). The remains of this building (5002) took the form of an irregular oval of tightly packed limestone rubble measuring 6.5 × 5.7 m on its exterior. It is not clear if it was a foundation supporting a superstructure or whether it was actually the collapsed remains of a stone structure. Four possible shallow post-holes (10008, 10009, 10010, 10011) identified cutting the top of 5002 could suggest the presence of an associated wooden superstructure, although their identification is by no means certain. Post-hole 5142 cut wall 5077, which may have acted as the base for a central support. This post-hole was located in almost the exact centre of the building. However, its small size (80 mm deep and 160 mm wide) suggests that it is unlikely to have played a major structural role. The stone building (5002) was found above layer 5402 (into which the phase 12 post-holes were cut). The only context separating 5002 and 5402 was a thin spread of clay (5129), which may represent a compacted surface or floor deposit relating to the earlier building. This absence of any other deposits suggests that the later building (5002) is a second phase of the earlier timber structure. It is unlikely that the stones represent the collapsed superstructure of the earlier building, as there does not seem to be a close

5m

5m earlier postrelationship 0between the distribution of the holes and the shape of the later building, although it is notable that the entrances into both structures are in broadly the same location. A terminus post quem for the stone structure is provided by an anonymous follis of AD 1030/5–1042 from 5079, a deposit below the building, as well as by its relationship with the 10th-century post-built structure. Its construction date is harder to determine although the limestone rubble of the building (5002) itself contained a limited assemblage of 10th- to 12th-century pottery and amphorae, as well as a few sherds of later medieval coarse ware. A further small group (around 25 sherds) of 10th- to 11th-century material was recovered from the overlying context 5040, which did not contain any later material. This limited evidence suggests that building 5002 probably dates to the 11th century. The fact that it is built directly above its 10th-century predecessor also supports this relatively early date. As noted in Chapter 3, Room 24 saw extensive use during the late antique period, indicated by the large number of pits and stake-holes cutting a trample layer (1300) above the mosaic pavement (3121). Room 24 was also used at some point during the medieval period although

144

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

dating evidence is very limited (Figs 4.38 and 3.53). A layer of mortar, rubble and clay (1391) developed over the industrial features, post-holes and stake-holes of phase 9 and contained very limited pottery dating from the 10th to 11th centuries as well as a single sherd from the late 15th century. A small curved structure, perhaps part of a small temporary shelter, was built above this layer indicated by post-holes 1394, 1396, 1398 and 1401 which varied in depth from 0.09 m to 0.26 m. It is also possible that some of post-holes recorded as cutting trample layer (1300; phase 8) could have been cut from higher up, and relate to later medieval activity and to the post-holes cutting (1391). Three larger post-holes or pits (1390, 1441 and 1443) (0.30–0.40 m deep) were also recorded cutting a clay layer (1402/1458) that post-dated layer 1391 noted above.15 The dating of these features is uncertain and it is possible that they date to the 16th century or later. Alternatively the sherd of late 15th-century pottery noted in layer 1391 may well be intrusive, and the features are broadly contemporary with the other phase 13 activity in the south wing.16 The coin hoard by the city wall The end of this period was marked by the deposition of a hoard of 94 billon trachea of Manuel Comnenus (1143–1180) and Isaac II Angelus (1185–1195) in an angle of the city wall (Fig. 4.33). The hoard, found in a small trench excavated to investigate the relationship between two stretches of the late antique wall, was presumably buried in a bag and had become slightly dispersed. It was found in a humic horizon around 0.70 m below the present ground surface, and just below the present topsoil. The hoard may have been deposited prior to the crossing of the Straits of Corfu by the fleet of the Fourth Crusade in 1203 (Papadopoulou 2005; in prep.). Summary/interpretation The occupation of the triconch area between the early 11th and late 12th centuries seems to represent a continuation of that which occurred during the 10th century, in which the levelled walls of the former domus were used as foundations for post-built structures and fairly rudimentary dry-stone buildings. These were noted particularly in the south and west wings. As noted above, the central courtyard remained free of buildings and instead saw extensive deposition of refuse (of which the former buildings continued to be relatively clear). The character of the material changed, however, with an increased presence of Aegean table wares (a pattern also noted across the Adriatic at Otranto) and (with the exception of the hoard noted above) a steady decline in the number of coins present in the archaeological sequence (Papadopoulou in prep.; Vroom in prep.). The occupation and status of the inhabitants of the triconch area in this period remain elusive. Although the buildings show little trace of grandeur or pretension, and are effectively little more than huts, as noted above their occupants had access to imported fineware pottery. It is clear that neither material culture nor architecture can be taken in isolation as an indicator of social status.

Phase 14: Early 13th to 14th century (Fig. 4.39) Reduced activity From the 13th century there is little sign of new buildings being erected in the triconch area, with the possible exception of a post-built structure in the north wing. However, it is clear that significant activity is occurring in the immediate vicinity of the triconch, evidenced by quantities of ceramics, glass and coins recovered particularly from the courtyard area and by a large rubbish pit dug in the north wing. With the exception of the pit excavated in the north wing, no closed or stratified contexts relating to this phase were recovered. The west wing (Rooms 27, 29) In Room 27, a single grave was inserted, cutting into the 10th-century (phase 12) wall (3260) as well as probable late antique levels (5315, 5338 and 5339) that post-dated the abandoment of circular structure 5336 (Figs 4.40, 4.41). This grave (5317) contained a sub-adult. It was oriented east–west, truncated in the east and contained a lining of limestone slabs. The skull, which had been crushed, rested on a limestone pillow. The arms were by its side, flexed at the elbow to rest on the lower chest and due to the truncation the feet were missing. The pottery found within the fill (5320) was dated to the 5th/6th century, to the medieval period and to the early Venetian period. The fact that this grave is likely to be later medieval in date highlights some of the uncertainty surrounding the graves that have been placed in phase 12. Further activity in the west wing during this phase took place in Room 29, which contained a pit (1418/1432) (Fig. 4.42). The pit was large (1.5 × 1.05 × 1 m) and partly cut the west wall of the room (1420). It contained a single sherd of 13th- to 14th-century pottery and three medieval sherds. It was cut into layer 1427, which also contained a single sherd of 10th- to 12th-century pot. Layer 1427 is also recorded as being cut by two post-holes (1425 and 1461) respectively 0.50 m and 0.38 m deep, and a third pit (1463), which was 0.40 m deep. Due to the lack of dating evidence it is hard to determine when these features were cut, but it is probably that they belong to this period or later. The north wing (Room 17) The latest phase of activity in the east part of Room 17 involved the digging and subsequent filling of a large pit (1186) (Gilkes and Lako 2004: 173–74) (Figs. 4.23, 4.43). The pit truncated graves 1191 and 1199 and cut the possible door-blocking 1196 and the layer 1193. It was sub-circular, measuring 1.70 × 1.50 × 0.65 m. The clay and rubble fill (1185) contained 294 sherds of pottery dating to the 13th–15th centuries as well as bone, glass, an iron fitting (SF 0093), and copper alloy objects (SF 0094, 0095). The dating of the ceramic assemblage in this pit suggests that the deposit primarily dates to the latter part of this period (i.e. the late 14th or even early 15th century) (Vroom 2004).

1243

1255 1257

5338 5318

1418 1420

1427

1186

1196

1249 1262 1253

4  The medieval occupation of the triconch area

1461 1463

1425

1247 1251

1245

3260 5319 5317

5276

1766

5007

0 0

10m

20m

145

Figure 4.39. Phase 14 (early 13th to 14th century) features in the triconch

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

146

5336

5338

3260 5318

5315 5319 5317

5339

0

2m

Figure 4.40. Grave 5317 in Room 27 (dating to the 13th century or later) showing relationship with 1409 and 3260

In the west part of Room 17, nine post-holes (1243, 1245, 1247, 1249, 1251, 1253, 1255, 1257, 1262) were found when the rubble make-up (1223) for the late antique mortar surface (1222) (see Chapter 2, phase 4) was removed (Figs 4.43, 4.44). These varied from 50–240 mm in depth although, again, the excavators considered it likely that they had been cut from higher up and not seen because of the similarity between the dark humic post-hole fills and the topsoil.17 The post-holes formed an L-shape representing two sides of a structure, measuring approximately 5 m and 3.5 m in length, although it is unlikely that the complete length of the shorter side was fully revealed. Six of the post-holes were in pairs of one large and one small, with the shape of the smaller post-holes suggested that the posts in these were slightly angled towards the larger and acted as supports or braces for the larger posts. The date of this structure is not clear, although the excavators considered that it was probably contemporary with the pit in the same room. Certainly no earlier medieval material was found within the room, suggesting that it probably does date to the 14th century or later, although strictly speaking it could date to any time from the 5th century onwards. The south wing (Rooms 18, 24, 28, 35) From the 13th century onwards little activity can be detected in the south wing, although it is possible that some of the features suggested above to be 10th to 12th century in date may belong to this later period. The later

Figure 4.41. Grave 5317 in Room 27 (probably dating to the 13th century or later).

medieval period saw the continued development of black humic soils across the south range. These may partly be the result of periodic flooding and partly the result of the dumping of material in marshy areas to facilitate access to the city walls and the channel-side. It was almost impossible to distinguish individual horizons in these black soils. This deposit was excavated across the southern range as 1574 and 5007 (which can probably be equated to the ‘lower topsoil’ 1114 excavated elsewhere across the triconch). It contained a limited quantity of ceramics dating from the 13th–16th centuries (with a slight weighting towards the earlier part of the period), as well as ten residual late Roman sherds, 38 13th/14th-century sherds and 23 15th/16th-century sherds. It is likely that 5007 represents the continuous build-up of deposits throughout this period, although it is possible that individual horizons were present but were not detected by the excavators (Figs 4.45, 4.46). The central courtyard (16) Considerable quantities of material of the 13th century and later were found in many of the medieval contexts excavated within the courtyard area, although the greater part of the ceramics from these contexts dated to the 10th–12th centuries as noted above. Above deposit 5301 (which overlay the D-shaped

4  The medieval occupation of the triconch area

147

1461

1425

1461

1425

1463 1463 1418 1418

0

5m

0

5m

West-facing section through Room 29 1463 29 West-facing 1462 section through Room North North

1462 1354 1354

1461

1461 Interpretation Interpretation North

1427 14641463 1416/5329 1427 1464 1416/5329 1439/5331

South

1421

1417 1417

1439/5331 pit cut

10th- to 12th-century rubble layer

pit cut

10th- to 12th-century rubble layer

0

1421

0

1m 1m

South

1421

North

South

1421

post-hole

late 6th-century late 6th-century early 5th-post-hole demolition demolition late 6th-century late 6th-centurylate 6th-century century wall early 5thdemolition floor make-up demolition century wall late 6th-century floor make-up

South

0

early 5thcentury wall early 5th1m century wall

Figure 4.42. Room 29 (plan and section) showing 13th- to 14th-century pit 1418/1432, post-holes 1425, 1461,1m 1463 and 0 underlying late antique horizons

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

148

1243 1243

1186

1196

1245 1186 1245

1255 1255

1251

1257

1196

1247 1249

1247 1249 1251

1262

1262

1257

1253

1253

0

0

5m

5m

Figure 4.43. Post-holes (cut from unknown higher level) in west part of Room 17

Figure 4.44. Post-holes in west part of Room 17, looking northwest (2 m horizontal scale) structure 5321 discussed in phase 12/13 above) was a similar deposit of humic soil and rubble (5276 and equivalents) (Fig. 4.28), which included 51 13th- to 14thcentury sherds and 122 15th-/16th-century sherds as well as 346 sherds datable to a broad medieval period.18 The

deposit also included coins of AD 976–1030, and 1075– 1080, as well as two coins of 1519–1630. This deposit may therefore represent part of a process of gradual deposition that continued into the post-medieval period. It is also possible that a stone-lined grave (1766) found

4  The medieval occupation of the triconch area

149

East-facing section across Room 18

South

North

5007 1574 1368

1606 5057=1676

5000

1m

0

Interpretation medieval build up of deposits

13th- to 16th-century build up of deposits

South

North

wall wall late 6th-century trample layer

mid - late 6th-century industrial refuse

0

1m

Figure 4.45. Later medieval horizons 1574 and 5007 in Room 18

Figure 4.46. Later medieval horizons 1574 and 5007 in Room 18 (looking west). The dark late medieval horizons found through much of the triconch area can be clearly seen in the upper part of the section overlying the late antique horizons below

William Bowden, Amy Culwick, Karen Francis, Oliver Gilkes, Kosta Lako and Jonathan Price

150

in the east portico dates to this period (Figs 4.47, 4.48). The east–west orientated grave contained the skeleton of an adult male of 27–35 years (1764). It post-dated layer 2071/2077, which contained a small and very mixed assemblage of pottery of which the latest sherd dated to the 13th century. However, this dating must remain very tentative.19 Summary/interpretation From the 13th century onwards, therefore, there seems to have been a reduction in activity in the triconch area, although the rubbish-filled pit in the north wing points to domestic occupation in the near vicinity of the triconch in the late 14th or early 15th century. This domestic

1766

0

1766

1m

occupation was of a relatively high status, or at least the inhabitants had access to a wide range of imported table wares. The partially revealed post-built structure in the north wing may also be related to this occupation, although in the absence of stratified dating evidence this can at best be a guess based on the fact that the post-holes were so shallow that they were probably cut through the lower levels of topsoil. Isolated burials were interred in the east portico and the west wing, although dating of these graves is difficult. Nonetheless, their proximity to the present ground surface suggests that they must belong to the later part of the archaeological sequence in the triconch. It seems that the central courtyard continued to be used for the occasional disposal of rubbish, although it is likely that occupation in the immediate area of the triconch was less intensive (perhaps as a result of the rebuilding of the wall circuit and the closing of the postern gate in the Merchant’s House area). However, given the mixed nature of the medieval courtyard deposits, it could legitimately be argued that they all relate to a single episode of dumping/ levelling in the 13th–14th centuries or later, and that the earlier material (discussed above in phase 12/13) derives from deposits elsewhere in the town (cf. Chapter 10). It is also possible that the ground-level was artificially raised in the area of the south wing.

Phase 15: 15th century and later

0

5m

Figure 4.47. Grave 1766 in the east portico

There is no discernable activity dating to this period in the triconch, with the exception of some limited quantities of pottery occurring in the upper horizons in the central courtyard area and elsewhere. Material relating to this period in fact seems to be concentrated in the area of the Merchant’s House (see Chapter 6).

Figure 4.48. Detail of grave 1766 (left) (1 m scale) with overall location (right)

4  The medieval occupation of the triconch area There are a few examples of early Venetian pottery (late 14th to 16th centuries) in the deposits from the courtyard (5276, 5322, 5519), together with a 16th- to 17th-century coin (Fig. 4.28). Earlier deposits in the courtyard (1500, 5301) also contained some (probably intrusive) pottery and a 15th-century coin. Three 15th- and 16th-century coins were also recovered from the topsoil. Venetian pottery was also recovered from topsoil/subsoil levels (5007, 5017) in Room 18 (together with a 16th-century coin, in Room 19), (as an intrusive find in wall rubble 5002) and Room 27 (as a probable intrusive find in grave fill 5320). Two 15th- to 16th-century coins were also found as unstratified finds from Room 24. Pottery finds from both the triconch and Merchant’s House area shows a sharp drop from around AD 1550 onwards, with only a few fragments from the postmedieval period occurring in topsoil contexts. Only one coin definitely post-dates the mid 16th century, although six Venetian coins could theoretically be as late as 1630. Summary/interpretation It is clear that the triconch area was largely unoccupied from the early 15th century onwards, with sporadic pottery finds perhaps attesting to the use of the area for occasional rubbish disposal or, alternatively, to cultivation with pottery finds resulting from manuring. The use of the area for agriculture is also suggested by the disturbed nature of the medieval horizons in the triconch, particularly within the central courtyard area.

5 6

7 8 9 10

11 12 13

Notes 1

2

3

4

The date ranges of these contexts are as follows: 1699 (includes multiple deposits): 7th, 10th–11th; 2052: 8th/9th, 10th–13th), 2071: late 7th/8th, 10th–13th; 5276: 8th/9th, 12th–16th), 5322: 8th–early 16th; 5359: 7th/8th, 10th–12th; 5361/5399: 8th/9th, 10th–12th; 5362: 5th–6th, late 7th/8th, 5539: 7th/8th, 9th/10th, late 11th–12th; 5540: 8th/9th–early 14th. Deposit 1409 contained several sherds of residual 5th/6th-century pottery and one 10th- to 12th-century sherd, as well as an Anonymous follis of 1030/35–1042. Context 1411 contained residual 6th-century pottery and one 8th- to 9th-century sherd; 26 10th- to 12th-century sherds; and 18 13th- to 14th-century sherds, together with a follis of Nicephorus II (AD 963–969). The site records suggest that 1411 extended over the well fill (1560), although the pottery dates suggest that this is unlikely. As with many of the post-holes noted in the triconch excavations, the similarity of the post-hole fills to the surrounding deposits meant that it was frequently unclear as to whether the posts were actually cut from the level at which they were recognised by the excavators. These post-holes were only seen after a short hiatus in excavation when the area was cleaned again. As with other post-built structures on the site, the excavators had

14

15 16 17

18

19

151

considerable difficulty distinguishing the black post-hole fills from the surrounding context, which was almost identical. It is quite likely, therefore, that other post-holes relating to this structure existed in the northern half of Room 19 but were missed. The later AMS date for skeleton 1512 suggests that the individual lived one or more generations later. The relationship between 1152 and 1193 was destroyed by the late medieval pit (1186) but the excavator was convinced that they were the same deposit. However, 1193 does not contain the huge quantities of finds that characterise 1152. John Mitchell was also convinced that the earrings from 1225 (SF 0498, 0499; Guest et al. 2004, 293–94) date to the 10th century rather than the 6th century, based on their similarity to later earrings and a lack of late antique parallels, and was proved correct by the radiocarbon dates. Deposit 1114 produced a very mixed assemblage of finds and is unreliable as a stratigraphic indicator. Grave 1265 probably belongs to this same 10th- to 11th-century period. DNA links between the skeletons will be discussed in full by Fenton et al. (in prep.). The finds assemblage from 5355 was ambiguous although both Paul Reynolds and Joanita Vroom consider it most likely to be a medieval deposit. The deposit (5361/5399/5539), which was found to cover the courtyard area when it was excavated in 2003, was equated with similar deposits noted in smaller trenches dug in earlier seasons (1545/1578/1807/1987/3068). Other finds recovered included a worked antler handle (SF 2700). Deposit 5301 is equated with earlier numbers 1469, 1699, 1708, 2053, 3247, 5335. The relationship between these post-holes and the underlying graves is ambiguous. In fact, 2052 appears to be a coherent 10th- to 12th-century context, while 5276 contains much later material dating from the 12th–16th centuries. However, there is no reliable way of deciding which of these contexts is the same as 1411. See also note 2 on the finds from 1409 and 1411. These features were only seen in section, although the bases were later recorded in plan. Four sherds of intrusive early Venetian pottery were also found in 1392, a late antique layer below 1391. The topsoil in Room 17 was excavated as two layers (1114 and 1112). It was postulated that the post-holes cut through 1114 but were covered by 1112, although no difference in finds between these layers could be discerned. Deposit 5276 is equated with previously excavated deposits 1411, 1500, 1553, and 2052. When first noted in a small trench adjacent to the west wing, 1500 was interpreted as a cobbled surface. However, when the deposit was excavated over a wider area, no coherent surface could be identified. Although 5276 was excavated and recorded separately from the underlying deposit 5301, the excavators were of the opinion that they were both part of the same deposit. See also note 12 on these deposits. The top of this grave was revealed during Lako’s earlier excavations.

5  The Merchant’s House and the city wall (5th to 7th centuries) William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Introduction The so-called Merchant’s House complex (plot 1) was excavated in the 2001 and 2002 field seasons. It lies immediately west of the Triconch Palace complex described in the preceding chapters, occupying an adjacent plot overlooking the Vivari Channel to the south (Figs 5.1, 5.2). It clearly forms a complex that is separate from the Triconch Palace and its earlier phases, and this separation is seemingly maintained throughout the Roman and late antique periods, and arguably into the medieval period also. This clear division was recognised early in the excavations, when three adjacent waterfront building-plots were identified, of which the Merchant’s House formed one (Bowden, Hodges and Lako 2002). The archaeological sequence identified in the Merchant’s House also clearly follows a separate trajectory to that of the Triconch Palace and consequently it has been decided to maintain this division in structuring the final report. As noted in Chapter 1, the name ‘Merchant’s House’, which was adopted early in the excavations, is slightly problematic since the commercial nature of this complex remains ambiguous. Nonetheless, the fact that the builders of the city wall in the early 6th century seem to have deliberately avoided the area, while leaving it accessible via a gate, suggests that it was considered important to maintain access between the Merchant’s House and the Vivari Channel. This access, moreover, was maintained for more than 700 years until the gate was finally blocked in the 13th century. It is wholly unlikely that the area’s function remained unchanged during this period (if it ever was commercial in the first place) but for reasons of simplicity the designation ‘Merchant’s House’ has been retained throughout. This chapter deals with the occupation of this area up until the early 7th century, including the evidence relating to the construction of the late antique city wall, while the medieval sequences are described in the following chapter. The phasing corresponds to that of the Triconch Palace

as outlined in Chapter 1, although it should be noted that the sequence prior to AD 400 (phases 1 and 2) presents a relative rather than an absolute chronology. This is because no coherent deposits earlier than the start of the 5th century were recovered; although, the relationships between the excavated structures allow the identification of the plans of the earliest buildings. Phases 1 and 2 have therefore been combined with structural changes described as a series of sub-phases (a-f). Similarly, the Triconch Palace phases 3–5 (the early to mid 5th century) are treated as a single phase for the Merchant’s House area, as the clear changes detectable in the Triconch Palace in this period are not apparent in the Merchant’s House. Nonetheless, for purposes of cross-referencing between the different excavation areas, the structure of the Triconch Palace phasing has been retained here.

Phase 1–2: 4th century and earlier Construction of earliest elements of the Merchant’s House Phase 1–2a (Fig. 5.3) The earliest structures in the complex were identified by topsoil clearance at the northernmost part of the site. This revealed a number of wall tops, just below the ground surface, although the excavations ceased before stratified deposits were reached (Fig. 5.4). The earliest of these walls (5700, 5701, 5703, 5705, 5706, 5707), built of mortared limestone blocks, formed parts of what was clearly a multiphase building of which the most coherent element was a small square structure (Room 48) formed by walls 5700 and 5701, with interior measurements of 4.8 x 5 m. The east–west wall (5703) indicates the presence of a further room or building to the north. Phase 1–2b (Fig. 5.5) At an unspecified date, a further right-angled wall (3209) was added, abutting wall 5703, to create Room 49 in the

5  The Merchant’s House and the city wall (5th to 7th centuries)

153

Merchant's House

Triconch Palace

20m

0 Figure 5.1. The Merchant’s House in relation to the Triconch Palace

Room 48 Room 48 Room 44

Room 46 Room 41

Room 43

Room 49

Room 46

Room 44

Room 49

Room 41

Room 50 Room 50

Room 37 Room 37

Room 43 Room 39 Room 47 Room 47

Room 39

Room 38 Room 38

Room 45 Room 45 0 0 Figure 5.2. The Merchant’s House with designated room numbers

10m 10m

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

154

5707

5700

5706

5705

Room 48

5703

5701 0

5m

Figure 5.3. Phase 1–2a (4th century and earlier) features in the Merchant’s House

Figure 5.4. The wall tops of Room 48, revealed by clearance of topsoil (looking southeast)

angle formed by wall 5703 and Room 48. It is possible that the east door in Room 48 was blocked at this time (5702), although access may also have been maintained between these two spaces. Room 49 was slightly smaller than Room 48, measuring 4 × 4 m. Like Room 48, its purpose is unknown but its size does not indicate a room of any pretension. Phase 1–2c (Fig. 5.6) The building was subsequently extended to the south with

the addition of Room 46 (of similar size to Room 49), formed by wall 3204. Contemporaneously, or shortly after, Room 37 was created by the construction of a series of piers (3055, 3207 and 3206), suggesting an arcaded front (Fig. 5.7). The piers were constructed using well-cut blocks of limestone, in a style not dissimilar to the 2nd-century walls 1058 and 1282 in the Triconch Palace. The eastern side of Room 37 seems to have been left partly open, with only a short length of north–south wall (3208) present. It seems likely that there were one or possibly two large openings

5  The Merchant’s House and the city wall (5th to 7th centuries)

Room 48 Room 48 5702 5702

155

5703 5703

Room 49 Room 49

3209 3209

0 0

5m 5m

Figure 5.5. Phase 1–2b (4th century and earlier) features in the Merchant’s House

Room 48 Room 48 Room 49 Room 49 Room 46 Room 46 5708 5708

3204 3204

Room 37 Room 37

3208 3208

3206 3206 3055 3055

3207 3207 0 0

Figure 5.6. Phase 1–2c (4th century and earlier) features in the Merchant’s House

5m 5m

156

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Figure 5.7. Room 37, looking south (with early piers labelled)

on the east side of Room 37, each measuring almost 4.5 m in width. This width suggests that these openings would have had wooden door lintels, rather than arches. This arrangement gives the impression that Room 37 was a workshop or store rather than residential in nature. There was no obvious sign of doors or thresholds within the openings to the south and east, although that is not to say that they were never present. It is likely that Rooms 37 and 46 where planned as a part of the same building project, probably linked to the development of the northern range of the complex. The walls share a characteristic construction of small to medium size rectangular limestone blocks bonded with firm yellow to pink yellow sandy mortar, suggesting they may be part of the same construction phase. Phase 1–2d (Fig. 5.8) At this stage, if not earlier, the long wall (3393) that forms the western boundary of the Merchant’s House complex must have been constructed. This wall (Fig. 5.9), which was later included within the fabric of the city wall, is undated apart from the fact that it predates the construction of Room 41/44 that clearly post-dates Rooms 37 and 46 described above. The only reason for suggesting that it was built later than these structures is that it seems to represent the first extension of building southwards towards the Vivari Channel, which may possibly be associated with an extension of the shoreline. The absence of doorways within this wall indicates that there was no access between the Merchant’s House area and the unexcavated area to the

west, reinforcing the impression that wall 3393 represents a boundary between the two properties. Room 41/44 (walls 3300, 3417 and 3438) clearly postdates wall 3393 (which is overlain by wall 3438), while wall 3300 clearly abuts the southwest pier of Room 37 (3206). Room 41/44 was only partially excavated, but clearly continued the line of the frontage of Room 37.1 Two entrances led to the south, of which one was marked by a threshold block (Fig. 5.10). Walls 3300 and 3417 were built of poorly finished limestone blocks, bonded with a light grey/white mortar, and in this sense appear similar to a short wall or pier (3205) added to the south face of wall 3204, which narrowed the entrance into Room 37 (see Fig. 5.7). A similar pier (3459), bonded with the same grey/white mortar as 3205 and 3300, abuts wall 5708 and is also likely to belong to this phase of construction. Room 41/44 is thus an extension of the structure comprising rooms 37, 46, 48 and 49. The open area (39) to the south of the range of buildings represented by Rooms 37 and 41/44 remained largely unoccupied in this period, although it may have been surfaced with a hard gritty silt/clay layer (4133) that was revealed in a small trench cut around the point where room 38 was abutted by the city wall (see below). A single coin dated to the 4th/5th century was found in this context, although by itself this is insufficient as dating evidence. There were no deposits separating the surface 4133 from the construction of the city wall and so it is likely that this remained the principal surface of the open area (39) for some considerable time.

5  The Merchant’s House and the city wall (5th to 7th centuries)

157

Room 48 Room 49 Room 46

Room 44

3204

3459 5708

3438

Room 41

3417

3205

Room 37

3300 3206

3393

Room 39 4133

0

5m

Figure 5.8. Phase 1–2d (4th century and earlier) features in the Merchant’s House

East-facing elevation of wall 3393 North

South 3393 City Wall 3128

obscured by tree roots wall 3438 wall 4040

threshold 4162

0

Figure 5.9. East-facing elevation of wall 3393, which dates to the 4th century or earlier

5m

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

158

Figure 5.10. Threshold in south wall of Room 41 (between walls 3300 and 3417) (30 cm scale)

Phase 1–2e (Fig. 5.11) The complex then underwent further alteration with the construction of Room 38, a long rectangular structure that extended south towards the Vivari Channel. Prior to the construction of Room 38, the spaces between the piers of Room 37 were blocked (with walls 3056, 3089, 3212) turning Room 37 into an enclosed space that was only accessible from Room 41/44 (Fig. 5.7). Probably at the same time, Room 50 (a small annex to Room 37) was created by the construction of walls 5704 (only revealed in plan) and 4035. The latter extended halfway across the gap between the east wall of Room 49 (3209) and the new wall 5704, creating the doorway between Rooms 37 and 50. The blocking of the spaces between the piers of Room 37, and the construction of Room 38 may have been part of

Room 48 Room 49

3209

Room 50

Room 46

Room 44

4035

5704

3438 3417

Room 41

Room 37

3208

3454

Room 43

3089 3206

3453

3393

4190

3212

3047 4192

3207

3055

3056

Room 39 4133

3266 3057/3273

Room 38

3228

0

Figure 5.11. Phase 1–2e (4th century and earlier) features in the Merchant’s House

5m

5  The Merchant’s House and the city wall (5th to 7th centuries)

159

Figure 5.12. Room 38, which probably predates the construction of the bath-house of the domus, visible on the left-hand side of the picture (looking south)

the same overall building campaign. Room 38 (Fig. 5.12) (comprising walls 3047, 3057/3273, 3228, 3266) measured 16.4 × 7.8 m. Walls 3057 and 3047 respectively abutted piers 3055 and 3206, which formed the southern corners of Room 37. The western wall of Room 38 (3047) also abutted wall 3212 that blocked the gap between piers 3206 and 3207. The foundation trench (4150) for the west wall 3228 of Room 38 was recorded running parallel with the wall, cut through the surface (4133) of the open area (39) (Fig. 5.13).2 Two doors (later blocked) allowed access between Room 38 and the open space to the west (39). A further door (also later blocked) allowed access from the east. The structure was built using mortared limestone blocks in a variety of shapes, suggesting that some may have been quarried from other structures. Room 38 is of particular note in terms of what it tells us about the relationship between the area of the domus and the area of the Merchant’s House. This is discussed further below, but the doorway in the eastern wall of Room 38 clearly suggests that Room 38 was built prior to the erection of the domus bath-house in the early 5th century. A separate section dug against the eastern wall of Room 38 (3057) found ceramics dated to the early 5th century overlying the footings of the building (deposit 3302), also suggesting that Room 38 was constructed prior to this period. The southern part of Room 38 was later reused in the 12th to 13th century (Chapter 6 phase 13) as the base of a tower.

On the west side of the open space or courtyard (39), a small rectangular room (43) (measuring about 4.3 × 5.3 m) was also created at probably the same time as Room 38 (Figs 5.14, 5.15). It was built in the angle created by wall 3393, which defined the western boundary of the area, and the south wall of Room 41/44 (3417 and 3438). An entrance existed in the south wall (between walls 4190 and 4192), although this was subsequently blocked by wall 4193 during or prior to the creation of Room 40 in the 6th century (Figs 5.15, 5.16).3 A second entrance was located on the east side of the room, between walls 3454 (which abutted wall 3417 of Room 41/44) and wall 3453. The structure was built using rectangular or sub-rectangular blocks of limestone bonded with coarse grey/white mortar. The alignments of Rooms 38 and 43 could suggest that they are part of the same building phase and represent the southwards extension of the Merchant’s House complex into the open area to the south, which may have been progressively reclaimed from the marshy edge of the Vivari Channel. Phase 1–2f (Fig. 5.17) The latest element in this early structural sequence was Room 47, which was added to the south side of Room 43. Its eastern wall (4105) clearly abuts wall 3453 of Room 43. Room 47 also clearly predates the construction of the city wall (3128) in the late 5th century, which truncated the southern end of the room (see below). The

3208

3205

Room 37

3300

3089 3206

3393

3212 3207

3047

3055

3056

3266

3266

3057/3273

4150

Room 38

4150 4133 3228

4133 3228 0

3228

1m

0

5m

Figure 5.13. Foundation cut (4150) of wall 3228, cutting surface 4133

Figure 5.14. The Merchant’s House under excavation with Room 43 in the immediate foreground (looking southeast)

3438

3417

3454

Room 43

3393

4190

3453

4192 3253

0

2m

4194 West

3393

4162

4190

East

4193

4192

0

3253 4076

1m

Figure 5.15 Plan of Room 43 showing original southern entrance and elevation showing entrance blocked by wall 4193

Figure 5.16. Southern wall of Room 43 (walls 4190 and 4192) with blocking 4193, overlain by later walls of Room 40

162

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Room 48 Room 49 Room 50

Room 46

Room 44

Room 41

Room 37

4101

3208

Room 43 3453

Room 39 Room 47

4105

4133

Room 38

0

5m

Figure 5.17. Phase 1–2f (4th century and earlier) features in the Merchant’s House

construction technique adopted within this wall was slightly haphazard, using limestone blocks of variable size mixed with fragments of tile bonded with white mortar that also included pieces of shell, metal and ceramics. The eastern wall (4105) is all that remains of Room 47 in its original form. No deposits inside Room 47 can be securely related to wall 4105 because the construction of wall 3253 against the western (interior) face of wall 4105 has obscured any relationship between wall 4105 and any associated floor levels. The earliest possible floor within Room 47 was represented by a level of orange/red mortar

(4197) that may represent the remains of a floor, revealed in a section excavated adjacent to the city wall overlying a layer of dark silty clay (4198). This was overlain by late 5thcentury deposits 4038 and 4039 (see below). Although the dating of Room 47 remains unclear, the fact that its position and plan clearly reflects that of Room 38, suggests that it is probably likely to belong to this pre 5th-century phase. Room 47 appears to have only been accessible via the southern doorway of Room 43, although a narrow and erratic section of wall 4105 could suggest a blocked doorway or step from the open space to the east (39).

5  The Merchant’s House and the city wall (5th to 7th centuries)

163

Figure 5.18. Room 37, showing the levelled remains of wall 4101 (2 m scale)

It was also probably around this point (or perhaps slightly earlier) that Room 37 was subdivided by the insertion of an east–west wall (4101) (Fig. 5.18). It was keyed in to wall 3208 at its east end, although its west end was not revealed. Its construction created a small rectangular space that remained connected with the smaller room (50) to the north. The original dimensions of Room 37 were restored, however, by the beginning of the 5th century when 4101 was demolished (see below phase 3–5). Summary/interpretation As noted above, the chronology of this early structural sequence is relative and can only be said to predate the early 5th century AD (or indeed the later 5th century in the case of Room 43 and Room 47), the date to which the earliest secure contexts can be ascribed (see below). However, this relative sequence does give us some indication as to the early development of the Merchant’s House, in particular in terms of its expansion to the south and its relationship with the area of the domus to the east. The nature and function of these earliest structures are obscure although it appears likely that they were part of a larger complex of buildings that lies unexcavated beneath the woodland to the north. No obvious floor levels were identified and no traces of wall plaster or veneers survived within any of the structures. The latter could indicate that these buildings served a commercial or industrial function, or alternatively that they were service rooms for an unexcavated residential complex to the north. It seems

unlikely, however, that the service area of a residential building would be placed on the river frontage that would normally be used for more public areas of the house. The gradual extension of this complex with a succession of small rooms may point therefore to commercial or industrial activity that was expanded in a piecemeal fashion over time. A major change occurred with the southwards extension of the complex towards the Vivari Channel, with the construction of Rooms 38, 43 and 47. These represent a significant change in alignment that was presumably dictated in part by the western boundary wall (3393), which certainly predates Room 43. This extension may reflect the reclamation of marshy ground at the edge of the channel. It is striking, too, that the original southern extent of the complex, prior to the construction of the western boundary and Rooms 38, 43 and 47, reflects the southern extent of the domus to the east represented by the long gallery (18) and the apsidal reception room (24), prior to the later construction of the domus bath-house (see also Chapter 8). This could suggest that the channel-side was originally much closer to the 4th-century, and earlier, domus. The construction of Room 38 is also notable in terms of the relationship between the Merchant’s House and domus areas (Fig. 5.19). As described in Chapter 2, it is likely that there was a track leading to the channel-side between the domus and the Merchant’s House, part of which was revealed as a ‘hollow way’ to the north. This track may have provided access to the Merchant’s House, first via the open-fronted building (37), the east side of which may have

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

164

Room 24

Room 37

Room 24

Room 37

Room 18

Room 18

Room 38

0

0

20m

5709

Room 24

om 37

Room 24

Room 37

Room 18

Room 18

Room 28

8

Room 38

0

20m

Room 36

0

20m

Figure 5.19. Interpretative phase plan showing changes in access to eastern part of the Merchant’s House up to c. AD 400

20m

5  The Merchant’s House and the city wall (5th to 7th centuries) opened directly onto the track and later via the doorway in the east wall of Room 38. Subsequently, access to this area was wholly blocked by the owner of the domus, who erected a wall (5709) between the apsidal reception room (24) and the east wall (5704) of the Merchant’s House. This presumably occurred at the same time that the domus bathhouse was erected (probably c. AD 400). It seems unlikely that the area in front of the long gallery of the domus (18) would ever have been accessible from this track (if the channel itself did not extend right to the gallery), although no evidence exists to confirm this. It seems probable therefore that this track was solely intended to provide access to the eastern side of the Merchant’s House. The fact that the owner of the domus was subsequently able to close off access to the eastern side of the Merchant’s House to build the bath-house could suggest that both the domus and the Merchant’s House were owned by the same individual. This will be examined in more detail in Chapter 9. Certainly, however, with the construction of Room 38 and the blocking of the arcaded front of Room 37, the focus of the Merchant’s House appears to be very much on the channel-side. The open area (39) maintained a direct relationship with the Vivari Channel although its purpose is unknown. The narrow doorways into Rooms 38 and 41/44 would perhaps argue against a commercial function as the unloading of bulk cargoes would ideally require vehicular or pack animal access. An alternative interpretation is that the complex was focused on the exploitation of the fish and shellfish of the lake, which was certainly the case at a later date. However, the difficulty of reaching the floor levels of these early structures means that this must remain in the realms of conjecture.

Phases 3–5: Early 5th century (AD 400–450) Early floors and surfaces (Fig. 5.20) The earliest stratified deposits found in the Merchant’s House area belong to this period. They relate primarily to a series of floors and rough surfaces within Room 37, which post-date the earlier blockings and sub-divisions of phase 1–2e and f. Room 37 The stratigraphically earliest deposits exposed in Room 37 suggest that by the early 5th century the room was being used for utilitarian purposes if this had not been the case before. A deposit of silt and ash, with frequent charcoal (4113=4103), was found to the north of wall 4101, which had probably been levelled by this time. It is possible that this deposit relates to fires within the room, alternatively it may represent the dumping of ash from elsewhere. Ceramics from 4113 date to the late 3rd/early 4th and early 5th centuries.4 This ashy layer was overlain by the remains of a hard opus signinum floor (4211) composed of pink and orange mortar mixed with pebbles. It survived only intermittently, but was noted in patches across the excavated eastern part of the room (Figs 5.21 and 5.22).

165

The floor was sealed beneath a thick layer of mottled grey-green clay (4100), which was markedly thicker at the south end of the room and which overlay the demolished cross-wall (4101). It may represent collapsed pisé walls, or alternatively was intended as a solid base to take the rough floor surface that overlay it (4057/4059). It contained a mixture of ceramics dated to the 2nd/3rd, late 4th/early 5th centuries and probably intrusive early 6th-century pottery, together with animal bone, vessel glass and marble fragments. Other dating evidence for the deposit was provided by six coins, four dating from the 4th century and two to the late 4th to early 5th century. Part of a curved copper alloy bracelet or fibula brooch (SF 2535) was also retrieved. Immediately overlying this thick deposit of clay was a thin (30–40 mm) and compact layer of white and yellow mortar (4057/4059), which was recorded over the majority of the room and which may represent the fragmentary remains of a floor. Its patchy condition suggests either that it had seen heavy use or had been destroyed during preparatory work for a later floor (4056, see phase 6 below). The pottery within 4057 dated to the early 5th century. Four coins were also found, all dating to the 4th century. On the south side of the room, between the later piers 3192 and 3214, a dark grey deposit of clay (4157) was recorded above 4057 and contained early to mid 5th-century pottery together with animal bone and mussel shells. It is also possible that the probable well (3191) in the southeast corner of Room 37 was cut during this period, although the only secure dating relates to the fact that it was clearly out of use by the late 5th century when it was sealed beneath a later floor (4056) (see below phase 6). The well was a circular structure, 1.62 m across, built in mortared limestone rubble (Fig. 5.23). Little excavation was carried out on its interior due to the high level of the ground water, but a well seems the most likely explanation. The proximity of the brackish Vivari Channel means that the water is unlikely to have been particularly palatable and consequently it is possible that the structure represents the head of a small underground cistern. Summary/interpretation During the early to mid 5th century, Room 37 of the Merchant’s House saw seemingly continuous activity represented by the sequence of floors in this room. The floors themselves seem utilitarian at best and their repeated replacement indicates heavy use. Although it seems unlikely that this was a residential space, it should be noted that by the mid 5th century the occupation of the adjacent domus appears to be at a similarly basic level and we should consider that domestic and artisanal activities are likely to have been combined in a single building in both areas (see Chapter 9). It should also be remembered that this succession of floors at least partly overlaps with the two most grandiose construction phases of the adjacent domus (phases 3 and 4) when the small peristyle courtyard was added around AD 400 followed by the addition of the triconch triclinium

Room 48 Room 49 Room 50

Room 46

Room 44

Room 41

Room 37

4101 3191

Room 43

Room 39 Room 47

4133

Room 38

0

Figure 5.20. The Merchant’s House showing features from phases 3–5 (AD 400–450)

Figure 5.21. Section through Room 37 (shown in Fig. 5.22), looking west

5m

3209

3210 4113 = 4103 3209

3210 4113 = 4103 Room 37

4101

4211

Room 37

4101

3214

3191

4211 3214

3191

3207 3207

mortar

0

0

mortar

2m

2m

East-facing section through southern end of Room 37

East-facing section through southern end of Room 37 South

North

South

North

3209 3207

3207

3214

3210

3214 4056

Interpretation

4057/4059

4100

4211

4211

4101

South

4056

Interpretation

4100

4057/4059

4101

South later floor

early wall

early 5thcentury floor

mid to late 5th-century pier

early wall mid to late 5th-century pier

later floor

mortar floor

4113=4103

mortar floor late 4th-/early 5th-century make-up layer/collapsed pisé wall

0

1m

early wall mid to late 5th-century pier

0

4th-century or earlier division wall

early 5thcentury floor

3210

North

ash and charcoal deposit

late 4th-/early 5th-century make-up layer/collapsed pisé wall

1m North

4211

4211

3209

0

4113=4103

ash and charcoal deposit

4th-century or earlier division wall

Figure 5.22. Section through floors and later deposits in Room 37

1m

early wall mid to late 5th-century pier

0

1m

168

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Figure 5.23. Well 3191 in Room 37, which dates to before the late 5th century, cut by pier 3192

and the enlargement of the peristyle perhaps around AD 420. However, we should probably be wary of seeing a connection between the construction activities in the two adjacent buildings, as all the excavations of Roman buildings in Butrint reveal that building activity was almost continuous.

Phase 6: Mid to late 5th century (AD 450–500) Construction of upper storey and later floors and surfaces (Fig. 5.24) Later in the 5th century new floors were inserted into the rooms of the north range (37 and 41/44), while the western range of the complex was extended with the construction of Room 47. Within Room 37 the well was sealed and piers were inserted, possibly to support a second storey. It was also possibly in this period that an internal division was created in Room 38. Elsewhere, poor quality surfaces suggest fairly utilitarian use. However, some of the finds from the various rooms indicate that before the Merchant’s House was abandoned at the end of the 5th century (see below Phase 7) the occupants had access to some luxury items. Room 37 (Fig. 5.25) During the second half of the 5th century, Room 37 was radically altered with the insertion of six large masonry piers (3192, 3210, 3213, 3214, 3215, 5709) set in lines of three on the inside of the north and south walls. The piers were square and of approximately equal size (varying from 0.66–0.69 m in width). They were solidly built, using

roughly squared limestone blocks separated by regular string courses of reused tile and bonded with a hard pinkish yellow mortar. One of the piers (3192) physically cut the masonry of the well (3191) (Fig. 5.23), clearly showing that well had fallen out of use by this time. Given that these piers only survive to a relatively low height it is difficult to judge their purpose but the most likely explanation seems that they were intended as supports for a second floor, probably simply rising vertically to support joists. It is also possible that the piers terminated at their present height and supported a raised floor (for example for a granary), although this is perhaps less likely. At some point a post was cut into the top of wall 3212, perhaps to prop up a collapsing wall or to further support the upper floor. A second post-hole was cut through the wall on the north side (3209), just behind the central pier 3210, but neither feature can be phased confidently. It is likely to have had the same function as the other post-hole. Following the insertion of the piers, the earlier mortar floors of Room 37 were replaced with a more durable surface (4056), which abutted the piers (Fig. 5.25. This surface comprised a compacted layer of clay set with polychromatic marble fragments, together with a few tiles and flat limestone pieces. It was much thicker at its southern end, becoming very thin in the north and contained ceramics dating to the early to late 5th century. Fifteen coins were also found, six of which dated to the 4th century, four to the 4th/5th centuries, and five to the 5th century. This relatively close date range suggests that the coins are not residual and thus support a mid to late 5th-century date for the floor. Small finds from the deposit included a copper alloy fibula brooch pin (SF 2624).

5  The Merchant’s House and the city wall (5th to 7th centuries)

169

Room 48 Room 49 3209 5709

4061

Room 41

Room 37

4090 3213

Room 43 4190 4193

Room 50

Room 46

Room 44

3212

4192

3210 3215 4054 4056

4058 3214 4141 3192

4203

4136

Room 39 4039

4038

Room 47

3275

Room 38

0

5m

Figure 5.24. The Merchant’s House showing features from phase 6 (AD 450–500)

Although over 36 kg of marble were recovered from the surface, the veneer laid on top of the clay was fragmentary and did not cover the whole of the room interior. Elsewhere the surface consisted mainly of limestone fragments, including one large slab. The marble was clearly derived from polychrome wall veneers, with the fragments presumably broken off during robbing of complete sheets. Floor 4056 also sealed the well 3191 in the southeast corner of the room and some marble fragments were found slumped into the subsiding backfill of the well. A fragment of an

inscription was found on one piece of marble re-used in the floor. Another fine, curved marble piece was probably part of a sigma (table) (SF 1349; Fig. 5.26). This was found to the east of pier 3214 together with a concentration of ceramic material and glass (Fig. 5.26). Several fragments of vessel glass were found including the edge of a dish with carved decoration in the form of a star, a column, an arch and possibly a face (Fig. 5.36). It has been suggested (Bowden and Mitchell 2007, 471) that this material and the marble veneers may derive from the collapse of an upper

170

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

post-hole 5709

3209 3210

3215 4054

Room 37 4056 3213

4058 3214

4141 3192

post-hole 3212 4136 0

2m

Figure 5.25. Room 37 in phase 6 (AD 450–500)

Figure 5.27. Tiles against wall 3089, possibly to support a dolium (30 cm scale)

Figure 5.26. Fragment of marble sigma and other finds in situ against pier 3214

storey, which was the focus of residential occupation, while the ground floor was used for storage. It seems, however, more likely that the marble was in situ and was used for the rough flooring described above. Four complete red floor tiles were found together in a square pattern against the eastern edge of the room (wall 3089) (Fig. 5.27). An iron nail was found driven through the narrow gap between the tiles and a second was located immediately adjacent to the north of the tiles. These tiles

were part of the floor layer 4056, but perhaps were intended to take a dolium or other storage vessel, as identical tiles were located beneath dolium 4054, the base of which was found in the northeast part of Room 37. A second dolium (4141) and a LR2 amphora, dating to the mid 5th century (4136) were located between piers 3192 and 3207 on the south side of the building. Fragments of other amphorae were also found along the south side. The dolium 4054 (Fig. 5.28) was set in a concave sub-circular depression (4053)

5  The Merchant’s House and the city wall (5th to 7th centuries)

Figure 5.28. Dolium 4054 in northeast corner of Room 37 (30 cm scale) sunk through floor 4056. Only its base survived, filled with sterile yellow silt (4055). Tiles from the floor were found, perhaps having been used to support the vessel, around its base. The second dolium (4141) (Fig. 5.29) survived in a similarly truncated fashion, also set in a shallow circular hollow (4140). The two dolia and the amphora were found in situ. Pieces of other amphorae were found nearby, including examples from Gaza, Tunisia and Samos.5 The mid 5th-century amphora date is confirmed by closely dated ceramics from the overlying deposit (4018). A terminus ante quem for the occupation of this room was provided by a sandy clay deposit (4018) up to 0.15 m deep that overlay the floor (Fig. 5.22). Pottery from this deposit dated to AD 460–475. The composition date of 4018 and that of overlying rubble layers also suggests that there was a hiatus of about 25 years between the last occupation of the room and its demolition or collapse. Perhaps the most notable find from this deposit (4018) was a lathe-turned ivory gaming piece (SF 2447) (Fig. 5.35). It was decorated with a carved cross and with Coptic motifs and inlaid with a paste decoration.6 The deposit also included 17 copper alloy coins lost or discarded as it accumulated over the floor. These comprised five coins of the 4th century; five ascribed a 4th/5th century date; five dating to the 5th century; and two to the 5th/6th century. Nearly 13 kg of marble veneer fragments were recorded under this context, but they almost certainly derive from the surface of the underlying floor (4056). Room 38 It is possibly at this stage that an internal division was inserted into Room 38, creating a smaller rectangular room within the earlier structure. The southern limit of this room is represented by an east–west wall (3275), which abuts the eastern wall of Room 38 (3057) and runs for 4.5 m before a return runs north for around 1.10 m, then appears to terminate (Fig. 5.30 and Fig. 5.12). The wall was built of reused material and was bonded with earth and a small

171

Figure 5.29. Dolium 4141 and amphora close to south wall of Room 37

Figure 5.30. Wall 3275, built within Room 38 prior to AD 550

amount of mortar. A further section of wall, around 1 m long, on the same alignment (4203) abuts the southern face of the central pier (3207) of Room 37. It is possible that this represents the northern end of wall 3275. A possible section of the same wall (3237) was revealed in a section excavated across Room 38, although this may also represent a later blocking (see phase 8 below) (Fig. 5.64). The dating of this wall is very difficult. Although the part of the wall revealed in section (3237) can be reasonably dated to the first half of the 6th century (see below), it is also likely that the west wall of Room 38 (3266) had been demolished by this time as had Room 37.7 The south wall of Room 37 must have been at least partially standing, however, when wall 4203 was erected. All that can be reasonably said is that this structure probably dates to earlier than AD 550, but it seems most likely that it belongs to the period prior to the construction of the city wall and thus before the point when much of the Merchant’s House was abandoned and demolished.

172

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Room 41/44 Probably during the second half of the 5th century, surfaces were laid within Room 41/44 that are thought to be contemporary with the mid to late 5th-century floor surface in Room 37 (4056) described above. A surface (4061) of irregularly shaped flat limestone slabs set in hard pink and white mortar was noted on the west side of the room (Figs 5.31, 5.32). Some of the slabs had apparently subsided under the weight of the masonry block 3436 built as a stair block for the city wall (see below). No dating evidence was recovered from 4061 but it was overlain by a deposit (4019) dated to very late 5th/early 6th century, while its relationship with the city wall is clearly indicative of a date prior to the early 6th century. This surface was equated with a rough floor of fragmented tiles and limestone slabs and fragments set in a brown clay matrix (4090), excavated in a narrow section on the east side of the room.8 A late 4thcentury coin was recovered from this surface, providing a terminus post quem for its creation, while it was overlain by a relatively coherent later 5th-century context (4033) that included a glass lamp (SF 2421) and a late 5th-century coin.9 These surfaces were the earliest deposits encountered in Room 41/44. Room 47 As noted above (phase 1–2f) no deposits inside Room 47 can be securely related to wall 4105 because the construction of wall 3253 against the western (interior) face of wall 4105 has obscured any relationship between wall 4105 and possible associated floor levels. However, in the late 5th century a new floor was laid in Room 47, above the possible earliest surface (4197) described above (phase 1–2f). This was represented by a layer of clay and pebbles (4038) that extended across most of the interior of Room 47 to a depth of about 80 mm (Fig. 5.33). This may itself have been a rough floor although it is more plausible that it served as a bedding layer for a pinkish-yellow plaster floor (4039) that was immediately above 4038. This surface was fragmentary or missing in places, especially around the edges of the room.

Crucially, both the plaster surface (4039) and the underlying clay and pebble layer (4038) were seen to continue beneath the later wall 3253 (built against the interior face of Room 47’s eastern wall 4105). This suggests that these early surfaces relate to wall 4105. At the north end of the room the clay and pebble layer (4038) abutted the earlier wall (4190) that formed the south side of Room 43.

Figure 5.31. Surface 4090/4061 in Room 41/44 (looking west)

Room 44 3436

4061

Room 41 4090

Room 43 0

Figure 5.32. Surfaces 4061 and 4090 in Room 41/44

5m

5  The Merchant’s House and the city wall (5th to 7th centuries)

173

Figure 5.33 Room 47 with surfaces 4038 and 4039 visible

Figure 5.34. Blocking wall 4193 in doorway between Rooms 43 and 47

Layer 4038 produced a mixed assemblage of pottery dated to the early to mid 3rd, late 4th- early 5th and the late 5th century, suggesting a deposition date in the late 5th century. Coins were recovered from both 4038 (one of possible 4th/5th century date) and 4039 (two of possible 5th/6th century date), but all are too corroded to be more certain of their date. This evidence certainly indicates that the latest surface of Room 47 dates to the later 5th century.

It many have been at this point that the doorway between Rooms 43 and 47 was blocked by masonry 4193, which was built using very irregular limestone blocks bonded with earth (Figs 5.15, 5.16, 5.34). The date of this blocking is unknown but occurred prior to the construction of Room 40 in the 6th century (or may have been part of the construction of Room 40). How Room 47 was entered after this door was blocked is unclear as there is no other visible means of ingress.

174

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Figure 5.36. Carved glass bowl fragment from deposit 4018 in Room 37, dating to AD 460–475

Figure 5.35. Ivory gaming piece from deposit 4018 in Room 37, dating to AD 460–475

Summary/interpretation The archaeology of the Merchant’s House area in the second half of the 5th century presents an intriguing picture, particularly in relation to the contemporary occupation of the domus area, with which, unsurprisingly, it shares a number of similarities. The most striking alteration was the apparent creation (or possibly reinforcement) of an upper storey within Room 37. This second storey, supported by well-built masonry piers, represents the earliest example of such a feature within the Triconch Palace area, although upper floors are subsequently clearly attested in both the domus and Merchant’s House areas. It is of note that the insertion of the piers into Room 37 broadly coincides with the appearance of silt layers within the long gallery (Room 18) of the domus that are thought to represent periodic inundations from the channel and it is tempting to link the creation of the upper floor with rising water levels. This will be further examined in Chapter 9.

The rich assemblage of marble veneer fragments used to surface the ground floor of Room 37 may well have derived from the robbing of the abandoned rooms of the domus complex, although it is of note that the quantity and variety of marble found concentrated in Room 37 is unparalleled in the domus itself, where no trace of marble veneers was found. Indeed the other material from Room 37 (including the ivory gaming piece and the carved glass bowl fragment) (Figs 5.35, 5.36) stands out from the rest of the Triconch assemblage in terms of the standard of living it represents. While there is no clear evidence that this material was in use within Room 37 (as opposed to deriving from the quarrying of the domus), this concentration does appear significant and it is tempting to see continued aspirations towards aristocratic living within this building, perhaps focused on the upper storey (Bowden and Mitchell 2007 and Chapter 9). Nonetheless, despite the use of fragments of marble veneers, the surfacing of the ground floor of Room 37 remains makeshift, although we should not overlook the effort that the owners had made to collect the veneers in the first place. Despite the presence of occasional high-status objects, the use of Room 37 appears similar to the contemporary occupation of the west wing of the Triconch Palace, with the presence of substantial pithoi set into the corners of the room. There seems little to argue against an interpretation of domestic occupation in Room 37 (rather than explicitly

5  The Merchant’s House and the city wall (5th to 7th centuries) mercantile activity). However, the standard of construction in Room 37 is noticeably higher than seems to be the case in the west wing of the Triconch (for example in the enigmatic circular structure that occupies Room 27): the piers, for example, are well mortared constructions that use tile string courses. The insertion of a masonry partition in Room 38, possibly creating a smaller internal rectangular room may also belong to this period, although the dating of this structure remains enigmatic. No occupation surfaces were identified dating to earlier than AD 500 that might be associated with this structure and its purpose remains unclear. There is little indication as to the nature of activity in the other rooms of the Merchant’s House that provided evidence of occupation in this period. The new surfaces identified within Rooms 41/44 and 47 appear utilitarian in nature, and the rooms may have been used in connection with activities associated with the channel but this must remain purely speculative. It is of note, however, that no traces of hearths were identified (in contrast to the rooms in the west wing of the Triconch Palace), which perhaps supports a suggestion of non-domestic activity in these other rooms.

Phase 7: Early 6th century (c. AD 500–525) Construction of the city wall and abandonment of the Merchant’s House (Fig. 5.37) Towards the end of the 5th century the Merchant’s House was abandoned. Its walls and roofs were robbed for building materials or otherwise allowed to collapse. These events continued into the 6th century, around the same time as a new defensive enclosure wall was constructed around the lower city (Andrews et al. 2004).10 The Merchant’s House was included within the walls, with the western boundary wall adopted as part of the new defences, and a new wall built across the property’s southern water frontage. The abandonment of the complex and the construction of the defences may be linked, but one of the most striking features of the Merchant’s House is the fact that the city wall apparently doglegs around the complex, while a gate was left in the wall at this point that apparently remained open until the 13th century. The city wall assimilated elements of many existing buildings into its circuit, with some parts of these earlier buildings included within the defended area while others were left outside the wall line. This pragmatic technique of wall-building, utilising sections of earlier buildings wherever possible, makes dating the construction of the wall circuit very difficult. In the Merchant’s House area, much of the defence is composed of the west wall of Room 47 (3393) and the southern end of Room 38 (3228), making excavations in these areas of no use for the purpose of dating the city wall. Nonetheless, the evidence recovered from excavations adjacent to the newly built sections of the wall (3128 and 3088) support a date of construction in

175

the first half of the 6th century, perhaps around AD 525. Indeed phases 7 and 8 as a whole in the Merchant’s House area, although forming a clear stratigraphic sequence, can really only be dated with any confidence to a broad period of c. AD 500–550. Room 41/44 It seems likely that the construction of the city wall caused the abandonment and partial demolition of Room 41/44. A deposit of broken roof tiles (4078) (which also included animal bone and glass) was dumped in the southeast corner of Room 41/44 on top of the poor tile and limestone surface 4090, and extended through the doorway into Room 37 (Fig. 5.38). Subsequently the entrance between Rooms 41/44 and 46 was largely blocked by the collapse of masonry from the western side of the connecting doorway. An intact block of masonry (4095) composed of rectangular limestone blocks and slabs set in hard white and pink mortar, corresponded to the fabric of the doorway (Fig. 5.39). To the south of this doorway a similar section of collapsed intact wall masonry (4084) fell on top of the rough floor (4090), sinking into the floor under its own weight. The area between these two fallen blocks of masonry was filled by a layer of predominantly pink mortar (4079) with limestone and tile rubble, which extended into Room 46. The colour and composition of this deposit indicates that it was derived from the destruction of the southern wall of Room 46. In the southwest corner of Room 41/44, however, a rectangular block of masonry (3436) was constructed on top of the rough stone floor 4061, abutting the room’s west wall (3438) (Figs 5.40, 5.41). It was constructed using roughly dressed and coursed blocks bonded by a mixture of earth and mortar and surrounding a poured core of rubble and mortar. It is clearly part of the construction of the city wall and may represent either a reinforcement of a structurally vulnerable angle of the wall or a stair block, which will be discussed in more detail below. A probable date for the construction of this section of the wall was provided by a dark soft sandy layer (4019) that extended over the whole of the western part of Room 41/44 and which apparently post-dated masonry block 3436. To the east of 4022 in Room 41, the same, or a very similar, deposit was recorded as 4081 and the two are stratigraphically equal (Fig. 5.42). The ceramic assemblage from 4019 is dated to the first half of the 6th century, while that from 4081 contained material dating from AD 460–75 to the 6th century.11 Rooms 43 and 47 On the western side of the Merchant’s House complex, a curious arrangement was adopted in which the long western boundary wall (3393) formed the basis of the new defence. The width of this wall was increased by adding a substantial new wall (3434) to its outer (west) face (Figs 5.40, 5.43). This new section of masonry had a maximum width of 1.20 m, but gradually narrowed as it followed wall 3393 towards the channel-side. It met the face of wall 3393

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

176

Room 48

Room 49 3436

Room 50

Room 46

Room 44 4022

Room 41

Room 37

3393

Room 43

3434

Room 39 4039

Room 47 4038 4188 4155

4156 4159 4185

3128 3088

Room 38

3228

3230 0

Figure 5.37. Phase 7 (early 6th century) features in the Merchant’s House

5m

5  The Merchant’s House and the city wall (5th to 7th centuries)

177

Figure 5.38. Demolition deposits in Room 41/44, looking east. This demolition may be associated with the construction of the city wall (2 × 2 m scale)

at the point where a substantial new section of east-west wall (3128) was constructed abutting the other (east) face of wall 3393 (Figs 5.37, 5.44, 5.45). This new section of wall was a much more significant construction measuring 1.64 m in width and built of limestone blocks of varying size bonded with a white gravelly mortar. It was built in quite clearly defined courses. Wall 3128 was effectively an L-shaped construction built against the earlier north-south boundary wall (3393). It is likely that this formed a corner buttress or possibly part of a projecting tower defending the gateway into the Merchant’s House complex. The presence of a tower is suggested by the fact that part of this wall appeared to extend westwards beyond the line of wall 3393 although this section was not fully defined. The new city wall (3128) truncated the southern end of Room 47 by c. 5 m, with Room 47’s east wall (4105) sliced through by the construction trench for the city wall foundations (Figs 5.46, 5.47, 5.48). This sharp-sided curvilinear cut (4156/4159) was observed running from east to west along the edge of the city wall (3128).12 It was cut through the mortar floor 4039. The cut ran for 3.70 m within Room 47 and for 1.60 m through the open area 39 and extended outwards to a maximum width of 0.40 m in the centre. It was not possible to excavate to the full extent of the foundation cut because of high groundwater. The excavations did reveal, however, a limestone foundation (4041) set in yellow mortar that was laid to take the city wall (3128). The foundation consisted of regularly sized blocks c. 0.20 × 0.10 × 0.03 m of which three courses were exposed. Following the curve of the trench cut, the

footings curved progressively inwards to the line of the wall. Three short parallel lines of building material (4155) were packed against the edge of the construction cut at its widest point presumably to lend greater solidity to the foundations (Figs 5.46, 5.47). The trench was backfilled to foundation level with green-grey silty clay (4160), which contained some mortar chunks, probably from construction of the footings themselves, and pottery and glass sherds. The pottery could go as late as the mid 6th century, although residual 2ndand 3rd-century sherds were also recovered. Although the diagnostic sherds are limited, this assemblage suggests a date in the first half of the 6th century for the construction of the late antique city defences.13 As noted above, the east wall of Room 47 (4105) was physically cut away by the construction trench (4159) for the city wall foundations. Conversely, the later east wall of Room 40 (3253) was built over the foundation trench and physically butted the city wall (Fig. 5.48). The construction of Room 40 will be discussed below (phase 9). Room 47 was presumably abandoned as a result of the construction of the city wall (if it was not previously disused). A burial (4188) was found interred inside an amphora in the southwest corner of the room, aligned with its west wall (3393) (Figs 5.49, 5.50). The vessel was a Gazan amphora dating from the late 5th century. It was set in a shallow grave cut through the mortar floor (4039) and contained a neonatal skeleton (4044). The body was supine with arms and legs extended. No accompanying grave goods were found in the grave fill (4043). The grave was immediately overlain by deposits

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

178

associated with the later 6th-century building Room 40 (3105). The upper portion of the amphora was largely missing and the remaining part bore scrapes (possibly shovel marks) from antiquity, which may relate to later clearing of the area prior to the construction of Room 40.

Figure 5.39. Detail of collapsed or demolished section of wall (4095) in door between Rooms 41 and 46 (looking west) (1 m scale)

The open area (courtyard) 39 The city wall also cut through part of the open courtyard area of the Merchant’s House (39), which was left accessible via a gate. The continuation of the cut (4156/4159) for the section of wall to the west of the gate (3128) was noted running parallel with the wall for c. 1.60 m, extending c. 0.40 m into the gate, and cutting through the compact gritty clay deposit 4133 described above (phase 1–2d). It curved in towards the line of the wall at its east end and had a maximum width of 0.32 m. The stone footings for the wall (4041) were overlain by a silty clay fill (4177) (equating with the foundation trench fill 4160 described above), which contained limited pottery dated to the 3rd and 5th centuries.14 The eastern section of wall 3088 ran for 4.74 m until it met the west wall of Room 38 (3228) (Figs 5.51, 5.52). It was constructed in a similar fashion to the west section (3128), measuring 1.58 m in width. It survived to a height of around 1.20 m. In contrast with wall 3128 the foundation levels of this section of the wall were not revealed. The next section of the defense was formed by the southern end of Room 38, which was left standing. This southern section of Room 38 presumably functioned as a makeshift tower defending the gate. The walls of Room 38 are not substantial enough to support a major structure and it is possible that a wooden superstructure formed a second storey. The east side of Room 38 was abutted by the long

3438 3436 3434

Room 44 4061

4022

Room 41 3393

0

5m

Figure 5.40. A possible stair block or reinforcement (3436) built as part of the city wall in the early 6th century

5  The Merchant’s House and the city wall (5th to 7th centuries)

179

length of wall (3230) that extended along the south side of the Triconch Palace and truncated Room 36 of the small bath-house of the domus (Fig. 9.4). The southern exterior faces of both sections of the city wall (3088 and 3128) showed clear signs of having been worn by the waters of the Vivari Channel (Fig. 5.53). The mortar had been eroded away by the water and stones had also washed out creating deep voids in the faces of the walls. It is thus clear that the wall was built right up against the waters of the channel. The gate The gate through the city wall and into the Merchant’s House area was 1.90 m wide, with the gateway itself paved with a well-made surface (Figs 5.54, 5.55 and 6.22). This was composed of a layer of sand and fragmented building rubble (4149) lain in the mouth of the gateway on top of the earlier surface (4133). On top of this a compacted stone and tile surface (4139) covering an area 1.80 × 0.95 m was deposited inside and in front of the gate. Within the gateway itself a pavement of small flat limestone slabs (4147) was laid in white mortar on top of this initial surface (4139).15 Above this a raised mortar and limestone step or threshold block (4148) was constructed, presumably acting as a stop for a door that opened inwards, although it is not known whether this was an original feature. A slot in the west wall (3128) was presumably intended to receive a door bar. Summary/interpretation The construction of the city wall clearly caused a significant reorganisation of the Merchant’s House area, and, apparently, caused the abandonment of at least some of its rooms. Certainly there is little sign of contemporary activity and it is of note that the earliest burial in the

Figure 5.41. The city wall in the northwest corner of the Merchant’s House, showing possible stair block 3436 abutting the earlier wall (3438) of Room 44

3438 3436

Room 44

3434

4022 4019 4081

Room 41 3393

0 Figure 5.42. Deposits 4019 and 4081 in relation to possible stair block 3436

5m

180

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Figure 5.43. The city wall on the west side of the Merchant’s House (looking south), showing composite construction using the earlier walls 3393 and 3438

Merchant’s House area dates to this period, reinforcing the picture of an area that was sparsely occupied. Although the wall seemingly respects the area of the Merchant’s House, this may have been opportunistic, with the wall builders taking advantage of the existing structures to build a wall that followed the channel edge. This used the natural defence of the channel and it is clear that the water lapped against the foot of the wall in certain periods. Certainly, the idea that the builders of the wall deliberately avoided the Merchant’s House is not sustained by the archaeology, which suggests that the buildings were significantly affected by the construction of the wall. The presence of the gate, however, is worthy of note, as is the fact that it opened onto the open courtyard area (39) that had formed the link between the Merchant’s House and the Vivari Channel throughout the history of the complex. Although many parts of Butrint’s late antique wall circuit remain difficult to define, gates were seemingly infrequent or at least no others have been conclusively identified (with the exception of the Hellenistic Lake Gate and Lion Gate, which apparently remained in use). This suggests that the Merchant’s House was recognised as an important means of access to the Vivari Channel, although this cannot be taken to imply continuity of ownership or function of the Merchant’s House complex itself. The form of the wall is quite singular. As noted by Andrews et al. (2004), the late antique wall was created by linking together parts of existing buildings with new lengths of masonry. This is particularly evident on the stretch of wall between the Merchant’s House and the Venetian Tower (Andrews et al. 2004, 129–31) and is clearly demonstrated at the Merchant’s House itself. On the west side of the complex a very awkward arrangement was adopted in

Figure 5.44. The city wall (3088, 3128) on the south side of the Merchant’s House, looking south

5  The Merchant’s House and the city wall (5th to 7th centuries)

East

181

West

3128

3393

3253

4159

0

tile

2m

Figure 5.45. Elevation of the north (interior) face of the section of city wall (3128) between the gate and wall 3393

4041

4155

4156

4105 4159 4185

3128 Position of the door bars

3088 3228 0

5m

Figure 5.46. The city wall showing construction cut 4156/4159

which the new stretch of wall was progressively narrowed to the point where it met the pre-existing structure. This method of construction had numerous points of weakness and would have been unlikely to provide much resistance to any sustained attack. The wall builders must have primarily relied on the Vivari Channel as the principal line of defence. Certainly the wall is a poor construction by comparison with other urban fortifications of Epirus Vetus and Epirus Nova, such as those of Onchesmos (Saranda) (Lako 1984; Bowden 2003, 91–93), Scampis (Elbasan) (Cerova 1993), Bylis (Muçaj 1990), Dyrrhachium (Gutteridge, Hoti and Hurst 2001) and Nicopolis (Hellenkemper 1987; Bowden 2003, 89–90), or even with the western defences of Butrint

itself (Bianchi et al. 2009; Molla et al. forthcoming). The form of the wall indicates that there cannot have been a continual wall walk as much of the wall was simply too narrow to sustain one. It is possible that a stair block existed on the corner of wall on the west side of Room 41/44, which would have allowed access onto a platform on this corner, although this ‘stair block’ may simply have been reinforcement for this corner of the structure. The gate into the Merchant’s House area was relatively small, although large enough to allow the passage of people and a limited amount of goods. It is hard, however, to imagine it being used to unload significant cargoes. As in the previous period there is no indication that the courtyard

182

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro area (39) was accessible to wheeled transport from the landward side, which would presumably mitigate against the use of the area as a delivery point for significant quantities of goods. The gate would have been wide enough to allow small rowing boats to be dragged through for secure storage away from the strong currents of the channel, although the door and threshold (if present) would have impeded this somewhat. Regarding its date, the available evidence suggests the city wall was built in the first half of the 6th century, possibly around AD 525. This is evident from the ceramics in the construction cut of the wall itself and from late 5thto mid 6th-century pottery from deposits that accumulated after the possible stair block was built. The ceramic assemblages contemporary with the wall construction and

Figure 5.47. Construction cut 4156. The fill of the cut is visible as a darker line in the soil in line with the edge of the possible packing material 4155 (1 m scale)

Figure 5.48. The relationship between walls 4105 and 3253 and construction cut 4159. Wall 4105 is clearly cut by the city wall (3128) and its foundation trench whereas wall 3253 clearly abuts the city wall

Figure 5.49. Burial 4188, using late 5th-century amphora (30 cm scale)

5  The Merchant’s House and the city wall (5th to 7th centuries)

3393

183

4039 4038

4188 4044

4188 4044

0

1m

0

5m

Figure 5.50. Location of burial 4188

West

East

3228

tile

3088

0

Figure 5.51. Elevation of the north (interior) face of the section of city wall (3088) between the gate and wall 3228

2m

184

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Figure 5.52. The relationship between walls 3088 and 3228 (Room 38) (1 m scale)

post-dating it have features in common with the major context from the north wing of the triconch (1152) dating to AD 525–550 (discussed in detail in Reynolds 2004). The fact that there is quite an extensive sequence of deposition belonging to c. 525–550 following the construction of the wall (in particular the well 4117, described below, which post-dates the wall but which is seemingly abandoned by 550) suggests that the wall construction is most likely to date to the start of this period. The later date for the Butrint walls may tempt some to associate its construction with the programme of fortification recorded by Procopius (Buildings IV) as taking place during the reign of the Emperor Justinian (527–565). Given the problems with this text, the fact that Butrint is not mentioned by Procopius does not in itself mitigate against a Justinianic date for the walls (Bowden 2006 with references). However, the fortification of the cities of Illyricum seems to have been an ongoing process from the reign of Theodosius II (408–450) onwards (Bowden 2003, 85) and the Butrint walls could equally belong in the reign of Justin I, or the last years of the reign of Anastasius.

Phase 8: Early to mid 6th century (c. 525–550) Continued demolition and shellfish processing (Fig. 5.56) Following the construction of the city wall, the demolition and robbing of the Merchant’s House continued through the second quarter of the 6th century, although traces of other sporadic activity can be detected. This included the use of the area for the processing of mussel catches (evidenced by a hearth in Room 41/44 and the development of a shell midden in Room 38). The open area (courtyard) 39 Following the completion of the city wall, a deposit (4051) was either laid or accumulated in the courtyard (39) (Fig. 5.57). It is recorded lying against both sections of the city wall (3088 and 3128). It comprised grey-green silty sand

Figure 5.53. The southern face of the city wall (3128) showing damage from waters of the Vivari Channel (1 m scale)

Figure 5.54. The gate in the city wall, showing pavement, 4147, threshold 4148, medieval deposit 3233, and medieval blocking 3242

with abundant grit and was interpreted as a natural deposit that had been affected by water action (either rainfall or inundation from the channel), although presumably it also functioned as a makeshift surface. Pottery from the deposit was dated from the early 5th century to the first half of the 6th century. It also included fragments of marble, glass and slag, two copper alloy objects (SF 2595 and SF 2607) and a glass bead (SF 2607). This deposit was overlain by a series of rubble and mortar layers that probably represent the deliberate demolition of the buildings surrounding the yard. At the north end of the yard several spreads of mortar were recorded (3117, 3155, 3384, 3385, 4107). Each is distinguished by the colour of the mortar, reflecting the different mortar mixes that were used in different buildings. They may represent successive demolition of the buildings fronting the yard, followed by the cleaning of blocks to be used elsewhere (represented by mortar and limestone chippings in 4107). The distinctive pink mortar of 4107 probably represents the dismantling of Room 43 whilst the others may be from the tearing down

5  The Merchant’s House and the city wall (5th to 7th centuries)

4139

185

4149

4133 4133 4147 3128

4148 3088

tile mortar

0

2m

Figure 5.55. The gate area showing surfaces 4133, 4149, 4139 and 4147

of Room 41/44. It was impossible to define the sequence of these deposits and it is likely that the demolition of these two rooms was broadly contemporaneous. Whether the buildings were wholly demolished or whether some foundations remained extant is not clear. The west wall of Room 47 (4105) was apparently demolished after Room 43, as deposit 4107 was overlain by demolition debris (3477) apparently deriving from the destruction of wall 4105 (Fig. 5.58). The mixture of clay and stones in 3477 may indicate that wall 4105 was built using both masonry and clay (pisé). Large quantities of rubbish became incorporated into the deposit; animal bone, glass, iron and worked stone. The small finds included a fragment of decorated bone inlay (SF 2557), while six coins were also found, dating from the mid 5th to 5th/6th centuries.16 In the southeast corner of the open area 39, a layer of tumbled or demolished masonry rubble and clay (4106), which overlay the water-affected silt (4051), probably derives from the wall of Room 38 (3228). The clay element of this deposit may also reflect the use of pisé walls, probably constructed on top of stone walls that stood to at least first floor level. It contained pottery dating to both the 5th and the mid to late 6th centuries. To its north an almost identical deposit (4114) probably represents the destruction of wall 3266 belonging to Room 38; it contained ceramics dating to the mid 6th century AD or slightly later. Both destruction deposits (4106 and 4114) were sealed by an extensive brown silt deposit (4050). Following these episodes of demolition, the yard area began to be used for the processing of catches of mussel

shells. These are either shelled uncooked when landed or cooked on an open fire for easy shelling and later reheating. Both processes can be observed on the shores of Lake Butrint today. Along the western side of the yard, a layer full of mussel shells (4089) was found above the demolition deposit (3477) (Fig. 5.58). This deposit (4089) was around 0.10 m thick and contained ceramics dating to the late 4th/early 5th century and to the first half of the 6th century (as well as animal bone and charcoal). A date in the first half of the 6th century suggests that the deposit accumulated following the abandonment and robbing of the principal buildings but before the later 6th-century building was erected. It was sealed by the medieval (10th- to 11th-century) road (3304). Around the same time, a small well was constructed close to the northwestern edge of courtyard 39. The circular structure (4117) (diameter c. 1.50 m, height 0.40 m) was made with limestone bonded together with white mortar (Fig. 5.59). It was only partially excavated because of the high water table. It cuts the layer (4051) that accumulated immediately following the construction of the city wall (see above) and did not contain the demolition debris that are related to the gradual destruction of the walls in this area. It may belong to the same phase as the mussel shell layer (4089) and was perhaps used to wash the mussels during their preparation. The lower fill of the well (4115) was a loose black and white/grey gravel deposit with crushed shells, limestone and tile fragments. The later fill (4108) was immediately below the medieval road (3304) and it

186

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Room 48 Room 49

4019 4082 4028

Room 43

4021

4081

Room 41

4024 4023

3385

4029 4094

3384 4107

Room 37

3090

3117 3116=3155

4117 3477

Room 50

Room 46

Room 44

Room 39 4114

4089

4051

4106

Room 38

0

Figure 5.56. Phase 8 (early to mid 6th century) features in the Merchant’s House

5m

5  The Merchant’s House and the city wall (5th to 7th centuries)

4107

3385

3385

3384

3384 3117

4107 3155

3117

3477

Room 39 4114

4114 4051

4089

4051

4106

4106

0

3116

4117

Room 39 3477

187

5m

Figure 5.57. Possible demolition levels (3117, 3155, 3384, 3385, 4107) in the area (39) in front of the gate together with possibly naturally accumulating deposit 4051

0

5m

Figure 5.58. Mussel shell-rich layer 4089, clay floor 3116 and well 4117 in the area (39) in front of the gate

did not contain shell fragments. Pottery from this context has been dated to the second quarter of the 6th century, suggesting the abandonment of the area and the end of the use of the well by AD 550. On the northern side of the open area 39 a possible clay floor (3116) was noted, overlying the mortar deposits 3117 and 3155 noted above (Fig. 5.58). This possible floor lay beneath rubble deposits (3104=3188) containing material dating to AD 525–550, while ceramics from the clay floor itself suggest a date in the first half of the 6th century. This clay floor seems therefore to belong to a period immediately following the demolition of the rooms surrounding area 39, and may be related to by the use of the northern part of the courtyard by those involved in mussel processing. Room 37 Within the former two-storey room (37) described above, a thick layer of sand/silt, mortar tiles and stones (3134/4034) was deposited above layer 4018 (the deposit containing the ivory gaming piece, which seemed to represent the end of occupation in Room 37 c. AD 460–475) (Figs 5.60, 5.61).

Figure 5.59. Probable 6th-century well (4117), apparently abandoned by AD 550 (see also Fig. 5.44)

188

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Figure 5.60. Tile deposit 3134/4034 (probably representing demolition) dating to the first half of the 6th century in the eastern half of Room 37, looking northeast (2 m scale)

The later deposit (3134/4034), which contained ceramics dating to the first half of the 6th century, was thicker on the east and west sides but much thinner in the centre and the southwest corner of the excavation. It was up to 0.51 m deep at the north end against the walls but dipped down to the south becoming much thinner (0.2 m). Several fragments of pythos jar were found within the deposit, which may derive from the storage vessels set into the marble floor (4056) within Room 37 (see above). Deposit 3134/4034 does not represent a roof collapse as most of the tile present is in small fragments, and instead may relate to the piecemeal demolition of Room 37 during the first half of the 6th century (indicated by pottery within the deposit). The final demolition of Room 37 is represented by a substantial layer of stone, mortar and tile rubble (3090), which was particularly concentrated on the north side of the room (Figs 5.61, 5.62). The abundant mortar inclusions may indicate cleaning of stonework for re-use elsewhere. Pottery from this deposit dates to AD 500/525–550 and the deposit may be analogous to the major tile dump/midden deposit 1152 from the north wing of the Triconch Palace (see Chapter 3, phase 8; Gilkes and Lako 2004, 170–71; Reynolds 2004, 228) On the west side of the room, within this demolition deposit (3090), or cut into it, were three partially complete ceramic vessels (4021, 4023, 4024), of which 4021 and 4023 may represent the use of the room for sporadic burial, although no human remains were identified (Fig. 5.61). This may be the result of disturbance or the poor survival of neonatal bones, or alternatively because the vessels were simply discarded rather than being used for burial.

A 6th-century LR1 amphora (4020) was recorded in a cut (4021) although the excavated limits of the cut were dubious as the area was considerably disturbed. Fragments from a Samos amphora were also collected under this number. The vessel was interpreted as a burial urn although some fragments of bone within the vessel proved to be animal. To its south a Samos amphora (4023) was investigated. It was highly fragmentary and contained no human bones. It too can be dated to no closer than the 6th century. The third vessel (4024) was a much smaller two-handled vessel of probable 6th-century date which was found among numerous animal bones and other pottery fragments mixed with mortar. Its size indicates that it could not have been used for burial. The widespread presence of amphora burials elsewhere in the Triconch Palace area in this period could support the interpretation of the amphorae as part of that cemetery phase, although the absence of human remains means that this is uncertain. There are no stone-lined graves in the immediate area. The vessels are “earlier rather than later” in the 6th century in the opinion of Paul Reynolds, suggesting that the structures in this area were demolished no later than 550. Room 41/44 As noted above, the deposits in the open area 39 indicate that Room 41/44 was demolished on the first half of the 6th century. Around this time, or perhaps slightly earlier, a small hearth (4082) was cut through the accumulated deposit (4081) (noted above as probably contemporary with the construction of the city wall) immediately inside

5  The Merchant’s House and the city wall (5th to 7th centuries)

189

4021

4024 4021

4023 3090

4024

4023

Room 37 Room 37

3090

0

0

2m

2m

East-facing section Room 37

East-facing section Room 37 South

North

North

South 3207

3207

3214

Interpretation

4018

South early wall

4018

4056

mid to late 5th-century pier

4056

4057/4059

late 5th-century abandonment deposit 4057/4059

late 5th-century abandonment deposit

mid to late 5th-century pier

early wall

3210

3214

South

Interpretation

3209

3090

mid to late 5th-century floor

mid to late 5th-century floor

3134

3090

early 6th-century 3134 deposit demolition 4100

early 6th-century early 5th-century demolition deposit floor

4101

4100

mid 6th-century demolition deposit

0

North

4101

mid 6th-century demolition deposit

late 4th-/early 5th-century make-up layer/collapsed pise wall

3209 3210

1m

0

1m North

early wall mid to late 5th-century pier

4th-century or earlier division wall

early 5th-century late 4th-/early 5th-century floor make-up layer/collapsed pise wall 4th-century or earlier division wall

0

1m

early wall mid to late 5th-century pier

0

Figure 5.61. Section through Room 37 deposits plus plan showing location of section and locations of fragmentary ceramic vessels in layer 3090

1m

190

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Figure 5.62. Tile, mortar and rubble layer (3090) of c. AD 525–550 in Room 37, which may represent the final demolition of the room (2 × 2 m scale)

3436

Room 44

4082

4022 4019

4081 4082

0

1m

Room 41

0

5m

Figure 5.63. 6th-century hearth 4082 in Room 41

the threshold of the room (3419). The hearth took the form of a simple semi-circular scooped pit (0.50 × 0.40 × 0.13 m) edged on its east and west sides with flat retaining stones and tile fragments and by the threshold to the south (Fig. 5.63). Its fill (4083) contained abundant mussel shell fragments, charcoal and small quantities of animal bone and vessel glass. The pottery in the hearth can be dated to no closer than the 6th century. Its use may be contemporary with the development of the spreads of mussel shells in courtyard 39 described above.

Room 38 This room was partly excavated in a trial trench in 2001 and 2002, cut through the north end of the structure between walls 3234 and 3057. This revealed a complex sequence of probable surfaces and deposits of mussel shells, most of which dated to the first half of the 6th century (Fig. 5.64). The earliest deposits in the sequence (3086 and 3211) are sticky clays that may be either trample soils on top of a surface below, or actually represent beaten earth floors in their own right. The earliest of these putative floors (3086) is dated to 500–550. The later of the two (3211)

5  The Merchant’s House and the city wall (5th to 7th centuries)

191

Room 37 Room 37

3234 3234

Room 38

3057

Room 38

3057

3051=3245=3236 3051=3245=3236

0

0

2m

South-facing section through Room 38 South facing section through Room 38 WestWest

East

4204 4204

3051 3051

32343234 3101 3101

3211 3140 3211 3140

Interpretation Interpretation

3085 3085

3237 3237

demolition

West

demolition rubble rubble

West dumping level dumping level

masonry

masonry partition partition

2m

East

30573057

3238 3084 3074 32363236 3086 3086 3238 3084 3074 1m 1m 0 0 mid 10th- to mid 10th- to 12th-centurybuild up

demolition 12th-century build up East demolition East rubble

rubble

3236 mortar floor mussel 7th-century and later 3236 mortar floor 6th-century mortar floorprocessing deposit mussel/charcoal deposit mussel 7th-century and later blocking wall 6th-century processing deposit early to mid 6th-centurymussel/charcoal 4th-century deposit early to mid 6th-century blocking wall or earlier wall floor/trample deposit floor/trample deposit 4th-century early to mid 6th-century 1m 0 early to mid 6th-century mortar floor

floor/trample deposit

floor/trample deposit

or earlier wall 1m 0

Figure 5.64. Section through Room 38, showing a series of deposits and floors that developed from the first half of the 6th century

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

192

4028

4029

4094

Figure 5.66. Early to mid 6th-century mortar floor (4029) in Room 43, looking south (1 m scale)

Room 43

tile mortar floor

0

2m

Figure 5.65. Room 43, showing probable early to mid 6th-century mortar floor (4029) and a mid 6th-century tile dump (4028)

was overlain by successive dumps of material (3101 and 3085). The latter of these (3085) was a dark silt layer rich in mussel shells, which represents the earliest example of mussel processing in Room 38, although the sparse pottery did not indicate a date beyond the 6th-century period given also by the date of the earliest surface below (3086). This deposit of mussel shells was sealed beneath thin and discontinuous mortar floors (3084 and 3140), suggesting that the building was being reused in the way that was also apparent in the abandoned rooms of the Triconch Palace to the east. A possible masonry partition (3237) constructed of reused limestone slabs and blocks bonded with clay, on top of one of these floors (3084) is also reminiscent of similar structures seen within the triconch area, although it may also be a late blocking associated with a structure represented by walls 3275 and 4203 (see phase 6 above). This phase was evidently short-lived as a dump of tile and limestone rubble (3238), also containing significant quantities of mussel shell, abutted this partition (3237) while a layer of tile fragments and demolition debris (4204) accumulated on top of the latest floor surface (3140). These demolition and rubble layers were then sealed by a thick and extensive deposit of mussel shells and charcoal (3074), probably dating primarily to the medieval period (see Chapter 6, phases 11–12).17 Room 43 Room 43 seems to have remained at least partially extant during this period although Room 41/44 (which formed the north wall of Room 43) was probably partly demolished.

The earliest deposit exposed in Room 43 was a compacted brown rubble layer (4094) with large quantities of tile fragments and mortar lumps. It was not excavated but ceramics from its surface dated to the late 5th to mid 6th century, while seven coins dated from the 4th to 6th centuries.18 Abraded pieces of animal bone and fragments of vessel glass suggest that it may have been redeposited from elsewhere. A compacted white mortar floor (4029) was subsequently laid over the rubble deposit (4094) (Figs 5.65, 5.66). This floor was intermittently evident across Room 43, which presumably reflects wear caused by usage. It apparently respects the steps that originally led into Room 47 and was cut by the later post-holes that supported the staircase of the late 6th-century building, Room 40. Aside from residual 3rd-century pottery, the floor contained material dating to the 5th/6th century, together with coins dating from possibly the 4th to 6th century.19 Floor 4029 represented the last occupation surface of Room 43 before the construction of Room 40. A tile dump (4028) was found above mortar floor in the northeastern corner of Room 43, probably representing the roof collapse after the last phase of use of the building around the mid 6th century (Fig. 5.65). Summary/interpretation During the first half of the 6th century, the buildings of the Merchant’s House underwent a process of piecemeal demolition, indicated by extensive rubble layers in Rooms 37 and 38 and in the open area 39. These seem to parallel similar activities occurring to the east in the abandoned rooms of the Triconch Palace. This demolition activity should also be seen, however, in the context of other activity taking place in Butrint, where the first half of the 6th century saw significant construction activity in the form of Christian buildings such as the Baptistery and Great Basilica (see Chapter 9). The possible presence of amphora burials in Room 37 is also paralleled much more convincingly within the Triconch Palace. While it is

5  The Merchant’s House and the city wall (5th to 7th centuries) possible that some of this material was being quarried for the construction of the city wall, in the Merchant’s House area at least much of this demolition appears to post-date the construction of the wall. It seems that Room 37 was occupied until the last quarter of the 5th century, and probably remained standing up to about 500, before the roof was presumably robbed of useful tile and the walls were dismantled during the first half of the 6th century. This demolition may also have served the purpose of creating an open area behind the city wall, which was presumably required for defensive purposes although the fact that Room 43 apparently remained standing suggests that this process was by no means rigorous. As in the adjacent Triconch Palace, the impression given is a complex one of demolition and opportunistic quarrying coupled with the use (perhaps seasonal) of the area for processing catches of mussels. In Room 38 mussel shells were identified in three successive deposits, the last of them marking the apparent abandonment of the room itself. A poor mortar floor and the construction of a rough masonry partition, subsequently sealed by the last shell midden, show, however, that some of this occupation was slightly more permanent. A layer of mussel shells denotes similar activity in the courtyard (39) where a well also suggests more permanent occupation (as does the continued supply of Mediterranean goods represented by amphorae and finewares). A hearth constructed in front of the entrance of Room 41/44, however, seems unlikely to relate to anything more than opportunistic reuse. Interestingly the evidence for mussel processing did not extend into Room 43, where a mortar surface may date to this period.

Phase 9: Mid to late 6th century (AD 550–575) A new two-storey building (Fig. 5.67) In the second half of the 6th century, the Merchant’s House was once again used as an area for domestic activity with the building of a new two-storey structure (Room 40) within the area that had previously formed Rooms 43 and 47. Room 40 The inaccurately named Room 40 is in fact a two-storey structure built within the footprint of the demolished Room 47, with an external staircase in what was Room 43 (which may have been at least partially standing). The southern and western walls of the new structure were effectively formed by the angle of the late antique city wall (wall 3128, which was newly built for the city wall, and wall 3393, which had previously formed the western boundary of the Merchant’s House before being incorporated within the city wall) (Figs 5.68, 5.69). A new north–south wall (3253) was built immediately alongside the remains of wall 4105 (the earlier wall of Room 47). The new wall was constructed on top of the remains of Room 47’s earlier mortar floor (4039) or on the make-up level (4038) at points where the mortar floor didn’t survive. At its south end it overlies the foundation offset (4041) of

193

the city defences and abuts the city wall (3128) itself (see above Fig. 5.48). At its north end it abuts east–west wall 4192 (the earlier north wall of Room 47). The new wall was reasonably well constructed though bonded with earth rather than mortar. It has five post-holes cut through its top, which are thought to be of medieval date (see Chapter 6 phase 13a). It is likely that its upper levels were composed of pisé, evidenced by the considerable quantity of clay found within deposits found within the building that probably relate to its collapse (see below 3105). A further new wall (4194) formed the northern end of the new building. The latter was built over the top of the earlier northern wall of Room 47 (4192) and the rough blocking of the door between Rooms 43 and 47 (4193). The fact that this new northern wall is recorded as having been bonded with mortar rather than earth could suggest that it belonged to a different phase than wall 3253, although the two walls appear bonded in photographs. A single well-dressed quoin block (presumably spolia) is visible on the corner of wall 4194, suggesting that some care may have been taken with the appearance of the structure (Fig. 5.70) Within the earlier Room 43, a masonry base (3232), almost certainly for an external staircase, was built against the new north wall of Room 40 (3253) (Fig. 5.71). It was made of unfinished rough limestone and tile fragments bonded with clay and measured 2.40 × 0.75 m. It survived to only 2–3 courses in height but the possible remains of steps can be seen on the eastern end. Six substantial posts were erected around this masonry base, evidenced by post-holes arranged with one at each end (4098, 4130) and four in an approximate row to the north (4096, 4085, 4111, 4120). The post-holes were all positioned either c. 1 m or 0.50 m away from the masonry base and are interpreted as supporting an upper platform at the top of the staircase. The post-holes were all circular in plan, with diameters of 0.26–0.36 m and depths of between 0.18 m and 0.50 m (Fig. 5.72). The fills varied although three of them (4096, 4120 and 4130) had limestone and/or other materials (e.g. tile fragments and pottery) used as packing around the edges. Another post-hole (4202) (measuring 0.26 m wide and 0.34 m deep), may also have belonged to the group. It was located directly against the new north wall of the building (4194) to the south of post-hole 4130. It may be a later reinforcement for the building although stratigraphically there is nothing to differentiate it from the remaining post-holes. All the post-holes cut the earlier mortar floor of Room 43 (4029), with the exception of post-hole 4130 that cuts the layer underneath the floor (4094). The mortar floor (4029) is dated to the 5th/6th century while the rubble underneath the mortar (4094) dates to the late 5th to mid 6th century (see above). This indicates that the building dates to the mid 6th century or later.20 The fills of the postholes themselves contained 5th- to 6th-century material. There were no doors or access points facing eastward into the courtyard (39) through the new wall (3253). Instead the only entrance into Room 40 was via the stone steps in the north wall leading from the earlier Room 43. Two

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

194

Room 48 Room 49 Room 50

Room 46

Room 44

Room 41 Room 37

Room 43

3232 4162 4194 4039

Room 39

4040

Room 40 3253

Room 38

0

5m

Figure 5.67. Phase 9 (c. AD 550–575) in the Merchant’s House showing location of new two-storey structure (Room 40). Post-holes are numbered in Fig. 5.69)

5  The Merchant’s House and the city wall (5th to 7th centuries)

4096

195

4085 3232

4111

4098

4120

4130 4162

4194 4202

3393 4039

4105 4040

4038

3253

Room 40

3128

0

2m

Figure 5.68. Overall view of Room 40, with Room 43 in foreground (looking south) (2 × 2 m scale)

Figure 5.69. Two-storey structure (Room 40) dating to c. AD 550–575

Figure 5.70. Northeast corner of Room 40 showing quoin on corner of wall 4194, looking southwest

Figure 5.71. Stair block (3232) at north end of Room 40, looking south

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

196

S

N

N

4096 S

N

S

4085

S

4111

N

S

N

4098

4120 S

N

4130 0

1m

Figure 5.72. Profiles of post-holes 4085, 4096, 4098, 4111, 4120, 4130, possibly supporting a balcony at the top of an external staircase

No trace of a new floor surface was found within the interior of Room 40 and it seems likely that the earlier mortar floor of Room 47 (4039) was opportunistically reused within the lower level rooms, with earlier detritus perhaps removed to this level (possibly indicated by the damage to the earlier amphora burial (4188) noted above). On the west side of the room was a lump of creamy white mortar (4040) (1.0 × 0.50 × 0.10 m) (Fig. 5.75). Its function or purpose is obscure, although it may represent the solidified contents of one or two sacks of lime that were left in the room rather than a coherent structure.

Figure 5.73. Mortar lump 4040, which may represent the solidified remains of bags of lime stored in the lower floor of the two-storey building (Room 40)

steps (4162) lead up out of Room 40 into Room 43 (Fig. 5.15). They are composed of unfinished limestone pieces of various shapes and sizes bonded together with mortar and earth. The bottom step is essentially a continuation of Room 43 wall 4190, perhaps trimmed down to this height to create the step while the second step is a new build, set back from the inner line of wall 4190. The top of the second step is level with the earlier mortar floor of Room 43 (4029).

Summary/interpretation Probably around the middle of the 6th century a two-storey structure was built in the angle of the city wall. It was a seemingly sturdy building with a tile roof (see below) although its new eastern wall may have been at least partly built in pisé. The upper floor was accessed via an external staircase and it is possible that the lower floor was used primarily for storage, although the door into this lower floor was probably too narrow for the passage of livestock. The discovery of glass, lamps, metalwork and fineware pottery within the remains of the collapsed roof could suggest that this material was from residential space on the upper floor. The presence of this material also indicates that the occupants of what appears to be a relatively modest structure continued to have access to imported goods and fineware pottery. The move towards upper storey living (discussed in

5  The Merchant’s House and the city wall (5th to 7th centuries) detail in Chapter 9) may in part be related to localised circumstances and the increased risk of flooding from the Vivari Channel although the use of upper storeys became increasingly widespread in late antiquity (see also Bowden and Mitchell 2007). There is some indication also that a two-storey building was erected in Room 18 of the southern wing of the domus in the early 6th century (see Chapter 3, phase 7). The function of this building remains open to interpretation. Its position adjacent to the gate within the city wall, and the fact that its builders were allowed to use the city wall itself to form the south and western sides of the structure, could suggest that it had some sort of official or military function. Alternatively its construction at this location could indicate that by the middle of the 6th century civil authority was so weakened that private individuals were able to co-opt key locations within the city for their own purposes (see also Chapter 9).

Phase 10: Late 6th to early 7th century (c. AD 575–650). Abandonment and collapse of Room 40 (Fig. 5.74) The two-storey building (Room 40) was abandoned probably around AD 575–600, and the structure subsequently collapsed. The Merchant’s House area was not abandoned definitively, however, as mussel processing continued to be practiced to a certain extent. Room 40 The lifespan of this two-storey structure was probably relatively short as it seems to have gone out of use in the late 6th century. Overlying the mortar floor (4039) and lying against wall 3253 was a mixture of collapsed roof tiles, occasional limestone and clay (3105) (Figs 5.75, 5.76, 5.77). Although nearly all were broken, many of the tiles were complete or almost complete, suggesting that the roof collapsed rather than being demolished. The deposit was relatively uneven, increasing from 0.10 m thickness at the north end up to 0.40 m thick at the south end. This apparent roof collapse (3105) has yielded a considerable amount of dating evidence including 6 coins of which the latest was a follis of Justin II dated 574/575.21 An assemblage of vessel glass was found, mostly from a 1 m wide strip against the west wall, as were a number of fragmented amphorae, dating mainly to the 6th century. This suggests that this material was originally stored against the west wall and was crushed when the roof collapsed. Numerous finds were discovered in front of the threshold, including an oil lamp, metalwork, and the follis noted above. The deposit also contained a bronze vessel handle in the form of a leaping hunting dog (SF 2492), which probably dates to the 4th century suggesting a long period of curation by the owner.22 Close to the surface of this deposit (3105) a broken ARS plate (Hayes 105) dated to AD 575–600 was found, which

197

together with the follis of Justin II, suggests that the collapse and abandonment of the two-storey building occurred during the last quarter of the 6th century. Following the collapse of the building the remains were probably left to weather, an episode represented by a layer of tile, clay and silt (3108) that overlay the more coherent roof collapse level below (3105) (Fig. 5.77). On the northern exterior of the building there was a gradual accumulation of material, at least some of which may have been dumped from elsewhere. All of the staircase post-holes and the staircase masonry 3232 belonging to Room 40 were sealed by a relatively thick, mixed deposit of clay, limestone fragments and pebbles (4001) (in Room 43) (Fig. 5.78). A significant 5th-century component within the ceramic assemblage suggests that at least some of this material may have been brought in from elsewhere and dumped in what was now an abandoned building. The deposit contained bone, glass, iron finds, and marble fragments as well as several fish hooks (SF 2465) and a copper alloy brooch or strap fastener (SF 2306). The fish hooks may indicate the continued use of the area by fishermen or may have been introduced from elsewhere in the same way as other elements of this assemblage. The final collapse of the walls of the building is probably represented by a layer of limestone blocks and tile fragments in brown clay and sand (4000). It was thicker along the north edge of Room 40. Some of the stone blocks may also be from the masonry staircase base. Pottery from this deposit again dates to the 5th century, and so must include residual and re-deposited material. The open area (courtyard) 39 In the second half of the 6th century or later, a small group of hearths were cut through the silt deposit (4050) along the east side of the open space in front of the gate in the city wall (39), close to the west wall of Room 38 (Fig. 5.79). The northernmost hearth (4118) consisted of a shallow sub-circular hollow (c. 0.85 × 0.50 × 0.15 m) located immediately to the south of Room 38’s blocked northern doorway (3234) (Fig. 5.80). It contained dark grey sandy silt (4119) into which were set flat pieces of limestone and fragments of tile. To the south a further hearth (4052) was placed directly on top of the silt deposit (4050) in front of the southern doorway of Room 38. In turn, the doorway was blocked (probably at the same time as the hearths were in use) by an earth-bonded limestone wall (3265) that also post-dated the silt deposit (4050) (Figs 5.81). The height of the west wall of Room 38 (into which this blocking was inserted) is unknown, as the building seems to have been partially demolished by this stage. The hearth (4052) took the form of a rectangular arrangement (c. 0.30 × 0.25 m) of small, flat limestone slabs. Scorching and reddening of the stones and the presence of charcoal and mussel shell fragments between them suggest that it was used for small-scale cooking. The third hearth (4122) (Fig. 5.82) was located further south in the angle between the wall of Room 38 (3228) and the city wall (3088). It was a

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

198

Room 48 Room 49 Room 50

Room 46

Room 44

Room 41 Room 37 Room 43

3234

Room 39

3105

4118

Room 40 4050 4052

3265

4122

Room 38

0

Figure 5.74. Phase 10 (c. AD 575–650) features in the Merchant’s House

5m

5  The Merchant’s House and the city wall (5th to 7th centuries)

199

4162

4194 3253

3393

3105

Figure 5.75. Layer of roof tiles (3105) probably representing the collapse of the two-storey building (Room 40) after AD 575 (looking north) (2 m scale)

sub-oval hollow (0.84 × 0.45 × 0.20 m) between the two walls loosely filled with limestone rubble and finished with a flat surface of stones (3471). These hearths can only be said to date to later than the mid 6th century (the date of the underlying silt deposit (4050) and the deposits immediately underlying it (4106 and 4114) noted above). They may be contemporary with the two-storey building or they may relate to the periodic use of the area for the preparation and cooking of mussels. It is possible that the large midden of mussel shells (3020) (which eventually covered this area of the Merchant’s House complex) may have begun to develop in this period. Although this midden seems to primarily date to the 10th century and later, sherds of late 6th- and 7th-century pottery within it indicate that it began to develop in late antiquity (see Chapter 6). However, as in the triconch area, material dating from the mid 7th to mid 9th centuries is almost wholly absent. Summary/interpretation Occupation of the two-storey building ceased during the last quarter of the 6th century, evidenced by a coin of Justin II

tile mortar

0

Figure 5.76. Layer of roof tiles (3105) probably representing the collapse of the two-storey building (Room 40) after AD 575

dating to 574/575 and an ARS plate of 575–600 that were the latest datable finds in the deposits associated with the building’s collapse. It is notable that the roof collapsed and that the tiles were not salvaged for use elsewhere, perhaps indicating a more generalised decline in activity within Butrint at this time. It is of course tempting to associate this abandonment with instability caused by the documented attacks of Slavs in 586/7 (Bowden 2003, 196–200 with references; Curta 2004) although the effect of these incursions is notoriously hard to quantify (see also Chapter 9). Following the abandonment of this building the area seems to have been used periodically for the preparation and cooking of mussels, an activity that probably continued throughout the early Middle Ages. This led to the development of a substantial midden of mussel shells that eventually covered much of the area immediately in front of the gate in the city wall (which remained open

2m

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

200

West-facing section through Room 40 North

4075

3381

South

4047

3128 3105

3108

3421

tile

0

1m

Interpretation North

post-hole

medieval industrial deposit

South

post-hole

city wall late 6th-century roof collapse

grave cut

0

later 6th-century abandonment

1m

Figure 5.77. Section showing the deposits in Room 40 following its abandonment

4118

4050

4001

3265 4052

tile mortar

4122 0

2m

0 Figure 5.78. Room 43, showing 4001, which covered the remains of the stair block (3232) of the 6th-century two-storey structure and which filled the associated postholes

2m

Figure 5.79. Hearths (4118, 4052, and 4122), dating to the mid 6th century or later on the interior of the gate

5  The Merchant’s House and the city wall (5th to 7th centuries)

201

Figure 5.81 Door blocking 3265, looking east (1 m scale). Note problems of high ground water

Figure 5.80. Hearth 4118 (mid 6th century or later) (1 m scale)

until the 13th century). The Merchant’s House did not see the interment of burials (particularly amphora burials) that characterise the final late antique phases of the Triconch Palace and it seems possible that the differences in use (and therefore perhaps ownership) between the two plots of land persisted even to the early 7th century. The history of the area between the 7th and mid 9th centuries remains obscure, however, with material culture of this period remaining elusive in both the Triconch and Merchant’s House.

Figure 5.82. Hearth 4122 (mid 6th century or later), looking south (2 m scale)

202

William Bowden, Ylli Cerova, Andrew Crowson and Emanuele Vaccaro

Notes 1

Room 41/44 is in fact a single space, but was designated as two separate rooms early in the excavation because of the presence of a medieval wall (4022) that divided the area. This division has been retained in order that this report is consistent with the excavation archive. 2 The construction cut (4150) was examined from the point where wall 3228 in Room 38 was abutted by the city wall (3088) up to the southernmost blocked doorway in Room 38. Excavation of the construction cut was limited to a depth of 80 mm because of the height of the groundwater table. Unsurprisingly it had steep sides, but little else about its character could be discerned and the footings of the wall were not fully revealed. 3 Wall 4190 was found below the steps (4162) and the northern wall (4193) of Room 40. 4 The latest date for this context is yielded by a large piece of ARS 50B or 64. 5 It is worth noting the presence of a fragment of a Tunisian Keay 8B, that does not accord with a mid to late 5th-century date of this phase. This type is documented in context 1152 of the Triconch Palace which has been dated to AD 525–550 (Reynolds 2004). This fragment is probably intrusive, and it can be confirmed by the late 5th-century date of the context above (4018). 6 Speculation that this was the earliest chess piece in the West caught the imagination of the world’s media in the summer of 2002, but is sadly probably unfounded! 7 The problems relating to this dating illustrate the problems caused by trying to associate deposits excavated in separate seasons in limited sections. While theoretically perfectly possible, in practice the results are often ambiguous. 8 The excavation in the part of this room designated as Room 41 (to the east of the medieval wall 4022) was limited to a 2 m wide section through the stratigraphy below a topsoil level 4033. A height difference of only 50 mm separated surface 4090 (excavated to the east of wall 4022) from 4061 (excavated separately to the west of wall 4022). 9 4033 was a number ascribed to finds from the cleaning of a layer below the topsoil (which had been removed in a previous season). The dating of these finds (of which the pottery was ascribed a date of AD 460–475) suggests that these cleaning finds probably represent a reasonably reliable context. 10 Previous publications relating to the city wall have placed it in the late 5th century (e.g. Bowden et al. 2002). However, revised dating of the ceramics from the foundation trench and overlying deposits suggests that it is later in date. 11 Dating evidence from context 4019: ARS 87B: first half of the 6th century; imported cooking pot Reynolds (1993) 6.6: late 5th century to 550; ARS 91B: 5th century; small ARS 91, cf. ARS 92, occurred in major deposit 1152 (Reynolds 2004);

12 13

14 15

16 17 18 19 20

21

22

LRC 1C?: rouletted wall: 5th century; Samian amphora, band rim, internal lid seat: 5th to mid 6th century; Gazan, steep rim, late 5th to mid 6th century; LRA 2 rim, fairly large, cf. late 5th century–c. 550 examples; casseroles with everted plain rim: usually 6th century. Context 4081: LRC 3E/F: late 5th/early 6th century; LRC 3C: c. 450–480; large LRA 2 button base: 5th century; ‘witch hat’ amphora lid: 5th-6th century; Late ARS D tall foot, from a large dish, cf. ARS 106: 6th century. (Paul Reynolds pers. comm and in prep.). The cut was numbered 4156 in Room 47 and 4159 in area 39 to the east of wall 4105 but both numbers represent the same feature. Dating evidence from 4160 comprised: ARS D base, small diameter, like a bowl, with a ring foot (too small a diameter for ARS 87, but possibly similar date – late 5th century550?); Fulford Casserole 35 rim: late 5th to mid 6th century paralleled in context 1152 (dating to c. 525–550) (Reynolds 2004); amphora handle, probably a Cretan amphora like 5383.1 (amphora burial): 6th century likely; thin-walled clibanus lid flange and rim: 5th–6th century (Paul Reynolds pers. comm. and in prep.). The sole 5th-century piece was a floor sherd of possible ARS C (perhaps ARS 84). The pavement was set flush against wall 3128 though for obscure reasons stopped several centimetres short of wall 3088 to the east. Pavement 4147 appears to have been laid immediately after the walls were built as the construction cuts for the walls physically lie directly beneath the stones. Pottery from 3477 was dated to the late 6th/early 7th centuries, although this may well be intrusive resulting from post-holes cut through the deposit in the medieval period. This layer probably equates to the extensive mussel shell midden 3020, which may represent seasonal mussel processing throughout the early Middle Ages. The coins from layer 4094 dated respectively to the 5th century (2), 5th/6th century (4) and 4th-6th century (1). Coins from floor 4029 date to the 4th/5th century (1), late 5th century (2) and 4th–6th century (1). Confusingly the post-holes are all recorded as being sealed by deposit 4000/4001 in which the material predominantly dates to the 5th century. It seems most likely that this 5thcentury material is residual or is redeposited. The remaining coins date to: earlier than the 4th century (1), 4th century (1), 5th century (1), first half of the 6th century (2). The follis of Justin II was originally thought to come from a deposit underlying the threshold of the building (hence the late 6th-century date given to the structure in Bowden and Mitchell 2007) but the excavators ultimately considered that this deposit was part of the collapse of the building. A copper alloy weight with punched dots (SF 2049) was also found in a deposit (3115) that was identified above 3105 at the north end of the building against wall 3253, but which was almost certainly part of 3105.

6  The medieval occupation of the Merchant’s House William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

Introduction The area of the so-called Merchant’s House, of which the late antique phases were described in the preceding chapter, saw intermittent occupation throughout the Middle Ages. From the 7th until the early 11th century, this occupation appears to have consisted primarily of seasonal use associated with the shelling and preparation of mussels. It was probably also used for the landing of fish and other produce from Lake Butrint and the Vivari Channel, which have left less tangible archaeological traces. During the 11th and 12th centuries, in common with the area of the former Triconch Palace to the east, there was quite intensive occupation and the construction of new post-built structures that used the levelled walls of the late antique buildings as foundations. During the 13th to 14th centuries, the gate in the city wall, which had remained open since late antiquity, was finally blocked. Following the blocking of the gate, thick black humic soils began to develop across the area in a process that lasted until the end of the 16th century. These soils were probably used for cultivation. The area was also used for a small number of burials during the 12th to 14th centuries. In the late 14th century the remains of one of the late antique structures were converted into a tower on the city wall. The Merchant’s House area provides an interesting counterpoint and source of comparison for the medieval phases within the adjacent Triconch Palace area. Together they indicate a region of the town that was peripheral from the 7th to early 11th centuries, before undergoing an intensive period of use and activity, which lasted for approximately 200 years, before ultimately returning to its peripheral state. The medieval phases of the Merchant’s House proved to be extremely complicated as the level of post-depositional disturbance (by later activity, agriculture and tree roots)

meant that a certain amount of intrusive material was introduced into what were probably earlier contexts. Equally the quantity of residual material in later contexts was correspondingly high. As elsewhere in this volume, areas of potential ambiguity or confusion are acknowledged where they occur.

Phases 11–12: Early 7th to first quarter of 11th century Mussel processing (Fig. 6.1) From the mid 7th until the early 11th century, sporadic mussel shell processing occurred within the Merchant’s House area with varying degrees of intensity, generating large shell middens within Rooms 38, 40 and the open area 39. Distinct phases of processing are suggested though their precise dating is still unclear. The hearths of the later 6th to early 7th century (see Chapter 5, phase 10) may relate to the initial growth of these middens, while later fire pits hint at periodic activity through much of the 7th/8th and earlier 9th centuries. From the later 9th century, coins and ceramics appear with greater frequency, perhaps the result of refuse dumping. This activity may be contemporary with the construction of a new sequence of hearths. Throughout this period, the Merchant’s House, with immediate access to a city gate, was an ideal location to land and process mussels from the adjacent Vivari Channel. The open area (courtyard) 39 As noted in phase 10, shell midden (3020) seems to have begun to develop during the later 6th or early 7th century (Fig. 6.2). The bulk of the midden, however, was generated during phases 11, 12 and 13a in two distinct episodes, finally extending across the majority of the courtyard’s area. It covered an area of 12.5 × 6.5 m and reached a maximum depth of c. 0.5 m.1 Medieval finds from the

204

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

Room 48 Room 49 Room 50

Room 46

Room 44

Room 41 Room 37 Room 43

3381

Room 39

Room 40

3074

3020

3351

4074 3310 3308

3297

3342 3296 3349 3285 3287 3011 3294 3318 4006 4008

Room 38

0

Figure 6.1. Phase 11–12 (early 7th to first quarter of 11th century) features in the Merchant’s House

5m

6  The medieval occupation of the Merchant’s House

205

Figure 6.2. The shell midden (3020) under excavation (looking south)

3296

3342

3349

3285

3297

3294

3287

3011 3318

4006

4008 3228

Figure 6.4. Features cut into shell midden (3020), looking north (2 × 2 m scale). At least some of these were cut from a higher level

3088

tile fired clay platform

0

2m

Figure 6.3. Hearths 3294, 3296, 3318, 4006 together with associated post-holes, cut into shell midden 3020

midden include a miliaresion and follis of Basil I dated 867–886 and 868–70, and a fragment of 8th-/9th-century cooking pot. Late 10th- to 11th-century pottery and a 10thto 11th-century copper alloy earring (SF 2042) were also recorded. Large quantities of animal bone from the midden and a worn limestone spindle whorl (SF 2297) also suggest the disposal of domestic refuse here. Some later medieval pottery dating to the 13th century is probably intrusive. The first identifiable medieval phase generated midden material to a depth of at least 0.25 m. A second later phase is indicated by the insertion of four possible hearths cut into the surface of the midden. They are all sub-oval in plan and unlike the phase 10 examples they are lined in

stone, tile or clay and measure c. 0.6 × c. 0.2 m (3294, 3296, 3318 and 4006) (Figs 6.3, 6.4). An associated ash layer (3297) was noted to the west. No dating evidence was recovered from these hearths. These features were apparently housed within a lean-to structure represented by a row of post-holes to the north (3011, 3285, 3287, 3296, 3342 and 3349), averaging 0.15–0.30 m in depth and 0.16–0.28 m in width, and by the southeast corner walls of the courtyard (3088 and 3228). A small cobbled surface (4008), which may be associated with these features, survived adjacent to the city wall (3088). A spread of ash (3250) on the cobbled surface probably derived from one of these hearths. Room 38 To the east of the courtyard a further large shell midden (3074) was partially excavated within the confines of the earlier Room 38. It originally extended for c. 15 m north–south and for c. 4.85 m east–west and is of identical

206

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

Room 43

3381

Room 40

3351

3310

4074 3308

0

2m

Figure 6.5. Shell midden 3381 in Room 40 with post-holes 4074, 3351, 3310, 3308

composition to the midden of the courtyard (3020) and that of Room 40 (see below). Like the courtyard midden it probably began to develop in the 6th or 7th century. Certainly pottery of the 6th century was present in the deposit, although it may be residual. Medieval dating evidence for the midden (3074) in Room 38 (where it was only revealed in a narrow section across the room) is represented by a single complete local cooking pot of 8th-/9th-century date, confirming the evidence from the courtyard for mussel processing in this period (Fig. 5.64). Rooms 40 and 43 A third shell midden (3381) occupied the entire interior of Room 40 and overlapped slightly into the courtyard, extending 8 m north–south and 3.5 m east–west (Fig. 6.5). Four post-holes with post packing were noted that seem to coincide with the eastern extent of the shell midden in the southern area of the room (3308, 3310, 3351, 4074). These posts represent an ephemeral structure, which probably stood for at least part of the time that the midden was being created and which thus prevented its further spread to the east. The midden was devoid of dating material though its

stratigraphic position is contemporary with those of the courtyard and Room 38. Summary/interpretation The level of activity within the Merchant’s House area from the 7th until the early 11th century is difficult to establish, but it seems likely that the presence of the gate within the city wall made this a convenient location to land and process mussels. This processing, which often involved the cooking of the mussels on small temporary hearths prior to shelling, led to the development of a midden of mussel shells that ultimately extended across much of Area 39 and the levelled remains of Rooms 38 and 40. The frequency and intensity of this activity is difficult to assess, and it is impossible to know whether it continued on an annual basis from the 7th century onwards. Certainly from the late 9th century onwards the level of material culture present in terms of coins and pottery clearly increased but whether this actually reflects burgeoning activity or merely the increasing presence of archaeologically identifiable material is difficult to say. From the later 9th and 10th centuries we see additional refuse appearing in this area although not to the extent that

6  The medieval occupation of the Merchant’s House

207

Room 48 Room 49 Room 50

Room 46

Room 44

Room 41 Room 37 Room 43 4098 4036 4011

4130 4076 4062

4025 4045 4064 4068 4013 4066 4016

4109 3312 3253 4070

Room 40

Room 39

3409 4196

3216 3359

3299

4185

4088 3320 4212

3304 3413=3197

3314 3282 3304

4181 4183

3305 3283

Room 38 3051

tile

0

Figure 6.6. Phase 13a (Early 11th century (after AD 1025) to 12th century) features in the Merchant’s House

5m

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

208

4196

4185

4181 4183 0

2m

Figure 6.7. Post-holes (4181, 4183) and pits (4185, 4196) in Room 38. Date uncertain but probably 11th to 12th century

we see in the area of the Triconch Palace during the 10th century, suggesting that the Merchant’s House remained a peripheral area in Butrint during this period.

Phase 13a: Early 11th century (after AD 1025) to 12th century

Figure 6.8. Post-holes (4181, 4183) and pit (4185), looking west. These features were recognised as cuts in the shell midden 3074, although may have been cut from a higher level

New post-built structures (Fig. 6.6) Probably during the later 10th or 11th centuries, a cobbled surface was created in front of the gate, which may conceivably be part of a road running north towards the former forum area. The processing of mussels seems to have continued after the creation of this road as some of the shell deposits described above are recorded as postdating it. During the 11th or 12th centuries a strongly built rectangular post-built dwelling of possibly two storeys was constructed above the remains of Rooms 40 and 43 (sealing the deposits of mussel shells) and against the city wall, suggesting that the latter was still upstanding or had been newly rebuilt. A foundation deposit containing a complete cooking pot of the 11th to 12th century provides a likely date for this structure. In the adjacent courtyard (39), a long north–south wall was constructed, subdividing the area, while a series of deposits developed or were dumped over the remains of Room 38. Room 38 At around this time Room 38 and the ground to the east saw the deposition of layers of clay, silt and small pieces of rubble and tile (3051=3196=3236 =3239=3245=3246=4168) (see Fig. 5.64). There is little indication that these deposits were

intended to form a coherent surface. They contained pottery dated to the mid 10th to the 12th century. A group of two post-holes (4181, 4183) (0.28 m wide, 0.27 m deep and 0.29 m wide, 0.08 m deep) and two shallow pits (4185, 4196) found cut into the earlier mussel shell layer (3074) may belong to this period. They were only recognised at the level of the earlier mussel shell deposit but their shallow depth and the presence of mid to late Byzantine pottery in the fill (4191) of pit 4185 suggest that they were cut from a higher level. No obvious structure could be discerned (Figs 6.7, 6.8). The open area (courtyard) 39 Probably during the later 10th or 11th centuries, open area 39 was divided by a new low dry stone wall. It is unclear as to what extent the walls of Room 38 were still standing as they are thought to have been partly demolished in the mid to late 6th century (see Chapter 5 phase 9). The dry stone wall (3216=3283=3299; hereafter 3216), c. 0.45–50 m in width, was constructed of uneven courses of limestone blocks bonded in clay and followed a sinuous north–south course for 14 m (Fig. 6.9). It was founded upon a bedding layer (3359) comprising stone and tile

6  The medieval occupation of the Merchant’s House

Figure 6.9. Wall 3216 and cobbled surface 3305, looking north from city wall (2 × 2 m scale)

209

fragments, which contained 10th- to 11th-century pottery as well as residual 5th- and 7th-century material. Three gaps within this wall mark the site of possible doorways. The northernmost of these was a relatively well-constructed door 0.75 m in width while the central and southern gaps were 1 m in width, the latter opposite the doorway of the building on the west side of the Merchant’s House and adjacent to the city gate. This low wall seems unlikely to have supported any superstructure and it seems most likely to have served to delineate the cobbled surface to the west. The area to the east of the wall was surfaced with beaten earth (3413=3409=3197; hereafter 3413), sealing the remnants of shell midden (3020) and bedding layer (3359). The surface was composed of compacted sandy clay with large quantities of limestone chips. Limited ceramics of the 10th to 11th century were recovered from one section of this floor (3413) with some early 14th-century fragments (from 3197) considered intrusive. Before the cessation of the mussel processing described above, a cobbled surface was laid within the western half of the courtyard in an area delineated by the long dividing wall (3216) (Fig. 6.10). The surface seemingly extended through the entrances in this wall. The earliest level was a layer of rounded and flat limestone fragments (3304), containing 10th- and 11th-century pottery, which was later replaced by a second surface of stone and tile fragments (3305).2 The surface extended for over 14 m from the threshold of the city gate at the south towards the northern limit of the excavations. It was 4.5 m wide adjacent to the

3409

3409

3359

3359

3304

0

3305

3413=3197

10m

0

3413=3197

10m

Figure 6.10. Plans showing extent of successive cobbled surfaces 3304 and 3305 and 11th- to 12th-century post-built structure in Room 40

210

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

4098 4098

4130 4130

4036 4036

4076 4076 4062 4062

4011 4011

4064 4064

4025 4025

4045 4045

4013 4013

4016 4016 3312 3312

4088 4088 3320 3320

4068 4068 4066 4066

3216 3216 3314 3314

4109 4109 3253 3253

Structural Structural posts posts

4070 4070

3282 3282

Partition Partition posts posts Roof Roof posts posts Foundation Foundation deposit deposit

0 0

door door

0 0

2m2m

2m2m

Figure 6.11. 11th- to 12th-century post-built structure above Room 40 gate, narrowing to c. 4 m to the north. This may represent a yard surface, or alternatively a metalled road running from the gate towards the former forum area.3 The earliest level of this road or yard surface lay beneath some of the mussel shell deposits noted above (3020) and it seems likely that the use of the gate for landing and processing mussels continued after its construction. It was noted that the area immediately to the north of the gate was free of mussel shell deposits, suggesting that the access was kept clear. Rooms 40 and 43 Perhaps slightly later, a sturdy rectangular post-built dwelling was constructed within Rooms 40 and 43 almost exactly following the footprint of the earlier house of the mid 6th century (see Chapter 5, phase 9). Unlike the surface to the east, it apparently sealed the deposits of mussel shells within Room 40, suggesting that it slightly post-dated the cobbled surface. It measured 8 m in length and was 3.5 m wide, defined on the east by a series of post-holes and by the city wall on the west and southern sides. Access was through a door in the east wall. Internal posts supported a

probable lean-to roof that projected beyond the east wall, supported by an external row of smaller posts, allowing rainwater to fall onto the courtyard’s cobbled surface (Figs 6.11, 6.12). The east wall was represented by seven structural postholes (4062, 4064, 4066, 4068, 4070, 4076, and 4109), c. 0.25–0.35 m in width and varying in depth from 0.12 m to 0.30 m. A 1.8 m wide gap between the southern posthole (3282) and the city wall, represented by the partial removal of wall (3253), indicates the site of a doorway to the interior, opposite the southern door of the courtyard shed and adjacent to the city gate. The east wall supports are paired to a series of slightly smaller external post-holes (3320, 3282, 3314, 4088, and 3216) cut into the levelled top of the east wall of Room 40 (3253). The relationship between these external posts and the line of post-holes immediately inside wall 3253 is not clear although it is possible that the outer posts formed the eastern limit of the building’s roof. The roof was presumably of lean-to construction built against the city wall, allowing rainwater to fall onto the cobbled road to the east. It is clear, however, that a number of reconstructions

6  The medieval occupation of the Merchant’s House

W

E 4036

S

N

4130

N

E

W

4098 E

W 4011 S

S

211

4076

N

S

N

W

4062 E 4088

4045 W

W

E

4025 W

N

3320 N NW

E

S

SE 4068 4066 3216

4016 W

4064

S

E

4013 W

E

NW

E

N

3308 N

S N 4070

SE 3314

S

S 4109

3282

0 0

2m Profiles

Figure 6.12. Profiles of the post-holes of the 11th- to 12th-century structure above Room 40

1m

212

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani been domestic or commercial/official (given its immediate proximity to the gate). This increase in activity is similar to that noted within the Triconch Palace area (see Chapter 4, phases 12 and 13) and is associated with an apparent intensification of occupation within the walled city, which was accompanied by the establishment of more formalised boundaries and divisions within the town (see Chapter 10).

Phase 13b: 12th to early 13th century Construction of new structures, occupation, and gradual decline (Fig. 6.14)

Figure 6.13. 11th to 12th-century cooking pot from pit 4025, interpreted as a foundation deposit of the post-built structure in Room 40

are possible, and indeed the two lines of post-holes may belong to different phases of the same structure. The southwest angle of the city wall functioned as the west and southern walls of the dwelling, indicating the defensive wall was still upstanding in this area or had been rebuilt. Two post-holes found against the west city wall (4016 and 4098), (c. 0.20 m deep and 0.25 m wide), provided additional support for the superstructure of the dwelling. In line with these supports, a small pit (4025) contained a complete cooking pot of the 11th to 12th century, which has been interpreted as a foundation deposit indicating a date of construction for the dwelling (Fig. 6.13). A similar foundation deposit containing a cooking pot of the 10th to 11th century was excavated from a medieval dwelling in the former Triconch Palace (Chapter 4, phase 12 Room 19). Within the interior of the new building, four centrally placed post-holes, (3312, 4011, 4013 and 4045) (c. 0.20 m deep and c. 0.25 m wide), may have acted as central supports for the building’s superstructure and/or mark the site of an internal partition wall. Summary/interpretation During the period between AD 1025 and the 12th century, the occupation of this area of Butrint appears to have become slightly more formalised and coherent. The earliest evidence of this was a cobbled surface laid down on the interior of the gate. This may represent a road running from the gate towards the interior of the city. This putative road was delineated to the east by a long wall, which probably also defined a yard space to the east. Probably at around the same time, or slightly later, a substantial structure was built above the foundations of the mid 6th-century Room 40. The building contained no obvious indication as to its function, which could have

From the 12th until the early 13th century, the Merchant’s House saw relatively limited activity, with occupation probably mainly relating to the presence of the city gate. Within the courtyard, one or more irregularly shaped timber buildings were constructed to the east of the dividing wall (3216). Within the southeast corner of Room 40, the timber building was replaced by a sequence of four successive hearths, sheltered by a small post-built structure. The open area 39 It was probably during this phase that a possible subrectangular timber building or buildings were constructed to the east of wall 3216 (see phase 13a). Twelve post-holes were identified cut into the earlier beaten earth floor of this area (3413) and the earlier layer 3359 (Fig. 6.15 and also 6.4). These post-holes (3138, 3144, 3147, 3149, 3151, 3153, 3289, 3291, 3293, 3346, 3426 and 3428) were an average of 0.30 m wide and 0.20 m deep, and most retained some evidence of packing stones (Fig. 6.16). One interpretation of these post-holes (and perhaps the most likely) sees one structure formed by the northern group of post-holes (3138, 3144, 3147, 3149, 3151, 3153, 3426 and 3428), and a second formed by the southern group of post-holes (3289, 3291, 3293 and 3346) (Fig. 6.17a). The northern building would have been a sub-rectangular hut of 2.5 × 1.5 m, while that to the south was a somewhat larger structure 2.5 m long and with northwest and southeast walls of 2.5 m and 1.7 m respectively. An alternative interpretation could see a single northwestsoutheast orientated building, 9 m in length and 4.20 m in maximum width narrowing to 2.20 m at its northern extent, where it follows the curve of the wall (3216), and 1.5m at its southern end which turns markedly to the east (Fig. 6.18b). Its western side would have comprised post-holes 3138, 3289, 3291, 3426 and 3428, and the eastern side would have comprised post-holes 3151, 3153, 3293 and 3346. The putative northern end is represented by a single post-hole (3144). Towards the northern end, two post-holes, c. 0.25 m in average width and depth without post packing (3147, 3149), may indicate the site of a possible partition. In both reconstructions the buildings seemingly retained the earlier beaten earth surface (3413). Perhaps contemporaneously the northern doorway of the existing low boundary wall (3216) was blocked by the insertion of

6  The medieval occupation of the Merchant’s House

213

Room 48 Room 49 Room 50

Room 46

Room 44

Room 41

Room 37

Room 43 Room 39

3393

3216 3298

3424

3428 3153 3426 3144 3151 3147 3138 3149 3359

3381

3413 3305

Room 40 3322

4002 4047

3289

3308 4072 3357 3450

3291

4004

3352 3452

3346

3270

3293

3265

3464 3423

3228

3128 3088

Room 38

3088 0

tile

5m

Figure 6.14. Phase 13b (12th to early 13th century) features in the Merchant’s House

wall 3298 constructed from uncoursed limestone blocks with occasional tile, bonded with a mixture of mortar and clay. A new hearth was built in the southeast corner of this area, in the angle between the city wall (3088) and the remains of the wall of Room 38 (3228) which was probably standing to a considerable height at this point (as its southern part formed part of the city wall). The new hearth was constructed in the same location as the earlier

hearths of phases 11 and 12 (the 7th to late 10th centuries). A compact layer of clay and occasional limestone chips (3464=3009) acted as a base for a circular spread of tile and limestone c. 1 m in diameter (3466=3423) (Fig. 6.18a). The interior was resurfaced successively by a layer of beach rolled pebbles (3422) (Fig. 6.18b), a scorched dome shaped clay surface (3424) (Fig. 6.18c) and a clay surface with tile fragments (3202) (Fig. 6.18d). A thin ash spread (3429)

214

NE SE

3426

3216 3298

3151

3138

3149

NW 3144

3426

3153

3144

3428

NW

SE

S

S

N

3147

N

3151 3153 3138

3359

SE

NW SE

3413

NW

3289

3346

3149

3147

3346

3289

W SE

E

NW

3293

3293

3291

0

3265 0

2m

Figure 6.15. 12th- to early 13th-century structures to east of wall 3216

3291

0

2m Profiles

1m

Figure 6.16. Profiles of post-holes of 12th- to early 13th-century structures to east of wall 3216

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

3428

SW

6  The medieval occupation of the Merchant’s House

3144 3144

3144 3144

3428 3428

3428 3428 3153 3153 3151 3151

3426 3426 3138 3138

3298 3298

215

3149 3149

3153 3153 3151 3151

3426 3426 3138 3138

3147 3147

3298 3298

3149 3149

3346 3346

3147 3147

3346 3346

3289 3289

3289 3289

3293 3293

3293 3293

3291 Structural Structural posts posts3291

3291 Structural Structural posts posts3291 0 0

2m2m

Partition Partition posts posts

0 0

Figure 6.17. Alternative reconstructions of the post-built structures to east of wall 3216

indicates its final phase of use, and was found to contain ferrous hammerscale suggesting that the hearth was used for blacksmithing. The date of this hearth is uncertain. The latest clay surface within it (3202) contained a limited quantity of late 13th- to early 14th-century pottery as well as intrusive late Roman material. The overlying mortar mixer (3199, see below) also contained 13th- to 14th-century pottery within its fills.4 No clear dating evidence was recovered from these features, with the only indication of date being the terminus ante quem provided by the dark soils of the 13th and 14th century that overlay this area (see phase 14a below). Room 40 The timber building that had been built during the 11th to 12th century above the levelled foundations of Room 40 (phase 13a) probably went out of use and was demolished during the later 12th or early 13th century. Subsequently a succession of four hearths was created within the southeast corner of the building, with a small associated post-built structure acting as a shelter or windbreak. This small structure is represented by seven post-holes, the arrangement of which suggests a partial repair during the course of its occupation (Fig 6.19).

Post-holes (3308, 3328, 3452, 4002 and 4047) are 0.30– 0.50 m in diameter and 0.30–0.40 m in depth and formed a small but quite strongly built pentagonal structure c. 1.5 m in length and width. The wide spacing of its eastern-most post-holes (3308 and 3452) perhaps indicates an opening at this point. A small hearth (3357), c. 0.60 m in length and 0.40 m in width, was located in the centre of this structure. The hearth retained traces of a stone lining. All of these features were cut into the earlier shell midden 3381.5 This structure was subsequently partially rebuilt, indicated by a succession of three intercutting hearths (3270, 3352 and 3450) above post-hole 3452 and truncating its fill (3451) (Fig. 6.20). This reconstruction saw the probable insertion of two further post-holes (4004 and 4072), one of which (4004) cut the earlier hearth (3357). The fill (4073) of post-hole 4072 contained considerable quantities of hammerscale suggesting that the earlier hearth (3357) was used for blacksmithing, with the surrounding structure acting as a shelter or windbreak for the hearth. The new post-holes created a smaller sub-rectangular structure 1.5 × 1 m. The first of a new sequence of hearths (3270) was cut adjacent to this structure. It had steep sides with a flat base, c. 0.60 × 0.60 m and was filled with charcoal-rich silty clay (3269). It was replaced by hearth

2m2m

216

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

3464

3464 3422

3423

3423

3228

3228

3088

a

3088

0

2m

b

3464

0

2m

3464 3424

3424

3202 3228

3228 3088

3088

c

0

2m

d

0

2m

Figure 6.18. Successive phases of hearth 3064 (13th century?) adjacent to city wall

(3450) positioned closer to the entrance to the west; this too was steep sided with flat base c. 0.75 m in length and 0.35 m in width filed with a silty clay rich in charcoal (3439). The latter hearth was abandoned by the dumping of rubble layer (3353) within its interior providing the foundation for the construction of the final hearth of the sequence directly above (3352). It was 0.60 m in length and c. 0.30 m in width and was of careful construction with a lining of sticky clay (3267) and a central pedestal composed of three tile

fragments. The burnt clay lining and ash within the hearth attests to frequent fires heating a container positioned upon the pedestal. A complete vessel (3268), dating to the 13th to 15th century, was found smashed above the pedestal and lining of the hearth, indicating its final phase of use. This sequence of hearths and post-holes was subsequently covered by dark humic soils in the 13th to 14th century (see phase 14).

6  The medieval occupation of the Merchant’s House

217

3328

3328 3308 4002 4047

4072

4002 3357

4047

3452

3450

3253

0

tile

3308

4004 3270 3253

3352

2m

0

tile

Figure 6.19. Sequence of post-holes and hearths in Room 40

South-facing section through hearths 3270, 3352 and 3450 West

East 3352 3267 3450

3253

3269

3439 3451 3452

tile

3270 0

50cm

Figure 6.20. Section through hearths 3270, 3352 and 3450

Summary/interpretation From the 12th century, there seems to be a slight reduction in activity in the Merchant’s House area in terms of scale and coherence. Within Room 40 a series of hearths were created of which the earliest is indirectly associated with blacksmithing (evidenced by hammerscale in a post-hole associated with a later hearth). These hearths were surrounded by small structures that presumably acted as shelters or windbreaks for them. The areas enclosed (a maximum of 1.5 × 1.5 m) were too small to be used by people at the same time as the hearths, and we can perhaps envisage a blacksmith occupying an advantageous location by the gate. To the east of the long north–south wall (3216) a series of post-holes indicated the presence of one or more timber structures, while another hearth also showed evidence of the presence of blacksmiths. There is little indication that these

were substantial buildings, and we can perhaps envisage some small-scale occupation of the area beside the gate, perhaps small stalls and blacksmiths’ workshops serving the needs of those entering and leaving via the gate. Dating evidence associated with the structures in Room 40 and the open area (39) is limited although a terminus ante quem is provided by the complete vessel found in the latest hearth in Room 40 (dating from the 13th to 15th century) and dark humic horizons of the 13th to 14th centuries that seal much of this activity (see below).

Phase 14a: 13th and 14th century Burials, the blocking of the gate and the creation of agricultural ground (Fig. 6.21) During the 13th century, the gate in the city wall was blocked, severing the long-held links with the Vivari

2m

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

218

3023 3018 4015 3397

Room 39

3029=3028

Room 40

3130 3199 3288

3128 3242 3088

3288

0

5m

Figure 6.21. Phase 14a (13th and 14th century) features in the Merchant’s House

Channel. This was immediately followed by the dumping of soils within the interior, converting the site into agricultural ground where evidence of cultivation is apparent. This area interior saw the dumping of homogenous black soils to c. 1 m in depth; domestic and farm refuse would be dumped throughout the remainder of the 13th and 14th centuries. The development of these black soil horizons seems to coincide with a phase of burial when six graves were inserted across the area of the Merchant’s House.

The open area 39 In the area of the city gate, a large rubble layer composed of limestone and tile fragments (3233=3130) was deposited over the threshold of the gateway and extended 4.6 m to the interior. This deposit contained ceramics of the mid 13th to 14th century and covered the hut and hearths of Room 40 and much of the latest surface of the cobbled yard (3305), on which a billon trachy dated to 1230–1237 was recovered. Subsequent to this deposition, the late antique gate was

6  The medieval occupation of the Merchant’s House East

219

West 3242

3088

3128

3233

4148

3305 4051

tile

3304 0

4148 1m

Figure 6.22. Elevation of the 14th-century blocking wall (3242) inserted into the gate on the city wall, showing underlying 13th- to 14th-century deposit 3233

blocked by the insertion of a large wall, 1.90 m in length and 1 m wide (3242). It was a roughly coursed limestone structure, with infrequent use of tile as packing, bonded with a white gritty mortar (Fig 6.22 and Fig. 5.55).6 A small mortar-mixing pit (3199), constructed in the angle beneath the city wall (3008) and wall 3228, may be associated with this blocking (and perhaps contemporary repairs to the superstructure of the city wall). The mortar mixer was constructed above the hearth (3464=3009) described above. Its sub-rectangular concave base, 1.7 long and 1.2 m wide, retained a lining of gravelly white mortar c. 0.10 m thick (3008). Its fill (3007) contained 13th- to 14th-century pottery, and it can be reasonably interpreted as having been used to mix mortar for the blocking of the gate and associated repairs to the walls. Following the blocking of the gate, the area of the Merchant’s House was buried beneath a thick level of black humic soil (Fig. 6.23). This type of deposit had been encountered elsewhere in the Triconch Palace excavations and had proved impervious to stratigraphic excavation in that it was impossible to visually identify any change within the horizon, which was up to 1 m deep in places. In the Merchant’s House area, it was therefore decided to excavate this deposit in 5 × 5 m gridded spits of c. 0.10 m depth, in order to determine how, and over what period of time, these deposits had formed.7 In the event, the ceramic sequence recovered suggested that these deposits had developed between the 13th and 16th centuries, with the later (early Venetian) material dominating in the upper spits and largely absent from the lower levels. This suggests a cultivated soil developing over time through deposits of manure and household rubbish, with the cultivation causing some intrusive material in the lower deposits. The earliest of these deposits formed over the rubble layer (3233=3130) and clearly post-dated the wall blocking the gate (3242). The blocking itself was immediately post-dated

by a rubble deposit (3062), which was overlain by dark humic soils around 0.40 m deep containing predominantly 13th- to 14th-century pottery (3014, 3053, 3063 and 3076). This humic level also extended north (3049, 3059) and east (2073, 2097) into the centre of the open area (39) and also covered the mortar mixer (3199) described above (deposits 2080, 2084, 2091). A 13th-century billon trachy was noted in 2091. These soils continued west above the former Room 40 (3028, 3029), where they were cut by burials (3018, 3023 and 3060) (see below). Within the area of the later tower (Room 45) a further deposit (4135) is likely to belong to this period (see below).8 Burials in the Merchant’s House During the period when the black humic levels described above were developing, a small number of burials were inserted into the area of the Merchant’s House (Fig. 6.24). The grave cuts of these burials were difficult to distinguish within the black soils, and in some cases the burials are recorded as being below the later levels of these medieval horizons. One of these burials (3060) was inserted adjacent to the city wall on the immediate exterior of Room 40, probably cut into one of the black soil horizons (3028), which lay above the rubble deposit beneath the city gate blocking (3130) (Fig. 6.25). It was apparently overlain by one of the later black horizons of late 14th- to 16th-century date (2070) (see below). Neither the grave cut nor fill could be distinguished in the black soils. The burial was placed in a supine position orientated east–west with the left arm across the chest. The body was between 13–15 years of age and of indeterminable sex. The absence of the skull, feet bones and left arm may be due to cultivation, as indicated by the presence of disarticulated human remains in the soils above (2065). AMS radiocarbon dating produced a calibrated date of AD 1280–1410.

220

3102/3137 3201

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

3124

3173

3200

3083

3066 3166

3087

3220

3167

3365

3081

3067/3050 3366

2076 2079 2086

2078 2087

3029

3049

3091

2068 2069 2070 3028

2081

2089 2094 3000

2090

3059

2098

2057

2066

2061

3024

2064

3025

3241

2097

3017

2099 3006 3013

3034

2067 2073 2080

2092

2060

3004

2054 2062 2063

2075 2082 2085

2056

3003

2093

3015

2059 2065 2072

3014 3053

2088

2084

3063 3076

2091 3001

2055 2074 2083 3005 3021 3038

3133=4135

3196

3044 3092

0

5m

Figure 6.23. Simplified plan showing sequences of spits of late medieval and early Venetian soils excavated in grid squares across Merchant’s House area

6  The medieval occupation of the Merchant’s House

221

3023

3018 4015 3396

3397

3029=3028

3023 3060

3018 4015 3396

3397

3060 0 0

1m

0

1m 2m

Figure 6.24. Late medieval burials in Room 40

A series of burials was also interred above the remains of Room 40. Four burials (3018, 3023, 3397 and 4015) were inserted through a deposit (3091) that overlay the post-holes of the 11th to 12th-century timber building. All the burials were orientated north–south, parallel to the adjacent city wall, and were interred with apparent respect for each other. Later cultivation significantly damaged all the skeletons. Grave 3397 contained a young male (3396) in his early 20s, buried in a supine position with left hand across chest and right arm placed on left elbow with head facing west

(Fig. 6.26). Both legs were flexed to the west with feet together. It was placed parallel and immediately adjacent to the western city wall. Iron nails (SF 2191) along with a copper alloy object (SF 2183) were noted near the skull, though their association with the skeleton is unclear. Possible cultivation damage was evident on the knee joints and pelvic bones whilst less than 50% of the skull was preserved. The lower part of the grave cut (3397) was located (sub-rectangular in plan, 1.35 m long, 0.59 m in width, with a flat base and gently sloping edges) cutting through context 3091. A concentration of limestone and

222

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

Figure 6.25. Late medieval burial 3060 (1 m scale)

infrequent tile was noted c. 0.25–0.30 m above the western extent of the burial, which may represent the remains of a cover. To the immediate east of skeleton 3396, and parallel to it, skeleton 4015 was also interred in a supine position (Fig. 6.27). The skeleton was that of a young male aged 14–20 years. The left hand was placed across chest and right arm placed on left elbow, while the body had fully extended legs that terminated immediately before a post-hole (4045). Within the grave fill (3420), a small unidentified iron object (SF 2333) was found near the area of the head with several pieces of iron, possible nails. The absence of a skull and damage to the knee joints and pelvic bones indicate truncation from later cultivation. A possible cut (3421) was noted, 0.80 m in length and c. 0.50 m in width. As with grave 3396, a concentration of limestone and infrequent tile noted above the burial may represent the remains of a cover. A fragmentary skeleton (3018) represented a further burial between graves 3396 and 4015. It was placed in a supine position and was aged between 12–14 years, although sex could not be determined. It was located at a slightly higher level than burials 3396 and 4015, resulting in heavy truncation from later cultivation with only rib cage, back of skull and fragments of upper arms preserved in situ. No grave cut could be identified for this burial. AMS radiocarbon dating produced a calibrated date of AD 1120–1260. A further burial (3023) was interred to the north of these burials, in the northwest corner of Room 40. This was the oldest of the group, a male between 25–30 years of age.

Figure 6.26. Late medieval burial 3396 (1 m scale)

Unusually it was placed in a crouched position on the left side facing east, although, like the other burials, it was parallel to the city wall. Its higher position, like burial 3018, has resulted in it being heavily disturbed by later activity with the majority of its right side absent. A moderate amount of rubble was found above the burial, suggesting (like 3396 and 4015) that a rough cover or rubble mound may have been placed over the grave. AMS radiocarbon dating gave a date range of AD 1020–1210, with the stratigraphic position of the burial suggesting the latter part of this period. Burials 3018 and 3023 are recorded as underlying one of the black humic soils (2090), which contained a billon trachy dated 1221–1254 and an early 14th-century coin (c. 1310), together with pottery of the same period. Within the black soil horizon above Room 40 (3029) disarticulated human remains in all probability derive from the above-mentioned burials and indicate the heavy truncation of the burials, although it should be noted that the burials are recorded as being cut into this layer.9 Disarticulated human remains were also recovered from deposits on the southern exterior of the city wall (2055, 2056, 2057, 2060, 2066 and 3092). These may also derive from the Room 40 burials. The damage to these burials, primarily to their knees, pelvic bones and skulls, with the higher burials heavily truncated, suggests intensive cultivation within this area.

6  The medieval occupation of the Merchant’s House

223

Figure 6.27. Late medieval burial 4015 (1 m scale)

A further burial (4138) was noted at the southern end of Room 38, where it was disturbed by the building of a later tower. It was apparently beneath a black humic horizon (4135) within which the disarticulated skeleton was found across an elliptical area 1.20 × 1.40 m, with the skull reinterred in a small tile-lined box (see phase 14b and Fig. 6.30 below). Summary/interpretation During the 13th to 14th centuries a significant change appears to have occurred within the Merchant’s House area, apparently initiated by the blocking of the gate in the city wall which had remained open for the previous 700–800 years. The blocking of the gate was followed by the development of black humic horizons across the area, above the earlier buildings. The quantities of ceramics, animal bone and other materials within these deposits indicate that they are derived from the dumping of domestic waste and probably represent the formation of cultivated garden soils. This suggests that occupation of the area continued close by, probably to the north (as also suggested by similar deposits in the Triconch Palace – see Chapter 4, phase 14). At the same time as the black soil horizons were developing, a small number of burials were interred in the area. The relationship between the burials and the black soils was difficult to determine, as the burials were apparently cut into the lower horizons of these cultivated soils and overlain by later similar deposits. The burials were extensively disturbed, probably by later cultivation, and grave cuts could rarely be distinguished. The presence of concentrations of rubble in the soil above the burials suggests that the graves were original capped or marked by small mounds of stones, which became buried as the soils continued to develop across the area. The locations

of these graves were subsequently forgotten resulting in the disturbance by ploughing. The concentration of four graves in the northwest corner of Room 40, could suggest a family group, deliberately placed on the edge of the cultivated area. DNA analysis of the bones also suggested that this was a possibility although the evidence was not conclusive (Fenton et al. in prep.). The implication of these changes is that the Merchant’s House (like the Triconch Palace) was becoming a more marginal area in this period, used for cultivation rather than occupation. However, the presence of significant quantities of imported ceramics, including glazed table wares, in this period (Vroom in prep.) is indicative of sustained nearby occupation.

Phase 14b: Late 14th century The creation of a new tower (Fig. 6.28) The end of the 14th century saw the creation of a new tower in the remains of the late antique Room 38, and possibly the contemporary repair of the city wall. This activity seems to have had little effect on the rest of the Merchant’s House area, where cultivation and the dumping of domestic refuse seems to have continued unabated. Rooms 38 and 45 The southern end of Room 38 was converted into a tower (Room 45) by the insertion of an east–west wall (3229), which was cut through the black soil level 4135 noted above (Figs 6.29, 6.30).10 This wall was 6 m in length, and abutted the existing wall (3228) to the west and the door blocking (3274) to the east. Only a single course of the superstructure survived, with the remainder of the extant structure belonging to the foundation, which was 0.45 m deep. The foundation was constructed from

224

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

4199

4138 3229 4165 4153

3160

4200 4134 4135 3228

0

5m

Figure 6.28. Phase 14b (late 14th century) showing new tower on city wall

uncoursed limestone bonded with a mortar and earth mixture with occasional use of tile between the stones. Its width of 0.80 m suggests that it supported a substantial superstructure. The single course of the superstructure that remained suggested a wall of c. 0.60–0,70 m in width with a substantial foundation offset to the south. The groundfloor door into the tower was presumably through this wall although the structure did not survive to a sufficient height to show its location.

In the interior of the tower a large masonry pier (4134), 1.15 m square, was inserted to support stairs leading to an upper level. Only the foundation of the pier survived, constructed in coursed limestone with tile fragments as levelling courses with a limestone and tile rubble core bonded in a gritty white mortar. The pier was cut through the thick level of black soil (4135) and disturbed the burial (4138) noted above. The skull of the burial was re-interred in a small box structure composed of tile fragments (4153)

6  The medieval occupation of the Merchant’s House 3228

225

3229 4199

4165 4153 3228

4200 3229

3160

4134

4165 4153

4200

4135

3160

4134

3228

4135 3228 0

tile 0

tile

2m 2m

North

South North

South

2055

3058

2055

3058 3061 3061

2074 2083 3229

3229

mortar tile

2074 2083 3005

3005

3021 3038

3228

3044 3021 3092 3038 31273044 3092

3127

3228

1m

0

mortar tile

0

1m

Figure 6.29. New tower on city wall, including west-facing section

(Figs 6.31, 6.32). Similar treatment of a displaced skull has also been noted at Corinth (Barnes 2003, 437, fig. 26.3). The position of this re-interred skull, at approximately the same level as the top of the surviving pier foundation, gives some indication of the ground floor level of the tower, as the skull was clearly reburied beneath the floor surface. Similarly a large block of fallen masonry (3160) found in the southwest corner of the tower (below 3044 and 3038, the latter an equivalent deposit to 4135) must have lain beneath the existing deposits when the tower was created. A deposit (3092) beneath this fallen masonry has a bone join

with layer (2091) within the open area 39, which overlies the phase 14a mortar mixer (3199) described above, and which contains pottery dated to the mid 13th and early 14th century (Figs 6.29, 6.23). A large mortar-mixing feature located to the immediate north of the tower within Room 38 is probably associated with the construction of the tower. A well-defined and compact layer of lime (4199) was positioned above a late medieval deposit (4168) above the phase 13a (11th- to 12thcentury) deposit (3051). The lime deposit was 2.3 m long, 2 m wide and c. 0.15 m in maximum depth (Fig. 6.33).

226

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

Figure 6.30. Overview of late 14th-century tower, looking west

4165 Area of disturbed bone (4138)

3228

4134

4153

4161

3229

3127

0

1m

3274 3230

3160 3088

0

5m

Figure 6.31. Plan showing location of skull placed in box (4153) with associated spread of disarticulated bone (4138)

6  The medieval occupation of the Merchant’s House

227

Figure 6.32. A skull, displaced by the construction of the tower, which has been subsequently placed in a box made of tile

Figure 6.33. Mortar mixer 4199, possibly associated with the construction of the late 14th-century tower, looking southwest (2 m scale)

It is possible that other sections of the city wall in the Merchant’s House area were repaired during this period. Spreads of rubble were noted against the city wall on the west side of Room 40, within the upper levels of the black humic soils (2070=2090) (Fig. 6.23). These deposits, which overlay the graves noted above, produced material dating to the late medieval and early Venetian periods. They may represent either collapsed material from the walls or waste material from repairs. On the exterior of the city wall to the south, further rubble layers (2064=2066) were found that may represent collapsed material from the walls or indicate the dumping of waste material from the rebuilding of the city wall (Fig. 6.23). Both contained late medieval and early Venetian ceramics spanning the 13th to 16th centuries, as did the overlying deposits (2056=2057=2060=2061) that also contained quantities of rubble. Two later 14th-century coins (1343–1400 and 1370–1382) were recovered respectively from contexts 2060 and 2056.

Phase 15: 15th to late 16th/early 17th century

Summary/interpretation Around the end of the 14th century, a new tower was built on the city wall. The foundations of a former late Roman building were used to create a projecting tower with at least two floors and an internal staircase. Other parts of the wall circuit may have been repaired at this time. The context for this construction seems likely to be the town becoming a Venetian possession in 1386. The castellanus of Butentro was permitted to repair his palace and his house in 1387, while castrum Butentroi is noted as being in need of repair in 1394. The refortification of the Butrint peninsula was also discussed in 1387 (Soustal 2004, 25). The construction of the tower may well be related to the need to protect and monitor the lucrative fish traps of the Vivari Channel. There is no indication, however, as to the nature and longevity of its occupation. Its relationship to the former gate is likely to be fortuitous, as it seems clear that the gate was blocked considerably earlier than the construction of the tower.

The open area (courtyard) 39 and Room 40 The black humic soils that apparently began to develop across the Merchant’s House area from the 13th to 14th centuries, continued to build up through the 15th and 16th centuries, presumably augmented by refuse, manure and night soils from nearby houses. These soils were almost certainly cultivated, evidenced by the mixed materials found within them. Early Venetian material was found from a level of 0.50–0.70 m below the ground surface, although it may be intrusive within the lower excavated spits (see Fig. 6.23).11 A coin of 1519–1630 was recovered from layer 2070 in the south part of Room 40, while the ceramic sequence from an equivalent layer in the north part of Room 40 (2087) extended into the first quarter of the 17th century.

Continued cultivation and deposition of domestic refuse (see Fig. 6.23) The cultivation and dumping of domestic refuse seemingly continued uninterrupted within the Merchant’s House area through the 15th century until the early 17th century. This was accompanied by the further development of up to 0.70 m of black humic soils, containing considerable quantities of early Venetian pottery and coins. The majority of early Venetian finds from the Triconch Palace and Merchant’s House areas are in fact concentrated in the area of the Merchant’s House (perhaps connected with use of the tower described above). Rubble layers within the tower and along the western city wall indicates the defences of the site were in a ruinous state by early 17th century with rare loss of ceramics and coins occurring into the early modern period.

Room 45 The occupation span of the late 14th-century tower described above seems to have been limited. The only sign

William Bowden, Andrew Crowson, Matthew Logue and Alessandro Sebastiani

228

of occupation was a small indeterminate rubble feature (4200) to the immediate east of the pier foundation (4134). It was 0.75 m across and c. 0.30 m in height composed of limestone and tile rubble (Fig. 6.29).12 The interior of the tower was filled by c. 1 m of soils (2055, 2074, 2083, 3005) that contained ceramics dating from the 13th to 16th centuries.13 Within the upper levels of this infill (in layers 2055 and 2074) fragments of over 32 Venetian glass vessels were uncovered. The upper level (2055) and an underlying level (3132) contained significant quantities of rubble suggesting that the tower may have been in a state of collapse by the 16th century. Similar deposits were excavated in gridded spits to the north and east of the tower (2099, 3003, 3004, 3006, 3013, 3015, 3017, 3050, 3067, 3081,) (see Fig. 6.23). All contained ceramics from the 13th to 16th centuries. Summary/interpretation The later 15th to early 16th century sees a continuation of cultivation and deposition of domestic refuse within the Merchant’s House area indicating contemporary occupation within the vicinity. The large quantities of 15th- and early 16th-century ceramics within the upper levels of the soils attest to a population able to access Venetian table wares (see Vroom in prep.). The apparent concentration of this material in the area of the Merchant’s House by comparison with the Triconch Palace, suggests that this concentration may be associated with occupation of the tower. The apparent collapse of the tower by the early 17th century, however, coincides with the last recorded activity within the Merchant’s House with only sporadic casual loss thereafter.

6

7 8

9

10

11

Notes 1 2 3

4 5

This midden, which was excavated as a single context, seems likely to represent a long-term process, which became clear when the finds were analysed. The date of this second surface (3305) is unclear. It contained pottery of the mid 13th to 14th century, which may derive from the overlying black cultivated soil (see below). The excavation records seem quite clear that this surface underlay at least some of the mussel shell deposits recorded as 3020, suggesting that the use of the area for mussel processing continued after the construction of the road/ surface and the building, the latter of which sealed the earlier mussel deposits in Room 40. This late medieval material may be intrusive, deriving from the overlying layer (2091), which contained material from the 11th to the 14th century. It is only the association of this structure with the hearths (of which the latest dates to the 13th to 15th century) that places it in this phase rather than earlier.

12 13

At probably the same time, the gate to the east of the Triconch complex (gate 2) was substantially remodelled. A postern gate was blocked and the main gate was blocked with the addition of a new tower, although there is no guarantee that these two blockings took place simultaneously. The dating of both is indeterminate with deposit 1033 (abutting the blocking of the postern) containing a mixed assemblage dating from the 12th to 15th century and the building of the tower datable to the phase of fortifications labelled “Medieval 2” (Andrews et al. 2004, 137–43; Gilkes and Lako 2004, 168 and also Bianchi et al. 2007). These spits were predominantly excavated by trowel. Deposit 4135 appears to be a coherent late medieval deposit although it was equated with a number of contexts (2074, 3021, 3038, 3133, 3195) that appear to be early Venetian (late 14th to 16th century). Disarticulated human bone was also recovered from context 3129, a probable late antique layer below the mussel shell midden (3381) (see phases 11–12 above), suggesting a considerable degree of disturbance in these deposits. During the excavation it was not recognised that almost all of 3229 was a foundation cut from a high level rather than a 0.45 m high wall abutted by a series of internal deposits. Reconstructing the sequence, however, it seems that the wall was cut through 4135 and a series of equated deposits (3021, 3038, 3133, 3195). All of these deposits, with the exception of 4135, contained late medieval and small quantities of early Venetian pottery, suggesting that the tower was built at the very end of the 14th or early 15th century. A billon trachy of 1244–1254 was recovered from layer 3133. As described above, these soils were excavated in 0.10 m spits in an attempt to discern a depositional sequence. Early Venetian material was recovered from spits: in front of the gate from (earliest to latest) 2098, 2092, 2085, 2082, 2075, 2072 (=3000), 2065 (=2094), and 2059; from the eastern exterior of Room 40 from 3000, 2094, 2089, 2086, 2079; and from Room 40 from 2070 (=2087?), 2078, 2069 and 2068. Within spits 2072, 2075 and 2086 coins dating to 1343–1471, 1462–1471 and 1536–1537 were recovered. On the eastern side of the open area 39, on the southeast corner, early Venetian material was recovered from spits 2073, 2067, and 2063 (earliest to latest), and on the central eastern area from spits 2093=2067, 2063, 2088=2062, 2081=2054. The excavators tentatively suggested that 4200 was a possible hearth, although no sign of burning was evident. It appears most likely to be a misleadingly coherent dump of rubble. As noted above (note 9), deposits 3021 and 3038, which were equated with 4135 (thus predating the construction of the tower), and 3044, which was underneath 4135, all contained small quantities of late 14th- to 16th-century ceramics. This is probably intrusive, given the difficulty of distinguishing these layers, but it remains possible that the tower is later than end of the 14th century, although given that the infill and collapse of the tower seems to date to the 15th or 16th centuries the window for construction seems short.

INTERPRETATION AND DISCUSSION

7  The mosaic pavements and painted walls of the domus John Mitchell

The houses of the upper echelons of society became major sites for the display of identity and social standing in the late Roman period. Recent studies have drawn attention to the phenomenon of the domus, which throughout the Roman world grew in the size of its footprint, a vivid expression of the status and power of the social elite, with an increased emphasis given to the spaces provided for the reception of guests (Baldini Lippolis 2001; Bowden and Mitchell 2007, 455–6; Ellis 1991). A major peristyle was the central focus, apparently open to visitors of all categories, assuming something of the role that the atrium had played in houses in Italy in the early Empire. From this central hub all the apartments of the house, private and public, were accessible. Indeed, from their layout, it would appear that houses of this period were even more open and transparent than they had been in Rome and Pompeii in the 1st centuries BC and AD (Wallace-Hadrill 1989, 46; see also Chapter 8 this volume). Decoration, carved and moulded architectural embellishment, wall-paintings and floor mosaics, furniture and statuary all played their parts in creating an ensemble of effect. In this, particular emphasis was given to entrances, to the more public thoroughfares within the house, to open courts and to the rooms in which guests were entertained, in particular the dining rooms, the triclinia and oeci, which, to judge from their disposition and growing physical presence, took on an increasingly important role as theatres of reception and entertainment in the liturgy of hospitality in late antiquity. In the triconch domus at Butrint, the best preserved elements of this decorative ensemble are the floor mosaics (Fig.7.1). These belong to phases 2 and 3, dating to between the 3rd and very end of the 4th centuries. In the first decades of the 5th century, when the house was abandoned as a functioning elite residence, these pavements would all have been visible in four of the functioning spaces within the complex in which the floor-surfaces are preserved. Room 18: the east vestibule and long gallery The earliest of these four spaces is Room 18 comprising

the east vestibule and the adjoining long gallery, which runs along the southern side of the domus, with views to the south over the Vivari Channel (Figs 7.2–7.6). The vestibule is an almost square space, measuring 4.42 m east–west × 4.14 m north–south. On excavation, much of the surface of the mosaic was found to be covered with a hard white mortar, which was not removed; however, enough of the pavement was visible to reveal the design (Figs 7.2, 7.7, 7.8). The central field of the floor was filled with a complex composition of spaced swastika-meander with single returns, the spaces staggered and containing squares (Prudhomme 1985, pl. 192c). The design is laid in framing bands of polychrome simple guilloche, in counterchanged colours (red, pink, white/yellow, ochre, white), sandwiched between superposed black right-angled isosceles triangles against white. The two visible squares each contained a comic mask, one female and one male. Both are set against brilliant white grounds. The female mask is strictly frontal, with long hair, in two shades of blue-grey with streaks of black and white, and grey lined with white on the brow, sweeping away symmetrically to fall down on either side of the face (Fig. 7.9). The flesh of the cheeks and chin is pink, with striking red accents on the brow, a wide nose framed in red, modelled in blue grey and lit with white, great round black eyes and a large rectangular black open mouth. This is probably the character of a girl, a leading female protagonist from New Comedy (Bieber 1961, 156, fig. 562–5), although it could also represent a heroine from Tragedy. The male mask is of a vigorous old man, with yellow flesh creased in white, long white hair streaked with blue-grey tightly brushed back from the brow, a prominent bulbous nose, long beetling black eyebrows, large round black eyes, a great oval black shouting mouth and a curling rectangular pectinate beard hanging from his chin (Fig. 7.10). This is the mask of one of the principal characters from the repertoire of New Comedy, apparently the hegemon presbutes – the choleric leading old man or father, with his angry raised eyebrows, typically paired with the girl, his daughter – or possibly the hegemon makropogon

0

0

Figure 7.1. Triconch domus, plan showing surviving mosaic pavements

0

10m Figure 7.2. Room 18, east vestibule and long gallery, plan with mosaic pavements

20m

Figure 7.3. (Plate 1) Room 18, east vestibule, vertical projection (MZ)

Figure 7.4. (Plate 2) Room 18, long gallery, eastern end, vertical projection (MZ)

234

John Mitchell

Figure 7.5. (Plate 3) Room 18, long gallery, central section, vertical projection (MZ)

Figure 7.6. (Plate 4) Room 18, long gallery, western end, vertical projection (MZ)

–­ the conciliatory phlegmatic long-bearded leading man – types which are described by the late 2nd-century AD Greek lexicographer Julius Pollux in his Onomasticon (Pollux, Onomasticon IV, 144, p. 244; Bieber1961: 93–4, 155, figs. 324–8, 330–7, 562–4; Robert 1911: 77–80). To judge from what survives of the scheme, it is likely that the complete central field of the floor contained five such squares, four masks facing inwards, set around a central

panel that would have contained a fifth motif, possibly a mask or a full figure of Dionysos/Bacchus, the god who presided over dramatic performance in the Greco-Roman world, or a similarly appropriate subject. This central area is surrounded by a sequence of frames (Figs 7.11–7.13). First, a wide band with a continuous orthogonal pattern of tangent sequences of affronted paired peltae in switching orientation alternating with

7  The mosaic pavements and painted walls of the domus

Figure 7.7. (Plate 5) Room 18, east vestibule, view to north (MS)

Figure 7.8. (Plate 6) Room 18, east vestibule, detail of central panel with masks (MS)

235

236

John Mitchell

Figure 7.9. (Plate 7) Room 18, east vestibule, female mask (MS)

Figure 7.10. (Plate 8) Room 18, east vestibule, male mask (MS)

Figure 7.11. (Plate 9) Room 18, east vestibule, south-eastern corner (JBB)

sequences of rosettes, saltires of spindles, all in black on a white ground, with a little yellow saltire cross in each of the resultant poised concave squares. This is bordered by a trichrome outlined round-tongued double guilloche,

in counterchanged colours, red, pink, white/yellow, ochre, white (Prudhomme 1985, pl. 74c); then an elegant trichrome band of consoles in lateral perspective in white and light and dark pink, outlined in black, with a black crosslet on

7  The mosaic pavements and painted walls of the domus

237

Figure 7.12. (Plate 10) Room 18, east vestibule, south-eastern corner, to east, showing painted wall-plaster (JBB)

0

0.5m

Figure 7.13. Room 18, east vestibule, south-eastern corner, plan of mosaic pavement

238

John Mitchell

Figure 7.14. Room 18, long gallery, view to west (JBB)

Figure 7.15. (Plate 11) Room 18, long gallery, eastern end, to north (JBB)

the display-end of each console (Prudhomme 1985, pl. 99j). This in turn is framed by a narrow saw-tooth fillet in white and black (Prudhomme 1985, pl. 10g), and then by plain bands of white and pink. Finally, there is an outer border of undulating ivy trail with heart-shaped leaves, parti-coloured red and black, in the scrolls, growing from a black stem against a white ground (Prudhomme 1985, pl. 64d). Inner and outer bands of red frame this scroll against the white ground. The configuration in the central field, the spaced swastika-meander containing squares bearing motifs in staggered sequence, is a design deployed quite commonly in mosaic pavements, generally assigned to the 2nd and 3rd centuries. Examples using an almost identical vocabulary include a floor that has been dated to the first half of the 2nd century at Saint-Romain-en-Gal, in Gaul, with busts of a silenus, Pan, maenads and satyrs set around a full-length standing figure of Bacchus, in the five squares (Lancha

1981a, 232–6, pl. CXXVIII–IX; Ling 1998, 64, fig. 44), and a 3rd-century pavement from Utica, dismembered elements of which are now on display on the north stairs at the British Museum, London (GB1859.4–2.108, BM Cat. Mosaics 6). Closely related variants are to be found in a late 2nd-century floor at Lyons, in which Bacchus seated on his panther in the central square is surrounded by the four seasons (Stern 1967, 45–7, pls XXIX–XXXI) and a pavement in the House of the Fishermen at Utica (Alexander, Besrour and Ennaifer 1976, no. 278a, 29–30, pl. XVII). The adjoining long gallery extended for 30 m to the west, with its long southern front, almost certainly fenestrated, overlooking the water (Fig. 7.14). The pavement here has much the same chromatic value as the floor of the vestibule, with white and black, bright brick-red, pink and various shades of blue-grey, but the design is different (Figs 7.15, 7.16). Here the central panel, a great elongated rectangle

7  The mosaic pavements and painted walls of the domus

0

239

1m

Figure 7.16. Room 18, long gallery, western end, plan of mosaic pavement

that runs the whole length of the room, carries an intricate design in black and white, with large octagons, tangent on the diagonal, interspersed with open swastikas (Figs 7.17, 7.18). Each octagon is subdivided into a complex configuration of triangles and squares, alternating black and white, to form an eight-pointed axially emphasized star: a central square is divided into four sub-squares – two on one diagonal are white charged with a small saltire cross made up of five black tesserae, while the opposing two are parti-divided into triangles; these are framed by a further sequence of triangles, large white ones on the principal axes and three smaller ones on the diagonal axes. The axial faces of the octagons are filled with black to generate the arms of the open swastikas that punctuate the design. This central tableau is surrounded by a band of trichrome three-strand guilloche, the strands counterchanged red, pink, white/blue, light blue-grey, white (Prudhomme 1985, pl. 72) (Fig. 7.19). Then, against the walls of the gallery is an outer border of running ivy-trail, in detail a little different from that in the vestibule, with prominently

waisted leaves parti-coloured red and pink in the scrolls; as in the vestibule, this is framed by narrow bands of red. The floors of these two adjacent spaces were laid on separate beds of similar make-up; first two foundations, an underlying statumen or rudus of cocciopesto, sand, broken tile and poorly burnt lime, then above this a more finely textured nucleus consisting of a mixture of powdered stone with a little crushed tile and lime to form a whitish mortar; and finally a bedding layer, a fine mortar of lime and powdered stone, into which the tesserae were set. In the gallery the rudus was more reddish in colour than in the vestibule, containing more crushed brick. The small mortared tesserae in the vestibule formed a flexible membrane and on excavation were found to be detaching from the underlying bedding layer. In the gallery, on the other hand, the tessellatum was well attached to the bedding layer and showed no signs of coming loose. This difference may possibly be related to the differing sizes of the tesserae used in the two areas, smaller in the vestibule and larger in the gallery. The tesselatum, the mosaic surface, in both

Figure 7.17. (Plate 12) Room 18, long gallery, detail edge of central panel and border (MS)

Figure 7.18. (Plate 13) Room 18, long gallery, detail of central panel (MS)

Figure 7.19. (Plate 14) Room 18, long gallery, detail of border (MS)

7  The mosaic pavements and painted walls of the domus

241

Figure 7.20. Room 18, east vestibule, east wall, reconstructed painted decoration

areas is expertly set, closely structured and tight; competent professional work in which the stones are laid in patterned ordered sequence without intervening gaps (Figs 7.18, 7.19). The tesserae used are small and regular in shape, for the most part of stone – white, black, reddish-pink, light pink and ochre – with glass tesserae set here and there to bring life to the surfaces. In the vestibule the stones are of two grades: around the perimeter very small, c. 5 × 5 mm, and further in somewhat larger, c. 7 × 7 mm, starting in the band of peltae and four-spindled rosettes. The mosaic surfaces in the two adjoining spaces are closely related in design and lay and there is no reason to believe that they are not contemporary. Both the eastern vestibule and the long gallery had painted decoration and on excavation passages of plaster still adhered to the lower surfaces of the walls, painted with a simple but effective scheme of illusionistic columns rising from a panelled stylobate. The walls of this range are constructed of local limestone carefully cut to rectangles, well-coursed, and finished with falsa cortina pointing. Their interior plastered finish is in two layers, a lower ariccio some 15 mm thick and an upper intonaco with a thickness of about 8 mm. The fugitive remains of painted decoration was preserved in the vestibule, on all three walls, covering the full length of the short eastern wall, and extending along the southern wall for 3.97 m and along the northern wall for 1.5 m. On the short eastern end wall, the lowest 225 mm is painted to resemble a dado revetted with panels of polished stone, alternately yellow and apparently plain white (Figs 7.12, 7.20). The panels are divided by vertical black lines 5 mm wide. The northern end of this end wall is obscured

by a later stairblock but the surface at the southern end was more or less intact and visible. Starting from the southeast corner, there is a plain panel which extends for 0.218 m, then a black vertical dividing line, followed by a second, yellow, panel, 0.62 m long, a second vertical black dividing line, and a then further plain panel that is cut by the later masonry insertion. This panelled dado is crowned by a sequence of narrow horizontal bands, each c.10 mm wide, in black, light sky blue, plain plaster, light blue again, then plain. From this columns rose, painted a deep red, perhaps to suggest porphyry. In the southern corner of the east wall a red vertical band extends c. 0.10 m each side of the corner. The moulded base and lower shaft of a red column rise at a point 0.87 m out from the corner-band along the east wall. The maximum width of this column base is 0.24 m, the shaft itself being 0.15 m wide. The ground either side of the column appears to be a dark grey-blue, but the evidence from the north wall suggests that the plaster here may have been left essentially unpainted. On the north wall, a later stairblock still partly covers the dado-level of the wall, but the plaster immediately above is preserved. Here a vertical red band, representing the corner-pilaster, extends 0.176 m from the corner. Then, 0.867 m along the wall to the west rises the shaft of a red column, 0.176 m in diameter. The ground between the corner and the column is plain unpainted plaster, except for a 28 mm band of light blue that runs horizontally across the wall about 0.40 m above the level of the floor of the room, perhaps to suggest a vista out to the open sky. On the opposite southern wall a strip of plaster is preserved at the very bottom. This shows the continuation of the dado with a succession of panels, alternately yellow

242

John Mitchell

Figure 7.21. (Plate 15) Room 24, reception hall, view to north (JBB)

and either plain or a very light sky blue, as on the east wall separated by vertical black lines. Starting at the southeastern corner, the first panel is plain and extends for 0.22 m, bounded by a vertical black line. This is followed by a yellow panel, in places discoloured to pink, about 0.64 m long. Next is a shorter panel, c. 0.46 m in length, which seems to be a very pale sky-blue, then a 0.68 m yellow panel, and a short 0.265 m pale sky-blue panel. The following two panels are yellow and pale sky-blue and have a total length of 1.195 m. Their dividing line has not survived but to judge from the surviving areas of coloured plaster it must have been more or less at the mid-point. The final preserved panel is yellow. These scant remains show that the room was originally decorated with restrained elegance, with illusionistic columns rising from a polychrome stylobate. Between the columns, occasional horizontal bands on a plain ground may have given the effect of a view out to the open sky. It is likely that this scheme, preserved in the vestibule, continued along the walls of the long gallery, where it would have harmonized with the fenestration of the long southern wall, which may well have opened into a columnar screen for some of its length. The make-up of these floors, which lie within the levels of the seasonally changing water-table and are now often submerged, were never examined by excavation and there is no independent stratified evidence for the date of the mosaics. However, the relationship of these rooms to subsequent, more easily datable parts of the house and the designs used both give reason to think that this southern range was constructed and decorated at some time in the course of the 3rd century AD.

Room 24: the reception hall The doorway at the western end of the long gallery leads into a large apsidal room, which must have served as the principal reception room of the domus, possibly also as a dining room, in phase 2 of the house (Figs 7.21–7.24). The room started life as a rectangular space, 6.65 m long and 7 m wide, with two doors, a wide main entrance in the south wall, 3.66m wide, and the 2.235 m wide door towards the southern end of the east wall, providing access from the long gallery. Just inside the southern entrance an elegant octagonal fountain was set into the floor, revetted inside with slabs of Proconnesian marble and framed with a surround of eight polished stone slabs each cut to a trapezium. These framing elements have all been detached and alienated, but fragments of slabs of Docimium/pavonazzetto and africano marble of an appropriate thickness found in the room may derive from this surround. In a second phase (phase 2b), the room was lengthened by 1.55 m and an apse was added at the northern end. The overall interior length was now increased to 11.60 m and the apse was 4.7 m wide. It was probably in this reconstruction that the principal southern door was narrowed to 1.77 m, that the secondary door in the east wall was closed down to 1.27 m, and that the fountain received a new bottom, reducing its depth to 0.15 m. In a third phase the southern door, the original main entrance, was blocked completely, leaving only the reduced eastern door as a means of access to the room. The walls were plastered and painted, in both phases, although few traces of pigment were preserved, and the floor was laid with a fine mosaic pavement. In the first rectangular phase of the room, the mosaic was composed

Figure 7.22. (Plate 16) Room 24, reception hall, southern end in vertical projection (MZ)

Figure 7.23. (Plate 17) Room 24, reception hall, northern end and apse in vertical projection (MZ)

John Mitchell

244

0

5m 0

5m

Figure 7.24. Room 24, reception hall, plan of mosaic pavement

with the medallion that formed the focal part of the composition, somewhat awkwardly displaced to the north to accommodate the octagonal fountain; and in a second phase a further section of pavement was set in the extension and in the apse to the north. The first-phase mosaic, in the southern and central parts of this room, is a large and complex composition of numerous contrasting patterns, for the most part abstract designs, set in panels or framing bands, in sequence from the centre to the edges of the room. The circular central motif is not preserved. This was surrounded by at least three concentric rings, regular zig-zags, white and black and red and black. This whole circular composition has an

overall diameter of c. 1.335 m., and is set within a large square, which in turn is poised lozenge-wise within a larger squarish rectangle (Figs 7.25, 7.26). The inner square is outlined with two rows of black tesserae and has a white ground. In the corners are small chalices, open to the centre, outlined in black and filled with red, grey and black, and from these plants trails issue, filling the spandrels with spiraling conical formations. The outer ‘square’ (2.19 × 2.01 m) has elaborate angle-motifs in its four triangular corners. The two that are partly preserved differ in form; in the northeast corner there is a large rounded vessel with a tall red conical lid, from which black stems with curling tendrils issue to fill the white ground (Fig. 7.27); while in

7  The mosaic pavements and painted walls of the domus

Figure 7.25. (Plate 18) Room 24, reception hall, central feature of pavement, to north (JBB)

0

1m Figure 7.26. Room 24, reception hall, central feature of pavement, plan

245

246

John Mitchell

Figure 7.27. (Plate 19) Room 24, reception hall, northeastern corner of central feature (MS)

0

1m

Figure 7.29. Room 24, reception hall, eastern side, plan of mosaic pavement

Figure 7.28. Room 24, reception hall, eastern side (JBB)

the southwest corner a large gourd-like vase, with vertical red and pale blue flutings, also puts forth stems, with at least one large black heart-shaped leaf. Flanking this large central square to the east and the west are long rectangular panels filled with rows of superposed red, black and pale blue triangles, all outlined in white (Prudhomme 1985: pl. 11h) (Figs 7.28–7.31. These panels are c. 20.5 mm long and c. 290 mm wide. Next is a broad framing band which runs round the eastern, southern and western sides of the square together with these flanking panels, with an orthogonal scheme, in black on white, of tangent peltae in alternately upright and recumbent confronted pairs – a so-called running-pelta pattern

7  The mosaic pavements and painted walls of the domus

Figure 7.30. (Plate 20) Room 24, reception hall, western side (JBB)

0

247

(Prudhomme 1985, pl. 222d). This border varies in width; on the east side 0.74 m wide, on the west 0.625 m, and on the south 1.805 m. The southern band accommodates the octagonal fountain in a somewhat irregular fashion. The three arms of pelta-pattern and the central square are framed next by a band of expanded bead-and-reel motif, made up of large white lenticular elements set against a black ground; 0.18 m wide on the eastern side, 0.19 m on the south, 0.17 m on the west, and 0.18 m on the north. This is bounded by two rows of white tesserae, which in turn are framed by a band of ornate chequer-board, in which the little squares are alternately saltire-divided with opposing triangles of black and pink to form alternately upright and recumbent hourglasses, and white with a little saltire cross formed of five pink tesserae (Prudhomme 1985, pl. 120g). The four sides of this band also vary in width: 0.34 m on the eastern side, 0.375 m on the south, 0.49 m on the west, and 0.515 m on the north. This is contoured all round by two rows of white tesserae (0.2 m), and then by a border of superposed isosceles triangles, white on black, 0.19 m wide on the eastern side, 0.17 m on the south, 0.18 m on the west, and 0.185 m on the north, followed by a further double row of white tesserae. The outermost border of this floor, some 0.40 m wide, consists of rectangular panels bearing a series of repeating motifs. The leitmotif here is a white panel spanned by a

1m

Figure 7.31. Room 24, reception hall, southwestern sector, plan of mosaic pavement

248

John Mitchell

Figure 7.32. (Plate 21) Room 24, reception hall, detail of western outer border (MS)

Figure 7.33. (Plate 22) Room 24, reception hall, detail of western outer border (MS)

large reddish-pink lozenge that carries a prominent black heart-shaped leaf (Fig. 7.32). This occurs every other panel and alternates with a succession of designs, arranged more or less in regular succession, which include: an elaborate chequer-board with squares with counter-set pairs of black and white triangles, hour-glass fashion, alternating with squares with a little black cross set on a white ground; a reticulate scheme with a chequered cross in black and

red or pink tesserae set within each little field; a scheme with four adjoining white four-petalled spindle-rosettes against a pink ground with a poised chequer-board of red and white tesserae in the interstices between the rosettes (Fig. 7.33); a variant of this with a black ground and poised serrated squares, pink towards the exterior and white towards the centre; four affronted pairs of black peltae alternately upright and recumbent on a white ground; an open reticulated design with a diagonal net in black on a white ground, with little poised squares outlined in black and red set within each field of the net; a simplified variant with a single diagonal black cross on a white ground, with red triangles defined by red tesserae rising from each side, and a single black tessera at the base of each triangle; and a chequer-board in which each little square consists of opposed pairs of black and white triangles, hour-glasses, switching in orientation from square to square. In a second phase, when the room was enlarged to the north and the apse added, the mosaic pavement was extended into these areas (Fig. 7.34). In the zone immediately in front of the apse, a broad uneven band divided into three panels spans the width of the room (Fig. 7.35). The central panel here (2.20 × 1.705 m) carries a reticulate design of black four-petalled spindle-rosettes, on a white ground, with elaborate crosses, each composed of four blue glass tesserae and nine tesserae of brick-red terracotta in the interstices. This is flanked by two panels (2.35 × 1.705 m on the east; 2.45 × 1.705 m on the west) of running peltae ornamented with little rectangles of red tesserae (occasionally blue or black and red) at the points where the tips of four peltae touch (Prudhomme 1985, pl. 222e). Together these three panels form a kind of threshold to the apse.

7  The mosaic pavements and painted walls of the domus

0

1m

Figure 7.34. Room 24, reception hall, northern extension and apse, plan of mosaic pavement

The apse itself has a semi-circular motif at the mid-point of its cord, now badly damaged, framed by an arching band with a kind of broad bead-and-reel motif in white and blue lenticular elements against red, and crowned by a little cross in blue glass tesserae in the bead at the top of the arch (Fig. 7.36). This arch frames a large frontal female mask, very similar to the one preserved in the western vestibule of Room 18, with straight hair sweeping down away from the brow, strong defining lines and light on forehead and nose and striking large round black eyes. Only the upper part of the face is preserved. The colours used are black, white, pink, brick-red and turquoise, brilliant blue and light green glass. The main field of the apse was filled with a striking composition of black and white parti-coloured overlapping scales or leaves, which issued fan-wise from the central mask in serried ranks, gradually increasing in size from the centre to the periphery (Prudhomme 2002, pl. 332a) (Fig. 7.37). This is a design that was quite widely adopted for the pavements of apses in the 2nd and 3rd centuries and continued in use into the late Roman period (e.g. Germain 1969, 79, no. 98, pl. XXV – Timgad). The wall of the apse does not describe a regular semi-circle and the space between the outer edge of this regularly described central Figure 7.35. (Plate 23) (right). Room 24, reception hall, view across room to east in front of apse (MS)

249

250

John Mitchell

Figure 7.36. (Plate 24) Room 24, reception hall, hub of apse with frontal female mask (MS)

composition and the wall is filled first with a crescentshaped band carrying a scrolling ivy trail with black and pink parti-coloured leaves that increase dramatically in size towards the centre; and then by a far narrower outer gusset right against the wall, a succession of small panels with simple geometrical compositions in black and white, a rosette of four spindles spanning a rectangle divided into four, a disc within a lozenge and a small ivy leaf (Fig. 7.38). The original rectangular mosaic surface in this room was supported by an underlying statumen of debris and soil. The overlying rudus, 80–100 mm thick, containing ceramic aggregate, was inferior in composition to that of the adjacent long gallery. The nucleus supporting the tesserae, which consisted of powdered stone, lime and sand, was here very thin, only 10–14 mm. Later post-holes sunk through the floor-surface allowed water to infiltrate the statumen, resulting in shrinkage and differential subsidence and the present uneven surface of the pavement. In the subsequent floor of the added apse the make-up was a little different. The statumen of this floor had a large proportion of soil with a poor aggregate of stones and broken tile, while the rudus, 50–60 mm thick, was deficient in lime and so lacked strength. As a consequence, this layer just reposed on the underlying statumen without bonding. The nucleus in the apsidal area, c. 35 mm thick, with inclusions of crushed Figure 7.37 (Plate 25) (left). Room 24, reception hall, detail of northeastern shoulder of apse (MS)

7  The mosaic pavements and painted walls of the domus

251

Figure 7.38. (Plate 26) Room 24, reception hall, detail of northern edge of apse (MS)

stone and sand, was similarly lacking in lime and so also weak and unstable. Here the poor quality of the statumen has led to the complete loss of the overlying nucleus and tesserae in many areas. A sherd from a Gazan amphora, probably of 4th-century date, possibly later, retrieved from the statumen in the area of the apse provides an approximate index for dating the floor. The tesserae are of local stone, black, white, pink, red and grey, and vary considerably in size. In the original area they tend to have slightly domical upper surfaces and measure between c. 5 × 5 and 9 × 9 mm, in the added strip in front of the apse, c. 9 × 8–10 × 9 mm. In the apse the stones that make up the overlapping scales are relatively large, while those used in the female mask at the hub of the apse are small, 4 × 5 / 5 × 5 mm; here tesserae of glass and bright brick-red terracotta were freely included for chromatic emphasis. The laying and the chromatic design of the original floor in this room are generally similar to the pavements in the gallery and the east vestibule. The tesserae are closely laid and create quite tight internal formations in their setting (Figs 7.19, 7.32). However, the setting is less tightly structured than in the pavement of the long gallery. In the reception room lines of tesserae have a tendency to veer from the straight and where a field of parallel rows meets a diagonal the junction tends to be fudged and filled with small sub-triangular scraps of stone; that is, not cleanly managed with adjacent lines of full rectangular tesserae in the colours of the two abutting

fields, as is the case in the gallery. The pavements in the gallery and the reception room are certainly not the work of the same team. An interval of time may well have elapsed between the laying of the two floors. The second-phase surfaces in the reception room are less disciplined and looser in their lay than the original floor. In the added strip that runs between the original rectangular surface and the apse, with the panels of running peltae and four-spindled rosettes, the setting is much less even than in the original floor of the room and the tesserae are irregular in shape with jagged edges; they are laid with little attention to order and there are considerable gaps between stones. In the apse there are the same irregularly shaped tesserae, with large spaces between stones. Here the tessellatum is laid in vertical rows but the lines have a tendency to wander. There must have been a palpable interval between the two main phases in this room. The chromatic tenor of this pavement in its original state, to a degree, would appear to continue that of the east vestibule and long gallery. Much of the mosaic is in black and white, with frequent accents in pink and some grey-blue. The floor is also laid in a manner that recalls the gallery pavement, with the tesserae set quite closely together in sequences which by and large make up coherent patterned structures. The floors in the two adjacent spaces are in harmony and work well together. Nothing can be said about the nature of the painted decoration of the walls in this room.

252

John Mitchell

Figure 7.39. Room 25, west vestibule, view to west (JBB)

Room 25: the western vestibule A third space with a preserved floor in mosaic was the western vestibule (Room 25). This is a roughly square space measuring 5.57 m (east–west) by 5.17 m (north–south), an entrance hall, with columnar tribelon openings in all four of its sides; it served the new principal entrance laid out in the major phase 3 remodelling of the house (Figs 7.39–7.41). Unlike the complex floor of the reception room (Room 24), the composition here is relatively simple and unified. Virtually the whole area is covered by one design, a ‘carpet’ of running-pelta pattern, an orthogonal scheme of tangent peltae in alternately upright and recumbent confronted pairs, the pairs alternately pink and grey-blue on a white ground (Prudhomme 1985: pl. 222d) (Fig. 7.42). These form whirling ‘propellers’ with blades of pink on one axis and pale blue on the other. This central area is framed by a single line of black tesserae, then by two lines of black and another line of white. There follows an 110 mm wide band with a median line and undulating wave running its length, the resultant semi-circles alternately pink and greyblue, the residual areas white. The outer edge of this band is defined by a single line of black tesserae, followed by two of white, two of red, two more of white, and then a single line of black. All this, in turn, is surrounded on all four sides by an outer frame, 0.46 m wide. This carries a reticulate design with a sequence of irregular octagons, intersecting on the long sides, forming in each octagon a central square and four elongated hexagons (Prudhomme 1985, pl. 28b) (Figs 7.43, 7.44). Pink saltire crosses with splayed arms are set in each square and in the hexagonal

surrounding fields. This frame is followed by a single line of white tesserae, three rows of pink, and then by a plain white outer border. The columnar screen on the east side of the room functioned as the principal entrance to the peristyle and to the interior of the domus in phase 3. Here, just in front of the threshold, on the western margin of the mosaic floor, bordering directly on the outer reticulated frame, an inscription was laid in red tesserae, to be seen by guests as they entered the house (Fig. 7.45). The preserved text is too fragmentary to be read in its entirety but has been deciphered as recording the owner of the house, with his name and his title, which implies senatorial status (see Chapter 2). The central section of this inscription is preserved to a length of c. 1.07 m. The letters, in red tesserae, c. 7 × 6 mm, on a white ground, were originally c. 130 mm. tall, although none is now preserved beyond 110 mm. The surviving letters may be transcribed as: [...]ẠṂΑΡ[ω ]ωΛΑΝ [.]| [.]O | Λ [... and have tentatively been read as: ...]ẠṂαρίῳ τῷ λανπ[ρ]οτά[τῳ|.. giving a name or title ending in …arius and a rank, to lamprotato, equivalent to that of clarissimus, of senatorial status (Charlotte Roueché, pers. comm.). The inscription appears to be cut by the marble stylobate that forms the threshold to the western portico of the peristyle. However, elsewhere the pavement in this room respects the phase 3 architectural fabric and it does not seem possible that it could be a left-over from a previous

Figure 7.40. (Plate 27) Room 25, west vestibule in vertical projection (MZ)

0

2m

Figure 7.41. Room 25, west vestibule, plan of mosaic pavement 0

2m

Figure 7.42. (Plate 28) Room 25, west vestibule, detail of central field (JBB)

Figure 7.43. (Plate 29) Room, 25, west vestibule, detail of border and central field (JBB)

Figure 7.44. (Plate 30) Room 25, west vestibule, detail of border (MS)

Figure 7.45. (Plate 31) Room 25, west vestibule, inscription at threshold to peristyle (MS)

7  The mosaic pavements and painted walls of the domus

255

phase. The present situation may have resulted from some early adjustment made to the stylobate or possibly from shifting foundations and subsidence. The statumen supporting this pavement consists of good stone aggregate set without apparent order or system in a soil matrix. The rudus above was only 30–40 mm thick, a poor mortar with ceramic powder, with scant lime and consequently very crumbly. The overlying nucleus is similarly thin, 10–14 mm, together with the tesserae c. 20 mm. This, like the rudus, is a poor mortar lacking a good binder, with rather less ceramic powder than the rudus. The poor quality of the make-up has resulted in the deterioration of the floor and the rather precarious state of its surface. The tesserae of this pavement are laid in a semblance of ordered rows but in fact with extreme irregularity so that there are barely three adjacent stones in the floor that are linearly aligned (Fig. 7.44). They come in all manner of differing shapes and sizes. A new colour aesthetic is introduced in this floor. Unlike the long gallery and the first-phase floor of the reception room, tesserae here are loosely set so as to form marks of colour, rather than to complement and enhance a precise linear design. The peristyle western portico (Room 16) Polychrome mosaic must originally have covered the floors of all four of the porticoes of the peristyle of the domus in phase 3, in the final years of the 4th century. To judge from the surviving surfaces, each of the walks was treated as an independent entity, and the panels of mosaic did not fill each arm in a completely regular and symmetrical fashion, although the bold acanthus scroll that frames the two surviving panels may have been designed to continue round and bring some coherence to the four walks of the court. The mosaicists had a limited range of coloured tesserae at their disposal: pink, pale blue, white, red, black and a bright orange-red. The floor of the west walk, immediately beyond the columnar entrance from the west vestibule, was laid for much of its length as one great panel, which extended from the southern end as far as the opening of the north portico (Figs 7.46–7.52). This field was framed by a double border; first a simple two-strand guilloche of pink and light blue ribbons lit with white, bordered by thick and thin bands of black; and secondly by an outer running acanthus scroll of great vigour on a white ground (Fig. 7.53). This scroll consists of an undulating stem, which forms a horn-like calyx in every arc, with a great curling claw-like serrated leaf in each hollow. The horns are variously coloured pink contoured in red or black, dark grey-blue contoured in red or light blue, and light grey-blue edged in black, all lit with white, usually following an alternating red/blue sequence. The leaves are a deep grey-blue with a black and a white line of tesserae contouring their closed outer sides. The elaboration of this acanthus scroll varies in the different parts of the portico. On the western side of the walk, directly in front of the columnar screen of the entrance from the west vestibule and running southwards, dramatic flowers

Figure 7.46. Room 16, west portico of peristyle, view to south (JBB)

of pink, yellow and black grow on long stems issuing from the horns around which ribbons wind, green-grey against pink and red against blue. However, in the northern end of the portico the flowers are less elaborate and at the far northern end of the panel they are omitted altogether (Figs 7.54 and 7.86). On the east side of the west portico, the side opposite the entrance, the execution is more restrained, pink and blue horns alternate, but detail and ornamentation is less elaborate than in the facing border, and the ground seems to be a greenish ochre. However, at the far southern end of the walk there is even more elaboration and ornamentation than on the western side by the entrance (Figs 7.53, 7.55, 7.77 and 7.80). Here more colours are introduced and in the southwestern corner the sequence is broken by a yellow horn, now decayed, with pink circling ribbon. Across the southern end of the portico the inner guilloche border is more complex too, a three-ply (rather than a two-ply) braid with strands in red, blue and yellow, and each point of intersection emphasized by a white tessera outlined in black. To judge from the borders of the western portico, therefore, particular emphasis in design and colour was given to the southern end of the space, while the visual effect was more generally restrained towards the northern end.

256 John Mitchell

Figure 7.47. (Plate 32) Room 16, west portico of peristyle, northern end, north panel in vertical projection (MZ)

Figure 7.48. (Plate 33) Room 16, west portico of peristyle, northern end, in vertical projection (MZ)

7  The mosaic pavements and painted walls of the domus

Figure 7.50. (Plate 35) Room 16, west portico of peristyle, southern end, in vertical projection (MZ)

257

Figure 7.49. (Plate 34) Room 16, west portico of peristyle, central section, in vertical projection (MZ)

John Mitchell

258

Figure 7.52. (Plate 36) Room 16, west portico of peristyle, main panel, to south (JBB)

0

2m

Figure 7.51. Room 16, west portico of peristyle, plan of mosaic pavement

The design of the great inner panel, which extends over much of the western portico, consists of a diagonal reticulate pattern of tangent eight-lozenge stars forming large poised squares and small orthogonal squares (cf. Prudhomme 1985, pl. 173b and e) (Fig. 7.56). The cells, which make up the pattern, alternate pink and grey-blue outlined in white. The smaller orthogonal squares have an additional inner black contour. The design may be read as a reticulation of interacting cuboid saltires, in ambiguous perspectival projection, with the cuboid arms of adjacent saltires sharing terminals – the small orthogonal squares – enclosing large poised squares. It is these large poised squares that constitute the dominant accents on this stretch of pavement. They carry a variety of motifs and patterns: a reticulated grid, overlapping semi-circular scales, peltae and crescents in various formations, a series of crosses, cruciform knots of Solomon, individual birds in profile, and one single large eye. In general, squares with the same or related motifs are laid in linear sequences running west–east across the short axis of the portico. The poor state of preservation of the floor, considerable areas of which are now bereft of tesserae, means that the precise distribution of all of these designs cannot be reconstructed

7  The mosaic pavements and painted walls of the domus

259

Figure 7.53. (Plate 37) Room 16, west portico of peristyle, border in southwestern corner (MS)

Figure 7.54. (Plate 38) Room 16, west portico of peristyle, border in northeastern sector (MS)

Figure 7.55. (Plate 39) Room 16, west portico of peristyle, border in southwestern sector (MS)

in its entirety. However, the broad sense of their disposition is clear. The western portico of the peristyle served in effect as the entrance hall to the phase 3 domus; it was on this surface that a guest would set foot on crossing the vestibule and passing through the tribelon screen of columns into the inner court. In considering the design of the pavement in its detail, therefore, it makes sense to start at the centre, with the floor directly in front of the entrance, then to cast the eye to the left (that is to the north) and finally to turn to the southern end of the space. The floor-surface immediately in front of the columnar entrance from the west vestibule has suffered more damage than any other part of this portico; for the most part the mosaic here has been destroyed (Figs 7.57, 7.58). However, in the four poised squares of which something survives, which lie adjacent to each other in a north–south sequence just in from the threshold, four birds are depicted, all in profile facing to the left. These are all brightly coloured, blue with red details, white, pale green and pale blue, and

Figure 7.56. (Plate 41) Room 16, west portico of peristyle, southern end, rows 27–30, to west (MS)

John Mitchell

260

Figure 7.57. (Plate 40) Room 16, west portico of peristyle, central section immediately in front of the entrance from the west vestibule (JBB)

9 0

10

11

12

13

14

15

16

17

18

19

20

1m

Figure 7.58. Room 16, west portico of peristyle, central section, drawing

stand among plants with blue leaves and red flowers. In these squares, liberal use is made of glass tesserae. It is only here in this portico that birds are represented, and it seems reasonable to suppose that the other squares in this central area contained similar subject matter, probably lines of birds running in west–east sequence in neighbouring squares across the portico. The northernmost of the four

(row 11) is a goose-like bird with red legs and beak and a white body shaded in pale blue (Fig. 7.59); of the next (row 13) only the head is preserved, outlined in pale green, and the top of a tree or plant; the third (row 15), perched on a leafy green tree, has a bright blue body outlined in purple porphyry stones, with a yellow wing contoured in red, a red blaze on its tail and red legs (Fig. 7.60); the fourth (row 17)

7  The mosaic pavements and painted walls of the domus

261

is pale blue with a wing striated in black, pink and white and red legs, with three large red flowers above its back and one at its feet. All are set against pink ground. This central section is bounded to the north, for the most part, by in-line sequences of three squares bearing the same or related designs running west–east across the portico. None of the motifs in the first row (row 10) has survived but the three squares in the next line (row 9) all contain a configuration of four fronted white peltae on a pink ground, the points of their backs tangent with the sides of the square (Figs 7.61–7.64). A small black cross is set at the mid-point. The designs in the outer squares of

the following sequence to the north are similar (row 8), in one a pair of dark grey-blue crescents back-to-back on one axis and in the other a pair of curtailed white peltae also back-to-back, on a pink ground. The middle square is exceptional and contains two large eyes superimposed, or possibly one all-seeing eye shown simultaneously on two axes (Figs 7.65, 7.66). On the west–east axis the eye

Figure 7.59. (Plate 42) Room 16, west portico of peristyle, central section, row 11, to east, goose-like bird (JBB)

Figure 7.60. (Plate 43) Room 16, west portico of peristyle, central section, row 15, to east, bird on tree (MS)

Figure 7.61. (Plate 44) Room 16, west portico of peristyle, view from centre looking north (JBB)

262

John Mitchell

Figure 7.62. (Plate 45) Room 16, west portico of peristyle, northern end, to north (JBB) is outlined in red and contoured in pale blue and purplish pink inside, while the eye on the other axis is defined in yellow contoured in black. At its centre is an interior eye outlined in black with a pale blue pupil in a white iris, set vertically with respect to the dominant outer red eye. The red eye is oriented so as to meet the glance of someone looking the length of the portico. Proceeding northwards, the next sequence of poised squares (row 7) contains crosses formed of two interlacing continuous oval ribbons, so-called knots of Solomon. The ribbons forming these knots are trichrome, pink, red and white, set against a black ground. These are followed by a sequence of panels (row 6) with overlapping white scales, contoured in black (Fig. 7.67); and then by squares (row 5) with a chequer-board grid in pink and white; followed by a further sequence (row 4) with large white four-petalled rosettes, with serrated triangles coloured black and white, chequerboard-fashion, rising from the sides of the squares between the petals. The three preserved squares in the penultimate row (row 3) have three different designs, the first a four-spindled white rosette with opposed pink and blue scale motifs between the petals, the second a field of contiguous repeating rows of superimposed chevrons with arms alternating white and pale blue, forming a zig-zag pattern (Prudhomme 1985, pl. 9g), and the third a field of somewhat irregular zig-zag rainbow pattern, in which each row is formed of lines of contiguous black, pink, blue, white and pale orange tesserae, not always in full sequence. Each square in the final row (row 2) contains a pair of fronted white peltae with little pink semi-circles emerging from the frame on the counter-axis (Fig. 7.68). Figure 7.63 (Plate 46) (left). Room 16, west portico of peristyle, northern end, east side, to north (MZ)

7  The mosaic pavements and painted walls of the domus

1 0

2

3

4

5

6

7

8

263

9

10

1m

Figure 7.64. Room 16, west portico of peristyle, northern end, plan of mosaic pavement

Figure 7.65. (Plate 47) Room 16, west portico of peristyle, northern end, rows 7–9, to south, with knots of Solomon

264

John Mitchell

Figure 7.66. (Plate 48) Room 16, west portico of peristyle, northern end, row 8, to north, eye (JBB)

The half-squares cut by the northern frame of the panel (row 1) are either horizontal dentate bands of black, white, pink and pale blue or nested chevrons laid similarly with poised tesserae in the same colours. The pavement to the south of the central section is more varied, more colourful and generally brighter than the north end of the walk (Figs 7.69–7.72). The two sequences immediately to the south of the southernmost of the four preserved birds have been destroyed (rows 18–19). Only one square survives in the next sequence (row 20), with a pair of white affronted peltae heavily contoured in black and little white lunettes rising from the edges of the frame on the other axis, against a pink ground – this is the same motif as the sequence at the extreme far north end of this panel (Fig. 7.73). The following line (row 21) consists of knots of Solomon, with looping ribbons in red

Figure 7.67. (Plate 49) Room 16, west portico of peristyle, northern end, rows 5–6, to south, with overlapping scales and chequerboards (JBB)

Figure 7.68. (Plate 50) Room 16, west portico of peristyle, northeastern corner, row 2, to south, with fronted peltae (JBB)

Figure 7.69. (Plate 51) Room 16, west portico of peristyle, view from centre. to south (JBB)

Figure 7.70. (Plate 52) Room 16, west portico of peristyle, southern end, to south (JBB)

Figure 7.71. (Plate 53) Room 16, west portico of peristyle, southern end, to west (JBB)

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

0

Figure 7.72. Room 16, west portico of peristyle, southern end, plan of mosaic pavement

1m

266

John Mitchell

and pink contoured in white against a black ground; two are preserved. Two of the three squares in the next line survive (row 22), each with a group of five contiguous white rosettes sharing slender foliate petals, forming an orthogonal grid, with crosses of five poised red tesserae in the intervals (Fig. 7.74). Next (row 23) are three squares with fields of contiguous repeating rows of superimposed chevrons with arms alternating white and pale blue/pale blue and dark blue, and red and pink in succession, forming a zig-zag pattern, with the darker combination in the median square; then a line (row 24) of three squares with a diagonal chequerboard grid, pink and white. In the following sequence (row 25) two prominent cruciform compositions flank a median square with a field of bold black, red, white, pink and pale blue zig-zags running on

Figure 7.73. (Plate 54) Room 16, west portico of peristyle, southern end, rows 20–21, to south, with fronted peltae and knots of Solomon (JBB)

the west–east axis. The cruciform devices consist of an orthogonal square within the large poised pink square, the smaller square heavily contoured in black and divided into four quarters, pink contoured in red, with a central stubby white saltire cross in each and an overlying larger saltire cross described by single lines of dark blue tesserae spanning the quarter square. The next row (row 26) consists of squares each of which is filled with a field of horizontal dentate bands, aligned on the north–south axis, made up of small contiguous poised squares in red, pale blue, white, black and pink (Fig. 7.75). By contrast there is variety in the following row (row 27). In the square adjacent to the west wall, a complex composition of nested crosses, with white and pale blue against pale pink dominating in a diffuse and ethereal outer cross, and dark blue defining a taught linear interior cross at the centre, set against pink and strong red and forcefully contoured in black. A field of zigzag chevron bands with arms light blue and dark blue, and red and pink, in alternation, fills the median square in this sequence; and the third contains a field of fronted red peltae forming a central Catherine-wheel on a pale blue ground. The next sequence (row 28) is a row of knots of Solomon, with red and dark pink ribbons lit with white along one edge, set against black (Fig. 7.76); then (row 29) three squares each with a pair of fronted pink peltae filling the full field with concave triangles outlined in red and blue rising from the frame into the resulting spandrels. In the remaining five rows at the far south end of the walk, to judge from what survives of the rather damaged surface, the disposition of motifs is more random, and sequences of identical devices are abandoned in favour of constantly changing variety (Fig. 7.77). In the first of these rows (row 30), the square closest to the west wall contains

Figure 7.74. (Plate 55) Room 16, west portico of peristyle, southern end, rows 22–26, to south (JBB)

7  The mosaic pavements and painted walls of the domus

267

Figure 7.75. (Plate 56) Room 16, west portico of peristyle, southern end, rows 24–28, to south (JBB)

Figure 7.76. (Plate 57) Room 16, west portico of peristyle, southern end, rows 26–30, to south (JBB)

a large rosette with four white foliate petals against pink with little stepped features between the petals; and the second has a chequerboard field of little squares in white, pink, pale blue and red crossed with white, in a now unintelligible sequence; the third square in the row is lost. The westernmost square in the following row (row 31), for the most part destroyed, was divided into a field of triangles, red and pink, dark and pale blue, forming an interior square with strongly defined orthogonal and diagonal cruciform axes, probably with a dominant cross element; the next carried a field of rainbow zigzags in red, white, black and

pale blue; and the third square is now lost. In the third sequence (row 32), the westernmost motif consists of four fronted and contiguous pink peltae contoured in red against white, enclosing a serrated square, black and white at the centre and pale blue and pink at the edges (Fig. 7.78). The second square in this row has a striking configuration of four parallelipipeds in cruciform array, with angled surfaces in red, pink and white/pale blue, pale yellow and white, all with a little square of red tesserae centred on their white tops, so as to suggest perspectival recession in a black field (Fig. 7.79). Again the eastern square in this row is missing.

268

John Mitchell

Figure 7.77. (Plate 58) Room 16, west portico of peristyle, southern end, rows 35–30, to north (MS)

Figure 7.78. (Plate 59) Room 16, west portico of peristyle, southwestern corner, rows 31–33, to east, with fronted peltae and dentated cross (MS)

Figure 7.79. (Plate 60) Room 16, west portico of peristyle, southern end, rows 32–34, to south, with perspectival parallelipipids and rainbow zigzags (JBB)

A large cross in red and pink poised squarelets, serrated in black, spans the western square of the fourth row (row 33), the intervals between the arms filled with stepped nested chevrons in pink, red, pale blue, dark blue and white (Fig. 7.78); the central square is destroyed; and the eastern one holds the remains of a rosette with four foliate white petals and little black sprigs growing up into the intervening spaces. In the final sequence of squares (row 34), although there is now much disturbance, it is clear that colouristic brilliance prevailed. In the centre was a knot of Solomon with ribbons in red and white and blue and white; to the west orthogonal serrated bands and lines formed nested frames both cutting and contained within the outer poised contour, red, white black, dark blue and pink, and at the

centre a focal chequerboard in black and white; and to the east there was a simple field of polychrome linear zig-zag, red, white and black. The resulting half-squares that finish off this end of the long rectangular tableau (row 35) are also exuberantly bright in their colouring: pink peltae contoured in red against a white ground, to the west; sun-rays of little poised squares in red, white, black and pale blue in the middle; and a nested step-configuration in red, pink, blue, white and black to the east (Figs 7.80, 7.81). At the north end of this great panel a reticulate band running across the walk, west–east, forms the threshold to a second mosaic tableau beyond (Figs 7.47, 7.82–7.84). This strip consists of an orthogonal pattern of irregular octagons adjacent and intersecting on their longer sides, to

7  The mosaic pavements and painted walls of the domus

Figure 7.80. (Plate 61) Room 16, west portico of peristyle, far southern end, rows 35–34, to north (MS)

269

Figure 7.81. (Plate 62) Room 16, west portico of peristyle, southern end, row 35, to north, half poised square with stepped design (MS)

Figure 7.82. (Plate 63) Room 16, west portico of peristyle, northern end, looking north over north panel (MS)

form squares and oblong hexagons (variant of Prudhomme 1985, pl. 166e), each resulting cell charged with a swastika. Immediately beyond, to the north there is a further large tableau of mosaic, of which only the western section is now preserved. The design is set at right-angles to the panel in the west walk, and once ran out to the east, down the north walk of the phase 3 peristyle. Like its companion in the west walk, this large panel is encompassed on its western and northern sides by the continuation of the exuberant acanthus scroll with its claw-like curling leaves that united these two floor-spaces into one continuum. However, this panel has its own inner frame, a powerful running wave motif, in pink and white, which frames the field on its southern, western and northern sides and

presumably continued down round its lost eastern extension. The design of the central tableau of this area is quite simple, yet complex in its visual outcomes: an orthogonal pattern of adjacent irregular octagons, forming poised squares, each square framing a prominent pink disc contoured in black, the octagons bearing alternately reversed inscribed swastikas (cf. Prudhomme 1985, pl. 166,e. For variants of this design, see Guimier-Sorbets 1983, pls CXXV and CXXXVI) (Figs 7.85, 7.86). The simple underlying scheme of continuous octagons is complicated and controverted by the linear joining of the arms of the swastikas that results in a dominant overlay of tangent interlocked rectangles with lateral projections. This devise transforms a relatively simple scheme into one

270

John Mitchell

0

1m

Figure 7.83. Room 16, west portico of peristyle, northern end, north panel, drawing

Figure 7.84. (Plate 64) Room 16, west portico of peristyle, northern end, border between main and northern panels, to west (JBB)

7  The mosaic pavements and painted walls of the domus

271

Figure 7.85. (Plate 65) Room 16, west portico of peristyle, northern end, northern side of north panel, to west (JBB)

Figure 7.86. (Plate 66) Room 16, west portico of peristyle, northern end, field of north panel, to west (JBB)

in which the generating configuration tends to be lost sight of and enters into constant shuttling equivocation with a secondary superimposed linear order. In this portico the statumen consists predominantly of stones and broken tiles; these are set on edge, possibly to assist drainage. The overlying rudus, which is quite thin, 50–70 mm, penetrates and bonds with the statumen; this contains an aggregate of river stones, broken tiles, shells, ceramic powder and river sand but the binder is poor and weak, only scant lime of indifferent quality. The nucleus also is extremely thin, 20–25 mm together with the tesserae it supports. This is a mortar made up of crushed stone, ceramic powder and river sand. As noted in Chapter 2, a sondage into the statumen of this floor, in one of the areas now devoid of tesserae, revealed a stamped base of a ARS dish (ARS 67 or similar), datable to between c. AD 360 and the early 5th century, a fragment of a cooking vessel of later 4th- to 5th-century type, a large fragment of marble furniture with promegranates carved in flat relief, possibly from the display front of a dining stibadium, probably work of the mid 4th century, vessel glass and fragments of painted plaster. This material suggests a rough date towards the end of the 4th century for this mosaic (and indeed other evidence described in Chapter 2 suggests a date of around AD 400). The pavement of the adjoining vestibule is likely to have been laid at more or less the same time. The lay of this floor is similar to that of the adjacent western vestibule. The tesserae are set close but extremely irregularly. No attempt was made by the mosaicists to create structured sequences in the setting of the stones and there are no regular lines. The configurations of linear devices like the four spindle-petals of a rosette or the bands of a knot of Solomon are not laid out with precision; petals do not meet at a central point and the bands that constitute the potent knots describe irregular courses. Tesserae are deployed in loose array to create areas of colour rather than to define and articulate graphic designs. As in the

adjacent west vestibule, the aesthetic here is predominantly a colouristic one; however, the intense density of the polychrome accents in the peristyle shows a marked increase on the floor of the vestibule.

Analysis The four surviving mosaic pavements of the domus, alongside the painted decoration of the rooms, would have served, together with other now lost appropriately appointed spaces in other parts of the house, as coherent decorative ensembles in successive phases of its history. From the 3rd-century phase 2 building, the eastern vestibule and the long gallery (Room 18) and the reception room (24) are preserved. The visitor, passing through the main entrance to the house, would have been confronted by two joining spaces decorated with restrained elegance and measured grandeur, the walls painted to resemble an airy loggia, with fictive red columns rising above a stylobate painted in imitation of panels of polychrome marble, yellow alternating with pale blue or plain unpainted plaster (Figs 7.12, 7.20). In the long gallery, in which the scheme evidenced in the vestibule very likely continued, painted columns would doubtless have complimented a loggia of real columns in the southern wall, with views out over the water, across the Vrina Plain to the distant range of hills. In the vestibule, the pavement has a central field laid out with a complex swastika-meander design, incorporating expertly characterized masks from the repertoire of New Comedy, including two principal protagonists, a young girl paired with her characteristically choleric father (Figs 7.2, 7.3, 8). These with two other masks would have surrounded a focal subject in the middle of the floor, possibly a fulllength figure of Dionysos, lord of the theatre. The masks of Tragedy and New Comedy appear as an occasional element in the decorative repertoire of houses throughout the Roman world from the 2nd century BC through to the 4th century AD. They could play either a

272

John Mitchell

major or a subordinate role in the iconographic composition of the mosaic floor of a room within a house (Lancha 1981b, 124–40 – Vienne (Isère); Lavagne 2005, 240–5, pls LXXVI–LXXXVII – Aix-en-Provence; Werner 1998, 99–141 – Rome; Kraus and von Matt 1975, pl. 103; Andreae 2003 – Pompeii; Foucher 1965 – Sousse; Salzmann 1991, 449–52, abb. 18–20 – Pergamum), or they could feature in schemes of painted decoration (Kraus and von Matt 1975, pl. 81; Barbet 1985, fig. 27; Ling 1991, figs 26, 227h-8), the most famous instance being the Room of the Masks in the House of Augustus on the Palatine Hill in Rome (Carettoni 1983, 23–6, colour pls A–C, E–H; Ling 1991: figs 33–4). Masks were reproduced in various media; they constituted the principal subject on a particular category of carved relief, the mask-relief, which enjoyed wide popularity under the late Republic and the early Empire (Kraus and von Matt 1975, 41 and 42; Cain 1988), and replicas in marble were set into the walls of domestic nymphaea and suspended between the columns of the peristyles of houses (Kraus and von Matt 1975, pls 82–3, 163). In all these cases the iconographic freight of the motif was probably not heavy. They made a probably not-too-grave proclamation of the cultural pretensions of the occupants of the house, and at the same time their grotesque, forbidding character, their black screaming mouths, would have figured as stereotypical apotropaic devices. Their reference to the practices of comedy, their gentle allusion to Dionysos, the god who presided over old traditions of cultural formation and expansive hospitality, and their fierce grimaces made them effective guardians at the door of the house, bringers of good fortune, a welcome to guests and a fright to inauspicious visitors (Cain 1988, 189). It is with something of the same sense that a wonderful polychrome garland with tragic masks was laid across the entrance to the House of the Faun at Pompeii (Andreae 2003, 218–45, figs 236–9; Beard and Henderson 2001, fig. 13g; Kraus and von Matt 1975, pls 39, 108). The mosaic in both the vestibule and the long gallery is laid with expert precision, a single panel filling each space (Figs 7.3–7.6). In both, black and white predominates in the main central panel, with a few colours introduced in the frames, pink, red and two shades of grey-blue, to establish contrast and to lend emphasis to the centre. In the vestibule this polychromy is more pervasive, invading the central panel in the running guilloches and in the masks that punctuate the design (Fig. 7.8), a strong presence in the outer border of double guilloche and the delicate consolefrieze, and dominant in the ivy trail that runs around the walls (Fig. 7.8). However, the two zones of polychromy are separated in this entrance-space by a broad intervening frame of running peltae and spindle-rosettes in unbroken back and white (Fig. 7.3, 7.7). So in the vestibule, a relatively complex and variegated floor greets the visitor, with figural elements, heightened colour accents and variety of design. This gives way in the long gallery to a more

restrained and dignified space, its pavement stretching away into the interior of the domus, the unvaried black and white of the long central field contrasting with subdued colour at the edges and complimented by soft red painted columns that transformed the walls into an open loggia (Figs 7.4, 7.6). The reception room (24), which in its initial form appears to belong to the same phase of construction, was designed to communicate directly with the long gallery and to function along with it (Fig. 7.21). The pavements of the east vestibule and gallery would have led directly from the eastern entrance to this dining room with its elegant octagonal fountain and elaborate floor. In the first phase of this room, in which it had a rectangular form, the pavement was laid out with a sequence of framing elements set around a central medallion, the centre of which is now destroyed (Fig. 7.22). The two squares which contain this disc are also in a poor state of preservation but vestiges of elaborate chalices and vases can be made out in their four corners, from which the stems of plants with flowers burst forth (Figs 7.25–7.27). The repertoire used by the mosaicists in this room for the most part consists of abstract ornament but these central elements may have been chosen for their appropriateness for one of the principal public rooms of the house, which most likely served as a hall for reception and audience (see Chapter 8). Central circular motifs, sometimes with figures in action, were not uncommonly deployed in reception rooms and in triclinia in Roman houses, where they would have underscored the constant motion that was a feature of the uses to which rooms of this kind were put; in audience halls accompanying the movements of the clients and friends who waited on the master of the house, and in banquetting rooms supporting the choreographies of the performers who entertained the guests and the movements of the servants who hurried constantly to and fro as they served the diners at their couches (Ellis 1997a). The chromatic tenor of this pavement in its original state, to a degree, would appear to continue that of the east vestibule and the long gallery. Much of the mosaic pavement is in black and white, with frequent accents in pink and some grey-blue. The floor is also laid in a manner that recalls the pavement in the gallery, with the tesserae set quite closely together in sequences that by and large make up coherent patterned structures. These adjacent spaces work together in harmony. This reference to the pavements of the vestibule and gallery was picked up in the extension to the floor in the reception room, when in a subsequent phase (phase 2b) the original rectangular space was enlarged to the north and a large apse added (Fig. 7.23). The focal motif in mosaic at the hub of this apse was the frontal mask of a female protagonist from New Comedy (Fig. 7.36). Although only the upper part of the head is preserved, it is clear that this is a repetition of the frontal female mask in the east vestibule (Fig. 7.9). In this way the floor of the vestibule was invested with renewed value, to start a line of cultural

7  The mosaic pavements and painted walls of the domus association that a guest could pick up and recognize at a subsequent point, while progressing through the house in the course of a visit. The mask in Room 24 would have provided a focal point of reference at formal receptions as well as at the less formal gatherings for which this room was designed. The other two surviving areas of mosaic pavement belong to a later phase, a radical reconstruction and refitting that took place probably at the very end of the 4th century. In both the new west vestibule and in the western portico of the peristyle the mosaics are laid in a very different way from those in the long gallery and the reception room. In these later rooms, the tesserae vary considerably in size and in shape and are set with only scant attention to sequence and structure. The contours of the design are approximate and often distorted, lines of tesserae waver and buckle; stones are used to shape, fill or mark their allotted spaces with areas or accents of colour. There is a focus on colour here, a de-emphasis of line, which is very much at variance with the earlier floors. In the west vestibule most of the floor is filled with a great central panel of running peltae, over which anyone entering the house would have had to walk (Figs 7.40 and 42). The peltae form two patterns, one of affronted pairs, the other of whirling ‘propellers’, which oscillate in front of the eye and never fully settle in the mind. Peltae, with their two profiles, one smoothly crescent the other sharply cusped, are potent motifs, commonly deployed on public monuments in Roman Italy and the inner provinces of the empire, visually arresting, always retaining some vestigial resonance of their origins in a type of leathercovered wicker skirmishing shield that was commonly associated with Amazons, the mythic warrior women of ineluctable fierceness and aggression. The attention of the visitor passing over this floor would have been taken and fascinated by this constantly changing pattern, drawn to understand and resolve the protean ambiguities in the design but inevitably foiled as the peltae swung to and fro between their alternate configurations. Likewise, in an apotropaic sense, this visually treacherous floor laid a field of spinning, shifting obstacles for any evil force which might think to float in through the entrance and invade the house. The inscription on the threshold to the peristyle can be understood in a similar way (Mitchell 2007, 287) (Fig. 7.45). On one level, and most obviously, it addressed the visitor, announcing the title and name of the owner of the house; on another, it may well have had a sentinel function. There was a tradition of deploying inscriptions at the doors of buildings and tombs, to safeguard the entrance and keep malevolent spirits at bay (Maguire, Maguire and Duncan-Flowers 1989, 16–18; Prentice 1906). At Butrint, the threshold inscription may well have had a force of this kind, confronting an unwelcome intruder with the name of the powerful proprietor of the house; here it would have had something of the same powers as the names, titles and images of emperors on the medallions and coins commonly

273

worn as pendants in this period to secure for the wearer the visible protection of a puissant patron (Maguire 1997). The pavement in the western portico of the peristyle, in effect the inner front hall to the house, which was laid in the same building-phase, continues and measurably enhances this iconography of address and captivation. The overall scheme of the floor is one designed to hold and tease the eye, a reticulate framework of tangent eight-lozenge stars forming interconnecting saltire-crosses, which meet in little squares, in such a way that their longitudinally divided arms shuttle between faceted projection and flat planarity (Fig. 7.56). This visually ambivalent matrix is inflected over its length by variation in colour and ornamental elaboration and by the positioning and concentration of particular motifs in the rows of poised squares that constitute the principal focuses of the pavement. The area immediately in front of the columnar entrance from the western vestibule is largely destroyed, but the few remaining areas show that bright red and blue glass tesserae were used in these squares, giving a particular brilliance to this part of the floor (Figs 7.49, 7.57–7.58). In the acanthus scroll that frames the central field, the curling horn-like calyces are alternately pink and blue, dramatically contoured in contrasting colours, with bold flowers growing on tall stems between each volute and gay ribbons winding round the horns. If one looks to the left, towards the northern end of this portico this polychrome brilliance diminishes (Figs 7.48, 7.62, 7.63); the squares are here duller and less differentiated in colour, with uniform motifs appearing in sequence, and in the border towards the northern end of the walk the colours are less varied and less intense, and the flowers between the volutes are reduced or completely absent. At the southern end of this portico, however, the opposite occurs (Figs 7.50, 7.69–7.71). Here colourful glass tesserae are introduced in some of the squares, and the motifs employed are striking for their variety, chromatic brilliance, inventiveness and complexity, with a different device selected for each of the surviving squares in the five southernmost sequences. The acanthus border at this southern end grows more exuberant, with bright contouring, flowers of varying light colours, prominent winding ribbons and at one point yellow being introduced for one of the scrolling calyces. There is a clear visual strategy here, to divert attention from the northern part of the portico, which presumably led to predominantly private apartments in the northern range of the peristyle, and to draw the eye and lead the mind towards the more public rooms in the southern range. Despite the fact that by this time the pavement at the western end of the gallery had been rather roughly cut up by an inserted drain (see Chapter 2), it would appear from the new mosaic in the peristyle that this space and the adjoining reception room may have continued to play a significant role in the public life of the house. A similar but more complex situation is to be found in the motifs in the poised squares. Directly in front of the entrance gaily-coloured birds predominated (Figs 7.49,

274

John Mitchell

7.58–7.60); it was only here that any living creature, plant or artifact was represented in the mosaic. Birds with bright plumage typically figure in late Roman and early Christian contexts as indicators of delight and leisured luxury, often with intimations of transcendence, of the delights of an ideal paradisiacal world. Whether reference should be seen here to souls in a Christian paradise is not made explicit, although there is little doubt that the occupants of the house at this time were practicing Christians. The tenor of the iconography is clear, in its declaration of a place of blessed peace and delight, a secure enclave removed from the immediate vicissitudes of everyday life (Maguire, Maguire and Duncan-Flowers 1989, 9–13). To north and south, this entrance area is flanked by squares containing a repertoire of motifs, which at first sight seem merely abstract and ornamental but on closer inspection engage the observer with their sequenced concentrations and insistent referential imagery. To the left, that is to the north, the rows beyond the birds (rows 9–7) contained particularly potent devices (Figs 7.48, 7.61–7.67); first a sequence of fronted peltae, next peltae and crescents together framing a median square with a glaring orange eye, then a line of knots of Solomon, and finally a sequence of dense overlapping leaves or tiles. Peltae, which, as we have seen, had been used to effect in the vestibule, recalled in their curiously rounded and sharply cusped profiles the shields of dangerous irregular troops and fierce Amazons; they formed a standard part of the repertoire of Roman mosaicists, who would often employ them on thresholds and other critical locations, where they marked the passage from one space to another or framed a central motif, as was the case in this very house in both the east and west vestibules (Rooms 18 and 25; Figs 7.3, 7.7, 7.8, 7.11–7.13, 7.40–7.43) and in the apsidal reception room (Room 24; Figs 7.22, 7.24, 7.25, 7.28–7.31). The knot of Solomon (Fig. 7.65) was a similarly potent device combining the protective powers of the cross and the knot, which was used throughout the Mediterranean in late antiquity, at times as little more than an ornamental device but often with clear apotropaic purpose (Maguire, Maguire and Duncan-Flowers 19989, 3–4; Sansoni 1998. See also Piccirillo 1993, 320–23; Kitzinger 1993, 4, fig. 1.2, and in general Trilling 1995, 70–6). Like peltae, overlapping leaves or scales are a motif that was widely used but their significance for contemporaries is not easy to pin down (Figs 7.63, 7.64, 7.67). However, whether this device should be identified as referencing a field of closely overlapping laurel-leaves of the kind commonly found on Jupiter-columns and other monuments of a sacral or celebratory kind in the Roman period (Bauchhnenss 1976: ills 13, 23, 40), or whether they should be read as tiles or scales, their sense would seem to be generally protective and empowering. The eye that glares out from the middle of this supporting field of puissant shapes is the Evil Eye, the eye of the envious, covetous, harmful gaze, which by sympathetic reversal could catch and hold any intruding force that might

seek to upset the spiritual or physical well-being of the house and its occupants (Figs 7.61–7.66). The eye, which offered protection against the effects of envy, was widely used on amulets and other artifacts of everyday life (Bonner 1950, 97–100; Dunbabin and Dickie 1983; Engemann 1975, 24–40, giving the principal earlier literature at 24 n. 13; Trilling 19995, 70). However, its use as the focal device on a mosaic pavement is not particularly common (cf. Engemann 1975, pls 10a – Jekmejeh near Antioch, 11a–b – Rome, 14c – Sousse); and its presence in the peristyle of the house at Butrint suggests that apotropaic aversion of harmful forces was a particular concern for the occupants of the house and that this may be a dominant subject of the iconography of the floor. Its particular function in this position would appear to be to control inappropriate entry to the private apartments in the northern range of the peristyle. The motifs deployed in the squares at the far northern end of this walk are less demonstrative: reticulate chequerboards, spindle-leaved rosettes, and polychrome zigzags (Figs 7.48, 7.64). However, the paired peltae that fill the three squares of the final row pick up the sense of the field surrounding the eye to lay down a powerful terminus for the whole panel and mark a threshold before the succeeding panel and the turn into the northern portico (Fig. 7.68). This sense of a threshold or boundary is emphasized by the band of intersecting irregular octagons, with a swastika in every cell, which runs across the width of the walk at this point, separating the long north–south panel from the intricate but iconographically less freighted west–east panel that runs down into the north walk of the peristyle (Figs 7.82–7.84). To the south of the birds in the entrance area of the west portico the array of motifs in the squares at first generally corresponds to the assemblage around the eye to the north, but as one progresses along the walk they develop very differently. Much of the mosaic has been destroyed at this point, but it seems that after the birds, peltae were followed by a row of knots of Solomon, then by sequences of interlocked spindle-leaved rosettes and fields of polychrome chevron-zigzags (Figs 7.50, 7.69, 7.72–7.74). However, here there is a strength in the definition of the motifs, a brightness in their high-lighting and a clarity in their coloration that contrasts sharply with the relatively subdued tones of the surface to the north. Moving southwards, there are at first intermittent rows with three similar configurations, diagonal chequerboards, knots of Solomon and pairs of fronted peltae (Figs 7.74, 7.76). However, on the whole, the assemblage of motifs is more various here than at the northern end of the portico, and in the final five rows, to judge from what survives, there was a systematic avoidance of such repetition, and a marked increase in variety and strong chromatic brilliance and contrast, with motifs more frequently set against a black ground to striking effect (Figs 7.71, 7.77). The eyes are here diverted by a veritable riot of colour and sharp dramatic incident.

7  The mosaic pavements and painted walls of the domus A new element is introduced in this southern walk, the cross. This features in a range of differing manifestations, some overt, others relatively covert. The first clear cruciform motif is deployed in the two outer squares of the eighth row beyond the birds (row 25), which flank a square of polychrome zigzags (Figs 7.72, 7.74–7.75). In these two fields a bright red cross within a red square powerfully contoured in black is complemented by blue saltires in its four quarters. The design stands out from the surrounding devices and immediately draws the eye. Two rows further on (row 27) a very different cross fills the western square, associated with a field of polychrome chevron zigzags and a propeller of whirling peltae in the eastern square (Figs 7.72, 7.75, 7.76). The western cross is a formation with spectral arms in white and pale blue against a pink ground; at its centre a fine blue cross, set against pink and then red within a poised square strongly contoured in black. The two very different contrasting crosses in this composition combine to telling effect. It should be noted that the whirling peltae in the eastern square in this row can also be read as a cross in rapid circular motion with sweeping arms. The following sequence (row 28) is made up of three cruciform knots of Solomon, particularly prominent with their red and pink ribbons lit with white and floating against a black ground (Fig. 7.72, 7.76). Next comes a row (row 29) of fronted peltae with prominent cusps rising up between them. Here the cusps on the peltae and the interstitial cusps on the opposed axis form a cross accent that could be picked up by someone aware of the unambiguous crosses on the floor (Figs 7.72, 7.76). Once the eye is attuned to recognizing and expecting crosses on the floor, it can discern cruciform accents in designs in which a cross is not the predominant visual element. Beyond this point there was at least one further overt cross, and probably two. The first is a square of which very little is preserved, two rows (row 31) beyond the peltae, which was composed of polychrome triangles traversed by strongly defined cruciform axes (Figs 7.72, 7.77, 7.78). The second, a further two rows on (row 33) in the southwestern corner of the field, is unmistakably a cross, spanning its square with bright red arms which are outlined with poised black tesserae to give a striking dentate contour (Figs 7.72, 7.77, 7.78). There are other unusually brilliant and eye-catching squares at this end of the walk, four fronted pink peltae outlined in red with a black and white chequerboard at their centre (Fig. 7.78), four parallelipipeds in dramatic perspective, red on one axis, blue on the other, emerging from a black ground (row 32; Figs 7.77, 7.79), and a striking knot of Solomon with gay ribbons in red and blue trimmed with white, also set with telling contrast against black (row 34; Figs 7.77, 7.80). In all of these and in the majority of the other designs deployed in this area the now attuned eye can quite readily discern cruciform accents. The peculiar insistence on the cross in this entrance walk of the peristyle indicates Christian reference. The apotropaic function of the cross, the most powerful protective device in the Christian

275

arsenal, was of course a constant of everyday life in late antiquity and the early Middle Ages (Brandenburg 1969; Engemann 1975, 42–8; Flint 1991, 173–8; Frankfurter 2004, 107–9; Kitzinger 1970; Maguire, Maguire and DuncanFlowers 1989, 18–22; Stevens 1977; Trilling 1995, 73). It seems likely that the bright crosses together with the more covert cruciform devices assembled towards the southern end of the west portico were designed to work alongside the empowered squares in the other areas of the mosaic to form an extensive configuration of emblems with apotropaic as well as ornamental value, designed to stake out and protect the entrance to this principal internal court, and symbolically ensure that no malign influence could cross the main western threshold into the peristyle and penetrate to the heart of the house. A similar strategy, in a different context, is be found in the final early 5th-century phase of the house, in which motifs of arguably apotropaic purpose were incorporated into the design of the stone grilles from the four main windows in the façade of the triconch dining hall, which formed part of a last major restructuring of the complex (Fig. 2.58). The overall design consists of a pierced screen of overlapping tiles/scales/leaves, incorporating at the top, in two cases, an upright cross-Chi Rho monogram, and in the other two windows, a six-pointed star. These could be interpreted as merely alluding to the Christian faith of the owner of the house; but the Chi Rho has an extensive history of use as an apotropaic symbol in late antiquity (Brandenburg 1969; Kitzinger 1970; Maguire, Maguire and Duncan-Flowers 1989, 21), and the eight- or six-pointed star is a device widely deployed on late Roman amuletic gemstones (for examples, see Michel 2001: passim. On the star as a talisman and symbol of health, see Walter 2000, 421, citing Perdrizet 1922, 35–8, and Vikan 1984, 78). On an active symbolic level, these emblematic stone screens would have functioned to protect the hall and the owner’s guests from evil spirits and demons that might slip through the open grilles of the windows with a breath of air or a shaft of light. However, there is another more generic level on which this pavement at the entrance to the house would have been designed to function. Alfred Gell and John Onians among others have drawn attention to the ways in which brilliant, complex and above all striking visual configurations, across time and cultures, have been used to surprise and psychologically gain advantage over – in Gell’s terms, to dazzle and to cast a spell over – the beholder, whether trading partner, opponent in war or simply domestic guest (Gell 1992; Gell 1998, 73–95; Onians 1994). For Gell, these mechanisms can be the highly ornate prow-board at the bow of a canoe in a Trobriand Island Kula trading flotilla or a complex ever-shifting design on an oriental carpet; for Onians, they are the Assyrian Assurnasirpal’s display of booty from his conquests in his palace at Nimrud, the monstrous mythical animals carved from the living stone for the wonderment of guests in Vincenzo Orsini’s garden of astonishment of the mid 16th century at Bomarzo, in the

276

John Mitchell

lower Tiber valley north of Rome, and the Wunderkammers of the early modern European elite. The mosaic floors of Roman houses can be seen as operating in similar fashion. To be effective, as Gell has argued, a visual display must be brilliantly polychrome so as to attract attention, complex so as to engage the eye and mind of the beholder, and ultimately resistant to resolution and mastery. This is precisely the effect achieved by the extraordinarily complex, varied and changing configurations on the pavement of the entrance portico of the peristyle in the house at Butrint. The differential brilliance of the designs seizes the eye, drawing the attention forward and then to the right, in the direction of the reception rooms in the southern water-range of the house. Thereon the beholder is engaged by the curious shapes of the individual motifs, their clusters and individual accents and the possible significations suggested almost subliminally by deeply embedded cultural reference. And finally he is caught, tied up, by an assemblage of imagery the precise sense of which cannot quite be comprehended, by individual motifs like the cruciform designs that morph between shapes, and by the very forms of the overall framework, the ambiguous visual play of elements that shuttle between one formation and another and fall in and out of planarity and apparent perspectival depth. The observer is dazzled, intrigued and confused, the guest is held, unable to comprehend fully and resolve what he sees, and so is constrained to break off in mid-gaze, still in a state of awe at the display and minded to return to settle the matter at a future convenience. In Alfred Gell’s reading, the mosaic is an enchanting and subtle social trap (Gell 1996). At the northern end of this entrance portico, the long complex endlessly variegated panel, at which we have been looking, is abutted by another panel of very different design (Figs 7.47, 7.82, 7.83, 7.85, 7.86). This is aligned west–east and once ran down along the northern walk of the peristyle. The scheme here is spare and uniform; nevertheless, similar principles underlie its conception. The basic pattern consists of an orthogonal design of adjacent octagons, forming poised squares, with swastikas in the octagons and pink discs in the squares. However, this basic structure is rendered intermittently irretrievable to the mind’s eye by the linear joining of the arms of the swastikas, which results in a dominant overlay of tangent interlocked rectangles with projecting ‘ears’ at each end. The observing eye switches between one configuration and the other, unable to hold them both simultaneously in view. As in the main panel in this walk, one is caught and bewitched by the mosaic; but the scheme here has nothing of the variety and colour of the floor by the main entrance to the house. There a bewildering panoply of bright imagery met the entering guest, while the pavement leading into the north range offered a far more subdued almost monochrome

carpet, which nonetheless could play equally bewitching if quieter tricks to divert the fancies of the occupants of the house as they moved about their quarters. The mosaic floors of the Roman domus, like its painted decoration, were designed not only to beautify its interior spaces with appropriate embellishment. To a greater or lesser degree, the rationale behind their imagery and their deployment was concerned with the effective visual articulation of the social structure and the functions of the house. In the triconch domus at Butrint, the four spaces with surviving mosaic pavement, at the time of their execution, would have functioned coherently within the complexion of the house at the various stages of its evolution. In its 3rd- and 4th-century phases, the mosaics in the old east vestibule and the long gallery (Room 18) would have impressed visitors with the elegant clarity of their design and the immaculate technical proficiency of their laying. The theatrical masks from New Comedy in the vestibule, with their reference to the pleasures of diverting entertainment and undertones of traditional cultural formation and pursuit, would have resonated with the single mask of a young female lead protagonist in the apse of the phase 2b reception room (Room 24). Subsequently, in the radically restructured phase 3 house, of around AD 400, a new visual structure was deployed to engage the visitor, in the west vestibule (Room 25) and in the entrance portico of the peristyle (Room 16). The linear clarity of the earlier mosaics gave way to a new colouristic aesthetic, in which a rich polychromy delighted and beguiled the eye of the beholder at the expense of graphic precision of design. Here the whirling peltae in the west vestibule set up an aura of movement and expectation, and then the complex assemblage of imagery in the entrance walk of the peristyle welcomed the guest with a promise of transcendent Christian paradisiacal peace and delight. At the same time, passage to the private apartments in the northern range was deterred by the eye and its drab accompanying symbols, while advance was encouraged southwards along the walk, where the brilliantly varied imagery and ornament and the complex visual ambiguity of the patterns promised radiant hospitality beyond the norms of expectation, a theatre of demonstrative teasing welcome, to which the visitor would be tempted time and again to return.1

Note 1

My thanks to Jacques Neguer for instruction on the technical aspects of the pavements and discussion of their make-up, to Elda Omari for acute observations and critical interventions on site, and for assistance of various kinds to Will Bowden, Jane Chick, Amanda Claridge, Katherine Dunbabin, Oliver Gilkes, Richard Green, Simon Greenslade, Richard Hodges, Sarah Leppard and Alessandro Sebastiani.

8  The domus and Triconch Palace as aristocratic residences William Bowden

Investigating the late antique house Recent decades have seen important developments in the study of housing in the early and later Roman world. Roman houses are now commonly interpreted as multifunctional spaces that frame finely nuanced interactions between members of different social groups. This view of the Roman house as a dynamic and highly flexible element of social relations in the Roman world now forms the broad frame of reference for the study of housing from the late Republic until late antiquity. The language of status within the Roman house drew heavily on the architecture and decoration used within public building. Using carefully constructed vistas and often subtle architectural and decorative devices, a hierarchy of space was constructed within aristocratic residences that functioned to reinforce the status relationship between the owner and his visitors. The uses of different spaces, however, and the values placed on them, were never fixed and immutable but changed according to the relationship between the users and their roles within social and domestic life. It is clear from contemporary texts that the language of private housing contained a myriad of social pitfalls for the unwary, and an awareness of its subtleties was an essential aspect of social mobility. The most oft-cited example of this occurs in Petronius’s Satyricon (32–49), in which the author lampoons the wealthy freedman Trimalchio who, despite the splendour of his house, has failed to master the etiquette of elite display. Equally important was the fact that luxury and grandeur in the context of private residences were not positives in themselves but could only be justified if they were in proportion to one’s role in urban public life and one’s role as a landowner and producer (Purcell 1995; Wallace-Hadrill 1994, 4–6). Consequently it was advantageous if these latter aspects could be alluded to in the architecture and decoration of the house itself. The archaeological aspect of this discussion, as its proponents freely admit, is based almost entirely around the evidence from the Vesuvian towns of Pompeii and

Herculaneum, where the degree of preservation allows us to formulate interpretations in a way that is denied to us elsewhere in relation to the early Roman period. However, even here we are essentially limited in that we only have one side of the multi-faceted story noted above. Without knowing more about the owner’s life outside the house it is difficult to ‘read’ the house in the way that the owner’s contemporaries would have been able to do. Similarly (with the notable exception of some of the Herculaneum houses), we lack any understanding of the furniture and fitments that would have played a key role in defining the wealth and status of the owner. We also lack the human furnishings of the house: the slaves who would be ever present and the clients and visitors themselves whose numbers and status would be directly indicative of the owner’s personal standing. Our interpretations of Roman houses are also drawn in part from the descriptions of Vitruvius and others who have provided us with a vocabulary with which to label the spaces that we encounter in the Campanian towns and in excavations elsewhere. These labels (tablinum, triclinium, cubiculum, etc.), while convenient, can be restrictive in that they can unintentionally limit our understanding of the multivariate functions of a given space (although most scholars fully accept this multiplicity of function). The labels become fixed through repetition, and what starts as a convenient though tentative means of referring to a particular area in a building eventually defines and restricts the ways in which we interpret it. The social mores relating to housing in the late Republic and early Empire were, if anything, exacerbated in the late Empire. This has also been the subject of intense scholarly interest. Simon Ellis and others have shown how means of status differentiation in the aristocratic houses of the earlier Empire became progressively more marked from the 3rd century onwards when we see the rise of what Ellis terms the ‘grand house’ (Ellis 1988; 1991; 1997a; 2000; 2007).1 As members of the upper classes placed less emphasis on their roles as magistrates and members of the curia, and

278

William Bowden

more on their roles as patrons, this was reflected in the increasingly elaborate design of private residences that supplanted the forum as the focus of elite display. The result of this process was that Roman towns became dominated by luxurious residences that could often take up entire city insulae. The buildings were focused around huge reception rooms, which utilised complex architectural forms derived in part from those used within imperial palaces. These rooms could include multiple apses and were often elaborately decorated with each area of the house functioning to highlight the status of the owner in relation to his visitors. Numerous examples of these houses have been excavated, notably in Italy and Sicily, Gaul, Spain, North Africa, the Balkans, Greece, and Asia Minor (Baldini-Lippolis 2001; Sodini 1995; Sodini 1997; Uytterhoeven 2007b). Not all late Roman houses were of this palatial form but even smaller houses reflected these architectural developments, leading some scholars (e.g. Ellis 2006) to postulate the idea of a ‘middle class’ in late antiquity, traceable in part through their housing. These middle class houses demonstrate aspirational features such as apsed triclinia and porticoes. Of course the ‘trickle-down’ of architectural features from the houses of the aristocracy was by no means a new phenomenon. Roman authors including Cicero and Tacitus comment on the competitive nature of house-building and the demands that this placed on the resources of the wealthy who were required to distinguish themselves from their peers and from those of lower status through the construction of ever-more opulent residences (Wallace-Hadrill 1994, 4–6). Despite the often sophisticated interpretations that have been placed on their remains, very few late antique aristocratic houses have been investigated in a way that most modern archaeologists would find satisfactory. Most of the frequently cited examples that appear in publications were excavated in earlier periods by archaeologists with different priorities, who sought material to illustrate a textbased narrative history of which the parameters were well established. The most prominent features of these houses (such as mosaics) were also suitable subjects for analysis within the art-historical framework that shaped the classical archaeological tradition in which the excavators operated. Bemoaning this state of affairs is of course well-trodden ground for late antique archaeologists, and indeed late antique archaeology as a sub-discipline has developed in part through a critique of these earlier techniques. While it is widely recognised that the data from many earlier excavations is unsatisfactory, until recently relatively little was being produced to replace it. The reasons for this are primarily financial and it is undeniable that excavation carried out to a respectable standard is expensive when attempted on a significant scale, although this in itself is not a reason for removing it from the repertoire of archaeological techniques. In fact, modern stratigraphic excavation is generally much better suited to finding secondary occupation – the levels of rubbish

dumping, minor industry and post-built structures – than it is at elucidating how a house was used in its primary phases. This is for the simple reason that most people swept their floors after their dinner parties and dumped their rubbish elsewhere. Portable furnishings and evidence of primary function do not survive in situ unless the house suffered a sudden catastrophe such as that which befell the Vesuvian cities or the shops at Sardis. Bereft of these furnishings it is often difficult to define the function of spaces or to differentiate between them in terms of the relative importance accorded to them by the owner(s) of the house. As a consequence of the above, although the academic discourse on the late Roman house is relatively well developed (most recently in the papers in Lavan, Özgenel and Sarantis 2007), it is worth considering that much of this discourse is based upon published plans derived from poor quality excavations. These excavations are often only published in a very limited way, with the writers concentrating primarily on art-historical detail, particularly mosaics. In this way, complex multi-phase buildings are reduced to a single plan that often appears much more coherent on paper than it does on the ground. Hundreds of years of occupation in which houses were repeatedly altered according to the changing needs and circumstances of their owners are compressed into a single image. Buildings become static representations of their excavators’ priorities, with excavations generally ceasing at points considered by the archaeologist to be of particular importance (for example when a particularly fine mosaic, stone pavement or complex architectural form is encountered). During the frequent reassessments of these buildings (reassessment of earlier data being the principal default approach of late antique archaeologists as noted above), these features can be interpreted in a way that ascribes an overall coherence to them that they may never have possessed in the past. Roles are ascribed to porticoes, audience halls, and triclinia based on these constructs, while elaborate routes of circulation are plotted between them (e.g. Ghedini and Bullo 2007). While not denigrating the value of this type of exercise, it is clear that overall models based on this type of data are open to question. This is perhaps the area where the Triconch Palace excavations can contribute to the debate on elite housing in late antiquity. While the major part of the archaeological sequence revealed belongs to the period after the Triconch had been abandoned as a high status residence, the sequence relating to the Triconch Palace and its predecessors suggests that one should be wary of interpreting these buildings on the basis of plans alone.

The Triconch Palace: architectural grandeur and architectural failure The plan of the Triconch Palace as revealed by the excavations contains a number of ‘set piece’ elements (the corridor and apsidal reception room of phase 2, the western entrance and

8  The domus and Triconch Palace as aristocratic residences peristyle of phase 3, and the triconch triclinium and enlarged peristyle of phase 4). The excavations showed that each of these features formed coherent architectural devices in their own right, but apparently failed to work with the set-piece structures that preceded them. Instead, each phase effectively negated the functions of what were previously the most opulent areas of the complex.

The phase 2 domus In its earliest phase, where we can discern clear elements of the plan (the phase 2 building of the 3rd and 4th century), the public focus of the building lay to the south. The visitor to the building entered through a door in the southeast corner of the complex and was immediately in a small but highly elaborate space, faced by the theatrical masks of the intricate mosaic on the floor, and the painted architectural vistas on the walls. From there they could proceed into the corridor or gallery, with its carpet of mosaic stretching into the distance and subsequently into the reception room beyond. (Fig. 8.1). At the same time access to the remainder of the house was restricted, although our understanding of the form of the rest of the house in this period is very limited. The use of a corridor rather than a peristyle to link the principal entrance of the house to the main reception room is not unusual. Although we do not know whether a peristyle and further reception rooms existed in the

279

Triconch Palace before the later structure of phase 3 (c. AD 400), it is clear that this arrangement, in which the audience hall was linked directly to the street by a corridor, was a deliberate device, which is readily paralleled elsewhere. The use of corridors in this way can be seen in particular in Asia Minor (Özgenel 2007) (Fig 8.2). At Ephesus, in the Villa over the Theatre (dating to the 5th century), a corridor 7 m wide and more than 17 m long led to a vestibule and from there into an apsidal audience hall, some 21 m in length. It is clear that the audience hall and the access to it, as in the case of the Triconch Palace, were kept separate from the remainder of the house and its peristyle with only limited access between them (Ellis 1988; Ellis 1997b). At Sardis in the late Roman town house, the street entrance opens into a small room furnished with benches, which allows access to a well-paved corridor, which in turn leads almost directly from the street to a square room, interpreted as an audience chamber. Access to the rest of house from this corridor is very restricted (Özgenel 2007, 267–68; Rautman 1995). At the late Roman villa at Halicarnassus, meanwhile, a similar arrangement is created by three axially aligned rooms, which lead to an apsidal audience chamber. Once again the route to the audience chamber does not require the visitor to enter the main body of the house (Ellis 1997b; Poulsen 1997). The dating of all these buildings is uncertain, although in all cases the corridor and audience chamber arrangement is thought to have been in place by the 5th century.

Figure 8.1. The long gallery (Room 18) looking towards the apsidal reception room (24) (JBB)

William Bowden

280

Triconch Triconch Palace, Palace, Butrint Butrint

Villa Villa over over thethe theatre, theatre, Ephesus Ephesus

0

0

20m 20m

0

0

20m 20m

Late Late Roman Roman town town house, house, Sardis Sardis

Roman Roman villa, villa, Halicarnassus Halicarnassus

0

0

20m 20m

0

0

20m 20m

Figure 8.2. Comparative plans of the phase 2 (3rd to 4th century) domus, the Villa over the theatre (Ephesus), the late Roman villa at Halicarnassus, and the late Roman town house at Sardis (after plans reproduced in Baldini Lippolis 2001) The arrangement whereby the main audience chamber is relatively accessible from the street, without visitors needing to enter the remainder of the house, is quite common (although by no means universal), oft-cited examples being that of the Palace of the Dux at Apollonia and the Bishop’s Palace at Aphrodisias (Ellis 1991; Polci 2003) (Fig. 8.3).2 The methods by which this was achieved varied, depending in part on the space and resources available to the owner. In the most opulent cases a separate peristyle could be employed, most famously at Piazza Armerina, although the associated rooms are usually interpreted as triclinia rather than simply as audience halls. In the case of the Triconch Palace, the long gallery fulfilled the function of leading the visitor to the audience hall, while at the same time allowing a view over the Vivari Channel. There may have been a small garden between the gallery and the channel, or alternatively the channel may have almost lapped against the southern wall of the gallery.

This is hinted at by the way that the southern wall of the gallery and the south wall of reception Room 24 reflect the southern limit of the earliest phase of the Merchant’s House (see Chapter 5). The later phases of the Merchant’s House to the south (Rooms 38, 39, 43 and 47) (and the later domus bath-house) may have been built on land reclaimed at the edge of the channel. This could suggest that the edge of the channel was originally much closer to the domus. Views over water had long been part of the architectural vocabulary of elite private building in Epirus (Bowden 2003a, 76–7; Bowden and Përzhita 2004a, 422) and indeed this was a common motif of status throughout the Roman period (Bergmann 1991). Painted depictions of villae maritimae show porticoes extending out into the sea and the creation of a direct relationship between building and water was clearly desirable, although whether this was actually the case at Butrint must remain a matter of supposition. The reception room itself clearly reflects the increasing

8  The domus and Triconch Palace as aristocratic residences

0 0

0

0 0

10m 10m 10m10m 10m

Bishop's Palace, Aphrodisias Bishop's Bishop's Palace, Palace, Aphrodisias Aphrodisias Bishop's Palace, Aphrodisias Bishop's Palace, Aphrodisias

281

0 0

0

0 0

10m 10m 10m10m 10m

Palace of the Dux, Apollonia Palace ofPalace thethe Dux, of the Apollonia Dux, Apollonia Palace Dux, Apollonia Palace of of the Dux, Apollonia

Figure 8.3. The Bishop’s Palace at Aphrodisias and the Palace of the Dux at Apollonia (after plans reproduced in Baldini Lippolis 2001)

value being placed upon this type of audience hall and indeed the changing nature of public rooms in the late Empire. In its earliest phases, the one known reception room at the Triconch Palace was a square room with a fountain that would have been clearly visible to those approaching it along the corridor. This type of square or rectangular reception room is a common feature of houses of the early to mid Empire, when the principal reception room of an aristocratic residence was the triclinium, which served as both a formal dining room and a reception room (Polci 2003). However, during the late Empire as patronclient relationships assumed greater hierarchical definition, separate rooms and new architectural forms evolved to accommodate the range of social occasions that structured the relationships between the various strata of Roman society. A particularly noteworthy feature at Butrint is the significant alteration that took place with the addition of the apse, perhaps in the 4th century. The conversion of rectangular reception rooms to apsidal spaces is not uncommon in late antiquity. Examples include the House of the Lycian Acropolis at Xanthos, which had an apse added at an unknown date between the 5th and 7th centuries (Manière-Lévêque 2007), and the Late Antique Residence

at Perge, which had an apse added to its reception room in the 5th or 6th centuries (Özgenel 2007, 254). Similarly, the rectangular reception room at the House of the Phoenix at Daphne (Antioch) also had an apse added in the 5th century (Ellis 2000, 64). This is an illustration of the increasing importance placed on the use of apsidal spaces in reception rooms. Apses were a feature of Roman domestic building from the reign of Nero onwards. In the mid and later Empire, however, apses were increasingly used in audience chambers and triclinia. The increased use of apsidal spaces in aristocratic houses is thought in part to mirror the growing importance of patron-client relationships within Roman society, forming a space in which the owner of the house or patron could be seated to meet his clients (a feature that seems to originate in the throne room of Domitian’s palace in Rome) (Ellis 1991). The increasing use of apses in triclinia also reflects changes in dining habits as the semi-circular dining couch (stibadium) seemingly replaced the traditional arrangement of three rectangular couches from the 3rd century onwards (Polci 2003), although at 4.7 m wide the apse in the Butrint reception room seems small for this purpose. The addition of the apse to the reception room at Butrint needs to be interpreted within this context (Fig. 8.4). In terms

282

William Bowden

Figure 8.4. Reception room (24) in which an apse was added to an earlier rectangular reception room

of its architectural form, the room could have functioned as either audience hall or triclinium, although its limited size suggests that it is unlikely that it could have accommodated the stibadium and sigma arrangement discussed below. In the early stages of the excavation it was interpreted as an audience hall that functioned in tandem with the threeapsed triconch triclinium following the model proposed by Ellis, in which the two different rooms served either for different types of social occasion or separate stages of the same social occasion (Ellis 1991). This arrangement of two or more reception rooms is well known from other late antique houses, with frequently cited examples at the Palace of Theoderic in Ravenna, the House of Bacchus in Djemila, the villa at Mediana in Niš, the House of the Hunt at Bulla Regia, and the House of the Protomes at Thurburbo Maius (see for example Ellis 1991; Ghedini and Bullo 2007; Polci 2003) (Fig. 8.5). The audience hall/triconch triclinium combination at Butrint seemed to sit comfortably within this model and was interpreted in this way in interim statements on the complex (e.g. Gilkes and Lako 2004, 166) and in secondary literature (Polci 2003, 88). Interpreted in this way, the audience hall could be used for the salutatio or ‘morning greetings’ and the triclinium used for grand dining. Alternatively, the audience hall would be used for pre-dinner drinks and entertainment. However, when the building was fully excavated and key questions of its phasing were resolved it emerged that it was unlikely that the two rooms could ever have functioned

together in this way. Although access via the eastern entrance to the gallery may have been maintained when the triconch and its twin north and south vestibules were built, the route between the eastern entrance and the city was blocked by the new buildings (although the eastern entrance could still be accessed by boat in the same way as the double-apsed vestibule of the triconch). More significantly, however, the grandeur of the gallery had been lost by this time. The mosaic was truncated by a large and roughly covered drain and probably covered by a rough mortar surface. It served mainly as an access route to the bath-house. In summary, therefore, the gallery and reception room apparently served as the principal public wing of the building in the 4th century. The gallery allowed easy access to the reception room that served as an audience chamber, this function being augmented by the subsequent addition of an apse. This embellishment reflects the owner’s desire to greet his clients in an environment suitable for the purpose. It is possible that the reception room also served as a triclinium, as we do not know whether there was a separate triclinium elsewhere in the house. However, the layout at Butrint seems to closely parallel the examples noted above in which a public room is located close to the street and could be accessed without visitors being allowed into the rest of the house. This does suggest that a separate triclinium may have existed elsewhere in the building, although as noted above the extant triconch cannot have fulfilled this role. This delineation of public and private space emphasised

8  The domus and Triconch Palace as aristocratic residences

283

0

0

20m

20m

House of House the Hunt, of theBulla Hunt,Regia Bulla Regia 0

0

20m

20m

House of House Bacchus, of Bacchus, DjemilaDjemila 0

0

20m

20m

Palace of Palace Theoderic, of Theoderic, Ravenna Ravenna

0

Mediana Mediana villa, Nis villa, Nis

0

0

20m

20m

0

20m

20m

House of House the Protomes, of the Protomes, Thurburbo Thurburbo Maius Maius

Figure 8.5. Houses with two reception rooms. Palace of Theoderic (Ravenna), House of Bacchus (Djemila), Mediana villa (Niš), House of the Hunt (Bulla Regia), House of the Protomes (Thurburbo Maius) (after plans reproduced in Baldini Lippolis 2001)

by the long gallery and apsidal reception room is continued within the rooms of the southern wing, by a quite subtle arrangement of doorways (Fig. 8.6). The only access to the rest of the house from the corridor was via the central room of the southern range (Room 21), where central doorways in the north and south walls allowed access between the gallery and the courtyard beyond. A door in the east wall of Room 21 allowed access to Room 20 (and to Room 19 beyond). Rooms 19 and 20 could only be entered via Room 21, as there was no direct access between these rooms and

the courtyard space beyond. By contrast the two western rooms of the southern wing (22 and 23) do not appear to have been accessible from Room 21, but were instead only linked to the central courtyard and the west wing. This suggests that what ostensibly appears to be a line of similar rooms may have served very different functions, with Rooms 19 and 20 being much more accessible to those using the long gallery and reception room and Rooms 22 and 23 having a more private role.

William Bowden

284

Room 27

Room 25

Room 26

Room 23

Room 22

Room 21

Room 20 Room 19

Room 24 Room 18

20m

0 Figure 8.6. Access routes and doorways in the phase 2 domus

The peristyle domus (phase 3) (Fig. 8.7) Around AD 400 the orientation of the house was radically altered by the addition of the western entrance (Room 25) and peristyle, with a large drain from the peristyle cutting across the mosaic of the corridor. It is possible that a previous entrance existed on the western side of the building, but the new entrance vestibule with its colonnaded screens was intended as the principal public entrance into the building, giving access to the new peristyle with its mosaic paved porticoes. As noted in Chapter 3, it may have been the destruction caused by a major earthquake that persuaded the owner of the building to radically rethink the whole design of the house. The mosaic of the western entrance was supported on a raft of rubble that separated it from the opus signinum floor of the earlier building beneath it by 0.50 m. The pavements of the peristyle were similarly were 0.44–0.50 m higher than those of the earlier long gallery (Fig. 8.8).

The deliberate raising of the floor in the western entrance must have been the result of necessity, as it would have been much easier to simply build from the level surface of the earlier floors, rather than have both the mosaic and the stylobates for the tribelons supported on an uneven platform of rubble. The new entrance is quite unusual in the degree of elaboration employed within it (Fig. 8.9). When it was first partially revealed in the excavations it was interpreted as the central part of a triclinium or reception room, with the tribelons giving access to three chambers that mirrored the lobes of the later triconch triclinium. However, the mosaic inscription immediately in front of the eastern tribelon was clearly intended to be read as one approached the peristyle from the west, announcing the name of the house owner to visitors. There would be little point in such an inscription that could only be read by those who were leaving a reception room. The use of the four column screens, or tribelons, in

8  The domus and Triconch Palace as aristocratic residences

285

Figure 8.7. (Plate 67) The phase 3 (c. AD 400) domus (Studio Inklink)

an entrance vestibule is quite a particular arrangement that finds relatively few parallels. However, an identical arrangement can be found in the Maison au triclinos at Apamea (Baldini Lippolis 2001, 137–139; Balty 1969) (Fig. 8.10). Here, following a reconstruction dated to after the third quarter of the 5th century, the street entrance led into a rectangular space from which tribelons gave access to rooms to the left and right, and onto the peristyle that lay immediately opposite the entrance. A similar vestibule was found in the North Temenos house at Aphrodisias (Smith and Ratté 1998), although in this instance the vestibule separates the street from an apsidal audience hall rather than a peristyle. Interestingly, both houses have been associated with individuals of high rank (although sometimes random associations with bishops or governors are common with many of the great peristyle houses of late antiquity). Given its colossal size this explanation is perhaps more convincing in the case of the Apamea house, associated by Balty with the governor of Syria Secunda. The apparent senatorial

rank of the owner of the Butrint house could also suggest a tentative association between this rather unusual type of entrance and the status of the house owner. The principal purpose of this elaborate entrance was to create a grand approach to the peristyle and also to create a view into the peristyle for those passing along the street to the west. No obvious sign was found of a further antechamber to the west of the vestibule, suggesting that the western tribelon could have opened directly onto the street. This would mean that the western entrance was an unusually open and very public space, suggesting again that the house owner had, or aspired to, a particular social status that required this. The interior of the western entrance is difficult to reconstruct. The column bases of the four tribelons were remarkably narrow, measuring only 0.18–0.20 m in diameter, suggesting that the accompanying columns were unlikely to be more than 2 m in height, with combined bases and capitals adding a further 0.30–0.40 m. No examples of

William Bowden

286

0

5m

0

5m

Existing floors & walls Level of peristyle pavement Existing floors & walls Level of peristyle pavement

0

20m

0

20m

Figure 8.8. Schematic cross section showing rise in relative levels between the peristyle and long gallery, associated with the rebuilt domus of phase 3 (c. AD 400) the capitals were found and it is unknown if they were of a crutch type (supporting arches) or whether they supported a flat architrave. The upper levels of the room must remain a source of conjecture, although it is possible that the tribelons supported high upper walls raising the interior height of the room in a fashion reminiscent of the classic atrium house (although in this case with a roof over the

central space).3 The original function of the two rooms that lay beyond the right and left tribelons must similarly be a matter of conjecture, although it is likely that at least one would have been occupied by a slave or freedman who would have controlled access to the house. These lateral rooms also respectively gave access to the north and south wings of the house.

inscription inscription inscription

0 5m 0 5m 0 5m Figure 8.9. The western entrance of phase 3 (c. AD 400), showing location of mosaic inscription

0 0

Maison au triclinos, Apamea aison au triclinos, Apamea

20m 20m

0 0

Maison au triclinos, Apamea Maison au triclinos, Apamea

0 0

North Temenos house, Aphrodisias North Temenos house, Aphrodisias

10m 10m

20m 20m

Figure 8.10. Comparative plans of Maison au triclinos (Apamea) and North Temenos house at Aphrodisias showing entrance vestibules reminiscent of that of the phase 3 (c. AD 400) domus (after plans reproduced in Baldini Lippolis 2001) 0 0

North Temenos house, Aphrodisias North Temenos house, Aphrodisias

10m 10m

288

William Bowden

Figure 8.11. The peristyle and porticoes of the phase 3 (c. AD 400) domus, looking west

The main function of the vestibule was to create a suitably imposing entrance to the peristyle, and the presence of the inscription immediately in front of the west portico of the peristyle leaves us in little doubt that this was the direction in which most visitors were expected to go. The inscription, although very fragmentary, appears to announce the name of the owner and the fact that he was a lamprotatos (the equivalent of the Latin clarissimus), i.e. of senatorial rank, although what this title actually signified in the context of late 4th- or early 5th-century Butrint is open to question. The peristyle itself was relatively modest in scale, its dimesions presumably restricted by the size of the actual property and the size of the courtyard that preceded it (Fig. 8.11). It measured slightly over 15 m across from the outer walls of the porticoes, and is comparable in scale to numerous other examples of peristyles from mediumsized houses across the empire, for example House A on the Athenian Agora (Frantz 1988, 38–39), and the Maison d’Amphitrite at Djemila (Sodini 1995, 178). While its porticoes were paved with quite elaborate geometric mosaics (see Chapter 7), the courtyard itself was paved with simple and irregularly cut limestone slabs. The surviving traces of two columns on the eastern side show that the columns were constructed from masonry (which would have been plastered) rather than being monolithic stone shafts as were utilised in the much larger triconch peristyle.

Surface water was drained from the new courtyard via the large drain (described in detail in Chapter 3). This drain was probably part of the laying out of the peristyle, evidenced by the fact that it ran beneath the wall that blocked the doorway between Room 21 and the courtyard. This door blocking was then abutted by the plaster floor make-up for the portico mosaic. It is this drain and the fact that it clearly truncates the mosaic in the long gallery (as well as a later mortar surface) that is one of the clearest indications that the gallery was redundant as a public space by the time the peristyle was built. It was probably also during this phase that the small bath-house adjacent to the Vivari Channel was constructed. From the 4th century onwards, small private bath-houses became increasingly common features of urban domus. This may have been partially because the larger public bath-houses were falling into disrepair or their water supply could no longer be guaranteed, but it was also due to an increased ambivalence towards public bathing that was inspired by Christian disapproval of the practice. While during the early Empire, a visit to the baths had been an important part of the daily social ritual, during late antiquity Christian leaders denounced the baths as places of immorality infested by demons (Baldini Lippolis 2001, 64–65; Yegul 1992, 314–17). Evidence suggests that the owner of the Butrint domus was Christian by the time that the phase 3 peristyle building was erected (see below and

8  The domus and Triconch Palace as aristocratic residences

289

Bath Bath Bath 0

Bath

0

10m

10m 0 House of the Hunt, Bulla10mRegia

0

10m

House to north-east of Octagon, Philippi House to north-east of Octagon, Philippi

Chapter 7 and Bowden 2008) and certainly the limited scale of the bath-house would be appropriate in a Christian context. Equally, the only access to the bath-house was via the corridor and apsidal reception room and it is unlikely that it would have been built in this location if these were still the main public areas of the building.4 The size of the bath-house clearly suggests that it was intended for use by one or two individuals at a time (together with one or more attendants), and can therefore be interpreted within the Christian context noted above. It is also intriguing that it was placed so close to the Vivari Channel and it is possible that the bath-house was supplied with water directly from the channel (and drained directly into it). The presence of the well/cistern in the middle of the peristyle shows that the house was not reliant on piped water (although this does not in itself indicate that such a supply was not present), while no trace of any water supply system in the form or lead or ceramic piping was found during the excavations. The distance from the main house can be explained by the need to minimise the fire risk associated with baths. The 6th-century architectural treatise of Julian of Ascalon (a compilation of laws relating to building promulgated in Palestine) required the furnaces and chimneys of private bath-houses to be located at least

House of the Hunt, Bulla Regia

Figure 8.12. Comparative plans of bath-houses from House of the Hunt (Bulla Regia) and the House to north-east of Octagon at Philippi (after plans reproduced in Baldini Lippolis 2001)

20 cubits (9.36 m) away if the house was located to the north or east (as in the case of the Butrint domus), had more than one storey, or had windows facing the bath (Baldini Lippolis 2007, 228; Hakim 2001, 11). While one cannot extrapolate laws from Palestine to Butrint, it is likely that similar legislation existed or that the issue was at least considered. Similarly sized bath-houses are known from many late antique domus, examples being found at the House of the Hunt at Bulla Regia (Baldini Lippolis 2001, 165–6), and at Philippi in the house to the northeast of the Octagon (4th-century phase) (Baldini-Lippolis 2001, 204–5) (Fig. 8.12). Elsewhere in Butrint, the slightly larger bath-house identified to the east of the Baptistery (which formed part of the Baptistery water supply) may well be part of a similar complex relating to a further substantial domus in the town (Bowden and Përzhita 2004b, 183–5), while on the Vrina Plain the major domus currently under excavation also contained a substantial bath complex (Greenslade forthcoming). It is possible that many further, similar bathhouses exist elsewhere that were detached from the main house as at the Butrint domus but which have not been detected by excavations that have not extended beyond the boundaries of the main house.

290

William Bowden

Figure 8.13. (Plate 67) The phase 4 (c. AD 420) Triconch Palace (Studio Inklink)

As noted in Chapter 5, the construction of the bath-house presumably blocked any access to the channel-side that had previously existed between the domus and the Merchant’s House, while the narrow gap between the apsidal reception room (24) and the Merchant’s House was also blocked by a wall. The construction of the bath-house may, therefore, represent an example of the owner of the domus being able to expand his property onto adjacent land (discussed further below in the context of the triconch triclinium). In sum, therefore, the new additions to the building (the western entrance, the peristyle and the bath-house) sat awkwardly with the most impressive elements of the earlier building (at least those which we are aware of). However, it is possible that these earlier parts (the gallery and reception room) were in a state of disrepair and affected by periodic flooding. This situation forced the owner (or his builder) to erect a new house on a higher level, probably creating a terrace of rubble from the earlier structure in the process.

Notwithstanding the opulent nature of the western entrance, the impression of the phase 3 domus is of a modest house, whose owner’s aspirations were restricted by both financial resources and the space available to him. This makes the colossal scale of the next phase all the more remarkable.

The phase 4 Triconch Palace Very shortly after the phase 3 peristyle and the western entrance were created, they were superseded by the colossal expansion of phase 4, which saw the peristyle tripled in size and the whole structure extended to the east with the addition of the triconch triclinium (Fig. 8.13). In addition, the location of the principal entrance was changed again with the blocking of the western entrance and the creation of new entrances to the north and south, which gave access to the eastern portico of the expanded peristyle. The building was also extended to the north with the construction

8  The domus and Triconch Palace as aristocratic residences

291

Phase 3 (c. AD 400)

Phase 4 (c. AD 420)

0

20m

Figure 8.14. Comparative plans of phases 3 and 4

of a whole new northern wing, made necessary by the construction of the north portico of the enlarged peristyle on the footprint of the earlier north wing. The time period that elapsed between the construction of the smaller peristyle and western entrance and their replacement is difficult to establish but was probably not more than 20 years and may have been considerably less. As noted above, it appears that the smaller peristyle was probably built around AD 400, while the larger peristyle, triconch and associated structures date to AD 400–420. The wholesale alteration of a building that had been totally

rebuilt only a few years earlier might seem odd, but, in fact, the rapid succession of building phases is characteristic of all the Roman houses excavated at Butrint where building phases can be closely dated. This was particularly apparent at the 1st- to 2nd-century villa of Diaporit, which seems to have undergone major alteration with every generation (Bowden and Përzhita 2004a, 422). Indeed, we might suppose that a continual stream of alterations to one’s house were important aspects of status display. The most striking feature of the new building is its sheer scale (Fig. 8.14). The addition of the new eastern wing and

292

William Bowden

the expansion of the house to the north effectively doubled the size of the complex. Whereas the earlier peristyle domus was a relatively modest residence, the new elements of the modified building were of a different ilk in terms of their size, the materials employed and their architectural unity. The reasons for this must remain a source of speculation. It is possible that the owner’s circumstances improved, or that the building passed into the hands of a wealthier individual, allowing the purchase of additional land and the purchase of labour and materials for the new buildings. However, we must also consider the role that changing circumstances in Butrint may have played in the process. The extension of the building to the east clearly demonstrates that in the early 5th century, the owner of the building was able to expand his house onto the adjacent plot, beyond the eastern boundary that had formed its limit for at least two centuries. In doing so he almost certainly blocked a road or alley, and built across a major town drain that ran from the so-called Gymnasium on the eastern edge of the forum. It is this that is perhaps the most significant aspect of the eastern extension, in that it suggests that the owner felt confident enough to build across what was presumably publicly-owned land. A similar expansion may have occurred to the north although this area remains unexcavated. The co-opting of adjacent land for one’s own house seems to have become an increasingly common phenomenon in the late antique town, although much of the legislation against the practice appears to refer to the small structures that were erected within the public monumental spaces of the earlier town. It is also clear that local magistrates were able to rule on a case-by-case basis, with the result that the upper classes had more or less carte blanche to act as they liked. The principal rationale behind much of the legislation was to preserve the appearance of the city. A law of AD 362 (Cod. Iust. 8.11.3 ) stated that if a person had built in a public area, without harming the urban decorum, “he should be considered the legal owner of the building and praised by the city as a contributor to its beauty” (Baldini Lippolis 2007, 200). Examples of aristocratic buildings encroaching on streets are relatively common and are known from Grumentum (Italy), where the major domus next to the theatre expanded over a street in the late 3rd or early 4th century (Baldini Lippolis 2001, 213); Rome, where the domus of Gaudentius blocked a street following a reconstruction in the late 4th or early 5th century (Baldini Lippolis 2001, 213); Ravenna, where the Domus dei Tappeti di Pietra was built across an earlier street (Augenti 2006, 194); Savaria (Pannonia), where the reception rooms of a large late antique residence are built across the street grid (Sodini 1997, 526); and Carthage, where in a situation analogous to Butrint, at the Maison du Triconque, the eponymous dining room encroached onto the adjacent street when it was erected in the first half of the 5th century (Baldini Lippolis 2001, 170–2). Further examples are noted by Baldini Lippolis (2007, 210–12) at Ptolemais, Thurburbo Maius, and Philippi.

The triconch triclinium and its north and south entrances are significantly different from the rest of the complex in that they are all of one unified construction, and do not incorporate any sections of earlier buildings. A single earlier wall is visible beneath the north conch, but there is no other trace of earlier structures suggesting that the area to the east of the road noted above may have been unoccupied or used as a garden or similar.5 It is clear that this new eastern wing and the enlarged peristyle were conceived as a single project. The new peristyle was almost three times the size of the peristyle of the previous phase (with internal measurements of 23 × 13 m excluding the porticoes). Its size is comparable with peristyles in some of the larger domus of late antiquity, being somewhat larger than those of the Palace of the Dux at Apollonia (Ellis 1985; Goodchild 1976) and the House of the Triapsidal Hall at Ptolemais (Ellis 1985). The large peristyle at Piazza Armerina is around 25% larger than the Triconch Palace peristyle but it is clear that the latter was a feature of considerable ambition (Fig. 8.15). The southern, eastern and western stylobates of the peristyle were built using substantial faced limestone blocks, which were presumably spolia, deriving possibly from a disused public building elsewhere in the city. It is likely that the colonnades came from a similar source, although since only two of the bases were found it is difficult to form a more accurate picture of their original source. However, the fact that the owner of the domus was able to expand the building dramatically, and at the same time was able to utilise materials from monumental structures elsewhere, is suggestive of fundamental changes taking place within the city in this period. The close dating of the triconch phase (to AD 400–420) and the small peristyle building (to around AD 400) could suggest that this urban change, occurred in this relatively short period, as it seems that the additional space and building materials were not available to the builder of the phase 3 domus but were available to the builder of the Triconch Palace shortly afterwards. However, we should be wary of looking for a single period or moment of ‘decline’ at Butrint, as it is becoming increasingly clear that many areas of the town underwent major changes much earlier (during the latter part of the 3rd century). Indeed the limited nature of 4th-century material across the town (see Reynolds in prep.) is indicative of a substantial contraction of occupation at the end of the 3rd century, considerably earlier than the period of major change at the Triconch Palace. The new construction again led to a substantial reorganisation of the entrances to the complex (Fig. 8.16). The western entrance with its inscription announcing the owner’s name became redundant and was blocked by the creation of Rooms 29 and 31. This was presumably possible because the blocking of the street to the east of the building made it possible to create a grand approach to the house from the centre of the town to the north. The new northern and southern entrances also allowed the triconch triclinium to function in the same way as the apsidal reception room (24) had done in the earlier phase, in that it could be easily reached by visitors who would be granted a view across the peristyle without being able to venture further into

8  The domus and Triconch Palace as aristocratic residences

Triconch Palace, Butrint

Piazza Armerina

Palace of the Dux, Apollonia

House of the Triapsidal Hall, Ptolemais

0

50m

293

Figure 8.15. Comparative plans of phase 4 (c. AD 420) Triconch Palace, Palace of the Dux (Apollonia), House of the Triapsidal hall (Ptolemais), and Piazza Armerina (other plans after those reproduced in Baldini Lippolis 2001)

Northern entrance from town 294

Western entrance blocked by Rooms 29 and 31

Room 29

Room 31

William Bowden

Southern entrance from Vivari Channel

0

Figure 8.16. Reorganisation of entrances in phase 4

20m

8  The domus and Triconch Palace as aristocratic residences

295

0

Palace of Lausos, Constantinople

0

0

20m

20m

House of Junius Bassus, Rome

20m

Palace of Galerius, Thessalonica

0

20m

House of the Seasons, Sbeitla

Figure 8.17. Comparative plans of double apsed vestibules. Palace of Lausos (Constantinople), Palace of Galerius (Thessalonica), House of Junius Bassus (Rome), House of the Seasons (Sbeitla) (after plans reproduced in Baldini Lippolis 2001) the house. This could suggest that the triconch served as an audience chamber as well as a triclinium, or indeed suggests that the triconch can be seen as an example of what Ellis (1991) classes as “grand dining rooms”. These were dining rooms that could be easily accessed from the street and formed the backdrop for very public dining in which wealth could be demonstrated to clients and peers. Ellis argues that these rooms were often adjacent to more private dining rooms, which does not seem to have been the case at Butrint, a point to which I shall return below. The entrance to the south was a particularly impressive double-apsed structure of a type well known from late antique palaces and houses, notably the Palace of Lausos

in Constantinople (Bardill 1997), the Palace of Galerius in Thessalonica (Sodini 1997, 457–8), the House of Junius Bassus in Rome (Baldini Lippolis 2001, 269) and the House of the Seasons in Sbeitla (Sodini 1995, 181) (Fig. 8.17). It is also commonly found in mausolea and at a slightly later date in churches. As noted in Chapter 3, the form of the room and the presence of a major wooden door on its southern, outer side, indicates that it was a major entrance (rather than forming access to private garden space). Those passing through it would enter directly into the peristyle, and in many cases would presumably proceed directly to the triclinium. This entrance was reached via one or two small

296

William Bowden

Figure 8.19. Mosaic depicting the Last Supper from S. Apollinare Nuovo, showing Christ in the position of honour for a diner Figure 8.18. Stibadium and sigma (after Åkerström Hougen 1974)

courtyards to the south, which are structurally later than the apsidal vestibule but which were probably part of the same overall design. The smaller courtyard gave direct access to the double-apsed vestibule, while a second, larger courtyard lay to the east (overlooked by the large window in the southern apse of the triclinium) linked to the smaller courtyard by a door. This larger courtyard may have been a small garden, or it may have formed an outer courtyard that in turn led to the smaller courtyard and the double-apsed vestibule. In either case, it is likely that one of these courtyards led to a small jetty or quay on the Vivari Channel, enabling visitors (and the owners) to arrive directly by boat, although the relationship between the courtyards and the channel was obscured by the subsequent construction of the city wall. This close relationship with the waterfront recalls that of earlier villae maritimae in which houses often projected into the water, and it is interesting to note this in a late antique context. Although late Roman villas and town houses do occur in waterfront locations, this aspect seems to be of less importance than it was in the late 1st century BC and early imperial periods, when numerous villas were built along the coastlines of the central Mediterranean (Lafon 2001). It is possible that a road existed along the channel-side, although there is very little space available unless the later city wall is built directly above such a road.6 The focal point of the new building was the triconch triclinium, described in detail in Chapter 3. Triconch reception rooms became increasingly common in the later empire, reflecting the need for rooms suitable for hosting the elaborate dinner parties that formed an important element within the patronage networks of late antiquity (Ellis 1991; 1997a; Dunbabin 1996; Lavin 1962; Morvillez

1995; Polci 2003). The triclinia of the early empire were rectangular rooms, with dining couches placed along three walls, an arrangement that is often reflected in the mosaic pavements of these rooms where the areas for the couches are left blank, resulting in the decorative panels forming a characteristic T-shape. Dining habits changed quite significantly, however, with the adoption of a semi-circular arrangement of dining couches, or stibadia, on which the diners would lie in a radiating pattern with their heads towards the centre (Fig. 8.18). Depictions of this process invariably (as might be expected) show diners supporting themselves on their left arms, leaving their right hands free for eating and drinking. Food would be placed on a small semi-circular central table or sigma. The most important guest would be positioned at the front left (to the viewer facing the apse) as can be seen in late antique representations of the Last Supper that show Christ in this position (Fig. 8.19). The other diners would be positioned in descending order of status, with the diner who held the lowest social position seated on the far side, directly opposite the guest of honour, a situation described in detail by Sidonius Apollinaris (Ep. 1.11.10) (Polci 2000, 151). The semi-circular stibadium used in this way actually made conversation between guests much more difficult than the earlier arrangement of rectangular couches. According to Polci (2000, 151) this was a reflection of the increasing importance of entertainment within the dining ritual, with the guest of honour now allowed the best view of the activities taking place in the open space in the centre of the triclinium. This arrangement necessitated very substantial buildings. The average male diner positioned on a stibadium would require a space around 1.25–1.50 m in length (multiplied by two at the chord of the apse where two diners would be seated opposite one another). The sigma would require additional space of around 1.25 m across judging from known examples, such as those from the Maison du Cerf

8  The domus and Triconch Palace as aristocratic residences at Apamea (Balty 1997). There would also need to be a passage behind the couches to allow individual diners to leave and return to their couches without disturbing fellow diners. When all these requirements are taken into consideration, it is clear that the substantial width of the apses of the triconch (between 6.08 m and 6.60 m) would, indeed, be required to make this dining arrangement comfortable.7 In addition, it should be noted that both the lateral apses in the triconch had doors opening on to them from the rooms to the east, suggesting that there must have been space left for a passage around the stibadia. The size of the apses suggests that 7 diners could have been accommodated with reasonable comfort in each, giving a hypothetical total of 21 dinner guests. The scale of the Butrint triconch is certainly on a par with the great triconch dining rooms of late antiquity, although, as noted above, had it been much smaller it would have been rather cramped for the diners. Its apses are slightly smaller than those of the Palace of Theoderic at Ravenna where the apses are around 7.5 m in width; slightly larger than those at the House of the Protomes at Thurburbo Maius (Baldini Lippolis 2001, 313; Sodini 1997, 182); and those of the triapsidal hall in the eponymous house at Ptolemais (Ward Perkins, Little and Mattingly 1986, 126–43). They are comparable in width to those of the triconch at Piazza Armerina, although the apses at the latter are significantly deeper than those at Butrint (Wilson 1983) (Fig. 8.20). As described in Chapter 3, the internal appearance of the triclinium would, in fact, have been slightly uneven. The principal windows in the room were in the southern apse, while the northern apse would probably have had no natural light, given the presence of additional rooms behind it. The eastern apse may have had small windows above the level of the four niches that were spaced at regular intervals around the wall. At the same time, the floor of the southern apse seems to have been at the same height as the central part of the room, unlike the floors of the northern and eastern apses that were slightly raised. The central apse of the Butrint triconch has a semicircular masonry foundation placed concentrically to the apse. This foundation is around 1m in width and previously was tentatively suggested to be the foundation of a stibadium (Gilkes and Lako 2004, 166), although it is very hard to see how this could be the case. As noted above, it is generally thought that stibadia took the form of individual wooden couches and, indeed, a masonry structure would be wholly impractical. The structure at Butrint, in any case, would not be wide enough to support a reclining diner. Instead it is likely to relate to secondary occupation of the Triconch and as such will be discussed in the following chapter.

The abandonment of the triconch As described in Chapter 3, it is in fact highly likely that the triconch triclinium never hosted a banquet of the type described above. Instead the evidence points to the whole

297

complex being abandoned as an elite dwelling prior to its completion. The evidence for this abandonment (described in detail in Chapter 3), rests on two principal factors: first, the lack of decoration relating to the final phase, and, second, the dating of secondary occupation within the west and south wings of the complex, which apparently commenced very shortly after the construction of the triconch triclinium. The lack of decoration relating to the final phase of the complex was noted in early reports on the excavations (Gilkes and Lako 2004, 167), as it was clear that the triclinium had no traces of flooring or wall coverings. It is possible, but highly unlikely, that these had been present but had been conclusively robbed out at a later date. Even if the room had been decorated with marble veneers these would have left some trace in the form of fixing holes and fragments of marble left at the base of the walls. Nonetheless, the discovery of the elaborate stone windows that must have belonged in the façade of the building clearly indicates that the shell of the building was complete. Probably the clearest evidence that the building remained undecorated came from the excavations of the peristyle in 2003 (Fig. 8.21). The mosaic of the western portico of the earlier, smaller, peristyle courtyard was discovered battered but broadly intact. It had clearly been intended that this mosaic should be retained within the western portico of the enlarged peristyle, as this was the only part of the new peristyle that retained the dimensions of its smaller predecessor. It is reasonable to assume that the southern portico of the smaller peristyle (which was partly included within the new plan) had been paved in this fashion, and that the pavement had been removed to facilitate the laying of a new mosaic that would extend the length of the new enlarged south portico. The mosaics in the other porticoes of the smaller peristyle had also been removed, presumably to reuse the tesserae in new pavements. It is interesting to consider that the builders did not feel that the (presumably geometric) southern portico mosaic could simply be extended along the length of the longer portico. It is possible that they felt they could retain the mosaic of the western portico, as that would be furthest from the new public areas of the enlarged complex and no longer visible to those entering the house. None of the other porticoes of the new peristyle had any trace of paving, with no sign of make-up layers or bedding for floors. It is therefore clear that this section of the building project was never completed. The date at which the triconch triclinium was built can be placed with a reasonable degree of reliability between AD 400 and 420. As the earlier (phase 3) domus (with the smaller peristyle and the western entrance) may well be as late as AD 400, it is likely that the triclinium and the larger peristyle may be closer in date to AD 420 unless we argue that the phase 3 domus was rebuilt almost instantly. Indeed, if we accept that the coins (dating to AD 408–423) from the levels cut by the north wall of the north wing are not intrusive, the grandiose complex of phase 4 that includes the triconch can be no earlier than 408. It is clear, however, that from quite an early date in the first half of the 5th century,

William Bowden

298

Triconch Palace, Butrint

Palace of Theoderic, Ravenna House of the Protomes, Thurburbo Maius

House of the Triapsidal Hall, Ptolemais

Triconch Palace, Butrint

Figure 8.20. Comparative plans of buildings with triconch dining rooms. Butrint, Palace of Theoderic (Ravenna), House of the Protomes (Thurburbo Maius), House of the Triapsidal Hall (Ptolemais), Piazza Armerina (other plans after those reproduced in Baldini House of the Triapsidal Hall, Ptolemais otomes, Thurburbo Maius Lippolis 2001)

Piazza Armerina

0

50m

8  The domus and Triconch Palace as aristocratic residences

299

Figure 8.21. Overall shot of small peristyle, showing western portico mosaic (JBB)

much of the west wing (Rooms 23, 25, and 26), were being used for secondary occupation. This included a series of hearths with associated deposits containing material dating from the late 4th and early 5th centuries.8 It could be argued that these hearths are associated with the construction phases of the triconch. We might imagine, for example, workmen cooking in rooms that were going to be renovated at a later stage. However, this evidence, together with the clear signs that the building was never completed, seems best interpreted as supporting the idea that the phase 4 complex with the triclinium and huge peristyle was begun around or before AD 420 but that the idea of using it as an elite residence was abandoned shortly after. Can we advance any explanation as to why this occurred? It is difficult to distinguish between causes that were local to Butrint and causes that were part of a more widespread process of social change. Over the course of the Butrint project, a number of suggestions were advanced, some of which made it into print. The first of these – that the triconch was abandoned because the construction of the city wall destroyed the relationship between the building and the Vivari Channel (Hodges et al. 1997, 223) – can definitely be discounted. Although it was a reasonable hypothesis, the city wall has now been conclusively dated to the first half of the 6th century, some 80 years or more after the cessation of construction at the triconch. More recently it has also been suggested (e.g. Bowden and Mitchell 2007) that the abandonment of the triconch project was due to repeated inundations of the site caused by rising sea levels and apparent subsidence within this area of Butrint (see above and Hernandez and Çondi 2008). This

remains possible and it is certainly likely that the site became increasingly susceptible to flooding in this period. It has already been noted that when the domus was rebuilt around AD 400, there was a clear attempt to raise its floor levels by almost half a metre by comparison with its predecessor, and it seems reasonable to surmise that this was partly an attempt to combat this problem. However, the argument that the construction programme was abandoned because of this problem must remain a tentative one. As the problem of water levels was seemingly an ongoing issue, it would be unlikely that the owner would commence the project at all if this was going to cause a significant problem. There are, of course, numerous other possible reasons for the cessation of building that are specific to the circumstances of the building’s owner. He may have run out of money or suffered some form of financial setback.9 He may have died and his heirs decided not to complete the project. He may simply have lost interest in it as a project and turned his attention to another residence. As I have argued elsewhere in relation to churches (Bowden 2001; 2003a) construction in itself does not seem to have been a particularly expensive undertaking in late antiquity until money was spent on lavish materials like marble (which is conspicuously absent from the triconch). Consequently, we should not overestimate the owner’s commitment to the project in terms of resources. Valuable items like columns and capitals were removed, perhaps by the owner himself (although we have no evidence as to when this robbing occurred). As I will discuss in the next chapter, there is certainly evidence that the Triconch Palace continued to be occupied in a coherent way through much of the 5th century, and it

300

William Bowden

is wholly possible that the owner or his descendants subdivided the complex and rented parts of it out to individual tenants, a process described in some detailed by Ellis (1988; 2000, 110–12). Ellis suggests that one explanation for this process may be the local curial classes disappearing in order to avoid crippling taxation, but there is simply no way of knowing. Certainly, the urban middle class became increasingly vulnerable to impoverishment (Brown 2002, 49–74), and this may perhaps have led to a new class of tenant in late antiquity. It is also conceivable that construction was abandoned because of changing social mores that rendered grandiose residences less important. The relatively early date for its abandonment might argue that local circumstances played a greater role than the more widespread cultural change that led to a drop in the number of luxurious private residences. This more widespread change seems to date to a slightly later period, starting c. 450–475 (Bowden 2003a; Ellis 1988; Ellis 2007, 12–13), and its causes remain obscure. As I have argued elsewhere (Bowden 2001; 2003a), the continued widespread construction of churches certainly indicates that there was no shortage of surplus resources available for construction. However, by the late 5th century these resources were no longer invested in grandiose private dwellings. At present the evidence from Butrint is insufficient to say whether any sort of more widespread change can be identified within the town around AD 420. Indeed, the first part of the 5th century in Butrint remains generally elusive apart from the Triconch Palace area, although as far as we can judge, the Merchant’s House area shows no sign of any sudden change in activity during the 5th century.

Was the triconch a church? Before turning to the complex phases of secondary occupation within the Triconch Palace and Merchant’s House areas that form the subject of the following chapter, it is important to consider briefly the question of whether the triconch triclinium could in fact be a church, as was suggested for the 60 years or so following its discovery. As was outlined by Gilkes and Lako (2004, 151), Luigi Maria Ugolini’s interpretation of the building as a Byzantine church (Ugolini 1937, 176) was followed by all the subsequent investigators (Anamali 1993: 470; Lako 1990; Meksi 1988, 207–8). This interpretation persisted until 1994 when the Butrint Foundation team offered the revised interpretation of the building as a palatial secular domus, which is the subject of this volume. Is it possible, however, that the final phases of the building (with the enlarged peristyle and the triconch itself) represent the conversion of a secular house into a church, as for example occurred on the Vrina Plain (Greenslade et al. 2006)? The evidence for the building being a church originally rested wholly upon the tri-apsidal plan of the main extant building, triconch churches being well attested throughout the early Christian Mediterranean (Deichmann 1954; Stollmayer 1999; Varalis 1999). Examples are present in

Albania at Butrint itself (recently discovered sections of the church on the acropolis), at Antigoneia (Varalis 1999; Mitchell 2006), Arapaj (Hidri 1983), Lin (Anamali 1993) and also in northwestern Greece at Paramythia (Pallas 1971, 236–7), and Dodona (Pallas 1971, 238–41) (Fig. 8.22). The Chi Rho monograms in the lunettes of the triconch windows could also be cited in support of such an interpretation. It could also theoretically be argued that the large contemporary peristyle was the atrium for such a structure, notwithstanding the fact that the principal entrances lie to the north and south of its eastern portico (an arrangement paralleled at Justinian’s church of St John at Ephesus). The case for an ecclesiastical interpretation for the building is weakened by the absence of Christian architectural sculpture (such as chancel screens or similar) or other decoration, although similarly plain buildings are also attested nearby at the Church of the Forty Martyrs (Mitchell 2004a) and Diaporit (Bowden and Përzhita 2004a). The most compelling evidence for a secular interpretation, however, is that relating to the date of the building. Although the dating of Christian buildings is notoriously problematic, often resting on circular arguments based on stylistic and typological considerations, it can be reasonably argued that most of the datable churches in Epirus date to the latter part of the 5th century until the mid 6th century (Bowden 2001; 2003a). Of the Butrint churches, those at Diaporit and Vrina have both been demonstrated to belong to the end of the 5th century (Bowden and Përzhita 2004a; Greenslade et al. 2006), while the Great Basilica and the Baptistery can be reasonably placed in the first half of the 6th century (Bowden and Mitchell 2004; Bowden and Përzhita 2004b; Mitchell 2004b). As has been clearly demonstrated, it would be difficult to place the construction of the triconch and peristyle much later than AD 420 which would make the triconch by far the earliest church in the town. Rather than being an unusually early church that was never finished and left undecorated, it is therefore much more logical to assume that the triconch was simply the latest phase of a long line of domestic buildings of increasing grandeur and complexity. The reasons for the abandonment of the project may never be found but the cessation of construction is by no means inexplicable. Instead, it can be seen as part of the picture of the changing mores of elite life in the later Roman Empire and of the changes that were occurring locally in Butrint itself.

Conclusion The triconch phase of the building is obviously that which dominates our interpretation of the complex. However, despite the grandiose nature of the triclinium, the design of the rest of the building in this phase remained conditioned by its predecessors. As with the previous phase of the building, ultimately the owner of the Triconch Palace failed to achieve a clearly articulated design. The component parts of the building sit rather awkwardly with one another,

8  The domus and Triconch Palace as aristocratic residences

Remains of He

301

llenistic city wa

Remains of He

ll

llenistic city wa

ll

Antigoneia Antigoneia

Paramythia

0

10m

0

10m

0

10m

Dodona Dodona

Lin

0

0

10m

0

10m

10m

10m 0 10m Figure 8.22. Comparative 0 plans of triconch churches. Antigoneia (after Mitchell 2007), Lin (after Anamali 1993), Lin Paramythia Paramythia, and Dodona (both after Pallas 1971)

with the earlier structures dwarfed by the massive, yet asymmetrical, peristyle and the vast triclinium. As with other earlier Roman villas, the Triconch Palace is ultimately the result of piecemeal construction over many generations, a factor that, to modern eyes at least, has a clear impact on the architectural coherence of the building. The Triconch Palace that we see at Butrint today and the structure that we see in plan here is perhaps therefore something of a chimera, a collection of opulent rooms and vistas of different periods that never really came together to form the sum of its parts. One wonders whether this factor contributed to the abandonment of the building programme of the final phase prior to its completion, the owner losing faith or interest in a building that would never fully realise the architectural model of elite life to which he aspired. Equally, however, one wonders whether the picture that we obtain from the Triconch Palace might not be replicated at many of the other iconic late antique domus sites that we know primarily as building plans, which become ever more coherent and canonical as they are redrawn and republished.

Notes 1

2

3 4

5

The literature on this subject is now vast. A comprehensive guide to the available bibliography can be found in Uytterhoeven 2007a and Uytterhoeven 2007b. For comparative plans of late antique houses see Baldini-Lippolis 2001; Sodini 1995; Sodini 1997. To avoid tedium for the reader, the traditional caveat “socalled” and inverted commas have been generally omitted from these and subsequent similar house names. This does not, however, imply acceptance of these designations. The western entrance has been depicted in this form in the reconstruction by Studio Inklink. Against this interpretation it could be argued that the baths would probably be used during the afternoon, after the morning salutatio was over and before any evening entertainment commenced. However, the argument for these areas being out of the public eye seems stronger. No excavation was carried out during the 1994–2003 excavations within the triconch itself since it had already been excavated to below floor level during earlier campaigns by the Albanian Institute of Archaeology, who believed that it was a church (Lako 1990). It is, however, possible that deeper excavations in this area would reveal further buildings.

302 6

7

William Bowden This certainly happened at Nicopolis where the southern section of the late antique wall was built directly above the decumanus maximus (Bowden 2003a, 89; Bowden 2007, 142–4). This raises some interesting questions about the use of the lateral apses of other ‘grand dining rooms’, such as the eponymous example at the House of the Seven Apses at Djemila, where the apses measure around 4.50 m across at

8 9

the chord. Their relatively narrow size suggests that dining in one of the apses in the Djemila house would be a rather cramped experience, and that once ensconced it would be very difficult for diners to move. These sequences are discussed in detail in Chapters 2 and 3. I am assuming a male owner, as we know that the owner around AD 400 (when the mosaic in the western entrance was laid) was male.

9  Urban change and the Triconch Palace site in the 5th to 7th centuries William Bowden

One of the most striking features of the archaeology of the Triconch Palace was the complexity of the deposits and structural remains relating to the period between the abandonment of the triconch construction project around AD 420 and the hiatus in activity that commenced around AD 650. These phases (5–10) were the subject of a great deal of attention, because they were seen as crucial to understanding the transitional period that was the original focus of the Butrint Project (Hodges et al. 2004: 8–11). They were also, in the main, easier to reach than the earlier deposits relating to the construction of the Roman phases, due to the high water table that hindered access to the earlier levels. Equally, with their post-holes, ephemeral surfaces and roughly built walls bonded with earth they were familiar territory to the excavation team, trained for the most part in the context of UK-based rescue archaeology. These factors led, perhaps, to the excavation creating an artificial separation between the ‘Roman’ phases of the domus and triconch (with their mortared walls, regular plans, columns and mosaics) and the ‘late antique’ phases that followed the abandonment of the triconch construction (with post-holes, trampled surfaces, etc.). This distinction, which has been perpetuated in this volume, also reflects a distinction between ‘Mediterranean’ and ‘northern European’ archaeological traditions, which with hindsight may have played a (generally positive) role in the excavation of the Triconch Palace, in that the background and research interests of those involved led them to focus on the complex late antique deposits rather than solely on the Roman structures. This separation, however, inadvertently perpetuates the erroneous idea that the hearths on top of mosaics are somehow less Roman than the mosaics on which they sit. Instead, it should be emphasised that most of the archaeological features discussed in this chapter are not phenomena associated with a post-Roman or early medieval town, but were occurring within the context of a town that remained in part a monumental city. The construction of

the triconch triclinium and its grandiose peristyle were abandoned at least 80 years prior to the construction of the late Roman town walls, and around 100 years before the construction of Butrint’s Baptistery and Great Basilica. The small buildings, industrial activities, and burials within the shell of the former Triconch Palace were taking place within the context of a dynamic and changing urban centre, rather than one that was in a steep and irreversible decline. Although many of the archaeological phenomena discussed below are of a type that has become very familiar from late antiquity – indeed, they form a key part of the paradigm of ‘urban transition’ that has become dominant in recent years – they also need to be understood on a local level. The archaeology of the Triconch Palace and Merchant’s House areas between the mid 5th and mid 7th centuries gives some important indications about the ways in which the topography of Butrint was changing, as well as hinting at some of the reasons for those changes. The discussion of these phenomena is treated thematically here, although these themes also fall into a broadly chronological scheme. This sees relatively coherent domestic occupation until the end of the 5th century, subsequent to which the nature of the area is changed by the construction of the city wall in the early 6th century. During the 6th century, activity becomes more varied and temporary in nature, with the area used for industrial activities, quarrying, rubbish-dumping and burial. In the latter part of the 6th century, we see more coherent building in the Merchant’s House, including a two-storey residence of a type that became increasingly common in late antiquity.

Subdivision and elite building at Butrint The phenomenon of ‘subdivision’ in peristyle houses and public buildings was first highlighted by Simon Ellis (1988) and has since become a cornerstone of debate about the changing nature of the late Roman town having

304

William Bowden

been documented throughout the empire (Brogiolo 1994; Saradi 1998; Uytteroeven 2007a, 45–46 with references). This process, in which roughly built walls appear in the urban landscape, creating small dwellings or workshops in previously grandiose houses or public buildings, has been interpreted in numerous different ways. Early explanations of subdivision were generally couched in pejorative terms, with these buildings ascribed to ‘squatters’ or barbarians camping within the ruins of Roman towns whose civilized grandeur they could not possibly comprehend. In the East, these structures were often associated with the ‘Islamisation’ of the cities, drawing on Jean Sauvaget’s famous schematic depiction of the change from the colonnaded street of the classical world to the Arab souk (Sauvaget 1934). It is now, however, clear that this type of construction appeared in many different cultural contexts in both urban and rural spheres (Lewit 2005). In many instances, for example the post-earthquake simple buildings in Hanghaus 2 at Ephesus (Ladstätter and Pülz 2007, 419) these buildings continue to be seen as representing the widespread impoverishment of the middle and upper classes, as is suggested in textual sources that highlight the difficulties faced by decurions on maintaining their houses (Saradi 1998, 22–23). However, this development also needs to be viewed in terms of changing attitudes towards urban space and the changing mores of middle and upper class domestic life (Thébert 1987). The uses of the different rooms of the Triconch Palace immediately after the abandonment of the phase 4 construction programme can be interpreted in the context of this phenomenon (Fig. 9.1). Between AD 425 and 450, the southern and western wings of the domus were occupied for domestic purposes and perhaps also served small-scale industrial functions (see below), although the nature of this occupation remains frustratingly opaque. Room 18 in the southern wing contained a hearth at one end, while numerous small post- and stake-holes attest to partitions within the room. It was extremely difficult to determine a pattern within these features and it is possible that they represent a succession of short-term occupations. The west wing (in particular Rooms 23 and 26) contained a much more clearly defined set of features and, with Room 25 (the former colonnaded entrance), formed a relatively coherent complex that may also have included Rooms 27, 28, 29 and 31. To the south of the former colonnaded western entrance, Room 26 contained two hearths of which one was a fairly substantial structure that may have in fact been a kiln or furnace. Beyond this lay a further room (23) with a large pithos, probably for water, lit with four oil lamps found within a later roof collapse deposit. A similar pithos was later placed in Room 26. This is most obviously interpreted as domestic occupation within the former domus. It is possible to imagine a dwelling with a kitchen utilising the ventilation provided by the open column screens of the western entrance. While clearly lacking the grandeur of its predecessor, it was almost certainly a perfectly functional dwelling, roofed with tiles

(to judge from a later roof collapse) and looking out onto the courtyard. It is also possible that there was more than one dwelling in this area, with shared cooking space within the colonnaded entrance. In the late 5th century, this complex was augmented with the enigmatic round structure that filled Room 27. Although this feature is difficult to interpret it seems probable that it was associated with the storage of agricultural produce or other foodstuffs on a scale that suggests some surplus. It is possibly comparable to the similarly enigmatic, but much smaller, round features excavated at the early Christian complex at nearby Diaporit where they dated to the first half of the 6th century (Bowden and Përzhita 2004a). The contemporary occupation in the Merchant’s House area appears to be broadly similar to that of the west wing of the Triconch Palace. Prior to around AD 450, the occupation of the Merchant’s House appears to be fairly utilitarian. The earliest datable deposits relate to rough floor surfaces in Room 37 that were seemingly replaced at frequent intervals suggesting heavy use. The small size of the buildings and their proximity to the Vivari Channel, could suggest a commercial or certainly utilitarian purpose. At the time of excavation, and in early interim reports (Bowden et al. 2002, 201–2), it was argued that this commercial activity persisted into the second half of the 5th century, when the building was seemingly converted into a two-storey structure, with storage space on a lower floor, and residential space on the upper storey. However, although it can be argued that the buildings pre AD 450 may well have had a commercial or industrial function, the later evidence for this commercial function is limited. In fact, from c. AD 450, the occupation of the lower floor of Room 37 of the Merchant’s House, with its two pithoi and in situ amphora and rough floor of marble veneers, appears very similar to the contemporary occupation in the west wing of the Triconch Palace. If anything the occupation of Room 37 seems to have greater pretensions to grandeur than can be seen in the similar structures of the west wing. The owners had gone to considerable effort to collect fragments of marble veneers (perhaps stripping them from the remains of the adjacent domus), while the piers that supported the upper floor (discussed below) were well-built constructions of mortared masonry. As noted in Chapter 3, the extent of subdivision in the Triconch Palace within the 5th century in fact remains open. The earlier doors appear to have been left open through much of the building, maintaining access between its different rooms and areas. Equally, although some areas of the building were clearly the focus of secondary occupation, which differed considerably from the earlier opulence of the building, the hearths, trampled dirt and wattle partitions were also absent from some areas. In particular the west portico (and the rest of the peristyle) appears to have remained clear of debris until the early 6th century, as did the apsidal reception room (24). In both these areas the earliest accumulations of rubbish do not appear to have occurred until AD 500 at the earliest.

Room 29 Room 27

Merchant's House

Room 31

Room 25

Room 26

Room 23

Room 21

Room 20

Room 19

Room 22 Room 37 Room 18

Room 28

Triconch Palace

20m

0

Phase 5 Room 27 Room 29

Merchant's House

Room 31

Room 25

Room 26

Room 23

Room 21

Room 20

Room 19

Room 22 Room 37 Room 18

Room 28

Triconch Palace

Phase 6

0

20m

Figure 9.1. The Merchant’s House and Triconch in phase 5 (c. AD 420–450) and phase 6 (mid to late 5th century)

William Bowden

306

Auxiliary room

Cour tyard

Cour tyard

Hall

Exedra

Exedra

Ornamental

Ornamental

fountain

fountain

Well

Well

Hall

Latrin e

Latrin e

Courtyard Kitchen Portico

furni

um furni Prae

Prae

Baths

Room

CourtyardRoom

Room

Vestibule

Service doorService door

Room

Stairway Room

Stairway Portico

Business

Baths

Storeroom

Room

um

Kitchen

Cellar

Cellar

Auxiliary room

Room

Room

Storeroom

Storeroom

Main hall

Main hall

Storeroom

Domus V

Domus V

Vestibulepremises

Business

premises

Main entrance Main entrance

4th century 4th century

Domus V

Domus V

Yard

Yard Tower

Stables

Kitchen Kitchen House 5 House 5

Stables

Kitchen

Kitchen

Public courtyard

Public courtyard House 6

House 6

Well

Well

House 3

s Kitchen

Stairway

House 2

House 2 Kitchen

Stable

s

Kitchen

Kitchen

House 7

House 7

Kitchen

Kitchen

Stairway

Kitchen

s

Stable

Stable s

House 3

Stable

Tower

Kitchen

House 4 Kitchen

Kitchen House 4 Kitchen

Reinforcement Reinforcement wall wall

The fact that the peristyle courtyard seems to have remained free of buildings or of build-up of rubbish until the 6th century adds to the impression of coherence within the area of the Triconch Palace. As noted in Chapter 3, it is possible that it remained a single dwelling with domestic occupation concentrated in the west wing and utilitarian activities concentrated in the south wing. An alternative interpretation could see a complex of dwellings surrounding a central communal space. This recalls the House of the Frescoes at Tipasa, a peristyle house divided into four apartments, perhaps in the 5th century (Ellis 1988, 568) and in particular the Casa de los mármoles, in the Morerías district of Mérida in Hispania (Fig. 9.2). Here a major peristyle house, mainly of the 4th century, focused on a large apsidal reception room, was apparently abandoned following an episode of destruction in the 5th century. In the late 5th to early 6th centuries the house was subdivided into up to seven separate dwellings based around the peristyle, which probably functioned as a communal courtyard (Alba-Calzado 1997; Arce et al. 2007, 311–13). Similar transformations have been documented elsewhere in Mérida, for example in the House of the Aqueduct of Tiermes and in the domus in Calle San Francisco of Astigi (where fires were lit on the earlier pavements), and at other cities in Hispania such as Valencia (Arce et al. 2007, 323). However, the only securely confirmed 5th-century domestic activity in the Triconch Palace itself is that in the west wing, although the north wing remains largely unexcavated and we do not know what was removed from the triclinium and the rest of the east wing by earlier excavations. The 5th-century occupants of the dwellings in the Triconch Palace and the Merchant’s House had access to a certain level of material culture, notably a delicately carved bone handle depicting a hand clutching a scallop shell (SF 2154) from Room 26 and the ivory gaming piece from Room 37 (SF 2447). Both dwellings also produced considerable quantities of glass, including glass lamp fragments (although it is not impossible that in Room 26 these were present as cullet to be melted down). While it would be unwise to overstate the original cost of these items or the prestige attached to them, they certainly suggest a standard of living well above the simply utilitarian. The overall impression of the 5th-century Triconch Palace and Merchant’s House, therefore, is not one of impoverishment and decay, although it certainly represents a dramatic change from the opulence of the earlier domus.

House 1

Kitchen House 1

Inundation and the city wall The relationship between the Triconch Palace and Merchant’s House areas and the Vivari Channel is likely to have been complex and dynamic. A key aspect of the Figure 9.2. Casa de los mármoles (Mérida) (after Alba

Late 5th/6th centuries Late 5th/6th centuries Calzado 1997)

9  Urban change and the Triconch Palace site in the 5th to 7th centuries status accruing to the domus would have derived from its channel-side location, and the activities of the Merchant’s House also seem to have been focused on the channel. The expansion of the Merchant’s House to the south and the creation of the domus bath-house may, in fact, have relied on the reclamation of land from the channel. The proximity of the channel, however, would have rendered this area vulnerable to any change in the water level and would have required regular maintenance of the waterfront. The ways in which the water levels may have changed in late antiquity are complex. There was widespread coastal uplift in the Mediterranean between the 4th and 6th centuries, in an episode known as the Early Byzantine Tectonic Paroxysm (Pirazzoli, Laborel and Stiros 1996). This caused uplift of between 0.5 and 1 m on the Ionian islands of Kephallonia and Zante, although it was up to 9 m on Crete. This sort of tectonic movement may result in a wide range of localised variations, however, and it seems this area of Butrint was lowered by earthquakes in the 4th century (see Chapter 2 and Hernandez and Çondi 2008). The skirt of land to the south of Butrint’s acropolis apparently slumped as a result of tectonic movement (evidenced by the 0.50 m drop on the south side of the forum pavement). This would have had significant implications for the waterfront in the Triconch area, rendering it increasingly vulnerable to seasonal inundation, which remains a problem today. The high groundwater is apparent to any visitor to Butrint who views the flooded theatre, but the effects are even more pronounced in the Triconch area. During the excavation from 1994 until 2003 it was evident that the area was relatively dry from late April until mid July, after which the groundwater rose to above the level of the mosaics in the long gallery (Room 18). The level of the channel itself can change by 0.5 m or more with the tide. As discussed in the previous chapter, it is possible that the water levels may have been one of the reasons that the Triconch building project was abandoned prior to completion and it may have affected the way that the building was used subsequently. Although the long gallery and south wing were used quite intensively the most clearly defined areas of later 5th-century domestic occupation were in the west wing and Room 37 of the Merchant’s House that occupy slightly higher areas in the complex and are unaffected by the seasonal rises. In the late 5th century the long gallery (Room 18) was apparently abandoned, perhaps as a result of inundation, represented by layers of sterile green silt that also partly extended into Rooms 21 and 22. Activity continued, however, elsewhere in the complex. An episode of abandonment is also apparent in the Merchant’s House at this time during the late 5th century. In both the Merchant’s House and the Triconch Palace, however, the abandonment of these areas would have facilitated the creation of an open area behind the city wall, which was built in the early 6th century and which fundamentally altered the relationship between the channel

307

and the town (see in general Andrews et al 2004; Bowden 2003a, 85–103). The relationship between the Triconch Palace/Merchant’s House and the city wall must remain a matter of speculation to some extent, but a number of aspects are worthy of comment. The line of the wall seemingly takes a dogleg to avoid the Merchant’s House and Triconch Palace (Fig. 9.3). There could be a number of reasons for this, not all mutually exclusive. First, it is possible that the owner(s) of the buildings carried sufficient local influence to avoid losing their properties to the construction of the city wall. The inclusion of a gate at this point (which would almost certainly benefit the owners/occupants of the Merchant’s House) may lend some circumstantial support to this hypothesis. Gates were by no means frequent in the late antique wall circuit, as far as we can tell, and the inclusion of one at this point may suggest some influence on the part of the house owner. Alternatively, the dogleg in the course of the wall may simply reflect the availability of convenient structures to be incorporated within the wall circuit. This seems to be a characteristic of the late antique channel-side walls of Butrint, which frequently included elements of earlier buildings (Andrews et al. 2004, 128–32). The new wall used the western wall of the Merchant’s House, thickening it with the addition of an awkward extra layer of masonry, while the southern part of Room 38 was allowed to protrude from the exterior of the wall. A new and more substantial section of wall then continued to the east, truncating the abandoned bath-house of the domus (Fig. 9.4). The need to follow existing buildings may relate to the third potential explanation for this sharp deviation in the course of the wall, in that the wall simply followed the edge of the channel, which may have been artificially altered at this point. As discussed in Chapter 5 (Phases 1–2), the layout of the Merchant’s House (Rooms 38 and 47) and the domus bath-house could suggest that land was reclaimed along the channel-side to create space for the southwards expansion of the buildings. If this was the case then the city wall may have also turned to follow this artificial river frontage. It seems most likely the route of the wall suited the wall builders rather than the owners of the Merchant’s House. This is supported by the fact that most of the buildings in the complex were demolished at this point. Indeed, Room 37, which had seen the most coherent domestic occupation, had been abandoned perhaps 50 years prior to the construction of the wall. It is certainly true that the builders of city walls had little compunction about making drastic alterations to urban topography, including the large-scale clearance of buildings in the path of walls, as well as in front of them and behind them (Christie 2001, 118–19). This is clearly indicated by the quantities of spolia contained within walls as well as by some examples of the dramatic effects of wall building, for example at Nicopolis, the provincial capital of Epirus, where the wall was built along two of the main axial streets (including the decumanus) (Bowden 2003a, 98;

William Bowden

308

Gate 6 Lion Gate

Gate 5 Lake Gate

Junia Rufina church

Gate 7

Gate 4

Acropolis Basilica

Great Basilica

West Gate Gate 8

Gate 3

Baptistery

Gymnasium church

Triconch Palace Western Defences

Merchant's House

Gate 2

Gate 1 Vivari Channel

0

200m

Figure 9.3. Plan of late antique Butrint showing known late antique gates

Bowden 2007, 142–44). A further striking example can be seen at Amphipolis, where the wall runs along the façade of Basilica A separating the body of the church from its atrium. While the Merchant’s House seems to have been largely demolished during or following the construction of the city wall, the effect on occupation across the Triconch area is more difficult to ascertain, although activity in

the west wing certainly ceased around AD 500, when the rooms became filled with dumps of broken roof tile. It seems likely, however, that the construction of the city wall significantly altered the nature of this part of Butrint. Subsequently the area changed into one of more liminal activity, used seasonally and on a temporary basis, while many of the building were quarried for building materials.

9  Urban change and the Triconch Palace site in the 5th to 7th centuries

309

Figure 9.4. The domus bath-house (Room 36) truncated by the city wall

Industrial activity and the question of ruralisation Late antiquity saw significant changes in the types of activity carried out within urban environments. Activities that had previously been carried out on the peripheries of towns, such as pottery manufacture, glass and metal working began to appear within both the former public areas of cities and within formerly grandiose residences. The laws that governed the location of these sorts of industries, requiring them to be sited in extra-urban areas, were still apparently in place, but seemed to have had little effect. The Urban Treatise of Julian of Ascalon, written in the mid 6th century, describes the rules and customs relating to building locations in Palestine (Saliou 1996; Zanini 2006, 398). It details proscriptions against the locating of noxious industries like tanning and cheese-making within the city walls, while industries requiring the use of fire, such as glass making and metal working were also limited to the peripheries. The only industry that was allowed within the town, according to Julian, was the working of gold and precious materials (Zanini 2006, 398). Whether or not these strictures were still followed to any extent in the East, it is clear that in many parts of Greece, the Balkans and Asia Minor, during the 5th and 6th centuries industrial activities were increasingly practised in areas formerly occupied by public buildings and luxury dwellings. What is less clear is the extent to which this reflects the new appearance of these activities within the urban area as a whole or simply their appearance in the

parts of towns that have been traditionally favoured by archaeologists (the monumental public centres and the houses of the elites). Examples are known from Philippi, where a luxury residence housed glass working by the early 5th century (after the house was abandoned in the second half of 4th century (Sodini 2007, 327)) and Delphi, where a major residence was transformed into a potters’ quarter in the 580s (Petrides 1997). As well as the appearance of extra-mural industries within the city, we also see an increasing presence of ‘rural’ activities within the urban areas, notably olive and wine presses, milling, and the keeping of animals. These activities are also accompanied by the creation of storerooms full of multiple pithoi suggesting the creation and storage of significant surplus. At Philippi, in the Bishop’s Palace, in the 6th century a two-storey building (discussed below) was erected in the former atrium of a peristyle house. A press was located in the ground floor, while an earlier small triclinium was converted into a storeroom containing 24 pithoi. Amphorae and pithoi were installed in rooms to the south of the atrium, together with a further press, suggesting that these too were used for production and storage (Sodini 1997, 460–1). Similar agricultural activity was clustered around the flanks of the Lechaion basilica in Corinth in seven houses constructed around the church, including some in the north portico of the atrium. The houses, which also recall the occupation of the Triconch Palace west wing and Room 37 in the Merchant’s House, included millstones as well as pithoi and hearths. The date of these buildings is difficult to

310

William Bowden

Figure 9.5. Late 6th- to early 7th-century house at Olympia (from Völling 2001)

establish although it is suggested that some (e.g. those in the north portico of the atrium) must postdate the destruction of the church, probably in the earthquake of 551. Other similar complexes are known from elsewhere in Greece (Sodini 1984, 370–2), for example at Olympia (Völling 2001) (Fig. 9.5), and rather earlier at Gortyn (Sodini 2003: 45), and from Asia Minor (Sodini 1997), for example from Sagalassos, where the so-called Palatial Mansion showed extensive agricultural usage, including troughs with holes for tethering cattle and the use of space for storing dried dung cakes (Vanhaverbeke et al. 2007, 504). The appearance of what are to all intents and purposes

villages over former urban areas is therefore emerging as a defining characteristic of late antique settlement. As with subdivision there are differing ways of interpreting these activities. Are we looking at a breakdown in the whole urban fabric or rational reuse of redundant areas of the town? This is particularly pertinent to the use of the Triconch Palace and Merchant’s House areas in the first half of the 6th century, in which the complex is turned over to the use of fishermen and blacksmiths, while other parts were quarried for building materials. At the same time, however, the presence of a masonry stair block in the long gallery suggests the creation of a two-storey building

9  Urban change and the Triconch Palace site in the 5th to 7th centuries

311

Merchant's House

Triconch Palace

Tile collapse Demolition spread

0

20m

Figure 9.6. Schematic plan showing late 5th- and 6th-century demolition across the Triconch Palace and Merchant’s House

(discussed in greater detail below). We also see the first appearance of burials in the abandoned rooms. This variety of activities is also contemporaneous with the construction of the town’s great Christian buildings, and with the town’s inhabitants’ continued access to products from across the Mediterranean. The end of the 5th century and the first half of the 6th century saw further demolition across the Merchant’s House area together with the collapse or demolition of the roofs of the west wing of the Triconch Palace, represented by thick dumps of smashed tile within the rooms (Fig. 9.6). The fragmentary state of these tiles suggests that they do not represent a roof collapse, but are instead dumps of broken tiles that could not be salvaged for use elsewhere. These dumps also extended into the western portico, suggesting that this area was losing its coherence, although it is of note that a well-built masonry wall was erected at this point to block the access between the courtyard and the south wing, indicating an attempt to differentiate between the activities in these areas. The absence of whole roof tiles suggests that this material was being removed for used elsewhere, perhaps in the Christian buildings that were being erected during this period (Bowden 2003a, 138). It should also be noted that the later stages of this demolition activity were occurring at the same time as the construction of the city wall and, certainly in the area of the Merchant’s House, there may be a link between the two. In the Merchant’s

House, however, the demolition seems to occur subsequent to the construction of the wall rather than before it. The south wing also contains evidence of small-scale industry from the early 6th century. A pit in the long gallery (18) containing the sawn-off neck of a Samos amphora dating to around AD 500 is probably associated with a blacksmith’s hearth, as this use of sawn-off amphora necks is characteristic of a later quite intensive use of the gallery by smiths. From fairly early in the 6th century, the apse of the reception room (24) was also being used by blacksmiths for a hearth or furnace that was rebuilt on a number of occasions, while huge numbers of post-holes cutting the early 6thcentury clay deposit above the mosaic of the reception room attest to structures that were frequently rebuilt (perhaps suggesting that they were temporary in nature). The south wing was also used for the indiscriminate dumping of mussel shells (Fig. 9.7). The mussels were being landed on the channel-side and shelled on the spot (much as they still are today), with the result that Rooms 18, 21 and 22 contained thick spreads of shells. Contemporary deposits of mussel shells also appear in Room 38 of the Merchant’s House and in the open area (39) in front of the gate in the city wall. The presence of these deposits is a clear indication that these are no longer maintained spaces, but rather show opportunistic reuse of an otherwise abandoned area. The same can be suggested in relation to the numerous

William Bowden

312

Merchant's House

Room 22

Room 21

Room 18

Room 39

Room 28

Triconch Palace

0

20m

Figure 9.7. 6th-century mussel shell deposits across the Triconch Palace and Merchant’s House

smithing hearths that characterise the occupation of the south wing of the Triconch Palace in the third quarter of the 6th century. As suggested in Chapter 3, the number of hearths and the apparent frequency with which they are rebuilt could suggest itinerant metal workers, returning to the same location at regular intervals and setting up on an otherwise largely abandoned area. The traces of hammerscale recovered from a number of these hearths show that these metal workers were clearly working with carbon steel, and probably using copper-based fluxes for welding. This would suggest that they were skilled specialist workers. The fact that it was possible to freely erect hearths within a partially ruined building, which was also being quarried for building material and being used for the dumping of rubbish and waste mussel shells, certainly suggests that the Triconch was increasingly peripheral to the occupied urban area. The quantities of household waste present, together with the burials discussed below, indicate a certain level of activity elsewhere, although the intensity of this occupation, particularly after 550, requires serious consideration.

Burials The appearance of intra-mural burials is one of the defining

features of the changes that affected the late Roman town, and one that has been widely documented elsewhere (Cantino-Wataghin 1999; Schachner 2007, 68–71 with extensive references). Consequently, the discovery of burials within the triconch area is of little surprise. However, the nature of the cemetery, with its very high proportion of infant burials, is worthy of comment (Fig. 9.8). The first burials within the Triconch area can be dated to the first half of the 6th century, appearing within Room 20 of the south wing and Room 29 in the west wing. A late 5th- or early 6th-century burial was also noted in Room 47 of the Merchant’s House. These were infant burials inserted within amphorae. A further stone-lined infant burial was interred in the long gallery (Room 18), at the end of the sequence of hearths discussed above, while two infant burials were interred in Room 20 (in an amphora and a tile grave respectively) and an infant was buried in an amphora in the apsidal reception room (24) From AD 575 onwards burials continued to be interred within the Triconch area, although it is of note that neither the courtyard nor the Merchant’s House were used for this activity (if one disregards the very inconclusive amphora burials from Room 37). Eight infant burials were found in Room 18 (one alla cappuccina, one amphora burial, one tile lined, four stone-lined and one in a simple cut), one

9  Urban change and the Triconch Palace site in the 5th to 7th centuries

Merchant's House

313

Room 25

Room 23

Room 20

Room 22 Room 21 Room 18

Room 40

Triconch Palace

0

20m

Figure 9.8. Distribution of late antique graves across the Triconch Palace and Merchant’s House area

stone-lined burial was found in Room 21, two neonate amphora burials in Room 22, together with a further infant in a simple cut, a further infant amphora burial in Room 24, and an infant burial in Room 25. Dating of infant graves tends to be more secure due to the fact that many are interred in amphorae. The latest is that from the apse of Room 24, interred in a Crypta Balbi 2 amphora of the mid to late 7th century. In total, therefore, we have a maximum possible 18 late antique burials within the entire excavated area, dating from a minimum period of 150 years between c. AD 500–650. This equates to approximately one burial every 8 years, far too low to reflect a frequently used cemetery. The predominance of infant and child burials present is very unusual while the diverse genetic groups present (with five infants or children displaying unique haplotypes) is also somewhat remarkable in such a small group of individuals (Fenton et al. in prep.). According to Todd Fenton, this genetic variability could indicate that a diverse group of individuals was using the area of the Triconch Palace as a burial ground, as opposed to a small number of families persisting over time. On this basis he suggests that during this time period this area of Butrint may have been acting as a migratory stop-over. It is tempting to suggest, therefore, that this burial data (in terms of both the genetic data and the low number of burials over a relatively long period)

supports the notion suggested above that the site was a stopping point for groups of skilled itinerant workers. The very high proportion of infant and child burials within the burial sample is intriguing and could be seen within the context of other cemeteries that seem to be exclusively for the interment of infants and children (Scott 1999, 114–23; Soren and Soren 1999, 477–82). As noted above, the dating associated with the adult graves is inconclusive, although the adult tomba a cappuccina in Room 18 is certainly most likely to be of late Roman date. However, the triconch cemetery differs from other infant cemeteries, for example that at Poggio Gramignano (Umbria), in that the age range of children at the triconch cemetery is rather more varied, with a significant number over one year in age, although, again, the most datable burials tended to be the youngest because of the use of amphorae. However, the relatively small number of burials over a lengthy period of time would probably argue against an interpretation of this area as a cemetery exclusively for infants and children, which would also explain the genetic diversity noted by Fenton.1 The distribution of graves is also intriguing, in that they are concentrated in the rooms of the former domus with none appearing in the former peristyle courtyard (16), and with only one noted in the Merchant’s House (although large parts of the Merchant’s House area were

William Bowden

314

Merchant's House

Merchant's House Room 37

Room 40

Room 37

Room 40

Room 18

Triconch Palace Room 18

Triconch0 Palace

20m

Figure 9.9. Location of two-storey late antique buildings in the Triconch Palace and Merchant’s House area

not excavated to late antique levels). While this might be explained in practical terms as a pragmatic use of ruined buildings, it may also reflect the desire to protect the graves from disturbance (either physical or supernatural). The apparent concentration of burials away from the Merchant’s House, however, may reflect the fact that during the second half of the 6th century the area on the interior of the gate saw significant permanent occupation, of a type that was seemingly absent from the Triconch Palace, in the form of a new and solidly built two-storey building.

The use of two-storey buildings The mid to late 6th-century building in the Merchant’s House is representative of the apparent change in domestic building that occurred from the second half of the 5th century that saw the increased use of upper storeys. The use of piers in Room 37 of the Merchant’s House in the second half of the 5th century, as noted in Chapter 5, apparently represents the earliest example of the creation of a two-storey building in the Triconch Palace area. Subsequently, two-storey buildings were created in Room 18 of the southern wing of the domus in the early 6th century (see Chapter 3, phase 7) and in Room 40 of the Merchant’s House in the mid to late 6th century (see Chapter 5, phase 9) (Fig. 9.9).

0 The adoption of two-storey structures is not readily20m attested elsewhere in Butrint, although it is highly likely that they were used in non-elite contexts from the early imperial period (particularly as terraced structures on the slopes of the acropolis). John Mitchell and I have suggested elsewhere (Bowden and Mitchell 2007, 471) that the appearance of this type of dwelling in the Merchant’s House represents an example of a phenomenon that is becoming increasingly well documented in late antiquity and the early Middle Ages, in which second storeys become a characteristic of elite residences. In the case of the 5th-century building in Room 37, while it seems hard to argue that a relatively small building of 7.74 × 7.5 m could represent any form of aristocratic residence in a Roman context, it is certainly the most sophisticated building within the whole excavated area of the Triconch Palace/Merchant’s House at this time. This apparent sophistication is also represented in the material culture recovered from the building, which included an ivory gaming piece (SF 2447; Fig. 5.33), a section of a marble sigma (SF 1349; Fig. 5.32) and high quality decorated glass (Fig. 5.34). The deliberate insertion of piers to convert a single storey building into a two storey structure is paralleled on a much grander scale at Nicopolis in the Episcopal complex

9  Urban change and the Triconch Palace site in the 5th to 7th centuries

315

A

A-A B B Room 40

B-B A 0

5m

Figure 9.10. Reconstruction of the mid to late 6th-century two-storey building from Room 40

to the south of Basilica B, where an audience hall had a series of massive piers inserted to support vaults and an upper floor (Papadopoulou 2007, 616–18), probably during the 6th century. A possibly analogous situation may also have existed at the Rruga Skenderbeu basilica in Saranda (20 km to the north of Butrint). Here a flight of steps in the northwest corner of the narthex, may relate to a similar two-storey building attached to the church, although galleries above the aisles may also be an explanation (Lako 1991). In the early 6th century there is evidence for the creation of a two-storey structure in the south wing of the Triconch Palace. The base of a masonry stairblock was constructed against the eastern wall of the eastern vestibule of the long gallery (Room 18), although it was not clear whether the

upper storey was constructed above the eastern end of the long gallery, or whether this was an external staircase leading to a second storey above Room 19. However, in the second half of the 6th century, a substantial building was erected in the southwest part of the Merchant’s House in the corner formed by the city wall (Fig. 9.10). The lower walls were built with stone, bonded with clay, while the upper walls were probably made of pisé. A staircase to the north was almost certainly external, leading to a platform supported by large posts, and the second storey itself. The Butrint two-storey structures are readily paralleled elsewhere in Greece and the Balkans in late antiquity. The appearance of a second storey is noted at Philippi in the Bishop’s Palace from at least the first half of the 4th century. Here an earlier house was divided into two

Western portico

Western rampart

Road

Eastern portico

William Bowden

316

Aqueduct Southern rampart 0

South gate

20m

Figure 9.11. Excavations in the southwest area of Caričin Grad (after Bavant 2007)

towards the mid 5th century, but in the last quarter of the 6th century the northern section of the house underwent a further major transformation, when a two-storey building was erected in the former atrium (Sodini 1997, 460–1). Olga Karagiorgou (2001, 197) also notes two-storey buildings at Nea Anchialos (with internal and external staircases). These are undated although they can be ascribed a general late antique date. Similarly at Corinth, the largest and most impressive of the houses (house 7) erected to the south of the Lechaion basilica was probably of two-storey construction (Sodini 1984, 370–2)

A particularly interesting parallel to the later 6th-century building from the Merchant’s House comes from a series of two-storey buildings excavated in the southwest part of Caričin Grad (Fig. 9.11). They were of two types, the first just a single building fronted by a closed rectangular yard with a portico on one or three sides, and the second a rectangular structure opening onto a yard and sometimes fronted by a portico. All the houses had external staircases of wood, sometimes with a stone base, and, as at Butrint, the lower walls were built from clay-bonded stone surmounted by cob-filled timber frame superstructures. Occupation was

9  Urban change and the Triconch Palace site in the 5th to 7th centuries on the upper floors with the lower floors probably used for stores and workshops (Bavant 2007, 369–70). Bavant cites examples of similar structures in Serbia, Macedonia, the Alpine regions and Bulgaria. The widespread adoption of upper storeys in late antiquity is increasingly recognised as a feature of housing of all social classes in late antiquity and the early Middle Ages. Certainly upper storey reception rooms are well documented in textual and archaeological sources in elite residences in Italy, most notably in Ravenna (Bavant 1989; Polci 2003 with references) and in the excavations of the forum of Nerva in Rome (Santangeli Valenzani 2004). In the Butrint examples, however, we may be seeing the adoption of these features by the middle classes. This emulation of elite architecture had commonly occurred within Roman domestic architecture since the late Republic (and was noted as a problem by Cicero (de Legibus 3.30–31) (Wallace-Hadrill 1994, 143) and was certainly an equally common feature within late antique residential building (Ellis 2007). Barbara Polci (2003, 104) suggests that this vertical growth of the city could in part have been a pragmatic solution to the decline in sewage and drainage systems. Given the apparent rise in water levels at Butrint it is possible that this could have been a factor in the triconch area, although two of the three examples of two-storey buildings are situated on slightly higher areas of the site. Equally, the adoption of small two-storey buildings in place of the peristyle houses of late antiquity appears to be increasingly ubiquitous and so we should also be seeking the explanation in wider social developments. The domestic building of more recent periods in Greece and the Balkans may well have its origins in some of these late antique and early medieval architectural developments. Sigalos (2004) gives examples of Middle Byzantine buildings from the 10th and 11th centuries, which, with their use of ground floor storage with pithoi, are reminiscent of the buildings that are emerging at the end of late antiquity. Two-storey buildings with a staircase to the residential first floor and an independent door for entering a ground floor with utilitarian functions are well documented in the 17th to 20th centuries in villages in Albania (e.g. Lazimi 1986; Riza 1978; Samimi 1978; Suli 1985; Zheku 1978).2

The 7th- to 9th-century hiatus The absence of mid 7th- to late 9th-century occupation in the Triconch Palace and Merchant’s House area is real. It is not an absence of material culture or the presence of forms of material culture that we were unable to recognise. Whatever the ambiguities of the archaeological sequence of the Triconch Palace were, this was not one of them. This has been reinforced by recent discoveries of 8th-century sequences in the Western Defences of Butrint and mid 9thto mid 10th-century phases on the Vrina Plain (Bowden and Hodges forthcoming; Hodges et al. 2009; Greenslade forthcoming). Material of this period was present in Butrint

317

but not in the Triconch Palace. The latest datable find from the Triconch was an amphora of Crypta Balbi 2 type of the mid to late 7th century used for an infant burial in Room 24. No coins later than the reign of Maurice (582–602) were found, although elsewhere in Butrint single coins of Phocas and Constans II have been noted (Cesano 1932, 74; Lako 1981, 127). Coins of Heraclius are conspicuously absent despite their presence elsewhere in Epirus, at Nicopolis and Kephalos (Bowden 2003a, 199–203 with references). The curtailment of occupation at the Triconch Palace may relate primarily to a long-term process of urban contraction rather than a sudden catastrophe, although the abandonment and collapse of the latest two-storey structure in the Merchant’s House (Room 40) in the last quarter of the 6th century is broadly contemporary with the documented Slav raids of 586/7 (Bowden 2003a, 196–200; Chrysos 1981; Curta 2004). These raids, recorded in the Chronicle of Monemvasia, together with the apparently more serious incursions of 614–616, recorded in the Miracles of St Demetrius, have caused considerable problems of interpretation for archaeologists in Epirus who have tended to treat them as a useful deus ex machina to explain any localised hiatus in the archaeological sequence. At the same time, however, the wider cultural implications of these incursions for the identity of the region’s inhabitants have been downplayed (Bowden 2003a, 24–33; Bowden 2003b, 58–9; Bowden and Hodges 2004). This forms an interesting counterpoint to the situation in Western Europe, where the immediate impact of violent invasion has been forsaken by scholars who focus instead on the long-term cultural transformations brought about by Germanic invasions (Ward-Perkins 2005, 5–10). At the Triconch Palace there is no evidence of violent destruction of the type claimed by Lako at Onchesmos (Saranda) (Lako 1991, 157) or by Ugolini for the baptistery at Phoenice (Ugolini 1932, 108). However, it would be obtuse to deny the potential effects of the instability of the late 6th and early 7th centuries on Butrint, given the descriptions of the woeful state of the towns of Epirus detailed in the letters of Gregory the Great. These record how by 604, the only remaining bishop on the mainland of Epirus was that of Nicopolis. Four further bishops were guests of a fifth (Alkison of Kerkyra), having been driven out of their sees by the invaders (Chrysos 1981, 74–6). Despite the huge areas excavated by the Butrint Project (Bowden and Hodges forthcoming), the 7th century remains almost wholly absent, with the second half of the 7th century entirely so.3 While future excavation may reveal sequences of this period (as with the fortuitous discoveries of late 8th century deposits in Butrint’s Western Defences noted above), there can be little doubt as to the scale of the contraction experienced at the town in this period.

Conclusion The archaeology of the Triconch Palace and Merchant’s House areas in the 5th to 7th centuries reflects a situation

318

William Bowden

that is becoming relatively familiar from late antique towns. From as early as AD 420, the complex was taken over by new and smaller dwellings within the shell of the earlier house. While these were clearly of a different scale to the earlier domus, we should not overstate the impoverishment of these dwellings, which were clearly coherent units in their own right and which, in the case of the two-storey structure in the Merchant’s House, possibly had aspirational elements and pretensions towards elite architectural practice. As noted at the start of this chapter, these buildings are appearing in the triconch area perhaps 75 years before the construction of the great ecclesiastical buildings of Butrint (Bowden and Mitchell 2004), so cannot be taken as any indication of a progressive generalised impoverishment within the town (just as the later appearance of the churches cannot be taken to indicate generalised prosperity) (Bowden 2003a, 105–59). A similar point is made by Tamara Lewit (2005, 255) in relation to the appearance of this phenomena at urban sites in Western Europe. At Barcelona, for example, subdivision of major houses, use of earth floors and spolia, and disuse of public areas coincides with the period when the city was an important episcopal centre and the residence of the Visigothic king. Similarly, at Mérida, the subdivision of major houses occurs contemporaneously with its greatest period of ecclesiastical eminence, while at Toulouse 5th-century subdivision coincides with largescale church building. This should serve as a salutary reminder of the multi-faceted nature of late antique urban change. Lewit interprets this new type of building as reflecting the changing mores of late antique society rather than impoverishment. Although the archaeology of the Triconch Palace

points towards a simplification of domestic architecture in the 5th and 6th centuries in Butrint, it was clearly still a community that remained in contact with changing Mediterranean supply networks throughout the 6th century (Reynolds 2004).4 Nonetheless, by the start of the 7th century these connections had dwindled almost to nothing and the town was no longer receiving coinage or using it for day-to-day exchange. The diminishing of Butrint’s connections with the wider Mediterranean world and the end of a monetary economy would have happened within people’s lifetimes. The parents of this generation of Butrint’s inhabitants, moreover, would have watched or participated in the erection of the Great Basilica and Baptistery. While the Triconch excavations present what is in some ways an increasingly familiar picture from the late antique Mediterranean, they offer a stark reminder that the process of ‘urban transition’ was one that meant real change that would have been readily apparent to the town’s inhabitants.

Notes 1 2 3

4

This issue will be examined in greater depth in the forthcoming publication of the human remains. These parallels and the accompanying bibliography were brought to my attention by Emanuele Vaccaro. The late 7th- to early 8th-century phase suggested by Paul Reynolds at the lakeside villa of Diaporit, based on the presence of unusual variants of Late Roman 13 amphorae (Bowden, Hodges and Lako 2002, 175–8; Bowden 2003a, 201–2), now appears less likely. The full late antique ceramic sequence from Butrint will be published by Paul Reynolds in a forthcoming volume.

10  From Roman insula to medieval quarter? Richard Hodges

Introduction For a thousand years after the 7th-century abandonment of the Triconch Palace and Merchant’s House areas, these channel-side plots were only episodically occupied. In this period no medieval equivalent of the Roman-period townhouses was ever to be erected here. Part of the reason for this was almost certainly that the area – as now – suffered seasonal flooding when the water table rose. The evidence of a seismic event at Butrint during the 4th century, clearly observed in the forum excavations to the north, as well as evidence of silts laid down in the immediate post-Roman period as a result of standing water in this part of central Butrint, all points to the problems inherent in permanently settling beside the Vivari Channel in the Byzantine period (Hodges 2008). The short abortive history of the Triconch Palace almost certainly shows that the situation was recognized as early as the 5th century. Nevertheless, the archaeology shows that the Triconch Palace area was not completely deserted during the thousand years spanning the 7th to 16th centuries. Quite the contrary, being beside the artery that connected the rich fishing in Lake Butrint and the Straits of Corfu, this area offered a premium location at certain times of the year.

Interpretating the archaeology of phases 11–15 The Triconch Palace and Merchant’s House areas were certainly abandoned by the mid 7th century. Other parts of Butrint almost certainly continued to be occupied on a small scale (cf. Bowden and Hodges forthcoming). As the Triconch Palace area was some distance from either the putative port on the western side or the old Roman roadbridge (close to the Great Basilica) further to the east, we should not be surprised (see Hodges 2008; Martin 2004). Essentially, this area was mid-way between the two main gates into Butrint, tucked behind a high wall and accessed by two minor gates. No evidence of any building as such existed in the

Triconch Palace area until the 10th century. There is evidence, however, of crushed mussel shells and accompanying hearths in some quantity inside the Merchant’s House gate (Gate 1). Fish-processing either on a regular seasonal scale, or conceivably on a larger scale over a shorter period, continued to occur here. Added to this, there are a few sherds of early medieval pottery as well as Otranto 1 type amphorae, like those now well-documented examples from the later 8th- or early 9th-century towers 1 and 2 in the Western Defences (Hodges et al. 2009). These might point to some occasional occupation. Whether the sherds are associated with the makeshift use of a building such as the triconch dining room itself, or a timber structure undetected in the excavations, or, indeed, were simply in domestic refuse within the later levelling deposits of phases 12–13 is impossible to say. The fortifications and the relationship of this area with the Vivari Channel are defining factors within the medieval history of the triconch area. The fortifications along the Vivari Channel were first erected in the early 6th century (Chapters 5 and 9; Andrews et al. 2004). The length of wall exposed defining the south side of this excavation area runs from the significant 22 m north–south diversion around the Merchant’s House area to Gate 2 immediately south of the triconch (that is, at the south end of the original (easternmost) plot that the Triconch Palace occupied). This early 6th-century stretch of wall included two gates referred to here as Gate 1, immediately in front of the Merchant’s House (described in detail in Chapters 5 and 6), and Gate 2, located east of the marine gate entrance of the Triconch Palace (Bianchi et al. forthcoming; Molla et al. forthcoming; cf. Andrews et al. 2004, 128–32; Gilkes and Lako 2004, 167–9) (Figs 10.1 and 11.3). Note should be taken of these gates; with the exception of Gate 4, just east of the Great Basilica, all other late antique gates provided major ingress into the town as opposed to ‘private’ access to the Channel. The late Roman defences appear to have been little altered when the refurbishment designated as Medieval 1 occurred, perhaps around the 11th century (Bianchi et al. 2007;

Richard Hodges

320

Burials

Room 17

Room 14

Room 27 Room 25

Room 31

Merchant's House

Chapel

Room 19

Room 40

Triconch Palace

Room 39

Gate 2

Burial

Burials

Gate 1 Mussel processing

0

20m

Phase 12

Merchant's House

Room 25

Room 31

Room 26 Room 24 Room 43

Chapel

Room 23

Room 39 Cobbled road or yard

Room R 40

Room 19

Gate 2

Triconch Palace

Gate 1 0

Phase 13

Figure 10.1. The Merchant’s House and Triconch in phases 12 and 13

20m

10  From Roman insula to medieval quarter?

321

Figure 10.2. The putative chapel within the remains of the triclinium complex (2 m scale)

Hodges 2008, 79–85; contra Andrews et al. 2004, 133–7). Unlike other parts of the circuit around the lower city, there is no clear evidence of repairs to the wall in the Triconch Palace area that can be securely related to this phase, although given that the walls survive to such a low level in this area it is quite possible that now vanished upper levels were repaired in this period. Whether or not this was the case, access to the Triconch Palace area apparently remained essential. Particular note should be taken of Gates 1 and 2 that were still in use, five hundred years or more after their construction (Fig. 10.1). Apart from Gates 4 and 9 (both close to the Great Basilica and possibly associated with a putative bishop’s residence), all the other entrances made within the Medieval 1 defences were major gates (see Fig. 11.3). The conspicuous rise in the volume of ceramics and coins belonging principally to the 10th to 12th centuries demonstrates beyond doubt the more intensive use of the area. It is not yet clear whether this preceded the refurbishment of the fortifications or accompanied these new works. Pertinent to this refurbishment are the remains of a surface that may represent a possible cobbled road leading from Gate 1 in the Merchant’s House area towards the direction of the old Roman forum. This unprepossessing road, along with the refurbishment of Butrint’s fortifications, the landscaping within the town described above, and the making of property boundaries attests to a new civic spirit. Two modest buildings belong to this era. Tellingly, both lie close to the gates: the sturdy, possibly two-storey, structure in Rooms 40/43 in the Merchant’s House area is by Gate 1, and the more modest post-built structure and its stone successor in Room 19 is not far from Gate 2. Proximity to

the Vivari Channel, accessed by these two gates, was almost certainly an important consideration in this reoccupation episode (see Plate 67). Cautiously we have been drawn towards identifying three different components to the triconch area in this period. These comprise: a possible chapel within the triconch dining room itself; an open space within the earlier peristyle courtyard, which saw the extensive dumping of 10th- to 12th-century material; and the modest buildings noted above. Let us look at each of these components in turn. The case for a small chapel within the southwestern Room 5 attached to the southern lobe of the triconch is tenuous (cf. Gilkes and Lako 2004, 170). First and foremost this is the best-preserved part of the triconch and much of the southwestern wall of the south lobe remains almost to eaves height, whereas the east and north lobes have been levelled. This suggests that the small room was functioning as some kind of structure when the rest of the Triconch Palace (perhaps along with the main palace buildings) was levelled. Some confirmation of this is to be found in the floor of roughly laid tiles found in earlier excavations by the Albanian Institute of Archaeology (Fig. 10.2). Removal of any stratigraphy covering this floor by earlier excavators effectively prevented any conclusive interpretation being reached. Certainly, as mentioned above, the association of this part of the abandoned Triconch Palace with graves found close to it led our Albanian colleagues to conclude that it was a church of some kind. The fact that mitochondrial DNA links exist between three of the skeletons found in the excavations attests to a common maternal relationship suggest a family group, which may have been buried in proximity to such a chapel. Similar small cemetery groups

322

Richard Hodges

Figure 10.3. Byzantine cross recovered from topsoil (SF 892)

resembling families from exactly this time have been found elsewhere at Butrint associated with the tiny chapel beside the well of Junia Rufina (Sebastiani 2008, 251–4) and the much reduced church on the Vrina Plain (concentrated in the chancel area) (Greenslade forthcoming).Further tenuous evidence for the presence of a chapel here comes in the form of the small damaged copper alloy cross (Fig. 10.3; SF 892) of later 10th- to 11th-century date found close by in topsoil 1114 (though similar, 6th-century small crosses are associated with households at Byzantine Hierapolis (Arthur 2006, 91–2, fig. 35) and there is no particular reason why such an object should be associated with a church. In sum, the interpretation of the triconch as a small chapel in use in the later 10th and 11th centuries rests on limited evidence. If such a building did exist, it was a modest chapel, with its altar placed in the southern apse of the room. Similar unprepossessing re-uses of earlier buildings as chapels are known at Butrint from this era. The east apsidal end of the late antique basilica found on the Vrina Plain appears to have been retained as a modest chapel between the 9th and 13th centuries, while its nave was made into a 9th- to 10th-century cemetery (Greenslade et al. 2006, 406; Greenslade forthcoming). One final and appropriate parallel is the re-use of the Roman and late Roman shrine of Junia Rufina, similarly just inside a gate – in this case the Lion Gate – for a chapel and cemetery (Sebastiani 2008). Nevertheless, by comparison with the plethora of private (urban) churches of Kastoria in northern Greece (Wharton-Epstein 1980), for example, these 10th- to 12th-century Butrint chapels were little more than private

shrines. In other words, churches constructed ex novo have yet to be identified as a feature of Butrint as it was renewed as a town in the early 11th century. The second component of this area is the apparent open space approximately equating to the large peristyle courtyard of the 5th-century Triconch Palace. The excavations produced no evidence of structures from this area, but a notable quantity of imported south Italian ceramics, especially Otranto types 1 and 2 amphorae, and small finds were found here, together with 12 folles of the 10th to 11th century, as well as a rare gold tetarteron of Basil II (976–1025). As noted below, it is unclear whether this represents the detritus of in situ occupation or commercial activity, or whether in fact this material was largely brought in from elsewhere in the town as part of a later levelling operation. The latter explanation finds some support in the mixed nature of the ceramic assemblage from these horizons and comparable deposits from elsewhere in Butrint, but this levelling operation must remain ultimately a matter of speculation. However, whether or not commercial activity was taking place at the Triconch Palace or elsewhere in the town, the material from these deposits provides a striking indication of the increasing volume of goods coming into Butrint between the 10th and 12th centuries. As noted in Chapter 4, the dominance of amphorae in these deposits by comparison with table wares is striking. A better understanding of the function and circulation of traded goods in this period may ultimately lead to a clearer understanding of the contemporary character of Butrint and the relationship

10  From Roman insula to medieval quarter? between the growth of imports and the development of new urban infrastructure at the town. The one structure associated with the area of the old palace peristyle – a D-shaped building based upon irregular limestone footings, 2.95 m across and 2.12 m wide – is difficult to interpret. The nearest, but improbable, parallel is the base of a 12th-century Seljuk tent found at Hierapolis, Turkey (Arthur 2006, 136–7, fig. 69). Only one dwelling was found within the Triconch Palace area in Room 19. This was located a short distance from Gate 2 in the fortifications to the Vivari Channel. This building was readily identified, but the many post-holes in the area immediately to the west in other deserted rooms of the old Triconch Palace may also relate to other contemporaneous structures. Yet more postbuilt structures may have existed in the area of the earlier south range of the domus. All in all it looks as though one or more simple buildings separated the peristyle area from the Vivari Channel. The building found in Room 19 was oval in shape, 5.5 m north–south and 2.8 m east–west at its widest point, a footprint of about 15 sq. m. The postholes were substantial – ranging between 0.32–0.54 m in diameter and set deep into the earlier deposits here. Posts of these dimensions could have supported a first floor, a feature found in the late antique buildings occupying the channel-side (Bowden and Mitchell 2007). Two steps on the south side, adjacent to the old Triconch Palace marine entrance, may form part of a makeshift stone stair block, but this is far from conclusive evidence that the building possessed a first floor (see Fig. 4.31). The structure, if it did in fact possess a first floor, is a more modest version of the dwelling found in Rooms 40/43 in the Merchant’s House sector (see below). However, as a single floor dwelling, with its distinctly oval form, the Room 19 structure is closer to the single-floored cabane, well-known in most parts of Italy from the 7th to the 10th centuries, often with a sunken floor (cf. Arthur 2004; Galetti 2001; Valenti 2004). It would have had walls of wattle and daub and, almost certainly, a roof of thatched reeds. Associated with it was a local cooking-pot of the 10th century in an apparent foundation deposit, confirming a phase 12 date for the structure. It was succeeded by a phase 13 version made in rubble, only the thin foundations of which survived. We can only assume that the stone building took more or less the same form as the timber one, but with rubble walls standing perhaps up to a metre or so. This latter building constitutes a small version of the familiar high medieval peasant dwellings found in southern Italy and the southern Balkans (cf. Sigalos 2004). The possible cobbled street inside Gate 1 was certainly a prominent feature of this period, although it is also possible that it was a narrow yard surface. A 14 m long tract of this putative road, approximately 4.5 m wide, was constructed here using small cobbles. Beside it, tucked inside the gate, was a sturdy rectangular post-built dwelling, occupying the footprint of the phase 9 building. It measured 8 m long (north–south) and 3.5 m east–west, approximately 28 sq. m in area. The remains of internal post-holes suggest a lean-to roof that projected beyond the east-facing wall, and

323

there was supported by a row of smaller posts. The robust construction suggests that the building had a second floor. In all, then, this building had potentially over 50 sq. m of floor space, with the lower floor being presumably intended for storage, although it can never have been terribly dry. Interestingly, as in the case of the Room 19 building, a small pit was found (4025) containing a complete 11thto 12th-century cooking pot. On the opposite side of the road or cobbled yard surface was a building, or pair of buildings, with beaten earth floor (sealing earlier middens of crushed mussel shells). The purpose of this building or buildings is a matter of conjecture but in the late 12th or early 13th century it was abandoned. Soon afterwards, small-scale smithing took place here, perhaps because it was a convenient location. But that too ceased promptly with the changes brought about in the 13th century. The archaeology of phase 14 (13th to 14th centuries), contra earlier interpretations (Gilkes et al. 2002), is strikingly ephemeral (Fig. 10.4). Building 40/43 had been abandoned. By this time the road or yard was covered with domestic refuse, in places up to one metre deep, and no longer used. The only significant investment appears to have been in a modest strengthening of the defences, now Butrint was in the hands of the Epirote Despots, by closing both Gates 1 and 2 (Andrews et al. 2004, 137–43, fig. 8.21; Bianchi et al. 2007). With no access to the Vivari Channel, these two areas appear to have been used as gardens, and for occasional burial. The domestic rubbish is perhaps associated with a household close by, while the putative chapel in the triconch dining room, or else another chapel hereabouts that has yet to be identified, might explain the burials. The single skull on the stone pillow, like a skeleton with a pillow found in the excavations at Junia Rufina (Sebastiani 2008, 255) and a similar occurrence at Corinth (Barnes 2003, 437, fig. 26.3) was almost certainly an earlier burial disturbed in the reconstruction of the tower noted below. The final significant episode here involved not any residential interests but a strengthening of the Vivari Channel fortifications. The new tower added to the wall circuit was well made with two floors and an internal staircase. This almost certainly belongs to the moment when Butrint became a Venetian possession in c. 1386–1387 (Soustal 2004, 25–6). Control of the lucrative fish traps had become of paramount importance, and soon led to the building of the Venetian Tower and Triangular Castle either side of the channel. Occupation in this strategically placed tower may explain the apparent concentration of later 15th- and 16th-century Venetian wares in the domestic rubbish found overlying the Merchant’s House rather than the Triconch Palace area. In sum, then, the excavations in the Merchant’s House and Triconch Palace areas reveal limited evidence of 7thto 9th-century occupation. In this period, as best we can presently tell, the kastron was first focused in the Western Defences before being transferred to the Vrina Plain (Hodges 2008). Mussel processing may have continued sporadically from late antiquity through the early Middle

Richard Hodges

324

Room 17

Room 29

Room 27 Burial

Merchant's House Burial

Burials R Room 40

Room 39 Refuse tips

Room 38

Triconch Palace

Mortar mixer

Gate 2 blocked

Pit

Gate 1 blocked

Tower

Phase 14

0

Phase 14b

20m

Figure 10.4 The Merchant’s House and Triconch in phases 14 and 15

Ages, but the evidence is not conclusive. By the 10th century, if not before, the channel-side area may have attracted at least periodic commercial activity involving initially substantial numbers of southern Italian amphorae. It seems that the nexus of the mussel processing was concentrated inside Gate 1, while the old peristyle was retained as a defined locus. Being an open yet contained space, readily accessed from the Vivari Channel, the peristyle would have certainly formed a suitable location for fairs or markets, although the evidence for this remains frustratingly thin and is based primarily on the fact that the peristyle was seemingly kept free of structures. The structural and material evidence suggests that Butrint underwent a significant period of development from the 10th to 12th centuries. The occupation of the triconch area was certainly much more intensive than previously and there was probably some refurbishment of the fortifications. But as a residential sector, unlike the Roman insula, it never grew into an urban quarter comparable with the mid Byzantine towns of, for example, Amorium (Lightfoot and Lightfoot 2007) or Pergamum (Buchwald 2007) at this time. Only two modest timber buildings, each close to a gate offering access to the Vivari Channel, date to this time too. Nevertheless, despite the simple character of the architecture, the refuse shows that these families had access first to Italian imported ceramics, and then from the later 11th century to glazed table wares from the Peloponnese and Aegean regions. This era evidently ended with the loss of Butrint

to the coalition involved in the Fourth Crusade if not before. A hoard of 97 Comnenan coins close to Gate 2 vividly benchmarks the significant changes the crusaders inadvertently brought about at Butrint (Fig. 10.5). By this time, perhaps stimulated by the constantly rising water table (cf. Bescoby et al. 2008), the nucleus of medieval Butrint had withdrawn to the slopes of the acropolis, with a ring of churches from the Gymnasium to the Baptistery church, aligned to the western gateway to the far west, and the Great Basilica to the east, denoting the southern limits of the residential dwellings. With Gates 1 and 2 now blocked, and access to the Vivari Channel no longer possible, allotments now occupied these areas inside the walls. The latest phase, attributable to the Venetian era, was essentially concerned with defending the precious fish traps here. The wealth of associated imported Venetian ceramics tellingly reveals that, although the buildings themselves were poor, those involved in fishing enjoyed apparently high standards as far as their table wares were concerned. It was certainly the value of the fish traps that led the Venetians to defend Butrint against all odds up until 1797, when they finally ceded Butrint and Corfu to the French under the Treaty of Campo Formio (Soustal 2004, 26).

Conclusion The archaeology of the Triconch Palace area in the medieval period is often difficult to interpret for three reasons. First,

10  From Roman insula to medieval quarter?

325

Figure 10.5. The Comnenan coin hoard found next to the city wall, perhaps deposited prior to the Fourth Crusade crossing the Corfu Strait in 1203 the imposing ruins of the Triconch Palace itself existed as a template determining the location and form of subsequent buildings. Second, the dumping of material (much of it containing domestic refuse) to form new levels – perhaps terracing to confront the rising water table – when mixed with alluvial deposits, and then churned up by human activity, have complicated the interpretation of the site. Several important levels show mixing of Roman and medieval material. Third, the preponderant type of building found hereabouts after AD 600 was constructed with timber posts (cf. Bowden and Mitchell 2007). Major stone-built dwellings are absent. Disentangling building phases in the churned up riverside levels, especially during phases 12–13 when several structures were doubtless rebuilt several times proved to be difficult. Let us look at each of these elements in more detail. First, let us consider the remaining importance of the Triconch Palace itself. It is clear that much of the upper parts of the Triconch Palace were levelled, possibly when the city wall was built alongside the Vivari Channel in the early 6th century. Significantly, though, the triconch dining room itself was not dismantled at this time, and indeed remains a prominent standing monument today. It seems, as though, there were two approaches to the building: one section – the dining room – being retained (or by dint of its more solid construction proving more resistant to casual quarrying); one section – the greater part of the house – being systematically robbed for building materials. In 1993 it was believed by our Albanian colleagues that the upstanding triconch was a church, an interpretation that, as discussed above, may be partly correct. Second, the dumping of material made the excavations

particularly difficult. Distinguishing one phase from another was far from easy. The mixing of large amounts of later Roman material with quantities of mid and late Byzantine ceramics and other objects was especially hard to track with certainty as the excavation proceeded. Only afterwards, with the catalogues of finds at our disposal, have we been able to make sense of these layers. Plainly the presence of 8th- to early 10th-century ceramics in small numbers suggests that the area was not entirely deserted at this time. However, were these potsherds disposed of here, or were these brought in as dumped material in the 10th to 11th centuries? Clearly, apart from the phases 12–13 dwelling found in Room 19 and the sturdier dwelling occupying Room 40/43 of the Merchant’s House belonging to the later 10th or 11th centuries, the sheer volume of coins and ceramics from the 10th to 12th centuries shows that the channel-side area was once again being actively occupied, although there remains the possibility that some of this material was moved from elsewhere in the town as part of an active policy of reclaiming the marshy channel-side through the dumping of extensive levelling deposits. Taken on its own the evidence for this possible levelling or terracing in the Triconch Palace area is not wholly convincing, as the deposits could also simply derive from intensive local occupation. However, much clearer evidence of comparable landscaping was discovered in the excavations of the forum, where the making of a terrace at the base of the steep incline of the acropolis hill, and covering the waterlogged areas immediately to the south, appears to date to around the later 10th or early 11th centuries. A terrace almost a metre high completely covered much of the earlier (ruined) Roman buildings on the north

326

Richard Hodges

side of the Roman forum (Hernandez and Çondi 2008, 289, n. 62). In effect, this terracing defined the topography of Butrint that endured until Luigi Maria Ugolini launched his excavations in 1928. Overlying the terracing at one point was a poorly made wall (forum 0141) made using Hellenistic wall blocks as orthostats. Importantly, within the rubble and earthen fill of the wall an anonymous folles of Basil II (976–1025) was discovered, which provides at least a terminus post quem for the its construction. This wall, one of several orthostat walls now identified at Butrint, is most likely to connote some kind of property division (Hodges 2008, 83). It is presently best paralleled by two examples found in recent Turkish excavations: the long standing wall found at Hierapolis, defining the new street frontage in the 10th century (cf. Arthur 2006, 47, fig. 12) and perhaps the powerfully-built enclosure wall made of spolia at Amorium of late 10th-/11th-century date (Lightfoot and Lightfoot 2007, 119–20). In sum, the terracing and orthostat walls, along with the creation of new fortifications enclosing the acropolis, and refurbishment of the late antique walls around the lower city (Bianchi et al. 2007), appear to belong to the revival of Butrint as a town in the early 11th century. It is tempting, then, to interpret the spreads of earth and rubbish found across the Triconch Palace area – in places up to half a metre deep – to the same urbanizing process. The intention, we might speculate, was to extend the terracing at the foot of the acropolis as far as the Vivari Channel. Most of all, the material raised the level of the hitherto slightly sunken peristyle and similarly lower southern range of the area. Unfortunately, however, distinguishing individual levels of occupation within these deposits, or separating in situ deposits from possible imported levelling material proved impossible. This brings us to the third point, the buildings of this period. Buildings of timber appear to have been a key element of the expedient use of building materials in late antiquity as well as in the middle Byzantine period. As we have seen, the late antique buildings from these areas – notably the phase 9 dwelling occupying Room 40 of the Merchant’s House area – appear to have established an architectural form that was maintained in the Middle Ages. Accordingly, timber elements played an important part in the 9th- to 10th-century minor aristocratic oikos on the Vrina Plain, made within a pre-existing stone building – the late antique Christian basilica (Greenslade et al. 2006; Greenslade forthcoming; Hodges 2008). Timber structures were certainly the precursors of stone buildings in Adriatic seaports such as Pescara, Italy (Staffa 2006), Ravenna (Augenti, Cirelli and Marin 2009, 141–3) and Stari Bar, Montenegro (Gelichi 2008, 17). The levelled walls of the Triconch Palace, as well as the levelled walls of the adjacent Merchant’s House, were punctured by numerous substantial post-holes along the tops of the remaining walls. These would appear to belong to medieval buildings that used the walls as secure

foundations. Although many of these post-holes could not be dated, the expedient use made of extant stone structures in the cases of the two principal buildings found in these excavations, in Room 19 and Rooms 40/43, suggest that many of the post-holes in the levelled wall tops are contemporary with these structures (which date to the 10th and 11th to 12th centuries respectively). Both buildings had timber frames, and traces of an external staircase may have existed for the building in Room 19 (cf. Augenti 2008, 183–92 for examples from Ravenna and Rome). Similar insubstantial buildings, normally with walls of reeds, clay and daub, and roofs of reeds, existed until the modern era, as Ugolini observed (1937). Neither building, though, could be described as a major town-house; none resemble the town-houses known from later medieval Greece or Montenegro (Gelichi 2008; Sigalos 2004). Major houses of this period in the southern Balkans and southern Italy were almost certainly built in uncoursed rubble. The fortified residence re-occupying the Acropolis Basilica at Butrint is likely to be just such a building (cf. Bowden and Mitchell 2004: 111; Hodges 2008, 81). A small corpus of similar buildings has now been identified further afield. A building of similar proportions and expedient construction to Butrint’s acropolis residence has recently been found (though is as yet unpublished) immediately west of the west entrance to the Roman amphitheatre at Durrës, ancient Dyrrhachium (Santoro, Hoti and Sassi 2008, 741–3). Yet another example is the house with a triclinium made of spolia to the east of the Peirene Fountain at Corinth (Scranton 1957, 39–41, 128–31, pl. 3.4). Smaller town-houses have also been found in the excavations at Corinth (cf. Scranton 1957, 15). These and others from Greece, including Athens, have been conveniently analysed by Sigalos (2004). Similar Byzantine residential building, including courtyard houses, has been identified in Turkey at Amorium (Ivison 2000; Ivison 2007), Hierapolis (Arthur 2006, 111–14, 131–4) and Pergamum (Buchwald 2007, 68– 9). Courtyard houses are also known from contemporary descriptions of Bari’s town-houses in the 10th to 12th centuries (Skinner 1998), Naples (Skinner 1994), Rome and Ravenna (Augenti 2008, 183–92). At Butrint, however, apart from the upstanding triconch – associated with which there is no medieval rebuilding – there is no obvious stone built town-house in the lower city conforming to these Byzantine architectural types. Plainly, the channel was not viewed as the premier place to construct substantial dwellings, as it had been for example in the Roman era and would be again in the later Venetian age (cf. Crowson 2007, 62–3). The fact that these timber structures, with the exception of the Room 19 building, were not remodelled in stone suggests that the channel-side area as a whole remained peripheral to the new town. Instead, the main nucleus was probably concentrated on the acropolis and its steep flanks, especially the northern slope, set back from the intractable problems of waterlogging in the lower city.

11  Conclusion Richard Hodges

When we launched the Butrint Foundation project in 1994, in common with our Albanian colleagues, we were tempted to interpret the ancient and medieval city through the prism elegantly fashioned by Luigi Maria Ugolini during the course of his large-scale excavations spanning the years 1928–36. In this view Butrint occupied a knoll with a skirt of surrounding lower ground beside the Vivari Channel that more-or-less remained an urban nucleus from Hellenistic to Venetian times. In short, Butrint at its zenith was marked by the line of its late antique defences alongside the Vivari Channel. This interpretation took almost no account of the considerable changes to the surrounding landscape and in particular to the changing water table, not to mention the considerable extant traces of buildings on the Vrina Plain. The latter has now been shown to be a major suburb of the Roman town and the administrative centre of the 9th- to 10th-century town. Our decision to excavate the Triconch Palace area was made for two reasons. First, this looked to be the site of a major early Byzantine residence, of a type that was increasingly being recognised as a key aspect of late antique aristocratic identity. Second, in 1994 this area seemed as though it had always been to a greater or lesser degree an occupied sector within the long life of the Roman and medieval town. After major surveys and excavations, these two assumptions required significant adjustment. First, the Triconch Palace was one of several later Roman town-houses located alongside the Vivari Channel. Second, the topographic history was much more complex, given the particular history of the water table at Butrint. Not unexpectedly, revisions to Ugolini’s paradigm have become both necessary and inevitable. In the course of the era since we began the Triconch Palace excavations the nature of Byzantine urbanism itself has come to be questioned. The assumption of urban continuity between late antiquity and the middle to high Byzantine periods can no longer be taken for granted (Brandes 1999; Haldon 1999; Ivison 2000; 2007). Instead, archaeologists and historians, with access to limited data, have begun to describe phases of ‘ruralisation’ in late

antiquity, and to call for a re-examination of the binary division between town and country between the 7th and 11th centuries. To what extent, too, have scholars defined towns by their administrative or defensive functions? Finally, as in northwest Europe, archaeological categories of trading places may well have existed that are not readily identified in the texts (Skre 2008). In short, with more archaeological interventions the history and form of Byzantine urbanism is being re-shaped. The stereotyped understanding of urban history can only be confronted by grasping a chronologically well documented sample of an urban area, and treating it as a sample, of which interpretation will be subject to further examination and revision as the opportunities arise. In this case, we believe, thanks to the Triconch Palace excavations, we have begun to shape a different history of urban process at Butrint. This history has been subject to the regional and extra-regional circumstances that have always governed the operations of this place, including its very particular and always fluid environmental circumstances. Let us begin by summarising this environmental context that undoubtedly played a major part in determining the topographic history of the Triconch Palace area. The slightly elevated outer margins of the skirt of land, which the Triconch Palace and the earlier Roman channel-side town-houses occupied, is likely to gain its character from underlying gravel formations (Fig. 11.1 – see colour plate 68). The lower-lying tract of land between the line of the south Hellenistic fortification and this slight ridge is likely to have undergone gradual infilling since the Hellenistic period, becoming dry land by the 1st or 2nd century AD. Tectonically, the area is undergoing gradual downlift, leading to a relative rise in water levels over time (Bescoby forthcoming). It is increasingly apparent that the immediate area experienced subsidence as a result of seismic activity in the 3rd or 4th centuries, resulting in the tilting of the forum pavement (Hernandez and Çondi 2008, 290). The rising water table clearly posed a problem in late antiquity, indicated by the deposits of what appear to be water-borne silts in the long gallery and the fact that the

328

Line of Section

ue

tiq

n eA

Triconch Palace

t La of all e n Li ty W Ci

12th century - present

Vivari Channel

6th - 12th century 11th Century levelling/ dumping deposits

2nd - 5th century ?Gravel formation

4th Century make up layer

0

40

80 metres

Figure 11.1 (Plate 68). Schematic cross section between the Forum and the Triconch Palace, showing development of deposits from 2nd to 12th centuries AD (D. Bescoby)

Richard Hodges

Forum

Line of Hellenistic wall circuit

Acropolis

Inactive Faultline

N

11  Conclusion floor levels of the later peristyle house seem to have been deliberately built at a higher level (see Fig. 8.8). By the 11th century the ground level was raised by a series of major deposits that may have been brought in specifically to counter the effects of the waterlogging caused by the higher water. If this was the case, the earth was presumably brought from areas of Butrint not subjected to seasonal waterlogging, possibly on the higher ground at the base of the acropolis slope. Alternatively, these deposits may have developed as a by-product of a relatively extended period of domestic rubbish dumping. However, perhaps as a result of the higher winter water table, the small, later medieval community ultimately retreated from the water’s edge returning to the extent of occupation during the later Hellenistic period (that is, little more than a narrow terrace extending out from the foot of the south-facing acropolis slope). In short, the intensive occupation of the skirt of land reaching to the edge of the Vivari Channel was a feature of the Roman town, and ultimately a short-lived feature of the new 11th-century town. Now let us examine the topographic history of the Triconch Palace area in the context of the changing topographic history of Butrint itself. The archaeology of the Triconch Palace area (including the Merchant’s House) mirrors that of the Vrina Plain, at least for the 2nd to 5th centuries, as modest town-houses evolved into large waterside properties (cf. Crowson and Gilkes 2007; Greenslade et al. 2006; Greenslade forthcoming; Hansen 2009, 81–87). This, then, begs the question as to whether the initial occupation of this channelside area began on the north side as soon as the Roman colony was founded in the Augustan period with small cottages similar to those discovered only in a fragmentary state below the later houses on the Vrina Plain (Greenslade forthcoming; Hansen 2009, 41–45). We might therefore deduce, that whatever form any early properties took, these were combined and aggrandised to become the domus described in Chapter 2 – the precursor of the Triconch Palace itself (cf. Hansen 2009, 44). The domus was evidently owned by a grandee whose name survives in part in a mosaic inscription that greeted visitors to the building as they entered off the alleyway from the west. The inscription can be read as ...]ẠṂαρίῳ τῷ λανπ[ρ]οτά[τῳ|.. and is interpreted as a name or title ending in ]ario and the rank lamprotatos (equivalent to the Latin clarissimus), someone of senatorial rank, thereby announcing the owner’s name and status to a visitor. It would be a mistake, however, to envisage an uninterrupted evolution from a colonist’s cottage to a 4th-century domus. Plainly the prosperity of Butrint suffered in the later 3rd century and it would be premature to generalize as yet how the channel-side property owner confronted the evident economic and seismic perturbations at this time. What is clear is that around AD 420, the owner of the domus was able to embellish his property, dramatically enlarging the peristyle and expanding onto adjacent plots to the east and west.

329

The focal point of the new complex was a lavishly conceived triconch dining room, designed as a location in which the owner of the house could entertain his guests in an atmosphere where their relative status was clearly defined. The aggrandisement continued the division of the building into public and private areas, and the triconch triclinium was seemingly intended to be accessible from the exterior of the complex on the north and south sides. To the south, in the architect’s uncompleted vision, the visitor may have been able to enter the palace directly from the Vivari Channel. This relationship with the waterfront was clearly intended to be important, judging from the insertion of an elegant new marine entrance. The ability to control the surrounding environment in terms of commanding sweeping vistas (often over water) was of great significance in the ideology of Roman residential housing. That the owner of the Triconch Palace was able to block streets to the west and the east (where a further road is thought to have existed) indicates both the changes that occurred in the topography of the town during late antiquity and the ability of this proprietor, at least, to manipulate his environment. The conspicuous deployment of spolia, fine stone probably removed from earlier Roman public buildings, provides another insight into the aspirations of this owner. As far as we can tell, intriguingly, spolia was not available to the builders when the smaller peristyle domus was built around AD 400, which could suggest significant changes in the public areas of the city between 400 and 420. Remains of four elegantly carved stone window frames with Chi Rho motifs show that the building was virtually completed and roofed when it was strangely abandoned. It also reveals the owner’s overt Christianity leading to speculation amongst Albanian archaeologists like Neritan Ceka that this was a bishop’s palace (Ceka 2005, 289; cf. Marano 2007), although Christian imagery was commonly employed in elite residences in late antiquity (Bowden 2008). For sure, the proprietor was a major figure in the early 5th-century town, but his town-house aspirations were on a similar scale to the proprietor of the Vrina Plain house from exactly this time (Greenslade forthcoming) (Fig. 11.2). More importantly, it is likely that several other dwellings may have lain alongside the channel as far west as the major bath-house excavated beside the Venetian Tower, which may have belonged to a private residence. Remains of major structures were incorporated in the later 5th-century fortifications (Andrews et al. 2004, 128–32, figs 8.4, 8.5; Martin 2004, 84–95), although their precise function remains unknown. Unlike the Vrina Plain where a large 5th- to 6th-century church was made within the earlier dwelling (Greenslade et al. 2006; Greenslade forthcoming), the Triconch Palace area was occupied by less imposing structures. Now, the shell of the palatial building was used for other purposes, both domestic and industrial. Most of the new buildings were combinations of timber and drystone or clay-bonded walling. None lasted more than a generation before being reconstructed. Were these seasonal occupants, or simply

Richard Hodges

330

20m

0

Triconch Palace Triconch Palace Triconch Palace

0

0

20m

0

20m

0

20m

0

0 20m

20m

Triconch Palace

Vrina Plain house

Vrina Plain house

Vrina Plain house Vrina Plain house

0

20m 20m

Figure 11.2. Comparative plans of the Triconch Palace and the Vrina Plain house

town-dwellers who coped with Butrint’s rising water? The answer unfortunately eluded us in the excavations. During the early 6th century Butrint was enclosed by a new city wall running along the water’s edge (Fig. 11.3). This separated the Triconch Palace area and its less impressive vernacular dwellings from the waterfront.

The construction of these defences represented a major investment for the Buthrotan community at this time (see Andrews et al. 2004; Bianchi et al. 2009; Molla et al. forthcoming). The defences appear to have been the work of different gangs, perhaps making use wherever possible of extant building materials. The Vivari Channel walls are of

Gate 6 Lion Gate

Gate 5 Lake Gate

Junia Rufina church

Gate 7

Gate 4

Acropolis Basilica

Great Basilica

West Gate Gate 8

Gate 3

11  Conclusion

Baptistery

Gymnasium church

Triconch Palace Western Defences

Merchant's House

Gate 2 Vrina Plain basilica

Gate 1 Vivari Channel

0

331

Figure 11.3. Plan of the walled area of Butrint (c. 6th century), with gates numbered

200m

332

Richard Hodges

far inferior construction to their western (landward facing) counterparts. It seems likely that expedience determined their architectural form: local materials were quarried from the upper fabric of the deserted Triconch Palace. Despite the construction of the walls, access to the waterway appears to have remained as important as ever with the presence of two gates making the fortifications potentially very vulnerable. A simple gateway (Gate 2) existed to the east of the former eastern boundary wall of the triconch area, and a simple gateway and quay existed in front of the Merchant’s House (which, too, was probably deserted by this time) (Gate 1). The significance of these gates should not be understated. Only three such ‘private’ gates provided access to the Vivari Channel. One directly in front of the Merchant’s House; one very close to the defunct Triconch Palace; and one north of any residence associated with (or the precursor of) the Great Basilica. The axial alignment of Gate 1 to the Merchant’s House could suggest the proprietor’s close involvement with channel traffic, although this needs to be reconciled with the apparent abandonment of the buildings themselves. As a result, the Triconch Palace area continued to be occupied by simple buildings, most of which probably belonged to fishermen. These unprepossessing buildings, though, tell only half the story. Indeed, we might be forgiven for assuming from the dwellings that a crushing poverty had overwhelmed Butrint. However, the huge quantities of finds associated with these households, well dated by numerous coins and pottery, lead us to conclude otherwise. Large amounts of imported amphorae as well as fine table wares in ceramics and glass distinguish the rubbish middens. These show Butrint’s trade connections with eastern, primarily Aegean and Asia Minor, and lesser Cretan and Levantine, sources, which are a characteristic feature of early 5th- to late 6th-century trade. Western imports were from relatively close regional sources in this period, comprising primarily those of Tunisia (table wares and amphorae), even when Roman Africa was under Vandal rule, and, from the late 5th century, Apulia (amphorae). Up to AD 550 African and Aegean-Levantine products were roughly on a par, but after 550 the shift to a more eastern-dominated and still abundant supply of imports of all classes is clear (Reynolds 2004; Reynolds in prep.). From the mid 6th century, the character of the area altered: simple graves mainly of infants, some buried in transport amphorae, were now associated with the dwellings. Urban life, it seems, was now taking a new character, in some respects rural, and involved a low-cost habitation form that would be replicated again in the 11th century. In sum, the rhetoric of the great channel-side buildings gave way to modest architectural forms, made with readily available materials. Yet, the material culture indicates the continued importance of traded commodities in the daily life of this community. By the early to mid 7th century Butrint probably amounted to little more than a modest village within the ruins of the Graeco-Roman city. Coins and potsherds of

this era are largely absent from the Triconch Palace area. Of course, this begs the question of what happened to the population. There is no evidence of any conflagration created by invaders. On the contrary, the archaeology suggests the city declined over 50–75 years to become no more than a village in size. The environmental conditions were changing, but not dramatically. The archaeology simply demonstrates a marked desertion of the area. Perhaps a few of its erstwhile citizens had either sought refuge on the acropolis above or else had migrated to either Corfu or adjacent hilltops, away from the vulnerable coast (Bowden 2003a, 195–211; cf. Bowden and Hodges forthcoming; Hodges 2008). A few sherds of 7th- to 9th-century pottery from the Triconch Palace (and Merchant’s House) excavations do not equate to urban continuity. These can be explained as either episodic occupation or, more probably, material brought here from elsewhere in Butrint along with the later levelling deposits. In this respect, the Triconch Palace area appears to be a reflection of most parts of the lower city. The continued occupation of two towers in the Western Defences until c. AD 800 provides an important index for identifying Butrint’s early mid Byzantine occupation. As the intense but limited occupation in these two towers shows, Butrint was not entirely abandoned. Instead, it seems likely that certain well-preserved late antique dwellings continued to be occupied, possibly with associated post-built structures. These, though, we may surmise, were tiny oases within the largely deserted townscape (Bowden and Hodges forthcoming; Hodges 2008; Hodges et al. 2009) (Fig. 11.4). A substantial volume of late 9th- to 10th-century ceramics as well as coins indicates that the use of the Triconch Palace area was changing by the later 9th century. Again, we need to keep in mind that these ceramics and coins might have been brought here with levelling-up soil. By this time, though, Butrint was once more active as a central-place. Coins as well as seals from c. 840–950 from the Vrina Plain household indicate that its administrative epicentre – a minor aristocratic oikos (Magdalino 1984) – lay outside the walled area (Bowden and Hodges forthcoming; Greenslade et al. 2006; Greenslade forthcoming; Hodges 2008). This nucleus appears to have comprised a dwelling (within a former late antique basilica), a possible chapel, associated post-built structures, kilns and a cemetery (Fig. 11.5). A notable feature of this household is the presence of many Otranto 1 transport amphorae. The same amphorae occur in the Triconch Palace area. Mussel processing remains by the gate of the Merchant’s House area is the other significant feature of this area in the 9th to 10th centuries. Given the absence of structures, it is tempting to interpret these remains as evidence of seasonal re-use of the channel-side for fish processing and limited commercial activity of a periodic nature. Was this an emboropanegyri, an occasional market or fair, where local fish products were exchanged for wines and utilitarian items from southern Italy (cf. Dunn 2007, 109 on the importance of Macedonian wetlands for taxable fish products in the

11  Conclusion

Western Defences

Triconch Palace Merchant's House Vrina Plain oikos

Vivari Channel

0

333

Figure 11.4. Butrint and the Vrina Plain showing locations of 7th- to 9th-century occupation traces

200m

Richard Hodges

334

Ruined bath house

Kiln

Entrance

Kiln

Hall

Chapel

Workshops? Cemetery Post hole

0

10m

Burial

Figure 11.5. The Vrina Plain aristocratic oikos of the mid 9th to mid 10th century

middle Byzantine period)? The conjunction of ceramics and coins of this period at Corinth, in the absence of buildings, led Guy Sanders (2003, 397) to this conclusion. More to the point, it is consistent with the growing archaeological corpus of seasonal central-place markets (sometimes misleadingly termed productive sites) from this era in northwest Europe (Loveluck and Tys 2006; Pestell and Ulmschneider 2003; Skre 2008, 337–8). Dagfinn Skre defines central-place markets as ‘aristocratic residences with political, religious and juridical functions for their surrounding territories’ (2008, 337), associated with which is evidence of crafts and inter-regional trade. In Skre’s typology, these are seasonal trading places, different in character from seasonal local markets where long-distance traded goods are rare. He distinguishes these two types from seasonal nodal markets – landing and beach markets connected to long-distance trade networks, and from towns with permanent infrastructure and both long-distance and local trade networks (Skre 2008, 338). The apparent desertion of the (aristocratic) household on the Vrina Plain appears to coincide with the renewal (over several generations) of Butrint as a permanent town (to

adopt Skre’s typology) during the later 10th and early 11th centuries (Fig. 11.6). A new ring of fortifications around the acropolis as well as the refurbishment of the lower ring of defences running along the Vivari Channel are the most apparent elements of this revival (Hodges 2008; cf. Whittow 1995). Judging from a closure wall connecting the acropolis to the Lake Gate as well as the four gates in the north acropolis defences, new emphasis was undoubtedly being placed on the north slope of the acropolis (Andrews et al. 2004, 133–7; Martin 2004, 97–99). Unfortunately, almost no archaeological investigations have taken place. This needs to be borne in mind as we focus once again on our excavations beside the Vivari Channel. The modest reoccupation of the Triconch Palace and Merchant House areas belongs to this town-making exercise. The coins show a spike in activity here just as the coin finds from the Vrina Plain cease. Note should be taken, though, that the volume of Otranto amphorae increased, suggesting that commercial operations simply intensified at the channel-side area. To this period we now ascribe not only the intermittent refurbishment of the late antique channelside fortifications (so-called Medieval 1) (see Hodges 2008)

Gate 5 Lake Gate

Gate 6 Lion Gate Chapel Gate Gate 7 North Gate Gate

Gate

Gate

Gate 4 Great Basilica

Acropolis

West Gate Gate 8 Gate 3

11  Conclusion

Triconch Palace Chapel?

Western Defences Merchant's House

Gate 2 Chapel

Gate 1

0

335

Figure 11.6. Butrint in the 11th century following the construction of major new fortifications

200m

336

Richard Hodges

and a possible cobbled road leading to Gate 1, but at least two post-built dwellings, several graves and perhaps a small chapel (in a lobe of the old triconch reception room) besides the continuity of mussel processing. The precise topography of this area can only be sketched with difficulty. Within the fortification with its late antique gateways, the peristyle of the old Triconch Palace appears to have remained an open space, perhaps for periodic fairs, although this remains a matter of speculation. Significantly, a rare tetarteron of Basil II (976–1025) in mint condition was found here. So, too were two denari of Lucca and a coin from the bishopric of Le Puy, perhaps associated with the First and the Third Crusades respectively. A damaged, but elegantly decorated small bronze cross was also found here (Fig. 10.3). As noted above, this confined space, was filled with tips of rubbish that raised the ground level above the winter water table, making it habitable at least for part of the year. This could be interpreted as part of a huge operation extending as far back as the old forum area, perhaps intended to facilitate renewed urban living. By the beginning of the new millennium, the refuse of commerce is more variegated than that associated with the 9th- to 10th-century oikos from the Vrina Plain. A century later, by the end of the 11th century, the ubiquitous south Italian Otranto amphorae appear to have been eclipsed by increasing numbers of glazed table wares from the Aegean region and even occasional Fatimid bowls from Egypt (Vroom in prep.). This change to the axes of seaborne trade, noted at the Triconch Palace area, appears to be confirmed by a single period site of likely early 12thcentury date on the Vrina Plain, where in an excavation of an isolated household located 500 m north of the old Vrina Plain nucleus, Byzantine glazed wares occur but Otranto amphorae are notably absent (Hodges 2008). As a result of these excavations we may speculate that the two apparently conflicting 12th-century descriptions of Butrint by Ibn al-Idrisi and Benedict of Peterborough may be two sides of the same story. In the mid 12th century alIdrisi describes Butrint as a small, well-populated town with markets (Soustal 2004, 23). The archaeological evidence confirms this: the acropolis and its steep north- and southfacing flanks, if it was terraced as we now believe, together with seasonal fishermen, amounted to a small town. Then again, the castellum desertum described by Benedict in 1191 (Stubbs 1867: 205), as likely as not, tells us that after the Normans expelled the Byzantines in the 1080s, and for long periods controlled the Ionian islands, no Byzantine authority maintained the acropolis castle. This cannot be confirmed because the possible castle overlying the Acropolis Basilica at the east end of the hilltop was comprehensively excavated by Ugolini, whose record no longer exists (Hodges 2008). Similarly, any castle that might have existed at the western end of the hill was obliterated by later medieval construction as well as the making of Ugolini’s new castle-museum in the 1930s. Al-Idrisi’s description may offer a clue to the late medieval occupation of the lower city of Butrint and to the

Triconch Palace area in particular. Butrint’s fortifications were once more refurbished, probably in the 13th century when the town was in the hands of the Epirote Despots, and subsequently strengthened still further under the Angevins (Andrews et al. 2004; Hodges and Molla 2009). But significantly both gates were blocked in the riverside section investigated in the Merchant’s House and Triconch Palace areas. This may offer a clue to why there was no further investment in the infrastructure and housing in the lower city. The absence of dwellings beside the Vivari Channel, notwithstanding ample ceramic finds, must be explained. The urban community, if it existed on any scale, appears to have occupied the north and south slopes of the acropolis, with the main nucleus on the north side. Presently we can only speculate that medieval housing occupied the upslope area behind the Hellenistic wall and possibly extended out across the uppermost tract of the lower skirt of land to ecclesiastical points such as the monastery in the Gymnasium area and the small church beside the Baptistery (cf. Bowden and Përzhita 2004b, 193–9). Beyond, within the Triconch Palace and Merchant’s House areas, occupation appears absent, though immense amounts of later medieval ceramics and sporadic coins attest perhaps to nearby housing or the continued use of this channel-side for allotments. The 14th-century tower added to the city walls was perhaps the most significant new construction in this area after AD 1204. Its primary purpose was, of course, to offer protection to the fish traps and the fishermen, but we may surmise from similar later medieval occupation of the two towers excavated in the Western Defences of Butrint, that the tower provided ample space as a residence. To summarize, the Triconch Palace excavations have revealed the installation of channel-side properties as the town took shape around an enlarged Roman civic centre. But with the decline and abandonment of this civic centre, the channel-side properties seemed to have grown in scale and ambition before their abandonment as prestige residences in the early 5th century (at least in the case of the Triconch Palace). From this moment the dwellings alongside the channel took a more pragmatic form, a situation that persisted into the early 6th century when the fortifications were erected. Although extensive this enceinte lacked the structural coherence and grandiosity of contemporary fortifications at Nicopolis and nearby Onchesmos (Saranda). As Will Bowden (2006, 280) has recently written: Fortifications … were readily co-opted into a narrative of urban revival. They were significant in terms of urban identity and were a commonly used artistic device to depict cities throughout the Roman period (La Rocca 2000). Walls were expensive to build by comparison with other construction projects and moreover required significant authority in order that individual property that lay on the proposed route of a wall could be successfully expropriated.

11  Conclusion

337

Phoenice

SARANDA (Onchesmos) Church of the Forty Martyrs

0

5km

Butrint

Xarra

Figure 11.7. Saranda (Onchesmos) in relation to Butrint

Mursia Taking Butrint’s new defences as a measure of the community’s perceived status, it is possible that Butrint was now eclipsed by Onchesmos where a more coherent, multi-towered enceinte was constructed at the same time as the great sanctuary around the Church of the Forty Martyrs was made above the port (cf. Bowden and Mitchell 2004; Hodges 2006, 218–20; Mitchell 2004) (Figs 11.7 and 11.8). Onchesmos, with the construction of the new hilltop sanctuary of metropolitan proportions, may well have become the pre-eminent centre hereabouts, with Phoenice obtaining modest new defences (perhaps as a town on the coastal road, cf. Bowden 2006) and, interestingly, Corfu remaining undefended. Saranda, it appears, now possessed a spiritual centre to parallel the new shrine at S. Michele on M. Gargano in northern Apulia (Castelfranchi and Mancini 1994; Trotta and Renzulli 2003), and had appropriated Butrint’s Hellenistic fame as a place of pilgrimage. Unlike great administrative towns such as Dyrrhachium or Nicopolis, Butrint was essentially a port – a secure stopover – with a rich hinterland as opposed to the gateway

Malathrea

into a region. The residents of the Triconch Palace area in Çiflik earlier and later Roman times were presumably landowners with fishing rights, whose wealth was contingent upon the e Aitoit regional demand for these local products. We can point to an adequate parallel for this a millennium later when, in the 17th and 18th centuries, Venetian Butrint supplied Corfu with impressive quantities of fish caught in the weirs. So, we may assume that the rhetoric of the Roman town-houses in the 2nd to 5th centuries was based upon a combination of classical residential values and the private trading of fish and agrarian products. By contrast, in the 11th century the possible cobbled street in front of the Merchant’s House gate (1) and refurbishment of the fortifications suggests an economic impetus to develop a riverside quarter. Although not densely inhabited, there appears to have been the will to revive Butrint’s urban topography. By the 13th century, though, this impetus had apparently diminished. Seen through such a prism of connectivity, it is the architectural remains that perhaps appear baffling. Those remains from the Roman period are grand, and indeed

Richard Hodges

338

SARANDA

Proteichisma

Apsidal building Cemetery Basilica Late-antique house Cemetery Building with mosaic

Odeion Warehouse

0

Other buildings

200m

Figure 11.8. Late antique Onchesmos

the inscription in the mosaic of the 4th-century domus presupposes the presence of a grandee. This is not the case, for example, for the two modest 11th-century dwellings excavated here. Perhaps the prosperity of Butrint in the Roman era – best illustrated by its channel-side townhouses – was more evident because in the course of the five centuries between the ages of Augustus and Anastasius it not only rivalled, but possibly eclipsed, the urban centre of Corfu (at Palaiopolis) where little has been found thus far in terms of elite residences. This was certainly not the case in middle and later Byzantine times when Corfu re-established itself as an administrative, and eventually, commercial hub, just off the Balkan coast. As a result, Butrint invariably became its appendage, rather than a centre as such. This relationship merits closer scrutiny. The Triconch Palace excavations have undoubtedly revealed the tentative evolution of an urban quarter, though the story is possibly not quite as we might have envisaged it when we began these excavations in 1994. Certainly,

these excavations have shed light on Byzantine urban history; but here, too, the picture we have unearthed is more episodic and aligned to the changing rhythms of Butrint’s Adriatic Sea connections. As a town, Butrint, it seems, was seldom at stasis and was often a place where a visitor in ancient or medieval times would have marvelled at the amount of construction in progress. In this it resembles modern Albanian ports, like nearby Saranda as we have experienced it during the course of these excavations. This, then, encourages us to conclude that the new Roman colony founded by Caesar then re-founded by Augustus, like the 11th-century mid Byzantine town, were benchmark episodes in a long history in which the town was often rural in character. We might safely conclude that as a Byzantine and later medieval port, Butrint sustained an important role as a central-place, prized for its fish, but as a town it only briefly re-discovered the ambition of its Augustan architects.

Bibliography

Åkerström Hougen, G. 1974. ‘The calendar and hunting mosaics of the Villa of the Falconer in Argos. A study in early Byzantine iconography’, Acta Instituti Atheniensis Regni Sueciae 23: 58–74. Alba Calzado, M. 1997. ‘Ocupación diacrónica del área arqueológica de Morería’, in P. Mateos Cruz, M. Alba Calzado, and J. Márquez Pérez (eds) Mérida: Excavaciones arqueológicas, 1994–1995: 285–315. Mérida, Consorcio Ciudad Monumental Histórico-Artística y Arqueológica de Mérida. Alexander, M. A., S. Besrour and M. Ennaifer 1976. Utique, les mosaiques sans localisation precise et El Alia. Tunis, Institut National d’Archéologie et d’Art. Anamali, S. 1993. ‘Architettura e decorazione tardoantica’, XL corso di cultura sull’arte ravennate e bizantina: colloquio dell’Associazione Italiana per lo Studio e la Conservazione del Mosaico e seminario internazionali sull’Albania dal tardoantico al medioevo, aspetti e problkemi di archeologia e storia dell’arte: 447–75. Ravenna, Edizioni del Girasole. Andreae, B. 2003. Antike Buntmosaiken. Mainz am Rhein, Verlag Philipp von Zabern. Andrews, R., W. Bowden, O. Gilkes and S. Martin 2004. ‘The late antique and medieval fortifications of Butrint’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 126–50. Oxford, Oxbow Books. Arce, J., A. Chavarría and G. Ripoll 2007. ‘The urban domus in late antique Hispania: examples from Emerita, Barcino and Complutum’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity: From Palaces to Shops (Late Antique Archaeology 3.2): 305–36. Leiden – Boston, Brill. Arthur, P. 2004. ‘From vicus to village: Italian landscapes, c. AD 400–1000’, in N. Christie (ed.), Landscapes of Change. Rural Evolutions in Late Antiquity and the Early Middle Ages: 103–34. Ashgate, Aldershot. Arthur, P. 2006. Byzantine and Turkish Hierapolis. An Archaeological Guide. Istanbul, Ege Yayinlari. Augenti, A. 2006. ‘Ravenna e Classe: archeologia di due città tra la tarda Antichità e l’alto Medioevo’, in A. Augenti (ed.), Le città italiane tra la tarda Antichità a l’alto mediovo. Atti del convegno (Ravenna, 26–28 febbraio 2004): 185–217. Florence, All’Insegna del Giglio. Augenti, A. 2008. A tale of two cities. Rome and Ravenna between 7th and 9th century AD, in S. Gasparri (ed.), 774. Ipotesi su una transizione: 175–98. Turnhout, Brepols. Augenti, A., E. Cirelli and D. Marin 2009. ‘Case e Magazzini a Classe fra VII e VIII secolo’, in G. Volpe and P. Favia (eds), Atti del V Congresso Nazionale di Archeologia Nazionale: 138–44. Florence, All’Insegna del Giglio. Baçe, A. and V. Bushati, V. 1989. ‘Vështrim mbi banesën

prehistorike dhe qytetare në Iliri dhe Epir’, Monumentet 1: 5–48. Baldini-Lippolis, I. 2001. La domus tardoantica. Forme e rappresentazioni dello spazio domestico nelle città del Mediterraneo (Collana di studi e scavi del Dipartimento di Archeologia, Università degli studi di Bologna). Imola, University Press Bologna. Baldini Lippolis, I. 2007. ‘Private space in late antique cities: laws and building procedures’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity: From Palaces to Shops (Late Antique Archaeology 3.2): 197–237. LeidenBoston, Brill. Balty, J. C. 1969. ‘L’édifice dit au triclinos’, in J. Balty (ed.), Apamée de Syrie. Bilan des recherches archéologiques 1965–1968: 105–16. Brussels, Centre belge de recherches archéologiques à Apamée de Syrie. Balty, J. C. 1997. ‘Mosaique et architecture domestique dans l’Apamée des Ve et VIe siècles’, in S. Isager and B. Poulsen (eds), Patrons and Pavements in Late Antiquity: 84–110. Odense, Odense University Press. Barbet, A. 1985. La peinture murale Romaine: les styles décoratifs Pompéiens. Paris, Picard. Bardill, J. 1997. ‘The Palace of Lausus and nearby monuments in Constantinople: a topographical study’, American Journal of Archaeology 101: 67–95. Barnes, E. 2003. ‘The dead do tell tales’, in C. K. Williams II and N. Bookidis (eds), Corinth: The Centenary 1896–1996 (Corinth XX): 435–46. Princeton, The American School of Classical Studies at Athens. Bauchhnenss, G. 1976. Jupitergigantensäulen. Stuttgart, Gesellschaft für Vor- und Frühgeschichte in Württemberg und Hohenzollern. Bavant, B. 1989. ‘Cadre de vie et habitat urbain en Italie centrale byzantine (Vie-VIIIe siècles’, Mélanges de l’École française de Rome 101: 465–32. Bavant, B. 2007. ‘Caričin Grad and the changes in the nature of urbanism in the central Balkans in the 6th century’, in A. G. Poulter (ed.), The Transition to Late Antiquity in the Danube and Beyond: 337–73. Oxford, Oxford University Press/British Academy. Beard, M. and J. Henderson 2001. Classical Art: From Greece to Rome. Oxford, Oxford University Press. Bergmann, B. 1991. ‘Painted perspectives of a villa visit: landscape as status and metaphor’, in E. Gazda (ed.), Roman Art in the Private Sphere. New Perspectives on the Architecture and Decor of the Domus, Villa, and Insula: 49–70. Ann Arbor, University of Michigan Press. Bescoby, D. 2007. ‘Geoarchaeological investigation at Roman

340

Bibliography

Butrint, in I. L. Hansen and R. Hodges (eds), Roman Butrint: an Assessment: 95–118. Oxford, Oxbow Books. Bescoby, D. J. forthcoming. ‘Landscape and environmental change: new perspectives’ in R. Hodges, I. L. Hansen and S. Leppard (eds), Butrint 4: The Archaeology and Histories of an Ionian Town. Oxford, Oxbow Books. Bescoby, D., J. Barclay and J. Andrews 2008. ‘Saints and Sinners: a tephrochronology for late antique landscape change in Epirus from the eruptive history of Lipari, Aeolian Islands’, Journal of Archaeological Science 35: 2574–79. Bianchi, G., N. Molla, M. F. Paris and F. Venturini 2007. Archaeological Survey of the Fortifications at Butrint (Albania), unpublished report for Butrint Foundation. Bianchi, G., N. Molla, M. F. Paris and F. Venturini 2009. ‘Politische edilizie e modi di costruire a Butrinto (Albania) nella Tarda Antichità’, in G. Volpe and P. Favia (eds.), Atti del V Congresso Nazionale di Archeologia Medievale: 727–34. Florence, All’Insegna del Giglio. Bieber, M. 1961. The History of the Greek and Roman Theater. Princeton N.J., Princeton University Press. Bonner, G. 1950. Studies in Magical Amulets, chiefly GraecoEgyptian, Ann Arbor and London, University of Michigan Press. Bowden, W. 2001. ‘A new urban élite? Church builders and church-building in late-antique Epirus’, in L. Lavan (ed.), Recent Research in Late-Antique Urbanism (Journal of Roman Archaeology Supplementary Series 42): 61–72. Portsmouth (R. I.), Journal of Roman Archaeology. Bowden, W. 2003a. Epirus Vetus: the archaeology of a Late Antique Province. London, Duckworth. Bowden, W. 2003b. ‘The construction of identities in postRoman Albania’, in L. Lavan and W. Bowden (eds), Theory and Practice in Late Antique Archaeology (Late Antique Archaeology 1): 57–77. Leiden, Brill. Bowden, W. 2006. ‘Procopius’ Buildings and the late antique fortifications of Albania’, in L. Bejko and R. Hodges (eds), New Directions in Albanian Archaeology: Studies Presented to Muzafer Korkuti: 277–86. Tirana, ICAA. Bowden, W. 2007. ‘Nicopolis: the ideology of the late antique city’, in K. Zachos (ed.), Nicopolis II. Proceedings of the Second International Symposium on Nicopolis: 135–49. Preveza, Municipality of Preveza. Bowden, W. 2008. ‘Cristianizzazione e status sociale nell’ Epirus Vetus tardoantico: le evidenze archeologiche’, in G. Cuscito (ed.), La Cristianizzazione dell’Adriatico (Antichità Altoadriatiche 66): 301–32 Trieste, Editreg SAS. Bowden, W., R. Hodges O. Gilkes and K. Lako, 2000. ‘Late Roman Butrint, Albania: survey and excavations, 1994–98’, Archeologia Medievale 27: 241–57. Bowden, W. and R. Hodges 2004. ‘Balkan Ghosts? Nationalism and the question of rural continuity in Albania, in N. Christie (ed.), Landscapes of Change: Rural Evolutions in Late Antiquity and the Early Middle Ages: 195–222. Aldershot, Ashgate Publishing. Bowden, W. and R. Hodges forthcoming. ‘An ‘Ice Age settling on the Roman Empire’: post-Roman Butrint between strategy and serendipity’, in N. Christie (ed.), Urbes Extinctae. Aldershot, Ashgate Publishing. Bowden, W., R. Hodges and K. Lako, with contributions from D. Bescoby, A. Crowson, O. Gilkes, S. Martin, J. Mitchell, L. Përzhita and P. Reynolds 2002. ‘Roman and late antique Butrint: excavations and survey 2000–2001’, Journal of Roman Archaeology 15: 199–229. Bowden, W. and J. Mitchell 2004. ‘The Christian topography

of Butrint’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 104–25. Oxford, Oxbow Books. Bowden, W. and J. Mitchell 2007. ‘The Triconch Palace at Butrint: the Life and Death of a Late Roman Domus’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity. From Palaces to Shops: 455–74. Leiden, Brill. Bowden, W. and L. Përzhita 2004a. ‘Diaporit: excavating the landscape of Roman Epirus’, Journal of Roman Archaeology 17: 413–33. Bowden, W. and L. Përzhita 2004b. ‘The Baptistery’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 176–201. Oxford, Oxbow Books. Brandenburg, H. 1969. ‘Christussymbole in frühchristlichen Bodenmosaiken’, Römische Quartalschrift für christliche Altertumskunde und Kirchengeschichte 64: 74–138. Brandes, W. 1999. ‘Byzantine towns in the seventh and eighth centuries – different sources, different histories’, in G. P. Brogiolo and B. Ward-Perkins (eds), The Idea and Ideal of the Town Between Late Antiquity and the Early Middle Ages: 25–57. Leiden, Brill. Brogiolo, G. P. (ed.) 1994. Edilizia residenziale tra V e VIII secolo. IV Seminario sul Tardoantico e l’Altomedioevo in Italia Centrosettentrionale (Montebarro-Galbiate, 2–4 settembre 1993). Mantova, SAP Società Archeologica S.r.l. Brogiolo, G. P., N. Gauthier and N. Christie (eds) 2000. Towns and Their Territories between Late Antiquity and the Early Middle Ages. Leiden – Boston – Cologne, Brill. Brogiolo, G. P. and B. Ward-Perkins (eds) 1999. The Idea and Ideal of the Town Between Late Antiquity and the Early Middle Ages: 25–57. Leiden-Boston-Cologne, Brill. Brown, P. 2002. Poverty and Leadership in the Later Roman Empire. Hanover and London, University Press of New England. Buckwald, H. 2007. ‘Byzantine Town planning – does it exist?’, in M. Grünbart, E. Kislinger, A. Muthesias and D. Stathakopoulos (eds) Material Culture and Well-Being in Byzantium (400– 1453): 57–74. Vienna, Verlag der Österreichischen Akademie der Wissenschaften. Budina, D. 1988. ‘Butrinti pararomak’, in N. Ceka (ed.), Butroti: Permbledhje Studimesh: 6–107. Tirana, Akademia e Shkencave e RPSSH Qëndra e Kërkimeve Arkeologjike. Cain, H.-U. 1988. ‘Chronologie, Ikonographie und Bedeutung der römischen Maskenreliefs’, Bonner Jahrbücher 188: 107–221. Cantino-Wataghin, G. 1999. ‘The ideology of urban burials’, in G. P. Brogiolo and B. Ward-Perkins (eds), The Idea and Ideal of the Town Between Late Antiquity and the Early Middle Ages: 147–80. Leiden – Boston – Cologne, Brill. Carettoni, G. 1983. Das Haus des Augustus auf dem Palatin. Mainz am Rhein, Verlag Philipp von Zabern. Carver, M. 1993. Arguments in Stone. Archaeological Research and the European Town in the First Millennium. Oxford, Oxbow. Castelfranchi, M. F. and R. Mancini 1994. ‘Il culto di San Michele in Abruzzo e Molise dalle origini all’altomedioevo (secoli V–XI)’, in C. Carletti and G. Otranto (eds), Culto e insediamenti micaelici nell’Italia medridionale fra tarda antichità e medioevo. Atti del convegno internazionale di Monte Sant’Angelo 18–21 November 1992: 507–51. Bari, EdiPuglia. Ceka, N. 2005. The Illyrians to the Albanians, Tirana, Mijeni. Cerova, Y. 1993. ‘Elemente arkitektonike të Skampis në shek. VI’, Iliria 23 (1–2): 273–84.

Bibliography Cesano, S. L. 1932. ‘Monetazione e circolazione sul suolo dell’antica Albania’, Atti e Memorie dell’Istituto Italiano di Numismatica 2: 47–98. Christie, N. 2001. ‘War and order: urban remodelling and defensive strategy in late Roman Italy’, in L. Lavan (ed.), Recent Research in Late-antique Urbanism (Journal of Roman Archaeology Supplementary Series 42): 106–22. Portsmouth (R. I.), Journal of Roman Archaeology. Christie, N. and S. T. Loseby (eds) 1996. Towns in Transition. Urban Evolution in Late Antiquity and the Early Middle Ages. Aldershot, Scolar Press. Chrysos, E. 1981. ‘Συμβολή στήν Ίστορία τής ‘Ηπειρου κατά τήν πρωτοβηζαντινή εποχή (Δ-ΣΤ αι)’, Ήπειροτικά Χρονικά 23: 6–111. Crowson, A. 2007. Butrinti Venecian/Venetian Butrint. Rome, Società Tipografica Romana. Crowson, A., and O. Gilkes 2007. ‘The archaeology of the Vrina Plain: an assessment’ in R. Hodges and I. L. Hansen (eds), Roman Butrint: an Assessment: 119–64. Oxford, Oxbow Books. Curta, F. 2004. ‘Barbarians in Dark-Age Greece: Slavs or Avars’, in T. Stepanov and V. Vachkova (eds), Civitas Divino-Humana. In Honorem Annorem LX Georgii Bakalov: 513–50. Sofia. Centăr za izsledvaniia na bălgarite Tangra TanNakRa IK. Deichmann, F. W. 1954. ‘Cella trichora’, in Reallexikon für Antike und Christentum Band 2: col. 944–54. Stuttgart, Anton Hiersemann. De Rijk, P., in prep. ‘Metal working at Butrint. Analysis of slag and wasters’. Dunbabin, K. M. D. 1996. ‘Convivial spaces: dining and entertainment in the Roman villa’, Journal of Roman Archaeology 9: 66–80. Dunbabin, K. M. D. and M. W. Dickie 1983. ‘Invidia rumpantur pectora: the iconography of Phthonos/Invidia in Greco-Roman art’, Jahrbuch für Antike und Christentum 26: 7–37. Dunn, A. 2007. ‘Rural producers and markets: aspects of the archaeological and historiographic problem’, in M. Grünbart, E. Kislinger, A. Muthesias and D. Stathakopoulos (eds), Material Culture and Well-Being in Byzantium (400–1453): 101–10. Vienna, Verlag der Österreichischen Akademie der Wissenschaften. Ellis, S. P. 1985. ‘The ‘Palace of the Dux’ at Apollonia and related houses’, in G. Barker, J. Lloyd and J. Reynolds (eds), Cyrenaica in Antiquity (British Archaeological Reports International Series 236): 15–27. Oxford, British Archaeological Reports. Ellis, S. P. 1988. ‘The end of the Roman house’, American Journal of Archaeology 92: 565–76. Ellis, S. P. 1991. ‘Power, architecture and décor: how the late Roman aristocrat appeared to his guests’, in E. K. Gazda (ed.), Roman Art in the Private Sphere. New Perspectives on Architecture and Decor of the Domus, Villa, and Insula: 117–34. Ann Arbor, University of Michigan Press. Ellis, S. P. 1997a. ‘Late-antique dining: architecture, furnishing and behaviour’, in R. Laurence and A. Wallace-Hadrill (eds), Domestic Space in the Roman World: Pompeii and Beyond (Journal of Roman Archaeology Supplementary Series 22): 41–51. Portsmouth (R. I.), Journal of Roman Archaeology. Ellis, S. P. 1997b. ‘Late antique houses in Asia Minor’, in S. Isager and B. Poulsen (eds), Patrons and Pavements in Late Antiquity: 38–50. Odense, Odense University Press. Ellis, S. P. 2000. Roman Housing, London, Duckworth. Ellis, S. P. 2006. ‘Middle class houses in late antiquity’, in W. Bowden, A. Gutteridge and C. Machado (eds), The Social and Political Archaeology of Late Antiquity (Late Antique

341

Archaeology 3.1): 413–37. Leiden, Brill. Ellis, S. P, 2007. ‘Late antique housing and the uses of residential buildings: an overview’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity: From Palaces to Shops (Late Antique Archaeology 3.2): 1–22. Leiden – Boston, Brill. Engemann, J. 1975. ‘Zur Verbreitung magische Übelabwehr in der nichtchristlichen und christlichen Spätantike, Jahrbuch für Antike und Christentum 18: 22–48. Fenton, T. W., J. Beatrice, L. Jenny, J. Wankmiller, M. Mutolo, C. Rauzi, and D. Foran in prep. ‘The Human Skeletons from the Triconch Palace and Merchant’s House’. Flint, V. 1991. The Rise of Magic in Early Medieval Europe. Princeton, Princeton University Press. Foucher, L. 1965. La Maison des Masques à Sousse, Fouilles 1962– 1963. Tunis, Imp. du Secrétariat aux Affaires Culturelles. Frankfurter, D. 2004. ‘The binding of antelopes: a Coptic frieze and its Egyptian religious context’, Journal of Near Eastern Studies 63: 97–109. Frantz, A. 1988. The Athenian Agora Volume XXIV: Late Antiquity: A.D. 267–700. Princeton, The American School of Classical Studies at Athens. Galetti, P. 1998. Abitare nel Medioevo. Forme e vicende dell’insediamento rurale nell’Italia altomedioevale. Florence, Le Lettere. Galetti, P. 2001. Uomini e case nel Medioevo tra occidente e oriente. Roma/Bari, Laterza. Gelichi, S. 2008. A Town Through the Ages. The 2006–2007 Archaeological Project in Stari Bar. Florence, Insegna del Giglio. Gell, A. 1992. ‘The technology of enchantment and the enchantment of technology’, in J. Coote and A. Shelton (eds), Anthropology, Art and Aesthetics: 40–63. Oxford, Clarendon Press. Gell, A. 1996. ‘Vogel’s net: traps as artworks and artworks as traps’, Journal of Material Culture 1, no. 1: 15–38. Gell, A. 1998. Art and Agency: An Anthropological Theory. Oxford, Clarendon Press. Germain, S. 1969. Les mosaïques de Timgad: étude descriptive et analytique. Paris, Editions du Centre National de la Recherche Scientifique. Ghedini, F. and S. Bullo 2007. ‘Late antique domus of Africa Proconsularis: structural and decorative aspects’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity: From Palaces to Shops (Late Antique Archaeology 3.2): 337–66. Leiden – Boston, Brill. Ghislanzoni, E. 1962. La villa romana in Desenzano. Milan, Treccani. Gilkes, O. and K. Lako 2004. ‘Excavations at the Triconch Palace’ in R. Hodges, W. Bowden and K. Lako (eds) Byzantine Butrint. Excavations and Surveys 1994–99: 151–75. Oxford, Oxbow Books. Gilkes, O., A. Crowson, R. Hodges, K. Lako and J. Vroom 2002. ‘Medieval Butrint: excavations at the Triconch Palace 2000 and 2001’, Archeologia Medievale 29: 7–12. Goodchild, R. G. 1976. ‘The Palace of the Dux’, in J. Humphrey (ed.), Apollonia, the Port of Cyrene: Excavations Conducted by the University of Michigan, 1965–1967 (Libya Antiqua Suppl. IV): 245–65. London, Society of Libyan Studies. Greenslade, S. forthcoming. ‘The Vrina Plain settlement between the 1st–13th centuries AD’, in R. Hodges, I. L. Hansen and S. Leppard (eds), Butrint 4: The Archaeology and Histories of an Ionian Town. Oxford, Oxbow Books. Greenslade, S., R. Hodges, S. Leppard and J. Mitchell 2006. ‘Preliminary report on the early Christian Basilica on the Vrina plain, Albania’, Archeologia Medievale 33: 397–408.

342

Bibliography

Guest, P., J. Mitchell, E. Nallbani and S. Gongecaj 2004. ‘The small finds and coins’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 293–304. Oxford, Oxbow Books. Guidobaldi, F. 1986. ‘L’edilizia abitativa unifamiliare nella Roma tardoantica’, in A. Giardina (ed.), Societá romana e impero tardoantico: 165–235. Rome, Edizioni Laterza. Guimier-Sorbets, A. M. 1983. ‘Le méandre à pannatons de clef dans la mosaïque romaine’, in Mosaïque: Receuil d’Hommage à Henri Stern: 195–213. Paris, Éditions Recherches sur les Civilisations. Gutteridge, A., A. Hoti and H. Hurst 2001. ‘The walled town of Dyrrachium (Durrës): settlement and dynamics’, Journal of Roman Archaeology 14: 388–410. Haldon, J. 1999. ‘The idea of the town in the Byzantine empire’, in G. P. Brogiolo and B. Ward-Perkins (eds), The Idea and Ideal of the Town Between Late Antiquity and the Early Middle Ages: 1–23. Leiden – Boston – Cologne, Brill. Hales, S. 2003. The Roman House and Social Identity. Cambridge, Cambridge University Press. Hansen, I. L. 2009. Butrinti Helenistik dhe Romak/Hellenistic and Roman Butrint. Rome, Società Tipografica Romana. Hakim, B. S. 2001. ‘Julian of Ascalon’s treatise of design and construction rules from sixth-century Palestine’, Journal of the Society of Architectural Historians 60: 4–25. Heather, P. 1990. ‘Senators and Senates’, in A. Cameron and P. Garnsey (eds), Cambridge Ancient History Vol XIII. The Late Empire A.D. 337–425: 184–210. Cambridge, Cambridge University Press. Hellenkemper, H. 1987. ‘Die byzantinische Stadtmauer von Nikopolis in Epeiros. Ein kaiserlich Bauauftrag des 5. oder 6. Jahrhundert’, in E. Chrysos (ed.), Nicopolis I. Proceedings of the first international symposium on Nicopolis: 243–51. Preveza, Municipality of Preveza. Hernandez, D. R. and Çondi, Dh. 2008. ‘The Roman forum at Butrint (Epirus)’, Journal of Roman Archaeology 21: 275–92. Hidri, S. 1983. ‘Resultate germimesh ne baziliken e Arapajt (1980–1982)’, Iliria 13 (1): 233–9. Hodges, R. 2006. Eternal Butrint. A UNESCO World Heritage Site in Albania. London, General Penne Publishing. Hodges, R. 2007. Saranda. Ancient Onchesmos. A Short History and Guide. Tirana, Mjgeni. Hodges, R. 2008. Shkëlqimi dhe rënia e Butrintit bizantin/The Rise and Fall of Byzantine Butrint. Rome, Società Tipografica Romana. Hodges, R., W. Bowden, O. Gilkes and K. Lako 2004. ‘Introduction’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 1–19. Oxford, Oxbow Books. Hodges, R., S. Kamani, M. Logue and J. Vroom 2009. ‘The sack of Butrint, c. AD 800’, Antiquity 83 (320) (http://antiquity. ac.uk/projgall/hodges/) Hodges, R. and N. Molla 2009. ‘The Fortifications at Butrint’, Expedition 52: 18–20. Hodges, R., G. Saraçi, W. Bowden, P. Chiles, O. Gilkes, K. Lako, A. Lane, S. Martin, J. Mitchell, J. Moreland, S. O’Hara, M. Pluciennik and L. Watson 1997. ‘Late-antique and Byzantine Butrint: interim report on the port and its hinterland (1994– 95)’, Journal of Roman Archaeology 10: 207–34. Hodges, R., W. Bowden and K. Lako (eds) 2004. Byzantine Butrint. Excavations and Surveys 1994–99. Oxford, Oxbow Books. Hodges, R. and J. Vroom 2007. ‘Late antique and early medieval ceramics from Butrint’, in S. Gelichi and C. Negrelli (eds) La

circolazione delle ceramiche nell’Adriatico: 375–88. Mantua, Societá Archeologica Padana. Hodges, R., S. Kamani, M. Logue and J. Vroom, forthcoming. ‘The Western Defences’, in R. Hodges, I. L. Hansen and S. Leppard (eds), Butrint 4: The Archaeology and Histories of an Ionian Town. Oxford, Oxbow Books. Hounslow, M. and A. Chepstow-Lusty 2004. ‘Holocene environmental change at Butrint: a preliminary evaluation of alluvial sediments using archaeomagnetic dating’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 396–97. Oxford, Oxbow Books. Hurst, H. and S. P. Roskams 1984. Excavations at Carthage i/1. The Avenue du President Habib Bourguiba: the site and finds other than pottery. Sheffield, The British Academy. Ivison, E. 2000. ‘Urban renewal and imperial revival in Byzantium (730–1025)’, Byzantinische Forschungen 26: 1–46. Ivison, E. 2007. ‘Amorium in the Byzantine Dark Ages (seventh to ninth centuries)’, in J. Henning (ed.), Post-Roman Towns, Trade and Settlement in Europe and Byzantium, Vol. 2 Byzantium, Pliska, and the Balkans: 25–59. Berlin, Walter de Gruyter. Karagiorgou, O. 2001. ‘Demetrias and Thebes: the fortunes and misfortunes of two Thessalian port cities’, in L. Lavan (ed.) Recent Research in Late-antique Urbanism (Journal of Roman Archaeology Supplementary Series 42): 182–215. Portsmouth (R. I.), Journal of Roman Archaeology. Karaiskaj, Gj. 1980. ‘Muret rrethuese të Butrintit në mesjetë (shek. VIII–XV)’, Monumentet 20: 5–36. Karaiskaj, Gj. 2009. The Fortifications of Butrint. London/Tirana, Butrint Foundation. Kitzinger, E. 1970. ‘The threshold of the holy shrine: observations on floor mosaics at Antioch and Bethlehem’, in P. Granfield and J. A. Jungmann (eds), Kyriakon. Festschrift J. Quasten: 639–47. Münster, Verlag Aschendorff. Kitzinger, E. 1993. ‘Interlace and icons: form and function in early Insular art’, in R. M. Spearman and J. Higgitt (eds) The Age of Migrating Ideas: Early Medieval Art in Northern Britain and Ireland: 3–15. Stroud, Alan Sutton. Kraus, T. and L. von Matt 1975. Pompeii and Herculaneum: The Living Cities of the Dead. New York, Harry N. Abrams. Krautheimer, R. 1986. Early Christian and Byzantine Architecture. Harmondsworth, Penguin. Ladstätter, S. and A. Pülz 2007. ‘Ephesus in the late Roman and early Byzantine period: changes in its urban character from the third to the seventh century AD’, in A. G. Poulter (ed.) The Transition to Late Antiquity in the Danube and Beyond: 391–433. Oxford, Oxford University Press/British Academy. Lafon, X. 2001. Villa maritima. Recherches sur les villas littorales de l’Italie romaine (Coll. EFR 307). Rome, École française de Rome. Lako, K. 1981. ‘Rezultatet e gërmineve arkeologjike në Butrint në vitet 1975–76’, Iliria 11 (1): 93–154. Lako, K. 1984. ‘Kështjella e Onhezmit’, Iliria 14 (2): 153–205. Lako, K. 1990. ‘Gërmimet arkeologjike të vitit 1990: Butrint’, Iliria 20 (1): 269–70. Lako, K. 1991 ‘Bazilika Paleokristiane e Onhezmit’, Iliria 21 (1–2): 123–86. Lancha, J. 1981a. Receuil général des mosaïques de la Gaul. Xe supplément à Gallia. III. Province de Narbonnaise. Paris, Éditions du Centre de la Recherche Scientifique. Lancha, J. 1981b. ‘Les mosaïques de la rue des colonnes’, Fondations Eugène Piot. Monument et mémoires publiés par l’Academie des inscription et Belles-Lettres 64, 84–140.

Bibliography La Rocca, C. 2000. ‘L’affresco con veduta di città dal colle Oppio’, in E. Fentress (ed.), Romanization and the City: creation, Transformations, and Failures (Journal of Roman Archaeology Supplementary Series 38): 57–71. Portsmouth (R.I.), Journal of Roman Archaeology. Lavagne, H. 2005. Receuil général des mosaïques de la Gaul. Xe supplément à Gallia. III. Province de Narbonnaise, 3., Partie Sud-Est. Paris, Éditions du Centre de la Recherche Scientifique. Lavan, L. 2001. Recent Research in Late-antique Urbanism (Journal of Roman Archaeology Supplementary Series 42). Portsmouth (R. I.), Journal of Roman Archaeology. Lavan, L, L. Özgenel and A. Sarantis (eds) 2007. Housing in Late Antiquity: From Palaces to Shops (Late Antique Archaeology 3.2). Leiden – Boston, Brill. Lavin, I. 1962. ‘The House of the Lord. Aspects of the role of palace triclinia in the architecture of Late Antiquity and the early Middle Ages’, Art Bulletin 44: 1–27. Lazimi, L. 1986. ‘Vështrim urbanistiko-arkitetonik i fshatit Dhërmi-Vlorë’, Monumentet 2: 117–140. Leake, W. M. 1835. Travels in Northern Greece. London, J. Rodwell. Lewit, T. 2005. ‘Bones in the bathhouse: re-evaluating the notion of squatter occupation in 5th–7th century villas’, in G. P. Brogiolo, A. Chavarría Arnau and M. Valenti (eds), Dopo la fine delle ville: evoluzione nelle campagne tra VI e IX secolo: 251–62. Mantua, SAP Società Archeologica s.r.l. Liebeschuetz, J. H. W. G. 2001. The Decline and Fall of the Roman City. Oxford, Oxford University Press. Lightfoot, C. and M. Lightfoot 2007. Amorium. An Archaeological Guide. Istanbul, Homer Kitabevi. Ling, R. 1991. Roman Painting. Cambridge, Cambridge University Press. Ling, R. 1998. Ancient Mosaics. London, British Museum Press. Loveluck, C. and D. Tys 2006. ‘Coastal societies, exchange and identity along the Channel and southern North Sea shores of Europe, AD 600–100’, Journal of Maritime Archaeology 1: 140–69. Magdalino, P. 1984. ‘The Byzantine aristocratic oikos’, in M. Angold (ed.), The Byzantine Aristocracy IX to XII centuries: 112–37. Oxford, British Archaeological Reports. Maguire, E. D., H. P. Maguire and M. J. Duncan-Flowers 1989. Art and Holy Powers in the Early Christian House. Urbana and Chicago, University of Illinois Press. Maguire, H. 1997. ‘Magic and money in the early Middle Ages’, Speculum 72: 1037–54. Manacorda, D. 2001. Crypta Balbi. Archeologia e storia di un paesaggio urbano. Milan, Electa. Manière-Lévêque, A.-M. 2007. ‘The house of the Lycian acropolis at Xanthos’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity: From Palaces to Shops (Late Antique Archaeology 3.2): 475–94. Leiden-Boston, Brill. Marano, Y. A. 2007. ‘Domus in qua manebat Episcopus: Episcopal residences in northern Italy during Late Antiquity (4th to 6th c. AD)’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity. From Palaces to Shops (Late Antique Archaeology 3.2): 97–130. Leiden-Boston, Brill. Martin, S. 2001. The Butrint Management Plan 2000–2005. London, Butrint Foundation. Martin, S. 2004. ‘The topography of Butrint’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 76–104. Oxford, Oxbow Books. Meksi, A. 1988. ‘Ndërtimet e kultit në Butrint’, in N. Ceka (ed.),

343

Butroti: Permbledhje Studimesh: 199–210. Tirana, Akademia e Shkencave e RPSSH Qëndra e Kërkimeve Arkeologjike. Michel, S. 2001. Die magischen Gemmen im Britischen Museum, 2 vols. London, British Museum Press. Mitchell, J. 2004a. ‘The archaeology of pilgrimage in late antique Albania: the Basilica of the Forty Martyrs’, in W. Bowden, L. Lavan and C. Machado (eds) Recent Research on the Late Antique Countryside (Late Antique Archaeology 2): 145–86. Leiden, Brill. Mitchell, J. 2004b. ‘The mosaic pavements of the Baptistery’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 202–18. Oxford, Oxbow Books. Mitchell, J. 2006. ‘Strategies for salvation: the triconch church at Antigoneia and its mosaic pavement’, in L. Bejko and R. Hodges (eds), New Directions in Albanian Archaeology: Studies Presented to Muzafer Korkuti: 261–76. Tirana, ICAA. Mitchell, J. 2007. ‘Keeping the demons out of the house: the archaeology of apotropaic strategy and practice in late antique Butrint and Antigoneia’, in L. Lavan, E. Swift and T. Putzeys (eds) Objects in Context and Use: Material Spatiality in Late Antiquity (Late Antique Archaeology 5): 273–310. Leiden – Boston, Brill. Morvillez, E. 1995. ‘Les salles des réception triconques dans l’architecture domestique de l’antiquité tardive’, Histoire de l’Art 31: 15–26. Molla, N., M. F. Paris and F. Venturini forthcoming. ‘Material boundaries: the late antique defences’, in R. Hodges, I. L. Hansen and S. Leppard (eds), Butrint 4: The Archaeology and Histories of an Ionian Town. Oxford, Oxbow Books. Muçaj, S. 1990. ‘Sistemi fortifikues i qytetit të Bylisit në antikitetin e vonë’, Iliria 20 (1): 169–200. Onians, J. 1994. ‘“I wonder …” A short history of amazement’, in J. Onians (ed.), Sight and Insight. Essays on Art and Culture in Honour of E. H. Gombrich at 85: 11–33. London, Phaidon. Ostrogorsky, G. 1959, ‘Byzantine cities in the Middle Ages’, Dumbarton Oaks Papers 13: 47–66. Özgenel, L. 2007. ‘Public use and privacy in late antique houses in Asia Minor: the architecture of spatial control’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity: From Palaces to Shops (Late Antique Archaeology 3.2): 239–81. Leiden-Boston, Brill. Pallas, D. 1971. ‘Epiros’, in M. Restle and K. Wessel (eds), Real Lexikon zur Byzantinischen Kunst 2: 207–334. Stuttgart, Anton Hiersemann. Papadopoulou, B. 2007. ‘Βασιλική Αλκίσωνος. Η ανασκαφική έρευνα των τελευταίων ετών’, in K. Zachos (ed.), Nicopolis B. Proceedings of the Second International Nicopolis Symposium (11–15 September 2002): 609–37. Preveza, Actia Nicopolis Foundation. Papadopoulou, P. 2005. ‘Un trésor monétaire des trachéa de billon du XIIe siècle provenant de Butrint (Albanie)’ in C. Alfaro, C. Marcos and P. Otero (eds), XIIIe Congreso Internacional de Numismatica, Madrid 2003: Actas – Proceedings – Actes I: 1231–1236. Madrid, Ministerio de cultura. Papadopoulou, P. in prep. ‘The middle and late Byzantine, medieval and early modern coins from the Triconch Palace’ Perdrizet, P. 1922. Negotium Perambulans in Tenebris. Étude de démonologie gréco-orientale. Strasbourg and London, Istra, Oxford University Press. Pestell, T. and Ulmschneider, K. (eds) 2003. Markets in Early Medieval Europe. Trading and ‘Productive’ Sites, 650–850. Windgather Press, Macclesfield.

344

Bibliography

Petrides, P. 1997. ‘Delphes dans l’antiquité tardive: premiére approache topographique et céremologique’, Bulletin de Correspondance Héllenique 121: 681–95. Piccirillo, M. 1993. The Mosaics of Jordan (ACOR Research Publications 1). Amman, American Center of Oriental Research. Pirazzoli, P. A., J. Laborel and S. C. Stiros 1996. ‘Earthquake clustering in the Eastern Mediterranean during historical times’, Journal of Geophysical Research – Solid Earth 101: 6,083–97. Polci, B. 2000. Palace and Hall in the Mediterranean Basin between Late Antiquity and the Early Middle Ages, Unpublished PhD Thesis (University of East Anglia 2001). Polci, B. 2003. ‘Some aspects of the transformation of the Roman domus between late antiquity and the early middle ages’, in L. Lavan and W. Bowden (eds), Theory and Practice in Late Antique Archaeology (Late Antique Archaeology 1): 79–112. Leiden, Brill. Popovič, V. 1984. ‘Byzantins, slaves et autochones das les provinces de Prévalitene et Nouvelle Épire’, in V. Popovič (ed.), Villes et peuplement dans l’Illyricum protobyzantin. Actes du colloque organisé par l’École Française de Rome (Rome 12–14 mai 1982): 181–243. Rome, École Française de Rome. Potter, T. 1995. Towns in Late Antiquity: Iol Caesarea and its Context. Oxford, Oxbow Books. Poulsen, B. 1997. ‘The city personifications in the late Roman villa in Halikarnassos’, in S. Isager and B. Poulsen (eds), Patrons and Pavements in Late Antiquity: 9–23. Odense, Odense University Press. Prentice, W. K. 1906. ‘Magical formulae on lintels of the Christian period in Syria’, American Journal of Archaeology 10: 137–50. Prudhomme, R. 1985. Le Décor Géométrique de la Mosaïque Romaine: Répertoire Graphique et Descriptif des Compositions Linéaires et Isotropes. Paris, Picard. Prudhomme, R. 2002. Le Décor Géométrique de la Mosaïque Romaine II Répertoire Graphique et Descriptif des Décors Centrés. Paris, Picard. Purcell, N. 1995. ‘The Roman villa and the landscape of production’, in T. J. Cornell and K. Lomas (eds), Urban Society in Roman Italy: 151–179. London – New York, Routledge. Rautman, M. L. 1995. ‘A late Roman town house in Sardis’, in E. Schwertheim (ed.), Forschungen in Lydien (Asia Minor Studien 17): 49–66. Bonn, Rudolph Habelt. Reynolds, P. 2002. ‘The pottery’, in W. Bowden, R. Hodges and K. Lako, ‘Roman and late-antique Butrint: excavations and survey 2000–2001’, Journal of Roman Archaeology 15: 200–29, at 221–9. Reynolds, P. 2004. ‘The Roman pottery from the Triconch Palace’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 224–69. Oxford, Oxbow Books. Reynolds, P. in prep. Butrint: the Roman and Late Antique Pottery. Rich, J. (ed.) 1992. The City in Late Antiquity. London, Routledge. Riza, E. 1978. ‘Banesa popullore në fshatrat e rrethit të Vlorës’, Monumentet 15–16: 185–219. Robert, C, 1911. Die Masken der neueren attischen Komoedie. Halle a S., Max Niemeyer. Saliou, C. 1996. Le traité de’urbanisme de Julian d’Ascalon. Droit et architecture en Palestine au Ve siècle. Paris, De Boccard. Salzmann, D. 1991. ‘Mosaiken und Pavimente in Pergamon. Vorbericht der Kampagnen 1989 und 1990’’, Archäologische Anzeiger: 433–56.

Samimi, G. 1978. ‘Të dhëna mbi arkitekturën e banesës popullore në krahinën e Skraparit’, Monumentet 15–16: 139–61. Sanders, G. 2003. ‘Recent developments in the chronology of Byzantine Corinth’, in C. K. Williams II and N. Bookidis (eds), Corinth: The Centenary 1896–1996 (Corinth XX): 385–400. Princeton, The American School of Classical Studies at Athens. Sansoni, U. 1998. Il nodo di Salomone: simbolo e archetipo d’alleanza. Milan, Electa. Santangeli Valenzani, R. 2004. ‘Abitare a Roma nell’alto medioevo’, in L. Paroli and L. Venditelli (eds), Roma dall’antichità al Medioevo II. Contesti tardoantichi ed altomedievali: 41–59. Milan, Electa. Santoro, S., A. Hoti and B. Sassi 2008. ‘L’anfiteatro di Durazzo. Studi e scavi 2004–2005’, Annuario della Scuola Archeologica Italiana di Atene LXXXIII, s.III,5, t.II, 2005: 717–806. Saradi, H. 1998. ‘Privatization and subdivision of urban properties in the early Byzantine centuries: social and cultural implications’, Bulletin of the American Society of Papyrologists 35 (1–2): 17–43. Sauvaget, J. 1934. ‘Le plan de Laodicée-sur-Mer’, Bulletin d’Etudes Orientales 4: 81–114. Schachner L. A. 2006. ‘Social life in late antiquity: a bibliographic essay’, in W. Bowden, A. Gutteridge and C. Machado (eds), Political and Social Life in Late Antiquity (Late Antique Archaeology 3.1): 41–93. Leiden – Boston, Brill. Scott, E. 1999. The Archaeology of Infancy and Infant Death. Oxford, Archaeopress. Scranton, R. L. 1957. Corinth, vol. XVI. Medieval Architecture in the Central Area of Corinth. Princeton, The American School of Classical Studies at Athens. Sebastiani, A. 2008. ‘Butrinto. Relazione preliminare dello scavo presso il Pozzo di Iunia Rufina’, Archeologia Medievale 35: 243–61. Sebastiani, A. forthcoming. ‘The medieval church and cemetery at the Well of Junia Rufina’, in R. Hodges, I. L. Hansen and S. Leppard (eds), Butrint 4: The Archaeology and Histories of an Ionian Town. Oxford, Oxbow Books. Sigalos, L. 2004. ‘Middle and late Byzantine houses in Greece (tenth to fifteenth centuries)’, in K. Dark (ed.), Secular Buildings and the Archaeology of Everyday Life in the Byzantine Empire: 53–81. Oxford, Oxbow. Skinner, P. 1994. ‘Urban communities in Naples, AD 900–1050’, Papers of the British School at Rome 62: 279–300. Skinner, P. 1998, ‘Room for tension: urban life in Apulia in the eleventh and twelfth centuries’, Papers of the British School at Rome 66: 159–76. Skre, D. 2008, ‘Post-substantivist towns and trade AD 600–1000’, in D. Skre (ed.), Means of Exchange. Dealing with silver in the Viking Age: 327–41. Aarhus, Aarhus University Press. Small, A. M., and R. J. Buck (eds) 1994. The Excavations of San Giovanni di Ruoti, Vol. 1., The Villas and their Environment. (Phoenix Supplementary vol. 33). Cheektowaga, University of Toronto Press. Smith, R. R. R. and C. Ratté 1998. ‘Archaeological research at Aphrodisias in Caria, 1996’, American Journal of Archaeology 102: 225–50. Sodini, J.-P. 1984. ‘L’habitat urbain en grèce à la veille des invasions’, in V. Popovič (ed.), Villes et peuplement dans l’Illyricum protobyzantin. Actes du colloque organisé par l’École Française de Rome (Rome 12–14 mai 1982): 341–96. Rome, École Française de Rome. Sodini, J.-P. 1995. ‘Habitat de l’antiquité tardive’, Topoi 5: 153–61.

Bibliography Sodini, J.-P. 1997. ‘Habitat de l’antiquité tardive (2)’, Topoi 7: 435–57. Sodini, J.-P. 2007. ‘The transformation of cities in late antiquity within the provinces of Macedonia and Epirus’, in A. G. Poulter (ed.), The Transition to Late Antiquity in the Danube and Beyond: 311–36. Oxford, Oxford University Press/British Academy. Soustal, P. 2004. ‘The historical sources for Butrint in the Middle Ages’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 20–26. Oxford, Oxbow Books. Soren, D. and N. Soren 1999. A Roman Villa and a Late-Roman Infant Cemetery: Excavation at Poggio Gramignano, Lugnano in Teverina (Bibliotheca Archaeologica 23). Rome, “L’Erma” di Bretschneider. Staffa, A. 1991. ‘Scavi nel centro storico di Pescara. 1: primi elementi per una riconstruzione dell’assetto antico ed altomedievale dell’abitato di ‘Ostia Aterna-Aternum’, Archeologia Medievale 17: 201–57. Staffa, A. 2006. ‘I centre urbani dell’Abruzzo adriatico fra tarda antichità ed altomedioevo’, in A. Augenti (ed.), Le città italiane tra la tarda antichità e l’alto medioevo: 345–476. Florence, All’Insegna del Giglio. Stern, H. 1967. Receuil général des mosaïques de la Gaul (Xe supplément à Gallia. II). Province de Lyonnaise I, Lyon-Paris, Éditions du Centre National de la Recherche Scientifique. Stevens, W. O. 1977. ‘The cross in the life and literature of the Anglo-Saxons’, in W. O. Stevens and A. S. Cook, The AngloSaxon Cross: 1–109. Hamden, Conn., Archon Books. Stollmayer, I. 1999. ‘Spätantike Trikonchoskirchen – ein Baukonzept?’, Jahrbuch für Antike und Christentum 42: 116–57. Stubbs, W. (ed.) 1867. Gesta Regis Henrici II. Benedicti Abbatis. The Chronicle of the Reigns of Henry II. and Richard I. (1169–1192) Known Commonly under the Name of Benedict of Peterborough (RBS 49). London, Longman and Co. (reprinted 1965). Suli, L. 1985. ‘Arkitektura e banesës popullore në rrëzomën e Delvinës’, Monumentet 1: 105–35. Thébert, Y. 1987. ‘Private life and domestic architecture in Roman Africa’, in P. Veyne (ed.) A History of Private Life from Pagan Rome to Byzantium: 313–410. Cambridge (M. A.), Harvard University Press. Trilling, J. 1995. ‘Medieval interlace ornament: the making of a cross-cultural idiom’, Arte Medievale 9: 59–86. Trotta, M. and A. Renzulli 2003. ‘La caverna di S. Michele al Gargano: funzione d’uso e funzione monumentale delle fabbriche antistanti all’imboccatura’, in R. Fiorillo and P. Peduto (eds), III Congresso Nazionale di Archeologia Medievale: 736–40. Florence, Insegna del Giglio. Ugolini, L. M. 1932. Albania antica 2: l’acropoli di Fenice. Rome/Milan, Scalia. Ugolini, L. M. 1937. Butrinto: il mito di Enea. Gli scavi. Rome, Istituto Poligrafico dello Stato. Uytterhoeven, I. 2007a. ‘Housing in late antiquity: thematic perspectives’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity. From Palaces to Shops (Late Antique Archaeology 3.2): 25–66. Leiden, Brill. Uytterhoeven, I. 2007b. ‘Housing in late antiquity: regional perspectives’, in L. Lavan, L. Özgenel and A. Sarantis (eds), Housing in Late Antiquity. From Palaces to Shops (Late Antique Archaeology 3.2): 67–96. Leiden, Brill. Valenti, M. 2004. L’insediamento altomedievale nelle campagne toscane. Paesaggi, popolamento e villaggi tra VI e X secolo. Florence, All’Insegna del Giglio.

345

Vanhaverbeke, H., F. Martens and M. Waelkens 2007. ‘Another view on late antiquity: Sagalassos (SW Anatolia), its suburbium and its countryside in late antiquity’, in A. G. Poulter (ed.), The Transition to Late Antiquity in the Danube and Beyond: 611–48. Oxford, Oxford University Press/British Academy. Varalis, Y. 1999. ‘Deux églises à choeur triconque de l’Illyricum oriental. Observations sur leur type architectural’, Bulletin de Correspondance Hellénique 123: 195–225. Völling, T. M. 2001. ‘The last Christian Greeks and the first pagan Slavs’, in E. Kountoura-Galake (ed.), The Dark Centuries of Byzantium: 303–23. Athens, Hellenic Research Foundation von Hesberg, H. 1994. Formen Privater Repräsentation in der Baukunst des 2. und 1. Jahrhunderts v. Chr. Cologne, Böhlan. Vikan, G. 1984. ‘Art, medicine and magic in early Byzantium’, Dumbarton Oaks Papers 38: 65–86. von Hesberg, H. 1994. Formen Privater Repräsentation in der Baukunst des 2. und 1. Jahrhunderts v. Chr. Cologne, Böhlan. Vroom, J. 2004. ‘The medieval and post-medieval fine wares and cooking wares from the Triconch Palace and the Baptistery’, in R. Hodges, W. Bowden and K. Lako (eds), Byzantine Butrint. Excavations and Surveys 1994–99: 278–92. Oxford, Oxbow Books. Vroom, J. in prep. Butrint: the Medieval and Post-Medieval Pottery. Wallace-Hadrill, A. 1988. ‘The social structure of the Roman house’, Papers of the British School at Rome 56: 43–97. Wallace-Hadrill, A. 1994. Houses and Society in Pompeii and Herculaneum, Princeton, Princeton University Press. Walter, C. 2000. ‘Some unpublished intaglios of Solomon in the British Museum, London’, in C. Walter (ed.), Pictures as Language: How the Byzantines Exploited Them: 415–23. London, Pindar Press. Ward-Perkins, B. 2005. The Fall of Rome and the End of Civilization. Oxford, Oxford University Press. Ward-Perkins, J. B., J. H. Little and D. J. Mattingly 1986. ‘Town houses at Ptolemais: a summary report of survey and excavation work in 1971, 1978–1979’, Libyan Studies 17: 109–54. Werner, K. E. 1998. Die Sammlung antiker Mosaiken in den vatikanischen Museen. Vatican City, Monumenti e Gallerie Pontificie. Wharton-Epstein, A. 1980. ‘Middle Byzantine churches of Kastoria: dates and implications’, Art Bulletin 62: 190–207 Whittow, M. 1995. ‘Rural fortifications in western Europe and Byzantium, tenth to twelfth century’, Byzantinische Forschungen 21: 57–74. Wickham, C. 2005. Framing the Early Middle Ages. Oxford, Oxford University Press. Wilkinson, D., 1992. ‘Summary and discussion’, in D. Michaelides and D. Wilkinson (eds), Excavations at Otranto. Volume I: the excavation: 42–58. Galatina, Congedo. Williams, C. K. II and N. Bookidis (eds), 2003. Corinth: The Centenary 1896–1996 (Corinth XX). Princeton, The American School of Classical Studies at Athens. Wilson, R. J. A. 1983. Piazza Armerina. London, Granada. Yegul, F. 1992. Baths and Bathing in Classical Antiquity. New York/London, Architectural History Foundation/MIT Press. Zanini, E. 2006. ‘Artisans and traders in the early Byzantine city: exploring the limits of archaeological evidence’, in W. Bowden, A. Gutteridge and C. Machado (eds) Social and Political Life in Late Antiquity (Late Antique Archaeology 3.1): 373–411. Leiden – Boston, Brill. Zheku, K. 1978. ‘Banesa fshatare në fshatrat kodrinore përreth qytetit të Tiranës’, Monumentet 15–16: 163–84.

Index

acropolis 1, 2–3, 307, 314, 324, 325, 326, 329, 332, 334, 336 Acropolis Basilica 300, 326, 336 Acropolis Castle 336 Adriatic Sea vii, 1, 2, 144, 326, 338 Aegean Sea 324, 332, 336 Africa 332 Agora, Athenian vii House A 288 Aix-en-Provence 272 al-Idrisi, Ibn 336 Albania viii, 1, 2, 5, 300, 317 Alkison (bishop) 317 alleyway (see road) Amorium 324, 326 Amphipolis (Basilica A) 308 amphorae 304, 309, 332, 336 6th-century 197 10th- to 12th-century 143 burials 88, 90–95, 105, 112, 115, 117, 177–178, 188, 192, 196, 201, 202, 312–313, 332 Calabrian 112, 115 Coptic 171 Cretan 91, 202 Crypta Balbi 115 Crypta Balbi 2 313, 317 Gazan 88, 171, 178, 251 Late Roman 1 188 Late Roman 2 170 Late Roman 13 318 medieval 60, 125 Mediterranean 193 necks in furnaces 78, 98–99, 311 Otranto 334, 336 Otranto 1 319, 322, 332 Otranto 2 322 Samos 78, 171, 188, 202, 311 southern Italian 139, 324 Tunisian 171, 332 ‘witch hat’ lid 202 Anastasius (emperor) 184, 338 Angevins 336 Antigoneia (triconch church) 300 Antioch 274 Apamea Maison du Cerf 296–297 Maison au Triclinos 285 Aphrodisias Bishop’s Palace 280 North Temenos House 285 Apollonaris, Sidonius 296

Apollonia (Palace of the Dux) 280, 292 Apulia 332, 337 Arab 304 Arapaj (triconch church) 300 Asia Minor 278, 279, 309, 310, 332 Astigi (domus in Calle San Francisco) 306 Athens 326 Augustus (emperor) 338 Balkans 278, 309, 315, 317, 323, 326, 338 Baptistery 95, 192, 289, 300, 303, 318, 324, 336 barbarians 304 Barcelona 318 Bari 326 Basil I (emperor) 205 Basil II (emperor) 322, 326, 336 Basilicata 66 bath-house, Triconch Palace (see also Room 28, Room 32 and Room 36) 33–37, 55, 60, 69, 115–116, 125, 159, 163, 165, 179, 280, 282, 288–290, 301, 307 bath-house by Venetian Tower 329 Benedict of Peterborough 336 blacksmiths (see smithing) Bomarzo 275 bridge 319 British School at Rome viii Bulgaria 317 Bulla Regia (House of the Hunt) 282, 289 burials (cemeteries, graves, skeletons) 3, 56, 87, 88, 90–95, 96, 105, 109–111, 112–115, 117, 118, 119, 125, 129–131, 132, 144, 148–150, 151, 177–178, 188, 192, 201, 203, 218, 219–223, 224–225, 228, 303, 311, 312–314, 317, 321–322, 323, 332 Buthrotum (see Butrint) Butrint (Buthrotum) vii, 3 Butrint Foundation vii, viii, 1, 300, 327 Butrint National Park 8 Byzantines 336 cabane 323 Caesar (emperor) 338 Campania 277 Caračin Grad 316–317 Carthage vii Maison du Triconque 292 cemeteries (see burials) Cherchel vii Christ 296 Chronicle of Monemvasia 317 Church of the Forty Martyrs 300, 337

Index Cicero 278 de Legibus 317 circular structure 69–73, 83, 119, 144, 175, 304 city wall 3, 5, 6, 37, 56, 76, 82, 86, 87, 96, 125, 131, 139, 144, 146, 152, 156, 159, 162, 171, 172, 175–184, 190, 193, 196, 199, 201, 202, 203, 205, 206, 208, 210, 212, 213, 219, 221, 222, 223, 227, 296, 299, 303, 306–308, 311, 315, 319–321, 323, 325, 326, 330–332, 334, 336 clarissimus 252, 288, 329 coarse wares 143 Comnenan coins 324 Comnenus, Manuel 144 Constans II 317 Constantinople (Palace of Lausos) 295 Corfu (Kerkyra) 317, 324, 332, 337, 338 Corinth vii, 225, 323, 326, 334 Lechaion basilica 309, 316 Peirene fountain 326 corridor (see Room 18) Crete 307, 332 Crusades 144, 324, 336 curia 277 Daphne (House of the Phoenix) 281 Delphi 309 Diaporit 1, 11, 12, 291, 300, 304, 318 dining couch (see stibadium) dining room (see triconch triclinium) Djemila House of Bacchus 282 House of the Seven Apses 302 Maison d’Amphitrite 288 Dodona (triconch church) 300 double apsed vestibule (see Room 11) drain city 13, 14, 40, 41, 50, 95, 117, 292 fountain, Room 24 16 peristyle courtyard Room 16 to long gallery Room 18 30–32, 33, 37, 55, 67, 273, 282, 284, 288 Durrës (Dyrrhachium) 2, 181, 326, 337 Dyrrhachium (see Durrës) Early Byzantine Tectonic Paroxysm 307 earthquake 37, 284, 304, 307, 309, 319, 327, 329 Egypt 336 Elbasan (Scampis) 181 emboropanegyri 332 entrance, Triconch Palace east 12, 14–16, 22, 24, 37, 58, 85, 282 north 39, 47, 48, 50–52, 55, 282, 290, 292, 300 south (marine entrance) 3, 47, 48–50, 55, 78, 282, 290, 292, 295–296, 300, 323, 329 west 14, 22, 24–28, 29, 37, 39, 47–48, 50, 53, 55, 61, 64, 65, 66, 69, 76, 83, 85, 255, 259, 273, 278, 284–288, 290, 291, 292, 297, 301, 302 Ephesus Church of Saint John 300 Hanghaus 2 304 Villa over the theatre 279 Epirote Despots 323, 336 Epirus 280, 300, 307, 317 Nova 181 Vetus 181

347

Fatimid (Caliphate) 336 finewares 139, 144, 193, 196, 290 flooding 66, 67, 76, 82, 139, 146, 174, 197, 290, 299, 306–308, 317, 319, 327–329 Florence viii forum 1, 37, 208, 210, 292, 307, 319, 321, 325–326, 327, 336 fountain, Room 24 16, 19, 23, 55, 242, 244, 247, 272, 281 furnaces (see ovens) gardens 5, 11, 14, 19, 22, 23, 24, 41, 48, 139, 223, 280, 292, 295, 296, 323 Gargano, Monte 337 gates in city wall 5, 82, 87, 96, 150, 152, 175, 178, 179, 180, 181, 182, 197, 199, 203, 206, 208, 209, 210, 212, 217, 218–219, 223, 227, 228, 307, 311, 314, 319, 321, 323, 324, 332, 334, 336, 337 Gaul 278 Germanic invasions 317 Gortyn 310 graves (see burials) Great Basilica 8, 95, 192, 300, 303, 318, 319, 321, 324, 332 Greece 2, 278, 300, 309, 310, 315, 317, 322, 326 Gregory the Great (pope) 317 Grumentum 292 Gymnasium 13, 292, 324, 336 Halicarnassus (late Roman villa) 279 hammerscale (see smithing) hearths 28, 53, 58, 60, 61–64, 66, 67, 69, 74, 76, 78, 79, 88, 96, 102, 175, 184, 190, 193, 197, 199, 203, 205, 212, 213–216, 217, 218, 228, 299, 303, 304, 309, 311, 312, 319 Heraclius (emperor) 317 Herculaneum 277 Hierapolis 322, 323, 326 Hispania 306 ‘hollow way’ (see trackway) Illyricum 184 Institute of Archaeology (Tirana) vii, 20, 39, 56, 301, 321 Institute of World Archaeology viii Integrated Archaeological Database 6 Ionian islands 2, 307, 336 Ionian Sea vii, 1 Ireland 7 Isaac II Angelus (emperor) 144 Isère 272 ‘Islamisation’ 304 Italian Archaeological Mission vii, 3, 20, 39, 56 Italy 231, 273, 278, 292, 317, 323, 326, 332 Jekmejeh 275 Julian of Ascalon (Urban Treatise) 289, 309 Justin I (emperor) 184 Justin II (emperor) 197, 199, 202 Justinian (emperor) 184, 300 Kastoria 322 Kephallonia 307 Kephalos 317 Kerkyra (see Corfu) kilns 53, 96, 304, 332

348

Index

Lake Butrint vii, 2, 3, 185, 203, 319 Lake Gate 180, 334 lamprotatos 252, 288, 329 late antique houses apsidal spaces 281–282 audience and reception chambers 279–282 building phases 291 bath-houses 288–289 corridors 279 encroachment on public space 292 entrance vestibules 279, 285, 295 luxurious residences 278 mosaics 278, 296 peristyles 288, 292 porticoes 278, 280 sigma and stibadiium 282, 296–297 triclinia 278, 280, 281, 296 triconch apses 297 Le Puy (bishopric) 336 Levantine 332 Lin (triconch church) 300 Lion Gate 180, 322 London (British Museum) 238 long gallery (see Room 18) Lucca 336 Lyons 238 Macedonia 317, 332–334 marine entrance (see entrance, Triconch Palace, south) Maurice (emperor) 317 Mediana villa 282 Mediterranean Sea vii, 1, 2, 5, 7, 139, 193, 274, 296, 300, 303, 307, 311, 318 Mérida 318 Casa de los marmoles 306 House of the Aqueduct of Tiermes 306 Morerías, district 306 Miracles of St Demetrius 317 Montenegro 326 mosaics, Triconch Palace inscription 25–28, 252, 273, 284, 288, 292, 329, 338, 329, 338 masks 231–234, 249, 251, 271–273, 276, 279 portico 30, 32, 37, 40, 55, 82, 85, 255–271, 273–276, 284, 288, 297 Room 18 14, 22, 23, 30, 32–33, 37, 55, 117, 231–242, 271–272, 274, 276, 279, 282, 288, 307 Room 24 19, 20, 23, 33, 37, 143, 242–251, 271, 272–273, 274, 276, 311 Room 25 22, 25–28, 30, 61, 64, 65, 69, 83, 119, 252–255, 273, 274, 276, 284, 302 mussel shells and processing 5, 60, 62, 67, 69, 79, 87, 88, 90, 95, 96, 99, 102, 109, 115–116, 117, 118, 119, 165, 184, 185–187, 190, 192, 193, 197, 199, 201, 202, 203–206, 208, 209, 210, 228, 311, 312, 319, 323, 324, 332, 336 Naples 326 Nea Anchialos 316 Nero (emperor) 281 Nicopolis 181, 302, 307, 317, 337 Basilica B 314–315 bishop 317 Nimrud (palace of Assurnasirpal) 275

Normans 336 North Africa 278 Norwich viii Olympia 310 Onchesmos (see Saranda) open area (see Room 39) Orsini, Vincenzo 275 Otranto (see also amphorae) 144 ovens (furnaces) 61–4, 69, 78, 82, 88, 96–102, 105, 109, 115, 304, 311 Palaiopolis 338 Palatine Hill (House of Augustus, Room of the Masks) 272 Palestine 289, 309 Pannonia 292 Paramythia (triconch church) 300 Peloponnese 324 Pergamum 272, 324, 326 Perge (Late Antique Residence) 281 peristyle courtyard, Triconch Palace (see also Room 16) 3, 5, 11, 14, 19–20, 22, 24, 25, 29, 32, 33, 37, 39–41, 48, 50, 51, 52, 55, 56, 65, 66, 74, 76, 78, 81, 85, 86, 87, 104–105, 109, 117, 118, 119, 131, 132–136, 139, 141, 144, 146–150, 151, 165–168, 252, 259, 269, 273, 274, 275, 276, 279, 283, 284, 285, 288, 290, 291, 292, 295, 297, 299, 300, 301, 303, 304, 306, 311, 312, 313, 321, 322, 323, 324, 326, 329, 336 Petronius (Satyricon) 277 Pescara 326 phase 1: 3rd century and earlier (c. AD 100–250) 11–14 phase 1–2: 4th century and earlier 152–165 phase 1–2a 152 phase 1–2b 152–154 phase 1–2c 154–156 phase 1–2d 156 phase 1–2e 158–159 phase 1–2f 159–163 phase 2: 3rd to 4th century 14–20 phase 2b 20–24 phase 3: Early 5th century (c. AD 400) 24–37 phases 3–5: Early 5th century (AD 400–450) 165–168 phase 4: c. AD 420 37–55 phase 5: Early to mid 5th century (c. AD 420–450) 56–66 phase 6: Mid to late 5th century (AD 450–500) 66–76, 168– 175 phase 7: Early 6th century 76–87, 175–184 phase 8: Early to mid 6th century (c. AD 525–550) 87–96, 184–193 phase 9: Mid to late 6th century (c. AD 550–575) 96–105, 193–197 phase 10: Late 6th to early 7th century (c. AD 575–650) 105–117, 197–201 phase 11: Mid 7th to early 10th century 119 phases 11–12: Early 7th to first quarter of 11th century 203– 208 phase 12: Early to late 10th century 119–132 phase 12–13: Early 10th to 12th century 132–139 phase 13: Early 11th to late 12th century 139–144 phase 13a: Early 11th century (after AD 1025) to 12th century 208–212 phase 13b: 12th to early 13th century 212–217 phase 14: Early 13th to 14th century 144–150 phase 14a: 13th and 14th century 217–223

Index phase 14b: Late 14th century 223–227 phase 15: 15th to late 16th/early 17th century 150–151, 227–228 Philippi 292 Bishop’s Palace 309, 315–316 house to the northeast of the Octagon 289 Phocas 317 Phoenice 317, 337 Piazza Armerina 280, 292, 297 pits 53, 59, 61, 62, 67, 76, 78, 79, 85, 88, 96, 98, 99, 109, 117, 118, 119, 129, 143, 144, 146, 150, 190, 203, 208, 212, 219, 311, 323 Poggio Gramignano 313 Pollux, Julius (lexicographer) 234 Pompeii 231, 272, 277 House of the Faun 272 portico, Triconch Palace 29–30, 32, 39–40, 41, 48, 50, 52, 55, 66, 74, 76, 78, 81, 82, 85, 86, 87, 95, 96, 104, 109, 119, 132, 150, 252–271, 273–276, 284, 288, 290–291, 292, 297, 300, 304, 311 post-built structures (post-holes, stake-holes) 5, 48, 53, 55, 56, 58, 60, 61, 62, 64, 66, 67, 73, 76, 78–79, 83, 85, 87, 88, 90, 95, 96, 109, 116, 117, 118, 119, 123–125, 131, 136–139, 141, 143, 144, 146, 150, 151, 168, 192, 193, 197, 202, 203, 205, 206, 208, 210–212, 215–216, 217, 221, 250, 278, 303, 304, 311, 321, 323, 324, 325, 326, 332, 336 post-holes (see post-built structures) potentiores 22 Procopius 184 Ptolemais 292 House of the Triapsidal Hall 292, 297 range, Merchant’s House north (see wing, Merchant’s House, north) west (see wing, Merchant’s House, west) range, Triconch Palace north (see wing, Triconch Palace, north) south (see wing, Triconch Palace, south) west (see wing, Triconch Palace, west) Ravenna 317, 326 Palace of Theoderic 282, 297 Domus dei Tappeti di Pietra 292 reception room (see Room 24) road (alleyway, street) Merchant’s House 5, 185, 208, 210, 212, 321, 323, 336, 337 Triconch Palace 12, 13, 14, 20, 25, 48, 55, 285, 292, 296, 329 Rome vii, 115, 231, 272, 274, 276, 326 Crypta Balbi vii, 115 Domitian’s Palace 281 domus of Gaudentius 292 forum of Nerva 317 house of Junius Bassus 295 Room 1 41 Room 5 47, 321 Room 6 47 Room 7 47 Room 8 47 Room 9 50 Room 10 52, 53 Room 11 (double apsed vestibule) 39, 47, 48, 51, 52 Room 12 50, 136 Room 14 50, 51–52, 74–76, 95, 130 Room 16 (peristyle courtyard) 29–32, 109, 139, 146–150, 255–271, 276, 313

349

Room 17 51–52, 74, 76, 95, 131, 144–146, 151 Room 18 (corridor, long gallery) 11, 14–16, 19, 22, 23, 24, 32, 33, 37, 55, 56–60, 64, 66, 67, 69, 76, 78–79, 82, 85, 87–8, 96–99, 102, 103–104, 105, 112, 115, 116, 117, 123, 125, 136, 138, 139, 151, 163, 165, 174, 197, 231–242, 249, 250, 251, 271–272, 274, 276, 278, 279, 280, 282, 283, 284, 289, 290, 304, 307, 311, 312, 313, 314, 315, 327 Room 19 11, 14, 19–20, 22, 24, 33, 55, 67, 76, 78, 81, 85, 88, 109, 118, 119, 123, 125, 131, 136, 139, 142–143, 151, 283, 315, 321, 323, 325, 326 Room 20 11, 14, 19–20, 22, 24, 33, 67, 76, 81, 85, 88, 90, 109, 112, 116, 283, 312 Room 21 11, 14, 19–20, 22, 24, 33, 55, 56, 58, 60, 66, 67, 76, 81–82, 86, 87, 88, 90, 96, 99–102, 104, 109, 112, 116, 283, 288, 307, 311, 313 Room 22 11, 14, 19–20, 22, 24, 30, 32, 33, 53, 55, 56, 60, 67, 76, 78, 82, 83, 86, 88, 104, 105, 111, 112, 117, 136–139, 283, 307, 311, 313 Room 23 11, 22, 24, 33, 55, 60–61, 64, 67, 76, 82–83, 88, 90, 112, 117, 125, 129, 136–139, 283, 299, 304 Room 24 (reception room) 14, 16–19, 20–22, 23, 24, 35, 37, 55, 58, 66, 76, 78, 82, 86, 87, 88, 96, 104, 105, 115, 117, 125, 139, 143–144, 151, 163, 165, 242–251, 252, 271, 272–273, 274, 276, 278, 279, 280–283, 289, 290, 292, 304, 311, 312, 313 Room 25 20, 22, 24, 25, 41, 53, 55, 60, 62, 64, 69, 83, 85, 109, 117, 118, 119, 121, 123, 131, 139, 141, 252–255, 273, 274, 276, 284–288, 299, 304, 313 Room 26 20, 22, 24, 25, 28, 33, 53, 55, 60–64, 69, 76, 82, 83, 109, 118, 121, 139, 141, 299, 304, 306 Room 27 20, 22, 25, 28, 60, 64, 69–73, 76, 83, 109, 119, 121, 123, 131, 144, 151, 175, 304 Room 28 (bath-house) 35–37, 55, 60, 88, 116, 125, 304 Room 29 11, 47–48, 52, 53, 64, 66, 73–74, 76, 90, 105–109, 116–117, 144, 292, 304, 312 Room 31 11, 47–48, 52, 64, 66, 73, 74, 76, 83, 95, 104, 119, 123, 131, 141, 292, 304 Room 32 (bath-house) 60, 88, 116 Room 35 125 Room 36 (bath-house) 33–35, 60, 69, 116, 179 Room 37 154–156, 158–159, 163–171, 172, 174–175, 187, 188, 192, 193, 304, 306, 307, 309, 312, 314–315 Room 38 33, 37, 158–159, 162, 163–165, 168, 171–172, 175, 178–179, 184, 190–192, 193, 197–199, 202, 203, 205–206, 208, 213, 223–227, 228, 280, 307, 311 Room 39 (open area) 156, 159, 162, 165, 177, 178–179, 180, 181, 184–187, 190, 192, 196, 197–199, 202, 203–205, 206, 208–210, 212–215, 217, 218–219, 225, 227, 228, 280, 311 Room 40 159, 173–174, 177, 192, 193–196, 197, 202, 203, 206, 208, 210–212, 215–216, 217, 218, 219, 222, 223, 227, 228, 314, 315, 317, 321, 323, 325, 326 Room 41/44 156, 158, 159, 165, 168, 172, 175, 181, 184, 185, 190, 192, 193, 202 Room 43 162, 163, 173, 175–178, 185, 192, 196, 197, 206, 208, 210–212, 280, 321, 323, 325, 326 Room 45 (tower) 5, 159, 219, 223–228 Room 46 154, 156, 175 Room 47 159–162, 163, 168, 172–174, 175–178, 185, 192, 193, 196, 202, 280, 307, 312 Room 48 152–154, 156 Room 49 152–154, 156 Room 50 158, 163

350

Index

Sagalassos (Palatial Mansion) 310 Saint-Romain-en-Gal 238 San Giovanni di Ruoti 66 San Michele (shrine) 337 Saranda (Onchesmos) 181, 317, 337, 338 Rruga Skenderbeu basilica 315 Sardis 278 late Roman Town House 279 Savaria 292 Sbeitla (House of Seasons) 295 Scampis (see Elbasan) seismic activity (see earthquake) Serbia 317 Sicily 278 sigma 169, 282, 314 skeletons (see burials) slag (see smithing) Slavs 199, 317 smithing (blacksmiths, hammerscale, slag) 60, 88, 96, 99, 105, 117, 215, 217, 310, 311, 312, 323 Sousse 272, 274 Spain 278 stake-holes (see post-built structures) Stari Bar 326 stibadium (dining couch) 32, 41, 42, 271, 282, 296, 297 Studio Inklink viii Straits of Corfu vii, 2, 3, 144, 319 street (see road) Syria Secunda 285

Ugolini, Luigi Maria vii, 8, 47, 118, 300, 317, 326, 327, 336 UK 7, 303 Umbria 313 Unesco 2 University of East Anglia viii University of Siena viii urban residential change burial practice 311–313 city walls 307–308 industrial activity 309–310 subdivision of buildings 303–306, 318 two-storey buildings 314–317 Utica (House of the Fishermen) 238

table wares 139, 150, 223, 228, 322, 324, 332, 336 Tacitus 278 theatre 1, 11, 307 Theodosius II (emperor) 181 Thessalonica (Palace of Galerius) 295 Thurburbo Maius 292 House of the Protomes 282, 297 Tiber valley 276 Tipasa (House of the Frescoes) 306 Toulouse 318 Tower Merchant’s House (also see Room 45) 177, 179, 203, 219, 223–228, 323, 336 Western Defences 119 trackway (‘hollow way’) 11, 14, 24, 163–165 Treaty of Campo Formio 324 Triangular Castle 3, 323 triconch triclinium (dining room) 3, 20, 37–39, 40, 41–47, 48, 55, 56, 60, 76, 86, 87, 95, 96, 165, 279, 282, 284, 290, 292–295, 296–297, 299, 300, 301, 303, 306, 319, 323, 325, 329 church 3, 47, 118, 300, 321–322, 323, 325, 329, 336 Trobriand Island 375 Tunisia 332 Turkey 323, 326 two-storey buildings Merchant’s House 117, 131, 168, 170, 174, 179, 188, 193–196, 197, 199, 208, 303, 304, 314–316, 317, 318, 321 Triconch Palace 76, 78, 85, 125, 131, 197, 310–311, 314–316, 321

Water Gate 8 wells Merchant’s House 165, 168, 169, 185–187, 193 Triconch Palace 29, 118, 119–121, 151, 289 Well of Junia Rufina 322, 323 Western Defences 119, 317, 319, 323, 332 wing (range), Merchant’s House north 156, 168 west 168 wing (range), Triconch Palace east 32, 33, 39, 55, 56, 291, 292, 306 north 12, 32, 39, 40, 50–52, 55, 56, 74–76, 95, 105, 118, 125, 129–131, 144–146, 150, 184, 188, 273, 274, 276, 286, 290–291, 297 south 14, 19–20, 24, 30, 32–33, 56–60, 64, 66, 67–9, 74, 76, 78–82, 85, 86, 87–90, 96–104, 109–115, 116, 117, 118, 119, 123–129, 131, 136–139, 142–144, 146, 150, 197, 276, 283, 286, 297, 304, 306, 307, 311, 312, 314, 315, 323, 326 west 20, 22, 24, 25, 28, 33, 39, 56, 60–64, 69–74, 76, 82–85, 86, 90–95, 105–109, 116, 118, 119–123, 131, 139–142, 144, 150, 175, 283, 297, 299, 304, 306, 307, 308, 309, 311, 312

Valencia 306 Vandals 332 Venetian Tower 8, 181, 323, 329 Vienne 272 villae maritimae 280, 296 Visigothic king 318 Vitruvius 177 Vivari Channel vii, 1, 2, 3, 5, 8, 14, 23, 29, 37, 41, 48, 55, 82, 86, 87, 95, 105, 139, 146, 152, 156, 158, 159, 163, 165, 175, 177, 179, 180, 181, 182, 197, 203, 217–218, 227, 280, 288, 289, 296, 304, 306–307, 319, 321, 323, 324, 325, 326, 327, 329, 330–332, 334, 336 Vrina vii Vrina Plain 1, 271, 289, 300, 317, 322, 323, 326, 327, 329, 332, 334, 336

Xanthos (House of the Lycian Acropolis) 281 Zante 307

Plates

Plate 1. Room 18, east vestibule, vertical projection (MZ)

Plate 2. Room 18, long gallery, eastern end, vertical projection (MZ)

352

Plates

Plate 3. Room 18, long gallery, central section, vertical projection (MZ)

Plate 4. Room 18, long gallery, western end, vertical projection (MZ)

Plates

Plate 5. Room 18, east vestibule, view to north (MS)

Plate 6. Room 18, east vestibule, detail of central panel with masks (MS)

353

Plates

354

Plate 7. Room 18, east vestibule, female mask (MS)

Plate 8. Room 18, east vestibule, male mask (MS)

Plate 9. Room 18, east vestibule, southeastern corner (JBB)

Plates

Plate 10. Room 18, east vestibule, southeastern corner, to east, showing painted wall-plaster (JBB)

Plate 11. Room 18, long gallery, eastern end, to north (JBB)

355

356

Plates

Plate 12. Room 18, long gallery, detail edge of central panel and border (MS)

Plate 13. Room 18, long gallery, detail of central panel (MS)

Plate 14. Room 18, long gallery, detail of border (MS)

Plate 15. Room 24, reception hall, view to north (JBB)

Plates

Plate 16. Room 24, reception hall, southern end in vertical projection (MZ)

Plate 17. Room 24, reception hall, northern end and apse in vertical projection (MZ)

357

358

Plates

Plate 18. Room 24, reception hall, central feature of pavement, to north (JBB)

Plate 19. Room 24, reception hall, southeastern corner of central feature (MS)

Plates

359

Plate 20. Room 24, reception hall, western side (JBB)

Plate 21. Room 24, reception hall, detail of southern outer border (MS)

Plate 22. Room 24, reception hall, detail of southern outer border (MS)

360

Plates

Plate 23. Room 24, reception hall, view across room to east in front of apse (MS)

Plate 24. Room 24, reception hall, hub of apse with frontal female mask (MS)

Plates

Plate 25. Room 24, reception hall, detail of northeastern shoulder of apse (MS)

Plate 26. Room 24, reception hall, detail of northern edge of apse (MS)

361

362

Plates

Plate 27. Room 25, west vestibule in vertical projection (MZ)

Plate 28. Room 25, west vestibule, detail of central field (JBB)

Plates

Plate 29. Room 25, west vestibule, detail of border and central field (JBB)

Plate 30. Room 25, west vestibule, detail of border (MS)

Plate 31. Room 25, west vestibule, inscription at threshold to peristyle (MS)

363

364

Plates

Plate 32. Room 16, west portico of peristyle, northern end, north panel in vertical projection (MZ)

Plate 33. Room 16, west portico of peristyle, northern end, in vertical projection (MZ)

Plates

Plate 34. Room 16, west portico of peristyle, central section, in vertical projection (MZ)

365

366

Plates

Plate 35. Room 16, west portico of peristyle, southern end, in vertical projection (MZ)

Plates

Plate 36. Room 16, west portico of peristyle, main panel, to south (JBB)

Plate 37. Room 16, west portico of peristyle, border in southwestern corner (MS)

367

368

Plates

Plate 38. Room 16, west portico of peristyle, border in northeastern sector (MS)

Plate 39. Room 16, west portico of peristyle, border in southwestern sector (MS)

Plate 40. Room 16, west portico of peristyle, central section immediately in front of the entrance from the west vestibule (JBB)

Plates

369

Plate 41. Room 16, west portico of peristyle, southern end, rows 27–30, to west (MS)

Plate 42. Room 16, west portico of peristyle, central section, row 11, to east, goose-like bird (JBB)

Plate 43. Room 16, west portico of peristyle, central section, row 15, to east, bird on tree (MS)

370

Plates

Plate 44. Room 16, west portico of peristyle, view from centre looking north (JBB)

Plate 45. Room 16, west portico of peristyle, northern end, to north (JBB)

Plates

371

Plate 46. Room 16, west portico of peristyle, northern end, east side, to north (MZ)

Plate 47. Room 16, west portico of peristyle, northern end, rows 7–9, to south, with knots of Solomon and eye (MS)

372

Plates

Plate 48. Room 16, west portico of peristyle, northern end, row 8, to north, eye (JBB)

Plate 49. Room 16, west portico of peristyle, northern end, rows 5–6, to south, with overlapping scales and chequerboards (JBB)

Plates

Plate 50. Room 16, west portico of peristyle, northeastern corner, row 2, to south, with fronted peltae (JBB)

Plate 51. Room 16, west portico of peristyle, view from centre looking south (JBB)

373

374

Plates

Plate 52. Room 16, west portico of peristyle, southern end, to south (JBB)

Plate 53. Room 16, west portico of peristyle, southern end, to west (JBB)

Plates

375

Plate 54. Room 16, west portico of peristyle, southern end, rows 20–21, to south, with fronted peltae and knots of Solomon (JBB)

Plate 55. Room 16, west portico of peristyle, southern end, rows 22–26, to south (JBB)

376

Plates

Plate 56. Room 16, west portico of peristyle, southern end, rows 24–28, to south (JBB)

Plate 57. Room 16, west portico of peristyle, southern end, rows 26–30, to south (JBB)

Plates

377

Plate 58. Room 16, west portico of peristyle, southern end, rows 35–30, to north (MS)

Plate 59. Room 16, west portico of peristyle, southwestern corner, rows 31–33, with fronted peltae and dentated cross (MS)

Plate 60. Room 16, west portico of peristyle, southern end, rows 32–34, to south, with perspectival parallelipipids and rainbow zigzags (JBB)

378

Plates

Plate 61. Room 16, west portico of peristyle, far southern end, rows 35–34, to north (MS)

Plate 62. Room 16, west portico of peristyle, southern end, row 35, to north, half poised square with stepped design (MS)

Plates

379

Plate 63. Room 16, west portico of peristyle, northern end, looking north over north panel (MS)

Plate 64. Room 16, west portico of peristyle, northern end, border between main and northern panels, to west (JBB)

380

Plates

Plate 65. Room 16, west portico of peristyle, northern end, northern side of north panel, to west (JBB)

Plate 66. Room 16, west portico of peristyle, northern end, field of north panel, to west (JBB)

Plates

381

Plate 67. Reconstructions of the Triconch Palace (2004). Top: c. AD 400; middle: c. AD 420; bottom: c. AD 1000–1200 (Studio Inklink)

Inactive Faultline

4th Century make up layer

11th Century levelling/ dumping deposits

Forum

Line of Hellenistic wall circuit

0

2nd - 5th Century

6th - 12th Century

40

80 metres

?Gravel formation

Triconch Palace

t La of all e n Li ty W Ci

ue tiq n eA

Vivari Channel

Line of Section

Plate 68. Schematic cross section between the Forum and the Triconch Palace, showing development of deposits from 2nd to 12th centuries AD (D. Bescoby)

Acropolis

N

382 Plates