Archaeological Sciences 1995: Proceesings of a Conference on the Application of Scientific Techniques to the Study of Archaeology [1 ed.] 9781785708060, 9781785708053

Citation preview

Archaeological Sciences

1995 Proceedings of a conference on the application of scientific techniques to the study of archaeology Liverpool, July 1995

Edited by Anthony Sinclair Elizabeth Slater John Gowlett

Oxbow Monograph 64 1997

First published in the United Kingdom in 1997. Reprinted in 2017 by OXBOW BOOKS The Old Music Hall, 106-108 Cowley Road, Oxford OX4 1JE and in the United States by OXBOW BOOKS 1950 Lawrence Road, Havertown, PA 19083 © Oxbow Books 1997

Paperback Edition: ISBN 978-1-78570-805-3 Digital Edition: ISBN 978-1-78570-806-0 (epub)

All rights reserved. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical including photocopying, recording or by any information storage and retrieval system, without permission from the publisher in writing.

For a complete list of Oxbow titles, please contact: UNITED KINGDOM Oxbow Books Telephone (01865) 241249 Email: [email protected] www.oxbowbooks.com UNITED STATES OF AMERICA Oxbow Books Telephone (800) 791-9354, Fax (610) 853-9146 Email: [email protected] www.casemateacademic.com/oxbow Oxbow Books is part of the Casemate Group

Preface Prior to 1978 the International Symposia on Archaeometry had been held in Britain but once these meetings started to be organised at overseas venues an important focus in Britain for inter-disciplinary debate was lost. In an attempt to fill this gap, a conference on the application of science within archaeology was held in Glasgow in 1987. This conference was the fifth in a biannual series since that date. We are very pleased that the Liverpool conference appeared to offer further confirmation of the value attached to such meetings, particularly as a platform for students and young research workers. The series continues, and the next meeting will be held in Durham in September 1997. Although the venues for this conference are British, the subject is international, and we were especially pleased to welcome delegates from the USA, Canada, Russia, Portugal, Italy, Germany, France, Greece, Hungary, Denmark and Eire. Many people were involved in the organisation of the conference. We would like to thank all the students and members of staff of the Department of Archaeology who devoted much time to ensure the success of the meeting, and particularly the other members of the organising committee, especially Christine Roberts, Jenifer Mirdamardi and David Bell. We are especially grateful to the Council for British Archaeology, Barrington Instruments, Earth Science Systems Ltd., Geometries Europe, Positioning Resources and the University of Liverpool for their generous financial support. Finally we would like to thank all the delegates and speakers for their contributions, and we offer our appreciation to all those who allowed their papers to be included in this volume. All the authors and referees are to be congratulated for their adherence to the strict deadlines which has allowed the rapid publication of these proceedings.

Anthony Sinclair, Elizabeth Slater and John Gowlett Department of Archaeology, University o f Liverpool, October 1995

Archaeological Sciences 1995 Organising Committee John Gowlett Alf Latham Greg McIntosh John Shaw Anthony Sinclair Elizabeth Slater Joan Taylor

Contents The Scientific Analysis of Artefacts: Petrography R.V. Davis Implement petrology: 50 years of service to the archaeological com m unity.................................................................................. 1 P. Kayani and G. McDonnell The potential of scanning electron microscope techniques for non-destructive obsidian characterisation and hydration rim dating research............................................................................................................................................................ 5 R .A .Ixer Detailed provenancing of the Stonehenge dolerites using reflected light petrography: a return to the lig h t....................... 11

The Scientific Analysis of Artefacts: Glass N. Earl and D.E. Watkinson Assessment of post-excavation systems for archaeological glass using FT-IR m icroscopy..................................................... 19 I. Merchant, J. Henderson, D. Crossley and M. Cable Medieval glass making technology: the corrosive nature o f g la s s ............................................................ .................................... 31 L. Blek and G. Gilmore Non-invasive revelation o f precious metal decoration in glass b ead s............................................................................................ 38 P. Mills and G.A. Cox Medieval windows: what can chemical analysis tell us? ................ .................................................................................................. 43 P.T. Nicholson Early glass and glazing in Egypt: new excavations at Tell el-A m arna......................... ................................................................ 49

The Scientific Analysis of Artefacts: Ceramics N. Brodle The chemical analysis of archaeological ceramics at the inter-regional level: the case of early Iron Age C yprus................................. ......................................................................................................................... 55 C.A. de Domingo and A.W. Johnston Trade between Kommos, Crete and East Greece: a petrographic sudy of Archaic transport amphorae................................ 62 M. Eiland Petrography of Parthian ceram ics....... ............. ....................................................... ................................ ............................................ . 69 M. Hughes, H. Blake, J. Hurst and T. Wilson Neutron activation analysis of Italian Maiolica and other Medieval Italian ceram ics............................................................... 77 L. Joyner Byzantine and Frankish cooking wares at Corinth, Greece: changes in diet, style and raw material exploitation............. 82 IK . Whitbread, R.E. Jones and J. Papadopoulos The early Iron Age kiln at Torone, Greece: geological diversity and definition of control grou p s................ ....................... 88 J. Williams and D. Jenkins Provenancing of prehistoric and Roman pottery using petrographic and mineralogical analysis: case studies from North W a le s................................................................................................................................................................ 92

The Scientific Analysis of Artefacts: Metallurgy J. Bayley and S. Butcher The composition and decoration of Roman B rooch es............... ................ ....... .................... ....................................................... 101 J. Bayley and K. Eckstein Silver refining: production, recycling and assaying.................... ............ .............................................. ......................................... 107 S. Bean The application o f metallurgical analyses in numismatics: some cautionary t a le s .......... ................................................

112

D. R. Griffiths, A.M. Feuerbach and J.F. Merkel Early Islamic steel manufacture in M erv............................................................. .............................................................................. 116 C. Mortimer and M. Stoney A methodology for punchmark analysis using electron m icroscopy.......... ................................................ .............. ............. .

119

E. C. Joel, J.J. Taylor, R.A. Ixer and M. Goodway Lead isotope analysis and the Great Orme m in e....... ...... ...... ............... ............. ........................................................................... 123 A.I. Seruya and D.R. Griffiths Ageing processes in gold-copper-silver allo y s........................................ ..................... .................................................................... 132 S. Srinivasan The composition of bronze slags from Kayladi, South India and implications for the problem o f tin in South Indian b ron zes......................................... ...................... ....................... .................................................................................. 136 J.J. Taylor, R.J. Watling, C.A. Shell, R.A. Ixer, R.J. Chapman, R.B. Warner and M. Cahill From gold ores to artefacts in the British Isles: a preliminary study o f a new LA-ICP-MS analytical approach....... .

144

Chronological Studies S. Barnett Luminescence dating of first millenium be ceram ics................................................................................................. .............. .

155

J. Rees-Jones Optically stimulated luminescence (OSL) dating of the Uffington White H orse...................................................................... 159 L. Shepherd Optical dating of archaeological ceram ics............................ .............. ......................................................................... ............. ....... 163 J. Csapo, Zs. Csapo-Kiss, L. Kolto and S. Nemethy Age estimation of materials with high Keratin content based on cystine and cysteic acid con ten t.................................

170

T. Brown, A.G. Latham, J.A.J. Gowlett and S. Luboga Uranium-series dating of fossil Nile Oyster from a Palaeolithic site, Mweya, Uganda....... ............. ............ .............. .

174

M. Farid Khan and J.A.J. Gowlett Age-depth relationships in the radiocarbon dates from Sanghao Cave, Pakistan..................................................................... 182 H. King Uranium-series dating of flowstones from the Great Orme m in e........................................................... ..................................... 188 Zs. Csapo-Kiss, J. Csapo, L. Kolto and S. Nemethy A possible err or in amino acid d atin g.......... .................... ....................... ...................................................... ....... .......................... 193 R. Curl and A.Latham Bayesian estimation of age differences using stratigraphic order........... ....................... ................ ....... .................................... 194

Dendrochronological Studies N. Bonde Dendrochronological datings of the viking age ship burials at Oseberg, Gokstad and Tune, Norway: some new resu lts.............................................................. .............................................. ......................................................................... 195 N. Bonde, I. Tyers and T. Wazny Where does the timber come from? Dendrochronological evidence o f the timber trade in Northern Europe: 14th to 17th Centuries A D ................................................................... ............................ ................................................ ........ .

201

C. Groves The dating and provenancing of imported conifer timbers in England: the initiation of a research project...................... 205

The Study of Ancient Environments I. Mainland A qualitative approach to dental microwear analysis....................................................................................................................... 213 R. Tipping, S. Carter, D. Davidson, D. Long and A. Tyler Soil pollen analysis: a new approach to understanding the stratigraphic integrity o f data............................... ..................... 222 C. Caple and D. Dungworth Investigations into waterlogged burial environments............. ...................... ................................................................................... 233 G.H. Endfield Myth, manipulation and myopia in the study of Mediterranean soil erosion............................................................................. 241 A. Krahtopoulou Agricultural terraces in Livadi, Thessaly, G reece............................................................................................................................ 249 T. Clare The palaeohydrology of a river basin in the English Lake District: an interim v ie w ............................................................

260

A. Clapham, T. Clare and D. Wilkinson A plant macrofossil investigation of a submerged forest............... ................................................................................................. 265 S. Gonzalez, D. Hubbart and G. Roberts Holocene development of the Seton Coast: a multidisciplinary approach to understanding the archaeology................... 271 D. Jenkins Phosphorous redistribution in cave sediments from the Lower Palaeolithic cave site of Pontnewydd............................... 282 Y. Ivaschenko The ecological crisis on the North Black Sea coast in the 2nd to 1st centuries B C ................................................................ 287 M. Madella Phytolith analysis from the Indus Valley site of Kot Diji (Sind, Pakistan)................................................................................ 294 A. Pantazidou, M. Roussomoustakaki and P. Theolakis The temple at Epikourious Apollo (Greece): consequences of the protection work on its microflora........................ ....... 303

A. Pantazidou, M. Roussomoustakaki and C. Urzi The microflora of the Milos Catacombs (G reece)........................................ .............. .................................................................... 308 A. Sazanov Palaeogeography and palaeoecology o f the Crimea in Ancient and Medieval tim es............................................................... 313 G. Turner-Walker and C.J. Scull Microfauna in Anglo-Saxon graves: entomological evidence at Boss Hall and the Butter Market, Ipswich..................... 320 L. Willies and J. Maskall The historical and archaeological “sp in-off’ from heavy metal environmental research in the Peak D istrict................... 328

Remote Sensing C. A. Shell Appropriate geophysics and excavation strategy: from mud brick to masonry in the East Mediterranean reg io n .......... 333 A. Schmidt and A. Marshall The impact o f resolution on the interpretation of archaeological prospection d a ta ........ ........................................................ 343

The Analysis of Human Remains V. Carolan, M.L. Gardner and A.M. Pollard Age determination via measurement of amino acid racemization in dental tissu e................................. .................................. 349 D. Lucy and A.M. Pollard Chemical study of degraded dentine by resolution enhanced infra-red spectroscopy.............................................................. 358 M.P. Richards and G.J. van Klinken A survey of European human bone stable Carbon and Nitrogen isotope v a lu e s................ ............................................

363

G. Panagiaris, A. Hassiacou and S.K. Manolis The ecological influence on human microevolution and culture: a study o f ancient Greek pouplations around Mount P in dos...................... ....................................... ....................... ........ ...................... ..................................... ................... 369 M. Tiley-Baxter A Taphonomic study of the human remains from Christchurch, Spitalfields...................... ............................ ............. ....... . 377

Perspectives on Human Evolution and Early Hominid Artefacts S.A. Andresen, D.A. Bell, J. Hallos, T.R.J. Pumphrey and I.A.J. Gowlett Approaches to the analysis o f evidence from the Acheulean site o f Beeches Pit, Suffolk, E ngland..................... .

389

J.L. Cormack Are the Sangoan and the Axheulean ‘industrial complexes’ distinct? ............................ ......................... ..................... ............ 395 R.H. Crompton and J.A.J. Gowlett The Acheulean and the Sahara: allometric comparisons between North and East African sites ........................................... 400 R. H. Crompton and L. Yu Running before they could walk? Locomotor adaptation and bipedalism in early hom inids................................................ 406 S. Crompton Technology and morphology: does one follow the other?.................. ........................ ...........................................................

413

P.S. Quinney and M. Collard Sexual dmorphism in the mandible of Homo neanderthalensis and Homo sapiens: morphological patterns and behavioural im plications........................................................................ ....... ................................. .............................

420

I. Spears and R.H. Crompton The mechanical significance o f “thick” enamel in early hominoids........................................................................................... 426 T. Weber Shape and size in the study o f older Palaeolithic stone flake inventories.......................... ................................ ...................... 434 B. Wood and M. Collard Grades and the evolutionary history of early African hom inids...................................................... .......................................

445

Implement Petrology: 60 Years of Service to the Archaeological Sciences R.V. D avis Chair o f CBA Implement Petrology Committee. Cumboots Old Farmhouse, Scalby, Scarborough, North Yorks, YO!3 OPO

Abstract The paper briefly sets out the contribution which implement petrology has made, and continues to make, to archaeological enquiry. The Implement Petrology Committee (IPC) has its origins in pioneer work by Keiller, Piggott and Wallis (Keiller et al. 1941) in 1936 which was later incorporated into the Council for British Archaeology at its inception in 1945. IPC members have been, and still are, at the forefront of petrological research and development on prehistoric stone tools. Enduring problems are recognised, previous successes acknowledged, and future directions explored.

IPC Mission The IPC is responsible for the British Implement Petrology Programme which was one of the earliest in Europe to analyze prehistoric artefacts. It demonstrated that Neolithic societies were involved in large scale extraction, manufacture and trade in stone tools. Members o f the Committee have carried forward this work continuously for the past sixty years through a series of individual and collaborative research programmes involving laboratory-based investigations, replicative studies, and fieldw ork at quarry sites and manufacture sites. They have also supervised research on stone axes and their petrology; it has been particularly rewarding for the Committee to invite former research students to full Membership. Justifiably, the Committee feels a strong sense of continuity and purpose, a tradition which it is eager to further in the spirit of its founders whilst remaining responsive to contemporary needs and circumstances.

The Past The petrographic analysis of stone implements has been carried out since the 1930’s when the museums of South Western England combined to organise a systematic survey of the Neolithic stone axes in their collections. Five reports illustrate the origins of the work of the present Committee (Keiller, Piggott and Wallis 1941; Stone and Wallis 1947, 1951; Evens et al. 1962; Evens, Smith and Wallis 1972). More recently, the Committee has published its results in two CBA Research Reports (Clough and Cummins 1979, 1988), providing petrological identifications, distribution maps, regionally based articles and a database for over 7,600 artefacts in the United Kingdom. These represent the great bulk o f axes in museum collections and the lists are a major resource for any prehistorian seeking to understand the dynamics and mechanics o f production and exchange within British Neolithic society. Many petrographic analyses appear as notes in published excavation reports, others as appendices to major surveys (for example, Davis in Cherry and Cherry 1987), others within the fabric of textbooks (for example, Bradley and Edmonds 1993), others as captions in museum

displays. Since 1952 the organisation o f the continuing work o f sectioning and identifying axes nationally, and keeping up to date with new finds, has been the chief responsibility o f the IPC. In the past the Committee thin-sectioned all axes and similar implements in non-flint materials available to it, but now work concentrates more on well provenanced and dated artefacts. This work has revealed the large-scale production and widespread dispersal o f these implements and led to the recognition of sophisticated exchange systems throughout Neolithic Britain.

The Present With the publication of its second Research Report (Clough and Cummins 1988) the Committee’s work might be said to have been achieved. However, it would be truer to say that this publication was only a foundation for a more refined analysis and a more subtle understanding o f the petrological evidence; many of the difficulties o f analysis and interpreta­ tion discussed by Davis (1985) still remain. On the petro­ graphic side there is a need to re-assess the original groupings and re-consider many of the ungrouped rocks; on the archae­ ological side the typology and context of axes as cultural markers (Wright 1977) need to be more closely examined. The work of the Committee, therefore, is entering a new and important phase of consolidation and discrimination. A database, such as the Committee has provided, gains in value as its completeness increases and its maintenance must be a matter o f steady, if undramatic work. The Comm ittee provides a network o f archaeologists and petrologists throughout the country. Anyone, excavator, curator or amateur fieldworker who finds an axe or other stone implement, is asked to contact the Archaeological Recorder for their region who can then arrange the petro­ logical analysis o f the specimen. A current list o f regional Archaeological Recorders and Petrologists can be obtained from the CBA. Unfortunately a small charge has to be made to over the cost of sectioning. In the case o f well-dated, excavated material, which is particularly desired, this should be a charge on the post-excavation account.

R. V. Davis

2

Since 1992 the activity of the Committee has divided into three spheres: petrology (routine investigations and training workshops), monitoring the database (upgrading records), and custodianship (storage o f paper and thin section archives). The Committee will continue to collaborate in all these areas with other organizations, such as university departments and museums services, the Lithics Society, the Irish Stone Axe Project and the Scottish Stone Axe Project. The Committee is also looking to extend its UK network overseas, across Europe and elsewhere. The Committee’s work will continue to generate publications and valuable information for archaeologists and the general public. A major goal is the publication of a definitive Petrological Atlas of Prehistoric Stone Implements which will consist of descriptions and photomicrographs in transmitted and reflected light of the 34 axe source rocks. It is our intention to follow up this Atlas with an electronic version suitable for home access using PC software, and on the Internet. Optical microscopy still remains the central basic scientific technique used by the IPC for the initial analysis of stone implements. But the Committee has always been keen to promote appropriately the use o f more sophisticated geochemical analytical techniques to help solve petrological and archae­ ological problems. For example, as long ago as 1981 the Committee was using XRF and XRD data to help differentiate between similar rocks from widely dispersed implements; fifteen years ago the Committee introduced the use o f backscatter SEM techniques and microprobe analysis to investi­ gate the sourcing and distribution of axes in the North West English Highland Zone (Fell and Davis 1988). More recently, the Committee has begun to explore the use of reflected light microscopy to help characterise opaque minerals.

The Future Naturally, owners would prefer to avoid damage to their im plem ents, indeed access to som e exceptionally fine specim ens has been denied in the past because o f the destructive (but now effectively repairable) cutting or coring techniques used to obtain a rock sample for analysis. So the developm ent o f non-destructive techniques (W illiamsThorpe e ta l 1995, and see also Kayani and McDonnell, this volume) is of particular interest to the Committee. Similarly, the C om m ittee looks forward to future collaboration (Crompton, this volume) in the application of the GRF-ND Landmark Analysis Programme to study the technology and morphology of roughout stone tools. An illustrative list of future IPC-related research and development includes: -

-

-

a study of axes and other artefact types using portable XRF and magnetic characterisation. This is pioneering research which aims to assist in the reinterpretation of current theories concerning the procurement, dispersal and use o f these important cultural markers in prehistory building on the application o f the recently-developed laser microprobe argon-argon techniques to give a geological date for a rhyolitic tuff stone axe fragment from the Stonehenge environs. The method requires only milligramme-sized samples and gives dates of sufficient accuracy to aid in provenancing artefacts to sources, as well as information on the heating histories of samples IPC data is held on computer by English Heritage (and the CBA). Access to the database is normally through the IPC, not because we wish to restrict its use, but

-

because, like all specialist databases, guidance about its use is advised, especially when its data is to be interpreted by researchers not familiar with its scope and limitations. Work which aims at improving data quality and ac­ cessibility by developing a PC-compatible facility will continue, facilitating improved local access to regional and national implement petrology data-sets using stand­ alone PCs and the Internet. Textual and numeric data will be available for individual implements and, in some cases, a photomicrograph o f the rock in thin section, location map and scale drawings w ill be provided. historically, research on prehistoric stone implements from SW England has formed the backbone o f the development o f implement petrology nationally. The present reinterpretation o f implement petrology data in light o f recent geological research on the origin and distribution o f Cornish Greenstones and extrusive igneous rocks (Floyd, Exley and Styles 1993) suited to stone tool manufacture will continue.

T h e G raidlw yd A xe F actory: a case-stu d y The IPC poster illustrated a range o f IPC activities o f which excavation and petrographic research at the Graiglwyd Axe Factory, Penmaenmawr by two IPC members, D. Jenkins and J. LI Williams, is illustrative o f work associated with the IPC. The stone axe factories o f Penmeanmawr are the source for the most important stone implement group in Wales (Group VII). They exploit the eastern chilled margin (pyroxene granophyre) o f a large intrusion (hypersthene microdiorite). The Graiglwyd site was in fact one o f the first o f its type to be surveyed and examined in the British Isles. The discovery was made in 1919 by Hazzledine Warren and excavations were conducted by him over the next two summer seasons during which five prehistoric working floors and numerous sites with struck stone flakes were identified. Unfortunately few records of the excavations themselves remain and no substantial investigations have since taken place. The present investigation therefore arose partly from an acute awareness o f our own ignorance o f the Graiglwyd axe making com plex, the realisation that one o f the most important Neolithic sites in Britain had been for so long ignored and neglected, and partly to assess the claims made by some archaeologists that the greater part o f the site had been destroyed by quarrying. Warren’s principal site on the northern scree slope was re-excavated in 1993 when an undisturbed baulk was located under a wall containing a roughout, abundant flakes and small amounts o f charcoal; dates are awaited. A detailed EDM survey was recently carried out which identified several new areas o f potential quarrying/axemanufacture. Of these the most significant site is a quarry in a rock outcrop on the crest o f Graiglwyd, associated with flakes and roughouts. There are also high concentrations of struck flakes in adjacent mound-like features which appear to be unique to Graiglwyd. This site, on which excavations were started in 1993, is the first o f its type to be identified at Graiglwyd, since all the others investigated by Warren exploited loose scree rather than rock outcrop. Other aspects o f axe-quarrying, such as variations in petrology, the weathering o f the rock, and the palaeo-environm ental

Implement Petrology

0 G

3

50m H

Figure 1. Possible early quarry sites at the Graiglwyd axefactory. (Repr. by permission o f the Gwynedd Archaeological Trust).

context, are also currently under investigation. The IPC believes that such reinvestigations will lead to a cautious reinterpretation of archaeological theories taking account o f new evidence properly interpreted by those with appropriate knowledge, skill and understanding of the data and how its was gathered. Parallel investigations by IPC members concern Group XII and the Stonehenge ‘Bluestones’, Group I and the Cornish ‘Greenstones’, the Irish Group IX porcelanite, and the Scottish Group XXII riebeckite felsite. The outcomes of these and other investigations will be communicated, in time, in the usual way. In conclusion, implement petrology was one of the earliest scientific material analysis programmes set up, well before the main archaeological science programme began in the 1960’s. It provided a powerful indicator of the non-parochial

nature o f British Neolithic societies. The British Implement Petrology Programme, together with the work o f Professor Goit in Brittany, acted as the model for European-wide petrological analysis o f Neolithic and Bronze Age stone artefacts. Analyses are under way in Mediterranean coastlands (variscite,jadeite), central Europe (amphibolite, schist, basalt), north-east Europe (serpentine), the Low Countries (jadeite, phanite), Scandinavia (dolerite), France and neighbouring countries (dolerite, fibrolite); and the Scottish and Irish Stone axe surveys are each adding valuable new information to the archaeological record. Members o f the IPC have provided generations o f archaeologists with primary research data for more than sixty years. Although membership of the IPC is by invitation only, the Committee welcomes expressions o f interest in its work from archae­ ologists and petrologists in UK and abroad.

R. V. Davis

4 Acknowledgements

Identification of Stone Axes. Proceedings of the Prehistoric

The author wishes to gratefully acknowledge the help received from Pat Phillips, Joan Taylor and other IPC colleagues and especially to David Jenkins and John Williams who provided the Graiglwyd case study material reported in this paper, and to the Gywnedd Archaeological Trust.

Society 29: 209-266

References Bradley, R. and Edmonds, M. 1993. Interpreting the Axe Trade: Production and Exchange in Neolithic Britain. Cambridge, Cambridge University Press. Clough, T.H.McK and Cummins, W.A. (eds) 1979. Stone Axe Studies, Volume 1. London, CBA Research Report No. 23. Clough, T.H.McK and Cummins, W.A. (eds) 1988. Stone Axe Studies Volume 2. London, CBA Research Report No. 67. Davis, R.V. 1985. Implement Petrology: the state of the art some problems and possibilities, in TheArchaeologist and the Laboratory (ed. P Phillips). London, C.B.A. Research Report No. 58, pp. 33-35. Davis, R.V. 1987. An analysis of prehistoric lithic material fromthe limestone uplands of Cumbria between Shap and Ash Fell. In

Prehistoric Habitation Sites on the Limestone Uplands of Eastern Cumbria (eds J. Cherry and P.J. Cherry). Cumberland & Westmorland Antiquarian and Archaeological Society Research Series 2, pp. 77-84 Evens, E.D., Grinsell, L.V., Piggott, S. and Wallis, F.S. 1962. Fourth Report of the Sub Committee of the South-Western Group of Museums and Art Galleries on the Petrological

Evens, E.D., Smith, I.F. and Wallis, F.S. (1972). Fifth Report of the Subcommittee of the South-Western Group of Museums and Art Galleries on the Petrological Identification of Stone Axes. Proceedings of the Prehistoric Society 38: 235-275. Fell, C.I. and Davis, R.V. 1988. The petrological identification of stone implements from Cumbria, in Stone Axe Studies (eds T. Clough and W. Cummins). London, C.B.A. Research Report No. 67, pp. 71-77. Floyd, P.A., Exley, C.S. and Styles, M.T. 1993. Igneous Rocks of South-West England, Geological Conservation Review Series, Chapman and Hall. Keiller, A., Piggott, S. and Wallis, F.S. 1941. First Report of the Subcommittee of the South-Western Group of Museums and Art Galleries on the Petrological Identification of Stone Axes, Proceedings of the Prehistoric Society 7: 50-72 Stone, J.F.S. and Wallis, F.S. 1947. Second Report of the Subcommittee of the South-Western Group of Museums and Art Galleries on the Petrological Identification of Stone Axes, Proceedings of the Prehistoric Society 13: 7-55 Stone, J.F.S. and Wallis, F.S. 1951. Third Report of the SubCommittee of the South-Western Group of Museums and Art Galleries on the Petrological Identification of Stone Axes, Proceedings of the Prehistoric Society 17: 99-158. Williams-Thorpe, O., Potts, P.J., Webb, P.C. Tindle, A.G. and Jones, M.C. 1995. Non-destructive, field-portable methods of analysis for archaeological stone. Paper presented at the confer­ ence, on Archaeological Sciences 1995, Liverpool, July 1995. Wright, R.V.S. (ed.) 1977. Stone Tools as Cultural Markers: Change Evolution and Complexity. Canberra, Australian Institute of Aboriginal Studies, Prehistory and Material Culture Series No. 12.

The Potential of Scanning Electron Microscope Techniques for Non-destructive Obsidian Characterisation and Hydration Rim Dating Research Peter I. Kayani and Gerry McDonnell Department o f Archaeological Sciences, University o f Bradford, Bradford BD7 1DP

Abstract The scanning electron microscope (SEM) is one o f the most versatile, and widely available, instruments for non­ destructive microstructual analysis. This paper assesses the potential o f a number o f SEM techniques that can be used to analyse archaeological obsidians, with particular reference to determining the morphological, chemical and crystallographic characteristics of included microcrystalline phases present in obsidian, together with their relative and absolute abundances and sizes and also the resolution of hydrated glass (perlite) zones. Preliminary results on obsidians from known archaeological sources in the Mediterranean area are reported. These results suggest that SEM-based approaches offer a powerful alternative to existing non-destructive characterisation methods, particularly for combined obsidian characterisation and hydration rim research.

Introduction In a recent paper, Williams-Thorpe (1995) identified one of main reasons for the limited progress in Old World obsidian lithic studies since the 1980s as arising from the lack of a rapid, low -cost, generally available and non-destructive sourcing method which would allow provenancing to become a routine part of post-excavation work. In addition, the lack o f use o f obsidian artefact hydration rim dating in the Mediterranean and the Near East compared to the U.S.A was also highlighted (Willams-Thorpe 1995: 219). This paper explores the potential o f the scanning electron microscope for non-destructive obsidian characterisation and for combined characterisation and hydration rim dating research. The only reliable high-precision method for whole rock rare earth element (REE) analysis that can be used on asreceived artefacts rather than samples is non-destructive instrumental activation analysis (INAA) (Kayani et a l 1994). However, non-destructive INAA requires highly specialised, and therefore, generally inaccessible, instrumentation (Warren 1973). The most widely available non-destructive method for w hole rock REE analysis is w avelengthdispersive x-ray fluorescence (WDS-XRF) analysis (Kayani et al. 1994). However, high-precision WDS-XRF is de­ pendant on sampling several grams of an artefact (Bouey 1991; Williams-Thorpe 1995) and is, therefore, generally inappropriate for analytical studies o f archaeological obsidian (Kayani et al. 1994). Attempts to develop alternative non-destructive and lowcost instrumental methods for whole rock analyses have only been partially successful (Kayani et al. 1994). Work at Bradford University in the 1980s using natural obsidian j3 particle emission (Leach and Warren 1981) and whole rock magnetic analysis (McDougall et al. 1983) indicated that these parameters showed significantly less between-source variation than reported for REE concentrations (e.g. Aspinall

etal. 1972; Hallam et al. 1976; Francaviglia 1984) resulting in relatively poor source discrimination, and in some cases, source overlap. Such methods are only likely to be useful in terms o f reducing the number o f whole rock REE analyses required (Williams-Thorpe 1995) A more promising technique, recently pioneered by Burton and Krinsley (1987) for obsidian from the South­ western United States, is back-scattered electron (BSE) petrography. As BSE-petrography requires only a con­ ventional scanning electron microscope (SEM), one o f the least specialised and therefore generally available analytical instruments, it is among the most accessible and non­ destructive char-acterisation methods available to geo­ archaeologists.

Background Although amorphous on a macro scale, obsidian typically contains a significant com ponent o f m icroscopic and submicron sized crystallite phases (Burton and Krinsley 1987). These crystallites comprise o f minerals incorporated from surrounding rocks, primary phenocrysts and microphencrysts developed in the magma chamber, post-eruption crystallites that precipitated as a result of the rapid cooling o f the magma or in the subsequent glass, and the products of sub-solidus processes such as devolatilzation and de­ vitrification (Kirkpatrick 1975; Swanson 1977). The numbers, sizes and textures o f these crystallites result from a unique set o f kinetic factors such as the degree o f superheating (which affects the number of seed nuclei and melt polym erisation) and supercooling (which affects chemical and thermal diffusion) (Kirkpatrick 1975; Swanson 1977) and are therefore likely to characteristic o f a particular obsidian source. Obsidian surfaces in contact with atmospheric water vapour gradually hydrate with time. A thin hydrated-glass

Peter 1. Kayani and Gerry McDonnell

6

rim is formed at a rate mainly determined by the composition of the obsidian and average ambient temperature. The thickness of the hydration rim is therefore time-temperature dependant and so the measurement of the hydration rim (using artefact thin-sections) formed on obsidian artefacts from the same site gives a relative indication of time since last working. Using known-date and known-composition standards (i.e. artefacts which have been dated by a absolute method such as radiocarbon dating) average ambient temperatures for a site can be determined and therefore absolute dates can be obtained from known-composition obsidian artefacts (e.g. Ericson 1988).

W hole Rock Analysis SEM’s have been increasingly used in recent years for the xray microanalysis of synthetic glasses using either WDS or EDS detectors (e.g. Henderson 1988; Verita et a l 1994). The advantages of SEM-WDS or SEM-EDS are principally the ability to distinguish and analyse, or exclude, inclusions, opacifiers and weathered areas (Verita et al. 1994). Additionally, elemental analysis using the SEM is more rapid than other conventional methods and is also non-destructive in that it allows multiple analyses to be made on the same sample (Verita et a l 1994). The WDS and EDS detectors have significantly different analytical performances for the analysis of glasses. It is known that EDS detectors suffer from inferior peak resolution and overlap (e.g. between aluminium and magnesium) compared to WDS detectors resulting in poorer precision (Verita et al. 1994). Minimum detection limits for EDS detectors are significantly worse than those for WDS detectors due to a poorer signal-to-noise ratio. While generally performing better than EDS detectors, WDS detectors suffer from the need to use a higher electron beam energy that results in the migration of sodium ions away from the point o f analysis, a potential source of error in analyses (Henderson 1988; Verita et al. 1994). The poor sensitivities of both WDS and EDS detectors to REE’s (Verita et al. 1994), found to be the most effective elemental discriminators o f obsidian sources (e.g. Francaviglia 1984), is probably the main reason for the limited use o f SEM-based microanalysis for obsidian characterisation studies. Sample masses generally preclude obsidian traceelement analysis using the SEM. An alternative approach is to use minor and major element concentrations to distinguish between obsidian sources. Except for the element’s alu­ minium and magnesium, EDS and WDS detectors are

broadly comparable in terms of precision and accuracy (Verita etal. 1994). However, the reported extent o f obsidian between-source variation in minor and trace elements is significantly less than for the REE’s (Bird et al. 1983) and, therefore, often do not give unambiguous source dis­ crimination (Francaviglia 1984: 330; Tykot 1994). However, major element microchemical analysis in the SEM can be useful as an initial screening procedure in distinguishing between the three broad types o f obsidian, alkaline, calcalkaline and peralkaline. A lso, sin ce the rate o f the development o f hydration rims is dependant on obsidian composition, the major element chemistry o f an artefact is required in order for an absolute date to be determined by this method (Ericson 1988). SEM-based major element analysis, therefore, has a potential role in obsidian hydration rim dating.

Backscattered Electron Atomic Number Contrast (BSE-Z) Imaging BSE-petrography uses reflected (typically 10-30 keV) (e.g. Burton 1987; M eeks 1988) high-energy electrons, the number o f which, collected by a BSE detector in an SEM, is proportional to the mean atomic number o f the phase under the incident electron beam. This dependence on mean atomic number produces atomic number contrast images where the grey-scale o f the image is a function o f relative atomic number differences between phases. A relatively high mean atomic number phase such as iron oxide therefore appears ‘brighter’ than a phase o f relatively low mean atomic number such as quartz in the same image (e.g. Figure 1). This compositional imaging ability o f BSE allows the study of chem ical and textual relationships am ong and within minerals on a centimetre to micrometre scale. The back-scatter coefficient (?]) is defined as the ratio of the total BSE current received by the BSE detector to that of the incident SEM electron beam and for pure elements its relationship to atomic number (Z) for elements greater than Z=10 is given by equation 1 and is graphically illustrated in Figure 2.

EQUATION 1:

?7 = [ ¥ ■ - ? ]

for Z > 1 0

For single phase specimen’s (i.e. single crystals) 7] is constant and the image consists o f uniform contrast. For poly phase specimens (i.e. polycrystalline materials), rj is variable and the image consists o f different contrasts with higher atomic number phases being brighter.

Table 1: Typical major element compositional ranges fo r obsidians from different sources world-wide and the associated measurement precision o f SEM-EDS and SEM-WDS (modified from Cann 1983 and Verita et al. 1994). Element

Typical Compositional Range (wt. %)

EDS Precision (± 2cr @ 10 wt. %)

WDS Precision (± 295%) of the BSE signal is a result of atomic number contrast. However, when imaging crystalline phases, a contribution to the total BSE signal is made by the interaction between incident electrons and the structure o f the crystal. This is known as the electron channelling effect and is a result of the range o f electrons within a target specim en (i.e. 1 -5 pm is now pseudomorphed by white T i0 2 minerals. Relict titanomagnetite is absent, it being totally pseudomorphed by 4 -2 0 pm diameter crystals o f titanite with white internal reflections. Trace amounts of chromite form rounded to euhedral crystals and show slight alteration to pale-coloured, poorly characterised ferritchromit rims. Sulphides, in m inor amounts, are widely distributed, either as discrete, coarse­ grained crystals or more often as smaller disseminated grains, forming sulphide-rich patches, Pyrrhotite ± pentlandite exsolution flames (SH43, 65, 61) and chalcopyrite are more abundant than pyrite and marcasite which may be alteration products after pyrrhotite. The majority o f the sulphides, especially the sulphide-rich patches, are associated with white mica, epidote and quartz which themselves are part of the metamorphic suite o f minerals. Euhedral rhombic titanite is present in the matrix o f the rocks. The petrography o f the SH33 group is remarkably consistent, i.e. it defines a very ‘tight’ group, with only SH61 showing minor differences, for example in having very coarse-grained chromite 300 pm, and pyrite up to 200 pm in diameter.

SH42, SH44 and SH45 SH42 differs from group SH33 in one significant respect, namely the absence of thin ilmenite oxidation-exsolution lamellae in altered magnetite/titanomagnetite. In all other respects it is similar. M acroscopically both SH44 and 45 are unspotted dolerites and under the microscope they form a loose group

14

R A . Ixer

with som e sim ilarities to SH42. W hilst much o f the petrography o f SH44 and SH45 is similar to that of group SH33, the rocks display the following significant differences. Titanomagnetite/magnetite has altered to white or pale, peach-coloured titanite devoid o f ilm enite oxidationexsolution lamellae (SH44) or locally showing very densely packed lamellae (SH45). Chromite is absent from SH44 or small and very rare in SH45 and titanite forms small, up to lOpmm diameter, anhedral grains rather than euhedral rhombic crystals. Ilmenite, altered ilmenite and sulphides have the same mineralogy and textures and approximate grain size as those seen in group SH33, other than a relative paucity o f pyrrhotite in SH44 and 45 and the presence of cubanite exsolution lamellae in chalcopyrite in SH44.

SH62 Petrographically SH62 show s the greatest number o f differences from group SH33. Titanomagnetite has altered to dense, white titanite that carries very rare, altered ilmenite oxidation-exsolution lamellae. Relict ilmenite is absent and ilmenite has altered to orange- and white-coloured TiOa minerals and titanite.

Dole rites fro m Preseli Outcrops Dolerites taken from the Preseli Mountains outcrops (figs 1 and 2) show a wider range in their detailed petrographical characteristics than do the Stonehenge monolith dolerites. However, it must be re-emphasised that both monoliths and outcrop samples share enough petrographical characteristics to group them all as Group XIII. In hand specimen Carn Ddafad-las (CM 12), and Craig Talfynydd (T F 2,3) are unspotted dolerites, Carn Marchogion (CMA1) is ‘less spotted’ and Cam Menyn (CM2) is a weakly spotted/unspotted dolerite. At all other outcrops spotted dolerites were sampled (Thorpe et al., 1991).

Carn Menyn (CM1, CM2, CM11) and Carn Gyfrwy (CM10) Although the outcrops lie within one kilometre of each other, there are significant petrographical variations between these dolerites especially in the textures and alteration of their iron-titanium oxides. Millimetre-sized titanomagnetites carry two generations o f ilm enite oxidation-exsolution lam ellae along (111), (CM 10, CM11). The thinner lamellae (>2 pm) have altered to white T i0 2 minerals, whilst the thicker, 5 -2 0 pm but up to 40pmm in CM 10, are replaced by brown-coloured T i0 2. In CM1, however, the oxidation-exsolution lamellae are thin and rarely present and are absent from CM2. The titano­ magnetite host is altered to loosely-packed white titanite except for CM2 where fine-grained ilmenite is present in darker cores within pale titanite pseudomorphs after titanomagnetite and may be relict ilmenite from internal composite crystals or late-stage neomorphic ilmenite. In all the sections magnetite and tabular ilmenite form internal and external composite crystals. The alteration of ilmenite varies between the outcrops but in all cases relict ilmenite is present. Relict ilmenite is surrounded by bluecoloured T i0 2 (rutile); that in turn is altered to titanite (CM1, CM2); or is accompanied by a highly anisotropic? carbonate phase and then alters to orange-coloured T i0 2 (CM 11); or is accom panied by fine-grained haematite (CM 10). The

presence o f haematite as part o f the alteration o f ilmenite is unique to CM 10 in the present study despite, elsewhere, it being a very common and widespread mineral replacing ilm enite (Ixer 1990). The fine-grained nature o f the haematite-Ti02 intergrowths suggest that they may be a more altered form o f ‘speckled ilmenite’. All the rocks, except for CM1, carry large, rounded to euhedral chromite grains that show dissolution features and ferritchromit alteration rims. All carry up to 300 pm diameter disseminated sulphide patches comprising chalcopyrite or chalcopyrite and pyrite, these patches are associated with secondary silicates, including quartz and titanite. Small chalcopyrite or hexagonal pyrrhotite grains ± pentlandite exsolution flames cement acicular box-work T i0 2 crystals within altered ilmenite or form discrete inclusions in epidote. Chalcopyrite alters to digenite, covelline and ‘idaite’ and pyrrhotite to pyrite/marcasite along (0001); all the sulphides are oxidised to limonite. Discrete rhombic crystals o f titanite are present in all the dolerites.

Carn Goedog (CGD 1, 2), Carn Breseb (CBR1) and Carn Bica (CB1) These three outcrops, each approximately one kilometre apart (fig. 1), share many petrographical features. Millimetre diameter titanomagnetite has altered to white or pale, lime-green coloured titanite although the cores of some altered titanomagnetites are darker and carry small magnetite/ ilmenite grains (CBR1). Ilmenite oxidationexsolution lamellae, 1-2 pm in thickness and altered to white T i0 2 and accompanied by thicker lamellae that have altered to pale orange T i0 2 minerals, are common except for Cam Bica (CB1). Here thin lamellae are rare to absent. A ll the rocks carry the same internal and external composite ilmenite-titanomagnetite intergrowths and most show the same type and degree o f ilmenite alteration. Relict ilmenite, altering to ‘speckled ilmenite’, is surrounded by pale-coloured T i0 2 minerals accompanied by a highly anisotropic mineral and these in turn alter to a triangular mesh- work o f pale orange, acicular T i0 2 crystals, all this is enclosed within a titanite rim. In CDG2, relict ilmenite is absent, it is totally pseudomorphed by orange-coloured T i0 2 minerals and titanite. Chromite, up to 100 pm in diameter, is present in CB 1 but is smaller in CBR 1 and CDG1/2, but in all cases shows oxidation to ferritchromit. Disseminated chalcopyrite or hexagonal pyrrhotite grains form sulphide patches associated with secondary silicates and both minerals cement acicular T i0 2 crystals within altered ilmenite. Chalcopyrite alters to spionkopite and pyrrhotite to limonite along (0001) planes. Euhedral, rhombic crystals o f titanite are present in all the rocks and trace amounts o f pyrite are present in CBR1 and CGD1.

Carn Marchogion (CMA2), and Craig Talfynydd (FT 2, 3) Despite the distance between the outcrops o f about two kilometres, their rocks share many characteristics. In all sections magnetite, now altered to white titanite, carries rare to very rare 1 -2 pm oxidation-exsolution lamellae now altered to white T i0 2 minerals; a second generation o f very rare, thicker lamellae, some o f which retain relict ilmenite, is present in CMA2 and TF2. Ilmenite and titanomagnetite form internal and external composite mixed crystals. Relict ilmenite alters to T i0 2 minerals and then to titanite in CMA2 and TF2 but the more

Detailed Provenancing of the Stonehenge Dolerites extensive alteration sequence o f ilmenite to ‘speckled ilmenite’ to pale T i0 2 to orange T i0 2 plus a titanite rim is seen in the rocks from Craig Talfynydd (TF3). Chromite is present in these same rocks but not seen in CMA2. A ll the sections show abundant sulphides notably hexagonal pyrrhotite often intergrown with chalcopyrite. Although disseminated sulphide patches associated with later metamorphic minerals are present, the majority o f the sulphides are discrete and coarse-grained. Pyrrhotite ± pentlandite exsolution flames is intergrown with lesser amounts of chalcopyrite and pyrite and all are oxidised to limonite. Titanite is present as small, poorly crystalline aggregates and as euhedral rhombs (TF2, 3).

Cam Ddafad-las (CM 12) Three polished thin sections were made and they show some petrographical variation. Titanomagnetite has altered to dense, white titanite, some with darker coloured cores. White T i0 2 pseudomorphs after thin ilmenite oxidation-exsolution lamellae are locally present but more commonly they are very rare or absent. Ilmenite shows the typical alteration sequence of relict ilmenite, altering to T i0 2 minerals and a highly anisotropic phase, themselves altering to orange-coloured T i0 2 and enclosed within a titanite rim. Rare, partially dissolved, chromite is present. Small disseminated chalcopyrite and/or pyrrhotite grains form sulphide patches. Both sulphides are present within epidote and both are altered to limonite. Titanite forms discrete rhombic crystals.

Petrographical Summary In terms of their petrographical features the outcrop dolerites can be grouped by selected characteristics as follows: 1. The presence/absence, thickness and density of ilmenite oxidation-exsolution lamellae in titanomagnetite. Two generations of lamellae are present and common in CBR 1, CDG1 and 2, CM 10 and CM11; lamellae are locally common to absent in CM1 and CM 12; lamellae are very rare in CMA2, TF2 and 3 and CB1 and are absent in CM2. In CM 10 the degree o f developm ent o f the oxidation-exsolution lamellae is particularly extensive and some o f the lamellae are notably broader than in any other section. 2. The degree and type of ilmenite alteration. Outcrops that show the complete alteration sequence from relict ilmenite to ‘speckled ilmenite’ to pale T i0 2 minerals ± an anisotropic phase to orange T i0 2 and a titanite rim are CB1, CDG1 and 2, CM10, CM11, CM 12, and TF3; those showing the alteration sequence relict ilmenite to T i0 2 minerals to titanite are CBR1, CM1, CM2, CM 12, CMA2 and TF2. CM 10 is unique in having haematite as part of the alteration sequence, and CDG2 in having no relict ilmenite. 3. Chromite is absent only from CM1 and CMA2; coarse­ grained sulphides, especially hexagonal pyrrhotite, are abundant in CMA2 and TF2; and CM1 carries abundant pyrite.

15

Discussion Although based upon empirically derived, minor, petro­ graphical variations, namely the presence/absence of, and textural intergrowths between, the iron-titanium-chromium oxides and their alteration products, detailed reflected light petrography is shown to be sensitive enough to subdivide samples within a single overall lithology - in this example an altered ophitic dolerite. This subdivision would be even easier if reflected light is supplemented by transmitted light petrography using the same principles but different minerals. Ixer (1994), working on basaltic rocks, proposed that primary igneous petrographical features, namely mineral species and their textural intergrowths, are more sensitive indicators of local conditions than those associated with alteration and weathering and so should be given greater importance when determining provenances. The present study confirms this proposal and suggests that it is true for coarser grained rocks and further, that metamorphic characteristics (as far as opaque petrography is concerned) are also less sensitive than primary igneous ones. Multiple sections from Carn Ddafas-las (CM 12) show there are significant petrographical variations within an outcrop and that these are comparable to inter- outcrop variation. Thorpe et al. (1991) reported similar inhomo­ geneities from Carn Gyfrwy (CM 10) over 900 m123in terms o f its geochemistry and Carn Menyn in terms o f its spotted to unspotted characteristics. Multiple sections o f SH45, however, showed few differences but were only sampling centim etre-size chips. Inter-outcrop variation must be accounted for by dedicated sampling, probably on the scale o f metres. Altered ophitic dolerite is a very common rock type and its opaques are illustrated in a number o f texts, notably Haggerty (1976a, b) and Ixer (1990), although opaques in metamorphosed dolerites are less w ell described (Ixer 1990). Comparison of the Stonehenge-Preseli outcrop dolerites with other altered dolerites show a number of unusual features, in particular the almost total absence of haematite (except for CM 10) as an alteration product o f ilmenite (or titano­ magnetite); the type and degree o f ilm enite alteration especially the developm ent o f the crystallographically orientated network o f T i0 2 crystals; the degree of dissolution o f chromite and the amount and grain size o f the sulphides, most notably hexagonal pyrrhotite. It is these features that unify all the samples into a single group, Group XIII. Whilst accepting and acknowledging the limitations imposed on the petrographical data due to imperfect sampling o f outcrop inhom ogeneities, a preliminary comparison between the Stonehenge monoliths and Preseli outcrops will be made. Group SH33 (SH33, 37, 43, 49, 65, 67 plus 61) are a petrographically ‘tight’ group; characterised by an abundance o f two generations o f ilmenite oxidation-exsolution lamellae in titanomagnetite, by the complete alteration sequence of relict ilmenite to white/orange T i0 2 within a titanite rim and the presence o f chromite. A comparable petrography is seen in rocks from Cam Goedog (CDG1), Cam Gyfrwy (CM 10) and Cam Menyn (CM11). However, CM10 is unique in having haematite present and in the intensity o f the oxidationexsolution lamellae within titanomagnetite and so differs significantly from Group SH33. Thorpe et al. (1991) have suggested that the probable geological source for members o f group SH33 were the East

16

R A. Ixer

Preseli outcrops Carn Menyn - Carn Gyfrwy or Cerrig Marchogion - Cam Geodog. This suggestion, largely based upon lithogeochemistry, is in accord with the petrographical data, but these new data further constrain the source to Cam Goedog (CDG1) or the CM 11 outcrop at Cam Menyn. Cam Gyfrwy (CM 10) is less likely and the petrographical match of group SH33 and Cerrig Marchogion is quite poor, as are outcrops CM1 and CM2 from Carn Menyn. Stonehenge SH44 and SH45 are characterised by titanomagnetites with ilmenite oxidation-exsolution lamellae being common locally but generally rare to absent. Chromite is also absent or very rare and ilmenite shows the complete alteration sequence from relict ilmenite to pale to orange T i02 in a titanite rim. Dolerites from Cam Bica (CB1) or Carn Ddafad-las (CM 12) are the closest petrographical match. Carn Bica, however, has coarse-grained altered chromite whereas at Cam Ddafad-las chromite is fine-grained or absent. Thorpe et al. (1991) suggest that SH44 and SH45 were from Cam Ddafad-las and opaque petrography supports this view. Stonehenge monolith SH42 is similar to group SH33 but lacks ilmenite oxidation-exsolution lamellae within altered titanomagnetite. Ilmenite alteration and the presence o f chromite are the same as group SH33. Petrographically, the closest similarities are with Carn Bica (CB1), Cam Ddafadlas (CM 12) or Carn Marchogion (CMA2), although the abundance of sulphides in CMA2 lessens this match. On lithogeochemical grounds Thorpe et al. (1991) suggested SH42 came from Cam Breseb (CBR1), but despite their strong geochem ical similarities the petrographical sim ­ ilarities between SH42 and Carn Breseb are poor, for exam ple Carn Breseb rocks have abundant oxidationexsolution lamellae. This sample constitutes a contradiction between the opaque petrography and geochemistry since Carn Bica and Carn Ddafad-las have significant chemical differences from SH42 (Thorpe et al. 1991). Monolith SH62 is difficult to match petrographically to any of the sampled outcrops. Ilmenite oxidation-exsolution lamellae in altered titanomagnetite are rare and it is the only monolith to show total replacement of ilmenite by T i0 2 minerals and titanite. Carn Gyfrwy (CM 10) and Cam Goedog (CGD2) dolerites also have no relict ilmenite but at Cam Gyfrwy haematite, uniquely, is present as part o f the alteration of ilmenite and Cam Goedog dolerite is char­ acterised by abundant oxid ation-exsolution lam ellae. Monolith SH62 plots within the centre of the SH33 group in terms o f its geochem istry as shown in Figures 9 and 10 (Thorpe et al. 1991), close to the dolerites from Cam Gyfrwy and Carn Goedog, hence the lithogeochemical provenance is probably correct. The opaque petrography mismatch may reflect insufficiently detailed sampling o f the monolith or outcrops, to cover all of the petrographical variations.

Conclusions 1. Ore petrography reconfirms that the Stonehenge bluestone dolerites come from the Preseli dolerite outcrops and belong to the Implement Petrology Group XIII. It strongly suggests that the monoliths come from a limited number o f outcrops or one that shows a wide range of petrographical characteristics. 2. Monoliths SH33, 37, 43, 49, 65, 67 plus 61 form a

generally very tight petrographical and geochemical group o f specimens suggesting they all come from a small area of no more than a few hundred square metres from Cam Goedog or Cam Menyn. Both petrographically and geochemically SH61 can be seen to be slightly different from the other members o f the group. 3. Both lithogeochemistry and petrography suggest that monoliths SH44 and SH45 come from the outcrop at Carn Ddafad-las. 4. Lithogeochemistry suggests that SH62 belongs to group 33 but petrographically no good match is available. 5. Lithogeochemistry suggests that SH42 has come from Carn Breseb outcrops but petrography suggest Carn Bica or Cam Ddafad- las.

S u ggestion s fo r F u rth er R esearch This preliminary work strongly suggests that with dedicated outcrop sampling, perhaps using ten metre spacings or less, lithogeochemistry together with petrography should be able to relate monoliths to discrete outcrops or to specific areas within an outcrop. Sampling of the other twenty-one (unsampled) Stone­ henge bluestone monoliths should be completed, following the same methods as those o f Thorpe et al. (1991), in order to determine the full extent of group SH33 or the presence o f any other petrographical-geochemical groups. Sampling and total petrographical examination o f preselite erratics, and resampling o f all group XIII material found in good archaeological contexts should be undertaken. Prior to this work, the petrographical study o f Group XIII artefacts has been restricted to transmitted light work using thin sections with cover slips. These cannot be used for detailed reflected light studies. If resampling o f artefacts is not possible, then limited ore petrography can be done using the existing thin sections although all mineralogical data are lost and much textural information becom es ambiguous and indistinct. A re- examination o f all Group XIII thin sections for their opaque petrography and matching this to the present study, may assist in determining the existence and extent o f subgroups within the dolerites. Indeed, initial results from work in progress from south­ west England’s axe-head collection suggest group SH33 to be very significant numerically. Only on completion o f all of this can the current debate on the removal and transport mechanism of the Stonehenge dolerite bluestones be dragged/carried forward.

Acknowledgements Drs R.S. Thorpe, O. Williams-Thorpe, V. Davies and R. Bevins are thanked for supplying or loaning dolerites from Stonehenge and/or Preseli and for encouraging discussions and critical improvements to the manuscript. Professor D.D. Hawkes is thanked for providing polished thin section facilities in the School o f Earth Sciences, Birmingham University. Mrs June Andrews is thanked for typing. The Palaeologos research fund is also acknowledged.

R eferen ces Atkinson, R.J.C. 1979. Stonehenge. Harmondsworth, Penguin Books. Berridge, P. 1994. Cornish axe factories: fact or fiction. In Stories

Detailed Provenancing of the Stonehenge Dolerites in Stone (eds N. Ashton and A. David). Lithics Studies Society Occasional Paper No. 4. pp. 45-56. Bevins, R.E. 1978. Pumpeyellite-bearing basic igneous rocks from the Lower Ordovician of north Pembrokeshire. Mineralogical Magazine 42: 81-83. Bevins, R.E. 1982. Petrology and geochemistry of the Fishguard Volcanic Complex, Wales. Geological Journal 17: 1-21. Bevins, R.E., Lees G.J. and Roach, R.A. 1989. Ordovician intrusions of the Strumble Head-Mynydd Preseli region, Wales: lateral extensions of the Fishguard Volcanic Complex. Journal o f the Geological Society o f London 146: 113-123. Bevins, R.E. and Rowbotham, G. 1983. Low-grade metamorphism within the Welsh sector of the paratectonic Caledonides. Geological Journal 18: 141-168. Clough, T.H Mck. and Cummins, W.A. (eds) 1979. Stone Axe Studies. London, Council for British Archaeology, Research Report No. 2. Clough, T.H. McK. and Cummins, W.A. (eds). 1988. Stone Axe Studies (Volume 2). London, Council for British Archaeology, Research Report No. 23. Clough, T.H. Mck. 1988. Introduction to the regional reports: prehistoric stone implements from the British Isles. In Stone Axe Studies ( Volume 2) (eds T.H. Clough and W.A. Cummins). London, Council for British Archaeology, Research Report No. 23. pp. 1-11.

17

Haggerty, S.E. 1976(a). Oxidation of opaque mineral oxides in basalts. In Oxide Minerals (ed. D. Rumble). Reviews in Mineralogy 3. Mineral Society of America, pp. 1-100. Haggerty, S.E. 1976(b). Opaque mineral oxides in terrestrial igneous rocks. In Oxide Minerals (ed. D. Rumble). Reviews in Mineralogy 3. Mineral Society of America, pp. 101-300. Ixer, R.A. 1990. Atlas o f Opaque and Ore Minerals in their Associations. Milton Keynes, Open University Press. Ixer, R.A. 1994. Does ore petrography have a practical role in the finger printing of rocks? In Stories in Stone (eds N. Ashton and A. David). Lithics Studies Society Occasional Paper No. 4. pp. 10-23. Stone, J.F.S. and Wallis, F.S. 1951. Third report of the sub­ committee of the south- western group of museums and art galleries on the petrological determination of stone axes. Proceedings o f the Prehistoric Society 17: 99-158. Thomas, H.H. 1923. The source of the stones of Stonehenge. Antiquaries Journal 3: 239-60. Thorpe, R.S., Williams-Thorpe, O., Jenkins, D.G., Watson, J.S. with Ixer, R.A. and Thomas, R.G. 1991. The geological sources and transport of the Bluestones of Stonehenge, Wiltshire, U.K. Proceedings o f the Prehistoric Society 57: 103- 157. Williams-Thorpe, O. and Thorpe, R.S., 1990. Millstone pro­ venancing used in tracing the route of a fourth-century B.C. Greek merchant ship. Archaeometry 33: 115- 137.

Assessment of Post-Excavation Systems for Archaeological Glass Using FT-IR Microscopy N J . Earl1 and D.E. W atkinson2 1 CRBE, The Nottingham Trent University, Burton St., NG1 4BU, Nottingham; 2Conservation Section, HISAR, University o f Wales College o f Cardiff, Cathays Park, Cardiff.

Introduction Since Roman times, glass has been in common use for vessel­ making and for use in windows. Its popularity is largely due to the qualities o f transparency and relative unreactivity. Its deposition in the soil may either be due to natural wastage or deliberate; for instance significant quantities o f Mediaeval stained window glass entered the archaeological record with the destruction of ecclesiastical buildings. It is principally the fate o f this material which provides information about, for instance, technology, colorants and provenance, with which this paper is concerned. Upon excavation, M ediaeval window glass is often opaque, with black surface layers around a sliver o f remaining “pristine” glass, and crumbles easily. In the majority of instances the glass will disintegrate if water, which is acting as a consolidant, is removed (Alten 1988). A ctive conservation approaches are usually either to consolidate the glass with aqueous resin emulsion, or to replace the water with acetone and use an acrylic resin; silanes may be used as a coupling agent. The size o f the assemblages retrieved, however, means that the elapse o f a time period, sometimes extending to years, is common before treatment. N o standard storage method has been established and experience has shown that significant further deteri­ oration occurs during storage. The aim of this research was therefore to examine the effects on glass of methods already in use as well as novel methods based on modifying the glass chemistry. In order to understand the changes that the various systems are likely to induce, it is necessary to consider the nature o f the corrosion processes to which glass is subject. A glass is a material cooled too quickly to crystallise, producing a thermodynamically unstable solid (Paul 1990). The energy required for crystallization is high enough for kinetic stability, although deterioration w ill occur in sufficiently aggressive environments or over long time periods. Glasses have no fixed chemical formula, resembling solid solutions o f several different components; stability is heavily dependent on composition. In this context, ‘glass’ refers to multi-component silicate glasses. The three major components are network formers, fluxes and stabilisers. Network formers give rigidity, silica (S i0 2), probably usually obtained from sand being the most significant; the higher the proportion, the greater the stability. Greatest stability occurs when silica constitutes above 66% o f the molecules present, corresponding to the point below which every silicon atom is associated with a non-silicon atom as its second neighbour. At this point the glass assumes a metasilicate composition (El-Shamy 1973). It is generally

accepted that the three-dimensional network is composed of S i0 4 tetrahedra linked together by Si-O-Si bonds, known as bridging oxygens, randomly arranged. Fluxes reduced the melting point, the most common being soda (Na2C 0 3) and potash (K2C 0 3). High proportions o f the latter were introduced to Mediaeval glass, by adding furnace ash (Turner 1956). On heating, C 0 2 was released, leaving the monovalent metal oxides the addition o f which breaks up the three dimensional structure. The Si-O-X bond is weaker than the Si-O-Si bond because it is more easily polarised. A divalent metal oxide, most commonly CaO and MgO, was often present and acted as a stabiliser to reduce the solubility of the binary, soda/potash silica glass. In Mediaeval glasses the percentages are high. The levels o f metal oxides in many Mediaeval glasses are so high that the glasses lack the critical amount o f silica required for optimum stability. There can be a reaction between water and the glass surface, resulting in the formation o f a silica-enriched layer. Initial signs are iridescence and dulling. It was demonstrated (Emsberger 1980) that this process involves the exchange o f alkaline oxide molecules with water molecules and the breaking o f the non-bridging oxygen bond, ie., Si-O-X + H2G = Si-O-H + X-OH During burial, the formation o f this layer may continue almost indefinitely so that only a sliver o f the original glass remains in the centre or no original glass survives. The formation o f a crust, which is little more than silica gel, makes the glass fragile and subject to physical disintegration. The tension at the interface between the unaltered and altered glass may cause spalling o f the surface, especially when subjected to the pressures o f salt crystallisation or repeated hydration-dehydration or freeze-thaw cycles. El-Shamy and Douglas (1972) have demonstrated that the kinetics are initially second order (linear to the square root o f time) because, in order to react with the pristine glass, the water must first pass through the altered layers. Where the alkali ions produced by the reaction remain within the system, the reaction becomes first order, (linear with time) as a sufficiently high pH (approximately 9) arises for the Si-O-Si bonds to be broken, so that silica passes into solution (El-Shamy, Lewins and Douglas 1972; Smets and Tholen 1985), ie., Si-O-Si + OH- — > Si-O- + Si-OH The enhanced porosity o f the network allows the removal of the larger alkaline earth metal ions. This w ill rarely occur in

N.J. Earl and D.E. Watkinson

20 Oxide S i02 a i 2o 3 Fe20 3 p 2o 5 CaO MgO Na20 k 2o PbO MnO CuO CoO

-

Glass 1 69.8 2.2 0.19

Glass 2 53.0 2.7 0.37

-

-

8.0 0.76 18.0 0.85

21.1 7.4 2.7 11.3 0.03 0.56 0.06 0.87

-

0.29 -

-

Glass 3 49.9 3.5 0.48 2.1 24.8 8.6 0.7 11.2 -

-

Figure 1. Ballidon Compositions by M olar Percentage

an archaeological environment because the diffusion of alkali ions away from the glass surface, through the soil, prevents a sufficiently high pH being reached. In closed storage environments, the alkali w ill remain in the immediate environment and loss of silica from the network can occur, thus causing deterioration beyond that which occurred in the burial environment. The greater the surface area of the glass, the greater the potential for alkali ions to be extracted into the surrounding water; the smaller the volume of this water, the more concentrated the resultant solution. A range of corrosion products derived from soil ions and gases can form, as the ions are able to react with the altered surface, even if not directly with the unaltered glass. Corrosion crusts may contain carbonates and phosphates (Cooper, Cox and Perutz 1993) and black layers are due to ions such as manganese and iron in the matrix (Alten 1988; Newton 1971; Cox and Khooli 1992).

-

-

Accelerated corrosion methods could only imitate, rather than reproduce, the corrosion occurring on burial. More than one type o f corrosion is found on archae­ ological glass and, at best, it would be possible to imitate only one of these. The relative efficiency o f the storage systems should be the same on pristine surfaces or those which have already undergone some deterioration. (The possible exceptions are any chemical systems, which have their basis in reacting with an already corroded layer).

Moreover, preliminary experiments showed that im ­ mersion in deionised water resulted in detectable corrosion, at least on the Mediaeval compositions, within a period o f weeks at ambient conditions.

Storage E x p e rim e n ts Im m ersion M ethods The majority of the methods in current use can be divided into two broad categories; the first o f which are based on immersion. Experiments comparable to variations on these system s were designed. In addition, the effect o f low temperatures both in water alone and in water/ethanol mixtures was examined. A decrease in corrosion rate would be expected with a drop in temperature, and ethanol at ambient temperatures has been recommended as a storage system (Newton and Davison 1989) because it replaces the water in waterlogged glass. Further, ethanol does not freeze in a domestic freezer which is important as waterlogged archaeological glass is fragile and may break under the pressures imposed by freezing. For the immersion system s the appropriate storage solution (fig. 2) was placed in polyethylene tubes with a glass sample in duplicate and the tubes were sealed.

Preparation of Glass Samples The glass used for experimentation was newly made and standardised as the variations inherent in archaeological glass would have nullified any conclusions. Differences in the results could thus be attributed directly to the properties of the storage systems, rather than differing glass composition, burial conditions or the presence o f microbiota. Three compositions of glass used in a long-term experiment to examine glass deterioration during burial at Ballidon in Derbyshire (Fletcher 1972) were selected. Glass 1 is a typical Roman composition, Glass 2 is a typical Mediaeval com­ position and Glass 3 is a composition similar to that of the “Linen-Smoother” found at Hangleton which formed 500 surface layers (Newton 1971). From the compositions it would be expected that Glass 1 should respond least to adverse storage conditions because of its high silica content, whereas Glass 2 and Glass 3 should both be affected, the latter demonstrating the greater response. The glasses were manufactured at the Department of Engineering Materials at the University of Sheffield. They were formed into blocks and lapped to a 600 SiC grit finish mechanically. It was decided not to subject the glass samples to artificial deterioration, prior to their use in the storage experiments for the following reasons:-

Additional errors would be introduced into the sample comparability.

H igh R H System s The second category o f methods is based on high Relative Humidity. Two variations were examined. The first (IX) was designed to be comparable to sealed bags containing small amounts of water in plastic boxes containing water. Samples were inserted into individual polyethylene (plastazote) cut-outs in unsealed polyethylene tubes and doused with deionised water. The tubes were floated upright in a polyethylene box containing water which was then sealed. The RH in the box was so high that droplets o f moisture condensed on the lid o f the box.

Experiment I II

III IV V VI VII VIII

Storage Solution Deionised Water Fitted into Cut-Out in Polyethylene Foam (Plastazote) Strip + Deionised Water Deionised Water 50/50 vol./vol. Deionised Water l Ethanol Ethanol Deionised Water 50/50 vol/vol Deionised Water / Ethanol Ethanol

Environment Ambient Ambient

Refrigerator Refrigerator Refrigerator Freezer Freezer Freezer

Figure 2. Immersion Systems

Assessment o f Post-Excavation Systems A second system commonly found is the placing o f glass on damp tissue paper or other absorbent substance. Samples were placed on damp acid free tissue paper so that one side was fully in contact with the paper in individual polyethylene tubes (X).

Chemical Environments There has been a significant body of work done on the behaviour of glass in various chemical environments, usually for the purposes o f industry (e.g. Jantzen and Plodinec 1984, Sales et al. 1984, Harvey, Litke and Larocque 1992) and biomedical research (e.g. Hench and Clark 1978). Chemicals can decelerate corrosion in a number of ways, including buffering so that a pH rise is prevented, slowing down the removal of ions from the matrix, and forming passivating layers. There has been little application to archaeological glass and there are clearly limitations on the use o f chemical on excavated material. For instance, change in appearance, especially where the surface is decorated is undesirable. Further, ions derived from impurities, incorporated into its surface during its lifetime or from the burial environment, may interfere with mechanisms effective on pristine glass. There has, however, been transfer o f industrial research in other archaeological materials and a range of chemical treatments, suggested by the literature, was accordingly examined (fig. 3). The appropriate storage solution (fig. 3) was placed in polyethylene tubes which were sealed.

IR Analysis Infrared Reflection Spectroscopy (IRRS), using Fourier Transform Infrared (FT-IR) with a microscope attachment, was chosen to monitor the experiments. Infrared spectro­

Experiment

Expected Effect

scopy can monitor changes occurring in the glass surface, for instance in the Si-O-Si and Si-O-X bonding, during corrosion. The reflectance function renders it non-destructive so that one o f each set of duplicate samples was periodically examined and returned to storage environments. The major advantage o f FT-IR over conventional infrared spectroscopy is that all wavelengths are examined simultaneously; the reduction in time enables the collection and averaging of multiple scans, improving the Signal:Noise ratio. It is also practical to examine more than one area. The use o f the microscope attachment enables areas as small as 10pm x 10pm to be pinpointed so that the corrosion in areas of, for instance, pitting or iridescence can be distinguished from that of the bulk matrix. FT-IR was done using a SpectraTech Research Plan Microscope attached to a Nicolet 510 Fourier Transform infrared spectrometer over the range 4 0 0 0 -7 5 0 c m 1. A working area of 100pm x 120pm was routinely selected and 250 scans were collected at a resolution o f 4 c m 1. Several spectra o f each characteristic area on a piece of glass were obtained to avoid errors due to focusing on a flaw or an unobserved corrosion feature. The significance o f the various peaks found in glass reflection spectra has been extensively examined in the literature (Sanders, Person and Hench 1972; Sanders, Person and Hench 1974; Clark, Ethridge, Dilmore and Hench 1977; Hench, Newton and Bernstein 1979). In pure vitreous silica the only bond present is the bridging oxygen Si-O -Si, producing one absorbance peak in the region detectable by FT-IR microscopy. This is the stretching vibration o f the SiO bond, located at 1100 c m 1. The addition o f alkali ions and the formation o f non-bridging oxygens produces a net decrease in local symmetry and the creation o f two different Si-O environments, splitting the peak into two.

Solution

Deionised Water Formn. of 0.5M A12(S04)3 aluminosili cate complex surface layer XII) Introduction of Ca2 (Oka Formn. of 2.5 M Ca(N03)2 andTomazawa 1980) (Adams 1984) Ca(OH)2 O.lMTris/HCl buffer XIII) pH7 Buffer pH buffering XIV) pH8 Buffer XV) pH9 Buffer (El-Shamy, Lewins and Douglas, 1972, Barkatt et al., 1993) 1M NaH2P04.2H20 , XVI) Introduction of PO^2 (Hench Formn. 4:1 mixture 1M KH2P04:1M NaH2P04 "dual and Clark 1978) 1M KH2P04 protective film layer", calcium phosphate over Si-rich layer. Approx. 1M sodium silicate XVII) Introduction of Si (Newton Common Commercially supplied potassium silicate. ion effect and Seddon 1992

Control

XI) Introduction of Al Clark 1978)

3+

21

(Hench and

Figure 3. Chemical solutions and their possible reaction with glass surfaces.

22

N.J. Earl and D.E. Watkinson

Figure 6. Glass 3 Original

Assessment of Post-Excavation Systems The Si-O-Si peak in an alkali environment occurs at around 1020-1050cm*1, lower than in vitreous silica due to the weakening in the interaction between structural units caused by the decrease in the number of Si-O-Si bonds. The other peak, typically between 920 cm*1 and 950-1, arises from the stretching vibration formed by the coupling assigned to the non-bridging silicon-oxygen vibration, (Si-O-X). It occurs at lower wavenumbers than the Si-O-Si peak, because it is weaker. Infrared reflectance is proportional to the concentrations of the vibrational species and so the location and intensity o f the Si-O -X peak are linearly related to the molar percentage of R20 and RO in the glass. Large amounts o f RO produced strong ionic bonds, giving a high intensity in the Si-O-X region, while the Si-O-Si peak becomes broader and less intense as the concentration o f alkali present increases. Also, for any given concentration o f Na20 the replacement of CaO for S i0 2 results in the shifting o f the SiO-Si peak to lower wavenumbers. This shows that the changes are a function of the amount o f Si present rather than just the increase in the alkali content. Scratches, pits and deposits reduce the intensity of both the Si-O-Si and the Si-O -X peaks which cannot therefore be used for the quantitative analysis of glass. They do not, however, cause a shift in location, and thus where a change in wavenumber occurs, it indicates that the composition has changed. Higher alkali concentrations occasion further changes in the peak positions as the composition o f alkali-metasilicate is approached. The Si-O-Si peak is at 1000 cm*1and the nonsymmetrical Si-O-X peak is at 850 cm*1. The presence o f two non-bridging oxygens in a tetrahedron causes coupling between the symmetrical and non-symmetrical vibrations. This can produce two small peaks on either side of 900 cm*1, partially filling in the valley between the Si-O-X and the SiO-Si peaks and creating a new shoulder beneath the Si-O-X peak at 820 cm*1. The nature of the changes resulting from corrosion is dependent on glass com position (Hench, Newton and Bernstein 1979). The initial deterioration process for less durable composition involves proton-alkali exchange at the

Expt.

glass surface, producing a decrease in the Si-O-X peak intensity and a shift from around 920 to 850 cm*1. A satellite peak apears at 960 cm*1 as a shoulder on the Si-O-Si peak, attributed to the non-bridging oxygen stretch in an alkali environment, but the intensity and location o f the Si-O-Si peak remain constant. As alkali depletion increases, the SiO-X peak m oves to even low er wavenumbers before complete elimination, while the Si-O-Si peak broadens and shifts to a higher wavenumber. Although the increase in wavenumber is continuous until it reaches a maximum, the amplitude o f the peak may eventually decrease slightly, probably attributable to loss o f silica from the matrix because o f increased porosity. More durable compositions behave rather differently. There is no minimum between the Si-O-Si and Si-O-X peaks in the pristine glass due to stronger interactions between the non-bridging modifier bonds and the silicon-oxygen bridging network bonds. Upon corrosion, the changes are substantially less pronounced. A small dip appears at 870 cm*1 and both the Si-O-Si and the Si-O-X peaks decrease in intensity, showing that no silica rich layer has been formed. The spectra produced from the pristine glasses used in this research are shown below (figs 4, 5, 6). The durability o f Glass 1 can be seen in the absence o f a minimum between the Si-O-Si and Si-O-X peaks. The presence o f distinct SiO-Si and Si-O-X peaks in Glasses 2 and 3 shows that they are intrinsically less durable.

Results At the conclusion o f the experimental period (15 months for physical variables and 9 months for chemical variables), there was little visible change to Glass 1 except for some slight crystalline deposit. In contrast, Glasses 2 and 3, particularly the latter, exhibited a number o f phenomena such as iridescence, a variety o f different deposits, surface loss and the formation of silvery crazed layers. It was found that the position, intensity and number o f peaks in the FT-IR spectra varied with the appearance o f the glass surface (figs 7, 8, 9, 10, 11, 12).

Notes

Wave-No.

Notes

Wave-No.

Notes

Between 877 and 882

VAV.

955-962

Orig. glass I II

1051.7

III 6 ninth only IV V VI VII

Between 1051 and 1056 Not examined Not examined Between 1046 and 1055 Between 1040 and 1059

v.s., b.

Between 981 and 998 Between 955 and 998

VIII

Between 1044 and 1053

v.s., b.

Between 965 and 993

IX

Between 1047 and 1070, usually 1059

v.s., b.

X

Between 1055 and 1074

b., v.s.

938-955, sometimes resolves to 2 peaks at 943 and 931 913-943

1059-1071 1055-1056

Additional Bands

Si-O-X

Si-O-Si Wave-No.

23

v.s, v.b. v.s., v.b. Same position as in orig. glass

932-955 932-962

v.s., v.b. v.s., v.b.

Between 959 and 967

v.s., b.

2° peak oil* Si-O-Si

v.s., b. resolves in some cases to 2 noisy peaks s.b.

b., v.s.

Key to Diagrams: a = periodically examined sample; b = undisturbed sample; T = Top surface; B = Bottom surface

Figure 7. Glass 1 FT-IR Data Physical Environments, Immersion and High R.H.

24

N.J. Earl and D.E. Watkinson

Expt.

Notes

Orig. glass

Si-O-Si Wave-No 1021-1033

I. a and b

1017-1028

v.s, v.b.

H. Ta

1017-1024

n.Tb

1012-1021

IIBa

Shiny, plated area

Between 1071 and 1083

SI. sharp’ed. Similar position to in orig glass Intensity varies

Si-O-X Position 924 (High Pt.)-939 Between 858 and 901 Between 870 and 901 or between 847 and 877 Between 847 and 901 Between 854 and 873

Wave-No.

w., b. sh. or absent

Additional Bands Position

i) Between 1248 and 1272 ii) Between 1395 and 1491

II Bb

Visibly unaltered area Patchy wliite areas -sparse

Patchy white areas- dense

III a. and b. IV a. and b. V a and b. VI a and b.

vn

Both sides of a and one side of b One side of b

VIII a and b IX

X Ta

Between 1009 and 1024 1066-1078

v.s., b.

Between 875 and 885

v.s.,b.

As orig.

Usually between 1021 and 1043 but as low as 1005-1017 As for sparse coverage

v.s., b.

Between 901and 920

Between 1066 and 1071 N/A N/A 1024-1028 Between 1007 and 1028 Between 992 and 1011 Between 1017 and 1036 Between 1017 and 1040- high pt. usually at 1028

Clear blue/ purple areas feature) and shiny areas

1066-1090

Tb

Clear blue areas and white spots

Between 1017 and 1080

Bb

(white patches, clear blue areas and small thick areas of white deposit)

Bet. 1044 and 1087

w., b. w.b.

2 variations in position

w,, b. sh.

v.s., slip.

i) Between 1472 and 1474 ii) 944 iii) Between 986 and 1012 iv) 1248-1260

Patchy white areas

Notes

Shifted to 854 and sharp’ed when density of deposit increased v.s.,b.

i) ii)

v.s.,b.

1200 region

v.s, shp sh/peak often present Sometimes present

Sometimes present 1250 andB. bands 1400-1470

B. band

854-855 1200 region 854-855 1085-1087

Absent

v.s.,b

SI. less intense relative to Si-O-X.

v.s.,v.b. v.s., b.

SI. more intense relative to Si-O-X in shiny areas than clear ones Broader

b.

Between 85 land 889

Usually between 908 and 920 but as high as 932 and as low as 896 847-877

Between 875 and 882, and between 854 and 858 Between 871 and 881, and bet. 851 and 858

v.s., slip. v.w., off Si-O-Si peak

Decrease in intensity relative to Si-O-Si

932 Between 913 and 933 Between 907 and 938 Between 930 and 945

v.w, shp. off Si-O-X B. band

Between 879 and 885

w., shp. 2 peak off Si-O-Si

Between 879 and 885

w., shp. 2 peak off Si-O-Si

v,s.,b.

854-858

w. Sharp’ed and stronger with presence of crystals

Broader and less well defined, in some cases has 2 constituents (between 854 and 858 and between 875 and 878)

1200 region

w.b. sh

Split into 2 slip peaks,

1456-1474

b.

Split into 2 peaks

1417-1542

B. band

v.s.,v.b.

Figure 8. Glass 2 FT-IR Data Physical Environments

1400 region v.w., v.b. feature

Assessment o f Post-Excavation Systems

Expt.

Si-O-Si Wave-Nos. 1010-1040 (high pt- 1024-1028) 1028-1040

Orig. glass

Notes v.s.,b

Si-O-X Wave-Nos. 920-939

Sharper.

866-877

Notes v.s.,b Intensity reduced relative to Si-O-Si

I

25

Additional Bands Wave-Nos.

1200 region

Notes

w.,b. Not directly assignable

1400 region

Between 1021 and 1044

II Ta

H.Tb

Irid. and non-iiid. areas

Between 1028 and 1040

water-mark

Between 1033 and 1059 Between 1037 and 1059

silver crazed layer HBa

Yellow irid. Blue irid.

HBb.

Cloudy area Visibly unalter-ed glass Silver crazed layer

Between 1012 and 1047 (high pt. 1028) Between 1040 and 1055 Between 1059 and 1071 1028-1043 1028-1043

v.s., b.

v.s, b.

as above as above

v.s., b. v.s., b.

i) 889

to 2 noisy BetweenResolves 866peaks w. and b

ii) Between 828 and 858 i)Between 863 and 877 ii) Between 827 and 854 as above i)Between 932 and 948 ii) Between 875 and 889 iii)Absent Between 870 and 882

v.s. SI. sharpd.

Between 858 and 877 Between 839 and 889 Between 863 and 872

v.s. SI. sharpd.

i) Between 873 and 894

v.s., b.

1230-1260 region

w.,b. Not directly assignable v.w.

i) w., b. ii) w.sh. Not always present as above as above

1200 region

sh.

1400 region

Noisy band

as above v.w., b. v.w., b. v.w., b. Significant decrease in intensity relative to Si-O-Si v.w.,b. resolves to 2 peaks

ii) Between 835 and 847 Irid. Ill a and b IV a andb

V (6 mths.only) VI a and b VII a and b

vm

a b

1028-1043 Between 1066 and 1071 Usually between 1033 and 1059, though extends as low as 1017 when peak is v.b. 1028-1033 Between 1005 and 1028 Between 1012 and 1043 Between 1005 and 1031 as above

v.s. SI. sharpd. s.,b. v.s.,b.

v.s.,b. v.s., b. v.s,v.b.

v.s.,b. v.s.,b.

Between 858 and 877 Between 851 and 889 Between 901 and 932

941-947 Between 901 and 951 Between 896 and 924 excptnlly extends as low as 877 Between 932 and 951 Between 892 and 951

w., b., shallow. v.s.,b.

v.s.,b.

v.s.,v.b.

v.s.,.v.b. v.s.,v.b.

Figure 9. Glass 3 FT-IR Data Physical Environments

Continued overleaf

N J. Earl and D.E. Watkinson

26

Expt.

Si-O-Si

Orig. glass DC

DCcont.

Feature­ less areas TandB, a andb b) rough vertical stripe with shiny deposit Ta rough, si. shiny area Tb water-mark without shiny deposit Tb misted area Ta

irid. areas and.watermark + sliiny spots Tb irid. areas Bb

X

Notes v.s.,b

Wave-Nos. 920-939

Notes v.s.,b

Wave-Nos.

Notes

v.s., v.b.

Between 901 and 934

v.s., v.b.

Between 877 and 882

w. “blip”, not always present

1009-1047

v.s., v.b. similar to orig.

Between 870 and 882

v.s., b., sharpened to a point

1047-1059

v.s., b.

Between 870 and 882

Sharpened to a point

1047-1059

v.s., b.

Between 901 and 913

v.s., b.

Between 875 and 889

weak 2° peak off Si-O-X

Usually between 1017 and 1036, though exception-ally as low as 1009 Bet. 1024 and 1044

v.s.,b.

Between 858 and 870

Shift from orig and diminished relative to Si-O-Si

approx. 828

Poorly defined and meiges with cut-off

SI. sharpd

Between 870 and 889

Diminished relative to Si-O-Si almost disappears in brighter irid. areas and watennark shp Intensity variessometimes comparable to Si-O-Si, sometimes virtually absent Diminished to sh.

1200 region

w.b. peak, appears siliceous

1400-1600

b., noisy bands.

Between 1018 and 1050

white layer Ba

Between 1025 and 1043

Glass beneath white layer Ba

Between 1040 and 1064

Ta andTb

Additional Bands

Si-O-X

Wave-Nos. 1010-1040 (high pt- 1024-1028) Between 1017 and 1047

Between 870 and 875

Similar to orig., si. sharpd Sharpd

Between 859 and 900 i) Between 910 and 943 ii) Between 858 and 877

1059-1074

Between 855 and 901 but usually 875-881

i) Of similar intensity to Si-O-Si peak ii) Diminished in intensity to pt. of disappearing b. and rounded. Comes to shp. peak Intensity of both bands appear to vary proportionally with sharpness of Si-O-X, so prob. assoc. + OH bonds. B. bands prob. attrib. to -OH bonds

Ba

Between 1028 and 1064

s.,b.

Between 870 and 873

Bb non irid

Between 1028 and 1043

s.b.

Between 870 and 873

Bb irid

Between 1016 and 1037

s.,b.

Between 870 and 872

Glass beneath white deposit

Between 1047 1064

v.s.,b.

Between 874 and 882

Significant diminution relative to Si-O-Si or s,v,shp Significant dimunition relative to Si-O-Si v.s.,v.shp.

Various

Diminished relative to Si-O-Si

Figure 9. Glass 3 FT-IR Data Physical Environments

Cannot be assigned because of Restralilen Banding.

Assessment of Post-Excavation Systems

27

Expt.

Si-O-Si

Pristine Glass

Wave-No. 1051.7

Notes v.s, v.b

Wave-No. 955-962

Notes v.s., v.b.

HO 2

Between 1055 and 1066

b., v.s.

Varies between 939 and

b.,v.s.

Si-O-X

962 XI

Between 1065 and 1069

b., s.

XH

1055-1062

b., v.s.

xni

1055-1066 Betweenl055 and 1064 Betweenl053 and 1062

b., v.s. b., v.s. b., v.s.

XIV XV

Varies between 928 and 943 Varies between 936 and 962. 931-948 Between 939 and 955 i) 939-951 ii) between 936 and 941 and between 945 and 949

XVI XVH

iii) between 924 and 927 and between 939 and 940 943-955 Varies between 943 and 986 (most peaks cover between 20 and 30 wavenos.)

b.,v.s. b., v.s.

1055-1066 Varies between 1028 and 1052 (high pt. usually 1047)

b., s. b., v.s. b.,v.s. b., v.s b., v.s., may resolve to 2 noisy peaks

b.,v.s. b.

Figure 10, Glass 1 Chemical Variables- FT-IR Data Expt. Orig. glass

Wave-No. 1021-1033

Notes v.s., v.b.

Wave-No 924 (High pt.)-934

Notes v.s., v.b.

Wave-No

Notes

H,°

Bet. 1028 and 1033

b., s.

Bet. 861 and 863

w., sometimes

I) Bet. 1226

I) sh

absent

and 1242

ii) Bet. 1450

XI

1084-1090, sh 1213-14

Sharper

920-935

XH

iandii) 1082-1085, sh. 1210-1215

Sharper, resembles vitreous silica peak

i)Varies between 896 and 901 (Non irid.)

xni

i,ii,andiii) between 1078 and 1090, sh 1230

Sharper

ii) i) (non-irid)

w,b. Intensity reduced relative to Si-O-Si Intensity relative to Si-O-Si, esp. for irid (ii)

and 1652

broadness varies

ii) 800

2 peak off Si-O-X (irid)

. 877 (irid.) between 920 and v.w. 951sh

ii) 900-931 (irid.) XIV

XV

Between 1082 and 1090

i) between 1047 and 1064 ii) (irid) 1062-1064

Sharper

b., v.s.

iii) Absent i) Between 901 and 932 ii) (irid.) Between 889 and 913. i) Between 842 and 858 and between 870 and 877 ii) (irid) Between 847 and 858 and between 766 and 785

XVI

1094-1097, s.sh 1233-1236

Sharper, similar to vitreous silica

b.w., almost sh. in some places

Between 1213 and 1241

i) w. andv.w (2nd peak not always present)

Between 1206 and 1229

s.sh.

939-955

w.b. Possibly remnant of Si-O-X

ii) 2nd peak poss. assoc.+ ring structure

Absent

"blip" 1000 XVH

1017-1040

920-941

Figure 11. Glass 2 Chemical Variables- FT-IR Data

NJ. Earl and D.E. Watkinson

28

Expt.

Si-O-Si Wave-Nos.

Notes

Si-O-X Wave-Nos.

Notes

1010-1040 (High pt. 10241028) Between 1029 and 1044

v.s, v.b.

920-939

v.s., b.

v.s, v.b.

870-871

w„b.

XI Clear surface

1071-1078

Sharper

850-900

Intensity reduced relative to Si-O-Si

XI(Sunken area in glass middle)

Between 1059 and 1083 (generally 1072-1078)

Between 847 and 894

X ll

Between 896 and 901

XIII

1071-1072 + triplet 10391040, 1025-1027, 10121015. 1046-1053

SI. sharpened

XIV XV

Between 1040 and 1053 between 1028 and 1044

Sharper b., v.s.

XVI

i) and ii) 1097-1102, sh. between 1225 and 1239.

i) and ii) Sharper, similar to vitreous silica.

XVII

986-1028

Orig Glass HO

iii) 1058 b., unusual shift to lower wave-nos

1132-1148

Varies between 858 and 889

Intensity reduced relative to Si-O-Si

Between 835 and 873 (high pt. usually 858)

Intensity reduced relative to Si-O-Si, sometimes only sh. i. ii) w, b, probably assoc. + Si-O-X

1). 932-951 ii) . 932-951 and varying between 835 and 877. iii) 920-940 896-939

Additional Bands Notes WaveNos.

Similar shape and intensity to Si-O-Si

b

Figure 12. Glass 3 Chemical Variables- FT-1R Data

Discussion As expected, the preeminent factor determining how the glass is affected by differing environments appears to be its initial composition. Thus, the spectra from Glass 1, predicted as stable, varied little. In contrast, Glasses 2 and 3, which were less stable showed marked differences in their corrosion behaviour dependent on the environment. In immersion system s at ambient temperatures, similar spectra were obtained right across the surfaces not in contact with any other material, suggesting even corrosion. The spectral changes were those associated with initial corrosion i.e. diminution o f the Si-O-X peak relative to the Si-O-Si peak and a shift to lower wavenumbers (fig. 13). In some cases the presence o f a peak in the 820 region suggested that the glass surface was approaching a metasilicate composition. There was little change in the position of the Si-O-Si peak. Even where the glass surface showed no visual phenomena associated with corrosion, such as iridescence, the changes in the spectra indicated that some deterioration had occurred. The spectra from the samples immersed at lower tem­ peratures indicated less corrosion. While significant spectral changes were, however, still arising from the surface o f the samples which had been stored in a refrigerator, the spectra from those stored in the freezer were very similar to those o f the pristine glass (e.g. fig. 14). This applied equally to those in water, ethanol and ethanol/water mixtures, even though samples stored in water alone were broken. The spectra from the surfaces of the samples subjected to high RH and those in contact with the polyethylene foam in immersion systems varied across the glass surfaces, sug­ gesting localized corrosion. In some areas, especially those characterized by bright iridescence or the formation of crazed silver layers, this appeared to be quite severe as the Si-O-Si bond had moved to higher wavenumbers, close to the region associated with vitreous silica. The Si-O-X peak was, however, still usually significant (fig. 15). It has been suggested (Isard and Patel 1981) that this is because the

nature o f the deterioration processes is somewhat different from those involved in immersion. The spectral data from the samples stored in the different chemicals was, on the whole, in good agreement with the visual phenomena observed. For instance, the spectra from the samples stored in A12(S 0 4)3 solution (fig. 16) indicate severe corrosion, including some dissolution o f the silica matrix. In fact there was a hole in the middle o f Glass 3 and Glass 2 dissolved com pletely, aside from som e small fragments. Also o f interest was the inhomogeneity in the spectra across a surface in contrast to immersion in deionised water alone. This suggested that even slight inhomogeneities in the glass surface would corrode differentially to the bulk matrix o f the glass. The only chemical environment which appeared to be a marked improvement on storage in deionised water alone was silicate solution where the spectra from both Glass 2 and Glass 3 were similar to those from the pristine glass (fig. 17).

Complementary Techniques Analysis o f leach solutions for silica and Scanning Electron Microscopy (SEM) with EDXA were used in conjunction with FT-IR microscopy. Generally, the results concurred. An interesting and important exception applied to the samples stored in silicate solutions. Both FT-IR microscopy and the EDXA suggested that glass surface had changed little from the pristine. SEM, however showed that all three compositions had been severely etched. This has significant implications for decorated glass surfaces.

Conclusions It has been demonstrated that FT-IR microscopy is suitable for detecting corrosion occurring on glasses of archaeological composition without accelerated conditions and can be used to distinguish between the effects o f different environments.

Assessment of Post-Excavation Systems

Figure 13. Glass 3 in deionised water, (shows separation o f peaks and diminution o f S i-0-X peak, indicating leaching)

29

Figure 14. Glass 3 in deionised water, freezer (shows little change from original)

Figure 15. Glass 2 resting on damp tissue paper (shows movement o f Si-O-Si to higher wavenumbers, indicating deterioration)

Figure 16. Glass 2 in aluminim sulphate solution (shows movement o f Si-O-Si, indicating deterioration)

Figure 17. Glass 2 in silicate solution (shows little change from original)

30

N.J. Earl and D.E. Watkinson

Spectral data can be used to elucidate the nature o f the corrosion that is occurring both on the bulk matrix o f the glass and in areas where particular corrosion phenomena are exhibited. While it is advisable to use another technique such as SEM in conjunction in case there are corrosion phenomena with no associated compositional change, FTIR microscopy is a useful and enlightening technique.

References Adams, P.B. 1984. Glass corrosion. A record of the past? A predictor of the future? Journal o f Non-Crystalline Solids 67: 193-205. Alten, H.I. 1988. Changes in waterlogged medieval window glass in Material Issues in Art and Archaeology (eds. Sayre, E.V., Vandiver, P.B., Druzik, J and Stevenson, C.), Materials Research Society Symposium Proceedings 123: 279-284. Barkatt, A., Sang, J.C., Jakubik, R.F. and Saad, E.E. 1993. Oscillations in the dissolution kinetics of silicate glass in trisbuffered aqueous media. Journal o f Non-Crystalline Solids 155 (2): 141-148. Clark, D.E., Ethridge, E.C., Dilmore, M.F, and Hench, L.L. 1977. Quantitative analysis of corroded glass using infrared shifts. Glass Technology 18 (4): 121-124. Cooper, G.I., Cox, G.A., and Perutz, R.W. 1993. Infrared microspectroscopy as a complementary technique to ElectronProbe Microanalysis for the investigation of natural corrosion on potash glasses. Journal of Microscopy 170: 111-118. Cox, G.A., and Khooli, A.R. 1992. The natural corrosion of glass: the formation and structure of plugs. Glass Technology 33 (2): 60-62. El-Shamy, T.M., and Douglas, R.W. 1972. Kinetics of the reaction of water with glass. Glass Technology 13 (3): 77-87. El-Shamy, T.M., Lewins, J., Douglas, R.W. 1972. Dependence on pH of the degradation of glass by aqueous solution Glass Technology 13: 81-87. El-Shamy, T.M. 1973. The chemical durability of K20-CaO-MgOS i02 Glasses Physics and Chemistry o f Glasses 14: 1-5. Ernsberger, F.M. 1980. The role of molecular water in the diffusive transport of protons in glasses. Physics and Chemistry o f Glasses 21 (4): 147-149. Fletcher, W.W. 1972. The chemical durability of glass. A burial experiment at Ballidon in Derbyshire. Journal o f Glass Studies 14: 149-51.

Harvey, K.B., Litke, C.D., and Larocque, C.A.B. 1992. The dissolution of a simple glass. Part 4. Behaviour in saturating and saturated systems. Physics and Chemistry o f Glasses 33 (2): 43-50. Hench, L.L., and Clark, D.E. 1978. Physical chemistry of glass surfaces. Journal o f Non-Crystalline Solids 28: 83-105. Hench, L.L., Newton, R.G., and Bernstein, S. 1979. Use of infrared reflection spectroscopy in analysis of durability of medieval glasses, with some comments on conservation procedures. Glass Technology 20 (4): 145-148. Isard, J.O. and Patel, R.W. 1981. A comparison between weathering and water leaching tests on glasses of simple composition. Glass Technology 22 (6): 247-250. Jantzen, C.M., and Plodinec, M.J. 1984. Thermodynamic model of natural, medieval and nuclear waste glass durability. Journal o f Non-Crystalline Solids 67: 207-223. Newton, R.G. 1971. The enigma of the layered crusts on some weathered glasses, a chronological account of the investigations. Archaeometry 13 (1): 1-9. Newton, R.G., and Davison, S. 1989. Conservation o f Glass, London, Butterworths. Newton, R.G., and Seddon, A.B. 1992. The durability of silicate glass in the presence of a saturated leachant Corrosion Science 33 (4): 617-626. Oka, Y. and Tomozawa, M. 1980. Effect of alkaline earth ions as an inhibitor to alkaline attack on silica glass. Journal o f NonCrystalline Solids 42: 535-543. Paul, A. 1990. The Chemistry o f Glass 2nd Edition London, Chapman and Hall. Sales, B.C., White, C.W., Begun, G.M., and Boatner, L.A. 1984. Surface layer formation on corroded nuclear waste glasses. Journal o f Non-Crystalline Solids 67: 245-264. Sanders, D.M., Person, W.B., and Hench, L.L. 1972. New methods for studying glass corrosion kinetics. Applied Spectroscopy 26 (5): 530-537. Sanders, D.M., Person, W.B., and Hench, L.L. 1974. Quantitative analysis of glass structure with the use of infrared reflection spectra. Applied Spectroscopy 28 (3): 247-255. Smets, B.M.J., and Thoen, M.G.W. 1985. The pH dependence of the aqueous corrosion of glass. Physics and Chemistry o f Glasses 26 (3): 61-63. Turner, W.E.S. 1956. Studies in ancient glasses and glassmaking processes. Part 4. The chemical composition of ancient glasses. Journal o f the Society of Glass Technology 40: 162-186.

Medieval Glass-making Technology: the Corrosive Nature of Glass Ian M erchant,1 Julian Henderson,2 David C rossley3 and M ichael Cable4 1Archaeology and Pre-history Department, University o f Sheffield; 2Archaeology Department, University o f Nottingham; 3 Adult Continuing Education Department, University o f Sheffield; 4 Engineering Materials Department, University o f Sheffield

Abstract The technology of Medieval glass production has previously been investigated by analysing the glass, or the ceramic crucibles, separately. The approach in this study was different, with the analysis o f interactions between clay crucible fragments and the glass adhering to them. This has made it possible to understand the corrosive and erosive effect the glass has had on the clay, and emphasises the material property requirements o f the crucible. In this study an interfacial zone, between the glass and clay, was observed by scanning electron microscopy. The Medieval site chosen for this project was Blunden’s Wood, in The Weald, Surrey. The date o f production has been established to be 1330 AD by magnetic dating, and, from analysing the glass, three different compositions of glass were defined, based on levels o f magnesia, alumina and potash. Descriptions o f the glass, clay and technology of Medieval glass production will be used primarily as an introduction to the complex nature of the reactions o f materials at elevated temperatures. Some of the crucibles appeared to be pitted, while others were not. An investigation was undertaken to ascertain the reasons for these differences, with conclusions leading to a better comprehension o f the nature of the attack of the glass on the crucible. On the basis of this, a comparison with twentieth century products was made, in an attempt to suggest the reasons why Medieval and modern crucible fabrics may be resistant to the attack of glasses. This study o f the corrosive effect of glass on clay may aid the reconstruction o f part o f the process o f glass technology in Medieval England, and as a result, contribute to our understanding o f glass-making sites and the context in which they lie.

Introduction Studies o f the English Medieval glass industry show records of over forty glasshouses in the Weald of Surrey and Sussex (Winbolt 1933; Kenyon 1967) between 1300 to 1620 AD. After this date the Wealden sites ceased to exist, as coal became the only source o f fuel for the furnaces. This change in fuel source was due to a ban on the use of wood in Wealden glass furnaces, as dictated by the terms of Sir Robert Mansell’s monopoly. Frequently, the technology o f glass production is investigated by analysing either glass or ceramic materials separately. The aim of this work is to study any interaction between the glass and the ceramic medium in order to create an understanding of the property requirements of crucibles, especially where liable to corrosion and erosion by molten Medieval glass. The other interface of interest is that between the glass and the environment, though many studies o f the weathering of glass have already been made (Hench and Clark 1978; Newton and Davison 1989). One advantage o f choosing the interfacial zone between the crucible and glass for analysis is that, theoretically, this zone becomes a fresh surface when the crucible is sectioned. It may not have undergone much weathering, even if it has undergone chemical alteration, and can provide a very good ‘snapshot’ o f glass-m aking technology, relatively un­ disturbed by time. Analysis of the interface may show the

penetrative and erosive effects o f glass while held at high temperatures.

Medieval Glass and the Weald Analysis o f small samples o f glass reveal many facts about Medieval glass technology. The formation o f glass from raw materials results in specific physical properties, and these characteristics can form the basis of a technical interpretation within an archaeological context (Henderson 1989). In general, the composition may be able to confirm the type o f glass, Pre-Medieval, Medieval or Post-M edieval. Glass technology during the English M edieval period can be characterised by the dominance of potash as the source o f alkali. Prior to this, English glass was based on soda-limesilica compositions o f a so-called ‘Roman’ type. The glass composition also enables glass properties to be established. For example, if viscosity is measured, the melting conditions, the working and annealing temperatures, refining behaviour and devitrification rate can be determined (Brill 1988; Doremus 1973). The glass under investigation originates from the Wealden glasshouse at Blunden’s Wood and dates from the second quarter of the fourteenth century. The reason for choosing this site is twofold. Firstly, there is an abundance o f both glass and crucible pieces available, and secondly, the

32

Ian Merchant, Julian Henderson, David Crossley and Michael Cable

production o f glass on the site has been dated magnetically to 1330. The pottery excavated on the site, Cheam ware and dark green glazed ware, basically suggests a date of 1350 AD, corroborating the view that Blunden’s Wood glasshouse was in operation during the second quarter o f the fourteenth century.

C eram ic M aterial and W eald en C lay Properties which improve chemical resistance of ceramics rely on a high bulk density and low apparent porosity (Whittaker 1993). The corrosion resistance o f refractory ceramics used for glass containment depends also on the temperature and composition o f the attacking medium. Today, to provide the ideal properties, ceramic materials are fusion cast, slip cast or isostatically pressed, and may contain form ulations o f zirconium oxide, alumina, silica and chromium oxide (Busby and Shaw 1989). For areas where bricks are used (e.g. regenerative systems) basic refractories such as magnesia, magnesia-chromic oxide and magnesiaalumina, may be chosen. These have good corrosion resistance, especially to alkali attack, and high thermal efficiency (Whittaker and Lacey 1990). Medieval glassmakers, however, would have had to rely on a readily available supply of ceramic material, i.e. clay. Clay sources would have been critical in the location of the glasshouse, which is presumably one reason why the English Medieval glass industry in the Weald is found on a bed o f clay (Thurrell etal. 1968). Wealden clay is characterised as a brick-making clay, not requiring the same tolerance as refractory clays, but in this context it must have been able to withstand the rigours of glass-m aking. The com position is based on kaolinite minerals, with additions o f chlorite, illite, quartz and organic material. The physical and chemical properties are difficult to quantify, but the clays generally contain 68% silica, 17% alumina, 6% iron oxide and other trace components and are a dark red-brown colour at 1200°C (Worrall 1975). Examination o f ceramic fragments permits the iden­ tification o f the material used in Medieval furnaces and crucibles. A nalysis o f the ceramic pieces, such as the appearance of the grains, may suggest the quality of clay material, or even the process temperature. A structure with good bonding is required, at the outer face as well as the front, working face, and this may suggest the need for pre-heating of crucibles prior to use. The small furnace, kiln C, at Blunden’s Wood may have been used for this purpose (Wood 1965). Chemical analysis of a number of crucible fragments from the site allows determination of the composition, and to some degree, the phases of the clay, as used for the production o f M edieval ceram ic m elting pots. A kiln temperature can be estimated, and a source o f clay may be located. The origin of the clay from Blunden’s Wood is not clear, although it has been suggested that the clay was not local to the site (Wood 1965). However, the clay used in the crucibles was able to withstand the temperatures reached in the glass-making furnace, regardless of the character of Wealden clay.

S cien tific A n alysis o f the G lass

L Technique o f Analysis Over 50 specimens, representing the range o f glass excavated from the site, were mounted in epoxy resin and polished to a surface depth of 1 micron, before coating with a thin layer o f carbon. Glass and ceramic bodies were analysed using a ‘Camscan’ scanning electron microscope (SEM), with an attached energy-dispersive spectrometer (E D S). EDS analysis can produce results which answer most tech­ nological questions, but certain problems associated with overlapping peaks and background scattering must be taken into account (Verita et a l 1994). For each analysis an accelerating voltage o f 20 kV and working distance o f 30 mm between the specimen surface and microscope detector was used. Cobalt was used to calibrate the microscope, and the results were normalised to 100%.

ii. Glass Compositions Three different glass compositions can be identified, (fig. 1). These compositional categories may be attributed to the intentional supply o f certain glass com p ositions, the experim entation o f new batches, or erroneous, failed syntheses remaining on site prior to being used as cullet. Results were checked against Corning B standard glass. Major components were found to fall within 5% o f the quoted values, while minor components were within 10%. These results compare favourably with those previously quoted for Blunden’s Wood (Vose 1980; Mortimer 1993).

Hi. The Interface Between Glass and Clay Examination o f the interface between the glass and clay may help to contribute to the picture o f Medieval glass production as far as glass technologists and archaeologists are concerned. This interior surface is o f particular interest, as weathering may often be limited to the outer surface. Compositional analysis o f the interface may provide valuable information about diffusion patterns and inter­ actions between the crucible and the glass, and those components migrating from the clay into the glass melt. Often, new materials awaiting use in modem furnaces have research carried out on the interaction between the ceramic and glass bodies (Duvierre et al. 1993), but this is not the case with ancient fragments. This study will redress this,

Type A

Type B

n=3G

n=8

Type C rt= 1 2

O x id e

W t%

S t .D e v

W t%

S iD e v

W t%

S iD e v

M gO

4 .6 7

0 .3 9

4 .7 4

0 .3 8

2 .3 4

0 .3 3

A I2 0 3

0 .3 4

0 .1 7

0 .5 3

0 .2 5

4 .3 3

0 .8 8

S i0 2

5 8 .6 3

1.7 6

4 5 .8 6

1 .3 3

6 0 .5 9

1 .9 4

P205

3 .3 5

0 .2 6

6 .2 9

0 .3

1 .5 4

0 .8 9

K20

1 2 .1 5

1.01

2 3 .9 6

1.1 6

17.51

1.51

CaO

16.51

1 .1 8

1 4 .7 9

0 .7 6

8 .3 4

0 .8 2

M n203

1.71

0 .1 4

2 .0 5

0 .3 3

0 .7 4

0 .0 9

Fe203

1.1 2

0 .1 7

1.1 7

0 .1 8

2.3 1

0 .8 2

Na20

0 .7 5

0 .2 6

1 .3 9

0 .0 2

0 .6 8

0 .1 4

T i0 2

0.21

0 .0 7

0 .2 9

0 .0 3

0 .4 2

0 .0 9

Figure I. Chemical composition o f glasses from Blunden's Wood

Medieval Glass-making Technology

33

Interface

Figure 2. A diagram o f the interface between glass and clay and trends will be discussed. A diagram of the interfacial zone can be seen in figure 2. The depth of glass penetration in the clay may demonstrate the type of attack. A narrow interfacial zone may indicate formation of a dense front face, causing erosion as the layer breaks away into the molten glass. A thick interfacial zone may suggest a molten glass passing through the porous body of the clay, becoming cooler as it does so, and solidifying at a certain point. A temperature gradient can be experimentally constructed, and therefore, an estim ation o f furnace temperature can be made. Modern refractory bodies are designed to withstand corrosion and erosion by any molten, or semi-molten material with which they come into contact. Thus, the ability o f the ceram ic material to resist the attack may govern the econom ics o f the process, and the behaviour o f the refractories is vital in furnace design, construction and operation (Cable 1968). Medieval glass-makers may have located their glasshouse near a source of the clay used in furnace structures and crucibles, though this was not always necessarily true. The glass/clay interface informs archaeologists about many aspects of glass technology, all within a cross-section of approximately 1000 microns, or 1 millimetre. This small size o f the material required makes for an invaluable form of analysis.

Plate 1. A back-scattered electron image o f the surface o f a crucible

Plate 2. A back-scattered electron image o f the interface iv. Electron M icrographs Secondary electron im ages provide little detail o f the interface. However, back-scattered electron images (BSE) allow the atomic contrast between the phases to become clearer. The interfacial zone was measured; the thickness being between 130 and 200 microns. Plate 1 shows the BSE for an area containing a small pit of glass where the possibility of erosion, or corrosion, may be. The homogeneity o f the glass is in direct contrast to the crystalline structure o f the clay. Linking the two materials is a dense layer comprising components of clay and glass. At higher magnification, as shown in plate 2, a line o f bright crystals can be seen, running parallel with the surface o f the glass, at a mid-point o f the interfacial layer. At very high m agnifications, the crystals are found to be rounded formations sintering together. The smallest growths are about 1 micron in diameter, rising to amalgamated forms in the region of 10 microns. The crucible material consists o f silica grains sitting in a matrix of glassy clay. Within the matrix small lath-like precipitates can be seen, high in silica and alumina, as typified by the structure of mullite.

v. Change in Chemical Composition Across the Inter­ fa ce Changes in chemical composition at regular intervals across the interface between the glass and clay were determined by EDS analysis using a scanning electron microscope. The results can be seen in figures 3 to 6. The line of bright crystals, as seen in Plate 2, is taken to be the mid-point o f the interfacial layer, the ‘O’ position, positive values are towards the glass/air surface, and negative values are towards the back face of the crucible. The silica level (fig. 3) remains relatively constant throughout the glass. It begins to rise within the interfacial zone, then levels o ff slightly, before rising to a steady value within the clay crucible. The level o f potash (fig. 3) remains constant till the surface o f the crucible is reached, before falling to 2-3% in the clay. The mobile ions o f potassium penetrate nearly 1000 microns, or 1 millimetre, into the clay. The easy diffusion o f these mobile ions may alter the diffusion patterns of other ions, and may explain the inflexion curve o f the silica line.

34

Ian Merchant, Julian Henderson, David Crossley and Michael Cable

Calcium oxide and magnesia diffusion patterns (fig. 4) follow similar trends. That is, they show reasonable stability in the glass, a sudden decline, and little penetration into the clay. There is an indication o f a freezing point, where the calcium oxide and magnesia suddenly stop penetrating the crucible. A defined peak of iron oxide (fig. 5) occurs at the mid­ point between the glass and clay, in addition to a small peak of titanium dioxide (fig. 5). The composition o f the bright crystals, as seen in the back-scattered electron images, is around 50% iron oxide, 40% titanium oxide and 8% alumina, confirming the presence o f these peaks. Early examination of glass-making sands in Britain revealed the presence o f ilmenite (FeT i03) and rutile (T i0 2) (Boswell 1917), and since silica grains have been identified within the clay, it is probable that the crystal contaminants present at the interface point derive from sand used in the crucible, and may be responsible for a reduction in refractory performance, though, of course, it is not possible to ascertain the true life expectancy o f these Blunden’s Wood crucibles. The alumina curve (fig. 6) shows a reaction taking place within the clay, between the interface and 1000 microns into the crucible. The alumina composition appears to gradually increase from the outer face of the glass to the interface, rapidly rising at the border, reacting within the clay, and finally, levelling off towards the inner face o f the crucible. This shows similarities to observations made of Gallo-roman glasses, where a small transfer of material over the interface leads to a little contamination of the glass melt (Saleh et al. 1972; Velde 1990). Diffusion of alumina from the clay into the glass is taking place, in opposition to the movement of the alkali ions.

vi. D ifferences Between Types o f Crucible Surfaces Some crucible fragments are pitted, as though the surface has been attacked by the glass. Others have a flat surface, with a layer o f glass adhering to the clay, showing little sign of corrosion. Examples of these two forms of surface have been investigated to ascertain any differences in the form of corrosion. As before, changes in chemical composition were measured at regular intervals through the pitted or flat surface using EDS analysis. The results can be seen in figures 7 to

10. A significant difference between surfaces can be seen in the silica curves (fig. 7) where the thinner layer of flat glass has allowed weathering of the glass to take place, denoted by the relatively higher silica content. The level o f silica returns to a standard clay value by about 400 microns. The silica curves follow the same patterns for both types o f surface. The potash curves (fig. 8) show an increased level within the pitted surface. This may begin to explain the different attack on the surfaces. The higher level of mobile alkali ions penetrates into the ceramic structure, disrupts it, and erodes the clay. The corrosion by the alkali component may cause sand grains within the clay structure to be released into the melt, increasing the silica level in the glass. The stability of the alumina (fig. 9) over the flat surface may also explain the small level of attack by the glass. The clay structure is vital in halting the corrosion caused by the glass, so, if the clay matrix is able to remain intact, the crucible will be able to withstand the erosive effect o f the

glass. Conversely, the defined peak across the interface of the pitted surface shows that the alumina rich matrix is diffusing out, and therefore, the surface is being destroyed, causing the pitted erosion. The small fall in calcium oxide content (fig. 10) over the flat surface can be linked to the alumina curve, showing the transfer o f these oxides from crucible to glass (A120 3) and glass to crucible (CaO) is lower than that observed for the pitted surface. This is confirmation o f a reduced level o f corrosion.

Conclusions This study has shown the benefits that can be obtained by analysing the interface between glass and clay in Medieval crucible sherds. Small fragments, less than one millimetre across, allow excellent examinations to be made where little archaeological evidence remains. The presence o f mullite laths in the clay matrix may suggest that the temperature o f the glass melting process was in the region of 1400°C, as dehydration and crystal formation o f heated kaolinitic clay leads to the formation silicon spinel at 925°C, and pseudo-mullite at 1100°C before finally stabilising into m ullite crystal laths at 1400°C (Worrall 1975). The crucibles did show signs o f erosion, from the presence of the pits, and the diffusion of alumina from the clay to the glass. It is not possible to predict how well the crucibles were able to cope with the temperature, or corrosive and erosive attack by the glass, as the original wall thickness is not known. The glass composition is typical of Medieval potash glass, the major components being silica, potash, calcium oxide, and with traces o f iron oxide causing a green tint. The source o f sand is unlikely to be identified, though analysis o f the trace elements will enable the sand source to be characterised (Henderson 1985) and could possibly enable the location o f the deposit to be found, eventually. Further compositional analysis will be required. It can be assumed that the potash derives from plant ash, though the exact plant is harder to define. The high lime content may have come from the plant ash, though it is conceivable that chalk or limestone was added to the melt. Chalk can be found in the Downs surrounding the Weald, while limestone can be found in the parish o f Kirdford, south-east from Blunden’s Wood. It is unlikely that the W ealden M edieval glassm akers were adding colouring agents to their glass, as the type o f material they were producing is low to medium quality vessel glass. Therefore, the manganese and titanium oxides probably derive from impurities in the sand or ash. The reducing agent, manganese dioxide, may have derived from the plant ashes, and would have been largely responsible for trans­ forming iron (II) oxide into iron (III) oxide (Schreurs and Brill 1984). Knowledge o f chemical compositions is important in the determination of many glass properties. Models can be developed, giving a viscosity profile, which can provide details about the melting temperatures and characteristics of the glass (Bottinga and Weill 1972; Shaw 1972). The models are as accurate as viscosity measurements on small fragments o f glass, and are non-destructive, making them ideal for the archaeologist.

Medieval Glass-making Technology

Figure 3. Change in concentration o fS i0 2and K?0 over the interface

35

Figure 4. Change in concentration o f MgO and CaO over the interface

10 8 0 X 1 d

6

e /

*— — ----------------• -------• ------ • -------* --------------• — - —« — -500

-3500 -3000 -2500 -2000 -1500 -1000

------500 • --------• -------• 1000 1500

0

Distance from irrterface/micron

_____________________— TiQ2

Fe2Q3_________________

Figure 5. Change in concentration o f T i0 2and Fe20 3 over the interface

Figure 6. Change in concentration o f Al20 3and Na2Q over the interface

Figure 7. Change in concentration o f S i0 2 over different surfaces

Figure 8. Change in concentration o f K20 over different surfaces

36

lan Merchant, Julian Henderson, David Crossley and Michael Cable

Figure 9. Change in concentration o f Al20 3 over different surfaces

Figure 10. Change in concentration o f CaO over different surfaces

Investigation o f the interface adds information about Medieval glass technology. The attack seems to produce a diffusive layer, which may extend over time while held at the melting temperature. Extension of the interface may result in silica grains from the crucible becoming part of the glass melt. This may increase the silica level in the glass, and significantly, cause erosion o f the crucible. The infiltration o f impurities into the crucible can greatly reduce the temperature capacity o f the clay, so the elimination o f potash and soda from the crucible is paramount in the chemical stability o f the ceramic material (Douglas and Frank 1972). Diffusion o f silica into the Medieval clay body is very similar to observations made in a modem, high technology zirconia-alumina-silica material, though the small peak is present at the edge of the glass/interface region (Cable 1994). This shows that clay used for Medieval pots had mechanisms for resisting attack by molten glass similar to today’s materials. Investigation of the interface must be taken in conjunction with analysis of the glass and crucible fragments excavated from the glass-making sites. However, this study does show that the analysis o f the interface can enhance the picture obtained, and provide much information from very small fragments.

Cable, M. 1994. Attempting to Predict the High Temperature Corrosion of Oxide Ceramics by Molten Silicates. In Corrosion o f advanced ceramics: measurement and modelling (ed. K.G. Nickel), Kluwer Academic Publishers, pp. 285-296. Doremus, R.H. 1973. Glass Science. London, John Wiley and Sons, pp. 101-113. Douglas, R.W. and Frank, S. 1972. A History o f Glass-making, London, Foulis Co, pp. 96-132. Duvierre, G., Sertain, E. and Reber, A. 1993. Advantages of using High Zirconia Refractories in Lead Crystal Glass Electric Furnaces, Glass Technology, 34: 181-186. Hench, L.L. and Clark, D.E. 1978. Physical Chemistry of Glass Surfaces, Journal o f Non-Crystalline Solids, 28: 83-105. Henderson, J. 1985. ‘The Raw Materials of Early Glass Production’, Oxford Journal o f Archaeology, 4: 267-291. Henderson, J. 1989. The Scientific Analysis of Ancient Glass and its Archaeological Interpretation. In Scientific Analysis in Archaeology (ed. J. Henderson), pp. 30-62. Kenyon, G.H. 1967. The Glass Industry o f the Weald, Leicester, Leicester University Press. Mortimer, C. 1993. Analysis of Glass and Glassworking Waste from the Collections of Guildford Museum, Ancient Monuments Laboratory Report 106/93. Newton, R. and Davison, S. 1989. Conservation o f Glass, London, Butterworth and Co. Saleh, S.A. George, A.W. and Helmi, F.M. 1972. Study of Glass and Glass-making at Wadi-el-Natrun, Studies in Conservation, 17: 143-172. Schreurs, J.W.H. and Brill, R.H. 1984. Iron and Sulphur Related Colours in Ancient Glasses, Archaeometry 26: 199-209. Shaw, H.R. 1972. Viscosities of Magmatic Silicate Liquids: An Empirical Method of Prediction, American Journal o f Science, 272: 870-893. Thurrell, R.G., Worssam, B.C. and Edmonds, E.A. 1968. Geology o f the Country around Haslemere. London, HMSO, pp. 3262. Velde, D. 1990. Alumina and Calcium Oxide Content of Glass Found in Western and Northern Europe, First to Ninth Centuries, Oxford Journal o f Archaeology, 9:-l 05—117. Verita, M., Basso, R., Wypyski, M.T. and Koestler, R.J. 1994. Xray Microanalysis of Ancient Glassy Materials: A Comparative Study of Wavelength Dispersive and Energy Dispersive Techniques, Archaeometry, 36: 241-251. Vose, R.H. 1980. Glass. London, Collins. Whittaker, I.R. 1993. ‘Fusion Cast Refractories for the Glass Industry’, Glass Technology, 34: 129-135.

References Boswell, P.G.H. 1917. ‘British Glass-sands: their Location and Characteristics’, Journal of the Society of Glass Technology. 1(1): 3-61. Bottinga, Y., and Weill, D.F. 1972. ‘The Viscosity of Magmatic Silicate Liquids: A Model for Calculation’, American Journal o f Science, 272: 438-475. Brill, R.H. 1988. Scientific investigations of the Jalame Glass and Related Finds. In Excavations at Jalame, Site o f a Glass Factory in Late Roman Palestine (ed. G.D. Weinberg), University of Missouri Press, Columbia, pp. 257-294. Busby, T.S. and Shaw, G.B. 1989. Special Refractories: Properties and Applications, Glass Technology. 30: 117-119. Cable, M. 1968. The physical chemistry of glass-making. In Eighth International Congress on Glass, pp. 163-178.

Medieval Glass-making Technology Whittaker, I.R. and Lacey, J. 1990. Basic Refractories in the Glass Industry, Glass Technology, 31: 172-176. Winbolt, S.E, 1933. Wealden Glass. Hove, Cambridges. Wood, E.S. 1965. A Medieval Glasshouse at Blunden’s Wood,

37

Surrey Archaeological Collections, 62: 54-79. Worrall, W.E. 1975. Clays and Ceramic Raw Materials. London, Applied Science Publishers Ltd, pp. 42-78; 147-154.

Non-invasive Revelation of Precious Metal Decoration in Glass Beads Leo B iek 1 and Gordon G ilm ore2 1 City University, c/o 15 Belsize Square, London NW3 4HT; 2 Nuclear Training Services Ltd., 6 Alexandra Road, Grappenhall, Warrington WA4 2EL

In trod u ction The Cambridge Museum o f Archaeology and Anthropology houses the Horace Beck Collection o f beads. In the course of cataloguing its European Section two groups o f beads were felt to invite special attention. All the beads were unremarkable except that one or two of them looked as if they just might be ‘gold-in-glass’ beads. These are known from the Mediterranean area as far back as the first century BC but continue forward into the early medieval period (Boon 1977, B ieketal. 1985). Their section resembles that o f a ‘hollow Swiss roll’. The genuine ones were made by drawing out a clear tube about 5 mm in diameter, applying gold (or silver) leaf, and coating it with a further layer o f clear glass. The tube was then crimped at intervals to form a succession o f spheroids which was finally separated into (usually) double or triple ‘segmented beads’, but mostly singletons. The two groups in question (No. 1895 47.1967 and No. 1901 47.1967i) were stated to have been derived from Corsica, and were part o f a collection which included some distinctive ‘eye beads, probably Punic’. They could thus be prehistoric but the date is o f secondary importance since the method o f manufacture remained the same. In the African Section of the Beck Collection there is a group of silver-in­ glass beads from a Greco-Roman site in Egypt, preserved in pristine condition, clearly showing what such objects originally looked like.

E xam in ation In some cases (Biek et al. 1985; Bayley, forthcoming) it has been found that no precious metal had ever been in­ corporated. Instead the interface between the two glass layers showed only a carpet of bubbles. Such a feature clearly scatters light falling on it, so producing a highly reflective surface mimicking a metallic layer but forming a ‘no-goldin-glass’ bead. In northwest Europe, weathering will often create a very similar appearance and it is then not possible to decide from microscopical inspection which type of bead is involved. This also applied to almost all of our Corsican examples and the situation was further compounded by spurious red, brown, green and blue colorations in many cases (fig. 1). All 45 beads were examined without preparation under a binocular microscope (up to x40). In a very few cases, only, did it prove possible to recognise metallic areas with some confidence, and these appear to be true gold- or silver-in­ glass beads (plate 1, plate 2). The absence o f a clear ‘collar’

residue from crimping, such as is commonly seen, remains a slight puzzle. Previous experience (Biek et al. 1985) had shown that neutron activation analysis could provide unambiguous results and the facilities available at the Universities’ Research Reactor (URR), Risley (now de­ commissioned), were used to good effect on the Corsican beads.

A nalysis The beads were analysed individually and non-destructively by neutron activation analysis using the URR. In order to limit the activation o f long lived radionuclides a short irradiation time, one minute, was chosen. The weight varied from 0.05 g for a bead fragment to 0.9 g for a large bead. To compensate for this the reactor power was varied from 225 kW to 20 kW with most beads being irradiated at 50 kW (a mixed thermal and epithermal flux o f 2 .4 x l0 !l cnr2.s''). Gamma-ray spectra o f each bead were measured, using a Ge(Li) detector o f resolution 1.85 keV at 1332 keV, after one minute decay and again after 48 hours decay. From these spectra seventeen element concentrations were derived, in particular gold (via 198Au) and silver (via 108Ag). (The other elements, Al, Mg, Ca, Ti, V, Mn, Co, Cu, Sr, Sn, Sb, Ba, In, Na and As, can have relevance to bead studies but are not o f immediate interest here.) The measurements were calibrated by reference to NBS Certified Reference Materials and synthetic standards prepared from Specpure oxides and evaporated BDH Standard solutions. The limit o f meas­ urement for gold was about 0.1 microgram o f gold per bead compared to the content o f a gold-in-glass bead which might be some hundreds of micrograms. The limit o f measurement for silver was somewhat poorer, 30 or so micrograms, but still well below the content o f a silver-in-glass bead.

R esults The results are most impressive. Of the 45 beads as many as 28 contained more than 300 ppm o f gold or silver (fig. 1). Discrimination is remarkable; for example, compare bead 37 (1930 ppm o f silver and 8 ppm o f gold) with bead c l (no detectable silver, but 730 ppm of gold). Of the other elements measured, average figures for sodium were 11- 12%, antimony 500-4000 ppm and manganese 6000-9000 ppm. It is hoped that this method, or a similar one, used elsewhere may lead to further painless ‘unmasking’ o f precious metals in glass beads.

Concentration in ppm

Figure 1. Concentrations o f silver and gold in potential silver-in-glass and gold-in-glass beads. (cl-c7: Corsica 1895 47.1967, 1-38: Corsica 1901 47.1967i)

Non-invasive Revelation o f Precious Metal Decoration

'O

Plate 1. c3, silvery grey, Ag 1500 Au 3; c4, lustrous golden, Ag < 300 Au 1690; 2, dull pinkish yellow, Ag 310 Au 780; 3, opalescent silvery yellowish, Ag 240 Au 1280; 8, dull grey with pink patches, Ag 1600 Au 6; 21, dark mauve red opaque, Ag 120 Au 678; 25, opaque dull pink with red patches, Ag 1330 Au 14; 37, opaque dark red with bright patches, Ag 1930 Au 8. All ca. xlO. (Photos: Gwil Owen, Cambridge University)

Non-invasive Revelation o f Precious Metal Decoration

41

Plate 2. 3, as in plate 1, Extensive bubble formation on part o f surface where it masks yellowish reflective interior. 21, as in plate 1, Uniformly dark smooth surface in contrast with white opaque porous section along break, ca. x25

42

Leo Biek and Gordon Gilmore

R eferen ces Bay ley, J., forthcoming. Examination of the glass beads. In S.M. Hirst, Anglo-Saxon cemetery no. 2 at Mucking excavated by M.U and W.T. Jones. Biek, L., Bayley, J. and Gilmore, G.R., with contributions by H.P. Rooksby, J. Henderson and S. Wyles. 1985. Scientific

examination of the glass beads. In S.M. Hirst, An Anglo-Saxon Cemetery at Sewerby, East Yorkshire. York: University Archaeological Publications 4, pp. 77-85. Boon, G.C. 1977. Gold-in-glass beads in the ancient world. Britannia 8: 193-207.

Medieval Windows: What Can Chemical Analysis Tell Us? P.J. M ills and G.A. Cox Department o f Physics, University o f York, Heslington, York, Y01 5DD

Abstract This paper will compare the decay of Medieval stained glass windows from two English churches: St. Mary’s, Fairford and St. Mary and All Saints, Checkley. The glass in both churches has been chemically analysed using Electron-Probe Microanalysis, X-ray powder diffraction and Fourier Transform Infrared spectroscopy. Whether the window orientation plays a part in influencing the extent o f decay will be considered.

Introduction It has been known for a long time that glass is susceptible to attack by aqueous solutions and atmospheric agents. It was established as early as the late sixteenth century that panels of glass should not be stored face to face as this damaged the surface. It is important to note that if the surface is scratched or damaged during manufacture, these areas act as primary sites for attack. Medieval window glass is typically of the potash-lime-silica type, otherwise referred to as fo re st glass. This has a very different composition to the earlier Roman and the later Renaissance and modern glass. Roman glass was of the soda-lime-silicate variety, the alkali coming from sources such as mined sodium salts. Later European material more typically contains mixed alkali sources and modern window glass is similar to its Roman counterpart. It is the necessary inclusion o f additives, whether deliberately or otherwise, that affects the durability and hence affects the stability o f the glass over the centuries. Other factors that are to be taken into account (other than composition) are the environment in which the glass was made, the surroundings in which it has been kept, the nature of the attacking agent and the age o f the glass. The environment to which a window has been exposed is difficult to monitor and indeed little or no information is available on such. For example, panels could have been moved from church to church, moved within the church (reversing or transferring north to south etc.) or replaced. Late replacements o f individual pieces o f glass are normally easily identified by their chem ical com position; con ­ temporary replacements can be difficult to identify. There is good evidence to suggest that the positions of the panels o f stained glass, within the churches referred to in this paper, have not been altered since their installation. They have, however, been restored numerous times.

Background A great deal of research has been done on the subject o f the atmospheric attack of Medieval window glass. Perez-y-Iorba et a l (1975, 1978, 1980) and Bettembourg (1977) have studied the effect o f the atmosphere on French window glass whilst Fitz et al. (1984) and Fitz (1986) worked on German glass and Austrian glass was studied by Schriener (Schriener 1988). Previous work on English glass has been largely

centred on the glass o f York Minster (Cox et al. 1979; Gillies and Cox 1986, 1988; Newton 1988). The culmination o f the above work is that we now appreciate that water is the primary attacking agent and that even large quantities of such pollutants as sulphur dioxide will not cause damage unless water is present (Douglas and Isard 1949; Adlerborn 1971; Fitz et al. 1984). The mechanism o f decay (Paul 1990: 180) is thought to progress as follows: -

-

Water attacks the surface o f the glass producing a leached layer via an ion exchange reaction (by removing alkalis from the glass into solution and hydrogen ions replacing them). The solution is alkaline in nature and thus further attacks the glass network by breaking siloxane bonds. The products (generally expressed as hydroxides) can then react with gaseous oxides to form carbonates and sulphates.

There is some uncertainty as to whether they form sim­ ultaneously or whether carbonates form first and are subsequently converted into sulphates. Most sulphates are only weakly soluble, hence when water evaporates from the glass surface they are left behind as deposits. These are the decay products. Previous workers have identified amorphous silica, calcite, gypsum, syngenite, palmerite, epsomite and various unidentifiable phases (Perezy-Jorba et al. 1978; Gillies and Cox 1988a, 1988b). In the extreme form o f weathering, the decay products can cause the glass to become opaque and thus render the window less than functional. Some workers have claimed that the decay o f glass can be explained solely by bio-deterioration, although in England there is no evidence to suggest that micro-organisms are present let alone cause the decay (Krumbein 1991). Lichen has been observed on some French glasses and calcium oxalate was identified as an excretion product formed by bacteria (Perez-y-Jorba et al. 1980). No cases have yet been found in the U.K. where lichen is present on vitreous surfaces, although there are som e churches that have excessive algae and m oss growth on their north-facing windows. Also some fungal hyphae were identified at York Minster (Gillies and Cox 1988b). The decay o f glass can manifest itself in one o f three

44

P. J. Mills and G. A. Cox

general ways: apparently unweathered, pitted and crusted. The former is characteristic o f high soda high silica glass (Gillies and Cox 1986). The latter form is believed to be the final stage o f deterioration. Therefore decay is said to extend from isolated circular pits which vary in depth and diameter, to pits that merge together until the whole o f the surface has been affected. It is then that a decay crust builds up. Conservationists are faced with the problem of whether or not to remove these decay products. If they are removed then fresh glass is exposed to the environment, but if they remain they spoil the aesthetic beauty of the window. Leaving the products in place does not necessarily mean that the glass is protected. The sulphates do not form a barrier to the environment and so decay can proceed; they will, however, make any solution that forms less aggressive.

For each sample that had decay products, on the outer surface, a small amount was removed. This sample was analysed using Diffuse Reflectance Infrared Fourier Trans­ form spectroscopy (DRIFT) and then complementarily analysed using X-Ray Diffraction (XRD). The former provides information on functional groups present in the unknown, whereas the latter conveys information on the crystal structure. In the present application o f the techniques the unknowns are compared with standard minerals to improve the identification process. DRIFT is useful because it gives good results for a small sample; XRD gives a better indication o f the relative amounts o f individual compounds present in the mixture.

Discussion Location The specimens of glass described here are from St. Mary’s, Fairford, Gloucs. and St. Mary and All Saints, Checkley, Staffs. The Fairford (FAI) glass dates to c. 1500. The glass is of the later European type. The specimens from Checkley (CHK) date from 1320-1340 and the glass is typical o f the 12th-14th century. Each church provided 66 pieces o f glass for study whilst the windows were being restored by Keith Barley o f Barley Studios, Dunnington. The samples were selected as being representative of the most corroded glass present in the panels available. Importantly it was possible to select glass from both north- and south-facing windows, thus enabling orientational affects on the extent o f de­ terioration to be considered. In the case of Checkley there were 34 north- and 32 south-facing specimens, whereas the decay at Fairford was almost non-existent on the north side and therefore only 22 were selected from such panels.

Methods The chemical composition of the glasses was determined using a Cambridge Instruments scanning electron microscope (model 90B), in association with a Link Systems X-ray analyser (model A N 10000) and software. A small area o f glass (from under the leading) was carefully ground and polished to a 1 pm finish. Each sample was analysed five times by exposing it for 100 seconds to the beam to determine the major/minor components, some thirteen oxides in total. The weight percentage o f the oxides was then averaged and converted into molar percentages. Figure 1 provides a summary table of the compositions for each site. It can be seen that the amounts of silica and potash present in the two glass types are markedly different.

Location

It is possible to condense chem ical com positional in­ formation from values o f the oxides to three reduced variables (Illife and Newton 1974). These are collectively termed “ternary co-ordinates”. These co-ordinates represent the glass network formers (sSi02'), alkaline-earth oxides that modify this network f R O ’) and the effective alkali content (VR 2 0 ’). Figure 6 shows the type o f plot that can be produced from such data. It is clear that there are two principal groups. These groups are not due to the different ages o f the glass, as might be expected, but it appears that the blue/colourless/ruby glass from Fairford is o f a distinctly different composition to the remainder o f that group. In other words, the coloured glass at Fairford is not o f the expected late European type but is still that o f the Medieval period. This is probably indicative o f their having a common origin. An alternative way o f studying the stability o f glass (representing the general glass composition) is to calculate the free energy of hydration (Paul 1977) using the following formula; EQUATION AG=ZX(AGJ where X is the molar percentage o f the elemental oxide within the glass and AGX is the free energy o f hydration for the relevant silicate. The more negative the value the less durable the glass. Figures 2 -5 use this parameter as a means of comparing composition with the extent o f decay. In each case the markers identify different possible groups. The extent o f corrosion is determined by an arbitrary visual classification. The value 1 represents durable glass, 2 = isolated pits, 3 = merged pits, 4 = totally pitted surface and 5 represents one that has a thick opaque crust.

Ternary co-ordinates

mol% of oxides Na20

CaO

K20

MgO

Si02

A1203

P205

Fe203

MnO

‘Si02’

‘R20’

‘RO’

Fairford mean ± std n=66

2.3 ± 0.8

21.9 ± 3.4

5.4 ± 3.0

5.8 ± 1.6

59.7 ± 2.2

1.5 ± 1.4

1.3 + 0.3

0.2 + 0.1

0.9 ± 0.3

63.9 ± 2.6

6.2 ± 2.8

28.9 ± 2.3

Checkley mean ± std n=66

3.3 ± 1.2

15.9 ± 1.2

10.2 ± 2.1

10.9 ± 1.4

54.1 ± 1.9

0.9 ± 0.3

2.3 ± 0.4

0.2 ± 0.2

0.9 ± 0.1

58.6 ± 1.6

12.4 ± 1.2

27.9 ± 1.2

Figure 1. The chemical composition o f the glasses

45

Medieval Windows

▼

5—i

corrosion type

4-

ED

▼

© ▼©

-

CID

w w

^

wi0r cD 20 wt%) and low-tin (Sn 20 wt%) exploiting their specific properties such as musicality, golden lustre and non-toxicity to make thinly wrought beta bronze vessels (around 23 wt% tin), hot struck coinage, cymbals, bells and mirrors, is seen in a long tradition from the 1st millennium BC in the Indian subcontinent continuing into the present day (Srinivasan 1994, Srinivasan and Glover: in press). Microstructures of wrought and quenched high tin bronze o f around 23 wt% tin were found in artefacts examined by the author from a range o f contexts from the subcontinent such as vessel fragments from the Gandharan grave culture of Taxila (ca 1000 BC) (Courtesy: Glover, I. and Pakistan Archaeological Survey); bowl fragments from the Nilgiri megaliths and the Adichanallur burials o f the 1st millennium BC and two Chola platters o f around the 10th12thc. AD (Courtesy: Government Museum, Madras, India); and coins of the Vishnukundin period (ca 4th c. AD) (Courtesy: Krishnamurthy, R, Tamil Nadu Numismatic Society, India). Other analyses of Indian high-tin bronze artefacts are reported from Adichanallur by Paramasivan (1941); from the Eastern Indian Assur graves o f the early Christian era by Chakrabarti (1979); from Andhra Pradesh in South India in the early centuries AD o f the Satavahana period where coins o f high-tin bronze (around 17 wt% tin) are reported, and the Vishnakundin period (4th c. AD) where coins o f quenched beta bronze (with around 23 wt%) are reported (Sastry et al. 1983: 21-9; Vijaykumar et a l 1987: 137-47); from Kulpauk village, Warangal district, Andhra Pradesh o f the Kakatiya period where six 11th century temple bells o f high tin bronze are reported (Rao 1981: 27-35). Ethnographic and metallurgical studies undertaken by the author following field investigations in 1991 showed that wrought and quenched high-tin beta bronze vessels, spatulae, gongs and cymbals are still made in Palghat district in Kerala in southern India and in eastern India (Srinivasan 1994) and that cast high-tin bronze mirrors are still made of 33 wt% tin in a traditional workshop at Aranmula in Kerala as recorded by the author and I. Glover (Srinivasan and Glover; in press): these demonstrate the continuity o f hightin bronze techniques in the subcontinent which appear to be indigenous. 104 statuary images were analysed for composition by ICP-OES by the author with half sampled from the major collections o f the Government Museum, Madras and the other half from Victoria and Albert Museum. About 50 of these were bronzes earlier than the 11th century with an

Bronze Slags from Kalyadi and the Problem o f Tin in South Indian Bronzes average tin content o f over 5 wt% and the rest of the bronzes were o f the 12th-18th century with an average tin content of slightly less than 5 wt%. The highest amount of tin recorded is 15 wt% in an image of Bhudevi (acc. no. IM -137-1927, Victoria and Albert Museum) of about ca 900 AD. Earlier im ages were also alloyed with tin such as an Andhra Satavahana image from the Victoria and Albert Museum (acc. no. IS -8-1989) of ca lst-2nd c. AD analysed by the author with 10 wt% tin, while an Andhra Satavahana image from Amravati analysed by Swarnakamal (1980: 54) had 16 wt% tin. The alloying of tin o f 10 wt% would have improved the hardness and fluidity o f copper. However about 15 wt. % tin alloyed with copper is the limit for the formation of the solid solution o f tin in copper designated as the (a) alpha phase. At higher tin contents the brittle (a+8) alpha plus delta eutectoid forms which would have greatly embrittled the as-cast bronze (Reeves et al. 1953). Hence this limit does not appear to have been exceeded in the manufacture o f statuary bronzes. However higher tin bronzes were clearly used in South India, as mentioned before, under specialised conditions of rapid cooling and quenching, intended to reduce the brittleness of high-tin bronzes as illustrated in Srinivasan (1994) and Srinivasan and Glover (in press). The bronze bowls with 23 wt% tin of South Indian antiquity and those still made in Kerala are hot forged followed by quenching, while cast mirrors made in Kerala o f a specular alloy o f 33 wt% tin in copper are very thinly cast with rapid cooling to reduce the brittleness (ibid). From about the 8th—10th c., beginning with the Pallava and early Chola dynasties, large solid cast bronze images began to be made on a fairly large scale, with scores o f images reported from collections and shrines in South India, often o f a metre high. Taking all this evidence together there were clearly accessible and sub­ stantial supplies o f tin in ancient South India from early periods. However, given the relative scarcity of tin in India, the source of tin is not easily guessed.

Literary and Archaeological Evidence for Tin in South India and Other Parts of the Indian Sub­ continent, and for Trade in Tin Tin deposits, which are rather scarce in India, are never­ theless reported in small amounts in parts o f Northeast Bihar, Gujarat, Rajasthan and Karnataka (Hegde 1978:39-42) while the exploitation of stream tin would not leave traces behind. The ores in north-east Bihar, from Gaya, Ranchi and Hazaribagh are thought to average 1.87 wt% tin but may yield as much as 6.2 wt% tin (Reedy 1992: 244). Although the cassiterite hard rock occurrences in the Singhbhum and Bastar region in the Chota Nagpur plateau in North India are uneconomical in modern terms, the Singbhum area in general shows extensive evidence o f mining in antiquity (Chakrabarti

1985-6: 65-70, Hegde 1978). Brown and Dey (1955: 167) refer to the smelting of tin ore in 1849 in the village iron smelting furnaces in Parg, while M allet (1874: 2 3 -44) discusses the secretive smelting of tin in iron smelting furnaces by the Bastar tribals in Madhya Pradesh. Finds of high-tin bronze bowls are also reported from this area in antiquity in the tribal Assur graves o f Chota Nagpur (Chakrabarti 1979: 65-74) of about the 1st century BC. In South India, reports by Geological Survey o f India

137

indicate small occurrences of tin in Karnataka as alluvial tin and hard rock tin. Sparse alluvial tin ores are reported in the Dambal area ofDharwar district near Gadag (Mukherjee and Dhaneshwar 1978: 3 4 5 -3 5 0 ). C assiterite occurences are reported at Challi, Chitradurga district, Karnataka (Mukherjee and Dhaneshwar 1978:345-350) and in granites in Kavital, Raichur district (pers comm., Dr Raghunandan, Geological Survey oflndia-Bangalore), while the tourmaline­ bearing pegmatites of the auriferous Mangalur schist belt are reported to show tin traces (Dr Raju, Geologist, Hutti Gold Mines, pers. comm.). According to Kuppuram (1986: 90) the Nidugal Cholas ofHemavati who were feudatories of the Cholas in Karnataka are said to have collected a fee on all exported goods which included gold, lead and tin along with other articles such as sandal and areca. Significantly Malroney (1975: 26) also m entions a literary reference to the export o f tin by Solomon’s navy from the Karnataka coast along with other items including peacocks, iron, gold, silver, ivory and apes. Given the literary references to tin from Karnataka, the significance o f the sparse deposits mentioned earlier need to be kept in mind. O f course placer mining of tin would not leave behind many traces and it is noteworthy that there is strong evidence of much placer mining for gold from the regions where tin is reported. The whole area of Karnataka appears to have been quite extensively prospected in antiquity for gold, both placer gold and auriferous quartz, with radiocarbon dates for timber suggesting that the gold mines at Hutti were mined to a depth o f 600 feet by the mid to late 1st millennium BC (Radhakrishnan and Curtis 1991:23-4, Allchin 1962). Tamil Sangam literature (ca 3rd c. BC - 3rd c. AD) mentions a merchant who came from the gold bearing region or Tulunadu, the modern Tulu speaking region ofKonkan in coastal Karnataka (Kuppuram 1986:17). In this region alluvial gold is found in the auriferous Dambal hills of Gadag district, while some minor occurences o f alluvial tin are also reported in the Dambal and the Dhoni Nallah region as indicated by Foote (1874: 140) and Mukherjee and Dhaneshwar (1978). To quote Foote (1874: 140), he says the following o f the ‘other metals found in the gold sands’ o f the Dambal and Dhoni Nallah: ‘the stream-gold is found associated with a black sand consisting o f magnetic iron in minute octahedra and a black residue not affected by the magnet. In the sand washed in the Dhoni nullah I found several minute rounded grains o f a grey metal which on further examination proved to be metallic silver. A couple o f little spangles o f pale yellow ish silvery hue were also obtained which were doubtless electrum, the natural amalgam o f gold and silver; beside these there were a few minute grains o f bronze colour which on examination proved to be a mechanical mixture o f metallic copper and oxide o f tin.’ The Jalagarus o f the Dambal region are described in the nineteenth century accounts of both Foote (1874: 140) as a sect o f people who carried out alluvial washing and panning for gold after the rainy season when the heavier gold would get washed downstream from the auriferous hills. Hence it is not impossible that some placer tin may have been discovered and exploited in antiquity with all the prospecting for placer gold. Mining o f placer tin would leave few traces if it had been exploited to depletion, and indeed some tin traces are found in the alluvial gold bearing area o f Dambal, Gadag district (Foote 1874: 140).

138

Sharada Srinivasan

It is an interesting coincidence that finds o f gold ornaments have also been made along with the high-tin bronzes of the South Indian Iron burials of Adichanallur (Allchin and Allchin 1982: 337), and the Nilgiri megaliths (Leshnik 1974: 265). Allchin and Allchin (1982: 337) also point out that evidence of gold is found in the South Indian burial complexes, which they believe to have developed indigenously. Abundant alluvial iron deposits such as magnetite are also found in South India where there is significant evidence for iron and steel manufacture in the past. Smelting o f tin in iron furnaces akin to those described in eastern India could have possibly been used in the context of the placer tin deposits of Karnataka. However, based on the sparse tin reserves today, one may only speculate if such activity took place. As for other possible sources o f tin, there is the account in the Periplus of ca 1st c. AD which mentions that the important imports into Muciri, on the Malabar coast, and western India were lead, copper and tin (Warmington 1928: 267-9). Afghanistan is also thought to have been a likely source of tin for India in antiquity (Muhly 1985: 281). Stannite and cassiterite with from a few tenths of one weight percent tin up to 11 wt% are reported from parts o f Afghanistan along with small amounts o f alluvial cassiterite, while Misgaran in Herat province is reported to show evidence of early exploitation (Reedy 1992: 244). There is also evidence of maritime links with Southeast Asia, which is rich in tin, under the Chola kings of South India, Rajaraja, Rajendra Chola and Kulottunga Chola (10th12th centuries AD) and Southeast Asia has been postulated as a source of Indian tin (Ray 1990: 4). A Tamil inscription from coastal Nangoor, Sirkali district, which included Tiruvengadu, a major find spot of Chola bronzes, was found in Takuapa in Thailand (Coedks 1968: 107). A charter from Southeast Asia refers to the foundation of a Buddhist vihara or monastery at Nagapattinam by the Sailendra ruler o f the Srivijaya kingdom (i.e. Java, Sumatra and Kedah in the Malay peninsula), patronised by the Chola kings and several traders from Southeast Asia went to and from Nagapattinam (Coedes 1968: 141). Contacts between Burma and South India by the early historic period is indicated in the affinities of the Pyu culture o f Burma with the Andhra Satavahana styles. A Chinese account of the Burmese Pyu kingdom of ca 800 AD mentions that the Pyu houses were roofed with lead and tin shingles (Coedes 1968: 104). The Arab writer Abu Zayd in the second decade o f the 10th century wrote that Kalah (in Sumatra) was the centre of commerce for goods including tin along with ivory, ebony, sandalwood and others (Wheatley 1983: 236). However, there is as yet no totally unambiguous evidence for trade in tin between South India and Southeast Asia. Few medieval bronze images have been found in the tinrich Malay peninsula or from excavations at Palembang, Sumatra, which is thought to be have been the centre of Srivijaya (P. Yves Manguin, pers. comm.) nor from Burma compared to the other parts of Asia under discussion. There is other literary evidence of trade in tin such as the Geniza documents of Jewish merchants from Cairo from the 11th century of the Chola period for the brisk trade in metals from medieval South Indian ports. They mention that iron and steel, brass and bronze vessels were regular commodities traded out of South Indian ports and that copper, tin and old

bronze vessels were sent to South India where new ones were created and old repaired (Goiten 1963:196): indicating that the metalworkmanship of South Indian craftsmen was highly prized. Nevertheless there are som e indigenous sources o f tin in South India and som e evidence for indigenous exploitation of tin as mentioned earlier. Hence the finds o f bronze slags from Kalyadi, Karnataka discussed further are interesting in that they lend weight to the literary evidence for tin from South India.

S u rface F in ds o f B ro n ze Slags w ith 5 w i% T in fro m Kalyadi* Hassan District* K arnataka

Description o f M ineralisation and Old Workings In the vicinity o f Kalyadi in Arsikere district, large scale open cast workings were described as gold workings by Jayaraman in the late 1930’s, but were later identified as being copper workings in 1939 by Radhakrishna (1967:17). The copper ocurrence is located about 200 m south-east o f the village o f Kalyadi (13° 14': 76°9‘) in Arsikere Taluk, Hassan district on the eastern flanks o f a ridge (average elevation 880 m above m.s.l) locally known as Belligutti (silver mountain). This deposit is now being developed by Hutti Gold Mines through the Kalyadi Copper Unit and contains an estimated 10,000,000 tonnes o f ore-bearing rock containing 0.60 wt% copper (Radhakrishna 1967: 10). The area forms part o f the Archaen Complex o f Mysore o ff the Nuggihalli schist belt and consists ofDharwar metamorphics associated with ultrabasic intrusives amidst a peninsular suite of granite gneisses. The ores occurs in the form of thin and sporadic malachite encrustations and limonitic boxwork localised in cleaved and fractured quartzites and quartzschists with the lodes and foliation planes dipping N.E. at about 70°. The host rock quartzite is noted over a strike length o f about 1000 m along the N .W -S.E direction (Murthy, n.d., GSI-Misc. Pub. No. 13: 156-9). In 1991-2 preliminary surface surveys o f som e old workings and slag heaps in South India were made by the author to explore the possible base metal deposits for ancient South India, and Kalyadi was one o f the sites visited. The old workings at Kalyadi are found at a elevation of over 910 m on Belligutti hill and old trenches and workings are seen over a strike length o f 300 m. I acknowledge the assistance o f Mr A. Rahim, Geologist, Kalyadi Copper Unit in visiting the old workings and slag heaps. There is one long continuous trench o f length about 250 m and width ranging from 10-35 m. It has been extensively worked by the miners of old and shows four parallel workings. From here a long line o f workings extends continuously and then turns southwest along the drag fold axis. The old miners appear to have followed the ore shoots and also developed on the drag fold axis (Murthy GSI-Misc. Pub. No. 13: 160). A small working o f 3.6 m x 2.0 m x 3 was noticed in the northern portion with thick malachite and azurite stains and encrustations on joint planes. The lodes are concealed in the workings by highly silted material, which indicate that the workings had been developed to a depth and have collapsed. The drillings by the Kalyadi copper unit, which has a mine shaft barely 30 m away from these old workings, has not intersected the working.

Bronze Slags from Kalyadi and the Problem of Tin in South Indian Bronzes Archaeom etallurgical Evidence o f Smelting in the Region Considerable amounts of slag are found in the flanks o f the hill and valley south-west of the main working from 10 m below the trenched area, near the old mine shaft spread over an area of 200 m x 40 m, and also at Ranganahalli 2 -3 km away. The slags could be seen on the surface of several buried mounds along with numerous tuyere fragments. However no dumps o f waste rock and gangue are seen at all which may because the host rock of mica-schists themselves found use as a building material. Many of the slag pieces were broken as the area had been bulldozed by the Kalyadi Copper Unit which has a mine shaft close by. Several mounds of low height of up to three feet were found in the region and these had innumerable tuyere fragments with slag thickly adhering to them (fig. 1). The largest unbroken slags were oval, up to 20 cm in length, ten cm across and about three cm thick (plate 1, fig. 2). These slags could be described as tapped ‘plate’ slag (after Merkel 1983:428) with a flow texture, and they had a purplebrown sheen.

b) Plan

Figure 1. Sketch o f slagged tuyere fragment from surface collection at the old slag heaps o f Kalyadi, Hassan district, Karnataka, South India.

139

Local History o f the Mining Region South Karnataka came under important southern dynasties such as the Satavahanas, Western Gangas, Hoysalas and Cholas as attested by abundant copper plate grants and was of economic and strategic importance due to the presence of rich auriferous deposits in the Kolar region. Several important Jaina bronzes are known from this region, some of which are in shrines such as Sravanabelagola, an important Jaina centre right from Mauryan times (4th century BC) according to inscription. The Cholas took over this region from the Gangas by the early 11th century and their interests in the gold bearing region o f Kuvulava, or Kolar, one o f the Ganga capitals, is indicated in the several copper plate grants in the region (Kuppuram 1986: 3, 85). The Hoysalas rose to ascendancy in the 11th c. and their abundant intricately carved potstone temples such as at Belur and Halebid in Hassan district are no more than 30 km from Kalyadi and attest to the extensive quarrying o f the local ultramafic hornblende schists. A large number o f corundum workings have been found in the region o f Kalyadi, Harnahalli and Arsikere and indeed the abundant lathe-turned and polished potstone pillars o f the Hoysalas indicate the use o f corundum powder for polishing. Although the slag heaps have not been systematically excavated, a small trial excavation was made at a small mound close to the Kalyadi copper mine by Poonacha, K.R. and Rao, S.R (Poonacha; K.R. pers. comm.). The excavation report states that ‘the excavation was found to have three layers with the topmost being o f burnt blackish earth with the middle being composed o f large quantities o f slag, and the bottom containing charcoal, fragments o f red ware, slags and burnt earth. Thick potsherds used as refractories were also excavated in the exca v a tio n ’ (Thapar 1980:26). According to Poonacha (pers com m .) the pottery was believed to be of the Hoysala period although the charcoal was not dated. The ASI- Bangalore were not able to make this material available to me for study. In the 14th c. the region came under the Vijayanagara rulers of Bellary. Timber from an abandoned working (PRL-254) has been carbon dated to AD 1630 (320±90 BP) (Thapar 1980: 85).

Analysis of Slag Specimens from Kalyadi and Discussion Preliminary non-destructive investigation o f the slags by XRF showed traces of copper in slags from Kalyadi. Crosssections of two slag specimens which had the plate-like structure of tapped slag were subjected to SEM-EDS analysis (using a HITACHI 5 -5 7 0 with Link AN 10000). Most interestingly, and rather unexpectedly, quantitative EPMA analysis (using a JOEL Superprobe JXA-8600) indicated the microscopic prills contained were in fact o f alpha bronze with about 5 wt% o f tin in copper (plate la ) and micro­ analysis by EPMA on prills in another specimen o f the same slag further confirmed that the bronze prills had around 5 wt% tin (fig. 3). Apart from the bronze prills, optical microscopy showed that the cross-section of the slag had a deep grey-green matrix. This consisted o f duller green lathes, lighter green globules and dendrites, as well as droplets and globules with a white metallic lustre with shapes consistent with metallic iron. SEM-EDS indicated that the lathes were of fayalite (iron silicate). EPMA analysis of the globules seen

140

Sharada Srinivasan

a) Plan

b) Section

Figure 2. Sketch o f tap slag specimen from old slag heap fro m K alyadi showing size, curvature and plate-like structure, with a shape consistent with smelting slags.

Plate 1. Micro-structure o f section o f slag from Kalyadi, under back-scattered atomic number contrast mode on SEMEDS, showing bronze prill with 5 wt% tin (bright patch in centre).

under back-scattered atomic number contrast confirmed that they were o f metallic iron (plate 2). The bronze prills, globules and dendrites contained no sulphur. These slags clearly have a high content o f iron-rich constituents and oxides, and o f metallic iron, which is much more likely to be a feature o f smelting slags as globules o f metallic iron would only be formed at high smelting temperatures under highly reducing conditions, as agreed by N.J. Seeley, and I. Freestone, (pers comm.). Indeed the slags, being rather large o f up to 20 cm in diameter and plate-shaped (plate 1, fig. 2) and weighing about 1 kg, do appear to be smelting slags rather than casting slags as agreed by Griffiths, D (pers. comm.). Moreover although innumerable tuyere fragments were found around the site not a single crucible fragment was found: further suggesting that this was smelting slag rather than casting slag. Furthermore the alloying o f copper with tin metal would be expected to be done at the foundry or casting centre rather than at remote ore deposits. It is worth speculating whether some co-smelting process of copper and tin ores was taking place. Stannite (Cu2PeSnS4; steel grey) is a tin ore which contains both copper and tin and could result in bronze on smelting. Bachmann (1982: 26) mentions that when sm elting copper ores with a noticeable tin content, reduction o f tin is favoured in the presence of copper due to intermetallic affinity. He postulates that, consequently, the manufacture o f low tin bronzes was possible at an earlier time than the reduction o f tin ores to tin metal, by smelting mixed copper-tin ores. However the sample he discusses from Central Germany contained no

Plate 2. Micro-structure o f section o f slag from Kalyadi, under back-scattered atomic contrast mode on SEM-EDS, showing smooth edged globules which were found to be of metallic iron.

Bronze Slags from Kalyadi and the Problem o f Tin in South Indian Bronzes

Kalyadi Slat; Bronze £riil l Brooze Prill 2 Bronze ^rill 3 Bronze Prill 4 Bronze Prill 5 Bronze thrill 6 Fe Globule Fe Globule Irregular Globule Latbe Lame

Cu7o 64.214 63.996 71.209 72.1 68.168 70.015 0.621 0.019 0.001 O.60I 0.016

Zn%

5------0 0 0 0 0 0 0 0

m

r

0.062 0 6 0.00) m

r

mm

mm mm 5— 0.034

nr

m r - As% B IT - m r m r 5------ m r 5—

0.024 T O ... 0 0 5------6.012 0 m r 6------6 0 0.023 * 0 4.198 5.016 4.212 5.5 o r 6 0

b

m r0 7 ^ 1 ---------r a r m r n r -m r m r Concentrate 21.3 Gangue 0.59 oTOTT EWK31 f

F E L L IN G

I

GOKSTAD

........ ...

=1 3 3 3

[

grave chamber

C

i

c

OSEBERG

I-------

grave chamber C

C

l................... r~-------- C

[ 600

T

T “ 700

' i .T"

800

834

900910

Figure 3. Diagram showing the dating o f the samples from the grave chambers (The rectangles show the samples ’ position in time. Their length shows the number o f tree-rings preserved in each sample. Ten o f the samples from Oseberg had sapwood preserved (black). Five o f these even had the bark-ring preserved, which makes it possible to show that the burial chamber was constructed in the summer o f 834 AD).

obtain an estimated felling date around AD 900, presumably between 900 and 905 and this is also the date for the construction of the grave chamber and the burial at Gokstad. In this investigation a preliminary estimate o f the number of tree-rings in the sapwood o f oaks growing in Southern Norway has been calculated (Bonde and Christensen 1993). Finally, the local chronology from Tune spans the period from AD 709 to 892, and both samples have been dated. If we again correct for the missing sapwood, we obtain an estimated felling date around AD 910. Here too, this reveals the date o f the construction of the burial chamber and o f the burial itself. The dates for the three burials also apply to the other artifacts which the graves contain. The dating of the grave goods is ante quern, that is, that the various items in the finds were produced prior to the year in question. The exciting part was whether the Norwegians, in the light of these results, wanted to proceed with the second phase of the project - the dating of the ships. This meant that it was necessary to sample the ships on exhibition, that is saw pieces out of them. After a couple o f months of deliberations the go-ahead was given. The Oseberg ship is 24 yards long and five and a half yards broad and about 90% of the original ship is preserved. The carvings on the ship are the largest and most important example o f monumental decorative art from the early Viking Age (Plate 2). A dendrochronological investigation o f the ship would thus not only give the date of construction of the ship, but also date this very early example o f ornamentation. In all 12 samples were taken from the ship’s planking, six each from the port and starboard sides. They were taken deep under the ship from the planks closest to the keel, so that the operation was not visible to visitors to the museum. The planks were sawn through with a fine hacksaw and as mentioned earlier the samples will be returned and reinserted in the ship.

Eleven o f the samples have been dated of which 5 have sapwood preserved. Accordingly, we can estimate the felling dates to around AD 820, which also reveals the date for the construction of the ship. The Gokstad ship is app. 25 yards long and five and a half yards broad. A total o f 13 samples were taken, and all have been dated. As none of these samples have sapwood preserved, correction has to be made for the missing sapwood rings. We can therefore conclude that the construction o f the ship took place in the last decade o f the 9th century, presumably between 895 and 900. Finally, four samples were sawn from the Tune ship, which originally was around 21.5 yards long and 4.5 yards broad and 40-50% o f the ship is preserved. Here too the samples have been dated and combined to form a chronology. Sapwood was preserved on one o f the samples and the construction of the ship can be dated to ca. AD 910. Up until now it has been assumed that the ships in the burials at Oseberg and Tune were old and outdated ships, which had had a long and active life prior to their being used in the burials. The dates for the construction of the ships and for the mounds at each of the three sites show on the contrary that it was relatively new vessels which were used. In each case the ship was no more that 10-15 years old. In the Skuldelev find mentioned previously, where five vessels had been sunk in order to create a barrage, in­ vestigations have shown that all the ships were very old - in the region of 100 years and worn out - when they were abandoned at the start o f the 12th century (Olsen and Crumlin-Pedersen 1969. Bonde and Crumlin-Pedersen 1990. Bonde eta l. 1991). Tree-ring analysis o f ships components excavated in London’s medieval harbour area show that ships were as much as 80 years old before they were broken up and reused in the revetments along the river Thames (Tyers 1994). This is not surprising. When a ship has been broken up or otherwise abandoned then it is understandable

Niels Bonde

200

Oseberg Gokstad Tune

Ship Construction date

Burial Chamber Construction date

when archaeological finds, both new and old, are to be preserved for posterity.

c. AD 820 c. AD 895-900 e. AD 910

AD 834 c. AD 900-905 c. AD 910-915

Acknowledgements

Figure 4. Summary. Tree-ring dates fo r the construction o f the ships and the burial chambers at the three sites.

that these were old, worn out vessels, as much as 100 years old. W hereas it might be a little surprising that the investigations o f the Norwegian ships show that it was a modern and fully functional vessel that carried the deceased into the afterlife. The dating of the burials at the three sites also date the finds/artifacts which the burials contain. Archaeologists and historians who work on the various artistic styles of the Viking A ge have hereby been given strong fix-points to be used in the evaluation of finds from the early Viking Age. The precise dating of the Oseberg ship burial is however way out o f the ordinary. In a strange way it takes one much closer to the incident - as if one was looking over the shoulder o f the person in front. Here too it appears that the Oseberg find exceeds everything else. And - with the establishment o f the construction date for the Oseberg ship, with its rich ornamentation, new possibilities are opened up for the interpretation of this unique find, not just o f artifacts o f southern Scandinavian origin but also those o f Irish origin which are present in the find. The investigation of the finds from Oseberg, Gokstad and Tune are also remarkable in another respect. In all three cases we are dealing with a site which was excavated many, many years ago. The dendrochronological Laboratory at the National Museum o f Denmark has, in recent years, carried out a number o f investigations o f this kind. For example the stocks of two canons, salvaged in 1846-47 near the island o f Anholt in the Kattegat, have been investigated and dated. Today the canons stand in the Royal Arsenal Museum in Copenhagen (Bonde et al. 1993). The many oak coffins from the Bronze Age, which were excavated at the end of the last century, of which the majority is to be seen at the National Museum, have similarly been investigated with very impressive results (Christensen and Jensen 1991). A further example is the Nydam ship which was excavated in 1863 and is now on exhibition at the Archaologischen Landesmuseum in Schleswig (D) (Bonde 1990). It is thoughtprovoking that we can extract new and important information from objects which have stood for decades on exhibition or in stores in museums around the country. Objects which are well-preserved and well-stored but which, in many cases perhaps suffer from lack o f interest due to the public’s quite overwhelming obsession with new and spectacular finds. Everybody working in museums should keep this in mind

I would like to thank The University Museum o f National Antiquities in Oslo which made this investigation possible, in particular Evabeth Astrup and special thanks are due to dr.phil. Arne Emil Christensen, who was the projects omnipotent Norwegian anchor man. Translation: David Earle Robinson

References Bonde, N. and Crumlin-Pedersen 1990. The dating of Wreck 2, the Longship from Skuldelev, Denmark. NewsWarp, 7: 3-6. Bonde, N. 1990. Dendrochronologische Altersbestimmung des Schiffes von Nydam. Offa. 47: 157-168. Bonde, N., Christensen, K., Eriksen, O.H. and Havemann, K. 1991. Dendrokronologiske dateringsunderspgelser p i Nationalmuseet 1990 - Dendrochronological Dating at the National Museum of Denmark 1990, Arkceologiske udgravninger i Danmark 1990. pp. 226-242. Bonde, N., Bartholin, Th., Christensen, K., Daly, A. og Eriksen O.H. 1993. Dendrokronologiske dateringsunderspgelser pa Nationalmuseet 1992 - Dendrochronological Dating at the National Museum of Denmark 1992, Arkceologiske ud­ gravninger i Danmark 1992. pp. 305-21. Bonde, N. and Christensen, A.E. 1993. Dendrochronological dating of the Viking ship burials at Oseberg, Gokstad and Tune, Norway. Antiquity. 67: 575-83. Brpgger, A.W., Falk, Hj. and Schetelig, H. (ed.) 1917-28. Osebergfundet I-III, V. Kristiania (Oslo): Universitetets Oldsaksamling. Christensen, K. and Jensen, J. 1991. Egtvedpigens aider. Nationalmuseets Arbejdsmark. 10-19. Christensen, K. and Havemann, K. 1992. Modern Oak chronologies from Norway. Dendrochronologia. 10: 137-146. Christensen, A.E., Ingstad, A.I. and Myhre, Bj. 1992. Osebergdronningens grav. Oslo: Schibsted. Gokstadfunnet. Et 100-irs minne (Cetenary of a Norwegian Viking find The Gokstad excavations). Sandefjordmuseene Arbok 1979-80. Holmboe, J. 1917. Botanisk underspkelse av torven i Oseberghaugen. In Brpgger, A.W., Falk, Hj. and Schetelig, H. (ed.), Osebergfundet I: 201-5. Kristiania (Oslo), Universitetets Oldsaksamling. Marstrander, S. 1986. De skjulte skipene. Oslo: Gyldendal Norsk Forlag. Nicolaysen, N. 1882. The Viking ship discovered at Gokstad in Norway. Kristiania (Oslo. Alb. Cammermeyer. Olsen, O. and Crumlin-Pedersen, O. 1968. The Skuldelev Ships (II). Acta Archaeologica. 38: 73-174. Schetelig, H. 1917. Tuneskibet. In Norske Oldfund II. Kristiania (Oslo): Universitetets Oldsaksamling. Tyers, I. 1994. Dendrochronology of Roman and early medieval ships. In Peter Marsden. Ships o f the Port o f London (ed. P. Marsden). London: English Heritage, pp. 201-09.

Where Does The Timber Come From? Dendrochronological Evidence of the Timber TVade in Nothern Europe N iels B on d e,1 Ian Tyers2 and Tom asz W azny3 1 Natural Sciences Research Unit, National Museum o f Denmark; 2 Archaeological Science Research School, University o f Sheffield; 3 Faculty o f Conservation, Academy o f Fine Arts, Warsaw

Introduction Dendrochronology can provide more than datings. A lot o f additional information can be obtained from the samples: the most obvious concerns the quality o f the timber and information about conversion of timber etc. In favourable cases dendrochronology also provides information about the origin o f the wood - where the trees grew, i.e. the provenan­ ce of the wood. The question of provenance of the wood - or timber, which is the term we ought to use - is o f extreme importance in dendrochronological research on old shipwrecks and on wooden artifacts. Often the place where a large sea going ship was built will not be the same as where the wreck was found and excavated hundred o f years later. It is not always that we have such luck as with the Swedish man-of-war (the VASA) that sunk on its maiden trip just outside the shipyard in Stockholm where it was launched in 1628 AD. In the same way international trade can bring a work o f art hundreds o f kilometres from the workshop where it originally was produced - not to mention the problems o f

determining the place o f origin o f the material used in the making o f the object! An engraving from the beginning o f the 17th century (plate 1), illustrating the town o f Szczecin/Stettin, clearly demonstrates some o f the problems that the dendrochronogist has to take into account when he or she is doing research into material coming from ships or related to trade. On the right of the picture, transportation of wood on the river Oder is taking place. The wood was transported from inland Europe to the harbour o f the city by barges or by floating downstream. In the town it was turned into timber and placed in a timber yard for re-sale. The ‘Neu Zimmer H o ff - the New Timberyard was situated close to the river. The question is from which region/area did the timber actually originate and where did it go? On the left side o f the engraving, the large sea going ships are seen sailing in to the harbour to take on their cargo. Some took timber to the Netherlands, to England, to Scandinavia, etc. even as far as Portugal. Where were these ships built? In the countries just mentioned or in the shipyard

Plate 1. Engravering showing the town o f Szczecin/Stettin. Braunius and Hogenherg 1618/23. (Photo. The Royal Library, Copenhagen).

Niels Bonde, Ian Tyers and Tomasz Wainy

202

shown to the left in the picture with the timberyard beside it - the ‘Schifbauer Lastade’ and the ‘Delen H o ff. Where did these ships end up? On the bottom of the sea somewhere, or perhaps some were broken up and the timber reused in other constructions. Similar questions can be asked if you are examining wooden artifacts. The wood could have been turned in to planks or ‘wainscots’ and sold to a local workshop, which used the material to cover the walls in a house with panelling, or sold to a workshop where altars and crucifixes (poly­ chrome sculpture) were produced. Most likely the bulk of the timber was exported as semi-products, wainscots, boards, barrel-staves etc. for resale or use in workshops in the big commercial centres in northern Europe. To deal with the problem o f the origin of the timber the dendrochronologist has to establish a huge network o f chronologies covering all the possible regions o f origin of the material which he or she receives. The foundation stone o f dendrochronology is that trees living under the same conditions will grow in the same way, and the conditions in the centre of eastern Europe are certainly quite different from those in Scandinavia or England. Based on these facts, the tree-ring measurements make it possible to indicate the geographical region within which the trees that produced the timber used in various constructions originated. Dendrochronologists in Copenhagen (Denmark), Lund (Sweden), Hamburg (Germany), Warsaw (Poland), and Sheffield (United Kingdom) co-operate very closely and have access to all the chronologies produced in their laboratories. Furthermore, as all the master-chronologies for

Sample from painting (PL) Mid 15th cent. Sample from shipwreck (DK) End 14th cent.

5,92

Sample from door (GB) End 14th cent.

2,32

Sample from shipwreck (DK) End 14th cent.

6,71

Figure 1. Values fo r cross-dating (t-values) o f objects from northern Europe, t-values are a measure o f similarity between two tree-ring series which show how well the series cross-date. Values equal to or greater than 3.5 are regarded as significant indicators o f a likely match. For t-values, see Baillie and Pilcher 1973.

NL

GB

oak (Quercus sp.) in our region have, through the 1980s, been expanded to cover the last 15 centuries, the possibility of doing successful research on items connected to trade and communication in northern Europe has increased consider­ ably. The original premise for dendrochronological dating o f medieval oak artefacts was that the material was derived from trees o f more or less ‘local’ origin (Eckstein and Bauch 1974; Fletcher 1977). Subsequently it was realized that sequences for medieval objects from across Europe were derived from one apparently similar source which was not local to the findsites of most o f the material (Baillie 1984). As research has continued and the chronology network has widened it has become apparent that there is a new branch o f tree-ring studies developing which may be dubbed

dendroprovenancing! This new aspect may be illustrated by examining groups of objects such as barrels, ships, paintings, furniture, altars, chests, wainscots and room panelling recovered from archaeological sites, shipwrecks, houses and art-historical collections from all over Europe. Take for exam ple, a painting from Gdansk (Poland), a shipwreck from the sea around Denmark and a door excavated in London. The tree­ ring sequences from these objects cross-date (fig. 1). By contrast the sequences do not cross-date with the local chronologies from Denmark and England (fig. 2). This phenomenon is characteristic o f movable object types and is never seen, at least in western Europe, in baulks or large cross-section beams. It has until now been recognised in objects dating from the late 14th century to the mid 17th century and the timbers are most frequently found in plank form derived from exceptionally evengrown 200-400 year old oak trees. The first clues to this riddle were found in the extensive customs records in Poland, Denmark and on the western European seaboard (Bang 1922). These demonstrate that vast quantities o f oak planking were exported through a number o f ports in the eastern Baltic primarily through the agency o f the German Hanse and The Teutonic Order throughout the European Middle Ages and up to the middle o f the seventeenth century. Although the material was exported from the Baltic ports, much of it probably originated more than 500 km inland and was floated down the river systems to the ports where the ships loaded their cargo. The key scientific evidence, which confirms the docu­ mentary evidence, was obtained by the construction o f an oak tree-ring chronology from large beams from an area o f

S

DK

D

PL

panels/GB

1,17

0,54

1,49

5,96

7,05

Sample from painting (PL) Mid 15th cent.

1,70

Sample from shipwreck (DK) End 14th cent.

1,36

1,64

1,77

2,75

1,94

5,44

5,85

Sample from door (GB) End 14th cent

1,11

1,89

3,35

3,37

4,17

4,87

5,59

Figure 2. Values fo r cross-dating (t-values) o f objects with oak master-chronologies from northern Europe. Note that the distance from England to Denmark is c. 800 km (500 miles) and Denmark to Poland is c. 600 km (350 miles) whilst England to Poland is c. 1400 km (900 miles).

Dendrochronological Evidence of the Timber Trade in Northern Europe

203

SHIPPED F R O M

1560

—

1565

G D AN SK

1575

1570

lli

1580

k o n i c sSBERG

1585

m

1590

n

1595

COURILAND

1600

d

1605

1

O T H ER S

Figure 3. Diagram compiled from the tables in the Books o f the Sound Dues. The graph shows the relative quantity o f wainscots shipped from various places in the Baltic area from 1562 to 1605 (figures fo r 1570 to 1573 are missing). In 1562 Gdansk, Koningsberg and Courland were responsible fo r about 90% o f the total export o f wainscots with Gdansk as by fa r the most important with up to 80%. Over the years the importance o f Gdansk declined up to 1605 where the three places still represented about 90% o f the total export but now with one third each. This mean that if dendrochronological research is to be carried out on material deriving from the Baltic region around 1560 the most obvious place to start will be to lookfor chronologies covering the area(s) that delivered timber fo r export through the harbour in Gdansk. With materialfrom about 1600 the picture is more complicated, chronologies from many various geographical regions will have to be taken into account.

Polish eastern Pomerania some distance away from the Vistula River, thus avoiding the possibility of contamination by timbers that were part of the trade (Wazny 1990). This sequence can be used not only to date the types of objects mentioned above, but also to indicate that the timbers were derived from somewhere in or around this area (Klein 1986; Bonde 1992). Recent re-analysis of some elements of this group has indicated the presence o f multiple sources within the group, and thus the working model for the analysis of this type of material has changed into a third form. The initial assumption o f exclu sively local material was reinterpreted as an hypothesised single foreign source (Baillie et al. 1985; Eckstein eta l. 1986). This hypothesis has now been further refined in the light o f documentary evidence (fig. 3), internal tree-ring cross-matching and signature sequences, into one o f multiple export areas within a larger geographical zone (Wa ny and Eckstein 1987; Wa ny 1992). Further refinement will require an extensive programme of creating local oak ch ron ologies from a wide range o f ecological zones throughout Europe. This w ill maximise our ability to differentiate sources with a high degree of resolution, and

aid future dendroclimatological work which will also require the same chronology network. The work will have to begin by dismantling the large master chronologies built for dating purposes (Hillam and Tyers 1995). These should then be rebuilt using modern zoning/clustering criteria. This requirement for a higher resolution chronology network shows that European oak dating studies are reaching a new level o f maturity. This pattern will be repeated as work on other species moves from dating to other areas o f research. The ability o f dendrochronological methods to identify groups of material from large scale transportation is o f historical and cultural interest, but also offers a useful lesson for other workers in other parts o f the world. The European experience shows that an appreciation o f the history o f woodland exploitation and trade in any research area is required before localised tree-ring sequences from non-living material can be successfully and reliably constructed. The successful dating o f large groups o f later-medieval, postm edieval and recent timber structures w orld-wide w ill require not only a greater understanding o f local and larger trade patterns, but also the creation o f more finely resolved chronology networks.

204

Niels Bonde, Ian Tyers and Tomasz Wazny

Acknowledgements This research was in part supported by the EC Environmental Research Programme (contract: EV5V-CT94-0500. Tree­ ring evidence o f climate change in Northern Eurasia during the last 2000 years). Translation: David Earle Robinson

References Baillie, M.G.L. and Pilcher, J.R. 1973. A simple crossdating program for tree-ring research, Tree-Ring Bulletin, 33: 7-14. Baillie, M.G.L. 1984. Some Thoughts on Art-historical Dendro­ chronology, Journal o f Archaeological Science, 11: 371-393. Baillie, M.G.L., Hillam, J., Briffa, K.R. and Brown, D.M. 1985. Re-dating the English art-historical tree-ring chronologies, Nature, 315: 317-319. Bang, N.E. 1922. Tabeller over skibsfart og varetransport gennem 0resund 1497-1660 (Tables de la Navigation et du Transport des Marchandises passant par le Sund 1497-1660). Anden del. Kpbenhavn. Bonde, N. 1992. Dendrochronology and timber trade in northern Europe from the 15th to 17th century. In Tree Rings and Environment. Proceedings o f the International Dendrochronological Symposium, Ystad, South Sweden, 3-9 September 1990. (ed. Th. Bartholin, B. Berglund, Dieter Eckstein and F. Schweingruber) Lundqua Report, 34: 53-55.

Eckstein, D. and Bauch, J. 1974: Dendrochronologie und Kunstgeschichte - dargestellt an Gemalden hollandischer und altdeutscher Malerei, Mitt. Dtsch. Dendrol. Ges., 67: 234243. Eckstein, D., Wa ny, T., Bauch, J. and Klein, P. 1986. New evidence for the dendrochronological dating of Netherlandish paintings. Nature, 320: 465-466. Fletcher, J.M. 1977: Tree-ring Chronologies for the 6th to 16th Centuries for Oaks of Southern and Eastern England, Journal o f Archaeological Science, 4: 335-352. Hillam, J. and Tyers, I. 1995. Reliability and repeatability in dendrochronological analysis: tests using the Fletcher archive of panel painting data. Archaeometry 37: 395-405. Klein, P. 1986: Age determination based on dendrochronology, PACT, 13: 225-237. Wazny, T. and Eckstein, D. 1987: Der Holzhandel von Danzig/ Gdansk - Geschichte, Umfang und Reichweite, Holz als Rohund Werkstof, 45: 509-513. Wazny, T. 1990. Aufbau und Anwendung der Dendrochronologie fur Eichenholz in Polen. Dissertation, University of Hamburg. Wazny, T. 1992. Historical timber trade and its implications on dendrochronological dating. In Tree Rings and Environment. Proceedings of the International Dendrochronological Sym­ posium, Ystad, South Sweden, 3-9 September 1990. (Ed. by Th. Bartholin, B. Berglund, Dieter Eckstein and F. Schweing­ ruber), Lundqua Report, 34: 331-333.

The Dating and Provenancing of Imported Conifer Timbers in England: the Initiation of a Research Project Cathy Groves Archaeology and Archaeological Sciences Research School, Department o f Archaeology and Prehistory, Sheffield University, West Court, 2 Mappin Street, Sheffield SI 4DT

Abstract Dendrochronology in the British Isles has almost exclusively concerned itself with the analysis o f oak timbers, although there have been recent successes with other species, such as ash, elm, beech, and native sub-fossil pines. Conifers are frequently found in medieval and post-medieval contexts, but these have been virtually ignored, mainly because they have been widely imported throughout the last millennium. Close co-operation between laboratories throughout northern Europe has made it possible to identify and provenance imported oak timbers. Similar techniques are now being evaluated for their possible application to imported conifer timbers found in buildings and recovered from archaeological excavations.

Introduction Dendrochronology is well established as one o f the major scientific dating methods in archaeology (e.g. Baillie 1982). It is capable of providing dates accurate to the calendar year for wooden objects and is thus the most precise o f all dating techniques. Tree-ring dates are also completely independent o f other historical or archaeological dating evidence such as stratigraphical or typological considerations. The vast majority of structural timbers excavated over the last 25 years in the British Isles have been oak (Quercus spp). Consequently our broad chronological coverage is confined to this genus, and the technique has been mainly restricted to the provision o f dates for oak timbers. Oak has proven exceptionally good for the development o f dendrochronological analysis in Britain, not only because of sample availability, but also due to its longevity and the reliability and clarity o f its ring sequences which make it eminently suitable for the construction of long reference chronologies. Oak reference chronologies in the British Isles cover the last 7000 years, although the geographical spread and strength o f replication vary through time with many more chrono­ logies available for the historic period than the prehistoric. This in theory allows oak timbers from buildings and archaeological features to be precisely dated, though in reality successful analysis requires the availability o f reference data for the relevant region and period. Over the past few years interest in the possibility o f working with other species has steadily increased as various excavated features have yielded large quantities o f structural timber o f types other than oak, such as ash (Fraxinus excelsior L.), beech (Fagus silvatica L.), elm (Ulmus spp), and various conifers. Chronologies have been produced for ash from the neolithic Sweet Track (Hillam et a l 1990), beech from medieval London (Tyers pers comm), and elm from post-medieval Droitwich (Groves & Hillam forth­ coming). However, the relative scarcity o f non-oak timbers makes the production of long composite reference chrono­ logies extremely difficult. Instead, precise dates for these sequences have been obtained from comparisons with

contemporary oak chronologies, often from the same archaeological site. Studies o f living trees suggest that inter­ species comparison is reliable, at least for some species (e.g. Groves & Hillam 1988). Similar success has been achieved with prehistoric bog pines (Pirns sylvestris L.) which have been dated by comparison with bog oak chronologies in Ireland (Brown 1991; Pilcher et a l 1995)). Work on bog pines has also been undertaken in England, although as yet neither White Moss in Cheshire (Lagaerd 1992; Lagaerd et al. 1992) nor Thorne Moor in Yorkshire (Boswijk pers comm) have produced absolute dates. There has also been a rise in dendrochronological interest in the use o f imported timbers at least partly fuelled by the development and exchange o f the large network o f oak chronologies covering northern Europe that has occurred over the last decade. These data have allowed quantities o f imported oak timbers to be dated and have had the added bonus o f identifying the geographical region from which the timbers were derived (e.g. Baillie et a l 1985; Bonde and Crumlin-Pedersen 1990; Tyers 1994). The rate o f growth o f a tree is primarily dependent on climate, thus the variation o f tree-ring widths through time reflects the variation in climatic factors, such as temperature and rainfall, for the region in which the tree grew. Climatic conditions in England are different from those in Scandinavia and consequently tree-ring patterns or chronologies obtained from these geographical areas are specific to that region. This new sub-field o f dendrochronology, described as ‘dendroprovenancing’ (Bonde and Jensen 1995), reveals information concerning the origin o f timbers and therefore provides information which enhances our understanding o f past economies with respect to the trading o f timber.

Importation There is little doubt from the documentary evidence that the import o f timber into England has occurred on a regular basis throughout this last millennium. During the eleventh to eighteenth centuries the regions exploited for their timber

206

Cathy Groves

resources covered a large part of northern Europe, though the key supply area and types of timber changed through the centuries. By the early to mid nineteenth century, North America had become the key supplier o f timber, apart from specialist timber still required from the Baltic region (e.g. Dollinger 1970; Fedorowicz 1980). Imported oak has already been identified dendrochronologically (see above) but it was only one o f a number of timber types imported. Large quantities of conifers such as Scots pine (Pinus sylvestris L.) and common or Norway spruce (Picea abies Karsten) were also imported. Conifer timbers have previously been identified in various buildings and on a number o f archaeological sites but these have been largely ignored by British dendrochronologists. This is due, at least in part, to species such as pine, spruce and European larch (Larix decidua Mill.) not being native to England. Pine was present in England up to the Bronze Age, but apart from some possible isolated relict forests, it was not growing within this country until reintroduced as a timber tree by early modern forestry (e.g. Clapham et al. 1989). Spruce and larch were introduced in the early to mid sixteenth century and early seventeenth century respectively and had been recognised as having commercial value by the end o f the seventeenth century, whereas North American timber trees such as Sitka spruce (Picea sitchensis Carr) and Douglas fir (Pseudotsuga menziesii Franco) were introduced during the nineteenth century (Evelyn 1729; James 1990; 164,184). It is therefore possible that some local conifer timber was available as early as the seventeenth century and may well have become more readily available following the increase in plantations during the latter part of the eighteenth century. However documentary evidence indicates that England remained heavily dependent on imported timber throughout the m edieval and post-medieval periods. Consequently, although the presence o f locally grown conifer timbers is not impossible, the majority of conifer timbers recovered from archaeological excavations or present in buildings are probably imported. Conifers are used for routine dating purposes elsewhere in Europe (e.g. Bartholin 1993; Thun 1993). Norway and Sweden, for instance, lie at the periphery of the northern limits o f the natural distribution of oaks (Schweingruber 1993: 195), and therefore dendrochronologists have con­ centrated their efforts on the species o f conifers that were readily available as a construction material. Following the successful dating of imported oak timbers by comparison with reference data from other European countries, the logical progression is to apply the same techniques to assemblages of conifers which are likely to have been imported.

timber trading during the medieval and post-m edieval periods. In addition the project may reveal information concerning the production and utilisation o f timber from non-native species grown in England thereby enhancing our understanding o f the history o f forestry. One o f the immediate objectives was to determine whether it was possible to produce replicated chronologies from individual sites or whether substantial mixing had occurred at the point o f export or import. This could severely hamper the successful production o f chronologies if the timbers present in a single structure were from multiple diverse sources. The analysis o f Tilbury Fort, E ssex (described below), was therefore regarded as a pilot study to determine the viability o f the project. Once the ability to produce site chronologies was established it was necessary to set up a database o f existing conifer reference chronologies from the historic period. Included were data from species known to have been found in England and from countries known from documentary sources to have exported timber to England. The database will be extended using data from the analysis of sites in England. The following criteria are used to influence the selection o f sites, although these may be relaxed as the project progresses: 1. the availability of sufficient numbers o f timbers contain­ ing a minimum of 50 rings. 2. the availability of strong stratigraphic or documentary evidence indicating the approximate date. 3. the availability of documentary evidence relating to the likely provenance o f the timbers.

P relim in ary R esults

The Database In order to obtain relevant reference chronologies, colleagues in the dendrochronology laboratories in Copenhagen in Denmark, Hamburg in Germany, Lund in Sweden, and Trondheim in Norway were approached. Their co-operation enabled the database to be established. Data has also been acquired for American and European sites, including two modern chronologies from sites in Scotland, from the International Tree-Ring Data Bank, part o f the World Data Center, and contact has been made with dendrochronologists from the Baltic States o f eastern Europe where new chronologies are being developed. This component o f the project is therefore well advanced, although it is currently concentrated on Scandinavia and other countries around the Baltic sea, since evidence from the two largest sites so far included in the project suggests that the timber originated in this region (see below).

T h e R esearch P ro ject and its A im s The project was conceived following a visit by the author to archaeological excavations at Tilbury Fort, Essex, where most of the major structural elements were coniferous and would therefore have ordinarily been precluded from dendrochronological analysis in Britain. The primary aim is to extend the scope o f British dendrochronology so as to enable precise dating evidence to be obtained from a wider range o f timber sp ecies and hence provide a more comprehensive dating system for timber structures and artefacts. An important secondary aim is that the ‘dendroprovenancing’ of timber will enhance our knowledge o f

The Tree-ring Results Four sites, all incidently in southern England, have been selected for inclusion in the project. These are: Tilbury Fort, Essex; Millers House, London; Windsor Castle, Berkshire; and Wells Cathedral, Somerset. The analysis o f Tilbury Fort is basically complete, whilst that o f Millers House and Windsor Castle are in progress. The timbers o f W ells Cathedral have been assessed but have yet to be sampled. Further sites will be incorporated into the project as they emerge. The results so far are described site by site below.

The Dating and Provenancing o f Imported Conifer Timbers in England Tilbury Fort, Essex Excavations were undertaken at Tilbury Fort, Essex, prior to the commencement o f the restoration o f the outer defences o f the fort. The excavation was carried out during 1988-89 under the direction o f Peter Moore o f the Passmore Edwards Museum. Large quantities o f w aterlogged and semiwaterlogged wood associated with the fort and its defences were present on the tidal foreshore of the River Thames. Samples were taken from structures ranging in date from the mid seventeenth century, associated with the founding of the defences in circa AD 1670, to the late nineteenth century, by which time various modifications had been carried out. It was established that the majority of the major structural timbers were spruce and those suitable were submitted for full analysis. The criteria by which samples were selected as suitable were based on those usually applied to oak in Britain. A sample must have a clearly defined ring sequence and a minimum o f 50 growth rings, although the chances o f obtaining a date are enhanced with increasing numbers of rings. In addition it is significantly better to obtain multiple samples from a structure so that these can be incorporated into a composite site sequence which can then be used for dating purposes. A total o f 52 spruce samples were measured. Many of the structures produced only one or two suitable samples and attempts to crossmatch these were generally unsuccessful. However it was possible to produce a 158-year chronology incorporating data from thirteen timbers (fig. 1). This low percentage o f crossmatched timbers is probably due to the low numbers o f samples available from some structures combined with the problem o f the structures varying in date from the mid seventeenth century to the late nineteenth century. A ll but three o f the crossmatched timbers were from structure 1800 which is the foundation piling for the fifth bastion of the fort and consists of piling within a horizontal framework (fig.. 2). Two of the remaining three timbers were from structures closely associated with structure 1800, whilst one was unstratified. It is structure 1800 in particular that is thought to contain imported timber, as large quantities o f timber were shipped from Norway for the construction of the fort (Moore pers comm). Other dating evidence indicates

- f ---------— y------- " " T " '— -— r IS O ®

207

that the assembly o f this bastion probably began in AD 1670 but was abandoned within approximately five years. If this is correct, the Tilbury spruce chronology should span the sixteenth and seventeenth centuries. There are conflicting views concerning whether it is possible to crossmatch spruce chronologies with pine chronologies but tests on the database implied that this was on occasion possible, particularly when both types were present on the same site. The Tilbury Fort spruce chronology was therefore compared with both spruce and pine reference data spanning the last millennium. Initial efforts were concentrated on chronologies available from Norway (fig. 3). Dr Thun o f University of Trondheim, Norway, suggests that if the spruce timber found at Tilbury is from Norway, it is most likely to have originated from south-eastern Norway and may have been shipped from one o f the harbours in Oslofjord. The only spruce chronology available from this region of Norway is that o f Ording (1941) which unfortu­ nately does not extend back beyond AD 1690. No conclusive results could be obtained from the comparisons with other Norwegian chronologies, either spruce or pine. The search was extended to include chronologies from elsewhere in the Scandinavian/Baltic region (fig. 3) but to no avail. The Tilbury Fort spruce timbers remain undated at present but will be compared with new reference chronologies as they become available. This pilot study demonstrated that it was possible to construct replicated site chronologies from conifer assem­ blages, although in this instance it has not been able to provide precise dating evidence for the erection o f the various foreshore structures or indicate the origin o f the timbers.

Millers House, London The excavations, undertaken by the Passmore Edwards Museum, were carried out in advance o f the development of the site as an interpretative centre for the renovated Millers House. A number o f phases concerning the reclamation o f land and its subsequent development were recorded. Phases III, IV, and VI contained timbers potentially suitable for dendrochronological analysis. Phase III represents the early development o f the artificial island in the River Lea, although it is possible that some of the timbers may be shuttering

"1....................'"'I— m ....... .""g.....................i

1--- — T”

19 5

is e a i 'i

18®® p i le s t 128 [1 0 9

1142 1129 1112 [1 1 1 113 1130

1166

lies

1868 3079

E H

§1© c o n te x t: “—

--------------—

........ ......... ....T '.......

107 'ii ------------- ----i 1 *"..... —

i—

*---------- 1—

51

*---------- 1--------------— r - ------------ r — --------- - i ------ — —

|----------- --------- j-------------------»...............

11

------------- r............. f ..— ......."i

100

Figure 1. Bar diagram showing the relative positions o f the spruce ring sequences included in the 158-year site chronology from Tilbury Fort, Essex.

208

Cathy Groves

Figure 2. Structure 1800, the timber foundations fo r the fifth bastion, is the largest o f the various postmedieval structures found on the foreshore at Tilbury Fort, Essex.

Figure 3. The areas from which some o f the pine and spruce reference chronologies are derived.

The Dating and Provenancing of Imported Conifer Timbers in England

209

—------------ -— ..............i -----------:------------1------------- ---------1---------------------- 1---------------------- 1---------------------- 1----------------------- 1---------------------- ?---------------------- 1-----------------------r * --------------------j-----------------------j------------- --------- 1

P h ase HI

1108®

~1 1 *1 ,3 2

............................. ..................

__________ 11897 ll® 1 5

--------------- —

ll®84..................

j-------------------------------1-----------------------------------------------«

—

....

...........

m i? m

.

I,

............."........' I

jl®©7 1809____________

i --------------------1------------------------- -------------------- ” T“

,,

----------------- 1--------------*

1---------------------- 1

-------------- '---------------- “ “ a

Figure 4: Bar diagram showing the relative positions o f the pine ring sequences included in the 112-year site chronology from Millers House, London.

Figure 5: The location o f the fo u r sites in England and the five Scandinavian reference chronologies with which the Millers House chronology crossdates.

from an earlier waterfront o f the island. Phase IV consists of the raising and extension of the island to the north. The timbers in this phase are tiebacks designed to prevent erosion. It is possible that some may have been re-used. The phase VI timbers represent the wooden foundations beneath the brick walls o f the Millers House which, according to other evidence, was completed in AD 1766 (Sable pers comm; Hewett 1980: 255). These consist o f planking, laid over crossbeams, each o f which is laid over two vertical piles, driven into the bottom o f the foundation trench. The completion of the analysis of this group o f about 50 conifer timbers had been originally planned to coincide with the production o f this paper. Only 18 conifer samples have so far been analysed as, due to problems with other aspects o f the site’s post-excavation analyses, the second batch of samples has only just arrived at the time of writing. The samples have currently only been identified to genus level and are pine. Sixteen samples have proved suitable for full analysis: three from phase III; one from phase IV; and twelve from phase VI. The rings sequences from seven o f the samples, all from

phase VI, crossmatch and have been combined to form a 112-year pine chronology (fig. 4). This was compared with Scandinavian pine chronologies covering the last millen­ nium. Consistent results were obtained with several Nor­ wegian and Swedish chronologies when the Millers House chronology spanned AD 1651-1762 (fig. 5; fig. 6). The quality o f the m atches appear to be acceptable when employing standard dendrochronological techniques (Baillie 1982) but, as this is the first set o f imported conifer timbers to be dated in this country, the reliability o f the matches may be reviewed in the light o f results from other sites. The ring sequences o f several of the dated timbers appear to extend to immediately below the bark surface, although the poor condition o f the timbers makes this very difficult to recognise. If this premise is correct, it suggests that the seven dated phase VI timbers were all felled circa AD 1762, which ties in closely with the documentary evidence. It is anticipated that the analysis o f the second batch of timbers will improve the replication and possibly extend the Millers House pine chronology. This should have the effect o f increasing the quality o f the crossmatches with the Scandinavian chronologies and proving that the initial results obtained from the first batch of samples are correct. The production o f dates also allows the geographical region of origin to be determined and, according to the success o f the next stage of analysis, may illuminate any difference in sources o f timber imported during phases III, IV, and VI.

Reference chronology

Students t value

Norway:

5.20

Sweden:

Hurdal AD 1645-1981 (Briffa et al 1986) Jondalen AD 1605-1981 (Briffa et al 1986) Gotland AD 1124-1987 (Bartholin 1994 pers comm) Gravsten AD 1469-1840 (Bartholin 1994 pers comm) Gotaland AD 1636-1855 (Bartholin 1994 pers comm)

4.00 4.47 3.00 4.70

Figure 6: The Students t values (Baillie and Pilcher 1973) obtained from the comparison o f the Millers House pine chronology with various Scandinavian pine reference chronologies. For oak at least, t values o f over 3.5 are considered significant if the visual match is acceptable.

210

Cathy Groves

Windsor Castle, Berkshire Following the fire at Windsor Castle in 1992, a large-scale tree-ring project was initiated. This project is described in detail elsewhere (Hillam & Groves forthcoming) but as the fire damaged area contains both oak and conifers it also forms part o f this research project. The conifer timbers are associated with the major post-medieval alterations carried out for George IV by Sir Jeffry Wyatville in the early nineteenth century. The timbers appear to be badly burnt but closer examination shows that only the outer surface had been carbonised. None o f these conifer structures are to be retained, so complete slices will be made available for analysis. As yet only four conifer samples have been sent for analysis: two from the Kitchen; one from the Grand Reception Room; and one from the Crimson Drawing Room. These have been identified to genus level and are pine. All four samples were suitable for analysis and two o f the ring sequences, one from the Kitchen and one from the Grand Reception Room, appear to crossmatch. No attempts have yet been made to crossdate the W indsor pines with Scandinavian, Baltic or North American pine chronologies. This will be done once a better replicated chronology has been produced. However the length of the ring sequences (96-150 rings) suggests that these may be an extremely valuable group o f pine timbers for chronology building and dating.

Wells Cathedral, Somerset An assessment visit to Wells Cathedral was undertaken to determine the dendrochronological potential of the nave and south transept roof timbers. The initial request concentrated solely on the dating o f oak timbers from the various phases o f construction, repair, and modification. However, during the assessment, it was noted that there were quite substantial quantities of conifer timbers present, mostly associated with nineteenth century interventions. These appeared suitable for analysis and will therefore be sampled when the extensive program of oak sampling is carried out later this year.

Conclusions The setting up of the database has already been largely achieved, though it is recognised that this will remain an on­ going part o f the project as additional chronologies become available from the regions of interest. It is particularly hoped that a new initiative in Scotland on softwood timbers in buildings will compliment the research currently underway in England (Crone and Mills pers comm). Site master chronologies have been successfully produced for the two major sites currently analysed. This result already indicates that the potential difficulties due to possible multi­ source assemblages are less o f a problem than originally anticipated. The dating of one o f these site master chrono­ logies with Scandinavian reference chronologies clearly demonstrates the presence of imported conifer timbers. The analysis o f timber imported from northern Europe will continue and expand but as the Wells Cathedral and Windsor Castle projects proceed the problems associated with the change over to North American supplies and the increased possibility o f the presence of locally grown conifers will also begin to be addressed. The continuation o f work on the above sites provides a solid basis for the project but it is

hoped that sites yet to be included in the project will range in date throughout the medieval and post-medieval periods so as to allow as broad an information base as possible. The dendrochronological analysis o f medieval and postmedieval conifer timbers in England is clearly in its infancy. Although it is obviously necessary to analyse much greater numbers o f timbers, the indications from the project are that in the relatively near future it may well be possible to provide precise dating evidence from imported conifer timbers. This will extend the scope o f dendrochronology in England and add to the information concerning the provenance o f timber, timber trade and forestry practices during the last millennium.

Acknowledgements The research is funded by English Heritage. I would like to thank in particular Dave Brown, Jennifer Hillam and Ian Tyers for valuable discussions; Helen Hibberd and Ian Tyers for carrying out the initial assessment on the Millers House timbers; Peter Moore, Ken McGowan and Ken Sable, all o f the Passmore Edwards Museum, Peter Bird of Caroe & Partners, Wells, and John Pidgeon for information about the sites. Dr Thomas Bartholin, University o f Lund, Dr Niels Bonde, National Museum o f Denmark, Dr Terje Thun, University o f Trondheim and Dr Sigrid Wrobel, University o f Hamburg, have all kindly provided reference chrono­ logies. Reference chronologies have also been obtained from the International Tree-Ring Data Bank based in Boulder, Colorado, and funded by the National Geophysical Data Center (part o f the World Data Center). The production o f various chronologies incorporated into the database have been supported by the EC Environmental Research Pro­ gramme (contract: EV5V-CT94-0500. Tree-ring evidence o f climate change in Northern Eurasia during the last 2000 years).

References Baillie, M.G.L. 1982. Tree-Ring Dating and Archaeology. London, Groom Helm. Baillie, M.G.L. and Pilcher, J.R. 1973. A simple crossdating program for tree-ring research, Tree Ring Bulletin. 33: 7-14. Baillie, M.G.L., Hillam, J., Briffa, K.R. and Brown, D.M. 1985. Re-dating the English art-historical tree-ring chronologies, Nature. 315: 317-319. Bartholin, T. 1993, Dendrochronology in building investigations in Sweden. In Dendrochronology and the Investigation o f Buildings (eds O. Storsletten and T. Thun), Riksantikvarens Rapporter. 22: 14-17. Bonde, N. and Jensen, J.S. 1995, The dating of a Hanseatic cogfind in Denmark. In Shipshape, Essays fo r Ole CrumlinPederson. Vikingeskibshallen i Roskilde, 103-22. Bonde, N. and Crumlin-Pedersen, O. 1990. The dating of Wreck 2, the Longship, from Skuldelev, Denmark, News WARP. 7: 3 6 Briffa, K. R., Wigley, T.M.L., Jones, P.D., Pilcher, J.R. and Hughes, M.K. 1986. The reconstruction of past circulation patterns over Europe using tree-ring data, final report to the Commission of European Communities, contract no CL.lll.UK(H). Brown, D.M. 1991. Studies on Pinus sylvestris L. from Garry Bog, County Antrim. Unpublished MSc thesis, Queen’s University of Belfast. Clapham, A.R. Tutin, T.G. and Moore, D.M. 1989. Flora o f the British Isles, 3rd edn. Cambridge, Cambridge University Press. Dollinger, P. 1970. The German Hans a. London: Macmillan Evelyn, J. 1729. Silva, 5th edn. London.

The Dating and Provenancing o f Imported Conifer Timbers in England Fedorowicz, J.K. 1980. England's Baltic trade in the early seventeenth century. Cambridge, Cambridge University Press Groves, C. and Hillam, J. 1988, The potential of non-oak species for tree-ring dating in Britain. In Science and Archaeology, Glasgow 1987 (eds E.A. Slater and J.O. Tate). Oxford, British Archaeological Reports, British Series, 196: 567-79. Groves, C. and Hillam, J. forthcoming, Tree-ring analysis and dating of timbers. In Multiperiod Saltmaking at Droitwich, Hereford and Worcester - excavations at Upwich 1983-4 (ed J D Hurst). London, CBA Research Report. Hewett, C.A. 1980. English Historic Carpentry. London, Phillimore. Hillam, J. and Groves, C. forthcoming, Tree-ring research at Windsor Castle. In Tree Rings, Environment and Humanity, the proceedings o f the 1994 International Tree-Ring Con­ ference, Tucson, Arizona. Radiocarbon. Hillam, J. Groves, C.M. Brown, D.M. Baillie, M.G.L. Coles, J.M. and Coles, B.J. 1990. Dendrochronology of the English Neolithic. Antiquity. 64: 211-20. James, N.D.G. 1990. A history of English Forestry, 2nd edn. Oxford, Blackwell. Lageard, J.G.A. 1992. Vegetational history and palaeoforest reconstruction at White Moss, South Cheshire, UK. Un­

211

published PhD thesis, Keele University. Lageard, J.G.A. Chambers, F.M. and Thomas, P.A. 1992. Palaeoforest reconstruction from peat exhumations at White Moss, South Cheshire, UK. In Tree Rings and Environment. LUNDQUA report. 34: 172-76. Ording, A. 1941. Arringanalyser pa gran og furu (Growth-ring analyses in spruce and pine), Meddelelser Norske Skogfors0ksvesen. 1 (25): 104-354. Pilcher, J.R., Baillie, M.G.L., Brown, D.M., McCormac, F.G., MacSweeney, P.B. and McLawrence, A.S. 1995. Dendro­ chronology of subfossil pine in the north of Ireland, Journal o f Ecology. 83: 665-671. Schweingruber, F.H. 1993. Trees and wood in dendrochronology. Berlin, Springer-Verlag. Thun, T. 1993, Dendrochronology and building construction. In Dendrochronology and the Investigation o f Buildings (eds O Storsletten and T Thun). Riksantikvarens Rapporter. 22: 4951. Tyers, I.G. 1994, Appendix 6. Dendrochronology of Roman and early medieval ships. In Marsden P, Ships o f the Port o f London; first to eleventh centuries AD. English Heritage Archaeological Report. 3: 201-209.

A Q ualitative Approach to Dental Microwear Analysis Ingrid L. Mainland Department o f Archaeology and Prehistory, University o f Sheffield, Northgate House, West Street, Sheffield S4 1ET

Introduction: Recent Dental Microwear Research and Methodologies Recent research has established that microscopic tooth wear patterns, dental microwear, is correlated with diet in several modern mammal species. There is known to be a relationship between broad dietary adaptations and dental microwear in primates (Teaford and Walker 1984; Teaford and Runestad 1992), hyaenids, large canids and felids (van Valkenburgh et al. 1990), V iverridae (Taylor and Hannam 1987), Chiroptera (Strait 1993), ruminating and-non-ruminating ungulates (Walker et a l 1978; Solounias and Moelleken 1992; Solounias and Hayek 1993). More specific, seasonal dietary variations have been associated with microwear patterns in hyraxes (W alker et al. 1978) and Cebus nigrivittatus (Teaford and Robinson 1989). It is maintained that these relationships can be used to identify diet in individuals from archaeological or palaeontological contexts and several such studies have been undertaken on ancient humans and various extinct mammals (e.g. Teaford and Walker 1984; Grine 1986; Solounias and Moelleken 1992; van Valkenburgh et a l 1990; M olleson et a l 1993). Most microwear studies have employed similar methodo­ logies. A scanning electron microscope (SEM) is used to examine the enamel and dentine surfaces of actual teeth or o f replicas. Micrographs are taken at magnifications between 100-650x and are examined for variability in microwear patterning using one o f two general analytical approaches: quantitative, based on metrical data, or qualitative, based on descriptive information. Diet-micro wear studies typically em ploy quantitative techniques. Microwear defects are recorded for frequency and dimension, thus reducing visual data to a numeric format and allowing statistical manipu­ lation. This approach can claim objectivity and can be standardised between different researchers. It is, therefore, considered more appropriate for identifying pattern varia­ bility relating to diet (Gordon 1988; Teaford 1991). At its simplest level, qualitative analysis attempts to distinguish betw een various diets through the visual identification o f distinctive microwear features or enamel surface characteristics. Walker et al. (1978) were, for example, able to contrast the heavily striated surfaces of grazing hyraxes with the polished surfaces o f browsing species. Qualitative methods have not been favoured by researchers interested in the identification o f diet-microwear relationships; they are criticised for being subjective and are thought to have a limited application because ‘it is impossible to mentally store and access more than a few images at a time’ (Gordon 1988:1139). It can be argued, however, that these com m ents refer m ainly to the sim ple types o f descriptive analysis outlined above. More sophisticated methods of examining visual data are found and can go

some way towards reducing these criticisms. The approach of Puech et al. (1983) to hominid microwear, in which patterns are recorded for the presence and absence o f 19 different descriptive categories, demonstrates how a large volume of descriptive data can be observed without recourse to memory. Although not attempted by Puech, such presence/ absence data can also be quantified (e.g. Ryan and Johanson 1989; Harmon and Rose 1988). The visual categorisation of microwear patterns through presence/absence data or descriptive categories thus permits the use o f statistical techniques in qualitative studies, eliminating some o f the potential subjectivity. It has been demonstrated that a qualitative approach o f this type can be used to examine fairly subtle differences in diet or microwear pattern, particularly when combined with basic microwear defect statistics like pit or striation frequency (Bullington 1991; Harmon and Rose 1988; Kelley 1990). A qualitative approach to diet-microwear relationships has arguably been undervalued and neglected because o f a desire for the objective accuracy o f quantitative methods. There is, however, considerable scope for inaccuracies to enter into quantitative microwear statistics: SEM conditions, surface morphology o f the specimen, casting defects and digitising error will all contribute towards a loss o f accuracy (Gordon 1988). Furthermore, a degree o f subjectivity is present in quantitative analyses: decisions on where a feature is measured, feature definition (i.e. pits, small pits, etc.) or recording accuracy will each vary between researchers. It might, therefore, be suggested that analysts have been overconfident in assuming that quantitative studies represent a thoroughly objective approach to microwear. Quantitative techniques can also be criticised because they ignore a potentially valuable aspect o f microwear: the surface appearance and texture o f enamel surfaces. In several studies surface characteristics were found to be useful dietary discriminators: e.g., polished surfaces were characteristic of a browsing diet in hyraxes (Walker et al. 1978) and were related to a vegetal diet in gorillas (Ryan and Johanson 1989). Moreover, feature densities on their own can be less discrim inating than when com bined w ith descriptive information concerning their location or edge morphology (Harmon and Rose 1988; Kelley 1990). These variations are not detected in quantitative studies and theoretically, therefore, two microwear patterns with vastly different surface characteristics could appear similar in a metrical analysis. A qualitative approach has been developed as part o f a larger study into the relationship between diet and dental microwear in domestic sheep and goats (Mainland 1994). In the following article, this approach and its application is discussed with reference to the identification o f dietary trends in the dental microwear o f domestic sheep and goats.

214

Ingrid L. Mainland

D en tal M icrow ear in D om estic Sheep and G oats The overall aim was to establish whether dental microwear could be used to identify the diet o f domestic sheep/goats in the archaeological past; to this end, it was necessary to examine microwear patterns in modem populations with known diets. Research focused on three aspects of sheep/ goat microwear: (i) how do microwear patterns vary across the occlusal surface o f sheep/goat teeth, (ii) does the microwear o f fodder-fed sheep/goats differ from that o f grazing sheep/goats, (iii) do grazing microwear patterns vary with season or pasture type? These questions were explored using a combination of qualitative and quantitative analytical techniques; a qualitative approach was used to identify microwear patterning pertaining to occlusal location while diet-related micro wear patterning was identified qualitatively and quantitatively. This article considers only the qualitative analysis of diet-related microwear patterns in fodder-fed and grazing sheep and goats. For a discussion o f occlusal variation and quantitative diet-microwear relationships in sheep and goats, see Mainland (1994).

Sampling Methods: magnification level used, tooth and area o f enamel examined Araldite replicas (Araldite MY 753, hardener HY 956, CibaGeiby) were made o f the lower deciduous fourth premolar (dP4) and were examined for microwear using a scanning electron microscope (SEM). A single micrograph was taken at a magnification o f 640x from a specific area o f enamel on the anterior facing buccal enamel band o f the posterior cusp. The choice o f tooth was made largely on practical grounds: the dP4 was erupted and in wear in all individuals examined. The buccal enamel band was selected because a survey of intratooth variation in microwear pattern had demonstrated this area o f the tooth to be particularly suitable for the identification o f diet-microwear relationships in domestic ovicaprids (Mainland 1994). One micrograph was excluded from the analysis because o f a poor quality image. The total sample thus consists o f 50 individuals (leafy hay n=16, summer rough n=6, winter rough n=8, summer sown n=7, winter sown n=8, grassy hay n=5).

Recording Qualitative M icrowear Patterns M ethod ology: a Q u alitative A pp roach to D en tal M icrow ear A n alysis

The samples Grazing and grassy hay-fed sheep were provided by the Macaulay Land U se Research Institute, henceforth referred to as MLURI. Mandibles from leafy hay fed sheep and goats were collected from the village ofPlikati in the Epirus region of Greece. In all cases diet was known for at least three months prior to examination.

The grazing sample Twenty-nine sheep were grazed on two pastures in the Scottish borders: (1) a rough, indigenous grassland domin­ ated by Nardus stricta\ (2) a sem i-indigenous pasture reseeded with rye-grass (Lolium sp.). To allow the identi­ fication of seasonal trends in microwear pattern, thirteen sheep were examined in summer ( ‘rough pasture’ = 6, ‘sown pasture’ = 7) and sixteen in winter ( ‘rough pasture’ = 8, ‘sown pasture’ = 8). The grazing population can, therefore, be divided into four groups; summer rough pasture, summer sown pasture, winter rough pasture and winter sown pasture.

The grassy hay-fed sample Five wethers were kept at the MLURI animal house in Edinburgh for a period of three months. During this time they were fed an unspecified type of grassy hay, though rye­ grass is the most probable (I. Wright, pers. comm.).

Twenty categories were created to describe the type o f microwear present on the enamel surface o f sheep and goat teeth. These range from the identification o f feature type to an impression o f surface texture and can be broadly divided into four groups: (i)

presence/absence o f feature types: pits, parallel striations, non-parallel striations, pit lines (ii) description o f features: features with sharp/rounded edges, deep/shallow features, features with anteriorposterior orientation, features with bucco-lingual orientation, features with a definite shape, feature edges joining up (iii) texture/appearance of surface: surface porous, surface smooth/abraded, surface polished, surface flat/uneven, areas o f ‘empty’ enamel, underlying enamel structure partially visible, underlying enamel structure fully visible, furrowed surface (iv) presence/absence o f surface damage: cracked enamel surface, flaking enamel surface.

A photographic key was established to aid identification and reduce subjectivity (Mainland 1994; Plates 7.1-7.20). This is complemented by a full verbal description for each category (Mainland 1994: 121-124). A micrograph was recorded for the presence (1) or absence (0) of each of the categories listed above. The dietary provenance of each specimen was unknown during examina­ tion in the SEM and while recording individual micrographs.

The leafy hay-fed sample

Identifying D iet-Related M icrowear Patterning

Seventeen sheep (n=9) and goat (n=8) mandibles were collected from Plikati during April 1992. These animals had been fed a mixture of grassy hay and leafy hay supplemented with bran and grain during the previous winter (Halstead, pers. comm.). The grassy hay was derived either from locally available sown pasture (alfalfa) or naturally occurring meadow grassland. The leafy hay consisted mainly of Turkey oak (Quercus cerris). For simplicity the Plikati sample is henceforth referred to as leafy hay though it is recognised that the animals also consumed some grassy hay.

The relationship between the qualitative categories and diet was investigated statistically. The underlying hypothesis was that microwear patterns from sheep consuming similar diets would be similar in terms o f surface characteristics. The qualitative data set consists o f twenty variables per micro­ graph which list the presence or absence o f each o f the twenty categories described above. These data were initially summarised using simple descriptive statistics to give a general impression of the distribution o f the categories within the various dietary groups (not discussed here, see Mainland

A Qualitative Approach to Dental Microwear Analysis 1994). Multivariate techniques were then employed to identify more specific relationships. Although microwear studies have generally favoured univariate statistical techniques (e.g. Teaford and Runestad 1992; van Valkenburgh e ta i 1990), a multivariate approach is more applicable for the analysis o f microwear data where each case (i.e. micrograph) is usually represented by several variables, as differences between cases or groups of cases can be evaluated using all variables simultaneously and potentially useful correlations between variables are not ignored. Several recent quantitative studies have used multivariate techniques to identify dietary trends with some success (Lalueza et a l 1993; Solounias and Hayek 1993; Molleson et a l 1993) but such techniques have not previously been applied to qualitative data. Three multivariate statistical techniques were used; correspondence analysis (CA), canonical correspondence analysis (CCA) and discriminant analysis (DA). CA, CCA and DA facilitate the analysis of multivariate distributions by compressing “the information contained in a large number o f variables into a much smaller number of new variables, ideally only two or three” (Shennan 1988: 2 4 1 -2 4 2 ). These new variables are then used to create ordination diagrams in which patterning and trends within the data can be deter­ mined visually (Greenacre 1984; Shennan 1988). In terms o f microwear data, these techniques can be used to identify which teeth exhibit comparable microwear patterns, rep­ resented by proximity in a scatterplot, and which variables contribute most to this similarity or difference, indicated by their position in the plots (CA, CCA) and their contribution to the axes (DA). Atypical individuals can also be distin­ guished since they will occupy outlying positions. It is, therefore, possible to consider both within group and between group variation in microwear pattern as well as the interrelation between various microwear variables. The multivariate analysis of microwear patterns in the MLURI and Piikati sheep and goats was complicated by sample size. The number of cases in each dietary group is smaller than the number o f qualitative variables and any observed differences between diets could, therefore, be due to chance alone (N. Fieller pers. comm.; Perrins et a l 1992). This is a problem for DA and CCA where differences between groups are actively sought, but not for CA where there is no prior knowledge of dietary groups. To minimise the impact o f sample size, two strategies were adopted. First the number o f variables used was reduced by employing a Stepwise DA which selects the minimum number of variables required for a successful discrimination. This procedure has the additional advantage o f elim inating any variables which do not contribute to between group differences and which could potentially diminish the discrimination. Secondly, the results o f CCA and D A were compared with that given by CA; since CA is unaffected by sample size, there can be some confidence in accepting any differences identified between dietary groups using D A and CCA if they are found also in CA.

A nalytical Procedure DA and CCA A series of Stepwise D A was undertaken to determine how successfully the qualitative categories could distinguish between micrographs grouped according to the various pasture and fodder types (fig. 1).

215

The dietary groups were chosen to allow investigation of the following; (i)

the difference between individual dietary groups: all dietary groups (analysis a); grass-based diets only (analysis b) (ii) the difference between fodder and grazing: grazing versus leafy hay (analysis c); grassy hay versus grazing (analysis e) (iii) the difference between fodder types: grassy hay versus leafy hay (analysis d) (iv) the effects o f pasture type and season on grazing microwear: rough pasture versus sown pasture (analysis f); summer grazing versus winter grazing (analysis g); all grazing diets (analysis h). The variables selected by the Stepw ise D A were next analysed using CCA. This allowed the statistical significance of between group differences to be assessed using the Monte Carlo permutation test (p 0.05) (fig.l).

Summary The multivariate analyses o f qualitative microwear patterns in the dP„4 have identified differences between the microwear

o f grazing and fodder-fed sheep and goats (analyses c and d), between that o f sheep/goats fed different types o f fodder (analysis e) and between that o f sheep grazing different pastures in different seasons (analyses a, b, f, g and h). The Monte Carlo permutation test indicates that these differences are only statistically significant for four analyses: analysis a (identifying differences between the six individual dietary groups), analysis b (identifying differences between the grassy hay-fed and various categories o f grazing sheep), analysis d (identifying differences between grassy hay and grazing) and analysis e (identifying differences between leafy hay and grassy hay). Although significant, however, dietary groups are clearly separated from each other in only one of these analyses: analysis e, leafy hay v grassy hay. Intergroup overlap is evident in each o f analyses a, b and d. In the remaining analyses (c, f , g and h) comparable or better discriminations are given by randomly generated data and the separation o f leafy hay and grazing (analysis c), rough and sown pasture (analysis f), summer and winter grazing (analysis g) and the four individual grazing groups (analysis h) cannot, therefore, be accepted as reflecting actual differences in microwear pattern.

CA Four separate CA were undertaken using the same dietary groupings and variables as in analyses a, b, d and e to evaluate further the validity o f the trends identified as statistically significant by the Monte Carlo permutation test. The arrangement o f micrographs in each CA is similar to that evident in the DA and CCA using the same data. It seems probable, therefore, that the multivariate analysis o f microwear in the dP.4 has identified real differences between the m icrowear o f certain dietary groups and is not merely

A Qualitative Approach to Dental Microwear Analysis

217

Figure 2. Stepwise discriminant analysis on dP4 qualitative data. Analysis a: distinguishing all six dietary groups, (summer rough (+) and SROU; summer sown ( A ) and SSOW; winter rough (x) and WROU; winter sown (A ) and WSOW; grass hay (O ) and HAY; leafy hay (A) and LEAFY) (see fig. 1 fo r categories used) (first discriminant function plotted on horizontal axis, second discriminant function on vertical axis; • = centroid; a cluster o f points represents cases with identical co­ ordinates).

Figure 3. Stepwise discriminant analysis on dP4qualitative data. Analysis b: distinguishing the grazing and grass-hay fe d sheep, (summer rough (+) and SROU; summer sown ( A ) and SSOW; winter rough (x) and WROU; winter sown ( A ) and WSOW; grass hay (O ) and HAY) (see fig. 1 fo r categories used) (first discriminant function plotted on horizontal axis, second discriminant function on vertical axis; • = centroid).

Ingrid L. Mainland

218

. -

. 3

« s -

2

-

1

|

L

,, l l |

'S

0

1

A

A

2

3

FUNCTION 1

Figure 4. Stepwise discriminant analysis on dP4 qualitative data. Analysis (c): discriminating between grazing f # ) and leafy hay (A) (see fig. 1 fo r categories used) (large symbols indicate position o f centroids).

FUNCTION 1

Figure 5. Stepwise discriminant analysis on dP4 qualitative data. Analysis (d): discriminating between grazing ( # ) and grassy hay (O ) (see fig. 1 fo r categories used) (large symbols indicate position o f centroids).

Figure 6. Stepwise discriminant analysis on dP4 qualitative data. Analysis (e): discriminating between leafy hay (A) and grassy hay (O ) (see fig. 1 fo r categories used) (large symbols indicate position o f centroids).

A Qualitative Approach to Dental Microwear Analysis

219

Figure 7. Correspondence analysis on dP4 qualitative data. Analysis (e): identifying differences between leafy hay (A) and grassy hay (O). Figure illustrates the position o f the micrographs within the first (horizontal) and second (vertical) ordination axes (see fig. 1 fo r categories used, a cluster o f points represents cases with identical co-ordinates).

Figure 8. Canonical correspondence analysis on dP4 qualitative data. Exploring the relationship between the dietary groups leafy hay (A) and grassy hay (O ) and the qualitative categories flaking present (FKY) or absent (FKN), surface fla t (FLY) or uneven (FLN), porosity present (PRY) or absent (PRN), surface smooth (SMY) or abraded (SMN), polish present (POY) or absent (PON), furrowed prism relief present (UFY) or absent (UFN), features deep (DPY) or shallow (DPN), features with an anterior-posterior orientation present (APY) or absent (APN). Figure illustrates the position o f the micrographs and qualitative variables within the first (horizontal) and second (vertical) ordination axes (a cluster o f points represents cases with identical co-ordinates).

220

Ingrid L. Mainland

reflecting random patterning. The trends identified can be summarised as follows; 1. leafy hay separates from grassy hay (CA, CCA and DA analysis a and d) ( figs 6 and 7) 2. grassy hay can be partially separated from grazing (CA, CCA, DA analysis e) (fig. 5) (Mainland 1994; figs 11.4 and 11.11) 3. grassy hay separates from summer rough/summer sown/ winter sown (CA, CCA, D A analysis a and b) (fig. 3) (Mainland 1994; figs 11.2, 11.9, 11.10, 11.13) In addition, there is some indication that the microwear of winter rough differs from that of the other grazing groups and leafy hay (analysis a CA, DA, analysis b CA, CCA) and that it may actually be more comparable to that o f grassy hay-fed sheep (CA, analysis a and b). This is, however, difficult to assess conclusively because o f the rather variable behaviour of winter rough within the different analyses.

The relationship between the dietary groups and the qualitative categories 1. Leafy hay from grassy hay Grassy hay is characterised by ‘porous’ and polished surfaces in which the underlying enamel structure is visible, leafy hay by the absence o f porosity and by deep features (fig. 8). 2. Grassy hay from grazing The grassy hay-fed sheep are characterised by porous enamel surfaces while grazers exhibit surfaces which are abraded, have areas devoid o f features and in which features are deep and do not have an anterior-posterior orientation (Mainland 1994; fig. 11.4c). 3. Grassy hay from summer rough/summer sown/winter sown The separation of grassy hay-fed sheep from those grazing on the rough pasture in summer and the sown pasture in winter and summer can again be attributed to the presence of ‘porous’ surfaces in the grassy hay-fed sample (Mainland 1994; fig. 11.9c).

Discussion A relationship can be identified between diet and the qualitative microwear patterns evident in the dP4. The greatest contrast is found between individuals fed different types o f fodder and betw een fodder-fed and grazing individuals: the micro wear o f sheep fed grassy hay is clearly distinct from that o f sheep and goats fed leafy hay while grassy hay-fed sheep can be partially separated from grazing sheep. Micro wear patterns are little affected by pasture type or grazing season; no significant differences are found between the microwear of summer and winter grazing sheep or between those grazing on the rough and the sown pasture. The microwear o f sheep grazing the rough pasture in winter is, however, somewhat different from the other grazers and may indeed be more comparable to grassy hay. This may explain the partial nature o f the separation o f grassy hay from grazers when the grazers are considered as one group. It is difficult to explain why specific microwear patterns are associated with certain diets. Microwear formation is a poorly understood aspect of micro wear research and although several factors are known to contribute, including patterns of jaw movement, food type (abrasivity, acidity and texture)

and tooth morphology, it is almost impossible to isolate particular causes. In the grazing and fodder-fed sheep and goats examined, the most likely explanation for variation can be found in food acidity and plant phytolith content. The porous surfaces and prism relief (i.e. the exposure o f the underlying enamel structure) evident in grassy hay-fed sheep may indicate the effects o f acidity: these surfaces closely resemble enamel which has been soaked in acidic solutions (c.f. Mainland 1994; plate 7.12 and Meurman and Frank 1991; figure 4) while prism relief has been associated with an acidic diet in various species (van Valkenburgh et al. 1990; Ungar 1994). It can also be anticipated that phytolith levels and hence abrasive content will be higher in grazing/grassy hay than in leafy hay: the fodder used at Plikati consisted primarily o f low-phytolith accumulating dicotyledon species (Quercus serris!Alfalfa sativa/Compositae) whereas grasses are high-phytolith accumulating species (Piperno 1988). It is unclear, however, whether differences in abrasive content or, indeed, acidity would bring about the microwear observed since little data exists for the formation processes behind qualitative microwear patterns. Further­ more, the various diets examined have not been assessed for abrasive content or acidity. U ntil such questions are examined directly any explanations must remain speculative.

Conclusions The analysis o f qualitative microwear patterns in domestic sheep and goats demonstrates the potential o f a qualitative approach for the investigation o f diet-microwear relation­ ships. The techniques developed for ovicaprid microwear distinguish diets where differences in abrasive content, one o f the principal factors implicated in microwear formation, are thought to be large (i.e. leafy hay and grassy hay), as well as those where differences are more subtle (i.e. grassy hay and grazing). Furthermore, this approach separates diets indistinguishable using quantitative techniques and in some cases discriminates more clearly between diets which both techniques can distinguish: a quantitative analysis using the same micrographs from the dP4 was unable to identify any differences between grassy hay and grazing sheep while the separation o f leafy hay and grassy hay is less pronounced as in the qualitative analysis (Mainland 1994). For some dietary distinctions, however, quantitative techniques are more successful, such as distinguishing sheep grazing the rough pasture from those on the sown pasture or leafy hay-fed sheep/goats from grazing sheep. It is apparent, therefore, that a qualitative approach ought not to be seen as an alternative but as a complement to quantitative microwear studies: qualitative categories are essential for distinguishing diets where microwear differs only in terms of microwear feature morphology and/or enamel surface characteristics (e.g. grassy hay and grazing), qualitative categories for microwear which differs only in the frequency and/or dimensions o f micro wear features (e.g. leafy hay and grazing). Despite these optimistic conclusions, it can be argued that further research and analytical development is necessary before any dietary reconstruction is attempted. First, a rather limited range o f diets has been examined. Thus, although the microwear o f leafy hay-fed sheep and goats can be distin­ guished from grassy hay fed sheep, it is unclear whether grassy hay would also differ from animals consuming fresh browse or, indeed, how the microwear o f fresh browse would

A Qualitative Approach to Dental Microwear Analysis compare with that from leafy-hay or grazing. Furthermore, it is not known whether the microwear o f the leafy hay and grassy hay-fed sheep/goats w ill be representative of these two fodders in general. The sheep/goats from Plikati were fed a mixture o f leafy hay, legume hay and meadow hay and their microwear may well differ from that o f animals fed only the legume hay or meadow hay or leafy hay. Likewise, rye­ grass fed to the MLURI sheep may create different microwear than hay made from meadow grasses or legumes. Secondly, although data from primates and wild ruminants does suggest that the turn-over-rate of dental microwear is rapid, in some cases a matter o f hours, and consequently that microwear is likely only to reflect the very recent diet of an individual (Walker and Teaford 1989), an insight into the cumulative effects o f microwear in domestic ovicaprids is needed as this could potentially complicate dietary reconstruction. Finally, som e m ethodological developm ent is necessary. The quantitative and qualitative analytical approaches have yielded different, but compatible and complementary results and research must now be undertaken on the potential o f an approach which combines the most productive elements of both techniques. Such a combined approach underlines the need to select only the most informative variables for examination, both to minimise recording time and to avoid the risk that a high ratio of variables to cases will yield spurious results at the analytical stage. For the qualitative data, it is anticipated that future developments will focus on categories which cannot be detected through a quantitative analysis such as polish, porosity, smooth/abraded, prism relief (underlying enamel structure fully/partially visible, furrowed enamel), deep/shallow features, each of which were associated with specific diets in domestic ovicaprids.

Acknowledgements This research would not have been possible without the assistance o f the Macaulay Land Use Research Institute nor the kind co-operation o f the villagers o f Plikati. I would also like to thank Paul Halstead for both collecting the mandibles from Greece and commenting on this paper.

References Bullington, J. 1991. Deciduous dental microwear of prehistoric juveniles from the Lower Illinois River Valley, A m e ric a n J o u rn a l o f Physical A n th r o p o lo g y . 84: 59-73. Gordon, K.D. 1980. Dental attrition in the Chimpanzee P a n tro g lo d y te s verus: A Scanning Electron Microscope study. Unpublished Ph.D. thesis. Yale University. Ann Arbor, University Microfilms. Gordon, K.D. 1988. A review of methodology and quantification in dental microwear analysis. S c a n n in g M icro sc o p y. 2: 1139— 1147. Greenacre, M.J. 1984. T h eo ry a n d a p p lic a tio n s o f co rre sp o n d e n ce a n a lysis. London, Academic Press. Grine, F.E. 1986. Dental evidence for dietary differences in A u stra lo p ith e c u s and P a ra n th ro p u s: a quantitative analysis of permanent molar microwear. J o u rn a l o f H u m a n E vo lu tio n . 15: 783-822. Harmon, A.M. and Rose, J.C. 1988. The role of dental microwear analysis in the reconstruction of prehistoric diet. In D ie t a n d S u b siste n ce : C u rren t A rc h a e o lo g ic a l P ersp e ctiv e s (eds D.V. Kennedy and G.M. Lemoine). Proceedings 19th Annual Conference CHACM, pp. 267-272. Kelley, J. 1990. Incisor microwear and diet in three species of colobus. F o lia P rim a to lo g ia 55: 73-84.

221

Lalueza Fox, C. and Pdrez-Pdrez, A. 1993. The diet of the Neanderthal child, Gibraltar 2 Devils tower through the study of the vestibular striation pattern. J o u rn a l o f H u m a n E vo lu tio n 24: 29-41. Mainland, I.L. 1994. An evaluation of the potential of dental microwear analysis for reconstructing the diet of domesticated sheep and goats within an archaeological context. Unpublished Ph.D. thesis. University of Sheffield. Meurman, J.H. and Frank, R.M. 1991. Progression and surface ultrastructure of in vitro caused erosive lesions in human and bovine enamel. C aries R esea rch 25: 81-87. Molleson, T., Jones, K. and Jones, S. 1993. Dietary changes and the effect of food preparation on microwear patterns in the late Neolithic of Abu Hureya, Northern Syria. J o u r n a l o f H u m a n E vo lu tio n 24: 455-468. Perrins, J., Williamson, M. and Fitter, A. 1992. Do annual weeds have predictable characters? A c ta E co lo g ica 13: 517-533. Piperno, D. 1988. P h y to lith a n a ly s is . A n a r c h a e o lo g ic a l a n d g e o lo g ic a l p e rsp e ctive . London, Academic Press. Puech, P-F., Albertini, H. and Serratrice, C. 1983. Tooth microwear and dietary patterns in early hominids from Laetoli, Hadar and Olduvai. J o u rn a l o f H u m a n E vo lu tio n 12: 721-729. Ryan, A.S. and Johanson, D.C. 1989. Anterior dental microwear in A u stra lo p ith e c u s a fa rensis: comparisons with human and nonhuman primates. J o u rn a l o f H u m a n E v o lu tio n 18: 235268. Shennan, S. 1988. Q uantifying A rc h a eo lo g y. Edinburgh, Edinburgh University Press. Solounias, N. and Hayek, L.A.C. 1993. New methods of tooth microwear analysis and application to dietary determination of two extinct ungulates. J o u rn a l o f Z o o lo g y 229: 421-445. Solounias, N. and Moelleken, S.M.C. 1992. Tooth microwear analysis of E o tra g u s sa n sa n ie n sis, Mammalia Ruminantia: one of the oldest known bovids. J o u rn a l o f V erte b ra te P a la e o n to lo g y 12: 113-121. Strait, S. 1993. Molar microwear in extant small-bodies faunivorous mammals: an analysis of feature density and pit frequency. A m e ric a n J o u rn a l o f P h ysic a l A n th r o p o lo g y 92: 63-79. Taylor, M.E. and Hannam, A.G. 1987. Tooth microwear and diet in the African Viverridae. C a n a d ia n J o u rn a l o f Z o o lo g y 65: 1696-1702. Teaford, M.F. and Robinson, J.G. 1989. Seasonal or ecological differences in diet and molar microwear in C eb u s n ig rivitta tu s. A m e ric a n J o u rn a l o f P h ysic a l A n th r o p o lo g y 80: 391-401. Teaford, M.F. and Runestad, J.A. 1992. Dental microwear and diet in Venezuelan primates. A m e ric a n J o u rn a l o f P h ysic a l A n th r o p o lo g y 88: 347-364. Teaford, M.F. and Walker, A. 1984. Quantitative differences in dental microwear between primate species with different diets and a comment on the presumed diet of S iva p ith ec u s. A m e ric a n J o u rn a l o f P h ysic a l A n th r o p o lo g y 64: 191-200. Teaford, M.F. 1991. Dental microwear; what can it tell us about diet and dental function. In A d v a n c e s in D e n ta l A n th r o p o lo g y (eds M.A. Kelly and C.S. Larson). New York, Wiley-Liss, pp. 341-356. Ungar, P.S. 1994 Incisor microwear of Sumatran Anthropoid primates. A m e ric a n Jo u rn a l o f P h ysic a l A n th r o p o lo g y 94: 339363. van Valkenburgh, B., Teaford, M.F. and Walker, A. 1990. Molar microwear and diet in large carnivores: inferences concerning diet in the sabre-toothed cat, S m ilo d o n fa ta l i s . J o u r n a l o f Zoology 222: 319-340. Walker, A.C., Hoek, H. and Perez, L. 1978. Microwear of mammalian teeth as an indicator of diet. S c ie n c e 201: 908910. Walker, A. and Teaford, M.F. 1989 Inferences from quantitative analysis of dental microwear. F o lia P rim a to lo g ia 53: 177189.

Soil Pollen Analysis: A New Approach to Understanding the Stratigraphic Integrity of Data Richard Tipping,1Stephen Carter,2 Donald Davidson,1Deborah Long1and Andrew Tyler1 1 Department o f Environmental Science, University o f Stirling, Stirling FK9 4LA; 2 A.O.C. (Scotland) Ltd, 4 Lochend Road, Edinburgh EH6 8BR

Introduction Soils are very valuable repositories of information about natural and culturally modified environments as they can contain data on an enormous range of the major physical, chemical and biological processes affecting landsurfaces. Soils buried either by natural or anthropogenic processes and preserved in time should, then, present enormous potential for the reconstruction o f past environments. It is unsurprising that the archaeological community should have been particularly aware of the value of buried soils given the high frequency with which they are recovered from archaeological excavations, sealed by, for instance, field banks, walls and cairns. In part the potential for palaeoenvironm ental and palaeoeconomic reconstruction was realised at a compara­ tively early stage in the development o f archaeological science. Soils were recognised as foci for the deposition o f anthropogenic remains and objects (Biek 1969). Cornwall (1958) elucidated the significance for the archaeologist of pedogenic changes, and integrated this evidence with climatic and economic data. In equal measure, Dimbleby (1962) was quick to seize on the potential of archaeological sites to record the evolution o f soils through time. Micro­ scopic observation o f soil components led to the development of soil micromorphology (Brewer 1964). The biological components o f buried soils have provided an equally intense area o f enquiry (Dimbleby 1955). One advantage terrestrial soils have over more conventional substrates for biological preservation, waterlogged deposits, is their direct relevance to the land surfaces occupied by human communities. For example, non-marine mollusca provide evidence o f both natural environments (Kerney 1977) and human economic activities (Kerney 1966; Evans 1972) in calcareous soils. Pollen is also preserved in some soils, and can provide information on the nature o f the vegetation on and around a soil surface, and o f agricultural and other activities taking place (Dimbleby 1969, 1985). Preservation is often not ideal, and alkaline sediments and arid climates introduce severe limitations on the application o f the technique. N evertheless soil pollen analysis has become, in the British Isles largely through the energies o f Dimbleby, a staple element o f many environmental recon­ structions on archaeological sites. In continental Europe analyses o f soils buried by archaeological monuments have been pursued for many decades (e.g. W aterbolk 1950). D istinct and critical differences in the approach made to such analyses between those on the continent and those encouraged by Dimbleby and others in Britain emerged early in the history o f the

technique. Continental practitioners have been concerned to identify the vegetation and land use immediately prior to burial (Waterbolk 1956; Groenman-van Waateringe 1979; Casparie and Geoenman-van Waateringe 1980) from single pollen spectra pertaining closely to old ground surfaces. In contrast it has become customary in this country to approach buried soil profiles as pollen-stratigraphic records (Dimbleby 1957,1969,1985), e.g. a vertical series o f contiguous pollen samples. It is this innovation, or rather the assumptions behind it, that have led to controversies in the value and appropriateness o f soil pollen analysis. The major assump­ tion in this approach is that soil profiles maintain some form o f stratigraphic integrity.

Assumptions of Stratigraphic Integrity in Soil Pollen Analyses It is acknowledged (Dimbleby 1957,1985) that the biological and physico-chemical complexities in soil development mean that stratigraphic controls can never be as assured as in sediments such as lake muds or peats. The latter sediment types are deposited or accreted in stratigraphic order and through being waterlogged present far more inert environ­ ments with only lim ited amounts o f post-depositional disturbance. Terrestrial soils do not evolve in stratigraphic sequence, are continually evolving, are constantly prone to disturbance and have b iological and physical m ixing processes which lead in many soils to homogenisation o f different-aged components. Nevertheless Dim bleby has consistently argued that interpretative sense can be made o f pollen analyses from soil stratigraphies. He has based this opinion on a number o f fundamental assumptions; (a) that movement o f pollen within soils “is not the simple movement of the pollen grains as discrete particles through the interstitial spaces o f the soil” (1985: 2) but that pollen “is locked up in the soil, perhaps in some form o f humic complex” (1985: 3); (b) that pollen “movement is therefore extremely slow ” (1985: 3), perhaps c. 3 cm/100 years in most soils; These assumptions lead to the suggestion that “there is a crude stratification, with the oldest grains being the lowest in the profile” (1985:3). Figure 1 presents Dimbleby’s model o f pollen distributions within the soil profile, as concentra­ tions per unit o f sedim ent (APF or A bsolute P ollen Frequency) and as a percent o f the pollen sum. The topmost parts o f a profile contain the majority o f recently deposited pollen grains, although some movement o f this assemblage occurs dow n-profile. B ecause o f the slow downward

Soil Pollen Analysis

AP F

20

40

60

80

100%

Figure 1. Dimbleby’s model o f the theoretical distribution in soil o f different-aged pollen grains (from Dimbleby 1985).

m ovement o f pollen the lower parts o f the profile are dominated by ancient or residual pollen, but long-lived pollen assemblages can occur high in the profile through upward mixing processes. As concentrations (APF), the topmost spectra are richest as recent pollen is less prone to removal through decay and because downward movement has not attenuated the input as much as in older assemblages. Dimbleby’s model is constructed from the interpretation of buried soil profiles. It has rarely been tested empirically, except in the experimental approaches o f Walch e* al. (1970) or the observations o f pollen movement at the long-term earthwork investigations at Overton and Wareham instigated by D im bleby (Jew ell and Dim bleby 1966; Evans and Limbrey 1974; Crabtree 1990). D im bleby’s inferential approach (1985) itself makes certain assumptions, not least in the need to label particular pollen tax a as ‘recent’, ‘intermediate’ in age or ‘ancient’ in order to explain their vertical distribution. But the potential circularity o f such arguments is clear; e.g., what constitutes an ‘ancient’ taxon is its position in the profile. In similar vein certain pollen grains are identified as downwashed contaminants in lower parts o f soil profiles because they are assumed to be more recent than their stratigraphic position (e.g., Dimbleby 1962; Rankine et al. 1960). Interestingly Dim bleby finds no contradiction between this and the assumption that pollen grains form coherent assem blages in soils. The over­ whelming majority o f grains cannot be so readily classified, however, and, the near-complete uncertainty over what grains are contemporaneous with what others remains one severe difficulty in the model. There is currently no adequate test o f the ability of soil pollen analyses to provide palaeoecological interpretations of the calibre of more securely stratified sediments. This conclusion is not by itself a condemnation o f the range o f interpretations produced by Dimbleby and other workers from past work. It may be that Dimbleby’s inference-driven model is a close approximation to reality. However, without adequate testing we will not be able to know with any confidence how reliable these interpretations are.

223

to quantify the role o f earthworms in pollen incorporation, are innovative but handicapped by their analysis o f only one component. Soil profiles evolve by a very complex set o f processes, in which particular components can be most effective only at discrete periods. For instance, increasing so il acidification greatly inhibits the sign ifican ce o f earthworm activity (Keatinge 1983). Single-variable experi­ ments o f this kind can only explain a restricted range o f the explanatory factors involved in soil pollen movement, and in addition cannot reproduce the com plex interactions between different variables that represent the true nature of soil development. More interesting are experimental approaches which trace through time the movement o f pollen within soil profiles. These allow the elucidation of the end-result o f soil-forming processes. Of note here are the long-term ‘experiments’ at the Overton and Wareham earthworks (above). Comparable in temporal scope are the experiments into pollen deterio­ ration ofHavinga (1984). There are, however, two limitations to such approaches. Firstly, the duration o f these experiments (32 years at Wareham and Overton and 20 years in Havinga’s work) is insufficient to explore the implications for truly long-term pollen movements in soil profiles: if Dimbleby’s (1985) estimate of c. 3 cm/century are realistic for soil pollen incorporation then the duration of the earthwork experiments are clearly inadequate to draw any conclusions. Secondly, it is not possible to isolate the significant factors in soil pollen incorporation and movement. The ideal experiment would attempt to combine the best of these different approaches. It should aim to observe pollen incorporation and movement within naturally evolving soils, thus allowing the full complexity of soil-forming processes to be expressed. It should be capable of tracing the movement o f pollen grains over sufficiently long time periods to allow conclusions relevant to timespans appropriate to buried soil profiles. The experiment should allow the depiction o f soil pollen movement in a range of soil types, particularly those soils typically preserved beneath archaeological monuments. It should aim not only to elucidate the patterns of soil pollen movement but also to reconstruct the processes responsible for these patterns. And finally the experiment must demon­ strate whether particular, if any, soil types successfully reproduce sufficient stratigraphic integrity to be capable o f the palaeoecological interpretations derived from more conventional pollen sites. What follows in this paper is the outline of a test which, it is suggested, will be capable o f satisfying these stringent criteria. It will concentrate on one fundamental aspect of this test, in the verification o f one key criterion, and will endeavour to establish a firm experimental basis for future work. The paper should be seen as an interim report. No data are presented on soil profiles themselves; rather, this report will be concerned with the validation o f the approach, which is currrently being tested through the financial support o f a major research grant from the Natural Environment Research Council.

D evelop in g a T est o f S oil-S tratigrap h ic In tegrity

The Temporal Element in Soil Pollen Movements

It is unlikely that a true test of Dimbleby’s model o f pollen incorporation and translocation can be generated from further analyses o f buried soil profiles. The laboratory-based experiments o f Ray (1959) and Walch etal. (1970), designed

The critical aspect lacking from previous attempts to explore the behaviour o f pollen in soils is argued to be a timescale adequate to understand soil-forming processes over periods longer than decades. Walch et al. (1970) successfully

224

Richard Tipping, Stephen Carter, Donald Davidson, Deborah Long and Andrew Tyler

monitored the movement up- and down-profile o f radioactively-labelled pollen grains over a short period, and unique identifier pollen types {Lycopodium spores) were successfully added to the soil surfaces buried beneath the earthworks at Overton and Wareham (Evans and Limbrey 1974) and traced over almost three decades. It would be of considerable advantage to be able to utilise this approach, the identification o f marker pollen grains introduced at known ages, and to observe their movem ents over, if possible, several centuries duration. This NERC-funded proposal has sought to do this by the stratigraphic pollen analysis o f naturally evolving soil profiles which received pollen from arboreta, plantations of trees and shrubs both rare and exotic to an area, planted for aesthetic reasons, scientific study and for commercial purposes. The key importance of arboreta is that the pollen from exotic trees, planted at a known age, can in theory be used as clear stratigraphic markers within stratified pollenbearing deposits, and as tracers of pollen movement within less clearly stratified soil profiles. They serve as unique injections or spikes onto soil surfaces, and as such can be used to measure the degree o f mixing of different-aged pollen assemblages upon incorporation into the profile.

T h e R eq u irem en ts o f th e S tu dy Fundamental to such a study is the selection of an arboretum. The study site needs to satisfy the following demands : (a) a long history of planting o f non-native trees; (b) a secure chronology o f planting; (c) the planting of non-native trees that would be expected to flower in the British climate; (d) a good understanding o f pre-Plantation native woodlands and changes in their representation within the study region; (e) a range o f contrasting soil types within the pollen catchment of the arboretum, and (f) the presence of a securely stratified and dated peat or lake mud to serve as a ‘control’. At an early stage it was considered advantageous also to be able to determine the extent o f translocation o f other comparably known-age ‘labelled’ particles. The principal dated inorganic materials added to the soils are thought to be historic-age tephras (Dugmore 1989) and post-19th century Spheroidal Carbonaceous Particles (SCP’s) (Wik and Natkanski 1990). The last requirement, particularly that concerning the deposition o f tephra, restricts the search to northern Britain, and a study site in Scotland is preferred through the more limited range o f native tree taxa there, and the resultant increase in tree types that could be regarded as non-native but flowering. ‘Control’ sedimentary sequences are more common in the north of Britain with its wetter climate.

D aw yck B otan ic Gardens By far the best arboretum in northern Britain for our purposes is that at Dawyck, near Peebles in Borders Region, southern Scotland, some 45 km south o f Edinburgh. It is now owned by the Royal Botanic Gardens at Edinburgh. Dawyck is the best-researched arboretum in Scotland (Bown 1992). It has the earliest records for the planting of exotic (e.g., non­ native) trees in Scotland, commencing in the 1650’s with the introduction of horse chestnut {Aesculus hippocastanum). Planting was intensified in the late 17th and early 18th centuries following its purchase by the Naesmyth family,

and again after the 1830’s with the fourth Baronet, Sir John Murray Naesmyth. The diversity o f the collection was further enhanced by F.R.S. Balfour in the early 20th century. The gardens cover 24 ha, rising from c. 230 m OD beside the privately owned Dawyck House to around 280 m OD. Dawyck is one o f the coldest parts o f Britain through its inland situation and high altitude (Bown 1992). The climate is comparatively dry with annual rainfall totals averaging 890-1140 mm. This more continental climate favours the planting o f boreal species, with an abundance o f conifers, and cold-hardy taxa from North America and China are common. Yet even the Mediterranean maritime pine {Pirns pinaster) is hardy as far north as Edinburgh, and the com paratively thermophilous sm all-leaved lim e {Tilia cordata) produces pollen at the present day. Many o f the planted species have similar pollen grain morphologies. Figure 2a shows the thirteen different pollen taxa defined for 59 o f the 72 taxa with known planting dates, and Figure 2b shows the principal plant taxa that comprise these pollen types. Figure 2 also illustrates the planting chronology at Dawyck in two ways. Figure 2a details the first plantings o f the pollen taxa. Figure 2b shows the contributions through time o f individual plant taxa to the records for pollen taxa. This shows that, for example, horse chestnut {Aescylus) was planted only once whereas species of fir {Abies) were planted from the late 17th century onwards, though with increasing frequency this century. An idealised and grossly sim plified pollen diagram o f the sequence anticipated at Dawyck can be constructed (fig. 3) from the documentary record. The site lies firmly within a region o f primary deciduous woodland, o f oak {Quercus), elm ( Ulmus), ash {Fraxinus), hazel {Corylus) and rowan {Sorbus) (Birks et al. 1975; Bennett 1989). Alder {Alnus) and birch {Betula) were common in moister localities. Scot’s Pine {Pinus sylvestris) is not known in macrofossil form (Dickson 1988) or as pollen in sufficient numbers to imply local presence (Bennett 1984). Southern thermophilous taxa such as beech {Fagus) and lime {Tilia) are not native (Pigott and Huntley 1980; Birks 1989). There is substantial evidence to suggest that trees were very rare elements o f the southern Scottish landscape long before the introduction o f non-native trees to Dawyck. Very extensive and comprehensive deforestation in the late Iron A ge, continuing into the Rom ano-British period (Mannion 1978; Tight 1987; Tipping 1992; Dumayne 1993; Dickson 1993) left essentially treeless hills which were in general not reforested in subsequent periods (Tipping and Halliday 1994). Woodland was extremely scarce in the medieval period (Anderson 1967), and the description of Tweeddale by Pennecuik (1715: 4), o f “not one wood worth the naming in all this open and windie country”, could equally be applied to earlier times. The bedrock geology comprises acid Silurian mudstones and shales, but this uniformity is modulated by thin tills, thicker periglacial colluvial deposits, fluvioglacial gravels and alluvial silts. The altitudinal and hydrological variations provide a great range in soil types and properties, from freedraining acid brown forest soils to incipient podsols and gleys. During reconnaissance within the Gardens, however, it was recognised that extensive landscaping had led to much soil disturbance, and that liberal use o f fertilizers within the arboretum might have accelerated pedogenic processes. An overriding necessity o f the research design is for rates and

Soil Pollen Analysis m §

§■

225 m to o

§ f A E S C U LU S

s

1

(horse chestnut) Ifirsl

a p ie s

§—

w

1

IA R IX Tvo e

TIUA

t..............>.........— .. .. .

I__________________

(E uropean & other larches)

( lim e s )

P fffU S DtPLOXYLON

j__________________

(Pinaster and other pines)

A C E R S A C C H A R U M Type

... .11. .................. I____________ _ _ _

P IC EA

(S ugar m aple)

(spruces)

P O P U LU S

(Lom bardy poplar)

I-------------------------------

P IN US H APLOXYLON

.......... .......

}.....-

I-----------------------j

(Black and other pines)

(Lebanon cedar)

fr—..........- ' ....-..- ... .

TAXODIACEAE

I__________________

JUNIPERUS Type

lb™..................■'

AESCULUS

CEDRU S

TSUGA

(W ellingtonia type) (C ypresses)

(W estern hemlock)

A . h ip p o c a s ta n u m ’..-

~X alba I------------------- A . balsa m e a

IIA B IE S

_

A . p ro c e ra A . n o rm a n d ll ----- A . g ra n d ls I_______________ A. a m a b ilis ,_______________ A . va H chli i------------------------ A . c a p h a lo n lc a i------------------------- A . m a r im ii \______________ A. m a g n ifie s l------------------------ A. d e la v a y i {------------------------ A. faxo n ia n a I------------------------ A. h o m o la p is I----------------------- r A. n u m ld ic a

L a rix d e cid u a P s e u d o ts u g a ta x ifo lia

L A R IX T yp e

L. e urol& p is L. la rlc ln a L. o c c ld e n ta lis __ P. ta x ifo lia

fT IU A

I------------------------- F. co rd a ta j------------------------7~P. p in a s te r ___________ P. s y lv e s tris P. p o n d e ro s a \ ... I______________ P. n ig ra

PINUS DIP LO XYLON A C E R S A C C H A R U M Type

I-------------------------

A . s a c c h a ru m

I______________

P. g la u ca P. s itc h e n s ls I -. P. e n g e lm a n n ii I______________ P. b ra w a ria n a I------------------------- P. o b o va ta I______________ P .je z o e n s is l------------------------ P. m a ria n a i______________ P. o m o rik a I______________ P. b ic o lo r I______________ P. o ria n ta lis I______________ P. b ra c h y ty la 1------------------------- P. liU a g a is is l P .k o y a m a i 4______________ P. a s p e ru ta

PICEA

P. n ig ra P O P ULUS PINUS

I______________ P. tric h o c a rp a P. c e m b ra P. p ea ce

HAPLOXYLON I------------------------- C. a fla n tic a CEDRU S TAXO DIACEAE

I-----

C. a tla n tic a

I------------------------S e q u o ia d e n d ro n g ig a n ta u m |-------------------------C h am ae cyp aris la w s o n ia n a *...j

JU N IP E R U S Type

TSUGA

................

C. la w s o n ia n a

j------------------------ C. n o o tk a ta n s is I______________ Thuja s ta n d ia h il I------------------------- T. ko ra ie n a is j______________ C u p ra s s o c y p ris ie y ta n d ii I------------------------J u n ip a ru a c h in a n s is I------------------------ J. s c o p u lu ru m T. h e te ro p h y lla i______________ f. m e rta n sia n a I______________ T .s ie b o ld il

Figure 2. The chronological order o f planting o f pollen taxa with known planting dates at Dawyck (a); the botanical taxa with known planting dates that contribute to the pollen taxa, showing the frequency ’ o f planting o f individual tree types (b).

processes o f post-Plantation soil-formation to have been natural. Accordingly the search for undisturbed soil profiles was extended to areas outside the arboretum. The two require­ ments were (a) that the study site selected should be confined to an area within the pollen source area of trees at Dawyck, e.g., east and north-east, downwind, o f the gardens and (b) that undisturbed soils of contrasting properties be available

within a limited area. A distinct advantage o f the tree types planted at Dawyck is the predominance o f anemophilous pollinators. In particular the preponderance o f conifer species with their well-know n facility for long-distance aerial dispersal allows the assumption that areas outside the garden would still receive much o f their pollen from the arboretum. A second advantage was that with increasing distance it could be expected that the pollen ‘rain’ fom Dawyck would

226

Richard Tipping, Stephen Carter, Donald Davidson, Deborah Long and Andrew Tyler

% non-native arboreal pollen

Figure 3. Idealised pollen diagram o f planted non-native tree taxa at Dawyck as percentages o f non-native tree taxa. More types o f tree have been planted than are represented here : each curve represents an identifiable pollen taxon (see fig. 2); e.g., Firs includes all Abies species. The timescale is derived from the known age o f planting (see text fo r discussion).

become increasingly well-mixed. This means that providing the soil profiles sampled lay close to each other, the second criterion, then it could be expected that both relative (%) and possibly absolute (influx) pollen inputs to soil surfaces would be uniform. This was also to be tested by pollen analyses. The intention was to locate a study site within undisturbed pasture. About one kilometre east o f Dawyck House is the valley of the Lour Burn, an area of rough pasture and Calluna heath. A detailed soil survey of the valley confirmed that a range o f soil types could be found within a very restricted area o f the valley. Although known to have been occupied and farmed for crops in the late historic period (Pennecuik 1715; Buchan 1927; Dunbar and Hays 1960), there is good circumstantial evidence that cereal cultivation had ceased by the first appearance o f exotic tree pollen types, and all the soil types sampled have probably evolved naturally over the last few hundred years.

E stab lish in g a P ollen C h ron ology fo r L ou r A key assumption in moving away from Dawyck, albeit by

only one kilometre, is that the planting record at Dawyck will be replicated in the pollen record at Lour. This needed to be tested by pollen-stratigraphic analyses on a ‘master* sediment sequence at Lour. This record would then serve, supported by comprehensive dating , as a securely stratified ‘master’ pollen chronology in comparisons with adjacent soil profiles. Such pollen analyses were also necessary to authenticate the planting record constructed for Dawyck. Figure 3 is a simplistic construct, and makes no allowance for differences between arboreal pollen taxa in either the numbers o f plants introduced at any one time or changes through time, or the effects o f differing pollen production or pollen dispersal mechanisms or rates between taxa. It is also possible that trees o f a particular pollen taxon could have been planted prior to their known planting date; o f the 305 species listed as extant by Balfour (unpublished) in 1949, only 72 (23.6%) have known dates of planting (above). Equally, no allowance can realistically be made for delays between initial planting o f a taxon, obtained from the documentary record, and its first appearance in the pollen record.

Figure 5. Percentages o f arboreal pollen as % a.p. at the deep peat LP against age before present (BP) (for pollen sum see fig. 4).

% op

Figure 4. Pollen diagram o f all tree taxa calculated as % ap at the deep peat at Lour, plotted against depth below ground surface; SCP’s, tarn and sums attained are presented as numbers; total pollen concentrations are in grains per cc x 100. The ‘grains per traverse' curve depicts actual numbers in black, the hollow curve being a x 10 exaggeration.

Soil Pollen Analysis 227

228

TAXON

Acer campestre type Acer saccharum type Aescylus Aims Betula Cedrus Coniferales undiff. Abies Picea CorylusIMyrica Fagus Fraxinus Juniperus type Larix type Pirns undiff. P. diploxylon type P. haploxylon type Populus Prunus type Quercus Taxodiaceae Taxus Tilia Tsuga Ulmus

Richard Tipping, Stephen Carter, Donald Davidson, Deborah Long and Andrew Tyler

STATUS

native - planted non-native non-native native native non-native non-native non-native non-native native non-native native - planted native - planted non-native non-native non-native non-native native - planted native - planted native non-native non-native non-native non-native native - planted

YEAR OF PLANTING

LOUR PEAT

LOUR NINE

present

first app.

present

first app.

unknown 1755 1650

yes yes no

1900 1930

no yes no

1880

1870 1680 1680 1760

yes yes yes yes

1830 1860 1810 1810

yes yes yes yes

1870 1840 1850 1900

unknown 19th C. ? 1860 1730 1740 1740 1840 1760 1910

yes yes yes yes yes yes yes yes no

1820 15th C. 1850 1850 1800 1800 1940 1850

yes yes no yes yes n.d. n.d. yes no

1850 1860

1850 unknown 1730 1860 unknown

no yes yes yes yes

1890 1880 1850 1830

no no yes yes yes

1860 1840 1800

1880 1850 1850

n.d.: not determined.

Figure 6. Details o f (a) non-native and native tree taxa recorded at Dawyck; (b) dates o f planting at Dawyck (if known); (c) native tree taxa recorded as pollen at (i) Lour Peat and (ii) Lour 9; (d) the estimated age o f the first appearance o f pollen o f non-native and native planted tree taxa at the two Lour sites.

During soil-mapping a relatively deep (1.07 m) peat sequence was located in the middle of the Lour Burn study area. The basin is « 1 0 m in diameter, and this implies that the peat has the pollen recruitment characteristics o f a ‘small hollow ’ (Bradshaw 1988); that is, the site would reflect plant communities growing within tens of metres o f the site if surrounded by woodland. In the open landscape o f the present and recent past the pollen catchment can be anticipated to have been larger, reflecting in part the planting of trees at Dawyck itself. Given that all the soil profiles lie close to the peat and between this and Dawyck, no significant differences in pollen inputs should be seen between these different sites. This site is called Lour Peat (LP). A second shallower peat sequence (Lour 9) was also analysed to test the data from LP. The results from this site are strikingly similar to those from the fuller record at LP (e.g. fig. 6, below) and this site will only be referred to briefly. Comprehensive pollen analyses have been undertaken on the sediments of these small peat deposits. Figure 4 is not a complete pollen diagram from the peat LP, but presents data on the range and stratigraphic representation o f all tree taxa recorded, to serve both to explore the stratigraphic integrity of this ‘master’ sequence, and to analyse the extent to which the Lour Bum sites reflect the planting chronology at Dawyck. These analyses are supported by a chronology independent of that derived from tree-planting at Dawyck, by conventional and wiggle-matched ,4C dating, 210Pb dating, SCP and tephra dating. Figure 5 presents similar % data to

figure 4, but for the topmost c. 30 cm o f LP only, plotted against cal. age BP as estimated by counts o f SCP’s (fig. 4), the most reliable measure available to date. Numbers o f these particles are expressed as % particles per total pollen in figure 4. Although ages of (i) the first appearance and (ii) peak in SCP’s are usually defined by 2,0Pb dating, and so are not independent o f other methods, there is a good deal of consistency at Scottish sites (Wik and Natkanski 1990; Jones et a l 1993; Rhodes 1995) for (i) to have a date o f around 1900 AD and (ii) to date to c. 1940 AD. At the peat these points occur at (i) 13.25 cm and (ii) 6.25 cm (fig. 4).

The Stratigraphic Integrity o f the Deep Peat LP Much o f the work in comparing the ‘master’ sequence with soil pollen profiles hinges on the critical assumption that the peat deposits preserve pollen assemblages in true strati­ graphic order, without disturbance or distortion. There have been suggestions, for example, that in some peats there can be post-depositional vertical movement o f pollen (Clymo and Mackay 1987). The test o f this assumption comes with observing the order in which non-native trees planted at Dawyck occur within the peats. Pollen from these trees should not be recorded in sediments older than the date o f planting at Dawyck. This analysis assumes that the arboreal pollen (ap) sum is adequate to demonstrate absence o f evidence, which is o f course impossible in the true sense, but in a probabilistic

Soil Pollen Analysis sense can be supported to an extent. For example, below 24 cm a pollen sum of 200 arboreal pollen grains could not be achieved (fig. 4). This restricts the statistical support for the interpretations presented below, but in a practical way demonstrates the near-absence of tree pollen in these sedi­ ments, either contemporaneous or downwashed. The tree taxa in figures 4 and 5 can usefully be divided into three types governed by their origins: native taxa not planted at Dawyck; native taxa planted at Dawyck; non­ native taxa planted at Dawyck (fig. 6). The overwhelming majority o f non-native taxa, and native taxa known to have been planted at Dawyck, are not represented in pollen counts at L P below 25 cm, a time estimated at c. 1840 AD. Comparable results are seen at Lour 9 (fig. 6). Three such taxa are recorded below this depth (fig. 4); Pinus (both undiff. and P. Diploxylon type), Ulmus and Fraxinus, though not as more than one to two grains per count. These occurrences are not regarded as evidence for the down-washing of pollen because these taxa are in many ways ‘special cases’. Pinus is not native in southern Scotland but populations are extant in the Cairngorm Mountains (Steven and Carlisle 1959) and as remnants of the* original pine forests as far south as the head of Loch Lomond (Stewart et a l 1984). These provide sources o f contemporaneous Pinus pollen throughout the period covered by the pollen diagram. Ulmus is native to southern Scotland, and might be thought a poor indicator o f late historic tree-planting. There is, however, much palynological evidence (Mannion 1978; Tight 1987; Tipping 1995) for Ulmus not to have grown in the region in abundance for most o f the historic period. Single grains recorded at Lour are derived either by long­ distance transport or from surviving but isolated trees. Although two species o f Fraxinus were planted at Dawyck, probably in the late 19th century (Gourlay 1908), pollen grains earlier than this, as early as the 15th century, are derived from native trees of F. excelsior (ash). More significant is the observation that only one true non-native tree taxon occurs in peat earlier than its known date of planting, Cedrus (fig. 6), apparently found as early as c. 1830 AD. However, this small age-difference may prove insignificant with more detailed dating. A single SCP is found at LP in peats dated to c. 1830 AD, but sporadic finds o f single SCP’s in 19th century contexts are noted by Wik and Natkanski (1990). In conclusion, it is very likely that this peat accumulation has acted as a secure stratigraphic context over at least the last c. 250 years. This is a very significant result since it is on this basis that the ‘master’ sequence can be used to observe and calibrate distortions o f this secure pollen stratigraphy w ithin the actively evolvin g soil pollen stratigraphies.

Representation o f Tree Taxa at Lour A large number o f non-native and native planted taxa at Dawyck (fig. 6) are recorded at Lour; 85 % o f these taxa at LP and 67 % at Lour 9. Of those taxa known to have been planted in the years preceding the mid-18th century AD Aescylus is not at all recorded at Lour, and Abies, Acer saccharum type, Larix type and Pinus (undiff. and P. diploxylon type) are only recorded much later than their planting dates. Tilia is very rarely recorded despite an avenue o f lime trees extending closer to Lour than the Gardens themselves. Similar delays between planting at Dawyck and

229

first appearance at Lour are generally less obvious in tree taxa planted later than the mid-18th century. Nevertheless, a lag between planting and pollen representation is clear within the 20th century, where there is no expansion o f Picea pollen despite the planting of commercial plantations a few hundred metres west and upwind o f the peat. It would be dangerous to make too much o f these apparent delays since the majority of tree taxa are recorded at very low frequencies. The arboreal pollen sum, although reason­ ably high, is not rigorous enough to provide assurance that the first record o f a taxon is coincident with its first appearance in the pollen ‘rain’. This is clearly a major problem in using some o f these taxa as ‘tracers’ in soil profiles. Only a few non-native and native planted taxa are recorded at percentages high enough to be able to confidently determine their ‘rational limit’, the “point at which the pollen curve begins to rise to sustained high values” (Smith and Pilcher 1973; 904). These taxa are: Pinus undiff. and Pinus diploxylon type (undifferentiated Pinus grains are probably to be equated with P. diploxylon type), Ulmus and Fagus, and possibly a taxon combining counts of Coniferales undiff., Abies and Picea. What appears from these analyses is the recognition that the time-interval over which planted tree taxa are represented at Lour is markedly shorter than anticipated from con­ struction of the Dawyck planting chronology. A 300-350 year history of pollen ‘injections’ was hoped for, but the earliest pulse at Lour, o f Pinus, is dated at c. 1800 AD, the rational limit of Ulmus is seen at c. 1840, and o f Fagus at broadly the same time. The first appearance o f Coniferales might post-date these, but seemingly not significantly. There are, therefore, at best two pulses of planted taxa at Lour, and these are only separated by c. 40 years or so. It might be safest to regard these as indistinguishable, and to accept that one early 19th century pulse is recorded. These data seem to confirm a concern that the Lour pollen record does not reflect the full intricacy and detail o f the Dawyck planting chronology. One reason for this might be the distance that separates Dawyck from Lour, in that many pollen grains released from trees at Dawyck may not have been dispersed the one kilometre to Lour. However, it may also be that a comprehensive depiction o f the documented Dawyck planting record by palynology is an unreasonable expectation. To test these two interpretations, preliminary pollen analyses were undertaken on a sedimentary sequence close to Dawyck Gardens, immediately north o f Dawyck House, in an abandoned meander o f the River Tweed (fig. 7). These few counts demonstrate that the palynological representation o f non-native trees at Dawyck itself is less than could have been expected from the documentary record. Although most of the thirteen non-native tree pollen taxa known to have been planted at Dawyck (fig. 2) are represented (missing taxa are Pinus haploxylon, Juniperus type (Chamaecyparis), Tsuga and Prunus type), many are represented by only single or a few grains. This is a closely similar result to the analyses at Lour. Exotic trees contribute an average of 60 % ap at this site compared to 6 5 -7 0 % ap at the deep peat LP. From this it is argued that it is very unlikely that distance from the arboretum is a serious weakness in the applicability o f the Dawyck planting chronology. Instead the deficiencies appear either to be inherent in the nature of the pollen record or specific to the processes of tree planting at Dawyck.

Richard Tipping, Stephen Carter, Donald Davidson, Deborah Long and Andrew Tyler

230

v?

DAWYCK ABANDONED MEANDER : % arboreal pollen

•o

v0?J

^

V ^

^ T * '" '

^

v$>A

j A; xcA

x‘o ° ^A'

^

,c3

2.95) were concentrated by centrifugation in tetrabromoethane, identified by polarising light microscopy and the relative abundances of the non-opaque minerals obtained by counting some 500 grains. Weathering history is manifested in the relative depletion and etching o f less stable minerals, and the latter was also checked by Scanning Electron microscopy (“SEM”) of selected grains o f individual mineral species. The clay fraction, by virtue o f its greater surface area, is more vulnerable to mineral dissolution and alteration. Samples 90%) and indicating a provenance derived exclusively from the Lower Palaeozoic mudstones o f the Hiraethog Moors to the West, in the headwaters o f the Aled. In the overlying Intermediate Beds and Lower Breccias the assemblage changes to resistate minerals (zircon, rutile, tourmaline etc.), a provenance characteristic o f Carboniferous strata outcropping to the North, together with traces o f material (staurolite, kyanite) derived from the Irish Sea Till. In the overlying Postglacial beds the provenance is a mixture o f these two sources (Jenkins 1984). Grains o f relatively unstable minerals, however, such as chlorite and clinopyroxenes are relatively fresh in the basal beds but deeply etched in the Intermediate Beds/Lower Breccias. This is consistent with the former originating under glacial conditions and the latter coming from an environment which had undergone prolonged and/or intense weathering such as might have operated during an Interglacial Period. This pattern is repeated in the mineralogy o f the clay fraction with the relatively unstable chlorite and derived vermiculite depleted in the “Interglacial Beds” and replaced by traces o f smectites. Despite an intensive search, however, using magnetic and density fractionation, neither crystalline ferric hydrous oxides nor bioliths could be detected. A fuller account o f the mineralogy and its palaeo-environmental implications is in preparation and will be reported in the quaternary literature.

C hem ical A nalysis In an attempt to characterise the pigmenting materials in the Intermediate Beds further, samples from the sediment column were extracted by pyrophosphate, oxalate and dithionite/ citrate treatments (McKeague et a l 1971) and the extracts analysed by Atomic Absorbtion Spectrophotometry (“AAS”) for their Fe, Al and Mn contents. These extractants are used to differentiate bew een Fe an A l that is organically complexed (pyrophosphate), is in amorphous inorganic form (oxalate) and is weakly crystalline (dithionite/citrate). Values

David A. Jenkins

284

Values in wt. %

Upper

0.01

0.1

1

0.01

0.1

1

10

Ss&Gr

Ponding Lower Breccias

Inter­ mediate Beds

Lower Ss&Gr IP

A*p ^*0 Al10,000 ppm) within them and this material can also be dispersed by gravity, wind and water, or by unwary farmers by spreading or roadmaking. Smelting produces slag (also much used at one time on roads), furnace-material residues (both of which again frequently contain percentage levels o f lead) and fine particulate material or fume which was deposited in, or issued from, the furnace or chimney, and can be found dispersed over very wide areas. Research into heavy metal contamination has focussed on a wide variety of objectives, some of immediate obvious value, such as into deaths o f animals and possible effects on humans, some with a more academic thrust, using the area as a field laboratory to establish the parameters whcih affect, for instance, the m igration o f lead in different g e o ­ environments. The number of enquiries received at the Peak District Mining Museum in recent years suggest a consider­ able expansion of interest in lead contamination as a means of training students in field techniques from a surprisingly wide range o f subject disciplines, though there are very few from historians and archaeologists who must rank as amongst the principal potential beneficiaries of such work! The death or ill-health of animals is a common experi­ ence in areas such as the Derbyshire Peak. They can often be related to water polluted by working on active mining sites, or to disturbed soil, or the grazing of grass early in the season or before the root systems are well enough developed to prevent tearing-up of roots and soil. Poisoning through contact with materials on smelting sites is another problem particularly as contamination levels are often locally higher, and animals seem to like the sweetness of lead compounds and will lick lead-glazing on furnace bricks, and fume in collapsed condenser flues. In one instance, at Riber near Matlock, soil contaminated by bole smelting some 500 years ago appears to have been naturally washed down the hillside and was used to form the banks of a silage clamp. This was picked up by the fork lift o f a tractor after tearing o f the polythene membrane used in its construction and led to several deaths of cows. Evidence for the presence o f lead in such cases results from p o st mortem examination by veterinary establishments (Sutton Bonington for Derbyshire), and the subsequent investigation of the source. It has led to the discovery of several smelting sites, for instance the Marsh Farm Cupola at Castleton and has helped confirm others such as the site of the bole at Wirksworth Bole Hill. Chickens are particularly susceptible to particulate lead in grit which they pick up, and in the 1960s, when chicken-keeping in rural surroundings was near universal, asking if chickens were kept usually provided sufficient evidence to confirm whether a lead site nearby was likely or not. Nowadays this cheap and cheerful investigative system is unfortunately less available. The earliest non-veterinary-scientific work within the Derbyshire area to come to the first author’s notice was a botanical study c.1970 on the Magpie Mine site at Sheldon (unpublished, pers. comm Dr D. Thornton then o f Salford University), in which small rectangular areas within a ruined building had a flora distinctive from the remainder. These may relate to the positions of wooden dressing tubs in which lead ore was dressed, but so far as we are aware this type o f research has not been carried further, though metallophytic plants are sufficiently well researched (Buchanan 1992) for such microstudies. Botanical research o f this type has

particular value in not causing disturbance o f the soil, which at Magpie, and many such sites, will require Scheduled Monument Consent. Disturbance o f the soil, if not done especially carefully, may also lead to animal poisoning, and frequently has. At about the same time the then Institute o f Geological Sciences published the first o f its Regional Geochemical Reconnaisances (Nichol et al 1970), on the Derbyshire area, which included most o f the contaminated area o f the Peak District. Their technique, now well established, depended on the assumption that stream sediments were an approx­ imate composite sample o f all the rocks and soil upstream of the sampling point. It was an attempt to provide a means of locating otherwise unknown mineral deposits. In practice the level o f contamination from already exploited veins and from smelting activities almost totally masked the original purpose, but provided the then best available source o f data on the distribution o f smelting sites (W illies 1970). A substantial number o f the 240 or so highest levels o f contamination could be related to known or suspected smelting sites, and the results were a fundamental tool in the still-continuing search for more (Willies 1990; Crossley and Kieman 1992). Geochemical reconnaisance based on streams was particularly suitable for detecting the nearby presence o f water power sites, but less effective in detecting hilltop sites as used in bole-smelting and for some o f the cupolatype furnace sites, though Nichol et a l included follow-up studies of bole and cupola sites at Ashover. This study also suggests it is possible to discriminate between mine and smelter contamination, where that might be a problem, by use o f lead/zinc ratios. In work embracing several mine/ smelting contaminated areas o f Britain, Bradley (1989) attempted to relate pedogenic processes over time as a means o f “fingerprinting” the source o f sediments in rivers. This may be useful for single site contamination, but the multi­ period nature and other complexities o f most mining areas in Britain may severely limit this as a useful technique. Wild, Eastwood and Newman (1984) used a randomised sampling technique in which a single spot sample was taken and analysed from each kilometre grid square within the rural areas o f the North East Derbyshire District. Some 25% (92 sq km) of the District was found to have lead levels elevated above 250 ppm, up to 16460 ppm. Two o f the authors (Eastwood and Wild 1986; Wild and Eastwood 1992) subsequently related this to smelter locations. Quayle (in an unpublished investigation), with one of his students, sampled an area near Wessington Hay where a smelting site was suspected from documentary evidence, but not precisely located. They analysed soil for lead from different depths (20 cm depth apart in each hole) along two axes, NE/SW and NW/SE from the centre o f the suspected area at 50 m intervals. The sampling from different depths helped solve the problem o f soil wash down a steep slope. Allowing for the probable skew due to the prevailing wind direction, the resulting iso-line plot clearly pinpointed the site, and subsequent search found lead slag. The first author c. 1980 (unpublished) used a Proton Magnetometer to investigate suspected (from slag scatters and absence o f vegetation) bole sites on Beeley Moor, near Matlock using 500 mm intervals. This was encouraging in that a number o f probable small hearths were located, some o f which closely related to a scatter o f heat-affected gritstone fragments. A more modem and less cumbersome instrument has recently been used by

The “Spin-Off”from Heavy Metal Research in the Peak District M cDonnell et al. (19 9 2 ) for a sim ilar purpose. The combination o f these techniques has clear application for regional, local and site investigations. The most substantial investigations into the problems o f contamination has com e from post-graduate and post­ doctoral studies by Imperial College, London. Several have attempted to evaluate problems in villages and parishes such as Winster, near Matlock (Merry 1988; Elliot 1989; CotterHowels 1991), and a major study in heavy metal migration was completed in 1993 (Maskall and Thornton 1993), and these collectively have a mass o f information about the area. The greatest archaeological value comes from studies of smelter sites, both in their location, and in the distribution of residues about them, since mining sites can generally be easily located by reference to spoil heaps and vein occur­ rence. However, C otter-H ow ell’s W inster-based study, which includes examination o f mine sites, showed the major contaminant there was not the sulphide or the common secondary-oxidation carbonate and sulphate minerals, but instead was pyromorphite, lead chloro-phosphate (see also Cotter-Howells and Giddens 1990). At Winster Pitts mine (which she refers to as the adjacent Drummers Venture), the soil-lead was predominantly in the form o f pyromorphite (Cotter-Howells 1991). This has led to the identification o f unusual features on the site as specifically relating to the mid-19th century working of pyromorphite in soil and sub­ soils above mineral veins, this being sold as “brown linnet ore” to the smelters using the Spanish slag-hearth furnace (W illies 1995). So far as is currently known, this is a relatively rare occurrence of pyromorphite in economic quantities, the cause o f which is not fully understood, but may relate to a phosphorous rich bed in now eroded overlying shales. No other remaining site in the Peak has quite so many surviving ore-preparation features as seen at Pitts Mine, which might w ell be due to the unusual nature o f the mineralisation rather than the vagaries o f archaeological survival. In the investigation o f migration o f lead in different rock and soil conditions Maskall and Thornton deliberately chose a variety o f smelting sites as dated sources o f contamination produced in reasonably well defined conditions, and was unusual in such studies in deliberately using specialist historical and archaeological input from its commencement. Scheduled monument consent was obtained where necessary, and care was taken on all sites to protect archaeological integrity as far as possible. A wide range o f sites were first investigated for their suitability, by taking samples at different depths on a 10 m grid pattern around the suspected or known site. This showed lead contamination levels up to, and no doubt beyond, the detection limit of 30,000 ppm (3%). At surface, this level of contamination usually leads to areas of nil vegetation, or at best the struggle for survival o f a metailophytic plant such as Minuatia verna. This effect can, however, be masked by down slope migration o f less contaminated material, and buried contaminated soil was recovered using a hand augur. Amongst other useful data, the hand-augur revealed the unsuspected survival o f a condensing flue at Stonedge Cupola. The survey, which deliberately covered different geological environments, also investigated suspected bole sites on “Bull Hills” south o f the orefield on Permo-Triassic strata (unconsolidated sands and pebbles) near Ashbourne, and succeeded in positively identifying two such sites. One of these was found, by

331

auguring, to have a substantial bank o f charcoal/slag material, which was otherwise totally masked by the grass sward. Archaeologically this is now identified as probably the second-best surviving bole-smelting site in the area, and it is hoped a more detailed geochem ical survey w ill eventually be carried out there. It is perhaps useful to point out that even in Derbyshire, bole hills were not solely used for lead smelting, and Quayle (1993) found that Sheldon Bole Hill, nearly at the centre o f the lead mining field, was most likely so-named from iron-making! Some half-dozen sites were chosen from the wide sample for shallow drilling using a highly specialised coring drill, to depths o f up to 10 metres. Its most obvious archae­ ologically value was the revealing o f depths and character o f material on waste heaps, and that it provided further analyses. At Harewood Cupola for example: “Essentially, the profile consists o f a substantial layer of waste about 80 cm thick overlying a felspathic sandstone ... The waste comprises a slag-rich layer between 0 -1 5 cm, a cinder-rich layer between 15 and 80 cm and a band o f coal ash between 80 and 85 cm” (Maskall and Thornton 1993: 62). Similar shallow drilling incidently revealed that at Stonedge Cupola the chimney was carefully founded on a steeply dipping lens of gritstone whilst the very close-by dam and wheelpit area was excavated within soft shales.

Conclusions Environmental research using sophisticated equipment and techniques is commonly undertaken on former industrial sites, particularly for contamination from heavy metal and materials such as arsenic. It can accidentally or purposely be used for the archaeological detection and investigation of sites, and can provide a great deal o f data about residues and their concentrations. Few archaeological investigations can afford the level o f expertise and equipment involved, and site investigations should actively seek out results o f environmental and related investigations, and where feasible, seek to direct them towards suitable areas for investigation. Active collaboration between historians and archaeologists and environmental scientists of many disciplines can be a useful two-way process. Substantial data are also available from veterinaries, and from botanists, and the inclusion o f the latter in an investigation o f contaminated or abandoned industrial sites is probably one o f the most cost-effective additions to an archaeological team in appropriate areas.

References Bradley, S.B. 1989. Incorporation of metalliferous sediments from historic mining into river floodplains. Geo Journal. 19.1: 5 14. Buchanan, P.T. 1992. Metalliferous plant communities: the flora of lead smelting in the upper Nent Valley. In Boles and Smeltmills (eds L. Willies and D. Cranstone), Historical Metallurgy Society, Peak District Mining Museum, Matlock Bath, Derbyshire, pp. 58-61. Crossley, D. and Kiernan, D. 1992. The lead smelting mills of Derbyshire. Derbyshire Archaeological Journal CXII: 6-48. Cotter-Howells, J. 1991. Lead Minerals in soil contaminated by mine-waste: Implications fo r human health. PhD. Thesis, Centre for Environmental Technology and Department of

332

Lynn Willies and John Maskall

Geology, Imperial College of Science Technology and Medicine, University of London. Cotter-Howell, J. and Giddens, R. 1990. Pyromorphite: a secondary lead mineral. Bulletin of the Peak District Mines Historical Society, 11.1: 21-22. Eastwood, I and Wild, M. 1986. Industrial archaeological geochemical prospecting: a scientific aid to the study of industrial history. Hallamshire Historian 1: 17.23. Elliot, Guy 1989. A study o f the behaviour and bioavailability of lead on sites contaminated by past mining and smelting activity in Derbyshire. MSc. Thesis, Centre for Environmental Technology, Imperial College of Science and Technology, University of London. Kiernan, D. 1989. The Derbyshire Lead Industry in the Sixteenth Century. Derbyshire Record Society, Chesterfield. Maskall, J. and Thornton, I. 1993. Heavy metal migration in soils and rocks at historical smelting sites. Final Reoprt. Global Environment Research Centre and Environmental Geo­ chemistry Research Group, Imperial College, London McDonnell, G., Dockrill, S., Heron, C., Starley, D. and Tirpak, J. 1992. Geophysics and slag analysis at Grinton Smeltings. In Boles and Smeltmills (eds L. Willies and D. Cranstone) Historical Metallurgy Society, Peak District Mining Museum, Matlock Bath, Derbyshire, pp.48-50 Merry, Graeme. 1988. A study o f contamination o f agricultural soils by past lead mining and smelting activities in the southern Peak District. Msc. Thesis, Centre for Environmental Tech­ nology, Imperial College of Science and Technology, Univer­ sity of London. Mott, R.A. 1967. Lead Smelting in Derbyshire. Bulletin o f the Historical. Metallurgy Group 1.8: 6-11.

Nichol, I., Thornton, I., Webb, J.S., Fletcher, W.K., Horsnall, R.F., Khaleele, J. and Taylor, D. 1970. Regional geochemical reconnaisance o f the Derbyshire area. Institute of Geological Sciences 70/2. HMSO. Quayle, D.V. 1993. Bole Hill, Sheldon: evidence of iron-making. Bulletin o f the Peak District Mines Historical Society 12.2: 48-49. Robey, J.A. 1994. Smelting Mills in the Upper Dove Valley. Bulletin o f the Peak District Mines Historical Society 12,4: 12-13 Wild, M. and Eastwood, I. 1992. Soil contamination and smelting sites. IN Boles and Smeltmills (eds L. Willies and D. Cranstone) Historical Metallurgy Society, Peak District Mining Museum, Matlock Bath, Derbyshire), pp. 54-57. Wild, M.» Eastwood, I. and Newman, A.P. 1984. Background distribution o f some heavy metals in soils o f North East Derbyshire, England. Sheffield City Polytechnic Report No. 1/84. Willies, L. 1970. Regional geochemical reconnaisance of the Derbyshire area (Review). Bulletin o f the Peak District Mines Historical Society. 4,4: 337-8. Willies, L. 1990. Derbyshire lead smelting in the eighteenth and nineteenth centuries. Bulletin o f the Peak District Mines Historical Society. 11,1: 1-19. Willies, L. 1995 (in press). A Derbyshire lead mining landscape: washing floors at Winster Pitts. In Industry and Agriculture: Post-Medieval landscapes in Northern England. Report o f the conference held 31st o f March to the 2nd o f April at Durham. (Eds. D. Cranstone and L. Willies.) Historical Metallurgy Society and the Society for Post-Medieval Archaeology.

Appropriate Geophysics and Excavation Strategy: from Mud Brick to Masonry in the East Mediterranean Region C.A. Shell Department o f Archaeology, University o f Cambridge, Cambridge CB2 3DZ

Abstract Geophysical techniques have an important role to fulfill in the the overall strategy and planning for the excavation of an archaeological site, especially where significant logistics are involved. Not only does geophysical investigation reveal, in appropriate circumstances, the extent o f the near-surface survival o f a site, different techniques in combination can give insight into the nature and form of the surviving evidence. Knowledge o f the spatial layout o f a site, and often its internal structure, is, in itself, significant evidence. Geophysical survey also can considerably enhance the efficiency of excavation by pointing to where structural relationships may be explored and the most effective use made of available resources. As illustration, the strategy for detecting mud brick structures at the tells of £atalhoyuk and Oylum in Turkey using magnetic gradiometry is contrasted with the integrated survey of the Roman town at Knossos, Crete, where magnetometric, resistivity and topographic survey were employed in combination, followed by an experimental investigation with ground penetrating radar of areas selected from the results of these established techniques, prior to excavation.

Introduction In the development o f scientific approaches to archaeology, geophysical techniques have long been recognised as valuable contributors, but their full application to archae­ ological problems and their integration from the outset into the research design o f field projects involving excavation has lagged well behind the effort to develop and test new instrumentation. Theoretical physical models of relatively simple anomalies can be generated, but are seldom useful analogues for the complex structures and inter-relationship of features and stratigraphy encountered in the real archae­ ological world. Advances in understanding the origin of anomalies are being made, of which the discovery by Fafibinder and Stanjek (1993) of a bacterial magnetite from the decay of organic material in archaeological sites is a singular example. Geophysics in archaeology is now half a century old, and should be considered to have come of age with the publication of Scollar et a l (1990). The development o f both instrumentation and digital techniques for processing and presenting data can be expected to move on apace, but the integration o f geophysics into field research and the building o f a body of model case histories, where geophysics has played an integral role in both the research design and evolution o f a project, need also to advance. Rapid survey techniques and digital data logging now allow large area surveys to be undertaken, and pave the way for the complete study o f not just the structure o f small sites but also their immediate context. Geophysical survey can assist in the strategic planning of an excavation. Its integration within a project can provide for the geo­ physicist an immediate reification of the interpretation of the geophysical contribution and, more importantly, a measure of the applicability of the techniques employed for the type of site, the ground conditions, and nature o f the features associated with it. David (1995) makes a first

synthetic step in this direction in a review o f the selection o f geophysical techniques for archaeological field evaluation in the UK. In this paper are discussed the selection o f techniques and the degree o f their success in three surveys undertaken in Greece and Turkey, two o f which have been tested by subsequent excavation. The work forms part o f a larger assessm ent o f both established and novel geophysical techniques in an integrated approach.

Techniques The three surveys undertaken were each associated with a particular problem. The applicability o f the established techniques o f resistivity and magnetic gradiometry were assessed with commonly used equipment, the Geoscan Research RM15 resistivity meter in the twin-probe con­ figuration, and the Geoscan Research FM36 fluxgate vertical gradiometer. The gradiometer was in all cases used in its most sensitive mode, recording to 0.1 nT. This is limited, however, by a measured standard deviation o f no better than ± 0.3 nT because o f fluctuation in the stability o f the instrument’s electronics. Data were recorded for both instruments on survey grids based on 20 m square units. Gradiometer readings were taken at 0.25 metre intervals along transects 0.5 metres apart. Resistivity measurements were made at 1.0 metre intervals in both directions on the identical grid. D avid (1995) proposes that the maximum reading separation for a magnetic site evaluation survey should be a 0.5 m separation between readings on transects 1.0 m apart. From this study it clear that the reading separation should be no more than 0.25 m on transects 0.5 m apart if detailed fine structure o f a site is to be resolved. The differing physical origins o f the signals for complementary resistivity and fluxgate gradiometer

334

C. A. Shell

surveys broaden the basis for interpretating the anomalies encountered. The opportunity was also taken to assess ground probing radar on the site of Roman Knossos, using a GSSI SIR-10 radar with 200, 300 and 500 Mhz antennae, with a scan interval in the region o f 0.01 metres on 1.0 metre separated transects. For this technique there is always a trade­ off between the higher resolution of the radar profile obtained from using a higher frequency antenna and the depth to which the signal penetrates (Atkin and Milligan 1992). The ground and clim atic conditions o f the eastern Mediterranean dictate a more selective approach than might be necessary in more temperate countries for both resistivity and radar surveys. It is well established that resistivity anomalies may be detected in both waterlogged and very dry soil conditions (Hesse 1978), but optimal conditions for the best differentiation o f features may only occur in drier climates in a short period in Spring when differential drying enhances resistivity contrast. This was the case with the survey at Knossos, where this period occurred over three weeks in late March and early April. David (1993: 40) suggests the corresponding optimal conditions for resistivity survey in Britain are met in late Summer/early Autumn, but this can depend very much on both the overall weather pattern and the nature o f the soil. In 1995 on the free draining chalkland in Cambridgeshire, conditions by mid June were w ell past their best for resistivity survey. In contrast, magnetic anomalies detected by the fluxgate gradiometer are singularly insensitive to changing ground conditions and, undisturbed, remain consistent from one year to the next. For ground sounding radar, moist soil may reduce pen­ etration through attenuation of the signal by dissolved solute, but in dry ground the reflecting strength of the interface between features may be less well defined, as it depends inter alia on the difference in electrical capacitance across the interface.

The Site Surveys Surveys were undertaken at two tell sites in Turkey at £atalhoyiik, Konya province, and Oylum, Gaziantep province, and, in contrast, on the site of the Roman Town at Knossos, Crete. At Knossos the survey formed the basis for deciding the position of the initial trenches in the excavations that follow ed, whereas at (Jatalhoyiik the survey was integrated from the beginning into the overall project.

Qatalhdyiik Tell The neolithic tell of Qatalhoyiik near Qumra on the Konya plain o f Turkey is well known for its early wall paintings and elaborate shrines in abutting houses built from mud brick (Mellart 1975). The beginning o f the new Qatalhoyiik Project, under the direction of Ian Hodder, provided the opportunity to evaluate whether a fluxgate gradiometer could detect mud brick structures if they were present within the detection depth of the instrument. If successful, the survey would give an indication of the extent and form o f the buildings over a large area of the East Mound, and enable comparison with the houses excavated by Mellart on the west side of the tell. As part of the initial work on the site a

detailed contour survey o f the East Mound was linked with the surface clearance o f the topsoil in a set o f 10 metre squares that were enlarged if necessary, at 50 metre intervals over the mound to reveal the top o f surviving walls and the nature o f the infilling sediment (fig. 1). This gave an immediate, direct, comparison for checking the success of the magnetometry in detecting mud brick walls. The tell at £atalhoyuk consists of a succession o f mud brick structures partly infilled with demolished mud brick walling and midden material, prior to the building of a further house on these foundations. The mud brick was man­ ufactured from the Konya plain alluvium or site deposits derived from it. It is evident that not only would there be an insufficient resistivity difference between the mudbrick itself and the infilling material, especially in dry conditions, to make a resistivity survey worthwhile, but that there would be present very few interfaces of the type that would strongly reflect the beam o f ground penetrating radar. In contrast, the mud brick may be made from alluvium or an alluviumoccupation soil mixture that is different magnetically from the adjacent infilling material. Similarly, the occupation soil was often burnt or contained ashy material, and as such, was likely to vary from one side of a mudbrick wall to the other, and so define an interface visible in a magnetic survey where the wall itself may not be detected. This proved to be the case. Figure 2 is an overall view o f the magnetic survey. On the north side o f the main mound there are extensive amorphous regions that correspond to dumps o f ashy material from a later phase o f use o f the mound. These obscure any underlying walls (fig. 3). However, on the northern extension o f the mound there is significant patterning with some suggestion o f oriented sections of early walling (fig. 4). The impressions are nevertheless diffuse and difficult to define with a reading range that is no more than 4 to 6nT for a 0.5 nT instrument. There are small areas o f high signal that probably correspond to distinct burning o f the fill, but may in some cases mark the position of hearths or ovens o f the type Mellart found set into the walls o f rooms. It is here that the advantages of combining the magnetic survey with the surface clearance com e to the fore. Subsequent clearance o f the topsoil from the centre o f the northern area revealed a complex o f house structures and the first small lane or alley between houses to be discovered at £atalhdyiik (fig. 5). Superimposing this plan on the magnetic image (fig. 6) shows the clear correspondence between magnetic features and the walls and fills, some burnt, in the cleared area, allowing the reasonable inter­ pretation o f adjacent linear features as evidence for similar house structures, such as those abutting the scraped area to the west at (1015, 1185). The magnetic survey, whilst in many cases only able to hint at the presence o f house structures in several parts of the East Mound, did show clearly the very extensive later ashy deposits covering much o f the north side of the central part o f mound. It usefully complemented the technique of selective area clearance o f the topsoil, which made more certain the identification o f features in the magnetic survey. In a number o f areas a clear contrast was found in the degree o f burning o f the fills of adjacent houses. Conversely, the magnetic survey pointed to one or two areas, such as at (1055,1130), where the patterning was different, suggesting rectilinear mud brick structures o f unusually small size and form, suitable for further investigation.

Appropriate Geophysics and Excavation Strategy Oylum Tell Oylum Tell, one of the largest prehistoric mounds in Turkey, situated south o f Gaziantep near the border with Syria, is a multi-period site under investigation by Prof. Engin Ozgen, Director General of Monuments and Museums. At the southern summit of the mound, above the modern village o f mud brick houses, are a small number of dressed stone blocks o f local basalt that could indictate the presence o f a significant Hittite structure. At the invitation o f Prof Ozgen, a small fluxgate gradiometer survey was undertaken on the site. Whilst the mound consists principally o f weathered mud brick and occupation material o f low magnetic signature, the basalt blocks have a high thermo-remanent magnetisation from the time o f their solidification, associated with the magnetite and other iron minerals present in the rock. The orientation of the characteristic peak and trough signal of magnetised material for each boulder will depend entirely on the orientation in which each basalt block lies. As such they are in high contrast with the magnetic signal expected from the mud brick. No doubt the basalt blocks would also contrast resistively with the surrounding sediment, especially in damp ground, but the fluxgate gradiometer could be expected not only to detect their strong magnetic signal but also possibly discover mud-brick structures in the mag­ netically quiet areas away from the basalt. The magnetic survey revealed a concentration o f basalt blocks in the southern part of the survey area which tend to form a rectilinear arrangement. This is difficult to interpret in both the unprocessed and contrast stretched images (fig. 7a). It may define a platform area. The arrangement o f the boulders suggests their use as a footing for walls, but it remains a matter of debate as to whether they define a monumental structure. Further enhancement of the contrast in the magnetically quiet north west comer of the survey area pointed to the presence of basalt blocks at depth, the fields from which obscured any possible evidence for plain mud brick features. Contouring the magnetic image (fig. 7b) may give a better impression of the rectilinear structures. At the west side an arc of basalt boulders form a revetment to the top of the mound. The Oylum survey provides an interesting contrast with Qatalhoyiik, the signals from the small basalt boulders creating strong individual signatures which in close prox­ imity are difficult to interpret, but nevertheless appear to define the remains of a rectilinear structure. The randomly oriented high peak and trough signals o f the boulders also obscure any signal that may otherwise have marked out surviving mud brick features.

Roman Knossos To the north o f the Minoan Palace at Knossos in Crete lie the remains of the prosperous Roman town, which include, in an area o f the former Evans estate, evidence for mon­ umental buildings (Hood and Smyth 1981). The possible threat to use this land for tourist parking provided the reason for a geophysical survey of an area of 2.5 ha, giving the opportunity to investigate the value of combined magnetic and resistivity surveying, as well as to evaluate ground­ sounding radar’s ability to detect Roman masonry walls and related features. A detailed topographic survey (fig. 8) was also undertaken to assist the geophysical interpretation. The

335

survey revealed not only the north-south terracing in the field but also high concentrations of stone, a clearance cairn, and the outline o f a modern shallow irrigation system. The overall picture from the resistivity survey o f this area (fig. 9) shows large linear features which correspond to the position o f fragments o f Roman concrete wall in the field, with probably a roomed building to the east o f the centre of the surveyed area. In contrast, the fluxgate gradiometer survey o f the same area (fig. 10) shows finer detail with the major wall positions defined by characteristic linear peaktrough signals, which may result from chemical remanent magnetisation during the setting o f the cement used in their construction. The rectilinear pattern o f the roomed house is well defined. The grid o f the modern irrigation system is also detected with a reverse trough-peak signal in the northward direction due to the survival o f the shallow channels in this system. The value o f the detailed topographic survey is immediately apparent, allowing these features to be readily identified and allowed for in the interpretation of the other magnetic anomalies. At the north o f the survey area are a pair of major walls meeting at a right angle which on subsequent investigation by the Knossos 2000 Project proved to be one to two metres thick, and still extend to a depth o f five or more metres. The peak-trough magnetic signal for these walls is interrupted in places. This probably corresponds to where the wall has been robbed away but a rubble fill remains in the robber trench to give still a high resistivity signal. To the east of the northward wall are two strong magnetic anomalies that coincide with the interior spaces o f a possible two roomed building that is defined in the resistivity survey but is not visible magnetically. For interpretation the value o f integrating resistivity and fluxgate gradiometer surveys is best illustrated by features in the south-east of the northern area (fig. 11). Adjacent to a roomed structure is a three metre square ‘so lid ’ high resistance feature, which corresponds with an area o f low magnetic signal. Its interpretation could either be as a solid, low magnetic, block of limestone, perhaps a statue base, or a cistern still surviving as a hollow space. Subsequent excavation by the Knossos 2000 Project showed it to be the latter, probably receiving its water supply from the Acropolis hill just to the east. As a second stage in the project, ground-sounding radar surveys were carried out over interesting anomalies defined in the conventional geophysical surveys to test the tech­ nique’s ability to detect features in the prevailing limestone and lim estone-derived sedim ents. Sets o f orthogonal transects one metre apart were measured and the adjacent profiles compared to look for continuity in the position of strongly reflecting anomalies. A typical profile is shown in figure 12. Relatively few strong reflectors were detected, but here can be seen to correspond to high resistances, while showing little relation to the magnetic profile. The relation­ ship with the resistivity survey was further investigated by plotting the position o f the strong radar reflections on the resistivity survey for the roomed building (fig. 13). It can readily be seen that the reflectors correspond to the high resistivity positions, but are intermittent in defining them. This may again relate to the degree to which the original wall survives undamaged. Where it has, a strong radar reflection may be expected, whereas the rubble backfill o f a robber trench will tend to scatter the radar signal rather than strongly reflect it. The evaluation o f the radar results is still

336

C. A. Shell

at a preliminary stage, but it is clear by comparison with the evidence from the excavation o f this area, in which floors were recovered, that the radar was not able to define horizontal surfaces in the conditions which prevail at Knossos.

Discussion At £atalhoyiik the magnetic survey, using the fluxgate gradiometer close to its detection limits, revealed the nature of the near-surface evidence, and especially showed the extent of the later ashy dumps on the north side o f the centre of the mound. This area could be eliminated from the search for neolithic buildings surviving close to the surface. Mud brick walls were detected on the northern part of the mound, but the patterning o f structures was more obviously demonstrated by the variation in the degree o f burning, and hence magnetic enhancement, of the fills of adjacent rooms. The area clearance o f the topsoil and the magnetic survey in combination gave a much more detailed picture of the surface archaeology o f the mound. The magnetic survey at Oylum defined with difficulty the probable basalt footings of mud brick structures, but no direct evidence o f mud bricks walling could be extracted from the data because of the high signal associated with each basalt block. Nevertheless an area was defined which would warrant further investigation by excavation. The integrated survey at Knossos provided not only an enhanced understanding from the combination o f resistivity and magnetic data of the manner in which the roman concrete walls of the buildings survived, but also formed a clear basis for deciding where the trenches of the Knossos 2000 Project would give the maximum information in the excavations that followed. A ground plan o f the Roman town was recovered for this area, and the best locations could be selected for testing ground sounding radar on the site. Given the expense o f radar surveys, not least in the post-processing needed for the data, its use on large sites such as Roman Knossos must be to target particular features already defined by resistivity or magnetic survey.

Conclusion From this short paper it is hoped that the benefits to be gained from com bining both resistivity and fluxgate gradiometer survey are apparent, and that the selective use of a single technique can also provide significant evidence where a careful evaluation of the conditions suggest it may be worthwhile. Ground sounding radar needs further investigation, especially in ground conditions similar to those encountered at Knossos. Integration of geophysical surveying

into a larger project design for the study o f particular site is the only way to maximise the contribution geophysical techniques can make to field research and through this to the overall further development o f archaeology as a discipline.

Acknowledgements The surveys at £atalhoyuk were supported by funding from the ^atalhoyuk Trust and the University o f Cambridge, and I am grateful to Prof. Dr Engin Ozgen, Director General of Monuments and Museums, Ministry o f Culture, Turkey, for the opportunity to undertake the survey at Oylum, Gaziantep. Dr Rifat Ergeg, Director of the Gaziantep Archaeological Museum, kindly provided every assistance and hospitality during my visit. The K nossos survey was supported financially by the British School at Athens in advance o f the Knossos 2000 Project, the Society o f Antiquaries o f London Research Fund, and the University o f Cambridge, and undertaken in collaboration with GR-Geotech, directed by Prof. Stavros Papamarinopoulos of the University o f Patras. Am ongst the many who assisted at Knossos, Kathryn Roberts, Marina Papaiannou, N eil Brodie, Richard Jones and Mark Corney deserve particular mention for their contributions. Support funding is also gratefully ac­ knowledged from the McDonald Institute for Archaeological Research and the University o f Cambridge for the field evaluation o f new geophysical surveying techniques.

References Atkin, M. and Milligan, R. 1992. Ground-probing radar in archaeology - practicalities and problems. The Field Archae­ ologist, 16: 288-91. David, A. 1993. Some current applications of geophysical survey for the evaluation of archaeological sites in Britain. Archaeologia Polona, 31: 33-50. David, A. 1995. Geophysical survey in archaeological field evaluation. Research and Professional Services Guideline no. 1. London: English Heritage. Fassbinder, J. W. E. and Stanjek, H. 1993. Occurrence of bacterial magnetite in soils from archaeological sites. Archaeologia Polona, 31: 117-28 Hesse, A. 1978. Manuel de prospection geophysique appliquee a la reconnaissance archeologique. Centre de Recherches sur les Techniques Greco-Romaines, 8. Dijon: University de Dijon. Hood, S. and Smyth, D. 1981. Archaeological Survey o f the Knossos Area. British School at Athens Supplementary Vol. 14. London, Thames and Hudson. Mellart, J. 1975. The Neolithic o f the Near East. London: Thames and Hudson. Scollar, I., Tabbagh, A., Hesse, A. and Herzog, I. 1990. Archae­ ological Prospecting and Remote Sensing. Cambridge Uni­ versity Press: Cambridge.

Appropriate Geophysics and Excavation Strategy

337

CATALHOYUK (EAST MOUND)

1300

1250

1200

1150

1100-

1050

1000-

950

900 KEY

pH]

MeHaarfs trenches

850 Modem buildings :

|

Fence

|

scraped areas

800 850

900

950

1000

1050

1100

1150

1200

Contours at 0.5m intervals Figure 1. Contour plan o f the East Mound at (fatalhdyiik, showing the extent o f the fluxgate gradiometer survey (after plan by T. Pollard and R. Matthews)

Appropriate Geophysics and Excavation Strategy 12 10

339

r-----------",.....-:-.,-----------;--I'-----~~=_=___IIIInIlIr:~ ~---......._~-___,.

1190

"

4 METRES

o

1170 1020

1040

10

1060

Figure 5. Plan ofthe cleared area on the northern summit ofthe East Mound at Catalhoyiik, showing the extensive mud brick rooms and a passage way. Burnt areas are indicated by stippling (plan by R. Matthews).

1000

1010

1020

10:30

1040 metreseast

1050

1060

1070

Figure 6. Combined plan and fluxgate survey for the northern summit of the East Mound at Catalhiiytik, showing the correspondence between the magnetic features and house structures and their fills. The magnetic survey is displayed in grey scale contrast with -4 nT to +4 nT black to white range.

Appropriate Geophysics and Excavation Strategy

341 N

~

1020

I

1000

Figure 9. Northern part of the resistivity survey at Knossos showing both walling and major structures. Survey units 20 metres square. Contrast stretched image.

100

120

140

160

180

200

Figure 10. Northern section of the fluxgate gradiometer survey, showing finer detail, but with features less obvious than in the resistivity survey.

Impact of Resolution on the Interpretation of Archaeological Prospection Data A. Schmidt and A. Marshall University o f Bradford, Department o f A rchaeological Sciences, Bradford BD7 1DP, U.K.

A b stract Geophysical surveys are widely used for archaeological prospection. As interpretation o f results depends on the quality o f obtained data, the influence o f spatial resolution has to be investigated. Strategies are deduced from sampling theory to resolve anomalies associated with archaeological features. A comparison o f various field surveys with different grid dimensions demonstrates the importance o f sufficiently dense data recording.

In trod u ction Any excavation of archaeological sites will destroy the buried features under investigation. In addition, the threat to buried archaeological features is ever increasing (e.g. by extensive building and developm ent schem es). Therefore, non­ destructive and effective techniques are essential for the evaluation o f buried ancient remains. Due to the ongoing development o f lightweight and robust field-instruments, geophysical methods are nowadays widely used for such archaeological prospection. Data can be analysed so as to direct excavations towards focal points of archaeological interest thus minimising the costs, time and damage while retaining a maximum o f information. Likewise, a thorough geophysical survey may yield sufficient information to leave a site intact and preserve it for the future. The rapid cover of large areas aids the recognition o f buried archaeological sites in advance o f major building schem es and thus facilitates the preservation or rescue o f the national heritage. However, all this depends crucially on the extraction of archaeologically relevant information from geophysical data. Currently data are either displayed as a grey-scale image onto which a skilled interpreter can then draw the location of buried features, or it is subjected to extensive mathematical inversion schemes that may suffer from the underlying illconditioned geophysical problems and are very demanding and time consuming in terms o f computing power: “[with the final inversion system] it should be possible to process data as rapidly as it can be produced from the field survey [i.e. within days]” (Allum et al. 1995). In what ever way the interpretation is done it will depend on the quality o f recorded data. Only if field data are collected in accordance with the expected size and nature o f archae­ ological features can a reliable interpretation o f geophysical data be achieved. Therefore, a priori knowledge of the site is required to adjust geophysical surveys as necessary. However, since such a priori knowledge is often very difficult to obtain one must generally aim for a high quality o f collected data. The issue o f data quality from geophysical surveys has various aspects. It is important to measure the investigated geophysical properties (e.g. ground resistivity or the earth’s magnetic field) as accurately as feasible, taking natural

limitations into account. It is, therefore, o f great importance that the chosen techniques match the archaeological, geological and environmental conditions o f a site (see e.g. Clark 1990: 124). For each chosen technique the design of instruments has to be carefully adjusted to record data with appropriate precision. For example, Scollar et al. (1990: 440) discussed the magnetic noise that arises on archae­ ological sites. This has direct implications for the choice o f sensitivity at which a magnetometer shall be operated. Another aspect o f data quality arises directly from the nature o f the investigated subject. For buried archaeological features spatial information is of great importance and hence the location o f each measurement has to be recorded precisely. Sampling strategies have been investigated for random measurement positions. This work concentrates on data that have been recorded on regular grids, consisting o f lines with equally spaced recording stations. The implications o f grid dimensions (inter-line spacing and station separation) will be discussed. Relevant archaeological features may be missed or misinterpreted if inappropriate grid dimensions are chosen. The first section w ill summarise the relevant parts o f sampling theory while the second section will demonstrate the importance of high spatial resolution for selected field surveys. This can be seen as a contribution towards the discussion about best field practice that was initiated by the publication o f guidelines by English Heritage. It is stated that for magnetometry surveys “... the maximum acceptable separation between instrument readings is 1.0 m x 0.5 m ...” while for resistivity surveys “... the reading interval is standard at 1.0 m x 1.0 m ...” (David 1995: 18, 27).

Sampling Theory In order to estimate the necessary spatial resolution for geophysical surveys, information about the shape o f typical anomalies is required. Figure la show s the m agnetic anomalies o f a buried cube (1 m x 1 m x 1 m) as measured with a fluxgate gradiometer in Britain, traversing South to North. Each trace represents a different burial depth, starting at 0.25 m (biggest anomaly) and increasing for each trace by 0.25 m. The dependence of the peak width from the burial depth is shown in figure lb for cubes o f different sizes (side

344

A. S ch m id t a n d A. M a rsh a ll

length 0.5 m, 1 m and 2 m, respectively). It can be seen that the width as measured at half the peak height (□ ) as well as the width defined by the separation of southern and northern zero-crossing (A) are linear functions o f the depth. Archae­ ological features are mainly buried at shallow depths for which the half-width is approximately given by the size o f the cube (W idth/Size=l, see figure lb) while the zero-width can be estimated as twice the size (fig. lb). A rough estimate for the required sampling interval can be derived from a probabilistic approach (Florsch and Hulot 1995 and pers. comm.). It is based on the probability to “hit” the anomaly o f a buried feature with a single measurement. Assuming that a single measurement, recorded between the southern and northern zero crossing (i.e. within the zerowidth wz), can still be associated with the anomaly (which may in practise be very difficult) the probability p m for “hitting” the anomaly with a survey grid of spacing a is given by K

P'" ~

•

W*

4 ■a2

- 4 - 3 - 2 - 1 0

1

2

3

This probability is also referred to as “search effort” (Kyrala 1964). For the investigated cube (wz * 2 x size, see figure lb ) the anomaly will be “hit” by any grid with a spacing smaller than 1.8 times the size o f the buried object. Neither the condition o f “hitting” the anomaly with only one reading nor the identification o f an anomaly through a value that is recorded close to a zero crossing are suitable for a reasonable interpretation o f a magnetometer survey. Therefore, this estimate should only be regarded as an upper limit. A better estimate can be derived using the so called “sampling theorem”. Given a physical property (e.g. the earth’s magnetic field) that varies continuously in space, the sampling theorem states that the continuous change can be recovered from discrete readings (e.g. taken at grid stations) if the separation o f readings does not exceed the Nyquist wavelength XN = V2Xmin (Press et al. 1992: 500). Here Xmin is the smallest wavelength to be found in the spectrum o f the continuous data. If only the position o f the peak is sought for an anomaly the half-width o f the peak (wh~ size, see figure lb ) can be regarded as upper limit for !4Xmjn (the top o f the peak is considered to approximate a “half sine wave”). The sampling

4

P osition [m ]

D e p th /S iz e Figure 1, a & b

Figure 2, a & b

Impact o f R eso lu tio n

345

Figure 3

rate may, therefore, not exceed the size of the objects under investigation. However, it is desirable not only to recover the peak of an anomaly but to have some information about its shape. It has been shown (Schmidt and Sheen 1995) that the power spectrum o f gradiometer data can be described as F (X) oc exp ( - depth * 2 n / X) where X is the wavelength. It can then be calculated that for the recovery of the spectral component that contributes 4.3% of the maximal component the sampling rate may not be bigger than the depth at which the feature is buried. As a “rule-of-thumb” it may, therefore, be concluded that the sampling interval should neither exceed the size nor the depth of expected features. Since the depth is often given by the topsoil thickness (ca . 0.3 m) a sampling interval of 0.5 m x 0.5 m may be recommended. It is worth noting that these values were derived from the information content of recorded data. Hence, image processing (“filtering”) can not overcome these limitations.

Field Data The implications o f sampling theory, as outlined in the previous section, are best understood by investigating data from field surveys. Geophysical measurements have been recorded with a Geosean FM18 fluxgate gradiometer and an RM4/DL10 resistivity meter connected to a 0.5 m twin-

probe frame. These instruments are commonly used for field surveys and are recommended by English Heritage (David 1995: 17, 27). However, a smaller sampling interval o f 0.5 m x 0.5 m was chosen (the enclosure at Temple Guiting, fig. 4, was indeed sampled at 0.25 m x 0.25 m). Magnetometer measurements have been recorded with the instrument stationary, i.e. not walking. From these recorded data subsets at wider intervals have been extracted for comparison. All data were displayed using the Contors computer program (Cheetham eta l. 1989) with bi-cubic interpolation between recorded readings and without further data treatment (e.g. no filtering). All surveys are aligned such that North is at the top o f the display. At Hazleton, Gloucestershire, two middle N eolithic chambered tombs (long barrows) have been found and partially excavated (Saville 1990). A 0.5 m x 0.5 m resistivity survey was carried out over the nearly undisturbed southern mound. Figure 2a shows the results, indicating kidney shaped quarries at the NE and SW o f the barrow as well as a chamber in the middle (dark area). A narrow semi-circular feature crosses the mound in the North. It coincides with the boundary o f a grassy area and is probably related to modem ploughing. Of great interest is the rectangular structure at the NW of the mound indicating a forecourt. This is a clear indication o f human activity connected with the burial mound. The important archaeological implications o f this

346

A. S c h m id t a n d A. M a rsh a ll

0

10

20m

Figure 4 find will be published elsewhere. A comparison with data from a 1 m x 1 m grid (fig. 2b) is very instructive. While bigger features (e.g. the quarries and the chamber) are still visible, the plough marks and the forecourt can not be resolved. This clearly demonstrates how a coarse sampling strategy may miss archaeologically relevant features. The survey in figure 2b does not even indicate that parts o f the site should be investigated with closer station spacing. Similar conclusions can be drawn from the magnetometer survey of a mid Iron Age enclosed farmstead at Guiting Power, Gloucestershire (The Park, fig. 3). Obviously, the high resolution data (0.5 m x 0.5 m, fig. 3a) are much crisper than the coarser data (1 m x 1 m, fig. 3b). This assists in the archaeological interpretation o f the site since features can be identified more easily. In addition, certain features are missing in the coarser survey (e.g. the linear feature running SE in the southern part, fig. 3a[l], a rectangular feature east of the main ditch, fig. 3a[2] and a curved feature in the NW part, fig. 3a[3]). For the archaeological assessment o f the site as a whole this additional information may well be of importance. The interior o f an Iron Age ditched enclosure at Middle Ground, Temple Guiting, Gloucestershire (fig. 4), was surveyed with a gradiometer at 0.25 m x 0.25 m intervals

(fig. 4a). The whole site was additionally surveyed at 0.5 m x 0.5 m (fig. 4c). Comparing this with data from a lm x lm grid it is clear that only at a resolution o f 0.5 m x 0.5 m can the pits in the middle and SE be undoubtedly identified as such. The highest resolution (0.25 m x 0.25 m, fig. 4a) reveals further features that can otherwise not be seen (e.g. linear feature connecting pits, fig. 4 a [l]). For this survey the effect o f non-isotropic sampling was investigated (0.25 m x 1 m, fig. 4b). Operating a Geoscan fiuxgate gradiometer in its “walking mode” allows to retain an inter-line spacing of lm while taking samples at station separations o f only 0.25 m. This is recommended by English Heritage (David 1995: 18) since survey speed and high resolution - at least in one direction - can be combined. In figure 4b the outcome of such survey was simulated with data extracted from the original survey, using a linear interpolation between adjacent lines (this can be improved by using more sophisticated schemes like bi-cubic spline interpolation). A comparison shows that the result is much more like the 1 m x 1 m survey (fig. 4d) than the 0.25 m x 0.25 m survey (fig. 4a). Even the data recorded at 0.5 m x 0.5 m (fig. 4c) are better defined. Data recorded under the recommended field conditions (i.e. recording while walking instead o f stationary measurements as in figure 4) will suffer from poorer positioning accuracy

Im p a c t o f R eso lu tio n

347

Figure 5

and will, therefore, show worse results. At a site of an Iron Age ditched enclosure near Guiting Power, Gloucestershire (The Bowsings, fig. 5), an area was subjected to repeated magnetic measurements under various recording and surveying conditions (fig. 5a to 5e). While figures 5a, 5b and 5c were recorded with 0.5 m x 0.5 m station separation, figures 5d and 5e were measured at 1 m x 1 m intervals. For figure 5a, 64 readings were averaged by the instrument at each position to obtain a measurement while for figures 5b to 5e only one reading was taken into account. For figures 5c and 5e data were recorded while walking along the lines. It can be seen that the data measured with the highest effort (fig. 5a; 0.5 m x 0.5 m, average of 64 readings, stationary recording) show the tentative hut circle in the NW comer most clearly. It is interesting to note that the change of station separation (figs 5a to 5c versus 5d and 5e) has a much greater influence on data quality than any change of the recording technique (figs 5a to 5c).

Conclusion It was shown that high spatial resolution is o f great importance for the quality of geophysical survey data. This can be seen from a theoretical point of view considering the information content o f discretely sampled data as well as from a practical point o f view comparing survey results obtained at different sampling intervals. It is obvious that mathematical inversion schemes work better if data quality is high. For manual interpretation, however, it may be argued that a skilled interpreter may be able to cope with data that

have been recorded on a coarse grid. It was demonstrated in this article that coarse sampling, although revealing wide ditches and massive foundations, may miss other features important for archaeological interpretation. It is often assumed that a fast and coarse sampling may indicate areas of interest for subsequent measurements at closer intervals. Survey results from Hazleton (fig. 2) showed that this is not necessarily the case: the important forecourt would have been missed by a 1 m x 1 m survey without any such indication. It should also be noted that archaeological features can often be detected due to their spatial extension. A linear sequence of high or low readings may suggest a ditch or foundation while a patch o f distinct readings may indicate an assembly o f pits. In order to clarify such structures sufficient spatially related measurements are required. A single reading for a feature is not enough. O bviously, the demand for high spatial resolution conflicts with the time pressure under which many surveys have to be carried out. Therefore, the trade-off between survey effort and possible archaeological interpretation has to be considered carefully. There may be cases in which the coverage o f a large area seems to be o f greater significance than a high spatial resolution. However, it is important to be aware o f the serious consequences a coarse sampling may have for the analysis o f a site. Solutions may be expected from improved instrumentation. With a recently introduced multiplexing system for Geoscan resistivity instruments several adjacent measurements can be recorded at a time. The successful use o f arrays o f five fluxgate gradiometers

348

A. S ch m id t a n d A. M a rsh a ll

with a separation of 0.2 m has been reported by Panitzki et a l (1995). This indicates that more high resolution data may be available in the future that were still recorded with a good survey rate.

Acknowledgements Parts of the calculations o f this work (fig. la) were performed using the MODELLER2 program, written by N. Sheen (University o f Bradford, Department of Archaeological Sciences and Geophysical Surveys of Bradford). The authors highly appreciate discussions with Dr A. Aspinall and P.N. Cheetham (University o f Bradford, Department of Archae­ ological Sciences) which have had an important impact on the final article. The comments and valuable information from Dr A. David and P. Linford (English Heritage) were a great help.

References Allum, G.T., Aykroyd, R.G. and Haigh, J.G.B. 1995. Restoration of magnetometry data using inverse-data methods. In Pro­ ceedings o f Computer Applications and Quantitative Methods in Archaeology, Leiden, April 1995. in press. Cheetham, P.N., Haigh, J.G.B. and Ipson, S.S. 1989. The archaeological perception of geophysical data. In Archae­

ological Sciences 1989 (eds P. Budd, B. Chapman, C. Jackson, R. Janaway and B. Ottaway). Oxford, Oxbow Books. Clark A.J. 1990. Seeing Beneath the Soil, Prospecting methods in archaeology. London, Batsford. David, A. 1995. Geophysical survey in archaeological field evaluation. English Heritage Research and Professional Services Guideline No 1. Florsch, N. and Hulot, C. 1995. Spatial aliasing errors in archaeological magnetic prospection. 1 Annales Geophysicae. 13, Suppl. I: C91. Kyrala, A. 1964. Basic probabilistic methods in geological search. Geophysics. 29: 105-108. Panitzki, M., Graber, M. and Stiimpel, H. 1995. Geoelectric and geomagnetic prospecting of an archaeological site in Kusakli (Turkey). Annales Geophysicae. 13, Suppl. I: C94. Press, W.H., Teukolsky, S.A., Vetterling, W.T. and Flannery B.P. 1992. Numerical Recipes in C, 2nd edition. Cambridge, Cambridge University Press. Saville, A. 1990. Hazleton North, Gloucestershire, 1979-82: The excavation of a Neolithic long cairn of the Cotswold-Severn group. English Heritage Archaeological Report no 13, Historic Buildings and Monuments Commission for England. Schmidt, A. and Sheen, N. 1995. Depth Estimation of Archaeolog­ ical Features by Power Spectrum Calculations of Geomagnetic Surveys. In Proceedings o f Archaeometry '94 (ed. A. M. Ozer), Ankara, May 1994. in press. Scollar, I., Tabbagh, A., Hesse, A. and Herzog, 1.1990. Archaeolog­ ical prospecting and remote sensing. Cambridge, Cambridge University Press.

Age Determination via Measurement of Amino Acid Racemization in Dental Tissue Vikki A. Carolan1, Michael L.G. Gardner2 and A. Mark Pollard3 1 School o f Science and Mathematics, Sheffield Hallam University, Sheffield, SI 1WB, UK; 2 Department o f Biomedical Sciences, University o f Bradford, Bradford, BD7 1DP, UK 3 Department o f Archaeological Sciences, University o f Bradford, Bradford, BD7 1DP, UK.

Abstract Study o f in vivo amino acid racemization in dental tissue provides a method o f age determination based upon chemical measurement. A procedure employing reversed-phase high performance liquid chromatography (RPHPLC) has been developed to allow the separation and quantification o f various derivatized amino acid enantiomers, obtained from dental collagen. This technique has been developed using modem samples and is to be applied to ancient samples. Data obtained and problems encountered are presented.

Introduction It is well known that in metabolically stable proteins, such as those found in the eye lens (Masters 1977), in dental tissue (Helfman and Bada 1975; 1976), in certain parts of the brain (Man 1983) and in vertebral discs (Ritz and Schlitz 1993), a slow amino acid racemization reaction takes place in vivo. This reaction proceeds throughout the lifetime of the host (and usually at a reduced rate after death as a result o f reduced temperature) and hence, in recent samples at least, the degree o f racemization can be related to the chronological age o f the host. This relationship has application to both forensic (e.g. Ogino et al. 1985; Ogino and Ogino 1988; Ritz e t a l 1990; 1993) and archaeological (Masters and Zimmerman 1978; Shimoyama and Harada 1984; Gillard et a l 1991) ageing problems, although the latter requires a separate estimation o f the post-mortem contribution. Since Helfman and Bada (1975) first reported the study o f amino acid racemization in the tooth enamel o f living humans, there have been numerous investigations of the racemization process and the techniques required to measure it in many matrices, including human dental tissue. On reviewing the literature on the study of amino acid race­ mization in dental tissue it has become apparent that no uniform procedure has been utilised by all research groups, and that the methods of sampling the specimens, and the analytical methods themselves, can affect the data obtained. Researchers have studied amino acids derived from all tooth types: this requires careful consideration of dental development to obtain accurate estimates o f dental age. The sampling procedures have also varied greatly: Ohtani et a l (1988) and Ohtani and Yamamoto (1991; 1992), for example, have studied whole teeth; Helfman and Bada (1976), Ogino etal. (1985), Ogino and Ogino (1988), Masters (1983; 1986) and Ritz et al. (1990; 1993) have looked mainly at crown dentine; Shimoyama and Harada (1984), Ritz et al. (1993) and Mdmstad et al. (1994) have looked at root dentine; Helfman and Bada (1975) and Ohtani and Yamamoto (1992) have studied enamel; and Gillard et al. (1990) and Saleh et

al. (1993) have taken specific sections o f a tooth e.g. vertical sections, horizontal sections etc. As teeth form from crown to root over a number of years more accurate results may be obtained by study o f a specific section o f the tooth, rather than obtaining an average value from a tooth which may have taken 5 -8 years to develop. Different fractions of the dental material have also been studied: initial work concentrated on the ‘collagen’ fraction which was obtained by acid hydrolysis of the complete dental sample. This more probably contains both collagenous and non-collagenous proteins. Other studies have looked at specific fractions o f the dental tissue. For example, Masters (1985) looked at racemization rates in the non-collagenous phosphoprotein fraction, and postulated that amino acids in this fraction racemize at a faster rate than collagen amino acids. Studies o f bovine incisors (Takagi et al. 1988) have indicated that there may be a difference in the relative amounts o f non-collagenous and collagenous protein in the root and crown o f the teeth. If this is similar for human teeth this may also affect the results obtained. The present work has looked at the fraction obtained by acid hydrolysis o f the demineralized tooth sample. This should contain primarily acid insoluble collagen, but may also contain non-colla­ genous proteins. In the literature, hydrolysis has usually been carried out using 6M HC1, at 1 0 0 -1 10°C for 6 or 24 hours. This has been deemed to give either complete (24hrs) or partial (6hrs) hydrolysis. In this study we have investigated a complete range o f hydrolysis times with respect to release o f amino acids and introduction o f induced racemization and shown that there is little change in amino acid release between 6 and 24 hours hydrolysis. Induced racemization does however increase over that time. In addition to studying each step o f the sampling, demineralization and hydrolysis procedures to optimise and standardize the methods employed, we have also investigated various methods of quantifying the amino acid enantiomers released from the dental tissue. Early studies (e.g. Masters 1983) generally desalted the amino acid hydrolysate on an

350

V ikki A. Carolan, M ic h a e l L. G. G a rd n er a n d A. M a rk P o lla rd

ion exchange column, often then separating aspartic acid from the mixture by anion exchange, derivatizing it using Lleucine-A-carboxyanhydride and performing the analysis on a traditional amino acid analyser. Later work (e.g. Shimoyama and Harada 1984; Ritz e ta l 1990; 1993) has primarily utilised esterification and acetylation using trifluoroacetic anhydride to derivatize the amino acid samples, followed by analysis using chiral gas chromatography (GC). This requires several manipulation steps, derivatization and expensive chiral GC columns and equipment. Only a few investigations utilising separation by liquid chromatographic techniques have been reported (e.g. Gillard et al. 1990; 1991 and Mornstad et al. 1994). Gillard etal. (1990; 1991) separated aspartic acid from the total acid hydrolysate by ion exchange, before derivatization with L-leucine-iV-carboxyanhydride, follow ed by separation on a C l 8 RP-HPLC colum n. Mornstad et al. (1994) derivatized the complete hydrolysis mixture with a chiral fluorescent reagent (O-phthaldialdehyde / iV-acetyl-L-cysteine) to allow detection using an HPLC system equipped with a fluorescence detector, but only separated the enantiomers of aspartic acid. In this work, we have developed a method based upon RP-HPLC to allow the total or partial separation of amino acids in a complete acid hydrolysate, facilitating simultaneous measurement of the enantiomeric ratio o f several amino acids. The method, based upon that o f Bruckner et al. (1991), utilises a chiral, fluorescent derivatizing reagent (0-phthaldiaidehyde / Nisobutyryl-D-cysteine) to form stereoisomeric dipeptides, which are then separated on a C18 RP-HPLC column using gradient elution.

Materials and Methods M aterials All reagents, acids, solvents etc. used were high analytical grade. HPLC grade water was produced using a Millipore Alpha Q water purification system (Millipore (UK) Ltd, Watford, Herts, UK). Amino acids used for preparation of standard solutions were obtained from Sigma Chemicals Ltd. (Poole, Dorset, UK). Standard solutions of individual amino acids were prepared by dissolution, in water, o f the appropriate mass o f amino acid to give a 0.02M solution. These individual solutions were then used to make up mixtures containing L-amino acids, D-amino acids and Dand L- amino acids at various enantiomeric ratios. Figure 1 gives details o f the teeth studied. These were all freshly extracted, upper 1st premolars, stored at -18°C for up to one year prior to analysis. Where possible, non-carious teeth were used; carious or filled teeth were sampled carefully to avoid possible contamination from bacteria or inclusion of secondary dentine. Any adhering organic matter was removed from the surface of the teeth; they were then soaked in sodium hypochlorite solution (~13% free chlorine) for 30 minutes, followed by thorough rinsing in HPLC grade water.

M ethods D etails o f sam pling, dem ineralization and hydrolysis methods are discussed at length in ‘results - method development’.

Ch romatog raphy: Reverse-phase high performance liquid chromatographic

(RP-HPLC) analysis was achieved on a modular Perkin Elmer HPLC system. It comprises: PE Series 410 pump, capable o f delivering a gradient mobile phase; PE ISS 200 autosampler, capable o f auto-derivatization and auto­ injection; PE LC oven 101, for temperature control o f the column; PE diode array 135 detector, operating in single channel mode at a wavelength o f 340nm; and a PE Nelson 1020 integrator for data acquisition, storage and mani­ pulation. Column Conditions: a 250mm x 4mm column was used, packed with Hypersil ODS (5pm) preceded by a 10mm x 4mm guard column packed with the same material (Shandon Scientific, Runcorn, UK). The columns were thermostatted to 25°C. Sample Number

Age of Donor

Condition

E180 224(1) 224(2) 57 88(1) 88(2) 65 67 E078 E083 E172 E212 E194 15 E056 64 E003 E191 68 121 150 115 14

11 11 11 13 15 15 18 18 23 25 29 30 35 38 43 46 47 47 49 60 62 69 72

Non-carious Non-carious Non-carious Non-carious Non-carious Non-carious Non-carious Non-carious Small filling Small caries Small caries Small caries Decay / filling Small caries Decay / filling Non-carious Small caries Small filling Non-carious Small filling Small caries Small filling Decay / filling

.

Figure 1. Details o f teeth used in production o f age versus %D-amino acid calibration curves

Mobile Phase: Eluent A was prepared by dissolution o f 6.26g sodium acetate trihydrate (HPLC Grade, Fisons, Lough­ borough, UK) in 1980 ml HPLC Grade water (Alpha Q Water System, Millipore, UK). The pH was adjusted to pH 6.2 by addition o f 10% (v/v) acetic acid (Fisons, Lough­ borough, UK) and the solution made up to 2 litres with water. The eluent was filtered through a 0.45pm filter (Millipore, UK). Eluent B was prepared by the addition of 600ml methanol (HPLC Grade, Fisons, Loughborough, UK) to 50 ml acetonitrile (HPLC Grade, Fisons, Loughborough, UK). Helium was constantly passed through the eluents to facilitate degassing. A linear gradient was applied for 80 minutes, at a flow rate o f Iml/min, from 0 to 57% B. Automated sample derivatization: derivatization o f the samples was carried out according to B r u c k n e r ^ /. (1991). Briefly: 260mM N-isobuty ry1-L or D-cysteine (IBLC l IBDC) (Fluka Chemicals, Switzerland) and 170mM O-phthaldialdehyde (OPA) (Sigma Chemicals, Poole, UK) in 1M potassium borate buffer, pH 10.4 (Fluoraldehyde reagent, Pierce, Rockford, IL, USA) were used as the derivatizing reagent (designated IBLC/OPA or IBDC/OPA). Using the

A g e D eterm in a tio n via M ea su rem en t o f A m in o A c i d R a cem iza tio n

derivatizing program of the ISS 200 autosampler: 25pl 0.4M sodium borate buffer, pH 10.4 (Hewlett-Packard, Waldbronn, Germany) was placed in a micro vial, 5pl IBLC/OPA or IBDC/OPA was added, followed by 10pl of amino acid sample. The autosampler then mixed the reagents twice and allowed the reaction to stand for two minutes. A 20pl aliquot was then injected onto the column.

Results M ethod D evelopm ent In the current work, the analytical methods employed in the measurement o f amino acid racemization have been studied and modified in order to improve the reliability o f the technique. Areas targeted initially for improvement included: development of more precise sampling procedures, taking into account dental physiology and archaeological pre­ servation; improvements and greater standardization o f demineralization, collagen extraction and hydrolysis pro­ cedures, to yield sam ples show ing a consistent, and minimized, degree of induced racemization; development of a relatively simple analysis technique, based upon RPHPLC, to allow simultaneous measurement of the en­ antiomeric ratio o f several amino acids, allowing a degree of internal calibration of the method and the possibility of estimation of the post-mortem contribution to racemization. Alternative methods of measuring enantiomeric ratios of amino acids employing stereospecific amino acid oxidases and a combined 2D-TLC/HPLC technique have also been investigated.

351

through the co-operation of a number o f local dentists. So far this has allowed the analysis in triplicate o f 23 upper 1st premolars, with a donor age range of 11-72 years.

Improvements to Sample Preparation Procedure and Minimization o f Induced Racemization There are two steps in the procedure o f obtaining amino acids from dental samples: demineralization to remove enamel and mineralized dentine and reveal the collagen template; and hydrolysis to break down the collagen proteins into their constituent amino acids. Both steps have the potential to induce artefactual racemization o f amino acids. The demineralization procedure has been studied, in some detail, with respect to the time taken, agitation o f sample and solution, use o f HC1 or EDTA and the effect o f pre­ drying the samples. Pre-drying of samples allows an accurate measurement of the weight loss during demineralization and may also promote more efficient demineralization. By continuous rolling o f the sample with the demineralization reagents a more rapid and uniform initial demineralization is achieved, this is illustrated in figure 2. A weight loss o f about 80%, representative o f complete demineralization is observed after 48 hours continuous rolling in 1M HC1. In order to ensure complete demineralization o f the horizontal section taken as a sample from the tooth, the whole tooth is

Development o f Sampling Procedures After study of the literature, it was concluded that using the same type of tooth throughout will reduce errors in estimation o f the exact age o f dentine. Considerations o f dental physiology and typical archaeological dentitions led to the conclusion that the use of upper (left and right) first pre­ molars is most appropriate. This limits differences due to different dentine developm ent rates, blood supply and variable external temperature fluctuations. The effect o f caries was also considered: due to the possibility o f contam ination by amino acids from bacteria and the formation of secondary dentine at carious sites, it was decided to avoid carious samples where possible. Where it was impossible to avoid the use of carious or filled teeth, sampling was carried out with extreme care to avoid inclusion o f any contaminated material. As dentine is formed from crown to root in a series of concentric cones, over a number of years, it is expected that an age gradient should be observed from crown to root. Preliminary investigations, taking 1mm horizontal sections from crown to root o f a single tooth, after removing the secondary dentine, gave no conclusive evidence for a gradient. A more refined sampling procedure (such as that employed by Saleh et a l 1993) may be required to reveal this gradient. It was decided, therefore, that rather than using a whole tooth, a horizontal slice ~lm m thick would be taken at approximately the crown/root junction, just above the pulp chamber. Enamel is removed by a reproducible demineralization method (see below), and any pulp or secondary dentine present is removed using a scalpel and tweezers, prior to hydrolysis. A large collection of freshly extracted human teeth o f known age has been built up

— ■— —

0.5MEDTA: static

— 0.5M EDTA; with rolling

----®—

1M HC1; with rolling

—

1MHC1: static

Figure 2. Plot o f demineralization time versus weight loss, to show the effect o f using 0.5M EDTA or 1M HCl to demin­ eralize tooth samples. The effect o f rolling the samples with the demineralizing reagents is also illustrated.

352

Vikki A. C arolan, M ich a e l L. G. G a rd n er a n d A. M a rk P o lla rd

demineralized for 36 hours in 20ml 1M HC1, leaving it soft enough for accurate sectioning, and the horizontal section taken is then demineralized for a further 12 hours in 20ml of fresh 1M HCL. After dem ineralization the remaining collagen template is dried to constant weight. The use o f 0.5M EDTA as a demineralizing agent was also studied. Even with continuous agitation of the sample with the demineralizing reagent, it was shown that more than seven days were required for complete demineralization (fig. 2). As the amount of induced racemization observed was no less than that when 1M HC1 is used, EDTA was not employed in the demineralization process. Hydrolysis is carried out in 500pl, 6M HC1, in evacuated sealed tubes, at 110°C. The optimum hydrolysis time was investigated with respect to the proportion o f amino acids released and the degree of induced racemization. Dental samples and standard L-amino acid solutions were hy­ drolysed for 1, 3, 6, 9, 12, 15, 18, 21 and 24 hours. It was shown both by HPLC and amino acid analysis that the amount o f most amino acids released from the dental collagen increased significantly from 1 to 6 hours hydrolysis, but did not increase significantly between 6 and 24 hours hydrolysis. Figure 3a shows the amount o f aspartic acid, serine and glutamic acid released (in nanomoles) from 200 nanomoles o f dental tissue (moles calculated from wet mass and average molecular weight of 100 amu), after various hydrolysis times, measured by amino acid analysis. As certain amino acids may be released preferentially, a standard hydrolysis time is necessary to account for this possibility. The amount of induced racemization was also studied, by measurement of the %D-enantiomer formed, after hydrolysis, in a standard L-amino acid solution. This was shown to increase, almost linearly, with increasing hydrolysis time (fig 3b), therefore a relatively short hydrolysis time was preferable. HPLC analyses o f the various derivatized hydrolysates showed extra small peaks from those samples hydrolysed for short periods o f time. Although these were

not identified categorically, it is thought they may be attributable to small peptides, which have not been fully broken down into their amino acid constituents. Taking into account amino acid release, induced racemization and HPLC analyses o f the samples an optimum hydrolysis time of 9 hours was chosen.

Development o f a Method to Measure the Enantiomeric Ratio o f Several Dental Amino Acids Previous methods of measuring amino acid racemization have mostly involved separation of the amino acid o f interest (usually aspartic acid) by ion exchange, follow ed by derivatization and analysis by either GC or HPLC (Carolan et al. 1996). Although these techniques have been reasonably successful it is postulated that the ability to measure several amino acid ratios from the same sample will enhance the reliability o f the estimate o f age at death given by the technique. There are various methods available to separate en­ antiomeric mixtures o f amino acids: these include the use of chiral columns, in both HPLC and GC; the addition o f chiral components to the m obile phase in an LC system; or derivatization with a chiral reagent followed by separation by either GC or HPLC, utilising a set o f sophisticated separation conditions. The use o f chiral additives in the mobile phase is often a laborious and expensive method of separation. Additionally, these systems do not tend to give optimum separation of a full range o f amino acids, but are successful for just one or two pairs. The same is largely true o f the chiral columns available from many manufacturers: although excellent separation o f a single pair o f enantiomers is achieved conditions do not tend to exist for the simultan­ eous separation o f several pairs o f amino acid enantiomers. Following extensive investigations o f the chiral eluents and columns available, methods utilising chiral derivatization follow ed by reversed-phase HPLC using a non-chiral, standard column were investigated.

Hydrolysis Time (hours)

Figure 3a. Plot o f the amount o f aspartic acid, serine and glutamic acid released (in nanomoles) from 200 nanomoles o f dental tissue (moles calculatedfrom wet mass and average molecular weight o f 100 amu), after various hydrolysis times, measured by amino acid analysis.

Figure 3b. Plot o f %D-aspartic acid versus hydrolysis time on degree o f induced racemization in a standard solution (0.02M) o f L-aspartic acid.

A g e D eterm in a tio n via M ea su rem en t o f A m in o A c i d R a cem iza tio n

353

29

minutes Figure 4. Chromatogram o f standard racemic acid solution, derivatized with OPA/IBDC, separated on a C l8 RP-HPLC column. Chromatographic conditions detailed in 'methods \ 1, D-Asp; 2, L-Asp; 3, D-Glu; 4, L-Glu; 5, D-Ser; 6, L-Ser; 7, D-Thur; 8, Gly; 9, 10, D-His, L-Thr; 11, L-His; 12, D-Ala; 13, D-Arg; 14, L-Arg; 15, L-Ala; 16, D-Tyr; 17, H-Arg; 18, L-Tyr; 19, D-Val; 20, D-Met; 21, L-Met; 22, L-Val; 23, D-Phe; 24, D-Ile; 25, L-Phe; 26, D-Leu; 27, L-lle; 28, L-Leu; 29, D-Lys; 30, L-Lys.

The method o f Nimura et al. (1980), utilising glucopyranosyl isothiocyanate (GITC) as the chiral dedvatizing reagent, was studied at length. This employed a standard C l 8 column, combined with an elaborate gradient elution system using phosphate buffer and methanol mobile phases. Separation of several amino acid standard pairs was achieved and the method was shown to give accurate reliable results for the measurement o f a small amount o f the D-amino acid enantiomer in an excess of the L-enantiomer. Unfortunately, problems were experienced with co-elution o f amino acids o f particular interest e.g. threonine and aspartic acid enantiomers. Although this rendered the method o f little use in the separation required it indicated the possible inter­ ference of threonine with aspartic acid which is likely to have contributed to some o f the anomalous results obtained in previous studies. A second method o f derivatizing amino acids with a chiral reagent was then investigated, involving the use of 0-phthaldialdehyde and AMsobuty ry1-L or D-cysteine (OPA/ IBLC or OPA/IBDC) as the derivatizing reagent and a simpler gradient mobile phase o f methanol/acetonitrile and sodium acetate buffer. The m ethod had been shown previously to be applicable to complicated amino acid mixtures such as those from food samples (Bruckner et al. 1991). A fluorescence detector is usually utilised in the experiment but because of the relatively large samples available from the collagenous proteins a U V detector provided adequate sensitivity. (The use of a fluorescence

detector was investigated and did show much greater sensitivity and associated accuracy. This detector may be necessary, therefore, when sm aller sam ples are being investigated, for example from non-collagenous proteins etc.) Conditions for the separation o f mixtures o f standard amino acids were optimised to give total or partial separation o f the full range o f amino acids in 80 m inutes (see ‘methods’). A typical chromatogram of a mixture o f standard racemic amino acids, derivatized with OPA/IBDC is shown in figure 4. The availability o f the two enantiomers o f the derivatizing agent i.e. L- and D- isobutyryl cysteine, allows the choice o f elution order o f the amino acid diastereomers produced in the reaction. It was found that utilising the D-enantiomer o f A-isobutyryl cysteine, to cause the D-amino acid enantiomers to elute prior to the L-amino acid enantiomers gave the most accurate results: the elution of the small D-peak prior to the large L-peak ensured that the D-peak was not ‘lost’ in any tailing o f the L-peak. Separation of standard mixtures o f amino acids containing 50:50 mixtures o f L- and D- enantiomers was facilitated with relative ease. However, dental samples are more likely to have a ratio o f less than 5% D- to 95% L-. Consequently studies o f the reproducibility and accuracy of the measure­ ment o f small amounts o f D-enantiomers in an excess o f the L-enantiomer were carried out and shown to be perfectly adequate: a standard amino acid mixture containing 99:1 L:D Asp was derivatized (O PA /IBD C ) and analysed

354

V ikki A. C arolan, M ich a el L. G. G a rd n er a n d A. M a rk P o lla rd

repeatedly, giving the following data: (n=6) mean %D-Asp = 1.18; cv = 2.91 %. A similar experiment was carried out on a dental hydrolysate, giving the following data: (n=5) mean %D-Asp = 3.13; cv = 4.60%. The derivatization method utilising OPA/IBDC was then applied to dental collagen hydrolysates and the chro­ matographic conditions optimised to give separation o f as many amino acids as possible. Problems with co-elution o f some enantiomers do occur, but, it is currently possible to measure simultaneously the enantiomeric ratio of (at least) three amino acids, namely aspartic acid, serine and glutamic acid. Hydrolysates produced from the tooth samples detailed in figure 1 were analysed using this method. A typical chromatogram is shown in figure 5. Measurement of peak areas for the D- and L- enantiomers o f aspartic acid, glutamic acid and serine allowed the calculation of the percentages of D- and L-amino acids for each o f the tooth samples detailed in figure 1. These percentages were used in the production o f three comp­ limentary calibration curves relating %D-amino acid to age (figure 6a-c). A linear (or near linear) relationship is seen in all cases, but the increase in %D-glutamic acid is minimal over the range investigated. This is to be expected due to the slower rate o f racemization o f glutamate. Both aspartic acid and serine show an increase in %D of several percent, over the age range studied. In addition to the analysis of collagen hydrolysates, the derivatization and analysis technique has been applied to the residual 1M HC1 solutions used to demineralize the tooth samples. Preliminary investigations indicate that there are amino acids present in this mixture, in measurable quantities.

However, it will be necessary to utilise a fluorescence detector in further analyses, to increase the sensitivity o f the technique sufficiently to allow accurate measurement o f these small amounts.

Discussion Age determination from measurement o f racemization of aspartic acid in dental tissue has been shown to be a valuable technique in the field o f forensic science (e.g. Masters 1986; Ohtani et al. 1988; Ohtani and Yamamoto 1991; 1992). These data indicate that it is possible to obtain several complimentary calibration curves relating formation o f the D-enantiomer o f an amino acid to the age o f the donor, when studying modern dental tissue. With further opt­ imization o f the chromatographic conditions it may be possible to extend this measurement to include, for example, leucine and valine. This may allow a degree o f internal calibration o f the method, and hence an improvement in the accuracy o f the data acquired and consequently in the estimation o f age. These investigations have shown that it is possible to achieve good separation o f dental amino acids utilising a relatively cheap and simple RP-HPLC method. Standardized sample preparation methods are required to reduce, as far as possible, the variables in the technique. The methods developed may now be applied to known age archaeological specim ens. H ow ever certain extra considerations must be made when dealing with ancient material. In previous investigations (Gillard et a l 1990; 1991), it was amply demonstrated that the ratio o f D/L

Figure 5. Chromatogram o f a dental hydrolysate, derivatized with OPA/IBDC, separated on a C18 RP-HPLC column. Chromatographic conditions detailed in ‘methods ’. Boxed region shown in expansion.

A g e D ete rm in a tio n via M ea su re m en t o f A m in o A c i d R a cem iza tio n

355

Age (Years)

Figure 6a. Plot o f %D-aspartiic acid versus age o f donor.

Figure 6b. Plot o f %D-serine versus age o f donor.

Figure 6c. Plot o f %D-glutamic acid versus age o f donor.

aspartic acid in dental collagen from human first premolars gives a very good prediction of the age of the individual, when using modern (extracted or fresh post-mortem) teeth. However, when the same technique was applied to the known-age dental remains from Spitalfields, where several hundred burials dating largely to the 18th C. AD were carefully excavated and studied (Molleson and Cox, 1993), the results, in common with other ageing techniques, showed some significant discrepancies (Gillard et al. 1991). It is clear from the data of Gillard et al. (1991), and forensic work carried out by other investigators (e.g. Masters 1986)

that the aspartic acid technique may partially fail when it is applied to tissue containing (potentially) degraded collagen. It has been postulated that this problem may be caused by selective microbial attack on the collagen, possibly via collagenase activity, resulting in fragmentation o f the molecule (Child et al. 1993) and subsequent variation in the rate o f terminal amino acid racemization (Smith and Evans 1980). This remains to be demonstrated. Therefore before the methods developed in the present work can be applied with com plete confidence to archaeological material, consideration o f the state o f the material must be made.

356

V ikki A. C arolan, M ich a e l L. G. G a rd n er a n d A. M a rk P o lla rd

Further investigation o f the amino acid content o f the demineralization solution, utilising a fluorescence detector, is needed to establish whether these are due to noncollagenous proteins or acid-soluble collagenous peptides. It has been postulated that certain fractions o f the dental proteins undergo racemization at a faster rate than collagen amino acids e.g. phosphoproteins (Masters 1985). More recently, it has been suggested by Ritz et al. (1994; 150) that ‘aspartic acid racemization is rapid in (acid soluble) non-collagenous proteins or peptides o f dentin but proceeds only slowly or not at all in the (acid insoluble) collagenous protein fraction’. It is essential therefore to study the racemization of amino acids from different protein fractions of dental tissue, i.e., acid soluble and insoluble collagen, and acid soluble non-collagenous proteins (phosphoproteins, glycoproteins, etc.), to assess where the most intact proteins are to be found and also where the fastest racemizing amino acids are located. Clearly there is considerable scope for improvement in our understanding o f how amino acids racemize in vivo in different components of mineralized tissue, and how post mortem degradation o f the proteins affects the rate o f racemization. However, when these investigations have been achieved, the methods of sample preparation and analysis developed in the present work should allow accurate measurement of the enantiomeric ratios of several dental amino acids.

Conclusions This work has shown that each step of the procedure to obtain and quantitatively measure the enantiomers o f amino acids from dental tissue must be studied, carefully optimized and standardized to allow reliable age at death determinations to be made. Samples must be chosen which display the least varia­ bility: selection o f same tooth type (ideally upper, (left and right) first premolars, to account for variations in dentine formation, blood supply and typical archaeological den­ titions; horizontal section of coronal dentine taken from just above the pulp chamber, to reduce variability due to the time of formation o f the sample, contributions from materials other than primary dentine etc. Demineralization and hydrolysis must both be standard­ ized to reduce the effect of induced artefactual racemization, to maximise the removal o f mineral content and release of amino acids. Demineralization in 1M HC1, for 48 hours, with constant agitation o f sample and demineralization solution is ideal. The optimum hydrolysis time has been shown to be 9 hours: this gives optimum HPLC analyses, whilst releasing a high percentage of amino acids and keeping induced racemization to a minimum. Separation o f amino acid enantiomers by chiral derivatization with IB D C /O PA , follow ed by RP-HPLC currently allows measurement o f three pairs o f amino acid enantiomers and production o f three age versus %D-amino acid calibration curves. With further optimization of the chromatographic separation techniques it may be possible to gain data from other amino acids, for example leucine and / or valine. This work has looked at the acid insoluble collagenous proteins o f modern dental material, with no sp ecific procedures for separating other collagen ous or non­

collagenous proteins present. It is suggested, from these data and study o f the literature dealing with amino acid racemization in both dental tissue and bone, that to fully understand amino acid racemization, and hence to fully optim ize age at death determinations from amino acid enantiomeric measurement, it is necessary to separate the various types o f protein present in mineralized tissue i.e. separation of acid insoluble collagen from acid soluble noncollagenous material such as phosphoproteins and glyco­ proteins. The separation o f the various protein fractions may also improve results obtained from archaeological material, where it is thought that degradation (probably microbial) will cause breakdown of various fractions of the proteins.

Acknowledgements This project is supported by NERC-SBAC grant GR3/09561.

References Briickner, H., Wittner, R. and Godel, H. 1991. Fully automated high-performance liquid chromatographic separation of DLamino acids derivatized with 0-phthaldialdehyde together with iV-isobutyryl-cysteine. Applications to food samples, Chromatographia. 32: 383-388. Carolan, V.A., Gardner, M.L.G. and Pollard, A.M. 1996. Measure­ ment of amino acid racemization in dental tissue as a method of determining age: a study of the techniques employed, Pro­ ceedings International Symposium on Archaeometry 1994. In press. Child, A.M., Gillard, R.D. and Pollard, A.M. 1993. Microbiallyinduced promotion of amino acid racemization in bone: isolation of the microorganisms and the detection of their enzymes, J. Arch. Sci. 20: 159-168. Gillard, R.D., Pollard, A.M., Sutton, P.A. and Whittaker, D.K. 1990. An improved method for age at death determination from the measurement of D-aspartic acid in dental collagen, Archaeometry. 32: 61-70. Gillard, R.D., Hardman, S.M., Pollard, A.M., Sutton, P.A. and Whittaker, D.K. 1991, Determinations of age at death in archaeological populations using the D/L ratio of aspartic acid in dental collagen. In Archaeometry ‘90 (eds. E. Pernicka and G.A. Wagner). Basel*. Birkhauser. pp. 637-644. Helfman, P.M. and Bada, J.L. 1975. Aspartic acid racemization in tooth enamel from living humans, Proc. Natl. Acad. Sci. USA. 72: 2891-2894. Helfman, P.M. and Bada, J.L. 1976. Aspartic acid racemization in dentine as a measure of ageing, Nature. 262: 279-281. Man, E.H. Sandhouse, M.E. Burg, J. and Fischer, G.H. 1983. Accumulation of D-aspartic acid with age in the human brain, Science. 220: 1407-1408. Masters, P.M., Bada, J.L. and Zigler, J.S. 1977. Aspartic acid racemization in the human lens during ageing and in cataract formation, Nature. 268: 71-73. Masters, P.M. and Zimmerman, M.R. 1978. Age determination of an Alaskan mummy: morphological and biochemical cor­ relation, Science. 201: 811-812. Masters, P.M. 1983. Stereochemically altered noncollagenous protein from human dentin, Calcif. Tissue Int. 35: 42-47. Masters, P.M. 1985. In vivo decomposition of phosphoserine and serine in noncollagenous proteins from human dentin, Calcif. Tissue Int. 37: 236-241. Masters, P.M. 1986, Amino acid racemization dating - a review. In Dating and Age Determination o f Biological Materials (eds M.R. Zimmerman and J.L. Angel). London: Croom Helm, pp 39-58.

A g e D eterm in a tio n via M ea su rem en t o f A m in o A c i d R a cem ization

Molleson, T. and Cox, M. (with Waldron A.H. and Whittaker, D.K.) 1993. The Middling Sort, The Spitalfields Project Volume 2 - the Anthropology. Council for British Archaeology Research Report 86, York. Mornstad, H., Pfeiffer, H. and Teivens, A. 1994. Estimation of dental age using HPLC-technique to determine the degree of aspartic racemization, J. Forensic Sciences. 39: 1425-1431. Nimura, N., Ogura, H. and Kinoshita, T. 1980. Reversed-phase liquid chromatographic resolution of amino acid enantiomers by derivatization with 2,3,4,6-tetra-0-acetyl-/?-D-glucopyranosyl isothiocyanate, J. Chromatogr. 202: 375-379. Ogino, T., Ogino, H. and Nagy, B. 1985. Application of aspartic acid racemization to forensic odontology: post mortem designation of age at death, Forensic Science International. 29: 259-267. Ogino, T. and Ogino, H. 1988. Application to forensic odontology of aspartic acid racemization in unerupted and supernumary teeth, J. Dent. Res. 67: 1319-1322. Ohtani, S., Kato, S., Sugeno,H., Sugimoto, H., Marumo, T., Yamazaki, M. and Yamamoto, K. 1988. A study on the use of the amino acid racemization method to estimate the ages of unidentified cadavers from their teeth, Bull. Kanagawa Dent. Coll. 16: 11-21. Ohtani, S. and Yamamoto, K. 1991. Age estimation using the racemization of amino acid in human dentin, J. Forensic Sciences. 36: 792-800. Ohtani, S. and Yamamoto, K. 1992. Estimation of age from a tooth by means of racemization of an amino acid, especially

357

aspartic acid - comparison of enamel and dentin, J. Forensic Sciences. 37: 1061-1067. Ritz, S., Schlitz, H.-W. and Schwarzer, B. 1990. The extent of aspartic acid racemization in dentin: a possible method for a more accurate determination of age at death?, Z. Rechtsmed. 103: 457-462. Ritz, S., Schiitz, H.-W. and Peper, C. 1993. Postmortem estimation of age at death based on aspartic acid racemization in dentin: its applicability for root dentin, Int. J. Leg. Med. 105: 289-293. Ritz, S. and Schiitz, H.-W. 1993. Aspartic acid racemization in intervertebral discs as an aid to postmortem estimation of age at death, J. Forensic Sciences. 38: 633-640. Ritz, S., Turzynski, A. and Schiitz, H.W. 1994. Estimation of age at death based on aspartic acid racemization in noncollagenous bone protein, Forensic Science International. 69: 149-159. Saleh, N., Deutsch, D. and Gil-Av, E. 1993. Racemization of aspartic acid in the extracellular matrix proteins of primary and secondary dentin, Calcif Tissue Int. 53: 103-110. Shimoyama, A. and Harada, K. 1984. An age determination of an ancient burial mound man by apparent racemization reaction of aspartic acid in tooth dentin, Chem. Lett. 10: 1661-1664. Smith, G.G. and Evans, R.C. 1980, The effect of structure and conditions on the rate of racemization of free and bound amino acids. In Biogeochemistry o f the Amino Acids (eds D.E. Hare, T.C. Hoering and K. King). New York: Wiley. Takagi, Y., Nagai, H. and Sasaki, S. 1988. Difference in noncollagenous matrix composition between crown and root dentin of bovine incisor, Calcif.Tissue Int. 42: 97-103.

The Chemical Study of Degraded Dentine by Resolution Enhanced Infra-red Spectroscopy D. Lucy and A.M. Pollard Department o f A rchaeological Sciences, University o f Bradford, Bradford, West Yorkshire, BD7 1DP

Abstract Areas of dentine from Medieval human teeth which had previously been identified as having been severely penetrated by saprophytic fungi (Dye et al., 1995) were separated from relatively unaffected areas. The (v4 phosphate domain, 1653 c m 1 amide I bands and (v2 carbonate domains were compared using Fourier self deconvolution enhanced FTIR spectroscopy. The only differences detected in this study between heavily affected dentine and normal dentine have been in the 1653 cm'1 band, identified as being due to peptide carbonyl (amide I) stretch, and linked to the state o f preservation o f hard tissue protein matrix; and an enhancement of the 603 c m 1 to 564 c m 1 peak ratio with degradation o f this band. Although peaks in this region of the spectrum are due to tightly bound HP042_ ions, the increase is thought to be attributable to a stretching band o f more labile PG43~ ions, the peak o f which was not resolved at 610 cm*1, underlying the 603 cm*1 peak. Any increase in the splitting factor is thought to be a result o f this change.

Introduction During work on the application of dentally based forensic age estimation techniques to Medieval human skeletons Lucy et al. (1995) found the internal dentine structures o f externally well preserved teeth to have been heavily affected by what turned out to be a network o f tunnels created by fungi similar to those described by Poole and Tratman (1978). The histological characteristics of the more heavily affected dentine were that it was structureless, did not display the normal birefringence o f dentine, and did not display characteristic fluorescence o f modern dentine. It also displayed a slight but noticeable pink tinge. Towards the

edges o f the heavily affected areas could be seen tunnels parallel to, and also orthogonal to, the remaining tubule structure (Plate 1). It was concluded that the heavily affected areas were the end result o f the tunnelling seen in the mildly affected areas. All teeth inspected had heavily affected areas, generally adjacent to the pulp chamber and the surface of the root, and also areas which appeared to have been entirely unaffected, mainly in the coronal dentine where the dentine had been protected to some extent by the enamel. In an attempt to understand the nature o f the chemical changes associated with the histological features it was decided to use infra-red spectroscopy to investigate any

Plate 1. Photomicrograph o f a 200 (1 section o f dentine field width about 1 mm. The exterior surface o f the tooth root is towards the top o f the figure, the pulp chamber towards the bottom. Dark areas (labelled A) are areas o f dentine which have been heavily affected by fungi. The lighter areas (labelled B) are not visibly affected at all. On the periphery o f the darker areas are areas with some penetration (labelled C) in which the course o f individual tunnels can be seen running parallel to, and orthogonal to, the dental tubules.

C h em ica l S tu d y o f D eg ra d e d D entine by R eso lu tio n E n h a n ce d In fra -red Sp ectro sco p y

chemical alteration. This technique has been used variously to examine organic matter (DeNiro and Weiner 1988), crystalline structures (Pleshko et al. 1992) and substitutions in apatite (LeGeros 1981) in mineralised tissues, and was therefore felt to give the most flexibility to observe gross chemical change in archaeological specimens.

width o f 20 and smoothing factor o f 9. Sample details and numeric results are presented in figure 1.

Results The full range spectra (4000 cm*1 to 400 cm*1) did not show any o f the expected changes between those areas o f dentine which were heavily affected and those which demonstrated little alteration. In fact neither showed any gross difference from modern dentine (fig. 2). Particular areas of the spectrum which had been used to examine specific aspects o f mineralised tissue were then examined. The amide I bands occupy that part of the spectrum around 1653 cm*1 and are primarily the result of peptide carbonyl stretching attributable to collagenous and non-collagenous matrix proteins (Pleshko et a l 1992). Amide II bands could not be further examined as these lie in the 1400 cm*1to 1600 cm*1 region which has strong v 3 C 0 32~ bands underlying it. Above baseline areas were measured for the 1653 cm*1peak, between 1753 cm*1and 1587 cm*1, using these wavenumbers as reference points for the baseline. In general the modem dentine samples had a greater peak area than even the unaffected Medieval dentine samples. The samples taken from heavily affected regions o f dentine had smaller 1653 cm*1 peak areas than the unaffected areas. Enamel samples had the lowest 1653 cm*1 peak areas of all the samples (see fig. 3). The v4 P 0 43~ antisymetric bend domain has been used to provide information about the state o f crystallinity in mineralised tissue. Termine and Posner (1966a) found that localised anisotropic electric fields provided by highly crystalline apatites split this absorption band into a well resolved doublet from the poorly resolved peak found in amorphous calcium phosphates. They devised a splitting

Experimental Procedures Ten human teeth from British Medieval burial sites were obtained, and three modem human controls. Areas o f heavily affected and unaffected dentine were separated from each of the Medieval teeth with a diamond edge cut saw mounted in a dental drill. Normal dentine was extracted from the modem teeth. Enamel was also obtained from five o f the Medieval teeth. All dentine and enamel was ground in an agate ball mill until sufficient powder of a fine particle size (less than 63 j i ) was obtained. Between 0.9 mg and 1.2 mg of dentine or enamel powder was added to between 299 mg and 301 mg of previously ground dry potassium bromide. These two constituents were then ground together for five minutes in an agate ball mill, and pressed into 9 mm disks under a ten tonne ring press. Spectra were obtained from a Perkin-Elmer 1700-X infra­ red spectrophotometer. Initial spectra were gathered for all samples in the 4000 cm'1 to 400 cm'1 wavenumber range using an instrumental resolution o f four wavenumbers and an air background. Further spectra were collected from the 630 cm*1 to 540 cm*1 wavenumber range at a resolution of one wavenumber and an air background, and also from the 900 cm*1 to the 820 cm*1 range, again using a resolution of one wavenumber and an air background. Both of the latter ranges o f the infra-red spectrum were subject to Fourier self-deconvolution (Kauppinen et a l 1981) with a band

(m g)

S am p le

P otassium

(m g)

D e s c rip tio n

brom ide

T o o th

T o o th

c o n te x t

num ber

A rc h a e o lo g ic a l

position

S ite

359

A b o v e b a s e lin e a re a (1 7 5 3 - 1 5 8 7 /c m )

6 0 3 -5 8 4 c m -1

6 0 3 - 5 8 4 c m -1

8 7 8 - 8 7 3 c m -1

p e a k ra tio

splitting fa c to r

p e a k ratio

0 .8 6

Lin c o ln P ip e lin e

W LM 92S216

U7

1

u n a ffe c te d

2 9 9 .6 0

1 .0 0

1 5 .8 0

1 .2 7

1 0 .2 6

Lin c o ln P ip e lin e

W LM 92S216

U7

t

h e a v ily a ffe c te d

3 0 0 .4 0

1 .2 0

1 4 .8 4

1.31

1 0 .1 0

0 .7 1

L inc oln C a th e d ra l

W E B A 93 S 293

U3

2

h e a v ily a ffe c te d

3 0 0 .0 0

0 .9 0

2 0 .4 0

1 .3 2

9 .8 9

0 .6 7

Lin c o ln C a th e d ra l

W E B A 93 S 293

U3

2

u n a ffe c te d

3 0 0 .6 0

0 .9 0

2 3 .6 1

1 .2 9

9 .6 8

0 .8 0

Lin c o ln P ip e lin e

W LM 92S245

U7

3

h e a v ily a ffe c te d

2 9 9 .5 0

0 .9 0

1 2 .7 9

1 .3 3

1 0 .0 4

0 .7 2

L incoln P ip e lin e

W LM 92 S245

U7

3

u n a ffe c te d

2 9 9 .8 0

1 .2 0

2 0 .4 0

1 .2 6

9 .9 6

0 .7 6

L incoln C a th e d ra l

W E B A 93 S 296 W E B A 93S296

L7

4

u n a ffe c te d

3 0 0 .0 0

0 .9 0

1 .3 2

9 .9 6

0 .8 7

L7

4

h e a v ily a ffe c te d

2 9 9 .9 0

0 .9 0

2 1 .3 6 1 2 .3 4

1 .4 0

9 .4 0

0 .6 8

2 9 9 .5 0

0 .9 0

1 0 .6 3

0 .8 4

Lin c o ln C a th e d r a l Lin c o ln C a th e d r a l

W EB A 93S296

4

enam el

2 0 .0 6

1 .1 4

C h ic h e s te r

CH86S260

5

u n a ffe c te d

3 0 3 .3 0

0 .8 0

2 2 .2 6

1 .1 9

9 .6 0

0 .6 9

C h ic h e s te r

C H 86S260

5

h e a v ily a ffe c te d

3 0 0 .3 0

1 .1 0

2 1 .9 9

1 .2 6

9 .7 7

0 .7 1

C h ic h e s te r

C H 86S260

L7

5

enam el

3 0 0 .1 0

1 .0 0

8 .8 0

1 .3 8

1 0 .6 6

0 .6 7

D e n b e ig h P rio ry

6

h e a v ily a ffe c te d

2 9 9 .6 0

1 .0 0

1 4 .6 3

1 .3 3

9 .9 6

0 .7 2

O e n b e ig h P rio ry

6

u n a ffe c te d

3 0 0 .5 0

1 .1 0

2 5 .5 0

1 .1 6

0 .6 7

L incoln C a th e d r a l

W EB A 93S293

L6

7

h e a v ily a ffe c te d

2 9 9 .1 0

0 .9 0

1 3 .3 2

1 .3 2

9 .7 1 1 0 .2 4

L incoln C a th e d r a l

W EB A 93S293

L6

7

u n a ffe c te d

3 0 0 .4 0

0 .9 0

2 1 .1 9

1 .3 0

9 .4 2

0 .6 4

Lin c o ln C a th e d r a l

W EB A 93S299

6L

8

h e a v ily a ffe c te d

3 0 0 .8 0

1 .1 0

1 5 .3 2

1 .3 5

1 0 .0 3

0 .6 5

L incoln C a th e d r a l

W EB A 93S299

6L

8

u n a ffe c te d

3 0 0 .3 0

0 .9 0

2 2 .5 7

1 .2 7

9 .8 3

0 .6 3

L incoln C a th e d r a l

W EB A 93S299

6L

8

enam el

2 9 9 .0 0

0 .9 0

3 .3 3

1 .3 0

1 0 .4 6

0 .7 5

Lin c o ln P ip e lin e

W LM 92S245

U7

9

h e a v ily a ffe c te d

3 0 0 .1 0

1 .0 0

9 .6 7

1.31

1 0 .2 6

0 .7 6

L in c o ln P ip e lin e

W LM 92S245

U7

9

u n a ffe c te d

2 9 9 .0 0

1 .2 0

1 7 .9 7

1 .2 8

1 0 .0 4

0 .8 3

L in c o ln P ip e lin e

W LM 92 S245

U7

9

enam el

3 0 0 .3 0

1 .2 0

4 .2 7

1 .2 4

1 0 .8 0

0 .8 6

L in c o ln P ip e lin e

W LM 92 S243

U7

Lin c o ln P ip e lin e

W LM 92 S243

U7

L incoln P ip e lin e

W LM 92S243

U7

M o d e rn s a m p le

none

to to to tt

M o d e rn s a m p le

no n e

M o d e rn s a m p le

none

0 .6 7

h e a v ily a ffe c te d

3 0 0 .3 0

1 .0 0

1 6 .5 8

1 .3 2

9 .9 7

0 .7 4

u n a ffe c te d

2 9 9 .5 0

0 .9 0

1 9 .5 6

1.31

9 .9 9

0 .7 5

enam el

3 0 0 .9 0

1 .0 0

3 .4 6

1 .2 5

1 0 .8 9

0 .7 6

u n a ffe c te d

2 9 9 .2 0

0 .9 0

2 2 .5 8

1 .1 3

8 .8 7

0 .6 3

12

u n a ffe c te d

3 0 2 .6 0

1 .0 0

2 4 .0 7

1 .1 0

9 .1 8

0 .6 8

13

u n a ffe c te d

3 0 0 .1 0

1 .0 0

2 3 .0 3

1 .1 2

8 .9 1

0 .7 3

F igure 1. L ist o f a ll sp ecim ens a n d data in this study.

360

D. L u cy a n d A .M . P o lla rd

Figure 2. Infra-red transmittance spectra o f the three types o f dentine covering the 4000 cm'1 to 400 c m 1 wavenumbers.

m o d e rn den tin e affe cted dentine u n a ffected dentine

e n am e l

S a m p le ty p e

Figure 3. Box and whiskers plot o f the 1653 c m 1peak area fo r all samples by sample type.

Figure 4. Deconvoluted ( V4 phosphate domain showing 603 cm'1 and 584 cm'1 peaks. 603 c m 1 to 584 cm'1 peak ratio r=(A-B)/(C-B), splitting factor s=A-B+C-B.

Figure 5. Scatter plot o f 603 c m 1 to 584 c m 1 peak ratio against 1653 cm'1 peak area.

Figure 6. Scatter plot o f 603 cm'1 to 584 cm '1 splitting factor against 1653 cm'1peak area.

C h em ica l S tu d y o f D eg ra d e d D entine b y R eso lu tio n E n h a n ce d In fra -red S p ec tro sc o p y

361

D iscussion

Figure 7. Deconvoluted v2 carbonate domain showing 878 cm’1 and 873 cm'1 peaks. 878 cm'1 to 873 cm'1 peak ratio r=(B-A)/(C-A).

Figure 8. Scatter plot o f 878 c m 1 to 873 c m 1 peak ratio against 1653 cm'1peak area.

factor which could be related by X-ray diffraction studies to the proportion of highly crystalline to amorphous calcium phosphate. Deconvoluted spectra in which the 603 c m 1peak was clearly separated from the 564 c m 1peak were obtained on our spectrometer. They were characterised in terms o f both the ratio o f the intensities of the 603 c m 1 and 564 cm'1 peaks, and their respective splitting factors as recorded by Weiner and Bar-Yosef (1990), the exact method being given in figure 4. It was found that there was a relationship between the ratio of the 603 cm'1 height to 564 cm'1peak height, and the 1653 cm'1peak area. This is represented shown in figure 5. figure 6 is a similar graph representing the relationship between the 603 c m 1/ 564 c m 1splitting factor and the 1653 c m 1 peak area. The v 2 carbonate bending band in the 875 cm*1 region is also resolvable using self-deconvolution into a triplet with peaks at 873 cm'1, 866 cm'1 and 878 cm*1 respectively. The 873 c m 1 band is associated with carbonate ions located at P 0 43- sites (type B carbonate). The 878 cm*1 band is associated with carbonate ions substituting for OH ions (type A carbonate) (Rey et al. 1989: 160). The 866 cm*1 peak was not resolved in this study. The relative peak intensities o f type A to type B carbonates were characterised from the (2 region as shown in figure 7. They too were then plotted against the 1653 c m 1 peak area, as shown in figure 8.

The reduction seen in the 1653 c m 1 peak area between modern dentine, unaffected Medieval dentine and affected M edieval dentine which has been heavily tunnelled is unambiguously due to a lower abundance o f peptide carbonyl (Pleshkoefa/., 1992). This implies a reduction in the number o f peptide bonds, and taken as a general reduction in total protein concentration. This peak has been taken to be a good indicator o f the general organic preservation in bone (DeNiro and Weiner 1988: Weiner and Bar-Yosef 1990), and this reduction must be linked to the action o f the tunnelling saprophytic fungi, using fractions o f the organic matter in dentine as part o f their metabolic process. It is not possible to say at this stage which organic fraction is mainly involved. The evidence from the (v4 phosphate bands is more difficult to interpret, figure 5 clearly shows some sort of inter-relationship between the 603 c m 1 and 564 cm*1 bands and the general condition of the dentine (taken here to be indicated by the 1653 cm*1 peak area). In this case it would seem to be an increase in 603 c m 1 relative to 564 cm*1 with decreasing 1653 cm'1intensity (Pearson r=-0.6947; d.f.=21; pp A S 20

9

30 14

40 15

50 14

60 15

Figure 4. Graph o f pelvis disintegration by age: the 5 pelvic measurements

70 14

80 13

90 11

M a ry T iley-B a xter

380

variables (crypt type, coffin type, coffin position, reposition­ ing and gap between death and burial) to obtain a viable sample. It was more efficient to double sample size by studying only the left side of the Spitalfields skeletons. At least one measurement was taken from every bone (except teeth and ribs): see appendix A for exact site areas. These measure­ ments produced a “disintegration index” (DI) for every skeleton: the total score from each measurement. This can be analysed by skeletal part to study differential preservation, as can be seen in figure 1. This shows disintegration by body part for the 77 named individuals studied. In order to continue sampling in the most efficient manner, the pelvis was selected for continuing study as it has the greatest variation in disintegration across its 5 measurements: AC (acetabulum), PS (pubic symphysis), IT (ischial tuberosity), SN (sciatic notch) and AS (auricular surface) (plate 1). In figure 2 skeletal disintegration (SDI) has been plotted against the disintegration o f the pelvis (PDI). As can be seen, there is a general correlation o f decay.1 A further 168 pelvises were studied, making 245 in total, each providing 5 measurements (AC, PS, IT, SN and AS).

Analysis and Results Age A ll named individuals could be exactly aged by parish records. To make the data more manageable age bands were used, as below: Age band

Actual age

Age band

0 1 2 5 10 15 20

0-6 m 6m -18 m 18 m-2 yrs 2-7 yrs 8-12 yrs 13-17 yrs 18-24 yrs

30 40 50 60 70 80 90

Actual age 25-34 35-44 45-54 55-64 65-74 75-84 85-94

yrs yrs yrs yrs yrs yrs yrs

Pelvis disintegration (PDI) was plotted by age to investigate whether skeletal disintegration was related to age. Disintegra­ tion was studied as distinct from fusion. As can be seen in figure 3, skeletal disintegration is highest amongst the young. As mentioned earlier, the bones of infants and juveniles are less mineralised than mature adults’. Bones develop as ossified growth centres surrounded by cartilage. Gradually the bone increases until it has overtaken cartilage development. Long bones grow from the metaphysis to the epiphysis: when these two meet, epiphysial union has occurred (generally between the ages of 12-32: Brothwell 1981). In an unfused state bones are more vulnerable to decay. Investigating the difference in bone survival by age further, the data were plotted such that for each age band the average disintegration can be seen for each o f the 5 pelvic measurements (figure 4). It now becomes clear that the bones most vulnerable to disintegration are the unfused ischium and pubis (AC, PS, IT: see plate 2). Small and cartilaginous, calcification increases through childhood and fusion at puberty, when as can be seen, pelvic disintegration follows a more predictable pattern (plate 2).

Plate 2. Partially fused pelvis (Skeletal data: S t no. 2145, sex: X, crypt type: 6, coffin type: 1, coffin position: 3, status: s, Gap: X.) (Copyright NHM 1995). Here the pubis and ischium are partially fused, while still unfused to the ilium. Held together by cartilage during life, the non-osseus parts have all decomposed. Exposed developing bone surfaces are rich in minerals and prone to attack. Sex “There are nearly always more males than females in skeletal collections from archaeological sites ... This has been explained in part by the comparatively rapid disintegration o f lightly built female skeletons” (Walker et al. 1988: 186). In order to explore whether this occurs at Spitalfields, the pelvic disintegration index o f fem ales and m ales was compared (figure 5). It was found that differences were minimal: average PDI being 5.9 and 5.8 respectively (confirming findings from Molleson & Cox 1993). With osteoporosis commonly blamed for poor female skeletal survival (Walker et al. 1988), sex and age were plotted out (figure 6). Small sample size explains marked differences in pelvis disintegration amongst the children. In the older generations the 80 age-grouped women seem to have abnormal bone disintegration. However, it is within one standard deviation o f the average, so is statistically insignificant.

Environment: crypt type The 22 crypts excavated at Spitalfields were grouped into 8 “types” by similar environment, to establish what effects these different environments would have on bone survival (see figures 7 & 8). Those conditions enabling good preservation o f the skeleton also favour good preservation o f the body. A number o f individuals were so w ell preserved that on excavation, autopsies were carried out and the remains then cremated. By plotting the overall % o f cremations in each crypt and comparing them to % bone disintegration, the importance o f environmental factors should become clearer (fig- 9). As can be seen, crypt type 8 has the environment most suited to high bone and body preservation: no outside walls or ground floor; it is the most insulated. In contrast, crypt

A T a p h o n o m ic S tudy o f the H um an R em a in s fr o m C hristchurch, S p ita lfie ld s

381

£

+ SEX SAM PLE

^

S IZ E

109

105

Figure 5: Graph o f pelvis disintegration by sex (with standard deviation)

|ffi

m w AGE

SAMPLE • F SIZE •M

0 8 7

1 5 7

2 1 3

5

10

15

3

2

3

1

1

3

m svs 20

30

40

3

6

8

6

8

7

90

50

60

70

80

9

8

5

8

7

7

9

5

4

5

F ig u re 6: G raph o f p elvis d isin teg ra tio n b y sex: bro ken b y a g e

M a ry T iley-B a xter

382

type 3 has the highest skeletal disintegration and one o f the lowest proportions of cremation. With waterproofed brick floors and outside walls, crypt type 3 is vulnerable to extremes of temperature and dampness, both deciding factors in bone preservation: they weaken the protein-mineral bonding of the bone, thus “markedly enhance its degrada­ tion” (Von Endt & Ortner 1984: 252). Type Crypt Name A 1 2 3 4 5 6 7 8

NC, LL, PV, EP, SC LE, LC, SE, LV, LW PK, HV LN, LS LP UN, US UC, NE, UE NP, SV

B

c

D

E

F

X X X X X X

G

H

X X

X

X

X

X X X X

X X

X

X X X

Figure 7: Table o f crypt types (A= outside walls, B= inside walls, C=in tower, D - first floor, E= earth floor, F= 5' brick floor, G= waterproofed brick floor, H= underneath church steps).

They were not air-tight, however: church sextons were legally obliged to perforate them (Molleson, pers. comm). By sitting in its own acidic juices, bones will be vulnerable to chemical attack. Apatite will dissolve from the bone structure, recrystalising with increased volume as brushite and causing bone to explode and disintegrate (Piepenbrink 1986). W ood coffins provide better drainage, allow ing the skeleton to dry out, becoming vulnerable to physical or biological dangers such as traffic vibrations, microorganisms and insect attack. Those remains from minimal or non-existent coffins have, paradoxically, low disintegration. Originally interred in coffins, these have decayed away, leaving the skeletal material uncontained. If the bones have survived well, they are clearly identifiable as from a discrete burial; the remains o f an individual. If, however, the skeletal material has decayed into bonemeal, its identification as the entire remains from one burial (rather than part o f one, or the collective remains o f more than one) must be more doubtful. In contrast, remains within coffins, however decayed into bonemeal, are contained and therefore more clearly identifiable (see plate 3). A s can be seen in figure 10, there is actually little difference in bone survival by coffin type.

Environment: coffin type There were in essence 5 coffin types found at Spitalfields:

Environment: coffin position

1 2 4 6

Coffins were stacked in piles in the crypts. Unfortunately the information as to how the coffins were arranged (which were underneath/ in the middle/ on top) was inaccessible. Horizontal coffins provide more static environments for decay: remains in sloping or vertical coffins would m ove as they rotted. Movement during putrefaction increases decay as new areas are exposed to bacteria (Mant 1987). Coffins were generally inserted horizontally, and it is mostly with

8

single wooden coffin double wooden coffin single lead or lead with wooden outer coffin wood inner with lead outer, sometimes with extra outer layer of wood no coffin remains found, or only wooden lid

It is known that lead coffins prevent drainage of body fluids.

F ig u re 8: P la n o f cryp ts fr o m C hristchurch, S p ita lfie ld s (F rom R ee ve a n d A dam s, 1993: 27).

A T a p h o n o m ic S tu d y o f the H u m a n R em a in s fr o m C hristchurch, Sp ita lfield s

C. TYPE 1 SAMPLE SIZE- SAMPLED 89 CREMATED 18

2 27 1

3 24 0

4 18 1

5 17 0

6 28 0

7 24 3

383

8 19 14

Figure 9: Graph o f % cremated by crypt with % disintegration by crypt

Repositioning The significance o f repositioning on skeletal disintegration depends on the state o f decay o f the body. Some coffins were rearranged before the crypts were closed up in 1867, by which time undoubtedly the remains were skeletonised, hence less vulnerable to attack. Once skeletonised little damage can be done to the bones unless the coffins were repositioned by some force: as these were fairly heavy it seems unlikely. Unrearanged coffins are deemed “primary”, moved ones “secondary” (figure 12).

Gap Between Death and Burial To enable easier manipulation of the data, days between death and burial were banded. s m 1

Plate 3: bonemeal recoveredfrom coffin (Skeletal data: Sk. no. 2105, sex: X, crypt type: 6 , coffin type: 1, coffin position: 1, gap: X.). I f the coffin containing individual 2105 had not survived, there might have been doubts as to whether the bonemeal came from one or more skeletons.

insertion of a new inhumation once space was decreasing that coffins were upended: “other” is due to coffin stacking; those underneath giving way under pressure (figure 11).

short medium long

0 -5 days 6 -1 0 days 11+ days

It can be seen in figure 13 that skeletal disintegration decreases as the gap between death and burial increases. One explanation is that this is related to age: 59% o f the sample buried soonest are under 17 years o f age, therefore most vulnerable to degeneration. The proportion o f children decreases as the gap between death and burial increases. An alternative explanation lies with how the body was treated between death and burial. After death “in a warm dry atmosphere the tissues o f the body quickly start to dry ... Embalming counteracts post-mortem dehydration... The body tissues begin to decompose within a few hours after death. Their complex organic compounds are gradually

384

PB.VISDiSmiEGRAIiONM)0(

Mary Tiley-Baxter

Figure 10: Graph o f pelvis disintegration and coffin type (with standard deviation)

COFFIN POSN. SAMPLE SIZE

HORIZONTAL 212

OTHER 21

VERTICAL 13

Figure 11: Graph o f pelvis disintegration and coffin position (with standard deviation)

PELVIS DIS8IIEGRM10N INDEX

A Taphonomic Study o f the Human Remains from Christchurch, Spitalfields

4SAMPLE SIZE

93

H153

Figure 12: Graph o f pelvis disintegration and primary/ secondary status (with standard deviation)

P3.VIS DiSMTEGRATlOK NDEX

1(JT

----------------------1------------------------1----------------------- 1----------------------- 1----------------------- 1----------------------- 1 ® m s_ GAP SAMPLE SIZE

0-5 days 32

6-10 days 75

11+ days 16

Figure 13: Graph of pelvis disintegration and gap between death and burial (with standard deviation)

385

Mary Tiley-Baxter

386

PRESERVAUCti OF W .T5. SKELETONS

Figure 14: Graph o f % survival o f Spitalfields skeletons compared to % survival o f West Tenter Street skeletons

Figure 15: Scatter graph to compare skeletal survival at Spitalfields and West Tenter Street

A Taphonomic Study of the Human Remains from Christchurch, Spitalfields broken into simpler compounds which, in the form of gases and fluids, eventually escape; only the bony framework remains” (Poison & Marshall 1975: 316-9). A putrefying body smells: distressing the relatives. Individuals with delayed burial would go through the hands of the undertaker (Reeve and Adams 1993: ch. 4) and measures to delay decay. There is very little known o f what happened in this regard, but “no bluebottle eggs were found associated with any corpse” (Molleson, pers. comm): some preservation tech­ nique was practiced. These skills would also seem to aid skeletal preservation.

Interpretation Age The sample from Spitalfields reflect the general assumption (Walker et al. 1988) that children’s bones are less likely to survive that adults’. The increase of disintegration with greater age shows mineral loss in the bone leading to a weakening in the structure and increasing fragility Sex “the com paratively rapid disintegration o f the poorly calcified females” (Walker et al, 1988; 186) is not in evidence here (the average PDI was 5.9 (F)and 5.8(M)).

387

Comparison with Waldron’s Work on West Tenter Street Human Remains As the structure o f this research was developed from T.W aldron’s skeletal work from the R om ano-British cemetery at West Tenter Street, the two sets o f data have been compared (figure 14). The Spitalfields sample were all in coffins in crypts: the West Tenter Street sample were buried in the ground over 1000 years earlier. The scale o f the y-axis has been inverted from measuring disintegration to preservation for each skeletal element. As can be seen, there are similar survival patterns between the different skeletal elements. These patterns must reflect “survivability” (robustness in order to survive and be in recognisable form on recovery) in skeletal elem ents. Similarities between the two sites are strong as both had whole interment: differences can be attributed to length of interment as well as the environmental factors o f crypt as opposed to soil. The only place where preservation overlaps is at the petrous temporal area. This appears to be low in the Spitalfields sample where many skulls were whole and it was impossible to gain an accurate reading. The similarity in preservation can also be seen in figure 15.

Acknowledgements Environment: crypt Temperature variation seems to be the most important factor in crypt type: thus skeletons in crypts with outside walls or on the ground floor are more vulnerable to decay that those in crypts on the first floor with no outside walls.

Coffin Type, Coffin Position, Repositioning These all made little difference overall among the range o f variables, and when F- tested were found to be statistically insignificant. (Scoring a significance of 0.047, 0.463 and 0.160 respectively.)

Gap Between Death and Burial Increased disintegration for those buried within 5 days is logical, when it is remembered that 59% were children with bones found to be more vulnerable to decay. In similar manner it makes sense that those left longest before burial are almost all adult, having been subject to any preserving techniques.

The Relative Importance of Different Taphonomic Variables Age seems to be the most important factor, with children’s bones surviving least well. Crypt type is also very significant, in how cushioned the remains are from extrem es o f temperature and damp. The gap between death and burial could also be important: more research is needed to discover how it relates to age and mortuary practices. In contrast, sex, coffin type, coffin position and reposi­ tioning o f remains are here much less significant.

Thanks to Prof. I. Hodder, Archaeology, Cambridge; Miss T. Molleson, Palaeontology, NHM; Mr P. Crabbe, Photo­ graphy, NHM,

Notes 1. The scales are of different calibration as the y-axis shows the scores of 5 pelvic measurements / individual (maximum 50) and the x-axis shows th scores of 42 skeletal measurements (not inlcuding pelvis) / individual (maximum 420).

References Brothwell, D.R. 1981 (3rd Ed). Digging up bones. Oxford, British Museum (Natural History) and Oxford University Press. Mant, A.K. 1987. Knowledge aquired from post-War exhumations. In Death, Decay and Reconstruction: Approaches to Archae­ ology and Forensic Science (eds A. Boddington, A.N. Garland and R.C. Janaway). Manchester, Manchester University Press, 65-80. Molleson, T. and Cox, M. 1993. The Spitalfields Project Vol 2: The Anthropology. London, CBA Research Report 86. Piepenbrink, H. 1986. Principles of dead bone decomposition and diagenetic consequences. In Innovative Trends in der Prahistorischen Anthropologie (ed. B. Herrmann) Band 7: 15-21. Poison, C.J. and Marshall, T.K. 1975 (3rd ed). The Disposal o f the Dead. London, English Universities Press. Reeve, J. and Adams, M. 1993. The Spitalfields Project Vol I: The Archaeology. London, CBA Research Report 85. Von Endt, D.W. and Ortner, D.J. 1984. Experimental effects of bone size and temperature on bone diagenesis, Journal o f Archaeological Science 11: 247-253. Waldron, T. 1987. The relative survival of the human skeleton: implications for palaeopathology. In Death, Decay and Reconstruction: Approaches to Archaeology and Forensic Science (eds A. Boddington, A.N.Garland and R.C.Janaway). Manchester, Manchester University Press, 55-64. Walker, P.L., Johnson, J.R. and Lambert, P.M. 1988. Age and sex biases in the preservation of human skeletal remains, American Journal o f Physical Anthropology 76: 183-188.

Mary Tiley-Baxter

388

Appendix Scapula

Skull 1FR 2PA 30 40C 5PT 6MA 7ZY 8MX 9MH 10MR 11MB

Frontal bone Parietal bone Occipital bone Occipital condyle Petrous temporal Mastoid Zygoma Maxilla Mandible head Mandible ramus Mandible body

Vertebrae 12CV 13TV 14LV 15SV

Cervical vertebrae Thorasic vertebrae Lumbar vertebrae Sacral vertebrae

Clavicle 16CLP Clavical proximal 17CLD Clavical distal

Pelvis 18AC 19PS 20IT 21SN 22AS

Acetabulum Pubic symphysis Ischial tuberosity Sciatic notch Auricular surface

23SCB 24GC 25CO 26ACR

Scapula body Glenoid cavity Coracoid Acromion

Sternum 27SM 28SB

Sternum manubrium Sternum body

Leg and foot 29FP 30FD 3 IP 32TP 33TD 34FIP 35FID 36CA 37TA 38T 39MT

Femur Femur distal Patella Tibia proximal Tibia distal Fibula proximal Fibula distal Calcaneum Talus Tarsals Metatarsals

Arm and hand 40HP 41 HD 42RP 43RD 44UP 45 C 46MC

Humerus proximal Humerus distal Radius proximal Radius distal Ulna distal Carpals Metacarpals

Approaches to the Analysis of Evidence from the Acheulean Site of Beeches Pit, Suffolk, England S.A. Andresen, D.A. Bell, J. Hallos, T.R.J. Pumphrey and J.A.J. Gowlett Department o f Archaeology, University o f Liverpool, Liverpool L69 3BX, UK

We give an account here of the various analytical approaches which we have applied or are applying to archaeological material recovered from Beeches Pit, a well-preserved Middle Pleistocene site in Suffolk (which is discussed in environmental and archaeological terms in Preec eetal. 1991 and Bridgland et al. 1994; and archaeologically at greater length in Gowlett et al. in prep). Archaeological excavations have been carried out in summer seasons from 1992 to 1995 (directed by JAJG), and a second phase o f the work is planned for 1996. The site o f Beeches Pit can be seen in two contexts: 1) that o f the local British Sequence, where there is a potentially excellent record through the last half million years, but where little archaeological evidence can be ascribed with certainty to a fully temperate climate. 2) in the general frame of hominid evolution, where there is a stone age record o f around 2.5 million years, but where there is little direct archaeological evidence that behavioural patterns at 0.5 million years ago were much different from those a million or more years earlier. Thus at Beeches Pit we can address questions that are important in terms o f building up a regional record, but also general research questions such as those addressed by Isaac (1981), which remain as relevant here as on much older sites. In this paper we give instances o f our approaches, and attempt to review such research questions, first providing brief details o f the site.

The site Beeches Pit is a nineteenth century brickpit, on land that is now forested, 8 km to the north-west o f Bury St Edmunds. It became known as a Pleistocene exposure late in the 19th century (Skertchly 1877), and more recently has been established to represent interglacial deposits belonging to the Middle Pleistocene. Geological and environmental work has been reported by Preece et al. (1991) and Bridgland et al. (1994) The pit is about 70 metres by 50 metres in extent, and exposes sediments around the margins to a depth o f about 4 metres. A pattern o f test trenches, both geological and archaeological, suggests that the most promising concentra­ tions o f archaeological material are in the north-west sector o f the pit, in an area where there are calcareous (tufaceous) sediments. The local sequence starts with chalk bedrock and then a diamicton believed to have been deposited during the Anglian glaciation (Preece et al. 1991). In the eastern side o f the pit geological trenches have shown the presence of

fluviatile gravels. There follows a phase o f deposition of calcareous, tufaceous sediments, occasionally with small blocks or flags o f purer tufa, but generally as a series o f pale-coloured clays. In the NW of the site, in the 10 metre square designated AF, geological cuts 2 -4 exposed several units o f clays. These sections suggested to Preece et al. (1991) that the organic clay stratigraphically overlay the tufaceous sediments, although there could be some interdigitation. The archaeological trench AF linking these cuts has confirmed that the dark organic clay (unit 4) and the underlying brown clay (unit 3) overlie the calcareous clays, which are here fronted by a steep slope (fig. 2). It seems likely that unit 2 also is later than the calcareous clays, but the excavation is not yet completed in its lower part, and it is possible that unit 2 runs into the tufaceous sediments. We postulate that the steep slope o f the calcareous clays, and hence the slope o f the overlying deposits, are an original feature, rather than postdepositional warping, because the slope is steeper than any other interface between units. It is possible that the steep front is an erosional feature, and that rises and falls o f water level have encouraging very local slumping and mixing o f sediments. Artefacts have been found in the clays, concentrated particularly in and around the dark horizon, Unit 4, which includes nodules o f burnt flint, and bones, some o f them also apparently burnt. Unit 4, and underlying clay units, can be follow ed downslope towards the west to an area where deposition may have taken place under water, judging from the faunal remains and lithology. In this case, the locality would represent a channel bank, and it seems likely that primary occupation or activity took place on its side, away from the water. Unfortunately, most o f the original occupation surface has probably been planed off by later erosion: the strati­ graphic evidence for this is a layer o f clay with flints (unit 7), possibly a solifluction deposit, which overlies the whole area, and appears to have truncated the clays. In this area the number o f artefacts relatively restricted: there is a preponderance o f small flakes, a few medium­ sized flakes, and absence of shaped tools. About 20 metres further east, another trench, AH, reveals a variation o f the stratigraphy. The trench is at present 5 metres by 3 metres, and extends from the old geological cut 5, which had revealed the presence o f artefacts. A lens or layer of darker sediment about 50 cm thick has been found stratified between layers o f pale calcareous clays. The dark clay unit varies laterally: it becomes reddish-brown to the north, and appears to grade into the pale yellowish calcareous clays on the northern e d g e . Within this unit there is a dense

390

S.A. Andresen, D.A. Bell, J. Hallos, T.R.J. Pumphrey and J.A.J. Gowlett

Figure 1. General plan o f Beeches Pit showing the extent o f the 19th century pit. Test trenches have delim ited the north west sector as the area most rewarding fo r archaeological excavation.

concentration o f flint artefacts, mainly debitage, but including at least one classic hand-axe and several other bifaces. The debitage amounts to several hundred pieces. There are far fewer pieces in the northern row o f metres squares, so that as in AF few if any artefacts occur actually within the tufaceous sediments. The layer slopes gently towards the south-west, and it may be possible to follow it further along the floor o f the pit. Other artefact localities may well exist outside the boundaries of the pit, but it seems unlikely that they will be found to the south or west, because the geological trneches (Bridgland et al. 1994) show that the chalk is relatively high. On the basis o f the geological and environmental work, coupled with the archaeology, the site can be interpreted in general terms: a glacial channel gradually silted up in the succeeding interglacial, so that there was a pond, o f undocumented dimensions. This proved attractive to animals and humans, probably over a considerable period o f time, since the stratigraphy suggests that the dark lens in AH is unlikely to be the same as Unit 4 in AF. The size of the pond is a point of interest: early on, when tufa was being deposited it may have been more extensive than during the deposition o f unit 4. Then, some mlluscan taxa indicate that the pool was small, possibly prone to drying out in the summer (Preece et al. 1991; Bridgland et al. 1994). Although the archaeological evidence may represesent sporadic bursts of activity, at least some o f it seems clearly linked with units 3 and 4.

Examples of Investigations In the initial work, a variety o f approaches has been employed with the aim o f interpreting the archaeological data. The site presents the preliminary difficulty that the different kinds of evidence tend to be spatially separated, although their state o f preservation is very good.

Thus one o f the key matters to establish at Beeches Pit is the relationship between the different sets o f preserved material: the artefacts and other large inorganic clasts, the microfauna and macrofauna; and how these fit into the general sedimentological scheme. First appearance indicates some pronounced separations or localisations on the site: artefacts tend to be on gentle slopes at the edge of a channel. Preserved macrofauna is on steeper slopes, extending down towards the channel floor, but microfauna appears to be preserved more widely.

A rtefact taphonomy In excavation we have tried to retain all stone finds, regardless o f origin, in the hope that this will be experimentally useful. An investigation (by SA) has shown that more than half o f the stones recovered are natural clasts mainly o f flint. In some cases we cannot easily determine whether there has been modification by human agency, but this is not a major problem. We intend that relationships between the two sets o f material can be investigated. Already the artefact data from the two localities have been analysed, and size distributions calculated, so that comparisons can be made with other sites and contexts (fig. 3). This single category o f evidence would suggst that the Beeches Pit artefacts have not been highly disturbed, because a large proportion o f the artefacts is less than 2 cm in length (over 70% in both AF and AH). W e have yet to determine whether the excess o f 1-2 cm pieces over 0 -1 cm pieces is primarily a result of taphonomic winnowing, or o f recovery procedures.

A rtefact refitting studies Refitting studies o f artefacts cast another light on degree o f disturbance, and also allow aspects o f past human behaviour to be studied directly. The flints at Beeches Pit are in sharp

Approaches to the Analysis o f Evidence from Beeches Pit, Suffolk

Geol.Cut2 53E

AF

391

AH

Topsoil

Quarry backfill/ slopewash

//

' / Calcareous ' tufaceous /Y

Dark grey silty clay

/ / Calcareous tufaceous f'f

clay^

//

^ //

Figure 2. Schematic sections o f the archaeological trenches. The layer marked (2 \ unshaded, contains bone, and may be largely redeposited from the calcareous tufaceous clay.

condition, and sometimes preserve flaws or colourings which can help in finding refits. Two large flints from the second season’s excavation in AH could be linked because they both shared a particular cortical slot on one side, and also a greyish mark w ithin the fabric o f the flint. Subsequently the butt o f a biface was found, and then its refitting tip-part. One o f us (TRIP) has carried out a detailed study o f refitting possibilities within AH excavation, with the result that more than fifty conjoins have been found in total (fig. 4). Most o f these are over a distance o f less than three metres. The refits conform, roughly speaking, with a surface aligned with the dip o f the sediments, but suggest a certain amount o f vertical movement, of the order o f 10-30 cm. A plot o f the refits in a rose diagram does not yet indicate conclusively whether pieces have travelled further in one direction than another.

Artefact edge-damage We thought initially that a proportion of the flakes at Beeches Pit were likely to have had intentional retouch. A study (carried out by JH) suggests, however, that intended retouch are very scarce. Edge damage can also be incurred through trampling, or natural p rocesses operating within the sediments. The study was intended to distinguish between these, using criteria of experimental researchAt present we believe that four flakes exmained so far have modification intentional retouch in two cases, and utilisation traces. Minor natural damage has been incurred by many other other specimens, including the tip part of the biface mentioned above, and seems to be more important than trampling.

Figure 3. Size distributions o f debitage from the localities o fA F and AH.

392

SA. Andresen, DA. Bell, J. Hallos, T.R.J. Pumphrey and J A J . Gowlett

80E

Figure 4. Plot o f refitting artefacts from AH excavation, showing where the two parts o f a biface were found.

Artefact m orphology It is plain from this that there are relatively few shaped tools at Beeches Pit. Eventually, however, there may be enough bifaces to allow useful com parisons o f technique and dimensions with other sites. For the moment the material itself encourages us to look at processes and decision paths in the manufacture of the material, rather than morphology. It seems likely that bifaces were made by at least three procedures: the modification of large flake blanks struck specially for the purpose; the working of large nodules o f flint; and the working o f natural tabular blanks. There are already indications, even from a few specimens, o f a wide range o f outcomes from the bifacial working routines. These include ‘classic’ bifaces (the one found in two parts, and two bifaces recovered in earlier work at Beeches Pit); ‘other’ or ‘non-classic’ bifaces, which may represent ad hoc working for a particular task; and discoids, represented by one small specimen found in 1995.

Fauna The fauna was first studied systematically by Simon Parfitt o f the FAU at the Institute o f Archaeology, UCL, who continues to have a major input to the project. Faunal material has been uncovered from both o f the areas excavated during the recent fieldseasons, although the macrofauna is limited to a series o f sandy silty clays in area AF, and is concentrated in units 2 -4 . In general the bone is very fragmentary in nature and during excavation it appears to be mainly cortical, although medullary/spongy bone has been recovered during wet sieving. The condition o f the bone appears to be to some degree dependent upon the integrity o f the element: complete elements or isolated fragments o f cortex being quite robust, whilst other superficially complete or near complete elements are extremely friable, including their cortical element. This

has implications for recovery bias, but in origin is likely to be a function o f local variations in burial conditions and o f calcification/decalcification. There are varying degrees of mineralisation apparent, with the bone from the lower units being much more heavily mineralised on the extreme western fringe o f the excavations. None o f the specimens appear to be heavily abraded. The question o f rounding has not been addressed yet, but a number o f bones exhibit features indicative o f a period o f sub-aerial weathering, such as linear cracks and exfoliation. Fractures are consistent with both green bone and dried bone damage. An updated species list is not yet available, but specimens recently found are known to include deer and bear. The microfauna, although not restricted to the same sediments, is most heavily concentrated in the same zones o f area AF (a point o f interest on its own). Teeth are the most common element type found, although post-cranial elements have been identified (Preece et al. 1991; B ell nd). The faunal material is being studied as by DAB as part o f a doctoral project in an attempt to identify which depositional agencies were involved in their collection, with as much stratigraphic and areal control as possible.

Sedimentology The faunal taphonomic studies are taking place in parallel with sedim entological analyses involving the study o f particle size, sediment chemistry and magnetic character­ isation to provide as accurate a sedimentological context for the bone data as possible. Gross sedimentological structures are also being accurately recorded. It is hoped that this work will allow the sediments containing archaeological material to be tied in accurately to an environmental record which has been assessed for its formation biases. As a preliminary analysis o f the small mammals has highlighted an apparent transition from a predominantly

Approaches to the Analysis of Evidence from Beeches Pit, Suffolk wooded environment to open grassland (Preece etal. 1991; S. Parfitt pers. comm.), the taphonomic analysis o f the sediments will be used to assess whether the pattern does reflect variation in the local environment or if it is in some way related to a change in collection agent.

Fire History Many burnt flints have been observed in both areas AF and AH. In AF the most clear examples occur in the dark organic Unit 4, and a little below it. Sometimes nodules o f flint have been burnt right through, so that they are red and have crackled or spalled into small pieces. In AH burning is also common, but there are fewer reddened pieces. S. Haritou (University o f Liverpool) has carried out a preliminary study of their magnetisation. This produces the interesting finding that some pieces may have been burnt in more than one orientation. An extended study is taking place to explore further the temperatures o f burning, and the question of whether pieces were burnt on more than one occasion (which have been calculated independently in some cases for TL studies: N. Debenham, pers. comm.). There are no indica­ tions at present to show whether the fire events were natural or o f human origin.

Research Questions The first and most obvious point o f the research is to set the site into the local record o f the British Middle Pleistocene. A more general series o f archaeological research ques­ tions in palaeoanthropology was set out by Isaac in papers in the 1970s. His later work centred on the ideas o f home bases and food-sharing, and led to vigorous debates concerning the context and association o f materials (e.g. Isaac papers: Isaac 1989). Recently Potts (1994) has suggested that debate has become too much concentrated on polarised issues, and has set out a new set of questions for palaeoecological investigation. Inspection o f these shows that there is a good deal o f concordance with issues outlined by Isaac (e.g. 1972), who also emphasised that explanations were not likely to be mutually exclusive (points discussed in Gowlett 1996). Thus we abstract here a set o f questions for study: 1) Location and density o f sites: can w e accumulate evidence for land-use and ecology, especially in relation to water and vegetation zones? 2) site integrity: to what extent do the artefacts and bone remain as deposited by hominids? 3) site sizes: how big were they, how long occupied, and at what season(s)? D oes any evidence relate to community size and organisation? 4) internal structure: is there any evidence for structural features or separations of activities which can inform about social behaviour? 5) resource transport: how are the sites located in relation to material resources, and how far were these transported, for modification or use? 6) artefact dynamics: can we disentangle threads or chains of activity, and in particular examine to what extent artefacts were made elsewhere or discarded elsewhere? 8) faunal potential: what relationships or overlaps can be traced, with either carnivores or possible prey species?

393

The exploratory phase o f archaeological investigation has tackled questions 1,2, and 6 fairly directly, touching also on 3, 4 and 8, which can be addressed in further work. Such points outline the aim to investigate the extent o f the human factor, within a natural envelope. The questions above seem to be as relevant in Europe, around half a million years ago, as in Africa around 2 million years ago. Very few additional human capabilities have been precisely documented in the intervening period. Homo had become relatively large and powerful; bifaces were certainly being made; there was possibly fire control - but even this is not agreed. Thus research questions remain essentially as for earlier times, with addition o f questions as to how adaptation was made to temperate latitudes, with seasonal variations o f temperature and resources. The extent o f interglacial occupation by archaic Homo sapiens, and the d ifficulties o f operating in a closed environment are questions which have been tackled by Gamble (e.g. 1986), and Roebroeks et al. (1992). Beeches Pit makes it reasonably certain that there was interglacial occupation in Britain, probably in isotope stage 11 ut because o f the changes o f environment recorded by our colleagues, it is too early to say whether this occurred in more than one setting, i.e. in both open and closed environments. The site preserves enough lines o f evidence to offer a focus for other research: e.g. experimental taphonomic studies by carried out independently by P. Quinney; bone preservation studies by the Research Laboratory for Archaeology at Oxford, dating studies by T. Atkinson, P. Roe, and N. Debenham and pollen sampling by H. Roe.

Conclusions Beeches Pit, as observed by Bridgland et al. (1994), is one o f the very few Acheulean sites in Britain which can be ascribed to occupation in fully temperate conditions. The first phase o f archaeological work at Beeches Pit has been pursued along a number of lines. These have begun to address our principal research questions, but do not equate with them one-to-one. They have addressed the most basic issues o f investigating stratigraphy, and establishing the integrity o f preservatiion. It is fairly unusual for geological and environm ental work to have preceded archaeological excavation by quite such an extent: we have had the benefit o f this, and it also seems likely that the archaeological results w ill pose further questions for research in the other disciplines. It is hoped to carry out further archaeological investigations, with expanded trenches, so as to address the more refined questions - especially about site ecology and artefact dynamics - as directly as possible.

Acknowledgments We are most grateful to the Forestry Commission, especially Mr W.J. McCavish; to Mr and Mrs J. Browning; to English Nature, especially Mr N eil Glasser; Mr Edward Martin o f Suffolk Archaeology; and to Richard Preece, Simon Lewis, John Wymer, David Bridgland, Simon Parfitt and Nick Ashton for help and advice. The fieldwork was supported by the University o f Liverpool Research Development Fund and SACOS. The excavation crews and helpers o f 1992-95 included:

394

S.A. Andresen, D.A. Bell, J. Hallos, T.R.J. Pumphrey and J.A.J. Gowlett

Kath Day, Paul Ogilvy, Jenny Woodcock, Tim Holt-Wilson, Gary Newman, Kit Owen-Smith, Simon Kershaw; Adam Beaumont, Caroline Birch, Kristine Birchall, Trevor Brown, Shirley Crompton, Nick Danks, Caroline Delahunty, Paul Dowman, Julie Dowman, Faye Halliday, Sue Haritou, Mike Hewison, Chris Jones, Alan Lovell, Huw Nicholas, Helen Odozi, Tom Pumphrey, and Matt Sparkes.

References Bridgland, D.R., Lewis, S.G. and Wymer, J.J. 1994. Middle Pleistocene stratigraphy and archaeology around Mildenhall and Icklingham, Suffolk: a report on a Geologists’ Association field meeting, 27 June 1992. Proceedings o f the Geologists* Association. 106: 57-69. Gamble, C.S. 1986. The Palaeolithic settlement o f Europe. Cambridge: Cambridge University Press. Gowlett, J.A.J. 1996. The frameworks of early hominid social systems: how many useful parameters of archaeological evidence can we isolate? In the Archaeology o f human Ancestry: Power, Sex and Tradition (eds J. Steele and S. Shennan). London: Routledge. Gowlett, J.A.J., Andresen, S.A., Bell, D.A., Hallos, J. and

Pumphrey, T.R.J. in prep. Beeches Pit: paper for submission to Geoarchaeology. Isaac, B. (ed.) 1989. The archaeology o f human origins: papers by Glynn Isaac. Cambridge: Cambridge University Press. Isaac, G.L1. 1972. Early phases of human behaviour: models in Lower Palaeolithic archaeology. In Models in archaeology (ed. D.L. Clarke). London: Methuen, pp. 167-199. Isaac, G.L1. 1981. Archaeological tests of alternative models of early hominid behaviour: excavation and experiments. Philo­ sophical Transactions o f the Royal Society o f London, Series B. 292: 177-188. Potts, R. 1994. Variables versus models of early Pleistocene hominid land use. In Early hominid behavioural ecology (eds J.S. Oliver, N.E. Sikes and K.M. Stewart). Journal o f Human Evolution 27: 7-24. Preece, R.C., Lewis, S,.G., Wymer, J.J. Bridgland, D.R. and Parfitt, S. 1991. Beeches Pit, West Stow, Suffolk. In Central East Anglia and the Fen Basin. Field Guide (eds S.G. Lewis, C.A. Whiteman and D.R. Bridgland) Quaternary Research Associ­ ation, London, pp. 94-104. Roebroeks, W., Conard, N.J. and van Kolfschoten, T. 1992. Dense forests, cold steppes, and the Palaeolithic settlement of northern Europe. Current Anthropology 33: 551-586. Skertchly, S.B.J. 1877. The geology o f the Fenland. Memoir of the Geological Survey of Great Britain. HMSO, London.

Are the Sangoan and Acheulean ‘Industrial Complexes’ Distinct? Julie L. Cormack Department o f Anthropology, University of California, Berkeley, California 94720-3710, U.S.A.

Abstract This paper tests the hypothesis that certain core-based implements generally labelled as Sangoan are larger in size and different in form to similar implements labelled as belonging to the Acheulean Industrial Complex. The dataset that is used is from museum collections. As a result, it suffers from sample biases such as availability o f certain specimens, historical collecting biases, and interobserver biases in the use o f particular typological labels. Nevertheless, the sample is more representative than any that could be used in originally characterising the Acheulean and Sangoan. The sample of 158 implements uses McBrearty’s (1988) definition o f the Sangoan that is based on four types: picks, choppers, core scrapers, and Clark’s (1963) coreaxes. Considering these caveats, nonparametric statistical analysis o f this dataset does not support the presence o f a separate ‘Sangoan phase’ within or after the Acheulean Industrial Complex.

Introduction: the S an goan an d A ch eu lean The purpose o f this paper is to compare statistically the size and feature variation between heavy duty tools associated with the Sangoan and similarly typed tools linked with the Acheulean Industrial Complex. Although both o f these industrial traditions are associated with the African Early Stone Age, the Acheulean is known more widely for its ubiquitous bifaces (handaxes and cleavers), whereas the Sangoan is largely synonymous with ‘heavy duty’ imple­ ments. In 1962, Maxine Kleindienst recognised five types o f heavy duty tools: picks, core scrapers, trimmed pebbles or chunks, choppers, and stone balls. The interpretation generally applied today is that several of these artefacts, specifically choppers, coreaxes, picks, and core scrapers belong to this group and when co-occurring they are referred to as Sangoan. Figures 12 and 13 (pages 96 and 97) in Clark and Kleindienst (1974) illustrate som e o f these tool types. Choppers, which are quite varied in shape, do not have a regular plan form. They can be bifacially flaked and often leave remains o f the cortex. Coreaxes, originally recognised and defined by Desmond Clark in 1963, are usually identified by their convergent or parallel edges. They are bifacially flaked objects with an unworked butt end. Picks are generally viewed as large tools with a prominent point and often an unflaked ventral surface, but they can also have a ridge across the dorsal surface. This latter type, a ‘rostrocarinate’ form was originally defined by Lankester in 1919 as: elongate with a central carina (or keel) and a flat, ventral, and often unflaked reverse face. For (core) scrapers, Klein­ dienst (1962) and Mary Leakey (1971) describe the most prominent feature o f this tool type as its high-backed nature with flaking usually carried out from a flat ventral face. These four implement types are considered typically heavy duty, and together they are referred to as Sangoan, a term noted by McBrearty (1988: 389) as ‘a very loosely defined archaeological entity.’ The name Sangoan comes from the

type site, Sango Hills located along the western shores of Lake Victoria in Uganda where these larger and chunkier implements were first found in the 1920s by Edward James Wayland, then Director o f the Geological Survey (Smith and Wayland 1923). In his studies o f East African materials, van Riet Lowe (1952) noted that the most outstanding tool of these assemblages was the elongate rostrocarinate or trifaced pick. The idea o f large crude forms, especially picks, comprising the Sangoan is persistent in today’s definition, although evidence particularly from the Sangoan horizon (B2/IV) at Kalambo Falls, which contains 74.8% light duty tools, suggests the inclusion o f smaller artefacts as well (Clark 1964).

The Dataset The dataset for this work comprises 158 artefacts from several African sites (Figure 1). Representing the Sangoan are objects from Nsongezi, Kalambo Falls, and the type site o f Sango Hills. Two picks are also identified from m iscel­ laneous sites in Uganda. Representing the Acheulean are various artefacts labelled from Nsongezi, Olduvai Gorge, Isimila, Kilombe, and Olorgesailie. This small dataset which com es from museum collections is not a random sample, and as such represents the difficulties which archaeologists must consider when dealing with highly selective museumbased collections. The sample is biased because o f certain limitations o f museum collections. First, only some objects (e.g. those stored within a research collection) may be readily available for study. Second, the availability o f objects in today’s museums is limited to what the original collector felt was important to collect, to catalogue, and to preserve. Third and especially relevant in this study, is the bias of how objects are perceived and labelled within a particular typological framework. Artefacts included in this Sangoan subset were those objects either labelled Sangoan or contained in boxes identified as Sangoan. There is certainly

396

Ju lie L. C orm ack

Artefact Type choppers coreaxes picks scrapers

Nsongezi

Sango Hills

2 2 9 4

6

Kalambo Falls Uganda 3 1 11 19

17 12

Total 11 3 39 35

2

A. Sangoan Artefact Type choppers coreaxes picks scrapers

Nsongezi Olduvai Gorge 3 2 1 1

3 1

Isimila 6 3 16 16

Kilombe Olorgesailie 17

1

Total 27 5 20 18

B. Acheulean

Figure 1. Frequency o f artefacts by site.

a

potential for ‘error’ using this approach; it should be emphasised, however, that this paper gives preliminary results to the question o f whether the Sangoan and Acheulean ‘Industrial Com plexes’ are distinct(ive) or not. It is certainly based on a dataset at least as representative as that originally used to make the Acheulean/Sangoan distinction. More substantial comments are expected to be added to this discussion by future consideration o f a less-biased and more contextually-relevant dataset, using artefacts from better defined archaeological assemblages. It would be valuable, for example, to reconsider this question in light o f evidence from, say, Kalambo Falls with its in situ Sangoan and Acheulean assemblages.

Results Two nonparametric statistical tests were applied to the data­ set. Chi-squared tests were used to compare nominal (or categorical) variables between the Acheulean and Sangoan, whereas the Mann-Whitney test compared median values o f ratio data such as length, breadth, and thickness measure­ ments. Figure 2 lists the results o f the chi-squared tests. Note the

Variables blank form plan-view shape cross-sectional shape roundness surface texture raw material working face number working edge type presence of bifurcation presence of secondary trim presence of fractures presence of central keel presence of battering

X2 Value 11.36 4.75 12.81 1.22 2.54 50.52 0.47 2.93 0.26 0.18 4.51 0.01 1.89

DF 2 2 6 1* 1 4 1 2 1 1 1 1 1

p Value 0.003*** 0.092 0.045*** 0.268 0.110 0.000*** 0.489 0.230 0.608 0.666 0.033*** 0.889 0.168

* In cases where only two categories are considered, Cramer (1994) recommends the use of Yales’ correction for continuity. Thus, in these cases where there is only 1 degree of freedom, the X2 and p values listed are calculated with this correction in mind. *** Significant at the alpha level 0.05 F ig u re 2. R esu lts o f X 2 tests.

p values marked with three asterisks in the far right-hand column. For this dataset, four variables were statistically significant: blank form, cross-sectional shape, raw material, and the presence o f fractures. Thus, in these cases, where there is only one degree o f freedom, the chi2 and p values listed are calculated with this correction in mind. Each o f the four statistically significant results will be discussed briefly.

Blank Form The statistically significant p value infers a ‘real’ difference in blank form use between the available Acheulean and Sangoan artefacts. This statement is supported by my observation and a general perception that A cheulean assemblages are more ‘flake-based’. Surprisingly in this sample, three often Acheulean flake blanks were recognised as picks and seven as scrapers.

Cross-sectional Shape The p value o f 0.045 is less than the alpha level and, therefore, represents a statistically significant difference between the observed and expected frequencies o f crosssectional shapes of artefacts in this dataset sample. A better contextually defined sample would probably clarify the ‘true differences’ between these assemblages.

Raw material. Industrial complex and raw material is highly statistically significant; a result which indicates a ‘real’ difference in raw material usage. This observation is clear when one considers the extensive use of quartz/ite in the supposed Sangoan (the abbreviation ‘quart/zite’ is used to represent either quartz or quartzite). In this sample, the total was 93% (Cormack 1994).

Fractures The frequency o f fractures is statistically significant. Although there is no direction in this result, in 1967, Glen Cole stated that, step flaking, or in the terminology used here - fracturing, was a trait more characteristic o f Sangoan than Acheulean artefacts.

Are the Sangoan and Acheulean Industrial Complexes ’ Distinct?

Variables blank form Acheulean Sangoan raw material Acheulean Sangoan presence of fractures Acheulean Sangoan

X2 Value

DF

397

p Value

39.77 25.95

6 6

0.000*** 0.000***

10.34 4.86

6 6

0.111 0.561

4.00 5.81

3 3

0.260 0.121

*** Significant at the alpha level 0.05

Figure 3. Statistically significant results. □ Sangoan A breakdown of three o f the statistically significant results is shown in Figure 3. Blank form still produces a statistically significant result which suggests that in ~]h the Acheulean and Sangoan samples, artefact type can be distinguished by the use o f particular blank forms. For the test o f raw material, even though quartz/ite is used overwhelm ingly in the Sangoan, chi2 tests for each ‘industrial com plex’ do not produce a statistically significant result, thus suggesting that for both samples, tool type is not dependent upon raw material. The low chi2 values in the third comparison for each sample also suggest that the presence o f fractures is independent o f tool type. Average values for the measurements o f the four tool types are illustrated in figures 4 to 7. Note in all four bar graphs, that in a comparison of length, breadth, thickness, and total perimeter across the four tool types, artefacts identified with the Sangoan are consistently larger in size for coreaxes. Results o f the Mann-W hitney tests which compare medians in these dimensions are given in fig. 8. There are no statistically significant differences between choppers identified as Sangoan and those identified as Acheulean. N one o f the two-tailed probabilities for the sample o f coreaxes is statistically significant either. However, it should

Artefact Type Jr

□ Acheulean □Sangoan

Figure 6. Average thickness comparison (N=148)

Figure 4. Average length comparison (N=158)

Artefact Type

__________ □Acheulean □Sangoan

Figure 5. Average breadth comparison (N=158)

Artefact Type

□ Acheulean □ Sangoan

Figure 7. Average total perimeter comparison (N-112)

Julie L. Cormack

398

Variables choppers (N=38) length breadth thickness total perimeter (N= 18) coreaxes (N=8) length breadth thickness total perimeter (N=7) picks (N=59) length breadth thickness total perimeter (N=50) scrapers (N=53) length breadth thickness total perimeter (N=37)

M-W U Value

p Value

93 104 136.5 26.5

0.076 0.158 0.703 0.237

4.5 3 6.5 4

0.392 0.250 0.785 0.857

266.5 232 284.5 198

0.047*** 0.011*** 0.090 0.043***

270 266.5 240.5 95

0.397 0.361 0.161 0.020***

*** Significant at the alpha level 0.05

Figure 8 . Results o f Mann- Whitney tests. be stressed that this result for coreaxes is based on very small samples: Acheulean (5), Sangoan (3) and it can therefore be suggested that this test may be inappropriate. It is proposed that the comparison o f average values in the bar graphs which show the Sangoan with smaller dimensions to the Acheulean is probably an effect of this small sample size compounded by the fact that the eight coreaxes come from four different archaeological sites. This, however, empha­ sises the relatively small number of these artefacts described, and their limited value for distinguishing between the Acheulean and Sangoan. In contrast, the differences in picks are statistically significant in length, breadth, and total perimeter and therefore indicate that this artefact can be distinguished on the basis o f industrial complex. Although this two-tailed result does not provide a direction o f difference the bar graphs showed that the Sangoan averages were consistently larger than the Acheulean values. Scraper measurements are only significantly different in total perimeter. Thus, the overall results o f Mann-Whitney tests can be stated as follows. The lack of difference in Acheulean and Sangoan choppers, although based on a smallish sample, provides a strong suggestion that choppers cannot be used as a diagnostic element in establishing the two ‘industrial com plexes’. In fact, since choppers are found throughout the Early Stone Age (particularly in the Oldowan) there is no case for isolating them as exclusively a Sangoan ‘type’. Although coreaxes have been thought to represent a unique Sangoan type it is hard to confirm this on the basis o f the small group examined here. Evidence o f ‘coreaxe-like’ forms from Acheulean horizons at Olduvai Gorge, Isimila, and perhaps Kilombe (Gowlett 1994, pers. comm) clearly demonstrates the need for a larger and better sample. Picks on the other hand, have always been viewed as a typical Sangoan form although Isaac (1977) and Biberson (1961) have given evidence at Olorgesailie and in North A frica, resp ectively, which places these tools in the Acheulean. The results o f the Mann-Whitney tests, which

gives statistically significant results for three o f the four measurement values, supports the observation that Acheulean picks are significantly different to Sangoan picks. The variable thickness is not significantly different at the 0.05 alpha level. Thus, it could be initially suggested that picks should remain a characteristic type o f the Sangoan, but that more comparative studies should be conducted on the Acheulean forms. Considering that scrapers are a highly variable group usually defined by edge type and related traits, it is surprising that o f a maximum sample o f 53, only total perimeter shows a statistically significant difference between the Acheulean and the Sangoan. Like choppers, this implement is not solely characteristic o f the Sangoan since the general form, ‘core scraper’ is found from the Oldowan onwards. For these two reasons - lack o f a more detailed type definition and persistence throughout the Early Stone Age - scrapers should not be isolated as a distinguishing feature o f the Sangoan.

Discussion and Conclusion Clearly from this work the ‘Sangoan’ can only be distin­ guished on the basis o f four possible criteria: differences in blank form usage, an increased use o f quartz/ite; the presence o f larger picks; and a difference in the frequency o f fractures. It has been demonstrated that quartz/ite is not associated with particular implements and can therefore be suggested that it is more likely that emphasis on this raw material is a result o f availability. The increased number o f fractures in the Sangoan compared to the Acheulean is probably also correlated with difficulty in using quartz/ite. Of the four tool types considered, picks seem to be ‘typical’ Sangoan forms, but their occurrence in the Acheulean must still be considered. There may be, for instance, a case for saying that Sangoan picks exceed in weight and size anything known in the Acheulean. As for coreaxes, these objects have often been mislabelled as handaxes, picks, or even scrapers. Their role in the question o f Sangoan versus Acheulean assemblages is important and we, as archaeologists, must consolidate their typological description. The term, ‘Sangoan’ is often linked with the phrase, ‘heavy duty’ since both refer to the larger and bulkier objects in a lithic tool kit. But here the results show that in a comparison o f Sangoan artefact measurements to Acheulean tools, there is not much statistically significant difference. In addition, the tool types that have been considered are not isolated to one geographical region. For instance, picks and coreaxes have been found in South and W est African contexts (Hanish 1958; Davies 1967,1976; Beaumont 1995 pers. comm). Also, choppers and scrapers are not restricted spatially sin ce they are known to occur beyond the boundaries o f Africa. Recommendations o f the Burg Wartenstein conference (Bishop and Clark 1967) defined an industry as represented by ‘all the known objects that a group o f prehistoric people manufactured in one area over some span o f time.’ Since it has been shown that typologically the Sangoan could include picks and that this ‘industry’ is not limited to a single temporal or spatial framework, we should question the use o f the Sangoan as a precise industrial marker. It is proposed, therefore, that the term Sangoan be dropped as an ‘industry’ (or ‘industrial complex’) and thatthe four tool types described here be retained as categories o f heavy duty artefacts. Just

Are the Sangoan and Acheulean ‘Industrial Complexes ’ Distinct? as importantly though, these tools should be viewed as a contributing and functional component of the archaeological lithic tool kit.

Acknowledgments My research was conducted on lithic collections from the following institutions: the Museum o f Archaeology and Anthropology, Cambridge University; the Donald BadenPowell Quaternary Research Centre and Pitt Rivers Museum, Oxford University; Franks House, The British Museum, London; and the Field Museum o f Natural History, Chicago. I would like to acknowledge, with sincere thanks, the people who provided me access to these collections and those who offered assistance, in particular D. Phillipson, R. Inskeep, D. Roe, N. Ashton, and G. Cole. Thanks especially to John Gowlett who graciously read this paper at the Archaeological Sciences Conference. Statistical consulting was provided by S. Byrne (University of Liverpool Computer Laboratory) and M.J. Platt (University of Liverpool, Department of Public Health). Financial support was provided by the Social Sciences and Humanities Research Council o f Canada; the B o ise Fund, Oxford U niversity; and the B ill Bishop Memorial Trust.

References Biberson, P. 1961 .Le Paleolithique inferieur du Maroc atlantique. Rabat, Publications du Service des Antiquites du Maroc. Fascicule 17. Bishop, W.W. and Clark, J.D. 1967. Discussion on Terminology, pages 861-875 and Recommendations with French translations, pages 870-901. In: Background to Evolution in Africa (eds. Bishop, W.W. and J.D. Clark). London: University of Chicago Press. Clark, J.D. 1963. Prehistoric Cultures of Northeast Angola and Their Significance in Tropical Africa. Lisboa: Companhia de Diamantes de Angola. Publicacoes Culturais No. 62. Clark, J.D. 1964. The Sangoan Culture o f Equatoria; The Implication o f its Stone Equipment. Diputacion Provincial de Barcelona, Instituto de Prehistoria y Arqueologia. Monografias IX pp. 309-325.

399

Clark, J.D. and Kleindienst, M.R. 1974. Chapter 4. The Stone Age Cultural Sequences: Terminology, Typology and Raw Material. In: Kalamho Falls Prehistoric Site. II. The Later Prehistoric Cultures, (ed. J.D. Clark). Cambridge: Cambridge University Press, pp. 71-105. Cole, G.H. 1967. The Later Acheulian and Sangoan of Southern Uganda. In: Background to Evolution in Africa (eds. Bishop, W.W. and J.D. Clark). London: University of Chicago Press, pp. 481-528. Cormack, J.L. 1994. Early Stone Age Heavy Duty Implements o f Africa. Unpublished PhD Thesis. University of Liverpool. Cramer, D. 1994. Introducing Statistics fo r Social Research. Stephy-Step Calculations and Computer Techniques Using SPSS. New York: Routledge. Davies, O. 1967. West Africa Before Europeans: Archaeology and Prehistory. London: Methuen and Company Limited. Davies, O. 1976. The ‘Sangoan’ Industries. Annals, Natal Museum 22(3): 885-91 1. Hanish, H. 1958. A possible Sangoan from East London. South African Archaeological Bulletin 13: 159-160. Isaac, G.L1.1977. Olorgesailie. Archaeological Studies o f a Middle Pleistocene Lake Basin in Kenya. Chicago: University of Chicago Press. Kleindienst, M.R. 1962. Components of the East African Acheulian Assemblage: An Analytic Approach. In: Actes du IVe Congres Panafricain de Prehistoire et de Petude du Quaternaire (eds. Mortelmans, G. and J. Nenquin). Tervuren, Belgique:Musee Royal d’Afrique Centrale, pp. 81-111. Lankester, R. 1919. On the Discovery of a Novel Type of Flint Implements Below the Base of the Red Crag of Suffolk, Proving the Existence of Skilled Workers of Flint in the Pliocene Age. Philosophical Transactions o f the Royal Society o f London 202: 283-336. Leakey, M.D. 1971. Olduvai Gorge. Volume 3. Excavations in Beds I and II. Cambridge: Cambridge University Press. McBrearty, S. 1988. The Sangoan-Lupemban and Middle Stone Age Sequence at Muguruk Site, Western Kenya. World Archaeology 19(3): 388-420. Smith, R.A. and Wayland, E.l. 1923. Some primitive stone implements from Uganda. Occasional Paper No. 1, Geological Survey of Uganda. van Riet Lowe, C. 1952. The Pleistocene Geology and Prehistory o f Uganda. Part ii. ColchestenAuthority of the Ugandan Government.

The Acheulean and the Sahara: Allometric Comparisons Between North and East African Sites R.H. Crompton1and J.A.J. Gowlett2 1 Department o f Human Anatomy and Cell Biology, University o f Liverpool, Liverpool L69 3BX, UK; 2 Department o f Archaeology, University o f Liverpool, Liverpool L69 3BX, UK

Bifaces or hand-axes, are the hallmark of the Acheulean, and represent one o f the most impressive, and certainly the most long-lived, design form in prehistory. The variation in bifaces, coupled with their long persistence and the reoccurrence o f traits, has puzzled many prehistorians, who have commented on the strange ‘variable sameness’ o f the record, and on its enigmatic forms (e.g. Isaac 1977; Wynn 1995). In previous work we have shown that some o f the variation in form is linked with size (Crompton and Gowlett 1993; Gowlett and Crompton 1994). We used multivariate allometry as a tool for exploring and accounting for the variation. Our initial work was confined to East African sites. We showed first that allometric patterns were marked and generally consistent within the Kilombe site complex. Then we showed, by introducing the sites o f Kariandusi and Kapthurin, that allometry was widespread in East African assemblages, and that it also had a long time duration - at least from one million years down to about 250,000 years ago. We now extend this work to a comparison between East and North African material (fig. 1).

The Dataset The work is based on two assemblages from the Casablanca area of Morocco from sites first described by Biberson

(1961a and b; fig. 2). The first collection from the STIC site is estimated to be around 700 ka old (Raynal and Texier 1989; Clark 1992), on grounds o f its stratigraphic position and faunal assemblage. Raynal and Texier (1989) reported the discovery of a new site in a comparable stratigraphic position in level L in the nearby Thomas 1 quarry. Both come from the beginning o f the Amirian phase, and the two assemblages resemble one another much more than material from another well-known site at Tighenif (Ternifine) in Algeria, which may be somewhat later. The Casablanca sites are overlain by the massive consolidated dune (H) o f the Amirian. Stalagmite in an erosive karstic feature within this has been dated by U -series to 409 ± 88 ka, broadly confirming the antiquity o f the sites (Raynal and Texier 1989). The STIC site is thus one o f the oldest in northern Africa, and may approach Kilombe and Kariandusi in age (Crompton and Gowlett 1993; Gowlett and Crompton 1994). Measurements had been taken (by JAJG) and were available from 300 o f the 540 biface specimens collected by Biberson. The finds are similar in size range to the lava series from Kilombe or Kariandusi (fig. 3), but the level o f standard­ isation is poor, reflected in far lower correlations o f the major dimensions (fig. 4). Compared with Kilombe and Kariandusi the finds are also relatively pointed and very thick - the mean for thickness/breadth (T/B) in the series is 0.60. There appears to be a continuous gradation from

Figure 1. Map o f selected East and North African Acheulean sites

The Acheulean and the Sahara

401

Figure 3. Acheulean bifaces: sketches o f two specimens from the STIC collection, madefrom cobbles and retaining cortex.

Biberson noted the work in the contemporary Grotte du Rhinoceros by P. Mieg de Boofzheim and C. Plessis, who described a Mousterian with bifaces. The bifaces thus belong to a final Acheulean which might also be regarded as earliest Middle Palaeolithic. Similarly early dates for the first Middle Palaeolithic are increasingly accepted in the Middle East and Europe (Mercier etal. 1995). The Cunette artefacts may be roughly the same age as the Kapthurin collection which we studied in East Africa (Gowlett and Crompton 1994), though in general form and dimensions they resemble more the far older obsidian series from Kariandusi (fig. 4). Before introducing the North African bifaces to the analysis, we framed the following surmises and hypothesis:

Figure 2. Map and sections o f the Sidi Abderrahman localities after Biberson (1961a)

ordinary bifaces to thicker specimens sometimes classified as picks (the status of the latter is considered by Cormack 1994, this volume). The second collection from Sidi Abderrahman Cunette is from a much later period. 128 specimens were measured. These come from a part o f the Sidi Abderrahman quarry called ‘the Cunette’, in which there was a quarry front nearly 700 metres long (fig. 2). The ancient consolidated sand dune (H) was eroded during one or more high sea levels, so that several caves formed in it. It was later degraded and buried by terrestrially-derived sediments. D 2 is the upper layer in this series, which Biberson marks both in the Grotte de POurs and in the Grotte des Littorines. The finds are believed to come from excavations o f 1957-58 at the Cap Chatelier, in the middle part of Cunette, and are ascribed by Biberson to the final Acheulean (his stage VIII). The sediments are a brecciated pink limestone, up to 6 m deep. This layer overlies the 18-20 m beach, but is cut by a later high sea-level, locally termed the Ouljian, which possesses U-series dates o f 12 0 -1 4 0 ,0 0 0 (Clark 1992; Hublin 1985). The sites therefore almost certainly antedate the last interglacial. The caves may have been cut in isotope stage 7, 9 or 11, suggesting an age in the range 170-360 ka for the sediment­ ation and artefacts (see also Clark 1992). In discussion

1. the Cunette results might be similar to the Kariandusi obsidian series; 2. the STIC pattern was unpredictable: the programme works from linear relations between variables, and where correlations are low, results might not be (formally) significant. The STIC collection was potentially interest­ ing because it has the same size range as Kilombe and Kariandusi, but a very different shape range.

Length (L) Breadth (B) Thickness (T)

STIC (N-276-229) 161 ± 29 91 ± 17 53 ± 13

Cunette D2 (N=128-105) 114 ±31 79 ± 19 36 ± 9

Correlation L&B T&B

0.66 0.31

0.88 0.61

Kilombe EH

Kariandusi lava

Length (L) Breadth (B) Thickness (T)

(N=106-95) 153 ±30 94 ± 18 43 ± 10

(N=100-l 10) 163 ± 24 94 ± 11 49 ± 9

Kariandusi obsidian (N=60) 123 ± 27 79 ± 13 37 ± 7

Correlation L&B T&B

0.83 0.46

0.59 0.17

0.77 0.68

Figure 4. Table of the biface data: basic measurements

R.H. Crompton and J.A.J. Gowlett

402

The groups o f sites are 5,000 km apart. If similar allometric adjustments were shown, it would demonstrate similar basic constraints in the Acheulean over a far wider area.

T he A n alyses Details o f the analytical procedures have already been given (Crompton and Gowlett 1993; Gowlett and Crompton 1994). The main analysis is o f multivariate allometry, which depends upon Jolicoeur’s multivariate generalization o f the allometry equation, (y=bxa)(Jolicoeur 1963). It is based on the angular relationships o f co-ordinates for each measure to the first principal component of the covariance matrix. This allows us to consider departures from isometry in each of the measured variables, without having to rely on any o f them individually as indicating ‘size5. In our original study we used artificially generated sets o f isometrically and allometrically varying specimens in order to be sure that we were deriving a coefficient by an appropriate extraction technique (Crompton and Gowlett 1993). We have also used principal components analysis to probe the structure of the datasets, and discriminant analysis to determine the extent to which bifaces are characteristic o f a particular assemblage.

M u ltivariate A llom etry R esu lts Results o f the Multivariate allometry study are given in fig. 5. The coefficient value o f 1.0 indicates isometry. Higher values indicate positive allometry, and values lower than 1.0, negative allometry. The following summary points can be made:

1. Both the early STIC series and late Cunette series show marked patterns o f allometry, confirming its importance through the time range already demonstrated in East Africa. 2. The pattern o f allometric adjustment is remarkably similar to that seen in East Africa, without any anomalies such as observed in bifaces from Kilombe Area Z. 3. The pattern is less markedly present at Sidi Abderrahman Cunette, which tends more towards isometry than any other series which we have examined. This may be because o f (a) high geometric standardisation in this series and/or (b) the small size o f most specimens (mean 114 mm), which would automatically eliminate the adjustments seen in very large bifaces (although the series included a small number o f very large specimens). 4. The pattern at STIC closely resembles that at Kilombe and in the Kariandusi lava series. This is interesting first because the pattern is still present even though the material is poorly standardised; second, it is still there even though the STIC assemblage contains far more pointed specimens, and lacks the cleavers found at Kilombe and Kariandusi. Thus the pattern which has been isolated is not simply a ‘cleaver factor5, such as might arise if small specimens tended to be hand-axes and large specimens tended to be cleavers. The datasets treated here present some considerations additional to those o f our previous work. The TA measure­ ment (thickness near the tip, at 0.8 length from base) had been taken only for some STIC specimens and few from Cunette. This was unfortunate since at Kilombe it was found to be a key discriminator in showing that the very thick bifaces from area Z had relatively thin tips: i.e. probably for

Kariandusi obsidian

Cunette Kilombe MM

~®~ Kariandusi Lava Kilombe KZ S TIC

Significantly different from isometry: KaObs: BA (