144 29 73MB
English Pages 777 Year 2017
The Oxford Handbook of
ARCHAEOLOGICAL CERAMIC ANALYSIS
The Oxford Handbook of
ARCHAEOLOGICAL CERAMIC ANALYSIS Edited by
A L IC E M . W. H U N T
1
3 Great Clarendon Street, Oxford, ox2 6dp, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © Oxford University Press 2017 The moral rights of the authorshave been asserted Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2016944779 ISBN 978–0–19–968153–2 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.
Acknowledgments
This Handbook would not have been possible without the hard work and expertise of its contributors. I also owe a debt of gratitude to Hilary O’Shea, Charlotte Loveridge, Annie Rose, Michael De la Cruz, and the rest of the OUP team for helping to bring this volume to fruition with minimum stress and maximum enjoyment. Special thanks to Jeff Speakman and the Center for Applied Isotope Studies, University of Georgia, for a publication subvention that allowed us to include the colored plates. Many of the individual contributors wish to thank various colleagues and associates for reading and commenting upon their contributions, and sharing unpublished materials; space limitations preclude acknowledging each individual by name, and so consider this a heartfelt, if general, round of thanks and appreciation to all involved behind the scenes.
Contents xi xix xxi xxiii xxv
List of Figures List of Tables List of Plates List of Abbreviations List of Contributors
PA RT I I N T RODU C T ION 1. Introduction to the Oxford Handbook of Archaeological Ceramic Analysis Alice M. W. Hunt 2. History of Scientific Research Michael S. Tite
3 7
PA RT I I R E SE A RC H DE SIG N A N D DATA A NA LYSI S 3. Designing Rigorous Research: Integrating Science and Archaeology Jaume Buxeda i Garrigós and Marisol Madrid i Fernández
19
4. Evaluating Data: Uncertainty in Ceramic Analysis Roberto Hazenfratz-Marks
48
5. Statistical Modeling for Ceramic Analysis Gulsebnem Bishop
58
6. Recycling Data: Working with Published and Unpublished Ceramic Compositional Data Matthew T. Boulanger
73
PA RT I I I F O U N DAT IONA L C ON C E P T S 7. Ceramic Raw Materials Giuseppe Montana
87
viii Contents
8. Ceramic Manufacture: The chaîne opératoire Approach Valentine Roux 9. The Organization of Pottery Production: Toward a Relational Approach Kim Duistermaat
101
114
10. Provenance Studies: Productions and Compositional Groups Yona Waksman
148
11. Mineralogical and Chemical Alteration Gerwulf Schneider
162
12. Formal Analysis and Typological Classification in the Study of Ancient Pottery Daniel Albero Santacreu, Manuel Calvo Trias, and Jaime García Rosselló
181
13. Fabric Description of Archaeological Ceramics Ian K. Whitbread
200
14. Analytical Drawing Prabodh Shirvalkar
217
PA RT I V E VA LUAT I N G C E R A M IC P ROV E NA N C E 15. Petrography: Optical Microscopy Dennis Braekmans and Patrick Degryse
233
16. Ceramic Micropalaeontology Ian P. Wilkinson, Patrick S. Quinn, Mark Williams, Jeremy Taylor, and Ian K. Whitbread
266
17. Electron Microprobe Analysis (EMPA) Corina Ionescu and Volker Hoeck
288
18. Isotope Analysis Bettina A. Wiegand
305
19. X-Ray Powder Diffraction (XRPD) Robert B. Heimann
327
Contents ix
20. X-Ray Fluorescence-Energy Dispersive (ED-XRF) and Wavelength Dispersive (WD-XRF) Spectrometry Mark E. Hall
342
21. Handheld Portable Energy-Dispersive X-Ray Fluorescence Spectrometry (pXRF) Elisabeth Holmqvist
363
22. Particle Induced X-ray Emission (PIXE) and Its Applications for Ceramic Analysis Marcia A. Rizzutto and Manfredo H. Tabacniks
382
23. Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) and Laser Ablation Inductively Coupled Plasma-Mass Spectrometry (LA-ICP-MS) Mark Golitko and Laure Dussubieux
399
24. Instrumental Neutron Activation Analysis (INAA) in the Study of Archaeological Ceramics Leah D. Minc and Johannes H. Sterba
424
25. Synchrotron Radiation Alan F. Greene
447
PA RT V I N V E ST IG AT I N G C E R A M IC M A N U FAC T U R E 26. Ethnography Kent D. Fowler
469
27. Experimental Firing and Re-firing Małgorzata Daszkiewicz and Lara Maritan
487
28. Fourier Transform Infrared Spectroscopy (FT-IR) in Archaeological Ceramic Analysis 509 Shlomo Shoval 29. Raman Spectroscopy and the Study of Ceramic Manufacture: Possibilities, Results, and Challenges Jolien Van Pevenage and Peter Vandenabeele 30. X-Radiography of Archaeological Ceramics Ina Berg and Janet Ambers
531 544
x Contents
31. Organic Inclusions Marta Mariotti Lippi and Pasquino Pallecchi
565
PA RT V I A S SE S SI N G V E S SE L F U N C T ION 32. Formal Typology of Iberian Ceramic Vessels by Morphometric Analysis Ana Luisa Martínez-Carillo and Juan Antonio Barceló
585
33. Mechanical and Thermal Properties Noémi S. Müller
603
34. Assessing Vessel Function by Organic Residue Analysis Hans Barnard and Jelmer W. Eerkens
625
PA RT V I I DAT I N G C E R A M IC A S SE M B L AG E S 35. Typology and Classification Eugenio Bortolini
651
36. Direct Dating Methods Sophie Blain and Christopher Hall
671
Index
691
List of Figures
3.1 Diagram flow of the states of ceramics from manufacture to the archaeological record.
20
3.2 Bar chart of Hispanic Terra Sigillata from Tritium Magallum recovered at Baetulo, Tarraco, and Ilerda, classified according the range of estimated equivalent firing temperatures.
28
3.3 Bar chart of Hispanic Terra Sigillata from (a) context LL85b context and (b) context TV83, and (c) a bivariate diagram of integrity (H2) vs fragmentation (FI).
37
3.4 Scatter plots of evenness for Hispanic Terra Sigillata from (a) context LL85b and (b) context TV83, and (c) evenness of the rarefaction experiment. A bar chart of the richness after the rarefaction experiment is presented in (d).
40
3.5 Binomial probabilities for n = 15 and p = 0.1.
41
5.1 Bar charts describing the (a) distribution and (b) relative frequency of vessel types in an assemblage.
61
5.2 Pie charts describing the relative distribution of vessel types in funerary assemblages from Athens and Sparta.
61
5.3 Histogram of amphora capacity measurements from a hypothetical shipwreck.
63
5.4 Common shapes of data distribution.
63
5.5 Bimodal distribution of mineral inclusions in a ceramic fabric.
64
5.6 Stem-and-leaf plot (worked example) of vessel weights.
65
5.7 Back-to-back stem-and-leaf plot comparing cooking pot volumes from two sites. 65 6.1 Timeline of a selection of former and current nuclear archaeometry laboratories, and estimates of the total numbers of archaeological specimens analyzed. Data compiled primarily from vol. 49(2) of Archaeometry. 77 7.1 “Integrated approach” for characterizing and sourcing ceramic raw materials.
89
7.2 Examples of primary and secondary clays: (a) kaolinite deposits in the crater of Mount Gibele at the volcanic island of Pantelleria (Italy); (b) outcrop of Upper Miocene marine clays in southern Sicily.
91
7.3 Brick and roof tile makers in western Sicily traditionally using NaCl as deflocculating agent.
97
8.1 Classification chart of roughing out and preforming techniques.
105
8.2 Diagnostic features taken into account for reconstructing an Early Bronze Age chaîne opératoire from the site of Tell Arqa (Lebanon).
106
xii List of Figures 8.3 Example of technostylistic trees obtained after classifying ceramic assemblages according to the concept of chaîne opératoire.
108
9.1 Entanglements of the materials used to make a carinated bowl.
127
9.2 Chaîne opératoire for a Middle Assyrian carinated bowl.
130
9.3 Entanglement of the life-history of carinated bowls, from production until deposit in the archaeological record.
132
10.1 Local reference samples, late Byzantine workshops, Thessaloniki, Greece.
151
10.2 Beirut medieval wares: main compositional groups as determined by hierarchical clustering analysis, and corresponding wares.
155
10.3 Beirut medieval wares: binary plot iron–silicon (top) and histogram of Mahalanobis distances (bottom).
157
11.1 Correlation of barium and phosphorus in Roman and Germanic pottery from two sites in Germany.
170
11.2 Leaching of calcium in two samples of calcareous pottery.
173
12.1 Summary of the different levels to approach pottery form and typological analyses discussed in the text.
182
12.2 Isomorphic relation between the decorative motifs recorded on Late Iron Age pottery and bronze discs in Mallorca (Spain).
187
12.3 Format translation related to hybridization phenomena between Punic wheel-thrown vessels and hand-made indigenous pottery in the Late Iron Age in Mallorca (Spain).
194
14.1 Art-historical period pottery illustrations.
218
14.2 Steps of traditional pottery illustration (part 1).
220
14.3 Steps of traditional pottery illustration (part 2).
221
14.4 Various aspects of pottery illustration.
222
14.5 Steps of new pottery illustration (part 1).
225
14.6 Steps of new pottery illustration (part 2).
226
14.7 Steps of new pottery illustration (part 3).
228
15.1 A standard polarizing light microscope with rotating sample stage. The camera and imaging software are essential tools for data output.
235
15.2 High relief and cleavage of a pyroxene mineral.
238
15.3 Photomicrograph of a Late Roman, quartz-tempered cooking vessel from Carthage (Tunisia).
258
16.1 The earliest known image of a microfossil.
267
16.2 Examples of the main microfossil groups that may be found in ceramics.
268
16.3 Examples of microfossils in ceramic matrices that can be applied in provenance studies.
273
16.4 Degradation of microfossils during firing.
274
List of Figures xiii 17.1 (a) Origin of BSE, SE, and characteristic X-rays emitted by the interaction between the focused electron beam and sample. (b) BSE image of a ceramic sample from Ibida (Roman-Byzantine period). (c) SE image of the same.
290
17.2 BSE images of various compounds of a ceramic body.
294
17.3 X-ray maps of Cucuteni ceramics (Copper Age).
295
18.1
Distribution of 87Sr/86Sr ratios in ceramic samples from archaeological sites of
18.2
Distribution of 87Sr/86Sr and εNd values in ceramic samples from Turkey,
different regions.
Greece, Bulgaria, and China.
315 316
18.3 Pb isotope data of lead-based glazes from New Mexico, USA, China, and various European locations.
319
19.1 (a) Permitted electron transitions to generate X-rays of the K series. (b) Interpretation of X-ray diffraction as the result of simple reflection. (c) Seemann–Bohlin focusing geometry for Debye–Scherrer and Straumanis methods. (d) Bragg–Brentano focusing configuration for scintillation counter (powder diffractometer) method.
329
19.2 X-ray diffraction charts of an archaeological calcareous illitic clay from Otterbach, Palatinate, Germany.
335
19.3 X-ray diffraction charts of archaeological ceramics buried under arid and humid conditions.
337
19.4 X-ray diffraction chart (CuKα) of stoneware from Sawankhalok, Thailand (14th-15th centuries ad).
338
21.1 SEM-BSE micrographs of heterogeneous ceramic matrices showing sand- temper. 364 21.2 Correlation of pXRF net peak area values to quantitative NAA data.
366
21.3 PCA biplots of the variance–covariance matrix of the pottery samples measured by pXRF and INAA.
370
21.4 PCA biplots with density ellipses for clusters indicated by low-dimensional pXRF data and high-dimensional ICP-OES and NAA datasets.
371
22.1 Parameters and coordinates of the experimental PIXE geometry.
384
22.2 Typical thin film yield curve of the São Paulo PIXE system with a Si(Li) X-ray detector and a 55 μm thick beryllium X-ray filter.
386
22.3 General view of LAMFI with the ion sources on the right and the 5SDH accelerator tank in the center.
387
22.4 The external beam setup at the LAMFI with the different detectors assembled. The lower part of the figure shows in detail the assembly of the coupled detectors.
388
22.5 Chimu ceramics at the Museum of Archaeology and Ethnology of the University of São Paulo (MAE-USP) collection.
390
xiv List of Figures 22.6 (a) PIXE spectra of two ceramic vessels (3635 and 3601) of the Chimu culture; (b) comparison of the concentrations values obtained by PIXE at different points, analyzed in the two ceramic pieces.
391
22.7 (a) Graph of the correlation between Al, Si, K, Ti and Fe elements giving rise to 4 large groups; (b) graph of the correlation between Al, Si, K and Ti elements, suggesting the origin of three major groups.
392
22.8 Sun ray plots of the elements Al, Si, K, and Ti showing the correlations in the groups and the respectively vessels in each group.
393
23.1 An ICP-MS laboratory, with both laser ablation and liquid sampling equipment visible.
401
23.2 Schematic diagram of plasma-ion source and common mass spectrometer types. 402 23.3 Examples of calibration lines for Fe, Rb, and Pr using NIST610, 612, and 679 as SRMs. 407 23.4 Surface geology of the Sepik coast of Papua New Guinea, with ceramic and clay sampling locations indicated.
416
23.5 Backscatter SEM-EDS image of a sherd cross-section from Wom (Papua New Guinea) analyzed by LA-ICP-MS.
417
23.6 Discriminant function plot summarizing the results of LA-ICP-MS analysis of all ceramic and clay pastes analyzed from the Sepik coast of Papua New Guinea. 418 24.1 Schematic overview of neutron activation and subsequent decay.
426
24.2 The core of a research reactor provides a high-intensity source of neutrons to activate samples.
428
24.3 The build-up and decay of radioactivity in an isotope according to its half-life and decay constant λ. 431 24.4 Complex gamma-ray spectrum resulting from the irradiation of an earthenware vessel from Oaxaca, Mexico.
432
24.5 Ceramic composition groups defined for the central Valley of Oaxaca, Mexico, for the Late-Terminal Formative period. 440 24.6 (a) The industrial potteries and bottle kilns of North Staffordshire, England, c. 1900 (from an historic postcard labeled “Fresh Air from the Potteries”). (b) Each pottery manufacturer developed one or more paste recipes, with a distinctive trace-element composition.
443
25.1 Schematic of the Australian Synchrotron’s small-and wide-angle X-ray scattering beamline by David Cookson and Jonathan de Booy.
448
25.2 One potential schematic of an experimental setup for SR pottery analysis, here micro-X-ray fluorescence, showing the positioning of the SR beam, sample, and solid-state detector.
450
25.3 Reflected light micrograph, SEM micrograph, and µ-XR diffractogram of the yellow-glazed area of a sherd from Mission San Luis (Florida, USA) collected during the analysis of American majolica by SR.
452
List of Figures xv 25.4 The three microregions of interest on a “black gloss” or “red figure” potsherd, as investigated by Walton and colleagues with a diverse set of SR methods.
456
25.5 An example of a high-temperature XRD pattern collected during “live” time- resolved temperature experimentation.
458
26.1 The domains of ethnographic and archaeological inquiry into pottery manufacturing practices and the two principle areas of contribution ethnographic observations and explanations may lead to the understanding of past ceramic production systems.
470
26.2 The distances travelled to ceramic resources (clay and temper) by potters from 117 communities.
475
26.3 Range of firing temperatures and soaking times for firings in West Africa.
479
26.4 Models of the technical and social influences on ceramic manufacturing based upon African case studies.
481
27.1 Examples of open firing and of a traditional updraft kiln.
491
27.2 Sketch of a typical firing diagram used when programming an experimental firing in a laboratory with a furnace.
492
27.3 Bar charts showing the changes in the mineral composition as a function of the firing temperature.
494
27.4 Determination of the original firing temperature using the K-H method (Teq 900–1000°C).
500
27.5 Two samples made of the same marly clay re-fired at 1200°C.
501
28.1 Curve-fitted FT-IR spectra of the ceramic material of representative Iron Age pottery from Levantine sites.
513
28.2 FT-IR spectra and second-derivatives of the spectra of the ceramic material of representative Bronze Age pottery from Canaanite sites.
514
28.3 FT-IR spectra and second-derivatives of reference standards of firing silicates.
516
28.4 FT-IR spectra and second-derivatives of reference standards of carbonate and silica minerals.
517
28.5 FT-IR spectra and second-derivatives of reference standards of minerals typically found in initial or unfired raw materials.
518
28.6 FT-IR spectra and second-derivatives of reformed minerals in pottery.
521
28.7 FT-IR spectra of paints and pigments.
524
29.1 Energy-level diagram explaining the different types of scattering.
532
29.2 Raman spectrum of calcite (CaCO3). 533 29.3 Formation and decomposition of calcite (CaCO3). 537 29.4 Plot of the polymerization index as a function of the main Si-O stretching component wavenumber.
539
30.1 Characteristic X-ray features of the main pottery forming techniques.
548
30.2 Inclusion alignment in a clay coil.
548
xvi List of Figures 30.3 Xeroradiographs of two stirrup jars showing both the blue color and edge enhancement inherent to the technique.
551
30.4 Examples of types of radiographic equipment available. (a) Portable medical Sirio 110/100 CR system; (b) Faxitron single-cabinet X-ray unit.
552
30.5 Radiographs of a Middle Minoan amphora (BM registration number G&R1906,1112.90) from the British Museum.
554
30.6 Radiographs of a bell-shaped handled cup (Middle Minoan I) from Knossos (BM registration no G&R1950,1106.16).
557
30.7 Radiographs of a Mycenaean krater (BM registration number G&R1898,1201.112) taken from the side (a) and above (b).
560
31.1 Fan-shaped phytolith from modern rice (1 mm).
567
31.2 Thin section of rice-tempered ceramic from Sumhuram, Sultanate of Oman.
568
31.3 Shell-tempered ceramic from Sumhuram, Sultanate of Oman.
570
31.4 Fracture surface of a potsherd from Sumhuram, Sultanate of Oman.
571
31.5 Renaissance ceramic tempered with wool (polarized 1 mm).
572
31.6 Rice-tempered ceramic from Sumhuram, Sultanate of Oman: epidermis of the rice husk with evident silicized tubercles (SEM).
574
31.7 Karnataka rice winnowing in South India.
577
32.1 Map of the study area.
592
32.2 Profiles of the eleven classes of vessels determined by preliminary descriptive analysis.
593
32.3 Erosion, dilation, opening and closing characteristic curves of a profile of the database, using an isotropic (circular) structural element, normalized by the area of the profile.
595
32.4 Segmentation of a profile into rim, body, and base.
595
32.5 A given ceramic and the ten most similar shapes, with the measure of the distance.
598
33.1 Typical load-displacement curves for different types of fracture observed in archaeological ceramics: (a) unstable; (b) semi-stable; (c) stable.
614
34.1 Schematic overview of different organic residues in archaeological ceramics, and techniques frequently employed to investigate these (34.1).
626
34.2 Schematic flow diagram showing the position of organic residue analysis in anthropological and archaeological studies.
628
34.3 The general principles of “microscopic (a)” and “molecular (b)” methods for organic residue analysis.
629
34.4 Schematic of the chromatography arrays used on mass spectrometers to separate the sample into its components and convert them into a form fit to enter the ion source.
631
34.5 Schematic representation of the antibody–antigen reaction.
633
List of Figures xvii 35.1 Two graphical representations of Petrie’s chronological ordination of archaeological types.
654
35.2 An example of frequency seriation.
656
35.3 The original diagram by which Krieger explained in detail the procedure for a correct application of his typological method.
658
35.4 Some quantifiable formal dimensions in pottery.
662
35.5 An example of categorical classification of pottery surface decoration from a Central European Neolithic context.
664
36.1 Schematic of the archaeological clock and radiation doses used in luminescence dating.
673
36.2 TL glow curve on quartz grains with the various characteristic peaks and a total OSL decay curve, made of its different components. The surface under the peaks or signal is function of the number of traps occupied before the stimulation. 674 36.3 TL glow curve on quartz grains with the various characteristic peaks and a total OSL decay curve, made of its different components.
674
36.4
677
Luminescence reader.
36.5 Microbalance data obtained on a Werra earthenware specimen excavated at Enkhuizen (1979).
685
List of tables
1.1 Analytical methods included in this Handbook, and which of the four primary research questions they contribute toward answering.
5
4.1 Uncertainty values calculated for a ceramic chemical group from the Central Amazon.
54
5.1 Frequency distribution table describing a small assemblage of Greek pottery.
60
5.2 Random number table (excerpt).
67
7.1 Chemical composition of the “average terrigenous marine clay” (after Clarke, 1924) and of some representative Italian and Greek clays.
94
11.1 Chemical alteration of Roman sigillata from Arezzo, Lyon, and La Graufesenque found at Velsen and Nijmegen in the Netherlands.
169
15.1 Mineral identification table derived from the integrated information of various geological handbooks. They represent a concise list of characteristics in thin section.
242
17.1 Mean atomic number for common minerals in ceramics.
292
17.2 Selected electron microprobe analyses (in mass%) and calculated structural formulae for illite (illitic matrix), albite (Ab), plagioclase (excluding albite), alkali-feldspar, K-feldspar, and muscovite.
297
19.1 Chemical composition of calcareous illitic clay from Otterbach, Jockgrim, Palatinate, Germany.
334
19.2 List of measured diffraction angles °2θ of mullite and quartz, d values calculated according to the Bragg equation.
334
23.1 Comparison of measurements on NIST679 (brick clay) by LA-ICP-MS, MD-ICP-MS, and other bulk techniques.
411
23.2 Comparison of measurements on New Ohio Red Clay (NORC) by LA-ICP-MS, MD-ICP-MS, and other bulk techniques.
413
24.1 The sensitivity of INAA in the analysis of ceramics varies by element, from percent level to parts-per-trillion.
425
27.1 Basic atmospheric conditions during firing in antiquity.
489
29.1 Based on the Ip ratio, the processing temperature of glass structures can be estimated.
538
30.1 Exposure times and kV for clay objects using a Faxitron cabinet X-ray machine with a 0.5 mm focal spot, 60 cm focus-to-film distance, 3 mA, and Agfa Structrex D4 Film.
556
xx List of tables 32.1 Vessels in the study by context.
591
32.2 Classification rates obtained by the method with one, three, and five suggestions and one, three, and five neighbors.
597
32.3 Normalized confusion matrix resulting from the application of the method to the database, using two subprofiles: rim and the combination of body with base.
597
32.4 Normalized confusion matrix resulting from the application of the method to the database, using the rim and body subprofiles.
598
33.1 Examples of material requirements placed on different ceramic products.
606
34.1 Schematic overview of the characteristics of selected techniques to investigate organic residues in archaeological ceramics.
627
List of plates
1 Experimental firings of different clay raw materials (Sicily, Italy).
2 Photomicrographs of thin sections of medium-high fired sherds with calcite (incompletely crossed polarized light, width of field 0.7 mm).
3 Late Bronze Age pottery sample 133 in (a) hand specimen and (b and c) thin-section photomicrograph in plane polarized light.
4 Photomicrographs of thin sections with a basalt–andesitic volcanic rock fragment under plane polarized light (a) and crossed polarized light (b).
5 Photomicrographs illustrating microstructural changes with increasing ceramic firing temperature. (For explanations see text in Chapter 27.)
6 MGR-analysis of six ceramic sherds, carried out to determine original firing temperatures. Sample 1 represents a briquette cut from a model ceramic sample; samples 2–6 represents archeological samples. (For explanations see text in Chapter 27; photographs were taken with a macro lens by M. Baranowski.)
7 Structural MGR-analysis of two ceramic samples (Photographs taken under a reflected light microscope). A = pottery fragment originally fired at Teq < 700°C; B = pottery waste, fragment over-fired at 1100–1150°C. (For explanations see text in Chapter 27; photographs were taken with a macro lens by M. Baranowski.)
8 MGR-analysis (800°C, 900°C, 1200°C) of five samples belonging to two MGR-groups—a fact which only becomes apparent after re-firing at 1200°C. (For explanations see text in Chapter 27; photographs were taken with a macro lens by M. Baranowski.)
9 Examples of matrix types (samples re-fired at 1200°C) of non-calcareous sherds (iron-rich red-firing or iron-poor whitish-firing) and of calcareous pottery (yellowish-greenish firing). (For explanations see text in Chapter 27; photographs were taken with a macro lens by M. Baranowski.)
List of abbreviations
AAS Atomic Absorption Spectroscopy BSE Backscattered Electrons EDS Energy Dispersive Spectrometry EMPA Electron Microprobe Analysis ICP Inductively Coupled Plasma INAA Instrumental Neutron Activation Analysis LA Laser Ablation MGR Matrix Groups by Re-firing MS Mass Spectrometry OES Optical Emission Spectrometry OM Optical Microscopy OSL Optically Stimulated Luminescence PIXE Particle Induced X-Ray Emission ppb parts per billion (10-9) PPL Plane Polarized Light ppm parts per million (10-6) RHX Rehydroxylation SE Secondary Electron SEM Scanning Electron Microscopy SR Synchrotron Radiation TL Thermoluminescence WD Wavelength Dispersive XPL Crossed Polarized Light XRD X-Ray Diffraction XRF X-Ray Fluorescence δ stable isotope ratio expressed relative to a standard ε measure of isotopic composition relative to a mantle reservoir
List of Contributors
Daniel Albero Santacreu is Assistant Lecturer in Prehistory and Archaeology at the University of the Balearic Islands (Spain). He has developed archaeometrical and technological analysis of hand-made prehistoric pottery vessels from the Balearic Islands, Sardinia, Andalusia, and Ghana. His current research concerns the role of technology in the interpretation of ceramics and the application of concepts such as agency, habitus, technological choices, and identity in the study of ancient societies. His most recent publication is Materiality, Techniques and Society in the Pottery Production (De Gruyter Open). Janet Ambers is a scientist in the Department of Conservation and Scientific Research at the British Museum. She currently works mostly on the imaging of museum-related materials, with a specific interest in radiography, and on the analysis of museum objects using various techniques but with an emphasis on Raman spectroscopy. Her particular interests include pigment analysis with particular emphasis on the palettes of Ancient Egypt and the Middle East, the identification of gemstones, jades, and geological materials by Raman spectroscopy, and the radiography of ceramics and other similar materials. She also has a professional interest in all forms of archaeological dating and analyses of human remains. Juan Antonio Barceló is an Associate Professor of Prehistory at the Universitat Autonoma de Barcelona (Spain). He carries out specialized research in archaeological techniques and theory, developing computer applications in archaeology, notably in the domains of spatial analysis, statistics, artificial intelligence, modeling, and computer-aided visualization. He is the director of the Quantitative Archaeology Laboratory (). He has directed and participated in archaeological projects in Spain, Portugal, Italy, Syria, Nicaragua, Ecuador, and Argentina. Hans Barnard is Adjunct Assistant Professor at the Department of Near Eastern Languages and Cultures and Assistant Researcher at the Cotsen Institute of Archaeology, both at UCLA. He has worked on sites in Armenia, Chile, Egypt, Iceland, Panama, Peru, Sudan, Syria, Tunisia, and Yemen as archaeological surveyor, photographer, and ceramic analyst. Currently he is involved in research projects investigating the interaction between the Tiwanaku and Wari polities in the Vitor Valley (near Arequipa, Peru), and between the Phoenician and Roman empires in Zita (near Zarzis, Tunisia). Ina Berg is Senior Lecturer in Archaeology at the University of Manchester, UK. Her main areas of research are ceramic studies, the archaeology of Bronze Age Greece, the Cyclades in particular, and island studies. While interested in all aspects of ceramics, her current research focus is predominantly on the application of X-radiography to pottery to explore forming techniques as a means to understanding past potting traditions, workshop organization, learning networks, and knowledge transfer.
xxvi List of Contributors Gulsebnem Bishop is a full-time CSIT faculty member and a program leader for the Doctoral Program in Information Technology at Stratford University, VA. She holds a doctorate degree in Computer Science and Information Systems from Pace University, NY, where she was able to combine her two passions: computing and archaeology. Her interests are data analysis and interpretation, database development, and administration, as well as future of data analysis. She has worked at a number of not-for-profit organizations such as the American Museum of Natural History, the Human Rights Campaign, and the National Cathedral in New York City and Washington, D.C., specializing in database systems development, administration, and management. Sophie Blain was a FNRS researcher at the University of Liège, Belgium. Her researches focused on medieval building archaeology and more particularly on dating methods applied to building materials, such as dendrochronology on wood beams and luminescence (TL/ OSL) dating methods on ceramic building materials and mortar. She worked on a number of medieval churches in south-eastern England, northern France, Belgium, and northern Italy. Eugenio Bortolini is currently a member of CaSEs Research Group (Complexity and Socio-Ecological Dynamics) as Research Fellow at the Department of Archaeology and Anthropology, IMF-CSIC (Spanish National Research Council, Barcelona, Spain) and Visiting Research Fellow at the Department of Humanities, Universitat Pompeu Fabra (Barcelona, Spain). His main research interests and topics include the adoption and transmission of cultural variants, the coevolution of culture and genes (Dual Inheritance Theory), the development and application of quantitative methods in archaeology, archaeological theory, and the prehistory of the Arabian Peninsula. Matthew T. Boulanger is a Lecturer in the Department of Anthropology at Southern Methodist University and a Research Associate at the Archaeometry Laboratory at the University of Missouri Research Reactor. He has published in Antiquity, American Antiquity, Journal of Archaeological Science, and Archaeometry. His research interests include compositional analysis of archaeological materials, digital data management and preservation, evolutionary archaeology, experimental archaeology, and Paleoindian lithic technology. Dennis Braekmans is Assistant Professor at the Faculty of Archaeology, Leiden University, and Department of Materials Science, Delft University of Technology. He runs the laboratory for ceramic studies, where mineralogical and geochemical laboratory analysis is linked with production studies, mechanical research, and ethnoarchaeological observation. The current research focus is geared to ancient ceramics from North Africa, the Eastern Mediterranean, and the Near East. Teaching activities include materials science, archaeometry, and experimental archaeology. Jaume Buxeda i Garrigós is Lecturer in Archaeology and Director of the Cultura Material i Arqueometria UB (ARQUB) research unit at the Universitat de Barcelona. His recent work is mainly related to research projects in historical archaeology (ARCHSYMB and TECNOLONIAL) designed to deepen our knowledge in aspects related to the interaction, influence, and cultural change during the colonization process, through the archaeological and archaeometric study of technical/technological impact, and issues of technical/technological traditions and change. Other research interests are classical archaeology, the effect of weathering, and the role and treatment of compositional data in archaeological research.
List of Contributors xxvii Manuel Calvo Trias is Lecturer in Prehistory and Archaeology at the University of the Balearic Islands (Spain). He is head of the ArqueoUIB Research Group and PI of a research project focused on the Balearic Islands Prehistory, as well as the project Archaeology in the Upper White Volta basin (north-east of Ghana). His research is centered on the analysis of material culture and technology. He is co-author of “Acción técnica, interacción social y práctica cotidiana: propuesta interpretativa de la tecnología” (Trabajos de Prehistoria 71) and “Ceramic Transactions in a Multi-Ethnic Area (Upper East Ghana)” (Applied Clay Science 82). Małgorzata Daszkiewicz works in collaboration with Warsaw University of Technology and, since 1994, with Arbeitsgruppe Archaeometrie (FU Berlin). In 1998 she set up her own company—ARCHEA—as a laboratory for archaeometric analysis and research, participating in international projects with many institutions. In 2009 and 2012 she was Senior Fellow at the Cluster of Excellence TOPOI at FU Berlin, where she is now a part-time employee. Her main research interests are in determining the technology and provenance of archaeological ceramics (using MGR-analysis, thin sections, WD-XRF and pXRF) as well as devising techniques for the classification of bulk ceramic finds and the development of methods for forming techniques and functional properties (she has created a joint databank with Gerwulf Schneider of c.30,000 WD-XRF analyses). Patrick Degryse is Professor of Archaeometry at the Department of Earth and Environmental Sciences and Director of the Center for Archaeological Sciences at the Katholieke Universiteit Leuven (Belgium). His main research efforts focus on the use of mineral raw materials in ancient ceramic, glass, metal, and building stone production, using petrographical, mineralogical, and isotope geochemical techniques. He teaches geology, geochemistry, archaeometry, and natural sciences in archaeology, is an A. von Humboldt Fellow and European Research Council Grantee, and is active in several field projects in the eastern Mediterranean. Kim Duistermaat is guest researcher at the Faculty of Archaeology, Leiden University. Her interests include the organization of pottery production, and the study of seals and sealings. Kim has worked and lived in Syria and Egypt and is currently resident in China. Laure Dussubieux is a Laboratory Scientist at The Field Museum’s Integrative Research Center. Jelmer W. Eerkens is a Professor of Anthropology at the University of California, Davis, where he runs the Archaeometry Laboratory. His research focuses on small-scale societies, the evolution of material technologies, and the reconstruction of ancient subsistence and settlement systems. Towards this end, he has applied a number of archaeometric techniques, including gas chromatography-mass spectrometry, instrumental neutron activation analysis, stable isotope analysis, and luminescence dating to ceramic assemblages from Western North America and Peru. Kent D. Fowler is an Associate Professor of Anthropology at the University of Manitoba in Winnipeg, Canada. Since 1996, his research interests have centered on the ceramic traditions of southeastern Africa from archaeological and ethnographic perspectives. His archaeological research has focused upon ceramic manufacture and use in first millennium ad farming societies of South Africa. He has conducted long-term ethnographic research in Zulu and
xxviii List of Contributors Swazi potting communities, examining how technical and social factors influence variation in manufacture. His recent research interest is in how petrographic and chemical analyses of ceramics are affected by potters’ behaviors and the use of ceramics. Representative publications include “Zulu pottery production in the Lower Thukela Basin, KwaZulu-Natal, South Africa” (2008); “Clay acquisition and processing strategies during the first millennium AD in the Thukela River basin, South Africa: An ethnoarchaeological approach” (2011); “Ceramic production in Swaziland” (2014); “Zulu ceramic production in the Phongolo River Basin, South Africa” (2015). Jaime García Rosselló is Lecturer in Prehistory and Archaeology at the University of the Balearic Islands (Spain). His ethnoarchaeological research focuses on the analysis of pottery forming methods in several potter communities from Chile, Ecuador, Egypt, Tunisia, Morocco, and Ghana. He is co-author of Making Pots: El modelado de la cerámica y su potencial interpretativo (BAR International Series 2540, Archaeopress) with Manuel Calvo. Mark Golitko is a visiting Assistant Professor in the Department of Anthropology at the University of Notre Dame, and a Research Associate at the Field Museum of Natural History in Chicago. He utilizes chemical methods to explore how human social networks evolve in response to environmental and social forces. His field and laboratory research has spanned several regions, including Papua New Guinea, Europe, and the Americas. His publications include “Mapping prehistoric social fields on the Sepik coast of Papua New Guinea: ceramic compositional analysis using laser ablation-inductively coupled plasma-mass spectrometry” (2012, Journal of Archaeological Science) and the book LBK Realpolitik: An Archaeometric Study of Conflict and Social Structure in the Belgian Early Neolithic (2015, Archaeopress). Alan F. Greene is a Postdoctoral Scholar in the Department of Anthropology and Stanford Archaeology Center at Stanford University. His field research based in Armenia focuses on the habitual economic transactions, production chains, and value transformations of Bronze Age and Iron Age communities in the South Caucasus (c.3500–800 bc). He is also co-director of the Making of Ancient Eurasia (MAE) Project, a research consortium involved in the radiographic, tomographic, and synchrotron radiation-based analysis of ceramic materials from across the Eurasian landmass. He is co-editor, with Charles Hartley (University of Chicago), of a forthcoming volume on the structural analysis of archaeological pottery. Christopher Hall is Professor Emeritus and Professorial Fellow in the School of Engineering, University of Edinburgh. He is a Fellow of the Royal Society of Edinburgh, and a Fellow of the Royal Academy of Engineering. His research interests are in the chemistry and physics of materials used in construction engineering. Recent work includes synchrotron-based diffraction methods, scanning-probe microscopy of mineral/water reactions, and engineering analysis and chemistry in archaeology and building conservation. Publications include Polymer Materials (1981, 2nd edn 1989); Water Transport in Brick, Stone and Concrete (with W. D. Hoff, 2002, 2nd edn 2012); Materials: A Very Short Introduction (2014); and numerous scientific papers in professional journals. Mark E. Hall is Assistant Field Manager of the Black Rock Field Office in the Winnemucca District of the Bureau of Land Management. His research and field work has spanned several geographic areas, from California and the Great Basin, Ireland, Japan, Mongolia, and
List of Contributors xxix the Russian steppes. He is the author of more than fifty research papers, many focusing on north-east Asian pottery production. Roberto Hazenfratz-Marks is a collaborator of the Archaeometry Group at the Nuclear and Energy Research Institute in São Paulo, directed by Dr. Casimiro S. Munita. He concentrates on the analysis of archaeological ceramics and geological material from Central Amazon. He has experience with instrumental neutron activation analysis, X-ray diffraction, optically stimulated luminescence dating, electron paramagnetic resonance, and multivariate data analysis. He is also Professor of Engineering. Robert B. Heimann is Professor Emeritus of Applied Mineralogy and Materials Science at Technische Universität Bergakademie Freiberg, Germany. From 1979 onward he worked in Canada as a research associate (McMaster University), visiting professor (University of Toronto), staff geochemist (Atomic Energy of Canada Limited), and research manager (Alberta Research Council). From 1993 to 2004 he was a full professor at TU Bergakademie Freiberg in Germany. He has authored and coauthored more than 300 scientific publications including several books, and in 2001 was awarded the Georg-Agricola Medal of the German Mineralogical Society (DMG). Research activities and interests are in advanced ceramics, single crystal growth, and thermal spraying of environmental and biomedical coatings, as well as archaeometallurgy and archaeoceramics. Volker Hoeck is a retired professor. He was Professor of Geology at Paris Lodron University of Salzburg, Austria, and Associate Professor at Babeş-Bolyai University of Cluj-Napoca, Romania. His research interests include geochemistry and petrography of magmatic and metamorphic rocks from the Alps, Carpathians, Dinarides, and Central Asia, as well as petrography and geochemistry of archaeoceramics. His recent publications are “Insights into the EPR Characteristics of Heated Carbonate-rich Illitic Clay” (2014, Applied Clay Science); “Burnishing versus Smoothing in Ceramic Surface Finishing: A SEM Study” (2015, Archaeometry), and “Geochemistry of Neogene Quartz Andesites from the Oaș and the Gutâi Mountains, Eastern Carpathians (Romania): A Complex Magma Genesis” (2014, Mineralogy and Petrology). Elisabeth Holmqvist works as an Academy of Finland Postdoctoral Fellow at the Department of Philosophy, History, Culture and Art Studies, University of Helsinki. Her research interests deal with ancient technologies and exchange systems, and geochemical provenancing of archaeological materials by ED-XRF, SEM-EDS, PIXE, ICP-MS, and INAA. Holmqvist has worked on archaeological ceramics from the Near East, Europe, and Latin America. In her PhD (UCL Institute of Archaeology, 2010), Holmqvist concentrated on Byzantine–Islamic pottery manufacture and trade in the Near East, while her current research focuses on Scandinavian and Baltic potting traditions and inter-regional ceramic exchange from the Neolithic into medieval times. Alice M. W. Hunt is a Assistant Research Scientist at the Center for Applied Isotope Studies, University of Georgia. Her PhD in archaeological materials analysis developed cathodoluminescence spectrometry of quartz as a method for differentiating raw material sources in fine-grained ceramics. Currently, her research focuses on developing analytical calibrations and protocols for bulk chemical characterization of cultural materials (ceramics, anthropogenic sediments, copper alloys, and obsidian) by portable XRF. Recent publications include
xxx List of Contributors “Portable XRF analysis of archaeological sediments and ceramics” (Journal of Archaeological Science, 2015) and a monograph Palace Ware across the Neo-Assyrian Imperial Landscape: Social Value and Semiotic Meaning (E. J. Brill). Corina Ionescu is Professor at Geology Department at Babeş-Bolyai University of Cluj- Napoca, Romania, and Associate Professor at Kazan (Volga Region) Federal University, Tartarstan, Russia. Her research interests focus on archaeoceramics and ophiolite petrology and geochemistry. Her recent publications include “Burnishing versus Smoothing in Ceramic Surface Finishing: A SEM Study” (2015, Archaeometry); “Insights into the Raw Materials and Technology Used to Produce Copper Age Ceramics in the Southern Carpathians (Romania)” (2016, Archaeological and Anthropological Sciences). Marisol Madrid i Fernández is a Researcher at the Cultura Material i Arqueometria UB (ARQUB) research unit at the Universitat de Barcelona. Her work focuses on the application of analytical techniques to the study of archaeological materials, especially ceramics. She has participated in more than forty research projects highlighting the current TECNOLONIAL project on historical archaeology and archaeometry. She is author/co-author of more than fifty publications and has been co-organizer of two scientific international conferences. She has had longlasting activity in the field of classical archaeology in several excavations, especially in the research project at the Roman town of Cosa, Italy, leading the study of Roman pottery and its archaeometric characterization. Marta Mariotti Lippi focuses on archaeobotany, palynology, and reproductive biology. She has carried out archaeobotanical investigations at sites in Italy (Pompeii and Vesuvian area, Paestum, Tuscany) and abroad (Russia, Czech Republic, Jordan, Lybia). Since 2001 she has cooperated in research projects in the “Land of Frankincense” UNESCO site of Sumhuram/ Khor Rori, and at Salut, Sultanate of Oman. Lara Maritan was, between 1999 and 2005, a visiting scholar at the University of Glasgow (UK) and the University of Cardiff (UK), funded by scholarships from the Gini Foundation, the Italian Society of Mineralogy and Petrology (SIMP), and the Accademia Nazionale dei Lincei/Royal Society. From 2003 she was a postdoctoral research assistant at the University of Padova before joining the faculty in the Department of Geosciences as an assistant professor in georesources and minero-petrographic applications for the environment and cultural heritage (GEO/09) in 2007. Ana Luisa Martínez-Carillo is a researcher at the Research Institute of Iberian Archaeology, University of Jaén. She is a specialist in the application of new technologies in archaeological analysis, in particular in ceramic studies, and has also participated in several regional, national, and European research projects focused on three-dimensional modeling, integration of datasets, and on-line dissemination of the cultural heritage. Leah D. Minc is an Associate Professor in the College of Liberal Arts at Oregon State University, and INAA Research Coordinator at the Oregon State University Archaeometry Laboratory. Giuseppe Montana has been Associate Professor at the University of Palermo since 2005. His research activity covers topics in the field of mineralogy and petrography applied to cultural heritage (archaeological ceramics, natural and artificial stones). Significant recent
List of Contributors xxxi publications include “Characterization of Clayey Raw Materials for Ceramic Manufacture in Ancient Sicily,” Applied Clay Science, 53 (2011): 476–488; “An Original Experimental Approach to Study the Alteration and/or Contamination of Archaeological Ceramics Originated by Seawater Burial,” Periodico di Mineralogia, 83 (2014): 89–120; “Different Methods for Soluble Salt Removal Tested on Late-Roman Cooking Ware from a Submarine Excavation at the Island of Pantelleria (Sicily, Italy),” Journal of Cultural Heritage, 15(2014): 403–413. Noémi S. Müller is Scientific Research Officer at the Fitch Laboratory of the British School at Athens. She has held postdoctoral positions at NCSR Demokritos, where her research focused on mechanical and thermal properties of archaeological ceramics, as well as in Nicosia, Cyprus, and Barcelona, Spain. She is interested in applying analytical methods to investigate inorganic archaeological artifacts and materials, focusing on the study of provenance and technology and with a special interest in archaeological ceramics. Her research also examines the affordance of utilitarian ceramics, focusing on cooking vessels, using material testing to explore the influence of technological choices in manufacture on material properties. Pasquino Pallecchi is Adjunct Professor at the University of Florence and Head of the Laboratory of Archaeological Heritage in Tuscany, Italy. Patrick S. Quinn is an archaeological materials scientist working on ceramics and related artifacts from a range of different periods and regions including prehistoric and later Britain, pre-contact California, and the prehistoric Aegean. His early research focused on the occurrence of microfossils within ancient ceramic pastes and their research potential in terms of pottery provenance and technology on Minoan Crete. He subsequently worked on the palaeontology and ecology of microfossil-producing organisms before rejoining the University of Sheffield, then the Institute of Archaeology, University College London, as a permanent member of the research staff, undertaking research, teaching, and consultancy on archaeological ceramic analysis. Marcia A. Rizzutto has been a Professor at the Physics Institute of the University of São Paulo since 2001. She is also the coordinator of the Research Center of Applied Physics to the Study of Artistic and Historical Cultural Heritage of the University of São Paulo. Since 2003 she has been devoted to the use of applied physics to the study of cultural heritage objects connected with different areas such as archaeology, history, art history, paleontology, chemistry, conservation, and restoration. As coordinator of the research group she has a wider investigation program using non-destructive physics analyses in different museum collections of the University of São Paulo, in partnership with teachers/researchers from the museum’s institutions. Valentine Roux is Director of Research at the CNRS, Nanterre, France. Her work combines methodological research on technical skills and an anthropological approach to ceramic assemblages, archaeological research on the history of technology and people in the Southern Levant, and ethnoarchaeological research in India on specialization and diffusion of potting techniques. Selected recent publications include a handbook, in collaboration with M.-A. Courty, entitled Des céramiques et des hommes: Décoder les assemblages archéologiques (2016); an edited issue of Journal of Archaeological Method and Theory (2013,
xxxii List of Contributors 20/2), with Courty, entitled “Discontinuities and Continuities: Theories, Methods and Proxies for an Historical and Sociological Approach to Evolution of Past Societies”; and an edited issue of Paléorient (2013, 39/1) with Braun, entitled “The Transition from Late Chalcolithic to Early Bronze in the Southern Levant: Continuity and/or Discontinuity?”. Gerwulf Schneider has since 1975 been involved in archaeometric research and the teaching of archaeometry, from 1980 to 2003 at the Institute of Inorganic and Analytical Chemistry (Arbeitsgruppe Archaeometrie), and currently as a research associate at the Cluster of Excellence Topoi at the Free University of Berlin. His main focus is on chemical and mineralogical analysis of archaeological ceramics and the interpretation of data in archaeological terms. His research interests encompass Roman pottery in Germany, Hellenistic to Late Antique pottery in the Mediterranean region, and various projects on Neolithic to medieval pottery in Europe, Mesopotamia, and Sudan (he has created a joint databank of c.30,000 WD-XRF pottery analyses with M. Daszkiewicz). Prabodh Shirvalkar is currently working as an associate professor in the department of Archaeology, Deccan College Post Graduate and Research Institute, Pune, India. His research interest focuses on the Harappan Civilization, particularly in regards to ceramic technology, provenance, and trade networks. At present, he is working on creating a model for the rural economy of Harappans. In addition, he has expertise in field archaeology and the application of processualism. Shlomo Shoval is Professor of Earth Sciences at the Open University of Israel. He is also Guest Professor at the Institute of Earth Sciences of the Hebrew University of Jerusalem and Visiting Scientist at the Institut Lumière Matière of Université Claude Bernard Lyon-1, France. He is an expert in the analysis of archaeological ceramics, ceramic raw materials, and clay minerals by infrared spectroscopy (FT-IR) and other scientific methods (LA-ICP-MS, XRF). Among his current research projects are studies of the technologies used in manufacture and in paint decoration of Levantine Bronze and Iron Age ceramics. Johannes H. Sterba has been working in neutron activation analysis for more than a decade, starting with geological samples and soon moving to archaeologically relevant materials such as pumice, obsidian, and later ceramics. He is interested in the statistical analysis of the gathered data and in the intercomparability of different analytical methods. Recent publications include “NAA and XRF analyses and magnetic susceptibility measurement of Mesopotamian cuneiform tablets” (Scienze dell’Antiquita, 2011); “Raising the temper—µ- spot analysis of temper inclusions in experimental ceramics” (Journal of Radioanalytical and Nuclear Chemistry, 2011); and “Volcanic glass under fire—a comparison of three complementary analytical methods” (X-Ray Spectrometry, 2013). Manfredo H. Tabacniks is a full professor at the Physics Institute of the University of São Paulo. In 1994 he served as a postdoctoral researcher for two years at the IBM Almaden Research Center, San Jose, California, on the application of ion beam methods (PIXE and RBS) for the analysis of thin films. Since 1996 he has been head of the Ion Beam Analysis facility of the Institute of Physics at USP. His main research interests deal with ion beam methods for advanced material analysis and for the modification of materials. Jeremy Taylor took up a Leverhulme Research Fellowship at the School of Archaeology, University of Leicester, in 1999, before being appointed a Lecturer in Archaeology in 2001.
List of Contributors xxxiii He is currently a director of the major field project at Burrough Hill, Leicestershire. His research interests center on social change in Iron Age Britain and the Western Roman provinces through study of their rural landscapes, and on interrelationships between theory and method in survey-based archaeological research, such as geophysics, geochemistry, and aerial survey. Michael S. Tite was, before retiring, the Edward Hall Professor of Archaeological Science and Director of the Research Laboratory for Archaeology and the History of Art at the University of Oxford, where he is now Professor Emeritus and Fellow of Linacre College. Formerly he served as Keeper of the Research Laboratory at the British Museum, and has been a Fellow of the Society of Antiquaries since 1977. The underlining theme of his research during the past thirty years has been the study of the technology involved in the production of (1) faience and related early vitreous materials from Egypt and the Near East, and (2) glazed pottery from the Bronze Age through the Roman period in the Near East, Europe, the Islamic world, and China. Jolien Van Pevenage is currently a PhD candidate in the Raman Spectroscopy Research Group. She is interested in the application of Raman spectroscopy to the study of ceramic artifacts for their identification and classification in order to define, for example, the origin or the composition of the materials and production techniques. Her research combines the use of Raman spectroscopy with X-ray fluorescence spectroscopy and advanced data processing methods. Her work is presented at international conferences and is published in leading scientific journals. Peter Vandenabeele was appointed as research professor in the Department of Archaeology at Ghent University, where he applies his analytical skills to the study of archaeological and artistic objects. His research focuses mainly on the application of Raman spectroscopy in art analysis. He has authored more than a hundred research papers on Raman spectroscopy and its application in archaeometry, along with several book chapters and conference presentations. Recently, he has published, together with Howell Edwards, Selected Topics in Analytical Archaeometry (RSC Publishing, 2012). Yona Waksman is a Senior Researcher at the French National Research Center (CNRS) in Lyon. She specializes in archaeometric approaches to medieval ceramics in the Eastern Mediterranean and the Black Sea, through provenance and technological studies. Her publications include major sites such as Constantinople/Istanbul, and lay new foundations for the investigation of economic, cultural, and social phenomena in the Byzantine world and the medieval Middle East. Ian K. Whitbread read Archaeology and Geology at the University of Bristol before undertaking a research fellowship at the Fitch Laboratory, British School at Athens, Greece. Formerly he was a Principal Research Scientist at the Center for Materials Research in Archaeology and Ethnology, Massachusetts Institute of Technology, USA, and then Director of the Fitch Laboratory. He joined the School of Archaeology and Ancient History, University of Leicester, in 2001. His current research interests lie in the analysis of ancient ceramic materials with respect to issues of trade/exchange and the socially embedded nature of technology. Bettina A. Wiegand has long-term experience in isotope geochemistry at the University of Goettingen, Germany, and Stanford University, USA. Application of isotope methods
xxxiv List of Contributors to various research fields include ceramic provenance studies, human migration studies, and hydrogeological and environmental research. Related publications include “Strontium Isotopic Evidence for Prehistoric Transport of Gray-Ware Ceramic Materials in the Eastern Grand Canyon Region, USA,” Geoarchaeology 26 (2011): 189–218; and “Reconstructing Middle Horizon Mobility Patterns on the Coast of Peru through Strontium Isotope Analysis,” (Journal of Archaeological Science 26 (2009): 157–165). Ian P. Wilkinson was a principal scientist with the British Geological Survey before retiring, and is now an Honorary Research Associate at both the BGS and the University of Leicester. He is a specialist in Mesozoic and Cenozoic-Recent ostracods and foraminifera, working on mapping, hydrocarbons, geohazards, and palaeoenvironmental and archaeological projects. Although focusing principally on the UK, he also has experience in, for example, Ecuador, USA, Antarctica, Hong Kong, the Persian Gulf, Papua New Guinea, Armenia, and Russia. Mark Williams is Professor of Geology at the University of Leicester, where he researches climate and environmental change reflected in the fossil record. He has worked at the Universities of Frankfurt, Lyon, and Portsmouth, and for the British Geological Survey and the British Antarctic Survey. He has a strong interest in the geology of the Arabian Peninsula and has consulted for Saudi Aramco.
Plates EXPERIMENTAL CLAY TEST-PIECES Leather-hard state (moisture content 15–20 weight)
Color by Munsell code
After firing 900 °C (oxidizing atmosphere)
Color by Munsell code
2.5Y 6/4
10R 5/6
2.5Y 6/3
2.5YR 5/8
5Y 5/2
10R 5/8
5Y 6/4
10R 6/8
2.5Y 5/3
10R 5/8
2.5Y 5/2
10R 5/8
Plate 1 Experimental firings of different clay raw materials (Sicily, Italy).
(a)
(b)
(c)
(d)
Plate 2 Photomicrograph of thin sections of medium-high fired sherds with calcite: (a) homogeneously distributed secondary calcite from recarbonatization after gehlenite in calcareous pottery (sigillata from Arezzo); (b) secondary calcite in calcareous pottery (Eastern Sigilata A) is dissolved at the surface where environmental acid solutions had access, as, for example, on the upper right site where the protective gloss is missing; (c) precipitated calcite in non-calcareous North Mesopotamian metallic ware; (d) calcite scale below the red gloss of a sherd of Tripolitanian sigillata from Carthage. (Photomicrographs by G. Schneider; incompletely crossed polarizers, width of field 0.7 mm.) (a)
(b)
(c)
Plate 3 (a) Late Bronze Age pottery sample 133, hand specimen of the fabric in fresh break, width of field is 15 mm (photograph I. K. Whitbread); (b) Late Bronze Age pottery sample 133, photomicrograph of the fabric in plane polarized light, width of field is 4.65 mm (photograph I. K. Whitbread); (c) Late Bronze Age pottery sample 133, photomicrograph of the fabric under crossed polars, width of field is 4.65 mm (photograph I. K. Whitbread).
Plate 4 Photomicrographs of thin sections with a basalt–andesitic volcanic rock fragment under plane polarized light (a) and crossed polarized light (b); (c) pale green pleochroic colors of a pyroxene mineral in plane polarized light (field is 1.8mm); (d) moderate (low second order) interference colors of an amphibole (field is 1.2mm); (e) secondary calcite formation in pore space; (f) test-fired ceramic with grog under crossed polarizers. A clear demarcation can be detected between grog and the matrix (field is 1.8mm).
Plate 5 Some microstructural changes with increasing firing temperature involving inclusions, pores, and groundmass. Photomicrographs in polarized transmitted light, parallel polars of: a-d) mollusk shells with still preserved (a) and completely obliterated (b) internal structure, pores with irregular (vughs, c) and spherical (vesicles, d) shape, and optically active (a, c) and inactive (b, d) groundmass, as a consequence of the firing temperature of the same base clay material; e–f) micritic limestone grains with still-preserved crystalline optical behavior (a) and completely obliterated (b) from the firing. Scanning electron microscope back scattered images of an illitic–chloritic clay in which the groundmass is still composed of well-defined grains at 1050°C (g), whereas at 1100°C it is partially melted, forming an amorphous phase, which determines bridging between grains.
Original sample
Sample after refiring 400 °C
600 °C
700 °C
800 °C
900 °C
1000 °C
1100 °C
1
2
3
4
5
6
Plate 6 MGR-analysis of six ceramic sherds, carried out to determine original firing temperatures. (For explanations see text in Chapter 27.)
(a)
(b)
Plate 7 Structural MGR-analysis of two ceramic samples. A = pottery fragment originally fired at Teq < 700°C; B = pottery waste, fragment over-fired at 1100–1150°C. (Photographs taken under a reflected light microscope.)
Original sample
Sample after refiring 800 °C
900 °C
1200 °C
1
2
3
4
5
Plate 8 MGR-analysis (800°C, 900°C,1200°C) of five samples belonging to two MGR- groups—a fact which only becomes apparent after re-firing at 1200°C.
8
SN
ovF
ovM
sMLT
MLT
FL
sovM (temper MLT)
Plate 9 Examples of matrix types (samples re-fired at 1200°C) of non-calcareous sherds (iron-rich red-firing or iron-poor whitish-firing) and of calcareous pottery (yellowish- greenish firing). (Photographs by M. Baranowski.)
Pa rt I
I N T RODU C T ION
Chapter 1
Introdu ct i on to t he Oxford Ha ndb o ok of Archaeol o g i c a l Ceramic Ana lysi s Alice M. W. Hunt Ceramic is one of the most complex and ubiquitous archaeomaterials, occurring around the world at prehistoric through industrial sites and used to fashion everything from residences and technological installations to utilitarian wares and decorative/votive figurines. It is not simply the range of cultures and functions that ceramics serve but the diversity in materials and manufacture technology that makes archaeological ceramic analysis as challenging as it is essential. In this volume we address the sociocultural, geochemical, and mineralogical complexity inherent in archaeological ceramic analysis and provide insight into the uncertainties by providing concrete guidelines for designing rigorous research strategies and developing sophisticated and answerable anthropological research questions. Part II is dedicated to issues related to designing ceramic research and evaluating the varied types of data this research generates. Jaume Buxeda i Garrigós and Marisol Madrid i Fernández (Chapter 3) outline the two essential types of research for archaeological ceramic analysis, advancing the discipline and answering archaeological questions, and provide tools and guidelines for approaching each. In Chapter 4, Roberto Hazenfratz Marks discusses how to identify and report the uncertainty inherent in ceramic analysis with particular emphasis on interpreting geochemical data. In Chapter 5, Gulsebnem Bishop discusses the types of data generated in archaeological ceramic analysis, the strengths and weaknesses of each, and the appropriate model/statistical tools for describing and analyzing this data. Many of the most interesting and important anthropological questions involve comparing ceramic datasets, often from different excavations or laboratories and separated by decades. In Chapter 6, Matthew Boulanger offers insight into working with these datasets and how to preserve data for future analysis. Part III, Foundational Concepts, provides a detailed discussion of, and recommends best practices for, the definition, description, and illustration of archaeological ceramics. In Chapter 7, Giuseppe Montana defines ceramic raw materials and describes their
4 ALICE M. W. Hunt compositional and physical properties, such as plasticity and swelling, and how these properties are manipulated by human behavior. Chapters 8 and 9 both investigate the social and economic organization of ceramic manufacture: Valentine Roux (Chapter 8) evaluates ceramic manufacture using the chaîne opératoire approach, while Kim Duistermaat (Chapter 9) offers a relational approach to ceramic manufacture based on network analysis models. Both approaches discuss how to locate the technical behaviors associated with ceramic manufacture within the socioeconomic and cultural constraints in operation. Yona Waksman tackles ceramic provenance in Chapter 10, providing unambiguous definitions and practical guidelines for forming production and/or compositional groups within a ceramic assemblage. In Chapter 11, Gerwulf Schneider details the geochemical and mineralogical uncertainty created by alteration of archaeological ceramics during their use-life and as a result of post-depositional processes, and discusses methods for identifying these changes during analysis. The last three chapters in Part III provide practical skills, definitions, and guidelines for the description and illustration of archaeological ceramic artifacts. Daniel Albero Santacreu, Manuel Calvo Trias, and Jaime Garciá Rosselló (Chapter 12) discuss formal classification and analysis of ceramics, offering universal definitions and insight for the interpretation of formal data, while remaining sensitive to the practical and cultural factors influencing vessel shape and size. Ian Whitbread (Chapter 13) similarly establishes guidelines for best practice in describing ceramic fabrics both in hand specimen (macroscopic analysis) and thin section (microscopic analysis), along with methodologies for preparation of samples and reporting/ publishing fabric descriptions. In Chapter 14, Prabodh Shirvalkar provides a primer for constructing accurate and informative analytical illustration of archaeological ceramics. The remainder of the volume is dedicated to the analytical techniques used in archaeological ceramic analysis and is organized broadly by anthropological questions. There are many ways in which these techniques could be categorized (see Chapter 35 on Typology and Classification), especially since each technique or method provides data relevant to more than one line of inquiry and most anthropological questions require more than one type of data (see Table 1.1). However, for reasons both practical and functional we present each analytical technique according to the fundamental anthropological question to which it most often or significantly contributes. These four fundamental research areas are Provenance (Part IV), Manufacture (Part V), Function (Part VI), and Date (Part VII). Each technique-or method-specific chapter includes its scientific and/or theoretical background, discussion of practical issues, such as cost and sample preparation, and case studies emphasizing the utility of the technique in addressing anthropological questions of provenance. When possible, guidelines for best practice of collecting and interpreting the data are provided. Ceramic provenance is typically evaluated using compositional data. In Part IV, bulk chemical, phase, and mineralogical analysis, as well as micropalaeontological analysis, are discussed. Chemical methods include chapters on isotope analysis by Bettina Wiegand (Chapter 18), X-ray fluorescence by Robert Heimann (Chapter 19) and Mark Hall (Chapter 20), handheld portable energy-dispersive X-ray fluorescence spectrometry by Elisabeth Holmqvist (Chapter 21), particle induced X-ray emission spectrometry by Marcia Rizzutto and Manfredo Tabacniks (Chapter 22), inductively coupled plasma-mass spectrometry by Mark Golitko and Laure Dussubieux (Chapter 23), neutron activation analysis by Leah Minc and Johannes Sterba (Chapter 24), and synchrotron radiation by Alan Greene
Introduction 5 Table 1.1 A nalytical methods included in this Handbook, and which of the four primary research questions they contribute toward answering Provenance
Manufacture Technology
Function
EPMA
X
X
X
Ethnography
X
X
X
Experimental Firing/Re-firing
X
FT-IR
X
ICP-MS/LA-ICP-MS
X
INAA
X
Isotope Analysis
X
Mechanical Properties
X
X
Micropalaeontology
X
Morphometric Analysis
X
X
X
Organic Inclusions
X
X
X
X
X X
Organic Residue
X
Petrography
X
X
PIXE
X
X
pXRF
X
Raman Spectroscopy
Date
X
RHX Dating
X
TL Dating
X
Typology
X
X
X-Ray Radiography
X
X
XRD
X
X
XRF-EDS/WDS
X
X
X
(Chapter 25). Mineralogical and phase analysis methods include chapters on petrography by Dennis Braekmans and Patrick Degryse (Chapter 15), electron probe microanalysis by Corina Ionescu and Volker Hoeck (Chapter 17), and X-ray diffraction by Robert Heimann (Chapter 19). Micropalaeontology (Chapter 16) is written by Ian Wilkinson, Patrick Quinn, Mark Williams, Jeremy Taylor, and Ian Whitbread. In Part V, methods and techniques on ceramic manufacture include chapters on ethnography (Chapter 26) by Kent Fowler, experimental firing and re-firing (Chapter 27) by
6 ALICE M. W. Hunt Malgorzata Daszkiewicz and Lara Maritan, X-ray radiography (Chapter 30) by Ian Berg and Janet Ambers, and organic inclusions by Marta Mariotti Lippi and Pasquino Pallecchi (Chapter 31). Chapters on Fourier transform infrared spectroscopy (Chapter 28) by Shlomo Shoval, and Raman spectroscopy (Chapter 29) by Jolien Van Pevenage and Peter Vandenabeele, are also included in Part V because they can provide valuable information about mineralogical alteration in ceramic fabrics and surface treatments. Vessel function can be assessed from the formal attributes and performance characteristics of a vessel or artifact, as well as any organic residues preserved on their surfaces. Therefore, Part VI includes discussion of morphometrics (Chapter 32) by Ana Martinez- Carillo and Juan Antonio Barcelo, mechanical properties (Chapter 33) by Noémi Suzanne Müller, and residue analysis (Chapter 34) by Hans Barnard and Jelmer W. Eerkens. Part VII investigates the direct and indirect or relative dating of ceramics, and includes chapters on typology and classification (Chapter 35) by Eugenio Bortolini, and direct dating techniques of luminescence and rehydroxylation dating in a chapter by Sophie Blain and Christopher Hall (Chapter 36).
Chapter 2
History of Sc i e nt i fi c Researc h Michael S. Tite Aims The primary aim of scientific research into archaeological ceramics is the investigation of the overall life-cycle of surviving ceramics starting with their production and continuing through their distribution to their use, reuse, and ultimate discard (Tite, 1999, 2008). The first step involves reconstruction of the production, distribution, and use of ceramics. The second step then involves interpretation of this reconstructed life-cycle in order to obtain a better understanding of the behavior of the people who produced, distributed, and used these ceramics. Reconstruction of the production technology of archaeological ceramics involves the investigation of raw materials, tools, energy sources, and techniques used in the procurement and preparation of the clay, forming of the pot, and its surface treatment, decoration, and firing. Such information can be inferred from the observed macrostructure, microstructure, and chemical and mineral/phase compositions of the ceramics. The reconstruction of distribution (i.e. provenance studies) involves trying to establish, on the basis of thin-section petrography and/or chemical composition, whether ceramics were locally produced or imported, and if the latter, to identify the production center and/or source of the raw material. The determination of the use to which ceramic vessels were put involves their examination for surface wear and the presence of soot deposits, the analysis of surviving organic residues, and the investigation of their physical properties. In the interpretation of the reconstructed production technology of archaeological ceramics, the primary questions that need to be considered are why ceramics were first adopted for use in different parts of the world, and why, when adopted, a particular production technology was chosen. The interpretation of the reconstructed distribution of ceramics is concerned both with trying to determine patterns of trade or exchange away from any identified production center or source of raw material, and the underlying sociocultural reasons for that pattern.
8 MICHAEL S. Tite
Overview of History Leaving aside isolated examples of scientific research into ancient ceramics in the nineteenth and first half of the twentieth centuries, a coherent and continuing program of such research really only began in the 1950s. This occurred as part of the emergence of the overall field of archaeological science which, at that time, included the application of geophysical prospection and scientific dating methods to archaeology as well as the scientific study of the full range of archaeological artifacts (Aitken, 1961). It is generally considered that the starting point for present-day scientific research into archaeological ceramics was the groundbreaking volume by Anna Shepard (1956) entitled Ceramics for the Archaeologist. However, two further crucial, and more or less contemporary, developments were the founding in 1955 of the Research Laboratory for Archaeology and the History of Art at the University of Oxford with E. T. (Teddy) Hall as its first Director, and the beginning of instrumental neutron activation analysis (INAA) of ceramics at the Brookhaven National Laboratory (Long Island, New York) under the direction of (Ed) Sayre (Sayre and Dodson, 1957). The Oxford Research Laboratory, which was founded through the combined efforts of a physicist (Lord Cherwell) and an archaeologist (Professor Christopher Hawkes), went on to become a focus for archaeological science, or archaeometry as it was termed in Oxford, both in the United Kingdom (UK) and internationally. Thus, the publication Archaeometry, which was started in 1958 as the Bulletin of the Research Laboratory for Archaeology and the History of Art, went on to become an international research journal. Similarly, the training course in 1962, and subsequent annual reunions, organized for archaeologists who had purchased proton gradiometers from the Laboratory, went on to become the International Symposium for Archaeometry, initially held annually, but now held biennially, of which the 40th Symposium was held in Los Angeles in 2014. The continuing development of archaeological science, including scientific research into archaeological ceramics, has depended, first, on the provision of funding for research specifically into aspects of this new discipline, and second, on the inclusion of the teaching of archaeological science in university archaeology degree courses. Although such crucial developments have now occurred, to a varying extent, throughout the world, the UK has always been at the forefront in this respect. Thus, in 1976, the UK Science and Engineering Research Council established a Science-based Archaeology Committee which was provided with funds to support the development of new scientific techniques and approaches to the study of archaeological material, and such research funding has continued in the UK in one form or another up to the present day. Furthermore, during the 1970s, the University of Bradford introduced one-year MA and three-year BTech courses in archaeological science, and now, few students in the UK can emerge from an undergraduate archaeology course without some knowledge of archaeological science, and many can be classed as true archaeological scientists. As discussed in detail in the next section in this chapter, progress in the reconstruction of the life-cycle of archaeological ceramics has depended mainly on the development and availability of new methods of scientific investigation. In contrast, progress in the interpretation of the life-cycle of archaeological ceramics has been, in large part, the result of improved communication and collaboration between the scientists involved in the reconstruction,
History of Scientific Research 9 and archaeologists from across the discipline, including field archaeologists, theoretical archaeologists, and ethnoarchaeologists.
Reconstruction of the Ceramic Life-Cycle In her book Ceramics for the Archaeologist, Shepard (1956) provided a description of the raw materials and processes involved in the production of archaeological ceramics, together with a summary of their physical properties. She then went on to describe how the raw materials used might be identified and production processes revealed. The analytical methods available to Shepard were essentially limited to binocular microscopy, petrographic microscopy, and optical emission spectroscopy (OES). As evident from the contents of the current volume, the range of analytical tools available for the study of archaeological ceramics has increased dramatically during the subsequent fifty or so years (Pollard et al., 2007). Of these, scanning electron microscopy (SEM), a range of methods for the determination of chemical composition, and organic residue analysis, using gas chromatography in combination with mass spectrometry (GC-MS), have had wide-ranging applications. In addition, there are a number of new analytical techniques which, by their nature, have had a more limited application for the study of archaeological ceramics.
Scanning Electron Microscopy Examination of polished sections using SEM has provided extremely valuable information on the production technology of archaeological ceramics, supplementary to that provided by optical microscopy (Tite, 1992). This is, in part, because the SEM provides a higher magnification (typically used in the range x100 to x500). However, more importantly, either an analytical SEM with attached X-ray spectrometer or an electron microprobe (EMP) can determine quantitatively the chemical composition of the different phases or components present. For example, the extent of vitrification in earthenware bodies can provide an estimate of the firing temperature, and the composition of the high temperature phases observed in porcelain bodies can provide information about the raw materials used in their production. Additionally, in cross-section, it has been possible to determine, for the first time, the composition of a slip or glaze entirely separately from that of the underlying ceramic body. It has thus been possible to distinguish between alkali- lime, high lead, and lead-alkali glazes (Tite et al., 1998), and identify the different opacifiers and colorants used. During the 1980s, the application of SEM analysis to archaeological ceramics facilitated research into a wide range of ceramics (earthenwares, stonewares, porcelains, and quartz- paste bodies), which resulted in significant advances in our understanding of their production technologies. Pioneers in this research include Kingery and Vandiver (1986) at Massachusetts Institute of Technology, and Tite and colleagues (Tite, 1992) at the British Museum.
10 MICHAEL S. Tite
Determination of Chemical Composition Initially, OES, INAA and X-ray fluorescence spectrometry (XRF) were the primary analytical techniques used in the study of archaeological ceramics. Of these techniques, INAA and XRF have continued in use until the present day, but OES was replaced first by atomic absorption spectroscopy (AAS) in the 1980s, then by inductively coupled plasma spectrometry with OES (ICP-OES) in the 1990s, and finally, by ICP with mass spectrometry (ICP- MS) at the beginning of the current millennium. In comparison with INAA, ICP-MS has similar detection limits but can analyze for a much wider range of elements. In addition, ICP-MS does not require access to nuclear reactors, which are becoming less widely available. However, INAA, in using a powder sample rather than the acid dissolution required with ICP-MS, retains significant advantages in terms of easy interlaboratory comparisons and rapid sample preparation. The primary role of these techniques has been the determination of the chemical compositions of ceramics for provenance studies with the aim of establishing where the ceramics were produced. Initially, chemical analysis was regarded as an alternative to thin-section petrography, especially in the case of fine-grained ceramics or when only the more ubiquitous non-plastic inclusions, such as quartz and shell, were present. However, Shepard, in the Preface written for the fifth printing (1965) of Ceramics for the Archaeologist, was highly critical of the use of chemical analysis in isolation. As a result, a more fully integrated approach to provenance studies has been progressively developed (Heidke and Miksa, 2000). A crucial first step in any such study should now be to group the entire assemblage from a site into pottery types on the basis of observed style (e.g. shape, surface decoration) and temper type, as determined by examination with a low-power binocular microscope. Second, the local geology should be assessed, and fieldwork undertaken to identify and collect samples of locally available clays and sands. Only then should both thin-section petrography and chemical analysis be undertaken on representative sherds of each pottery type.
Organic Residue Analysis From the 1990s onwards, GC-MS has been extensively used for the analysis of the organic residues surviving in archaeological ceramics (Heron and Evershed, 1993). Lipids, present as fats, oils, waxes, and resins, are the most useful surviving organic compounds for residue analysis, and can be used to establish whether animal, vegetable, or fish products were the original contents of pottery vessels. As well as contributing to the investigation of past human diets, organic residue analysis is particularly valuable in helping to understand the reasons for adoption of the first ceramics (Craig et al., 2013). In addition, the distribution of the bulk quantities of lipids over the interior of vessels has provided information on how different cooking pots were actually used (Charters et al., 1993).
Analytical Techniques with More Limited Application In the context of microscopy, the very high magnification possible with transmission electron microscopy (TEM) has played a crucial role in understanding the complex processes
History of Scientific Research 11 involved in the production of lustre decoration on glazed Islamic ceramics (Perez-Arantegui et al., 2001). In the context of the determination of chemical composition, handheld XRF systems are being increasingly used in the analysis of ceramic glazes. Although these instruments only provide semi-quantitative data and do not analyze for the lighter elements, the particular importance of this technique is that it is entirely non-destructive, and the instrument can be brought to and used on archaeological sites and in museums where the ceramics are located. In the context of phase analysis, the well-established method of X-ray diffraction (XRD), which identifies both mineral phases surviving from the raw materials and those formed during firing the body, has continued to be used to provide information about the raw materials and firing temperatures employed in ceramic production (Heimann and Maggetti, 2014: 73). More recently, Raman spectroscopy, which is again non-destructive, has been increasingly employed to identify the pigments used to decorate ceramics, as well as the colorants and opacifiers used in ceramic glazes (Smith and Clark, 2004).
Interpretation of the Ceramic Life-Cycle In order to achieve satisfactory interpretation of the reconstructed life-cycle of archaeological ceramics, it is essential that collaboration between the scientist involved in the reconstruction and the archaeologist who provided the ceramics is successful. Unfortunately, there were some tensions between scientists and archaeologists when such research first began in the 1950s (Preface to fifth printing of Shepard, 1965), and by the early 1990s there was considerable criticism of archaeological science by archaeologists and curators, particularly in the context of artifact studies (Dunnell, 1993). However, by the end of the 1990s these tensions had significantly reduced, in large part as a result of a more sustained dialogue between archaeological scientists and archaeologists. Thus, a high proportion of university archaeology departments, and particularly those in the UK, now include archaeological scientists on their staff, and a significant number of PhD students are jointly supervised by an archaeological scientist and archaeologist. In addition, there are an increasing number of research excavations, such as Batán Grande in Peru (Shimada and Craig, 2013), and Çatalhöyük in Turkey, excavated since 1993 by Ian Hodder (), on which archaeological scientists are actively involved in the in situ examination of artifacts, the selection of samples for analysis back in the laboratory, and fieldwork to locate possible sources of the relevant raw materials, together with participation in experimental artifact replications. Fred Matson (1965), who introduced the concept of ceramic ecology, was one of the first scholars to emphasize that the effective interpretation of the life-cycle of archaeological ceramics requires an holistic approach. Such an approach takes into account the fact that production, distribution, and use are firmly embedded within the overall environmental, technological, economic, social, political, and ideological contexts which in turn both impinge on (i.e. constrain and drive) and are impacted by the production, distribution, and use of ceramics.
12 MICHAEL S. Tite Taking Matson’s ceramic ecology concept as their starting point, Sillar and Tite (2000) proposed a framework for the interpretation of the life-cycle of archaeological ceramics and, in particular, the factors that determine technological choice and change. For example, they believe that the first factors influencing technological choice are the availability and performance characteristics of the raw materials, tools, energy sources, and procedures for forming, surface treatment, and firing of the pottery. Secondary factors which influence technological choices include patterns of trade and exchange, which impact the scale of production and the extent of craft specialization (e.g. household, workshop, or factory production) (Peacock, 1982). The intended use of the ceramics can also influence technological choice by requiring specific performance characteristics, and hence, physical properties of the ceramics (Tite et al., 2001). When considering the factors that determine technological choice, it is important to adopt an emic approach in which one tries to understand what the artisans thought they were achieving through their particular technological choice, and what the people using the ceramics thought were their important properties. In this context, ethnoarchaeological studies (Longacre, 1991) have a definite contribution to make to archaeological ceramic analysis by exposing us to alternative ways of thinking about the world. Thus, ethnoarchaeology reminds us that the choice of raw materials and production techniques can express aspects of social identity and/or have ideological significance (Jones, 2000). Similarly, ceramics can be used to establish bonds between social groups, and that, for this purpose, visual and tactile characteristics can be more important than utilitarian properties, such as mechanical strength and thermal shock resistance. After production and technological choice, the second aspect of the ceramic life-cycle requiring interpretation is the distribution of ceramics away from identified production centers or sources of raw material. For such “provenance” studies to make a significant contribution to our understanding of trade and exchange, large-scale, long-term projects are essential. One example has been the study of Mayan pottery, which has involved more than 100,000 fabric characterizations by binocular microscopy and more than 10,000 INAA analyses (Bishop, 1994). Another example is the wide-ranging analysis, over a fifty-year period, of Minoan and Mycenaean pottery, which started with OES analyses (Catling et al., 1963) but subsequently made use of a fully integrated approach (Day et al., 1999). Both these studies have demonstrated that, by interrogating provenance data beyond simplistic classifications of local and non-local, they can provide significant insights into prehistoric social landscapes. Further, quantification of the fall-off in the quantity of a particular ceramic type with increasing distance from source can, in principle, be used to assist in distinguishing between the different modes of trade or exchange (e.g. down-the-line, middleman, or central place) although, in reality, an unambiguous result is rarely obtained (Renfrew, 1975).
Future Developments For the investigation of unglazed archaeological ceramics (i.e. essentially earthenwares), no major developments in the range of scientific techniques used are envisaged in the
History of Scientific Research 13 immediate future. Instead, the emphasis will continue to be on the employment of an holistic approach in which, as far as possible, production, provenance, and use of a complete ceramic assemblage are investigated. For the future, greater use of handheld XRF for body analyses is envisaged, together with more frequent fieldwork in an attempt to identify sources of raw materials. Crucial to such studies is the continuing improvements in collaboration between archaeological scientists and archaeologists through greater involvement of the former in fieldwork and excavation, and in defining the archaeological research questions. Additionally, using the methodology developed by Richard Evershed and colleagues at the University of Bristol (Evershed, 2008), unglazed ceramics will be increasingly exploited as a source of organic residues to detect the processing of different commodities and the investigation of dietary change. Further research into the factors determining the mechanical and thermal properties of unglazed ceramics is also envisaged (Heimann and Maggetti, 2014: 18). Conversely, it is in the investigation of glazes and high-gloss layers applied to earthenwares, stonewares, and porcelains that the more recent (and future) developments in methods of scientific examination are likely to have the greatest impact. In this context, the high X-ray flux generated by synchrotron radiation sources can be used for high-resolution XRF and XRD, as well as for both extended X-ray absorption fine structure (EXAFS) and X-ray absorption near-edge spectroscopy (XANES), both of which provide information on element valence state. Rutherford backscattering spectroscopy (RBS) using a high-energy ion beam can provide information on composition depth profiles. Following on from its use in the analysis of ancient glasses (Shortland et al., 2007), laser ablation ICP-MS (LA-ICP-MS), because of its low detection limits and wide element range, has considerable potential for supplementing EPMA analyses of ceramic glazes and highgloss layers, the same polished sections being used for both methods. However, because of their inhomogeneity, the method will have only limited use in the analysis of ceramic bodies, probable exceptions being some of the more homogeneous porcelain bodies (Wallis and Kamenov, 2013). Again, stable isotope analyses, previously used in the study of ancient glasses to provide information on the sources of the raw materials (Degryse and Schneider, 2008), could similarly be used in the study of ceramic glazes. Thus, strontium (Sr), neodymium (Nd) and oxygen (O) isotopes could be used to investigate the sources of the silica and alkali flux used to produce a glaze, and antimony (Sb) isotopes to investigate the source of antimony used in the production of lead antimonate yellow and calcium antimonate white opacifiers. For example, in the investigation of the production technology for Islamic lustre decoration, Pradell et al. (2008) have used micro-XRD to obtain information on the type and size of the copper (Cu) and silver (Ag) nanoparticles, XANES and EXAFS to determine the valence state of the Cu and Ag, and RBS together with TEM to determine the thickness of the lustre layer. Similarly, in the investigation of the production technology of Athenian red-figure ware, Walton et al. (2013) have used scanning transmission electron microscope (STEM) methods to determine the microstructure and chemistry of the gloss layer, XANES to determine the valency of the iron in the gloss layer, and LA-ICP-MS to search for compositional inhomogeneities within the gloss layer.
14 MICHAEL S. Tite
References Aitken, M. J. (1961). Physics and Archaeology (New York: Interscience Publishers Inc.). Bishop, R. L. (1994). “Pre-Columbian Pottery: Research in the Maya Region.” In: Scott, D. A. and Meyers, P. (eds), Archaeometry of Pre-Columbian Sites and Artifacts (Los Angeles, CA: Getty Conservation Institute), 15–65. Catling, H. W., Richards, E. E., and Blin- Stoyle, A. E. (1963). “Correlations between Composition and Provenance of Mycenaean and Minoan Pottery.” The Annual of the British School at Athens 58: 94–115. Charters, S., Evershed, R. P., Goad, L. J., Leyden, A., Blinkhorn, P. W., and Denham, V. (1993). “Quantification and Distribution of Lipids in Archaeological Ceramics: Implications for Sampling Potsherds for Organic Residue Analysis and the Classification of Vessel Use.” Archaeometry 35: 211–223. Craig, O. E., Saul, H., Lucquin, A., Nishida, Y., Taché, K., Clarke, L., Thompson, A., Altoft, D. T., Uchiyama, J., Ajimoto, M., Gibbs, K., Isaksson, S., Heron, C. P., and Jordan, P. (2013). “Earliest Evidence for the Use of Pottery.” Nature 496: 351–354. Day, P. M., Kiriatzi, E., Tsolakidou, A., and Kilikoglou, V. (1999). “Group Therapy in Crete: A Comparison between Analyses by NAA and Thin Section Petrography of Early Minoan Pottery.” Journal of Archaeological Science 26: 1025–1036. Degryse, P. and Schneider, J. (2008). “Pliny the Elder and Sr-Nd Isotopes: Tracing the Provenance of Raw Materials for Roman Glass Production.” Journal of Archaeological Science 35: 1993–2000. Dunnell, R. C. (1993). “Why Archaeologists Don’t Care about Archaeometry.” Archaeomaterials 7: 161–165. Evershed, R. P. (2008). “Organic Residue Analysis in Archaeology: the Archaeological Biomarker Revolution.” Archaeometry 50: 895–924. Heidke, J. M. and Miksa, E. J. (2000). “Correspondence and Discriminant Analyses of Sand Temper Compositions, Tonto Basin, Arizona.” Archaeometry 42: 273–299. Heimann, R. B. and Maggetti, M. (2014). Ancient and Historical Ceramics: Materials, Technology, Art, and Culinary Traditions (Stuttgart: Schweizerbart). Heron, C. and Evershed, R. P. (1993). “The Analysis of Organic Residues and the Study of Pottery Use.” In: Schiffer, M. B. (ed), Archaeological Method and Theory, vol. 5 (New York: Academic Press), 247–284. Jones, A. (2000). “Life after Death: Monuments, Material Culture and Social Change in Neolithic Orkney.” In: Ritchie, A. (ed), Neolithic Orkney in its European Context (Cambridge: The MacDonald Institute), 127–138. Kingery, W. D. and Vandiver, P. B. (1986). Ceramic Masterpieces—Art, Structure and Technology (New York: Free Press (Macmillan)). Longacre, W. A. (1991). Ceramic Ethnoarchaeology (Tucson: University of Arizona Press). Matson, F. R. (1965). “Ceramic Ecology: An Approach to the Study of the Early Cultures of the Near East.” In: Matson, F. R. (ed), Ceramics and Man. Viking Fund Publications in Anthropology No. 41 (Chicago, IL: Aldine), 202–217. Peacock, D. P. S. (1982). Pottery in the Roman World (London: Longman). Perez-Arantegui, J., Molera, J., Larrea, A., Pradell, T., Vendrell, M., Borgia, I., Brunetti, B. G., Cariati, F., Fermo, P., Mellini, M., Sgamellotti, A., and Viti, C. (2001). “Luster Pottery from the Thirteenth Century to the Sixteenth Century: A Nanostructured Thin Metallic Film.” Journal of the American Ceramic Society 84: 442–446.
History of Scientific Research 15 Pollard, M., Batt, C. M., Stern, B., and Young, S. M. M. (2007). Analytical Chemistry in Archaeology (Cambridge: Cambridge University Press). Pradell, T., Molera, J., Smith, A. D., Climent-Font, A., and Tite, M. S. (2008). “Technology of Islamic Lustre.” Journal of Cultural Heritage 9: e123–e128. Renfrew, C. R. (1975). “Trade and Interaction.” In: Sabloff, J. A. and Lamberg-Karlovsky, C. C. (eds), Ancient Civilizations and Trade (Albuquerque: University of New Mexico Press), 3–59. Sayre, E. V. and Dodson, A. W. (1957). “Neutron Activation Study of Mediterranean Potsherds.” American Journal of Archaeology 61: 35–41. Shepard, A. O. (1956). Ceramics for the Archaeologist. Carnegie Institution Publication 609 (Washington, D.C.: The Carnegie Institution). Shimada, I. and Craig, A. K. (2013). “The Style, Technology and Organization of Sicán Mining and Metallurgy, Northern Peru: Insights from Holistic Study.” Chungara, Revista Chilena de Antropología 45: 3–31. Shortland, A., Rogers, N., and Eremin, K. (2007). “Trace Element Discriminants between Egyptian and Mesopotamian Late Bronze Age Glasses.” Journal of Archaeological Science 34: 781–789. Sillar, B. and Tite, M. S. (2000). “The Challenge of ‘Technological Choices’ for Materials Science Approaches in Archaeology.” Archaeometry 42: 2–20. Smith, G. D. and Clark, R. J. H. (2004). “Raman Microscopy in Archaeological Science.” Journal of Archaeological Science 31: 1137–1160. Tite, M. S. (1992). “The Impact of Electron Microscopy on Ceramic Studies.” In: Pollard, A. M. (ed), New Developments in Archaeological Science (Proceedings of the British Academy, vol. 77) (Oxford: Oxford University Press), 111–131. Tite, M. S. (1999). “Pottery Production, Distribution, and Consumption—the Contribution of the Physical Sciences.” Journal of Archaeological Method and Theory 6: 181–233. Tite, M. S. (2008). “Ceramic Production, Provenance and Use—a Review.” Archaeometry 50: 216–31. Tite, M. S., Freestone, I., Mason, R., Molera, J., Vendrell-Saz, M., and Wood, N. (1998). “Lead Glazes in Antiquity—Methods of Production and Reasons for Use.” Archaeometry 40: 241–260. Tite, M. S., Kilikoglou, V., and Vekinis, G. (2001). “Review Article: Strength, Toughness and Thermal Shock Resistance of Ancient Ceramics, and their Influence on Technological Choice.” Archaeometry 43: 301–324. Wallis, N. J. and Kamenov, G. D. (2013). “Challenges in the Analysis of Heterogeneous Pottery by LA-ICP-MS: A Comparison with INAA.” Archaeometry 55: 893–909. Walton, M., Trentelman, K., Cummings, M., Poretti, G., Maish, J., Saunders, D., Foran, B., Brodie., M., and Mehta, A. (2013). “Material Evidence for Multiple Firings of Ancient Athenian Red-Figure Pottery.” Journal of the American Ceramic Society 96: 2031–2035.
Pa rt I I
R E SE A RC H DE SIG N A N D DATA A NA LYSI S
Chapter 3
Designing Ri g orou s Researc h Integrating Science and Archaeology
Jaume Buxeda i Garrigós and Marisol Madrid i Fernández Introduction Archaeology, as social science, is a factual science; it deals with people and social systems, natural resources, and artifacts, all of which must be considered concrete things, whether natural or artificial, living or inanimate, individual persons or social systems.1 Ceramic artifacts are one of the most important means (Neustupný, 1993) in the archaeological process of inquiry because of their ubiquity and abundance in the archaeological record. Thus, a large number of archaeological research questions are totally or partially based on the study of pottery. However, several of the important research questions need to be answered not on archaeological grounds, but by first considering the material nature of the ceramics under investigation. The previous statement illustrates the basic duality of archaeometric research. Even if the problems under inquiry are of an archaeological nature, the objects of analysis are of a material nature. Archaeological concepts do not necessarily map those of the material empirical reality under study. Ceramics, as artifacts, are concrete inanimate objects artificially (fire-induced) transmuted from the natural raw materials employed in their manufacture. All ceramics can be considered as composite materials; they are made from two or more constituent materials with significantly different physical and/or chemical properties that result in a material with characteristics different from the individual components (e.g. clayey matrix, aplastic inclusions, porosity). Some of those individual components might remain separate and distinct within the finished materials structure. Thus, ceramics are characterized by their main intrinsic and essential source properties of composition,2 structure,3 shape, and size, as well as by their main intrinsic essential derived properties, such as mechanical, thermal, color, and porosity.
20 J. Buxeda i Garrigós and M. Madrid i Fernández SIMPLE CERAMICS BODY
COMPLEX CERAMICS COMPOSITE DECORATIONS
Clay and aplastic raw materials Intrinsic essential source properties
Pigments, glazes, glosses and slips raw materials
Intrinsic essential derived properties
Composition Structure
Color Mechanical – Thermal properties Porosity
(Shape) (Size)
Intrinsic essential source properties Composition Structure
Intrinsic essential derived properties Color Mechanical – Thermal properties Porosity/Impermeability
(Shape) (Size) Paste preparation Forming (Surface treatments)
Decoration preparation
Unfired pottery (shaped paste) Intrinsic essential source properties
Intrinsic essential derived properties
Composition Structure
Color Mechanical – Thermal properties Porosity
(Shape) (Size)
Fired/Unfired pottery Intrinsic essential source properties Pre-firing decorations
Intrinsic essential derived properties Color Mechanical – Thermal properties Porosity
Composition Structure (Shape) (Size)
Multiple cycles
Firing
Post-firing pigments
Life finished pottery Intrinsic essential source properties
Intrinsic essential derived properties Color Mechanical – Thermal properties Porosity/Impermeability
Composition Structure (Shape) (Size)
Use and post-depositional time Formation processes
As received state pottery Intrinsic essential source properties Composition Structure
Intrinsic essential derived properties Color Mechanical – Thermal properties Porosity
(Shape) (Size)
Figure 3.1 Diagram flow of the states of ceramics from manufacture to the archaeological record. For the sake of simplicity, we divide ceramics into two general categories: simple and complex (Figure 3.1). A simple ceramic, even with added elements (spouts, handles, etc.), is formed only by the materials of the body and the simple decorative forming techniques (such as polishing or incising) made out of it (see e.g. Rye, 1981). Contrariwise, complex ceramics exhibit distinct parts (such as glazes or glosses) made of materials other than those of the body, called composite decorations. The material nature of ceramics, therefore, is that of a composite material, which can possess a complex structure composed of different parts made out of different materials for different purposes. The material nature of ceramics needs then to be studied either as the bulk composition of the body or as each of its different individual components (e.g. clayey matrix, inclusions). And the same can be said for every one of the other parts that exhibit
Designing Rigorous Research 21 complex ceramics (e.g. glazes, which are composed with vitreous matrix, inclusions, porosity, and pigments). This material nature can be addressed from the chemical, mineralogical (and vitreous), or petrographic level, not as independent information, but as necessarily related things.
Ceramics and the Archaeological Process In order to be able to set up a representational numerical model (Hand, 2004) of the material nature of ceramics for archaeometric research, we propose a general model for the empirical system of their life history (Figure 3.1). A simple ceramic flows in a linear sequence starting with the clays and aplastic raw materials that will be used to produce first a paste, and then the unfired ceramic, shaped through various forming techniques and, possibly, the addition of simple surface treatments. Paste preparation, forming, and surface treatments can be considered processes that transmute the original natural facts into the artificial, synthetic material; that is, the unfired ceramic.4 This process entails different degrees of complexity for different ceramics, from a simple event to a large and difficult collection of events, involving a variety of raw materials and auxiliary artifacts. Thus, ceramic complexity observed in the natural raw materials is largely culturally induced by humans. Unfired pottery is always a different shape and size than its raw material components, and the composition and structure may also be significantly changed. Thus, the derived properties of the ceramic are also accordingly different from those possessed by the raw materials. The firing process will again transmute the unfired vessel to ceramic (life-finished pottery) by firing-induced changes that will further affect its composition and structure, and derived properties accordingly. Changes in ceramics related to firing are dependent upon firing conditions, such as atmosphere, temperature, and type of pyrotechnical installation, in a way that several unfired ceramics sharing the same properties of composition and structure could end up with significant differences. Thus, one paste could end up as several different fabrics and,5 owing to different firing conditions, exhibit different mineralogical compositions, microstructures, and derived physical properties. Finally, ceramics in systemic contexts (Schiffer, 1987)—that is, in life assemblages (Orton, 2000)—will be used until discarded, at which point they enter into the archaeological context. Formation processes of the archaeological record, including weathering, may also significantly change the intrinsic essential source properties. Composition and structure of as-received-state pottery may have undergone major changes (e.g. Buxeda, 1999; Secco et al., 2011); shape and size, after breakage and loss of completeness, may become unknowable (e.g. Orton, 1993).6 Obviously, derived properties will change accordingly and/or may become non-determinable (e.g. the grayish color developed in ceramics after pyrite crystallization induced by sulfate-reducing bacteria as in Buxeda et al., 2005, or Secco et al., 2010). Production of a typical Roman amphora is a good example of a simple ceramic; whatever complexities exist in the manufacture process, paste preparation, forming, simple superficial treatments, and firing, it flows in a simple linear sequence (Figure 3.1, Simple Ceramics body flows in straight sequence from “Clay and aplastic raw materials” to “Life-finished pottery”). Other types of ceramics flow in sequences of increasing complexity in the sense that they include composite decorations. For example, unfired Terra Sigillata is completed
22 J. Buxeda i Garrigós and M. Madrid i Fernández with a simple decoration technique, the application of a layer that becomes a gloss after a single firing (Figure 3.1, Simple Ceramics body flow from “Clay and aplastic raw materials” to “Unfired pottery (shaped paste),” while Complex Ceramics–Composite Decorations flow from “Gloss raw materials” to “Unfired pottery,” pre-firing decoration is added, and then the complex ceramic is fired and flows to “Life-finished pottery”). The manufacture of some decorated Attic black gloss vases flows along a similar path to Terra Sigillata, but after firing, a second decorative cycle is employed, which adds several post-firing pigments but without a second firing (Figure 3.1, like Terra Sigillata, but after firing it flows through “Multiple cycles” again to Complex Ceramics–Composite Decorations and from “Gloss raw materials” to “Unfired pottery,” and then through “Post-firing pigments” to “Life-finished pottery”). Another type of pottery, majolica, has a more complicated production sequence. First, undecorated vessels are fired to produce what is called bisque pottery. Next, a decorative layer, usually including pigments and opacifiers, is applied to the bisque pottery in order to produce a decorated glaze during a second firing (Figure 3.1, like Roman amphora, but after firing it flows to Complex Ceramics–Composite Decorations and then flows from “Glaze raw materials” to “Fired pottery,” and through “Pre-firing decorations” it flows to a second “Firing” and then to “Life-finished pottery”). An even more complex production sequence is involved in the manufacture of lustre pottery; involving a second decoration cycle after the bisque firing and glazing of the vessel during which powdered metal is applied to the glazed surface, creating a metallic or iridescent sheen after a third firing (Figure 3.1, like majolica, but after the second firing it flows again to Composite Decorations from “Glaze raw materials” to “Fired pottery,” and, again, through “Pre-firing decorations” it flows to a third “Firing” and then to “Life-finished pottery”). In all of these examples, the finished pottery exhibits a body which is by itself a composite material. Moreover, complex ceramics also exhibit composite decorations with distinct layers of complex composite materials added to it; that is, the glosses, glazes, pigments, and metals. Minerals and microstructures formed at the interfaces among these distinct parts (body and composite decorations) also contribute to the complexity of these ceramics and provide valuable information about their manufacture and the sociocultural environment in which they were consumed. Human-induced complexity may, in turn, be related to social complexity and to the development of scientifically based technology. Thus, it can be hypothesized that higher complexity in ceramics manufacture implies higher social complexity, as well as the development of technology. At the same time, more sophisticated or complex production sequences may also imply increased standardization of the manufacture processes and, consequently, a decrease in the variability of the manufactured products.
Research Problems and Routine Problems in Archaeology and Archaeometry As discussed above, archaeometric analysis of pottery within archaeological studies is focused on the material nature of the artifacts. From an analytical perspective, ceramic analysis is an inverse problem; we start with as-received-state pottery (i.e. the excavated, post- depositional pottery), but all our research questions relate to earlier stages in its life history
Designing Rigorous Research 23 (life-finished pottery, unfired pottery, and raw materials) as well as the processes connecting those states (Figure 3.1). The methodological and theoretical difficulties arising from the inverse problem of archaeological ceramic analysis, together with the necessity of establishing and continuously developing critical background knowledge, methods, and theories, are archaeometric research problems. These problems are essential for archaeological materials analysis as a scientific discipline, but their solutions are not necessarily significant for archaeology in general. Crucial research problems for archaeological materials analysis include: • Development of new analytical techniques and methods and/or the development and application of existing techniques and methods to archaeological materials analysis. For example, the increasing basic research on mechanical and thermal properties of ceramics, the application of synchrotron radiation, or the use of portable X-ray fluorescence spectrometry (XRF) (Kilikoglou and Vekinis, 2002; Hein et al., 2008; Iñañez et al., 2013; Hunt and Speakman, 2015). • Development of relevant background knowledge to facilitate the reconstruction and interpretation of the processes involved in the manufacture, use, and post-depositional environment of ceramics. For example, studies about weathering and its effect on the chemical and mineralogical composition of archaeological ceramics (Buxeda, 1999; Secco et al., 2011). • Development of statistical methods and models for compositional data (Martín- Fernández et al., 2015). • Ethnoarchaeometric studies enabling archaeological materials scientists to observe directly the processes of pottery manufacture and more accurately reverse engineer archaeological ceramic production (David and Kramer, 2001; Buxeda et al., 2003). Archaeometric research problems enable the investigation of archaeometric routine problems. Routine problems, as opposed to research problems (Bunge, 1996: 81ff.), are well posed, their approach is well defined, and the solution can be partially foreseen, on the basis of the existing knowledge. Therefore, routine problems do not necessarily contribute significantly to the development of archaeological materials analysis as a scientific discipline, but they are essential for addressing relevant and important archaeological research problems. Archaeometric routine problems could not be answered without archaeometric research problems and their solutions, but the aim of routine problems is to be an active part of an archaeological research program. Archaeometric routine problems typically address two broad and related archaeological research questions: (1) identification of meaningful ceramic groups and provenance, and (2) aspects related to ceramic manufacture. Focusing now on archaeology, it is clear that nowadays it is considered an increasingly scientific discipline in terms of its methods and theories. However, it is still an undeveloped or underdeveloped scientific field, because of the uneven knowledge of its almost endless case studies from different places and times (in essence, the object of the study of archaeology is the entire material remains of mankind in every place on Earth and from every time). Thus, archaeometric routine problems may still contribute significantly to the advance of a particular case study when its archaeological background knowledge is still scarce by providing an insight into previously unforeseen knowledge. However, in archaeological issues already well known, the only significant scientific contribution is to deepen such knowledge with
24 J. Buxeda i Garrigós and M. Madrid i Fernández archaeological research problems that necessarily increase in degrees of complexity; that is, through large and well-designed archaeometric research programs.
Identification of Meaningful Ceramic Groups and Provenance The first step in identifying the provenance of archaeological ceramics is to objectively classify them into meaningful ceramic groups whose provenance can be identified in a second step. Groups are inferred from the material nature of the ceramics themselves, their intrinsic essential properties, and not from other relational and accidental properties, such as decorative motifs or themes, potters’ stamps, or the “criterion of abundance” (Harbottle, 1982) in the archaeological record. These latter properties may, however, play an important role in the background knowledge and evaluation of the archaeometric data. The duality of archaeometric research means that there is not necessarily a correspondence between the archaeological concept of a production center or workshop and the meaningful ceramic group in material terms. As will be seen in what follows, identification of ceramic groups and provenance ascription can be conducted mainly on the chemical and/ or petrographic levels. Chemical characterization allows ceramics to be mathematically represented in the space given by their elemental concentrations. In such space, the existing structure (i.e. the differential occurrence of points in this space, see Bishop and Neff, 1989) is assumed to represent the existing different meaningful ceramic groups. However, as illustrated in Figure 3.1, this mathematical space is that of the as-received-state pottery, in other words possibly weathered after alteration of these during its use-life and/or post-depositional processes. Meaningful ceramic groups are instead referred to the unfired pottery state (i.e. the mathematical space of their elemental concentrations before firing); and from which the life-finished pottery state is guessed not to differ in a significant way (i.e. in chemical terms, mainly changes related to loss on ignition, but also in volatiles; see Kilikoglou et al., 1988; Béarat et al., 1989; Cogswell et al., 1996). Unfired-pottery-state mathematical space is not the geographical space of archaeological production centers, because paste preparation can be a complex human-induced process of behaviors and expectations resulting in the procurement, combining, and processing of different raw materials in a single production center to create several different pastes, or contrariwise to create similar pastes in geographically distinct production centers. Meaningful ceramic groups defined in the elemental concentrations of mathematical space, when mapping the unfired-pottery-state mathematical space (i.e. when there is no significant effect from firing and weathering in the present-day chemical data), are the incertitude zone, or non-resolution space (adapted from Picon and Le Miere, 1987), and must be considered the smallest unit of a chemically based classification, referred to hereafter as paste compositional reference units (PCRU) (adapted from Bishop et al., 1982). After archaeological ceramics are classified into PCRUs, archaeological materials scientists can address the question of provenance. In order to do this, reference groups (RG), also called localized references (Picon and Le Miere, 1987), are needed for comparative analysis. RGs are PCRUs defined from the study of ceramic materials recovered at production centers (or workshops) and archaeologically identified, possibly together with extant raw materials used in the manufacturing process also recovered on the site. Relating meaningful ceramic groups (PCRUs) with production centers, RGs map groups from the elemental
Designing Rigorous Research 25 concentrations in mathematical space onto the geographical space where the production center is located. If we can assume that the data treatment of the as-received-state pottery elemental concentrations is close enough to what we could achieve if we could work straight on the elemental concentrations of the unfired pottery (i.e. if the firing and weathering effects are under control), provenance through chemical analysis is just a matter of relating ceramic groups (PCRUs) to RGs. Thus, the understanding of the paste preparation processes and the determination of composition of possible raw materials can be avoided. Reverse engineering the paste preparation process, including selection and treatment of raw materials, is a complex research problem by itself, the solution of which, while shedding light on a technical problem of ceramic manufacture, is not necessarily relevant for determining provenance. Thus, analysis of potential raw materials becomes a complementary practice, the purpose of which is to inform on technical aspects of the production sequence. However, raw-materials characterization is really advisable in those cases where production centers are not easily identified in the archaeological record; for example, because of the absence of kilns in pottery production (Hunt, 2012). It is important to highlight that provenance based on RG is a provenance ascription to actual production centers (or, more precisely, to the incertitude zone of the RG). Contrariwise, provenance based on raw materials is a provenance ascription to possible geographical (geological) source areas, not to actual production centers. At the petrographic level, since ceramics are composite materials, the study of non-plastic inclusions, as different individual components, might also enable provenance ascription in the unfired-state-pottery space. In general, extant inclusions can be understood as remnants of the raw materials not transmuted during manufacture and post-depositional processes. In this way, identifying inclusions is, in effect, identifying some of the original raw materials of the as-received-state pottery, partially canceling the inverse problem of archaeometric studies. However, since fabric characterization using petrography is not based simply on extant inclusions (Whitbread, 1995), identification of meaningful ceramic groups (or fabrics, in petrographic studies) is a far more complex problem because it is based on the whole intrinsic essential source properties of composition and structure at a petrographic level (i.e. including identification of extant non-plastic inclusions, frequencies, sizes and shapes, groundmass, vesicles, vughs, channels, concentration features, and the like). It is important to highlight that the state–space defined using petrographic methods is not the same as the state–space defined using bulk chemical methods. Mapping elements from petrographic onto the chemical state–space can be complicated; independent studies using petrographic and bulk chemical methods on the same samples may even lead to the identification of different meaningful ceramic groups. Petrographic and bulk chemical analyses are complimentary techniques, providing different but related types of information. Therefore, an integrated methodology is advisable. In short, archaeometric routine problems related to the classification and provenance of ceramics must, necessarily, be based on bulk chemical and/or petrographic characterization of as-received-state ceramics, and therefore subject to the problem of reverse engineering. The ultimate goal, however, is to be able to infer the compositions of archaeological ceramic assemblages (i.e. the diversity of meaningful ceramic groups in ceramic assemblages); a complex research goal in which archaeological inquiry totally subsumes archaeometric research. However, because of its importance and scope, this research aim will be considered in detail below.
26 J. Buxeda i Garrigós and M. Madrid i Fernández
Aspects Related to Ceramic Manufacture Archaeometric studies focused on understanding archaeological ceramic manufacture typically have one of the following fields of inquiry: a Research aims centered on understanding how life-fired pottery was made (i.e. the manufacture process). b Research aims related to the performance characteristics of life-fired pottery. c Research aims devoted to the study of manufacturing techniques/technological change. In order to design research focused on reconstructing how pottery was manufactured, we must understand that ceramic manufacture is a complex human-driven process related to social complexity and the development of techniques and scientifically based technology. This distinction between techniques and technology is significant; techniques is pre- scientific craftsmanship, while a technological process implies science-based research and development (Bunge, 1996: 197ff.). For example, ceramic manufacture processes can be placed on a continuum from earliest technical processes based on experience, to more recent technological processes based solely on knowledge. Technology implies the refinement of techniques using the results of pure and applied science (e.g. the new hightechnology ceramics). Reverse engineering the processes of ceramic manufacture (research aim a) is an inverse problem starting from the as-received-state pottery. Ideally, we would investigate ceramic manufacture by studying excavated production centers or workshops, where in addition to ceramics in different stages of the production process, extant raw materials may be found together with artifacts used in those manufacture processes (e.g. kiln structures, potters’ tools, etc.). Guessing the role played by each of these elements (ceramics, raw materials, and artifacts) in the ceramic manufacture process enables archaeological materials scientists to reconstruct the entire manufacture sequence; nevertheless, this reconstruction is a hypothesis which must be tested. An important consideration when attempting to identify the intended performance characteristics (Skibo and Schiffer, 2008) of the life-fired pottery (research aim b) is that most of the pottery recovered from a production center corresponds to waste materials, particularly over-fired ceramics or kiln wasters. These artifacts may obscure the identification of the intended performance characteristics of the finished pottery because they are or were considered defective in some way and were discarded. Therefore, it is better to study pottery from consumption (or reception) centers (e.g. towns, residences, storage facilities) to understand the intended or desired performance characteristics; vessels “in use” rather than waste vessels. Both research aims a and b, reconstructing manufacture processes and identifying performance characteristics, must be conducted on well-defined meaningful ceramic groups in order to contribute meaningfully to the complex empirical reality under study. Moreover, the number of individuals analyzed should be large enough that significant statistical inferences can be drawn from the data. The determination of the number of individuals that is large enough depends upon the structure of meaningful ceramic groups, as well as the number of different firing-induced fabrics existing within each meaningful ceramic group. In other words, ceramics at one production site may be stratified (in statistical sense) according
Designing Rigorous Research 27 to the existence of different meaningful ceramic groups. At the same time, each ceramic group may be stratified according to different fabrics. Sample selection and other sampling issues will be thoroughly discussed below. Several examples can illustrate the previous points. As a first case study, we consider research recently undertaken on materials from the Roman Terra Sigillata production center of Andújar, which clearly show how one workshop may end up with different reference groups (i.e. different meaningful ceramic groups) in the life-finished pottery stage of the manufacture process (Roca et al., 2013 and references therein). At this site, pottery and molds corresponding to different stages of production of the workshop were identified and sampled for archaeometric characterization. Bulk chemical analysis by X-ray fluorescence and the statistical data treatment of these compositional data, together with mineralogical analysis by X-ray diffraction, enabled the identification of six different pastes. One of them was a low calcareous paste that was used for the manufacture of cooking ware throughout the workshop’s life. The other five pastes were used in subsequent periods: three of them during the Julio-Claudian period, the other two during the Flavian period. The three pastes used in the Julio-Claudian period are calcareous, but with significant differences. The most calcareous one was used to produce painted Iberian pottery, coarse pottery, lamps, thin-walled pottery, and some forms of Italian Terra Sigillata-like pottery, together with a few examples of Hispanic Terra Sigillata molds. The other two pastes were only used to produce Hispanic Terra Sigillata pottery and molds. During the Flavian period, the two pastes in use exhibit a drop in the calcium content. In fact, one of them, only used for Hispanic Terra Sigillata pottery and molds, can be classified as low calcareous. The other one was used for the manufacture of Hispanic Terra Sigillata pottery and molds, but also for coarse pottery. These results illustrate the existence of different RG (i.e. localized meaningful ceramic groups in the mathematical space) in just one single workshop (i.e. one archaeological meaningful point in the geographical space), because of the existence of products for different purposes, and also diachronic changes in the manufacture process. The chemical analysis of several local clays (Picon, 1984) also suggests that at least four different raw materials could be exploited in order to produce those six different pastes. Thus, all the analytical evidence indicates great behavioral complexity related to the exploitation of raw materials and paste preparation processes. However, since the scope of this research project was the identification of the meaningful ceramic groups (six) in order to define their characteristics (i.e. the RG) to provide localized references for provenance studies, an in-depth study of locally available potential raw materials, and the actual manufacture processes leading to the different pastes, is not necessary. This subject remains as a subject by itself, albeit one related to identifying the technical/technological aspects of ceramic manufacture, not to provenance necessarily. An example of a rigorous and complete study that identified ceramic manufacture processes at a production center, including the necessary experimental tests, was conducted by Montana et al. (2007). As we have seen, production centers are ideal for understanding ceramic manufacture processes and establishing reference groups for provenance studies. Consumption centers, on the other hand, are ideal for the study of performance characteristics of meaningful ceramic groups; performance characteristics are given by the intrinsic essential derived properties of color, mechanical properties, permeability/waterproof qualities, and the like, resulting from the prepared paste and the firing process (Figure 3.1). One example is the Roman Terra Sigillata production center of Tritium Magallum, the biggest ceramic production center on
28 J. Buxeda i Garrigós and M. Madrid i Fernández the Iberian Peninsula identified to date. Despite the fact that an archaeological and archaeometric research project of the entire production center in order to know the processes of ceramic manufacture has yet to be conducted, its RG was chemically defined by Picon (1984) after analyzing a substantial sample of individuals from Tritium Magallum. The establishment of this RG has enabled its identification at several consumption centers throughout the Iberian Peninsula. Characterization of a large number of individuals, recovered from three different Roman cities in the Catalan area, facilitated identification of the preferred fabric manufactured by potters at Tritium Magallum, as well as the corresponding final color, an essential visual performance characteristic of these ceramics (Madrid, 2005; Buxeda et al., 2013; Madrid and Buxeda, 2014). Once the meaningful ceramic group was identified and linked to Tritium Magallum RG after the bulk chemical analysis by X-ray fluorescence, and the subsequent compositional data treatment, several fabrics were established, after the mineralogical analysis by X-ray diffraction according to the association of crystalline phases and their relation with mineralogical scales (Heimann, 1982; Maggetti, 1982). These fabrics were the result of three different ranges of estimated equivalent firing temperatures (EFT). EFT ranges affect the color of both ceramic paste and gloss, making it problematic to determine which color was intentional for the life-finished pottery. Therefore, a statistically significant sample of artifacts was considered with the aim of evaluating the distribution of ceramics Tricio-EFT Baetulo: Highly significant (χ2 = 21.79) Tarraco: Significant (χ2 = 5.63) Ilerda: Very significant (χ2 = 6.7)
Frequency
Baetulo: Not significant (χ2 = 2.99) Tarraco: Not significant (χ2 = 0.12) Ilerda: Not significant (χ2 = 0.19)
R3
Ilerda R2
Tarraco R1
Baetulo
χ2 = 4.38, df =4, p-value = 0.36
Figure 3.2 Multiple bar chart of Hispanic Terra Sigillata from Tritium Magallum recovered at Baetulo, Tarraco, and Ilerda, classified according the range of estimated equivalent firing temperatures (R1 to R3). R1 = 850–950°C, R2 = 950/1000°C, R3 = 1050–1150°C.
Designing Rigorous Research 29 from each EFT range for each city. As illustrated in Figure 3.2, the χ2 test for the contingency table for the frequencies of Terra Sigillata in the three cities, according to the EFT range, does not indicate significant differences in distribution (p-value = 0.36), suggesting that both variables, city and EFT range, are not associated. More detailed statistical analysis, the χ2 test among EFT ranges within each town, reveals no significant differences between the number of ceramics classed in ranges R1 (EFT 850–950°C) and R2 (EFT 950–1000°C) for the three cities. This means that Terra Sigillata fired at these EFT ranges are distributed in statistically similar proportions of the assemblage at these three sites. However, significant, very significant, and highly significant differences in distribution patterns are observed between Terra Sigillata with EFT ranges R2 and R3 (1050–1150°C) at Tarraco, Ilerda, and Baetulo respectively (Figure 3.2). These differences in distribution indicate that artifacts fired in EFT range R3 are significantly more widely consumed than those fired at either range R1 or R2. These results further enable us to infer that Terra Sigillata, fired at range R3, was the intended commercialized product manufactured at Tritium Magallum, even if Terra Sigillata fired at different EFT ranges and exhibiting different performance characteristics was also distributed at the same time. In contrast, the Gaulish Terra Sigillata workshop of La Graufesenque distributed ceramics during the same period as Tritium Magallum, and could be considered its main rival for the Terra Sigillata market on the Iberian Peninsula, but had a very different distribution pattern (see Madrid, 2005; Madrid et al., 2005; the former includes the comparison with La Graufesenque RG defined by Picon on Terra Sigillata from this production center). Characterization of a large number of individuals recovered from Emporiae, Baetulo, and Tarraco enabled the estimate of just one range of EFT (1050–1150°C), although several mineralogical fabrics were identified. This means that products distributed by La Graufesenque are more standardized than those of Tritium Magallum, exhibiting a greater homogeneity in the red color which characterizes this tableware. In addition to providing the red color, transverse rupture strength and the waterproof resistance of the gloss on Terra Sigillata are its most important intrinsic essential derived properties providing its performance characteristics. Research aimed at the evaluation of these properties must be conducted on a statistically significant number of samples recovered from consumption centers. Taken together, all the main intrinsic essential derived properties provide objective material-based criteria for quality, under which inferences about consumption can be hypothesized. Looking again at Terra Sigillata from the cities of Baetulo, Madrid, and Buxeda (2007) determined that there were several different grades or degrees of quality present after evaluating the transverse rupture strength, and the adherence and the sintering state of the gloss. The manufacture of different qualities of the same pottery type may indicate different expectations of the consumers about these vessels and their performance characteristics, and/or may be also related to different prices and uses of the pottery. Evaluating the quality of Terra Sigillata allows a deeper understanding of both its consumption and the sociocultural expectations of its production: more complex anthropological questions than trade and/or regional distribution patterns. The third aim (c) of ceramic manufacture studies is understanding and evaluating technical/technological change (Schiffer, 2011). These studies must be conducted on meaningful ceramic groups whose performance characteristics have been determined by previous studies, enabling the researcher to readily identify the intended life-finished pottery, the degree of standardization, and its quality in terms of consumption. Changes in the intrinsic essential
30 J. Buxeda i Garrigós and M. Madrid i Fernández source properties of composition, structure, and design (i.e. shape and size), and the intrinsic essential derived properties of the pottery, such as color, mechanical–thermal properties, and porosity, are potentially related to changes in performance characteristics and/or changes in the behavioral chain, such as acquisition of new knowledge, experience, or technologies, changes in the economic structure, and so on. For example, the Romanization process of the Catalan coast around the first century bc catalyzed drastic changes, impacting all aspects of daily life, from the abandonment of Iberian settlements after Roman towns were established, changes in land ownership and agricultural practices, and the end of Iberian pottery production. Roman pottery was manufactured locally in the newly established Roman production centers, in some cases in large quantities. Roman amphorae, for the distribution of wine, were mass produced in Catalonia, beginning with imitations of Italic Dressel 1 type amphorae and, subsequently, two new types unique to the Catalan area, Tarraconense 1 and Pascual 1. In order to understand these changes in Roman amphorae production on the Iberian Peninsula, finite element analysis (FEA) was conducted in order to compare the performance characteristics related to the design of the changing amphorae shapes and sizes (Vila et al., 2007). The FEA results suggest that the new designs, Tarraconense 1 and Pascual 1, improve the mechanical properties of the amphorae during ship transportation, possibly the most important performance characteristic for a cargo container. Until now, these new designs for amphorae manufactured in the Catalan area have been uniquely interpreted as an attempt to show a new local identity in wine production. Whether or not issues of identity expressed in design can still be conjectured, these new designs are, according to the results, actual technical improvements.
Formation Processes, the Archaeological Record, and Archaeometric Studies As stated above, one of the major aims of the archaeometric study of ceramics is to classify pottery on objective grounds using the intrinsic essential properties of its material nature. By first classifying pottery according to meaningful ceramic groups, researchers are able to determine how many and which meaningful ceramic groups are present in an archaeological pottery assemblage, and in what proportions; that is, the composition of this assemblage. A second analytical step is evaluating provenance or the place of origin of the ceramic materials used. At this point, it is important to clarify the meaning of several concepts related to the place where artifacts are found. First, the provenance of an artifact is the empirical fact of its physical location, and, by inference of association, its context; the basic unit of provenance, or context, in the archaeological record is the stratum (Lyman, 2012). A stratum is an empirical fact in the archaeological record comprising the smallest unit resulting from site formation processes (Schiffer, 1987). The materials belonging to each stratum form an assemblage. The assemblage associated with a stratum is the basic unit of material association in the archaeological record (Orton and Tyers, 1990). The importance of a context and the assemblage it contains is that its composition is believed to be related to different social facts about past living societies, such as activity type,
Designing Rigorous Research 31 social status, and exchange of goods, and potentially provide chronometric information. The latter is an inference derived from the most modern artifact in the assemblage, providing a terminus post quem for the formation of the stratum. Whether or not the chronologies of those artifacts are accurately known is a different archaeological problem, as is the life-span of those artifacts, something that can dramatically shift the terminus post quem. Contrariwise, the guessed social facts reflected in an assemblage are not necessarily directly observable. As has been acknowledged before (see e.g. Schiffer, 1987; Orton and Tyers, 1990; Lyman, 2008), archaeological assemblages result from living or systemic assemblages and formation processes of the archaeological record. It is not the scope of the present chapter to examine this problem thoroughly, but several points need to be highlighted. First, the assemblages directly related to the social facts of interest are the living or systemic assemblages not the archaeological assemblages. Second, archaeological assemblages, even if originated in those, are ultimately the result of discard and formation processes. Discard events may be complex and result in serious changes to the composition of the living assemblage. Hoards and funerary deposits can be archaeological assemblages intentionally created by and for a living society. Most archaeological assemblages, on the other hand, are the result of abandonment, discard, loss, and so on, obscuring and complicating the relationship between living assemblages and the materials entering archaeological ones. Third, even after archaeological assemblages are formed from materials departing a systemic context, new cultural and natural formation processes may further distort or enhance differences between the living and archaeological assemblages in a difference of degree. This is, for example, the case of a primary deposit that has been distorted by a posterior construction in the area. Thus, the archaeological assemblage is the as-received assemblage state, and reconstructing its relationship with a living assemblage is, again, an inverse problem. As a consequence, archaeometric research is, by necessity, conducted on archaeological assemblages; that is, on the as-received state of the archaeological record. The previous consequence has severe implications that may not have been sufficiently considered for archaeometric research. The definition of an archaeological assemblage is given by its link to an archaeological context, that is to a stratum. However, its intrinsic essential source properties are those of composition, fragmentation, and integrity. Composition of an archaeological assemblage results from the partition of a domain into mutually exclusive classes after a classificatory concept, such as ceramic classes or types. Compositions have parts and components as intrinsic essential source properties. Parts are the labels of each class or category of the partition; for example, Hispanic Terra Sigillata could be one of the parts of an archaeological assemblage. Components are the number or frequency measured for each class. Size of a composition, therefore, is an intrinsic essential derived property resulting from the sum of its components (Aitchison, 1986). Parts, components, and size can be considered characteristic or non-structural properties as opposed to the structural properties of composition, given by the intrinsic essential derived properties of richness and evenness, derived from components. Richness stands for the number of meaningful ceramic groups represented; that is, components whose frequency in the assemblage is higher than zero. Evenness stands for the distribution of the individuals among the meaningful ceramic groups in the assemblage: the greater the richness and evenness of the assemblage, the greater its diversity (Hurlbert, 1971). Fragmentation and integrity describe the state of ceramic artifacts extant in the archaeological assemblage as a result of formation processes. Ceramics are typically found broken
32 J. Buxeda i Garrigós and M. Madrid i Fernández and, even after refitting (or reconstruction), may remain incomplete. It has been suggested, therefore, that these properties could be used to shed light on the nature of formation processes responsible for an assemblage (see e.g. Orton and Tyers, 1990, which calls these properties brokenness and completeness). However, as discussed below, in our opinion fragmentation and integrity must be measured and used in order to evaluate the assemblages on the basis of their own properties. Thus, even if originated in living assemblages, studying archaeological assemblages linked to the contexts resulting from formation processes one must acknowledge that almost all intrinsic essential source and derived properties are related to the numerosity of ceramics classes (i.e. their frequencies) and their state in terms of recovering. Measuring numerosity or frequency is enumerating or counting the individuals present in an assemblage, frequently called quantification in the archaeological literature. Because of that, quantification methods are of primary importance in archaeometric studies. Several families of methods of measuring numerosity are used for archaeological ceramic analysis: sherd count; weight and related measures; estimated vessel-equivalents (EVEs); estimated vessels represented (EVREP), and pottery information equivalents (PIEs) standing for pseudo-count transformations (PCT). These methods and their properties have been widely discussed in the literature, for example Orton (1993); however, some aspects of quantifying archaeological ceramics still need to be considered. Following our aim to set up a representational numerical model for archaeometric research, it must be highlighted that the only representational numerical model for quantifying archaeological ceramic assemblages must be based upon sherd count and/or the maximum number of individuals (MxNI) (one of the EVREP measures): the first measure quantifies the number of actual fragments or sherds in the assemblage, while the second estimates the number of individuals in the assemblage after the refitting process, considering as one individual any sherd or groups of joined sherds that do not positively join any other sherd or group of sherds. Both measures are representational, since they correspond to facts of the empirical reality. All other measures are operational in the sense that they infer, at least partially, numbers of individuals on theoretical assumptions, not on the empirical reality. The use of the operational measures to quantify ceramic assemblages has two important consequences. First, most of these measures are based just on specific types of sherds, such as rims, handles, and bases. Therefore, not all sherds are equally weighted in the archaeological analysis, despite the fact that all sherds possess the intrinsic essential source properties of composition and structure needed for archaeometric characterization. In several situations we could even analyze individuals that were not counted at all, or we could even have more individuals characterized in one assemblage than individuals estimated by the archaeological study. For example, we could characterize archaeometrically a wall fragment sherd that would not be used at all in a minimum number of individuals estimation based on rim sherds. Similarly, there are even some assemblages that only provide wall fragments, or similar, but no rims, bases, or handles used in such estimations. In those cases, the minimum number of individuals estimate is 0 (or some scholars may apply a correction estimating 1 individual, see e.g. Raux, 1998). Whichever the case (0 or 1) we could take several samples from the extant wall sherds for archaeometric characterization with more individuals characterized than estimated by the archaeological study. Second, and partially related to the first, operational measures do not provide an appropriate sampling framework that can be used to design rigorous sampling. One method is to estimate the number of individuals, and
Designing Rigorous Research 33 another method, with no necessary relation to the previous one, is to sample a set of individuals to be archaeometrically characterized. Representational measures ensure that there is an appropriate sampling framework, based on the empirical reality that includes every individual. Thus, every sampled individual has been counted when measuring numerosity, and sampling fractions are meaningful. This situation does not preclude the use of operational measures of numerosity for other types of analysis (see e.g. Arcelin and Tuffreau-Libre, 1998) because there is no reason to perform just one of them, but in order to conduct a sampling for archaeometric research representational measures should be present. This is certain for all archaeometric studies, but it is compulsory for studies devoted to the deep complexities of the archaeological record. If we use the term “individual” for a vessel estimated from the archaeological assemblage, whichever the quantifying method in use, and “true individual” for an actual vessel in the living or systemic context, we can see that representational measures (sherd counts and MxNI) are similar in the sense that both of them either estimate the number of true individuals or overestimate it (especially sherd counts). For example, we have two sherds and each one belongs to a different vessel; the estimation of two individuals, by sherd counts or by MxNI, would recover the number of true individuals. If we have two sherds belonging to one individual but not joining, they would be counted as two different individuals, overestimating the number of true individuals. Even if both measures can end up in overestimations, only the maximum number of individuals (MxNI) should be used for sampling purposes in archaeometric studies, since joined sherds are known to be part of the same vessel, and this crucial information is retained. Nevertheless, the number of true individuals sampled for archaeometric studies will be either the number of individuals included in the sample or fewer. This is easily understood if we consider that the number of true individuals sampled is the number of individuals in the sample, unless one or more true individuals are sampled more than once, because their extant parts do not join and are considered as different individuals in MxNI terms. Strategies for avoiding sampling more than one true individual at a time, such as sampling specific types of sherds (e.g. pointed bases in amphorae studies), may imply a selection that may produce some uncontrolled bias. There is no simple solution to that problem; not for samples based on the maximum number of individuals, neither for samples based on any other measure. This situation becomes critical if no refitting process has been performed because there is no control at all on whether different sherds may come from the same true individual. By using representational measures we can also describe the empirical reality providing some measures for fragmentation and integrity. (1) Fragmentation can be measured (i.e. quantified) using the number of fragments and the MxNI as follows:
FI =
1 N 1 ∑ ln , N i =1 Fi
where FI is the fragmentation index, N is the MxNI, and F is the number of fragments in the ith individual. (2) Integrity can be measured as the loss of information due to the same formation processes responsible for the preservation of smaller parts or fragments of the original vessel. Such processes can induce a gradual loss of information for each of the individuals estimated with the MxNI. For one individual estimated by MxNI, the maximum
34 J. Buxeda i Garrigós and M. Madrid i Fernández information retained corresponds to a complete vessel (CV), while the minimum corresponds to an uninformative isolated sherd that cannot be identified or associated with the general classification of vessels, such as rim, wall, or base. Between these two extremes of retained information several categories are established for intermediate amounts of retained information. Those categories are established according to the class of pottery under study and the classification system in use, and will be explained in a Terra Sigillata case study later in this section. Integrity of the assemblage, therefore, is based on the loss of information for all MxNI in the assemblage. If all individuals estimated as MxNI are CV, evenness of the assemblage will be maximized. Otherwise evenness will decrease according to the degree of information lost. Evenness, therefore, is measured according to information entropy, also known as the Shannon index (Shannon, 1948), by using in the calculation logarithms to base 2 (therefore, information entropy –H2–will be measured in bit). Thus, both measures for fragmentation and integrity have a direct translation to the empirical reality of the archaeological assemblage under investigation. The preceding discussion is based upon the basic units of context and assemblage; that is, the stratum and the materials that it contains. Those basic units are empirical facts, but they are not necessarily themselves the object of interest for an archaeological research problem. Often, archaeological research questions require analysis of multiple units. In these cases, the archaeological context of interest is built according to the inferential concept of association which allows for the aggregation of two or more basic units; that is, more than one stratum. Aggregation of strata implies the aggregation of assemblages, which may lead to an estimation of MxNI less than or equal to the sum of the MxNI of the individual assemblages, since sherds or groups of sherds from different assemblages may join (providing further valuable information on formation processes). When the assemblage under study is quantified using representational measures a sampling strategy can be devised, if needed at all, on the existing sampling frame which is equally based on the empirical reality. In the case of a stratified assemblage, where the population under study contains different meaningful ceramic groups, a stratified sampling strategy is appropriate, taking a random sample of each stratum (in this case, a stratum is a sub- population of the assemblage or meaningful ceramic groups). However, prior knowledge of the stratification of an assemblage on an archaeometric basis is usually lacking. Instead, what is not usually lacking is an stratification on an archaeological basis, based on (1) relational and accidental properties (such as decorative motifs, potters’ stamps, or the criterion of abundance), together with (2) an estimation of the intrinsic essential source properties of composition and structure determined macroscopically or using a stereoscope microscope, and (3) some indications on intrinsic essential derived properties such as color. As will be seen in the example later in this section, the state space of such archaeologically based stratification is not necessarily the same as the one defined by the intrinsic essential source properties of composition and structure provided by the archaeometric research. Therefore, it is not appropriate to condition the whole sampling strategy using archaeologically based stratification. What is more appropriate is to perform a probabilistic sampling in two or more surveys (1) starting with either the archaeological stratification or without any stratification at all. Then, an initial sample is taken using a stratified or a random sampling design, respectively. After this initial sample is analyzed and the results evaluated, (2) a post-stratification (i.e. stratification after a first survey) can be designed, now on an archaeometric basis. At this time, more surveys can be performed if needed according to these new and subsequent
Designing Rigorous Research 35 post-stratifications, until the necessary (or the attainable) number of individuals has been sampled. Besides, the sampling strategy can be organized according to multiphase sampling. The phases used in this strategy are related with stratification in the assemblage, not in terms of different meaningful ceramic groups, but in terms of variability within a single meaningful ceramic group. Typically, this variation exists because of (1) differences in ceramic fabrics arising from firing conditions, (2) weathering processes (Buxeda, 1999; Secco et al., 2011), (3) the non-normal distribution of inclusions (Buxeda et al., 2003), and (4) other special problematics. This stratification is not known before archaeometric characterization; therefore in the first phase, every individual in the sample is analyzed for a particular set of properties (by applying the same analytical techniques); for example, chemical characterization by means of X-ray fluorescence and mineralogical characterization by means of X-ray diffraction. In the next phase of sampling, a subsample of individuals is taken from those included in the first phase of sampling, not from the entire assemblage, and they are analyzed for different properties using different techniques, such as microstructural characterization by means of scanning electron microscopy and optical microscopy on thin sections. Additional phases, analyzing different properties, such as thermal shock resistance and hardness, can be performed on subsequent subsamples of the previous phase. This multi-phase sampling strategy avoids the often expensive and time-consuming analysis of a large number of individuals using multiple techniques before the stratification of the assemblage is known. It also ensures that techniques are not randomly applied to different individuals. To illustrate the previous discussion, we look at two ceramic assemblages from the Roman city of Baetulo. This example will show how two different contexts can be studied and compared by using representational measures of quantification of the ceramic assemblage, together with the determination of fragmentation and integrity. In a second step, the archaeological stratification used at the initial survey of the sampling process will be explained and compared with the post-stratification based on the archaeometric research. Focusing our attention on the archaeological research problem of the Terra Sigillata distribution at different contexts of the Roman town of Baetulo, the Terra Sigillata of two different contexts, LL85b and TV83, was selected for archaeometric study (Madrid, 2005; Madrid and Buxeda, 2014). Both contexts were dated to the end of the first century ad and were inferentially formed by aggregating several archaeological strata. LL85b was interpreted as fill from different parts of a Roman domus. The entire assemblage of 191 fragments, after refitting, contained 152 MxNI. One hundred and five of these MxNI were classified archaeologically as Italian Terra Sigillata (ITS); three were classified as Unknown Terra Sigillata (UTS). Both ITS and UTS were considered a single category (U/ITS) because UTS is manufactured in shapes similar to ITS but with different visual characteristics, especially the color of the gloss (orange-colored instead of the characteristic reddish one). The remaining individuals were classified archaeologically as Gaulish Terra Sigillata (GTS) for 31 MxNI and Hispanic Terra Sigillata (HTS) for 13 MxNI. These three categories (or strata), U/ITS, GTS, and HTS, make up the archaeological stratification of the population of interest. After estimating MxNI, in a second step, each individual (in MxNI terms) was classified according to several categories defined on the basis of the portion of information retained of the complete vessel (i.e. a qualitative estimation of how much of the original vessel in the systemic context has been recovered in the archaeological one). The first category is
36 J. Buxeda i Garrigós and M. Madrid i Fernández the actual MxNI, because the refitting process could not demonstrate that sherds assigned to different individuals belong to the same vessel. In that sense, for all these individuals (as MxNI) we have, at least, the information that they may belong to different vessels. The second category, MxNINU (MxNI non-undetermined) is used to describe individuals that could be identified as a part of a vessel, such as sherds from the rim, wall, or base. All MxNI will again be included in this second category provided that this information is also retained. On the contrary, those individuals that do not keep this information will only appear in the first category (MxNI) but not in this second one (MxNINU). The third category, MxNIA (MxNI associated), is for individuals that can be further classified as (associated to) a particular class of vessel, such as cups or bowls but not to a specific type (i.e. all MxNINU that are associated to a particular class of vessel will be also included in this category). The fourth category, MxNII (MxNI identified), corresponds to those MxNIA individuals that can be classified according to specific vessel type, such as Consp. 22 cup, but not to a subtype. The fifth category, MxNIT (MxNI classified according to subtype), is used for MxNII individuals which can be classified into subtypes, such as Consp. 12.2; MxNII individuals whose type does not contain subtypes can also be classified as MxNIT, such as Drag. 36. The penultimate category, CP (complete profile), is used to refer to MxNIT vessels whose profile can be totally reconstructed after refitting, while the last one, CV, stands for complete vessels. In the hypothetical case that the assemblage under study was formed by ten unbroken and totally complete vessels, all the previous categories would account for ten individuals. The breakage process and the loss of integrity would change this ideal situation to just ten (or even more than ten) in the first category and decreasing numbers in the subsequent categories. A bar graph illustrating the loss of information for the total ceramic assemblage from LL85b by archaeological strata defined according to the classes of Terra Sigillata is shown in Figure 3.3a. The information entropy (H2) for each class of Terra Sigillata was also calculated. This graph shows that most of the pottery in the assemblage is archaeologically classified as U/ITS, a class considered older than the final chronology of the context. Moreover, the information entropies are considered as relatively similar, ranging from 2.05 bit (GTS) to 2.24 bit (HTS). These information entropies are considered as relatively low ones, far from the maximum attainable when facing seven categories (2.8073 bit, i.e. the logarithm to base 2 of 7), and reflect a substantial loss of information in the assemblage. In the previous ideal case with ten MxNI corresponding to ten complete vessels, the bar graph would show a maximum evenness (all seven categories up to ten), and information entropy would be the maximum attainable (2.8073 bit). Another context at Baetulo, TV83, after the archaeological analysis, is believed to result from the abandonment of a group of tabernae located in a building at one of the decumani of the city. The total number of fragments (or sherds) in the assemblage of TV83 was 259 and the MxNI estimated after refitting was 182. Six out 182 were UTS and 14 were ITS, bringing the total MxNI for class U/ITS to 20. GTS accounted for 99 MxNI, and HTS accounted for 63 MxNI. The bar graphs and information entropies (Figure 3.3b) clearly illustrate a different distribution pattern for Terra Sigillata in context TV83 than LL85b. U/ITS is not well represented in the TV83 assemblage, while GTS and HTS, the Terra Sigillata classes synchronic to the formation of the context, are abundant. Information entropies also present a wider range, from 1.88 bit (U/ITS) to 2.22 bit (GTS). It is important to highlight that U/ITS from the TV83 context exhibits the lowest information entropy (i.e. the maximum loss of
Designing Rigorous Research 37 information or the lower integrity), something that seems to be expected for pottery much older than the time of the formation of the context. The bivariate plot of integrity, measured as information entropy, against fragmentation, measured by the fragmentation index FI, is shown in Figure 3.3c. If we center our attention on context TV83, this plot clearly illustrates that U/ITS is lying far away from the rest, having both lower integrity and higher fragmentation than GTS and HTS. This combination of traits, low integrity, and high fragmentation is hypothesized for ceramics heavily distorted LL85b
(a)
Total (H2 = 2.18) U/ITS (H2 = 2.19)
MxNI MxNINU MxNIA
GTS (H2 = 2.05)
TV83
(b)
Total (H2 = 2.16) U/ITS (H2 = 1.88)
MxNII MxNIT CP CV
GTS (H2 = 2.22)
MxNI MxNINU MxNIA
MxNII MxNIT CP CV
HTS (H2 = 2.08)
HTS (H2 = 2.24)
Frequency
Frequency
(c)
0.00
Fragmentation
U/ITS HTS
–0.05
GTS
–0.10
LL85b U/ITS TV83
GTS
–0.15 HTS
1.8
1.9
2.0
2.1
2.2
2.3
H2
Figure 3.3 (a) Multiple bar charts for LL85b context. (b) Multiple bar charts for TV83 context. H2 = information entropy (in bit). U/ITS = Unknown and Italian TS, GTS = Gaulish TS, HTS = Hispanic TS. MxNI = maximum number of individuals, MxNINU = MxNI nonundetermined, MxNIA = MxNI associated, MxNII = MxNI identified, MxNIT = MxNI classified according to subtype, CP = complete profile, CV = complete vessel. (c) Bivariate diagram of integrity (H2) vs fragmentation (FI). Black triangle: total of LL85b. Black circles: strata of LL85b. Gray triangle: total of TV83. Gray circles: strata of TV83.
38 J. Buxeda i Garrigós and M. Madrid i Fernández by formation processes that lead to a serious loss of information and a low rate of refitting (i.e. a high fragmentation). Contrariwise, if we focus our attention on the LL85b context, U/ITS has integrity and fragmentation similar to that of GTS and HTS. This situation in the LL85b context can be related with the higher frequency that this U/ITS class presents in front of GTS and HTS. These results suggest several viable hypotheses of Terra Sigillata distribution at Baetulo that are not directly related to consumption, but to formation processes of both contexts. The assemblage from TV83 is dominated by the pottery types contemporary to the hypothesized time of deposition. Few sherds of the older U/ITS group entered into the assemblage, and they correspond to small, isolated fragments seriously distorted by the intensive action of various formation processes. In contrast, the assemblage from LL85b contains U/ITS as its most frequent class, and its integrity and fragmentation are similar to GTS and HTS, suggesting that the filling of this context included material from an older context, already containing the U/ITS, that was simply redeposited in this secondary context together with a few contemporary ceramics (GTS and HTS). This latter pattern suggests that the filling of context LL85b was intentional, removing a previous deposit, while that of TV83 resulted from the natural infilling of a context after abandonment. Summarizing, the archaeological study enabled us to inferentially define two contexts TV83 and LL58B dated back to the end of the first century ad aggregating, in both cases, several archaeological strata. The sherds of Terra Sigillata of each assemblage were then archaeologically classified into three categories (U/ITS, GTS, and HTS), leading to an archaeological stratification of both TS assemblages. Then, after a refitting process, MxNI was estimated and each individual was then classified on the basis of the portion of information retained of the complete vessel (from MxNI to CV). The calculation of the integrity, as information entropy (H2), and the fragmentation, after the fragmentation index FI, for each class of Terra Sigillata within each one of the analyzed contexts (LL85b and TV83) enabled the comparison of both contexts. At this point, a clear differentiation of two synchronic contexts in terms of their formation processes has been hypothesized. It seems that formation processes distorted in different ways the original vessels by acting on their integrity and fragmentation, which in turn affects the estimation of MxNI. All this information (estimation MxNI, archaeological classification, integrity, fragmentation, and formation processes) is relevant for interpreting archaeological contexts and must be taken into consideration before comparing them in terms of consumption. Moreover, the archaeological analysis conducted so far has also provided a valid sampling frame, based on a representational model of the empirical reality that considers the whole ceramic assemblage regardless of the information retained by different sherds (i.e. it is not just based on especial shapes such as rims, bases, etc.). All sherds have been used on an equal basis in the archaeological analysis and have probabilities above zero to be sampled for the archaeometric study. Thus, this sampling frame can be safely used to design the routine archaeometric study that will provide information on the main intrinsic and essential source properties of composition and structure that would enable the archaeometric classification of this Terra Sigillata (i.e. its post-stratification). It is after this post-stratification that we will get the necessary understanding of our ceramic assemblages in terms of meaningful ceramic groups, to deepen the knowledge of our archaeological research problem (see Research Problems). Considering the financial and practical limitations imposed in the study of this Roman town of Baetulo,7 the routine archaeometric study was performed on a sampling fraction
Designing Rigorous Research 39 of 9.9% at LL85b (15 out of 152 MxNI), and 10.4% at TV83 (19 out of 182 MxNI). The first phase of this multiphase sampling was devoted to chemical characterization by X-ray fluorescence analysis and mineralogical analysis by X-ray diffraction. In subsequent phases, as stated above, complementary techniques (scanning electron microscopy and different tests for the assessment of mechanical properties) were also used. In that way, different meaningful ceramic groups were identified conducting to a post-stratification of the population (i.e. the Terra Sigillata in the ceramic assemblages of the TV83 and LL85b contexts) now based on the intrinsic essential source properties of composition and structure. It must be highlighted that there exists an important difference between the archaeological stratification (i.e. the classification in U/ITS, GTS, or HTS) and the archaeometric post-stratification (i.e. the classification in meaningful ceramic groups). Archaeological stratification is conducted on the whole ceramic assemblage under study (i.e. all MxNI are archaeologically classified). On the contrary, archaeometric post-stratification is only conducted on the individuals included in the sampling (i.e. the individuals included in the archaeometric study). Unless all MxNI have been archaeometrically characterized, post-stratification of the population is a matter of inference based on the results for the sampling fraction. As will be seen in what follows, the post-stratification of the Terra Sigillata ceramic assemblage is a partition of this domain into different parts (i.e. the meaningful ceramic groups). The numerosities of each part (i.e. their frequencies) are the components. As stated at the beginning of this section, any part with a component higher than zero will contribute to the richness of the assemblage, while the distribution of these components will define the evenness of the assemblage. Both properties are intrinsic essential derived properties based on components and serve to measure the diversity of assemblages (Buxeda et al., 2013). Measuring and comparing diversities of both assemblages is thus the first step after the archaeometric post-stratification. The study of diversity after archaeometric post-stratification revealed a richness of five for LL85b (Figure 3.4a) and six for TV83 (Figure 3.4b), which means that archaeometrically five and six different Terra Sigillata meaningful ceramic groups were identified in each assemblage. As expected, the state–space of the archaeological stratification, based upon three macroscopic/archaeological ceramic groups, is different than the state–space of the archaeometric research, which identified twice as many strata (or groups, i.e. classes) in each assemblage. At LL85b, three Terra Sigillata meaningful ceramic groups were identified in the U/ ITS class: Arezzo, Production A, and Latium-Campania. Classes GTS and HTS correspond to La Graufesenque and Tritium Magallum production centers respectively. The evenness graph illustrates the information entropy for this context as 1.69 bit (72.7% of the maximum value for five classes) (Figure 3.4a). In the Terra Sigillata assemblage from TV83, two productions were identified within each of the three macroscopic/archaeological ceramic groups: Arezzo and Pisa productions were identified for the U/ITS class; La Graufesenque and Montans productions for GTS; and products from Tritium Magallum and the area of Tritium for HTS. In this context, the evenness graph illustrates the information entropy of this context as 1.93 bit (74.7% of the maximum attainable for six classes), a value quite similar to that of LL85b. However, a fair comparison between the contexts in terms of diversity requires a correction for the difference in sample size from the two contexts, using a rarefaction process simulating that all compared samples had the size of the smaller one (Heck et al., 1975). This process was performed by random sampling without replacement for 15 individuals out of the 19 at TV83,
40 J. Buxeda i Garrigós and M. Madrid i Fernández LL85b(n = 15) Richness = 5
(b)
TV83(n = 19) Richness = 6
14
14
12
12
10
10 Frequency
Frequency
(a)
8 6
8 6
4
4
2
2
0
0 Arezzo
La.Grauf.
Prod..A
Lat..cp.
Tritium.M.
LL85b(n = 15; R = 5) – TV83(n =1 9; R = 6) Richness = 5.176 ± 0.75
(c)
La.Grauf Tritium
(d)
Arezz
Pisa
Montans Area.of.T.
TV83(n =19; R = 6) Richness = 5.176 ± 0.75
14 400
12
Frequency
10
300
8 200
6 4
100
2 0
0 4 5 6 1 2 3 Resampling of 15 individuals 1000 times (± 1s)
4 5 6 3 Richness in resampling of 15 individuals 1000 times
Figure 3.4 (a) Evenness graph for LL85b. (b) Evenness graph for TV83. H2 = information entropy, H2% = relative H2. (c) Evenness graph of the rarefaction experiment. Gray: LL85b, Black: TV83 (with 1 standard deviation error bars). (d) Bar chart of richness after the rarefaction experiment. repeated 1,000 times. The results (Figure 3.4c) illustrate that the evenness graphs for the two contexts cannot be considered significantly different from the rarefied samples at two standard deviations. A bar graph of richnesses in the rarefied sample (Figure 3.4d) also demonstrates that there is not a significant difference in this richness after resampling. Therefore, although both contexts appear to be derived from different formation processes, the diversity of the assemblages from each context, based on the archaeometric study, do not differ significantly. Two complementary points to consider about inferring the composition of a target population based on the results of the archaeometric characterization need to be addressed. These two points examine, on the one hand, the probability that one meaningful ceramic group that is present in the assemblage is included in the sample for the archaeometric study; on
Designing Rigorous Research 41 the other hand, in a reverse process, the reliability of the estimation of relative frequencies in the whole assemblage after the relative frequency observed in the archaeometric sample. For the first point, let us imagine that LL85b contained a meaningful ceramic group making up 10% of the total assemblage, in MxNI. What is the probability that any of the individuals of this meaningful ceramic group is included in the sample for archaeometric study? The answer is that in a sample of 15 individuals (i.e. one like our actual sample for this context), such meaningful ceramic group could easily be represented from zero to three individuals, with an accumulated probability of 0.94; the higher probability being one individual (p = 0.34), followed by two individuals (p = 0.27) (Figure 3.5). However, there is still a probability of 0.21 that none of the individuals included in the sample will be from this 10% meaningful ceramic group. As we have seen in the previous paragraph, it is highly probable that a meaningful ceramic group accounting for 10% of the whole assemblage (i.e. the population) is represented in the sample (sample size = 15 individuals) for the archaeometric study. Even so, in 21 out of 100 cases none of the individuals would be sampled. For the rest (79 out of 100 cases) one, two, three, four, and even five individuals could be included in the sample (Figure 3.5). Thus, provided that this meaningful ceramic group is included in the sample, we would see in the post-stratification after the archaeometric study that this meaningful ceramic group accounts for different percentages of the sample, from 6.7% (when only one individual is included) up to 33.33% (when up to five individuals are included). These enormous discrepancies that would end with very different inferences about the assemblage arise because of the small sample size (15 individuals). Thus, what would be the necessary sample size to make sound inferences? Since our target meaningful
Binominal Distribution: binomial trials =15, probability of success = 0.1 0.35 0.30
Probability Mass
0.25 0.20 0.15 0.10 0.05 0.00 0
5
10 Number of Successes
Figure 3.5 Binomial probabilities for n = 15 and p = 0.1.
15
42 J. Buxeda i Garrigós and M. Madrid i Fernández group in the assemblage accounts for 10% of the individuals, to identify a meaningful production group with a true probability (p) of 0.1, with a confidence level of 0.95 (α = 0.05), and a desired precision of estimate (e) of ± 0.05, the sample size for the archaeometric study should be 73 individuals out of 152 for context LL85b (if the assemblage of the context under study could be considered a large population—in statistical terms—the sample size should increase up to 139 individuals). This number (73) is significantly larger than the actual number of samples (15) selected for the study. This smaller sample size is able to estimate a true probability (p) of 0.1, with a confidence level of 0.8 (α = 0.2), and a precision of estimate (e) of ± 0.1 (the sample size under such conditions, for a population of 152 individuals, would actually be 14; the sample size of 15 would be the actual number for a large population). As demonstrated above, the archaeometric results drastically changed the archaeological stratification of Terra Sigillata at Baetulo, facilitating classification of the pottery into specific meaningful ceramic groups that can, often, be identified with geographical provenances (in the sense expressed above as minimum areas of provenance attribution). In this example, Terra Sigillata identified as Arezzo, Pisa, La Graufesenque, Montans, and Tritium Magallum can be associated with specific workshops or production centers, thanks to the existence of already established RGs. In the other cases, such as the Production A, Latium-Campania, and the area of Tritium, the meaningful ceramic groups cannot be identified with specific workshops (i.e. with geographical provenances), but are identified and hypothetically related to geographic/geological areas based on several evidences (some of them archaeological ones such as the names in potters’ stamps, the existence of still not archaeometrically analyzed kiln wasters, analytical results on Terra Sigillata molds from the town of Pozzuoli in the area of Latium-Campania, etc.; some others are of geological nature, such as the volcanic inclusions in the Production A whose origin is presumed in the Bay of Naples). The small sample size for both contexts impacts the desired precision of inferences about the composition of the whole assemblages. Nevertheless, the existence of a valid sampling frame, together with the possibility of designing an appropriated sampling strategy, always under a representational model close to the empirical reality, can lead to a significant improvement of understanding for an archaeological research problem whose complexity increases as soon as we try to pass from a general insight (the Roman town of Baetulo as a unique context) to a more detailed analysis (consumption patterns in different contexts within the same Roman town). Before we finish this section, we must emphasize that the archaeological research problem is far from over. The previous example explains how to face the received state of the studied assemblages in order to get an in-depth knowledge of the composition of our contexts. As we have seen, the influence of formation processes is mandatory in the contexts we recover during the archaeological excavation. Thus, moving to a life or systemic context (i.e. making inferences on the leaving societies in terms, for example, of consumption patterns) is still an inverse and complex archaeological problem whose solution needs to be approached with all the information gathered so far.
Conclusions Archaeological research based on the study of pottery remains is an inverse problem that needs to address, first of all, the material nature of ceramics. As stated throughout this
Designing Rigorous Research 43 chapter, ceramics—simple or complex—are concrete inanimate things artificially transmuted from the raw materials employed in their manufacture. As composite materials, ceramics are characterized by their main intrinsic and essential source properties of composition and structure, together with shape and size, as well as their main intrinsic essential derived properties. This material nature of ceramics must be evaluated and understood from the archaeometric point of view (i.e. chemical, mineralogical—and vitreous—or petrographic level). Archaeometry, as a scientific discipline, needs to solve archaeometric research problems in order to establish and develop essential background knowledge, methods, and theories to conduct archaeometric routine problems; facilitating the routine application of well- established methods and techniques to generate the necessary data to address archaeological research problems. Archaeometric analysis of ceramics is also an inverse problem: we start with the as-received-state pottery, the end product of weathering and formation processes, but all of our research questions relate to the states of life-finished pottery, unfired pottery, and raw materials, and the manufacture processes connecting the three of them. Archaeometric routine problems aim at identifying meaningful ceramic groups and provenance, and aspects related to ceramic manufacture. Production centers and workshops offer an ideal framework in which to recover the processes of ceramic manufacture, with the necessary implication for technical/technological change, and to define reference groups (and fabrics) that will function as localized references for provenance studies. Contrariwise, consumption centers are ideal for addressing intended performance characteristics and the interaction of vessels with things and people in the behavioral chain of use, including technical/technological change, as well as archaeological research problems based on composition of assemblages. Questions about the composition of archaeological assemblages cannot be answered without understanding the formation processes which created the archaeological record. We have argued that a representational approach is appropriate in order to model the empirical reality in a numerical relational system, thus providing the necessary link between the archaeometric research and the archaeological as received-state empirical reality. Context and assemblage are the basic units of the representational approach, mirroring the basic units of site formation processes, and providing a framework for assemblage compositions based on archaeological stratification and estimation of components; that is, frequencies of individuals. Archaeological stratification can be based on relational and accidental properties of ceramics, together with an estimation of intrinsic essential source properties of composition and structure, and intrinsic essential derived properties determined or hypothesized by macroscopic and stereoscope microscope observations. In the future, portable analytical facilities and instruments may enable preliminary archaeometric stratification, prior to laboratory analysis, that would improve the initial sampling frame for the archaeometric study. Fragmentation and integrity must be considered, and proxy representational measurements have been proposed based on the loss of information. Finally, archaeometric results enable us to infer the composition of the as-received-state assemblage, creating a new stratification (a post-stratification) and facilitating the study of diversity. While archaeometric research problems must continue to be solved to push forward this scientific discipline, the routine application of archaeometric studies on the thousands of ceramic individuals, needed to generate the necessary data and the knowledge relevant to archaeology, make its application routine. But these archaeometric routine problems are the
44 J. Buxeda i Garrigós and M. Madrid i Fernández problems that would become significant in archaeological research, provided that they are included in and designed for real archaeological research problems.
Acknowledgments The ideas of this work have been developed in the frame of the TECNOLONIAL research project (HAR2008-02834 and HAR2012-33784) funded by the Ministerio de Economía y Competitividad (Spain). We are very grateful to the editor of this volume, Alice M. W. Hunt, for her valuable comments and corrections that have helped to improve the original text.
Notes 1. The philosophical foundation of this chapter draws on Bunge’s thought. Several concepts and ideas of his work (see especially Bunge, 1996) are used all through the text and will not be acknowledged individually. 2. All materials are composed of atoms (chemical level). Those atoms are arranged (in an ordered or unordered way) in crystalline or vitreous structures (mineralogical/vitreous level). Those mineralogical and/or vitreous structures are then arranged in different ways (petrological level).Composition can then be addressed at these different but necessarily related levels by appropriate methods. 3. How the components of this composite material are arranged (e.g. the microstructure of the clayey matrix, sorting, packing). 4. For a concrete thing its state or change in the state. 5. Fabric is the end product of a paste after firing. As such, a paste can end in more than one different fabrics. It must be taken into consideration that fabric is also used as a synonym of paste in petrographic analysis, which can introduce some misunderstandings. 6. Pottery is usually recovered as sherds, not as complete vases. Moreover, even after refitting processes vases are usually still incomplete. 7. The study of this Roman town of Baetulo included many more contexts besides the two here presented as examples. The total number of individuals included in the sampling for archaeometric analysis was 225 (Madrid, 2005).
References Aitchison, J. (1986). The Statistical Analysis of Compositional Data (London: Chapman and Hall). Arcelin, P. and Tuffreau-Libre, M. (ed) (1998). La quantification des céramiques. Conditions et protocole, Collection Bibracte, 2 (Gluxen-Glenne: Centre archéologique Européen du Mont Beuvray). Béarat, H., Dufournier, D., Nguyen, N., and Raveau, B. (1989). “Influence de NaCl sur le coleur et la composition chimique des pâtes céramiques au cours de leur cuisson.” Revue d’Archéométrie 13: 43–53.
Designing Rigorous Research 45 Bishop, R. L. and Neff, H. (1989). “Compositional Data Analysis in Archaeology.” In: Allen, R. O. (ed), Archaeological Chemistry IV. Advances in Chemistry Series 220 (Washington, D.C.: American Chemical Society), 57–86. Bishop, R. L., Rands, R. L., and Holley, G. R. (1982). “Ceramic Compositional Analysis in Archaeological Perspective.” In: Schiffer, M. B. (ed), Advances in Archaeological Method and Theory, vol. 5 (New York: Academic Press), 275–330. Bunge, M. (1996). Finding Philosophy in Social Science (New Haven, CT: Yale University Press). Buxeda i Garrigós, J. (1999). “Alteration and Contamination of Archaeological Ceramics: The Perturbation Problem.” Journal of Archaeological Science 26: 295–313. Buxeda i Garrigós, J., Cau Ontiveros, M. A., and Kilikoglou, V. (2003). “Chemical Variability in Clays and Pottery from a Traditional Cooking Pot Production Village: Testing Assumptions in Pereruela.” Archaeometry 45: 1–17. Buxeda i Garrigós, J., Cau Ontiveros, M. A., Madrid i Fernández, M., and Toniolo, A. (2005). “Roman Amphorae from the Iulia Felix Shipwreck: Alteration and Provenance.” In: Hars, H. and Burke, E. (ed), Proceedings of the 33rd Internaitonal Symposium on Archaeometry, 22– 26 April 2002, Amsterdam. Geoarchaeological and Bioarchaeological Studies 3 (Amsterdam: Institute for Geo-and Bioarchaeology of the Vrije Universiteit in Amsterdam), 149–151. Buxeda i Garrigós, J., Madrid i Fernández, M., Morán Álvarez, M., Payà i Mercé, X., and Pérez Almoguera, A. (2013). “La terra sigillata d’Ilerda, caracterització arqueomètrica i estudi històric-arqueològic de la seva producció i de la seva relació amb les ceràmiques engalbades.” In: Roca Roumens, M., Madrid i Fernández, M., and Celis i Betriu, R. (eds), Contextos ceràmics d’època altoimperial en el Mediterrani occidental (Barcelona: Universitat de Barcelona). Cogswell, J. W., Neff, H., and Glascock, M. D. (1996). “The Effect of Firing Temperature on the Elemental Characterisation of Pottery.” Journal of Archaeological Science 23: 283–287. David, N. and Kramer, C. (2001). Ethnoarchaeology in Action (Cambridge: Cambridge University Press). Hand, D. J. (2004). Measurement Theory and Practice. The World through Quantification (London: Arnold). Harbottle, G. (1982). “Chemical Characterization in Archaeology.” In: Ericson, J. E. and Earle, T. K. (eds), Contexts for Prehistoric Exchange (New York: Academic Press), 13–51. Heck, K. L., Jr, van Belle, G., and Simberloff, D. (1975). “Explicit Calculation of the Rarefaction Diversity Measurement and the Determination of Sufficient Sample Size.” Ecology 56: 1459–1461. Heimann, R. B. (1982). “Firing Technologies and Their Possible Assesment by Modern Analytical Methods.” In: Olin, J. S. and Franklin, A.D. (eds), Archaeological Ceramics (Washington, D.C.: Smithsonian Institution Press), 89–96. Hein, A., Müller, N. S., Day, P. M., and Kilikoglou, V. (2008). “Thermal Conductivity of Archaeological Ceramics: The Effect of Inclusions, Porosity and Firing Temperature.” Thermochimica Acta 480: 35–42. Hunt, A. M. W. (2012). “On the Origin of Ceramics: Moving toward a Common Understanding of ‘Provenance.” Archaeological Review from Cambridge 27: 85–97. Hunt, A. M. W. and Speakman, R. J. (2015). “Portable XRF Analysis of Archaeological Sediments and Ceramics.” Journal of Archaeological Science 53: 626–638. Hurlbert, S. H. (1971). “The Nonconcept of Species Diversity: A Critique and Alternative Parameters.” Ecology 52: 577–586.
46 J. Buxeda i Garrigós and M. Madrid i Fernández Iñañez, J. G., Madrid i Fernández, M., Molera, J., Speakman, R. J., and Pradell, T. (2013). “Potters and Pigments: Preliminary Technological Assessment of Pigment Recipes of American Majolica by Synchrotron Radiation Micro- X- Ray Diffraction (Sr- µXRD).” Journal of Archaeological Science 40: 1408–1415. Kilikoglou, V., Maniatis, Y., and Grimanis, A. P. (1988). “The Effect of Purification and Firing of Clays on Trace Element Provenance Studies.’ Archaeometry 30: 37–46. Kilikoglou, V. and Vekinis, G. (2002). “Failure Prediction and Function Determination of Archaeological Pottery by Finite Element Analysis.” Journal of Archaeological Science 29: 1317–1325. Lyman, R. L. (2008). Quantitative Paleozoology (Cambridge: Cambridge University Press). Lyman, R. L. (2012). “A Historical Sketch on the Concepts of Archaeological Association, Context, and Provenience.” Journal of Archaeological Method and Theory 19: 207–240. Madrid i Fernández, M. (2005). “Estudi arqueològic i caracterització arqueomètrica de la Terra Sigillata de la ciutat de Baetulo (Badalona).” Universitat de Barcelona, Barcelona: (last accessed March 9, 2014). Madrid i Fernández, M. and Buxeda i Garrigós, J. (2007). “Qualitat i consum ceràmic de la sigillata augustal. Noves vies d’estudi i interpretació a partir de l’arqueometria.” Empúries 55: 53–66. Madrid i Fernández, M. and Buxeda i Garrigós, J. (2014). “Hispanic terra sigillata Productions Documented on the Catalan Coast: Some Unexpected Results and New Issues.” In: Martinón-Torres, M. (ed), Craft and Science: International Perspectives on Archaeological Ceramics. UCL Qatar Series in Archaeology and Cultural Heritage 1 (Doha: Bloomsbury Qatar Foundation), 101–108. Madrid i Fernández, M., Comas i Solà, M., and Padrós i Martí, P. (2005). “Étude archéologique et analise archéométrique de la céramique sigillée sud-gauloise documentée a la ville romaine de Baetulo (Badalone, Barcelone).” In: Nieto, X. Roca, M., Vernhet, A., and Sciau, P. (eds), La difusió de la terra sigillata sudgàl·lica al nord d’Hispania (Barcelona: Monografies del Museu d’Arqueologia de Catalunya), 179–197. Maggetti, M. (1982). “Phase Analysis and Its Significance for Technology and Origin.” In: Olin, J. S. and Franklin, A. D. (eds), Archaeological Ceramics (Washington D.C.: Smithsonian Institution Press), 121–133. Martín-Fernández, J. A., Buxeda i Garrigós, J., and Pawlowsky-Glahn, V. (2015). “Logratio Analysis in Archeometry: Principles and Methods.” In: Barceló, J. A. and Bogdanovic, I. (eds), Mathematics and Archaeology (Boca Raton, FL: Science Publishers), 178–189. Montana, G., Fabbri, B., Santoro, S., Gualtieri, S., Iliopoulos, I., Guiducci, G., and Mini, S. (2007). “Pantellerian Ware: A Comprehensive Archaeometric Review.” Archaeometry 49: 455–481. Neustupný, E. (1993). Archaeological Method (Cambridge: Cambridge University Press). Orton, C. R. (1993). “How Many Pots Make Five? An Historical Review of Pottery Quantification.” Archaeometry 35: 169–184. Orton, C. R. (2000). Sampling in Archaeology. Cambridge Manuals in Archaeology (Cambridge: Cambridge University Press). Orton, C. R. and Tyers, P. A. (1990). “Statistical Analysis of Ceramic Assemblages.” Archeologia e Calcolatori 1: 81–110. Picon, M. (1984). “Recherches sur les compositions des sigillées hispaniques: Techniques de fabrication et groupes de production. Appendice II.” In: F. Mayet (ed), Les céramiques sigillées hispaniques. Contribution à l’histoire économique de la Péninsule Ibérique sous l’Empire
Designing Rigorous Research 47 Romain, vol. 1. Publications du Centre Pierre Paris 12. Collections de la Maison des Pays Ibériques 21 (Paris: Diffusion de Boccard), 303–317. Picon, M. and Le Miere, M. (1987). “Géochimie.” In: Minskovsky, J.-C. (ed), Géologie de la Préhistoire: Méthodes, Téchniques, Applications. (Paris: A.E.E.G.P), 883–901. Raux, S. (1998). “Méthodes de quantification du mobilier céramiques. État de la question et pistes de réflexion.” In: Arcelin, P. and Tuffreau-Libre, M. (eds), La quantification des céramiques. Conditions et protocole. Collection Bibracte 2 (Gluxen-Glenne: Centre archéologique Européen du Mont Beuvray), 11–16. Roca Roumens, M., Buxeda i Garrigós, J., and Madrid i Fernández, M. (2013). “El centro productor de Terra Sigillata Hispánica altoimperial de Andújar. Nuevos datos y algunas puntualizaciones a partir del estudio de un lote de moldes.” In: Roca Roumens, M., Madrid i Fernández, M., and Celis i Betriu, R. (eds), Contextos ceràmics d’època altoimperial en el Mediterrani occidental (Barcelona: Universitat de Barcelona), 429–481. Rye, O. S. (1981). Pottery Technology. Principles and Reconstruction. Manuals on Archaeology 4 (Washinton, D.C.: Taraxacum). Schiffer, M. B. (1987). Formation Processes of the Archaeological Record (Salt Lake City: University of Utah Press). Schiffer, M. B. (2011). Studying Technological Change: A Behavioral Approach (Foundations of Archaeological Inquiry) (Salt Lake City: University of Utah Press). Secco, M., Maritan, L., Mazzoli, C., Lampronti, G. I., Zorzi, F., Nodari, L., Russo, U., and Mattioli, S. P. (2011). “Alteration Processes of Pottery in Lagoon-Like Environments.” Archaeometry 53: 809–829. Shannon, C. E. (1948). “A Mathematical Theory of Communication.” The Bell System Technical Journal 27: 379–423, 623–656. Skibo, J. M. and Schiffer, M. B. (2008). People and Things. A Behavioral Approach to Material Culture (New York: Springer). Whitbread, I. K. (1995). Greek Transport Amphorae. A Petrological and Archaeological Study. Fitch Laboratory Occasional Paper 4 (London: The British School at Athens).
Chapter 4
Evaluating Data Uncertainty in Ceramic Analysis
Roberto Hazenfratz-M arks Introduction The uncertainty associated with a measurement can be defined as a parameter that characterizes the dispersal of the values attributed to a measurand (JCGM, 2012). Assuming the instrument is accurate and correctly calibrated, the uncertainty can be determined by the standard deviation of a group of measurements of samples. Uncertainty affects the way the repetition of the analyses leads to the same results, contributing significantly to the elemental variation observed in the context of ceramic analyses. If all sources of variation in analytical determinations could be reduced to a minimum, there would still be a non-reducible amount of error related to fluctuations associated with the measurement process (Bishop et al., 1990). Evaluating uncertainty in experimental results is an indispensable task because uncertainty can affect the statistical treatment and interpretation of data. The nature of the analytical data obtained in ceramic analysis can be as simple as the weight of pottery sherds or integral pieces, the determination of mineral phases, the concentration of a single chemical element, or, more commonly, the concentrations of many elements that need to be analyzed together (Baxter and Buck, 2000). This chapter focuses on sources of uncertainty in the chemical characterization of archaeological ceramics. Uncertainty, such as analytical error in the chemical analysis of archaeological ceramics, can have a drastic impact on the interpretation of a dataset. For example, in the simplest case in which the concentration of just one chemical element per specimen is of interest, uncertainty may affect direct comparison of that variable among samples. A high degree of uncertainty in the data may also hide patterns if, for example, the researcher plots a histogram in order to verify the existence of a uni-or multimodal distribution to assess clustering of samples regarding the element analyzed (Baxter and Buck, 2000). When two chemical elements are of interest, a bivariate case, uncertainties unrelated to the natural variability of raw materials could weaken or mask existing correlation between them, which could also mislead the archaeological interpretation. Finally, in the most frequent multidimensional case, where three or more chemical elements are of analytical interest, a high degree of uncertainty may
Evaluating Data 49 affect not only the low-precision elements but also the general correlation structure, which is an important feature in multivariate statistical analysis (Glascock, 2000). Ideally, uncertainty evaluation in ceramic analysis needs to address errors from all possible systematic or random sources, from those introduced during ceramic manufacture up to and including those which result from instrument error or analytical measurement in the laboratory. Frequently, adequate information about the parameters which could aid in the evaluation of analytical and/or instrumental error is not reported in archaeometric publications (Bishop et al., 1990). Possible reasons for this include the fact that experimental uncertainties are not used directly in many multivariate statistical analyses and the accurate numerical evaluation of all possible sources of errors is not a straightforward task; frequently, only the error derived from the counting statistics is considered for single measurements. A dataset with a high degree of uncertainty not derived from natural variability and cultural behavior may mislead the researcher in answering questions regarding “provenance,” dating, different pottery workshops, or choice of raw materials. The main purpose of analyzing uncertainty in “provenance” studies is to evaluate and ameliorate its effects on the statistical analysis of a chemical dataset (Harbottle, 1982). These effects also potentially influence the priorities of future researchers, whose aims and questions will be based on the initial misinterpretation of the data. In integrated programs of ceramic analysis, adopted increasingly since the 1980s and comprising the combination of different analytical techniques (Buxeda i Garrigós et al., 2001), assessment of uncertainty is crucial because it can lead to resolving ambiguous relationships among data obtained by different analytical techniques and prevent incorrect archaeological interpretation of the data.
The General Expression of Uncertainty Uncertainty in chemical compositional data derived from archaeological ceramics can be evaluated using the following equation, first proposed by Harbottle (1976):
ST2 = SN2 + SC2 + SS2 + SA2
where ST2 is the total uncertainty, SN2 is the natural uncertainty, SC2 is the uncertainty related to cultural aspects of ceramic manufacture, SP2 is the uncertainty related to post-depositional effects and alteration, SS2 is the sampling uncertainty, and SA2 is the uncertainty introduced by the analytical method of analysis. This equation attempts to account for the natural, cultural, and analytical sources of variability or uncertainties associated with a particular archaeologically meaningful chemical group of ceramic samples (Bishop, 2003). It is important to remember that this equation is mainly an intellectual exercise for addressing possible sources of uncertainty. It may be difficult and impractical to determine precisely all the terms in the general equation for a chemical group. However, the general expression of uncertainty can and should be used as a guide for identifying potential sources of uncertainty in archaeological ceramic analysis. Frequently, archaeometric studies are guided by archaeological questions. For example, the researcher may be interested in understanding ceramic manufacture behaviors and,
50 Roberto Hazenfratz-Marks therefore, will require the chemical data to reflect patterns regarding cultural choices, such as selection, mixture, and modification of raw materials, and minimize variation related to post-depositional phenomena (SP2 ), the sampling design (SS2 ), and the analytical procedure employed (SA2 ). In chemical characterization for provenance studies, the focus is the determination of elements measured at trace levels (in micrograms per gram, or even lower), such as rare earth elements (REEs) and transition metals. Owing to the low concentration of these elements in a ceramic matrix, minor contaminations and depletions can critically affect their detection, measurement, and determination, and significantly increase their uncertainty. To give a general idea of the typical total uncertainty values for chemical groups of archaeological ceramics, researchers at Brookhaven found coefficients of variation ( sT2 / mean) in the range of 4–9% for elements such as Sc, La, Ce, Hf, Th, and Fe, in an assemblage of 63 late Bronze Age pottery sherds from Greece and 10–20% in ceramics from Mesoamerica (Bieber et al., 1976). Uncertainty or variance in ceramic chemical groups can be related to the geochemical origin of the raw materials and/or pottery manufacture behaviors, such as the addition of tempering materials (Harbottle, 1976).
Natural and Cultural Variance (S and S ) 2 N
2 C
For any given element in a ceramic matrix, there is expected variation or uncertainty related to geological, geochemical, and weathering processes affecting the raw materials used in pottery manufacture, SN2 (Bishop et al., 1982). Depending upon the region studied, the range of chemical compositional variation may be inferred from the geochemical literature. In other cases, prior chemical analysis of the regional argillaceous sediments could provide insight into the elemental occurrence and variation in available raw materials, although these studies would be unlikely to determine the total range of this variation accurately (Bishop, 2003). In “provenance” studies of ceramics it is important that there is sufficient natural elemental variation in the raw materials or ceramic pastes to form discrete groups during statistical analyses; that is, the between-source variation exceeding the within-source variation, known as the provenance postulate (Weigand et al., 1977). The elemental variation introduced during ceramic manufacture is regarded as the cultural imprint of the artifact, SC2 (IAEA, 2003). One of the cultural imprints of potential interest is the addition of non-plastic inclusions that do not naturally occur in the raw materials. The use of temper by potters can vary significantly, potentially diluting important trace elements and enhancing others. Even if the purpose of the analysis is not to investigate technological choices regarding the use of temper, the uncertainty introduced by tempers should be considered anyway because of its effect in the elemental concentration data, or even in the mineralogical analyses (Summerhayes, 1997). The simplest solution would be to assess the elements most affected by tempering and exclude them in the multivariate analysis. There are statistical procedures developed to consider shifts in the original concentrations, such as the best relative fit (Harbottle, 1976), the modified Mahalanobis filter, which considers the precision of measurements (Beier and Mommsen, 1994), and concentration ratios (Buxeda i Garrigós, 1999; Dias and Prudêncio, 2008). These procedures facilitate the differentiation of chemical patterns which are not strongly different and improve the geographical resolution for
Evaluating Data 51 pottery “provenance” determination (Mommsen and Sjöberg, 2007). Grog, crushed pottery sherds used as temper, may further attenuate the distinctiveness among chemical groups (Neff et al., 1989). Even for well-established paste recipes, compositional differences occur between ceramic objects and batches of objects manufactured in the same workshop, depending on the control of ceramic production parameters. These variations, related to the mixing of different clays in varying proportions, pre-treatment of a clay paste in order to select particle sizes of interest, the influence of firing on the chemical profiles, and so on, may increase SC2 . These behavioral choices are often conditioned by geographic distance to suitable natural resources and functional, social, religious, and economic requirements (Bishop, 2003).
Post-Depositional Variance (S ) 2 P
It is important to assess possible changes in composition and microstructure during burial, prior to data interpretation in ceramic analysis (Tite, 2009). Alteration may occur as leaching or enrichment of certain elements, the most susceptible of which are alkali and earth alkali metals, such as Na, K, Ca, Rb, and Cs, although other elements may also be altered. Many of these chemical and mineralogical alterations occur as a result of in-burial interaction of ceramic materials with rainfall and groundwater and depend on the technological aspects of ceramic production, the type of raw material used, firing temperature, and atmosphere, as well as conditions of the burial context, such as pH, proximity to water sources, and rain regimes. Each of these factors can lead to alteration in the elemental profiles of archaeological ceramics, increasing compositional variability, chemical and mineralogical, as a result of the formation of new minerals and destruction and alteration of others. Chemical and mineralogical alteration can affect the formation of chemical reference groups expected to reflect the chemical signature of an area or pottery workshop (Buxeda i Garrigós et al., 2001; Schwedt et al., 2004).
Sampling Variance (S ) 2 P
Sampling is one aspect of research design that allows the investigator control over the representativeness and size of the population selected for analysis, in order to adequately address a specific archaeological question. The geographic scale of the research, from investigation of a single site to a regional survey or interregional comparison, will influence sampling strategy and population size for adequate characterization of ceramic groups (Bishop, 2003). The ideal sample population is hardly ever analyzed in archaeometric studies. Most frequently, samples are taken opportunistically from museum collections or active excavations, which may vary in their representativeness of time and space of the ancient people under study (Bishop, 2012). The reason for this is not necessarily the lack of knowledge about the best strategy, but that sampling is usually restricted by practical aspects of the research design, such as which specimens archaeologists and museums make available for chemical and mineralogical analysis, budget, and cost of analyses.
52 Roberto Hazenfratz-Marks In terms of uncertainty, inadequate sample populations lead to badly characterized chemical groups, or even unrealistic ones. However, determining the adequate size of a sample population is not a straightforward task since little or no knowledge about the samples, in terms of elemental and mineralogical composition, may be known a priori (Baxter and Buck, 2000). Nevertheless, the researcher should be aware of this fact and assess how it can affect uncertainty in establishing chemical reference groups or other aspects of analysis. Even in the case of a biased sampling, useful analytical results for archaeological interpretation may be generated (Bishop et al., 1988). The dangers of forming chemical groups based on the analysis of an inadequate sample population are most critical for coarse-ware ceramics. The poor mixing and heterogeneity of raw materials could result in the poor distribution of inclusions in the ceramic body, yielding different chemical compositions for different parts of the artifact. In such cases, and if the vessel cannot be adequately sampled, the analyst may obtain results that are not representative of the whole ceramic object (Baxter and Buck, 2000), increasing the uncertainty of data. To avoid this problem, a microstructure or petrographic analysis prior to any chemical analysis is helpful in order to assess the quality of mixing of the raw materials. The combined error regarding sampling from a ceramic body and other analytical uncertainties can be determined by reproducibility experiments, combining measurements of aliquots from the same ceramic object (Bishop et al., 1980).
Analytical Variance (S ) 2 A
Variation and uncertainty related to analytical method is the most controllable source of uncertainty in archaeological ceramic analysis. For the archaeologist, it is important to obtain high-quality data, which, in terms of accuracy, involves the minimization of analytical uncertainties that could obscure interesting patterns that answer the archaeological questions formulated. Analytical quality control can be used to identify possible analytical inadequacies, which can be corrected for or used to eliminate a chemical element from consideration (Bode and Dijk, 1997). Frequently, owing to restriction in the availability and/or destructive sampling of ceramic material it is not possible to analyze samples in duplicate or triplicate. Therefore, careful analytical quality control using certified reference materials (CRMs) must be maintained to minimize analytical uncertainty. The uncertainty introduced by the analytical method should be lower than the natural and cultural uncertainty in the sample population (Harbottle, 1976). Chemical characterization of ceramics in most cases involves trace element determination; therefore, limit of detection (LOD) and limit of quantification (LOQ) of an instrument and/or calibration influence its ability to accurately measure these elements and, ultimately, the differentiation of ceramic pastes with only minor chemical differences. When preparing samples for analysis, it is important to remove residual material, such as soil and organic matter, in order to avoid adding chemical “noise” to the data (Baxter and Buck, 2000). For example, the removal of outer surfaces from shreds before sample preparation reduces the chance of contamination (IAEA, 2003). In activation analysis studies, the analytical uncertainties may include errors due to contamination in the aliquot extraction from a ceramic body, inadequate sample
Evaluating Data 53 homogenization, weighing errors, impurities in the irradiation vials, the suitability and homogeneity of reference materials, particle flux variations, absolute quantity of the analyte, counting statistics, gamma spectrum processing, sensitivity, selectivity, spectral interferences, decay and background corrections, and counting geometry (Harbottle, 1976; Bishop et al., 1990; Bode and Dijk, 1997; Bishop, 2003). Prior to any application of statistical techniques, the first stage in data analysis should be a visual inspection of columns in a dataset, representing different chemical elements determined, in order to identify obvious discrepancies or errors. This preliminary evaluation helps reduce analytical uncertainty, since these outliers would increase the dispersion parameters of the samples and reduce the homogeneity of chemical groups. Concentrations below the detection limit of the analytical technique for a specific chemical element should be considered missing values. They may represent a kind of uncertainty if an imputation method is used. On the other hand, by excluding samples to eliminate such a problem, one may lose information about the data structure, as that type of missing data appear in a non- random fashion. If the quantity of missing values is above critical limits employed, such as 10% or 20% of the number of samples, it would be advisable to exclude the element (Baxter and Buck, 2000). One common approach for determining the term SA2 and controlling the analytical uncertainty is the analysis standard or certified reference materials (SRM/CRM) simultaneously with the unknown ceramic samples. Reference materials are designed to have no natural variability or sampling error, and the concentration of the chemical elements of interest in the matrix is known from certificates. Furthermore, poor analytical precision can prevent the comparison among datasets (Bishop et al., 1990). Tools as control charts for identification of non-conformances, z-scores, and u-scores are common approaches for the analysis of analytical results of standard reference materials (Bode and van Dijk, 1997). One means of quantifying sources of errors other than the counting statistics is by subtracting the latter out from the coefficient of variation of samples of a reference material (Blackman and Bishop, 2007). Finally, some caution is advised in physicochemical characterization studies of archaeological ceramic materials. While analytical instruments and calibrations may provide measurements statistically significant for modern engineering applications, where the ceramic production parameters are more controlled, these measurements may not be statistically significant for the cultural context under study, the identification of differences in manufacture behavior, or evaluating relevant socioeconomic aspects of pottery production (Rice, 1987: 327).
Example As mentioned above, it may be difficult to calculate precisely each source of variability in the general equation of uncertainty. In practice, it is easier to use the equation as an instrument to evaluate possible sources of variation in a specific archaeometric study. However, in some cases, it is possible to estimate values for each source of uncertainty and have a quantitative idea of the overall reliability of the data.
54 Roberto Hazenfratz-Marks Table 4.1 Uncertainty values calculated for a ceramic chemical group from the Central Amazon Element
sT (%)
sP2 (%)
s S2
s A2 sT2 (%)
(s
2 N
)
sP2 sT2 (%)
+ sC2 sT2 (%)
Some elements of interest La
11
≈ 0
≈ 0
10
90
≈ 0
Lu
12
≈0
≈0
41
59
≈0
Yb
11
≈0
≈0
33
67
≈0
Cr
12
≈0
≈0
25
75
≈0
Eu
13
≈0
≈0
50
50
≈0
Fe
14
≈0
≈0
7
93
≈0
Sc
11
≈0
≈0
12
88
≈0
Th
11
≈0
≈0
32
68
≈0
Altered elements Na
67
≈0
≈0
1
35*
64
K
70
≈0
≈0
4
51*
45
Cs
50
≈0
≈0
4
36*
60
*Estimated by the dispersion of elemental concentrations in clay sources.
In the following example uncertainty is calculated for the chemical data, obtained by neutron activation analysis, of pottery from a large archaeological site in Central Amazon, Brazil, located in a geochemically and hydrologically dynamic area. Two chemical groups were determined using multivariate statistics; however, the following discussion focuses on the uncertainty in a single group. Analytical uncertainty was estimated for some of the elements of interest for which there was good analytical quality control using standard reference materials. In order to illustrate the effect of alteration on elemental concentration, the post-depositional uncertainty was calculated for three elements, Na, K, and Cs. These three elements are presented only for illustrative purposes and were not used in the definition of chemical groups. It is also important to note that the elements in Table 4.1 are only used to illustrate uncertainty in geochemical data for this particular sample and are not necessarily ideal elements for group determination in general; diagnostic elements vary according to the region, type of ceramic, and analytical instrument. In Table 4.1, the second column, ST , corresponds to the coefficient of variation calculated for this chemical group, which varies from 11% to 14% for the elements of interest. Uncertainty related to post-depositional alteration (third column) and sampling (fourth column) were not calculated. In the latter case, this is because the sample population derives from the most representative units of the excavation and different stratigraphic levels within these units, making it as representative of the ceramic archaeological record of the site as possible, reducing SS2 to zero. The proportion of total uncertainty represented by
Evaluating Data 55 the analytical uncertainty, (SA2 / ST2 ), is presented in the fifth column; analytical uncertainty was estimated by the coefficients of variation calculated for a standard reference material analyzed together with the unknown samples. In our test case, the proportion of uncertainty related to the analytical method can be as low as 7% for iron or as high as 50% of the total elemental variance for europium. The sixth column presents the amount of total uncertainty related to the combined natural and cultural uncertainty, (SN2 + SC2 ) / ST2 , which can be difficult to separate. For the elements of interest, this term was calculated by substitution of the other determined terms in the general equation; and for Na, K, and Cs it was approximated by the coefficients of variation for some suitable clay sources analyzed. Finally, the seventh column, SP2 / ST2 , represents the proportion of variance related to post- depositional effect; assumed to be zero for the elements used in the definition of chemical groups but for Na, K, and Cs this term was determined by substitution of the other estimated terms in the general equation. A comment about Na, K, and Cs elements that were affected by post-depositional phenomena according to previous studies: the analytical uncertainty for these elements is relatively low (maximum of 4% for K and Cs), but the proportion of post-depositional uncertainty is high, more than 50% for Na and Cs. Chemical groups differentiated on the basis of Cs in this region, therefore, could potentially be based on post-depositional alteration rather than raw material sources.
Concluding Remarks This chapter demonstrates the importance of evaluating all possible sources of uncertainties in ceramic analysis, as they can affect the statistical treatment and the archaeological interpretation of data. Even in the cases where it is not possible to quantify precisely all sources of variation in a dataset, it is important to consider their potential existence, so that the investigator can make the appropriate analytical and statistical choices to yield optimal data that reflect mainly natural and cultural aspects of archaeological interest.
References Baxter, M. J. and Buck, C. E. (2000). “Data Handling and Statistical Analysis.” In: Ciliberto, E. and Spoto, G. (eds), Modern Analytical Methods in Art and Archaeology. Chemical Analysis Series 155 (New York: John Wiley & Sons), 681–746. Beier, T. and Mommsen, H. (1994). “Modified Mahalanobis Filters for Grouping Pottery by Chemical Composition.” Archaeometry 36(2): 287–306. Bieber, A. M., Jr, Brooks, D. W., Harbottle, G., and Sayre, E. V. (1976). “Application of Multivariate Techniques to Analytical Data on Aegean Ceramics.” Archaeometry 18(1): 59–74. Bishop, R. L. (2003). “Instrumental Neutron Activation Analysis of Archaeological Ceramics: Progress and Challenges.” In: IAEA-TRS416—Nuclear Analytical Techniques in Archaeological Investigations (Vienna: IAEA), 35–44. Bishop, R. L. (2012). “Sources and Sourcing.” In: Nichols, D. L. and Pool, C. A. (eds), The Oxford Handbook of Mesoamerican Archaeology (USA: Oxford University Press), 579–587.
56 Roberto Hazenfratz-Marks Bishop, R. L., Canouts, V., Crown, P. L., and De Atley, S. P. (1990). “Sensitivity, Precision, and Accuracy: Their Roles in Ceramic Compositional Data Bases.” American Antiquity 55(3): 537–546. Bishop, R. L., Canouts, V., De Atley, S. P., Qöyawayma, A., and Aikins, C. W. (1988). “The Formation of Ceramic Analytical Groups: Hopi Pottery Production and Exchange, A.C. 1300–1600.” Journal of Field Archaeology 15(3): 317–337. Bishop, R. L., Harbottle, G., and Sayre, E. V. (1982). “Chemical and Mathematical Procedures Employed in the Mayan Fine Paste Ceramics Project.” In: Sabloff, J. A. (ed), Excavations at Saibal—Analyses of Fine Paste Ceramics, BNL-28105. doi: 10.2172/5258886. Blackman, M. J. and Bishop, R. L. (2007). “The Smithsonian- NIST Partnership: The Application of Instrumental Neutron Activation Analysis to Archaeology.” Archaeometry 49(2): 321–341. Bode, P. and van Dijk, C. P. (1997). “Operational Management of Results in INAA Utilizing a Versatile System of Control Charts.” Journal of Radioanalytical and Nuclear Chemistry 215(1): 87–94. Buxeda i Garrigós, J. (1999). “Alteration and Contamination of Archaeological Ceramics: The Perturbation Problem.” Journal of Archaeological Science 26: 295– 313. doi: 10.1006/ jasc.1998.0390. Buxeda i Garrigós, J., Kilikoglou, V., and Day, P. M. (2001). “Chemical and Mineralogical Alteration of Ceramics from a Late Bronze Age Kiln at Kommos, Crete: The Effect on the Formation of a Reference Group.” Archaeometry 43(3): 349–371. Dias, M. I. and Prudêncio, M. I. (2008). “On the Importance of Using Scandium to Normalize Geochemical Data Preceding Multivariate Analyses Applied to Archaeometric Pottery Studies.” Microchemical Journal 88: 136–141. Glascock, M. D. (2000). “The Status of Activation Analysis in Archaeology and Geochemistry.” Journal of Radioanalytical and Nuclear Chemistry 244(3): 537–541. Harbottle, G. (1976). “Activation Analysis in Archaeology.” In: Newton, J. W. A. (ed), Radiochemistry: A Specialist Periodical Report, vol. 3 (London: The Chemical Society), 33–72. Harbottle, G. (1982). “Provenance Studies Using Neutron Activation Analysis: The Role of Standardization.” In: Olin, J. S. and Franklin, A. D. (eds), Archaeological Ceramics (Washington, D.C.: Smithsonian Institution Press), 67–77. IAEA (2003). TRS416—Nuclear Analytical Techniques in Archaeological Investigations (Vienna: IAEA). JCGM (2012). International Vocabulary of Metrology— Basic and General Concepts and Associated Terms (VIM). Third Edition. JCGM 200:2012. Mommsen, H. and Sjöberg, B. L. (2007). “The Importance of the “Best Relative Fit Factor” When Evaluating Elemental Concentration Data of Pottery Demonstrated with Mycenaean Sherds from Sinda, Cyprus.” Archaeometry 49(2): 359–371. Neff, H., Bishop, R. L., and Sayre, E. V. (1989). “More Observations on the Problem of Tempering in Compositional Studies of Archaeological Ceramics.” Journal of Archaeological Science 16: 57–69. Rice, P. M. (1987). Pottery Analysis: A Sourcebook (Chicago: University of Chicago Press). Schwedt, A., Mommsen, H., and Zacharias, N. (2004). “Post- Depositional Elemental Alterations in Pottery: Neutron Activation Analyses of Surface and Core Samples.” Archaeometry 46(1): 85–101. Summerhayes, G. R. (1997). “Losing Your Temper: The Effect of Mineral Inclusions on Pottery Analyses”. Archaeology Oceania 32: 108–117.
Evaluating Data 57 Tite, M. S. (2009). “Mastering Materials.” In: Cunliffe, B., Gosden, C., and Joyce, R. A. (eds), Oxford Handbook of Archaeology (New York: Oxford University Press). doi: 10.1093/ oxfordhb/9780199271016.001.0001. Weigand, P. C., Harbottle, G., and Sayre, E. V. (1977). “Turquoise Sources and Source Analysis: Mesoamerica and the Southwestern U.S.A.” In: Earle T. K. and Ericson, J. E. (eds), Exchange Systems in Prehistory (New York: Academic Press), 15–34.
Chapter 5
Statistical Mode l i ng f or Ceram ic A na lysi s Gulsebnem Bishop Introduction Analysis of archaeological ceramics is one of the most important and complex tasks performed by archaeologists. Information about manufacture, function, and date can be inferred from the shape, decoration, and fabric of archaeological ceramics, in addition to providing valuable information about technological and aesthetic changes over time. Ceramic analyses, whether morphometric, typological, or geochemical, generate large, complex datasets that need processing and interpreting before they can be used to answer archaeological questions. Statistical analysis is profoundly useful in processing large quantities of data to reveal overall trends and commonalities. In this chapter, we focus on statistical analysis of compositional and formal ceramic data. Before we go into the relevant statistical models in detail it is worth spending a little time defining “statistical analyses.” In the broadest sense, statistics is a discipline that deals with making sense of data. Data is factual information used to describe objects, processes, behaviors, and observations. As such, data exists in many formats. It can be qualitative or quantitative, numerical or non- numerical, discrete or continuous; however, only numerical data can be analyzed statistically. Qualitative or categorical data deals with descriptions or qualities; they can be observed but not measured. The texture and color of ceramic sherds are an example of qualitative data; they can be observed and described but not measured. Qualitative data can, however, be numeric. Nominal qualitative data is data in which a numeric code is used as a label for the quality or observation, such as the use of the Munsell Soil-Color Chart numbers to represent the color of ceramics. Qualitative data can also be ordinal; observations or qualities can be ranked or rated. Quantitative data, on the other hand, deals with quantity and can be measured. Measurement, from a statistical point of view, simply involves assigning numbers to objects or events. Therefore, measured data or quantitative data is always numerical or expressed as a number. Numerical data can be discrete or continuous. Discrete numerical data is used to describe objects and events that can be counted as distinct and separate; they are assigned
Statistical Modeling for Ceramic Analysis 59 whole or discrete numbers, such as the number of students in a class or the values generated when rolling a set of dice. Continuous numerical data is used to describe observations or qualities of an object or event that are measured on a continuum of values, such as height, weight, and length. Data is powerful. It is the raw material for all inference and interpretation. However, it is the understanding and analyzing of that data that provides knowledge. The first step in statistical analysis is to identify the problem or question that needs to be answered and gather data accordingly. Next, the data needs to be described. During this process, appropriate statistical models and analyses are used to identify patterns in the dataset. Finally, the statistics generated are used to make inferences about the data, predict outcomes, and/or answer the research question.
Statistical Models Statistical models are a mathematical representation of the data and/or its outcome. They help us organize and examine data objectively by describing the data visually, enabling us to see, literally, latent patterns and/or structures. The fundamental differences between qualitative and quantitative data require them to be described differently and will be discussed separately.
Describing Qualitative Data Qualitative data typically need to be converted into nominal data or categories which can be counted before they can be described. For example, vessel types, such as vase, jar, and bowl, are qualitative but the number of objects assigned to each category can be counted and, therefore, described statistically. The most effective ways of organizing and describing qualitative data are frequency tables, bar charts, and pie charts.
Frequency Tables One of the best ways to organize qualitative data is as a frequency table or frequency distribution list. In statistics, frequency refers to the number of times an event occurred in an experiment or study. A frequency table or frequency distribution list, therefore, is a simple table that records or describes how often an event, value, or category of interest occurred. Most often in archaeological ceramic analysis, frequency of a particular observation is only meaningful relative to the total number of observations or the relative frequency (number of events in a category/total number of events). In the small assemblage described in Table 5.1, the qualitative or categorical data, vessel type, has been used to organize the assemblage, which is then described by frequency and relative frequency. Frequency tables describe the distribution of observations or events in a study population. As such, they can be used to compare populations or particular categories/observations within those populations. In archaeological ceramic analysis, frequency distribution tables
60 Gulsebnem Bishop Table 5.1 Frequency distribution table describing a small assemblage of Greek pottery Vessel Type
Frequency
Relative Frequency
6
32%
Kylix
10
53%
Pyxis
3
16%
Amphora
are often the first step in making sense of a ceramic assemblage and can be used to determine activity area, by describing the relative frequency of different types of vessels in different parts of the site, or changes in occupation of a site over time, by describing the relative frequency of ceramic types in different strata.
Bar Charts Data described as a frequency distribution table can also be described as a bar chart or pie chart. Bar charts are one of the most common visual representations of data. Bar charts use rectangular bars, which can be vertical or horizontal, with lengths proportional to the values they represent to visually describe frequency and distribution of a dataset. For example, using the data described in Table 5.1, we can plot the frequency distribution of vessel types (Figure 5.1a) and the relative frequency of those types in the assemblage (Figure 5.1b). Bar charts, like frequency tables, can be used to describe the overall distribution of ceramics or ceramic types at an archaeological site or context, identify activity areas, and/or compare consumption of ceramic types across archaeological sites.
Pie Charts Pie charts are used to visually describe the relative frequency distribution of categorical data. The chart is a circle, representing the total number of observations in a study or test population, divided into segments representing the categories of data in the study population. The area of each segment is proportional to the number of observations in that category, relative to the whole population or its relative frequency. Therefore, pie charts always describe data as a percentage distribution and cannot be used to describe frequency counts. This makes pie charts particularly useful for comparing datasets with unequal population sizes; for example, the types of vessels recovered at two different sites or from two different contexts at the same site. In Figure 5.2, a pie chart is used to describe the relative frequency distribution of vessel types used in funerary assemblages in two contemporary Greek cities, Athens and Sparta. While this data could also be described using a frequency distribution table, the pie chart provides a simplified, visual representation of the data. We can easily identify that phiale were included in burials at both cities in similar proportions; however, lekythoi were favored in Athens while loutrophoroi were preferred in Sparta. Although this pattern would also be
Statistical Modeling for Ceramic Analysis 61 Frequency Distribution
(a)
Number of vessels
12 10 8 6 4 2 0 Amphora
Kylix
Pyxis
Relative Frequency Distribution
(b) Pyxis Kylix Amphora 0
0.1
0.2 0.3 0.4 Percentage of the total assemblage
0.5
0.6
Figure 5.1 Bar charts describing (a) the distribution and (b) relative frequency of vessel types in a ceramic assemblage. (Figure courtesy of A. M. W. Hunt)
Relative distribution of funerary vessels from Athens and Sparta Athens
Sparta 16%
5%
25%
Kernos 20%
15%
Lekythos Loutrophoros Phiale
20% 54%
45%
Figure 5.2 Pie charts describing the relative distribution of vessel types in funerary assemblages from Athens and Sparta. (Figure courtesy of A. M. W. Hunt)
present in the frequency distribution table, it is possible that it would be obscured by the frequency counts if, for example, the number of lekythoi in Sparta was bigger than the number in Athens. These assemblages can only be compared when their relative size is taken into consideration and described as relative frequencies. The strengths of pie charts include their ability to describe and compare datasets of different sizes and describe the overall composition of a population. The biggest weakness of
62 Gulsebnem Bishop pie charts is that only a few categories of data can be described at a time before the segments become too small to be useful. Similarly, pie charts can also only be used to compare datasets with the same categories of data.
Describing Quantitative Data Quantitative data are measurements that are already expressed numerically. However, continuous data, measurements which may take on any value within a range or interval, often need to be grouped into discrete categories before they can be described statistically (see Histograms). The simplest way to describe quantitative data is as a table of values or measurements. Even though these tables can be very insightful, they can be difficult to interpret, particularly if the dataset is large. Therefore, quantitative data are typically organized as tables and then described as histograms and bar charts, stem and leaf plots, box and whisker plots, and scatter plots.
Histograms and Bar Charts Histograms and bar charts are visually similar; both use rectangular bars to represent the frequency of an observation. However, histograms are used to describe continuous data and bar charts are used to describe discrete data. Discrete data are typically already categorical; they are counts or scores of observations. In fact, when qualitative data is organized into categories and quantified, it is converted into discrete data. The distribution of vessel types in an assemblage is an example of a discrete dataset. Continuous data are measurements which exit on a continuum, such as length or height. These data must be converted into discrete categories or groups before they can be described by a histogram. Typically, the categories used to describe continuous data as a histogram are intervals or ranges of measurement. All the observations contained by that interval are “counted” as part of that category. The capacity of wine amphorae recovered from a hypothetical shipwreck is an example of a continuous dataset that can be described using a histogram. Roman amphorae held approximately 27.8 liters and, according to Matheson and Wallace (1982), Rhodian amphorae held approximately 25.5 liters in 300 bc, 26.5 liters in 230 bc, and 24.4 liters in 200 bc. Understanding the distribution of amphora capacities could, therefore, tell us (a) whether the amphorae were from Rhodes and (b) how old the shipwreck was. By creating capacity categories or groups, from 23–23.9 liters, 24–24.9 liters, and so on, we can organize and describe the data graphically as a histogram (Figure 5.3). In this fictitious example, the histogram shows a clear trend in amphora capacity; the majority of the vessels hold between 25 and 25.9 liters. Therefore, the amphorae are probably from Rhodes and the hypothetical shipwreck dates to around 300 bc. It is also an example of a unimodal distribution. An important aspect of describing data using histograms and bar charts is its distribution behavior or the shape of the data’s distribution. Distribution behavior becomes extremely important, particularly in inferential statistics where results or outcomes are predicted from
Statistical Modeling for Ceramic Analysis 63 Amphora Capacity Measurements 14
Number of amphora
12 10 8 6 4 2 0
23–23.9
24–24.9
25–25.9 26–26.9 Capacity measurement (L)
27–27.9
28–28.9
Figure 5.3 Histogram of amphora capacity measurements from a hypothetical shipwreck. (Figure courtesy of A. M. W. Hunt)
Uniform
Skewed right
Symmetric
Skewed left
Figure 5.4 Common shapes of data distribution. (Figure courtesy of A. M. W. Hunt)
the distribution behavior of the data. Data distribution can be either unimodal, with one clear trend or mode visible in the data, or multimodal, two or more trends being visible in the data, and typically conform to one of four basic shapes, uniform, symmetric or normal, skewed left, and skewed right (Figure 5.4). The histogram in Figure 5.3 is an example of a symmetric, unimodal distribution. The most common multimodal distribution behavior, and an important one in archaeological ceramic analysis, is a bimodal distribution, where two trends are clearly visible in the data. In Figure 5.5, the size of mineral inclusions in a ceramic fabric is plotted as a histogram. The bimodal distribution of the inclusions suggests the addition of a mineral temper, in this case crushed limestone.
64 Gulsebnem Bishop Mineral Inclusion Distribution Number of mineral inclusions
25 20 15 10 5 0 –1––2
0– –1
1–0
2–1 φ size
3–2
4–3
8–4
Figure 5.5 Bimodal distribution of mineral inclusions in a ceramic fabric. (Figure courtesy of A. M. W. Hunt)
Stem and Leaf Plots Stem and leaf plots or stemplots provide another way of representing quantitative data graphically in order to visualize the shape of a distribution. Unlike histograms, stem and leaf plots retain the original numerical data to at least two significant units. Therefore, stem and leaf plots are able to display the relative density and shape of the data while retaining most of their raw numerical integrity. Stem and leaf plots are arranged like a table. The stem of the data, the first digit or digits of the numeral, is organized in a vertical column on the left, from the smallest to largest value. The leaves are usually the last digit or the digit(s) after a decimal point. Leaves are organized as horizontal rows, from smallest to largest value, and aligned with the appropriate stem. To illustrate, refer to the worked example in Figure 5.6 in which the continuous numerical data for vessel weight are listed as a table and corresponding stem and leaf plot. The first step in creating a stem and leaf plot is to organize the raw data (Figure 5.6a) from least to greatest (Figure 5.6b). Next, the data is rounded to the nearest whole number (Figure 5.6c) and broken into stems (tens digit) and leaves (ones digit) (Figure 5.6d). Stem and leaf plots always require a key or code, which explains the units/digits of the stem and leaves. Figure 5.6 demonstrates the value of stem and leaf plots quite nicely: we can easily identify that the majority of the vessels weigh between 6 and 8 lbs. Stem and leaf plots are particularly useful for large numeric datasets in which patterns are not easily discerned from standard tables, particularly when it is important to compare similar datasets without losing their numerical resolution. Comparative stem and leaf plots or back-to-back stem and leaf plots allow us to quickly identify similarities and differences in distributions between datasets. For example, if we wanted to understand the level of standardization in cooking pot production between two workshops, we could do a comparative stem and leaf plot of cooking pot volumes (Figure 5.7).
Figure 5.6 Stem-and-leaf plot (worked example) of vessel weights. (Figure courtesy of A. M. W. Hunt)
Figure 5.7 Back-to-back stem-and-leaf plot comparing cooking pot volumes from two sites. (Figure courtesy of A. M. W. Hunt)
66 Gulsebnem Bishop In a back-to-back stem and leaf plot, the stem is the same for both datasets and is plotted in the central column. The leaves are specific to each population being compared and are placed to either side of the stem, as illustrated in Figure 5.7. The back-to-back stem and leaf plot makes it easy for us to see that, while there is a high level of intraworkshop standardization, interworkshop standardization is fairly low; each workshop manufactures two sizes of cooking pot, but those sizes do not overlap, and vessels from Akko are typically more than 2 mL larger than those from Haifa.
Sampling In inferential statistics distribution lists are used when we do not have access to the whole group of material we are interested in learning about. For example, we might want to analyze ceramics from 120 sites in Greece but we might not have information about the entire assemblage for each of these sites. Instead of giving up the project, we can use statistical sampling to make inferences about the entire population from a smaller sample population. Sampling is the statistical selection of a subset of individuals from a population to estimate characteristics of the whole population. The first and most important step in statistical sampling is defining the problem or question and the population from which the sample will be drawn. In some cases, the population definition is obvious. For example, if we want to investigate standardization of Greek transport amphorae in the sixth–fifth centuries bc, our statistical population will be Greek transport amphorae from contexts dated between the sixth and fifth centuries. In other cases, the statistical population may be less clear and require careful consideration and definition. For example, if we want to investigate changes in potting technology over time, both the observations of interest (the aspect(s) of potting technology we are interested in) and the statistical population (the type(s) of vessels, the region and the time period we are investigating) need to be refined and more clearly defined. Often the problem and statistical population are partially defined by limitations inherent in the sample frame or list of individuals available to be sampled. In our Greek transport amphorae example, if our sampling frame contained only vessels from the fifth century bc, we would need to redefine our problem and statistical population to reflect this limitation in the data. In archaeology, we typically use probability sampling in which every individual in the statistical population has a chance (probability > 0) of being selected in the sample. Since the probability of an individual being selected can be accurately determined, probabilistic sampling can produce unbiased estimates of population totals by weighing sampled individuals according to their probability of selection. The three most useful types of probabilistic sampling for archaeologists are simple random sampling, stratified random, and sampling systematic sampling.
Simple Random Sample A simple random sample is one of the most straightforward sampling methods and can be done without replacement (once a sample has been selected it is removed from the statistical population), or, less commonly, with replacement (where the sample is returned to the
Statistical Modeling for Ceramic Analysis 67 Table 5.2 Random number table (excerpt) 40603
16152
83235
40941
53585
69958
73505
83472
55953
37361
98783
24838
39793
80954
76865 32713
60916
71018
90561
84505
53980
64735 85140
17957
11446
22618
34771
25777
27064 13526
39412
16013
11442
89320
11307
49396
39805
12249
57656 88686
57994
76748
54627
48511
78646
33287
35524
54522
08795 56273
61834
59199
15469
82285
84164
91333
90954
87186
31598 25942
91402
77227
79516
21007
58602
81418
87838
18443
76162 51146
58299
83880
20125
10794
37780
61705
18276
99041
78135 99661
40684
99948
33880
76413
63839
71371
32392
51812
48248 96419
75978
64298
08074
62055
73864
01926
78374
15741
74452 49954
34556
39861
88267
76068
62445
64361
78685
24246
27027 48239
65990
57048
25067
77571
77974
37634
81564
98608
37224 49848
16381
15069
25416
87875
90374
86203
29677
82543
37554 89179
52458
88880
78352
67913
09245
47773
51272
06976
99571 33365
33007
85607
92008
44897
24964
50559
79549
85658
96865 24186
38712
31512
08588
61490
72294
42862
87334
05866
66269 43158
58722
03678
19186
69602
34625
75958
56869
17907
81867 11535
26188
69497
51351
47799
20477
71786
52560
66827
79419 70886
12893
54048
07255
86149
99090
70958
50775
31768
52903 27645
33186
81346
85095
37282
85536
72661
32180
40229
19209 74939
79893
29448
88392
54211
61708
83452
61227
81690
42265 20310
48449
15102
44126
19438
23382
14985
37538
30120
82443
94205
04259
68983
50561
06902
10269
22216
70210
60736 58772
11152
38648
09278
81313
77400
41126
52614
93613
27263
99381 49500
04292
46028
75666
26954
34979
68381
45154
09314
81009 05114
statistical population after selection and may be selected again). The best way to talk about sampling is by working an example. Let us say we have 5,000 ceramic sherds from Area Q at a site and we want to send 100 to the laboratory for chemical analysis. To take a simple random sample, we assign sequential numbers to each of the 5,000 sherds (individuals) in the statistical population (N). Since the largest number in population N has four digits, it is helpful to list all the samples with four digits, such as 0001, 0002, 0003 … 5000. The sample population (n) is selected by using a random number generator (available in most statistical software packages) or a random number table (Table 5.2) to select four-digit numbers at random until 100 samples have been selected. The four-digit random numbers are selected by reading four adjacent terms together or by selecting every nth term to create groups of four. These random numbers can be read in any direction/combination and do not need to begin at the start of the table. For example, we will begin with the second row and select every fourth term to generate our four-digit random numbers. This means that, using Table 5.2, the first
68 Gulsebnem Bishop six random numbers generated are 4,596, 6,154, 5,031, 3,853, 5,423, and 1,771. Numbers 6,154, 5031, and 5,423 do not correspond to individuals in population N and so they are discarded. Individuals 4,596, 3,853, and 1,771 become part of the sample population n, and the process continues until population n contains 100 individuals.
Stratified Random Sampling Most archaeological research questions involve comparing datasets in which differences among populations are obvious or expected, such as vessel type or archaeological context. These differences define subpopulations within the statistical population N. To ensure a representative sample population n, it is necessary to ensure that each subpopulation is sampled. To do this, we use a stratified random sample. A stratified random sample is very similar to a simple random sample, except that the sample is selected in accordance with the relative proportion of that subpopulation to the total statistical population N. For example, Area Q at our hypothetical excavation has four contexts, each of which contains both plain and painted wares: context A contains 20% plain and 80% painted; context B contains 32% plain and 68% painted; context C contains 50% plain and 50% painted; and context D contains 60% plain and 40% painted. For simplicity, let us assume that each context has an equal number of individuals or 1,250 sherds. To take a stratified random sample we need first to separate the statistical population N into subpopulations according to context. Since we want a sample population n of 100 individuals, we need to select 25 sherds from each context. Next, we need to separate each subpopulation (context) into plain and painted wares. In context A there are 250 plain sherds (20%) and 1,000 painted sherds (80%). To ensure that the sample population of 25 shreds selected from this subpopulation is representative, we need to take 20% of the sample from the plain sherds and 80% of the sample from the painted sherds. For a sample size of 25, this means five plain and 20 painted sherds need to be selected. From here the process is the same as taking a simple random sample. Sequential numbers are assigned to the sherds in the subpopulation for context A and random numbers created until five plain and 20 painted sherds are selected. Each of the subpopulations is sampled according to the proportion of plain and painted ware it contains, generating a sample population n that is representative of the diversity/heterogeneity in the statistical population N. The advantages of stratified random sampling are that it improves the accuracy of a sample taken from a heterogeneous population, provides higher-resolution information about the statistical population, and is often more cost effective since fewer observations/individuals are required for a representative sample.
Systematic Sampling Systematic sampling is similar to simple random sampling in that every individual in the statistical population has a known and equal probability of selection and is most effective when the statistical population N is logically homogeneous because the sample units are uniformly distributed over the population. To take a systematic sample we must determine the sampling interval k. To arrive at k, we simply divide the population size N by the desired sample
Statistical Modeling for Ceramic Analysis 69 size n. In our example, the statistical population size is 5,000 sherds and we want to send a sample of 100 sherds for chemical analysis. This means that our sampling interval is 50; every 50th sherd in population N is selected for analysis.
Describing Complex Data Statistics can also be used to model or understand underlying structure in complex data in which there are multiple variables, such as chemical compositions. Although there are numerous statistical procedures and methods that can be used to describe complex data, in this section we focus on the six most commonly used by archaeologists: multi-response permutation procedure (MRPP), kernel density estimation (KDE), principal components analysis (PCA), t-tests, and chi-square tests. Each of these procedures is able to detect subtle differences among multivariable datasets, making them extremely valuable in archaeological research where the observations of interest are often subtle. Although these procedures involve complex mathematics and statistical theory, they are all available commercially in most statistical software packages, making them accessible to laymen.
Multi-Response Permutation Procedure (MRPP) MRPP is a non-parametric method that tests the hypothesis of no difference among pre- existing groups (Mielke, 1984). Put another way, MRPP test whether the observed variation among groups results from different populations of origin or arises from natural variation within a single population of origin. The null hypothesis in MRPP is that there is no difference between groups; that is, they have the same population of origin. MRPP discriminates between this “natural” and “artificial” variation by calculating the weighted mean within group Euclidean distance or delta (δ) for each pre-existing group and determining the probability that random Monte Carlo permutation (groups) of the individuals in all the groups would have a smaller δ than the observed δ. The acceptance or rejection of the null hypothesis is determined by generating a p-value; if the p-value is equal to or smaller than the significance level (a) the null hypothesis is rejected and the variation among groups is considered to result from different populations of origin. The major advantage of MRPP is that it is able to compare groups containing an unequal number of individuals because δ is a weighted mean distance. MRPP is also a more robust alternative to normal theory-based parametric methods, such as t-tests, because it does not require assumptions, such as multivariate normality and homogeneity of variances, that archaeological data are seldom able to meet.
Kernel Density Estimation (KDE) KDE is a non-parametric estimate of the probability density function of a random, continuous variable. KDE is related to histograms, which as we have seen are used to describe the distribution or density of a variable across a population. Histograms have a few flaws related
70 Gulsebnem Bishop to the arbitrariness of their parameters; namely, group width and starting position. In addition, because histograms describe only available data, there are often discontinuities, making it difficult to detect underlying structure in the data. KDE resolves at least two of these issues. First, KDE involves moving a smoothing function (the kernel) along the observations to generate a continuous probability curve that reveals and predicts data structure despite discontinuities. Second, by estimating results according to an average shifted histogram— that is, the average of several histograms based on shifts in group edges—KDE eliminates the arbitrariness of the steps between groups, which are often unrealistic in histograms. However, the bandwidth of the kernel or size of the groups impacts the shape of the pattern and, thus, the interpretation of the data. Wand and Jones (1995) discuss a variety of complex mathematical functions for determining the appropriate bandwidth for a dataset; however, Shennan (1997) argues that these methods are no more appropriate/successful than simply trying out bandwidths of several sizes and intuitively selecting the one with the best balance of “smooth” to “rough.” Examples of how KDE has been used in archaeology data can be found in Baxter and Beardah (1995a, 1997).
Principal Component Analysis (PCA) PCA uses orthogonal transformations to convert a set of observations that are potentially correlated into a set of linearly uncorrelated variables called principal components. These principal components are orthogonal; that is, uncorrelated with each other because they are eigenvectors of the covariance matrix. The number of principal components is always less than or equal to the original number of variables and are defined by the transformation in such a way that the first principal component has the largest possible variance (accounts for as much of the variability in the data as possible) and each subsequent component has the highest variance possible that is orthogonal to/uncorrelated with the variation described by the preceding components. PCA is primarily used as a tool for exploratory data analysis in order to reveal the internal structure of the data responsible for its variance. PCA results are often evaluated in terms of components scores or factor scores and component loadings. Component scores are the transformed variable values corresponding to particular data points and loadings are the weight by which each standardized variable in the original data should be multiplied to get the component score. By squaring the component loading of a variable, we are able to determine the percentage of the total variation in that variable that is accounted for by the component. In this way, PCA is able to provide high-resolution information about the structure and variation in complex and large datasets. In archaeological ceramic analysis, PCA can be used for many and diverse applications, from understanding variation in vessel shape (Shennan, 1997) to evaluating geochemical provenance groups. However, PCA has some limitations which can restrict its applicability for archaeological data. First, PCA is sensitive to the relative scaling of the original variables; every individual or sample in the population must have a value for every variable in order for PCA to correctly determine variation distributions/data structure. Second, PCA is a purely descriptive technique; PCA does not and cannot make predictions about what future data and observations will look like.
Statistical Modeling for Ceramic Analysis 71
T-Tests A t-test is any statistical hypothesis test in which the test statistic follows a Student’s t or normal distribution if the null hypothesis is correct. Typically, t-tests are used to estimate or determine whether two datasets are significantly different from one another. A Student’s t distribution is used instead of a normal distribution when the scaling term is unknown and must be derived or estimated from the data itself. There are two types of t-test: independent and paired. As the names indicate, an independent t-test is used when we want to compare two independent and identically distributed datasets, one from each of the populations being compared, and paired t-tests are used to compare paired data from the same population, such as repeated measures or before and after studies. When datasets are put through the t-test formula, a t-statistic is generated. The t-statistic is a measure of variance; more specifically, it is the ratio of an estimated parameter from its notational value and its standard error. Next, the p-value for the desired significance level, typically 5% or 1%, is calculated. The p-value is used to quantify the statistical significance of an observation or hypothesis and is defined as the probability of obtaining a result equal to or more extreme than what is actually observed. One of the limitations of t-tests is that failure to find a statistically significant difference between the datasets does not necessarily mean that the sample populations are the same. Another limitation, particularly for archaeological data, is that four assumptions about the data or conditions must be met for the test to be valid: (a) that one variable is continuous and the other is dichotomous; (b) the two datasets have equal variance; (c) the two datasets are normally distributed; (d) the observations are independent (except if running a paired t-test). In archaeological ceramic analysis, t-tests are most often used to compare macroscopic data, vessel morphology, or context (see Abramiuk, 2012). For example, t-tests can be used to understand the distribution of artifacts across a landscape. However, the limitations and assumptions required for valid t-tests seriously reduce their utility for archaeological data.
Chi-Square Test (or Chi-Squared Test) The chi-square test (χ ) is any statistical hypothesis test in which the test statistic follows a chi-square distribution if the null hypothesis is true. Chi-square tests are used to determine whether the difference between expected and observed frequencies for one or more variables is statistically significant. The most commonly used chi-square test in archaeology is the Pearson’s chi-square test, also known as the chi-square goodness-of-fit test or the chi- square test for independence. The Pearson’s chi-square test is able to evaluate large populations of unpaired data in order to determine the probability or likelihood that the observed difference between populations arose by chance (Drennan, 1996). Like the t-test, chi-square tests use a null hypothesis of “no difference” between datasets (observed and expected), which is accepted or rejected on the basis of the p-value for the desired level of significance. Also like the t-test, chi-square tests do not provide information about the strength of the relationship between expected and observed frequencies or the substantive significance of the relationship in the population. (Van Pool and Leonard 2
72 Gulsebnem Bishop (2011) discuss the difference between statistical significance and substantive significance.) Additionally, chi-square tests are sensitive to sample size, the calculated chi-square is directly proportional to the sample size regardless of the strength of the relationship between variables, and small expected frequencies in one or more of the cells in the dataset. These last two limitations are critical for archaeological datasets because the sensitivity of chi-square to sample size may make a weak relationship appear statistically significant in a large dataset and/or impact the validity of chi-square tests where one or more cells are less than five.
Conclusion In this chapter we tried to present an overall view of statistical data types, tests, and tools that can be used to describe and analyze archaeological ceramic data. Ceramic data can be numerical or non-numerical, discrete or continuous, quantitative or qualitative. By understanding our data, its strengths and limitations, we can develop a well-defined research question and sample population in order to describe that data and derive and predict meaningful outcomes. Tables and bar and pie charts can be used to describe and predict outcomes for non-numeric or qualitative data, and histograms, stem and leaf plots, and distribution analysis can describe and predict outcomes for quantitative data. Complex data can be described and modeled and outcomes predicted using multi-response permutation procedure (MRPP), kernel density estimation (KDE), principal components analysis (PCA), t-tests, and chi-square tests.
References Abramiuk, M. (2012). The Foundations of Cognitive Archaeology (Cambridge, MA: MIT Press). Baxter, M. J. and Beardah, C. C. (1995). “Graphical Presentation of Results from Principle Component Analysis.” In: Huggett, J. and Ryan, N. (eds), Computer Applications and Quantitative Methods in Archaeology 1994. BAR International Series 600. (Oxford: BAR International Series), 63–67. Baxter, M. J. and Beardah, C. C. (1997). “Some Archaeological Applications of Kernel Density Estimates.” Journal of Archaeological Science 24: 347–354. Drennan, R. D. (1996). Statistics for Archaeologists: A Common Sense Approach (New York: Springer). Mielke, P. W. (1984). “Meteorological Applications of Permutation Techniques Based on Distance Functions.” In: Krishnaiah P. R. and Sen, P. K. (eds), Handbook of Statistics, vol. 4 (Amsterdam: North-Holland), 813–830. Shennan, S. (1997). Quantifying Archaeology (Edinburgh: Edinburgh University Press). VanPool, T. L. and Leonard, R. D. (2011). Quantitative Analysis in Archaeology (Oxford: John Wiley and Sons). Wallace Matheson, P. M. and Wallace, M. B. (1982). “Some Rhodian Amphora Capacities.” Hesperia: The Journal of the American School of Classical Studies at Athens 51(3): 293–320 Wand, M. P. and Jones, M. C. (1995). Kernel Smoothing (London: Chapman & Hall/CRC).
Chapter 6
Recycling Data
Working with Published and Unpublished Ceramic Compositional Data Matthew T. Boulanger Introduction The Oxford English Dictionary defines “recycle” as the process of returning material to a previous stage in a cyclic process. In this sense, nearly all scientific pursuits involve recycling data: analyses generate data from which knowledge is produced, and that knowledge is used to formulate new analyses that in turn generate new data. Knowledge and data are entwined, each requiring the other to produce and sustain itself (Carraway, 2011). But as scientists interested in the formulation of knowledge we often lose sight of the importance of data, focusing instead on the resultant knowledge and neglecting the primary datasets used to produce that knowledge. This problem has been, and in some cases continues to be, endemic across scientific disciplines (e.g. Curry, 2011; Hanson et al., 2011). This chapter focuses on the fundamental nature of data recycling with respect to the analysis of archaeological ceramics. Though archaeological ceramic analysis is a small player in the Big Data movement, archaeology as a discipline is beginning to recognize that it suffers from many of the same problems of poor data-management practices being discussed by scientists as varied as ecologists, particle physicists, and paleontologists (e.g. Curry, 2011; Brewer et al., 2012; Uhen et al., 2013). Here, I focus specifically on one form of archaeological ceramic data: geochemical data generated to facilitate estimating or determining provenance, or geological origin of ceramic raw materials. Despite this focus, nearly all of the details of this discussion are applicable to other forms of ceramic data, be they quantitative or qualitative. Until relatively recently most ceramic-compositional data were generated by a limited number of individuals operating in specialized laboratories. Decreases in cost and physical size of analytical instruments, as well as the increased commercial availability of such instruments have resulted in the decentralization of data-generation capabilities (Shackley, 2010; Frahm and Doonan, 2013). Although I believe this shift to be positive, a historical perspective on the use of these technologies is necessary. In the past, specialized laboratories
74 Matthew T. Boulanger carried—and lived up to—the obligations of interlaboratory communication, collaboration, insurance of mutually interchangeable data, standardization to allow reuse (both internally and externally), and maintenance of databases (Harbottle, 1982b). Reuse of data in this environment was typically small scale and easily accomplished: the individual players knew each other, knew what data were being generated, and knew each others’ procedures so as to ensure mutual intelligibility. Regardless of its “potential to make very real changes in our discipline” (Shackley, 2010: 18), the currently increasing widespread adoption of instrumentation capable of generating compositional data has the very real potential of leading to what Speakman and Shackley (2013: 1435) refer to as silo science—a situation in which “each researchers’ data is self-contained, independent, and cannot be verified externally.” Silo science, as they mean it, is highly problematic for archaeology in general, and ceramic compositional analysis in particular, because data are generated without regard for precision, accuracy, or interobserver compatibility. As long as the data appear to be internally consistent with previous analyses on the same instrument, they are good enough. Under this perspective, the archaeological knowledge (i.e. from where a piece of obsidian came, or what compositional groups are made from pottery) is the goal, and chemical analyses are simply a means to get there. The problem, though, is that the data used to generate this knowledge can be neither replicated nor evaluated. At best, this results in additional time and effort to re- analyze the same or similar specimens. At worst, it means that the resultant knowledge can be neither confirmed nor refuted, and any evaluation of these data necessarily boils down to argument from authority or opinion.
An Historical Perspective on the Recycling of Data Analyses of archaeological ceramics have, in many ways, always rested upon a foundation of data recycling. Consider the case of the University of Michigan’s Ceramic Repository for the Eastern United States established by the National Research Council in the late 1920s and directed by Carl E. Guthe (Lyon, 1996; O’Brien and Lyman, 2001). James B. Griffin, one of the first graduate students to assist Guthe with the repository, was charged with bringing “some semblance of order into the heterogeneous prehistoric pottery material in the Midwest and in the eastern United States … by the study of material already out of the ground in Museums and private collections” (Griffin, 1976: 21, emphasis added; see also O’Brien and Lyman, 1998: 110–115). Individuals wishing to draw upon data housed in the repository could travel to the university or correspond with Guthe and/or Griffin. They could incorporate existing data into their own research as well as contribute new data to the repository. The goal of the repository was to provide a central hub for data storage and data dissemination. As successful as the Ceramic Repository and similar institutions were, their use was restricted by the need to travel to a specific location. This situation slowly began to change in the late 1960s, when computers made it possible to store and to analyze data digitally. However, computing centers were generally available only at large universities and government- funded national laboratories. Storage media— punch- cards and magnetic
Recycling Data 75 tape—required physical delivery, and neither media was amenable to easy exchange across long distances. Even if they had been, the end user would need to have the correct hardware to read them and knowledge of the correct sequence of control commands to access them. Nonetheless, the advent of digital data-storage set the stage for fundamental changes in how data could be stored, accessed, and reused. Concomitant with the development of digital computers was the development and refinement of physicochemical techniques aimed at the compositional analysis of archaeological pottery, driven in large part by nuclear research facilities throughout the Western world (see reviews by Sayre, 1963; Harbottle, 1976; Wilson, 1978; Harbottle, 1982a; Bishop and Blackman, 2002; Beaudry-Corbett, 2003; Speakman and Glascock, 2007). Nearly all compositional analyses performed after the mid-1960s have been directed at identifying the raw material provenance of archaeological material (Sayre et al., 1957; Emeleus and Simpson, 1960; Richards and Hartley, 1960; Sayre, 1963; Perlman and Asaro, 1969; Perlman et al., 1972; Harbottle, 1976; ). Many data were produced between 1960 and 1980, and most of these data were generated at a handful of laboratories specializing in bulk compositional analysis. In 1982, Harbottle suggested that until the mid-1970s most compositional studies of archaeological ceramics were “isolated, ‘self-contained’ usually even within the laboratory concerned” (Harbottle, 1982b: 68). Pottery analyses were usually undertaken at the level of individual projects or sites. Rigorous comparisons to databases—even within the same laboratory—were rarely undertaken except at the coarsest levels. There was little concern with ensuring long-term data preservation. I suspect that this isolation of data was, at least partially, the result of three factors. First, comparisons of raw data were often performed, at least initially, by hand. Second, laboratories were focused primarily on the novelty of methodology. And, third, not many data existed. Regardless of cause, this is precisely the situation described thirty years later by Speakman and Shackley (2013) as “silo science.” By the mid-1970s, this “isolation” of data was effectively ending. Sayre (1975) introduced a series of computer programs for multivariate compositional-data analysis; thus it was possible to analyze large data matrices without resorting to a slide rule or pocket calculator. Harbottle (1982b) estimated that data existed for upwards of 50,000 archaeological specimens worldwide; thus comparative data were available for many regions of the world. And, most importantly, nearly every laboratory producing ceramic-compositional data had established routine protocols involving the analyses of standard reference materials, and the mutual exchange of these standards with other laboratories. Over time, nearly all major laboratories converged on the use of internally developed standard materials (e.g. Perlman-Asaro Standard Pottery) (Yellin et al., 1978; Asaro and Adan-Bayewitz, 2007; Yellin, 2007), internationally available standard reference materials (Blackman and Bishop, 2007; Glascock et al., 2007; Harbottle and Holmes, 2007), and intercalibration among laboratories (Yellin et al., 1978; Ferreira et al., 1980; Yeh and Harbottle, 1986; Djingova et al., 1990; Tomlinson, 2002). All of these factors, as well as the newly available microcomputer, crystallized in the early 1980s, and archaeologists and ceramic analysts began addressing the need to establish one or more repositories to preserve and provide access to data being generated across the world (e.g. Matson, 1982a: 25–26, 1982b; Sayre, 1982). The Smithsonian Archaeometric Research Collections and Records (SARCAR) initiative (Bishop et al., 1983) was created as a potential solution for the accumulation—but general lack of rigorous integration and long-term preservation—of ceramic data in archaeometric laboratories.
76 Matthew T. Boulanger Like the University of Michigan’s Ceramic Repository, SARCAR was planned as a physical repository to which researchers could travel to access data. In part, this was because of limitations in data-storage technologies at that time. Large-capacity data storage was only available on magnetic media, and this was not only difficult to transport, it was (relative to later technologies) fragile both physically and in terms of its ability to retain data. Subsets of data could be sent to colleagues on 8 in., 5.25 in., or 3.5 in. diskettes; but again, these magnetic media were not particularly robust. Moreover, with capacities of 500 kilobytes, 800 kilobytes, and 1.44 megabytes respectively, none was capable of holding a very large dataset. Interlaboratory data exchanges typically involved printing tabular data at the source and subsequent transcription of these data at the destination (e.g. Tomlinson, 1997; Hein et al., 2001; French et al., 2008). Despite the difficulty of transferring and digitizing stacks of printed paper and the fragility of magnetic storage media, direct exchange and reuse of ceramic data showed a marked increase during the 1980s and early 1990s (e.g. Jones, 1986; Knapp and Cherry, 1994). In 1995, after several years of increased Internet commercialization, the National Science Foundation Network (NSFNet) transitioned into what became the backbone of today’s World Wide Web (WWW). The effect of the WWW on the reuse of archaeological data was profound because it provided the ability for researchers to have near-instantaneous access to data generated at other laboratories, regardless of the size of a database. In the mid-1990s, Hector Neff, then at the Archaeometry Laboratory at the University of Missouri Research Reactor (MURR), established a Web page containing datasets from published studies on the Internet.1 To the best of my knowledge, this was the first effort to provide a single source of compositional data for archaeological ceramic analyses, in open-access format, on the WWW. In the years since, the sophistication and scope of online data distributions have changed dramatically, moving away from individuals and individual laboratories toward large data warehouses. The ceraDAT database () developed by Hein and Kilikoglou (2011) is intended to provide a central repository for compositional data on archaeological ceramics from the Aegean and eastern Mediterranean. No similar database exists for the Americas. However, the Digital Archaeological Record (tDAR) () (Kintigh and Altschul, 2010; McManamon et al., 2010) is an online repository for all archaeological data and can serve as a distribution hub for compositional data (e.g. Boulanger, 2013). Unlike earlier repositories (e.g. the Michigan Ceramic Repository and SARCAR) these tools provide researchers with direct access to raw data at any time from nearly anywhere with access to the Internet.
Recycling Data in the Twenty-First Century Kintigh et al. (2014: 879) argue that “the greatest payoff [in archaeological knowledge during the twenty-first century] will derive from exploiting the explosion in systematically collected archaeological data that has occurred since the mid-20th century.” I agree, and the current push to implement policies and infrastructure for long-term data-management practices should provide benefits to archaeology in the future. An historical perspective suggests that recycling of ceramic analysis data has been a major objective throughout most of the twentieth century; yet what is fundamentally different in the twenty-first century is the almost
Recycling Data 77 limitless capacity to store data, and the ability for individuals to access those data instantaneously from nearly anywhere (e.g. Snow et al., 2006). Development of computer technologies for storing, analyzing, and sharing ceramic-compositional data has the potential to greatly increase the scale of data recycling. Adoption both of ethical obligations (e.g. principles six and seven of the Society for American Archaeology’s Principles of Archaeological Ethics, ) and funding mandates (e.g. the National Science Foundation Data Management Plan Requirements adopted in 2011, ) regarding data management and data sharing provide incentives and motivations for current researchers to ensure that their data are reusable over the long term. However, these programmatic statements do not address the large body of data generated during the latter half of the twentieth century that remains inaccessible—and largely unusable—in its current forms (Snow, 2010). Data must be transitioned into formats that are usable with modern technologies. If Harbottle’s estimation that 50,000 archaeological specimens were analyzed by 1982 is correct, all of these data were generated prior to the introduction of 3.5 in. floppy disks, MS- DOS, and the IBM Personal Computer. Nearly all of the laboratories that produced these data have now closed, and have been closed for over a decade (Figure 6.1). Interlaboratory data standardization must be established; that is, it must be demonstrated that legacy data are directly compatible with newly generated data. This may not be a problem for those laboratories that, while in operation, established such standardizations. Yet not all laboratories did so. Moreover, laboratory practices change over time, and intimate knowledge of these changes was often retained only by those individuals working within the laboratory. These
Brookhaven Berkeley Michigan Smithsonian/NIST MURR SLOWPOKE (Toronto) ITN, Portugal Hebrew Univ. Demokritos Univ. Manchester Univ. Sofia British Museum Budapest Bonn Texas ASM SLOWPOKE (Quebec) Argentina IPEN, Peru McMaster
Formal laboratory/Institute operating Reactor operating, occasional archaeom. research 500 archaeological specimens analyzed by NAA
1950
1960
1970
Oregon State
1980
1990
2000
2010
Figure 6.1 Timeline of a selection of former and current nuclear archaeometry laboratories, and estimates of the total numbers of archaeological specimens analyzed. Data compiled primarily from vol. 49(2) of Archaeometry.
78 Matthew T. Boulanger data must also be made available in a manner that ensures their perpetual availability, and this may require current researchers to reconsider long-held beliefs about ownership of data and intellectual property. I address each of these points in the following section, using my work on the ceramic-compositional database of the former archaeometry program at Lawrence Berkeley Laboratory to provide specific examples and direction.
The Lawrence Berkeley Laboratory Archaeometry Program Archaeometric research began during the mid-1960s at Lawrence Berkeley Laboratory (LBL) in the research group of Isadore Perlman, comprising himself, Frank Asaro, and Helen V. Michel. By 1967, Perlman and Asaro (1967) had developed a method involving neutron activation (NAA) and high-purity germanium detectors, and, shortly thereafter, they modified this method to include the use of an in-house multi-element reference material, Perlman-Asaro Standard Pottery (Perlman and Asaro, 1969), to ensure consistently precise elemental determinations during each assay. Perlman took early retirement from LBL in 1972, and in 1973 accepted dual positions in archaeology and chemistry at the Hebrew University of Jerusalem where he, with Joseph Yellin, began developing an NAA archaeometry laboratory (Yellin, 2007). Asaro remained at LBL and directed the archaeometry program from 1972 onwards. Although he officially retired in 1991, Asaro continued archaeometric research at LBL until his death in 2014. Between 1967 and 1989, the Berkeley archaeometry program analyzed over 10,000 archaeological specimens, most of which were assayed during the 1970s. Generation of compositional data at LBL involved the use of three different computer data-storage media and extensive use of paper records. Powdered archaeological specimens were prepared for irradiation, and descriptive and contextual data for each specimen were handwritten on loose-leaf paper stored in D-ring binders. IBM punch cards were prepared listing specimen identification number, weight, thickness, irradiation and decay times, counting times, and other pertinent information. These cards were fed into the LBL mainframe and used to process spectral data, stored on magnetic tape, from the five gamma-ray counts performed on each specimen. Processing of the spectral data resulted in the generation of elemental abundance data, which were printed out and quality checked by hand. After the data were quality checked, various statistical tests were performed both by hand and on the LBL mainframe, and compositional groups were formed. Once analyses were completed, data were stored digitally on magnetic-tape cartridges in LBL’s mass storage system. Printed copies of elemental abundances were also retained. Irradiation of specimens at the laboratory ended in 1989 with the decommissioning of the Berkeley reactor (Asaro and Adan-Bayewitz, 2007). Shortly before this time the LBL group began sending specimens to either the Reed Research Reactor or MURR for irradiation. After irradiation offsite, specimens were shipped back to LBL for counting. In 1990 and 1991 considerable effort was made to copy the LBL archaeometry database from obsolete mass storage cartridges to more than fifteen individual 3.5 in. floppy disks. Funding was scarce and computing infrastructure at LBL had changed dramatically by this time. Much of the effort to copy these data was personally funded by Helen Michel (F. Asaro, pers. comm., 2006).
Recycling Data 79
The Lawrence Berkeley Laboratory Archaeometry Archives In 2006, Asaro transferred a large collection of materials from the LBL program to the Archaeometry Laboratory at MURR (Asaro and Adan-Bayewitz, 2007). These archives consist of surplus specimens and powders, more than 5,000 pages of handwritten records, twenty volumes of 12 in. × 18 in. dot-matrix printouts of elemental abundances, and a COM microfilm archive of channel counts for (presumably) every gamma-ray count conducted at the laboratory. These documents represent the largest single archive of archaeological (and some geological) specimens irradiated and analyzed at LBL from 1967 until decommissioning of the LBL reactor in 1989. Although the laboratory performed gamma-ray counts on specimens irradiated elsewhere, these data are not present in the archive. The massive amount of compositional data in the LBL archive, and the presence of detailed descriptive data for nearly all of the analyzed specimens, provide significant research potential. Despite the effort to retrieve the elemental abundance data from antiquated magnetic tape during the early 1990s, the 3.5 in. floppy disks retained by the laboratory were found to be empty. Moreover, none of the handwritten descriptive and contextual information for analyzed specimens had ever been digitized. Thus, the only record of data produced at LBL was in printed form. Digitization of these data required the line-by-line transcription of more than thirty elemental abundance calculations for each specimen in addition to all identification numbers, descriptions, and archaeological contexts. Fidelity in transcription was ensured through third-party examination of each entry afterwards. Nearly three years of labor performed sporadically between various other obligations was required to transcribe all of the data.2 At some point in the digitization process it became apparent that many more specimens had been analyzed at LBL than those for which elemental abundance data were present. Once all of these records had been digitized I estimated that the archives sent to MURR contained less than half (roughly 40%) of the estimated 10,000 specimens analyzed (Boulanger and Glascock, 2009). Between 2009 and 2013 I was able to locate an additional 80% of these missing data within the records of former LBL collaborators and in published sources (Boulanger, 2013). In some cases, these were fortuitous discoveries. Michal Artzy completed her PhD dissertation research through LBL in the early 1970s and was hired there as a post- doctoral researcher. Before leaving LBL in the late 1970s, Artzy printed a large amount of elemental data from the LBL mainframe with the intention of continuing her research. More than thirty years later, she provided copies of these printouts for the digitization effort. In other cases, the discoveries were disheartening. One well-meaning colleague stated that he had been holding on to several binders of LBL data printed in the early 1970s. But, after hearing that MURR had received the archives of the LBL program, he had assumed them to be redundant and subsequently discarded them. Whether they were indeed redundant may never be known. If they were not, it is likely that these data can never be recovered. In 2011, a preliminary digitized LBL database containing all the data provided by Asaro was uploaded to tDAR to ensure long-term open access to these data. In 2014, an update (including all of the data acquired between 2010 and 2013) was made to these files (Boulanger, 2014). In addition to making the digital versions of these archives available, all the original paper records were transferred to the University of Missouri Anthropology Museum Support Center for long- term curation. Because nearly all the original
80 Matthew T. Boulanger collaborators with the LBL program are either retired or deceased, questions concerning ownership or intellectual property rights of these data are vague at best. Respecting the collaborative agreements that resulted in the generation of these data, it was decided to include the names of laboratory analysts (i.e. Perlman, Asaro, and Michel) as well as those of collaborating archaeologists as the “authors” of these data as listed on tDAR. Thus, any use or citation of these data will explicitly acknowledge the contributions of the individuals involved.
Looking Backward and Looking Forward Ceramic provenance studies have historically been situated on the fringes of the social and natural sciences (Dunnell, 1993; Killick and Young, 1997), yet our data are explicitly archaeological. Recycling data has been, and will continue to be, a fundamental aspect of archaeological ceramic analysis. In some cases, much work may be required before extant data can be incorporated into current and future projects (Kintigh, 2006; Snow, 2010). However, we should not shy away from this task. From an ethical standpoint, many of these data were generated with some form of public funds, and thus researchers have an obligation to ensure that these data are usable by future generations of scientists. From an economical perspective, reusing existing data is an efficient way of increasing sample size and the significance of research at minimal cost—consider that today’s costs of reproducing the entirety of the LBL database would be somewhere between US$300,000 and $1,000,000, and doing so would require at least three years of full-time exclusive work! Perhaps most significant, though, is the fact that the preservation and reuse of data encourages—and allows—a culture of replication, reproducibility, and standardization, which serves to decrease waste of resources and ultimately improve the quality and accuracy of archaeological knowledge (Ioannidis, 2014). In the United States, the scope and scale of regulatory compliance archaeology (cultural resource management) began growing exponentially in the 1960s. Much of this work focused on saving artifacts and archaeological data from proximate destruction by bulldozers, hydroelectric projects, and highway construction. Yet standards and guidelines for curating those artifacts and data were not established until decades later, when management of archaeological collections had reached a tipping point at which collections and data were being lost (Bawaya, 2007). Ceramic compositional data are reaching an analogous tipping point. I estimate that today, compositional data have been generated for well over 300,000 individual archaeological specimens by NAA alone, and yet no formal guidelines (e.g. Geboy and Engle, 2011) exist to assure the quality or long-term usability of these data. Moreover, no dedicated repository exists in which these data may be stored. Exacerbating this problem is the fact that data-storage technologies are changing at hyperexponential rates; media on which data were stored as recently as ten years ago are nearly obsolete today. Failure to keep pace with these changes has led to a massive backlog of legacy data. Unless these issues are addressed, this situation will only worsen as more people generate more data. Working to preserve legacy databases allows us to better recognize deficiencies in current practices and, perhaps most importantly, to identify how best we may ensure that our own data are usable by future generations of scientists.
Recycling Data 81
Notes 1. An archived version of this page from 1999 is available on the Internet Archive: (last accessed May 5, 2016). The current version of this page is available via the MURR Archaeometry Laboratory website: (last accessed May 5, 2016). 2. The project would likely have taken less time if dedicated external funding was available, but many granting agencies (e.g. the National Historical Publications and Records Commission) are prohibited from supporting projects involving the digitization of federally funded programs. Because LBL was supported by the US Atomic Energy Commission, the US Energy Research and Development Administration, and the US Department of Energy, digitization of records from the laboratory is ineligible for nearly all federal granting programs.
References Asaro, F. and Adan-Bayewitz, D. (2007). “The History of the Lawrence Berkeley National Laboratory Instrumental Neutron Activation Analysis Programme for Archaeological and Geological Materials.” Archaeometry 49: 201–214. Bawaya, M. (2007). “Curation in Crisis.” Science 317: 1025–1026. Beaudry-Corbett, M. (2003). “A Social History of Archaeological Materials Characterization Studies.” In: Van Zelst, L. (ed), Patterns and Process: A Festschrift in Honor of Dr. Edward V. Sayre (Suitland, MD: Smithsonian Center for Materials Research and Education), 19–25. Bishop, R. L. and Blackman, M. J. (2002). “Instrumental Neutron Activation Analysis of Archaeological Ceramics: Scale and Interpretation.” Accounts of Chemical Research 35: 603–610. Bishop, R. L., Olin, J. S., and Blackman, M. J. (1983). “SARCAR: A New Archaeometric Resource.” In: ICOM Committee for Conservation, 7th Triennial Meeting, Preprints (Paris: ICOM with the J. Paul Getty Trust), 84.8.1–84.8.2. Blackman, M. J. and Bishop, R. L. (2007). “The Smithsonian– NIST Partnership: The Application of Instrumental Neutron Activation Analysis to Archaeology.” Archaeometry 49: 321–341. Boulanger, M. T. (2013). “Salvage Archaeometry: Lessons Learned from the Lawrence Berkeley Laboratory Archaeometric Archives.” The SAA Archaeological Record 13: 14–19. Boulanger, M. T. (2014). “Lawrence Berkeley National Laboratory (LBNL) Nuclear Archaeology Program Archives.” On- line digital document collection available on tDAR: (last accessed May 3, 2016). Boulanger, M. T. and Glascock, M. D. (2009). “Salvage Archaeomery: Rescue, Preservation, and Dissemination of Geochemical Data.” Paper presented at the 2009 Annual Meeting of the Society for American Archaeology. Brewer, S., Jackson, S. T., and Williams, J. W. (2012). “Paleoecoinformatics: Applying Geohistorical Data to Ecological Questions.” Trends in Ecology and Evolution 27: 104–112. Carraway, L. N. (2011). “On Preserving Knowledge.” The American Midland Naturalist 166: 1–12. Curry, A. (2011). “Rescue of Old Data Offers Lesson for Particle Physicists.” Science 331: 694–695.
82 Matthew T. Boulanger Djingova, R., Kuleff, I., and Penev, I. (1990). “Instrumental Neutron Activation Analysis of Reference Materials for Archaeometric Investigations of Pottery.” Journal of Radioanalytical and Nuclear Chemistry 144: 397–406. Dunnell, R. C. (1993). “Why Archaeologists Don’t Care about Archaeometry.” Archeomaterials 7: 161–165. Emeleus, V. M. and Simpson, G. (1960). “Neutron Activation Analysis of Ancient Roman Potsherds.” Nature 185(4707): 196. Ferreira Marques, M. F. and Schubiger, P. (1980). “Comparison of Neutron Activation Analysis from Five Ceramic Standards of the C.R.A.M. CAEN and the Perlman/Asaro-Standard.” Journal of Radioanalytical Chemistry 55: 131–134. Frahm, E. and Doonan, R. C. P. (2013). “The Technological Versus Methodological Revolution of Portable XRF in Archaeology.” Journal of Archaeological Science 40: 1425–1434. French, E. B., Hoffman, S. M. A., Robinson, V. J., and Tomlinson, J. E. (2008). “The Perlman and Asaro Analyses of Late Helladic I-III Sherds from the 1963 Excavations: A Statistical Re- evaluation.” In: Taylour, W. D. and Janko, R. (eds), Ayios Stephanos: Excavations at a Bronze Age and Medieval Settlement in Southern Laconia. Supplementary vol. 44 (London: British School at Athens), CD118–CD123. Geboy, N. J. and Engle, M. A. (2011). “Quality Assurance and Quality Control of Geochemical Data: A Primer for the Research Scientist.” United States Geological Survey Open-File Report 2011–1187. Glascock, M. D., Speakman, R. J., and Neff, H. (2007). “Archaeometry at the University of Missouri Research Reactor and the Provenance of Obsidian Artifacts in North America.” Archaeometry 49: 343–357. Griffin, J. B. (1976). “A Commentary on Some Archaeological Activities in the Mid-continent 1925–1975.” Midcontinental Journal of Archaeology 1: 5–38. Hanson, B., Sugden, A., and Alberts, B. (2011). “Making Data Maximally Available.” Science 331: 649. Harbottle, G. M. (1976). “Activation Analysis in Archaeology.” Radiochemistry 3: 33–72. Harbottle, G. M. (1982a). “Chemical Characterization in Archaeology.” In: Ericson, J. E. and Earle, T. K. (eds), Contexts for Prehistoric Exchange (New York: Academic Press), 13–51. Harbottle, G. M. (1982b). “Provenience Studies Using Neutron Activation Analysis: The Role of Standardization.” In: Olin, J. S. and Franklin, A. D. (eds), Archaeological Ceramics (Washington, D.C.: Smithsonian Institution), 67–77. Harbottle, G. M. and Holmes, L. (2007). “The History of the Brookhaven National Laboratory Project in Archaeological Chemistry, and Applying Nuclear Methods to the Fine Arts.” Archaeometry 49: 185–199. Hein, A. and Kilikoglou, V. (2011). “ceraDAT—Prototype of a Web-based Relational Database for Archaeological Ceramics.” Archaeometry 54: 230–243. Hein, A., Beier, T., and Mommsen, H. (2002). “A Complete Chemical Grouping of the Perlman/Asaro Neutron Activation Analysis Database of Mycenaean and Minoan Pottery.” In: Kilikoglou, V., Hein, A., and Maniatis, Y. (eds), Modern Trends in Scientific Studies on Ancient Ceramics: Papers Presented at the 5th European Meeting on Ancient Ceramics. BAR International Series 1011 (Oxford: Archaeopress), 143–150. Ioannidis, J. P. A. (2014). “How to Make More Published Research True.” PLoS Med 11(10): e1001747. doi: 10.1371/journal.pmed.1001747. Jones, R. E. (1986). Greek and Cypriot Pottery: A Review of Scientific Studies. Fitch Laboratory Occasional Paper 1 (London: The British School at Athens).
Recycling Data 83 Killick, D. and Young, S. M. M. (1997). “Archaeology and Archaeometry: From Casual Dating to a Meaningful Relationship?” Antiquity 71: 518–524. Kintigh, K. (2006). “The Promise and Challenge of Archaeological Data Interpretation.” American Antiquity 71: 567–578. Kintigh, K. W. and Altschul, J. H. (2010). “Sustaining the Digital Archaeological Record.” Heritage Management 3: 264–274. Kintigh, K. W., Altschul, J. H., Beaudry, M. C., Drennan, R. D., Kinzig, A. P., Kohler, T. A., Limp, W. F., Maschner, H. D. G., Michener, W. K., Pauketat, T. R., Peregrine, P., Sabloff, J. A., Wilkinson, T. J., Wright, H. T., and Zeder, M. A. (2014). “Grand Challenges for Archaeology.” Proceedings of the National Academy of Sciences, 111: 879–880. Knapp, A. B. and Cherry, J. F. (1994). Provenience Studies and Bronze Age Cyprus; Production, Exchange and Politico- Economic Change. Monographs in World Archaeology No. 21 (Madison, WI: Prehistory Press). Lyon, E. A. (1996). A New Deal for Southeastern Archaeology (Tuscaloosa: University of Alabama Press). McManamon, F. P., Kintigh, K. W., and Brin, A. (2010). “Digital Antiquity and the Digital Archaeological Record (tDAR): Broadening Access and Ensuring Long-term Preservation for Digital Archaeological Data.” The CSA Newsletter 23(2): (last accessed May 3, 2016). Matson, F. R. (1982a). “Archaeological Ceramics and the Physical Sciences: Problem Definition and Results.” In: Olin, J. S. and Franklin, A. D. (eds), Archaeological Ceramics (Washington, D.C.: Smithsonian Institution), 19–28. Matson, F. R. (1982b). “Selection and Conservation of Study Materials and of the Data Obtained.” In: Olin, J. S. (ed), Future Directions in Archaeometry (Washington, D.C.: Smithsonian Institution), 128–131. O’Brien, M. J. and Lyman, R. L. (1998). James A. Ford and the Growth of Americanist Archaeology (Columbia: University of Missouri Press) O’Brien, M. J. and Lyman, R. L. (2001). Setting the Agenda for American Archaeology: The National Research Council Archaeological Conferences of 1929, 1932, and 1935 (Tuscaloosa: University of Alabama Press). Perlman, I. and Asaro, F. (1967). “Deduction of Provenience of Pottery from Trace Element Analysis.” University of California Lawrence Radiation Laboratory Report UCRL-17937. Perlman, I. and Asaro, F. (1969). “Pottery Analysis by Neutron Activation.” Archaeometry 11: 21–52. Perlman, I., Asaro, F., and Michel, H. V. (1972). “Nuclear Applications in Art and Archaeology.” Annual Review of Nuclear Science 22: 383–426. Richards, E. E. and Hartley, K. F. (1960). “Spectrographic Analysis of Romano-British Pottery.” Nature 185: 194–196. Sayre, E. V. (1963). “Methods and Applications of Activation Analysis.” Annual Review of Nuclear Science 13: 145–162. Sayre, E. V. (1975). “Brookhaven Procedures for Statistical Analyses of Multivariate Archaeometric Data.” Brookhaven National Laboratory Report No. BNL-21693. Sayre, E. V. (1982). “Preservation and Coordination of Archaeometric Data: The Whole is Greater than the Sum of its Parts.” In: Olin, J. S. (ed), Future Directions in Archaeometry (Washington, D.C.: Smithsonian Institution), 116–120. Sayre, E. V., Dodson, R. W., and Thompson, D. B. (1957). “Neutron Activation Study of Mediterranean Potsherds.” American Journal of Archaeology 61: 35–41.
84 Matthew T. Boulanger Shackley, M. S. (2010). “Is There Reliability and Validity in Portable X-ray Fluorescence Spectrometry (PXRF)?” The SAA Archaeological Record 10(5): 17–20, 44. Snow, D. R. (2010). “Making Legacy Literature and Data Accessible in Archaeology.” In: Frischer, B., Crawford, J. W., and Koller, D. (eds), Making History Interactive: Computer Applications and Quantitative Methods in Archaeology (CAA): Proceedings of the 37th International Conference, Williamsburg, Virginia, United States of America, March 22–26, 2009. BAR International Series S2079 (Oxford: Archaeopress), 350–355. Snow, D. R., Gahegan, M., Giles, C. L., Hirth, K. G., Milner, G. R., Mitra, P., and Wang, J. Z. (2006). “Cybertools and Archaeology.” Science 311: 958–959. Speakman, R. J. and Glascock, M. D. (2007). “Acknowledging Fifty Years of Neutron Activation Analysis in Archaeology.” Archaeometry 49: 179–183. Speakman, R. J. and Shackley, M. S. (2013). “Silo Science and Portable XRF in Archaeology: A Response to Frahm.” Journal of Archaeological Science 40: 1435–1443. Tomlinson, J. E. (1997). “Statistical Evaluation of the Asaro-Perlman Neutron Activation Data on Mycenaean Pottery from the Peloponnese.” Annual of the British School at Athens 92: 139–164. Tomlinson, J. E. (2002). “Comparison of the Results of Neutron Activation Analysis on Ancient Pottery at Two Laboratories: N.C.S.R. ‘Demokritos’ & the University of Manchester.” In: Kilikoglou, V., Hein, A., and Maniatis, Y. (eds), Modern Trends in Scientific Studies on Ancient Ceramics: Papers Presented at the 5th European Meeting on Ancient Ceramics. BAR International Series 1011 (Oxford: Archaeopress), 35–44. Uhen, M. D., Barnosky, A. D., Bills, B., Blois, J., Carrano, M. T., Carrasco, M. A., Erickson, G. M., Eronen, J. T., Fortelius, M., Graham, R. W., Grimm, E. C., O’Leary, M. A., Mast, A., Piel, W. H., Polly, P. D., and Säila, L. K. (2013). “From Card Catalogs to Computers; Databases in Vertebrate Paleontology.” Journal of Vertebrate Paleontology 33: 13–28. Wilson, A. L. (1978). “Elemental Analysis of Pottery in the Study of its Provenance: A Review.” Journal of Archaeological Science 5: 219–236. Yeh, S. J. and Harbottle, G. (1986). “Intercomparison of the Asaro-Perlman and Brookhaven Archaeological Ceramic Analytical Standards.” Journal of Radioanalytical and Nuclear Chemistry 97: 279–291. Yellin, J. (2007). “Instrumental Neutron Activation Based Provenance Studies at the Hebrew University of Jerusalem, with a Case Study on Mycenaean pottery from Cyprus.” Archaeometry 49: 271–288. Yellin, J., Perlman, I., Asaro, F., Michel, H. V. and Mosier, D. F. (1978). “Comparison of Neutron Activation Analysis from the Lawrence Berkeley Laboratory and the Hebrew University.” Archaeometry 20: 95–100.
Pa rt I I I
F OU N DAT IONA L C ON C E P T S
Chapter 7
Ce r amic Raw Mat e ria l s Giuseppe Montana Introduction Raw materials are the essential substances for the manufacturing of any secondary good. In ancient times, natural resources were the only available raw materials, preliminarily transformed, modified, or purified, or used directly without significant modification. The ceramic raw materials, being basically sandy and/or silty clays, are, in effect, comprised within the wider category of “geomaterials.” Geomaterials include all the materials of geological origin, such as rock, sediment, or soil, which could be advantageously utilized for the manufacture of artifacts in their natural state or after man-made processing. Consequently, a geomaterial by itself is somehow associated to a specific manufacturing procedure in order to be converted into a finished artifact. “Raw material” is used to refer to the substance(s) from which archaeological ceramics are manufactured because the term “clay” is problematic for several reasons. The term “clay,” in fact, can refer to a natural incoherent (or weakly coherent) fine-grained material deriving from the in-situ alteration of some types of rocks, or deposited during an erosion cycle (sediment). It can also be used as a descriptive term for small-size particles (less than 0.004 or 0.002 mm depending on the adopted classification) in the granulometric study of sedimentary rocks or soils. Clayey materials primarily contain “clay minerals,” which are mainly originated by the weathering of feldspars and by low temperature hydrothermal alteration of various crystalline rocks. Clay minerals are hydrous silicate minerals belonging to the group of phyllosilicates, and they are distinguished by a “sheetlike structure” with interlayer spaces along the crystallographic c-axis. They are essentially composed of Si, Al, Mg, O and H2O, frequently with appreciable amount of Fe and K. A major characteristic of clayey materials is the developing of plasticity when mixed with a suitable amount of water and hardening when air dried and properly fired. The manufacture of ceramic artifacts from a clay-based paste, known from the Neolithic, can be considered without doubt one of the most significant advances in human technological development. Any ceramic object represents the result of a well-structured production chain which starts with the localization and the exploitation of a suitable raw material and ends with the artisanship and craftsmanship of the potter. All these subsequent operations of shaping, drying, and firing will lead to a good finished product having specific
88 Giuseppe Montana morphological and functional characteristics, showing very different mechanical properties with respect to the original raw material. Selection and preparation of raw materials are important aspects of pottery making, but at the heart of this process is pyrotechnology. At temperatures around 600°C, evaporation of the structural water in the clay minerals triggers their collapse, and subsequent mixing with water fails to render it pliable. In case of the calcite-bearing or calcareous clays and raw materials, at temperatures around 850–900°C (temperatures typically reached even in antiquity), when the decarbonation of calcium carbonate is largely completed and the CO2 is driven off, new crystalline phases (anhydrous calcium aluminosilicates) become stable and grow up. The hardening reaction induced by the presence of carbonates in the paste thus occurs by means of high-T and low-P conditions. Mineralogical and structural transformations at high-T are essentially influenced by composition of the raw clay, its grain-size distribution, kiln temperature, and kiln atmosphere (oxidizing or reducing). As stated by Cultrone and co-workers (2001) concerning high-temperature Ca and Mg silicate formation in carbonate- rich materials (gehlenite, wollastonite, diopside, and anorthite), combined mass transport (viscous flow) and reaction-diffusion processes are involved. The development of reaction textures in consequence of the marked disequilibrium originated by the artificially generated pyrometamorphism could have some implications for understanding ancient ceramic technologies and discriminating raw clays sources. Over the last fifty years the importance of integrating typological classifications of archaeological ceramics with quantitative compositional data has increased. As stated by Tite (2008): “the primary aim of the application of the physical sciences to the study of ancient ceramics is to contribute to the reconstruction of their life-cycle from production through distribution to use, and then to help in the interpretation of this reconstructed life-cycle in terms of the behavior of the people involved.” Since the publication of the pioneering text by Shepard (1956), the conventional analytical techniques strictly belonging to the field of mineralogy and petrography have been widely and productively applied to the study of ceramic artifacts, supplying essential information concerning their composition and production or manufacture technology. Determining mineralogical and chemical resemblances between ceramic pastes could facilitate the identification of specific productive centers. Within a particular territorial and/or chronological context, these analytical studies may assist the reconstruction of the multifaceted empirical knowledge linked to the whole ceramic manufacturing process. The development of a well-established production standard is somehow related to the familiarity of the manufacturer with local geomaterials and pyrotechnical expertise. Therefore, experimental data should always be integrated into studies pertaining to archaeological ceramics. Several analytical methods applied to ceramic artifacts might help to answer some major questions of the archaeologist, confirming or rejecting his hypotheses, which generally have socioeconomic and/or historical implications. In general, the analytical procedures belonging to mineralogy, petrography, and geochemistry are considered to be very important tools for the studies of ceramic and related clayey raw materials. Distribution or “provenance studies” are intended to establish, on the basis of thin-section petrography and/or chemical composition, whether pottery was locally produced or imported. Provenance studies of archaeological ceramics are based on the assumption that pottery from specific locations exhibits compositions at least distinguishable from pottery produced elsewhere. This is because every raw clays source area should be characterized by definite compositional markers. The classification of a ceramic paste and its assignment to
Ceramic Raw Materials 89 a production center can be established more easily when ceramic sherds, kiln wasters (if available), and raw material are studied in combination and put side by side in terms of mineralogical and chemical composition, and critical textural aspects (i.e. packing and size distribution of aplastic grains) are considered and compared. To identify with a satisfactory level of accuracy the source of the clayey raw materials used in a given production center requires an “integrated approach” in which field geology, mineralogy, petrography, and chemistry are uniformly concerned (Figure 7.1). In any case, a good knowledge of the compositional and technological properties of the clayey raw materials available in that particular territory may be a fruitful approach for a more accurate determination of ceramic provenance. It should be considered, however, that in the case of the study of specific kiln sites or wider regional ceramic production and circulation it is important to establish the degree of compositional (mineralogy and chemistry) and/or textural variability in the clayey raw materials of the area. Certainly, this will help in understanding the intraproduction variability which is an important factor for the correct interpretation of the ceramic analytical data. It is thus not surprising that the study of ceramic raw materials has been increasingly regarded in archaeometric ceramic research, in both archaeological and ethnographic case studies, as the best starting point for identifying local paste recipes for pottery diachronically produced in any historical period. The subsequent interpretation of the ceramic “production chain” involves the consideration of further questions relating, for example, to the extent of technological specialization with the choices made in the selection of raw materials in a specific region, as well as mixing or tempering procedures and firing skills with reference to a given class of ceramic objects. It is a significant achievement to be able to identify whether a single clay or a mixture of clayey materials was consciously used to create a ceramic paste with the specific compositional and
CLAY DEPOSITS (POTENTIAL RAW MATERIALS CERAMIC PRODUCTION)
INTEGRATED APPROACH
PRODUCTION CENTERS • CIRCULATION/TRADE
FIELD SURVEY (ETHNOGRAPHY AND TRADITIONAL MANUFACTURES, GEOLOGY)
PETROGRAPHY (MINERALOGICAL AND TEXTURAL MARKERS)
CHEMISTRY (CHEMICAL MARKERS)
PROVENANCE
• IMPORTS • LOCAL MANUFACTURES
END-USE REQUIREMENTS • TEMPERING • MIXING • FIRING CONDITIONS (OXIDISING/REDUCING)
ACHIEVEMENT OF A REPRODUCIBLE STANDARD QUALITY
CROSS-CORRELATION ARCHAEOLOGICAL CERAMIC
Figure 7.1 “Integrated approach” for characterizing and sourcing ceramic raw materials.
90 Giuseppe Montana textural characteristics to produce, after drying and firing, finished objects with the looked- for desired qualities and performance characteristics. Once the “compositional reference groups” have been established, the model of exchange and trade for a given archaeological/chronological context may be very likely resolved, and the questions of where the ceramics were produced and/or where they were distributed and used will be answered clearly.
Mineralogical and Chemical Characteristics The raw materials for manufacturing pottery essentially include: (a) clays or clayey materials, the dominant or matrix material, and (b) temper, that is to say sand size grains (0.06– 2 mm) which can be arbitrarily added to the ceramic paste by the potter in order to control physical and mechanical properties or other desired characteristics in the finished products. Supplementary inorganic natural materials are necessary to impart a slip or a glaze, while the manufacture of bricks and tiles often requires organic additives (e.g. straw). As mentioned in the Introduction to this chapter, the term “clay” has three different connotations: (1) fine- grained and ordinarily loose geological material that develops plasticity when mixed with water; (2) a grain-size classification for the smallest particles (less than 0.002 mm) found in geological sediments; (3) the group of hydrous allumosilicate minerals named “clay minerals,” produced (by weathering or pedogenesis) from preexisting minerals and mostly characterized by a sheet-like crystalline structure. Clay minerals are ubiquitous in continental fluvial and lacustrine deposits, soils, weathering environments of silicate rocks, and marine sediments. Clay deposits are found for the most part on or near the surface of the Earth. In a straightforward categorization concerning the origin of clay deposits it may be recognized: “primary clays” which are formed in situ by the weathering of bedrocks characterized by high-alumina minerals such as feldspars (e.g. granite, arkoses, etc.); “secondary clays” mainly formed by means of fluvial transport and deposition of clay minerals and clay-sized particles produced by eroding pre-existing rocks and soils within a complete sedimentary cycle (Figure 7.2). The majority of natural clays are composed of heterogeneous mineral mixtures that can be roughly subdivided into clay and non-clay minerals, including poorly crystalline (low-range ordered or more simply amorphous) inorganic phases. Sporadically a single clay mineral may be predominant (e.g. kaolinite deposits). Accordingly, clays are ordinarily composed of mixtures of clay minerals and clay-sized or coarser crystals of other minerals such as quartz, feldspars, micas (muscovite and biotite), carbonate (chiefly calcite and dolomite), iron oxides, hydroxides (e.g. hematite and goethite), and a variety of parent bedrock fragments (lithoclasts). All the clay minerals belong to the family of “phyllosilicates” (from Greek φύλλον, leaf) which are distinguished by a composite layered structure leading to a silicon:oxygen ratio of 2:5 giving a flaking appearance. The most characteristic property of several clay minerals is the ability to adsorb water between the layers, resulting in strong repulsive forces and clay expansion. From the structural point of view, clay minerals consist of stacked tetrahedral and octahedral layers, which are kept together by cations, respectively coordinated by Si or
Ceramic Raw Materials 91
(a)
(b)
Figure 7.2 Examples of primary and secondary clays: (a) kaolinite deposits in the crater of Mount Gibele at the volcanic island of Pantelleria (Italy); (b) outcrop of Upper Miocene marine clays in southern Sicily.
92 Giuseppe Montana Al (tetrahedral layer) and by Al, Mg, Fe2+, Fe3+ (octahedral layer). They are classified first into “layer types,” differentiated by the number of tetrahedral and octahedral sheets that have combined, and then into “groups,” based on the type of isomorphic cation substitution that has occurred. Therefore, in the simplest way, clay minerals can be distinguished: (a) two- layer or 1:1 sheet silicate structures, consisting of an octahedral layer linked to a tetrahedral layer, or (b) three-layer or 2:1 sheet silicate structures, consisting of an octahedral layer sandwiched between two tetrahedral layers. The composite layers are stacked along the crystalline c-axis and linked by cations and/or water molecules placed in the interlayer spaces. Sheet silicates ordinarily exhibit a special form of polymorphism (the same chemical composition but diverse structural framework), defined as “polytypism,” which constitutes differences in the stacking of the tetrahedral and octahedral layers. The most common clay minerals belong to: (1) kaolinite and serpentine group (e.g. kaolinite, dickite, nacrite, halloysite, chrysotile, lizardite, antigorite); (2) illite-mica group (e.g. pyrophyllite, talc, illite, muscovite, biotite, phengite, glauconite, celadonite); (3) smectite group (e.g. montmorillonite, beidellite, nontronite, saponite, hectorite); (4) chlorite group (e.g. clinochlore, chamosite, nimite, pennantite). The kaolinite group is the most common of clay minerals with two-layer 1:1 crystalline structure (tetrahedral and octahedral combined layers). The general formula is Al4Si4O10(OH)8 and kaolinite, dickite, and nacrite are the principal polytypes. Halloysite is distinguished by the presence of one water layer in the interlayer structural space. The identification of polytypes, accomplished by means of X-ray powder diffraction (XRPD) methods, enable identification of the formation conditions of the deposit. The three principal minerals in the “serpentine group,” chrysotile, antigorite, and lizardite, have very similar chemical compositions, the general formula incorporating all three members being X6Si4O10(OH)8 (where X = Mg, Fe2+, Ni, Zn, Mn), but significantly different structures which attempt to minimize the mismatch between sheets of SiO4 tetrahedra and the octahedral sheets. The mica group is composed of 47 different mineral species which are all characterized by 2:1 tetrahedral-octahedral-tetrahedral sheet structure with interlayer cations and little or no exchangeable water. The most abundant minerals of the mica group are muscovite KAl2(AlSi3O10)(OH)2 and biotite K(Fe,Mg)3AlSi3O10(OH)2. Both muscovite and biotite are usually characterized by tabular crystals with prominent basal planes, and they are very common minerals in igneous (e.g. granite pegmatite, granodiorite), metamorphic (e.g. slate, phyllite, schist, gneiss) and sedimentary clastic rocks. The name “illite” is a general term for the clay mineral constituent of argillaceous sediments belonging to the mica group, having 2:1 composite layer structure. In the illite structure the substitution of Si4+ by Al3+ produces a net negative charge of about 0.7–1.0 per formula unit, which is equilibrated by the entry of K+ in the interlayer space, where water molecules may also be present. A general formula of illite could therefore be K(Al,Mg,Fe)2(Si,Al)4O10(OH)2. Glauconite is the Fe-rich illite, while celadonite is relatively more Mg-rich. Up to sixteen minerals belong to the so-called smectite group, a family of swelling 2:1 phyllosilicates (expandable along the crystallographic c-axis) having residual surface charge because of the isomorphous substitution in the octahedral layer (Mg2+, Fe2+, Mn2+ may substitute Al3+) and/or in the tetrahedral layer (Al3+ and sporadically Fe3+ may substitute Si4+). These substitutions lead to net negative charges in the clay structure which must be satisfied by the presence of charge-balancing cations within the interlayer. Monovalent cations such as Na+ cause more expansion than divalent cations such as Ca2+. Smectites are referred to as
Ceramic Raw Materials 93 “swelling” clay minerals because they may take up a notable amount of water in the interlayer space of their structure. The most frequent mineral in this group is montmorillonite, with a general chemical formula (Ca,Na)(Al,Mg,Fe)4(Si,Al)8O20(OH)4·nH2O. Other common smectite group minerals are beidellite, nontronite (dioctahedral), Mg-rich saponite, and Li-rich hectorite (trioctahedral). Montmorillonite is able to absorb in a stepwise manner up to four monomolecular sheets of water producing a maximum swelling along the crystallographic c-axis of around 10 Å which in turn leads to a complete basal (001) d-spacing of around 20 Å (thickness variation by almost 100%). Each adsorbed water layer separates the composite T-O-T sheets by almost 3 Å, which is the approximate diameter of a water molecule. The chlorite group is composed of twelve phyllosilicate minerals characterized by a structure consisting of T-O-T layers regularly alternating with a brucite-like octahedral layer of closely packed hydroxyls. The chlorites are primarily found in weakly metamorphosed materials deriving from the alteration of pyroxenes, amphiboles, and micas. They vary greatly in chemical composition owing to substantial cation substitution in both the octahedral and tetrahedral sites. Thus, a rather complicated general formula may be stated: A5-6T4Z18, (where A = Al, Fe2+, Fe3+, Li, Mg, Mn, or Ni; T = Al, Fe3+, Si; Z = O and OH). The most common species in the chlorite group are clinochlore (Mg, Fe2+)5Al(Al,Si3O10)(OH)8 and chamosite (Fe2+, Mg)5Al(Al,Si3O10)(OH)8. Major and trace elements in clay-sized sediments are mostly dependent upon the lithology of parent geomaterials and the complex of processes responsible for their formation. Clay materials vary greatly in their chemical composition. Some of them contain much iron (i.e. red clays), while others are rich in calcium carbonate (i.e. calcareous clays). Predictably, the chemical composition of clays is basically controlled by the relative abundances of clay minerals as well as non-clay minerals. These latter are mostly representative of the silt-and sand-sized detritic fractions (respectively 0.002–0.06 mm and 0.06–2.0 mm). Indeed it follows that a wide range of variation of the concentration of major elements may occur. In Table 7.1 the chemical composition of the “average terrigenous marine clay” (after Clarke, 1924) is reported together with those of some Italian and Greek clays, which have been here considered as representative examples of deposits traditionally used as ceramic raw material in the Mediterranean area. It is possible to highlight some remarkable differences in the bulk chemistry between the various clayey materials concerning especially the most abundant oxides (i.e. SiO2, Al2O3, Fe2O3, CaO). In general, clay materials for ceramic manufacture are characterized by the predominance of SiO2 (generally between 50% and 65%) that can be correlated to the abundance of detritic quartz grains (naturally present in the clay material) and other silicate minerals (e.g. feldspars, mica) or rock fragments (e.g. granite, gneiss, sandstone), which, on the whole, compose the coarser granulometric fractions (silt and sand), in addition to the SiO2 content of the clay minerals composing the finer groundmass. Another relevant compositional differentiation between the various ceramic raw materials is the amount of calcium oxide. In fact, according to the classification made by Tite and Maniatis (1975), the “calcareous clay” (CC) with a concentration of CaO greater than 5% weight can be distinguished from the “non-calcareous clay” (NCC) having CaO concentration lower than 5% weight. Of course, the CaO abundance is strictly associated with the presence of calcite, often due to the presence of calcareous microfauna (particularly in marine clays) and/or to detritic sand-sized fragments of limestone.
1,75 0,75 1,02 2,01 0,68 0,67 1,29 0,89 0,61 1,15
Montana et al., 2011 Grifa et al., 2009 Gliozzo et al., 2008 Gliozzo et al., 2005 De Bonis et al.,2013 De Bonis et al., 2013 Hein et al., 2004b Hein et al., 2004a Clarke, 1924
Na2O
Montana, 2010
Reference
2,38
2,21
7,20
3,10
3,81
2,13
3,86
3,36
2,95
3,09
MgO
SiO2
P2O5
18,89 62,60 0,23
22,09 59,99 0,17
13,31 53,82 0,20
19,96 61,88 0,15
14,75 54,94 0,13
13,26 54,79 0,23
19,30 60,00 2,49
14,96 57,67 0,14
17,45 57,22 0,13
15,71 56,63 0,19
Al2O3
Normalized value vs. L.O.I. n = number of samples analyzed belonging to the same geological context and considered for calculating the average values
Average Marine Terrigenous Clay
Quaternary red clayey alluviums (Crete, Greece) n = 18
Neogene marine clay deposits (Crete, Greece) n = 24
From Upper Cretaceous to Quaternary basinal sediments (Campania region, Italy) n = 7
Miocene–Pleistocene basinal sediments (Campania region, Italy) n = 22
Quaternary terraced fluvial deposits (Foggia, Italy) n = 16
Quaternary alluvial clay (Paola, Italy) n = 4
Middle Pliocene marine clay (Benevento, Italy)
Lower Pleistocene marine clay (Palermo, Italy) n = 9
Quaternary marine clay (Ischia, Italy) n = 2
Description of clay deposits
2,47
2,80
2,34
3,00
2,85
2,60
2,49
2,43
1,77
2,75
K2O
TiO2
2,24 1,39
1,69 1,24
14,24 0,89
1,85 0,93
15,50 0,74
21,55 0,59
4,89 1,03
14,15 0,69
12,43 0,93
14,73 0,71
CaO
0,13
0,15
0,09
0,09
0,12
0,13
0,15
0,07
0,02
0,11
8,09
9,06
7,03
7,75
6,48
3,99
7,40
5,51
6,38
5,46
MnO Fe2O3
Table 7.1 Chemical composition of the “average terrigenous marine clay” (after Clarke, 1924) and of some representative Italian and Greek clays
Ceramic Raw Materials 95 The concentration of Al2O3, as well as those of K2O and Na2O, in general, can be positively correlated to the relative abundance of clay minerals and/or to the quantity of monomineralic granules of feldspars and mica in the coarser granulometric fractions. The bulk concentration of Fe2O3 is controlled by the abundance of colloidal particles of poorly crystalline iron oxides/hydroxides in the clay materials. It is usually greater in the sedimentary deposits where particles with diameter less than 0.002 mm are on the whole predominant on silt and sand. MgO content is often associated with the presence of dolomite or Mg-bearing silicates in the sand fraction, while TiO2 content is related to the presence of rutile and leucoxene among the accessory heavy mineral and also to the abundance of 2:1 clay minerals. Both MnO and P2O5 are normally not abundantly concentrated in clays and are respectively linked to the presence of manganese oxides (e.g. pyrolusite) and calcium phosphate (apatite). Concerning trace elements, the chemical composition of primary and secondary clays is even more closely correlated than major elements to the parent geomaterial from which they have developed by weathering or to the pre-existing rocks or soils eroded and involved in the sedimentary cycle. For example, alluvial clay deposits in a territory characterized by basic volcanic rock outcrops are very likely richer in trace elements, such as Cu, Zn, Ni, Cr, and light rare earth elements LREE (La and Ce), that can be used as geochemical “markers” for a basic environment. Another example: during the weathering process vanadium is incorporated in the structure of clay minerals by replacement of aluminum in tetrahedral or octahedral positions. Therefore, somewhat higher vanadium concentrations should characterize clay deposits where the clay-sized particles are relatively more abundant than the silt ones. The concentration of Co is apparently positively correlated to the content of Fe in the clays because of their positive association in the geochemical cycle in general. High values of Ba can be associated to the content of feldspar minerals of the clay material (the geochemical affinity between K and Ba is well known), while a relatively higher Rb concentration could reflect the greater weights of the finest fractions (clay and silt) in the clay deposit and/or an abundance of mica flakes. Strontium is more often positively correlated to Ca concentration thus acting as marker of sedimentary marine calcite-rich clays. Clay deposits of the upper Tortonian age in Sicily underlying the Messinian evaporitic sequence, particularly, are characterized by more than 1,000 ppm of Sr. Mineralogical and chemical composition of a given clay raw material significantly influence its suitability for producing pottery, bricks, or tiles. For example, iron-rich and “calcareous clays” may be used in the manufacture of red colored and porous ceramic after firing at 850–950°C with oxidizing kiln atmosphere. Clays naturally containing volcanic minerals and rock fragments or artificially tempered with them are, theoretically, more suitable for producing purposely designed cooking ware, fired also at relatively lower temperatures (generally within 750–850°C). In fact, the presence of volcanic temper produces a ceramic paste more effectively resistant to thermal shock than quartz temper, which suffers the alpha–beta phase transition at 573°C that produces a rapid volume change (around 5%). The expansion and contraction of quartz grains due to repeated cooking cycles may produce defects in the ceramic object, such as veining, causing it to break more readily and thus limiting its usefulness. For comparable reasons, clay deposits full of swelling clay minerals (smectites) are seldom used “pure”; however, they are often mixed with poorly plastic clays in order to improve their performances. Smectite-rich clays profusely tempered with coarse sand are also attested to have been used locally in the Mediterranean area for producing bricks and/or tiles.
96 Giuseppe Montana
Relevant Physical Properties The common characteristic of all clay materials, plasticity when mixed with water, derives chiefly from the sheet-like structure of clay minerals and from their minute crystal size. As highlighted above, clay minerals, particularly those belonging to the smectite group (e.g. montmorillonite), are capable of changing their volume in one dimension by absorption or desorption of water. The volume change process is termed “swelling” if water molecules enter into the crystal structure or interlayer space, and “shrinkage” if water is released after dehydration. Moreover, owing to the ionic substitutions at tetrahedral and octahedral level, the surface of any individual clay particle (also indicated as “micelle”) is typically negatively charged, leading to the attraction/absorption of cations and water molecules. According to Van Olphen (1963), a structure of water molecules can be built up around the micelle particle, giving rise to a “double diffuse layer.” A relatively elevated concentration of high valence cations will support a thin “diffuse layer.” Conversely, low cation concentration of the aqueous solution or low valence cations will produce a thicker “diffuse layer.” As a direct consequence, when high valence cations (typically Ca2+ followed by Mg2+ and Al3+) are concentrated in the “diffuse layer,” clay particles can approach each other closely without repulsion because the Van der Waals forces of attraction exceed the repulsive forces. These “attractive” forces lead to the formation of clay particle aggregates which are inclined to precipitate—a phenomenon named “flocculation.” When clay particles are dispersed in seawater the same phenomenon can be observed owing to the high cation concentration of seawater. Important for our discussion of ceramic raw materials and their properties is that clay materials predisposed to flocculation often produce an excessively porous paste which will likely result in unwanted cracking and fracturing of the ceramic artifacts during drying and firing. Worral (1982) explains how the deflocculation of a clay with adsorbed Ca++ can be accomplished by replacing this cation with Na+. For this reason, in modern industrial ceramic production deflocculants (e.g. sodium metasilicate, sodium hexametaphosphate, sodium pyrophosphate), with the ability to sequester Ca2+ and alter the pH of water, are habitually used in order to provide pastes with satisfactory properties for casting and molding. Owing to the presence of a low- valence cation (Na+), clay particles are prevented from aggregating and a more dispersed solution/mixture is obtained. An interesting solution for overcoming the problem of flocculation, attested in several ancient Sicilian ceramic workshops and still used by some contemporary potters, involves the addition of NaCl to freshwater during the preparation of clay paste (or even the direct use of seawater), in order to avoid flocculation of the locally available calcium-rich clay materials (CaO concentration up to 20–25% weight) (Montana, 2011) (Figure 7.3). Another physical property of clays relevant for their use as ceramic raw materials is their particle-size distribution. One of the most common procedures for determining the particle- size distribution of a geomaterial involves separating the loose materials into a coarse (particles of between 2 and 0.06 mm) and a fine (particles < 0.06 mm) fraction using distilled water dispersion in a settling cylinder according to Stokes’ Law. The fine fraction suspension may be further separated into silt (0.06–0.002 mm) and clay (< 0.002 mm) via centrifugation at 2000 rpm and 4000 rpm, respectively. Particle-size distribution has important implications for both technological performance of the raw materials and their archaeometric identification (provenance). Grain size is an important consideration in the selection of raw materials and the preparation of a ceramic paste because the particle-size distribution of the paste affects the performance characteristics
(a)
(b)
(c)
Figure 7.3 Brick and roof tile makers in western Sicily, traditionally using NaCl as a deflocculating agent: (a) and (b) Spina’s Furnace at Gela (Caltanissetta); (c) Martorelli’s Furnace at Racalmuto (Agrigento).
98 Giuseppe Montana of the object during manufacture and, ultimately, its intended use. There are three important ranges of particle sizes each with a distinct role: the “texture fraction” composed by very coarse sand grains (1–2 mm) primarily used to give strength to structural clay objects; the “filler fraction,” composed for the most part of medium and fine sand grains (0.1–0.5 mm), used to control excessive shrinkage and cracking of the ceramic objects during drying and firing; and the “plastic fraction,” grain sizes below 0.06 mm, composed of both silt and clay particles, the latter being solely responsible for working and shaping the ceramic paste. In the manufacture of thin-walled fine ware using the potter’s wheel, the “plastic fraction” should predominate in order to facilitate the fluid movement of the paste. A ceramic raw material appropriate for the manufacture of coarse wares for daily use (e.g. bowls, cups, jars, basins, jugs) might be composed of: 0–10% in the texture fraction, 5–30% in the filler fraction, and 70–90% in the plastic fraction. It is important to remember that many natural clay deposits contain a sand fraction (0.06–2 mm) between 10–15 weight % and a significant part of the fine fraction (up to 85–90 weight %) is composed of coarse silt grains (0.02– 0.06 mm), which act as part of the filler fraction. The relative amount of clay-sized particles (< 0.002 mm) is variable in natural clay deposits, ranging between 35–60 weight %. As a result of their clay minerals content, ceramic raw materials possess some distinguishing characteristics. For example, colloidal clay mineral particles when mixed with water can easily slide over each other, giving rise to the so-called “plasticity” of clayey materials. Plasticity, according to Grim (1953), can be defined as the property of moistened fine- grained earths to be deformed under the application of pressure, with the deformed shape being retained when the deforming pressure is removed. Plasticity is acknowledged to be largely dependent upon water content, grain-size distribution, and the relative abundance of swelling-type clay minerals in the material. Estimation of the plasticity of a ceramic raw material can be determined by means of the empirical measurement of the “Atterberg limits,” according to the standard normative CEN ISO/TS 17892–12:2004, using a “Casagrande apparatus.” Atterberg limits, that is to say the plastic limit (Wp) and the liquid limit (Wl), define the increasing water content boundaries between non-plastic, plastic, and viscous fluid states. The plasticity index (Ip = Wl–Wp) defines the complete range of water content in which the clay material can be worked in the plastic state. The Atterberg limits of a clay material therefore represent its water content at critical stages of fine-grained sediment behavior and as such are mainly used in geotechnical engineering to provide empirical information regarding the reaction of soils to water. However, they have also been applied in the characterization of raw materials used in ceramic production, in order to evaluate material workability when plastic (Marsigli and Dondi, 1997). Slightly to moderately plastic clays exhibit values of the plasticity index (Ip) lower than 18%, while highly plastic clays are characterized by a plasticity index generally greater than 22%.
Experimental Testing: Why and How The compositional and textural characterization of clay raw materials has been increasingly recognized as a successful approach for identifying centers of ceramic production when carried out in parallel with the study of ceramic artifacts from the archaeological excavation of a site or sites in a particular region. Nevertheless, fundamental questions generally arise concerning the interrelation between raw materials composition and their selection criteria,
Ceramic Raw Materials 99 which may depend both on natural diversity and/or accessibility (vicinity to the production site) of available raw materials. At the same time, it is well known that there are inherent limitations in the raw materials themselves, related to their mineralogical and technological properties (e.g. grain size distribution, plasticity, linear shrinkage after drying and firing), which also affect their suitability for ceramic manufacture. Therefore, a compositional comparison between potential ceramic raw materials and the ceramic artifacts themselves should always be undertaken with an appropriate level of care. A detailed study of textural, compositional, and technological properties of local clay deposits, which were potentially employed as raw materials for ceramic production in antiquity, can be illuminating for archaeological/archaeometric case studies interested in understanding regional ceramic production and circulation. The outcomes of such a study include, but are not limited to: (1) construction of a database of local ceramic raw materials, highlighting the important technological aspects of each deposit relevant for ceramic manufacture and differentiated in terms of typological classes (forms), end use, and chronological contexts; (2) identification of pertinent markers for distinguishing local ceramic products from imports, by means of cross-correlation of compositional and textural data; (3) determination of regional production centers based on a comparison with local raw clays; and (4) construction of a prototype model of Territory Information System (TIS), which is considered to be a critical instrument for categorizing the so-called ceramic “reference groups” within the most important archaeological contexts in the studied area. Analysis of raw clays and clayey materials in order to evaluate their suitability for ceramic production will commonly include granulometric analysis, chemical analysis such as X-ray fluorescence spectrometry (XRF), inductively coupled plasma-optical emission spectrometry (ICP-OES), inductively coupled plasma-mass spectrometry (ICP-MS), instrumental neutron activation analysis (INAA), mineralogical analysis, X-ray diffraction (XRD), petrography/petrographic microscopy, and empirical tests for determining key properties, such as plasticity. Investigating random variation of compositional and textural properties of natural clays facilitates the constitutive modeling of the behavior of ceramic raw materials. Clay minerals characterization (type and quantity) and measurement of grain-size distribution will allow a comparative evaluation of the features influencing performance and failure of a ceramic raw material. After separation of the sediment fraction smaller than 0.002 mm, the semiquantitative composition of the clay minerals may be obtained by XRD analyses of the material. Moreover, experimental firings performed on test-pieces (clay briquettes) at various final temperatures may be carried out. Briquettes can be manufactured specifically starting from the collected clays (adding an appropriate amount of water until obtaining a workable paste at the plastic state) with the help of a wooden mold. These empirical experiments allow the researcher to simulate the ceramic end-products obtainable from a given raw material both in its natural state and according to various paste preparation protocols involving artificial mixing and tempering of raw materials. Linear shrinkage and color changes after firing of different raw clays can be directly measured and comparatively evaluated (Plate 1). Experimental test-pieces fired at higher temperature might also be used for obtaining “thin sections” for the petrographic study under the polarizing microscope. The “microscopic fabrics” of the raw clays (hypothetical ceramic raw material), fired at their natural state or after mixing and/or tempering experiments, could be profitably compared to the analogous results showed by the archaeological ceramic artifacts in order to individuate the clay source and appreciate any possible “receipt” modification attributable to specific end-use requirements or diachronic technological evolution.
100 Giuseppe Montana
References CEN ISO/TS 17892-12:2004. Geotechnical Investigation and Testing—Laboratory Testing of Soil—Part 12: Determination of Atterberg Limits. Clarke, F. W. (1924). “The Data of Geochemistry.” US Geological Survey Bulletin 770 (Washington): 785. Cultrone, G., Rodriguez-Navarro, C., Sebastian, E., Cazalla, O., and De La Torre, M. J. (2001). “Carbonate and Silicate Phase Reactions during Ceramic Firing.” European Journal of Mineralogy 13(3): 621–634. De Bonis, A., Grifa, C., Cultrone, G., De Vita, P., Langella, A., and Morra, V. (2013). “Raw Materials for Archaeological Pottery from the Campania Region of Italy: A Petrophysical Characterization.” Geoarchaeology: An International Journal 28: 478–503. Gliozzo, E., Fortina, C., Memmi Turbanti, I., Turchiano, M., and Volpe, G. (2005) “Cooking and Painted Ware from San Giusto (Lucera, Foggia): The Production Cycle, from the Supply of Raw Materials to the Commercialization of Products.” Archaeometry 47(1): 13–29. Gliozzo, E., Vivacqua, P., and Memmi Turbanti, I. (2008). “Integrating Archaeology, Archaeometry and Geology: Local Production Technology and Imports at Paola” (Cosenza, Southern Italy). Journal of Archaeological Science 35: 1074–1089. Grifa, C., Cultrone, G., Langella, A., Marcurio, M., De Bonis, A., Sebastian, E., Morra, V. (2009). “Ceramic Replicas of Archaeological Artefacts in Benevento Area (Italy): Petrophysical Changes Induced by Different Proportions of Clays and Temper.” Applied Clay Science 46: 231–240. Grim, R. E. (1953). Clay Mineralogy (New York: McGraw-Hill Book Co). Hein A., Day P. M., Cau Ontiveros, M. A., and Kilikoglou, V. (2004a). “Red Clays from Central and Eastern Crete: Geochemical and Mineralogical Properties in View of Provenance Studies on Ancient Ceramics.” Applied Clay Science 24: 245–255. Hein, A., Day, P. M., Quinn P. S., and Kilikoglou, V. (2004b). “The Geochemical Diversity of Neogene Clay Deposits in Crete and Its Implications for Provenance Studies of Minoan Pottery.” Archaeometry 46(3): 357–384. Marsigli, M. and Dondi, M. (1997). “Plasticità delle argille italiane per laterizi e previsione del loro comportamento in foggiatura.” L’industria dei laterizi, Gruppo editoriale Faenza Editrice (Faenza, Italy) 46: 214–222. Montana, G. (2010). La prima serie di analisi mineralogiche sulle anfore di Ischia. In: Olcese, G., Le Anfore greco italiche di Ischia: archeologia ed archeometria (Roma: Edizioni Quasar), 199–202. Montana, G. (ed) (2011). Le “argille ceramiche” della Sicilia occidentale e centrale. (Enna: Ilion Books). Montana, G., Cau Ontiveros, M. A., Polito A. M., and Azzaro, E. (2011). “Characterisation of Clayey Raw Materials for Ceramic Manufacture in Ancient Sicily.” Applied Clay Science 53(3): 476–488. Shepard, A. O. (1956). Ceramics for the Archaeologist (Washington, D.C.: Carnegie Institution of Washington). Tite, M. S. (2008). “Ceramic Production, Provenance and Use—a Review.” Archaeometry 50: 216–231. Tite, M. S. and Maniatis, Y. (1975). “Scanning Electron Microscopy of Fired Calcareous Clays.” Transactions and Journal of the British Ceramic Society 74: 19–22. Van Olphen, H. (1963). An Introduction to Clay Colloidal Chemistry (New York: Wiley). Worral, W. E. (1982). Ceramic Raw Materials. The Institute of Ceramics Series (Oxford: Pergamon Press).
Chapter 8
Ceramic Manu fac t u re The chaîne opératoire Approach Valentine Roux Introduction The term chaîne opératoire was coined almost fifty years ago by Leroi-Gourhan while seeking to characterize techniques: “Techniques are at the same time gestures and tools, organized in sequence by a true syntax which gives the operational series both their stability and their flexibility” (Leroi-Gourhan, 1964: 164). Rooted in French cultural ethnography which promoted the cultural dimension of material culture (Mauss, 1947; Maget, 1953; Haudricourt, 1987), the study of technical facts according to this concept gave birth to numerous studies in the domain of the anthropology of techniques under the leadership of R. Creswell, whose team was gathered around the journal Techniques & Culture. The definition of the chaîne opératoire used to be largely debated by both anthropologists and prehistorians (Karlin et al., 1986; Balfet, 1991). Nowadays, depending on authors, it describes the whole manufacturing process—defined as a series of operations that transform raw material into finished product, either consumption object or tool (Creswell, 1976: 13)—or part of the manufacturing process, which is then decomposed into several chaînes opératoires (Lemonnier, 1983). In archaeology, the worldwide success of the chaîne opératoire owes mainly to the results obtained in the 1980s and early 1990s by studies in the anthropology of techniques and ethnoarchaeology (e.g. Balfet, 1981; Creswell, 1983, 1996; Arnold, 1985; Longacre, 1991; Gallay, 1992; Gosselain, 1992; Dietler and Herbich, 1994; Lemonnier, 1992, 1993; Latour and Lemonnier, 1994; Sigaut, 1994). Although each of these studies focused on a different culture area, their conclusions can be distilled into a universal observation or general trend: there is a strong correlation between technological behaviors and social groups. Individuals tend to do as their group does, thus maintaining the diversity of cultural traits within their social group and making visible their social borders. Applied to archaeological assemblages, this correlation opened new avenues of research since it enabled researchers to view objects from a different perspective, as part of a social and technological process and therefore as significant of the social groups behind them. This chapter first addresses the cogency of the social dimension of the techniques since it represents the cornerstone of the technological approach. Description of the main features
102 Valentine Roux of the ceramic chaînes opératoires follows, completed by the procedure for highlighting diagnostic attributes. The methodology for classifying archaeological assemblage according to the chaîne opératoire approach is then precised. We conclude by mentioning the different domains of interpretation rendered possible by the analysis of ancient chaînes opératoires.
The chaîne opératoire and Its Social Dimension Clarifying the link between chaînes opératoires and social groups requires explaining why it is that techniques have an identity dimension. Explanation is to be found in the learning and transmission processes studied in the fields of experimental psychology and social anthropology. The results of these studies indicate that the mastery of technical practices corresponds to a process of inheritance which occurs both at the individual (the learning process) and collective (the transmission process) level according to both “bio-behavioral” and “anthropological” rules. At the individual level, it has been demonstrated that learning involves a tutor and a model (Reed and Bril, 1996; Bril, 2002a, 2002b). When the individual explores the task to be learned, he does it through the observation of a model which corresponds to the tutor’s way of doing it. The tutor’s role is to educate the learner’s attention and to orient his exploratory activities toward the model and intended outcome to be achieved. This guidance not only facilitates the learning process, it also participates directly in the reproduction of the task. This guidance is the key to cultural transmission. At the end of the learning process, the learned skills are literally “embodied” (Dobres, 2000; Gosselain, 2000; Ingold, 2001): (a) the learner has progressively acquired the perceptual-motor and cognitive skills necessary, proposed and demonstrated by the tutor, for making objects, but only these skills; (b) in the course of the apprenticeship, the learner has built up specific mental representations about the way to make objects. As a consequence, it will be difficult for the learner to perceive and manufacture objects in a different way than the one he/she learned, by virtue of the “bio- behavioral rules” which require a subject to learn not by innovating but by reproducing a model, therefore acting as true “fixing agent” of the cultural model. At the collective level, transmission occurs within groups made up of individuals linked by social ties. These ties determine the social perimeter into which ways of doing are transmitted, and by the same token, the boundaries beyond which there are other networks with other ways of doing (e.g. Mahias, 1993; Kramer, 1997; Stark, 1998; Bowser, 2000; Livingstone-Smith, 2000; Gosselain, 2002, 2008; Degoy, 2008). The “anthropological” rules which determine the transmission network of skills are the same as those that maintain the cohesion of the group and facilitate its reproduction. The nature and structure of the groups within which a “way of doing” is transmitted are highly variable. They can correspond to a band, a clan, a faction, a caste, a subcaste, a lineage, a professional community, an ethnic group, an ethno-linguistic group, a population, or a gender (exclusive transmission of a “way of doing” among women or men). In addition, the nature and structure of a group can change over time and the social boundaries be redefined. Thus a “way of doing” can be used at a time t by a small social group, and at a time t+1 by a larger social group, the social
Ceramic Manufacture 103 boundary delimited by the transmission network having evolved in the course of time. Moreover, the same community can include several transmission networks depending on the type of objects. For example, the manufacture of culinary pots may be the responsibility of women at the household level of production, whereas storage jars may be the responsibility of a few specialized men at the regional level of production. As a result, the historical dynamics at work will vary depending on types of objects, creating phenomena of arrhythmia (Perlès, 2013). But whatever the social boundaries, learning and transmission processes explain that technical traditions overlap with them: technology is always transmitted through tutors selected within one’s social group. The immediate archaeological implications are: (a) the chaînes opératoires are inherited ways of doing, that is technical traditions transmitted through successive generations; (b) the distribution of technical traditions indicate the social perimeters into which they have been learned and transmitted; (c) changes within technical traditions are the expression of culture histories and the factors affecting them (Shennan, 2013); (d) technical traditions situated in space and time can be powerful chrono- cultural markers, in particular when stylistic features are not significant (Roux et al., 2011; Ard, 2013); and (e) the combined study of technical processes and objects is necessary for an anthropological understanding of archaeological assemblages.
Describing the Ceramic chaîne opératoire Discussion of the ceramic chaînes opératoires involves two levels of description. The first describes the main actions which organize the successive transformations of the raw material into a finished product. They are: collecting and preparing the raw materials, fashioning, finishing, surface treatment, decoration, and firing. The order of these actions is universal given the properties of the material and the objective sought (making vessels). The second level describes the chaînes opératoires involved in each of these actions. It is at this level that technological behaviors or activities are highly variable. This diversity is determined by both cultural and functional factors. Preparing the raw material into a ceramic paste includes: drying, pounding, sorting, hydrating, adding temper, and wedging. Each of these behaviors is dictated by the potter’s natural and cultural environment, the inherent properties of the raw material, in terms of its qualities for making the desired finished products, the modifications of the raw material necessary to achieve the sought-after qualities of the finished products, and the potter’s cultural tradition. The chaîne opératoire related to the fashioning stage includes a series of operations which transform the clay paste into a hollow form and can be described in terms of technique, methods, gestures, and tools. Some important definitions are given below. Method: orderly set of functional operations undertaken to obtain the desired shape, starting from the raw material. It comprises phases, stages, and operations, each of which can be achieved through different techniques. There are three main forming phases: fashioning of the body (lower part, upper part), of the orifice (neck and rim), and of the base.
104 Valentine Roux The fashioning of the body can be divided into two stages; the forming of the roughout and of the preform. Roughout: hollow form which does not present the final geometrical characteristics of the container. A roughout is obtained by thinning operations. Preform: container with its final geometrical characteristics but whose surface has not been (or will not be) subjected to finishing techniques. A preform is obtained by shaping a roughout.
Technique: physical modalities according to which clay is fashioned. These modalities can be described on the basis of the following parameters:
(a) the source of energy (muscular energy vs. rotative kinetic energy); (b) the clay mass onto which the pressures are exerted (homogeneous vs. heterogeneous); (c) the type of force (pressure vs. percussion); (d) the type of pressure (discontinuous vs. continuous); (e) the degree of hygrometry of the clay paste (humid vs. leather hard vs. dry)
The two main families of fashioning techniques are distinguished by the source of energy involved: techniques not using rotative kinetic energy (RKE) and those that do. There are eight roughing-out techniques that do not use RKE. They are further differentiated as techniques which act on assembled elements (the coiling technique by pinching, crushing, and drawing, and the slab technique) and those that act on a clay mass (modeling by pinching and drawing, hammering, and molding). There are seven preforming techniques which do not use RKE. They are distributed between those that act on wet clay (scraping, preforming with continuous pressures, beating) and those that act on leather hard clay (shaving, “repoussage,” paddling, hammering). The roughing-out technique that uses RKE is wheel throwing. There are four preforming techniques that rely on RKE: wheel throwing, wheel coiling, wheel molding, and turning (shaving leather hard clay paste with RKE). In total, there are nine roughing-out and eleven preforming techniques which describe the ceramic fashioning process. These techniques are implemented according to methods, gestures, and tools whose description accounts for the diversity of the fashioning chaînes opératoires (Figure 8.1). The finishing techniques are achieved after the preforming stage and can act on wet (smoothing) or leather hard clay (brushing, smoothing on leather hard clay). Surface treatments transform the superficial state of the vessel and involve either rubbing the vessel (softening, burnishing/polishing, shining), or coating it (slips, glazes, organic materials, graphite, silica, carbon). The three types of decorative techniques are distinguished by dimensionality: low relief or one-dimensional decoration (painting); negative relief or recessed decor (impressed—rolled, simple, pivoting, embossed; paddled; incised—simple, pivoting, scratching, carving; or excised—excised, pierced); and two-dimensional or highrelief decors (applied elements or modeling). Firing is the final step in the manufacturing sequence. It is a major one since it is at this stage that the vessels are gaining their final physicochemical properties. The latter depend not only on the clay properties, but also on the firing parameters which include temperature, heating rate, time of exposure, and firing atmosphere. The firing techniques are distributed
Ceramic Manufacture 105
ROUGHING OUT With RKE
Without RKE on a clay mass
on assembled elements Coils
Modeling
Slabs
• pinching • crushing • drawing
Moldling
on a clay mass Hammering
Wheel throwing
• pinching • drawing
PREFORMING Without RKE Wet clay Pressure
Percussion
Continuous scraping
Beating
With RKE
Leather hard clay Pressure
Percussion
“Repoussage” Padding shaving hammering
Wheel throwing
Wet clay
Leather hard clay
Pressure
Pressure
Wheel coiling
Wheel molding
Turning
FINISHING
Figure 8.1 Classification chart of roughing out and preforming techniques. between two main families: those where the vessels are in contact with the fuel (open firings, walled firings) and those where they are not (kilns).
Identifying the Ceramic chaîne opératoire The technological reading of the clay pastes implies a comparison between (a) the final structural state of the clay material, characterized by petrofabrics, and (b) its initial structural state (the raw material), characterized by petrofacies, in order thereafter to unravel the technical process of transformation of the raw material (pounding, hydrating, adding temper, wedging, forming, firing). The interpretation of the initial structural state calls upon general reference data of geological facies and local paleogeographic data as well as physics of materials and experimental data in order to understand the structural transformations of the clay paste (Roux in coll. with Courty, 2016). The identification of the manufacturing process is a difficult exercise in the sense that each gesture produces features which can obliterate the previous features and that surface features are polysemic: not only the same surface features can be obtained by different techniques, but
106 Valentine Roux (a)
(b)
(c)
cm 3
0
(e)
(d)
cm 3
(f)
0
0
0
3 cm
(g)
3 cm
(i)
(h)
cm 3
0
0
(j)
(k)
0
3 cm
(l)
3 cm
Figure 8.2 Diagnostic features taken into account for reconstructing an Early Bronze Age chaîne opératoire from the site of Tell Arqa (Lebanon): (a) print on external base indicating a clay disk laid on a basalt support;(b) concentric over-thicknesses related to a coil laid on the clay disk; (c) view of the coil laid on the clay disk; (d) digital thinning prints at the junction base/body; (e, f) bumpy wall and fissures indicating discontinuous pressures on assembled elements; (g) oblique fissure indicating oblique junction of coils; (h) internal wall showing that the neck was made after the body and finished with a rotary movement; (i) external wall combed after fashioning the neck, while humid; (j) crisscross humid combing; (k) digital depressions indicating the hand support against the internal wall while combing; (l) regularizing the junction bottom/body when leather hard and while the pot was drying upside down.
Ceramic Manufacture 107 also the same technique can produce different surface features. This explains why ceramic chaîne opératoire analysis has taken a longer time to develop than lithic technology, whereas its basis was elaborated as soon as the 1960s (Franken, 1970, 1978; Rye and Evans, 1976; Van der Leeuw, 1977; Rye, 1977, 1981). Since the 1990s research has been conducted, calling upon both ethnographic and experimental data and considering both surface features and microfabrics (e.g. Pierret and Moran, 1996; Livingstone-Smith et al., 2005) (Figure 8.2). The principle is the one of “controlled analogy.” Attributes considered to be significant indicators of particular techniques are those whose formation has been explained and the univocal character of which has been demonstrated. For this purpose, experiments are carried out according to a protocol where only one parameter varies at a time. It is then possible to unravel the mechanisms explaining the formation of the attributes and to assess their diagnostic value. As a general rule, given the often polysemic character of the attributes, it is important to combine different scales of observation, from the naked eye down to the microscope (Roux and Courty, 1998). It is also important to combine different analytical tools (e.g. thin-section, X-ray analysis; Pierret et al., 1996; Berg, 2008).
Classifying Ceramic Assemblages According to the Concept of the chaîne opératoire Highlighting the ancient ceramic traditions constituent of archaeological assemblages requires not only deciphering the manufacturing process involved in the making of the ceramics, but also classifying the assemblages according to the chaîne opératoire approach. This is an original procedure implying a hierarchical classification including three successive sortings:
(1) The first sorting is by technical groups: they are defined by the manufacturing process as expressed by both the microfabrics and the surface features present on the inner and outer walls of the vessels (sherds or full vessels). (2) The second sorting is by technopetrographic groups; that is, by petrographic group within each technical group: it is done by reference to the classification of the petrofacies present on the site. Once the catalogue of these petrofacies is achieved, the sherds belonging to each technical groups are examined in order to identify the class of petrofacies they belong to and to characterize their petrofabrics in terms of technological transformation undergone by the raw material. It is at this stage that the modalities for preparing the clay paste are studied and the ensemble of the chaîne opératoire restituted, from the collection of the raw material to the firing. (3) The third sorting is by technomorphological and stylistic groups, that is by morphological and stylistic types within each technopetrographic group. It is at this stage that the functional categories of vessels made according to each technopetrographic group are characterized.
These successive and embedded sortings are meant to characterize the different chaînes opératoires present in the assemblage (the technopetrographic groups) and to link them to
108 Valentine Roux the intention of the potter (the finished products). Results can be visualized with the help of technostylistic trees of a dendrogram type (Figure 8.3). They offer a synoptic view of the different chaînes opératoires present in the assemblage and the finished products they are implemented for. They also allow us to discuss the nature of the technostylistic variability of the assemblage, whether functional or cultural, whether simple or complex in terms of sociological composition. The functional variability of the chaînes opératoires can be established when the function of the vessels determines the variability of the chaînes opératoires. When it does not, this is cultural by default. As an example, when a technopetrographic group is associated with a unique type of pot (e.g. cooking pot) and when this function explains the differences in the chaînes opératoires of these vessels and others in the assemblage, then we are in a situation where the variability can be interpreted in functional terms, as opposed to variability created by social or cultural borders. The sociological complexity underlying the variability of an assemblage can be established depending on its technopetrographic homogeneity or heterogeneity (Roux and Courty,
MOLDED
Scraped outside
COILED
Smoothed outside
GROUP PETRO-A
Scraped outside
Smoothed outside
Slipped outside
GROUP PETRO-B Paddled
Not paddled
GROUP PETRO-A
Burnished
Not burnished
GROUP PETRO-C
Figure 8.3 Example of technostylistic trees. The tree on the left gathers molded ceramics made up with same clay materials. The preforming techniques vary depending on function of ceramics (functional variability). The tree on the right gathers coiled ceramics whose preforming and finishing techniques co-vary with clay sources and relate to different functional categories (functional variability). Now the molding and the coiling techniques apply to the same functional categories, signaling therefore two technical traditions corresponding to two sociological groups.
Ceramic Manufacture 109 2007). Homogeneous assemblages are characterized by homogeneous technopetrographic groups with either low or high sociocultural variability. Homogeneous assemblages with a low sociocultural variability are characterized by only one technical tradition and the use of local clay sources and describe sites occupied by a homogeneous social group, which is a single group sharing the same way of doing. At the regional scale, the juxtaposition of simple homogeneous assemblages expresses the sociological mosaic of a region. Homogeneous assemblages with a high sociocultural variability are characterized by a few technological traditions and the use of one or several clay sources located in the neighborhoods of the site. They reveal sites with multiple social components whose sociological interpretation will depend on the petrographic, quantitative, and contextual data (e.g. urban, port, colonized, economic exploitation sites). Heterogeneous assemblages are made up of n technological traditions characterized by heterogeneous petrographic groups with either low or strong variability revealing a wide variety of clay sources distributed in the region (low variability) or even beyond in the macro-region (strong variability). Heterogeneous assemblages signal the presence of consumers originating from a wide regional area. The functional interpretation of the site will depend on petrographic, quantitative, and contextual data (consumer sites importing vessels from different places, marketplaces, gathering place including aggregation, pilgrimage, ceremonial sites, etc.).
Interpreting the chaînes opératoires Once the functional and sociological variability of ceramic assemblages is characterized, each chaîne opératoire can be studied from a socioeconomic, historical, and evolutionary perspective. On the synchronic axis, issues relate to the production, distribution, and circulation of ceramic vessels. They can be dealt with at the scale of the site, but it is at the macro-regional scale that an overview of the spatial distribution of technical traditions will emerge, benefiting indispensably from the comparative perspectives of multiple site analysis. Issues for modalities of production are restricted to homogeneous assemblages (those occupied by single or multiple groups of producers); issues for modalities of distribution and circulation of objects apply to both homogeneous and heterogeneous ones (those occupied by consumers exclusively). Distribution (direct or indirect) relates to the acquisition of the vessels within both social and economic frameworks. Circulation relates to the movement of the vessels in the course of their use by consumers. Modalities of distribution and circulation can be understood using two elementary mechanisms: first, that potters have an inherited way of doing, and manufacture vessels in response to a demand from all or part of their social group; and secondly, that the movement of containers combined with ways of doing, quantities, and morpho-functional types generates, at the regional scale, spatial distribution specific to the cultural components in operation (Gallay, 2007). The operation of these two mechanisms gives rise to three main zones or interaction spheres: central, peripheral, and remote. The first one designates both production and distribution zone of a particular ware or class of vessel. The latter two designate regions or social spheres into which ceramics are distributed or have circulated along with their consumers. The distinction can be made on
110 Valentine Roux the basis of the quantities of vessels, their form or type, and the recurrence of their presence through time. Non-recurrent anecdotal quantities of exogenous ceramic traditions and/or technologies indicate the circulation of containers. On the diachronic axis, the chaîne opératoire approach addresses the evolution of traditions and technologies over time, and by the same token the history of the social groups, by virtue of the transmission mechanisms that allow for an anthropological link between vessels. Both the chaînes opératoires and the finished product are considered, given that the dynamics of historical change may affect them differently depending upon their nature and context of production. These differential dynamics are the privileged witness of endogenous or exogenous evolutionary phenomena in relationship with both the producers and consumers. In concrete terms, the issue is first to identify patterns of cultural descent in the chaîne opératoire in order to establish whether there is filiation between the ceramic assemblages (Haudricourt, 1987; Creswell, 1996; Manem, 2008). When traditions are linked by inherited technical gestures, there is historical continuity. On the other hand, when traditions are not linked by inherited technical gestures, it indicates that social groups are not interconnected, and therefore there is potential for the emergence/expansion of new groups and/or the disappearance of previous groups (Roux, 2013). The next issue is to approach the historical dynamics behind the emergence of new technical facts. The complexity of “innovation” leads us to believe that it is not possible to order the different factors at work and, therefore, that the dynamic approach is probably most appropriate (Roux, 2003). Secondly, the issue is to examine the evolutionary forces underlying the historical dynamics. They are two categories of forces: those underlying the order of development of techniques in relationship with “the technical trend” (“la tendance,” thus named by Leroi-Gourhan, 1964), and those specifying the conditions for change. The former explains the way techniques evolved, with a general trend toward lower energy expenditure (Simondon, 1958; Leroi-Gourhan, 1964; Boëda, 2013). The latter relate to the context in which historical scenarios occur. For example, social mutations have been shown to be determinant for technological leaps; the diffusion of a technique has been shown to depend on the sociological structure of the potters, either homogeneous or heterogeneous (Creswell, 1993, 1996; Roux, 2010, 2013). Both categories of forces represent huge areas of research within evolutionary archaeology.
Conclusion Although the chaîne opératoire approach is now more than fifty years old, its operational dimensions have not finished surprising us yet. Not only does this approach enable us to identify ancient technical traditions and technosystems, it also enables us to follow the history of social groups by identifying patterns of cultural descent through the transmission of technical gestures. The huge heuristic character of the chaîne opératoire lies in its inherited character, which makes it both a social and a transmissional indicator. Its epistemological strength is its grounding in empirical data. Its ambition joins with evolutionary archaeology in seeking to highlight the general forces behind changes (Shennan, 2013). This
Ceramic Manufacture 111 qualitative approach, resulting in a measure of the phenomena under study, has thus a rich future ahead.
References Ard, V. (2013). “Ceramic Traditions and Cultural Identities: West- Central France during the Late Neolithic II Period (c. 3400–2900 cal. BC).” Oxford Journal of Archaeology 32(4): 367–389. Arnold, D. E. (1985). Ceramic Theory and Cultural Process (Cambridge: Cambridge University Press). Balfet, H. (1981). Production and Distribution of Pottery in the Maghreb. In Production and Distribution : A Ceramic Viewpoint (Oxford: B.A.R). Balfet, H. (1991). Observer l’action technique: des chaînes opératoires, pour quoi faire? (Paris: Editions du CNRS). Berg, I. (2008). “Looking through Pots: Recent Advances in Ceramic X-Radiography.” Journal of Archaeological Science 35: 1177–1188. Boëda, E. (2013). Techno-logique & technologie: une paléo-histoire des objets lithiques tranchants (s.l.: Archéo-éditions.com). Bowser, B. J. (2000). “From Pottery to Politics: An Ethnoarchaeological Study of Political Factionalism, Ethnicity, and Domestic Pottery Style in the Ecuadorian Amazon.” Journal of Archaeological Method and Theory 7(3): 219–248. Bril, B. (2002a). “Apprentissage et contexte.” Intellectica 2(35): 251–268. Bril, B. (2002b). “L’apprentissage de gestes techniques: ordre de contraintes et variations culturelles.” In: Bril, B. and Roux, V. (eds), Le geste technique. Réflexions méthodologiques et anthropologiques. Technologies/Idéologies/Pratiques (Ramonville Saint- Agne: Editions érès), 113–150. Creswell, R. (1976). “Techniques et culture, les bases d’un programme de travail.” Techniques et culture 1: 7–59. Creswell, R. (1983). “Transferts de techniques et chaînes opératoires.” Techniques et Culture 2: 143–164. Creswell, R. (1993). “Of Mills and Waterwheels.” In: Lemonnier, P. (ed), Technological Choices: Transformation in Material Cultures since the Neolithic (London and New York: Routledge), 181–213. Creswell, R. (1996). Prométhée ou Pandore? Propos de technologie culturelle (Paris: Editions Kimé). Degoy, L. (2008). “Technical Traditions and Cultural Identity: An Ethnoarchaeological Study of Andhra Pradesh Potters.” In: Cultural Transmission and Material Culture. Breaking Down Boundaries (Tucson: University of Arizona Press), 199–222. Dietler, M. and Herbich, I. (1994). “Ceramics and Ethnic Identity. Ethnoarchaeological Observations on the Distribution of Pottery Styles and the Relationship between the Social Contexts of Production and Consumption.” In: Courtin, J. and Binder, D. (eds), Terre cuite et Société. La céramique, document technique, économique, culturel. XIVe Rencontres Internationales d’Archéologie et d’Histoire d’Antibes (Juan-les-Pins: Editions APDCA), 459–472. Dobres, M. A. (2000). Technology and Social Agency (Oxford: Blackwell Publishers). Franken, H. J. (1970). “A New Approach to the Study of Pottery from Excavations.” Archeologia 21: 1–8.
112 Valentine Roux Franken, H. J. (1978). “The Analysis of Ancient Methods of Pot-Making.” Acta Praehistorica et Archaeologica Berlin (9/10): 77–78. Gallay, A. (1992). “A propos de la céramique actuelle du delta intérieur du Niger (Mali): approche ethnoarchéologique et règles transculturelles.” In: Gallay, A., Audouze, F., and V. Roux (eds), Ethnoarchéologie: Justification, Problèmes, Limites (Juan-les-Pins: APDCA), 67–90. Gallay, A. (2007). “The Decorated Marriage Jars of the Inner Delta of the Niger (Mali): Essay of Archaeological Demarcation of an Ethnic Territory.” The Arkeotek Journal (), 1(1). Gosselain, O. (1992). “Technology and Style: Potters and Pottery among Bafia of Cameroon.” Man 27: 559–586. Gosselain, O. (2000). “Materializing Identities: An African Perspective.” Journal of Archaeological Method and Theory 7(3): 187–217. Gosselain, O. (2002). Poteries du Cameroun méridional styles techniques et rapports à l’identité (Paris: CNRS Editions). Gosselain, O. (2008). “Mother Bella Was Not a Bella. Inherited and Transformed Traditions in Southwestern Niger.” In: Stark, M., Bower, B., and Horne, L. (eds), Cultural Transmission and Material Culture. Breaking Down Boundaries (Tucson: University of Arizona Press), 150–177. Haudricourt, A. G. (1987). La technologie science humaine. Recherches d’histoire et d’ethnologie des techniques (Paris: Editions de la Maison des Sciences de l’Homme). Ingold, T. (2001). “Beyond Art and Technology: the Anthropology of Skill.” In: Anthropological Perspective on Technology (Albuquerque: University of New Mexico Press), 17–32. Karlin, C., Pelegrin, J., and Bodu, P. (1986). “Processus techniques et chaînes opératoires: un outil pour le Préhistorien.” Bulletin de la Société Préhistorique Française 83(3): 66–67. Kramer, C. (1997). Pottery in Rajasthan. Ethnoarchaeology of Two Indian Cities. (Washington, D.C.: Smithsonian Institution Press). Latour, B. and Lemonnier, P. (1994). De la préhistoire aux missiles balistiques. L’intelligence sociale des techniques (Paris: Editions La découverte). Lemonnier, P. (1983). “L’étude des systèmes techniques, une urgence en technologie culturelle.” Techniques et Culture 1: 11–34. Lemonnier, P. (1992). Elements for an Anthropology of Technology (Ann Arbor: University of Michigan). Lemonnier, P. (1993). Technological Choices: Transformation in Material Cultures since the Neolithic (London and New York: Routledge). Leroi-Gourhan, A. (1964). Le Geste et la Parole. Technique et Langage (Paris: Albin Michel). Livingstone Smith, A. (2000). “Processing Clay for Pottery in Northern Cameroon: Social and Technical Requirements.” Archaeometry 42(1): 21–42. Livingstone-Smith, A., Bosquet, D., and Martineau, R. (eds) (2005). Pottery Manufacturing Processes: Reconstitution and Interpretation (Oxford: Archaeopress). Longacre, W. (1991). Ceramic Ethnoarchaeology (Tucson: University of Arizona Press). Maget, M. (1953). Guide d’étude directe des comportements culturels (Paris: Civilisations du Sud). Mahias, M. C. (1993). “Pottery Techniques in India: Technological Variants and Social Choice.” In: Technological Choices: Transformation in Material Cultures since the Neolithic (London and New York: Routledge), 157–180. Manem, S. (2008). Etude des fondements technologiques de la culture des Duffaits (âge du Bronze moyen) (Nanterre: Paris-X). Mauss, M. (1947). Manuel d’ethnographie (Paris: Payot). Perlès, C. (2013). “Tempi of Change: When Soloists Don’t Play Together. Arrhythmia in ‘Continuous’ Change.” Journal of Archaeological Method and Theory 20(2): 281–299.
Ceramic Manufacture 113 Pierret, A. and Moran, C. J. (1996). “Quantification of Orientation of Pore Patterns in X-Ray Images of Deformed Clay.” Microscopy Microanalysis Microstructures 7(5): 421–432. Pierret, A., Moran, C. J., and Bresson, L. M. (1996). “Calibration and Visualization of Wall- Thickness and Porosity Distributions of Ceramics Using X- Radiography and Image Processing.” Journal of Archaeological Science 23: 419–428. Reed, E. S. and Bril, B. (1996). “The Primacy of Action in Development. A Commentary of N. Bernstein.” In: Dexterity and Its Development (Hillsdale, NJ: Erlbaum Associates), 431–451. Roux, V. (2003). “A Dynamic Systems Framework for Studying Technological Change: Application to the Emergence of the Potter’s Wheel in the Southern Levant.’ Journal of Archaeological Method and Theory 10: 1–30. Roux, V. (2010). “Technological Innovations and Developmental Trajectories: Social Factors as Evolutionary Forces.” In: O’Brien, M. J. and Shennan, S. J. (eds), Innovation in Cultural Systems. Contributions from Evolutionary Anthropology (Cambridge, MA, and London: The MIT Press), 217–234. Roux, V. (2013). “Spreading of Innovative Technical Traits and Cumulative Technical Evolution: Continuity or Discontinuity?” Journal of Archaeological Method and Theory 20(2): 312–330. Roux, V. in coll. with Courty, M. A. (2016). Des céramiques et des hommes: Décoder les assemblages archéologiques (Paris: Presses Universitaires de Paris Ouest). Roux, V. and Courty, M. A. (1998). “Identification of Wheel-Fashioning Methods: Technological Analysis of 4th–3rd Millenium BC Oriental Ceramics.” Journal of Archaeological Science 25: 747–763. Roux, V. and Courty, M. A. (2007). “Analyse techno-pétrographique céramique et interprétation fonctionnelle des sites: un exemple d’application dans le Levant Sud Chalcolithique.” In: Recherches en archéométrie: la mesure du passé (Oxford: Archeopress), 153–167. Roux, V., Courty, M. A., Dollfus, G., and Lovell, J. (2011). “A Techno-Petrographic Approach for Defining Cultural Phases and Communities: Explaining the Variability of Abu Hamid (Jordan Valley) Early 5th Millenium cal. BC Ceramic Assemblage.” In: Rowan, Y. and Lovell, J. (eds), Culture, Chronology and the Chalcolithic: Theory and Transition. CBRL Levant Supplementary Monograph Series (Oxford: Oxbow Books), 113–132. Rye, O. S. (1977). “Pottery Manufacturing Techniques : X-Ray Studies.” Archaeometry 19: 205–211. Rye, O. S. (1981). Pottery Technology. Principles and Reconstruction (Washington, D.C.: Taraxacum Press). Rye, O. S. and Evans, C. (1976). Traditional Pottery Techniques of Pakistan: Field and Laboratory Studies (Washington, D.C.: Smithsonian Institution Press). Shennan, S. (2013). “Lineages of Cultural Transmission.” In: Roy, E., Lycett, S. J., and Johns, S. E. (eds), Understanding Cultural Transmission in Anthropology: A Critical Synthesis. Methodology and History in Anthropology (Oxford: Berghahn Books), 346–360. Sigaut, F. (1994). “Technology.” In: Companion Encyclopedia of Anthropology. Humanity, Culture and Social Life (London: Routledge), 420–459. Simondon, G. (1958). Du mode d’existence des objets techniques (Paris: Aubier). Stark, M. T. (ed) (1998). The Archaeology of Social Boundaries (Washington, D.C.: Smithsonian Institution Press). Van der Leeuw, S. (1977). “Towards a Study of the Economics of Pottery Making.” In: Ex Horreo (Amsterdam: University of Amsterdam), 68–76.
Chapter 9
T he Organi z at i on of P ot tery Produ c t i on Toward a Relational Approach Kim Duistermaat Introduction The organization of ancient pottery production has been a topic of interest in ceramic archaeology for decades,1 because organization and production are seen as sources of information on the economy and sociopolitical processes in society (Schortman and Urban, 2004; Costin, 2005). However, besides viewing organization as a proxy for larger‐scale economics or politics, it is equally interesting to study organization on its own merits (Kohring, 2012b). How people organize themselves in order to make pottery concerns the relations between the people who make pottery, the relationships between potters and pottery users, and the dynamics of power and authority between potters and others. How did people cooperate and communicate, and how did they control material, human, and spatial resources, as well as the products? And how is the organization of pottery production related to the organization of other activities? Archaeologists try to approach these questions by searching for links between the material and the social aspects of pottery production. We usually try to identify material correlates for predefined social structures, often based on ethnographical examples. In this chapter I argue that these strategies have a tendency to limit our view on the diversity of ancient organizational practice, because their top‐down perspective restricts the possibilities for identifying new, previously unknown ways of organizing. On the contrary, I advocate the development of a bottom‐up, relational approach, building on several recent developments in the study of technology and organization. Such an approach will enable us to think outside the predefined boxes of types, modes, and parameters of production, in order to access the large variety of ways in which people organized pottery production in the past. This chapter will first discuss several traditional archaeological approaches to the organization of pottery production, including ceramic ecology and typological approaches. My main focus is on what I see as a major issue: these approaches struggle to bridge the analytical divide between the material remains and the social structures (organization) they are
The Organization of Pottery Production 115 trying to identify. I do not present a complete chronological or historical overview: many of these approaches were developed in roughly the same period, have mutually influenced each other, and are still influencing the work of many pottery specialists today. I will proceed to discuss a number of approaches that focus on technology and human–thing relations, including social constructivist approaches, behavioral archaeology, and approaches influenced by actor‐network theory. The latter see “the social” not as a structure or framework that has left material traces; rather, the social is understood as an effect that comes about through the interaction of people, artifacts, materials, and animals. These perspectives do away with the gap between material and social. Not all of these “sociotechnical” approaches have yet been applied to the study of pottery production organization, but they jointly provide important tools and principles for future work. In the section entitled “Towards a Relational View of the Organization of Pottery Production”, I present some preliminary suggestions for a relational approach, using an archaeological case study as an illustration.
Ceramic Ecology Which variables, which people, things, materials, circumstances, institutions, and events, influence how pottery was produced in a specific case? At least since the work of Frederick Matson (1965) it has been clear that ceramic production should not be studied in isolation. Production is too strongly interlinked with consumption, distribution, and other processes in society to do so. We should study the organization of production as part of a larger whole, including the artisans and their social identities and roles, their technologies and the means of production, the objects themselves and their functions and meanings, the mechanisms for distribution, and the consumers (Costin, 2005: 1038–1039). That all these components together influence, and are influenced by, production has been central to pottery studies for a long time, especially since the articulation and spread of ceramic ecology, later combined with general systems theory (a historical review can be found in Kolb, 1989; see also Matson, 1965; Rice, 1984, 1987, 1996: 184–185; Van der Leeuw and Pritchard, 1984; Arnold, 1985; Pool, 1992; Pool and Bey, 2007: 17–20), and the “Leiden School” approach to pottery study (Loney, 2000; Van As, 2004). “Holistic ceramic ecology” presented a model in which the “pottery production system” is influenced by the physical, biological, human, and cultural environment, and by the economic, social, religious, and psychological subsystems (Kolb, 1989). The ceramic ecological approach explicitly aimed to focus on the relations between all “subsystems” and their constituent parts in the model. The aim to understand the relations between the environment, materials, social and economic factors, and the production and use of pottery, resulted in a boost in archaeometric, ceramic ethnoarchaeological, and experimental research. The approach encouraged archaeologists to work together with other disciplines, and to look at the total picture rather than one aspect of production only. This approach has in many different ways shaped archaeological ceramic research, and is still very much a visible influence in recent work (Pool and Bey, 2007: 19). Critics have accused the model of “techno‐environmental determinism” and of not focusing enough on understanding the exact nature of the relationships between the variables (Gosselain, 1998; Jones, 2004; Pool and Bey, 2007: 18; Arnold, 2008). The model groups social, economic, religious, and environmental factors in just as many subsystems, and
116 Kim Duistermaat keeps them apart from each other and from the pottery production subsystem. Although it acknowledges links between each of them, they are kept as separate analytical entities, forming a “frame” or “context” for pottery production. This creates an inferential gap between the data on the pottery and its manufacture, and the social, economic, or religious “frame” in which these data are supposedly set. We have to manage a jump from one system to the other, from the pots at hand to the “larger” context, but it is not clear how to do so. Moreover, in the ceramic ecological approach, materials, objects, and humans have rather passive roles to play. Human behavior is seen as an “adaptation” to the environmental, cultural, social, or economic context. Material properties and principles are often seen as main drivers behind technical choices, leaving “the social” in the background.
Typological Approaches Between the 1970s and the 1990s, many archaeologists used modes or types of production organization to describe and classify the organization of craft production. Viewing organization as part of the socioeconomic context or of political economy, they explicitly aimed to focus on the social, economic, and political factors relevant to production. In pottery studies, the most often‐used typologies include those of David Peacock (1982), Elizabeth Brumfiel and Timothy Earle (1987), and Sander Van der Leeuw (1977). Over the past two decades, critique on typological approaches has been fierce (e.g. Costin, 1991, 2004, 2005, 2007; Pool, 1992; Mills and Crown, 1995; Feinman, 1999; Clark, 2007a; Neupert, 2007; Pool and Bey, 2007; Shimada and Wagner, 2007; Arnold, 2012). A production type is a label that obscures the complex, continuously moving, multifaceted, and multilayered reality behind it. Types are limiting and prescriptive rather than helping us to understand. Many scholars of craft production have now abandoned their use; however, some current archaeological studies of pottery production still rely heavily on these classifications. Perhaps this is because of a tendency to search for simple ways to approach the enormous complexity of pottery production (Rice, 1996: 191; see also Costin, 1991, for a claim that identifying production organization is relatively easy). Unfortunately, the typological approach is still advocated in introductory texts (Tite, 2008; Orton and Hughes, 2013) and even in work that otherwise seems to have moved away from traditional typological thinking (e.g. Van der Leeuw, 2008: table 12.2). As in the ceramic ecology approach, a fundamental problem of the typological approach is that production types are based on supposedly universal links between variables. These models assume agreement between a limited number of organizational variables, such as output, intensity of production, economic dependence on the craft, density of production debris, spatial extent of the distribution of the products, and the size, elaboration, and context of production locations (Feinman, 1999). However, many of these links are assumed to exist, mostly based on ethnoarchaeological case studies, but most links have not actually been studied very well. Furthermore, the material correlates of these variables are mostly assumed, rather than studied. Even those relations between variables and archaeological evidence that have been studied in more detail, through ethnoarchaeology, experimental archaeology, or archaeometry, are not universal (such as the causal links between organizational variables and “standardization”; for recent critical discussions see Berg, 2004; Gandon et al., 2014; Hilditch, 2014; Kotsonas, 2014; Roux, 2003b). Production types based on these
The Organization of Pottery Production 117 links thus have limited relevance for real‐life cases. However, our aim in studying pottery production should not be to discuss the universality of the links assumed in the model, or the validity of the types of production. Our aim should also not be to classify our case in one or the other type. Rather, we should try to understand how—in our particular historical case—pottery production was organized and why in that manner, which links were there, and why. A related problem is that these typologies also more or less explicitly assume relations between types of production and processes on a larger scale, such as the emergence of social complexity or state formation. Archaeologists tend to classify their case in a production type as a stepping stone to approach larger social processes, connecting those modes of production that are perceived to be more complex to increasing sociopolitical complexity (David and Kramer, 2001: 304). Related to this is the obsession with the concept of “specialization” as a cause or indicator of social complexity and political power (perhaps most clearly voiced in Rice, 1981; for more reading on the problematic concept of specialization see also Schortman and Urban, 2004; Hruby and Flad, 2007; Menon, 2008; Day et al., 2010; Baysal, 2013). This fascination with “specialization” is rooted in eighteenth-century political economy and nineteenth-century cultural evolutionism, and is shaped by our own experiences with capitalism (Clark, 2007a; Arnold, 2008). These theories sketch a unilinear evolution of human organization from egalitarian to capitalist, and frame organization in terms of costs, economics, and efficiency. This has historically defined the ways we understand the relations between specialization and human organization (Patterson, 2005; Kienlin, 2012), and the ways we understand the relations between technology and society in general (Dobres and Hoffman, 1994; Rice, 1996: 180; Arnold, 2008: 2–3; Dobres, 2010: 105). However, we should question whether neoclassical economic theory applies to past societies, or whether these models are limiting our views on the past (Clark, 2007a; Baysal, 2013). More recently, it has been suggested that increasing sociopolitical complexity can better be characterized by a growing variety of simultaneously existing forms of production organization, not by the emergence of complex forms of organizations per se (Sinopoli, 2003: 21; Costin, 2004). Notwithstanding this nuance, it is unclear exactly how a type of organization such as “household production” (a label that covers a wide variety of cases and definitions), or a parameter such as “division of labor,” connects to specific forms of sociopolitical organization; the precise links between them are not clearly defined. We should question, not assume, the nature of relations between organizational practices, political power, and social inequality (Schortman and Urban, 2004; Day et al., 2010).
Cathy Costin and Christopher Pool’s Characterizational Approaches: Typologies in Disguise In an attempt to overcome the drawbacks of rigid and prescriptive typologies, Cathy Costin (1991) suggested that we should describe production organization with the help of four parameters, which are not static but can each independently vary between extremes across a sliding scale. Originally these parameters were: context (the nature of elite control over production), concentration (the relative geographical concentration), scale (the size and constitution of production units), and intensity of production (the degree to which production
118 Kim Duistermaat is part‐time or full-time). At almost the same time, Christopher Pool (1992; Pool and Bey, 2007) similarly proposed to study the “dimensions of variation” in production, consumption, and distribution. These included scale, intensity, size of production and consumption units, segregation of activities, location of production and consumption, the variability of products, and the range and direction of distribution. How these approaches differ is discussed in Mills and Crown (1995) and Pool and Bey (2007). Originally, Costin (1991) again proposed a typology, channeling the almost infinite number of possible combinations of the four parameters into eight types of production, ranging from “individual specialization” to “retainer workshop.” There were several critiques on this approach: first, many aspects of the parameters (such as the amount of time spent on production, or the amount of income raised by production) are difficult to operationalize in the archaeological record (Sinopoli, 2003: 17; Shimada and Wagner, 2007; Menon, 2008; Kelly, 2009). Second, the focus is very much on the extremes of a parameter, but if a specific case sits somewhere in between it is difficult to position it in the model (Kelly, 2009). Moreover, the approach focuses mostly on the division of labor and sociopolitical centralization, and on artificial dichotomies such as prestige vs. utilitarian. And it neglects issues such as the value of things, time, and skill (Clark, 2007a; Shimada and Wagner, 2007; Day et al., 2010). Again, each parameter conflates several aspects that are not necessarily causally related, thus reintroducing the problem with the assumed links between variables in the typological approaches mentioned above. For example, in Costin’s model, the parameter “context” includes not only the relation of potters to authorities, but also makes statements on the types of product (utilitarian or luxury), on the principles driving production (such as efficiency), on the nature of demand, on the quality of the products, and on the access to products (whether controlled by elites or not). In “scale,” the number of workers and the principles of their recruitment (kin‐based or not) are conflated, although they are in reality not necessarily related. It is clear that in order to understand how these various aspects are related in a specific case, we must study them separately rather than conflate them in parameters (Pool and Bey, 2007; Arnold, 2008). In her later publications Costin (2001, 2005) seems to abandon the eight types, stating that her approach of parameters varying along axes “eschews typology” (Costin, 2005). Furthermore, over the years Costin (2001, 2005, 2007) redefined the four parameters that are important for understanding production organization: they are now called the degree of elite control, the spatial organization of production units, the size of production units, and the relative amount of part‐time or full-time production (Costin, 2007; Pool and Bey, 2007; Arnold, 2008). Additional important variables are listed as well, including the types of production loci, the social relations of production, the composition of the work‐group, specialization, and the relations between producers and consumers (Costin, 2005). Costin’s approach had a large impact in the field of pottery studies. Among the effects were a much more focused consideration of the various variables at play, and a clear realization that most variables are relative rather than absolute and thus can best be studied in comparison to other data. Many of the parameters are still found to be useful. Nevertheless, Costin’s approach, too, is limiting our vision to the specific nature of the links between all variables involved in the organization of pottery production in a particular case. Both typological and characterizational approaches decide beforehand which variables are most likely to be linked together and how, which variables are most interesting to look at, and which predefined sociocultural forms they may represent. This is overly reductive (Olsen et al., 2012: 184).
The Organization of Pottery Production 119
A Focus on Technology and Human–Thing Relations The time has come to develop new strategies for studying the organization of pottery production in archaeology, to move beyond a search for the material traces of predefined forms of organization. Without disregarding the enormous wealth of information and useful concepts resulting from the approaches discussed above, I think a relational approach can be built from elements that are core issues in the approaches discussed below, all dealing with the study of technology and human–thing relations. It is time to consider “organization” as what has to be explained (the explanandum), rather than as a predefined category that explains the patterns in our data (the explanans; Latour, 2005). We should adopt a bottom‐up approach that meticulously studies the data at hand on their own merit. Further, we should carefully study which relations exist between our data in our specific case, without relying from the outset on assumed, but non‐universal, links. If our material conforms to such links, fine; if it does not, even better: we will have brought out new information. Moreover, we should take an holistic perspective and focus on the materials, spaces, the potter, the pots, the users, and the function or use of the vessels. A future strategy should be applicable in cases where direct evidence for production is missing, and it should be useful for all kinds of societies. I think elements from approaches such as social construction of technology (SCOT), cultural technology, behavioral archaeology, holistic approaches, symmetrical archaeology, and entanglement can all contribute to the development of such a strategy, even if they have their roots in different theoretical or disciplinary backgrounds (for short introductions see Hodder, 2012b). In this section, I briefly discuss their principles and those elements I think are useful, before presenting a brief case study as an illustration.
The Social Construction of Technology Social constructivist approaches were developed in sociology from the 1980s onwards. They reject explanations that attribute changes in technological practice only to their internal technical aspects, or to “black‐box” mechanisms such as “efficiency,” “market forces,” “adaptation,” or “progress” (Loney, 2000; Killick, 2004). For example, science, technology and society (STS) studies or social construction of technology (SCOT) studies (Bijker, 2010; Bijker et al., 2012[1987]) focus on how the choice for a particular technology is closely interwoven with the beliefs, social structure, and historical choices of various groups of people. Technology is socially constructed (Bijker, 2010). In the words of Dobres and Hoffman (1994: 247): “the relation between technology and society can be described as a ‘seamless web’ […] that dialectically weaves together social relations, politics, economics, belief systems, ideology, artifact physics, skill, and knowledge.” One of the notions in this approach that is useful for my purposes here is the idea of “relevant social groups” (Bijker, 2010). SCOT acknowledges that an artifact, or technology, is heterogeneous: there is not one “real” artifact but there are many forms of it. This pluriformity exists because different social groups attribute different values and meanings to an artifact. In order to understand the social construction of technology, we have to identify these social groups and their view on the technology
120 Kim Duistermaat under study (Bijker, 2010). Social constructivist approaches have been criticized for privileging humans over things, artifacts, and technology, and for lacking adequate understandings of materials and technological processes (Killick, 2004). SCOT is rarely explicitly quoted as a source of inspiration for pottery studies (however, see Jeffra, 2011a; Murphy and Poblome, 2012).
Cultural Technology and chaîne opératoire A social constructivist perspective that has become widely adopted in archaeology in general, and in archaeological pottery studies in particular, is that of technological choices or cultural technology (Lemonnier, 1992, 1993, 2012; Pfaffenberger, 1992; Dobres and Hoffman, 1994; Dobres, 2000, 2010). This approach claims that most components of a technique—the choice of materials, the energies transforming materials, the choice of tools, the specific gestures, and the knowledge and skills involved—are determined by social factors (Lemonnier, 1992, 2012). Techniques are never only ways to make things in a utilitarian sense; techniques are also ways to fulfil social, political, religious, and symbolic needs (Gosselain, 2011). One of the most important analytical tools offered by this approach is the use of the chaîne opératoire, or operational sequence, as a methodology to study social factors in technological choices (for a detailed discussion see Roux, 2011; Roux, Chapter 8, this volume). Another important idea is that technological choices in one technique may be paralleled in other techniques, together forming a technological system (Lemonnier, 1992, 2012). Pottery making and other crafts can share pervasive beliefs and practices, gestures, technical knowledge, or resources (such as clay or fuel). Relations between crafts may also concern the existence of similar objects in other media (Knappett et al., 2010), the shared use of space, a similar consumer group, or other crafts and activities carried out by potters during the day or year. Ideally then, pottery should not be studied in isolation from other crafts (Sinopoli, 1998, 2003; Sillar, 2000; Sillar and Tite, 2000; Killick, 2004; Sofaer, 2006; Brysbaert, 2007; Shimada and Wagner, 2007; Michelaki, 2008; Brysbaert and Vetters, 2010; DeMarrais, 2013; Goldstein and Shimada, 2013). In ceramic studies the popularity of the chaîne opératoire approach has resulted in an increased focus on topics such as the social identity of potters, communities of practice, mobility and interaction of potters, skill, apprenticeship and learning strategies, and technological innovation and change (some recent examples are Gosselain, 1998, 2000, 2011; Roux, 2003, 2011; Berg, 2007; Michelaki, 2008; Brysbaert and Vetters, 2010; Jeffra, 2011a; Kohring, 2012a; Sofaer and Budden, 2012; Abell, 2014; Hilditch, 2014). These are all crucially important topics for understanding the organization of production. However, there seems to be less interest in studying organization. This is perhaps because in practice the chaîne opératoire approach as applied in archaeological pottery studies is mostly limited to the study of the producers and the production stage of ceramics: the preparation of raw materials, and the shaping, decoration, and firing stages, including the used tools, firing installations, and spaces (Sillar, 2000; Skibo and Schiffer, 2008; Hodder, 2012a). However, the organization of pottery production is influenced not only by decisions made during the production of pots, but also by considerations related to their distribution and use. After production, pots become tools (Braun, 1983), components of other techniques such as storage, food preparation, transport, and burial, and therefore pots become part of the technological choices
The Organization of Pottery Production 121 of users in their activities. Since all techniques can be studied with the chaîne opératoire approach, the method may be applied to the study of the whole life-cycle of a pot or an assemblage, from raw material selection, production, distribution, use, breakage, repair, and reuse, to discard, and to identify the social groups involved in these processes (Lemonnier, 1993, 2012; Naji and Douny, 2009; Knappett, 2012).
Behavioral Archaeology One of the more vocal advocates of the need to focus on the interaction between people and things is behavioral archaeology. Developed since the 1970s by Michael Schiffer and colleagues, behavioral archaeology claims that it is impossible to directly observe “social processes,” such as organization, since these are theoretical constructs. We can only study behavior, since behavior is composed of people–object interactions which leave traces (Schiffer et al., 2001; Schiffer, 2007, 2011; Skibo and Schiffer, 2008; LaMotta, 2012). Behavioral approaches try to achieve a full understanding of a particular technology by studying, in minute detail, several core aspects. Well known is the focus on cultural deposition processes and site formation processes, a topic much neglected in other approaches to pottery production. Other core components of the approach are the behavioral chain and the life history (of objects or of technologies; Hollenback and Schiffer, 2010); activities and interactions; technical choices; and performance characteristics and compromises (Skibo and Schiffer, 2008; Hollenback and Schiffer, 2010). Behavioral approaches promote an integrated study of the complete life history of a technology or artifact, from procurement, production, use, reuse, and repair, to discard and deposition. Each link in this behavioral chain is an activity, an interaction between people and things. Behavioral chain analysis specifies all components of these interactions, such as the location, frequency, other artifacts, external influences, techno‐communities, and cadena. The concept of cadena is used to indicate all social groups interacting with the artifact during its behavioral chain. A cadena can be homogeneous, when all members appreciate the same performance characteristics of an object, or heterogeneous, including many different (and sometimes conflicting) demands on performance characteristics. The cadena and all activities in a behavioral chain provide input to the technical choices an artisan will make during production, as an artisan weighs the effects of his choices on the various performance characteristics and demands (Walker and Schiffer, 2006; Schiffer, 2007; Skibo and Schiffer, 2008). The concept of cadena is comparable to the “relevant social groups” in the SCOT approach, but later publications suggest that a cadena not only includes people but may also contain objects and materials, treating people and things as socially equivalent or symmetrical (Walker and Schiffer, 2006; Schiffer, 2007; Skibo and Schiffer, 2008; Hollenback and Schiffer, 2010; a cadena also resembles the “entanglements” of Hodder, 2012a; see also below). Behavioral approaches see technical choice as a (conscious or unconscious) decision based on the (utilitarian, symbolic, ritual, etc.) use of the pot, while the technological choice of the cultural technology approach concerns the (conscious or unconscious) adoption of a practice based on the experiences and background of the potter (such as social identity, community of practice, and learning patterns). Technical and technological choice are thus two complementary sides of the process of making things, which both may offer useful insights in the organization of production. For Schiffer’s views on the differences and similarities
122 Kim Duistermaat between concepts of behavioral archaeology and cultural technology, such as behavioral chain vs. chaîne opératoire, technical choices vs. technological choices, and techno‐communities vs. communities of practice, see Schiffer (2007) and Skibo and Schiffer (2008). One of behavioral archaeology’s main attractions for the study of the organization of pottery production is the focus on the people involved in all stages in the life of an artifact, which all may influence production, production decisions, and social relations during production. Behavioral archaeology moreover offers a practical methodology to study these processes. It actively advocates an integrated use of archaeology, ethnoarchaeology, experimental archaeology, and archaeometry. Behavioral approaches have been said to be overly utilitarian and functionalist (Gosselain, 1998) or materially determinist (Hodder, 2012a: 229), emphasizing things over people (Webmoor, 2007). Although in principle they claim not to favor utilitarian, material‐ based, or functional perspectives over “non‐utilitarian” social or ritual explanations (Skibo and Schiffer, 2008: 25; Hollenback and Schiffer, 2010: 318–319; Schiffer, 2011), in practice the approach is often understood as such. This is not in the least owing to the insistence that we have to identify utilitarian performance characteristics first, before thinking about possible non‐utilitarian characteristics (as advocated in Skibo and Schiffer, 2008: 26; contra Dobres, 2000, 2010). Olsen et al. (2012: 186) furthermore object to the fact that behavioral archaeology puts the relational properties of things (performance characteristics) second to their “intrinsic” properties. Others point out that behavioral approaches do not pay enough attention to the deep history of the involvements of people and things (Webmoor, 2007; Hodder, 2012a), and portray artisans as “engineers” doing tests and solving technical problems (David and Kramer, 2001: 141). An additional problem, in my view, is behavioral archaeology’s focus and reliance on predefined universal or nomothetic principles and assumptions about the relations between material traces and social processes. For behavioral archaeology they are not only the ultimate aim of our efforts but also an indispensable tool needed to bridge the gap between the archaeological record and our interpretation of it (as, e.g., in Schiffer et al., 2001; Walker and Schiffer, 2006; Schiffer, 2011). However, these principles and laws are obscuring our view of the actual associations between people and things (Gosselain, 1998; see also below).
Holistic Approaches to the Organization of Craft Production An approach specifically developed to study the organization of craft production, and combining methods and insights from ceramic ecology, behavioral archaeology, chaîne opératoire, and constructivist perspectives, is the “holistic” approach developed by Izumi Shimada (Shimada and Wagner, 2007; Shimada and Craig, 2013). This ambitious approach explicitly looks at the whole craft production process, from raw material acquisition to product use and recycling, including both the material–technological and the social–ideological components of a craft production system, while trying to avoid modern preconceptions and analytical distinctions (Shimada and Wagner, 2007). It has four major components: (1) a regional, multi‐site, and diachronic scope to clarify the environmental, historical, and social contexts of craft production and the distribution and use of its products; (2) the study of production sites, aimed at understanding the complete production process; (3) close interdisciplinary cooperation between complementary specialists; and (4) the integration of
The Organization of Pottery Production 123 archaeometry, experimentation, and ethnoarchaeology (Shimada and Wagner, 2007). The focus in the holistic approach is very much on the detailed study of direct evidence for production, although in principle this approach can be applied as well to assemblages that lack such evidence. Recent studies adopting an holistic approach to pottery production organization (including study of the environment, production, production locations, distribution, function, and use), whether or not explicitly following Shimada’s framework, are Day et al. (2006, 2010), Duistermaat (2008), Gagné (2012) and Greene (2013). Such holistic studies have been successful in bringing out the nuances and complexities of the (organizational) relations between all actors involved in craft production. They also clearly show that the study of craft production is far from an easy matter, and ideally involves a long‐term commitment of an interdisciplinary team.
The Ontological Turn Understanding the relationship between the social and the material, and developing theory and method to bridge the gap and understand the one by studying the other, has been a long-standing issue in archaeology (Olsen, 2003, 2010; Hicks, 2010); many would perhaps say that this is what archaeology is all about. A major transdisciplinary ontological turn that has been taking shape since the 1980s promotes a radically different perspective: the dualism between social and material, between humans and things, is not a given, but a construct of modernist thinking that we should let go of (Olsen, 2003; Walker and Schiffer, 2006; Witmore, 2007; Hicks, 2010; Watts, 2013). This perspective is strongly influenced by actor-network theory (ANT), a sociological approach originating in STS studies and developed since the 1980s (Law, 1992; Latour, 2005). In archaeology, approaches influenced by ANT that are relevant for my argument here include symmetrical archaeology (Shanks, 1998; Olsen, 2003, 2007, 2012, 2013; Webmoor, 2007; Olsen et al., 2012) and entanglement (Hodder, 2011, 2012a). Symmetrical concepts and ideas are also influencing behavioral archaeology (Walker and Schiffer, 2006; Schiffer, 2007; Skibo and Schiffer, 2008; Hollenback and Schiffer, 2010). ANT‐inspired approaches share an understanding of the social not as a category separate from the material, as a larger context or framework behind the material world. The “social” or “society” is not something written in or embodied by things; rather, things are an inseparable part of its constitution (Shanks, 1998; Olsen, 2010). “The social” is an interactive effect, emerging during the mutual interaction between humans, nature, things, animals, and so on. People, society, technology, and material culture are continuously coproducing each other, rather than one being embedded in the larger context or framework of the other. In order to see the effect we call “organizing,” we have to reassemble the associations and interactions between all these actants, while treating people and things symmetrically, without any a priori ontological or analytical distinction between the two. In this perspective there is no gap to bridge between the social and the material (Webmoor, 2007: 572); rather, the “materials of past (and present) societies are not seen as an epiphenomenal outcome of historical and social processes […] but actually as constituent— even explanatory—parts of these very processes” (Olsen, 2010). The lack of living people as a source of information in archaeology, as opposed to ethnography, is not seen as hindering
124 Kim Duistermaat or complicating the study of “the social” through material remains: “We uphold a materialist outlook—you do not have to talk to people to find out how they conceive of the world, because something of the way people operate, work, and do is wrapped up in their achievements. People are so involved with the world of material goods that we can put to one side the old split between mind and matter, beliefs and the material world that may leave traces for the archaeologist to work upon” (Olsen et al., 2012: 167). This is not only relevant for archaeology: students of contemporary organizations increasingly turn to study materials and technology, irrespective of the fact that they can directly observe and interview organizational members (Orlikowski and Scott, 2008; Orlikowski, 2010; Leonardi et al., 2012; Carlile et al., 2013; Humphries and Smith, 2014). In this respect, organizational research is now starting to look at “things” using concepts and methods that have been used and developed in archaeology for decades. Bruno Latour proposes that “reassembling the social” is best done by looking at situations of innovation, at the places where things are made (such as an artisan’s workshop), at situations of breakdown and failure, and through looking at the history of technology (Latour, 2005). This, and the focus of ANT on technology, power, and organization, makes this approach especially interesting for those studying the organization of pottery production. However, there are as yet few pottery studies adopting ANT‐inspired or symmetrical approaches (examples are Watts, 2008 (cited in Watts, 2013); Jervis, 2011, 2013; Stockhammer, 2012; Van Oyen, 2013).
Towards a Relational View of the Organization of Pottery Production The approaches mentioned in the previous paragraph relate to, and differ from, each other in complex ways (Coupaye and Douny, 2009; Hicks, 2010; Hodder, 2012a; Ingold, 2012). A discussion of their compatibility or comparability falls beyond the scope of this chapter. In this section I propose these approaches may be combined to study pottery production organization. Examples of other strategies combining elements of these approaches are Hilditch (2008) and Jeffra (2011a).
Tracing Entanglements As Latour (2005) suggested, “the social” (such as organizational practices) can be reassembled by empirically following, tracing, the numerous interactions between all human and non‐human “actants.” Similar strategies are employed by symmetrical archaeology using the term “rearticulation” (Olsen et al., 2012: 176) or entanglement (Hodder, 2012a). Interactions between these actants lead to the formation of actor‐networks (Law, 1992; Latour, 2005), assemblages (Shanks, 1998; Alberti et al., 2013; Fowler, 2013), or entanglements (Hodder, 2012a; see Fowler, 2013, for a more elaborate discussion of the differences and similarities between these concepts). Organization should be understood as a process, as emerging from such entanglements. There exist no such entities as “household production” or “attached
The Organization of Pottery Production 125 production” which are out there and can be discovered, or which can be used to explain archaeological data. Rather, we have to explain how organization becomes, how it is a continuously emerging effect of the entangled and enmeshed relations and interactions of heterogeneous actants, including people, objects, tools, materials, spaces, and forces (Law, 1992; Hernes, 2008; Jervis, 2011; Humphries and Smith, 2014). In the case of pottery production, the actants are many, and may include anorganic materials (clay, water, rocks, metal), plants (as fuel, organic temper, resins, rope and textiles, wood), people (potters, users, authorities, children, middlemen; including their skills, needs, demands, and identities), animals (transport animals, cattle and sheep providing dung for fuel and temper, bone tools, and hair as tools and temper), technologies (clay extraction, paste preparation, shaping, decorating, firing, transporting, cooking, storage, distribution, burial), architecture, places, and spaces (fields, mountains, the workshop and its location and layout, roads, places where pots are used), concepts, interests, feelings, and opinions (efficiency, aesthetics, magic, value, gender, norms), forces, energies, processes, and reactions (gravity, pressure, speed, oxidation, weight, temperature, time, decomposition). Each of these actants can, in their turn, be seen as entanglements, networks of relations. For example, a “potter” is a complex meshwork of interactions and associations between a human being, clay, tools, technology, knowledge, skills, other people’s opinions about “potters,” the community, and more. Recurrent interactions between all these may lead to a stable state that presents itself in daily life as a single entity, recognized as a “potter” (Law, 1992; Hernes, 2008; Michelaki, 2008; Orlikowski and Scott, 2008; Budden and Sofaer, 2009; Jervis, 2011; Fowler, 2013). One of the more challenging aspects of a research project is to determine which entanglements we decide to see—for analytical purposes—as such a black‐boxed entity, and which entanglements we aim to “untangle” by following the interactions between all the actants involved (Latour, 2005; Hernes, 2008: 7–8). This depends on our research questions. Research questions should avoid a top‐down approach, avoid trying to fit archaeological evidence in (and searching for evidence of) a priori existing analytical distinctions (Shimada and Wagner, 2007), meta‐narratives, frameworks, concepts, and models, such as “the emergence of complex societies,” “craft specialization,” or “modes of production” (Olsen et al., 2012: 175–176, 190). This does not mean that larger-scale questions are irrelevant. I am also not claiming that variables that are thought to influence craft organization, such as task divisions, specialization, or intensity of production (cf. Costin, 2005; Van der Leeuw, 2008: figures 12.2–12.8) are irrelevant. Rather, I suggest that we should not let these constructs lead our way, determining from the start which associations are worth tracing. I propose that it is important to question, rather than assume, the existence and specific nature of these variables and the relations between them in each particular case, and to also actively search for any associations that do not fit these a priori frameworks. We have to adopt a bottom‐up approach (Fahlander, 2013; Mímisson and Magnússon, 2014), focusing on practice and process, and systematically following the “networks of empirical, statistical, metaphorical, narrative, conceptual, causal and systemic association” in our data (Olsen et al., 2012: 176), while using our creativity and trying to think beyond our usual assumptions. Such a relational approach will also enable us to see how organization and complexity is apparent on any scale, and how we can approach larger‐scale issues through the detailed study of materials on the micro‐level (Day et al., 2006; Kohring, 2011, 2012b; Mímisson and Magnússon, 2014).
126 Kim Duistermaat I propose four interlinked and overlapping strategies to study the organization of pottery production (Hodder, 2012a: 204–227; Humphries and Smith, 2014). First, it is important to attend to the material properties of the actants involved in pottery production, and to what they do: how they constrain, afford, or influence (organizational) practices. Secondly, using the chaîne opératoire approach, we can map the sequences, activities, and entanglements of pottery production and its organization, identifying all actants and establishing what they do. Thirdly, the chaîne opératoire approach may be used to follow the biography or life‐history of our material. Last, we can trace the spatial aspects of these entanglements, including the location of materials and production, users and use activities, and the distribution and circulation of pottery. We can also trace the various temporal dimensions of the entanglements at various timescales (Hodder, 2012a). These approaches may be combined to bring the relevant social groups or cadena into view. All strategies are interlinked, and have no particular order or sequence (and most probably will be performed simultaneously). They provide different lenses one can use to look at the same material, to bring different aspects of it into focus. Together, they can be used to map entanglements and situate them in space and time. I will briefly describe these strategies in more detail below, each time using a Middle Assyrian “carinated bowl” from Tell Sabi Abyad, Syria (c.1200 bc), as an example of an actant under study (all information is based on Duistermaat, 2008).2 I chose this particular shape because it has become almost iconic for the supposedly standardized, centralized, state‐controlled mass production of pottery in the Middle Assyrian period, a view which I find overly simplifying and unhelpful for understanding Middle Assyrian craft production. Moreover, my choice for a vessel instead of a find from the pottery workshop we also excavated at Sabi Abyad, will hopefully show that many useful insights can be gained in the absence of direct evidence for production. I chose to draw “tanglegrams” (Hodder, 2012a) as a visual support of my point, but this actually may not be the most practical solution. Of course, tracing entanglements should not be limited to one bowl, but should include the whole ceramic assemblage (Roux, 2011; Van Oyen, 2013). It should also include tools, spaces, materials, texts, images, and any other actants involved in the organization of pottery production, as much as are available.
First Strategy: Tracing Material Properties A first step in tracing entanglements involves tracing materials, and studying how the physical properties of materials affect the organization of production (Jones, 2004), what these materials do, and what happens to them during their life (Ingold, 2012). We can look at materials from at least three perspectives. The first perspective concerns the physical nature and properties of the materials involved in the actant under study, and their interrelations or entanglements with other materials and actants both within the object itself and outside it (Figure 9.1). Our bowl was made of calcareous clays with vegetal inclusions. Possibly, animal dung was used as temper material. If we focus on tracing the entanglements of dung, this opens up a range of connections to other actants and processes, such as animals, the plants they ate, agriculture, procurement of amounts of dung, drying times and places for dung cakes, seasonal activities, and the use of dung in other activities (e.g. as fuel, or as a component of plaster; cf. Sillar, 2000; Goldstein and Shimada, 2013). Of course, detailed
The Organization of Pottery Production 127 potting cooking
smithing hearths, kilns
fields
excrements smell fertility
building
season dung as fuel
architecture
dung in plaster
plants
increase workability decrease plasticity reduce weight increase porosity rough vessel surface
smell
sheep sheep pen
sheep herding
sheep dung
brackish water? farming
salt
pits and bins
salt supply? reduce lime spalling
fields
soils
wadi
calcareous clay
river
dust
houses storage jars
water
wells? building not very plastic high shrinkage risk of lime spalling
river writing
brewing increase plasticity
cooking
dust and dirt soil, land fertility
Figure 9.1 Entanglements of the materials used to make a carinated bowl. understandings of materials will also provide information on the spatial dimensions of entanglements, for example when establishing the geographical source of materials. Secondly, we can study what materials do: how they interact together and how their interaction results in constraints and affordances for action of other actants, affecting the material engagement between material and potter (Malafouris, 2008), the chaîne opératoire, and the organization of work. Materials do not have intrinsic properties that are waiting to be brought out by people; rather, these properties and affordances are the result of the interaction between actants in specific situations (Knappett, 2004; Hodder, 2012a; Jones and Alberti, 2013: 24) and have functional as well as representational aspects (Gosselain, 2011). In our case of dung temper, these interactions may result in specific technologies for mixing clay and dung, the smell of the clay body, effects of the dung temper on increased workability of the otherwise rather
128 Kim Duistermaat short clay body, effects of vegetal inclusions on coping with shrinking problems, and the behavior of the clay body during firing and its relation to kiln technology and firing skill. A third perspective focuses on what happens to materials during activities after their production: how they perform and affect use, but also how they change, decay, and disintegrate during their life‐history. In our case most of the dung will have burnt out during the firing stage. However, the dung‐containing fabric of the bowl will have particular qualities including fabric porosity and vessel weight. Together with size and shape, this fabric will have effects on, for example, tensile strength and mechanical shock resistance of the bowl, breakage patterns and frequency, or the resistance to hot contents. This particular fabric will also change during its lifetime; for example, as a result of its interaction with acid contents. Apart from the functional consequences of material properties, these properties also constrain or afford social practices (Jones, 2004). In our case (although this has not been studied yet), we may, for example, wonder whether any material properties of dung, such as its smell, may relate to cultural connotations of dung (Sillar, 2000); if these were negative, perhaps we can link them to the choice of dung‐free fabrics for drinking goblets, the preference of the high elite for glass and metal drinking vessels, or the low social status of the potters and their profession. The entanglements of the calcareous clay and other components of the fabric can be traced in similar ways. A second aspect of material properties is form; the specific size and shape of our vessel. Again, form is a bundle of connections among processes, uses, techniques, and performances (Olsen et al., 2012: 191). We are dealing with a small bowl with a flat base, a simple rounded rim, and a flaring, lightly carinated wall. The bowl is somewhat slanted to one side and the base is cracked. The surface is left untreated and undecorated. Comparisons with similar bowls show that ours belongs to the middle one of three loosely defined size groups. Form and material affect interactions between actants in specific situations, resulting in particular constraints and affordances (Knappett, 2004; Gosden, 2005; Hodder, 2012a). For example, our bowl affords the holding and taking out of food, drink, and other materials, it fits in a hand, the carination and surface prevent slipping and facilitate grip, it can easily tip over, it can be stacked (but the irregular slant prevents a high stack), it is lightweight, it is not very watertight, it fits the mouth of large jars as a lid, it can hold c.0.3 liters, it is not particularly beautiful, and it is very recognizable (and regarded as an archaeological “type‐fossil” for the Middle Assyrian period). This specific form is closely tied to the way in which it was made, and to the ways it was expected to perform during use. It is also closely related to bowls made in the same shape, but from bronze. Bronze bowls were expensive and rare. Still, these bowls shared the shape of our everyday pottery bowl, which was found in huge quantities in a large variety of contexts. This suggests that the carinated shape is not only recognizable for us archaeologists but carried meaning for its users as well. Meaning and value may also be accessed through its quick and rather careless shaping and finishing, and through their relative uniformity.
Second Strategy: Tracing chaînes opératoires Using the second lens to look at our bowl, the aim is to study in detail its chaîne opératoire and forming techniques, using a variety of low‐tech and high‐tech methods of analysis
The Organization of Pottery Production 129 (e.g. Smogorzewska, 2007; Bouzakis et al., 2011; Berg, 2013), in order to trace the entanglements of the production process. A close mapping of all sequences and activities related to the making of our bowl will enable us to identify the actants related to each step of production (such as the potter, assistants, the clay and other ingredients, tools and firing installations, and spaces). We can study how materials and their properties affect the chaîne opératoire, and what happens to them during the work. Questions are provoked about the kind and size of spaces and tools needed, and about access to materials and spaces. A thorough study of the chaîne opératoire and traces on the vessels (such as fingerprints, mistakes and corrections, or differences in skill levels) provides information on the probable number of people at work, on task segmentation, repetition and serial work, and on the possible involvement of children or trainees (Crown, 2007; Joy, 2009; Sofaer and Budden, 2012). Moreover, through this approach, we can come closer to gestures, knowledge, and skills. These are important aspects when addressing questions about the intensity of production or the output per potter, the relative skill level of the potter, or the amount of time spent on production (Roux, 2003a, 2003b; Crown, 2007). We can focus on the potter’s social identity and status, and on communities of practice. The chaîne opératoire approach also helps to assess time, including issues such as seasonality and the involvement of people in other activities, time‐flow of the work, and the minimum amount of time needed to complete the work. This is crucial information if we want to estimate the intensity and output of production. The chaîne opératoire opens up possibilities for comparisons with other crafts and activities using similar technologies, gestures, tools, spaces, or materials, producing similar products, or dealing with similar user groups (Sofaer, 2006; Brysbaert and Vetters, 2010). Regarding our interest in the organization of production, we may want to pay special attention to those parts of the chaîne opératoire that involve communication between people involved in production, and communication and cooperation with other human actors (such as neighbors, users, suppliers of materials, or authorities). We should also consider how techniques, tools, infrastructure, and spaces afford organizational practices. Figure 9.2 presents a basic chaîne opératoire for our carinated bowl. In order to keep the image readable, I have listed the various materials, places, and activities that are part of the entanglement in separate boxes rather than in a tanglegram. Also, in order to focus on organizational practices, I have marked those steps that are likely to have involved task divisions or the help of assistants, as well as those steps that likely involved communication and cooperation with people outside the workshop (see also Sofaer and Budden, 2012). It appears that our bowl was made by a skilled potter and at least one assistant, who were able to throw vessels from the cone using the rather short local clay, and fire them with modest firing losses. Communication and cooperation with others was mostly needed for the acquisition of raw materials and tools, and perhaps for kiln building. Despite skill levels, or perhaps we should say enabled by them, our potter was focusing on output and speed, and less on quality and aesthetics. Among the interesting and unexpected entanglements of our bowl in production are the frequent links to scribes. In one phase of the site, the pottery production takes place in the courtyard of a scribe’s house (identified by texts found there). Furthermore, cuneiform writing and seal impressions occur sporadically on our type of bowls, and unfired waste fragments from pottery production are found in recurrent association with unfired cuneiform tablets and clay sealing fragments. These entanglements show that the cooperation and social relations between potters and scribes may have gone beyond the sharing of raw materials.
130 Kim Duistermaat
ACTIVITIES
clay digging
collecting sheep drug
crushing
crushing
days
take out large particles mixing clay, water and dung SPACES field river sheep pen mixing pit or place kneading place place to put ready cones wheel emplacement drying space/shade firing location storage space workshop house scribe’s house aboandoned space/ruins
farming grinding cooking basketmaking grinding tool making writing seal use
potting assisting kiln building sheep herding
fetching water
days - weeks resting kneading
store unused clay
shaping into cones put cone on the wheel spin the wheel miniutes
throw a bowl from the cone use thread to cut place bowl aside move bowl to dring space
days
collect/make bricks
dry
collect clay prepare plaster
hours
repeat steps
MATERIALS clay water dung digging tools baskets water skins or jars grinding tools stick to stir liquid clay lubricant potter’s wheel stik to spain the wheel bowl of water thread kiln fuel gypsum bitumen basalt thread reeds wind direction wood sunshime rainfall temprature seasons
build kiln
move bowl to kiln
collect fuel
firing
recycle damaged bowls
fetch water repair cracks
hours
move bowl to storage
Figure 9.2 Chaîne opératoire for a Middle Assyrian carinated bowl with a flat base. Gray circles indicate likely situations where communication with “outsiders” is necessary. Open circles indicate likely moments of task division and the presence of assistants.
Third Strategy: Tracing Biographies The “biography” approach has enjoyed some popularity in the past decades (Hicks, 2010). Some studies focus on the meaning or significance of objects to people (as in Kopytoff, 1986; Gosden and Marshall, 1999), some on the technical and functional changes of objects during their use‐life (as in behavioral chain analysis, see above; see also Peña, 2007). Others call for more attention to the literary techniques of biography writing (Burström, 2014), discuss the long‐term life‐history or evolution of a particular technology (Roux, 2010, 2013; Laneri, 2011), or investigate the extension of an object’s life‐history into the present (Shanks, 1998; Holtorf, 2002). I do not intend to sketch a linear or chronological life‐course for our bowl, starting with production, through use, maintenance, and reuse, and ending with discard. Of course, all these stages in the “life” of our bowl are important to consider, but the sequential aspect (was it first used during meals, and later as a lid, or the other way around?) is
The Organization of Pottery Production 131 often hard to reconstruct. My focus is mostly on the relations between objects and people during the post‐production part of the bowl’s life‐cycle: these relations determine the use, value, and meaning of our vessels and directly influence production (Clark, 2007a). We will follow the bowl as it gathers entanglements of heterogeneous actors (Shanks, 1998). The purpose is to consider “the range of interactions between people and objects and [to explore] how multiple forms of agency emerge through them” (Jervis, 2013: 219). Recurrent associations with other pottery vessels, with other objects and materials, with places, and with people create durability in the social assemblages in which our bowl was a participant (Jervis, 2011, 2013; Zedeño, 2013). Therefore, I will adopt a “relational” perspective and attempt to map the heterogeneous events and actions that our bowl (and similar bowls) was participating in (Joy, 2009), in order to understand the variety of activities and actants. These different activities, interactions, and resulting social assemblages (or cadenas) may have affected the organization of pottery production in different ways (Walker and Schiffer, 2006). The mapping of interactions is based on traces of use and repair and residues of contents (Skibo, 2013), specific find contexts, and the appearance of this particular shape in other materials, contemporary images, and texts. The same approach should, of course, also be applied to any available direct evidence for production, including spaces (e.g. Papadopoulos and Sakellarakis, 2013, using computer simulation to study the affordances of a room identified as a pottery workshop), architectural features, and tools (studying tool manufacture, acquisition, and style, e.g. Gosselain, 2010; Ramón and Bell, 2013), use‐wear on potter’s tools (e.g. Torchy and Gassin, 2010; Van Gijn and Lammers‐Keijsers, 2010), and tool provenance (e.g. Murphy and Poblome, 2012; Fiaccavento, 2013: 85). It is crucial to link these studies to the chaîne opératoire studies of the pottery assemblage. The resulting tanglegram (Hodder, 2012a) in Figure 9.3 presents several instances of our bowl’s “cumulative” biography. Our bowl has now become a tool in other technologies, such as cooking, food preparation, storage, and burial. The entanglements of each of these technologies can be traced again by using the chaîne opératoire approach, following the courses of action resulting from these associations (Sillar, 2000; Jervis, 2013); Figure 9.3 shows only the very start of such entanglements (cf. also fi gure 3.5 in Hodder, 2012a). As Figure 9.3 shows, our bowl was a multipurpose bowl mainly used for the presentation and consumption of food and drink. As such, the bowl played a role in re‐enacting and maintaining social relations, traditions, and feelings of “home,” through specific ways of sharing meals. Perhaps these ways were similar to the modern Middle Eastern “mezzeh,” where multiple small bowls containing different kinds of food are placed in the middle of a group of people, rather than each person having their individual plate. The connections with “brewing” raise questions about connections between potting and brewing, especially in the light of a contemporary text from the site suggesting that the brewer was on occasion in a position to order the potter to produce vessels (Wiggermann, 2008). Perhaps this lead can be followed further by attempting a better identification of the vegetal fibers in the pottery fabric. Did the potters indeed use animal dung as temper, or did they use the waste of the brewing process, so that the brewer was not only a user but also a supplier to the potter? Our bowl had additional roles in craft production, storage, and ritual activities such as burial. Comparisons of this tanglegram with that of other types of small bowls reveals that other small bowls were never used in burials, nor as lamps or jar lids, although they do afford such uses. This, combined with the rather rare occurrence of maintenance and repair of carinated bowls, raises
132 Kim Duistermaat cooking everybody
kitchen house
meals
food
storage vessels, sacks, baskets, etc.
short-term dry storage
banquets elites
food serving
open vessels scoop
bronze drinking bowls drinking
strainer? potting
gypsum
water
bitumen
jar lid
scraping tool wick oil
storage grain
large jars burial gift
oil lamp
darkness
brewing
beer
processing
pavement pit lining
garbage
offering
beer
burial jar ritual afterlife
temple
filler
built environment
deposition
secondary use
fracture
reuse
use
repair
repair
production
building
Figure 9.3 Entanglement of the life-history of carinated bowls, from production until deposit in the archaeological record. The size of the circles indicates the relative importance of this use. Gray circles indicate secondary uses after fragmentation beyond repair. questions about the values and meanings our bowl had for the community living at the site. The use of specific vessels in burials and during shared meals touches on the creation of identity and community, issues that were of special importance in a settlement that was founded by Assyrians in “hostile” territory, as part of a hegemonic strategy to incorporate the region into the Assyrian empire. The variety of our bowl’s biography also illustrates its multiplicity, and this opens up ways to investigate the composition of the various “relevant social groups” interacting with our bowl. As these groups are directly contingent on decisions concerning design and technology, they are of interest for the study of the organization of pottery production.
The Organization of Pottery Production 133
Fourth Strategy: Locating Entanglements in Space and Time Tracing the spatial dimensions of the entanglements of materials, production, and biography will enable us to access issues of source, distribution, circulation, and deposition of materials and objects (Hodder, 2012a; Jervis, 2013). We can look at source areas of materials and tools, helping to plot the location of production regionally, and bringing people– landscape interactions into focus (Druc, 2013; Michelaki et al., 2015). At different scales, we can study the spatial dimensions of the chaîne opératoire, the specific layout of the workshop, and the location of production in relation to the site (Stissi, 2012). Also important is the distribution of vessels after production: were they exchanged locally or further away, through which mechanisms, and which actants are involved? Each use also has its own spatial dimensions. At our site, we were fortunate to find the production locations where our pottery was made, including workshop areas, drying areas, and kilns. We were able to study the spatial organization of production activities in detail, both in the workshop as well as in relation to the rest of the site. But even if that is not the case, careful plotting of spatial dimensions may yield interesting understandings on the movement of materials, tools, products, and people. Again, we should study these processes through the whole life‐cycle of the actants involved in production. One example of the spatial dimensions of our bowl is its regional distribution. Sabi Abyad was a fortified estate founded in order to incorporate the region into the Assyrian empire, and to exploit its agricultural resources. Texts suggest that there were numerous settlements in the close surroundings of our site, housing local non‐Assyrian inhabitants who were dependent on the Assyrian administration. However, our typical carinated bowl was only sporadically found at such sites. This raises questions on how dependent the local population really was, and on the apparent lack of active attempts to “Assyrianize” the local population by encouraging the use of “Assyrian” vessels. Moreover, if pottery was not distributed among dependents, this informs us about the relatively small size of the user group for whom our potter produced, putting doubts on the idea that production was a full-time affair (Duistermaat, 2015). Locating entanglements in time can also be done on several scales (Gosden, 2005; Hodder, 2012a). On one scale, there is “operational” time: the time and sequence that builds up each activity that is part of making or using our pottery vessels. There is often a specific order for doing things, and there are constraints and demands on time in each sequence. One can think of drying time needed before firing a vessel, or of the need to finish an operation before the wheel loses momentum. Moreover, certain cooking or storage techniques require vessels to be in use for considerable durations of time, while other uses result in quick fragmentation. The temporal perspective also brings in concerns of seasonality, and the simultaneous involvements in other tasks and crafts. On a second scale, we can consider the life‐history of artifacts. This does not only concern the various uses an artifact had, but also the recycling, inheriting, or purposeful destruction or deposition of artifacts. Assemblages are never homogeneous in age: some pots will be brand new, while others will have been used and reused for decades. This is not only relevant for chronological purposes, but also directly impinges on production organization, affecting aspects such as replacement needs and rates, and output volumes. On a third scale, we can look at the historical developments and changes of techniques and organization. The very important and currently much studied concept of innovation and technological change is crucial in this respect. Historical
134 Kim Duistermaat patterns—in the form of knowledge and experience—may constrain the adoption of new practices. For example, the use of a certain technique or production organization can complicate the adoption of a new one (Arnold et al., 2007; Van der Leeuw, 2008; Jeffra, 2011a), or a certain layout of a building may dictate future locations of walls (Hodder, 2012a).
Analyzing Entanglements and Reassembling Organization Despite the hard work of tracing the innumerable entanglements of all possible actants we have at our disposal, one could argue (as do Jones and Alberti, 2013: 27) that this does not immediately bring forth an understanding of organizational practice. Nevertheless, in the process of doing so, we have gained an amazingly detailed understanding of our pottery, its production, and related materials and people. In itself, this is already much more than we ever could have learned from classifying our case based on a limited number of predefined criteria, as in typological approaches. We have now identified the relevant actants (including people and non‐humans) and traced their relations. But the relations between the actants should not simply be lines. It is more productive to view these relations as actions expressed with verbs, such as “use,” “produce,” “depend on.” To understand organizational practices, we should look at what these actants do: what are the actions they perform together and on each other, and how do they affect organizational practices? Which actants are “mediators,” influencing and consolidating roles, relationships, communication, control mechanisms, decisions, and power, and how do they do so (Latour, 2005: 37–42)? We should also look for patterns, recurrent actions, and routines (Olsen et al., 2012). Partly, these questions can be tackled through archaeometry or experimental archaeology. For Latour, the clue to reassembling the social lies in the process of writing (Latour, 2005: 121–140; see also Burström, 2014). He sees the writing process as crucial, because the social will appear only through a well‐written account. He defines a good account as a narrative in which every participant is a mediator, is doing something. The quality of an account is measured according to the number of actors the scholar is able to treat as mediators, without taking the shortcuts provided by concepts such as, for example, “efficiency.” I think it is also interesting to see if we can approach the relations between actants in a more formal way. Any tracing of actants will quickly result in a large and varied collection of heterogeneous connections between heterogeneous actants, which need to be analyzed for significant relational patterns. Would it be possible to approach these entanglements with techniques from the quickly growing field of network analysis and computer applications in archaeology (Knappett, 2011; Hodder and Mol, 2015)? Organization studies adopt such a formal approach with the concept of “narrative network” (Pentland and Feldman, 2007) or “action network” (Pentland et al., 2010), in order to visualize and analyze these patterns and routines to understand organizational practices. In a narrative network, each action between actants is a “narrative fragment” (e.g. “the potter forms a vessel on the wheel”). Fragments are linked together in coherent sequences (narratives), much like a chaîne opératoire (e.g. “potter waits for assistant to place clay on wheel—potter forms vessel on the wheel—potter cuts vessel from wheel head and puts it aside”). A narrative is different from the perspective of each of the multiple actants (humans and non‐humans alike). All narratives together, and the links between them, form the narrative network which characterizes that particular
The Organization of Pottery Production 135 organizational routine (e.g. “throwing a vessel on the wheel”). The narrative network can be visualized in a graphical image. Analysis of the network, and comparisons with networks for similar situations elsewhere in place or time, can yield information on which actants and actions affect organizational change the most (Pentland and Feldman, 2007; Hayes et al., 2011; Pentland et al., 2012). Narrative networks and organizational routines can also be analyzed and compared through a variety of statistics for network analysis (Pentland et al., 2010), and through agent‐based simulation (Gao et al., 2014). Of course, this approach cannot be transferred from organizational studies to archaeology as is, but I think it is worthwhile to explore the possibilities it offers for the analysis of archaeological material. As yet, archaeological applications of computer techniques in the study of craft production organization are rare: Brysbaert et al. (2012) discuss how process mining techniques can be used to analyze chaînes opératoires, and perhaps ontological datasets will be key in exploiting the strength of computers to search for meaningful patterns in our entanglements (Hong et al., 2013). An example of the application of agent‐based simulation to the organization of salt mining in Hallstatt can be found in Kowarik et al. (2012); while Rouse and Weeks (2011) use agent‐based modeling to study production specialization in Bronze Age Arabia. This is not the place to present a full discussion of network approaches and related computer techniques, nor of their compatibility with the approaches discussed in this chapter. Useful introductions are published elsewhere: for discussions of formal network analysis in archaeology, see Brughmans (2010, 2013, 2014), Knappett (2011, 2013), Östborn and Gerding (2014), Peeples and Roberts (2013); for introductions to complexity theory and modeling, see Bentley and Maschner (2007), Kohler (2012), Kohler and Van der Leeuw (2007); for introductions to simulation and agent‐based modeling, see Barton (2014) and Lake (2013).
Conclusions In this chapter I have discussed two major traditions in the study of pottery production organization: ceramic ecology and typological approaches. Despite their major contributions to our understanding of pottery production, there are two important shortcomings. First of all, there is an analytical gap between pottery production and the larger social or economic “context.” It is often not clear how a certain type of production is linked to these larger‐scale concepts. The nature of the relations between organizational practices, power, and social inequality should be the subject of our inquiries, not part of the typological label used as explanation. Secondly, typologies link variables such as output, intensity, economic dependence, or labor divisions, while these links should be questioned in each particular case. These issues, as well as the more recent development of approaches focusing on technology and people–thing relations, suggest that the time has come to develop new strategies to study the organization of pottery production. I suggest that such strategies can be built from elements offered by SCOT, cultural technology, behavioral archaeology, holistic approaches, symmetrical archaeology, and entanglement perspectives. I provided brief summaries of each of these different approaches. An approach to the organization of pottery production should view organization as a process, emerging from the specific interactions between people, materials, objects, animals, and so on. It should study organizational processes first on their own merits, rather
136 Kim Duistermaat than as a proxy for the larger sociopolitical or economic context or complexity. We should follow the evidence: organization should be explained from the relations emerging from our data, not used as a label to explain our data. We should adopt an empirical, bottom‐ up perspective, focusing on relations and actions. We should incorporate the multitude of factors influencing organizational processes in an holistic perspective, and allow for active roles of people, materials, and things in a symmetrical manner. This will result in an acknowledgment of the unique historical quality of each case, and in an appreciation of variation in organizational practices rather than a search for universal principles. The approach should ideally also cover the study of relations with related crafts and containers in other media, breaking down traditional boundaries between material categories in field projects. As a limited illustration of these points, I proposed to study the organization of pottery production applying the concept of entanglement through four related strategies, focusing on materials, chaînes opératoires, biographies, and placements in space and time. These strategies can be used to trace the entanglements of all vessels, tools, materials, spaces, and so on, relevant to the organization of pottery production, while keeping an open view on relations that do not fit our traditional concepts. They will also yield information on the relevant social groups or cadenas involved, and how they relate to production decisions. A careful analysis of this multitude of relations will allow us to “reassemble” organizational practices. I used the example of a small bowl to illustrate each strategy. The analysis of these entanglements can take the form of carefully written narratives. I also suggested that it would be worth exploring the possibilities offered by formal network analysis and computing technology. All this will only be possible by fully integrating archaeology, experimental archaeology, and archaeometry, and by enlisting the expertise of different specialists, something that is increasingly done (Pollard and Bray, 2007). Tracing entanglements and analyzing their patterns will be a laborious, time‐consuming project, but I am positive that our analytical methods and techniques are capable of making such a project both feasible and worthwhile. A new strategy does not need to discard all previous insights, but can build on them. Through mapping entanglements of materials, chaîne opératoires, and life‐histories, and by placing them in space and time, many of the variables important for understanding organizational practices (e.g. those listed by Costin, 2005) will come into view. However, by carefully tracing entanglements, we can approach each variable independently, without any preconceived typologies, predefined links between variables, or a priori assumptions on organization. Tracing entanglements is a way to systematically and consciously explore relations and associations in our data, without following only those paths prescribed by models. This may yield new and unexpected understandings and avenues for research. The results of such a study will not yield a cover‐all label to characterize production organization, but rather a detailed and animated narrative. This will not render cross‐cultural comparisons impossible, only more laborious. In any case, I think it is an illusion to think that specific cases grouped under the typological label of, for example, “individual workshop organization” or “attached production” have more in common or are better comparable than cases described in detailed narratives. In conclusion, the time is right to develop new approaches to the study of pottery production organization. A broad variety of theoretical perspectives and practical methods are being developed, including those that promote a radically new perspective on people, things, technology, and their mutual relations. Analytical techniques and methods, both in
The Organization of Pottery Production 137 archaeometry and in computer science, have reached unprecedented levels of precision and strength, and have opened up a wide range of possibilities for studying pottery. I am confident that these developments will contribute to exciting new approaches and discoveries in the field of craft organization.
Notes 1. I do not use the phrase “craft specialization” here. The use of the word “specialization” where actually “organization” is meant, even in basic textbooks (Orton and Hughes, 2013), has caused a lot of confusion and discussion and should be avoided (Clark, 2007a, 2007b vs. Costin, 2007; Hendon, 2007; Smith, 2004: 82–83). Organization and specialization are different processes with different causes and dynamics (Neupert, 2007). 2. I used a typology based on vessel shape. However, for the approach proposed here a typology based on forming techniques and fabric, rather than shape, would have been more useful (Jeffra, 2011b; Roux, 2011). For more reading on categories and typologies, see Fowler (2013), Jervis (2011), Lucas (2012), Shanks (1998), Van Oyen (2013), and Zedeño (2013).
References Abell, N. (2014). “Migration, Mobility and Craftspeople in the Aegean Bronze Age: A Case Study from Ayia Irini on the Island of Kea.” World Archaeology 46(4): 551–568. Alberti, B., Jones, A. M., and Pollard, J. (eds) (2013). Archaeology after Interpretation. Returning Materials to Archaeological Theory (Walnut Creek: Left Coast Press). Arnold, D. E. (1985). Ceramic Theory and Cultural Process (Cambridge: Cambridge University Press). Arnold, D. E. (2008). Social Change and the Evolution of Ceramic Production and Distribution in a Maya Community (Boulder: University Press of Colorado). Arnold, D. E. (2012). “The Social Evolution of Potters’ Households in Ticul, Yucatán, Mexico, 1965–1997.” In: Douglass, J. G. and Gonlin, N. (eds), Ancient Households of the Americas. Conceptualizing What Households Do (Boulder: University Press of Colorado), 163–188. Arnold, D. E., Hutter Wilson, J., and Nieves, A. L. (2007). “Why Was the Potter’s Wheel Rejected? Social Choice and Technological Change in Ticul, Yucatán, Mexico.” In Pool, C. A. and Bey III, G. J. (eds), Pottery Economics in Mesoamerica (Tucson: University of Arizona Press), 59–85. Barton, C. M. (2014). “Complexity, Social Complexity, and Modeling.” Journal of Archaeological Method and Theory 21: 306–324. Baysal, E. (2013). “Will the Real Specialist Please Stand Up? Characterising Early Craft Specialisation, a Comparative Approach for Neolithic Anatolia.” Documenta Praehistorica 40: 233–246. Bentley, R. A. and Maschner, H. D. G. (2007). “Complexity Theory.” In: Bentley, R. A., Maschner, H. D. G., and Chippendale, C. (eds), Handbook of Archaeological Theories (Lanham, MD: AltaMira Press), 245–270. Berg, I. (2004). “The Meanings of Standardisation: Conical Cups in the Late Bronze Age Aegean.” Antiquity 78(299): 74–85.
138 Kim Duistermaat Berg, I. (2007). “Meaning in the Making: The Potter’s Wheel at Phylakopi, Melos (Greece).” Journal of Anthropological Archaeology 26: 234–252. Berg, I. (2013). “The Potter’s Wheel in Mycenaean Greece: A Re‐Assessment.” In: Graziadio, G., Guglielmino, R., Lenuzza, V., and Vitale, S. (eds), Φιλική Συναυλία Studies in Mediterranean Archaeology for Mario Benzi. BAR International Series 2460 (Oxford: Archaeopress), 113–121. Bijker, W. E. (2010). “How Is Technology Made?—That Is the Question!” Cambridge Journal of Economics 34: 63–76. Bijker, W. E., Hughes, T. P., and Pinch, T. (eds) (2012) [1987]. The Social Construction of Technological Systems. New Directions in the Sociology and History of Technology. Anniversary Edition (Massachusetts: Massachusetts Institute of Technology). Bouzakis, K. D., Pantermalis, D., Efstathiou, K., Varitis, E., Paradisiadis, G., and Mavroudis, I. (2011). “An Investigation of Ceramic Forming Method Using Reverse Engineering Techniques: The Case of Oinochoai from Dion, Macedonia, Greece.” Journal of Archaeological Method and Theory 18: 111–124. Braun, D. P. (1983). “Pots as Tools.” In: Morre, J. A. and Keene, A. S. (eds), Archaeological Hammers and Theories (New York: Academic Press), 107–134. Brughmans, T. (2010). “Connecting the Dots: Towards Archaeological Network Analysis.” Oxford Journal of Archaeology 29(3): 277–303. Brughmans, T. (2013). “Thinking through Networks: A Review of Formal Network Methods in Archaeology.” Journal of Archaeological Method and Theory 20: 623–662. Brughmans, T. (2014). “The Roots and Shoots of Archaeological Network Analysis: A Citation Analysis and Review of the Archaeological Use of Formal Network Methods.” Archaeological Review from Cambridge 29(1): 18–41. Brumfiel, E. M. and Earle, T. K. (1987). Specialization, Exchange and Complex Societies (Cambridge: Cambridge University Press). Budden, S. and Sofaer, J. (2009). “Non‐Discursive Knowledge and the Construction of Identity. Potters, Potting and Performance at the Bronze Age Tell of Százhalombatta, Hungary.” Cambridge Archaeological Journal 19(2): 203–220. Burström, N. M. (2014). “Things in the Eye of the Beholder: A Humanistic Perspective on Archaeological Object Biographies.” Norwegian Archaeological Review 47(1): 65–82. Braun, D. P. (1983). “Pots as Tools.” In: Morre, J. A. and Keene, A. S. (eds), Archaeological Hammers and Theories (New York: Academic Press), 107–134. Brysbaert, A. (2007). “Cross‐Craft and Cross‐Cultural Interactions during the Aegean and Eastern Mediterranean Late Bronze Age.” In: Antoniadou, S. and Pace, A. (eds), Mediterranean Crossroads (Athens: Pierides Foundation), 325–359. Brysbaert, A., Bocchi, L., and Tuosto, E. (2012). “Relating Archaeological chaîne opératoire and Process Mining in Computer Science.” Archeologia e Calcolatori 23: 165–186. Brysbaert, A. and Vetters, M. (2010). “Practicing Identity: A Crafty Ideal?” Mediterranean Archaeology and Archaeometry 10(2): 25–43. Carlile, P. R., Nicolini, D., Langley, A., and Tsoukas, H. (eds) (2013). How Matter Matters. Objects, Artefacts and Materiality in Organization Studies (Oxford: Oxford University Press). Clark, J. E. (2007a). “In Craft Specialization’s Penumbra: Things, Persons, Action, Value, and Surplus.” In: Hruby, Z. and Flad, R. (eds), Rethinking Craft Specialization in Complex Societies: Archaeological Analyses of the Social Meaning of Production (Archaeological Papers of the American Anthropological Association) 17(1): 20–35.
The Organization of Pottery Production 139 Clark, J. E. (2007b). “Who’s Minding Production?” Author reply, in: Hruby, Z. and Flad, R. (eds), Rethinking Craft Specialization in Complex Societies: Archaeological Analyses of the Social Meaning of Production (Archaeological Papers of the American Anthropological Association) 17(1): 182–185. Costin, C. L. (1991). “Craft Specialization: Issues in Defining, Documenting, and Explaining the Organization of Production.” In Schiffer, M. B. (ed), Archaeological Method and Theory, vol. 3 (Tucson: University of Arizona Press), 1–56. Costin, C. L. (2001). “Craft Production Systems.” In: Feinman, G. and Price, T. (eds), Archaeology at the Millennium: A Sourcebook (New York: Kluwer Academic/Plenum Press), 273–327. Costin, C. L. (2004). “Craft Economies of the Ancient Andean States.” In: Feinman, G. M., Nicholas, L. M., and Skibo, J. M. (eds), Archaeological Perspectives on Political Economies (Salt Lake City: University of Utah Press), 189–223. Costin, C. L. (2005). “The Study of Craft Production.” In Maschner, H. D. G. and Chippindale, C., Handbook of Archaeological Methods. (Lanham: AltaMira Press), 1032–1105. Costin, C. L. (2007). “Thinking about Production: Phenomenological Classification and Lexical Semantics.” In: Hruby, Z. and Flad, R. (eds), Rethinking Craft Specialization in Complex Societies: Archaeological Analyses of the Social Meaning of Production (Archaeological Papers of the American Anthropological Association) 17(1): 143–162. Coupaye, L. and Douny, L. (2009). “Dans la trajectoire des choses. Comparaison des approches francophones et anglophones contemporaines en anthropologie des techniques.” Techniques & Culture 52–53: 12–39. Crown, P. L. (2007). “Life Histories of Pots and Potters: Situating the Individual in Archaeology.” American Antiquity 72: 677–690. David, N. and Kramer, C. (2001). Ethnoarchaeology in Action (Cambridge: Cambridge University Press). Day, P. M., Relaki, M., and Faber, E. W. (2006). “Pottery Making and Social Reproduction in the Bronze Age Mesara.” In: M. H. Wiener, Warner, J. L., Polonsky, J., and Hayes, E. E. (eds), Pottery and Society. The Impact of Recent Studies in Minoan Pottery. Gold Medal Colloquium in Honor of Philip P. Betancourt (Boston: Archaeological Institute of America), 22–72. Day, P. M., Relaki, M., and Todaro, S. (2010). “Living from Pots? Ceramic Perspectives on the Economies of Prepalatial Crete.” In: Pullen, D. J. (ed), Political Economies of the Aegean Bronze Age. Papers from the Langford Conference, Florida State University, Tallahassee, 22–24 February 2007. (Oxford: Oxbow Books), 205–229. DeMarrais, E. (2013). “Understanding Heterarchy: Crafting and Social Projects in Pre‐Hispanic Northwest Argentina.” Cambridge Archaeological Journal 23(3): 345–362. Dobres, M.‐A. (2000). Technology and Social Agency (Oxford: Blackwell). Dobres, M.‐A. (2010). “Archaeologies of Technology.” Cambridge Journal of Economics 34: 103–114. Dobres, M.‐A. and Hoffman, C. R. (1994). “Social Agency and the Dynamics of Prehistoric Technology.” Journal of Archaeological Method and Theory 1(3): 211–258. Druc, I. (2013). “What Is Local? Looking at Ceramic Production in the Peruvian Highlands and Beyond.” Journal of Anthropological Research 69: 485–513. Duistermaat, K. (2008). The Pots and Potters of Assyria. Technology and Organisation of Production, Ceramic Sequence and Vessel Function at Late Bronze Age Tell Sabi Abyad, Syria (Papers on Archaeology of the Leiden Museum of Antiquities PALMA 4) (Turnhout: Brepols Publishers).
140 Kim Duistermaat Duistermaat, K. (2015). “The Pots of Assur in the Land of Hanigalbat—The Organization of Pottery Production and Hegemonic Practices in the Far West of the Middle Assyrian Empire.” In: Düring, B. (ed), Understanding Hegemonic Practices of the Early Assyrian Empire. Essays Dedicated to Frans Wiggermann (PIHANS 125) (Leiden: Nederlands Instituut voor het Nabije Oosten), 125–152. Fahlander, F. (2013). “Sherlock against Lestrade: A Study in Scale.” In: Bergerbrant, S. and Sabatini, S. (eds), Counterpoint: Essays in Archaeology and Heritage Studies in Honour of Professor Kristian Kristiansen (Oxford: Archaeopress), 637–641. Feinman, G. M. (1999). “Rethinking Our Assumptions: Economic Specialization at the Household Scale in Ancient Ejutla, Oaxaca, Mexico.” In Skibo, J. M. and Feinman, G. M. (eds), Pottery and People. A Dynamic Interaction (Salt Lake City: University of Utah Press), 81–98. Fiaccavento, C. (2013). “Potters’ Wheels from Khirbet al‐Batrawy: A Reconsideration of Social Contexts.” Vicino Oriente 17: 75–103. Fowler, C. (2013). “Dynamic Assemblages, or the Past Is What Endures: Change and the Duration of Relations.” In: Alberti, B., Jones, A. M., and Pollard, J. (eds), Archaeology after Interpretation. Returning Materials to Archaeological Theory (Walnut Creek: Left Coast Press), 235–256. Gagné, L. A. (2012). “Middle Cypriot White Painted Ware: A Study of Pottery Production and Distribution in Middle Bronze Age Cyprus.” PhD thesis, University of Toronto. Gandon, E., Casanova, R., Sainton, P., Coyle, T., Roux, V., Bril, B., and Bootsma, R. J. (2011). “A Proxy of Potters’ Throwing Skill: Ceramic Vessels Considered in Terms of Mechanical Stress.” Journal of Archaeological Science 38(5): 1080–1089. Gao, D., Deng, X., and Bai, B. (2014). “The Emergence of Organizational Routines from Habitual Behaviours of Multiple Actors: An Agent‐Based Simulation Study.” Journal of Simulation 8: 215–230. Goldstein, D. J. and Shimada, I. (2013). “An Experimental and Ecological Approach to Modeling Ancient Fuel Use.” In Damblon, F. (ed), Proceedings of the Fourth International Meeting of Anthracology. BAR International Series 2486 (Oxford: Archaeopress), 121–132. Gosden, C. (2005). “What Do Objects Want?” Journal of Archaeological Method and Theory 12(3): 193–211. Gosden, C. and Marshall, Y. (1999). “The Cultural Biography of Objects.” World Archaeology 31(2): 169–178. Gosselain, O. P. (1998). “Social and Technical Identity in a Clay Crystal Ball.” In: Stark, M.T. (ed), The Archaeology of Social Boundaries (Washington, D.C.: Smithsonian Institution Press), 78–106. Gosselain, O. P. (2000). “Materializing Identities: An African Perspective.” Journal of Archaeological Method and Theory 7(3): 187–217. Gosselain, O. P. (2010). “Ethnographie comparée des trousses à outils de potiers au sud du Niger.” Bulletin de la Société préhistorique française 107(4): 667–689. Gosselain, O. P. (2011). “Technology.” In: Insoll, T. (ed), The Oxford Handbook of the Archaeology of Ritual and Religion (Oxford: Oxford University Press). Greene, A. F. (2013). “The Social Lives of Pottery on the Plain of Flowers: An Archaeology of Pottery Production, Distribution, and Consumption in the Late Bronze Age South Caucasus.” PhD thesis, University of Chicago. Hayes, G. R., Lee, C. P., and Dourish, P. (2011). “Organizational Routines, Innovation, and Flexibility: The Application of Narrative Networks to Dynamic Workflow.” International Journal of Medical Informatics 80: xxx.e161–xxx.e177.
The Organization of Pottery Production 141 Hendon, J. A. (2007). “Production as Social Process.” In: Hruby, Z. and Flad, R. (eds), Rethinking Craft Specialization in Complex Societies: Archaeological Analyses of the Social Meaning of Production (Archaeological Papers of the American Anthropological Association) 17(1): 163–168. Hernes, T. (2008). Understanding Organization as Process. Theory for a Tangled World (London and New York: Routledge). Hicks, D. (2010). “The Material‐Cultural Turn: Event and Effect.” In: Hicks, D. and Beaudry, M. C. (eds), The Oxford Handbook of Material Culture Studies (Oxford: Oxford University Press), 25–98. Hilditch, J. R. (2008). “Reconstruction of Technological Choice, Social Practice and Networks of Exchange from a Ceramic Perspective in the Middle Bronze Age Cyclades.” PhD thesis, University of Exeter. Hilditch, J. R. (2014). “Analyzing Technological Standardization: Revisiting the Minoan Conical Cup.” In: Kotsonas, A. (ed), Understanding Standardization and Variation in Mediterranean Ceramics. Mid 2nd to Late 1st Millennium BC. Babesch Supplements, vol. 25 (Leuven: Peeters), 25–37. Hodder, I. (2011). “Human‐Thing Entanglement: Towards an Integrated Archaeological Perspective.” Journal of the Royal Anthropological Institute (N.S.) 17: 154–177. Hodder, I. (2012a). Entangled. An Archaeology of the Relationships between Humans and Things (Chichester: Wiley‐Blackwell). Hodder, I. (ed) (2012b). Archaeological Theory Today. Second Edition (Cambridge: Polity Press). Hodder, I. and Mol, A. (2015). “Network Analysis and Entanglement.” Journal of Archaeological Method and Theory (online first: doi 10.1007/s10816-015-9259-6). Hollenback, K. L. and Schiffer, M. B. (2010). “Technology and Material Life.” In: Hicks, D. and Beaudry, M. C. (eds), The Oxford Handbook of Material Culture Studies, 313–332. (Oxford: Oxford University Press). Holtorf, C. (2002). “Notes on the Life History of a Pot Sherd.” Journal of Material Culture 7(1): 49–7 1. Hong, Y., Solanki, M., Foxhall, L., and Quercia, A. (2013). “A Framework for Transforming Archaeological Databases to Linked Ontological Datasets.” In: Contreras, F. and Melero, F. J. (eds), Fusion of Cultures. Proceedings of the 38th Annual Conference on Computer Applications and Quantitative Methods in Archaeology, Granada, Spain, April 2010. British Archaeological Reports S2494 (Oxford: Archaeopress), 569–574. Hruby, Z. and Flad, R. (eds) (2007). Special Issue: Rethinking Craft Specialization in Complex Societies: Archaeological Analyses of the Social Meaning of Production (Archaeological Papers of the American Anthropological Association) 17(1). Humphries, C. and Smith, A. C. T. (2014). “Talking Objects: Towards a Postsocial Research Framework for Exploring Object Narratives.” Organization 21(4): 477–494. Ingold, T. (2012). “Toward an Ecology of Materials.” Annual Review of Anthropology 41: 427–42. Jeffra, C. D. (2011a). “The Archaeological Study of Innovation: An Experimental Approach to the Pottery Wheel in Bronze Age Crete and Cyprus.” PhD thesis, University of Exeter. Jeffra, C. D. (2011b). “An Examination of Emic Possibilities: Experimental Archaeology and Cypriot Ceramic Typology.” In: Scarcella, S. (ed), Archaeological Ceramics: A Review of Current Research. BAR International Series 2193 (Oxford: Archaeopress), 103–116. Jervis, B. (2011). “A Patchwork of People, Pots and Places: Material Engagements and the Construction of ‘the Social’ in Hamwic (Anglo‐Saxon Southampton), UK.” Journal of Social Archaeology 11(3): 239–265.
142 Kim Duistermaat Jervis, B. (2013). “Objects and Social Change: A Case Study from Saxo‐Norman Southampton.” In: Alberti, B., Jones, A. M., and Pollard, J. (eds), Archaeology after Interpretation. Returning Materials to Archaeological Theory (Walnut Creek, CA: Left Coast Press), 219–234. Jones, A. (2004). “Archaeometry and Materiality: Materials‐Based Analysis in Theory and Practice.” Archaeometry 46(3): 327–338. Jones, A. M. and Alberti, B. (2013). “Archaeology after Interpretation.” In: Alberti, B. Jones, A. M., and Pollard, J. (eds), Archaeology after Interpretation. Returning Materials to Archaeological Theory (Walnut Creek, CA: Left Coast Press), 15–35. Joy, J. (2009). “Reinvigorating Object Biography: Reproducing the Drama of Object Lives.” World Archaeology 41(4): 540–556. Kelly, G. O. (2009). “Craft Production and Technology during the Iron Age to Early Historic Transition at Kodumanal, Tamil Nadu.” Tamil Civilization 23: 1–14. Kienlin, T. L. (2012). “Beyond Elites: An Introduction.” In Kienlin, T. L. and Zimmermann, A. (eds), Beyond Elites. Alternatives to Hierarchical Systems in Modelling Social Formations (Bonn: Rudolf Habelt Verlag), 15–32. Killick, D. (2004). “Social Constructionist Approaches to the Study of Technology.” World Archaeology 36(4): 571–578. Knappett, C. (2004). “The Affordances of Things: A Post‐Gibsonian Perspective on the Relationality of Mind and Matter.” In: Demarrais, E., Gosden, C., and Renfrew, C. (eds), Rethinking Materiality: The Engagement of Mind with the Material World (Cambridge: McDonald Institute for Archaeological Research), 43–51. Knappett, C. (2011). An Archaeology of Interaction. Network Perspectives on Material Culture and Society (Oxford: Oxford University Press). Knappett, C. (2012). “Materiality.” In: Hodder, I. (ed), Archaeological Theory Today. Second Edition (Cambridge: Polity Press), 176–195. Knappett, C. (2013). “Why Networks?” In: Knappett, C. (ed), Network Analysis in Archaeology. New Approaches to Regional Interaction (Oxford: Oxford University Press), 3–15. Knappett, C. and Malafouris, L. (eds) (2008). Material Agency. Towards a Non‐Anthropocentric Approach (New York: Springer). Knappett, C., Malafouris, L., and Tomkins, P. (2010). “Ceramics (as Containers).” In: Beaudry, M. C. and Hicks, D. (eds), The Oxford Handbook of Material Culture Studies (Oxford: Oxford University Press), 582–606. Kohler, T. A. (2012). “Complex Systems and Archaeology.” In: Hodder, I. (ed), Archaeological Theory Today. Second Edition (Cambridge: Polity Press), 93–123. Kohler, T. A. and Van der Leeuw, S. E. (eds) (2007). The Model‐Based Archaeology of Socionatural Systems (Santa Fe, NM: SAR Press). Kohring, S. (2011). “Social Complexity as a Multi‐Scalar Concept: Pottery Technologies, ‘Communities of Practice’ and the Bell Beaker Phenomenon.” Norwegian Archaeological Review 44(2): 145–163. Kohring, S. (2012a). “Conceptual Knowledge as Technologically Materialised: A Case Study of Pottery Production, Consumption and Community Practice.” In: Stig Sørensen, M. L. and Rebay‐Salisbury, K. (eds), Embodied Knowledge: Historical Perspectives on Belief and Technology (Oxford: Oxbow Books), 106–116. Kohring, S. (2012b). “A Scalar Perspective to Social Complexity: Complex Relations and Complex Questions.” In: Kienlin, T. L. and Zimmermann, A. (eds), Beyond Elites. Alternatives to Hierarchical Systems in Modelling Social Formations (Bonn: Rudolf Habelt Verlag), 327–338.
The Organization of Pottery Production 143 Kolb, C. C. (1989). “Ceramic Ecology in Retrospect: A Critical Review of Methodology and Results.” In: Kolb, C. C. (ed), Ceramic Ecology, 1988: Current Research on Ceramic Materials. BAR International Series 513 (Oxford: British Archaeological Reports), 261–375. Kopytoff, I. (1986). “The Cultural Biography of Things: Commoditization as Process.” In: Appadurai, A. (ed), The Social Life of Things. Commodities in Cultural Perspective (Cambridge: Cambridge University Press), 64–91. Kotsonas, A. (ed) (2014). Understanding Standardization and Variation in Mediterranean Ceramics. Mid 2nd to Late 1st Millennium BC. Babesch Supplements, vol. 25 (Leuven: Peeters). Kowarik, K., Reschreiter, H., and Wurzer, G. P. X. (2012). “Modelling Prehistoric Mining.” Mathematical Modelling 7(1): 17–29. Lake, M. W. (2013). “Trends in Archaeological Simulation.” Journal of Archaeological Method and Theory 21(2): 258–287. LaMotta, V. M. (2012). “Behavioral Archaeology.” In: Hodder, I. (ed), Archaeological Theory Today. Second Edition (Cambridge: Polity Press), 65–87. Laneri, N. (2011). “The Life‐History of the Potter’s Wheel in the Ancient Near East.” In: Scarcella, S. (ed), Archaeological Ceramics: A Review of Current Research. BAR International Series 2193 (Oxford: Archaeopress), 64–72. Latour, B. (2005). Reassembling the Social. An Introduction to Actor‐Network‐Theory (Oxford: Oxford University Press). Law, J. (1992). “Notes on the Theory of the Actor‐Network: Ordering, Strategy, and Heterogeneity.” Systems Practice 5(4): 379–393. Lemonnier, P. (1992). Elements for an Anthropology of Technology. Anthropological Papers no. 88 (Ann Arbor: University of Michigan Press). Lemonnier, P. (ed) (1993). Technological Choices. Transformation in Material Cultures since the Neolithic (London and New York: Routledge). Lemonnier, P. (2012). “Technology.” In: Thieberger, N. (ed), The Oxford Handbook of Linguistic Fieldwork (Oxford: Oxford University Press), 298–316. Leonardi, P. M., Nardi, B. A., and Kallinikos, J. (eds) (2012). Materiality and Organizing. Social Interaction in a Technological World (Oxford: Oxford University Press). Loney, H. L. (2000). “Society and Technological Control: A Critical Review of Models of Technological Change in Ceramic Studies.” American Antiquity 65(4): 646–668. Lucas, G. (2012). Understanding the Archaeological Record (Cambridge: Cambridge University Press). Malafouris, L. (2008). “At the Potter’s Wheel: An Argument for Material Agency.” In: Knappett, C. and Malafouris, L. (eds), Material Agency. Towards a Non‐Anthropocentric Approach (New York: Springer), 19–36. Matson, F. R. (1965). “Ceramic Ecology: An Approach to the Study of the Early Cultures of the Near East.” In: Matson, F. R. (ed), Ceramics and Man (Chicago, IL: Aldine), 202–217. Menon, J. (2008). “Archaeological Problems with Specialization.” Studies in History 24(1): 137–157. Michelaki, K. (2008). “Making Pots and Potters in the Bronze Age Maros Villages of Kiszombor‐Új‐Élet and Klárafalva‐Hajdova.” Cambridge Archaeological Journal 18(3): 355–380. Michelaki, K., Braun, G. V., and Hancock, R. G. V. (2015). “Local Clay Sources as Histories of Human‐Landscape Interactions: A Ceramic Taskscape Perspective.” Journal of Archaeological Method and Theory 22 (3): 783–827.
144 Kim Duistermaat Mills, B. J. and Crown, P. L. (1995). “Ceramic Production in the American Southwest: An Introduction.” In: Mills, B. J. and Crown, P. L. (eds), Ceramic Production in the American Southwest (Tuscon: University of Arizona Press), 1–29. Mímisson, K. and Magnússon, S. G. (2014). “Singularizing the Past: The History and Archaeology of the Small and Ordinary.” Journal of Social Archaeology 14(2): 131–156. Murphy, E. A. and Poblome, J. (2012). “Technical and Social Considerations of Tools from Roman‐period Ceramic Workshops at Sagalassos (Southwest Turkey): Not Just Tools of the Trade?” Journal of Mediterranean Archaeology 25(2): 197–217. Naji, M. and Douny, L. (2009). “Editorial.” Journal of Material Culture 14(4): 411–432. Neupert, M. A. (2007). “Contingency Theory and the Organizational Behavior of Traditional Pottery Production.” In: Skibo, J. M., Graves, M. W., and Stark, M. T. (eds), Archaeological Anthropology: Perspectives on Method and Theory (Tuscon: University of Arizona Press), 138–162. Olsen, B. (2003). “Material Culture after Text: Re‐Membering Things.” Norwegian Archaeological Review 36(2): 87–104. Olsen, B. (2007). “Keeping Things at Arm’s Length: A Genealogy of Asymmetry.” World Archaeology, 39(4): 579–588. Olsen, B. (2010). In Defense of Things (Lanham, MD: Alta Mira Press). Olsen, B. (2012). “Symmetrical Archaeology.” In: Hodder, I. (ed), Archaeological Theory Today. Second Edition (Cambridge: Polity Press), 196–216. Olsen, B. (2013). “Reclaiming Things: An Archaeology of Matter.” In: Carlile, P. R., Nicolini, D., Langley, A., and Tsoukas, H. (eds), How Matter Matters. Objects, Artefacts and Materiality in Organization Studies (Oxford: Oxford University Press), 171–196. Olsen, B., Shanks, M., Webmoor, T., and Witmore, C. (2012). Archaeology. The Discipline of Things (Berkeley: University of California Press). Orlikowski, W. J. (2010). “The Sociomateriality of Organisational Life: Considering Technology in Management Research.” Cambridge Journal of Economics 34: 125–141. Orlikowski, W. J. and Scott, S. V. (2008). “Sociomateriality: Challenging the Separation of Technology, Work and Organization.” The Academy of Management Annals 2(1): 433–474. Orton, C. and Hughes, M. (2013). Pottery in Archaeology. Second Edition (New York: Cambridge University Press). Östborn, P. and Gerding, H. (2014). “Network Analysis of Archaeological Data: A Systematic Approach.” Journal of Archaeological Science 46: 75–88. Papadopoulos, C. and Sakellarakis, Y. (2013). “Virtual Windows to the Past: Reconstructing the ‘Ceramics Workshop’ at Zominthos, Crete.” In Contreras, F. and Melero, F. J. (eds), Fusion of Cultures. Proceedings of the 38th Annual Conference on Computer Applications and Quantitative Methods in Archaeology, Granada, Spain, April 2010. British Archaeological Reports S2494 (Oxford: Archaeopress), 47–54. Patterson, T. C. (2005). “Craft Specialization, the Reorganization of Production Relations and State Formation.” Journal of Social Archaeology 5(3): 307–337. Peacock, D. P.S. (1982). Pottery in the Roman World: An Ethnoarchaeological Approach (New York: Longman). Peeples, M. A. and Roberts, J. M. (2013). “To Binarize or Not to Binarize: Relational Data and the Construction of Archaeological Networks.” Journal of Archaeological Science 40: 3001–3010. Peña, J. T. (2007). Roman Pottery in the Archaeological Record (New York: Cambridge University Press). Pentland, B. T. and Feldman, M. S. (2007). “Narrative Networks: Patterns of Technology and Organization.” Organization Science 18(5): 781–795.
The Organization of Pottery Production 145 Pentland, B. T., Feldman, M. S., Becker, M. C., and Liu, P. (2012). “Dynamics of Organizational Routines: A Generative Model.” Journal of Management Studies 49(8): 1484–1508. Pentland, B. T., Hærem, T., and Hillison, D. (2010). “Comparing Organizational Routines as Recurrent Patterns of Action.” Organization Studies 31(7): 17–40. Pfaffenberger, B. (1992). “Social Anthropology of Technology.” Annual Review of Anthropology 21: 491–516. Pollard, A. M. and Bray, P. (2007). “A Bicycle Made for Two? The Integration of Scientific Techniques into Archaeological Interpretation.” Annual Review of Anthropology 36: 245–259. Pool, C. A. (1992). “Integrating Ceramic Production and Distribution.” In: Bey, G. J., III, and Pool, C.A. (eds), Ceramic Production and Distribution. An Integrated Approach (Boulder, CO: Westview Press), 275–313. Pool, C.A. and Bey, G. J., III (2007). “Conceptual Issues in Mesoamerican Pottery Economics.” In: Pool, C. A. and Bey, G. J., III (eds), Pottery Economics in Mesoamerica (Tucson: University of Arizona Press), 1–38. Ramón, G. and Bell, M. (2013). “Re‐Placing Plainware: Production and Distribution of Domestic Pottery, and the Narration of the Pre‐Colonial Past in the Peruvian Andes.” Journal of Anthropological Archaeology 32: 595–613. Rice, P. M. (1981). “Evolution of Specialized Pottery Production: A Trial Model.” Current Anthropology 22(3): 219–240. Rice, P. M. (ed) (1984). Pots and Potters: Current Approaches in Ceramic Archaeology (Los Angeles: University of California Press). Rice, P. M. (1987). Pottery Analysis. A Sourcebook (Chicago and London: University of Chicago Press). Rice, P. M. (1996). “Recent Ceramic Analysis: 2. Composition, Production, and Theory.” Journal of Archaeological Research 4(3): 165–202. Rouse, L. M. and Weeks, L. (2011). “Specialization and Social Inequality in Bronze Age SE Arabia: Analyzing the Development of Production Strategies and Economic Networks Using Agent‐Based Modelling.” Journal of Archaeological Science 38: 1583–1590. Roux, V. (2003a). “A Dynamic Systems Framework for Studying Technological Change: Application to the Emergence of the Potter’s Wheel in the Southern Levant.” Journal of Archaeological Method and Theory 10(1): 1–30. Roux, V. (2003b). “Ceramic Standardization and Intensity of Production: Quantifying Degrees of Specialization.” American Antiquity 68(4): 768–782. Roux, V. (2010). “Technological Innovations and Developmental Trajectories: Social Factors as Evolutionary Forces.” In O’Brien, M. J. and Shennan, S. J. (eds), Innovation in Cultural Systems. Contributions from Evolutionary Anthropology (Cambridge, MA: The Massachusetts Institute of Technology Press), 217–234. Roux, V. (2011). “Anthropological Interpretation of Ceramic Assemblages: Foundations and Implementations of Technological Analysis.” In: Scarcella, S. (ed), Archaeological Ceramics: A Review of Current Research. BAR International Series 2193 (Oxford: Archaeopress), 80–88. Roux, V. (2013). “Spreading of Innovative Technical Traits and Cumulative Technical Evolution: Continuity or Discontinuity?” Journal of Archaeological Method and Theory 20: 312–330. Schiffer, M. B. (2007). “Some Thoughts on the Archaeological Study of Social Organization.” In: Skibo, J. M., Graves, M.W., and Stark, M. T. (eds), Archaeological Anthropology: Perspectives on Method and Theory (Tucson: University of Arizona Press), 57–71.
146 Kim Duistermaat Schiffer, M. B. (2011). Behavioral Archaeology. Principles and Practice (London and New York: Routledge). Schiffer, M. B., Skibo, J. M., Griffitts, J. L., Hollenback, K. L., and Longacre, W. A. (2001). “Behavioral Archaeology and the Study of Technology.” American Antiquity 66(4): 729–737. Schortman, E. M. and Urban, P. A. (2004). “Modeling the Roles of Craft Production in Ancient Political Economies.” Journal of Archaeological Research 12(2): 185–226. Shanks, M. (1998). “The Life of an Artifact in an Interpretive Archaeology.” Fennoscandia archaeologica 15: 15–30. Shimada, I. and Craig, A. K. (2013). “The Style, Technology and Organization of Sicán Mining and Metallurgy, Northern Peru: Insights from Holistic Study.” Revista de Antropología Chilena 45 (1): 3–31. Shimada, I. and Wagner, U. (2007). “A Holistic Approach to Pre‐Hispanic Craft Production.” In: Skibo, J. M., Graves, M. W., and Stark, M.T. (eds), Archaeological Anthropology. Perspectives on Method and Theory (Tuscon: University of Arizona Press), 163–197. Sillar, B. (2000). “Dung by Preference: The Choice of Fuel as an Example of How Andean Pottery Production Is Embedded within Wider Technical, Social, and Economic Practices.” Archaeometry 42(1): 43–60. Sillar, B. and Tite, M. S. (2000). “The Challenge of ‘Technological Choices’ for Materials Science Approaches in Archaeology.” Archaeometry 42(1): 2–20. Sinopoli, C. M. (1998). “Identity and Social Action among South Indian Craft Producers of the Vijayanagara Period.” Archeological Papers of the American Anthropological Association 8: 161–172. Sinopoli, C. M. (2003). The Political Economy of Craft Production: Crafting Empire in South India, c. 1350–1650 (New York: Cambridge University Press). Skibo, J. M. (2013). Understanding Pottery Function (New York: Springer). Skibo, J. M. and Schiffer, M. B. (2008). People and Things. A Behavioral Approach to Material Culture (New York: Springer). Smith, M. E. (2004). “The Archaeology of Ancient State Economies.” Annual Review of Anthropology 33: 73–102. Smogorzewska, A. (2007). “Technological Marks on Pottery Vessels. Evidence from Tell Arbid, Tell Rad Shaqrah and Tell Jassa el‐Gharbi (Northeastern Syria).” Polish Archaeology in the Mediterranean 19: 555–564. Sofaer, J. (2006). “Pots, Houses and Metal: Technological Relations at the Bronze Age Tell at Százhalombatta, Hungary.” Oxford Journal of Archaeology 25(2): 127–147. Sofaer, J. and Budden, S. (2012). “Many Hands Make Light Work: Potting and Embodied Knowledge at the Bronze Age tell at Százhalombatta, Hungary.” In: Stig Sørensen, M. L. and Rebay‐Salisbury, K. (eds), Embodied Knowledge: Historical Perspectives on Belief and Technology (Oxford: Oxbow Books), 117–127. Stissi, V. V. (2012). “Giving the kerameikos a Context: Ancient Greek Potters’ Quarters as Part of the polis Space, Economy and Society.” In: Esposito, A. and Sanidas, G. M. (eds), “Quartiers” artisanaux en Grèce ancienne. Une perspective Méditerranéenne (Villeneuve d’Ascq: Presses Universitaires du Septentrion), 201–230. Stockhammer, P. W. (2012). “Performing the Practice Turn in Archaeology.” Transcultural Studies 1: 7–42. Tite, M. S. (2008). “Ceramic Production, Provenance and Use—a Review.” Archaeometry 50(2): 216–231.
The Organization of Pottery Production 147 Torchy, L. and Gassin, B. (2010). “Le travail de la poterie en contexte chasséen: des outils en silex pour la production céramique?” Bulletin de la Société préhistorique française 107(4): 725–735. Van As, A. (2004). “Leiden Studies in Pottery Technology.” Leiden Journal of Pottery Studies 20: 7–22. Van der Leeuw, S. E. (1977). “Towards a Study of the Economics of Pottery Making.” In: van Beek, B. L., Brandt, R. W., and Groenman‐Van Waateringe, W. (eds), Ex Horreo (Amsterdam: University of Amsterdam), 68–76. Van der Leeuw, S. E. (2008). “Agency, Networks, Past and Future.” In: Knappett, C. and Malafouris, L. (eds), Material Agency. Towards a Non‐Anthropocentric Approach (New York: Springer), 217–247. Van der Leeuw, S. E. and Pritchard, A. C. (eds) (1984). The Many Dimensions of Pottery (Amsterdam: University of Amsterdam). Van Gijn, A. and Lammers‐Keijsers, Y. (2010). “Toolkits for Ceramic Production: Informal Tools and the Importance of High Power Use‐Wear Analysis.” Bulletin de la Société préhistorique française 107(4): 755–762. Van Oyen, A. (2013). “Towards a Post‐Colonial Artefact Analysis.” Archaeological Dialogues 20: 81–107. Walker, W. H. and Schiffer, M. B. (2006). “The Materiality of Social Power: The Artifact‐Acquisition Perspective.” Journal of Archaeological Method and Theory 13(2): 67–88. Watts, C. (2008). Pot/Potter Entanglements and Networks of Agency in Late Woodland Period (c. AD 900–1300) Southwestern Ontario, Canada. British Archaeological Reports International Series no. 1828 (Oxford: Hedges). Watts, C. (ed) (2013). Relational Archaeologies. Humans, Animals, Things (London and New York: Routledge). Webmoor, T. (2007). “What About ‘One More Turn after the Social’ in Archaeological Reasoning? Taking Things Seriously.” World Archaeology 39(4): 563–578. Wiggermann, F. A. M. (2008). “Cuneiform Texts from Tell Sabi Abyad Related to Pottery.” In: Duistermaat, K., The Pots and Potters of Assyria. Technology and Organisation of Production, Ceramic Sequence and Vessel Function at Late Bronze Age Tell Sabi Abyad, Syria (PALMA 4) (Turnhout: Brepols), 559–564. Witmore, C. L. (2007). “Symmetrical Archaeology: Excerpts of a Manifesto.” World Archaeology 39(4): 546–562. Zedeño, M. N. (2013). “Methodological and Analytical Challenges in Relational Archaeologies: A View from the Hunting Ground.” In: Watts, C. (ed), Relational Archaeologies. Humans, Animals, Things (London and New York: Routledge), 132–153.
Chapter 10
Provenanc e St u di e s Productions and Compositional Groups Yona Waksman In memoriam Maurice Picon Initial Definitions and Principles “Provenance studies” is the common expression used by archaeological scientists to designate analytical investigations, that aim at identifying the place of manufacture of archaeological artifacts, or the location of the raw materials sources used to manufacture them. Such investigations are not new as, according to Harbottle (1976), they were introduced by Fouqué (1879), who studied ceramics found in Santorin petrographically in order to distinguish local artifacts from imports. They have been carried out by many researchers since (e.g. Catling et al., 1963; Shepard, 1963; Perlman and Asaro, 1969; Jones, 1986; Peacock and Williams, 1986; Picon, 1993; Mannoni, 1994; Schneider, 2000; Speakman and Glascock, 2007; see also Tite, Chapter 2, this volume, for further history of research). In spite of its widespread use and general acceptance, the term “provenance studies” may be ambiguous, if not misleading (Hunt, 2012). Therefore, we would like to begin by defining our terminology before addressing some of the issues related to “provenance studies.”1 This chapter mainly deals with ceramics considered representative for a given production, which is part of the output of a workshop or group of related workshops.2 A production may include different types of wares and last for variable periods of time, but it always corresponds to similar clayey material or ceramic paste. The characteristics of the paste are those of the final product; they integrate both those of the raw materials and their potential changes due to processing by the potters, subsequent changes being considered independently. A single pottery workshop may manufacture several different productions corresponding, for instance, to different functional categories, such as table and cooking wares, the latter having technical constraints requiring a more careful selection of raw materials (Picon, 1995; Tite et al., 2001). Conversely, it may be difficult to distinguish within an “area of uncertainty” the output of several workshops working in the same technical tradition, exploiting raw materials within the same geological formation presenting similar geochemical and mineralogical features, sometimes over long distances (Picon, 1993). The place of manufacture of an artifact is defined as its origin, and the location where it was recovered archaeologically as its provenance. Provenance studies, as defined here,
Provenance Studies 149 designate the procedures and reasoning that aim at attributing archaeological ceramics to their origin, and by extension to a predefined production, based on petrographic or chemical analysis.3 Provenance studies rely on the postulate that, within a given production, ceramics share similar features that differentiate them from ceramics belonging to other productions. Chemical and petrographic analyses of a representative sample for a production enable us to define it analytically as a compositional group, and to separate it from others, provided that they are not included in the same area of uncertainty. The comparison of the analytical features of ceramics of unknown origin with those of predefined productions enables us to test hypotheses of attribution, and to identify origins when these productions are localized. Localized productions stricto sensu correspond to workshops attested archaeologically, if not by their structures (e.g. pottery kilns), at least by evidence related to the production process, which often include reference samples of undoubted local origin (e.g. pottery wasters).4 Whenever such evidence is absent, field prospecting, geological maps and reports, and textual documents concerning craftsmanship may give indications about possible locations of the raw materials’ sources. In general, provenance studies require appropriate comparative data, usually organized in databases, in order to test hypotheses of origin based on archaeological criteria.
Implementation Contexts of Use and Related Issues Study of a Site The most common case is one of an archaeological site where no pottery workshops were found, and where ceramics were initially classified according to typology and fabric.5 Laboratories may be requested to test this field classification by defining productions according to analytical criteria, to determine which ceramic categories belong to the same production, and to evaluate the choice in raw materials. Whenever appropriate comparative data are available, issues concerning the local status of productions, the consumption of imported ceramics at the site, and its location within exchange and commercial networks can be addressed. These latter questions require reference groups stricto sensu to attribute ceramics to their origin, and more generally comparative compositional groups to attribute artifacts to predefined productions, and to determine whether the site is included in their respective area of diffusion. Compatibility with a local origin or, on the other hand, imported status might be inferred from local geological features illustrated on geological maps, especially when using petrographic analysis. However, local attribution usually requires either a thorough field survey of the area,6 to investigate available raw materials and their specificity (see, e.g., Gauss and Kiriatzi, 2011), or evidence of local production, which brings us to the next point.
Study of a Workshop Whenever evidence of pottery production is found, the corresponding reference group(s) may be defined, and the following issues addressed: which ceramic categories
150 Yona Waksman and types belong to the local repertoire? to the same local production? choice of specific raw materials and reasons for this choice (technical? cultural? etc.), chronological evolution in raw materials procurement or processing, specialization of production, and so on. Reference groups may also enable the identification of local wares which are not attested archaeologically (e.g. wares pre-or post-dating the available evidence of production, see for example below, Fatimid wares in “A Case Study: The Medieval Productions of Beirut”), provided that they correspond to similar choices in raw material procurement and processing. Investigations of the diffusion of pottery from a particular workshop require analysis of sherds from consumption sites and open another area of inquiry: which part of the production is exported? what is the area of diffusion of the workshop? is it concurrent with other workshops manufacturing similar wares? what is its part in the market, and how does it evolve? and so on.
Study of a Ceramic Type Another approach in provenance studies consists in investigating a specific ceramic type: was it manufactured in several workshops? can we distinguish the different productions? did some of the workshops dominate the market? what is their respective contribution to the procurement of a given site? are there geographical or chronological trends? was this type manufactured in specialized workshops, or in workshops manufacturing other types as well? was it diffused as byproduct of another ware? and so on. These different approaches in provenance studies are complementary but, in practice, are rarely addressed simultaneously. Most studies correspond to, or may be broken down into, one of these cases, which conditions the sampling strategy and the interpretation of data.
Sampling Sampling is an important part of the process in provenance studies, and determines, to a large extent, what may be expected from the results of analysis. The selection of samples for analysis is closely connected to the archaeological data (context and questions, see previous paragraph), the archaeometric data (especially availability of comparative analyses), and, obviously, the available material.7 We detail below the initial steps of sample selection leading to the analytical definition of productions; further discussion about sampling for chemical analysis is also provided.
Defining Unlocalized Production When defining productions with unknown geographic locations, we are generally evaluating whether or not one or more ceramic types with similar fabric belong to a single production. The samples, which may come from different sites, should be selected from among the most common examples of the ceramic types. A frequent mistake is selecting atypical samples whose attribution on the basis of macroscopic criteria is difficult. In practice, the “core” of the production has to be defined before variants and marginals can be considered.
Provenance Studies 151
Figure 10.1 Local reference samples, late Byzantine workshops, Thessaloniki, Greece. They include kiln furniture-tripod stilts, some still attached to ceramic bases (left), overfired wasters, unfinished biscuit-fired wasters (right).
Samples should also be selected among well-documented ceramics, in terms of archaeological context, drawing, and photographs. Sherds should have enough form and decoration to be representative of an identified typological category. Our ability to extrapolate the results obtained from a batch of samples to an entire ceramic production critically depends upon these factors.
Defining Localized Production When analyzing ceramics from a site where evidence of pottery production is available, it is essential to sample reference materials of undoubtedly local origin: pottery wasters rejected at various stages of the manufacturing process, such as unfinished or overfired wares;8 kiln furniture made with the same material as pottery, such as saggars, tripod stilts, kiln bars (Figure 10.1).9 Evidence of production may be difficult to discern in the archaeological record, especially in contexts that do not include structures related to pottery manufacture, such as kilns. In these cases, it is critical to carefully check the status of potential wasters, especially overfired sherds which may come from hearths or burned destruction layers and not from workshops. In order to complete the definition of a production, samples representative of the typological repertoire of finished products, whether shown by similar wasters to be local or still
152 Yona Waksman awaiting confirmation of their local status, should also be selected, using the same criteria as in the case of non-localized production. Common assumptions may sometimes be misleading, as shown by the examples of sherds found in the filling of a kiln representing a later depositional event and not corresponding to its output; and of imported ceramics present in workshop contexts, possibly coming from nearby warehouses. Additionally, finished products are not always present in workshop contexts in large quantities, requiring a second step in the sampling process. Pottery from consumption sites will be sampled both to study the diffusion of the production and to complement its definition. Consumption contexts may significantly contribute to our knowledge of its repertoire and dating, and to the building up of its typo-chronology (see the the section “A Case Study: The Medieval Productions of Beirut”). In instances where several productions were manufactured in the same workshop (different functional categories, evolution of clay procurement or processing in time, etc.), each production has to be defined independently. It is often necessary in these cases to supplement the initial sampling in order to obtain statistically significant compositional groups for each production. The number of samples required to define a production may vary according to the homogeneity of the raw materials, the degree of standardization of its processing, the coarseness of the fabric, and so on. An empirical rule of thumb for chemical analysis is to take a first sample of about 20 individuals and then determine if further sampling is requested. An iterative procedure should be carried out as much as possible: initial sampling, analyses, examination of the results, and further sampling whenever necessary. The latter may be needed, for instance, when too many marginals were present in the initial sampling, or in the case of heterogeneous material or poorly standardized processing, or to strengthen less well-defined groups in case of multiple productions. Correlation between the initial criteria (typology, fabric) and members of compositional groups should guide supplementary sampling. The attribution of ceramics to predefined productions is less demanding in terms of number of samples, as we are comparing them to a well-defined group to which they will add (or not). However, we usually consider a group of sherds rather than a single example, as the latter might present chance similarities or dissimilarities.
Productions and Compositional Groups Classifications and Attributions Depending upon the type of analysis, the initial analytical results typically consist of tables of quantitative data (chemical analysis) or descriptions based on mostly qualitative data (petrographic analysis).10 The next steps are generally carried out in two phases: classification and attribution. In the case of chemical analysis, the large number of variables requests the use of multivariate statistical tools such as cluster and discriminant analysis. In the classification phase, samples are clustered, according to analytical criteria, into groups considered homogeneous and distinct enough to correspond to different productions. The evaluation of the latter point is part of data interpretation, an important part of the work of archaeological scientists, as opposed to analysts. In the classification phase are determined: how many groups are present; what are their boundaries; should marginal
Provenance Studies 153 samples be considered part of a group or excluded; and, mostly, are the similarities and differences observed significant in terms of production. This interpretation is based on a number of factors, discussed in some detail in the section “Classifying and Defining Productions, Using Chemical Analysis”, most of which cannot be taken into account by statistical tools. Attribution supposes that the productions to which the samples under study potentially belong have been defined previously using archaeometric analysis. The attribution phase involves comparison of analytical characteristics, between samples to be attributed and predefined productions. Both cluster and discriminant analysis may be used with quantitative data, although the latter is theoretically better suited to studies in which there is a choice among a finite number of hypotheses. The combination of petrographic and chemical approaches may be useful, because these two approaches access different types of information (e.g. Day et al. 1999; Schneider, 2000).11 Petrographic data are more closely connected to fabrics and technological features on the one hand and to geological features on the other than chemical data. Petrographic data may indicate potential raw materials origin even when a priori hypotheses and reference groups are lacking. On the other hand, mineralogical associations and petrographic features may be too common to be indicative. Chemical analyses are usually more discriminating, and generate quantitative data which are better adapted to statistical analysis.
Classifying and Defining Productions, Using Chemical Analysis Chemical analysis is considered here as a mean to characterize ceramic productions by determining the bulk composition of ceramic pastes. Given that ceramic pastes are composite and heterogeneous by nature (see, e.g., Montana, Chapter 7, this volume), the sample taken from a sherd for analysis, and more precisely the volume analyzed, should be large enough to be representative of the material. In addition, because statistical analyses are generally used to compare the chemical data from the current investigation and previous chemical data stored in the laboratory database, it is essential that the analytical method used provides reproducible results. Homogenized samples (analyzed as powders, pellets, or beads) and stable experimental conditions are critical in this respect. A production is defined chemically by the distribution of elemental concentrations in ceramic pastes within a batch of representative samples. The assumption that distributions are normal, or log-normal, is often implicit in multivariate statistical treatments. This assumption is not always verified, especially for small batches of samples, or for productions corresponding to ellipsoidal rather than spherical distributions (see, e.g., “A Case Study: The Medieval Productions of Beirut”).12 A distribution is often summarized by its mean value (m) and standard deviation (σ), which characterizes the dispersion of values around m. Variable dispersions may be observed within a single production, depending on the element, the mineralogical features of the components, the geological context from which they derive, and so on. For example, the CaO content in calcareous pastes may vary by more than 10% in absolute weight percent; Ti concentrations above 1% TiO2 in kaolinite- rich pastes may be fairly variable too; in clayey materials related to ultrabasic contexts, concentrations of Cr and Ni may also vary widely from sample to sample within a range of high to very high values. For the same element, a large dispersion may be considered “normal” in one geochemical context and “abnormal” in another, in the latter case indicative for
154 Yona Waksman instance of more than one production or of a pollution. As a result, compositional groups may be defined at different levels of aggregation within the same hierarchical clustering analysis. Each chemical element, or group of related elements in Mendeleiev’s periodic table, may have a different behavior and be affected differently by a variety of factors. As far as we know, there is no general rule for these behaviors. Instead, from raw materials to statistical treatment of the data, there are combinations of geochemical, technological, and analytical parameters which archaeological scientists have gradually come to recognize and integrate in their interpretative schema. Lines of interpretation include, but are not necessarily limited to, the following (see also Maggetti, 1982): • geological features of the initial environment, whenever the latter is known, and generally speaking, the nature of the material as a geological material, which will condition its mineralogical and geochemical features • technological aspects: raw material processing (tempering, refining, mixing), shaping, decoration, firing, and so on. • alterations due to the use of the object • post-depositional alterations • analytical precision and biases related to instrumental or analytical aspects • biases introduced by statistical analysis • last but not least, archaeological data. In our opinion, the latter should be taken into account while interpreting analytical data, which does not mean that the archaeological scientist is influenced by possible expectations about the results. Although a limited number of parameters may be formalized and integrated in statistical analysis (e.g. Beier and Mommsen, 1994), a large part is left for the archaeological scientists to determine connections between the patterns observed in the data and their possible causes. In our opinion, which follows Picon’s, multivariate statistics should be considered as a guide to interpretation, and not a substitute for it. Statistical analyses are very useful for highlighting the structure of the data and provide convenient outputs for the graphical presentation of results, but it is always necessary to come back to the initial individual chemical compositions of the samples for interpretation.
A Case Study: The Medieval Productions of Beirut Large-scale excavations were carried out in Beirut after the civil war, prior to the reconstruction of its city center. Remains of pottery workshops from the Roman and Crusader periods were unearthed (François et al., 2003; Reynolds et al., 2008–2009), providing us with the opportunity to analytically define the corresponding productions. Research on medieval wares, carried out at the Laboratoire de Céramologie in Lyon,13 are summarized below as an illustration of different aspects of provenance studies.
Provenance Studies 155
Beirut red-paste wares
wasters and kiln furniture ceramics from kiln contexts
Beirut buff-paste wares
Figure 10.2 Beirut medieval wares: main compositional groups as determined by hierarchical clustering analysis, and corresponding wares. Local productions, identified thanks to reference samples stricto sensu (black dots), complemented with sherds coming from pottery workshop contexts (gray dots), correspond to different compositional groups and sub-groups. Two very different clayey materials were used: low to moderately calcareous for red-paste cooking and table wares (part of the classification shown at the top and bottom left), and highly calcareous for table wares of a different technical tradition (bottom right) (after Waksman 2011).
Definition of Reference Groups The first step in provenance studies in workshop contexts is sample selection for the definition of local reference groups. Beirut workshop contexts provided local reference samples, chosen among pottery wasters (biscuit-fired unfinished wares, over-fired wares) and kiln furniture (kiln bars). Additional sampling of the medieval ceramic corpus was conducted to include representatives of most pottery types present in large quantities in these workshop contexts. The classification according to chemical compositions of the samples distinguished two main groups; one containing several loosely defined subgroups (Figure 10.2). Local reference samples are spread among all the different groups and subgroups, confirming the local status of the whole sampling considered, with some noticeable exceptions such as a Cypriot import found in a workshop context (not included in the classification Figure 10.2). Two very different raw materials were used by the potters. One group is composed of low to moderately calcareous pastes, of variable chemical compositions but sharing a number of common features, such as low alkali and fairly high to high iron and titanium contents. In this specific case, compositions are too variable to be well represented by a single mean and standard deviation. The classification tends to split the samples in a large number of
156 Yona Waksman subgroups, but all these subgroups follow roughly two main geochemical behaviors, indicated by inter-element correlations and associated with their calcium contents (Figure 10.3, low CaO and mid CaO groups). Mahalanobis distances, which are able to take such correlations into account, may be preferred to Euclidean distances in this case. The clayey materials corresponding to Beirut red-paste group (Figure 10.2) were used to manufacture glazed cooking wares, for the less calcareous ones, and slipped and glazed table wares presenting with various decoration techniques (plain glazed, reserved-slip, slip-painted, and sgraffito wares). The other main chemical group corresponds to a production of buff-paste table wares following a completely different technical tradition, which is based on the association of highly calcareous pastes (around 30% CaO) and alkali or lead-alkali glazes (making possible the turquoise color). This technical tradition is more specifically related to the Islamic world. The first part of the study enabled us to constitute reference groups, and suggests that the choice of raw materials corresponds either to technical criteria (use of low-calcareous clays for cooking wares and of high-calcareous clays to favor bonding of the alkali and lead-alkali glaze) or to cultural ones (symbolic value of turquoise). It also demonstrated that the traditional typological categories (plain glazed, reserved-slip, slip-painted, and sgraffito wares), usually considered separately in archaeological publications, are in fact part of the same production, which represents a significant change in perspective when studying commercial fluxes.
Diffusion of Beirut Wares and Complements of Definition Beirut red wares were already known archaeologically, and partially defined analytically, before the Beirut workshops were unearthed (Stern and Waksman, 2003; Waksman et al., 2008). In these early studies,14 compositional groups corresponding to the, as yet unlocated, productions were defined using a sample from consumption contexts, and especially from Crusader Acre. They helped approaching the diffusion of Beirut wares, widespread in the Eastern Mediterranean and occasional in the Western Mediterranean and the Black Sea regions. Petrographic analysis carried out by Porat identified lower Cretaceous formations, extending through Lebanon, Israel, and Jordan, as the likely source of the clayey material used to manufacture these productions (Waksman et al., 2008). The discovery of the workshops, and subsequent inclusion of Beirut reference samples into the compositional groups defined by the early studies, set their exact location to Beirut (Waksman, 2002). Consumption sites also provided valuable information about the Beirut productions: precise chronological data; variability of the typological repertoire; evidence on which types from the production were preferentially exported; quantities of Beirut wares found in consumption contexts, and thus the relative part that Beirut played in the sites’ procurement; identification of other productions associated in the same trade networks; and so on. These “complements of definition” are, for instance, well illustrated by the contribution of Stern’s studies of the pottery of Crusader Acre to our knowledge of Beirut wares (Stern and Waksman, 2003; Stern, 2012). Stern established their typochronology during the Crusader period, and, through quantification of pottery finds, pointed out their massive presence in the region. Contrary to still common prejudices, cooking wares constitute the part of Beirut productions which experienced the largest diffusion.
Provenance Studies 157
14 12 10
Fe2O3 (%)
8 Low CaO 6
Mid CaO High CaO
4 2 0
0
20
40
60
80
100
SiO2 (%)
Number of samples
20
15
6 5
LEV212 BZY615 BYZ 94 LEV 29 BYZ 93 LEV 72 BYZ 95 LEV 77
3 1 0
0
BZY620 LEV233 BZY 96 BZY814 LEV 33 LEV 71 LEV334 LEV 34 BZY688 BZY619 LEV 79 LEV 78 LEV 30 LEV 73 LEV336 LEV321 LEV 14 BYZ816 LEV 76 LEV 28
LEV207 LEV209 LEV211 LEV208 LEV 2 LEV329 LEV 70 FBC 37 LEV 8 LEV 31 LEV210 LEV 78 LEV215 DYZ101 LEV337 LEV 74 BZY687 LEV332 LEV105 LEV333 BZY815
LEV213 LEV214 FBC 31 LEV103 LEV232
LEV548
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 Mahalanobis distance
Figure 10.3 Beirut medieval wares: binary plot iron–silicon (top) and histogram of Mahalanobis distances (bottom). The low-calcareous pastes show large variations in absolute concentrations values, but also strong inter-elements correlations, related to a same geochemical behavior and interpreted as a same production (top). Fatimid cooking wares coming from terrestrial contexts (in black) may be attributed to Beirut using discriminant analysis, as they fit the low-calcareous Beirut reference group (in gray), whereas the attribution of sample LEV548, coming from the Serçe Limanı shipwreck and altered by sea water, requests to come back to the initial chemical data (bottom).
158 Yona Waksman
Attribution to Beirut of Other Types Pottery production also existed in Beirut at periods which lack the corresponding archaeological evidence, such as workshops and kilns. The reference groups defined for the Crusader productions may be used to identify other local products, provided that their manufacture involved similar raw materials and processing. For example, Beirut Fatimid wares, which just precede the Crusader ones, can be positively identified as locally manufactured in Beirut by their chemical similarity with the Crusader reference groups (Waksman, 2011). Figure 10.3 presents the histogram of Mahalanobis distances between a sampling of Fatimid cooking wares and the Crusader low-calcareous reference group. Cooking wares from terrestrial excavations are well integrated in the reference group on the histogram, unlike examples of similar typology coming from the Serçe Limanı shipwreck (Bass and van Dorninck, 1978), one of which (LEV548) is shown in Figure 10.3. On closer examination, the shipwreck samples can be attributed to Beirut as well, provided that the chemical alterations associated with marine environment are taken into account (Pradell et al., 1996; Waksman, 2011). This is an obvious example of why coming back to the unprocessed chemical data is critical. The Serçe Limanı shipwreck, well dated thanks to glass stamps and coins, is an important chronological reference point for the Fatimid period. In the present case, it helped identify the Beirut Fatimid repertoire, including a type of sgraffito ware previously attributed to Egypt, Palestine, or North Africa (Jenkins, 1992; Mason, 2004). The results also suggested a Levantine origin for the main cargo of the shipwreck, consisting of glass cullets. Laboratory investigations of Beirut medieval productions provide a concrete example of the following aspects of ceramic provenance studies: • definition of local productions (in this case, multiple productions) and establishment of chemical reference groups, based on the analysis of pottery wasters and kiln furniture together with finished products. It is notable that these productions could be defined before the discovery of the workshops, which established their location, and that petrographic analysis indicated potential areas of origin • the diffusion study contributed to the typo-chronological definition of the productions and provided information about trade networks and fluxes • reference groups were also used to attribute to Beirut Fatimid wares whose local status was not attested archaeologically
Concluding Remarks Provenance studies of ceramics may appear a well-established research area. However, instrumental developments, in portable equipment (e.g. portable X- ray fluorescence devices), and in large research infrastructures (synchrotron), have stimulated new field and laboratory practices. The latter may provide some interesting results (e.g. Molera et al., 2013), but also urge us to remember fundamental constructs, such as the complex nature of
Provenance Studies 159 ceramic material and the range of parameters involved in the interpretation of analytical results, sampling procedures, and archaeological issues. Engagement with the latter is fruitful in provenance studies of ceramics owing to the need to build up the corpus of geochemical and petrographic data adapted to specific archaeological questions. Furthermore, unlike other materials, we cannot only reason in terms of localization of raw materials. The ubiquity of clays, their processing by potters, and the composite nature of ceramic pastes imply that we should rather think in terms of the final output; that is, the productions, defined by both their archaeological and archaeometric features. The development of provenance and technological studies of ceramics has generated large chemical and petrographic databases in different laboratories. An important issue currently facing the discipline concerns the status of archaeometric data and good practices for its sensible use, depending on the level of archaeological information recorded in archaeometry databases. This issue involves both the safeguarding of “historical” databases and the interconnection of ceramics databases containing complementary information.15 The building up of such networks, using IT and statistical tools of complex data modeling, sharing, and handling,16 may open new perspectives in the field.
Notes 1. An extended version may be found in Waksman (2014). 2. The term “workshop” does not suppose here a specific level of organization of the production unit (Peacock, 1982). The concept of “production,” however, implies a certain degree of standardization in raw material selection and processing. 3. We use here “chemical” analysis in the sense of “elemental” analysis. 4. Compositional groups including reference samples become reference groups. 5. i.e. the features of the paste as observed with the naked eye, using a hand lens, or under the binocular microscope; see, for example, Tomber and Dore (1998). 6. In most cases clay materials are expected to come from the close surroundings of a production site (Arnold, 1985), although some rare examples of clay transport over long distances are known (Ballet and Picon, 1987). 7. In practice, it often comes as a limitation when trying to define productions using material from a single site. 8. Overfired wasters are more likely to be chemically altered during burial (Picon, 1987), so that other reference samples should be preferred whenever possible. 9. Fragments of kiln walls were shown to be inadequate in the cases we examined, as different raw materials had to be selected for them. 10. Quantitative approaches to petrographic data are seldom used (Whitbread, 1991). 11. Other analytical methods, such as heavy mineral analysis and X-ray diffraction, may be useful as well. 12. In an n-dimension hyperspace, n being the number of elements analyzed. 13. Analysis is carried out by wavelength-dispersive X-ray fluorescence. Twenty-four elements are determined, seventeen of which, including major and minor elements (MgO, Al2O3, SiO2, K2O, CaO, TiO2, MnO, Fe2O3) and trace elements having various geochemical behavior (V, Cr, Ni, Zn, Rb, Sr, Zr, Ba, Ce), are used on a routine basis in clustering and discriminant analyses. Classifications are carried out by hierarchical clustering analysis on standardized data, using Euclidean distances and average linkage.
160 Yona Waksman 14. A first report appeared in 1999: Waksman, S. Y., Segal, I., Porat, N., Stern, E. J., and Yellin, J. (1999). “An Analytical Study of Ceramics Found in Crusader Acre: Levantine Productions and Imports from the Byzantine World” (Jerusalem, Geological Survey of Israel Internal Reports GSI/8/99). 15. See a selection of online resources in References. 16. Ongoing doctoral research is carried out on this subject by A. Öztürk at the University of Lyon.
References Arnold, D. E. (1985). Ceramic Theory and Cultural Process (Cambridge: Cambridge University Press). Ballet, P. and Picon, M. (1987). “Recherches préliminaires sur les origines de la céramique des Kellia (Egypte). Importations et productions égyptiennes.” Cahiers de la céramique égyptienne 1: 17–48. Bass, G. F. and van Doorninck, F. H. (1978). “An 11th Century Shipwreck at Serçe Liman, Turkey.” International Journal of Nautical Archaeology 7: 119–132. Beier, Th. and Mommsen, H. (1994). “Modified Mahalanobis Filters for Grouping Pottery by Chemical Composition.” Archaeometry 36(2): 287–306. Catling, H. W., Richards, E. E., and Blin- Stoyle, A. E. (1963). “Correlations between Composition and Provenance of Mycenaean and Minoan Pottery.” Annual of the British School at Athens 58: 94–115. Day, P. M., Kiriatzi, E., Tsolakidou, A., and Kilikoglou, V. (1999). “Group Therapy in Crete: A Comparison between Analyses by NAA and Thin Section Petrography of Early Minoan Pottery.” Journal of Archaeological Science 26: 1025–1036. Fouqué, F. (1879). Santorin et ses éruptions (Paris: Masson). François, V., Nicolaïdès, A., Vallauri, L., and Waksman, Y. (2003). “Premiers éléments pour une caractérisation des productions de Beyrouth entre domination franque et mamelouke.” In: Actes du VIIe Congrès International sur la Céramique Médiévale en Méditerranée (Athens: Caisse des Recettes Archéologiques), 325–340. Gauss, G. and Kiriatzi, E. (2011). Pottery Production and Supply at Bronze Age Kolonna, Aegina: An Integrated Archaeological and Scientific Study of a Ceramic Landscape (Vienna: Verlag der Österreichischen Akademie der Wissenschaften). Harbottle, G. (1976). “Activation Analysis in Archaeology.” Radiochemistry 3: 33–72. Hunt, A. (2012). “On the Origin of Ceramics: Moving towards a Common Understanding of ‘Provenance’.” Archaeological Review from Cambridge 27(1): 85–97. Jenkins, M. (1992). “Early Medieval Islamic Pottery: The Eleventh Century Reconsidered.” Muqarnas 9: 56–66. Jones, R. E. (1986). Greek and Cypriot Pottery. A Review of Scientific Studies. Fitch Laboratory Occasional Papers, 1 (Athens: British School at Athens). Maggetti, M. (1982). “Phase Analysis and Its Significance for Technology and Origin.” In: Archaeological Ceramics (Washington D.C.: Smithsonian Institution), 121–133. Mannoni, T. (1994). 25 Anni di archeologia globale, vol. 5: Archaeometria geoarcheologia dei manufatti (Genoa: Escum). Mason, R. B. (2004). Shine Like the Sun: Lustre-Painted and Associated Pottery from the Medieval Middle East (Costa Mesa, CA: Mazda Publishers, and Toronto: Royal Ontario Museum).
Provenance Studies 161 Peacock, D. P. S. (1982). Pottery in the Roman World: An Ethnoarchaeological Approach (London: Longman). Peacock, D. P. S. and Williams, D. F. (1986). Amphorae and the Roman Economy. An Introductory Guide (London: Longman). Perlman, I. and Asaro, F. (1969). “Pottery Analysis By Neutron Activation Analysis.” Archaeometry 11: 21–52. Picon, M. (1987). “La fixation du baryum et du strontium par les céramiques.” Revue d’Archéométrie 11: 41–48. Picon, M. (1993). “L’analyse chimique des céramiques: bilan et perspectives.” In: Archeometria della Ceramica. Problemi di Metodo, Atti 8° Simposio Internazionale della Ceramica (Bologna: Int. Centro Ceramico), 3–26. Picon, M. (1995). “Grises et grises: quelques réflexions sur les céramiques cuites en mode B.” In: Actas das 1as Jornadas de Ceramica Medieval e Pos-Medieval (Porto, Maio: Camara Municipal de Tondela), 283–292. Pradell, T., Vendrell- Saz, M., Krumbein, W.- E., and Picon, M. (1996). “Altérations de céramiques en milieu marin: les amphores de l’épave romaine de la Madrague de Giens (Var).” Revue d’Archéométrie 20: 47–56. Reynolds, P., Waksman, S. Y., Lemaître, S., Curvers, H., Roumié, M., and Nsouli, B. (2008– 2009). “An Early Imperial Roman Pottery Production Site in Beirut (BEY 015): Chemical Analyses and a Ceramic Typology.” Berytus 51–52: 71–115. Schneider, G. (2000). “Chemical and Mineralogical Studies of Late Hellenistic to Byzantine Pottery Production in the Eastern Mediterranean.” In: RCRF Acta 36 (Abingdon), 525–536. Shepard, A. O. (1963). Ceramics for the Archaeologist (Washington D.C.: Carnegie Institution). Speakman, R. J. and Glascock, M. D. (eds). Archaeometry 49(2). Stern, E. J. (2012). ‘Akko I, The 1991–1998 Excavations. The Crusader-Period Pottery. 2 vols. IAA Reports 51 (Jerusalem: Israel Antiquities Authorities). Stern, E. J. and Waksman, S. Y. (2003). “Pottery from Recent Excavations in Crusader Acre: Typological and Analytical Study.” In: Actes du VIIe Congrès International sur la Céramique Médiévale en Méditerranée (Athens: Caisse des Recettes Archéologiques), 167–180. Tite, M. S., Kilikoglou, V., and Vekinis, G. (2001). “Review Article: Strength, Toughness and Thermal Shock Resistance of Ancient Ceramics, and Their Influence on Technological Choice.” Archaeometry 43(3): 301–324. Tomber, R. and Dore, J. (1998). The National Reference Roman Fabric Reference Collection. A Handbook (London: Museum of London Archaeological Service). Waksman, S. Y. (2002). “Céramiques levantines de l’époque des Croisades: le cas des productions à pâte rouge des ateliers de Beyrouth.” Revue d’Archéométrie 26: 67–77. Waksman, S. Y. (2011). “Ceramics of the ‘Serçe Limanı type’ and Fatimid pottery production in Beirut.” Levant 43(2): 201–212. Waksman, S. Y. (2014). “Etude de provenances de céramiques.” In: Circulation des matériaux et des objets dans les sociétés anciennes (Paris: Archives Contemporaines), 195–215. Waksman, S. Y., Stern, E. J., Segal, I., Porat, N., and Yellin, J. (2008). “Some Local and Imported Ceramics from Crusader Acre Investigated by Elemental and Petrographic Analysis.” ‘Atiqot 59: 157–190. Whitbread, I. K. (1991). “Image and Data Processing in Ceramic Petrology.” In: Recent Developments in Ceramic Petrology. British Museum Occasional Papers, 81 (London: British Museum), 369–391.
Chapter 11
M ineral o g i c a l a nd Ch em ical A lt e rat i on Gerwulf Schneider Introduction Pottery is a very stable material compared to wood, metal, and glass. In consequence, sherds of archaeological ceramics have survived for up to 10,000 years while buried in the ground. This, in addition to its occurrence on nearly every archaeological site, is one of the reasons why pottery is important as a carrier of information about ancient cultures. Many ancient cultures are even named after their typical pottery type. On the other hand, archaeological sherds buried in the ground are part of the surrounding soil and will behave accordingly. The study of ancient pottery in the archaeo-ceramological laboratory is aimed at reconstructing the technology used in its manufacture and determining the date and place of its manufacture. The interpretation of analyses is based on the assumption that the potsherds have not been altered during deposition in the ground. That this possibility exists, however, was already mentioned by Sayre et al. (1957) when they used neutron activation analysis (INAA) to determine the provenance of archaeological potsherds from the Mediterranean region. The many analyses of archaeological ceramics done thereafter using NAA, optical emission spectrometry (OES), AAS (atomic absorption spectroscopy) and WD-XRF (wavelength dispersive X-ray fluorescence), and later also inductively coupled plasma emission or mass spectrometry (ICP-OES and ICP-MS), yielded a large body of evidence relating to possible post-depositional chemical and mineralogical changes in ceramic materials. Nearly everybody working on analysis of ancient ceramics has at least once been faced with the problem of alteration, and there are many published papers on this issue, which has also been discussed at various archaeometric meetings. Suggestions for further reading may be found in the articles referred to herein. Three approaches have been used to study the possible alteration of ceramic material during deposition in its burial environment (Schwedt et al., 2004). The most common approach is that of comparative studies. Products made by a single workshop within a fixed period should have the same composition given that material changes due to technical reasons can be excluded,
Mineralogical and Chemical Alteration 163 such as, for example, changing the recipe for pottery designed to serve a special purpose. The differing chemical composition of sherds buried in different environments can, thus, be interpreted in terms of chemical alteration (Freeth, 1967). A special case is presented when material from an excavated kiln can be analyzed (Buxeda i Garrigós et al., 2001). The advantage of this approach is that the effects of the depositional environment on many sherds of differing ceramic quality are considered. As another approach, profile studies can be employed in order to avoid the assumption that the examined sherds originally had the same composition. In this approach, different layers of a large sherd cut parallel to its surface are analyzed. Because the surface of the sherd is most vulnerable to environmental influences, it will show a different composition to that of the sherd’s core (e.g. Picon, 1976; Thierrin-Michael, 1992, Schwedt et al., 2004). The third approach is the experimental simulation of burial conditions to study pos sible changes (e.g. Segebade and Lutz, 1976, 1980; Franklin and Vitali, 1985; Bearat et al., 1992). However, the interpretation of the results here is limited by the question of whether short-term experiments on a few selected samples can really be compared with the post-depositional alteration of sherds varying in composition and ceramic quality and buried in very different depositional environments for thousands of years. Thus, experiments can only highlight some possible tendencies. It is very difficult to define general rules on how post-depositional alteration can be recognized and how severe its effects on a ceramic object can be. This is due to the large chemical and structural variability of ceramics and of the post-depositional environments encountered in different climates (e.g. humidity, temperature, pH, and redox conditions). Therefore, any generalization of results is limited. The original clay composition, preparation of the body, forming, and firing have a big influence on the size and number of closed and open pores within the ceramic body, and thus on the accessibility for soil solutions. Non-calcareous and calcareous ceramics behave very differently. The application of a slip or a glaze, as well as many other factors, must also be taken into consideration. Another aspect is the part played by the degree of vitrification of the sherd. A lot also depends on body composition and on temperature, atmosphere, and time of firing. Furthermore, post-depositional chemical changes are very much bound to the corrosion of glass as discussed, for example, by Freestone (2001). Products with the same composition and made at the same pottery workshop can differ significantly depending on firing in their resilience to weathering, because of the varying amount of glass in the sherds. When comparing results from different studies, another factor which must also be considered is how the samples were taken and prepared for analysis. It is common practice to remove a surface layer before drilling, or before powdering a fragment in a mill, for chemical analysis. In these instances, the issue of whether the post-depositional alteration effects on the chemical data are significant or negligible depends on the thickness of the removed surface layer. This may explain some contradictory experiences. In our laboratory we clean fragments before analysis by removing all surface layers of about 1 mm or more, if possible. Maybe therefore, among our 30,000 or so chemical analyses of archaeo logical ceramics by WD-XRF only a minority are seriously affected by post-depositional alteration effects. Non-destructive chemical analysis of the surfaces or old breaks of a sherd using a portable energy-dispersive XRF-analyzer will, however, be far more greatly affected by such alteration.
164 Gerwulf Schneider
Mineralogical Alteration Rehydration and Rehydroxylation During firing, clay loses water in several stages. Below 500°C mainly physically adsorbed water is lost (dehydration). Between about 500°C and 900°C the clay minerals more or less lose their structural water (dehydroxylation) forming new badly crystallized or amorphous phases. During burial and exposure to environmental humidity these mineral and amorphous phases can take up water again (Grim and Bradley, 1948). This results in an expansion and in an increase in weight of the ceramic. Incompletely destroyed clay minerals will be rehydrated and rehydroxylated in a humid surrounding. If the newly formed crystals are large enough, reflexes of clay minerals appear again in X-ray diffraction diagrams and may then be detected in pottery fired between 500°C and 1000°C. The most important implication of rehydration and rehydroxylation is that the abundance of clay minerals in ceramics does not exclude high firing temperatures. It must be mentioned, however, that the basal peak of the clay mineral illite may be seen in diffractograms of pottery fired to about 1000°C. In non-calcareous high-fired illitic clays, illite thus may appear together with mullite (Maggetti, 1982). An example of rehydration in low-fired non-calcareous Neolithic sherds from Switzerland was given by Maggetti (1982). The X-ray diffractograms frequently showed a large peak indicating the presence of very fine-grained clay particles of a mixed layer type that disappeared after re-firing above 300°C, which was well below the original firing temperatures. The disappearance, therefore, is an indication that these minerals were formed during burial. Rehydroxylation of clay minerals caused by environmental humidity was observed long ago (Grim and Bradley, 1948; Hill, 1953; Kingery, 1974; Hamilton and Hall, 2012). This uptake of water is very slow. It was, therefore, proposed that determining the quantity of water required for rehydroxylation be used as a means of dating fired ceramics (Zaun, 1982; Wilson et al., 2009). The problem is that even sherds made from the same clay behave differently depending on their original firing temperature and on environmental conditions. This may be partly overcome when each sample is refired at 500°C or 650°C (Bowen et al., 2011) to determine its original water content, and then the experimental gain of mass in a humid environment is detected for every individual sample (individual kinetic constant). The age of this sample can then be determined through extrapolation. There are many assumptions: the temperature of re-firing must be sufficient for a total dehydration and dehydroxylation without decomposing eventual carbonates; eventual contributions to the loss of weight by oxida tion of carbon or organic contents must be excluded; the influence of the environmental temperature and of other possibly varying burial conditions over a long time span must be corrected. It is, therefore, still an open question as to how reliable such data are and how much they depend on the assumptions made.
Formation of Gehlenite and Calcite Mineralogical changes occur in calcareous pottery fired above about 850°C. Calcium from the decay of calcite reacts with the new phases after decomposition of the clay
Mineralogical and Chemical Alteration 165 minerals to calcium silicates and calcium–aluminum silicates. In experiments, gehlenite, diopside, wollastonite, and anorthite are then observed as new phases by X-ray diffraction. The gehlenite problem, however, was that this mineral was not found in the Roman pottery which was supposed to have been made from the clay used in the experiments and fired at similar temperatures. This was explained by the breakdown of gehlenite as a metastable mineral to calcite during burial in humid climates (Maggetti, 1981). The decomposition of gehlenite to calcite was confirmed in experiments by Heimann and Maggetti (1981). In thin sections of medium-fired calcareous pottery (between about 850°C and 950°C) where all primary calcite is decomposed secondary calcite derived from gehlenite appears finely distributed within the matrix (Plate 2a). It explains the presence of calcite in calcareous pottery fired above 850°C. At lower firing temperatures, between 600°C and 800°C, secondary calcite is formed by recarbonatization of calcium oxides and hydroxides after the decomposition of primary calcite. In thin sections it can mostly be distinguished from primary calcite which survives in very low-fired pottery (below about 700°C depending on grain sizes and time). In high-fired calcareous pottery (above c.1000°C) in thin sections only precipitated calcite appears in open pores of a vitrified matrix. X-ray diffraction will detect calcite in all these sherds but only primary calcite proves a low firing temperature below 700°C or 750°C, depending on firing atmosphere and time.
Formation of Zeolites In experiments, zeolites are formed together with rehydroxylation (Hill, 1953) or with the breakdown of gehlenite (Heimann and Maggetti, 1981). The formation of small amounts of zeolites is common in altered sherds of high-fired calcareous pottery (e.g. Picon, 1991). This process has also been observed in experiments (Heimann and Maggetti, 1981). Analcime (or wairakite) derives from the decomposition of gehlenite and from the alteration of the glassy phase. This has been shown by X-ray diffraction of samples from sites in Switzerland (Maggetti, 1981), in Spain (Buxeda i Garrigós, 1999; Schwedt et al., 2006), and on Crete (Buxeda i Garrigós et al., 2001). The formation of analcime is connected with an enrichment of sodium derived from the surrounding soil and thus changing the chemical composition of the sherd significantly (e.g. Buxeda i Garrigós et al., 2002).
Alteration in Seawater Bearat et al. (1992) found some changes in experiments with low-f ired pottery where calcium was exchanged through magnesium in carbonate minerals. In lagoon-like conditions special alteration effects may appear, such as the formation of tiny crystals of pyrite at the expense of hematite, which was noted by Secco et al. (2011). The subsequent crystallization of jarosite and gypsum after pyrite was interpreted in terms of changing environmental burial conditions differing between the two sites examined in the aforementioned paper.
166 Gerwulf Schneider
Chemical Alteration Some of the chemical post-depositional alterations are clearly connected to mineralogical changes which could be detected by X-ray diffraction, but absorption as well as leaching primarily concern the badly crystallized phases after the decay of the clay minerals, glass, and other amorphous phases.
Absorption and Contamination Quantitative chemical analysis of the body of a sherd is mainly aimed at determining its provenance by comparison with chemical analysis of sherds of known origin. Therefore, possible post-depositional alterations must be taken into account. Rehydration and rehydroxylation only change the water content. However, of the elements determined by WD- XRF or NAA, it seems that significant absorption effects exist for a series of chemical elements, such as Ca, Sr, P, Ba, Fe, Mn, Na (and Cs), including possible post-depositional contamination by Cu, Zn, and Pb. Many of these effects depend on the high cation exchange capacity of ceramics, which may be of the same order as it is in clay (Hedges and McLellan, 1976). To remove possible contamination by soluble salts, the fragments to be prepared for analysis should be washed with distilled water in an ultrasonic device. Some of the mineralogical alterations change the bulk chemical composition of a sherd. This is the case, for example, when calcite is precipitated within the sherd or if zeolites are formed in high-fired calcareous pottery. Other post-depositional changes, such as, for example, recarbonatization, do not change the elemental chemical composition which is used to determine the provenance of sherds. The post-depositional alteration effects differ for non-calcareous and calcareous pottery and depend very much on firing temperatures. Therefore, the effects of absorption, as well as those of leaching, in overfired calcareous pottery may be very different from those in the lower-fired parts of the same sherd (Picon, 1991).
Water Rehydration and rehydroxylation result in a gain of weight because of the absorbed water. This may be easily determined as loss on ignition, for example, at 950°C. However, this is only a measure for the water content if there is no organic matter, no sulfur, and no carbonates in the ceramic matrix. In chemical analyses, the water content acts as a dilution agent and can be treated as such.1 For varying water contents, chemical data can easily be accounted for by comparing analyses on a water-free (ignited) basis, as is done in many laboratories.
Calcium The formation of calcium silicates and calcium–aluminum silicates, such as gehlenite, and recarbonatization does not change the initial calcium content. An example of homogeneously distributed secondary calcite from recarbonatization after gehlenite in
Mineralogical and Chemical Alteration 167 a calcareous sherd of Arretine sigillata is shown in Plate 2a. The calcite within another calcareous sherd of Eastern sigillata A (Plate 2b) very probably also originates from recarbonatization. During burial it was leached at the surface where environmental acid solutions had access as for example on the upper right hand site where the protective gloss is missing. The secondary calcite does not change the original calcium content of the samples (if not leached). On the other hand, calcite or gypsum crystallized within cracks and open pores of a sherd derived from precipitation by invading calcium bicarbonate-containing solutions will change the composition significantly. In Plate 2c of a high-fired sherd of non-calcareous North-Mesopotamian metallic ware the precipitated calcite appears as a fine scale in the matrix and in open pores. The precipitation here caused an increase to about 6 wt.% CaO in the core of this non-calcareous sherd which previously had less than 1 wt.% CaO detected from comparison to sherds of the same compositional group without secondary calcite. A special case of post-depositional alterations is illustrated by a sherd of Tripolitanian sigillata shown in Plate 2d. Here a thin scaly layer between body and gloss makes the sherd macroscopically appear to have a white engobe below the red gloss. For chemical analysis this layer, however, would be removed. All four examples of calcite are connected with medium-to high-fired sherds (c.800–1000°C). The presence of calcite in diffractograms therefore cannot be attributed to low firing temperatures. Gypsum in ceramics may be recognized by analyzing sulfur. In thin sections, the possible original gypsum content of the clay can be distinguished from secondary infiltrated gypsum by the “swallow-tailed” pores left by the original gypsum crystals. After firing, these characteristic pores are filled with very fine-grained secondary gypsum. Gypsum derived from precipitation will fill cracks and open pores with irregular shapes. In studies of Mesopotamian pottery these effects play an important role and certainly must also be taken into account in studies of pottery from other areas with similar environments. Besides the increase of calcium by infiltrated calcite or gypsum, the values of strontium and its relation to calcium may also be changed dramatically. Analyzing profiles of four sherds using an electron microprobe, Freestone et al. (1985) found an enrichment of calcium in the outer layers which was not connected to precipitation of calcite. This could be excluded because only the fine matrix was analyzed and thin sections were studied. Thus, the U-shaped profiles as found for calcium, phosphorus, and iron are clear indications of absorption within the matrix.
Phosphorus Another very common effect of contamination is the absorption of phosphorus. In archaeological sherds, P2O5 concentration is nearly always higher than in the clay from which the sherds have been made. Phosphorus contents above about 0.5 wt.% P2O5 generally indicate altered sherds, even if macroscopically they look intact. The fixation of phosphorus certainly represents the largest of the alteration effects. Phosphorus originates from the soil in which the sherds are buried. It can be particularly high in pottery in humid climates and in acid soils containing bones and organic waste, as is the case with many archaeological finds. In some cases, for example in weathered sherds of Roman Terra Sigillata, phosphorus can rise up to more than 10 wt.% P2O5 in the bulk analysis
168 Gerwulf Schneider of the body, and it would be even higher in the surface layers normally removed before analysis. In rare instances some of the phosphorus may also derive from former organic contents in the vessels (Duma, 1972; Dunnell and Hunt, 1990; Bearat and Dufournier, 1994). Phosphorus from temper containing crushed bones can easily be detected in thin sections, and it normally yields much higher contents of P2O5 which then are correlated to calcium contents. As phosphorus is not determined by NAA, this effect is only observed when other techniques, such as WD- XRF, are used for chemical analysis (Schneider, 1978; Rottländer, 1981–1983; Lemoine and Picon, 1982; Freestone et al., 1985; Picon, 1987; Walter and Besnus, 1989). The absorbed phosphorus is homogeneously distributed within the sherds and is not correlated to calcium or iron. Discrete grains or crystalline phosphate phases are not detected. Phosphorus, thus, must be bound to amorphous and organic phases (Collomb and Maggetti, 1996). Calcareous pottery seems to be more involved and the alteration is certainly connected to the ceramic quality of the sherds, particularly to their porosity and to their level of vitrification. The phosphorus absorption as a rule is connected with other chemical alterations and thus can be used as a marker for alteration. Elevated phosphorus in most instances is correlated with higher ignition losses (rehydroxylation), with higher barium contents, and, sometimes, with leaching of other elements.
Example: Roman Pottery at Different Sites Finds of Roman sigillata produced at three production centers: Arezzo, Lyon, and La Graufesenque, found in Velsen and Nijmegen in the Netherlands, were examined in terms of alteration. The samples represent medium-to high-fired calcareous pottery (c. 800–1,000°C), the three production centers represent sigillata of different quality and the different find spots represent different burial conditions. For each site the element concentrations for each of the three centers are averaged in spite of their large variation and are compared to the values of the respective reference groups. Arezzo is represented by forty-seven finds in Arezzo; for Lyon and for La Graufesenque data from Schneider (1978) have been used. In Table 11.1 the values are shown for loss on ignition (l.o.i.) and for the chemical elements associated with alteration effects. Compared to the reference groups the absorption of water (l.o.i.), phosphorus, strontium, and barium of the finds in Velsen may be regarded as negligible, but in the samples found in Nijmegen the increases are obvious. The effects are least pronounced in sigillata from La Graufesenque, known for its high quality. Analyses of pottery from the Roman legionary camp of Dangstetten (Schneider and Daszkiewicz, 2006) offer a similar example. Nearly all samples from this site showed elevated phosphorus contents. The effects on products from Lyon and from Arezzo again differ significantly. In sherds of sigillata from Lyon the average phosphorus content was 3.9 wt.% P2O5 (n = 19) and all samples had more than 1.3 wt.% P2O5. In sherds of sigillata from Arezzo the average was 1.3 wt.% P2O5 (n = 148). Here 121 of 148 samples (82%) had more than 0.5 wt.% P2O5. This may be compared to analyses of sherds of Arretine sigillata found in Arezzo where only six of 96 samples (6%) had more than 0.5 wt.% P2O5. The elevated phosphorus contents noted at Dangstetten, Nijmegen, and Velsen, as well as in Arezzo,
Mineralogical and Chemical Alteration 169 Table 11.1 Chemical alteration of Roman sigillata from Arezzo, Lyon, and La Graufesenque found at Velsen and Nijmegen in the Netherlands. The values for absorption are as analyzed. For leaching, the analyses results have been normalized without phosphorus to a constant sum of 100% Production site
find spot
l.o.i (wt%)
P2O5 (wt%)
Sr (ppm)
1.38
0.28
274
Ba (ppm)
absorption:
Arezzo (n = 47)
459
Arezzo
Velsen (n = 20)
1.07
0.19
282
367
Nijmegen (n = 20)
4.61
3.85
350
1069
Reference group
–
0.27
260
425
Lyon
Velsen (n = 7)
4.21
0.48
265
458
Nijmegen (n = 11)
8.15
7.20
435
1406
Reference group
–
0.30
354
386
La Graufesenque
Velsen (n = 19)
1.20
0.33
363
399
Nijmegen (n = 54)
3.55
3.32
440
729
CaO (wt%)
Na2O (wt%)
K2O (wt%)
leaching Production site
find spot
Rb (ppm)
Arezzo (n = 47)
9.61
0.77
2.63
134
Arezzo
Velsen (n = 20)
8.75
1.00
2.82
140
Nijmegen (n = 20)
8.35
0.82
2.69
134
Reference group
17.60
0.56
2.16
114
Lyon
Velsen (n = 7)
6.73
0.57
2.25
124
Nijmegen (n = 11)
6.86
0.50
2.19
131
Reference group
10.40
0.35
3.75
173
La Graufesenque
Velsen (n = 19)
11.10
0.37
3.67
165
Nijmegen (n = 54)
10.17
0.25
3.73
171
were mostly detected in samples with increased losses on ignition indicating rehydroxylation of weathered sherds.2
Barium and Strontium Together with elevated phosphorus contents, barium is increased in most cases as, for example, in Velsen and Nijmegen (Table 11.1). But this is not always the case and, depending on burial conditions, on many occasions barium is significantly raised without remarkable changes in phosphorus content. The uptake of barium has been described in various papers, the most recent by Golitko et al. (2012). Sometimes strontium shows a similar behavior to barium, but the maximal increase is lower. The explanation of this phenomenon, that barium originates from barite veins and strontium originates from a sedimentary environment
170 Gerwulf Schneider with high strontium contents, proposed by Picon (1985, 1987, 1991), seems not to cover all cases. The effect in calcareous pottery is generally greater than in non-calcareous pottery, but the degree of alteration seems much more connected to the ceramic quality and to the burial conditions. Figure 11.1 gives an example that shows the correlation of barium with phosphorus in non- calcareous Roman Terra Nigra (orange squares) and Germanic pottery (green triangles) compared to the local clay (pale re circles) from which the Roman and Germanic pottery in Ladenburg was made (Schneider, 2002). The higher-fired and finer Roman pottery from Ladenburg shows a good correlation of barium with phosphorus. The lower-fired Germanic pottery from Ladenburg has higher barium and phosphorus values. Other Roman non- calcareous pottery produced and found at three find spots in Speyer on the other side of the Rhine (rhombuses) shows a similar correlation but with another ratio of barium to phosphorus. The different ratio indicates that the post-depositional conditions were different,
2500
2000
Barium [ppm Ba]
Ladenburg
1500
1000
500 Speyer
0 0.00
0.50
1.00
1.50
2.00
2.50
Phosphorus [wt.% P2O5]
Figure 11.1 Correlation of barium and phosphorus in Roman and Germanic pottery from two sites in Germany. The analyses of clay supposed to have been used for the pottery in Ladenburg (pale circles) mark the starting point of the original composition of Roman Terra Nigra (squares) and Germanic pottery (triangles) in Ladenburg. In Speyer one of the three find spots is less influenced by absorption (empty rhombuses) than the other two find spots (pale and dark rhombuses). Note the different starting composition and correlation in Ladenburg and in Speyer.
Mineralogical and Chemical Alteration 171 and at one of the three find spots in Speyer (empty rhombuses) post-depositional alteration seems less pronounced than at the two others.
Sodium The possible formation of zeolites in high-fired calcareous pottery causes a possible gain in sodium content. In an example of calcareous pottery from Spain this could be up to three times the original value (Schwedt et al., 2006). In this study, sodium was significantly enriched in the surface layers of the higher-fired and analcime-containing sherds.3 In another example, given by Buxeda i Garrigós et al. (2002), the average sodium contents rose from 0.84 wt.% Na2O in the unaltered group to 1.75 wt.% Na2O in the analcime-containing group of the same Mycenaean pottery. In another study by Buxeda i Garrigós et al. (2001), sodium contents are also clearly correlated with the peak intensity of analcime in the altered samples. The samples presenting the highest analcime values also show the largest leaching effects of potassium and, to a lesser extent, of rubidium. These effects in high-fired calcareous pottery limit the use of sodium (potassium, rubidium) as a significant element for provenance deter mination, but from the author’s own experiences this concerns only a very minor proportion of archaeological pottery. In any case, the effect may be overcome when possible analcime contents are checked by X-ray diffraction, when sodium contents of sherds of different firing temperatures within an otherwise homogeneous group are compared, and when other indications of alteration may be observed. In overfired calcareous pottery, besides absorption of sodium, magnesium and strontium could also be enriched (Picon, 1991).
Caesium The absorption of caesium in low-fired (below 750°C) pottery was reported by Buxeda i Garrigós et al. (2001) when they studied the alteration effects of alkaline elements in various ceramics using NAA. This is likely to be related to the absorption capacity of the illites still present in the low-fired pottery. However, in another case, in high-fired calcareous Mycenaean sherds, caesium showed the opposite effect. It was clearly depleted in the surface layers (Schwedt et al., 2006).
Manganese and Iron Under certain circumstances manganese and iron may infiltrate buried sherds, leading to significantly increased concentrations of these elements. For manganese this has been described in several studies (Freeth, 1967; Dufournier, 1979; Walter and Besnus, 1989). The content of manganese is related to the porosity and to the loss on ignition of the sherds, indicating the alteration effect (Dufournier, 1979). It is enriched, for example, in the surface layers of amphorae studied by Picon (1987). In a single analysis of a sherd it cannot be determined whether high manganese contents represent the original values. Outliers in a series, however, may be due to post-depositional alteration. In certain burial conditions, iron could also be fixed at the surfaces and within cracks in the sherds, thus causing a significant enrichment of iron in the bulk chemical analysis. Mostly this is not visible macroscopically on the original sherds but it is observable in thin sections (e.g. Daszkiewicz and Prinke, 1999) or it can be made obvious by re-firing to about 1,100°C, when enriched iron contents appear as
172 Gerwulf Schneider red zones on the break or surface of the sherd, around open pores, or in cracks, in otherwise iron-poor pale sherds. Freestone et al. (1985) showed a U-shaped profile of iron using microprobe analysis in a study of sherds with elevated phosphorus contents at the surfaces.
Copper, Zinc, Lead, Silver Archaeological sherds are often buried in an environment featuring metals. In such cases corroded bronze may cause higher trace contents of copper, zinc, and lead in the surrounding soil as well as in the buried sherds. Such elevated trace contents were detected in pottery found in the mining area of Laurion in Greece (De Paepe, 1979). As the outer layers were clearly enriched it could only be explained by accumulation of these metals from aqueous solutions after burial. Similar effects have been shown in the author’s analysis of material from an excavated bronze casting workshop at Olympia. Here, nearly all pottery samples, as well as the surrounding marly clay in which the sherds were deposited, evinced elevated trace contents of copper. Accidental burial of a sherd in the vicinity of metal artifacts may thus also explain frequent outliers of one, two, or three of the elements copper, zinc, or lead in a series of analyses. Such outliers are mostly connected to altered sherds recognized by their high phosphorus content and/or high ignition losses. Arretine sigillata found in Dangstetten may serve as an example. Within the series of 148 analyzed samples at least 23 samples had anomalously high zinc contents. These high zinc contents were always connected to elevated traces of copper. Mostly, but not always, these sherds had higher phosphorus contents showing this element as a marker for alteration effects. Six outliers caused by lead contamination are not correlated with zinc or phosphorus. Adan-Bayewitz et al. (2006) reported an example of anomalously high traces of silver originating from in situ contamination after deposition. After ruling out other possibilities this seemed to be the most probable explanation of higher silver contents of sherds excavated in Jerusalem compared to other sites in Israel.
Special Cases of Deposition Sherds buried in seawater show elevated magnesium which, according to Lemoine et al. (1981), derives from a reaction of seawater with the glassy phase. In experiments, Bearat et al. (1992) found a significant decrease in calcium and an increase in magnesium, but no changes in sodium. The latter, however, does not seem to be excluded in other cases of burial influenced by seawater (e.g. elevated sodium contents at Velsen, see example in Table 11.1). Sherds buried in ash, for example, in a kiln or hearth, present another case. Such sherds could take up significant amounts of potassium, which is then enriched in the surface layers (Dufournier, 1979).
Leaching Calcium Calcite in sherds will be more or less slowly dissolved in acid soils. This is a well-known phenomenon and has been demonstrated in various studies of chemical alteration (e.g. Freeth, 1967). Analyses of profiles clearly point to reduced calcium content in the outer layers
Mineralogical and Chemical Alteration 173 (Thierrin-Michael, 1992; Schwedt et al., 2004; Schwedt et al., 2006). Extreme effects are presented by examples of coarse-tempered sherds where the coarse temper of crushed calcite was totally dissolved and could only be recognized by the large rhombohedral pores left behind. An analogous effect can be observed in shell-tempered pottery, where in thin sections only the characteristic pores left by dissolved crushed shell can be seen. Finely distributed calcite within the matrix derived from recarbonatization is dissolved starting from the surface or from cracks where acid soil solutions have easy access. Subsequently, a variably thick layer depleted in calcite appears below the surface, as can be seen in many thin sections of medium-to high-fired calcareous pottery. An example of Eastern sigillata A is shown in Plate 2c. These leaching effects, of course, will change the chemical composition.4 Leaching of calcium is connected mainly, though not exclusively, to the dissolution of carbonates. Therefore, firing temperatures, the amount of calcium in the glassy phase, and calcite contents will have large influences. Under certain burial conditions, deposition of calcite may predominate leaching, resulting in profiles showing enriched calcium in the outer layers (Freestone et al., 1985). Examples: In cross-sections of calcareous sherds re-fired at temperatures between 1,100°C and 1,200°C (MGR-analysis)5 the effect of leaching can be seen very clearly. The outer layer of a calcareous sherd from Tell Arbid (Iraq) exhibits the typical red color of non-calcareous ceramic after re-firing at 1200°C (Figure 11.2b). The effect of leaching, however, cannot be seen in the original sherd before re-firing (Figure 11.2a). Chemical analyses of the yellow and of the red parts in this example showed 19.2 wt.% and 7.6 wt.% CaO respectively. The routine analysis of a larger sample, as usual after removing surface layers, revealed a calcium content of 16.2 wt.% CaO, which thus is still lower than in the yellow part. 290 ppm of strontium in the yellow part was reduced by leaching to 208 ppm in the red part (routine analysis of the whole sherd after removing surface layers amounted to 262 ppm). It is important to note that not only the contents of calcium and strontium but also their ratio changed significantly. Normalization of the analyses to the same calcium contents showed that the other elements in this example have
(a)
(b)
(c)
1 cm
Figure 11.2 Leaching of calcium in two samples of calcareous pottery: (a) sample from Tell Arbil (Iraq) before re-firing; (b) other fragment of the same sample after re-firing at 1200°C; (c) fragment of a sample of Roman pottery from Aguntum (Austria) after re-firing to 1100°C showing leaching at the surfaces and at the old break (left) exposed to environmental influences. Photographs with macro-lens by M. Baranowski.
174 Gerwulf Schneider not been affected by leaching. This example is not an isolated case and most medium-to high- fired calcareous pottery studied in thin sections shows similar effects. Some samples where the leaching includes an old break of the sherd prove this effect to be caused by post-depositional alteration. It can be made visible by re-firing, as illustrated with a fragment of a Roman pottery from Aguntum (Austria) re-fired at 1100°C (Figure 11.2c). Leaching was also observed in the examples discussed with the absorption of phosphorous (Table 11.1, lower part).
These samples represent calcareous pottery in a humid climate and acid soils. All analyses were carried out after removing the outer layers of the fragments before powdering and, therefore, show reduced alteration effects. To account for the absorbed phosphorus contents the analyses have been normalized to a constant sum of 100% without P2O5. Comparing calcium contents at Velsen and Nijmegen with the reference values shows significant leaching of calcium at both find spots for products from Lyon and Arezzo. This, however, cannot be verified in this instance for other elements e.g. potassium and rubidium (which in other cases was found to be significant). At Velsen, elevated sodium contents must be explained by the influence of seawater at this coastal site, even though the small fragments for analysis were washed thoroughly in distilled water in an ultrasonic device. A special effect concerns strontium, which is enriched together with barium (Table 11.1) in the same instance when calcium is reduced or unchanged. This leads to ratios of strontium to calcium depending more on post-depositional alteration effects than on different raw materials. Considering only ten sherds (also used in Table 11.1) with phosphorus contents between 9 and 11 wt.% P2O5 shows the leaching effects more obvious and how these result in significantly higher values of the stable elements titanium, aluminum, and iron (chromium, nickel, zirconium). Silica seems not to belong to these stable elements and may be leached to some degree, as already hypothesized by Rottländer (1981–3).
Alkaline Elements The possibility of leaching of the alkaline elements in pottery during burial is well known. For potassium, it was already reported by Picon (1976). The leaching effect on potassium and barium is very marked in overfired calcareous pottery (Picon, 1991). Because of the negli gible effect for rubidium, the important ratio of rubidium to potassium changes very much. Experiments of leaching calcareous sigillata in neutral and acid solutions by Segebade and Lutz (1976, 1980) showed leaching for manganese, potassium, calcium, caesium, arsenic, rubidium, but not for sodium. Based on a large series of analyses by NAA, the leaching of all alkaline elements was discussed in several papers, including studies of profiles (Buxeda i Garrigós et al., 2002; Schwedt, 2004; Schwedt et al., 2004, 2006). The leaching effect depends mainly on the glassy phase and, therefore, it is significant especially in high-fired calcareous pottery. But this does not exclude the effects in low-fired and in non-calcareous pottery. The largest effect according to these studies seems to be for caesium followed by rubidium, potassium, and sodium. But the order of elements does not always have to be the same (e.g. Buxeda i Garrigós et al., 2001). It is important to note that the loss of potassium can disrupt luminescence dating, which is based on a constant dose rate from the radioactive elements in the sherds (Zacharias et al., 2005).
Mineralogical and Chemical Alteration 175 As a consequence it was proposed not to use, for example, sodium for provenance studies, but this cannot be confirmed from the author’s experiences with some 30,000 analyses. Certainly potassium and rubidium belong to the most significant elements in provenance studies.
Conclusions The major aim of most studies concerning post-depositional alteration has been to see how far alteration effects have an influence on the interpretation of pottery analyses in terms of provenance and technology. Most papers concern only selected aspects of absorption or leaching. The reported experiences also depend very much on the methods used for the analysis when the range of the considered chemical elements depends on the detection limits of the analytical technique. So phosphorus, and in some cases also calcium, is not taken into consideration when NAA is used. On the other hand, because of the typically low concentrations of caesium, its varying effects cannot be detected by WD-XRF. However, the combination of NAA and WD-XRF covers most chemical elements in archaeological ceramics down to less than 1 ppm (as would also ICP-MS), and it seems that the elements not discussed here are not affected by post-depositional alteration. The combination of chemical analysis with thin-section studies and/or X-ray diffraction analysis (XRD) has proven to be very helpful for the discovery and interpretation of alteration effects. Re-firing experiments6 make some absorption and leaching effects clearly visible. The physical integrity of a sherd, on the other hand, must not be a criterion for negligible post-depositional alteration. Another factor which limits the potential to generalize the results of multiple studies is that the effects depend on the composition and ceramic quality of the ceramic object and on its particular depositional environment. Chemical and mineralogical composition, prepara tion, and firing of the clay determine the open porosity and the amount of glass, resulting in different resistance against the attack of solutions during burial. In this respect, calcareous pottery and non-calcareous pottery behave very differently. For the alteration effects, the tendencies are similar but also sometimes contradictory. The magnitude of the effects depends very much on the specific circumstances of each case. Post-depositional alteration mostly concerns primarily the surface layers, being more or less thick, as many studies have shown. The mineralogical composition of the surfaces, however, may also differ from the composition of the core caused by other effects, such as higher temperature (as in the core) or efflorescence of salt during the drying phase of the ceramic object. These effects must also be considered. In consequence, samples used for the analysis of the mineralogical or chemical composition of a ceramic object should be taken from the core after having at least removed the surface layers of 1 mm or more all around. Even this will not be sufficient to eliminate all alteration effects which are observed in numerous analyses of such samples. Rehydroxylation, phosphorus contents, and many other effects concern the core of the sample as well. Here, a word should be said concerning non-destructive analy sis using portable energy-dispersive X-ray fluorescence (p-ED-XRF). In pottery, even the X-ray energy of the heavy elements will yield information only about a surface layer of less than approximately 0.5 mm. Therefore serious alteration and contamination effects must be
176 Gerwulf Schneider taken into account for the interpretation of the data. This may be avoided by measuring at fresh breaks, but then the analysis is not non-destructive. The largest and the most common of the chemical alteration effects is certainly the absorption of phosphorus. This element is therefore irrelevant for determining provenances, but it may indicate possible post-depositional alterations of other elements. Sulfur and chlorine are mostly lost during firing and are less significant for provenances, but they could also be indicators of secondary contamination. For the rest of the elements which can be determined with sufficient precision and accuracy, the effects are normally limited and tendencies are known from many analyses. Chemical alteration has to be taken into account at least for calcium, strontium, barium, and the alkaline elements, in some cases also for manganese, iron, zinc, copper, and lead. Experience, however, shows that the alkali and the alkaline earths, in spite of their unstable behavior, may well be significant in distinguishing reference groups and are certainly among the elements which should be analyzed. In any event, the idea that a general list could be made of diagnostic elements for determining provenance, which would be valid for all cases, should be rejected. Mineralogical post- depositional alteration does not generally influence provenance determinations, but must be taken into account when original firing conditions are reconstructed. This mainly concerns the abundance of clay minerals and calcite. Their abundance in a sample does not exclude a high firing temperature because of the possibility of rehydroxylation and recarbonatization. Calcite or clay minerals detected by X-ray diffraction (XRD) do not provide evidence of a low firing temperature. This can only be shown by the abundance of undecomposed primary calcite; for example, by thin-section studies. Last but not least, post-depositional alteration may help us to gain information about burial conditions, and may be a possible perspective for dating.
Notes 1. The full analysis of pottery samples, calculated as weight per cent of oxides, amounts to 100%. The non-determined major components, such as water and phosphorus (and sulfur), could be regarded as dilution effects and treated mathematically by calculating a best relative fit (Olin and Sayre, 1979; Beier and Mommsen, 1994). Full chemical data for major elements, including phosphorus, may also be normalized for comparison to a constant sum of 100% without loss on ignition and with or without phosphorus (and sulfur). In both methods the measured elemental concentrations are multiplied with a calculated constant factor for each sample. 2. The very variable losses on ignition at 950°C here are not connected to the age of the pottery, which is more or less similar for all sherds at all three sites. Besides, the effect of the carbonates and organic contents on losses on ignition must also be taken into account. It would be worth taking such samples as test objects for rehydroxylation dating. 3. There is also another effect which may cause enrichment of sodium in the surface layers and which in some cases must be taken into account. This is salt in the makeup water of the pottery which during drying is effloresced at the surface due to capillary action. This may cause a white surface layer to appear on high-fired calcareous pottery, which is sometimes erroneously interpreted as a white slip. This, however, is not an alteration effect from burial.
Mineralogical and Chemical Alteration 177 4. Lack of calcite in the surface layer of medium-to high-f ired pottery may not only be due to leaching. During firing, the surfaces could have been exposed for a short time to a higher temperature, or the surface layer could have had a higher salt content due to efflorescence during the drying stage. Both effects will result in different conditions for the formation of gehlenite in the surface layer. Thus, the amount of secondary calcite here could be less than that in the core. In thin sections this may present the same picture as leaching. The effect of a higher temperature on the surface can be observed in many thin sections of calcareous and of non-calcareous pottery and can be proved by determining the equivalent firing temperatures of the surface and of the core separately. 5. See Chapter 27 by M. Daszkiewicz and L. Maritan in this volume. 6. See Chapter 27 by M. Daszkiewicz and L. Maritan in this volume.
References Adan-Bayewitz, D., Asaro, F., and Giauque, R. D. (2006). “The Discovery of Anomalously High Silver Abundances in Pottery from Early Roman Excavation Contexts in Jerusalem.” Archaeometry 48: 377–398. Bearat, H. and Dufournier, D. (1994). “Quelques experiences sur la fixation du phosphore par les céramiques.” Revue d’Archéométrie 18: 65–73. Bearat, H., Dufournier, D., and Nouet, Y. (1992). “Alterations of Ceramics Due to Contact with Seawater.” Archaeologia Polona 30: 151–162. Beier, T., and Mommsen, H. (1994). “Modified Mahalanobis Filters for Grouping Pottery by Chemical Composition.” Archaeometry 36: 287–306. Bowen, P. K., Ranck H. J., Scarlett, T. J., and Drelich, J. W. (2011). “Rehydration/Rehydroxylation Kinetics of Reheated XIX-Century Davenport (Utah) Ceramic.” Journal of the American Ceramic Society 94(8): 2585–2591. Buxeda i Garrigós, J. (1999). “Alteration and Contamination of Archaeological Ceramics: The Perturbation Problem.” Journal of Archaeological Sciences 26: 295–313. Buxeda i Garrigós, J., Kilikoglou, V., and Day, P. M. (2001). “Chemical and Mineralogical Alteration of Ceramics from a Late Bronze Age Kiln at Kommos, Crete: The Effect on the Formation of a Reference Group.” Archaeometry 43: 349–371. Buxeda i Garrigós, J., Mommsen, H., and Tsolakidou, A. (2002). “Alteration of Na, K and Rb Concentrations in Mycenaean Pottery and Proposed Explanation Using X-Ray Diffraction.” Archaeometry 44: 187–198. Collomb, P. and Maggetti, M. (1996). “Dissolution des phosphates présents dans les céramiques contaminées.” Revue d’Archéométrie 20: 69–75. Daszkiewicz, M. and Prinke, D. (1999). “Archäokeramologische Untersuchungen zur Trichterbecherkultur in Kujawien (Zentralpolen).” Ethnographisch-Archäologische Zeitschrift 40: 299–336. De Paepe, P. (1979). “Chemical Characteristics of Archaic and Classical Coarse Wares from Thorikos, SE Attica (Greece).” Miscellanea Graeca 2: 89–112. Dufournier, D. (1979). “Deux exemples de contamination des céramiques anciennes par leur milieu de conservation.” Figlina 4: 69–83. Duma, G. (1972). “Phosphate Content of Ancient Pots as Indication of Use.” Current Anthropology13: 127–130.
178 Gerwulf Schneider Dunnell, R. C. and Hunt, T. L. (1990). “Elemental Composition and Inference of Ceramic Vessel Function.” Current Anthropology 31: 330–336. Franklin, U. M. and Vitali, V. (1985). “The Environmental Stability of Ancient Ceramics.” Archaeometry 27: 3–5. Freestone, I. C. (2001). “Post- Depositional Changes in Archaeological Ceramics and Glasses.” In: Brothwell, D. R. and Pollard, A. M. (eds), Handbook of Archaeological Sciences (Chichester: John Wiley & Sons), 615–625. Freestone, I. C., Meeks, N. D., and Middleton, A. P. (1985). “Retention of Phosphate in Buried Ceramics: An Electron Microbeam Approach.” Archaeometry 27: 161–177. Freeth, S. J. (1967). “A Chemical Study of Some Bronze Age Pottery and Sherds.” Archaeometry 10: 104–119. Golitko, M., Dudgeon, J. V., Neff, H., and Terrell, J. E. (2012). “Identification of Post- Depositional Chemical Alteration of Ceramics from the North Coast of Papua New Guinea (Sanduan Province) by Time-of-Flight-Laser Ablation-Inductively Coupled Plasma-Mass spectrometry (TOF-LA-ICP-MS).” Archaeometry 54: 80–100. Grim, R. E. and Bradley, W. F. (1948). “Rehydration and Dehydroxilation of the Clay Minerals.” American Mineralogist 33: 50–59. Hamilton, A. and Hall, C. (2012). “A Review of Rehydroxylation in Fired-Clay Ceramics.” Journal of the American Ceramic Society 95(9): 2673–2678. Hedges, R. E. M., and McLellan, M. (1976). “On the Cation Exchange Capacity of Fired Clays and Its Effect on the Chemical and Radiometric Analysis of Pottery.” Archaeometry 18: 203–207. Heimann, R. B. and Maggetti, M. (1981). “Experiments on Simulated Burial of Calcareous Terra Sigillata (Mineralogical Change). Preliminary Results.” In: Hughes, H. J. (ed), Scientific Studies in Ancient Ceramics, British Museum Occasional Paper 19 (London), 163–177. Hill, R. D. (1953). “The Rehydration of Fired Clay and Associated Minerals.” British Ceramic Society 52: 589–613. Kingery, W. D. (1974). “A Note on the Differential Thermal Analysis of Archaeological Ceramics.” Archaeometry 16: 109–112. Lemoine, C., Meille, E., Poupet, P., and Barrandon, J. N. (1981). “Étude de quelques altérations de composition chimique de céramiques en milieu marin et terrestre.” Revue d’Archéometrie (Suppl. 1981): 349–360. Lemoine, C. and Picon, M. (1982). “La fixation du phosphore par les céramiques lors de leur enfuissement et ses incidences analytiques.” Revue d’Archéometrie 6: 101–112. Maggetti, M. (1981). “Composition of Roman pottery from Lousonna (Switzerland).” In: Hughes, H. J. (ed), Scientific Studies in Ancient Ceramics, British Museum Occasional Paper, London, 33–49. Maggetti, M. (1982). “Phase Analysis and Its Significance for Technology and Origin.” In: Olin, J. S. and Franklin, J. D. (eds), Archaeological ceramics (Washington, D.C.: Smithsonian Institution,), 121–134. Olin, J. S. and Sayre, E. V. (1979). “Environmental and Technological Causes of Variations in the Absolute Concentrations of Elements in Ceramics.” In: Proceedings of the 18th International Symposium on Archaeometry (Bonn), Archaeophysika 10: 607. Picon, M. (1976). “Remarques préliminaires sur deux types d’altération de la composition chimique des céramiques au cours du temps.” Figlina 1: 159–166. Picon, M. (1985). “Un exemplaire de pollution aux dimensions kilométriques: la fixation du barium par les céramiques.” Revue d’Archéometrie 9: 27–29.
Mineralogical and Chemical Alteration 179 Picon, M. (1987). “La fixation du phosphore par les céramiques lors de leur enfuissement et ses incidences analytiques.” Revue d’Archéometrie 11: 41–47. Picon, M. (1991). “ Quelques observations complémentaires sur les altérations de composition des céramiques au cours du temps: cas de quelques alcalins et alcalino-terreux.” Revue d’Archéometrie 15: 117–122. Rottländer, R. C. A. (1981–1983). “Über die Veränderungen von Elementkonzentrationen in keramischen Scherben während der Bodenlagerung.” Teil I, Sprechsaal 114: 742–745; Teil II, Sprechsaal 115: 210–218; Teil III, Sprechsaal 116: 571–577. Sayre, E. V., Dodson, R. W., and Burr Thomson, D. (1957). “Neutron Activation Study of Mediterranean Potsherds.” American Journal of Archaeology 61: 35–41. Schneider, G. (1978). “Anwendung quantitativer Materialanalysen auf Herkunftsbestimmunen antiker Keramik.” Berliner Beiträge zur Archäometrie 3: 63–122. Schneider, G. (2002). “Chemische und mineralogische Zusammensetzung römischer und germanischer Keramik aus Ladenburg.” In: Lenz-Bernhard, G., Lopodunum III. Die neckarswebische Siedlung und Villa Rustica im Gewann “Ziegelscheuer”—Eine Untersuchung zur Besiedlungsgeschichte der Oberrheingermanen, Landesdenkmalamt Baden- Württemberg, Suttgart, Forschungen und Berichte zur Vor-und Frühgeschichte in Baden-Württemberg, Band 77 (Stuttgart), 617–644. Schneider, G. and Daszkiewicz, M. (2006). “Chemische Analysen zum Tafelgeschirr aus dem Militärlager von Dangstetten.” In: Roth-Rubi, K., Dangstetten III, Das Tafelgeschirr aus dem Militärlager von Dangstetten, Regierungspräsidium Stuttgart –Landesamt für Denkmalpflege (Stuttgart: Konrad Theiss Verlag), 167–193. Schwedt, A. (2004). “Untersuchung von (Spuren-) Elementkonzentrationsprofilen in archäologischer Keramik mittels Neutronenaktivierungsanalyse.” Dissertation, Mathematisch- Naturwissenschaftliche Fakultät, Rheinische Friedrich Wilhelms Universität, Bonn. Schwedt, A., Mommsen, H., and Zacharias, N. (2004). “Post- Depositional Elemental Alterations in Pottery: Neutron Activation Analysis of Surface and Core Samples.” Archaeometry 46: 85–101. Schwedt, A., Mommsen, H., Zacharias, N., and Buxeda i Garrigós, J. (2006). “Analcime Crystallization and Compositional Profiles— Comparing Approaches to Detect Post- Depositional Alterations in Archaeological Pottery.” Archaeometry 48: 237–251. Secco, M., Maritan L., Mazzoli, C., Lampronti, G. I., and Mattioli, S. P. (2011). “Alteration Process of Pottery in Lagoon-Like Environments.” Archaeometry 53: 809–829. Segebade, C. and Lutz, G. J. (1976). “Simultane instrumentelle Multielementbestimmung in antiker Keramik (Terra Sigillata) durch Aktivierungsanalyse mit hochenergetischen Photonen.” Journal of Radioanalytical Chemistry 34: 345–363. Segebade, C. and Lutz, G. J. (1980). “Photon Activation Analysis of Ancient Roman Pottery.” In: Slater, E. A. and Tate, J. O. (eds), Proceedings of the 16th International Symposium on Archaeometry and Archaeological Prospection (Edinburgh, 1976). (Edinburgh: National Museum of Antiquities of Scotland), 20–49. Thierrin-Michael, G. (1992). “Römische Weinamphoren—Mineralogische und chemische Untersuchungen zur Klärung ihrer Herkunft und Herstellungsweise.” Dissertation, Universität Fribourg. Walter, V. and Besnus, Y. (1989). “Un example de pollution en phosphore et en manganese des céramiques anciennes.” Revue d’Archéométrie 13: 55–64. Wilson, M. A., Carter, M. A., Hall, C., Hoff, W. D., Ince, C., and Betts, I. M. (2009). “Dating Fired-Clay Ceramics Using Long-Term Power Law Rehydroxylation Kinetics.” Proceedings of the Royal Society A 465: 2407–2415.
180 Gerwulf Schneider Zacharias, N., Buxeda i Garrigós, J., Mommsen, H., Schwedt, A., and Kilikoglou, V. (2005). “Implications of Burial Alterations on Luminescence Dating of Archaeological Ceramics.” Journal of Archaeological Science 32: 49–57. Zaun, P. E. (1982). “Influences of Soil Stratification on Prehistoric Pottery.” Neues Jahrbuch für Mineralogie Monatshefte 3: 106–118.
Chapter 12
Form al Ana lysi s and T y p ol o g i c a l Cl assificati on i n t h e St udy of Ancient P ot t e ry Daniel Albero Santacreu, Manuel Calvo Trias, and Jaime García Rosselló Introduction Clay is a highly malleable material which has been used for millennia to create vessels as well as an endless repertoire of artifacts. Form and decoration make up two of the most visible and accessible attributes of pottery, a fact which has probably precluded their thorough analysis and set the basis for the typological classification of archaeological pottery as early as the mid-nineteenth century.1 Broadly speaking, the study of pottery form and its classification can be considered on four analytical levels, which provide the tools needed to characterize vessels and incorporate them into multiple interpretative discourses (Figure 12.1). These levels cover from the most peculiar aspects to general ones, from the most analytical stages of research to the most interpretative arguments, as each of them aims at solving specific problems in the study of pottery form using, in turn, particular concepts and methodologies. Summing up, the many analytical levels may be organized as: (1) Description of pottery form. (2) Classification and elaboration of typological proposals. (3) Typological–interpretative tools. (4) Interpretative perspectives of typologies. Until the 1960s, these analytical levels were in step with the development of a theoretical basis regarding vessel description and typologies. However, since the 1970s, efforts have been mainly concentrated on promoting the methodological development of a kind of
Quantitative (e.g. metrical mathematical)
Class Type Variety Etc.
Polythetic/Monothetic
Paradigmatic/Taxonomic
Theorical/Practical
Intuitive/Objective
Pottery classification
Level 2
Variability symmetry/Shape/Size Isomorphism/Skeuomorphism
Interpretative tools
Level 3
Active
Non-active
Product Specialization Symbolic Approach Identity Style
Functionalist Approach
Fossil Index Seriation
Chrono-cultural Approach
Agency Ceramic Homology Habitus Hydridization
Technological Approach (multidimensional)
Interpretative perspectives
Level 4
Fractality/Homology (Extrinsic parameters)
Level 5
Figure 12.1 Summary of the different levels to approach pottery form and typological analyses discussed in the text. It includes the main concepts and tools in each level.
Qualitative (e.g. geometric, topological)
Formal descriptions (e.g. form-based analysis)
Level 1
Formal analysis and typological classification 183 grouping (Whittaker et al., 1998; Read, 2007), frequently considering the application of such methodology as the aim of the study (Dunnell, 1986; Read, 1989). This was the case for more systematized form analyses (e.g. form-based analysis) and their combination with mathematical and statistical protocols (e.g. discriminant analysis, principal component analysis, cluster analysis, curve analysis; Sheppard, 1971; Whallon and Brown, 1982; Read, 1989, 2007; Hendrix et al., 1996; Gilboa et al., 2004), which gave rise to the so-called electronic paradigm (Adams and Adams, 1991). Nevertheless, there has always been a minor interest in moving from the application of the different methods used to describe and classify pottery to the interpretative meaning of the typologies proposed (Sheppard, 1971; Read, 1989). Due to their growing marginalization, pottery typological and form studies are currently suffering in favor of archaeometric analyses, a revision and revalorization of the role played by the former is needed in order to restore them as potentially relevant tools to approach both technology and people in past societies. Furthermore, morphological analysis is considered to facilitate the generation of multidimensional and holistic interpretations of materiality. Consequently, it is assumed that vessel form is related to certain phenomena which are not evident in the technological studies that focus on vessel fabrics and forming. This chapter introduces a summarized revision of some of the theoretical–methodological aspects key to typology so as to comprehensively understand the main interpretative proposals which use this research tool, evaluating their particular aims, proceedings, and concepts. Finally, a revision of the use of pottery form analysis is proposed, as well as its incorporation into the theoretical and interpretative framework that is provided by the social theory of technology to elaborate explanations for the active role of vessel form and to develop typologies which include a new explanatory dimension.
Form Description and Classification Strategies Most of the classification efforts in the last decades have been devoted to the definition of strategies aimed at delving into the problems postulated by the first two analytical levels. A comprehensive revision of such research is far from the possibilities and aims of this chapter; instead, a short and general outline of the most important trends currently in use is provided. The first strategies for the study and description of pottery form and decoration were characterized by their limited, systematized, highly intuitive, eclectic, and subjective nature, supported by aesthetic assessments coming from the personal experience or perspective of the analysts. As a result, imprecise terminology usually based on morphofunctional criteria was used (Hendrix et al., 1996). Since the 1950s, however, these guidelines have been complemented by other methodological strategies—such as the form-based analysis (e.g. Sheppard, 1971; Ericson and Stickel, 1973; Hendrix et al., 1996), which was aimed at bringing greater objectivity to the description of pottery following a strict and systematized analysis of the form on the basis of geometrical models. Since then, morphometric quantitative analyses have been incorporated into morphological studies, significantly increasing the number of
184 D. A. Santacreu, M. C. Trias, and J. G. Rosselló attributes recorded in a pottery piece, as well as defining its form through the use of ratios, indexes, and mathematical models. Regarding the second level, the focus has been centered on defining typological classifications for pottery. Classifications in archaeology usually tend to organize the record into categories which share some internal coherence, depending on the similarities and differences present in the artifacts’ attributes. On the one hand, it implies the definition of categories and, on the other, the assignation of the individual pieces to such categories (Sheppard, 1971; Rice, 1987; Read, 1989). In pragmatic terms, both descriptions and classifications of pottery have to be systematic and coherent in order to promote the use of a standardized terminology and a typology devoid of subjective interpretations, so as to favor understanding amongst researchers (Whittaker et al., 1998). At this second level, typological classifications have originated either from proposals based on the intuitive researchers’ perceptions of the differences and similarities existing between ceramics (e.g. Krieger, 1944; Gifford, 1960; Rouse, 1960), or alternatively, on allegedly more objective methodologies which used mathematical and statistical tools for grouping. They are intended to inductively create replicated descriptions and classifications of the vessels which can simultaneously compare a broader number of attributes with better definition of both the data and the variables being analyzed. This kind of tool creates groups with a strong internal coherence and provides a less arbitrary boundary between categories (Whallon and Brown, 1982; Read, 2007). However, these grouping strategies are not devoid of problems, since the analysis frequently incorporates variables or attributes of the vessels which are not relevant for archaeological questions. Nevertheless, some authors such as Read (1989, 2007) considered that intuitive classifications could be even more informative and useful than some of the more objective ones. Typological groupings may also vary depending on the way the attributes are considered: they can either be paradigmatic (Whallon, 1972) or taxonomic (Read, 1989, 2007). In the for mer, the most frequent in multivariant statistical analyses, no hierarchy is postulated for the variables used in the classification, so all the attributes recorded for a vessel can be treated both simultaneously and independently. In the latter, the several attributes of a vessel are considered to have a different validity for determining pottery types; thus they have to be used in a sequential and hierarchical order according to many different criteria. Furthermore, pottery classifications might also depend on different ontologies. In this sense, a heated debate has been generated about the emic or etic nature of the typologies archaeologists create. Hence, it is possible to discriminate between theoretical (e.g. Krieger, 1944; Spaulding, 1953; Gifford, 1960; Rouse, 1960) and practical typologies (e.g. Hill and Evans, 1972; Adams, 1988; Adams and Adams, 1991; Kampel and Sablatnig, 2007). The former can be encompassed as emic classifications, closely related to so-called folk classifications or ethnotaxonomies (Kempton, 1981; Rice, 1987; Fowler, 2006), where both potters and other members of their community (i.e. consumers and non-consumers) are assumed to base their classifications on tangible aspects related to certain physical parameters of the materials (e.g. appearance, form, and size) as well as on intangible or cultural phenomena. In this context, many scholars understand the pottery classification process as proceeding from an inductive or theo retical nature which allows the discovery and/or replication of the natural types present in the artisan’s mind; a classification which underlies the data recorded. In this sense, the ceramic types defined are considered to carry an important cultural and historical meaning, and, consequently, to mirror the ideas and values of the people who made and used the artifacts.
Formal analysis and typological classification 185 As a counterpart to emic typologies, the classifications drawn by analysts from etic perspectives, also referred to as devised classifications, have dominated research since the 1960s, coinciding with the boom of processualist and positivist views in the typological study of archaeological pottery. Although these perspectives consider that the pottery found in archaeological sites was originally related to the rationality scheme of the potters and their communities, as well as to the functional, socioeconomic, and symbolic–ideological contexts which characterized the life-cycle of the vessels, it must be accepted that archaeological typologies designed in the present have little to do with folk classifications. The many complex terminological and classificatory shades used by the members of a certain culture to arrange a pottery assemblage in their minds may be impossible to perceive or replicate by foreigners. In this sense, a number of papers have proved (Weigand, 1969; Birmingham, 1975) that many of the classificatory elements frequently used by archaeologists (e.g. base, rim, lip form) have a weak connection with those used by ancient societies. Moreover, it should be considered that several of the elements people use to classify their surrounding material culture are difficult to see in the archaeological record. Practical perspectives argue that typologies imply an interpretative process, an analytical and creative operation, that originated in the ideational and conceptual realm. Similarly to the other attributes of pottery (e.g. fabric), they suppose an action which goes beyond the empirical world and is deeply influenced by the theoretical approach of the researcher. However, it triggers a certain degree of artificiality and arbitrariness in the classificatory pro cess as, for instance, the researcher has to decide which attributes, of the seemingly endless possibilities, should be measured and selected for formal comparison. In short, morphometric and typological analyses of pottery are considered to be born from the rational schemes of the scholars and to be aimed at structuring a specific ceramic universe. This idea invited some authors (Hayden, 1984; Rice, 1987; Whittaker et al., 1998) to make a conceptual distinction between classification and typology. The former is considered as an empirical grouping of objects based on their differences and similarities. Typology, on the other hand, implies a classification with a clear theoretical background as well as explicit and well-agreed norms or proceedings to solve specific problems. According to Rice (1987), although devised classifications and folk classifications originate from different and often opposed concepts and objectives, both can provide feedback and interesting incentives for pottery classification. Currently, the general option is an intermediate position, where typologies are accepted to be subjective but are also potentially suitable for going beyond the mere description or organization of the record and being used in an interpretative discourse. That is, the differences perceived at the etic level can be reflected at the emic level (Read, 2007). In any case, each and every different typological proposal defines its own concepts to orga nize the factual universe depending on its interpretative and instrumental objectives and the kind of record it works with. This is the case particularly for concepts such as mode (Rouse, 1960), type-variety (Gifford, 1960), and type (Dunnell, 1986), among many others. This diversity is a consequence of each typological classification strategy (i.e. the attributes selected and the grouping strategies used) being drawn, either consciously or unconsciously, for specific research objectives which vary depending on their intention of creating descriptive, comparative, or analytical typologies or, rather, dealing with chronological, cultural, functional or technological aspects by using such typology (Adams and Adams, 1991). Furthermore, as already noted by these authors, the aim of typology determines the kind of record to be selected (for instance, if it considers only complete pieces or also includes sherds).
186 D. A. Santacreu, M. C. Trias, and J. G. Rosselló
Typological-Interpretative Tools The third analytical level comprises a series of interpretative tools which transcend the typological classifications and morphological analyses described above. These tools, which are understood and used differently by each perspective, allow the researcher to go beyond the data and build coherent discourses about past societies. A large number of interpretative tools can be included at this level, some of which are the following: (a) Morphotypological variability. The morphological variability of the record can be accessed by organizing the pottery assemblage into different categories and observing the pieces’ differences and similarities in different chronological and spatial segments. This variability is informative of the routines and the repetition of actions followed in the production process. Thus, some of its aims are facilitating a formal comparison of pottery and identifying the existence of either variations or continuities in the record (Adams and Adams, 1991). As will be explained in the section “Pottery Typologies and Main Interpretative Proposals,” the variability present in any record has been interpreted in a myriad ways which are also related to different analytical scales (Ericson and Stickel, 1973; Dobres, 1999) and objectives. In most interpretations, such as culture historical and processualist typologies, macroscalar analyses are dominant as they consider extensive regional spaces and temporal segments (e.g. Hendrix et al., 1996), although microscalar analyses of variability can be also applied (Dobres, 1999). (b) Symmetry, size, and form as evidence of the potter’s expertise. The analysis of symmetry, form, and size of the individual pieces is fundamental for pottery studies. As well as descriptors of the vessels, the symmetry, the kind of forms modeled, or the size of the objects have been proposed as an indication, together with a number of attributes such as the fabric, surface treatment, or wall thickness, or the potters’ level of technical skill (Sheppard, 1971; Budden and Sofaer, 2009; Vidal, 2011). (c) Translation of the form (isomorphism and skeuomorphism). The concept of isomorphism implies the repetition of a given form in several objects, thus providing them with the same meaning. In these cases, any potential variation does not trigger changes in the behavior or relationship between the elements constituting the object; thus its structural relation remains constant (Samaniego, 2013). Isomorphism cases in pottery may be numerous and varied. The most common is the preservation of the form and the metrical proportions of the vessels when their size is modified. Another example may be the repetition of decorative patterns on different media (Figure 12.2). Skeuomorphism, a variant of isomorphism which has been also reinterpreted by the different schools of thought (see Frieman, 2010), refers to the translation of the form and other perceptive aspects of the ceramic universe to other technologies and vice versa. Well-known examples of skeuomorphism in pottery are the reproduction of forms and decorative patterns copying basketry, carpentry, or leather-work (Manby, 1995; Hurcombe, 2008), as well as the imitation of some types of metallic containers such as occurs in Etruscan bucchero pottery. The analysis of both isomorphism and skeuomorphism has a large interpretative potential when its identification is followed by an evaluation of the influence of these phenomena over society. Hence, it is
Formal analysis and typological classification 187
0
10 cm
Figure 12.2 Isomorphic relation between the decorative motifs recorded on Late Iron Age pottery and bronze discs in Mallorca (Spain). important to record whether the translation of the form is restricted to certain types or affects a wide variety of forms, which is the direction of the translation, whether the loan is limited to two technologies or affects more spheres, whether the replicated form shares the same contextual relationship in the many materials or is significantly different, and so on.
Pottery Typologies and Main Interpretative Proposals Both pottery typologies and interpretative tools are drawn, conceived, and used in a certain way depending on their ultimate aims. Interpretative strategies in archaeology have played a key role in the classificatory systems which structured and organized the real world. After briefly describing the use of form analysis and typologies by the different interpretative posi tion, a projection of the future of morphotypological analyses and their inclusion in more technological–social interpretations will be discussed.
Typologies and Chronological Frameworks Typological seriation strategies can be considered, together with stratigraphic principles, the first methodological tools promoting modern archaeology, because one of the first objectives of typological seriation was to organize and chronologically place the myriad of archaeological objects already recovered, which in the early nineteenth century were considered exclusively from a collector’s and antiquarian’s point of view (Trigger, 1989; Orton et al., 1993). The first modern typological proposals are therefore found at the very beginnings
188 D. A. Santacreu, M. C. Trias, and J. G. Rosselló of the discipline, in the development of seriation strategies according to raw material and stylistic criteria. Archaeological objects were not only classified and organized into types, but were also relocated in concrete chronological and stratigraphic sequences: this was the birth of relative chronology, the essential and only dating method available before the discovery of absolute dating. Thus, the late nineteenth century produced paradigmatic examples of pottery seriation (see Chapter 37, this volume) such as Smith’s study of terra sigillata (1854), Pottier’s Normand pottery (1867), or Plique’s research (1887), as well as Petrie’s work in Lachish, Palestine (1891), and Egypt (1890), where the ceramic types were identified in the stratigraphic sequence (Sinopoli, 1991; Orton et al., 1993; Hendrix et al., 1996). Originally, the association between typology, seriation, and stratigraphy rested on considering the ceramic type as the chronological reference for a culture, based on the principles which allowed the paleontological identification of fossils with geological strata (Adams and Adams, 1991; Orton et al., 1993). The definition of ceramic types from different sites provided cross-datings which were grouped into regional chronological sequences, the similarities between types representing temporal proximity (Trigger, 1989). This twofold nature of pottery typology (i.e. as stylistic-formal organization and chronological reference) was present as one of the main analytical strategies in most archaeological discourses until well into the twentieth century, and is still in use despite the conceptual redefinition of typology. Currently, multiple archaeological discourses still find this relationship, based on the concept of relative chronology and cross-dating, essential in many typological classifications. This association is clear, for instance, in the analysis of wheel-made seriated pottery, such as amphorae, thin-walled Roman ceramics, and Terra Sigillata.
Typologies and Cultural Frameworks Far from denying the chronological use of typology, culture history provided it with a new meaning. A century ago, the evident relationship between certain findings and concrete geographical areas was proposed, and the definition of cultural areas as being home to different human groups followed (Trigger, 1989). This threefold association among recurring objects, geographical areas, and cultural groups signaled an important qualitative and conceptual leap. Since then, types have not only provided a concrete chronology but also a regional and cultural perspective, and even an ethnic affiliation by studying the objects recovered (Childe, 1929). Culture history defined a new concept of culture, which was eventually integrated into archaeology as the archaeological culture and fossil directeur, which implied the relationship among the archaeological materials found in a particular place, their chronology, and a specific ethnic group or people (Childe, 1925, 1929; Kossina, 1926). This new perspective promoted typology to another dimension, as it fulfilled one of the objectives of the culture historical paradigm: the interpretation of the archaeological record as a mirror of nameless prehistoric peoples identified by the characterization of their archaeological cultures rather than as developmental evidence of their culture. Furthermore, diffusionism as an interpretative tool was used to explain cultural change, using materiality to define the origin, movement, and interaction of those peoples. In these interpretations, pottery gained protagonism owing to its identification with cultures and ethnic groups. Many typological strategies used for pottery analysis had an impact
Formal analysis and typological classification 189 on its cultural dimension (e.g. Krieger, 1944; Gifford, 1960; Rouse, 1960). Until the 1950s pottery typologies and the concept of fossil directeur were prevalent for the archaeological identification of ethnocultural groups. Such was the case of the debate on the integration of the Prague ceramic type into Slavic ethnicity, the association of Linearbandkeramik with the first Neolithic communities in Central Europe (Childe, 1929; Klopfleisch, in Hibben, 1958), and the first research on the ethnic group responsible for the Bell Beaker pottery (Castillo, 1928; Bosch Gimpera, 1940). Aiming at defining chronocultural typological entities, their analysis of pottery variability was focused on macroscales covering regional territories and large time periods. Furthermore, in order to determine the scope and distribution of cultural entities, diffusionism understood isomorphic phenomena as the imitation and subsequent copy of forms or other specific elements owing to the dominance or influence of one culture over another by means of trading (e.g. colonial), political, military, or other relations.
Typology and Functionality The New Archaeology, closely related to a functionalist view of society, considered pottery a product (Sackett, 1977; Binford, 1989) or tool (Braun, 1983) whose manufacture responded to known and preconceived needs. Thus, one aspect which characterized pottery life for them was its function regarding one or more ends (Rice, 1990); hence, pottery was designed following functional criteria (Sheppard, 1971; Smith, 1985; Rice, 1987; Orton et al., 1993). Its function determined or restricted pottery forms, so innovative forms may have responded to new needs, making them representative of human behavior. The definition of morphofunctional relationships demanded more systematic and (presumed) objective criteria, such as ethnographic analogy, to address vessel functionality: differentiating between the description of the form and the analysis of functionality (Birmingham, 1975; Henrickson and McDonald, 1985; Rice, 1987). This close relationship between form and function, supplemented by absolute dating which overcame the definition of chronocultural entities as the main interest of typologies, favored the development of new classificatory and interpretative strategies. This was the theoretical context at the peak of functional classifications characterized by organizing materiality from the presumed function of the artifacts, usually considering the morphological attributes inherent to the objects (Adams and Adams, 1991). Here, pottery classifications depended primarily on vessel form to establish functional categories. In such classifications, numerous parameters of the form (e.g. mouth width) were considered broad indications of its function, the kind of contents (i.e. liquid vs. solid), and their manipulation inside the container (Sheppard, 1971; Henrickson and McDonald, 1985; Rice, 1987; Sinopoli, 1991). Similarly, varieties of the form were also related to the use of certain culinary techniques (Rice, 1987). The link between the function and effectiveness of a form and the physical properties of the vessels was also addressed. In this sense, some morphological elements, such as the curved profile in a pot, were proved to maximize thermal shock resistance in cooking pottery (Woods, 1986) as well as impact resistance (Schiffer and Skibo, 1987), caloric efficiency, or thermal conductivity (Hally, 1986; Schiffer and Skibo, 1987).
190 D. A. Santacreu, M. C. Trias, and J. G. Rosselló In short, this functionalist interest pretended to understand the socioeconomic view of a society. For processualism, isomorphism did not necessarily imply the direct derivation or influence of one style on another; it may have responded to an autochthonous adaptation to a specific environment and economy (e.g. agriculture) together with the use of peculiar culinary practices (Sheppard, 1971), with the subsequent functional specialization of the whole materiality. Similarly, processualist typologies considered a limited variability in any specific form as the evidence for both specialization and a reduced number of potters. Actually, this school has always been interested in defining the degree of specialization and its links with more evolved behaviors, where potters produced better quality and technically more efficient products. This question was addressed using form and surface analyses, considering asymmetrical profiles the production of scarcely specialized potters regarding their mastering of technical gestures. Vessel size was also considered indicative of experience (Longacre, 1999; Brodà et al., 2009).
Typology, Textual Metaphor, and Identity The first post-processual paradigms stressed the active role of pottery forms and decorations as the material medium of a communicative event which was expressed with the symbols inherent to the objects. In this context, Hodder’s seminal work (1982) shook up archaeology in general and pottery studies in particular by highlighting the symbolic and ideologi cal aspects of material culture. According to his views, material culture and, consequently, pottery features were significantly constructed and should be considered an active element in the definition of societies. He also reinforced the idea that material culture was neither innate nor did it passively mirror society; on the contrary, it was created by people’s actions (Hodder, 1998). These ideas originated the textual metaphors held by interpretative archaeology (Hodder, 1991; Tilley, 1999). Under the influence of semiotics and hermeneutics, it understood the analysis of material culture and its interpretation as a communicational event full of signifier and signified elements. People acted in accordance with the social symbolic system and each individual in turn played an active role in his/her society. Thus, pottery forms and attributes were not just a neutral product but the embodiment of the symbolic connotations of a community in a certain place and time. Pottery, as a social product, reproduced the symbolic system of the society it was inserted in, and the similarities and differences embodied in the stylistic tendencies of formal analysis expressed a common rationality and emphasized the identity of a particular style against the rest (Prieto, 1999). This new association with symbolic and identitarian constituents explains the new uses of typology. In this view, the study of pottery style made visible identities related to status, cast, ethnicity, and genre, among others. Mahias (1993), for instance, documented the link between technological and stylistic variations and the caste-based social structure in India. Regarding ethnic identity, a number of papers integrated typological and technological analyses of pottery (Dietler and Herbich, 1989; Gallay and Huysecom, 1991; Hardin and Mills, 2000). The work of Gosselain (2000) exemplified the distribution and expansion of styles regarding local languages. Alternatively, Bowser (2000) observed that Achuar and Quechua women in Ecuador use pottery as a descriptor of political identity.
Formal analysis and typological classification 191
Typology and Technology: New Perspectives and Interpretative Possibilities of Social Dynamics In the last few decades, there has been a rise in the incipient application of typological strategies to technological analyses born from a clear integration with the social dynamics of the groups studied. This social perspective of technology has mainly been developed from anthropological views and focused its interest in the study of the materials and techniques associated with pottery-making. However, in addition to these issues, typological analyses are also a useful strategy for a social approach to technology. This analytical strategy has gained popularity since the 1980s, with the active participation of two schools from different academic backgrounds: the anthropology of techniques (e.g. Lemonnier, 1986) and the analysis of technology of the social agency theory (e.g. Dobres, 2000). The inclusion of this proposal in archaeological questions implies that the study of the objects—that is, elements made, used, exchanged, maintained, and abandoned in a social space, usually during daily activities—can lead to the complex social practice of the technological process and its connection with rationality schemes, social praxis, power relations, economic bases, material reality, and so on; all of them interpreted as parts of a whole and unable to be understood separately because they are mutually constructed. The technological process, similarly to other social activities, would originate in the daily and contingent praxis through habitus dynamics (Bourdieu, 1977) and agency (Barrett, 1994), in a web of relations between objects and people (Latour, 2008). It implies the incorporation of patterns culturally chosen through constant practice. These attitudes, elections, and perceptions of technical alternatives, embedded in social relations and configured by the habitus, may be perceived as natural and absolutely logical, besides any consideration of the efficiency of techniques and materials. Even if technological practice and tradition can be seen as predetermined and static, they imply relational and dynamic phenomena. They are an historical product which is active in the present, for technological practice is materialized in a series of learnt and interiorized dispositions which allow the reproduction of social structures and, at the same time, explain their changes through agency. Hence, technology becomes a complex cultural phenomenon, incorporated in historically contingent worldviews, interiorized social actions, and agency. Consequently, the study of technological processes has to transcend the analysis of the simple physical medium as it is intimately connected with social phenomena. Typological strategies and their interpretative contributions can be incorporated into this discourse as the analysis of pottery form, as far as it implies a collective technological choice in a specific social context, makes possible the identification of the conscious and unconscious schemes, and praxis of the technological process. Hence, typological strategies are understood as a valuable tool for the interpretation of technological–social dynamics. This implies considering pottery form and the typologies developed from it as an active element relevant for interpretation and thus different from passive views and typological mechanicist interpretations. In this interpretative context, vessels’ morphometry and typology, accepting pottery as a social rather than individual formalization, could be considered a priori as indirect evidence of certain technological praxis embodied with sociocultural connotations. However, such
192 D. A. Santacreu, M. C. Trias, and J. G. Rosselló evidence does not need to be defined by the structure of the classifications devised; that is, by the types, categories, or groups created, nor by the grouping strategies used (e.g. paradigmatic/taxonomic, intuitive/objective). So, from this perspective, typological analyses are mainly focused in taking advantage of the interpretative potential of third-level tools. Some examples may clarify their current use.
(a) Analysis of Microscale Knowledge Transfer and Agency This perspective considers the analysis of morphotypological variability at a microscale so as to deepen the dynamics behind two specific questions: knowledge transfer and agency. In clear opposition to processualist typologies, social technology considers that the variability present in the pottery forms produced by a community depends upon people’s interaction (either conscious or unconscious) with multiple and varied elements and values typical of their society. The presence of variability is frequently considered as evidence of a break with traditional learning patterns and the disintegration of the potters’ technological–formal schemes (García Rosselló, 2010). A second interpretative position in the analysis of variability is focused on the role played by individuals and their agency capacity. The idea of the individual has traditionally been uncomfortable for the main interpretative paradigms; clearly seen in the many typological grouping strategies which tended to establish highly standardized typologies insisting on the importance of making them consistent and homogeneous (Whittaker et al., 1998). In the search for consistent standardizations, “anomalous” cases were considered outliers—that is, unconnected with the norm determining the perceptive difference noted by the researchers—thus they were difficult to classify, understand, and explain. Since the active role of the individuals and their agency capacity has been made clear, the analysis of marginal forms or types validated in typological classifications has offered more interpretative flexibility to address different kinds of phenomena while explaining the complexity observed in material culture with more coherent discourses (Dobres, 1999). Regarding agency, it should be remembered that vessel form, as well as other highly visual attributes such as decorative motifs, is frequently a collectively perceived aspect. These more visible dimensions of pottery are precisely those favored by individuals to communicate messages to the rest of the community and define their social space (Herbich, 1987; Gosselain, 2008). It explains the higher degree of innovation in vessel form and decoration than in any other dimension of pottery, such as paste preparation or modeling techniques, as the latter are less visible to the rest of the community. A useful strategy to study the concept of variability and its interpretative potential in knowledge transfer dynamics, agency, and technological tradition could be their articulation in the type-variety system (Gifford, 1960). Although this system was developed as a mere taxonomic tool, some scholars (Rice, 1987; Sinopoli, 1991; Read, 2007) suggest its use to interpret cultural aspects and observe record variability. This typological system presents the advantage of simultaneously recording the cultural patterns shared and accepted by the whole society which originated the traditions materialized in the recurring material attributes (types), while also reflecting the variations in the artifacts as a consequence of an individual’s or small social groups’ actions (varieties). In short, this kind of conceptual tool becomes useful when materiality gains an active role in the configuration of society.
Formal analysis and typological classification 193 (b) Potter’s Expertise and Social Context of Production A second variable used by this school is the analysis of formal and perceptive aspects of pottery such as symmetry, size, or form complexity as representative of the potter’s profile. Far from considering the potter in evolutionary terms such as specialization and technical complexity, morphology is used to identify apprentices and consider questions such as learning, knowledge transfer, and the social interaction of pottery production. The identification of the potter’s expertise using these variables demonstrates the existence of processes made of repeated actions which produce certain standardization and technological knowledge transfer (Budden and Sofaer, 2009). Because pottery features, which result from the specific way-of-doing of each person, respond to social determinants and evidence the strategies of the technological habitus used by a potter to manufacture a product in a concrete social situation, their analysis would evidence the potter’s profile. His/her technical skill and its embodiment in certain morphological attributes of the pottery should be considered contingent. Thus, they can be evaluated only in the concrete contexts for pottery making. In this view, technical skill is not studied as the mere evidence of the potter’s technical knowledge per se or the existence or a specialized production, but rather as the response to a specific social context; and, consequently, it constitutes a key element to enlarge on the dynamics of a society.
(c) Isomorphism and Hybridization Social technology understands that the ultimate interpretation of isomorphism and skeuomorphism is determined, not by simple imitation, but by the integration of praxis into a contextual and social framework as well as into the technological and symbolic relationship between the different kinds of materials and objects participating in any society. Thus, their interpretation should consider that the loan of a form does not necessarily imply a functional or symbolic translation: it cannot be automatically inferred by the sole formal similarity of two objects, as only contextual analyses can determine the coincidence between function and meaning. Postcolonial perspectives, used in the analysis of material culture (Gosden, 2004; Van Dommelen, 2006), understand the translation of the form derived from the contact between different cultures as hybridization. This idea is based on the premise that intercultural contacts are never neutral and that the parts involved cannot be considered passive entities, but rather active agents. The complex phenomena of hybridization occurs where the material outcome of cultural contact is noticeably different from the original material culture of each culture because both groups have actively modified, reinterpreted, and hybridized practices, objects, and dynamics, giving rise to new contexts and meanings. In the case of the typological analysis of pottery, the translation of the form, even if it retains some reminiscence of the original, goes hand in hand with variations and reinterpretations which constitute a rupture in the structural relationship between the parts which have originally made the object. Hybridizations in pottery form are quite common in the Western Mediterranean during the Iron Age, particularly along the French coast (Dietler, 1997) and the Balearic Islands (Albero, 2011), resulting from the intensification of contact between indigenous and Greek or Punic communities. Although formal references can still be identified, these new hybrid types present their own peculiarities and characteristics which affect many aspects related to pottery, from the manufacture system (hand-made
194 D. A. Santacreu, M. C. Trias, and J. G. Rosselló
Figure 12.3 Format translation related to hybridization phenomena between Punic wheel-thrown vessels and hand-made indigenous pottery in the Late Iron Age in Mallorca (Spain). pottery continues) to structural or metric elements. The result is a hybridized and reinterpreted form, typologically different from the original schemes of both indigenous and foreign communities (Figure 12.3).
(d) Homology and Fractality Conceiving ideas and matter as a group of connected webs and nodes rather than separate elements (Latour, 2008) leads to Lemonnier’s concept of representation (1993). Hence, any technological action is related to a series of mental operations which are often unconsciously internalized by a habitus which has to be interpreted in the global technological scheme of the group and facilitates technological transfers and loans among crafts sharing the same scheme. Furthermore, the concept of representation, the mental models of the sequence and order of the action, is not exclusively related to a concrete technological action, but incorporates content and information of an ideological, social, and/or symbolic kind which function in a network of supra-technological meaning affecting the totality of the signification and semiotic schemes and models of the community. The idea of the transversal direction of the signification schemes and technological processes in a community suggested by Lemonnier (1993) leads to a fractal and homological approach to society. This approach is related to the formal view of the object analyzed from typological strategies. In this sense, the ultimate analysis of third-level tools is understood to be inserted in an holistic interpretative framework. Thus, the study of typological identities and correspondences which have been analyzed from third-level variables has to follow four analytical levels: (a) an intrinsic level, restricted to pottery form; (b) a second intrinsic level from a multidimensional view of pottery technology; (c) an extrinsic analysis covering the diverse technological fields; and (d) a further extrinsic analysis between technology and the remaining social spheres considering different scales, as they represent the many manifestations of the same phenomena.
Formal analysis and typological classification 195 This integral analysis is possible thanks to two strategies: the study of homology relations among different fields and the analysis of the presence of fractal patterns. The concept of homology, taken from Bourdieu’s perspective (1977), reveals structural similarities between different fields beyond their own peculiarities and dynamics. Applied to the social interpretation of technology and, specifically, to typological strategies, this concept of homology would refer to the similarities in the praxis and dynamics of the diverse technological fields of the society. As an interpretative strategy, homology aims at analyzing whether the behavior of each of the third-level tools used is restricted to the typological dimension of pottery or whether structural homologies are observed in other dimensions of the ceramic universe regarding the rest of the steps in the châine opératoire, such as raw material management, paste preparation, or forming activities. This analysis provides a multidimensional picture of pottery and identifies whether the same phenomena are materialized throughout the pottery-making process. Later on, based on the logics of the homology of fields, it is possible to evaluate if the dynamics identified in the pottery are present in other technologies (e.g. metallurgy, building, glass-making, basketry). The documentation of the same dynamics in other stages of pottery technology, as well as their identification in the rest of the technological manifestations, furthers the analysis of the typological differentiation, transcending the mere formal identification level to be inserted in the dynamics of the global technological scheme of the community. This holistic view would provide information regarding knowledge transfer dynamics, the social structures related to the learning process, technical skill, the social value of objects, and so on. The second tool which the comparison and signification aspects of third-level variables allow is the application of fractal strategies which try to record the existence of recurrent patterns or structures at several scales. Fractal patterns facilitate comparison between microscalar and macroscalar processes, for they are understood as manifestations of the same dynamics (Brown et al., 2005). A fractal interpretative framework constitutes an effective tool to integrate the typological data into another manifestation of scalar models which are expressed in the same way at upper levels, such as the technology or the social and ideological relations of the group analyzed. This strategy aims at confirming whether the typological patterns of pottery variability, appearance, and perception, as well as isomorphism and hybridization, now integrated into the technological praxis of the group and thus with ideological, social, or symbolic contents and information, are also present in other dimensions of the society. Documenting the patterns detected at these different scales helps us evaluate the significance level of the typological proposal and articulate more complex and coherent interpretative discourses beyond the classification strategy. It thus demands integrating the typological strategy first into technological questions and then expanding it to upper-level social and ideological discourses.
Conclusions This chapter has examined the wide variety of systems used by archaeology to analyze pottery form and decoration depending on the different philosophy, methodology, and interpretative option favored, originating quite diverse typologies which respond to simi larly differing objectives. Nevertheless, the several interpretative proposals should not be
196 D. A. Santacreu, M. C. Trias, and J. G. Rosselló considered exclusive but rather complementary: each proposal is important in the study of the form. For instance, without the chronological determination it would have been hard to deal with other aspects related to the use of ancient vessels. Nevertheless, once these temporal and cultural parameters have been established, it should be considered whether it is worth devoting so much effort to the description of pottery form and typological study. Taking this view, some scholars have stated that form analysis and typology development has turned into a “rite of passage” archaeologists have to go through. However, the description of vessel form and the typological classification of pottery should not nowadays be an end in itself in archaeological research. A thorough critical reflection is needed to evaluate the potential role and use of these kind of analyses in the study of pottery, both in historical and anthropological terms. This question is particularly critical in the case of prehistoric pottery, where a lack of systematization and large formal variability within a single assemblage could impair the development and application of reliable typologies able to achieve accurate seriations and chronological determinations.
Note 1. Although typological classifications of pottery may include technical aspects (e.g. fabric or forming technique), this chapter deals with the classical concept of typology; that is, the analysis and classifications focused on vessel form, size, and decoration.
References Adams, W. Y. (1988). “Archaeological Classification: Theory Versus Practice.” American Antiquity 62: 40–56. Adams W. Y. and Adams, E. W. (1991). Archaeological Typology and Practical Reality (Cambridge: Cambridge University Press). Albero, D. (2011). Caracterización tecnológica, social y adaptación funcional de cerámicas prehistóricas en el Oeste y Sureste de Mallorca (1700–50 BC): aproximación sincrónica y diacrónica a partir del estudio arqueométrico de pastas (Granada: Universidad de Granada). Barrett, J. C. (1994). Fragments from Antiquity: An Archaeology of Social Life in Britain, 2900– 1200 BC (Oxford: Blackwell). Binford, L. R. (1989). “Styles of Style.” Journal of Anthropological Archaeology 8: 51–67. Birmingham, J. (1975). “Traditional Potters of the Kathmandu Valley: An Ethnoarchaeological Study.” Man 10(3): 370–386. Bosch-Gimpera, P. (1940). “The Types and Chronology of Western European Beakers.” Man 40: 6–10. Bourdieu, P. (1977). Outline of a Theory of Practice (New York: Cambridge University Press). Bowser, B. (2000). “From Pottery to Politics: An Ethnoarchaeological Study of Political Factionalism, Ethnicity and Domestic Pottery Style in the Ecuadorian Amazon.” Journal of Archaeological Method and Theory 7(3): 219–248. Braun, D. (1983). “Pots as Tools.” In: Moore, J. and Keene, A. (ed), Archaeological Hammers and Theories(New York: Academic Press), 107–134. Brodà, Y., Cannavo, V., Govi, E., Levi, S., Marchetti, S., and Pellacani, G. (2009). “Bronze Age Terramare Pottery from Northern Italy: Exercises in Experimental Reproduction.” In: Biró,
Formal analysis and typological classification 197 K., Szilágyi, V., and Kreiter, A. (eds), Vessels: Inside and Outside (Hungary: Hungarian National Museum), 103–109. Brown, C., Witschey, W., and Liebovitch, L. (2005). “The Broken Past: Fractals in Archaeology.” Journal of Archaeological Method and Theory 12(1): 37–78. Budden, S. and Sofaer, J. (2009). “Non-Discursive Knowledge and the Construction of Identity Potters, Potting and Performance at the Bronze Age Tell of Százhalombatta, Hungary.” Cambridge Archaeological Journal 19(2): 203–220. Castillo, A. (1928). La cultura del vaso campaniforme. Su origen y extensión en Europa (Barcelona: Facultad de Filosofía y Letras). Childe, G. (1925). The Dawn of European Civilization (London: Routledge and Kegan Paul). Childe, G. (1929). The Danube in Prehistory (Oxford: Oxford University Press). Dietler, M. (1997). “The Iron Age in Mediterranean France: Colonial Encounters, Entanglements, and Transformation.” Journal of World Prehistory 11(3): 269–358. Dietler, M. and Herbich, I. (1989). “Tich Matek: The Technology of Luo Pottery Production and Definition of Ceramic Style.” World Archaeology 21(1): 148–183. Dobres, M. A. (1999). “Of Paradigms and Ways of Seeing: Artifact Variability as if People Mattered.” In: Chilton, S. (ed), Material Meanings: Critical Approaches to the Interpretation of Material Culture (Salt Lake City: University of Utah Press), 7–22. Dobres, M. (2000). Technology and Social Agency (Oxford: Blackwell). Dunnell, R. C. (1986). “Methodological Issues in Americanist Artifact Classification.” In: Schiffer, M. (ed), Advances in Archaeological Method and Theory (New York: Academic Press), 35–99. Ericson, J. E. and Stickel, G. (1973). “A Proposed Classification System for Ceramics.” World Archaeology 4(3): 357–367. Fowler, K. D. (2006). “Classification and Collapse: The Ethnohistory of Zulu Ceramic Use.” Southern African Humanities 18(2): 93–117. Frieman, C. (2010). “Imitation, Identity and Communication: The Presence and Problems of Skeuomorphs in the Metal Ages.” In: Eriksen, B. V. (ed), Lithic Technology in Metal Using Societies, vol. 67 (Højbjerg: Jutland Archaeological Society Publications), 33–44. Gallay, A., Huseycom, E., and Mayor, A. (1994). Peuples et céramiques du delta interieur du Niger (Genève: Departement d’Anthropologie et d’Ecologie de l’Université de Genève). García Rosselló, J. (2010). “Análisis traceológico de la cerámica. Modelado y espacio social durante el Postalayótico (s. V-I AC) en la península de Santa Ponça (Calvià, Mallorca).” Unpublished PhD thesis, Universitat de les Illes Balears, Mallorca. Gifford, J. C. (1960). “The Type-Variety Method of Ceramic Classification as an Indicator of Cultural Phenomena.” American Antiquity 25(3): 341–347. Gilboa, A., Karasik, A., Sharon, I., and Smilansky, U. (2004). “Towards Computerized Typology and Classification of Ceramics.” Journal of Archaeological Science 31: 681–694. Gosden, C. (2004). Archaeology and Colonialism. Cultural Contact from 5000 BC to Present (Cambridge: Cambridge University Press). Gosselain, O. (2000). “Materializing Identities: An African Perspective.” Journal of Archaeological Method and Theory 7(3): 187–217. Gosselain, O. (2008). “Thoughts and Adjustments in the Potters Backyard.” In: Berg, I. (ed), Breaking the Mould: Challenging the Past through Pottery. BAR International Series 1861 (Oxford: Archaeopress), 67–79. Hally, D. J. (1986). “The Identification of Vessel Function: A Case Study from Northwest Georgia.” American Antiquity 51(2): 267–295. Hardin, M. A. and Mills, B. A. (2000). “The Social and Historical Context of Short-Term Stylistic Replacement: A Zuni Case Study.” Journal of Anthropological Method and Theory 7(3): 139–163.
198 D. A. Santacreu, M. C. Trias, and J. G. Rosselló Hayden, B. (1984). “Are Emic Types Relevant to Archaeology?” Ethnohistory 31(2): 79–92. Hendrix, R. E., Drey, P. R., and Bjornar Storfjell, J. (1996). Ancient Pottery of Transjordan. An Introduction Utilizing Published Whole Forms: Late Neolithic Through Late Islamic (Berrien Springs, MI: Andrews University Press). Henrickson, E. F. and McDonald, M. (1983). “Ceramic Form and Function: An Ethnographic Search and an Archaeological Application.” American Anthropologist 85: 630–643. Herbich, I. (1987). “Learning Patterns, Potter Interaction and Ceramic Style among the Luo of Kenya.” African Archaeological Review 5: 109–136. Hibben, F. C. (1958). Prehistoric Man in Europe (Norman: University of Oklahoma Press). Hill, J. N. and Evans, R. K. (1972). “A Model for Classification and Typology.” In: Clarke, D. (ed), Models in Archaeology (London: Methuen), 231–274. Hodder, I. (1982). Symbols in Action (Cambridge: Cambridge University Press). Hodder, I. (1991). Reading the Past. Current Approaches to Interpretation in Archaeology (Cambridge: Cambridge University Press). Hurcombe, L. (2008). “Organics from Inorganics: Using Experimental Archaeology as a Research Tool for Studying Perishable Material Culture.” World Archaeology 40, 83–115. Kampel, M. and Sablatnig, R. (2007). “Rule Based System for Archaeological Pottery Classification.” Source Pattern Recognition Letters 28(6): 740–747. Kempton, W. (1981). The Folk Classification of Ceramics: A Study of Cognitive Prototypes (New York: Academic Press). Kossina, G. (1926). Ursprung und Verbreitung der Germanen in vor-und rühgeschichtlicher Zeit (Berlín-Lichterfelde: Irminsul/Schriften und Blätter für deutsche Art und Kunst). Krieger, A. D. (1944). “The Typological Concept.” American Antiquity 9(3): 271–288. Latour, B. (2008). Reensamblar lo social: Una introducción a la Teoría del Actor-red (Buenos Aires: Manantial). Lemonnier, P. (1986). “The Study of Material Culture Today: Towards an Anthropology of Technical Systems.” Journal of Anthropological Research 5: 147–186. Lemonnier, P. (ed) (1993). Technological Choices: Transformation in Material Cultures since the Neolithic (London: Routledge). Longacre, W. (1999). “Standardization and Specialization: What’s the Link?” In: Skibo, J. and Feinman, G. (eds), Pottery and People (Salt Lake City: University of Utah Press), 44–58. Mahias, M. C. (1993). “Pottery Techniques in India; Technical Variants and Social Choice.” In: Lemonnier, P. (ed), Technological Choices: Transformations in Material Cultures since the Neolithic (London: Routledge), 157–180. Manby, T. G. (1995). “Skeuomorphism: Some Reflections of Leather, Wood and Basketry in Early Bronze Age Pottery.” In: Kinnes, I. and Varndell, G. (eds), Unbaked Urns of Rudely Shape (Oxford: Oxbow), 81–88. Orton, C., Tyers, P., and Vince, A. (1993). Pottery in Archaeology (Cambridge: Cambridge University Press). Prieto, M. P. (1999). “Caracterización del estilo cerámico de la Edad del Bronce en Galicia: Cerámica campaniforme y cerámica no decorada.” Complutum 10: 71–90. Read, D. W. (1989). “Intuitive Typology and Automatic Classification: Divergence or Full Circle?” Journal of Anthropological Archaeology 8: 158–188. Read, D. W. (2007). Artifact Classification: A Conceptual and Methodological Approach (Walnut Creek, CA: Left Coast Press). Rice, P. M. (1987). Pottery Analysis: A Sourcebook (Chicago, IL: University of Chicago Press).
Formal analysis and typological classification 199 Rice, P. M. (1990). “Functions and Uses of Archaeological Ceramics.” In: Kingery, W. D. (ed), The Changing Roles of Ceramics in Society (Westerville, OH: American Ceramic Society), 1–10. Rouse, I. (1960). “The Classification of Artifacts in Archaeology.” American Antiquity 25(3): 313–323. Sackett, J. R. (1977). “The Meaning of Style in Archaeology: A General Model.” American Antiquity 42: 369–380. Samaniego Bordiu, B. (2013). “El esquematismo en el arte prehistórico de la Península Ibérica.” Electronic PhD thesis, Universidad Complutense de Madrid, Madrid. Schiffer, M. B. and Skibo, J. M. (1987). “Theory and Experiment in the Study of Technological Change.” Current Anthropology 28: 595–622. Sheppard, A. (1971). Ceramics for the Archaeologist (Washington, D. C.: Carnegie Institution of Washington). Sinopoli, C. (1991). Approaches to Archaeological Ceramics (New York: Plenum Press). Smith, M. F. (1985). “Towards an Economic Interpretation of Ceramics: Relating Vessel Size and Shape to Use.” In: Nelson, B. A. (ed), Decoding Prehistoric Ceramics (Carbondale and Edwardsville, IL: Southern Illinois University Press), 254–309. Spaulding, A. C. (1953). “Statistical Techniques for the Discovery of Artifact Types.” American Antiquity 18(4): 305–313. Tilley, C. (1999). Metaphor and Material Culture (Oxford: Blackwell). Trigger, B. (1989). A History of Archaeological Thought (Cambridge: Cambridge University Press). Van Dommelen, P. (2006). “Colonial Matters: Material Culture and Postcolonial Theory in Colonial Situations.” In: Tilley, C., Keane, W., Kuechler, S., Rowlands, M., and Spyer, P. (eds), Handbook of Material Culture (London: Sage), 104–124. Vidal, A. (2011). “Para aprender no hay edad: irregularidades frecuentes en la cerámica realizada por aprendices adultos.” In: Morgado, A., Baena, J., and García, D. (eds), La investigación experimental aplicada a la Arqueología (Ronda: Universidad de Granada, Universidad Complutense de Madrid, Asociación Experimenta, Ronda), 393–399. Weigand, P. C. (1969). Modern Huichol Ceramics, Mesoamerican Studies (Carbondale, IL: University Museum, Southern Illinois University). Whallon, R. (1972). “A New Approach to Pottery Typology.” American Antiquity 37(1): 13–33. Whallon, R. and Brown, J. (eds) (1982). Essays on Archaeological Typology (Kampsville, IL: Center for American Archaeology Press). Whittaker, J. C., Caulkins, D., and Kamp, K. A. (1998). “Evaluating Consistency in Typology and Classification.” Journal of Archaeological Method and Theory 5(2): 129–164. Woods, A. J. (1986). “Form and Function: Some Observations of the Cooking Pot in Antiquity.” In: Kingery, W. D. (ed), Ceramics and Civilization (Westerville, OH: American Ceramic Society), 157–172.
Chapter 13
Fabric Des c ri p t i on of Archaeol o g i c a l Ceram i c s Ian K. Whitbread Introduction Ceramics found in the archaeological record were produced to serve many functions ranging from the transportation and storage of goods, food preparation, cooking, and consumption (Orton and Hughes, 2013: 247) to supporting activities involving heat, such as hearths and crucibles, representative art, such as figurines and sculptures, and building materials (Mills, 2013). The diversity of these functions and the wide geographical distribution of ceramic production are reflected in the materials of which ceramics are made—their fabrics. Ceramic fabrics can be described in terms of their compositional and structural properties; more specifically, the arrangement, size, shape, frequency, and composition of the material constituents of the ceramic (Whitbread, 1995: 368). The term “fabric” is also used in reference to groups of ceramics that are characterized by having specific material properties in common (Tomber and Dore, 1998). Ceramic fabric descriptions aim to record, so far as possible, not only significant geologi cal properties of the raw materials but also the potters’ technological choices and actions (the chaîne opératoire; Sillar and Tite, 2000; Whitbread, 2001), such as preparation of the clay body, vessel construction, and firing. Ceramic fabrics are synthetic in origin (Rice, 1987: 3). Their constituents reflect the geological characteristics of the regions from which the raw materials were obtained. However, selection and processing of these raw materials are dictated by the agency of potters with the aim of preparing clay bodies possessing physical properties appropriate for the production methods and intended functions of specific end products. Preparing a clay body may have required only the simple combination of water with a natural clay, but potters often mixed different clays and matrix materials, removed inclusions by sieving or sedimentation, that is settling in water, or added inclusions through the process of tempering to optimize the physicomechanical properties of the vessel. There are many reasons why a potter might make such adjustments. Often raw materials are unsuitable for ceramic production in their original state (Nicklin, 1979). Furthermore,
Fabric Description of Archaeological Ceramics 201 particular physical properties can enhance the forming and drying process, or the functional performance of the resultant ceramic. For example, thick-walled vessels dry more quickly and evenly prior to firing if inclusions are present that “open” the clay (Gibson and Woods, 1990: 206). Ceramics that are repeatedly heated during use, such as cooking pots, benefit from the presence of inclusions to increase mechanical toughness and thermal shock resistance (Tite et al., 2001). However, clay body preparation also depends on the ways in which potters learn to engage with their materials and, in part, may reflect personal or cultural preferences (Gosselain, 2000). Construction methods such as pinching, coiling, slab construction, hammer and anvil, molding, and wheel throwing, used individually or in combination, may leave their imprint in the micromorphology of ceramic fabrics (Rye, 1981; Woods, 1985; Courty and Roux, 1995; Whitbread, 1996; Roux and Courty, 1998). The most distinctive fabric characteristics, however, are produced by firing. The temperatures, oxidizing and reducing atmospheres, and duration of firing affect the appearance and physicomechanical characteristics of ceramic vessels. In most archaeological examples, the matrix component was converted to terracotta or earthenware ceramic by firing it in temperatures up to 900–1000°C using a bonfire or kiln (Gosselain, 1992). Under these conditions the crystalline structure of clay minerals breaks down, as a result of the loss of structural water, and is converted to ceramic by the sintering and vitrification that subsequently develops (Shepard, 1956: 19–31, 49–94; Cardew, 1969: 61–68; Rhodes, 1973: 64–7 1; Rye, 1981: 29–40, 96–110; Rice, 1987: 93–94; Gibson and Woods, 1990: 24–56).
Description of Ceramic Fabrics Fabric descriptions need, so far as possible, to record the critical properties of ceramic materials that fulfil the requirements of three fundamental and related areas of archaeological interpretation (Whitbread, 1995: 366–378): characterization, technology, and provenance determination. Characterization is the process of defining groups or classes of similar ceramic fabrics that, in combination with additional evidence such as shape, decoration, function, or find context, can be used to examine patterns of ceramic production, distribution, and use. Technological studies investigate potting practices specific to one or more social groups, which may reflect sociocultural identities, boundaries, and the transmission of technical knowledge (Gosselain, 2000). Finally, composition of the raw materials can be used to study exchange and trade networks by determining the likely geological provenance of ceramics that have passed into areas of different regional geology (e.g. Fitzpatrick et al., 2003). At the most basic level of description archaeological ceramics are frequently divided into two broad types of fabric, coarse and fine. There is no universally recognized distinction between these types. In some situations “coarse ware” may simply refer to undecorated utili tarian ceramics in contrast to “fine ware,” which is often decorated, used for serving and display. From a materials perspective, however, coarse and fine fabrics are normally distinguished by whether they contain inclusions visible to the naked eye (coarse fabrics) or not (fine fabrics). Both coarse and fine fabrics can be described in hand specimen or thin section using polarizing microscopy. Sometimes more advanced analytical methods are
202 IAN K. Whitbread required to address questions of characterization, technology, and provenance, and ceramic fabric description is used in combination with these methods (Day et al., 1999; Stoltman and Mainfort, 2002). Generally speaking, however, thin-section petrography is better suited for analyzing coarse fabrics and their inclusions and bulk chemical analysis is usually more effective for fine fabrics. A fabric description is the analyst’s record of observations. So far as possible, it needs to be kept independent of interpretations concerning technology and provenance. This approach allows alternative interpretations to be proposed, tested, and amended without affecting data from the original fabric analysis. The process of interpretation draws upon evidence from fabric descriptions and characterization, but also additional sources such as regional geology, production methods, and geochemistry. This wider analytical process of description and interpretation constitutes ceramic petrology (Whitbread, 1995: 28–29). As fabric descriptions using polarizing microscopy are usually necessary to achieve adequate resolution for such interpretations, ceramic petrology commonly refers to such studies. The greatest challenge in describing ceramic fabrics lies in recording complex visual and physical information in ways that are meaningful and reproducible for archaeologists unfamiliar with the fabrics in question. A second major challenge is in defining recognizable, coherent, and sustainable fabric groups across different ceramic assemblages. Such assemblages may contain overlapping ranges of material variation and are commonly studied by pottery specialists with diverse research interests. Macro-and microphotographs are invaluable aids to fabric description but they also illustrate the difficulties outlined above. They are severely limited in value without appropriate interpretation. Their restricted field of view rarely conveys the full range of fabric variation encountered within a single sample, and one photograph cannot illustrate the variation within a group of samples. Finally, they are no substitute for the physical and optical tests carried out in hand specimen study or under a polarizing microscope. For this reason, comprehensive written records are essential. Training in mineralogy and petrology is also necessary to gain proficiency in polarizing microscopy and in determining the optical properties of minerals and rocks in thin section (Stoltman, 1989: 147; Nesse, 1991). Descriptions based on optical examination by hand specimen or polarizing microscopy both rely heavily on the observational skills of the analyst. As such they incorporate significant elements of interpretation and inevitably reflect the analyst’s range of experience in studying geological materials and archaeological ceramics.
Hand Specimen Studies Ceramics break with a brittle fracture to leave a fresh surface that can be readily studied in hand specimen. This analytical method lacks the accuracy and precision of laboratory-based techniques, such as chemical analysis or thin-section microscopy. However, it is not constrained by the limitations of physical damage and costs that accompany laboratory sample preparation. For this reason, hand specimen examination continues to be the most effective method for characterizing fabrics across ceramic assemblages from excavations and surveys (Moody et al., 2003; García-Hera, 2000; Pentedeka et al., 2010). It is often the first step in selecting samples for more advanced analytical methods. More importantly, where fabric
Fabric Description of Archaeological Ceramics 203 groups defined by thin section or chemical analysis possess characteristic features in hand specimen then it can be used to extend laboratory characterization from a set of samples to the ceramic assemblage as a whole (Gauss and Kiriatzi, 2011).
Surface Preparation Occasionally sherds of broken pottery display suitable breaks, but where these do not occur it is necessary to create a fresh break. The most effective tool is a pair of spring-loaded pliers that open wide enough to accommodate the thickness of the vessel wall. These can produce a clean break with little waste and minimal damage to the sherd. Pincers are effective with very soft fabrics, but can crush or crumble harder sherds. A small hammer and chisel is best used for producing fresh breaks on large objects such as brick, tile, or thick-walled storage jars. Using these methods to remove samples for thin section or chemical analysis has the advantage of leaving a fresh break for hand specimen examination.
Hand Specimen Description There are many methods of describing ceramic fabrics in hand specimen, reflecting the different aims and experience of pottery specialists and the materials they study. One widely used method was proposed by Peacock (1977), based on the study of Roman ceramics at Carthage in North Africa. It describes the key components of ceramic fabrics in hand speci men, but aims to be simple enough for rapid sorting of large ceramic assemblages (Tomber and Dore, 1998: 5–9). The method also forms the basis of fabric descriptions proposed by the UK Prehistoric Ceramics Research Group (2010). Color is a particularly distinctive property of ceramic materials in hand specimen. It does not feature in the definition of fabric given above because different firing conditions can produce a wide range of colors within and between ceramic objects composed of the same materials (Shepard 1956: 102–107; Beck, 2006). For example, there may be a different colored core visible in a fresh break, or the external surface of a vessel may have a different color from its interior (Velde and Druc, 1999: 122–126). The most frequent color is normally noted, along with the range of variation observed on several sherds. Where variation appears to be systematic it may reflect consistencies in the firing process and be a useful guide to the technological control maintained by potters (Frankel, 1994), but relatively minor or erratic variations should not be over-emphasized. The most widely used method of describing color is based on comparisons with colored chips in the Munsell® Soil-Color Chart (Munsell, 2009; Shepard, 1956: 107–113; Rice, 1987: 339–343). Each coded color chip provides a common reference for other researchers to consult and interpret in their own terms. The value of these charts lies in aiding precision (repeatability) rather than accuracy (closeness to the “true” color); however, both can be compromised as color perception varies between people and lighting conditions (Frankel, 1980; Gerharz et al., 1988). Groups of tablets are referred to by a color name which provides a structured reference with wider tolerance than individual chip codes. For this reason, Munsell® color names should be used rather than substituted with seemingly more “accurate” personal color descriptions.
204 IAN K. Whitbread The hardness of a ceramic fabric depends upon its composition and degree of vitrification and/or firing temperature. The Peacock system provides a rapid hardness assessment in terms of soft (scratched with a fingernail), hard (scratched with a steel knife), and very hard (not scratched with a steel knife). These divisions are simple and quick to apply, and can be equated to Mohs’ scale of mineral hardness. This is composed of ten minerals, from talc (1) to diamond (10), each harder than the next. Hardness points based on Mohs’ mineral scale can also be used to test pottery, but caution is necessary when interpreting results from any scratch tests as they may reflect bonding of the material rather than its hardness (Shepard, 1956: 113–117; Rice, 1987: 355–356). Mohs’ scale is an ordinal scale. Its numerical values should therefore be treated as labels of rank and not used in mathematical calculations (Curewitz, 2004). Handling ceramics is a tactile process and, while subjective assessments may not be reliably reproduced, the system recognizes variation in the “feel” (harsh, rough, smooth, soapy, or powdery) of ceramic fabrics, which can be a distinctive property when processing large quantities of material. The surface appearance of a fresh break is described by its fracture, ranging from conchoidal to smooth, hackly (angular facets), or laminated; although Shepard (1956: 137) noted that this property catches the eye but tends only to be significant across broad categories of fine and coarse fabrics. Peacock described the frequency of inclusions as: sparse, moderate, common, or abundant. These terms are subjective in the absence of independent reference points to maintain consistency in their application, such as the percentage boundaries outlined below for thin- section analysis. The mode (most frequent size visible) and the maximum grain diameter are usually adequate measures of inclusion size. Equally significant is the degree of grain-size sorting (well or ill) and rounding/angularity as assessed using comparator charts derived from sedimentary petrography (Boggs, 2009: figures 2.3 and 2.12). Peacock (1977: table 3) provides an inclusion identification table. With experience it is possible to identify some inclusions accurately in hand specimen but it is also very easy to make mistakes. Difficulties arise from the small size of most grains and their abraded surfaces. Inclusion color can also be misleading. For example, both quartzite and limestone can be white, requiring use of a steel knife (which scratches limestone but not quartz) or dilute hydrochloric acid (which fizzes if the material is calcareous) (Moody et al., 2003) to discriminate between them. If in doubt, it is more reliable to describe the appearance of inclusions first and then suggest possible compositions. Inclusion compositions can be verified by thin- section microscopy.
Hand Specimen Description Example This section illustrates the hand specimen fabric description for the Early Bronze Age sherd described in thin section below (for illustration of the sherd in hand specimen see Plate 3). Hand specimen fabric: Red, well sorted fine sand Number of samples: 1 Sample number: 133 Color: 2.5YR 5/8 red. Hardness: soft.
Fabric Description of Archaeological Ceramics 205 Feel: rough. Fracture: hackly Inclusion Frequency: c.20%. Sorting: well sorted. Maximum size: c.1 mm. Average size: c.0.5 mm. Rounding: Subrounded–subangular. Inclusion composition: predominant: pale yellowish gray rock fragments, rarely with foliation, phyllite(?); very few: white limestone (softer than steel knife blade), very dark gray quartz/feldspar(?) (harder than steel knife blade); rare: vughs
Thin-Section Studies Analyzing thin sections of ceramic fabrics under the polarizing microscope provides greater resolution than hand specimen study. Microstructures and the compositions of inclusions can, for the most part, be more accurately identified down to a grain size of 0.03 mm, which is the standard thickness of a thin section (Reedy, 2008: 109–231; Quinn, 2013: 39–68). Thin sections are studied by observing the optical properties of minerals, rock fragments, and fired clay with respect to the path of polarized light which is transmitted through the sample. Two types of polarized light are employed in combination. Plane polarized light (PPL) uses one polarizing filter, located below the microscope stage, and crossed polars (XPL) is applied when a second polarizing filter is inserted into the light path, above the sample, at 90° to the lower polarizer. In this latter case, all light from the lower polarizer is blocked by the upper polarizer and the image appears to be black if materials in the light path have no crystalline structure (e.g. glass) or where minerals are optically isotropic, having the same optical properties in all directions. However, most minerals are optically anisotropic, having different optical properties in different directions, and refract the polarized light passing through them such that they display a range of interference colors (Nesse, 1991).
Sample Preparation Samples need to be removed as chips that are large enough to be ground to a flat surface between 1 × 2 cm and 2 × 5 cm in area, depending on the dimensions of the ceramic objects and the size of the microscope slides. A larger surface area is more representative of a fabric, especially if it contains large or sparse inclusions, and leads to a more reliable analysis. Thin sections are usually prepared as vertical cross-sections through the wall of a ceramic vessel (Whitbread, 1996). Friable fabrics need to be impregnated with epoxy resin before grinding (Quinn, 2013: 23–33). Once mounted on microscope slides the samples are ground to a thickness of 0.03 mm, at which point most rock-forming minerals are translucent. Considerable skill is required to prepare thin sections that are evenly ground to the correct thickness. A sample should be large enough to allow at least two thin sections to be produced in the event that there are problems in sample preparation.
206 IAN K. Whitbread
Qualitative Thin Section Description As fired sedimentary material, ceramic fabrics have been likened to metamorphosed sedimentary rocks (Peacock, 1977), but there are significant differences. Sedimentary petrography is most effective when studying materials dominated by sand-sized (0.0625–2 mm) and coarser grains, whereas ceramics are mainly composed of fired clay-size particles or matrix, often comprising 70% or more of the sample. While petrographic methods inherited from geology are essential for describing igneous, metamorphic, and sedimentary rock and mineral inclusions (Williams et al., 1982), additional methods are required to address features arising from the technological processes of ceramic production. For example, micromorphology of the matrix material can provide important technological information on preferred orientations arising from vessel-forming techniques and on the degree of firing used to produce the ceramic. Ceramic fabrics can also contain anthropogenic inclusions not normally encountered in sedimentary deposits, such as grog (grains of pottery found within the ceramic, cf. Whitbread,1986; Cuomo Di Caprio and Vaughan, 1993), industrial waste such as slag from metal processing (Knowles and Quinlan, 1995), and agricultural waste such as chaff (Quinn, 2013: 161). Quantitative methods can be used to characterize ceramic fabrics (see below), but thin- section descriptions are usually qualitative, focusing on the observation of compositional and structural properties with supporting measurements, such as inclusion frequencies and grain-size ranges. Early studies often focused on inclusion compositions owing to their criti cal role in provenance determination and often incorporated interpretation within descriptions, which meant that technological assumptions were not always tested. For example, inclusions might be referred to as “temper” without an explicit justification for their being intentionally added by the potter. In contrast, the presence of crushed schist temper in Hohokam buff wares was argued to be temper by Miksa (2001) based on a range of explicit properties comprising grain size distribution, rock homogeneity, angularity, fractured grains, finely disseminated micas from the crushed schist, and occasional grains of volcanic rock and quartzite from implements used to do the crushing. Greater sensitivity to recording technological properties has been introduced by applying a more comprehensive and systematic method of description that allows analysts to record all features observed in a ceramic thin section. By enabling the characteristics and properties of ceramic fabrics to be isolated and recorded, their significance, in terms of technological and provenance determination, can be more effectively established rather than merely assumed. To this end, elements of sedimentary petrography and soil micromorphology, using the system proposed by Kemp (1985: 15ff.; derived from Bullock et al., 1985), are combined and modified for ceramic fabrics (Whitbread, 1995: 379ff.; see also Josephs, 2005). This approach adopts existing terminologies where appropriate but remains open to adaptation to accommodate new developments in ceramic petrology. It is most efficient to sort samples of ceramic fabrics into groups and/or classes with similar properties before describing them (Quinn, 2013: 73–79), rather than combining the descriptions of individual samples as a basis for constructing a classification. Groups may be based on broad geological characteristics, such as major rock or mineral components, while classes may constitute subdivisions based on secondary compositions or technological variation, such as grain size, sorting, or frequency. The classification is tested by attempting to reassign samples to alternative or new groups and classes based on further examination
Fabric Description of Archaeological Ceramics 207 of their compositional and micromorphological properties. The procedure ends when the classification achieves its greatest stability, and reassigning samples no longer improves the definition of groups or classes within the dataset. Fabric descriptions therefore usually refer to collections of samples rather than individual pieces. Various degrees of diversity are often observed between the fabric characteristics of individual samples within a fabric class. This is an important property, expressed in terms of the homogeneity of the class, which can arise from a number of factors. It may reflect materials variation inherent within the production of certain ceramics or possibly limitations in the sampling program for fabric analysis. Acknowledging this variation by defining the boundaries of class membership is an effective way to establish the properties that must be met for attributing future samples to a group or class, rather than relying on the description of a “typical” sample without any indication as to the range of variation that may be present between related fabrics. The frequencies of different components within a fabric can be estimated using point- counting or digital image analysis (see below). More commonly, however, estimates made using comparator charts are favored as a more rapid method of evaluating proportions based on illustrations of different percentage areas of the field of view (Matthew et al., 1991). These estimates have limited accuracy and precision, with a general tendency towards the over estimation of proportions, especially where fabric properties vary widely across the thin section or where inclusions are poorly sorted. One way of reducing such inherent errors is to allow a degree of tolerance by referring to frequencies in ranges. Frequency ranges derived from soil micromorphology are: predominant (>70%), dominant (70–50%), frequent (50– 30%), common (30–15%), few (15–5%), very few (5–2%), rare (2–0.5%), and very rare ( lamellae of albite(stripes in XP)
orthopyroxenes may show fine twinning or exsolution
pale green, peochroic to pink
first order Extinction is white to gray, parallel to the low order cleavage interference, up to first order yellow
andalusite, clinopyroxene
pleochroic halos around zircon inclusions, visible as burn marks in PP
colorless, altered giving grainy or grayish appearance in PP. Concentric compositional zoning possible (both visible in PP and/or Xp)
birefrigence is low, only white to gray low order interference colours, occasionally first order yellow
quartz, K-feldspar, Cordierite
Intergrowth of Na-rich feldspar with K-feldspar (microcline or orthoclase) = perthite. Alteration to a fine grained micaceous material, partly or entirely.
Near 90° cleavage in basal sections. Longitudinal sections show one cleavage.
large variety two cleavages possible but but often not also untwinned, well developed XP: diagnostic polysynthetic twins; stripes or simple twinning
n/a
n/a
Quartz
common mineral in sedimentary, metamorphic and igneous rocks
SiO2
low relief
anhedral, rarely n/a euhedral.
never cleavage, possible fractures
colorless
first order white wavy extinction Plagioclase, to gray, rare first (undulatory) Alkali feldspar, order yellow apatite, cordierite, beryl & scapolite
Rutile
common accessory TiO2 mineral in intermediate to mafic igneous rocks and in many metamorphic rocks
very high
possibly associated with biotite
n/a
n/a
usually red, red-brown, yellowish brown. possibly pleochroic. yellowish to reddish brown
extreme n/a birefringence, Interference colors may not show due to the strong color of the mineral. When present, they are pastels of very high order
Sanidine
widespread and common in a wide variety of volcanic, igneous, and metamorphic rocks, to a lesser extent in some immature sedimentary rocks (high temperature).
n/a
euhedral
simple twins
n/a
colorless, low generally clear birefringence; white to gray
Anorthite: CaAl2Si2O8; Albite: NaAlSi3O8; Orthoclase KAlSi3O8
n/a
subgrains by low grade methamorphism. Coesite as high pressure polymorph of quartz. Quartz does not alter
Hematite has n/a a deeper red color and a more irregular, platey, habit.
n/a
exsolved -> lamellae of albite (stripes in XP)
(continued)
Table 15.1 Continued ID
Occurrence
Composition
Relief (PP)
Appearance Twinning and habit
Cleavage
Color (PP)
Interference Extinction Colors (XP) Angle
Similar Minerals
Serpentine
altered ultrabasic rocks, especially from ophiolites
Mg3Si2O5(OH)4
low to moderate
irregular
occasional
yes ; chrysotile fibrous, others one perfect
pleochroic; colourless to pale green, often oxidizes to red in pottery
birefringence parallell low, mostly first order whites and grays
fibrous lower birefringence amphiboles; than micas, chlorite patchy and higher RI mixed with other minerals
Sillimanite
metamorphic mineral Al2SiO5 found in high-grade aluminous schists and gneisses. Polymorph of andalusite and kyanite
high relief
needle-like, fibrous, or bladed habit
n/a
n/a
clear
upper second order
n/a
n/a
n/a
Titanite
accessory mineral CaTiSiO5 in igneous and metamorphic rocks, less commonly found in sedimentary rocks
very high relief
wedge, n/a diamond, or distorted hexagonal shape, euhedral crystals
n/a
nearly colorless, sometimes brownish or gray due to high refractive index
very high n/a birefringence -; interference colors
n/a
n/a
Spinel
common isotropic minerals in aluminous Si-poor, metapelitic rocks
combination of endmembers: spinel (MgAl2O4), hercynite (FeAl2O4); gahnite (ZnAl2O4); galaxite (MnAl2O4)
n/a
n/a
n/a
commonly anhedral
highly variable isotropic
n/a
n/a
Staurolite
metamorphic mineral, common in medium- grade micaceous schists and gneisses
Fe2Al9Si4O23(OH)
n/a
often also contains many quartz inclusions
common twins
n/a
yellow, pleochroic
up to 1st order n/a red but often masked by the strong colour of the mineral
tourmaline, epidote
n/a
Talc
marbles or as an alteration product of mafic and ultra-mafic rocks
n/a
n/a
similar to white n/a micas
single
clear
high birifrigence, 3rd order interference
Difficult to n/a distinguish from muscovite and white micas
n/a
n/a
Remarks
Tourmaline
granites, higher grade Na(Mg,Fe,Mn,Li,Al) metamorphic rocks, and 3Al6[Si6O18](BO3) often as detrital grains 3(OH,F)4
moderate
Tremolite
marbles, talc, and forsterite
Ca2Mg5Si8O22(OH)2; n/a some Fe may substitute for Mg, forming solid solutions towards actinolite
Wollastonite
high-grade marbles, associated with diopside and garnet
CaSiO3
n/a
rare
very poor
Fe-rich blue n/a to yellow, Mg-rich yellow-pale yellow, Li-rich colourless- other
bladed, columnar, needle-like crystals
n/a
one cleavage longitudinal and two when viewed in end section (60°–120°)
n/a
n/a
moderate to elongate and high relief bladed
n/a
n/a
n/a
clear to pale green, slightly pleochroic more pronounced with greater Fe content
upper 1st to n/a 2nd lower order interference colours
n/a
n/a
colorless
low,1st order yellow
Ca-Mg Silicates n/a
n/a
252 Dennis Braekmans and Patrick Degryse found in archaeological ceramics, is mainly composed of fine-grained (600–650°C). Therefore, the ceramic body can be regarded as an artificial rock (Maggetti, 1986) formed through anthropogenic pyrometamorphism (Grapes, 2011). As with natural rocks, its composition can be studied by polarized light optical microscopy (OM) and X-ray diffraction (XRD) but, for detailed information, methods such as electron microprobe analysis (EMPA), scanning electron microscopy (SEM) (Freestone and Hunt, Chapter 31 this volume), Fourier transform infrared spectroscopy (FT-IR) (Shoval, Chapter 28 this volume), cathodoluminiscence (CL), or Mössbauer spectroscopy are required. Optical microscopy (Braekmans and Degryse, Chapter 15 this volume) is the basic method to study ceramics in thin sections; that is, their composition, the relationship between minerals, the distribution and grain size of inclusions, and the texture and degree of vitrification. However, the optical properties are often insufficient for a clear mineralogical determination, in particular when the minerals are very fine grained. Sometimes optical determination of minerals is problematic because many of the minerals present are covered or penetrated by iron oxide during firing, masking important optical features. Currently available chemi cal, structural and XRD databases are insufficient for identifying compounds and phases newly formed during firing, which have no natural equivalents, and/or minerals thermally modified beyond their usual composition. XRD complements OM by utilizing crystal structure (Heimann, Chapter 19 this volume). However, it does not allow the study of mutual relationships between minerals. Most of these issues can be successfully addressed by EMPA, an essential analytical method in all branches of geosciences. Since the advent of commercially available electron microprobes more than fifty years ago (Long, 1995), they have become a routine tool for detailed compositional characterization of materials. This is particularly important in
Electron Microprobe Analysis (EMPA) 289 the case of minerals forming solid solutions (e.g. feldspar) and for minerals with manifold substitutions (e.g. pyroxene and amphibole). Maggetti (1979, 1982) has emphasized the importance of mineralogical information in answering both provenance and technological questions for archaeometric studies. Electron microprobe is efficient in analyzing the matrix, the clasts, and the firing phases in archaeoceramics. In the following, we will review the physical principles of the electron microprobe and will present basic information about the equipment, samples, and analytical conditions. The potential outcomes of electron microprobe analyses for the study of archaeoceramics will also be addressed. For more details, the reader is referred to relevant literature; for example, Freestone (1982), Jansen and Slaughter (1982), Loretto (1984), Long (1995), Reed (1995, 2005), Lifshin and Gauvin (2001), and Goldstein et al. (2003).
The Physical Principles of Electron Microprobe Analysis Electron microprobe provides insight on the composition of mineral phases using a narrow electron beam to stimulate the emission of X-rays. Accelerated electrons hitting a solid target interact with the material in several ways and produce various signals (Figure 17.1a), among which the most important are the backscattered electrons (BSE), the secondary electrons (SE), and the X-rays (Reed, 2005). The latter include the characteristic X-rays and a continu ous X-ray spectrum. The scattering of the beam electrons inside the sample causes deflections and reflections in all directions. In the case of multiple deflections at a small angle or in the case of high-angle reflections, a fraction of the incident electrons exit the sample, giving rise to the backscattered electrons. The backscattered electrons come from a small volume, i.e. ~50–100 μm3, inside the sample. Their energy distribution depends on the atomic number Z of the stimulated atom (Reed, 2005). The secondary electrons are ejected upon the reaction of beam electrons with atoms near the surface of the sample or emerge from the BSE when they leave the surface. The SE have a lower energy than the BSE and provide data about the surface. However, as they are partly produced by BSE, the SE also convey some information about the chemistry. The continuous X-ray spectrum is generated when an incident electron interacts with an atomic nucleus in the sample. It may mask the characteristic lines of elements present in low amounts. The nucleus of an atom is surrounded by electrons in several levels (called shells), labeled with increasing distance from the nucleus as K, L, M, N, O, P and Q. An electron of a compound within the sample hit by an electron from the incoming beam is expelled and leaves a vacancy. Then, an electron from an outer shell fills in this vacancy and emits a characteristic X-ray radiation. If, for example, an electron fills in a vacancy left on K-shell, the according radiation is called K radiation. The energy and the wavelength of the X-ray radiation vary with the atomic number Z (Reed, 1995). The characteristic X-rays enable quantitative determination of the chemical composition of a small volume of the sample, whereas the BSE and SE are used for imaging (Figure 17.1b–c).
290 Corina Ionescu and Volker Hoeck (a)
Kfs
(b)
focused incident electron beam
Po
Bt
Si backscattered electrons secondary electrons
secondary electrons
Si-Ca
Ilt
Po
Po
Qz
characteristic X-rays
Po
Qz
Ms
Po Po
(c) sample surface sample
electron interaction volume within the sample
50 µm
Figure 17.1 (a) Origin of BSE, SE, and characteristic X-rays emitted by the interaction between the focused electron beam and sample. The electron interaction volume within the sample and the thickness of the electron beam are not at scale. Simplified drawing, based on a chart from ; (b) BSE image of a ceramic sample from Ibida (Roman–Byzantine period) containing quartz (Qz), muscovite (Ms) and (chloritized) biotite (Bt) embedded in a predominantly illitic (Ilt) matrix with some pores (Po). In the center there is a siliceous test (Si), with a Ca-rich siliceous core (Si-Ca); (c) SE image of the same area as in (b), showing the high topography of the clasts and the test, and the lower topography of the illitic parts of the matrix. Most of the small black parts from (b) are proved here to be areas with very low topography and not pores. Mineral abbreviations according to Whitney and Evans (2010) are used throughout this chapter.
Instrumentation The electrons are generated by heating a filament commonly made from La-, Ce-hexaborate, or tungsten. The filament acts as a cathode held at a negative voltage (15–30 kV) towards the anode. The electrons accelerate through an aperture in the anode and form the electron beam. This is focused by two or three electromagnetic lenses to a diameter of a few microme ters. The electron beam is visible when it hits luminescent material, such as scheelite or periclase. For monitoring the beam current, microprobes have a swing-in Faraday cage made of a conducting material. From the signals emerging from the electron beam–specimen interaction, the BSE and the SE are collected by special detectors. The electron microprobe is commonly equipped with four or five wavelength dispersive (WD) spectrometer, used for
Electron Microprobe Analysis (EMPA) 291 the quantitative analysis, and one energy-dispersive (ED) X-ray spectrometer, involved in the qualitative analysis. Each WD spectrometer contains a crystal of known lattice spacing d. Since one crystal with a certain d value can detect only a limited range of characteristic wavelengths, several crystals with different d values are necessary for analyzing the whole spectrum. The most used crystals are: TAP (thallium acid phthalate), PET (pentaerythritol), LIF (lithium fluo ride), ADP (ammonium dihydrogen phosphate), muscovite, quartz, and KAP (potassium acid phthalate). All X-ray radiations are focused by the spectrometers and subsequently collected by gas- filled counters. Ionization of the gas by interacting with the incoming X-ray radiations generates electric pulses, which are counted in order to measure the X-ray intensity. The ED spectrometer has a completely different design than the WD spectrometer. It is built up by a solid-state detector, a pulse height analyzer, and a multiple channel analyzer. The detector is made of a semiconductor crystal, mainly Si(Li) or high purity Ge. The detectors are equipped either with a thin polymer-based or Be window. The first can also detect elements below the atomic number of Na (11). The ED spectrometer shows all characteristic X-rays simultaneously for the whole range of energies of interest. The X-rays striking the detector generate an electric charge which is amplified by a transistor. The time during which the system needs to process the pulses and during which no other pulses are recorded is called “dead time.” Dead time increases with increasing input count rates. The multichannel analyzer displays the data as counts versus energy diagram (insert in Figure 17.2a). Apart from the high overall count rate, the advantage of ED spectrometry compared with WD spectrometry is the display of the whole range of energies produced by all elements occurring at the measured sample point. On the other hand, WD spectrometers have better resolution and a higher peak/background ratio. The WD spectrometers are also more precise and capable of detecting lower concentrations than the ED spectrometer. Commonly, the beam has 10–40 nA current and a diameter of ≤5 μm. The accelerating voltage may vary from 1 to 30 kV. A lower beam current is necessary for obtaining high- resolution imaging, whereas a higher beam current is used for detecting elements in low concentrations. The detection limits of the instrument depend upon equipment performance and specimen matrix, but are typically