Ceramics of the Indigenous Cultures of South America: Studies of Production and Exchange through Compositional Analysis [Illustrated] 0826360289, 9780826360281

Citation preview

C E R A M IC S OF T H E I N DIGE NOUS C U LT U R E S OF S OU T H A M E R IC A

Ceramics of the Indigenous Cultures of South America Studies of Production and Exchange through Compositional Analysis

Edited by

M i c h a e l D. G l a s c o c k , H e c t o r N e f f , a n d K e v i n J . Vau g h n

U n i v er sit y of N ew M e x ico Pr e ss · A l buqu erqu e

© 2019 by the University of New Mexico Press All rights reserved. Published 2019 Printed in the United States of America Library of Congress ­Cataloging-­in-­Publication Data Names: Glascock, Michael, editor. | Neff, Hector, editor. | Vaughn, Kevin J., editor. Title: Ceramics of the indigenous cultures of South America: studies of production and exchange through compositional analysis / edited by Michael D. Glascock, Hector Neff, and Kevin J. Vaughn. Description: Albuquerque: University of New Mexico Press, 2019. | Includes bibliographical references and index. Identifiers: LCCN 2018055939 (print) | LCCN 2018056153 (e-book) | ISBN 9780826360298 (e-book) | ISBN 9780826360281 (pbk.: alk. paper) Subjects: LCSH: Indian pottery—South America. | Indians of South America—Antiquities. | South America—Antiquities. | Archaeological chemistry. | Archaeometry. Classification: LCC F2230.1.P8 (e-book) | LCC F2230.1.P8 C47 2019 (print) | DDC 980/.01 —dc23 LC record available at https://lccn.loc.gov/2018055939

Cover photograph courtesy of Fotosearch Composed in Minion Pro

Contents •

List of Illustrations  ix Preface and Acknowledgments  xv Chapter One.  Compositional Analysis of Archaeological Ceramics  1

Michael D. Glascock Chapter Two.  Pottery Production and Consumption in the ­Andean-­Amazonian Frontier of Southwestern Colombia (2500–500 BP)  15

Hernando J. Gir aldo, Robert J. Speakman, Michael D. Glascock, and Alejandr a M. Gudiño Chapter Three.  Cultural Implications of Instrumental Neutron Activation Analysis of Ceramics from Palmitopamba, Ecuador  25

Ronald D. Lippi and Alejandr a M. Gudiño Chapter Four.  Fabric and Culture: Technological Change in Ecuadorian ­Finger-­Painted Pottery  37

Maria A. Masucci, Hector Neff, Michael D. Glascock, and Robert J. Speakman Chapter Five.  Crafting Beer Jars for the Inca on the North Coast of Peru  51

Fr ances Hayashida Chapter Six.  Early Horizon Cupisnique Ceramic Production in Pomac, North Coast of Peru: The Role of Archaeometry in Its Holistic Understanding  55

Izumi Shimada Chapter Seven.  Was Huacas de Sicán a Pilgrimage Center? Results from Compositional Analysis of Serving Vessels from the Great Plaza  59

Go Matsumoto

v

vi Contents Chapter Eight.  Chemical Characterization and Patterns of Ceramic Movement in the Cajamarca Region of Northern Peru  73

Jason L. Toohey Chapter Nine.  Inca Craft and Ritual Production: Compositional Analysis of Ceramic Pigments from the Temple of the Sun, Pachacamac  87

James A. Davenport Chapter Ten.  The Analysis of Inca Pottery from the Cuzco Region: Implications for the Provisioning of Ceramics for Machu Picchu and Other Inca Sites  97

Richard L. Burger, Lucy C. Salazar, and Michael D. Glascock Chapter Eleven.  1,500 Years of Pottery Production in the Nasca Region of Peru: Instrumental Neutron Activation Analysis from the Site of La Tiza  113

Christina A. Conlee, Matthew T. Boulanger, and Michael D. Glascock Chapter Twelve.  Wari Ceramic Production in the Heartland and Provinces  125

Patrick Ryan Williams, Donna Nash, Anita Cook, William Isbell, and Robert J. Speakman Chapter Thirteen.  Compositional Analysis of Prehispanic Pottery from the Dry Eastern Lowlands of Bolivia  135

William D. Gilstr ap, Emlen Myers, and Hector Neff Chapter Fourteen.  Pottery from Funerary Mounds along the Arid Atacama Desert Coast, Chile: Chemistry, Circulation, and Exchange between the Inlands and Coast during the Formative Period  147

Itací Correa Girrulat, Fr ancisco Gallardo, Mauricio Uribe Rodríguez, Ester Echenique, José Fr ancisco Blanco, Samuel Flewett, Matthew T. Boulanger, and Michael D. Glascock Chapter Fifteen.  A Compositional Characterization of Ceramic Production and Circulation during the Formative Period in Tarapacá, Northern Chile  161

Mauricio Uribe Rodríguez, Estefanía Vidal Montero, Michael D. Glascock, Andrew Menzies, Marcia Muñoz, and Cody C. Roush Chapter Sixteen.  Testing the Social Aggregation Hypothesis for Llolleo Communities in Central Chile: Style, Pastes, and Instrumental Neutron Activation Analysis of Ceramic Smoking Pipes and Drinking Pots  173

Fernanda Falabella, Silvia Alfaro, María Teresa Planella, Matthew T. Boulanger, and Michael D. Glascock Chapter Seventeen.  Recruited or Annexed Lineages: A Chemical Analysis of Purén and Lumaco Pottery and Clays  191

Leslie G. Cecil, Tom D. Dillehay, and Michael D. Glascock

Contents Chapter Eighteen.  Prestige Ceramics in Inca Qollasuyu: Production and Distribution of Imperial and Regional Ceramics in the Southern Andes  195

Verónica I. Williams, Terence N. D’Altroy, Hector Neff, Robert J. Speakman, and Michael D. Glascock Chapter Nineteen.  Social Interaction and Communities of Practice in Formative Period Northwestern Argentina: A Multi-­Analytical Study of Ceramics  209

Marisa Lazzari, Lucas Pereyr a Domingorena, Wesley D. Stoner, María Cristina Scattolin, María Alejandr a Korstanje, and Michael D. Glascock Chapter Twenty.  From the Mountains to the Yungas: Provenience and Distribution of Ceramics in Ambato Societies of the Andes of Argentina in the Fifth Century AD  215

Martin Giesso, Andrés G. Laguens, Silvana R. Bertolino, Matthew T. Boulanger, and Michael D. Glascock Chapter ­Twenty-­One.  Instrumental Neutron Activation Analysis of Archaeological Pottery from Mendoza, Central Western Argentina  221

Nuria Sugr añes, María José Ots, Michael D. Glascock, and Gustavo Neme Chapter ­Twenty-­Two.  Ancient Exchange Networks in the Central Amazon  231

Eduardo G. Neves, Casimiro S. Munita, Roberto Hazenfr atz, and Guilherme Z. Mongeló Chapter ­Twenty-­Three.  Ceramic Instrumental Neutron Activation Analysis Studies in South America  241

Hector Neff and Kevin J. Vaughn References Cited  249 Contributors 289 Index 293

vii

Illustrations •

Figu r es

Figure 3.6. Two reconstructed Inca vessels from a grave at the Yumbo cemetery . . . . . . . . . . . . . . . . . . . . . . . . . 30 Figure 3.7. A number of sherds of Cosanga pottery excavated and subjected to INAA . . . . . . . . . . . . . . 30 Figure 3.8. Map showing known distribution of Cosanga pottery at sites in northern Ecuador . . . . . . . . . . . 31 Figure 3.9. Scatterplot of PC#1 and PC#3 showing four compositional groups of pottery from sites in northern Ecuador . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Figure 3.10. Scatterplot of Cr and La concentrations showing the ellipses for the four compositional groups and the Yumbo and Inca samples from the grave at NL-­18 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Figure 3.11. Partially reconstructed unique vessel . . . . . . . . . . . . 33 Figure 3.12. Scatterplot of PC#1 and PC#2 showing four compositional groups and several unassigned ceramic specimens . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Figure 3.13. Scatterplot of PC#1 and PC#3 showing the four ceramic compositional groups and 40 raw clay samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Figure 4.1. Map of coastal Ecuador indicating collection region for ceramic and raw material samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Figure 4.2. Seriation of Complex I and Complex II Guangala ­finger-­painted ceramics and utilitarian wares from El Azúcar . . . . . . . . . . . . . . . 40 Figure 4.3. Complex I unslipped, ­finger-­painted vessel sample and El Azúcar Valley clay-­rich sediment . . . . . . . . 41 Figure 4.4. Complex I red-­slipped, ­finger-­painted vessel sample and recovery of chert and tuff fragments in river drainage float . . . . . . . . . . . . . . . 42 Figure 4.5. Complex II unslipped, ­finger-­painted vessel, forms, and surface burn patterns . . . . . . . . . . . . . . 42 Figure 4.6. Complex II red-­slipped, ­finger-­painted vessel, forms, and surface burn patterns . . . . . . . . . . . . . . 43 Figure 4.7. Principal components biplot for data from INAA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

Figure 1.1. Cluster analysis dendrogram showing the results for Inca pottery measured by Betty Holtzman at LBNL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Figure 1.2. Biplot of PC#1 and PC#2 for Inca pottery from Cuzco and the Pacific coast . . . . . . . . . . . . . . . . . . . 10 Figure 1.3. Biplot of PC#1 and PC#3 for Inca pottery from Cuzco and the Pacific coast . . . . . . . . . . . . . . . . . . . 10 Figure 1.4. Scatterplot of PC#1 and PC#2 showing the difference between Inca pottery from Cuzco and the Pacific coast . . . . . . . . . . . . . . . . . . . . . . . . . 13 Figure 1.5. Log-­log scatterplot of Cs and Eu concentrations showing the difference between Inca pottery from Cuzco and the Pacific coast . . . . . . . . . . . . . . 13 Figure 2.1. Location of the four valleys surveyed in the Upper Caquetá . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Figure 2.2. Small stone sculpture found in the Valencia Valley and ­corrugated-­style sherd collected from the Upper Caquetá . . . . . . . . . . . . . . . . . . . . . 18 Figure 2.3. Scatterplot of PC#1 and PC#2 showing seven compositional groups . . . . . . . . . . . . . . . . . . . . . . . . 20 Figure 2.4. Log-­log scatterplot plot of Cr and K in parts per million . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Figure 2.5. Raw clay samples plotted against confidence ellipses for core compositional groups shown in Figure 2.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Figure 3.1. Map of Ecuador showing western Pichincha research region and Palmitopamba . . . . . . . . . . . . 26 Figure 3.2. A Tsáchila family near Santo Domingo de los Tsáchilas around 1970 . . . . . . . . . . . . . . . . . . . . . . . . 26 Figure 3.3. Site of Palmitopamba occupies a high hill south of the town of the same name . . . . . . . . . . . . . . . . . 27 Figure 3.4. Stone foundation of a ­never-­finished Inca building . 28 Figure 3.5. Yumbo pottery is mostly plainware for domestic purposes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

ix

x Illustrations Figure 5.1. Location of study sites on the north coast of Peru . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Figure 6.1. An ­oblique-­angle photo of Kiln 38, Kiln Cluster 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 Figure 6.2. Three views of Kiln 8 that show its overall shape and construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 Figure 7.1. Map showing the area of the Lambayeque complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 Figure 7.2. Great Plaza and the trenches hitherto excavated . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 Figure 7.3. Biplot of the first two PCs from PCA of the 225 specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Figure 7.4. Dendrogram showing the results of hierarchical cluster analysis of the 225 specimens . . . . . . . . . . . 65 Figure 7.5. Biplot of the first two PCs from PCA of the 122 specimens belonging to Sicán BDP Groups 1 to 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 Figure 7.6. Dendrogram showing the nested structure of the assemblage of 122 specimens that belong to Sicán BDP Groups 1 to 3 . . . . . . . . . . . . . . . . . . . 67 Figure 7.7. Biplots of the first two PCs from PCA of the assemblage of 122 specimens . . . . . . . . . . . . . . . . . . 68 Figure 8.1. Map of the Cajamarca Basin indicating the location of Yanaorco . . . . . . . . . . . . . . . . . . . . . . . . . 74 Figure 8.2. Architectural plan of Yanaorco . . . . . . . . . . . . . . . . . . 76 Figure 8.3. The ceramic samples formed five analytical groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 Figure 8.4. Examples of Amoshulca Black Geometric, Amoshulca Black Geometric, and Cajamarca Semi-­Cursive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 Figure 8.5. Examples of Cajamarca Black and Orange style . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 Figure 8.6. Examples of Cajamarca Fine Black and a new Cajamarca Gray ware style . . . . . . . . . . . . . . . . . . . .81 Figure 8.7. Distribution of fine ware and utilitarian ware vessels by analytical group . . . . . . . . . . . . . . . . . . . . 81 Figure 8.8. General distribution of form types by analytical group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 Figure 8.9. Distribution of time-­diagnostic sherds by analytical group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 Figure 9.1. Location of Strong, Willey, and Corbett’s excavations at the Temple of the Sun, Pachacamac . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 Figure 9.2. Photo taken during Strong’s excavations showing technique used . . . . . . . . . . . . . . . . . . . . . . 90 Figure 9.3. Schematic profiles showing occurrence percentages of Inca, Inca-­Associated, Early Pachacamac, and all ceramics as identified by Strong et al . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 Figure 9.4. Log-­log scatterplot of the neutron activation results showing amounts of Cs and Sc . . . . . . . . . . 92 Figure 9.5. PCA of the LA-­ICP-­MS results . . . . . . . . . . . . . . . . . . 93

Figure 9.6. Photos of ceramics belonging to each INAA compositional group . . . . . . . . . . . . . . . . . . . . . . . . . 94 Figure 9.7. PCA of the LA-­ICP-­MS of black and white pigments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 Figure 10.1. Archaeological sites that provided the ceramics analyzed in this study . . . . . . . . . . . . . . . . . . . . . . . . 98 Figure 10.2. Biplot of PC#1 and PC#2 showing compositional groups . . . . . . . . . . . . . . . . . . . . . . . 101 Figure 10.3. Biplot of PC#1 and PC#3 showing compositional groups . . . . . . . . . . . . . . . . . . . . . . . 101 Figure 10.4. Scatterplot of PC#1 and PC#2 showing compositional groups . . . . . . . . . . . . . . . . . . . . . . . 102 Figure 10.5. Scatterplot of PC#1 and PC#3 showing compositional groups . . . . . . . . . . . . . . . . . . . . . . . 102 Figure 10.6. Scatterplot of PC#1 and PC#3 showing compositional groups . . . . . . . . . . . . . . . . . . . . . . . 103 Figure 10.7. Scatterplot of PC#1 and PC#2 showing the Group 2 and Pachacamac groups . . . . . . . . . . . . . 103 Figure 10.8. Machu Picchu ceramic paste groups . . . . . . . . . . . 104 Figure 10.9. Ceramic paste Group 1 vessel forms . . . . . . . . . . . . 104 Figure 10.10. Ceramic paste Group 2 vessel forms . . . . . . . . . . . 105 Figure 10.11. Patallacta ceramic paste groups . . . . . . . . . . . . . . . 105 Figure 10.12. Sacsahuaman ceramic paste groups . . . . . . . . . . . 106 Figure 10.13. Choquequirao ceramic paste groups . . . . . . . . . . . 107 Figure 10.14. Espiritu Pampa ceramic paste groups . . . . . . . . . . 108 Figure 10.15. Ceramic paste groups from Pachacamac and other sites in the Lurin Valley . . . . . . . . . . . . . . . . 109 Figure 11.1. Map of the Nasca drainage with La Tiza and other sites mentioned in the text . . . . . . . . . . . . . 114 Figure 11.2. Map of La Tiza with units sampled for INAA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 Figure 11.3. Biplot of first two PCs for the data set . . . . . . . . . . . 118 Figure 11.4. Scatterplot of logged elemental concentrations of Al and Sb values . . . . . . . . . . . . . . . . . . . . . . . . . 119 Figure 11.5. Examples of pottery from identified compositional groups from La Tiza . . . . . . . . . . . 121 Figure 12.1. Map of Wari sites mentioned in the text . . . . . . . . . 126 Figure 12.2. PCA for Moquegua and Wari heartland samples with elemental loadings . . . . . . . . . . . . . .127 Figure 12.3. Scatterplot of first two PCs for Moquegua and Wari heartland chemical groups . . . . . . . . . . . . . . 128 Figure 12.4. Scatterplot of first two PCs showing 90% confidence ellipses and unassigned sample. . . . . 128 Figure 12.5. Scatterplot of Sr and U showing the difference between Moquegua and Wari samples . . . . . . . . . 128 Figure 12.6. Scatterplot of Sc and Th illustrating Baúl Reference group, Mejia A, and Mejia E . . . . . . . . 129 Figure 12.7. Scatterplot of La and Cr distinguishing Baúl Reference group and Mejia A–E . . . . . . . . . . . . . . 129 Figure 12.8. Scatterplot of Cs and Al demonstrating separation of Baúl Reference group and Mejia G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

Illustrations Figure 12.9. Scatterplot of Eu and Hf differentiating the Wari-­1, Wari-­2, and Wari-­3 groups . . . . . . . . . . . . 130 Figure 12.10. Example photos of ceramic vessels from Cerro Baúl sampled via INAA . . . . . . . . . . . . . . . . . . . . . 132 Figure 13.1. Map of the Gas-­TransBoliviano pipeline ­right-­ of-­way and all sites mentioned in the text . . . . . . 136 Figure 13.2. PCA biplot of first two components illustrating three distinct compositional groups . . . . . . . . . . . 142 Figure 13.3. Scatterplot comparing logged elemental concentrations of Th and Sc . . . . . . . . . . . . . . . . . 143 Figure 13.4. Scatterplot comparing logged elemental concentrations of Fe and Sb . . . . . . . . . . . . . . . . . 143 Figure 14.1. Mound cemeteries from Loa River mouth, ­Tocopilla-­Hornitos coastal section, and Mejillones Peninsula . . . . . . . . . . . . . . . . . . . . . . . . 148 Figure 14.2. Loa Café Alisado examples . . . . . . . . . . . . . . . . . . . . 151 Figure 14.3. ­Quillagua-­Tarapacá examples . . . . . . . . . . . . . . . . . 152 Figure 14.4. San Pedro Negro Pulido and San Pedro Rojo Pulido examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 Figure 14.5. General map showing where pottery and clay samples were obtained . . . . . . . . . . . . . . . . . . . . . . 154 Figure 14.6. PCA plot of overall sample . . . . . . . . . . . . . . . . . . . 156 Figure 14.7. Hierarchical cluster analysis of quantitative elemental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 Figure 14.8. PCA considering only ­Quillagua-­Tarapacá pottery fragments and clay samples . . . . . . . . . . . 157 Figure 14.9. PCA considering only Loa Café Alisado fragments and clay samples . . . . . . . . . . . . . . . . . . 158 Figure 15.1. Map of the region, with Formative sites mentioned in the text . . . . . . . . . . . . . . . . . . . . . . . 162 Figure 15.2. Loa Café Alisado (LCA) fragment showing a detail of the “comma rim” . . . . . . . . . . . . . . . . . . . 163 Figure 15.3. Quillagua Tarapacá Café Amarillento (QTC) vessel with lid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 Figure 15.4. Pica Charcollo fragment with its characteristic scraped surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 Figure 15.5. Graph showing frequencies of ceramic types in each petrographic group . . . . . . . . . . . . . . . . . . . . 165 Figure 15.6. Biplot of the first and second PCs . . . . . . . . . . . . . . 166 Figure 15.7. Log-­log scatterplot of Th and Rb concentrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Figure 15.8. Scatterplot of the first and second PCs for the LCA specimens of the Early Formative period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Figure 15.9. Scatterplot of first and second PCs for the QTC / ​QRP types of the Late Formative . . . . . . . . 168 Figure 15.10. Results of the automated scanning electron microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 Figure 15.11. PCA for the samples of Tarapacá and Atacama . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 Figure 16.1. Map of central Chile showing the Llolleo sites mentioned in the text . . . . . . . . . . . . . . . . . . . . . . . 174

xi

Figure 16.2. Mapuche nguillatun ceremony in Carahue, early twentieth century . . . . . . . . . . . . . . . . . . . . . . 175 Figure 16.3. The La Granja site . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 Figure 16.4. Llolleo inverted T–type smoking pipes . . . . . . . . . 178 Figure 16.5. Llolleo drinking pots and wide-­mouth pots. . . . . . 179 Figure 16.6. Frequency distribution of smoking pipes and pottery sherds paste groups . . . . . . . . . . . . . . . . . . 184 Figure 16.7. Scatterplot of PC scores and loading vectors for the first two PCs calculated from smoking pipes and ceramic sherds . . . . . . . . . . . . . . . . . . . . 185 Figure 16.8. Log-­log scatterplot of Rb and Cs concentrations in the La Granja data set . . . . . . . 186 Figure 16.9. Distribution of INAA chemical groups for smoking pipes and pottery sherds . . . . . . . . . . . . 187 Figure 16.10. Scatterplot of PC scores for the first and second PCs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 Figure 16.11. Projection of smoking pipe samples from the La Granja database . . . . . . . . . . . . . . . . . . . . . . . . . 189 Figure 17.1. Ceramic samples from Huitranlebu, Purén, southern Chile, projected onto the first two PCs with vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 192 Figure 17.2. Plot of Cr and Th base-­10 logged concentrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 Figure 17.3. Ceramic samples and three clay samples projected onto the first and third PCs . . . . . . . . . 193 Figure 18.1. The Inca Empire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 Figure 18.2. The southern part of the Inca Empire . . . . . . . . . . . 197 Figure 18.3. Inca pottery from northwest Argentina . . . . . . . . . 198 Figure 18.4. Santamariano and Belén Black-­on-­Red pottery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 Figure 18.5. Yocavil Polychrome and Famabalasto Black-­ on-­Red pottery . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 Figure 18.6. Yavi Chico Polychrome, Inca Paya, and Polished Blackware ceramics . . . . . . . . . . . . . . . . . 200 Figure 18.7. Urcosuyo Polychrome and Pacajes pottery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 Figure 18.8. Plot of Sm and Eu base-­10 logged concentrations for pottery from northern Chile, the Titicaca Basin, northwest Argentina, and two unknown production areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 Figure 18.9. B ­ ivariate plot of PC#1 and PC#2 based on PCA of the northwest Argentina pottery group . . . . . 204 Figure 18.10. Discriminant analysis plot showing separation of all northwest Argentina pottery groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 Figure 18.11. Plot of Cr and Tb base-­10 logged concentrations showing separation of Groups 5 and 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Figure 18.12. Plot of Fe and Sb base-­10 logged concentrations showing separation of Groups 5 and 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

xii Illustrations Figure 18.13. Plot of Co and Mn base-­10 logged concentrations showing separation of Groups 4 and 7 . . . . . . . . . 205 Figure 19.1. Map of study area showing principal areas and sites mentioned in the text . . . . . . . . . . . . . . . . . . . 210 Figure 19.2. Bulk INAA data for ordinary wares, decorated wares, and Vaquerías, Condorhuasi, and intermediate fabric wares . . . . . . . . . . . . . . . . . . . . 212 Figure 20.1. Map of northwest Argentina showing the location of the Ambato Valley . . . . . . . . . . . . . . . . 216 Figure 20.2. Distribution of Ambato sites in the Ambato Valley . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 Figure 21.1. Map of Mendoza Province with archaeological sites from Tunuyán, Diamante, and Atuel River basins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222 Figure 21.2. Compositional group distribution on PC#1 and PC#2 for sherds from the Tunuyán River basin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 Figure 21.3. Compositional group distribution on PC#1 and PC#2 for sherds from the Diamante and Atuel River basins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227 Figure 21.4. Compositional group distribution on PC#1 and PC#2 for sherds from all groups from both areas, north and south of Mendoza Province . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228 Figure 21.5. Compositional group distribution on PC#1 and PC#2 for sherds from both areas . . . . . . . . . . . . . 228 Figure 21.6. Chemical groups in the north and south of Mendoza Province . . . . . . . . . . . . . . . . . . . . . . . . . 229 Figure 22.1. Map of the study area in the Amazon state, Brazil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234 Figure 22.2. Topographic map of the Lago Grande site . . . . . . . 235 Figure 22.3. Pottery sherds from Lago Grande and Osvaldo archaeological sites . . . . . . . . . . . . . . . . . . . . . . . . . 236 Figure 22.4. PCA scores for the elemental concentrations of pottery sherds from Lago Grande and Osvaldo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238 Figure 22.5. Distribution of Paredão and Manacapuru pottery sherds in the chemical groups defined for Lago Grande and Osvaldo . . . . . . . . . . . . . . . . 238 Figure 23.1. Map of South America showing the locations of archaeological sites corresponding to the individual chapters . . . . . . . . . . . . . . . . . . . . . . . . . 242

Ta bles Table 1.1. Concentrations of Elements in Standards and ­Quality-­Control Samples Used for Analysis of Pottery Samples at the Archaeometry Laboratory at MURR . . . . . . . . . . . . . . . . . . . . . . . . . 3 Table 1.2. Comparison of Results for Perlman / ​Asaro Standard Pottery and the Interlaboratory Conversion Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Table 1.3. Group Membership Probabilities for Site-­Based Groups with Outliers Removed . . . . . . . . . . . . . . . 11 Table 1.4. Group Membership Probabilities after Moving Samples to Their Best Compositional Group with Outliers Removed . . . . . . . . . . . . . . . . . . . . . . 12 Table 2.1. Counts of Pottery and Raw Clay Samples by Valley . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Table 2.2. Chemical Group Assignments by Valley . . . . . . . . . . . 19 Table 4.1. Ceramic Class Petrographic Characteristics . . . . . . . . 45 Table 8.1. Cajamarca and General Andean Chronologies . . . . . . 75 Table 8.2. Ceramic Types and Analytical Group Membership . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 Table 10.1. Samples from Cuzco (Sacsahuaman) Collected by Max Uhle and Selected by Betty Holtzman for Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 Table 10.2. Samples from the Lurin Valley and Pachacamac Selected by Betty Holtzman for Analysis . . . . . . . 100 Table 11.1. Chronology of the Nasca Region . . . . . . . . . . . . . . . . 114 Table 11.2. Count and Percentage of Sherds Assigned to the INAA Compositional Groups . . . . . . . . . . . . . . . . 120 Table 14.1. Calibrated Radiocarbon Dates Considered in this Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 Table 14.2. C ­ hrono-­Cultural Assessment and General Frequencies of Coastal Tumuli Pottery . . . . . . . . . 151 Table 14.3. Area, Sites, and Number of Samples . . . . . . . . . . . . . 155 Table 14.4. Provenance of Clay Samples for INAA . . . . . . . . . . 155 Table 15.1. Thermoluminescence (TL) Dates for the Different Ceramic Types . . . . . . . . . . . . . . . . . . . . 164 Table 15.2. Main Petrographic Groups . . . . . . . . . . . . . . . . . . . . . 165 Table 15.3. Number of Specimens Collected at Each Site and Included in the INAA . . . . . . . . . . . . . . . . . . . . . . . 166 Table 15.4. Modal Mineralogy of Samples 24–32 . . . . . . . . . . . . 170 Table 16.1. Thermoluminescence and Radiocarbon Dates for the La Granja Site . . . . . . . . . . . . . . . . . . . . . . . 177 Table 16.2. Morphological Types of Smoking Pipes . . . . . . . . . . 180 Table 16.3. Distribution of Decorated Smoking Pipe Fragments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 Table 16.4. Comparative Frequency of Smoking Pipe Colors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 Table 16.5. Number and Density of Smoking Pipe Fragments in Llolleo Sites . . . . . . . . . . . . . . . . . . . 182 Table 16.6. Smoking Pipes Paste Groups at the La Granja Site with and without Decoration . . . . . . . . . . . . . 182 Table 16.7. Distribution of Paste Groups of Smoking Pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 Table 16.8. Chemical Groups Determined by INAA and Mahalanobis Distances for Pottery and Smoking Pipe Samples . . . . . . . . . . . . . . . . . . . . . . 186 Table 16.9. Distribution of Chemical Groups at the La Granja Site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 Table 16.10. Mahalanobis Distance Calculations for Projection of Samples of Smoking Pipes . . . . . . . 189 Table 18.1. Ceramic Types by Site and Region . . . . . . . . . . . . . . 202

Illustrations Table 18.2. Summary of Ceramic Types Assigned to Compositional Groups based on Mahalanobis Distance Probabilities . . . . . . . . . . . . . . . . . . . . . . . 206 Table 20.1. Locations of Clay Sources . . . . . . . . . . . . . . . . . . . . . 218 Table 20.2. Number of Sherds Assigned to the Ambato Valley Reference Group and Other Sources . . . . 218

xiii

Table 21.1. Environmental and Archaeological Contexts of Sherds Elected for INAA . . . . . . . . . . . . . . . . . . . . 222 Table 21.2. Chemical Groups Identified by Archaeological Site . . . . . . . . . . . . . . . . . . . . . . . . . 226

Preface and Acknowledgments •

During preparation of this volume, we learned that Isabelle Druc had recently published Ceramic Analysis in the Andes (Druc, ed. 2015). Although there is some overlap, this book differs mainly by the greater number of examples, countries, and archaeological questions being investigated. For example, it has been our experience that many archaeologists struggle with the interpretation of ceramic compositional data because they are often large multivariate data sets that can be difficult to manage. As a result, one of the benefits we see in this volume is the number and diversity of studies presented that also cover a wide array of questions. Chapter 1 provides a detailed outline of the analytical procedures for INAA and discusses the most commonly employed approaches to interpretation of compositional data. Chapters 2 through 22 present the case studies from seven different countries: Colombia, Ecuador, Peru, Bolivia, Chile, Argentina, and Brazil. Chapter 23 reviews some of the methodological and substantive issues raised in several of the studies.

The inspiration for this volume came about while working on the database for ceramic samples analyzed by the Archaeometry Laboratory at the University of Missouri Research Reactor (MURR) over the past 30 years. The entire ceramics database contains data for more than 100,000 samples of pottery and clays from around the world that have been analyzed by instrumental neutron activation analysis (INAA) at MURR. We discovered that more than 6,000 samples have been analyzed for colleagues working on a variety of projects throughout South America. We recognized this as an opportunity to organize a symposium entitled “Ceramics of the Indigenous Cultures of South America” at the 2016 meeting of the Society for American Archaeology in Orlando, Florida, where the results from various projects were presented. The comparisons and contrasts between projects conducted in different parts of South America proved extremely interesting. Several of the presentations described the use of multiple methodological approaches; for example, by integrating the reliable compositional groups determined by INAA with information obtained from complementary techniques such as optical petrography, laser ­ablation-­inductively coupled ­plasma-­mass spectrometry (LA-­ICP-­MS), X-­ray diffraction (XRD), and Mössbauer spectroscopy to explain factors such as the geological environments and technological choices of the potters. As a consequence of the successful symposium, a majority of the participants agreed to contribute chapters to this volume. Furthermore, it was agreed that the data from all projects be made available to the public.

The editors wish to thank the authors for timely submission of their chapters. We also acknowledge Candis  C. Lindsey, who assisted by proofreading the individual chapters. And, finally, we acknowledge the National Science Foundation for a series of grants (numbers 9102016, 9503035, 9802366, 0102325, 0504015, 0802757, 1110793, and 1415403) to the Archaeometry Laboratory at MURR that made this work possible.

xv

1

Compositional Analysis of Archaeological Ceramics M i c h a e l D. G l a s c o c k

Introduction

In the middle of the twentieth century, archaeologists and chemists began making use of instrumental techniques capable of measuring the elements present in and on the surfaces of pottery while seeking information to answer archaeological questions. Compositional analysis produces chemical fingerprints by which artifacts with similar compositions might be grouped together and dissimilar artifacts might be assigned to different groups. By examining the compositional data and microscopic and other archaeological information (e.g., context, decoration, use) together, archaeologists are able to infer answers to their questions. Today, the application of compositional analysis to studies of ceramics and other artifacts has become almost routine. Although many analytical methods have been employed to study archaeological materials during the past century, the vast majority of compositional analy­sis investigations have used instrumental neutron activation analysis (INAA) for the purpose of performing bulk characterization of pottery pastes (Speakman and Glascock 2007). The reasons for choosing INAA are (1) the ease of sample preparation, (2) the entire analytical sample is analyzed, (3) a large number of elements can be measured with excellent accuracy and precision, (4) the possibility of automation facilitates efficient handling of large numbers of samples, and (5) the data from samples analyzed years apart and in different INAA laboratories can be compared with an acceptable

Ceramics (i.e., pottery) were the first synthetic materials created by humans and are among the most common artifacts recovered from archaeological sites around the world. The role of humans in the production, distribution, and usage of pottery makes it an ideal material for investigating questions regarding human behavior and interactions between prehistoric cultures. Some of the questions of interest to archaeologists are, Where did potters obtain their raw materials? How was pottery production organized? What does craft organization tell us about the social, political, and economic conditions under which they were made? What can the presence of pottery made at one site but also found at another tell us about trade and exchange? Technological investigations of ancient pottery began in the middle of the nineteenth century with studies of its visual and microscopic properties. T. W. Richards (1895) and his Harvard graduate students performed the first quantitative analysis of pottery using the wet chemistry gravimetric method. However, they concluded that the method was too laborious and not practical for studying large numbers of samples. Interest in scientific characterization of pottery grew during the 1930s and 1940s when Anna Shepard (1956) laid the foundations for ceramic petrography.

1

2 Gl a sco ck level of confidence. Although INAA has some disadvantages because certain elements are not sensitive by INAA (e.g., Mg, Si, and P), the samples become radioactive, and availability is limited, it can be argued that these disadvantages are relatively minor compared to the impact that INAA data has afforded to thousands of ceramic studies. When INAA is combined with a complementary technique such as petrographic analysis, an even more complete picture of the production, distribution, and consumption of pottery is often possible. The most significant alternative to INAA for bulk analysis of pottery is inductively coupled p ­ lasma-­mass spectrometry (ICP-­MS). ICP-­MS is capable of measuring more elements than INAA, but ICP-­MS requires the analytical sample be dissolved in strong acids under very high temperature and pressure in a microwave oven. In addition, the number of samples one can process in a ­single batch by ICP-­MS is limited (~10–20), expensive high-­purity acids are necessary, and more labor is required such that the total analytical costs are compa­ rable. For these reasons, ICP-­MS studies of pottery are less common than INAA. Two other analytical methods that have gained popularity in recent years are energy dispersive X-­ray fluorescence (ED-­XRF), primarily through use of por­table analyzers, and laser a­ blation-­inductively coupled p ­ lasma-­mass spectrometry (LA-­ICP-­MS). However, neither is a “true” bulk analytical technique. The popularity of the portable XRF spectrometer is due to the capability of performing a rapid, nondestructive analysis at archaeological sites and in museums. However, the identification of unambiguous compositional groups by XRF (i.e., either laboratory based or portable) is challenging because (1) the shallow penetration depth of X-­rays means the entire sample is not uniformly analyzed, (2) precision and accuracy for heterogeneous samples is limited, and (3) the number of discriminating elements measured is restricted (Speakman et al. 2011). Although LA-­ICP-­MS can measure a greater number of elements than XRF, the analysis takes place only on the surface (Wallis and Kemanov 2013). Instead of studying the pottery fabric, both analytical techniques are more successful when analyzing slips, paints, and glazes. Another application for LA-­ICP-­MS is for the examination of temper inclusions present in the pottery fabric (Stoner and Glascock 2012).

Sa mple Pr epa r ation of Pottery a nd Cl ays The sample preparation procedures employed by the Archaeometry Laboratory at the University of Missouri Research Reactor (MURR) and most other INAA laboratories are similar. Work begins by recording complete descriptions of the archaeological samples, including their field identification numbers, provenience, style, paste color, and so forth. For pottery samples submitted as sherds, two small fragments of approximately 1 cm2 in area (~0.5 g each) are removed from the original sherd. One fragment is retained for archival purposes, and the other is used to produce the analytical sample. All exposed surfaces of the analytical sample are scraped with a ­silicon-­carbide grinding tool to remove pos­sible soil contamination and decorations such as paint or glaze. The scraped specimen is then washed in deionized water, dried, ground into powder with an agate mortar and pestle, and stored in a glass container. For whole pots and other precious samples in museums that cannot be ground, powders are obtained by drilling with a ­tungsten-­carbide drill bit. Unfired clays are heated in a furnace at 700ºC for two hours, after which a portion of the clay is ground into a powder and stored in a glass container. The powdered samples are dried in an oven at 105ºC for 24 hours before they are ready for weighing. At MURR, two subsamples are prepared from each analytical sample for separate short and long irradiations. For short irradiations, 150 mg of powder is weighed into a clean, high-­density polyethylene vial. For long irradiations, 200 mg of powder is weighed into a clean, high-­ purity quartz vial. Weights are recorded to the nearest 0.01 mg. Both vials are sealed prior to irradiation. If less sample material is available, the sample weights are reduced proportionally. When the total amount of powder is less than 100 mg, the short irradiation sample is transferred into a quartz vial after it has decayed to a safe handling level. Standard reference materials made from the National Institute of Standards and Technology certified standard reference materials (SRMs) are similarly prepared. The primary standard for ceramic analysis is coal fly ash (i.e., SRM-­1633a or SRM-­1633b), with suitable concentra-



Compositional Analysis of Archaeological Ceramics

3

Table 1.1.  Concentrations of Elements in Standards and ­Quality-­C ontrol Samples Used for Analysis of Pottery Samples at the Archaeometry Laboratory at MURR STANDARDS Element

SRM-­1633a Coal Fly Ash

QUALIT Y CONTROL

SRM-­1633b Coal Fly Ash

SHORT HALF-­L IVES (T 1/ 2 < 24 HRS) Na (%) 0.165 Al (%) 14.1 K (%) 1.89 Ca (%) — Ti (%) 0.80 V (ppm) 300 Mn (ppm) 190 Ba (ppm) 1,320 MEDIUM HALF-­L IVES (24 HRS < T 1/ 2 < 7 DAYS) As (ppm) 145 La (ppm) 79.1 Lu (ppm) 1.075 Nd (ppm) 76 Sm (ppm) 16.8 Yb (ppm) 7.50 U (ppm) 10.3 LONG HALF-­L IVES (T 1/ 2 > 7 DAYS) Sc (ppm) 38.6 Cr (ppm) 193 Fe (%) 9.38 Co (ppm) 44.1 Ni (ppm) 130 Zn (ppm) 220 Rb (ppm) 134 Sr (ppm) 835 Zr (ppm) 240 Sb (ppm) 6.15 Cs (ppm) 10.4 Ce (ppm) 168 Eu (ppm) 3.58 Tb (ppm) 2.53 Hf (ppm) 7.29 Ta (ppm) 1.93 Th (ppm) 24.0

0.201 14.8 2 — 0.79 300 143 683

SRM-­688 Basalt Rock

New Ohio Red Clay

8.79

0.138 9.49 3.45 0.14 0.62 203 263 611

132 85.5 1.05 82 18.6 7.43 8.8

14.8 50.1 0.59 47 9.25 4.31 3.3

40.2 197 7.78 48.6 116 206 138 1,036 223 4.85 10.5 184 3.93 2.73 6.76 1.84 24.4

18.3 90 5.05 22.6 75 94 181 60 181 1.11 10.1 112 1.72 1.24 7.34 1.48 14.9

tions for all elements present in ceramics except calcium.1 For calcium determinations, the basalt rock ­standard ­(SRM-­688) is used. To monitor quality control of data collected at MURR, an in-­house reference material (i.e., New Ohio Red Clay) is used. The concentrations of elements in the standards and quality controls used at MURR are listed in Table 1.1.

Sa mple Ir r a diation a nd Measu r ement Following sample preparation, two irradiations and three measurements are conducted on each sample. As described in Glascock (1992), the short irradiation is carried out through the pneumatic tube irradiation system where the samples are sequentially irradiated in pairs for five seconds by a neutron flux of 8 × 1013 n cm−2 s−1.

4 Gl a sco ck After a 25-­minute decay, the radioactive ­samples are counted for 12 minutes by a pair of high-­purity germanium detectors. Gamma rays for up to nine elements are measured from the short irradiation samples, including Al, Ba, Ca, Dy, K, Mn, Na, Ti, and V. The long irradiation samples are ­bundled together in batches of 50 along with standards and quality controls. The bundle of ­samples and standards is subjected to a 24-­hour irradiation in a neutron flux of 6  ×  1013 n cm−2 s−1. The long irradiation samples are allowed to decay for seven days before the vials are washed and placed on an automatic sample changer to count for 30 minutes each to measure seven elements, including As, La, Lu, Nd, Sm, U, and Yb. After the counts are finished, the samples are allowed to decay for at least two more weeks before they are counted on the sample changer a final time for 2.5 hours each. The third count yields up to 17 elements, including Ce, Co, Cr, Cs, Eu, Fe, Hf, Ni, Rb, Sb, Sc, Sr, Ta, Tb, Th, Zn, and Zr. The concentration data from all three measurements are tabulated in parts per million for each element. After all ­quality-­control checks verify that the data are correct, a spreadsheet is created listing the samples in the form of a matrix with the analytical IDs in the first column and concentrations of elements in subsequent columns. A final step is to combine the chemical data with the descriptive data to produce a single database file and merge it with all previous pottery and clay analyses from MURR now numbering more than 7,500 for South America and 100,000 for the entire world. Although we attempt to measure about 33 elements in samples by INAA, a few of the elements may be below the detection limits for pottery coming from certain regions (e.g., As, Ni, Sb, and Sr), some elements have such low precision that they may be less useful for the statistical analysis (e.g., Nd, Tb, and U), and some elements may exhibit contamination from the researcher having used a ­tungsten-­carbide drill bit (e.g., Co and Ta). These elements should always be examined with caution.

Inter labor atory Compar ability Since the first analyses of pottery and clays were conducted in the middle of the twentieth century, dozens of laboratories have used INAA to analyze pottery. The INAA laboratories at Brookhaven National Laboratory

(BNL), Lawrence Berkeley National Laboratory (LBNL), University of Manchester, University of Michigan, and others produced compositional data for tens of thousands of pottery samples from all over the world. Most of these ­first-­generation INAA laboratories are no longer in operation. The investments in labor, neutrons, and other resources to generate the compositional data were ­significant—not to mention the fact that portions of the archaeological samples were consumed. In 2005, recognizing that the data collected in the f­irst-­generation laboratories might be lost, the Archaeometry Laboratory at MURR began a project to rescue the INAA data from the closed laboratories. Fortunately, a few of the retired scientists from BNL, LBNL, and University of Manchester (Asaro and Adan-­ Bayewitz 2007; Harbottle and Holmes 2007; Newton et al. 2007) were still living. They cooperated with MURR by providing access to their INAA data and other records that, after hundreds of hours of effort, facilitated creating a digital version of their data. Although some data could not be recovered, a majority of the data were rescued (Boulanger 2012). Data from the LBNL and University of Manchester laboratories can be downloaded from the Archaeometry Laboratory at MURR web pages: http:​ ​/​/ ​archaeometry​.missouri​.edu ​ /  datasets ​/ ​datasets​.html. Before a successful comparison of the data from different INAA laboratories is possible, information about the reference standards used in each laboratory is necessary. Largely for historical reasons, the various laboratories employed different reference standards to calibrate their data, which hinders direct comparison of compositional data. However, if the standards have been analyzed in each laboratory or if a sufficient number of similar archaeological samples have been analyzed in common, it is possible to establish interlaboratory conversion factors. The value of maintaining interlaboratory comparability is obvious, as it makes it possible to merge the data from different laboratories into a single large database. The LBNL was the first INAA laboratory to analyze pottery from South America. In 1968, Betty Holtzman, a graduate student at the University of California, Berkeley, obtained access to a unique collection of Inca and Wari pottery stored in the Lowie Museum. Holtzman used a ­tungsten-­carbide drill to remove powders from a total of 166 samples that she later analyzed by



Compositional Analysis of Archaeological Ceramics

5

Table 1.2.  Comparison of Results for Perlman/​A saro Standard Pottery and the Interlaboratory Conversion Factors ELEMENT

PERLMAN AND ASARO (1969)

MURR RESULTS (N = 19)

LBNL-­T O-­M URR CONVERSION FACTORS

Na (%) Al (%) K (%) Ca (%) Sc (ppm) Ti (%) V (ppm) Cr (ppm) Mn (ppm) Fe (%) Co (ppm) Ni (ppm) Zn (ppm) As (ppm) Rb (ppm) Sr (ppm) Zr (ppm) Sb (ppm) Cs (ppm) Ba (ppm) La (ppm) Ce (ppm) Nd (ppm) Sm (ppm) Eu (ppm) Tb (ppm) Dy (ppm) Yb (ppm) Lu (ppm) Hf (ppm) Ta (ppm) Th (ppm) U (ppm)

0.258 15.3 1.35 — 20.6 0.78 — 102 — 1.02 14.1 278 59 30.8 70 145 — 1.71 8.31 712 44.9 80.3 — 5.78 1.29 — 4.79 2.96 0.43 6.23 1.55 14.0 4.82

0.250 15.5 1.38 0.26 19.9 0.99 152 113 42 1.02 14.2 245 92 30.2 66 114 175 1.65 8.41 718 45.6 81.3 34.9 6.48 1.31 0.71 4.65 2.81 0.51 6.38 1.72 13.6 5.15

0.969 1.013 1.022 — 0.820 1.269 — 1.108 — 1 1.007 0.881 1.559 0.981 0.943 0.786 — 0.965 1.012 1.008 1.016 1.012 — 1.121 1.016 — 0.971 0.949 1.186 1.024 1.110 0.971 1.058

INAA using the Perlman / ​Asaro Standard Pottery for calibration (Perlman and Asaro 1969). Unfortunately, Holtzman’s untimely death ended the project, and the data were never published. Fortunately, as mentioned above, the LBNL pottery database was rescued. To make the LBNL data compatible with projects conducted at MURR, an intercalibration exercise was conducted in which the Standard Pottery was analyzed relative to the usual suite of standards employed at MURR. The resulting LBNL-­to-­MURR conversion factors are listed in Table 1.2.

Statistica l Inter pr etation of Compositiona l Data The interpretation of pottery compositional data has been discussed in detail by multiple authors (e.g., Baxter 1994; Bieber et al. 1976; Bishop and Neff 1989; Buxeda i Garrigós and Kilikoglou 2003; Glascock 1992, 2016; Harbottle 1982a; Kilikoglou et al. 2007; Neff 2000) and will only be summarized here. The primary goal is to identify distinct homogeneous groups within the multivariate database. The production locations for pottery are

6 Gl a sco ck inferred by ­comparing unknown samples to knowns (i.e., clay ­samples) or by indirect means such as the “criterion of abundance” (Bishop et al. 1982) or by other arguments based on geological or sedimentological characteristics (e.g., Steponaitis et al. 1996). The MURRAP software program used by the Archaeometry Laboratory at MURR adopted many of the statistical procedures from the articles mentioned above. The program can be downloaded for free from the website: http:​ ​/​/ ​archaeometry​.missouri​ .edu ​/ ​datasets ​/  GAUSS​_Download​.html. Initial hypotheses about compositional groups are derived either from non-­compositional data (e.g., archaeological context, decoration, or ceramic type) or from applying various ­pattern-­recognition techniques to the multivariate chemical data. Some of the more common ­pattern-­recognition techniques employed to examine multivariate data are cluster analysis (CA), total variation matrix (TVM), principal component analysis (PCA), and canonical discriminant analysis (CDA). Because every project has different questions and every data set is unique, there is no specific sequence by which ­pattern­recognition techniques should be used. In fact, the use of multiple approaches is often recommended. When the results of multiple approaches are found to be in agreement, greater confidence in the final interpretation is the result. All operations within the MURRAP program automatically transform the raw elemental data into logarithms of concentrations. As demonstrated by Aitchison (1999), the log transformation has the intent of creating a more symmetric distribution of the variables (elements) and more approximately equal variances. An additional advantage is that the transformation to logarithms compensates for the differences in magnitude between the major elements (i.e., %) and the trace elements (i.e., ppm) and makes data handling more convenient.

Cluster Analysis One of the procedures most often applied to multivariate compositional data is cluster analysis (CA) because it is rapid and easy to understand. The procedure groups samples according to their similarity to one another. The first step is to calculate the Euclidean distances between pairs of samples. The CA procedure in MURRAP employs a “bottoms up” approach toward building a hierarchy of clusters. Using this approach, every sample is assumed

to be the lowest member of a cluster. As the distances between pairs of samples are compared, those with the smallest distances between are merged together as one moves upward in the hierarchy. Samples separated by large distances are placed in separate clusters. In CA, a dissimilarity matrix is created in which the distances between all pairs of specimens are calculated using one of several possible distance measures (Sayre 1975). The most popular distance measure is the ­squared­mean Euclidean distance, where one calculates the distance between specimen j and specimen k according to the equation d 2jk =

1 n

n

Cij Cik

2

i=1

The scaling factor, n, corresponds to the number of elements actually determined. It removes the possible problem of missing values, since the measure averages over only those elements for which data are included. For an assemblage containing m samples, there are m(m – 1)/2 pairs of samples, and a d 2 value is calculated for each pair. The results are typically presented in the form of a dendrogram indicating the relationships between samples. Although CA is an efficient tool for displaying the visual relationships between samples, it has a number of weaknesses: (1) the solution is not unique and strongly depends on the choices made by the analyst; (2) groups are always created, even if none exist; and (3) CA fails to take into account the correlations between elements. The latter can be a serious problem when dealing with highly correlated data. Groups created from CA should always be considered tentative until validated subsequently by more robust methods of evaluation.

Total Variation Matrix The total variation matrix (TVM) method is an approach to compositional data originally described by Aitchison (1982) and subsequently employed by Buxeda i Garrigós and Kilikoglou (2003) to examine archaeometric data. TVM is used to evaluate the degree of variability within a 2 multivariate data set from the measured variance, m . Using Aitchison’s procedure, the variation matrix is constructed from all possible variances of the logratios for all n elemental concentrations. For example, the covariances are defined by



Compositional Analysis of Archaeological Ceramics ij, kl

= cov log

xj xi ,log xk xk

 for i, j,k,l = 1,…,n.

All diagonal elements of the variation matrix are zero, and the off-­diagonal elements of the variances are defined as

ij

=

ii, jj

= var log

xi xj

 for i, j = 1,…,n.

As a result, the covariances can be calculated from the logratio variances using the equation ij, kl

=

(

il

+

jk

ij

kl

)/ 2

and the total variation vt is given by i=n

vt =

i=1

j=n

2n

j=1 ij

A large value for vt indicates greater variation and suggests the data set may be polygenic (i.e., multiple subgroups), whereas a small value for vt indicates less variability and suggests a possible monogenic data set. Aitchison (1992) also demonstrated that the total variation is directly related to the Euclidean distances between all samples in the data set. In addition to total variation parameter, Buxeda i Garrigós and Kilikoglou (2003) have demonstrated that the variation matrix of the entire data set provides other useful information. For example, sources of variability within the data set can be revealed by examining the logratio variances, τ.i, for each element as a divisor. The parameter provides an estimate of each element’s contribution to the total variation. The information from total variation and logratio variances can be used to isolate samples affected by processes such as alteration and contamination.

Principal Component Analysis Principal component analysis (PCA) is a variable reduction technique often used to simplify the structure of compositional data. The goal is to explain as much of the variance as possible using the minimum number of variables. Although the PCA is not explicitly a ­group-­formation technique, it makes important

7

contributions toward ­finding groups and understanding the differences between groups. PCA performs an orthogonal transformation on the data to convert potentially correlated variables into a set of linearly uncorrelated variables called the principal components (PCs). The goal is to transform the original multivariate data into a new representative data set. However, PCA is only useful if the data are correlated in some way. For uncorrelated data, PCA offers no advantages. In general, archaeological and geological samples exhibit multiple correlated elements. The PCA transformation applies eigenvector methods on the v­ ariance-­covariance matrix to determine the directions and magnitudes of maximum variance in the data set. Assuming the compositional data matrix has m samples and n elements, the first step in PCA involves centering the data on the mean values for each element. This insures that the transformed data will also be centered in the same location but will have no effect on the spatial relationships between samples or the variances between elements. The first principal component (PC) is described by the linear combination of element concentrations according to the equation PC1 = a11C1 + a12C2 +…+ a1nCn where the coefficients a11, a12, . . . a1n of the eigenvector represent the weighting factors for each element that represent the correlation coefficients between the elements and the PCs. By definition, the weighting factors are constrained so that the sum of their squares must equal one: 2 2 2 a11 + a12 +…+ a1n =1

The second PC is calculated in the same way as the first PC, with the condition that it must be orthogonal to the first PC, and it must account for the maximum amount of remaining variance. This process is continued until the total number of PCs is equal to the number of elements, n. The reference axes produced by PCA create a new coordinate system that offers an improved perspective for viewing the data set. The position of every specimen in the elemental concentration space can be converted to its principal score in the new PCA space. Ideally, a greater proportion of the structure for the compositional data set under examination will be

8 Gl a sco ck explained by the lowest PCs. The results of PCA can be studied by inspecting a table of weighting factors or by viewing one or more scatterplots. The total number of unique two-­dimensional scatterplots possible for inspection is n(n – 1)/2; however, the scatterplots for the h ­ igher-­ ordered PC scores gradually explain lesser amounts of the variation. Although different criteria have been proposed for determining how many PCs should be investigated and how many should be ignored, one of the most common criteria is to include all PCs until the total percentage of variance explained reaches an acceptable level. For a majority of archaeological work, the Archaeometry Laboratory at MURR recommends explaining at least 90% of the variance. PCA can also be used to examine the basis for differences between groups. In R-­mode analysis, the weight on each PC can be used to display the scores for samples in the new PCA space. In Q-­mode analysis, the factor scores for the variables (elements) can be inspected as vectors. As described by Neff (1994), the multivariate analysis (MVA) method that performs both simultaneously is known as RQ-­mode PCA. When the RQ-­mode PCA technique is used, it is possible to display both samples and element vectors simultaneously on a single plot known as a biplot. The directions and lengths of the element vectors can be easily interpreted in terms of explaining which elements are responsible for differentiating compositional groups from one another and indicating the degree of correlation between elements. Examination of the element vectors can also be a useful tool for identifying element pairs with the greatest potential for use in scatterplots. Identifying scatterplots of element pairs showing that the differences between compositional groups are real and not a consequence of the PCA procedure is desirable.

Mahalanobis Distance Individual compositional groups are characterized by the location of their centroids and the unique correlations between the elements. The existence of correlations between the elements in geological and archaeological materials necessitates the use of Mahalanobis distance (MD) probability calculations to properly handle compositional data (Bishop and Neff 1989; Sayre 1975). The MD is defined as the squared Euclidean distance between a sample and a group centroid, divided by the

group variance in the direction of the sample (Sayre 1975). It is equivalent to measuring the number of standard deviations between a sample and the group mean along each principal component axis. Mathematically, the MD between specimen k and the centroid of group A can be written as 2 DkA =

n

n

Cik Ai Iij C jk A j . i=1 j=1

Ai and Aj are the mean concentrations of elements i and j in the group, and Iij is the ij element of the inverse of the ­variance-­covariance matrix. The MD is both unitless and ­scale-­invariant and accounts for all correlations between pairs of elements as derived from the off-­diagonal terms of the v­ ariance-­covariance matrix, which the s­ imple Euclidean distance does not. If all axes were rescaled such that they have unit variance, the MD would be equal to the Euclidean distance. Calculation of the probability that a particular sample belongs to a group is based not only on its proximity to the group centroid in Euclidean terms but also on the rate at which the density of samples decreases away from the centroid in the direction of the sample of interest. The significance of differences between two groups of specimens can be tested by Hotelling’s T 2 statistics (the multivariate equivalent of the Student’s t) calculated from D2 according to the equation T2 =

D2

1 1 + m1 m2

,

where m1 and m2 are the numbers of samples in the two groups. Hotelling’s T 2 statistic is equivalent to the MD for individual data points. Therefore, the probabilities of membership are easily calculated after transforming the T 2 statistic into the related F-­value by F=

[m1 + m2 − v −1] T 2 , [m1 + m2 − 2]n

where n is the number of elements. MD calculations can also be used to replace missing values (Sayre 1975). When the number of samples with missing values is modest, it is possible to calculate a replacement value for each sample relative to its ­presumed



Compositional Analysis of Archaeological Ceramics

compositional group based on minimizing the effect on the MD with and without the replacement value.

Group Validation The initial groups created from the examination of scatterplots or cluster analysis should be validated by using the MD to calculate membership probabilities for individual samples. Calculation of the MD from a sample to a group requires that the number of samples in the group exceeds the number of variables (elements, or PCs) by at least one. To avoid bias, individual samples should not be compared to a compositional group in which they are already a member. The solution to this problem is to use a jackknifing procedure by which membership probability for each sample is calculated by temporarily excluding the sample from the group to which it is being compared. For most accurate probability calculations, theoretical studies have shown that the s­ ample-­to-­element ratio, m ​/ ​n, should range from three to five, and the larger the better according to Foley (1972). Unfortunately, the m ​/ ​n problem affects most compositional analysis projects in archaeology because most archaeological sites lack the number of artifacts, or more likely, because archaeologists lack the financial resources necessary to analyze the number of artifacts needed to achieve the recommended ­sample-­to-­element ratio. The most common method for circumventing the small s­ ample-­to-­element ratio is to base MD measures on a reduced number of principal components rather than using the original element ­concentrations.

Classification and Discriminant Analysis Classification and discriminant analysis are techniques used to assign samples to sample groups and to validate those groups. As new archaeological samples are analyzed, the same procedures used to validate group membership are applicable to classifying (or assigning) the new samples to the existing compositional groups. After membership to an existing group is confirmed, it may be necessary to reevaluate the entire data set. If the new ­samples do not belong to existing groups, they may be outliers or representatives of yet to be identified groups. Canonical discriminant analysis (CDA) is a procedure for dimension reduction similar to PCA. CDA contrasts with PCA by extracting a new set of variables that

9

­ aximize the differences between two or more groups m instead of maximizing the total variance. CDA relies on the assumption that the pooled v­ ariance-­covariance matrix is an accurate representation of the total variance and covariance (Davis 1986). By definition, CDA also requires all of the samples to belong to one of the known groups. CDA cannot be used to find new groups. The CDA procedure constructs a series of canonical discriminant functions (one fewer than the number of groups) that for each group maximizes the likelihood for specimens to belong to their assigned group and minimizes the likelihood of belonging to all other groups. The main requirements for CDA are (1) two or more groups, (2) at least two samples per group, and (3) the number of variables must be at least two fewer than the number of samples. The individual discriminant functions are linear combinations of the original data that successively describe decreasing amounts of the separation between the groups. Two-­dimensional scatterplots of the discriminant functions are used to illustrate the success in separating groups by CDA.

Inca Pottery in the LBN L Data base Used for Illustr ation To demonstrate the statistical procedures in the MURRAP program, the data for 43 samples of Inca pottery analyzed by Holtzman at LBNL are used. The Inca-­style samples are from two different regions. Thirteen s­ amples came from the site of Cuzco-­Sacsahuaman (CUZ), and the remaining 30 samples came from the Lurin Valley along the Pacific coast. Of the latter, 14 samples came from the site of Pachacamac (PAC), and 16 came from other sites located throughout the Lurin Valley (LUR). To make data compatible with data generated at MURR, the conversion factors in Table 1.2 were applied to the LBNL data. A total of 26 elements were measured in common between the two laboratories. The MURRAP program performed a log base-­10 transformation on the Inca samples prior to a hierarchical c­ luster analysis. The results of the cluster analysis are shown in the dendrogram in Figure 1.1, where it appears that there are three clusters of samples. The Upper cluster contains all of the pottery from Cuzco along with one sample from Pachacamac and four from the Lurin Valley. The Lower cluster contains a mixture of samples

Figure 1.1. Cluster analysis dendrogram showing the results for Inca pottery from Cuzco (CUZ), Pachacamac (PAC), and Lurin Valley (LUR) measured by Betty Holtzman at LBNL. Three tentative clusters are apparent. Upper cluster has all Cuzco samples. Lower cluster has both Pachacamac and Lurin Valley, but none from Cuzco. The Outlier cluster samples are dissimilar from samples in the Upper and Lower clusters.

Figure 1.2. Biplot of PC#1 and PC#2 for Inca pottery from Cuzco and the Pacific coast. Vector lengths represent contributions of elements to the PCs. Groups are surrounded by 90% confidence ellipses. Outlier samples are individually labeled.

Figure 1.3. Biplot of PC#1 and PC#3 for Inca pottery from Cuzco and the Pacific coast. Vector lengths represent contributions of elements to the PCs. Groups are surrounded by 90% confidence ellipses. Outlier samples are individually labeled.



Compositional Analysis of Archaeological Ceramics

from Pachacamac and other sites in the Lurin Valley but no samples from Cuzco. Two possible Outlier samples (i.e., PAC-­04 and LUR-­07) are located at the bottom of the dendrogram. For convenience, the Upper, Lower, and Outlier clusters are tentatively labeled Cuzco, Coast, and Outlier. The Inca pottery data were then subjected to an RQ-­ mode PCA based on the v­ ariance-­covariance matrix for all 26 elements. The PCA transformation facilitated creation of the biplot for PC#1 versus PC#2 shown in Figure 1.2, where 60.6% of the variance is explained. A second biplot for PC#1 versus PC#3 shown in Figure 1.3 explains 56.7% of the variance. An examination of both biplots finds that the elements As, Cs, and Sb are heavily weighted on PC#1, Na and Ni are heavily weighted on PC#2, and Ca and Ni are heavily weighted on PC#3. Figures 1.2 and 1.3 also confirm that the Outlier samples are significantly different from the Cuzco and Coast groups. Mahalanobis distance–based probabilities for the original Cuzco and Coast sample groups were calculated using the jackknifing technique described earlier. Due to the small number of samples in both groups, the calculations were based on the first six PCs that explain slightly more than 90% of the variance. According to the probabilities shown in Table 1.3, samples LUR-­08 and PAC-­ 05 initially located in the Coast group have much greater probabilities of membership in the Cuzco group. After moving both samples to the Cuzco group, the Mahalanobis distance–based probabilities were recalculated, and the new results are shown in Table 1.4. The improved results are more robust with only sample LUR-­10 showing low probabilities of membership in both groups. The low probability was judged too low to justify moving the ­sample to the other group. Using the new sample grouping, a scatterplot of PC#1 versus PC#2 was generated as shown in Figure 1.4. The Cuzco and Coast groups are well separated. Finally, to show that the compositional groups are not a consequence of MVA, a log-­log scatterplot of elements Cs and Eu was generated as shown in Figure 1.5 showing no overlap between the 90% confidence ellipses for both compositional groups. Based on these results, it appears likely that pottery samples LUR-­08 and PAC-­05 were produced near Cuzco and later transported to the Lurin Valley region. Since Pachacamac was a well-­recognized religious center for the Inca, it is possible that the two pots were brought there for religious pilgrimage.

11

Table 1.3.  Group Membership Probabilities for Site-­Based Groups with Outliers Removed MEMBERSHIP PROBABILITIES(%) FOR SAMPLES IN GROUP: COAST ANID

Coast

Cuzco

Best Group

LUR-­01 LUR-­02 LUR-­03 LUR-­04 LUR-­05 LUR-­06 LUR-­08 LUR-­09 LUR-­10 LUR-­11 LUR-­12 LUR-­13 LUR-­14 LUR-­15 LUR-­16 PAC-­01 PAC-­02 PAC-­03 PAC-­05 PAC-­06 PAC-­07 PAC-­08 PAC-­09 PAC-­10 PAC-­11 PAC-­12 PAC-­13 PAC-­14

17.755 52.288 50.015 89.091 90.186 79.453 10.803 78.858 0.131 37.548 83.627 85.349 98.121 15.987 93.057 73.672 5.690 54.718 5.960 62.504 85.027 21.068 80.144 35.255 89.104 76.393 0.608 43.286

8.495 0.091 0.089 0.069 0.086 0.067 70.152 5.480 0.249 21.654 0.157 0.028 0.058 0.601 0.131 0.061 0.064 0.062 61.129 0.076 0.067 0.161 0.029 0.035 0.117 0.045 0.346 0.386

Coast Coast Coast Coast Coast Coast Cuzco Coast Cuzco Coast Coast Coast Coast Coast Coast Coast Coast Coast Cuzco Coast Coast Coast Coast Coast Coast Coast Coast Coast

MEMBERSHIP PROBABILITIES(%) FOR SAMPLES IN GROUP: CUZCO ANID

Coast

Cuzco

Best Group

CUZ-­01 CUZ-­02 CUZ-­03 CUZ-­04 CUZ-­05 CUZ-­06 CUZ-­07 CUZ-­08 CUZ-­09 CUZ-­10 CUZ-­11 CUZ-­12 CUZ-­13

11.219 15.938 0.049 13.156 0.265 0.603 10.664 5.496 3.255 6.722 0.622 40.451 1.878

79.227 90.831 0.652 3.800 15.942 44.043 79.036 51.615 35.218 90.058 54.941 46.664 81.710

Cuzco Cuzco Cuzco Coast Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco

Note: Results are based on the first six PCs. Probabilities calculated after removing each sample from group.

12 Gl a sco ck Table 1.4.  Group Membership Probabilities after Moving Samples to Their Best Compositional Group with Outliers Removed MEMBERSHIP PROBABILITIES(%) FOR SAMPLES IN GROUP: COAST ANID

Coast

Cuzco

Best Group

LUR-­01 LUR-­02 LUR-­03 LUR-­04 LUR-­05 LUR-­06 LUR-­09 LUR-­10 LUR-­11 LUR-­12 LUR-­13 LUR-­14 LUR-­15 LUR-­16 PAC-­01 PAC-­02 PAC-­03 PAC-­06 PAC-­07 PAC-­08 PAC-­09 PAC-­10 PAC-­11 PAC-­12 PAC-­13 PAC-­14

13.852 46.338 33.731 86.897 84.439 98.777 64.467 0.040 19.503 65.415 82.312 98.421 20.696 95.120 76.950 0.084 78.658 53.989 83.395 21.943 88.678 35.399 72.481 78.386 1.401 40.341

5.008 0.034 0.032 0.030 0.040 0.073 4.216 0.221 17.548 0.074 0.009 0.023 0.352 0.072 0.027 0.021 0.016 0.035 0.031 0.091 0.016 0.014 0.066 0.018 0.125 0.257

Coast Coast Coast Coast Coast Coast Coast Cuzco (not moved) Coast Coast Coast Coast Coast Coast Coast Coast Coast Coast Coast Coast Coast Coast Coast Coast Coast Coast

MEMBERSHIP PROBABILITIES(%) FOR SAMPLES IN GROUP: CUZCO ANID

Coast

Cuzco

Best Group

CUZ-­01 CUZ-­02 CUZ-­03 CUZ-­04 CUZ-­05 CUZ-­06 CUZ-­07 CUZ-­08 CUZ-­09 CUZ-­10 CUZ-­11 CUZ-­12 CUZ-­13 LUR-­08 PAC-­05

0.039 0.629 0.009 1.608 0.086 0.902 0.217 0.364 0.022 0.143 0.416 6.155 1.838 0.527 0.397

73.995 84.827 0.111 15.787 17.886 29.851 69.289 54.688 40.190 85.235 45.875 53.868 73.258 74.400 62.418

Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco Cuzco (moved from Coast) Cuzco (moved from Coast)

Note: Results are based on the first six PCs. Probabilities calculated after removing each sample from group.

Conclusions Compositional analysis of ceramics has a proven record of success for investigations into the role of humans in topics such as (1) the organization and standardization of pottery production, (2) changes in access to resources, (3) mobility patterns, (4) long-­distance trade and exchange, and (5) the nature and scale of interaction. Although a number of analytical techniques have been employed to characterize pottery, the analytical method with the longest and most successful history for bulk analysis of pottery is INAA. The advantages of INAA for bulk analysis of pottery are (1) the high precision and accuracy for a large number of elements with concentrations ranging from trace to major, (2) the ease of sample preparation, (3) the capability of being automated such that large numbers of samples can be analyzed efficiently, and (4) the ability to successfully merge compositional data collected years apart and in different laboratories into a single database. This chapter describes the INAA procedures at MURR employed by the authors of chapters 2–21. Because large quantities of data are created by compositional analysis, multivariate analytical methods are necessary to interpret the data and to make inferences regarding human behavior and interactions between cultures. Statistical methods such as cluster analysis, principal component analysis, total variation matrices, Mahalanobis distance–based probabilities, and discriminant analysis from the statistical sciences are used to facilitate the identification of compositional groups, to compare new samples to previously established groups, and to validate the results. Two-­dimensional and ­three-­dimensional scatterplots and biplots are valuable tools for displaying the results. The results should take into account the correlations between elements, possible alterations during production, and changes during use and postdepositional processes. In this chapter, the MVA procedures used at MURR have been explained and demonstrated using a modest data set of Inca pottery from the sites of Cuzco, Pachacamac, and the Lurin Valley. Many of the other chapters in this volume utilize one or more variations of these ­procedures.



Compositional Analysis of Archaeological Ceramics

13

Figure 1.4. Principal components scatterplot of PC#1 and PC#2 showing the difference between Inca pottery from Cuzco and the Pacific coast. Groups are surrounded by 90% confidence ellipses. Outliers are not shown.

Figure 1.5. Log-­log scatterplot of Cs and Eu concentrations showing the difference between Inca pottery from Cuzco and the Pacific coast. Groups are surrounded by 90% confidence ellipses. Outliers are not shown.

Ack now ledgments

Note

I would like to acknowledge the members of the archaeometry group at MURR who over the past 25 years have assisted in the compositional analysis of more than 7,500 samples of pottery for colleagues working in South America. This work was supported by the University of ­Missouri–Columbia and by a series of laboratory support grants from the National Science Foundation (numbers 9102016, 9503035, 9802366, 0102325, 0504015, 0802757, 1110793, and 1415403) to the Archaeometry Laboratory at MURR.

1. The Archaeometry Lab discontinued use of the coal fly ash standard (SRM-­1633a) in December 2009, when the last bottle was consumed. Since January 2010, all INAA analyses have used the newer standard (SRM-­1633b).

2

Pottery Production and Consumption in the ­Andean-­Amazonian Frontier of Southwestern Colombia (2500–500 BP) H e r n a n d o J . G i r a l d o , R o b e r t J . S p e a k m a n , M i c h a e l   D. Gl ascock, a n d A leja n dr a M. Gu diño

Introduction

l­owlands to the highlands. The type of raw materials or staples moving in the opposite direction is unknown. (2) The exchange of finished products, probably textiles from the highlands to the lowlands, and ceramic objects in the opposite direction. (3) The transference of esoteric knowledge from the lowlands to the highlands. Additionally, the movement of warrior communities’ populations from the lowlands to the mountainous region through invasions by has commonly been mentioned (Duque 1966). The migration of Andean communities to the lowlands is also hypothesized. The origin of these ­Andean-­Amazonian interactions in southwestern Colombia is unknown. Some scholars argue that there is archaeological evidence of these interactions observed in stylistic similarities in the material culture of the two regions during the Postclassic period (ca. AD 900–1550) (Uribe 1981:270–271), or even before (Duque 1966). However, the history of occupation of some groups in the foothills is not very old. The settlers of the foothills along the upper Caquetá River and the Sibundoy Valley were Quechua speakers. Meanwhile, their closest neighbors in the lowlands and highlands were Kamsá / ​Quillacinga speakers (Ramírez de Jara 1996:60–61). The settlement of the Quechua speakers, or Ingas, in the Amazonian foothills seems to be related to the progression of military expansion of the Inca Empire at the end of the fifteenth century AD (Ministerio de Cultura de Colombia 2010:36). This implies that the dynamic

The relations between Andean and Amazonian societies at the moment of the Spanish conquest in southwestern Colombia (sixteenth century AD) are described as dynamic, meaning that there was a constant circulation of goods, knowledge, and people between the two regions, with the foothill inhabitants working as intermediates (Renard et al. 1988:34). The role of the inhabitants of the foothills was not only economic, as specialized traders, but also to provide active mediation in the construction of the ­Andean-­Amazonian relations, merging cultural traditions and ethnic identities (Ramírez de Jara 1996). This role of foothill inhabitants obviously resulted from their geographic location. They were settled in or along natural routes connecting the Amazonian lowlands with the Andean highlands. For example, the settlements in the Sibundoy Valley were located on a strategic route connecting the Mocoa groups along the limits of the rain forest to the Pasto communities in the mountains (Ramírez de Jara 1996). The interactions between the Andean and Amazonian communities in southwestern Colombia imply the exchange of certain types of goods and ideas, which can be summarized by three types (Langebaek 1998): (1) The exchange of raw materials and staples, especially coca leaves, cotton, pepper, honey, and manioc from the

15

16

Gi r a l d o et a l .

relations described in the ethnohistoric accounts between Andean and Amazonian communities in southwestern Colombia were more recent than previously thought. The recent development of ­highland-­lowland relations produced by the colonization of the territory in the Amazonian foothills by Inga communities does not imply that there were no previous interactions between the two regions. On the contrary, some hypotheses of the development of social inequalities in the Andean area, specifically in the San Agustín region, after the first century AD rest on the existence of a dynamic exchange of goods, ideas, and people with Amazonian societies.

San Agustín Chiefdoms and Amazonia The epitome of ­Andean-­Amazonian interactions is seen in the development of the San Agustín chiefdoms in the highlands. These chiefdom societies were characterized by the construction of monumental stone sculptures found in elite class funerary mounds from 200 BC–AD 900 (Drennan 2000; Duque 1966; González 2007). Archaeological research in the area has found no evidence of control of economic resources by the elite (Drennan and Quattrin 1995) or coercive leadership (Drennan 2000). Due to the absence of archaeological evidence of economic or military power as the base of political leadership, it has been suggested that the San Agustín elite based their political power on the control of shamanic or religious knowledge. The link between San Agustín societies and the lowlands developed in different ways. Some scholars relate the iconography depicted in the stone sculptures with typical fauna of the Amazonian lowlands (Drennan 1995; Duque 1966; Gnecco 1996; Llanos and Ordoñez 1998), suggesting that the esoteric knowledge manipulated by religious leaders came from that region. Other scholars argue that the change in mortuary rituals in San Agustín was the result of invasions by lowland warrior communities and the migration of San Agustín communities to Amazonia after about the ninth century AD (Duque 1966). This theory is usually supported by the stylistic similarities in the ceramics of the two regions, specifically the presence of ceramics with corrugated exterior surfaces in San Agustín post-­AD 900 (Uribe 1981). Pottery with this type of decoration is very common in the Amazonian region of northern South America (Guffroy 2006). However, the “invasion / ​migration” hypothesis as an explanation of the change in mortuary ritual has been

rejected due to the continuity of the settlements (Drennan 2000). Nonetheless, scholars suggesting that the changes in mortuary practices were a local phenomenon do not reject some kind of interaction between the two regions due to the presence of that particular kind of ceramic. Even though these interpretations show contradictions, they imply an active circulation of goods, knowledge, and people between the Andean and Amazonian regions from the second century BC until the Spanish conquest in the sixteenth century AD. The links mentioned above have been suggested on the basis of ethnohistoric data or from “intuitive” inferences of the archaeological record, but little empirical evidence has been presented. Therefore, the evaluation of the suggested interaction between Andean (San Agustín) and Amazonian societies by observing patterns of pottery production and consumption in the communities located between (the Upper Caquetá) by using instrumental neutron activation analysis (INAA) is discussed in this chapter.

The Upper Caquetá Region The scenario or route through which the implicit interactions between San Agustín societies and those located in the Amazonian rain forest occurred was possible because of the Upper Caquetá (Figure 2.1). The Upper Caquetá is the name for the southeastern slope of the Colombian Massif (Macizo Colombiano), a complex Andean orographic system where some of the most important Colombian rivers (Magdalena, Cauca, Patía, and Caquetá) originate. This region presents outstanding intraregional differences in altitude, climate, vegetation, and orogeny. The northern part, 50 km long from north to south, from the Páramo de las Papas (3,000 m asl) to the Santa Clara Valley (1,000 m asl) has been exposed to powerful tectonic activity and affected by volcanic activity. The temperature and vegetation vary according to the altitude, from 12°C in the Páramo to 16°C in the southern limit. The southern region of the Upper Caquetá, 20 km in length from the Descanse Valley (850 m asl) to the Yunguillo Valley (600 m asl), is characterized by alluvial deposits forming broad and tall terraces of accumulation (Gnecco et al. 2001). Soils are less fertile than in the north, and temperature varies from 16°C to 24°C. Today, the Upper Caquetá is inhabited by different indigenous communities (Papallacta and Ingas) and peasant colonizers. Due to its geographic and historic relevance as



Pottery Production and Consumption in the ­Andean-­Amazonian Frontier

­ ediator between two macroregions, the upper Caquetá m River valley was the focus of the multidisciplinary Human Use of Landscape in the Upper Caquetá Project directed by Cristobal Gnecco, Gonzalo Buenahora, and Reinaldo García at the turn of the century. The project focused on changes in land use, settlement patterns, demography, social structure, and cosmovision of the communities s­ ettled in the different valleys crossed by the upper Caquetá River, from its first human occupation until the present (Gnecco et al. 2001). The archaeological part of the project’s main objective was to establish the differences in the settlement patterns between valleys according to their soil productivity and their change through time. To accomplish that objective, the project involved the survey of approximately 100 km2 focused on, but not restricted to, four valleys: ­Valencia—30 km2 (2,950 m asl), Santa Rosa— 15 km2 (1,700 m asl), Descanse—15 km2 (800 m asl), and ­Yunguillo-­San Carlos—7 km2 (650 m asl) (Figure 2.1). The systematic survey showed that the region was occupied from the first century AD until the sixteenth century. According to demographic estimates, the population density was very low in the entire region (Giraldo 2007). The majority of the population settled in dispersed houses. Nevertheless, in the Valencia and Santa Rosa Valleys, some hamlets were identified. The survey identified some stone sculptures and mounds (not dated) in the Valencia and Santa Rosa Valleys, but rougher and smaller than those observed in San

17

Agustín (Figure 2.2). The presence of stone sculptures in a site in southwestern Colombia is usually inferred as evidence of some kind of relationship between that site and San Agustín, independent of style, distance, or date (e.g., Cadavid and Ordoñez 1992; Ramírez de Jara 1996:62). Typological analysis of the potsherds indicated that the ceramic assemblages were very homogeneous throughout the region (Gnecco et al. 2001), a perspective not shared by other scholars. Llanos and Alarcón (2000), for instance, indicate the presence of at least three ceramic assemblages belonging to the same number of ethnic groups in the Upper Caquetá: one restricted to the Valencia Valley, another to the region between the Santa Rosa and Yunguillo Valleys, and another one in Mocoa, outside the study area of this chapter. Therefore, the Yunguillo Valley was the frontier between the San ­Agustín-­related communities and those from the rain forest (Llanos and Alarcón 2000:37). The differences between the Amazonian and Andean communities were easily identifiable in the ceramic assemblages, since the former had ­corrugated-­style pottery (Figure 2.2). Llanos and Alarcón (2000) also noticed some similarity in ceramic decorative styles between the Santa Rosa and San Agustín ceramic assemblages. This resemblance led them to propose a strong relationship between these two areas and the probable routes of connection. Ramírez de Jara (1996:128) and Llanos and Alarcón (2000) mention two different routes from which San Agustín inhabitants could have mobilized toward Amazonia through the

500 KM Quinchana

N

VALENCIA

VENEZUELA etá qu

Ca

PANAMA

ita es M

SANTA ROSA

To

r

ve

s

Ri

COLOMBIA

SAN AGUSTIN

AMAZONAS

ECUADOR

DESCANSE

Caquetá Rive r YUNGUILLO

Figure 2.1. Location of the four valleys surveyed in the Upper Caquetá. Map by Hernando Giraldo.

N PERÚ 20 KM

18

Gi r a l d o et a l .

Figure 2.2. Small stone sculpture found in the Valencia Valley (left). A ­corrugated-­style sherd collected from the Upper Caquetá (right). Photo by Hernando Giraldo.

upper Caquetá River: the first one crossing the Páramo, next to the Valencia Valley, and the second one through the La Candela Road, connecting San Agustín with the Santa Rosa Valley (Figure 2.1). As mentioned, the presence of sherds with corrugated surfaces in the San Agustín ceramic assemblages is often used to suggest a continuous interaction between the highlands and lowlands from AD 900 on, with the Upper Caquetá being a communication route. Interestingly, the systematic survey carried out by Gnecco et al. (2001) in the Upper Caquetá provided very low proportions of this kind of pottery. In the Yunguillo Valley, the “frontier zone,” the proportion of corrugated potsherds was 0.3% (Giraldo 2007). This low percentage should call into question the inferred robust interaction between the highlands and lowlands, especially if the exchange of such types of pots is the strongest evidence for those relationships. Besides, the decorative styles shared by the pottery from Santa Rosa and San Agustín regions are extremely widespread throughout southwestern Colombia, and the similarities observed in them are based on a very subjective method. Finally, it is not possible to establish if the pots were imported or only local imitations, independent of how similar these pottery styles look. Taking into account these observations, the Human Use of Landscape in the Upper Caquetá Project aimed to document the patterns of production and distribution of the ceramics in the Upper Caquetá by characterizing the chemical composition of individual ceramic specimens using INAA (Gnecco et al. 2001). By using INAA, it is possible to determine the loci of acquisition of raw materials and to define areas of ceramic production and exchange (Ashley et al. 2015; Bishop et al. 1982). Although the compositional analysis of ceramics was restricted

to samples from the four valleys mentioned above, the results could provide some indirect information about the relationship between Andean and Amazonian societies, and in this way, evaluate the role of the interactions between Amazonian societies in the development of San Agustín chiefdoms. Archaeologically, if the interactions between these two regions (Andes and Amazonia) were continuous and pottery (or its contents) was among the items exchanged, one should expect to see evidence of the movement of that type of good throughout the Upper Caquetá. On the other hand, a restricted movement of pottery in the Upper Caquetá would be an indication of low levels of interaction between the highlands and lowlands, and would undermine the hypothesis of influence of the rain forest communities in the emergence of social inequalities in San Agustín. In addition, one would expect that corrugated ceramics collected in the systematic surface survey would be nonlocal products; otherwise, it would be pos­sible to argue that this ceramic style in San Agustín or in the Upper Caquetá could be an imitation of a foreign style.

N eutron Activation A na lysis Sampling Analyses were conducted on a sample of 237 sherds and 9 raw clays collected from the Valencia, Santa Rosa, Descanse, and Y ­ unguillo-­San Carlos Valleys (Table 2.1). Sixteen specimens of the sample belonged to the corrugated style. They came from the Santa Rosa (n = 3), Descanse (n  = 2), and ­Yunguillo-­San Carlos Valleys (n = 11) (Table 2.2).



Pottery Production and Consumption in the ­Andean-­Amazonian Frontier Methods

a ­scatterplot of K and Cr base-­10 logged concentrations showing separation of groups Santa Rosa 3, Santa Rosa 4, and Valencia. Figure 2.5 projects data derived from analysis of raw clay against the 90% confidence ellipses derived for Santa Rosa 3, Santa Rosa 4, and Valencia shown in the previous figure. Mahalanobis distance–based probabilities of group membership indicate that samples CGG238, 240, 241, 242, and 243 from Descanse, Yunguillo, and Santa Rosa have high probability of membership in Santa Rosa 4. Likewise, CGG244, also from Santa Rosa, exceeds 1% probability of membership in Santa Rosa 3. CGG239, a sample from Descanse, has less than 1% probability of membership in any of the core groups but projects favorably with Santa Rosa 2 in multiple projections of the data. CGG245, a clay sample from Valencia, has a greater than 1% probability of membership in Santa Rosa 4, but this sample plots with the Valencia reference group in multiple projections of the data. A second sample from Valencia (CGG246) has extremely low probabilities of membership in any of the reference groups. Thirteen samples of corrugated ceramics were assigned to the Santa Rosa 4 compositional group. These samples were collected from the Descanse and ­Yunguillo-­San Carlos Valleys. The other three samples of corrugated ceramics collected from the Santa Rosa Valley present a different signature. One belongs to the Yunguillo compositional group, one to the Valencia group, and the remaining sample was unassigned.

The ceramics were prepared for INAA at the University of Missouri Research Reactor (MURR) Archaeometry Laboratory according to routine procedures as described in chapter 1 of this volume. Two irradiations and three gamma counts produced elemental concentrations values for up to 33 elements in most of the analyzed samples. Statistical analyses were subsequently carried out at MURR on base-­10 logarithms of concentrations for the concentration data using procedures also described in chapter 1.

R esu lts The ceramic samples separated into three core groups (Santa Rosa 3, Santa Rosa 4, and Valencia) and four smaller groups (Santa Rosa 1, Santa Rosa 2, Santa Rosa 5, and Yunguillo). Forty-­six ceramic samples and one sample of raw clay could not be assigned to any of the compositional groups. Figure 2.3 shows the first two components derived from principal component analysis (PCA) of the ­variance-­covariance matrix for 31 elements retained in the quantitative analysis presenting the seven compositional groups (Ni and As were deleted from consideration due to the large number of missing values). Figure 2.4 is

Table 2.1.  Counts of Pottery and Raw Clay Samples by Valley* VALLEY NAME

POTSHERDS

R AW CL AY

Valencia Santa Rosa Descanse Yunguillo-­San Carlos

70 73 49 45

2 2 4 1

19

Discussion Intraregional Interactions in the Upper Caquetá The chemical patterns observed in the Upper Caquetá are relatively difficult to discern as a consequence of the low compositional variability within the limits of the study

* San Carlos and Yunguillo are two localities located very close to each other in the same valley.

Table 2.2.  Chemical Group Assignments by Valley VALLEY

SANTA ROSA 1

SANTA ROSA 2

SANTA ROSA 3

SANTA ROSA 4

SANTA ROSA 5

VALENCIA

YUNGUILLO

UNASSIGNED

TOTAL

VALENCIA SANTA ROSA DESCANSE SAN CARLOS YUNGUILLO TOTAL

0 12 0 0 0 12

0 2 7 0 0 9

1 15 20 2 1 39

7 15 15 20 11 68

2 3 0 0 0 5

42 5 0 0 0 47

0 2 0 1 8 11

18 19 7 1 1 46

70 73 49 23 22 237

0.14 -0.10

-0.06

PrincipalPC Component 2 #2 -0.02 0.02 0.06

0.10

Santa Rosa 1 Santa Rosa 2 Santa Rosa 3 Santa Rosa 4 Santa Rosa 5 Valencia Yunguillo Unassigned

- 0.10 -0.06

-0.02

0.02

0.06

0.10

0.14

0.18

2.8

Principal 1 PCComponent #1

Figure 2.3. Scatterplot of PC#1 and PC#2 showing the seven compositional groups. Ellipses represent 90% confidence level for group ­membership.

2.2 2.0

Santa Rosa 4

1.6

1.8

Santa Rosa 3

1.4

Chromium (log base-10 ppm) Cr (log base-10 ppm)

2.4

2.6

Unassigned Samples

1.0

1.2

Valencia

3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

2.8

K (log base-10 ppm) Potassium (log base-10 ppm)

Figure 2.4. Log-­log scatterplot plot of Cr and K in parts per million. Ellipses represent 90% confidence level for membership in the groups.

Clay Samples

2.4 2.2

CGG241

1.8

2.0

90% Confidence Ellipse for Santa Rosa 3 CGG242 CGG238

1.6

CGG240

CGG243

CGG244

1.4

Cr (log base-10 ppm)

90% Confidence Ellipse for Santa Rosa 4

CGG239

CGG245

90% Confidence Ellipse for Valencia 1.2

Chromium (log base-10 ppm)

2.6

CGG246

3.2

3.4

3.6

3.8

4.0

4.2

K (log(log base-10 Potassium base-10ppm) pmm)

4.4

4.6

Figure 2.5. Raw clay samples plotted against confidence ellipses for the core compositional groups shown in Figure 2.4.



Pottery Production and Consumption in the ­Andean-­Amazonian Frontier

area. It is probable that several of the clays used to produce pottery were derived from alluvial sources, which would explain why clays and pottery from Santa Rosa, Descanse, and Y ­ unguillo-­San Carlos are similar. However, the compositional differences between Valencia and the other valleys allow building some relevant inferences on ceramic production and in i­nter-­and intraregional interactions in the Upper Caquetá. Pottery from Valencia seems to be principally a local product, with some pots coming from the lowest valleys. Due to the long distance separating the Valencia and Santa Rosa Valleys (20 km in a straight line), it seems unlikely that raw clay was the material moving from region to region, due to the constraints on raw material transport distances discussed by Arnold (1985). The low compositional variability in the clays in the foothills of the Upper Caquetá makes it impossible to ascertain the loci of production of the foreign pottery collected in the Valencia Valley, but it is significant that no sherds from the Yunguillo compositional group, which seems to be local to the valley of the same name, were found in the Valencia Valley. Nonlocal sherds collected from Valencia belong to the compositional groups Santa Rosa 3, 4, and 5 (Table 2.2), but no pattern can be identified in the foreign sherds (e.g., sherds with elaborate decoration). Consequently, the distribution of ceramics from the Valencia compositional group did not reach the lowest part of the foothills, only the Santa Rosa Valley. Although the sample is small (n = 2), most decorated sherds found in the Santa Rosa Valley (with red bands painted over a ­cream-­colored slip) belonged to the Valencia compositional group. This does not mean that all samples from Valencia were costly ceramics, but it indicates that the exchange between the two valleys was based not only on utilitarian vessels (or their contents) but also in material culture with important symbolic meaning. The results of the compositional analysis suggest a movement of pottery between neighboring valleys, specifically between Valencia and Santa Rosa, which is not extraordinary, given their close proximity. However, it seems that distances involved in the movement of pottery were not large enough to be observed, at least in the ­highland-­lowland direction, because of the lack of samples from the Valencia compositional group in the Descanse and Yunguillo Valleys. Although the data are not conclusive, they point to different dynamics of interaction than those commonly assumed between Andean and Amazonian communities in southwestern Colombia,

21

with a more spatially restricted circulation of goods, at least for those produced in the Upper Caquetá. There is another aspect of the relationship between Andean and Amazonian communities that the data from INAA can clarify. As mentioned, one piece of evidence commonly used to establish the interaction between highland and lowland communities is the presence of pottery with a corrugated exterior surface in San Agustín during the Postclassic period (ca. AD 900–1550) (Llanos and Alarcón 2000; Llanos and Ordoñez 1998; ­Reichel-­Dolmatoff 1975; Uribe 1981). The corrugated style is made by digital pressure in the joints of the rolls created for making vessels. This type of pottery technology is widespread in the Amazonian regions of Colombia, Ecuador, and Peru, post-­AD 1000 (Becerra 1998). According to some scholars (Llanos and Alarcón 2000; Uribe 1981), pottery with this style was transported from the lowlands to San Agustín following the Páramo or the La Candela routes, crossing through the Upper Caquetá. The former connects the Valencia Valley with the Quinchana site through the Páramo de las Papas. The latter connects the Santa Rosa Valley with the most outstanding site of the San Agustín culture, Mesitas. By using the INAA results, it is possible to evaluate the type of interaction involved in the two regions by identifying the loci of production of corrugated sherds. Were they brought to the study area by exchange, or were they produced locally by imitating foreign styles? Sixteen specimens with corrugated style were analyzed within the sample. They came from the Santa Rosa, Descanse, and Yunguillo Valleys. The entire assemblage from the Descanse and Yunguillo Valleys belonged to the compositional group Santa Rosa 4. This means that they could be produced locally in the foothills or in the Santa Rosa Valley, although we lack information on the compositional characteristics of the clay used for pottery production in the lowlands. On the other hand, the three samples collected from the Santa Rosa Valley were produced with different raw clays, from two or three areas: one fits with the Yunguillo compositional group, another fits the Valencia reference group, and the last one is unassigned. Despite the problems identifying the loci of production for most of the samples, the evidence shows that the provenance of corrugated ceramics was not unidirectional. Corrugated ceramics circulated both from the valley closest to the Amazonian border and from the valley next to the Páramo. The small sample size is not helpful for strong assertions, but the results suggest

22

Gi r a l d o et a l .

that the pottery with corrugated decoration in the Upper Caquetá is not necessarily the result of exchanged goods from the lowlands, but rather locally produced, implying a limited movement of goods between the valleys. In this sense, the observation of Llanos and Alarcón (2000:39– 40) that the “motifs” of the corrugated ceramics in San Agustín and the lowlands are not similar makes sense, despite claims to the contrary (Uribe 1981:271). To summarize, (1) there is evidence for a limited movement of ceramic vessels throughout the region, (2) corrugated pottery was also produced in the Valencia Valley, (3) the percentage of ­corrugated-­style potsherds collected from the systematic survey in the Upper Caquetá was insignificant (less than 0.03% of the ceramic assemblage), and (4) there are some stylistic differences between the corrugated ceramics found in San Agustín and those from the lowlands. These four factors point to very low levels of pottery movement between the Andean and Amazonian societies as observed in the communities located in between. These results present a significant counterpoint to understanding the development of the social trajectory of the San Agustín chiefdoms. They indicate that exchange (even indirectly) of goods, specifically pottery, toward San Agustín from the Amazonian region did not significantly impact the consolidation of political structures during the Postclassic period (AD 900–1550). Few distinctive “Amazonian” pots reached the foothills or the highest section of the Upper Caquetá. While some corrugated ceramics were produced in the valley next to the Páramo, their appearance in San Agustín should be explained by mechanisms other than long-­distance exchange or migration. The common assumption about the interaction between these two regions in that period seems to originate in the implicit association between the Inga traders and corrugated ceramics, even though they appear in the Colombian lowlands and foothills at very different periods of time, the end of the fifteenth century AD for the former, and the ninth century AD for the latter. Similar conclusions can be made for the Classic period in San Agustín (ca. AD 1–900), even though there is no distinctive “Amazonian pottery” next to the Upper Caquetá region before the ninth century AD. However, in that period, there is no mention of a San ­Agustín­Amazonia relationship based on ceramic similarity but rather on weak symbolic resemblances between the known ethnographic Uitoto mythology and San Agustín iconography (Duque 1966). More recently, González (2007:123)

has indicated that the long-­distance exchange of ceramic vessels and obsidian was pivotal for strengthening the regional political center of the Mesitas community in San Agustín during the regional Classic period. According to González (2007:123), the elite “consolidated their strategic preeminence by promoting local craft production and participating in exchange networks.” Unfortunately, González did not mention who the trade partners of the San Agustín elites were. Future research will indicate whether there was pottery movement between the San Agustín and Upper Caquetá regions.

Conclusions The studies in southwestern Colombia on the relations between Andean and Amazonian societies have been focused on the type of elements moving from the lowlands to the highlands, but not in the opposite direction. Ethnohistoric accounts are helpful for identifying some of the goods obtained by the inhabitants of the mountainous section of the Upper Caquetá from their counterparts in the lowlands, but these interactions seem to be part of very recent social dynamics related to the Inca expansion at the turn of the fifteenth century AD. Corrugated pottery seems to be one of the goods circulating from the two regions through the foothills, but small amounts of this type of ceramic have been observed in the Upper Caquetá, and INAA results do not support exclusive production of this type of pottery in the lowlands. Although there is a low compositional variability in the region from Santa Rosa to the foothills, it seems that the production of ceramics was local, and the movement of ceramic v­ essels was restricted to neighboring valleys. These results show that pottery was not exchanged between the San Agustín and Amazonian regions, although shorter distance pottery exchange between these two regions and the upper Caquetá valleys cannot be excluded. In addition, exchange of other products and raw materials could have taken place between these two regions, although archaeological evidence supporting such interaction has yet to be found. Some scholars (e.g., Duque 1966; Llanos and Alarcón 2000; Uribe 1981) have inferred a continuous flow of exotic materials between San Agustín and the Amazonian region (assuming an important role of Amazonian societies in the emergence and maintenance of San Agustín chiefdoms) by using as their strongest evidence similarities between some types of vessels between these two



Pottery Production and Consumption in the ­Andean-­Amazonian Frontier

regions. The results of the INAA of the ceramics and the lack of other types of archaeological evidence connecting the highlands and the lowlands do not support such a hypothesis. Nonetheless, they invite us to look for new and better evidence to support the possible existence of such relations. Future research on clay characterization in zones outside the Upper Caquetá, like the Quinchana and Mesitas sites in San Agustín and the flat rain forest, where the corrugated pottery is more common, would provide better evidence for the existence of exchange networks between  these regions using the Upper Caquetá as a direct route.

23

Ack now ledgments We would like to thank Cristobal Gnecco for allowing us to make public the results of his research, which was graciously funded by Colciencias (Colciencias 183–196), to Gabriela Cervantes for reading a previous draft of this chapter, and to the anonymous reviewers for greatly improving the manuscript. Analyses performed in the Archaeometry Laboratory at MURR were partially supported by a grant from the National Science Foundation (#1415403). The INAA data cited in this work is available from http:​ ​/​/ ​archaeometry​.missouri​.edu ​/ ​datasets ​ / datasets.html.

3

Cultural Implications of Instrumental Neutron Activation Analysis of Ceramics from Palmitopamba, Ecuador R o n a l d D. L i p p i a n d A l e j a n d r a M . G u d i ñ o

Introduction

notable type of site consists of petroglyphs pecked in river boulders in the Yumbo territory. Working together with the Tsáchilas, the modern descendants of the Yumbos, we have attempted to interpret these symbols but so far without interpretive consensus from the Tsáchilas. Most of the sites in the region were habitation sites, varying quite a bit in size. A few forts, or pucaras as they are known in Kichwa, the Inca dialect still widely spoken in Ecuador, have also been identified. This lengthy and very challenging stage of field research culminated with the publication of a monograph in Spanish (Lippi 1998) and a revised and condensed version in English (Lippi 2004a).

The Western Pichincha Project was begun by Lippi in 1984 as a multiyear archaeological exploration of the western slope of the Andes Mountains in northern Ecuador’s Pichincha Province (Figure 3.1). A regional survey over several years across diverse zones of approximately 6,000 km2 of rugged cloud forest and tropical rainforest resulted in the discovery of over 300 archaeological sites, which undoubtedly comprise only a small fraction of all sites in the region. Noteworthy types of Yumbo sites discovered by Lippi include dozens of tola (earthen mound) sites, mostly consisting of large rectangular mounds with a platform but also including both large and small circular or elliptical burial mounds. The distribution of tolas in parts of the Yumbo territory seems to mark ancient polities or social groups, some of which were mentioned by name by early Spanish chroniclers. The Yumbos are known from Spanish documents to have inhabited the western cloud forest at the time of arrival of the Spanish and to have conducted extensive trade with various highland towns in the vicinity of Quito (Salomon 1997:17– 26). Lippi discovered physical evidence of trade routes in the form of deeply eroded trails through the rain forest. These trails, locally known as culuncos, mostly date to the Yumbo and historic periods, though some almost certainly were in use many centuries earlier by Formative period peoples (possibly as early as 1500 BC). A third

The Y umbos a nd Incas in North w ester n Ecua dor While sites in western Pichincha spanned a few millennia, the majority were associated with the Yumbos, the Late period (ca. AD 900–1700) indigenous peoples of the western cloud forest. The Yumbos were documented meagerly by the Spanish, and they nearly disappeared by the late 1600s through catastrophic epidemics, migration, forced labor, and assimilation (Lippi 2004a:23–26; Salomon 1997). The most easily identifiable ­modern-­day descendants of the Yumbos are the Tsáchilas (formerly known in Spanish as the “Colorados”), an indigenous nation currently confined to seven small reserves immediately south of the

25

26

L i ppi a n d Gu di ño

Figure 3.1. Map of Ecuador showing the western Pichincha research region and the location of Palmitopamba. Map by Ronald Lippi.

Figure 3.2. A Tsáchila family posed for a portrait near Santo Domingo de los Tsáchilas around 1970. The Tsáchilas are the descendants, at least in part, of the Yumbos. Photo by Ronald Lippi.

western Pichincha research region and near the booming city of Santo Domingo de los Tsáchilas (Figure 3.2). Other Yumbo descendants may have migrated eastward across the Andes Mountains to be assimilated by ­Kichwa-­ speaking tribes such as the Quijos and possibly by other upper Amazonian people. This is harder to prove due to the careless proliferation of the term “Yumbo” by Spanish chroniclers in the sixteenth century and later, as pointed out by Porras G. (1974:165–175) and Salomon (1997:12). The imperial army of Tawantinsuyu (the Inca Empire) pushed northward into m ­ odern-­ day Ecuador from about AD 1475 to 1525 under the kingships of Tupaq Inca Yupanki and Wayna Qhapaq. The final northward push of the empire was led by Wayna Qhapaq but met prolonged resistance by a few northern highland chiefdoms usually referred to collectively as the Caranquis. Lippi has presented the hypothesis that the Yumbos were derived from the Caranquis or related Barbacoan peoples in the northern highlands of Ecuador and migrated to western Pichincha perhaps a thousand years ago (Lippi 2004b). It is believed that the IncaCaranqui War lasted at least a decade and ended around AD 1500 or later (Salomon 1980:219). While the Incas claimed lands to the west of the Ecuadorian highlands to the Pacific Ocean, this was mostly in name only; it is not at all clear that the Incas actually established hegemony or even occupied these



Instrumental Neutron Activation Analysis of Ceramics from Palmitopamba

lowland regions at the far northern end of their empire. The Spanish gained control of territories today comprising Ecuador beginning in 1534. The Spanish described an Inca road that went into the Yumbo territory and mention was made of an Inca general who claimed to have taken control of the area as the Inca army advanced northward toward Colombia in the late 1400s. Ethnohistorian Salomon (1997:23–25) found that the Yumbos occupied an anomalous position in the Inca Empire since they appear not to have been administratively integrated. He also found accounts of Inca nobility taking refuge in Yumbo country following the Spanish conquest. More recently, the Ecuadorian historian Estupiñán Viteri (2003) found a document alleging that Rumiñahui, the captain of the personal guard of Atahualpa, the Inca emperor captured and executed by the Spanish, fled to Yumbo country with some of Atahualpa’s heirs and possibly even with Atahualpa’s mortal remains, where Rumiñahui ordered the building of a fort from which a rebellion was to have been launched. After many years of surveying in western Pichincha, Lippi has catalogued only four Inca forts, and the largest of these is at Palmitopamba (NL-­20). This site may also be of historical significance in Ecuador, where Rumiñahui today is accorded the status of national hero (though he is vilified in Peru as a traitor in the Inca civil war). Estupiñán Viteri thought this connection was reasonable (Lippi et al. 2003) but subsequently hypothesized that the Inca site of Mallqui Machay farther south beyond the Yumbo territory may have been the aforementioned fort.1 Whether or

Figure 3.3. The site of Palmitopamba (NL-­ 20) occupies a high hill immediately south of the town of the same name, which is mostly hidden in a valley. In the background is Pichincha Volcano. The city of Quito lies at its foot on the far side some 45 km to the ­south-­southeast. Photo by Ronald Lippi.

27

not Rumñahui and the corpse of Atahualpa were ever at Palmitopamba, the site was clearly an important point of interaction between the Yumbos and the Incas west of the northern highlands in Ecuador.

Pa lmitopa mba a nd the Thr ee Cer a mic Assembl ages To augment sketchy information on the Inca presence in this tropical region, research shifted in 2002 to excavations at the site of Palmitopamba, which was discovered by Lippi in 1984 and recognized as a probable Inca fortress (pucara) near the northern boundary of the Yumbo territory (Figure 3.3). The site covers several acres and is centered on a very high, steep hill immediately south of the modern town of the same name. Thus, the Western Pichincha Project morphed into the Palmitopamba Archaeology Project, which continues to the present. While excavations over seven seasons since 2002 have indeed confirmed that the site served around the early 1500s at least very briefly as an Inca military site with some familiar as well as enigmatic stone features, it turns out that through most of its history, the site was an important Yumbo center. This can be seen mostly through the platform mound built on top of the hill with 3–4 m of fill carried up and deposited, as well as in the several terraces constructed on the north slope of the hill. The presence of a horizontal volcanic tephra (volcanic ash, sand, and pumice) layer visible in various excavation units and

28

L i ppi a n d Gu di ño

dating to about 900 years ago (a few centuries before the Incas marched north into Ecuador) demonstrates that this monumental earthmoving at the site was done by the Yumbos. Radiocarbon dates confirm that much of the Yumbo occupation preceded the Inca arrival by centuries. No Inca pottery has yet been found on the tola at the site summit, though admittedly excavations there have been limited so far, but it is found elsewhere at the site in modest quantities and is even more abundant farther down the hill. From the apparent absence of Inca pottery at the summit of the site, we infer the Incas respected the sacred precinct at the summit and mostly left it alone. The relative abundance of Inca pottery at historically known Inca forts in the adjacent highland region is notoriously low, typically representing less than 10% of all sherds recovered (Antonio Fresco, personal communication 1987), and that appears to be the case for Palmitopamba too. At Palmitopamba on the lower terraces, the Inca pottery is within the upper 50 cm and is mixed stratigraphically with Yumbo pottery, which continues down to a depth of a meter or more. The most significant Inca ruin on Terrace 4 of the Palmitopamba site is the foundation of a rectangular, stone masonry building, which was completely buried until excavated in 2007 and 2008 (Figure 3.4). In two subsequent seasons, other features associated with this unfinished building were excavated. All of the Inca stone features on Terraces 3 and 4 (well below the Yumbo tola at the summit) appear to have been abandoned prior to their completion. This leads one to infer that the Inca

presence at the site was abruptly terminated for reasons upon which one can only speculate. The arrival of Spanish troops in highland Ecuador might have been the precipitating factor. Whether or not Rumñahui and Inca troops were responsible for the incomplete works at Palmitopamba is even more difficult to know. We have also fallen short in determining just how long the Incas were present at the site. Ethnohistoric evidence mentioned earlier places troops under Inca command in the northern highlands no earlier than about 1490, around the time of the Inca-­Caranqui War, which is believed to have lasted a decade or so (Salomon 1980:219). Whether the first Inca incursion into western Pichincha was for trading purposes, to establish a western flank in the war, or for both reasons is uncertain. It is possible that the Inca presence occurred years later for economic reasons, or approximately in 1534, when Spanish troops moved northward into ­present-­day Ecuador under the command of Sebastián de Benalcázar. Salomon (1997:25– 26) and Estupiñán Viteri (2003) found historic evidence that the Incas fled into western Pichincha away from the Spanish and that some Incas remained among the Yumbos following Spanish domination in the highlands. Radiocarbon dating did not provide the precision needed to specify the time of arrival or the duration of the Inca presence at Palmitopamba. The dates most closely associated with the Inca structure cover far too broad a time range when calibrated; pertinent assays range from the mid-­1400s to the mid-­1600s (Beta Analytic 2008). The best determination on historical grounds is that the

Figure 3.4. The stone foundation of a ­never-­finished Inca building excavated on Terrace 4 at NL-­20. Terrace 4 is 43 m below the platform mound on the summit. Subsequent excavations revealed an enclosed patio (kancha) adjacent to the building. Photo by Ronald Lippi.



Instrumental Neutron Activation Analysis of Ceramics from Palmitopamba a

29

b

Figure 3.5a and b. Yumbo pottery is mostly plainware used for domestic purposes. The whole pot was a grave offering at NL-­18, while the sherds are typical Yumbo with more or less mica or pyrite inclusions. Photos by Ronald Lippi.

Incas were there for no more than a few decades (beginning around 1490 at the earliest) and possibly for as little as a few years (circa 1534) before the Palmitopamba site was abandoned by the Yumbos and the Incas alike. While this dating uncertainty is frustrating, one can still learn quite a lot with regard to the Yumbo-­Inca interaction. Despite the presence of Inca sling stone caches at the summit of the site, there is no clear evidence of hostilities with the Yumbo. The ammunition caches seem more symbolic than tactical. Stratigraphic excavations on lower terraces, as mentioned above, show an apparently peaceful coexistence of these two groups. It is important here to make brief mention of what constitutes Yumbo as opposed to Inca pottery. For the most part they are easily distinguishable to the naked eye. Yumbo pottery consists almost entirely of coarse but well-­ fired plainware with sand and pyrite or mica temper, with only occasional red slip, mostly unpolished, and often slightly irregular in shape (Figure 3.5). Vessels at Palmitopamba appear to have been used generally for cooking and food or drink preparation and consumption; this is plain domestic ware with relatively few diagnostic attributes that could be used for typology or seriation. Inca pottery, well-­known throughout the Inca Empire and described in many detailed archaeological studies from Peru, consists of very competently made ware coming in quite distinct forms, including aríbalos, pedestal bowls, plates with animal effigy appendages, and so forth. They are typically red-­slipped, smoothed, and polished vessels and are often elaborately painted. Useful sources on Inca pottery found in Ecuador include studies by Meyers (1976:Anexo:1–33), Idrovo Urigüen (2000:293–306), and Bray (2003b:165–209). The Yumbo

and the Inca assemblages at Palmitopamba are markedly different in vessel form, temper, finishing, and decoration. While our Inca collection from NL-­20 mostly consisted of small sherds, we recovered two fractured but complete Inca vessels as well as a few other Inca sherds from a grave at NL-­18 (Figure 3.6), which is a satellite site of NL-­20 and was primarily a Yumbo cemetery. One of the goals of the Palmitopamba project has been to study the Inca presence in the tropical forest habitat of western Pichincha and in particular the nature of the relationship with the resident Yumbo population. Research at this dual component site, which contains a long Yumbo occupation followed by a joint occupation by the Yumbos and the Incas, is revealing some details of that interaction. There is a third pottery assemblage at the site known as Cosanga (also known inappropriately as Panzaleo) ware (Figure 3.7). This assemblage consists of large vessels with very thin walls (often about 3 mm) and mica tempering. Typically, the surface is pinkish orange or gray, and the vessels often have a folded rim. There may occasionally be a simple red or white painted design or plastic decoration on the exterior. The pottery is easy to spot as an exotic ware at sites throughout the northern Ecuadorian highlands, where it is widely reported but occurs in low frequencies. Porras G. (1975:150–153) suggested that the ware (he called it “Cosanga-­Pillaro”) originated on the eastern slope of the Andes in or around the Cosanga Valley by about 400 BC. A millennium later it moved up into the northern Andes of Ecuador after ­hunter-­gatherer groups drove the Cosanga farmers out of the eastern mountain region. However, that proposal is speculative and some of his dates are suspect. Lumbreras (1990:42–60) disputes this purported eastern lowland origin.

30

L i ppi a n d Gu di ño

a

b

Figure 3.6a and b. Two reconstructed Inca vessels from a grave at the Yumbo cemetery of NL-­18 (very near NL-­20) include a pedestal beaker (or pedestal pot) and an aríbalos. Both are stylistically and technologically what we consider to be “locally made Inca.” Photos by Ronald Lippi.

Figure 3.7. A number of sherds of Cosanga pottery were excavated at NL-­20 and subjected to INAA. Photo by Ronald Lippi.

Bray (1995, 2003a:125) believes the presence of Cosanga pottery in the highlands is due to long-­term trading between the eastern lowlands and the highlands, quite possibly based on manioc chicha consumption and food provisioning for funeral rituals (Bray 1995, 2003a:138– 140), at least for the Cosanga vessels found in mortuary contexts. Somewhat different Cosanga vessels appear in

domestic contexts. As she points out, there are no pure Cosanga sites known in the highlands, only dozens of sites with a small percentage of Cosanga pottery (which Bray [2003a:119–140] describes in more detail). She agrees more or less with Porras G. (1975) that the Cosanga ware covers the better part of two millennia from perhaps 300 BC to the time of the Spanish conquest (Bray 2003a:125).



Instrumental Neutron Activation Analysis of Ceramics from Palmitopamba

Though there were stylistic changes over time, this was apparently an amazingly conservative artistic tradition. Bray (1995) had the Smithsonian Institution’s Conservation Analytical Laboratory conduct instrumental neutron activation analysis (INAA) on about a dozen highland Cosanga sherds and the same number from the eastern lowlands. The compositional similarity between the two groups was very high. She also had X-­ray diffraction analysis done to compare northern highland Caranqui pottery with northern highland Cosanga pottery and found the two samples to be starkly different. Observations of thin sections from the two collections showed an abundance of metamorphic rock in the Cosanga pottery, which was most likely obtained in the foothills of the eastern range of the Andes and would not have been readily available to the highland potters. The most likely conclusion from these three analyses is that the Cosanga wares in the northern highland sample originated in the eastern lowlands, possibly in or near the Cosanga Valley, which supports our use of the term “Cosanga” rather than “Panzaleo” (a historic highland ethnic group south of Quito). Other researchers cited by Bray (1995:144–145) have also given the Cosanga pottery an eastern origin based on the presence of metamorphic inclusions, despite Lumbreras’s objection (Bray 1995:144–145). While Bray and several others had previously documented the widespread occurrence of Cosanga pottery in the northern highlands, Lippi’s discovery of small quantities of the exotic ware at various sites throughout western Pichincha (Lippi 1998:321), though mostly concentrated in the higher elevations nearer the sierra, further extended the distribution of this presumed trade ware (Figure 3.8). Besides using this geographic data, we will continue to work to refine the somewhat inconsistent dating of Cosanga pottery by recovering it from very secure contexts at Palmitopamba.

INA A for Pa lmitopa mba Cer a mic a nd Cl ay Sa mples From very early on in the research at Palmitopamba, the following objectives were considered when selecting ­samples for INAA. (1) For the Yumbo pottery: (a) Determine to what extent it is a homogeneous assemblage or may represent diverse assemblages

31

Figure 3.8. Map showing the known distribution of Cosanga pottery at sites in northern Ecuador. Modified from Bray (1995:139) with the addition of the Western Pichincha research region and the legend. Map by Ronald Lippi.

by submitting a fairly large sample of what appear to be different Yumbo types. (b) Sample local clays to try to identify where Yumbo potters were obtaining their raw material. We have collected clay samples from 41 deposits within about 4 km of the NL-­20 site. (2) For the Inca pottery: Determine whether the Inca wares at the site were exotic ceramics from the Ecuadorian highlands or beyond or were locally made pottery. To this end, we submitted Inca sherds from Palmitopamba, from two other sites in western Pichincha, and several from the northern highlands for comparative purposes. (3) With respect to the Cosanga pottery: (a) Determine whether all Cosanga pottery at Palmitopamba is a homogeneous assemblage or whether some might be a local imitation of the exotic ware. (b) Determine to what extent Palmitopamba Cosanga pottery compares to Cosanga pottery

32

L i ppi a n d Gu di ño from the highlands and eastern lowlands. We obtained several highland sherds for comparative purposes but none so far from the eastern lowlands.

(4) Try to find a composition match for one unique vessel that differs from all other ceramics at the site by comparison with the INAA database for the Andes. In pursuit of these objectives, over the course of the three field seasons from 2002 to 2004, we submitted 94 sherds from Palmitopamba and elsewhere, including 28 Yumbo sherds, 36 Inca sherds, 29 Cosanga sherds, and one sherd from a unique vessel of unknown affiliation. We also submitted 40 raw clay samples collected near Palmitopamba. Following our excavation of a Yumbo-­Inca grave at the nearby cemetery site of NL-­18 in 2008, which contained Inca vessels, we submitted an additional six samples of Inca sherds from the grave. Results from the University of Missouri Research Reactor (MURR) were prepared in four installments following sample submissions and analy­ses (Descantes et al. 2004, 2005; Ferguson and Glascock 2009; Speakman and Glascock 2003). Four compositional groups of ceramics were identified, one consisting of both Yumbo and Inca sherds from Palmitopamba, another consisting of Cosanga sherds from Palmitopamba as well as from a few highland sites, and two groups of Sierra Inca sherds from four different Inca sites in the greater Quito area in the northern highlands of Ecuador (Figure 3.9). Two Inca sherds collected by Lippi during the western Pichincha surveys many years

ago from two small sites west of Palmitopamba were also submitted. One of them belongs to the Sierra Inca 1 group and the other is unassigned, making both of them different from the Palmitopamba Inca sherds. Of the six sherds submitted from the Yumbo-­Inca grave at NL-­18, two are believed to be Yumbo. One of them fell into the Palmitopamba Yumbo-­Inca INAA cluster, while the other one may belong to that same group but is not a close match. Sherds from the two Inca vessels (an aríbalo and a pedestal beaker) also fell into the Palmitopamba Yumbo-­Inca cluster with an Inca sherd from a presumed Inca vessel possibly belonging to the same cluster. A fourth Inca sherd from a pedestal base did not match that cluster or any other defined cluster from the four INAA studies performed (Figure 3.10). It should also be mentioned that one of the 14 Cosanga sherds from Palmitopamba falls into the Palmitopamba compositional group rather than the Cosanga group, which suggests the sherd may have been misidentified or possibly was a local imitation of Cosanga. All the highland Cosanga sherds, which were from nine different sites in the southern part of Quito to the Machachi Valley a little farther south, fell into the same compositional group and were indistinguishable from the previously tested Cosanga sherds from Palmitopamba (with the one exception noted). There were a few outlier sherds, including the one from the unidentified vessel (Figure 3.11). When the archaeologists at MURR used a larger sample base and did a Euclidean distance search for this sherd, they found the

PC # 3

Sierra Inca 2

Sierra Inca 1

Cosanga Palmitopamba

PC #1 Principal Component 1

Figure 3.9. Scatterplot of PC#1 and PC#3 showing four compositional groups (90% confidence level) of pottery from NL-­20 and various other sites in northern Ecuador. Vectors show elemental influences on the ceramic data. Unassigned samples are not shown.



Instrumental Neutron Activation Analysis of Ceramics from Palmitopamba

33

60

Cosanga

50 40

Sierra Inca 2

Figure 3.10. Log-­log scatterplot of Cr and La concentrations showing the ellipses (90% confidence level) for the four compositional groups and the Yumbo and Inca samples from the grave at NL-­18. The point labeled RDL139, which falls outside the Palmitopamba and other known compositional groups, was from an Inca pedestal plate.

La (ppm) Lanthanum (ppm)

30

Sierra Inca 1 20

Palmitopamba

10 9 8 7 6

10

10 closest matches coming from northern Peru, though the matches are far from conclusive evidence of a shared compositional group. Unfortunately, the Andean database is still extremely incomplete. The discreteness of the four compositional groups is somewhat less clear when the unassigned sherds are added to scatterplots of the data (Figure 3.12). On such plots there seems to be less difference between Palmitopamba and Sierra Inca 1 and slightly less between Sierra Inca 1 and Sierra Inca 2. This may be due to the shared geological substrate and similar parent materials for these areas. For our interpretation, we consider only the sherds that are indisputably within one of the groups. Also collected and submitted to MURR were 41 raw clay samples from sources within a few kilometers of Palmitopamba. The clay samples varied quite a bit in terms of mineral composition. A very few of them were somewhat similar to the Palmitopamba ceramic group and several of them were a little more similar (approximately 10% probability of membership) to the Sierra Inca 1 group. Figure 3.13 shows one principal components scatterplot. The clay samples tested appear not to represent the primary raw material sources for the four ceramic groups. Of the clay samples that are most similar to the Palmitopamba ceramics, one comes from a deposit that is about 2 km northwest of Palmitopamba, and two others came from deposits that are some 3 km east of Palmitopamba. In both cases, that is the ­straight-­line distance; the actual walking distance may be twice that far. Clays sampled closer to the site were not close matches with the ceramics.

20

30

40

50

60

70

80

90 100

Cr (ppm) Chromium (ppm)

Figure 3.11. Partially reconstructed unique vessel from NL-­20. Neither macroscopic analysis nor INAA gives any clues regarding the cultural affiliation of this vessel. Photo by Ronald Lippi.

MURR archaeologists combed their limited database for Ecuadorian ceramics to look for comparative data. Unfortunately, the Ecuadorian database is very small and comparisons attempted between Palmitopamba and south coastal sites in Ecuador are probably not meaningful. There is no reason, based on what Lippi has determined with regard to the origins and trade relations of the Yumbos, to believe that they were linked in any direct way with those sites. There might have been some very indirect link through obsidian trade, but that would not

PC # 2

34

L i ppi a n d Gu di ño

Palmitopamba

Cosanga

Sierra Inca 2

Sierra Inca 1

Clay Unassigned

PC #1 Principal Component 1

Figure 3.12. Scatterplot of PC#1 and PC#2 showing the four compositional groups (90% confidence level) and several unassigned ceramic specimens. Two clay specimens that were submitted earlier than the other 38 are also shown.

Sierra Inca 2

PC # 3

Sierra Inca 1

Cosanga

Palmitopamba

Principal Component 1 PC #1

explain any vague similarities in ceramic composition, which is assumed to be coincidental pending additional evidence.

Conclusions All the Yumbo pottery from Palmitopamba submitted so far falls into a well-­defined compositional group, suggesting that Yumbo pottery from Palmitopamba represents a single production center. This is useful information since

Figure 3.13. Scatterplot of PC#1 and PC#3 showing the four ceramic compositional groups (90% confidence level) and 40 raw clay samples (labeled points). Unassigned samples are not shown.

Yumbo studies are still in their infancy. It will be very interesting in coming years to add Yumbo pottery from other western Pichincha sites to see if it becomes possible to identify production centers or perhaps even social or political units within Yumbo country. So far, the clay sources for the Palmitopamba ceramics have not been identified. None of the Palmitopamba sherds analyzed appear to be a match for any of the many clay samples, even when minerals associated with ceramic temper are identified and their effect on bulk composition is taken into account. This may indicate simply that



Instrumental Neutron Activation Analysis of Ceramics from Palmitopamba

the correct source has not yet been located. There are undoubtedly many clay sources in the area we have yet to test. It is also possible that potters typically mixed clays from two or more sources. Determining mineralogical similarity between the Yumbo pottery and similar Caranqui pottery from the northern highlands, which could help to confirm or disconfirm the matter of Yumbo origins among Caranqui peoples, is now considered a low priority since it seems very unlikely that pottery was imported given that the Yumbos were well established in their own region for several centuries. Only if one could confidently isolate very early Yumbo pottery would it be useful to compare it to contemporary Caranqui ware. So far, it is not possible to distinguish very early Yumbo pottery at Palmitopamba. Work needs to continue toward the goal of identifying temporal indicators in this very plain, conservative ceramic tradition. The fact that the Yumbo and Inca pottery at the site of Palmitopamba appear to have been made from the same raw material is significant. From what is widely known about Inca imperial administration, the most reasonable interpretation is that the Inca troops who occupied Palmitopamba had the local Yumbo potters make Inca pottery to imperial standards in terms of vessel form, decoration, temper, and shape. We made inquiries about this with a few Inca scholars from the central Andes, and they were not at all surprised that local potters were making provincial Inca pottery. To what extent that has merely been the supposition suggested by early chronicles or other documents and to what extent it has been confirmed by mineralogical analysis is not known, but the Palmitopamba data do provide trace element support for the assertion. A remaining mystery about this similarity in the elemental analysis of the two wares is the obvious macroscopic difference; that is, the Yumbo pottery often contains a generous amount of sparkling pyrite or mica in the paste, while such is not found in the Inca ware. It is possible this “glitter” was added to traditional Yumbo wares during the Inca period to emphasize the difference between the two wares. Alternatively, perhaps the Yumbo potters were instructed not to include the sparkly temper for the Inca vessels; this is pure speculation. If there was such a recipe difference, it is interesting that it does not impact the bulk elemental signatures sufficiently to create a difference between Inca and Yumbo pottery. It is very interesting that the two Inca sherds found

35

several kilometers west of Palmitopamba did not match the Palmitopamba Inca pottery. One sherd fell within the Sierra Inca 1 group, suggesting it may have been part of a vessel transported from the highlands. It is interesting to note that an Inca specialist in northern Ecuador, Antonio Fresco, expressed his opinion (personal communication 1985) that the sherd was “imperial Inca” rather than “provincial Inca.” By “imperial Inca” he meant to suggest the vessel may have been transported northward from the core Inca area around Cuzco or from a well-­established Inca center in m ­ odern-­day Ecuador. He may have been wrong about that, but he was correct in inferring that the sherd was different from other Inca pottery from the mountain region. The other Inca sherd from the western mountain region did not match any of the other Inca pottery, and its place of origin is yet to be determined. Perhaps it was locally made by another Yumbo polity even farther west. Since the Inca presence at Palmitopamba was at most only a few decades prior to the Spanish conquest, and since there are reasons to believe some Incas may have remained in Palmitopamba to hide from the Spanish, a search for late pottery that might show Spanish influence has been conducted. None of the Inca pottery from Palmitopamba appears to reflect any Spanish influence, though macroscopic attributes may be more useful than mineralogical ones in solving this issue. There is one glazed Spanish sherd from the site in a stone feature apparently associated with the Inca building foundation. The function of this feature remains a mystery, but the presence of the majolica sherd implies a postconquest date for the feature. The Cosanga pottery, with one possible exception, forms a tightly defined compositional group and differs markedly from all other ceramics from Palmitopamba. This was to be expected given the widespread belief that Cosanga is a trade ware from the upper Amazon, a hypothesis not falsified by the Palmitopamba and highland data presented here. The one sherd from many that represents the vessel of unknown cultural affiliation at Palmitopamba remains unknown until somebody actually recognizes the pot upon close inspection or a much larger INAA database becomes available for the Andes. For now, that vessel remains unidentified. The similarity in composition between several ­Palmitopamba-­area clay samples and the Sierra Inca 1 group of pottery might be coincidental since it is hard to

36

L i ppi a n d Gu di ño

imagine a circumstance under which Inca administrators would ignore local clay deposits and send off to the western flank of the Andes for deposits to be carried on foot to the Andean highlands over very long and difficult rain-­ forest trails. To what extent highland and western mountain clay deposits are mineralogically similar is not clear due to an absence of studies. The parent material of nearly all these clays is primarily a volcanic conglomerate from the same western range volcanoes. The difference between Sierra Inca 1 and Sierra Inca 2 ceramic groups might be attributed to the existence of two different Inca pottery centers, though if such was the case, it appears that which pottery ended up at which site was not simply a matter of proximity. In fact, two of the four highland Inca sites have pottery from both composition groups. It could be that some pottery moved with imperial troops on the march and ended up at sites more distant from the production center. It is also possible that the two Inca ceramic groups were in fact from the same production center in the highlands, a center that may have relied on more than one raw material source. What is perhaps more disconcerting is that when unassigned sherds are added to the scatterplots, the Palmitopamba and Sierra Inca 1 groups, as previously noted, are not so discrete. The same may be said to a lesser degree between Sierra Inca 1 and 2 groups. This is probably to be expected given that the parent materials for many clays in the highlands (except in the eastern range of the Andes) and the western flank are similar. Given the usefulness of much of this data from the INAA, it is unfortunate that the databases for the central and northern Andes are so small, despite many years of concerted effort by MURR personnel to remedy this situation. It is a huge undertaking and will probably require additional decades of research.

Ack now ledgments All funding for the Palmitopamba Archaeology Project has been provided by the Butler Foundation, which is based in Concord, New Hampshire. We are enormously grateful to the trustees for their very generous, long-­term support. Authorization for archaeological research has been granted on an annual basis by the Instituto Nacional de Patrimonio Cultural del Ecuador. We are also grateful to residents of Nanegal Parish and especially to those of Palmitopamba and La Perla, who have welcomed us into their communities and several of whom have done outstanding work in the field and laboratory for us. Appreciation is also extended to local town officials in Palmitopamba and to members of the Parish Board of Nanegal for their support and interest in our project. The opportunity to have INAA performed by the MURR Archaeometry Laboratory at a very modest cost through National Science Foundation grant funding (#1415403) has been very valuable to our project. Finally, we wish to give special mention to the excellent archaeologists and students who have dedicated themselves to the project on a continuing or occasional basis over the years: Tamara Bray, Estanislao Pazmiño, Esteban Acosta, Byron Ortiz, Salomé Osorio, Christian Brito, Eugenia Rodriguez, David Brown, Brandon Lewis, and others for briefer periods. The INAA data cited in this work is available from http:​ ​/​/ ​archaeometry​.missouri​.edu ​/ ​ datasets ​/ ​datasets​.html.

Note 1. Estupiñán Viteri has been quoted several times in the press in Ecuador and has a book-­length manuscript nearly completed on this topic.

4

Fabric and Culture Technological Change in Ecuadorian F ­ inger-­Painted Pottery M a r i a A . M a s u c c i , H e c t o r N e f f , M i c h a e l D. Gl ascock, a n d Robert J. Spe ak m a n

Introduction

phase of southwestern coastal Ecuador (ca. 300 BC– AD 600) offers an opportunity to examine this issue. The project began with traditional ceramic typological and chronological analyses of a poorly defined prehispanic period in southwest coastal Ecuador (Masucci 1992). It has developed to a point where there is detailed data on ceramic technology and the organization of production as well as settlement, subsistence, and social organization (Masucci 1995, 2008; Reitz and Masucci 2004). Such a rich data set provides a valuable case study in which it is possible to ask questions about the relationship between fabric and culture, such as technology or material choices, and cultural meaning. How and in what ways are these ceramic data relevant to other cultural changes, shifts, developments, and questions of anthropological interest? Why did technological changes occur? The wider aim is to move beyond the technological and material patterns themselves to query and examine the range of human, social, ideological, and environmental interactions that affected the ceramics and their technology and style. The goal is to imbed, as Whitbread has suggested, the materials’ analysis within archaeological research (Whitbread 2001:449). How were the raw materials and therefore the choice of raw materials imbedded in the representation these materials had within the culture’s beliefs (Lechtman 1984; S­ illar 1996:259)? This type of research could then advance the applications of technological analyses while also p ­ roviding a richer understanding of the

Early studies of the prehispanic past of Ecuador established cultural units defined by differences in ceramic style, yielding a series of ceramic phases for each region (Evans and Meggers 1961). These phases were then organized within a broader cultural evolutionary framework of developmental stages. Much current research still references these early models, even as their usefulness is questioned. In particular, many of the ceramic phases are identified by a single or a small number of ceramic stylistic traits. Technological analyses and research on the organization of ceramic production have begun to broaden and deepen ceramic analysis with patterns correlated and integrated with data on settlement, subsistence, economy, interaction, and symbol systems (cf. Bray 1995; Hill 2009; Lippi and Gudiño 2006; Masucci and Macfarlane 1997). As ceramic analyses in Ecuador develop to match more intensively researched areas, particularly the central Andes, a key question arises. What is the relationship between fabric and culture? What does the broader data set and integrated analyses mean for larger questions of anthropological interest (e.g., movement of peoples, identity, social groups, alliances, and the relationship of technology and technological choice) to other facets of societal development and even ideology (cf. Lechtman 1984; Sillar 1996)? An integrated research program on the Guangala

37

38

M a succi et a l .

past s­ ocieties of the coast of Ecuador, which go beyond ceramic styles.

Ecua dor: The Gua nga l a The late prehispanic cultural phase of the southern littoral region of Ecuador, named by archaeologists as the Guangala phase, is relegated as the least complex of the coastal manifestations of the Regional Development period (500 BC–AD 500) of the Ecuadorian archaeological sequence (Masucci 2008). This is due to the absence of the traditional markers of increased sociopolitical complexity present in neighboring regions of the northern Andean area. In contrast to its neighbors, Guangala settlements were primarily dispersed farmsteads with no clear indicators of settlement or social hierarchies. The Regional Development period in the Ecuadorian coastal sequence was devised to mark a period of regionalization or rise of more regionalized cultural styles with definitions based almost wholly on decorated pottery (Evans and Meggers 1961). The period was also delineated to mark the appearance of evidence of more complex social constructions, most famously the Jama-­Coaque and La Tolita in the central coastal region (Masucci 2008). The Guangala phase of the Regional Development period is therefore seen as an exception. Why do some areas exhibit a set of patterns associated with increased social complexity such as mound building, site differentiation, differential access to exotics, and associated elite or high-­status goods (e.g., figurines seated with material accoutrements associated as high-­status symbols), while in adjacent regions, like the southern littoral region, they do not? Increasingly sophisticated models that have led to recognition of multiple trajectories and heterarchy rather than a focus on hierarchy avoid simply seeing such societies as failed experiments or to immediately assume that there was some resource or environmental deficiency that did not support complexity. Instead, there appear to be many paths and models for social change, and the northern Andean region has become a fertile area for examining these variations (Drennan 1996). In such studies, individual cases can be interesting, and different cultural trajectories to related factors can be observed. For example, the explanation for the rise of complexity during the Regional Development period has been stated as being due to increased interaction and control of key exotics such as Spondylus shell

(Meggers 1966). There is abundant evidence, however, of extensive ­inter-­and intraregional interaction among the Guangala region and the wider coastal region, Andean highlands, and as far as west Mexico (Burger et al. 1994; Hosler 1994; Masucci 1995; Masucci and Macfarlane 1997; Reitz and Masucci 2004). Therefore, even the explanatory framework of cause and effect for the societal changes during this period can be called into question. There is also no stasis or lack of change during the Guangala phase. Artifact style and technology, diet, resource use, and settlement are all in flux. This occurs as the establishment of local or regional complexity in neighboring regions and is evidenced by figures with local symbols of power, burials with high-­status goods, hierarchies in size, and investment in settlements (Masucci 2008; Reitz and Masucci 2004). The southern coastal Guangala region exhibits changes that may signal the local response to broader changes in the Andean region. The focus here is on the correlation of ceramic technological analy­ ses, mineralogical and elemental, with typological and functional analysis of Guangala ceramics and faunal and settle­ment data. This information can be used to demonstrate how this suite of analyses and the correlation of data can be useful for addressing larger issues of differences in developments and societal responses of late prehispanic northern Andean societies in this dynamic period of societal change and upheaval. It is also possible that one can arrive at a richer understanding of the relationship between technological choice and Guangala society.

Guangala Settlement, Subsistence, and Societal Change The Guangala phase of the southwestern littoral region of Ecuador, as with other coastal cultural phases, is defined by a change in ceramic style dated to about 300 BC–AD 600 / ​ 800. This period is also associated with the expansion of settlement and land use inland into previously sparsely populated secondary and tertiary valleys, suggesting that the ceramic changes are part of larger economic forces, population shifts, or possibly wetter conditions. The construction of water management features such as check dams and catchment basins is also part of this transition between what is labeled as the Late Formative to Early Regional Development period within the broader Ecuadorian prehispanic cultural period sequence. The coastal valleys characterized by Guangala pottery show a consistent pattern of dispersed ­settlements, suggesting



Fabric and Culture

39

Figure 4.1. Map of coastal Ecuador indicating collection region for ceramic and raw material samples. ­Chongon-­Colonche hill range is shaded, and El Azúcar lies directly to the south of the hill range (after Masucci et al. 2016).

f­ armsteads oriented to or adjacent to areas of tillable lowlands near seasonal watercourses and cultural features such as check dams and catchment basins. Typically, sites are represented today by shallow sheet midden deposits, although sites with deeper midden deposits indicating more continuous, longer term use have been found. One of these settlements was investigated in El Azúcar Valley beginning in 1986 (Figure 4.1; Masucci 1992). El Azúcar sites offered well-­preserved stratified midden deposits extending over most of the cultural phase. Radiocarbon assays date the midden deposits to between 60±150 BC and AD 370±80 (Reitz and Masucci 2004:64). A total of 27,362 sherds from three midden contexts were analyzed in the ceramic analyses (Masucci 1992, 2000). The most surprising findings of the typological and chronological analyses were changes over time in the form, surface treatment, vessel size, and paste of what had previously been lumped together as one enormous category of Guangala ­finger-­painted domestic pottery (Figure 4.2). F ­ inger-­width, fugitive, b ­ lack-­painted

­ ecorations over the body of coarse to moderately coarse d paste jars and plates are among the first markers for the Guangala ceramic complex identified by early archaeologists in Ecuador’s coastal region (Bushnell 1951:36–43). These were part of a suite of ceramic attributes that identify this ceramic style, and they are still one of the decorations most commonly used to identify Guangala sites and ceramics. They are one of the cultural trait markers for the time period, but little further attention is afforded to the vessels with these painted designs since they are seen as a useful general marker for the phase but otherwise chronologically and culturally insignificant. But in fact these vessels are not static regional time markers. Based on the surprising findings of the typological and chronological analyses, samples from the full range of stylistic and formal ceramic groups have been analyzed using thin-­section analysis. A subset of that sample was analyzed by instrumental neutron activation analysis (INAA) at the University of Missouri Research Reactor (MURR) by Hector Neff, Michael D. Glascock, and Jeff Speakman

40

M a succi et a l .

Figure 4.2. Seriation of Complex I and Complex II Guangala ­finger-­painted ceramics and utilitarian wares from El Azúcar. Complex I Early Guangala unslipped, thin-­walled, ­finger-­painted (Tomate) on the left and Complex II Middle Guangala thick red slipped, ­thick-­walled, ­finger-­painted (Lechuga) on the right. Photos by Maria Masucci.

(see chapter 1 for a description of the methodology). In addition, electron microprobe analysis was conducted at the Carnegie Institute, Washington, DC (Masucci and Macfarlane 1997; Masucci and Neff 1997).

Compositiona l A na lysis Petrographic Analysis The petrographic analysis focused on the attributes of microstructure, micromass, texture, and identification of mineral inclusions in the pastes and the grouping of ­samples that exhibited similar attributes (Courty et al. 1989; Folk 1974; Kerr 1977; Peacock 1977; Whitbread 1989). The specific attributes recorded for each thin section were (1) microstructure—shape, size, and arrangement of void spaces; (2) micromass—optical activity, homogeneity, identity of fine-­sized particles; (3) composition and ­texture—the identity, size, roundness, and sorting of the coarse inclusions in the fabric, estimation of frequency, modal percentages, and association of mineralogical constituents of the pastes, as well as the weathered state, degree of alteration, degree of vitrification, and orientation of minerals and void spaces; and (4) the arrangement and ratio of c­ oarse-­to-­fine-­to-­voids (Folk 1974; Pettijohn 1975;

Whitbread 1986, 1989). The estimates of frequency, modal percentage and sorting, and description of shapes are based on comparison with visual charts (Bullock et al. 1985; Folk 1974; Pettijohn 1975; Quinn 2013; Whitbread 1989). The ceramic sample for the petrographic analysis was selected to cover the range of paste compositions visible with the aid of a binocular microscope (10x–20x) as well as the major surface treatment and form groups from the chronological complexes defined in typological analyses and the ceramic seriation (Masucci 1992). Resource survey was also conducted to provide a foundation of local and regional raw materials and geology for the ceramic analyses (Masucci and Macfarlane 1997:767, Figure 1). To date, 348 ceramic thin sections, 62 sediment samples, and 39 geological rock samples have been analyzed. The thin sections are archived at Drew University, and the ceramic samples are stored in El Azúcar and type collections with the Instituto Nacional del Patrimonio Cultural, Region 5, Guayaquil, Ecuador. The ceramic thin-­section samples span from the Late Formative through contact period, but the majority (n = 188) of the sample is from the early to middle portions of the Guangala phase. The majority of the ceramic and raw material s­ amples (75%) are drawn from the ­Chanduy-­El ­Azúcar-­Río Grande Valley. The samples from outside this valley system were obtained during a regional resource survey or through



Fabric and Culture

contacts with other archaeologists. The sampling area therefore covers, although not in equal detail, both the southern Guayas coastal area and the area south and immediately north of the coastal C ­ hongon-­C olonche hill range (Figure 4.1). This is significant since this topographic division is the source and cause of a variation in raw materials creating identifiable differences in materials dependent on distance from the hill range and its outcrops (Masucci and Macfarlane 1997). Through thin-­ section analysis of the raw material samples, two compositional groups were discerned, defining a northern and southern clay resource provenance. The thin-­section analysis of the ceramic samples isolated six fabric classes (Masucci 1992; Masucci and Macfarlane 1997) distinguished primarily by composition but also by texture and micromass. The fi ­ nger-­painted, coarse paste vessels were grouped in two separate fabric classes based on differences in texture and composition (Figures 4.3–4.6). The groups correlated with the typological and chronological types and are detailed below. Although the fi ­ nger-­painted

41

samples grouped into two distinct fabric types, both compositions could be related to local raw materials.

INAA A subset of 249 ceramic samples and 54 sediment samples were submitted for INAA. Multivariate statistical analy­ ses were used to examine the data for archaeologically meaningful groups. Principal component analysis (PCA) biplots were used for pattern recognition, and Mahalanobis distances were calculated to test and to refine the potential groups identified via PCA (Neff and Glowacki 2002). Figure 4.7 shows a PCA biplot with confidence ellipses indicating the subgroup structure identified by these means. The majority of ceramic specimens combine into one main group, labeled Core, that is the most defensible, with a series of small divergent groups. These chemical group assignments were made based on all 32 elements (Figure  4.7). The Core group includes samples from all

Figure 4.3. Photomicrographs of Complex I Early Guangala unslipped, ­finger-­painted vessels, vessel form, and rim form profiles. Burn patterns and soot occur on the vessel base. Photos by Maria Masucci.

44

M a succi et a l .

­ olychrome painted vessels, as chemically distinct, and p their likely origin did not correspond to the location of their find site. Therefore, thin-­section analysis provided details for the observed typological, formal, and chronological distinctions in the domestic wares that are indistinguishable chemically and likely all of local origin but that demonstrated nonlocal production and movement of fine ware vessels, which by thin section are indistinguishable. Further details on these results will be summarized, specifically as they relate to the coarse fi ­ nger-­painted vessels and integrate with other related data on settlement and subsistence. These data will be presented according to the two primary chronological divisions defined through the seriation and stratigraphic analyses and supported by radiocarbon assays (Masucci 1992).

Complex I: Ea r ly Gua nga l a (ca. 200 BC –A D 200; Figu r es 4.3 a nd 4.4) Early Guangala was first defined by the presence of ceramics referred to as “red fi ­ nger-­painted jars” (Bushnell 1951; Paulsen 1970). The typological, formal, and chronological analyses of El Azúcar ceramics for this project demonstrated that between about 200 BC and AD 200, or the initial part of the ceramic phase, moderately coarse, unslipped jars and plates decorated with ­finger-­width, black painted lines are the predominant ceramic style, ranging from 45% to 70% of all sherds within excavation lots and 60% of all rims (326 / ​552 rim sherds; Masucci 1992:171, 427). Further, 81% of rims identifiable to form (326 / ​404 rim sherds) were of one form, which are ­round­bottomed, fl ­ aring-­collared jars with sharp, defined throat angles and a range of rim and lip forms (Masucci 1992:178, Table 47). The vessels had a wall thickness range of 0.3– 1.0 cm, rim diameter range of 6–46 cm (mode = 18 and average = 18 cm), and vessel aperture diameter range of 5–32 cm, with vessel height range of 10–45 cm. Therefore, this form had a wide range of sizes (Figure 4.3).

Burn Patterning The most notable feature in the analysis of these sherds was the high percentage of burned sherds. Pre-­breakage burning is restricted, however, to lower bodies and bases, medial angles, and sometimes the lip exterior edges (Figure 4.3). Burned surfaces on these sherds are charred and damaged, showing cracks and alteration of the surface and

a distinct layer of soot. Basal sherds consistently showed extremely heavy burning and charring. The location of the burned areas on these Complex I (Early Guangala) sherds corresponds to that described by Hally (1983:8–10) for vessels suspended over a fire (Masucci 1992:184). These Complex I vessels with ­finger-­painted decorations therefore were characterized by unslipped but wiped and smoothed surfaces and were the most commonly produced and used. A wide range of sizes—but a narrow range of forms—made up the domestic assemblage. Based on the heavily burned basal surfaces, the majority were cooking pots, although the sherds from the largest size range showed very rare evidence of burning and may have been stationary, possibly coming from storage vessels rather than cooking vessels.

Petrographic Analysis (n = 25; Table 4.1) Samples from Complex I ­finger-­painted sherds can be related within a single homogenous paste group with minor differences in texture from locally available sediments. The interpretation is that vessels are locally made with some cleaning of local clays or use of related sources distinct in texture (Masucci and Macfarlane 1997).

INAA Results (n = 13; Figure 4.7) The INAA sample for this type was a subset of 13 from the 25 examined through thin-­section analysis. All these samples from the Complex I fi ­ nger-­painted sherds are members of the Core group relatable to each other and the southern (local) sediment samples.

Other Studies Locally available iron oxides would have made a suitable paint for the fi ­ nger-­painted decorations. Refiring experiments indicate that the paints are mineral based rather than organic since passing 800°C, the painted decoration remains, although it does exhibit changes from shades of gray and black to red as the minerals fully oxidize with the increasing temperatures.

Subsistence Faunal and macrobotanical remains attest to a varied diet for the settlers of El Azúcar Complex I. They relied on domesticated and wild plant foods, domestic ­animals



Fabric and Culture

45

Table 4.1.  Ceramic Class Petrographic Characteristics COMPLEX AND CER AMIC T YPE

SORTING

PREDOMINANT INCLUSION GR AIN SIZE

INCLUSION GR AIN SIZE R ANGE (MM)

PREDOMINANT GR AIN ANGUL ARIT Y

PREDOMINANT INCLUSIONS (>1%)

COMPLEX I: EARLY GUANGAL A UNSLIPPED, ­F INGER-­PAINTED COMPLEX II: MIDDLE GUANGAL A RED-­S LIPPED, ­F INGER-­PAINTED

Moderate

Very fine sand to fine sand

0.05–0.4

Subangular; subrounded

Quartz; Plagio­ clase; Feldspar; Amphibole

Bimodal

Fine sand and coarse to very coarse sand

0.1–1.5

Subangular quartz and feldspar; Subrounded chert and igneous

Quartz; Plagio­ clase; Feldspar; Chert; Tuff; Mudstone

ACCESSORY INCLUSIONS (