Archaeological

Obsidian Studies
Method and Theory
ADVANCES IN ARCHAEOLOGICAL AND MUSEUM SCIENCE

Series Editors:
Martinj. Aitken, Oxford University
Edward V. Sayre, Smithsonian Institution
and R. E. Taylor, University of California, Riverside

Volume l PHYTOLITH SYSTEMATICS: Emerging Issues

Edited by George Rapp, Jr. and Susan C. Mulholland
Volume 2 CHRONOMETRIC DATING IN ARCHAEOLOGY
Edited by R. E. Taylor and Martin J. Aitken
Volume 3 ARCHAEOLOGICAL OBSIDIAN STUDIES: Method and Theory
Edited by M. Steven Shackley

A Continuation Order Plan is available for this series. A continuation order will bring delivery
of each new volume immediately upon publication. Volumes are billed only upon actual
shipment. For further information please contact the publisher.
 Archaeological
Obsidian Studies
Method and Theory

Edited by

M. STEVEN SHACKLEY
University of California
Berkeley, California

Published in cooperation with the

Society for Archaeological Sciences

SPRINGER SCIENCE+BUSINESS MEDIA, LLC

 Library of Congress Cataloging-in-Publication Data

Archaeological obsidian studies method and theory I edited by M.

Steven Shackley.
p. cm. -- (Advances in archaeological and museum science v.
3)
"Published in cooperation wlth the Socie'i.Y for AI'chaBOiuglcai
Sc i ences. "
Includes bibliographical references and index.
ISBN 978-1-4757-9278-2 ISBN 978-1-4757-9276-8 (eBook)
DOI 10.1007/978-1-4757-9276-8
1. Hydration rind dating. 2. ObSidian. I. Shackley, M. Steven.
II. Society for Archaeological Sciences (U.S.) III. Sertes.
CC78.7.A39 1998
930.1·028·5--dc21 98-2610
CIP

© 1998 Springer Science+Business Media New York

Originally published by Plenum Press. New York in 1998
Softcover reprint of the hardcover 1st edition 1998

http://www.plenum.com
10987654321

All rights reserved

No part of this book may be reproduced, stored in a retrieval system, or transmitted in any
form or by any means, electronic, mechanical, photocopying, microfilming, recording, or
otherwise, without written permission from the Publisher
 Consulting Editors

Arnold Aspinall Heather Lechtman

University of Bradford Massachusetts Institute of Technology
Bradford, England Cambridge, Massachusetts

Rainer Berger Christian Lahanier

University of California, Los Angeles Louvre Research Laboratory
Los Angeles, California Paris, France

Hisao Mabuchi
Jonathon E. Ericson
Tokyo National Research
University of California, Irvine Institute of Cultural Properties
Irvine, California
Tokyo, japan

Donald K. Grayson Robert Maddin

University of Washington Harvard University
Seattle, Washington Cambridge, Massachusetts

Garman Harbottle Patrick Martin

Brookhaven National Laboratory Michigan Technological University
Brookhaven, New York Houghton, Michigan

David Harris Frederick R. Matson

Pennsylvania State University
University of London
University Park, Pennsylvania
London, England

Vagn Mejdahl
W. David Kingery Rise National Laboratory
University of Arizona Copenhagen, Denmark
Tucson, Arizona

Pieter Meyers
B. Foss Leach Los Angeles County Museum of Art
Wellington, New Zealand Los Angeles, California
 jacqueline S. Olin julie K. Stein
Smithsonian Institution University of Washington
Washington, D.C. Seattle, Washington

Ernst Pernicka
Max Planck Institute Henrik Tauber
for Nuclear Physics
National Museum of Denmark
Heidelberg, Germany Copenhagen, Denmark

john R. Prescott
University of Adelaide
Adelaide, Australia Michael S. Tite
Oxford University
Oxford, England
Frank Preusser
Los Angeles, California

Giorgio Torraca
T. Douglas Price
University of Rome
University of Wisconsin, Madison
Rome, Italy
Madison, Wisconsin

Fran~ois Schweizer
Laboratory of the Museum of Lambertus Van Zelst
Art and History Smithsonian Institution
Geneva, Switzerland Washington, D.C.
 This volume is dedicated to
Sir Humphrey Davy, Martin Klaproth,
Michael Faraday, P. Robinson
and to all the other early natural scientists who set the course
for archaeological geochemistry, and this book
 Contributors
W. R. Ambrose • Division of Archaeology and Natural History, Research School
of Pacific and Asian Studies, Australian National University, Canberra,
Australia
]. R. Bird • Australian Nuclear Science and Technology Organization, Menai,
New South Wales, Australia
Geoffrey E. Braswell • Department of Anthropology, State University of New
York at Buffalo, Buffalo, New York
David A. Clague • Monterey Bay Aquarium Research Institute, Moss Landing,
California
Robert H. Cobean • Instituto Nacional de Antropologfa e Historia, Mexico
City, Mexico
M. Kathleen Davis • Northwest Research, Corvallis, Oregon
R. Fullagar • Division of Anthropology, Australian Museum, Sydney, Australia
Michael D. Glascock • Research Reactor Facility, University of Missouri, Co-
lumbia, Missouri
C. Gosden • Pitt Rivers Museum, Oxford University, Oxford, United Kingdom
R. C. Green • Department of Anthropology, University of Auckland, Auck-
land, New Zealand
Joachim H. Hampel • Department of Geology and Geophysics, University of
California, Berkeley, California
Thomas L jackson • Pacific Legacy, Aptos, California
james]. Mazer • Chemical Technology Division, Argonne National Labora-
tory, Argonne, Illinois
Barry E. Scheetz • Materials Research Laboratory, Pennsylvania State Univer-
sity, University Park, Pennsylvania
M. Steven Shackley • Phoebe Hearst Museum of Anthropology and Depart-
ment of Anthropology, University of California, Berkeley, California
]. Specht • Division of Anthropology, Australian Museum, Sydney, Australia
ix
X CONTRIBUTORS

Christopher M. Stevenson • Archaeological Services Consultants, Columbus,

Ohio
G. R. Summerhayes • School of Archaeology, La Trobe University, Bundoora,
Victoria, Australia
Timothy Teague • Department of Geology and Geophysics, University of Cali-
fornia, Berkeley, California
R. Torrence • Division of Anthropology, Australian Museum, Sydney, Australia
Robert H. Tykot • Department of Anthropology, University of South Florida,
Tampa, Florida
Marshall I. Weisler • Department of Archaeology, University of Otago, Dune-
din, New Zealand
 Series Foreword

This volume is the third in the Advances in Archaeological and Museum Science
series sponsored by the Society for Archaeological Sciences (SAS). The purpose
of this series is to provide summaries of advances in various topics in ar-
chaeometry, archaeological science, environmental archaeology, preservation
technology, and museum conservation.
The SAS exists to encourage interdisciplinary collaboration between
archaeologists and colleagues in the natural and physical sciences. SAS mem-
bers are drawn from many disciplinary fields. However, they all share a common
belief that physical science techniques and methods constitute an essential
component of contemporary archaeological field and laboratory studies.
The series editors wish to thank the reviewers of each of the chapters in
this volume for their excellent comments and suggestions. We also wish to
thank Chriss jones for her invaluable assistance in the preparation of the texts
for submission to the publisher.

xi
 Preface

As noted in the introductory chapter, this volume is the second major review
of research progress in the study of archaeological obsidian. An earlier book,
Advances in Obsidian Glass Studies: Archaeological and Geochemical Perspectives,
appeared in 1976. A comparison of the treatment of topics reflected in this
earlier work and that contained in this volume not only highlights important
advances in the quality and depth of research on archaeological obsidian over
more than a quarter of a century but also illustrates more generally some
characteristics of developments in the archaeological science field in general.
Both volumes address the two foci of research involving archaeological
obsidian-studies of the hydration process for chronometric applications and
chemical characterization as an approach to attempts to reconstruct prehistoric
regional trade and exchange relationships. In the 1976 volume, the examina-
tion of the physics and chemistry of the hydration process in obsidians was
considered as well as specific applications and problems of obtaining and
interpreting hydration measurements for chronometric purposes as applied to
archaeologically related obsidian in several regions-California, western Mex-
ico, japan, and New Zealand-were considered. Another set of chapters ad-
dressed various issues and problems that were then of concern in chemical
characterization applications for California, Mesoamerica (ancient Mexico),
New Zealand, and the Near East.
This book is focused on similar issues-obsidian hydration studies and
source characterization applications-applied to several different areas. The
chapters comprising this volume reflect the increasingly complex under-
standings of the geochemical properties of obsidian along with a more exacting
and precise appreciation of the physical processes involved in obsidian hydra-
tion. Also, concerns about the validity of compositional data obtained from
certain analytical strategies have become more nuanced. If conclusions about
temporal and spatial relationships using analytical data obtained from obsidian
are to be valid and generally useful as the basis of inferences concerning the
social or economic behavior of prehistoric populations, all components of the
xiii
xiv PREFACE

data sets used in the chain of inference and argument must be critically
evaluated. This is certainly evident in the treatments by the authors whose
contributions are included in this volume. It also might be noted that compar-
ing the list of contributors reflects the changes in research direction of those
involved in basic and applied research in this field over the years in that only
one author-W R. Ambrose-contributed a chapter both to this volume and
to the 1976 compendium.
The editor of this volume has noted that the most important goal of
assembling these contributions has been to facilitate communication between
those involved in basic and applied archaeological science studies-in this case
involving obsidian-and the archaeological professional whether in the acad-
emy or in the public archaeology sector. This is a laudable and commendable
goal. Unfortunately, there is a problem in that archaeology is not a single
discipline. In its institutional and conceptual development, archaeology, particu-
larly in the United States, has not evolved in the same manner as has, for
example, physics, chemistry, or geology. There has not developed an agreement
on a single model of what constitutes the basis of how archaeologists are to
"understand" the phenomenon they study since they cannot even agree on what
constitute the primary units of analysis. In light of the increasing heterogeneity
in conceptual frameworks of those engaged in professional archaeology, an
important task of those involved in archaeological science is to initiate and
institutionalize interactions with those archaeological colleagues for whom
"archaeology is science or it is nothing" to demonstrate that only with empiri-
cally based data can valid generalizations about what actually occurred in the
past be accomplished and why the humans exhibited the individual and collec-
tive behaviors reflected in the materials being examined by archaeologists.
At first blush, obsidian would seem to be an unlikely candidate for all of
this attention. But in the Sherlock Holmes tradition of archaeology, the smallest
piece of data is welcomed in attempting to piece together the set of complex
and often confusing lines of evidence that are used to trace and understand
changes in the behavior of our species.
R. E. Taylor
University of California
Riverside, California
 Contents

Chapter I • Current Issues and Future Directions in

Archaeological Volcanic Glass Studies:
An Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I

M. Steven Shackley

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Current Issues in Obsidian Archaeometry . . . . . . . . . . . . . . . . . . . . . . . . . 4
Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Chapter 2 • A Systematic Approach to Obsidian Source

Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I5

Michael D. Glascock, Geoffrey E. Braswell, and Robert H. Cobean

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Sourcing Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Previous Weaknesses in Interpreting Compositional Data . . . . . . . . . . . . 20
A More Systematic Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Multivariate Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Examples from the Murr Mesoamerican Obsidian Database . . . . . . . . . . 33
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
XV
xvi CONTENTS

Chapter 3 • Mediterranean Islands and Multiple Flows: The

Sources and Exploitation of Sardinian Obsidian 67

Robert H. Tykot

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Multiple Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
Multiple Flows: The Monte Arci Sources . . . . . . . . . . . . . . . . . . . . . . . . . 70
Analysis of Geological and Archaeological Material . . . . . . . . . . . . . . . . . 72
The Distribution of Sardinian Obsidian . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

Chapter 4 • Intrasource Chemical Variability and Secondary

Depositional Processes: Lessons from the American
Southwest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

M. Steven Shackley

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Regional Geologic History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Intrasource Chemical Variability: The Mule Creek Case . . . . . . . . . . . . . 89
Secondary Depositional Effects and Archaeological Problem Domains . . 95
A General Sampling Strategy for Recording Tertiary Sources in Arid
Environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
References 100

Chapter 5 • Characterization of Archaeological Volcanic Glass

from Oceania: The Utility of Three Techniques . . . . . . 103

Marshall I. Weisler and David A. Clague

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Macroscopic Source Identification of Volcanic Glass Artifacts from
Southeast Polynesia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Energy Dispersive X-Ray Fluorescence (EDXRF) Analysis of
Melanesian Obsidian .......................... ............. 109
Electron Microprobe Characterization of Hawaiian Volcanic Glass . . . . . 113
CONTENTS xvii

Summary and Conclusions .................................... 123

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

Chapter 6 • Application of PIXE-PIGME to Archaeological

Analysis of Changing Patterns of Obsidian Use in
West New Britain, Papua New Guinea . . . . . . . . . . . . . 129

G. R. Summerhayes,j. R. Bird, R. Fullagar, C. Gosden,]. Specht, and R. Torrence

Abstract ................................................... 129

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Previous Obsidian Characterization in Melanesia ................... 131
Demand for Finer Discrimination of Sources . . . . . . . . . . . . . . . . . . . . . . 133
PIXE-PIGME . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
Fieldwork at the West New Britain Sources . . . . . . . . . . . . . . . . . . . . . . . 140
Improved Discrimination between Obsidian Sources . . . . . . . . . . . . . . . . 144
Changing Patterns of Obsidian Distribution . . . . . . . . . . . . . . . . . . . . . . . 146
Conclusions ................................................ 153
References .................................................. 155

Chapter 7 • Factors Affecting the Energy-Dispersive X-Ray

Fluorescence (EDXRF) Analysis of Archaeological
Obsidian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159

M. Kathleen Davis, Thomas L.]ackson, M. Steven Shackley, Timothy Teague,

and]oachim H. Hampel

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Introduction ................................................ 160
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Results ................................................... . 167
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 78
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

Chapter 8 • Laboratory Obsidian Hydration Rates: Theory,

Method, and Application . . . . . . . . . . . . . . . . . . . . . . . 181

Christopher M. Stevenson, james]. Mazer, and Barry E. Scheetz

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
xviii CONTENTS

Principles of Glass-Water Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182

The Development of Hydration Rates at Elevated Temperature . . . . . . . . 183
Compositional Dependence of Glass Hydration . . . . . . . . . . . . . . . . . . . . 185
Measurement of the Hydration Rim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Measurement of Effective Hydration Temperature and Relative
Humidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
An Archaeological Case Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
Summary and Discussion ..................................... 201
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

Chapter 9 • Obsidian Hydration Dating at a Recent Age Obsidian

Mining Site in Papua, New Guinea . . . . . . . . . . . . . . . 205

W R. Ambrose

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
Calculated Obsidian Chronology of the GBJ Site . . . . . . . . . . . . . . . . . . . 216
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221

Chapter 10 • A 1990s Perspective on Method and Theory in

Archaeological Volcanic Glass Studies . . . . . . . . . . . 223

R. C. Green

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
A Personal Background Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
Volcanic Glass as a Category . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Spatial and Temporal Variation in "Source" Terminology . . . . . . . . . . . . . 226
Procedures and Techniques for Source Characterization . . . . . . . . . . . . . 228
Other Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Problem of Age Determination through Obsidian Hydration
Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233

About the Editor 237

Correspondence 239

Index ................................................. 241

 Chapter 1

Current Issues and Future

Directions in Archaeological
Volcanic Glass Studies
An Introduction

M. STEVEN SHACKLEY

ABSTRACT

Not since R.E. Taylor's 1976 edited volume, Advances in Obsidian Glass Studies:
Archaeological and Geochemical Perspectives has there been an attempt to update
the recent advances in archaeological obsidian studies. In these last twenty
years, the use of obsidian archaeometry has witnessed almost boundless expan-
sion, party due to the concomitant advances in computer technology and
instrumental chemistry, and partly due to a recognition by archaeologists that
archaeometrists provide much more than mere measurement. The primary
purpose here is to present the most recent advances in volcanic glass geochem-
istry and hydration analysis as a coarse gained perspective for glass studies in
archaeology for the next century. Not all the analytical techniques are used for
archaeological obsidian are discussed here, but the most commonly used instru-
mental techniques are presented in this volume.
Obsidian hydration studies have seen an equally rapid growth in the serve
of archaeological problem solving, albeit with a bit more argumentative reason-

M. STEVEN SHACKLEY • Phoebe Hearst Museum of Anthropology and Department of Anthro-

pology, University of California, Berkeley, California.

Archaeological Obsidian Studies, edited by Shackley.

Plenum Press, New York, 1998.
1
2 M. STEVEN SHACKLEY

ing. The papers focused on hydration issues in this volume present very current
methodological and theoretical studies that follow the high level of experimen-
tation typical of hydration work.
A secondary purpose of this volume overall, and probably the most
important goal, is to communicate these important advances to the users: the
archaeologist in academia and the private sector that depend on the validity
and reliability of archaeometric research often without questioning sampling
or analytic methods. For these reasons, this volume will hopefully act a guide
to the analytical techniques that are used by the archaeometrists, the sampling
methods used at the source localities in the field, and the problems faced in
both the field and the laboratory.

INTRODUCTION

In 1976, R.E. Taylor published the first comprehensive treatise on archae-

ological obsidian studies (Taylor 1976). At that time, with a few exceptions,
geochemical and hydration analyses of obsidian in the service of solving
archaeological problems was merely a decade old, and certainly not universal.
In the American Southwest, the early work by Boyer and Robinson (1956), an
interesting refractive index study of obsidian from northern New Mexico, was
essentially ignored for nearly 30 years (but see Green's chapter on its use in
New Zealand). With the exception of a few isolated but important obsidian
studies after this publication, Southwestern archaeologists continued to rely
on ceramics to deal with problems of exchange and interaction (Jack 1971;
Schreiber and Breed 1971). Even today, few archaeologists employ hydration
analyses in their research in the Southwest. In all the regions addressed in this
volume, research trajectories progressed in similar fashion. When Taylor pub-
lished the edited volume Advances in Obsidian Glass Studies: Archaeological and
Geochemical Perspectives in 1976, few of the authors in the book could have
predicted the interest that obsidian archaeometry enjoys in the last decade of
the 20th century.
In these last twenty years, the use of obsidian archaeometry has witnessed
almost boundless expansion, partly due to tHe concomitant advances in com-
puter technology and instrumental chemistry, and partly due to a recognition
by archaeologists that archaeometrists provide much more than mere measure-
ment. In the case of archaeological geochemistry, archaeologists have been
recently provided with instrumental techniques only dreamed of during the
birth of processual archaeology and its attendant emphasis on quantification
and scientizing of the discipline. In the last decade, organizations like the
Society for Archaeological Sciences (SAS) have grown tremendously after years
in the hands of a few dedicated archaeometrists and archaeologists. New
organizations devoted to specific aspects of archaeometry and geochemistry
ISSUES AND DIRECTIONS IN VOLCANIC GLASS STUDIES 3

such as the International Association for Obsidian Studies (lAOS) have arisen
and grown rapidly reflecting the truly international interest in archaeological
geochemistry and obsidian hydration studies.
This international cooperation and interest in obsidian studies is most
certainly reflected in these pages. While the focus of this volume is not to cover
every region of the world, providing laundry lists of source standard and
hydration rate data, a good portion of the globe is covered to some degree.
The primary purpose here is to present the most recent advances in
volcanic glass geochemistry and hydration analysis as a coarse grained prospec-
tive for glass studies in archaeology for the next century. Not all the analytical
techniques used for archaeological obsidian are discussed here, but the most
commonly used instrumental techniques are presented in this volume. Today
nearly every archaeologist who sends his or her samples to an analyst anywhere
on the globe will send it to a lab that uses x-ray fluorescence spectrometry
(XRF), neutron activation analysis (NAA), inductively coupled plasma-mass
spectrometry ICP-MS or PIXE/PIGME, particularly the former two. Because the
world has now shrunk to the level of instantaneous hyperspace communication
and frequent flier miles, virtually all of the volcanic glass analysts in the entire
world communicate frequently enough that the papers and authors here will
seem quite familiar, particularly to the archaeometrists.
Obsidian hydration studies have seen an equally rapid growth in the
service of archaeological problem solving, albeit with a bit more argumentative
reasoning. The papers focused on hydration issues in this volume present very
current method and theory studies that follow the high level of experimentation
typical of hydration work. One of these authors, Ambrose, was an author in
the 1976 volume and an early leader in the intrinsic hydration rate ideology.
A secondary purpose of this volume overall, and probably the most
important goal, is to communicate these important advances to the users; the
archaeologists in academia and the private sector that depend on the validity
and reliability of archaeometric research often without questioning our sam-
pling or analytic methods. For these readers, the volume will hopefully act as
a guide to the analytical techniques that are used by the archaeometrists, the
sampling methods used at the source localities in the field, and the problems
faced in both the field and the laboratory. How certain can we be that a piece
of debitage less than 7 mm in diameter and l mm thick is actually from the
source assigned by the analyst? How many samples are a minimum number to
characterize a source? How variable is obsidian source chemistry in a single
source 7 Is there an optimal method to measure hydration rims in humid
climates? All of these issues and many more are addressed in the volume
providing access to the method and theory in obsidian studies that will be
important in the next century.
In 1993, many of the authors here participated in a symposium invited
by the Society for American Archaeology and organized by the Society for
4 M. STEVEN SHACKLEY

Archaeological Sciences entitled Archaeological Science: Past Achievements/Fu-

ture Directions. At the time it seemed interesting that the SAA thought that there
was a need for such a symposium since archaeometry has not always faired well
with that venue. Inde~d. this volume is from a symposium that was rejected by
that very same annual meeting for reasons unexplained. It pointed out to many
of us that there is a rather extreme lack of communication (or even co-interest)
between what archaeologists that study stone tools think and do and what
archaeometrists who are interested in rocks think and do. Hopefully this
volume is one step toward solving this problem.

CURRENT ISSUES IN OBSIDIAN ARCHAEOMETRY

While this volume certainly will not solve all the problems currently
facing volcanic glass studies, there are a number of salient issues that continue
to strain the analytic and quantitative abilities of the archaeometrists. Most of
these issues will continue into the next century, and the authors hint at some
of the rising issues that will face 21st Century obsidian analysts. An important
point to raise at this juncture, is that most of the authors explicitly cited a
number of issues plaguing compositional and hydration analyses, without
deliberate communication between them during the construction of the vol-
ume. This could be either heartening or a matter of concern. Most probably it
is an effect of the very real issues facing the discipline, but all must be wary of
the creation of a normal science of obsidian archaeometry where analysts follow
an imaginary (or real) leader into an abyss of tautological analyses and thought.
We are hardly, however, at that point. A few of the authors actually take differing
or complementary positions on a number of issues, but all focus on the very
real need to integrate with archaeological problem solving, despite (or because
oO recent theoretical purterbations in Anglo-American archaeology (Gero and
Conkey 1991; Layton 1989; Pearce 1990; Preucell991; Redman 1991; Shanks
and McGuire 1996; Wandsnider 1992). Archaeometrists, including obsidian
analysts, are contracted by archaeologists employing emprical and post-proc-
essual frameworks, many of whom are quite theoretically hostile to one another
(see Preucell99l). Empiricists see the geochemical data or hydration meas-
urements as another aspect of scientific enquiry, while many anti-empiricists
or mildly post-processual archaeologists use the data as a means toward any
number of different ends. Statements such as "technology is not science" and
other logical inconsistencies, common among some post-processual archaeolo-
gists, would make most of the authors here recoil in horror. Given that the data
resulting from analyses such as those discussed here can be used in any number
of ways, the value of this volume is elevated. It is crucial that the issues
continuing in obsidian archaeometry be understood by an increasingly hetero-
geneous archaeology, not only so that we are all better informed, but so that
ISSUES AND DIRECTIONS IN VOLCANIC GLASS STUDIES 5

the interpretations be based on a valid and replicable science, despite that some
of the users are anti-science. Toward this goal the volume hopefully succeeds.

Systematic Approaches in Source Heterogeneity Studies

Systematic regional approaches to obsidian source characterization has
recently become increasingly important as we learn that what was previously
defined as a single source appears to be much more variable than originally
conceived. Working in any number of regions around the world, many of
the authors realize the problem and have devoloped a variety of strategies to
solve or mitigate them. Using NAA Glascock, Braswell and Cobean propose
a systematic approach to the characterization of Mesoamerican sources that
has great ramifications for archaeological problem domains. Mesoamerica
was one of the first regions of the world to take advantage of obsidian source
characterization, but this brought an early realization of the problems of
compositional analysis. Glascock et al. in this volume elucidate the weak-
nesses of previous compositional analyses, and propose a new systematic
method that increases the reliability and validity of the results that includes
systematic recording of source standard provenience and the interaction of
bivariate and multivariate statistical analyses. One important point noted by
Shackley, and implicitly by Tykot is that few archaeologists realize that the
elements useful in discriminating sources in one region may not be useful
in another, and that a handful of grab samples from a roadcut is rarely
sufficient to characterize source variability. These are issues that pervade the
chapters on chemical characterization, but all propose sampling and analytical
methods to deal with them.
Since the 1960s, obsidian procurement and distribution has been an
integral part of Mediterranean archaeological problem domains, but as Tykot
notes, has not received a detailed assessment in over 20 years (see the recent
attempt of Williams-Thorpes 1995 to do just this). Our concept of production
and distribution has changed (and improved) so much in the last several
decades, that these older models simply do not contain the explanatory ability
necessary to match current theory. Tykot focuses on these issues and applies a
systematic characterization program using ICP-MS, again focusing on the
intra-source variability inherent in so many sources of archaeological obsidian.

Secondary Deposition and Archaeological Research

The ignorance of obsidian studies in the American Southwest has brought
its own problems. Recent work to rectify this have located more than 30
sources, many over 15 million years old and subject to extreme erosion. In some
cases, the original obsidian zone on a dome or even the entire rhyolite dome is
gone (see Bouska 1993; Hughes and Smith 1993). Shackley, outlines a strategy
6 M. STEVEN SHACKLEY

to cope with issues of severe secondary distribution as well as intrasource

chemical variability in eastern Arizona and western New Mexico. As discussed
in the previous chapters, these have great ramifications for archaeological
problems of exchange, interaction, procurement and production. The character
and presence of secondary sources has only recently gained recognition as an
issue. Archaeometrists can lose sight of the very real fact that stone tool makers
are often not concerned with the location from which they procure raw
material, only that it be easily procurable (Church 1995;jackson 1989; Shack-
ley 1992, 1995; Wyckoff 1993). In Shackley's Southwestern case, obsidian
nodules are available in secondary contexts as much as 100 km from the
primary source, and in the late period, in a different cultural territory. It is no
longer enough to chemically characterize a source of obsidian by grabbing five
samples from a road cut and ignore the ultimate geological distribution of the
raw material.

Basaltic Glass
Recently, the methods used to characterize obsidian (silicic glass) have
been employed to characterize basaltic glass, particularly in Oceania. Weisler
and Clague compare the utility of EDXRF, electron microprobe, macroscopic
assignment for volcanic glasses in Oceania. Based on these experiments in this
very new territory, the authors note that what was once considered a rather
nondescript artifact class in formal terms, may, indeed, hold great value for
addressing questions of prehistoric long-distance interaction. Further, as many
have noted, the authors learned that a greater number of source samples are
needed to adequately characterize a single source, again an observation that all
the authors espouse.

Beyond Neutron Activation

While most compositional studies have relied on NAA or XRF in the
analysis of archaeological volcanic glass, Summerhayes et al. (this volume)
explore the utility of PIXE-PIGME in the service of explaining the variability of
the sources of archaeological obsidian in West New Britain, Papua New Guinea.
Like EDXRF it is a rapid, non-destructive technique perfect for the analysis of
museum specimens or where indigenous groups do not want the specimen
destroyed. As in the previous discussions, the authors and other Oceanic
researchers were finding that the data from the relatively small number of source
standards did not provide the resolution necessary to address current problems
in prehistory. In part to deal with these problems, a concerted effort was made to
increase the precision of PIXE-PIGME by controlling sample size, resolving
irradiation issues, and addressing some of the unassigned artifact issues. Their
instrumental technique promises to be a leader in future obsidian studies.
ISSUES AND DIRECTIONS IN VOLCANIC GLASS STUDIES 7

Obsidian Hydration: Problems and Prospects

As mentioned earlier, since hydration experiments deal directly with
chronology, the method has remained in rather constant contention throughout
its history. Stevenson, Mazer, and Sheets focus on some of these issues and
present some interesting solutions. A primary critique of obsidian hydration
dating is the lack of consistent comparisons with other chronometric data.
Stevenson et al. (this volume), utilizing high temperature experimentation in
the development of hydration rate constants based on intrinsic water content,
use case examples from Easter Island, Chile and Choeistaas Pueblo in Arizona.
Comparing radiocarbon and dendrochronology dates the authors are con-
vinced that the intrinsic method has merit particularly when problems of
calibration are solved.
A vexing problem in obsidian hydration in hot and humid climates is the
rapid hydration and aggressive weathering of glass surfaces. These two factors
combine to make hydration dating in the worlds tropical environments difficult
at best and impossible typically. Ambrose discusses a method to detect the
hydration rim by targeting protected fissures in the glass that have not been
subjected to these extreme weathering processes. Analyzing a data set from a
recent period mining site in Papua New Guinea, Ambrose is able to define a
chronology where radiocarbon or other chronometric techniques would be
impossible. His technique provides a viable alternative, and when combined
with long term laboratory based hydration rate determinations, promises to
increase the current value of obsidian hydration dating in all environments.

FUTURE DIRECTIONS

It could be a rather pretentious undertaking to predict the future of

obsidian archaeometry. I doubt seriously that the authors of Taylor's 1976
volume could have predicted the impact micro-computerization would have
on the field, or the issues that are presented today. Most of the problems likely
present in the near future are certainly here today, and while most were at least
delineated in this volume, a few deserve a bit more attention.

Which Instrument Is Best?

Just about the most frequently asked question by archaeology students
is: Which instrument is best to analyze my stone objects? The answer, unfor-
tunately, is: "It depends .... " Again, the problem design and the level of
precision needed to address that design will determine which instrument is the
best for a given project. The chapters in this volume present that data, but in
disparate form.
8 M. STEVEN SHACKLEY

With the general improvements in technology, instrumental techniques

in archaeological geochemistry have similarly improved. Almost all instrumen-
tal and empirical techniques have been used including density, magnetism,
atomic absorption, megascopic criteria, as well as the techniques discussed
here. As discussed above, three major instrumental methods dominate the field
today: NAA or INAA, XRF, PIXE-PIGME; with ICP-MS a close second (see Neff
and Glascock 1995). All of these methods have benefited from the revolution
in microprocessors and the attendant software explosion and are easier to use
and misuse today. There is not the space to detail the intricacies of each method
here. For archaeologists, Goffer (1980) and, to a certain extent, Harbottle
(1982) do this quite well. Glascock (1991) presents a good, though technical,
treatment of NAA (see also Neff and Glascock 1995, and Glascock et al., this
volume), and wavelength XRF and PIXE-PIGME are explained in detail in
Davis et al., this volume; and Summerhayes et al., this volume; see also jenkins
et al. 1995 for quantitative XRF applications).
There has been a certain level of mythology regarding the optimal analytic
instrument for geochemical studies of stone. The prevailing ideal seems to be
that NAA is the best way to go if funding is not a problem. NAA for most
elements is certainly accurate and it simply can detect more elements than the
other two (Neff and Glascock 1995; Glascock 1991). NAA, however, has two
primary shortcomings. It is in essence a destructive technique, and while the
material is not technically destroyed, depending on original size, the sample
may be broken into relatively small pieces and remain radioactive for many
years. Additionally, partly due to cost and partly due to the public fear of nuclear
energy, NAA is not readily available. Also for obsidian characterization, NAA
cannot analyze for Ba and Sr as accurately as the other two methods. These two
elements are important incompatibles in silicic melts and can be extremely
important in separating sources or dealing with intra-source issues (Hildreth
1981; Hughes and Smith 1993; Macdonald et al. 1987; Mahood and Hildreth
1983). NAA can, however, analyze for a number of other elements with great
precision.
For museum specimens and artifacts that are subject to repatriation, NAA
is not a feasible choice. If precision and accuracy are necessary, which can be
an issue in intra-source studies and where the possibility of long-distance
exchange is probable, then NAA will always provide the most efficient alterna-
tive. Additionally, if the sample is extremely small ( < 5 mm in largest diameter),
NAA may be the best analytical method. Other methods may require a mini-
mum sample size for optimal results (see Davis et al. this volume).
For a number of years now, Michael Glascock at the University of
Missouri Research Reactor Facility using NAA and Shackley at the Berkeley
Archaeological EDXRF Lab have been analyzing the same obsidian source
standards and artifacts simply as part of routine data sharing. It is remarkable
how similar the results of the two analyses are, for the 17 or more elements
ISSUES AND DIRECTIONS IN VOLCANIC GLASS STUDIES 9

measured in common. Hughes (1988) in his paper on the chemical variability

in the Coso Volcanic Field also obtained similar results in the mid-Z incompat-
ible elements between NAA and EDXRE Much of this is due to the extreme
accuracy of EDXRF in the mid-Z (and part of the high-Z) region which
fortunately contains some of the most sensitive incompatible elements for
discriminating obsidian sources, including Rb, Sr, Y, Zr, Nb , and Ba (see Giaque
et al. 1993) . Table 1.1 and Figure 1.1 exhibits the results from both instruments
for source standard splits from the Obsidian Butte source, a Quaternary glass
source located in Imperial Valley, southern California (see Hughes 1986).
Indeed, Hughes earlier analysis of the same elements is statistically similar.
Much of this paired accuracy is due to calibration to international standards
and the increasing analytical precision of todays instruments.What this all
means for an archaeologist is that any laboratory employing EDXRF, WDXRF,
NAA, PIXE-PIGME, or ICP-MS will provide valid and comparable results, given
a sufficient sample size, particularly for obsidian. The particular project con-
straints are now the main criteria for the selection of analytical methods.
Requiring that the laboratory supply analyses of international standards so that
user can assess instrumental precision should be part of the archaeologist's
evaluation of the data.

.,. .. ..•
~
I ~

--'"
• •

lj •
I •• ••
·--- --..-- r
~ - ·- •
I •
•

-~ .;..,
rl'
•
. •
~~
•• .&Thppm

..."- -
- -
500
400 •• LABORATORY
300 1 ••• • M=Missouri
200
100 j
0 • B=Berkeley

Figure l.l. Multivariate matrix plot of five measured elements byBerkeley (EDXRF) and Missouri
(NAA) labs for Obsidian Butte, California source standards taken from Table l.
 ......
c
Table 1.1. Quantitative Analysis of Source Standard Sample Splits from Obsidian Butte, California Analyzed by EDXRF
at Berkeley (B Prefixes) and by NAA at the University of Missouri Research Reactor Facility (M Prefixes)"

SAMPLE Mn Fe Zn Th Rb Sr Zr Ba La Ce Nd

BOB1 359.58 15916.91 67.66 27.24 140.11 19.40 304.86 491.90 80.42 168.83 97.36
MOB1 361.20 15467.40 56.40 19.63 135.40 0.00 325.20 483.10 54.17 123.85 63.68

BOB2 482.94 22987.09 70.73 21.02 132.31 53.72 458.93 586.22 59.43 129.78 72.96
MOB2 420.20 19932.20 62.30 17.74 123.00 0.00 473.50 531.00 43.54 98.18 51.05

BOB3 379.67 19659.40 67.35 21.65 120.93 48.81 427.46 624.12 65.36 138.85 65.19
MOB3 429.80 19769.80 45.30 17.83 121.00 20.10 468.60 518.00 44.68 99.61 55.95

BOB4 447.34 19953.92 79.36 30.83 152.94 31.24 355.47 528.54 73.77 145.60 82.83
MOB4 376.40 15969.80 52.80 18.90 130.20 0.00 349.70 467.50 50.89 114.55 45.82

BOB6 484.79 18983.53 83.71 29.87 152.35 26.07 342.78 490.20 85.05 run run
MOB6 357.00 14887.10 50.30 19.55 135.70 0.00 322.50 394.70 54.47 122.88 66.51

BOB? 459.91 21773.29 69.09 25.01 132.24 53.94 434.85 641.50 57.21 run run
MOB? 434.00 20087.70 59.80 17.85 119.20 0.00 495.70 575.10 43.75 98.36 54.39
~
BOB9 446.90 22111.66 67.57 22.37 137.63 53.91 446.03 646.63 run run run ~
MOB9 425.00 20200.10 59.70 18.00 122.10 28.10 474.30 539.70 44.60 100.28 53.81 i;l
BOB10 419.28 22161.28 64.98 20.31 131.00 54.82 441.86 604.41 63.57 131.32 65.20
MOB10 413.70 19444.90 56.30 17.44 118.30 0.00 450.20 524.50 43.82 96.52 52.91
~
~

aAllmeasurements in parts per million (ppm). Calculations of ppm from weight percent based on formula in Glascock ( 1991 ). Instrumental methods for EDXRF analysis discussed in
~
Shackley (1995): forNAA in Neff and Glascock (1995), Glascock (1991) and Glascock el a/. (this volume). nm =not measured.
s
~
ISSUES AND DIRECTIONS IN VOLCANIC GLASS STUDIES 11

Source Sampling and Secondary Depositional Effects

In the last two decades we have compiled a very complete database of

archaeological obsidian. While there are still regions that are not covered well
enough (i.e., East Asia, and South America) the world of unknown sources of
archaeological obsidian is rapidly shrinking. Based on the discussions in this
volume, it is apparent that the focus now is toward refining sampling at the
source, including the consistent mapping of any secondary depositional envi-
ronments. Granted, some Quaternary sources, and some environments do not
indicate erosional problems (Shackley et al. 1996; Summerhayes et al. this
volume), but many do, and consequently require transect and more intensive
survey. The advances in GIS will certainly aid these investigations in the future,
but ground truthing will always remain crucial in this endeavor. It is no
accident that many of the sources reported in Taylor (1976) now are charac-
terized by any number of sub-sources or even have been determined to be more
than one source.

Quantitative Methods and Data Reduction

While many of the chapters in this volume employ inferential and
multivariate statistical methods in the interpretation of geochemical and hy-
dration data, the sub-discipline as a whole has not critically evaluated these
methods (cf., Baxter 1994). The use of stepwise discriminant analysis between
two or more pre-determined groups is quite common, and is similarly reflected
in these pages (see Glascock et al. and Shackley, this volume). Unfortunately,
this is well known to produce over optimistic results, particularly when
numerical transformations of the raw data are implemented (Baxter 1994;
Burton and Simon 1996). This self-reflection appears to be occurring in ceramic
archaeometry, but has not yet appeared in obsidian studies, although it most
certainly should (Arnold et al. 1991; Bishop et a!. 1990; Neff 1994; Neff et al.
1988). Most recently, ceramic studies in the Aegean and the American South-
west have begun to question the assumption that group designations based on
multivariate statistical analyses (i.e. cluster, factor/principal components, and
discriminant) are necessarily the most perspicacious method to discriminate
groups (Day 1992; Day et al. 1996; Tsolakidou et al. 1996). In the Aegean case,
petrographic analyses of source clay, ethnographic studies of potter's behavior,
and comparisons between petrographic slides of archaeological pottery and the
sources, indicated that multivariate groupings of NAA data were most often
incorrect with respect to both source provenance and the production centers
(see especially Tsolakidou et al. 1996). While the problems inherent in the
heterogeneity of pottery are not an issue with obsidian, the possibility of
misassignment using exclusively multivariate analyses may be just as problem-
atical. While the studies in this volume commonly use a combination of
12 M. STEVEN SHACKLEY

bivariate plots and multivariate analyses to assign artifacts to source, it is clear

that in many cases the bivariate plots may be a more accurate reflection of
source heterogeneity, as well as a better media for source assignment (see Baxter
1992). Glascock et al. (this volume) feel quite comfortable with normalizing
data by normal or other log transformations, elimination of outliers and
reanalysis, and finally relying on cross-validation (jack knifing), although there
has been very little critical examination of this technique, particularly in
geochemistry. It is quite possible that, unlike the ceramic issues, this analytical
trajectory will be found to be quite adequate and parsimonious. Quantification
and data reduction will certainly be a topic of discussion in the next century;
possibly the most important topic.

World Wide Database and the Future of Obsidian Studies

At the time of this writing (1996), the International Association for
Obsidian Studies (lAOS) had established a database for all the known sources
of archaeological obsidian in the world on the net <http://www.peak.org/-
skinncr/s_home.htmb as an accessible web site. Given the huge endeavor, it
is woefully incomplete, but with continued input from researchers, this data-
base will allow anyone to access source descriptions, images, and data. The site
also serves as a forum for discussions on all the issues, discussed here, and
more. Regardless of an individual's personal feelings about this form of com-
munication, some version of it will be with the world for decades to come.
Interactive databases and sites for many aspects of science, including course
work, allow individuals from anywhere on earth with access to the web, the
ability to communicate on any issue. This more than any other factor will
condition the way we do, and especially, communicate science and obsidian
research. It is a reasonable prediction that the next "book" on obsidian research
will be electronic; and what international subject is more appropriate than the
scientific analysis of obsidian?

REFERENCES
Arnold, D.E., Neff, E.H., and Bishop, R.L. 1991 Compositional analysis and "sources" of pottery:
and ethnoarchaeological approach. American Anthropologist 93: 70-90.
Baxter, M.]. 1992 Archaeological uses of the biplota neglected technique? In Lock, G., and Moffet
] . eds., Computer Applications and Quantitative Methods in Archaeology, 1991, pp. 141-148.
BAR International Series S577. Oxford, British Archaeological Reports.
_ _ 1994 Stepwise discriminant analysis in archaeometry: a critique. journal of Archaeological
Science 21: 659-666.
Bishop, R.L., Cannouts, B., Crown, P.L., and DeAtley, S.P. 1990 Sensitivity, precision, and accuracy:
their roles in ceramic compositional data bases. American Antiquity 55: 537-546.
Boyer, W.W., and Robinson, P. 1956 Obsidian artifacts of northwestern New Mexico and their
correlation with source material. El Palacio 63: 333-345.
ISSUES AND DIRECTIONS IN VOLCANIC GLASS STUDIES 13

Bouska, V. 1993 Natural Glasses. New York, Ellis Horwood.

Burton, ].H., and Simon, A. W 1996 A pot is not a rock: a reply to Neff, Glascock, Bishop, and
Blackman. American Antiquity 61: 405-413.
Church, T. 1995 Comment on neutron activation analysis of stone from the Chadron formation
and a Clovis site on the Great Plains by Hoard et al. (1992).]ournal of Archaeological Science
22: 1-5.
Day, P.M. 1992 Ceramic production and distribution in Late Bronze Age East Crete: a study by
petrographic analysis. Paper presented at the 28'h International Symposium on Ar-
chaeometry, Los Angeles.
Day, P.M., Brodie, N.j., and Kilikoglou, V. 1996 The presumption of local provenance and other
deadly sins. Paper presented at the 30'h International Symposium on Archaeometry, Urbana.
Gero, j.M., and Conkey, M.W, eds. 1991 Engendering Archaeology: Women and Prehistory. Oxford,
Basil Blackwell, Ltd.
Giaque, R.D., Asaro, E, Stross, EH., and Hester, T.R. 1993 High-precision non-destructive X-ray
fluorescence method applicable to establishing the provenance of obsidian artifacts. X-Ray
Spectrometry 22: 44-53.
Glascock, M.D. 1991 Tables for Neutron Activation Analysis (3rd Ed.). Columbia, Research Reactor
Facility, University of Missouri.
Goffer, Z. 1980 Archaeological Chemistry: A Sourcebook on the Applications of Chemistry to Archae-
ology. New York, john Wiley and Sons.
Harbottle, G. 1982 Chemical characterization in archaeology. In Ericson,].E. and Earle, T.K., eds.,
Contexts for Prehistoric Exchange. New York, Academic Press: 13-51.
Hildreth, W 1981 Gradients in silicic magma chambers: Implications for lithospheric magmatism.
journal of Geophysical Research 86: 10153-10192.
Hughes, R. E., ed. 1984 Obsidian Studies in the Great Basin. Berkeley, Contributions of the University
of California Archaeological Research Facility 45.
Hughes, R. E. 1988 The Coso Volcanic Field Reexamined: Implications for obsidian sourcing and
hydration dating research. Geoarchaeology 3: 253-265.
Hughes, R. E., and Smith, R. L. 1993 Archaeology, geology, and geochemistry in obsidian
provenance studies.Inj.K. Stein,j.K and Linse, A.R., eds., Effects of Scale on Archaeological
and Geoscientific Perspectives. Geological Society of America Special Paper 283: 79-91.
jack, Robert N. 1971 The source of obsidian artifacts in northern Arizona. Plateau 43: 103-114.
jackson, T.L. 1989 Late prehistoric obsidian production and exchange in the North Coast Ranges,
California. In Hughes, R.E. ed., Current Directions in California Obsidian Studies. Berkeley,
Contributions of the University of California Archaeological Research Facility 48: 79-94.
jenkins, R., Gould, R.W, and Gedcke, D. 1995 Quantitative X-Ray Spectrometry. New York, Marcel
Dekker.
Layton, R., ed. 1989 Who Needs the Past?: Indigenous Values in Archaeology. London, Unwin Hyman.
MacDonald, R., G.R. Davies, C.M. Bliss, P.T. Leat, D.K. Bailey, and R.L. Smith 1987 Geochemistry
of high-silica peralkaline rhyolites, Naivasha, Kenya Rift Valley. journal of Petrology 28:
979-1008.
Mahood, G. A., and Hildreth, W 1983 Large partition coefficients for trace elements in high-silica
rhyolites. Geochimica et Cosmochimica Acta 4 7: 11-30.
Neff, H. 1994 RQ-mode principal components analysis of ceramic compositional data. Ar-
chaeometry 36: ll5-130.
Neff, H., Bishop, R.L., and Sayre, E.V. 1988 A simulation approach to the problem of tempering in
compositional studies of archaeological ceramics. journal of Archaeological Science 15:
159-172.
Neff, H., and Glascock, M.D. 1995 The state of nuclear archaeology in North America. journal of
Radioanalytical and Nuclear Chemistry 196: 275-286.
Pearce, S. 1990 Archaeological Curatorship. Washington, DC, Smithsonian Institution Press.
14 M. STEVEN SHACKLEY

Preucel, R.W, ed. 1991 Processual and Postprocessual Archaeologies: Multiple Ways of Knowing the
Past. Carbondale, Illinois, Center for Archaeological Investigations, Occasional Paper 10:
1-324.
Redman, C.L. 1991 In defense of the seventies. American Anthropologist 93: 295-307.
Schreiber, john P. and Breed, William]. 1971 Obsidian localities in the San Francisco Volcanic
Field, Arizona. Plateau 43: 115-119.
Shackley, M. S. 1992 The Upper Gila river gravels as an archaeological obsidian source region:
Implications for models of exchange and interaction. Geoarchaeology 7: 315-326.
_ _ 1995 Sources of archaeological obsidian in the greater American Southwest: An update and
quantitative analysis. American Antiquity 60: 531-551.
Shackley, M.S., Hyland, JR., and de Ia Luz Gutierrez, M. 1996 Mass production and procurement
at Valle del Azufre: a unique archaeological obsidian source in Baja California Sur. American
Antiquity 61: 718-731.
Shanks, M., and McGuire, R.H. 1996 The craft of archaeology. American Antiquity 61: 75-88.
Taylor, R.E. ed. 1976 Advances in Obsidian Glass Studies: Archaeological and Geochemical Perspec-
tives. Park Ridge, New jersey, Noyes Press.
Tsolakidou, A., Day, P.M., Kiriatzi, E., and Kilikoglou, V 1996 Group therapy in Crete: a comparison
between analyses by NAA and petrographic thin section of Early Bronze Age Pottery from
Knossos. Paper presented at the 30'h International Symposium on Archaeometry, Urbana.
Wandsnider, L. 1992 Quandries and Quests: Visions of Archaeology~ Future. Carbondale, Illinois,
Center for Archaeological Investigations, Occasional Paper 20: 1-273.
Williams-Thorpe, 0. 1995 Obsidian in the Mediterranean and the Near East: a provenancing
success story. Archaeometry 37: 217-248.
Wyckoff, D.G. 1993 Gravel sources of knappable alibates silicified dolomite. Geoarchaeology 8:
35-58.
 Chapter 2

A Systematic Approach
to Obsidian Source
Characterization
MICHAEL D. GLASCOCK, GEOFFREY E. BRASWELL, AND

RoBERT H. CoBEAN

ABSTRACT

A systematic approach to sample collection, chemical analysis, and statistical

evaluation of obsidian sources is recommended before significant numbers of
artifacts should be analyzed. Multivariate statistical methods based on Maha-
lanobis D2 enable robust rejection of erroneous source assignments and lead to
sourcing of artifacts with a very high degree of confidence. These multivariate
procedures can also assist in identifying and evaluating abbreviated analytical
methods that are more rapid and less expensive. Test cases based on obsidian
data from sources in Mesoamerica are presented.

INTRODUCTION

Archaeologists seeking to reconstruct past human behavior and interac-

tion often examine the materials that early peoples made and used. Lithic

MICHAEL D. GLASCOCK • Research Reactor Facility, University of Missouri, Columbia, Mis-

souri. GEOFFREY E. BRASWELL • Department of Anthropology, State University of New York
at Buffalo, Buffalo, New York. ROBERT H. COBEAN • Instituto Nacional de Antropologia e
Historia, Mexico City, Mexico.

Archaeological Obsidian Studies, edited by Shackley.

Plenum Press, New York, 1998.
15
16 MICHAEL D. GLASCOCK et al.

materials are undoubtedly the most widely studied of all archaeological mate-
rials because: ( 1) they are found practically everywhere that prehistoric peoples
lived or traveled; (2) they are well preserved over time; and (3) they occupied
an important role in the material culture of most prehistoric peoples. Although
nearly every type of stone was used by one prehistoric culture or another, the
lithic material providing archaeologists with the most definitive evidence
concerning contact between different cultures is the volcanic glass obsidian.
Obsidian, with a composition similar to rhyolite, is formed when highly
viscous molten lava cools rapidly such that the process of crystallization is
precluded. The glass is usually black or gray in color (although other colors
are possible) and is sometimes banded or streaky in appearance. As a glass,
obsidian has physical properties that readily yield sharp-edged, conchoidal-
shaped fractures. Its glassy consistency, predictable fracturing qualities, and
sharp working edge contributed to obsidian being a highly sought-after mate-
rial. Obsidian generally makes better edges than most other types of tool-mak-
ing stones. In regions where obsidian was available, it was often the most
commonly used material for making sharp-edged implements, weapons, and
jewelry. Obsidian raw materials and finished products were valuable commodi-
ties that were often exchanged over long distances from their sources.
There are two types of obsidian sources: primary and secondary. Primary
sources consist of lava flows or pyroclastic bomblets surrounding volcanic
cones. Secondary sources (or deposits) are the result of the erosional activities
of glaciers, streams, gravity or other geologic processes that involve the
transport of primary material. Due to the abundance of raw material, most
prehistoric quarries were located near the primary sources.
Obsidian artifacts have been used to examine resource procurement
patterns, manufacturing processes, distribution networks, and to develop
use-site chronologies (Elam 1993). In addition, hydration-rim measurements
of obsidian artifacts can be used to investigate chronological changes in
socioeconomic conditions (Freter 1992).
Clark (1981) identifies several attributes of obsidian that give it a unique
role in archaeology:
1. The number of sources is limited. Because nearly all pre-Tertiary obsidi-
ans have lost their glassy properties, obsidian is generally restricted
to the tectonic regions that experienced vulcanism since the begin-
ning of the Tertiary period. In addition, not all lava flows are suffi-
ciently rich in silica (e.g., Hawaii) for glass formation to take place.
The most important obsidian bearing regions are the Mediterranean,
eastern Africa, central Mexico, South America, western North Amer-
ica, japan, and New Zealand.
2. Obsidian objects are found in a variety of locations and contexts. Obsid-
ian artifacts are found in a much wider distribution area than that of
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 17

their sources. Obsidian artifacts are found in many villages, rural

locations, elite areas, and ceremonial sites that are often hundreds of
kilometers removed from their originating sources.
3. Vast amounts of debitage are found near the ancient quarries. The large
amounts of waste obsidian found near sources help archaeologists
locate prehistoric quarries and make it possible to estimate the
quantities of raw obsidian excavated and exchanged.
4. Artifacts retain many of their physical features from fabrication. Scars,
platform angles, and other distinguishing marks on artifacts help
archaeologists to investigate the production techniques, the skills of
prehistoric craftsmen, and the probable functions of the tools.
5. In most archaeological contexts, obsidian artifacts are indestructible.
Thus, archaeologists have confidence that artifacts they are investi-
gating closely resemble the tools or trinkets that prehistoric peoples
were using before they were either cached, lost, or discarded.
6. The fragile nature of obsidian means that it had a high replacement rate.
Examining the quantities of artifacts at use sites and the rate of
consumption provides information through which prehistoric eco-
nomic conditions can be estimated.
7. The chemical composition of obsidian at any particular source or flow is,
with few exceptions, homogeneous and different sources or flows are
compositionally different from each other. The fortunate circumstance
that obsidian, by virtue of having been molten during the period of
volcanic activity; is well mixed material means that sources are
typically homogeneous. The homogeneous character of most obsid-
ian sources and the fact that different sources are customarily dissimi-
lar in composition is a fundamental attribute of obsidian that enables
successful differentiation between sources.
8. A new obsidian surface will begin absorbing water gradually with time.
The hydration process offers archaeologists a technique, through
microscopic examination of obsidian thin sections, by which artifacts
can be dated and compared to one another to learn more about the
chronology of an archaeological site. The only qualification to this is
that because obsidian from chemically dissimilar sources and sub-
sources hydrates at different rates, each artifact must be traced to its
source before its date of manufacture can be determined.
In order to take advantage of these and other attributes of obsidian for
archaeological examination, archaeologists have explored a variety of methods
to characterize obsidian for sourcing purposes. For archaeological purposes,
adequate characterization entails identifying a property or set of properties that
are the same for obsidian specimens from the same source and different between
specimens from different sources. In other words, archaeologists seek to
18 MICHAEL D. GLASCOCK et al.

observe or measure attributes of obsidian that are both homogeneous within

sources and heterogeneous between them.

SOURCING METHODS

Archaeologists have long sought methods for obsidian sourcing that are
rapid, reliable, non-destructive, and low-cost. As a result, a variety of methods
have been explored to identify characteristics of obsidian that can be employed
to associate artifacts with sources.
Visual appearance is very likely one of the first properties of obsidian to
have been investigated as a method for sourcing (Fuller 1927). It is possible to
describe obsidian in terms of several visual traits, including: color in transmit-
ted light, color in reflected light, refractive index, success of fracture, opacity,
internal structure, and luster. However, appearance alone is not a reliable guide
by which the source of obsidian can be determined. Because assessment of the
visual attributes of individual obsidian specimens is a consequence of one's
subjective judgement, appearance is often incapable of being employed reliably
by different individuals as a method for obsidian differentiation. Assorted
colors (black, gray, clear, reddish-black, reddish-brown) and various degrees of
banding or mottling often occur within a single obsidian deposit. In spite of
these inadequacies, many modern-day archaeologists still employ visual meth-
ods with marginal to moderate success (Bettinger et al. 1984; Clark 1988;
Moholy-Nagy and Nelson 1990;Jackson and Love 1991). In general, only when
supported by other types of analyses can meaningful statements be made about
obsidian materials on the basis of appearance.
Several other previously investigated methods for obsidian sourcing
include: ( 1) density measurements (Reeves and Armitage 1973); (2) thermolu-
minescence (Huntley and Bailey 1978); (3) Mossbauer spectroscopy (Long-
worth and Warren 1979); (4) fission-track analysis (Duranni et al. 1971); (5)
measurement of magnetic properties (McDougall et al. 1983); and (6) meas-
urement of natural radioactivity (Leach et al. 1983). Although some of these
methods identified systematic differences between sources, the overlap be-
tween different sources is such that their reliability for differentiation purposes
is low, and none show much promise.

Compositional Analysis
The chemical composition of most volcanic obsidian ranges from about
70-75% Si0 2 , 10-15% Alz0 3 , 3-5% Nap, 2-5% KzO and 1-5% total Fez0 3
+ FeO. Peralkaline obsidians are typically higher in Fe composition than
rhyolitic obsidians. In addition, the intrinsic water content of obsidian ranges
from 0.1-0.5% and water content increases to about 3.5% by weight as
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 19

rhyolitic obsidian gradually transforms into a less useful form of the glass
known as perlite. Most of the remaining elements in obsidian are present in
concentrations of much less than 1% and are therefore referred to as trace
elements.
The success of chemical sourcing is due to the fact that these trace
element compositions may differ by one or two orders of magnitude between
sources, while within-source variation is usually much smaller. In a few cases,
within-source variation in elemental composition is considerable (e.g., Bow-
man et al. 1973). Nevertheless, the correlation in variation between certain
elements is so extraordinary that the ability to assign artifacts to the source is
just as certain as if the flow were homogeneous. Other sources have been
discovered where a collection of individually homogeneous and discrete flows
have been reported (e.g., Hughes 1988). Obsidian artifacts from these sources
can be assigned to specific subsources.
Over the past three decades, many analytical techniques have been used
to chemically source obsidian artifacts. These methods include: (1) optical
emission spectroscopy (OES), used to examine obsidian from the Mediterra-
nean (Cann and Renfrew 1964); (2) atomic absorption spectroscopy (AAS),
used to characterize obsidian from Alaska (Wheeler and Clark 1977); (3)
particle-induced X-ray emission spectroscopy (PIXE), used to study obsidian
from New Zealand (Duerden et al. 1984); (4) inductively-coupled plasma
emission spectroscopy (ICPES), used to analyze obsidian from New Mexico
(Stevenson and McCurry 1990); (5) X-ray fluorescence spectroscopy (XRF),
used to characterize obsidian from northern California (Hughes 1982); and (6)
neutron activation analysis (NAA), used to study obsidian from the highlands
of Guatemala (Asaro et al. 1978).
In order to be successful, a technique for chemical fingerprinting must
be quantitative, capable of simultaneously measuring several elements, sensi-
tive to the elements of interest, independent of sample matrix, and independent
of artifact size and shape. In addition, the choice of a method for analysis may
depend upon its availability, cost, speed, accuracy, ability to differentiate
between sources, existence of comparative data, etc. XRF and NAA have proven
to be highly cost effective and, therefore, are the methods most frequently used
to source artifacts.
In XRF, a beam of X-rays irradiates the specimen causing displacement
of atomic electrons from the inner energy levels. As electrons from outer
(higher) levels repopulate these vacant inner (lower) levels, energy is emitted
in the form of fluorescent X-rays. Because the energy levels are unique for each
element, these X- rays have characteristic energies that permit identification of
the element. By measuring the intensities of emitted X-rays, one can determine
the quantities of elements present in the specimen. In obsidian, the elemental
compositions of Na, K, Ti, Mn, Fe, Rb, Sr, Y, Zr, Nb, and Ba are frequently
determined by XRF
20 MICHAEL D. GLASCOCK et al.

The principles of NAA differ from XRF in that samples are irradiated by
thermal neutrons (usually from a nuclear reactor). During irradiation, a few
neutrons are captured by the nuclei of atoms in the specimen. This process,
called activation, causes some of the nuclei to become unstable. During and
after neutron irradiation, these unstable nuclei emit gamma rays with unique
energies at rates defined by the characteristic half-lives of the radioactive nuclei.
Identification of the radioactive nucleus is possible by measuring the gamma-
ray energies. Determination of their intensities permits quantitative analysis of
the elements in the sample. Use of NAA on obsidian specimens by a combina-
tion of two or three irradiation, decay, and measurement strategies enables
determination of the elements Ba, Ce, Cl, Co, Cs, Dy, Eu, Fe, Hf, K, La, Lu, Mn,
Na, Nd, Rb, Sb, Sc, Sm, Sr, Ta, Tb, Th, U, Yb, Zn, and Zr. If necessary, the
sensitivities of different subsets of these elements can be enhanced by changing
the irradiation, decay, and measurement schemes.
In general, XRF is more available, more rapid, and less expensive than
NAA. NAA offers a far greater number of elements, more sensitivity, superior
precision, and greater accuracy than XRE In XRF, the artifact can be studied
without destruction. On the other hand, NAA can be performed on extremely
small samples (-10 mg), meaning that only minor damage to valuable artifacts
may be required.

PREVIOUS WEAKNESSES IN INTERPRETING

COMPOSITIONAL DATA

A review of previous obsidian research reveals that, in some studies, only

a fraction of the information potentially available was recovered. In other
instances, poorly planned studies came to conclusions that were later refuted.
Most problems involved the unsatisfactory characterization of sources (i.e.,
failure to locate all possible sources, collection and analysis of too few speci-
mens from each source, poor descriptions of source areas, analysis of samples
from modern road cuts rather than prehistoric quarries, analysis of too-few
elements, and failure to identify the most critical elements useful for discrimi-
nating between particular regions or individual sources).
Another fallacy is the anticipation by a few archaeologists that the same
elements successful in differentiating the sources in one region will succeed
in others. Unfortunately, one cannot know in advance which elements will
be best or which analytical technique will be most efficient for a particular
region until after the sources have been thoroughly characterized. Other
problems pointed out by Hughes (1984) are due to the uncritical acceptance
of the results of artifact sourcing studies analyzed by inadequate statistical
procedures. These include various analytical or statistical shortcuts with
unproven reliability.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 21

Although many thousands of obsidian-source and artifact analyses have

been made throughout the world, no single databank exists that makes
available all of the compositional and descriptive data. Instead, individual
research groups have concentrated on sources in particular regions, some-
times in complete isolation from other groups working in the same or
neighboring regions. The use of different analytical calibrations by different
analytical laboratories makes data exchange problematic. As a result, ex-
changes of data and source samples between laboratories are a rare rather
than routine occurrence.
One example illustrating several of these problems is the NAA work of
Pires-Ferriera (1973) which dealt with a large collection of obsidian artifacts
found in the Valley of Oaxaca. In her study, Pires-Ferriera employed a procedure
that measured two short-lived elements (i.e., Mn and Na). Although her work
was mostly successful, Pires-Ferriera assigned a large fraction of artifacts to the
source at Altotonga which more recent data calls into question. From more
recent work reported by Cobean et al. (1991), we now realize that Altotonga
obsidian overlaps perfectly with the nearby Zaragoza source on the basis of Mn
and Na. However, both sources are easily differentiated by a number of other
elements measured in the Cobean et al. (1991) study. Recent research by Elam
(1993), reporting six elements from an analysis of -500 obsidian artifacts from
the Valley of Oaxaca, identifies Zaragoza as one of the most important source
supplying obsidian to Oaxaca. Not a single artifact in this recent study was
traced to Altotonga. Pires-Ferriera also suggested that several unassigned
artifacts came from an unknown Oaxacan source that she was unable to locate.
The existence of this hypothetical source is disputed by more comprehensive
data (Glascock et al. 1988) indicating that the unassigned artifacts very likely
came from sources in the state of Hidalgo that Pires-Ferriera did not adequately
characterize. Thus, on the basis of minimal data, questionable source assign-
ments for obsidian artifacts from Oaxaca were accepted for nearly twenty years.
A study of obsidian artifacts from San Lorenzo Tenochtitlan, Veracruz by
Co bean et al. (1971) also reported the similarity of the Otumba, Mexico source
to sources in Guatemala (over 1,000 km to the south) for elements commonly
measured by XRE Later, Charlton et al. (1978), using a small suite of elements,
found chemical similarity for the obsidian source at Paredon (-50 km from
Otumba) with the Otumba source. A larger suite of elements (Co bean et al.
1991) now enables identification of inter- and intra-regional differences be-
tween these and other sources. As a result, it is quite probable that the artifacts
identified as the Unknown A group in Co bean's earlier study were actually made
of Paredon obsidian.
Other examples of studies with inadequate source information involve
work in the Andean region of South America. There, hundreds of artifacts have
been analyzed and results suggest the existence of 25 or more sources (Avila-
Salinas 1975; Burger and Asaro 1977). Thus far, less than ten sources have been
22 MICHAEL D. GLASCOCK et al.

spatially located and chemically characterized in the entire Andes region. Thus,
a great deal of work still remains before substantial reconstruction of the
obsidian exchange networks in South America will be possible.
As Hughes (1986) points out, some researchers (e.g., Sappington 1981,
1984) have failed to consider the significance of particular elements and,
instead, assigned equal weight to all measured elements. Unfortunately, this
practice may result in an increased number of misclassifications. Hughes
suggests that a more critical approach to variable selection in multivariate
applications can reduce the number of errors. Finally, Leach and Manly ( 1982)
argue that the power of any sourcing algorithm to reject wrong answers should
be demonstrated.

A MORE SYSTEMATIC APPROACH

In order to make reliable source attributions, researchers must: (1) locate

each source precisely and obtain a representative range of source specimens
from all sources; (2) demonstrate that flake-quality obsidian can be obtained
from the source; (3) verify that the source was accessible to prehistoric peoples
and it was not exposed by more recent (historical period) mining or construc-
tion activities; (4) locate prehistoric quarries or identify materials at an archae-
ological site that originated from the primary source area or a secondary
deposit; (5) analyze the source specimens thoroughly to determine the number
of different chemical fingerprints represented; (6) correlate the source finger-
prints to specific source locations; and (7) show that artifacts assigned to a
particular source cannot be mistaken for those from another source. A system-
atic program of sampling, chemical analysis, and statistical evaluation is
required to accomplish these goals.

Sampling and Chemical Analysis

Before the benefits of artifact sourcing can be fully realized, it is necessary
to locate and sample all likely obsidian sources in the region(s) of interest. The
geographic extent of each primary source and the areal distributions of secon-
dary deposits should be determined. Intensive field surveys may be required to
determine the locations of obsidian outcrops and prehistoric quarries showing
evidence of utilization by ancient man. Use of a global positioning system ( GPS)
is highly recommended while conducting the survey and collecting specimens
for analysis, especially in those areas where maps are deficient. In addition,
because different archaeologists may have used various names to describe the
same source, clarification of name(s) with geographic coordinates is necessary
to avoid confusion by laboratory analysts and others who are unlikely to visit
the source.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 23

The number of source specimens collected and analyzed should be

sufficient to enable determination of the range of compositional variation in
the source. In general, we recommend that at least a dozen source samples be
analyzed, although for large and complex sources more than 100 specimens
may be desirable. If possible, specimens should be taken from the actual
locations of prehistoric quarries rather than from road cuts or other sites that
are the consequence of modern human activity. Additional specimens should
be collected beyond those slated for analysis in order to maintain samples for
future reference, exchange with other laboratories, and conduct possible hy-
dration-rate experiments. If the amount of compositional variation in particu-
lar sources proves to be greater than anticipated or if artifact analyses indicate
the existence of undiscovered sources, additional field surveys and collections
should performed (Shackley 1988).
A comprehensive data bank or source catalog should be compiled to
store all of the descriptive, physical, and chemical information gathered.
Previous efforts by Skinner (1983) and Hughes (1986) were an initial step
in this direction. More recently, members of the International Association for
Obsidian Studies (lAOS) have started collecting information for a comput-
erized database that would organize all known descriptive and chemical
information on obsidian sources. Informational categories urged for each
source include: (l) geographic location and name(s) of source; (2) type of
source (either primary or secondary); (3) geologic setting of the source area;
(4) evidence of quarrying or chipping activity and the quality of the obsidian
for tool making; (5) visual characteristics (both macroscopic and micro-
scopic) of the glass; (6) refractive index of the glass; (7) age of source and
method of determination (e.g., K-Ar, fission-track); (8) source specimens
were collected from flows, bomblets, or nodules; (9) names of persons who
are knowledgeable of the source area; (10) evidence from use sites of artifacts
that came from the source area; (11) types of chemical analyses performed;
(12) a tabulation of all compositional data (not only averages); (13) results
from hydration rate measurement investigations; (14) a listing of relevant
publications.
Source specimens should be analyzed comprehensively by all available
analytical techniques (preferably both NAA and XRF) and in sufficient num-
bers to establish the concentration ranges, means, and standard deviations for
each source. Use of well-known standards such as the National Institute of
Standards and Technology (NIST) SRM-278 Obsidian Rock and SRM-1633a Fly
Ash are recommended for calibration (Graham et al. 1982). In addition, routine
quality-control samples should be analyzed and rigorously monitored between
analytical runs in order to certify that the data are as accurate and precise as
reasonably possible (Glascock and Anderson 1993). These steps will insure
that data from separate batches and different laboratories are comparable and
bias is minimized.
24 MICHAEL D. GLASCOCK et al.

Statistical Evaluation of Obsidian Source and Artifact Data

Most researchers recognize the wisdom of collecting many source speci-
mens and analyzing for several elements to achieve a statistically significant
differentiation between two sources. Unfortunately, as Ward (1974) and, more
recently, Hughes (1986) have pointed out, it is difficult to visualize more than
two or three elements at once during statistical analysis. For example, low-di-
mensional plots (bivariate scattergrams and ternary plots) of the measured
elements are the usual way most researchers display and examine their data.
When considering a small number of sources, this technique may be satisfac-
tory. However, when the number of sources in a geographic region is quite large,
such as in central Mexico (-20 sources) or in the California-Oregon-Nevada
region ( 100+ sources), low-dimensional plots of raw elemental concentrations
are frequently inadequate. This can be particularly problematic in XRF where
the number of high-precision, element-concentration pairs is small. As a result,
the failure to consider multivariate methods of data analysis and the unsystem-
atic use of source and artifact data has led to a number of erroneous source
assignments.
One of the most notable examples of mistaken source assignments
concerning failure to use multivariate methods concerns the Chatham Island
artifacts. Using XRF data and simple univariate methods, Leach (1973) mis-
takenly concluded the artifacts to be of Mayor Island origin, located approxi-
mately 600 km to the west of the Chatham Islands. Subsequent use of more
comprehensive NAA data and multivariate methods found the Chatham Island
artifacts to be from Easter Island located almost 7,000 km to the east (Leach
and Warren 1981).
Some researchers (Perlman and Asaro 1969; Wilson 1978) have argued
against consideration of correlated elements because little extra discrimination
between sources is achieved by analyzing a number of correlated elements.
However, the concentrations of various elements in obsidian and other geo-
logical materials are far from independent from one another. Hence, the
practice of not considering correlated elements results in a loss of potentially
useful information. For example, Bowman et al. (1973) demonstrated that in
spite of the existence of extreme correlations in compositional data from Borax
Lake obsidian their data interpretations are just as definitive as if the correla-
tions had not existed. Multivariate statistical procedures described below
permit fully accounting for interelement correlation wit):wut sacrificing the
useful information from all measured elements.

MULTIVARIATE ANALYSIS

The main objectives of statistical analysis when applied to compositional

data are data exploration, hypothesis generation, hypothesis verification, and
SYSTEMATIC .4.PPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 25

data summarization. Ward (1974), Sayre (1975), Bieber et al. (1976), Leach
and Manly (1982), Bishop and Neff (1989) and others have pioneered the use
of the multivariate analytical techniques in order to achieve these objectives in
archaeometric studies. Before describing the statistical procedures employed
by the Archaeometry Laboratory at the Missouri University Research Reactor
(MURR), a brief review of the terminology and techniques employed in
multivariate analysis is presented. More detailed explanations are available in
Sayre (1975), Bieber et al. (1976), Davis (1986), and Bishop and Neff (1989),
and Neff (1994).

Sample Distributions and Outlier Specimens

Previous authors (e.g., Picon et al. 1975; Bieber 1976) have differed over
whether concentration data for geological materials are distributed normally
or lognormally. Fortunately, as Ahrens (1954) has noted, if the dispersion (or
spread) in the data is small compared to the mean concentration, the lognormal
distribution approximates a normal distribution. The assumption of lognor-
mality is also convenient because log transformations compensate for the large
differences in magnitude between the high-concentration elements (such as K,
Fe, and Na) and the trace elements (e.g., rare earths or REEs). Thus, log
transformation creates a convenient quasi-standardization (Sayre 1975) or
rescaling of the original data that facilitates the application of various multi-
variate methods.
Specimens that chemically differ from all others and have a disruptive
influence on the shapes of sample distributions are commonly known as
outliers. During the application of multivariate methods, preliminary screening
and group refinement procedures are used to eliminate statistical bias caused
by outlying specimens (Bishop and Neff 1989).

Euclidean Hyperspace, Variance-Covariance, and Correlation

Multivariate statistical techniques are based on the idea that each speci-
men can be considered a point in multidimensional space (hyperspace), with
the number of dimensions equivalent to the number of elements measured. In
such a hyperspace, groups of specimens with similar compositions form a cloud
of points surrounding a mathematically calculable "center of mass." An indi-
vidual compositional group is characterized by the location of its centroids and
by the unique hyperellipsoidal shape of the group distribution (Sayre 1975).
The shape is due to the correlations between different elements. Distinct source
groups are represented by two or more clouds that do not overlap.
As the dispersion in concentrations for an element increases, it becomes
more difficult to use the element to differentiate between source groups because
they will have an increasing tendency to overlap with one another. A parameter
26 MICHAEL D. GLASCOCK et al.

that quantifies dispersion in compositional data is the measured variance,

cr~ , in the mean element concentration. The measured variance represents the
sum of the natural variance in group composition, cr! , and the variance due
to sampling and analytical errors, cr;, , as defined by expression ( 1).

(l)

And it is equal to the square of the standard deviation, crm, of the group's mean
concentration. In analyses of obsidian source data by NAA, we have found that
about half of the elements are measurable with standard deviations of 2-5%.
just as the sample variance is a measure of the spread of concentration
values for an element around a central mean, the sample covariance is a
measure of the mutual variability of a pair of elements around their common
mean. The sample covariance between any two elements j and k can be
calculated from expression (2).

"" -
vjk - -
_.i-=='-1_ _ _ _ _ __ (2)
n-1

where cij and cik denote the respective concentrations of the jth and kth
elements in specimen I. The mean element concentrations are given by Ai and
Ak, and n is the total number of specimens under consideration. Although the
covariance is somewhat like a correlation coefficient, it has not been stand-
ardized for the magnitudes of the elements. Therefore, the covariance can take
on any value and is not limited to the range from -1 to +1.
For computational purposes, the covariances between elements are ar-
ranged in the form of a square matrix with one row and one column for each
element. The intersection of each row and column contains the covariance for
the corresponding pair of elements. The diagonal entries contain the variances,
since the covariance for an element against itself is the same as the variance.
Appropriately, the matrix is called the variance-covariance matrix.
Finally, the correlation coefficient, ri 1,, between any two elements j and k
is defined by the ratio in expression (3).

(3)

where cri and crk are the respective sample standard deviations, and crik is the
sample covariance between the elements. Correlation coefficients range in
value from +1 to -1, with +1 indicating a perfect relationship, 0 indicating no
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 27

relationship, and -l indicating an inverse relationship between elemental

compositions. For obsidian it is not uncommon to find correlation coefficients
of 0.9 or higher for certain elements (e.g., REEs).

Bivariate Analysis

As stated above, two-dimensional scattergrams are frequently used to

make visual distinctions between obsidian source-groups. In addition, highly-
correlated element pairs and outliers can sometimes be identified through
inspection of bivariate plots. Confidence ellipses (or probability intervals
surrounding source groups) are frequently calculated and drawn at a constant
Mahalanobis distance from group centroids. In regions where there are rela-
tively few possible sources, bivariate plots often permit unambiguous assign-
ment of artifacts to sources. Bivariate plots can be used to graph original
concentration data or linear combinations of the data such as principal com-
ponents or discriminant functions.

Cluster Analysis
Cluster analysis is often used in the initial inspection of obsidian-source
data because it is a rapid and efficient technique for evaluating relationships
between large numbers of samples, between which distance measures have
been calculated. Some investigators resort to cluster analysis routinely (Hur-
tado de Mendoza and jester 1978; Blackman 1984). Cluster analysis allocates
samples into distinct groups. The results are commonly presented as dendro-
grams showing the order and levels of clustering, as well as the distances
between individual samples.
Cluster analysis is based on a dissimilarity matrix in which the distances
between all pairs of specimens are calculated using one of several possible
distance measures (Sayre 1975; Harbottle 1976). One of the most frequently
employed distance measures is the squared-mean Euclidean distance (SMED),
defined between specimens j and h according to equation (4).

(4)

where the scaling factor, n, corresponds to the number of elements actually

measured.
Most clustering procedures employ Euclidean distances between individ-
ual samples in the calculation of similarity or difference. In doing so, they
assume that the elements under consideration are uncorrelated (hyperspheri-
cal). Consequently, dendrograms may accurately represent the differences
28 MICHAEL D. GLASCOCK et al.

between within-group members, but greatly distort the differences between

clusters of samples (Sneath and Sokal1973; Sayre 1975; Harbottle 1982). Thus,
the effectiveness of dendrograms as a means of faithfully illustrating differences
between groups is questionable. Dendrograms can be used as an initial step in
the identification of groups, but they are inadequate as a quantitative measure
of the ability to differentiate between groups. Fortunately, as illustrated below,
there are other procedures that enable a more reliable determination of both
within- and between-group differences.

Principal Components Analysis

One of the more widely utilized techniques for exploring underlying
relationships in multivariate data is factor analysis. In particular, the factor
analytical approach known as principal components analysis (PCA) is employed
when the main objective is to derive a set of uncorrelated linear combinations of
the original variables that retain the original information. If the original variables
are significantly correlated, a reduced number of principal components may be
adequate to describe the significant features of the original dataset.
In PCA, one performs a transformation of the data using eigenvector
methods to determine the direction and magnitude of maximum variance in
hyperspace (Davis 1986). Computation of the eigenvectors is derived from
either the variance-covariance or correlation matrix of the original variables.
The first principal component (PC) is oriented in the direction of maximum
variance in the data. The second PC lies in the direction of the maximum
remaining variance, with the additional constraint that it must be orthogonal
to the first PC. The third PC is orthogonal to the first two PCS, again
representing the direction of maximum remaining variance. The procedure
continues until the number of PCS is equal to the number of original dimen-
sions. All resulting PCS are then linear combinations of the original axes. The
loading factors (coefficients) used to construct the PCS can be used to identify
attributes important to the individual components.
The PCA technique can be used to describe the entire distribution of
elemental concentrations in a compositional dataset on a different basis. The
positions of samples in the original element concentration space are converted
into principal scores calculated in terms of the linear combinations of the
original data and are projected onto the new principal component axes. For
correlated data, a larger portion of the data structure (often >90% of the total
population variance) can be inspected from two- or three-dimensional plots
using the first few PCS than is possible with most bivariate plots of the original
elements. Thus, PCA eliminates repetitive data and finds a new basis which
facilitates dimensionality reduction. If PCA provides a satisfactory differentia-
tion between source groups, it is often possible to assign artifacts by projecting
their principal component scores against the principal axes.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 29

The PCA approach is widely employed in archaeometric studies of

ceramics (Neff et al. 1988) and manmade glasses (Baxter 1991) but has only
rarely seen application in obsidian studies (Stark et al. 1992). However, PCA
can be especially useful to understanding the chemical basis for group separa-
tion and in identifying original elements that by themselves may be satisfactory
to describe the most significant compositional differences between particular
obsidian source groups.
In their discussions of taxonomic methods, Cattell (1952) and Sneath
and Sokal (1973) point out that multivariate factor analysis can be used in two
ways. Associating variables over all specimens is called a Q-mode technique.
On the other hand, associating specimens over all variables is called an R-mode
technique. The emphasis of most PCAs has been the use of R-mode techniques
with displays of variable scores plotted in R-space to reveal the contributions
of individual variables. However, applications involving Q-mode techniques
on compositional data are becoming more common (Neff et al. 1988).
Recent articles by Baxter (1992) and Neff (1994) demonstrate a simulta-
neous RQ-mode PCA as a means of examining relationships between specimens
and between elements simultaneously. This technique of presenting R-mode
and Q-mode plots together is called biplotting (Gabriel197l). The simultane-
ous use of both Q-mode and R-mode plots allows the individual contributions
of all elements to group separation to be displayed.

Discriminant Fnnction Analysis

In multivariate analysis, the term discriminant analysis is frequently used
to describe two different processes: discrimination and classification. The proc-
ess of discrimination involves identifying a mathematical transformation of the
original variables that best reveals the differences between known groups of
observations. On the other hand, the process of classification involves the
categorization of a number of observations into known groups.
Discriminant function analysis (DFA) is a multivariate technique differ-
ing from PCA in that it is used to summarize relationships between groups.
DFA is based on the assumption that the pooled variance-covariance matrix is
an accurate representation of the total variance and covariance of the data
(Davis 1986). When employing DFA, one assumes that all specimens in the
dataset necessarily belong to one of the known source groups.
Two computational methods can be used to derive discriminant functions:
concurrent or stepwise. The concurrent method finds a set of linear combinations
of the original variables that best summarizes between-group differences by
considering all variables simultaneously, regardless of their individual discrimi-
nating powers. The concurrent method computes a series of discriminant
functions (one less than the number of source groups under consideration) that
minimize the ratio of the difference between group multivariate means to the
30 MICHAEL D. GLASCOCK et al.

multivariate variance within each group. As a result, the procedure maximizes

the total score of specimens belonging to their assigned group and minimizes the
total score of belonging to all other groups. The resulting discriminant functions
are linear combinations of the original data each successively describing lesser
amounts of separation between compositional groups.
In the stepwise method, the variables are entered into the discriminant
function one at a time on the basis of their discriminating power. The stepwise
method begins by identifying the single best discriminating variable. The initial
variable is then paired with all others in order to identify the next best
discriminating variable. This process is successively repeated for all remaining
variables. A reduced set of most discriminating variables, is often as good at
differentiating as the entire suite of variables. In their studies of obsidian from
western North America and the Maya lowlands, Nelson (1985) and Hughes
(1986) used stepwise discriminant analysis procedures available in the SPSS
statistical package.
Bivariate plots of discriminant functions are useful for visually displaying
group separation. In the same manner as PCA, one can employ discriminant
functions to calculate the discriminant scores for artifacts. By comparing
artifact scores with scores for source specimen groups, decisions regarding
source assignments for the artifacts can be made.

Mahalanobis Distances and Probability Calculations

Discriminant analysis for classification purposes and related techniques
is based on the standardized-squared distance or Mahalanobis distance. The
Mahalanobis distance (D 2) statistic is defined as the squared Euclidean distance
between a group centroid and an individual specimen divided by the group
standard deviation in that direction. Mathematically, the D2 from specimen k
to the centroid of cluster A is given by expression (5).

l l [cik -A;]· Iii · [ cjk - Aj]

n n

n~ = (5)
i=l j=l

where A. and A.J are the mean concentrations of elements I andJ. in the cluster
I

and LIJ is the il)"th element of the inverse of the variance-covariance matrix. The
Mahalanobis distances from a specimen to each of several alternative group
centroids can be used to estimate the relative probabilities of membership in
each group (Bishop and Neff 1989). TheD 2 statistic can also be used to measure
the separation distance between pairs of multivariate means in units of pooled
variance (Davis 1986).
As explained by Klecka (1980) and Hughes (1986), the classification
aspect of discriminant analysis is based on a number of statistical assumptions:
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 31

(l) there are two or more groups; (2) there are at least two observations in each
group; (3) the number of discriminating variables must be at least two less than
the total number of specimens; ( 4) the discriminating variables are measured
at the interval level; (5) the discriminating variables must not be linear
combinations of other discriminating variables; (6) each group is drawn from
a sample population with a multivariate normal distribution on the discrimi-
nating variables; and (7) the variance-covariance matrices for each group must
be approximately equal. Of these assumptions, the normality of multivariate
distributions and homogeneity of group variance-covariance matrices are the
most important and difficult to satisfy in practice (Sneath and Sokal 1973).
The assumption that each group is drawn from a multivariate normal
distribution is critical for tests of significance. Significance tests are computed
by comparing a statistic calculated for a sample relative to a theoretical
probability distribution for that statistic. If the sample population does not
satisfy this requirement, the distribution for the sample statistic will be different
from the theoretical distribution resulting in some inaccuracy. As long as the
deviation is minor, Lachenbruch (1975) has shown that the discriminant
analysis is not particularly sensitive. The assumption of multivariate normality
is also important for classification based on relative probability of group
membership. Probabilities are calculated from a chi-square distribution which
is appropriate only if the variables are normally distributed. If the normality
assumption is violated, probabilities for membership in some groups may be
overstated while the probabilities for other groups may be underestimated.
According to Ahrens (1954), the evidence from basic statistics (i.e., means,
ranges, etc.) seems to indicate that most geochemical data distributions are
approximately normal (or lognormal) and the multivariate normality assump-
tion is rarely violated.
The more serious assumption concerns the homogeneity of group vari-
ance-covariance matrices. If the matrices are not equal, distortions in the
classification equations will occur such that specimens have a greater likeli-
hood of being allocated into groups with greater dispersion as measured by the
determinant of the variance-covariance matrix (Klecka 1980). Leach and
Manly (1982) employed a power transformation to alleviate this problem in
obsidian data from the New Zealand and Pacific region. Fortunately, as Klecka
(1980) points out, for classification purposes violation of the homogeneity
assumption is only important when comparing probabilities that are nearly
equal in magnitude. As a result, cautious consideration of similar probability
values is mandatory when making classifications.
The D2 statistic, unlike the simple Euclidean measure, incorporates
information about the correlations between pairs of elements as derived by the
off-diagonal terms of the variance-covariance matrix. Thus, it permits calcula-
tion of the probability that a particular specimen belongs to a group based not
only on the proximity of the specimen to the group centroid, but also on the
32 MICHAEL D. GLASCOCK et al.

rate at which the density of specimen data points decreases away from the group
centroid in the direction of the specimen of interest.
Calculation of probabilities using D2 requires that the number of speci-
mens defining the source groups exceed the number of elements under consid-
eration by at least one. Probability calculations with small-sized groups suffer
from a problem that Harbottle (1976) calls stretchability because of the tendency
of individual specimens to stretch the envelope so as to include that specimen.
Excluding the specimen from the group before calculating its probability of
membership helps to eliminate this problem. Nevertheless, it is recommended
that each compositional group consist of three to five times as many specimens
as the number of elements under consideration (Harbottle 1976).
The significance of differences between two groups of specimens can be
tested by Hotelling's T2 statistic (the multivariate equivalent of the Student's t)
according to expression (6).

(6)

where n1 and n2 are the number of specimens in each of two groups. Hotelling's
T2 statistic is essentially equivalent to D2 for individual data points. If v is the
number of elements considered in the analysis, the T 2 statistic can be trans-
formed into an F ratio by expression (7).

(7)

This is easily transformed into a probability percentage using tables of the F

distribution.
Finally, as Sayre (1975) recommends, Mahalanobis distance calculations
are quite useful for handling missing data. When large numbers of specimens
are analyzed, there is almost a certainty that a few concentration measurements
will be missed for some specimens. This occurs most often when the group
mean concentration for an element is near the detection limit. For example,
the elements Sr and Ba are near or below the NAA detection limit for a number
of obsidian sources. Rather than completely eliminate specimens with occa-
sional missing values from consideration, it is possible to substitute by com-
puting a value that minimizes the Mahalanobis distance for that specimen from
the group centroid. As a result, the substitute concentration values for these
few specimens will have little effect on the statistical characteristics of the group
and enables them to be included in all group calculations.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 33

EXAMPLES FROM THE MURR MESOAMERICAN

OBSIDIAN DATABASE

To demonstrate the approach advocated here, we present examples from

our experience in building a database for obsidian source materials from
Mesoamerica. Braswell and Cobean were actively involved in acquiring the
source specimens and Glascock primarily handled the data analysis and statis-
tical interpretation. Subsequently, artifact analyses have been performed in
collaboration with more than two dozen different colleagues interested in
Mesoamerica. As a result of the source characterization work, nearly 4,000
source specimens and artifacts from Mesoamerica have been analyzed and
sourced in the MURR Archaeometry Laboratory

Obsidian Sources in Central Mexico

One of the most ambitious obsidian source characterization studies
carried out to date was initiated in 1980 by Cobean with support from the
National Science Foundation and the University of Missouri. In collaboration
with other archaeologists from Mexico's Institute Nacional de Antropologia e
Historia (INAH), Co bean collected more than 800 obsidian source specimens
(weighing 710 kg) from the volcanic sources located throughout central
Mexico shown in Figure 2.1.
The main purpose of Co bean's project was to establish a reference sample
repository and geochemical data bank for Mexican obsidian. In addition to
collecting samples for analysis at MURR and for future reference, portions of
the source specimens were sent to Fred Nelson (Brigham Young University) for
XRF analysis and Jonathon Ericson (California-Irvine) for hydration rate
measurements. A major goal was to increase the confidence with which
obsidian artifacts from archaeological sites could be associated with specific
source areas. A final goal of the project was to employ a detailed statistical
analysis to identify the most discriminating elements between regions, sources,
and, if possible, different flows or outcrops from the same source. Identification
of highly discriminating elements might facilitate the development of more
rapid and less expensive analytical procedures for artifacts that are as reliable
as the comprehensive analysis applied to characterize source specimens.
Several of the obsidian sources from which Cobean collected samples
extend over vast areas greater than 300 km 2 • An intensive field survey mapped
the obsidian flows, located prehistoric quarrying areas, and gathered other
essential information. Partial reports of this field work appear in Stocker and
Cobean (1984) and Cobean et al. (1991).
In following with the thorough design of the sampling strategy, a com-
prehensive analytical program using NAA evolved in the MURR laboratory
Sample preparation techniques (Glascock et al. 1988), standards for analytical
34 MICHAEL D. GLASCOCK et al.

0 500 1000 km

''
,' '
,' ' I

,' ' I
I

,' '
I
I
I \
'
,' ' I \ \

I
I \
' \

' ''
''
I
I \

' I \
''
ZacualtipanA
0 50 100 km
Tepalzingo
~ AATulancingo
Pachuca-1 ,-2,-3
Ucareo-2 A AAitotonga
Malpai~
Ucareo-1 A A Zaragoza
OtumbaA.
apecuaro I Paredon
Santa Elena Derrumbadas
MexicoC~ A Guadalu
~ctoria
OrizabaA--

Figure 2.1. Locations of obsidian sources in central Mexico mentioned in this study.

calibration (Graham et al. 1982), and quality control procedures (Glascock and
Anderson 1993) were developed in order to provide the most complete and
accurate analyses possible. In addition, database management and statistical
methods originally developed for compositional studies of ceramic materials
are being employed (Bishop and Neff 1989; Neff 1990).
MURR's analytical program for obsidian routinely measures 27 elements
in every source specimen by employing two irradiations and three measure-
ments. A five-second irradiation in a neutron flux of 8 x 10 13 n/cm 2/s is followed
by a 25-minute decay and 12-minute count to determine the short-lived
elements Ba, Cl, Dy, K, Mn, and Na. To measure the medium- and long-lived
elements, samples are irradiated for 70 hours in neutron flux of 5 x 10 13 n/cm 2/s.
The medium-lived elements (Ba, La, Lu, Nd, Sm, U, and Yb) are determined
by a 2,000-second count after 7-8 days of decay. The long-lived elements (Ce,
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 35

Co, Cs, Eu, Fe, Hf, Rb, Sb, Sc, Sr, Ta, Tb, Th, Zn, and Zr) are determined by a
10,000-second count after 4-5 weeks of decay
A total of 575 of the original800 Mexican specimens have been analyzed.
The most intensive analyses were made on specimens from the source regions
near Pachuca in the state of Hidalgo, Ucareo-Zinapecuaro in the state of
Michoacan, and the Pico de Orizaba-Guadalupe Victoria source area straddling
the border between the states of Veracruz and Puebla.

Obsidian Sources in Guatemala

Obsidian sources in the highlands of Guatemala were sampled by
Braswell who collected 137 samples from the seven areas shown in Figure 2.2.

0 200km

MEXICO

,
,, ,
,, ,'
,, ,
,

*
San Martin Jilotepeque A EIChayal
A Sansare

A Guatemala City A A Jalapa

San Bartolome
Milpas Altas
A Laguna de Ayarza

A
0 40km lxtepeque

Figure 2.2. Locations of obsidian sources in Guatemala mentioned in this study.

36 MICHAEL D. GlASCOCK ct al.

In addition, the longitude and latitude coordinates for each specimen were
recorded. Special emphasis was placed on obtaining specimens from the
complex series of obsidian outcrops and river cobble areas surrounding the
town of San Martin Jilotepeque, Chimaltenango (an area also known as Rio
Pixcaya to many archaeologists). More than 70 specimens from this source area
have been analyzed to date. A specific goal of Braswell's project was to
determine whether obsidian quarries that were near one another could be
adequately differentiated such that study of procurement trends and differences
between archaeological sites, regions, and time periods might be possible.
In addition to providing samples to build an obsidian database for the
Guatemalan sources, Braswell searched for a pair of sources suggested by
previous research on artifacts from the neighboring La Entrada region of
Honduras (Glascock and Aoyama 1990). Although Braswell's Guatemalan
search failed to locate the sources supplying La Entrada, a new obsidian source
was located near the village of Sansare. Details concerning the Sansare source
are reported in Braswell and Glascock (1992).

Statistical Analysis of Source Data

At MURR, we employ a series of computer-based procedures, written in
both FORTRAN and GAUSS languages, that facilitate the application of multi-
variate methods on personal computers. The GAUSS language routines written
by Neff (1990) replace an earlier series of Brookhaven National Laboratory
(BNL) programs by Sayre (1975) and Bieber et al. (1976). Our repertoire of
statistical capabilities also includes the PC-based program SYSTAT.
After the obsidian specimens are analyzed, the concentration data for
each is entered as a single record in a Master dBASE file with other analytical
and descriptive information for that specimen. From the Master file, it is a
simple matter to extract specimens as either individuals or groups for a
particular statistical analysis. First, individual specimens that appear to be
outliers rather than reliable core-group members are eliminated. If the number
of missing values for an element is quite large, the element may be omitted
from further consideration. If the number of missing values is small, then a
GAUSS procedure based on minimizing the Mahalanobis distances from group
centroid is used to calculate substitute values. In our mesoamerican database,
Sr is below our detection limit of -25 ppm in about half of the samples we
analyze, thus causing us to eliminate it from the list of elements we consider.
Dependent upon the particular source, Ba is occasionally below our detection
limit of -15 ppm and a missing value substitution is necessary. In either
instance, one can sometimes make a rapid evaluation of candidate sources
for many artifacts based on the presence or absence of Sr and Ba.
Before comparing artifacts to sources, it is necessary to demonstrate
the ability of the source database to answer source-related questions. For
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 37

example, can the source data be used to differentiate between geographic

regions, between the sources within a region, or between outcrops within a
source? How many different source profiles are represented in the mesoameri-
can database? Which elements or combinations of elements are best for
differentiating between sources? Can a limited but satisfactory suite of ele-
ments be identified that can be measured more rapidly and at lower cost that
works for a majority of artifacts? What is our degree of confidence in making
source assignments?
In order to answer these questions, the GAUSS language routines are
designed to establish chemical profiles for a collection of sources describing a
region, individual sources within a region, and different outcrops within a
source. Examples from MURR's mesoamerican database are presented here for
illustration.

Differentiation between Regions

Are the obsidian sources in different geographic subregions of Mesoamerica

more different than sources within each subregion?
In order to answer this question, the mesoamerican database was sub-
divided into nine geographic subregions with one or more obsidian source
located in each. Based on the geographic proximity of sources to one another,
the regions created were: Zaragoza (consisting of the Zaragoza and Alto tonga
sources); Orizaba (Pico de Orizaba, Guadalupe Victoria, and Derrumbadas
sources); Paredon (Paredon and Santa Elena sources); Otumba (Otumba and
Malpais sources); Tulancingo (Tulancingo and Tepalzingo sources); Pachuca
(all Pachuca subsources); Zacualtipan (Zacualtipan source); Ucareo (Ucareo
and Zinapecuaro sources); and the Guatemalan region (consisting of all
sources in the highlands of Guatemala). A total of 712 source specimens were
extracted from the Master dBASE file in order to create the tentative regional
groupings.
On the basis of their poor analytical precision, the elements Cl, Dy, K,
Nd, and U were omitted. Due to its large number of missing values, the element
Sr was also omitted. Concentrations for the remaining 2l elements were
transformed into their base-10 logarithms to compensate for the differences in
magnitude between major elements such as Fe and Na and trace elements like
the REEs. In order to identify the dimensions of greatest variance, the variance-
covariance matrix of the nine-region dataset was input to the GAUSS routine
designed for Q-mode PCA calculation.
Table 2.1 lists the calculated eigenvalues, percent variance described by
each principal component, and the cumulative variance from the PCA. Table
2.2 presents the eigenvectors (coefficients) for the first six linear equations
ranked from largest to smallest. As shown in Table 2.1, the first four PCS
describe over 96% of the variance-covariance structure of the dataset. It often
38 MICHAEL D. GLASCOCK et al.

Table 2.1. Eigenvalues and Percentage of Variance Explained by

Principal Components Calculated from the Variance-Covariance
Matrix of Concentration Data in Log Base-l 0 ppm for Nine
Sub-regions in Mesoamerica

Principal Cumulative
Component Eigenvalue %Variance %Variance

1.4934 65.58 65.58

2 0.3069 13.48 79.06
3 0.2722 11.95 91.01
4 0.1235 5.42 96.43
5 0.0253 1.11 97.54
6 0.0153 0.67 98.22
7 O.Ql18 0.52 98.73
8 0.0084 0.37 99.10
9 0.0055 0.24 99.34
10 0.0045 0.20 99.54
11 0.0034 0.15 99.69
12 0.0022 0.10 99.78
13 0.0017 0.08 99.86
14 0.0011 0.05 99.91
15 0.0007 O.o3 99.94
16 0.0005 0.02 99.96
17 0.0004 0.02 99.97
18 0.0003 O.Ql 99.99
19 0.0002 O.Ql 99.99
20 0.0001 0.00 100.00
21 0.0000 0.00 100.00

turns out that a transformation of the dataset using the first few PCS may
delineate the different source regions better than the original variables. For
example, Figure 2.3 presents a Q-mode plot of scores for the source specimens
in each region plotted on PCOl versus PC02. Except for the Otumba and
Guatemalan regions which overlap slightly, the source regions are well sepa-
rated on these dimensions. Many of the regions in Figure 2.3 consist of multiple
sources and with the exception of the Guatemalan and Zaragoza regions,
anticipated internal structure is not readily apparent.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 39

Table 2.2. Eigenvectors for the First Six Principal Components

Best Describing Nine Sub-regions in Mesoamerica

Element PC01 PC02 PC03 PC04 PC05 PC06

Ba -0.4402 0.6812 -0.0656 0.2856 0.2466 0.0617

La 0.1407 0.1954 0.2881 0.0277 -0.4028 0.0243
Lu 0.2910 0.0262 0.0590 0.0344 0.1797 -0.1913
Sm 0.2032 0.0837 0.1350 0.1396 -0.0842 0.1250
Yb 0.2968 0.0261 0.0655 0.0529 0.1825 -0.1264
Ce 0.1678 0.1697 0.2548 0.0426 -0.3257 0.0493
Co -0.2266 0.1971 0.5872 -0.6060 0.1012 -0.2226
Cs -0.0030 -0.1136 0.2293 0.2119 0.4822 0.2740
Eu 0.2028 0.4857 -0.2195 -0.2281 0.1870 0.2221
Fe 0.1403 0.1490 0.0712 -0.0353 -0.0240 0.0775
Hf 0.2996 0.1003 -0.0390 -0.0158 0.1017 0.0770
Rb 0.0672 -0.0760 0.1046 -0.0293 0.2991 0.0913
Sb 0.0101 0.0709 0.3689 0.4985 0.0676 -0.5310
Sc 0.0532 -0.1286 0.0617 -0.3099 0.1850 0.0559
Ta 0.2350 -0.0849 0.0227 0.0032 0.1845 -0.1539
Tb 0.2719 0.0430 0.0795 0.1284 0.0774 0.0683
Th 0.1023 -0.0770 0.2307 -0.0314 0.2959 0.1850
Zn 0.2559 0.1166 -0.1054 0.0778 0.0183 -0.0472
Zr 0.3441 0.2652 0.0194 -0.1090 -0.1634 0.0723
Mn 0.1017 0.1185 -0.3773 -0.2141 0.1629 -0.6121
Na 0.0306 0.0185 -0.0704 0.0069 -0.0143 -0.0319

Differentiation between Outcrops in Complex Source Areas

Do the obsidian outcrops that cover large geographic areas at San Martin
]ilotcpcquc, Sierra de Pachuca, and Ucarco-Zinapccuaro have multiple composi-
tiona! fingerprints? Which clements best illustrate these differences? How many
fingerprints arc present in the mcsoamcrican database?
Obsidian from the San Martinjilotepeque (SMJ) source area was obviously
important to the earliest peoples of southern Mesoamerica as it appears in
Formative period contexts at many archaeological sites. The source area is quite
large and raw obsidian is found in numerous surface outcrops shown in Figure
2.4. The outcrops at Pachay, Choatalum, and the site labelled Group 6 have the
greatest archaeological evidence for utilization during prehistoric times.
40 MICHAEL D. GLASCOCK et al.

...
0
Tuloncinqo reoion

ci

....0 Pachuco region

ci

0 Zocuoltipon region
0
N
0
0 ~
u ....
Q.0 Poredon reoion
ci
I

...
0
ci
I

<D
0
ci
I
•
"'O L-~--~--~--~--~~~~---L--~--~--~~--~--~
'i' -0.06 -0.04 - 0 .02 0 .00 0 .02 0 .04 0 .06 0 .06
PCOl
Figure 2.3. Q-mode bivariate plot of specimen scores on first and second principal components
illustrating differentiation between obsidian source regions in Mexico and Guatemala. Probability
ellipses at the 95% confidence level are included to indicate the different source regions.

The compositional data for seventy SMJ source samples were extracted
from the database and submitted to cluster analysis (Figure 2.5). Examination
of the clustering dendrogram suggests existence of six compositional sub-
groups in the SMJ source area. This observation is readily confirmed by the
bivariate plot of Cs versus Sc presented in Figure 2.6. A comparison with the
outcrop locations shown in Figure 2.4 indicates the geographical distribution
of fingerprints. In particular, the SMJ-1 subgroup fingerprint <~ppears in five
different areas (labelled Group lA through Group lE) and in three river cobble
samples representing a secondary deposit. The archaeological evidence to date
indicates that the vast majority of artifacts sourced to SMJ came from quarries
associated with the SMJ-1 fingerprint. Thus far, there are no confirmed assign-
ments of artifacts to any of the other SMJ subsources.
The Sierra de Pachuca source located ten kilometers from the city of
Pachuca, Hidalgo is one of the largest and most important in central Mexico.
Most Pachuca obsidian is readily characterized by a dark-green color. Pachuca
obsidian is of such high quality that it was one of the most widely utilized sources
in all of Mesoamerica, with Pachuca artifacts being found several thousands of
kilometers from the source area (Neff et al. 1993). Some variation in color was
noted in specimens collected by Cobean from the source area and it was
hypothesized that more than one compositional fingerprint exists. To examine
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 41

Group 2

San Martin
Jilotepeque

1
0 1 2km

Figure 2.4. Locations of obsidian ou tcrops in the vicinity of San Martinjilotepeque, Guatemala.

this hypothesis, the data for 167 specimens from the Pachuca source area was
extracted from the database. A cluster analysis dendrogram indicted three
possible subgroups (labelled Pachuca-1, Pachuca-2, and Pachuca-3 in Figure
2.1). Additional confirmation of these groups is shown in the bivariate plot of
42 MICHAEL D. GlASCOCK et al.

SMJ006-.
SMJ007-I
SMJOOS-'-.
SMJ014-'---.
SMJOlO-. I
SMJOll-I
SMJ013- I ---I
SMJ009-. I-.
SMJ069-'-' I
SMJ068-. I-.
SMJ073-'-. I I
SMJ070- 1 -. I I
SMJ071-I I I
SMJ067-. I-. I I
SMJ072-'-' I I I
SMJ022-. I-I I
SMJ024-'-. I I I
SMJ030-I-. I I I
SMJ034-' I I I I
SMJ028-. I-I I I
SMJ029-I I I I I
SMJ031-'-I I
SMJ025-I I
SMJOlS- I I I I
SMJ033-' I I
SMJ027-. I I
SMJ032- I-. I
SMJ008- '-------------------------.
SMJ043-. I------------.
SMJ045- •-. I
SMJ042- •-. I
SMJQ46- I------- •
SMJ020-. I----------.
SMJ021-'-.
SMJ047-I
SMJ019-' -----'
SMJ044- '--------' I
SMJ039-. I------.
SMJ041-'---. I
SMJ018-'-.
SMJ016-. I-------.
SHJ03B- I - . I
SMJ017-. I-'
SMJ040-'-'
SM.J048- '--------------------------------'
SMJ012-. I--------.
SMJ036-'-. I I
SMJ037-' ---.
SMJQJS- I----------------------------------------------- I
SMJ058-. I-------------.
SMJ064-I
SMJ066-I
SMJ065-'-.
SMJ060-I
SMJ059-I
SMJ062-'-.
SMJ063-I
SMJ061-'-. I
SMJ026- '--------------------------------------------------------.
SMJ049-. I-
SMJ053-'-.
SMJ054-I-.
SMJ057-' I
SMJOSS-'-.
SMJ051-. I-------.
SMJ052-.- I I
SMJ023- '-----.
SMJ056-' ------------------------------. I
SMJ050- '--------------------------'

Figure 2.5. Dendrogram showing cluster analysis results for obsidian source specimens from San
Martinjilotepeque based. Squared-mean euclidean distance calculations were based on log base-10
transformations of the 21 element concentrations. Six tentative subgroups are identified.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 43

...
,;

"'
N
SMJ - 6 &
~"
9
N
,;

SMJ-4

ec. N
A

0
~
(J
rn 0
,;
(Y
SMJ-2
"! .
SMJ-5

I)
.
SMJ-1
"'
-
~

- 2 .0 2 .4 2.8 3 .2 3 .6 4 .0 4 ,4
Cs (ppm)

Figure 2.6. Bivariate plot of Cs versus Sc for the compositional subgroups of the San Martin
Jilotepeque, Chimaltenango obsidian source. Probability ellipses at the 95% confidence level
indicate the locations of the subgroups.

Cs versus Eu for Pachuca obsidian shown in Figure 2. 7. Subsequent comparison

of specimens with Cobean's sample collection records has been used to relate our
chemical subgroups to geographic areas. All dark-green source material and
Pachuca artifacts sourced to date have been linked to the Pachuca-1 fingerprint.
A similar cluster analysis and bivariate plot inspection performed on
compositional data for source specimens from the Ucareo-Zinapecuaro region
confirm the existence of three compositional fingerprints (labelled Ucareo-1,
Ucareo-2, and Zinapecuaro in Figure 2.1). Abivariate plot of source specimens
from the Ucareo-Zinapecuaro region is shown in Figure 2.8. The archaeological
evidence supports the dominance of the Ucareo-1 subsource fingerprint rela-
tive to both the Ucareo-2 and Zinapecuaro subsources.
At first glance, the ability to source artifacts to specific outcrops or quarries
within a multi-fingerprint source would seem to have little importance with
respect to studies of long-distance trade. However, if raw material extraction is
focussed on a particular type of obsidian that is distributed widely (i.e, certain
colors or fracturing qualities) subsource identification may be extremely impor-
tant. For example, highly specific subsourcing has the potential of supplying
data on the organization of extraction activities and perhaps distribution of the
material (i.e., mining claims by individuals or groups). More importantly,
subsource identification is absolutely essential for obsidian-hydration studies.
44 MICHAEL D. GlASCOCK et al.

ao

@~--·
"'
N

... "
N

N
N

i.e 0
N

"
w
Q)

Poctluco- 1

"' Pochuco-J

Qv
1.8 2 2 2 .6 3 .0 3.4 38 42
Cs (ppm)

Figure 2.7. Bivariate plot of Cs versus Eu for the compositional subgroups of the Sierra de
Pachuca, Hidalgo obsidian source. Probability ellipses at the 95% confidence level indicate the
locations of the subgroups.

N
...
0

0
"'
0 Ucoreo- 1

"'ci

i.e "' Zinopecuaro

0
w" ---~

Q)
0
0

...
0 Ucoreo - 2
0

0
0
0 140 160 180 200 220 240 260 280
Mn (ppm)

Figure 2.8. Bivariate plot of Mn versus Eu for compositional subgroups of the Ucareo-
Zinapecuaro, Michoacan obsidian source. Probability ellipses at the 95% confidence level indicate
the locations of the subgroups.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 45

Ericson (1981) has shown that chemical differences in specimens from multi-
ple-fingerprint sources in California have a significant effect on hydration rates.
In each of the regions where multiple fingerprints are known, a more careful
analysis is necessary because misidentification of chemical subgroup for artifacts
also being dated may lead to erroneous dates. In addition, if certain subsources
were exploited at different times, then archaeologists may be able to use their
data to examine trends in obsidian procurement and exploitation.
Through application of the above mentioned PCA, cluster analysis, and
bivariate plotting techniques, thirty different fingerprints were discovered in
the regions of Mesoamerica under investigation. Eighteen sources are located
in central and eastern Mexico and twelve sources are located in the highlands
of Guatemala. The next step is to identify procedures through which the
individual sources can be differentiated from one another.

Distinguishing between Individual Sources in Mexico

Can we formulate a reliable procedure for differentiating between all sources

in central Mexico? What are its limitations? Does the procedure reject specimens
from the Guatemalan sources?
Since we previously demonstrated that most of the Mexican sources are
easily distinguished from the Guatemalan sources, the problem of differentiat-
ing between individual sources is separated into two portions (i.e., Mexico and
Guatemala). Data for the eighteen source groups involving 575 source speci-
mens from central Mexico was extracted from the Master dBASE file to form
individual source groups. Six outlier specimens from five different sources were
eliminated on the basis of a preliminary screening of elemental means, ranges,
and standard deviations.
To identify the dimensions of greatest variance and elements likely to be
important in differentiating between sources, an RQ-mode PCA was performed
on the remaining 569 specimens. The resulting Q-mode and R-mode plots for
PC01 versus PC02 are shown in Figures 2.9 and 2.10, respectively. Table 2.3lists
the individual eigenvalues, percent variance, and cumulative variance explain-
ing the PCA results for these data. With exception for the three subsources in the
Pachuca region, all of the central Mexican sources are separated at the 95%
confidence level on the basis of the first two PCS. Although not shown here,
further inspection of bivariate plots using other PCS finds the confidence ellipses
for the Pachuca sources are not separable until we include component PC05.
Inspection of the R-mode plot in Figure 2.10 finds that Ba, Co, Zr, Hf,
and the REEs are contributing most to separation between individual sources
on the first two principal components. On the other hand, Na and Rb are
contributing least to this view of the data. A bivariate plot of two high- precision
elements (Cs versus HD shown in Figure 2.11 offers the greatest visual
differentiation between confidence ellipses for the sources in central Mexico.
46 MICHAEL D. GLASCOCK et al.

""ci
0

"'
0
ci

...
0
Pica de Orizabo
Pochuco-1
ci

N
0
N
0
ci Cuodolvpe"'
u Victoria '-..J Altotonga
Q.
ucar••-o
CJ 0 Santa/h
0
0
ci
Oerr-umbodos Elena VPoredon
N
0 "alpaio 00
ci
I Otumbo C)
...
0
Zoro9ozo

ci ZocuolliponO
I

"'0ci
I -0.06 -0.06 - 0 .04 -0.02 0 .00 0 .02 0.04 0 .06 0 .06
PC01

Figure 2.9. Q-mode bivariate plot on the first two principal components showing 95% confidence
ellipses for obsidian sources in central Mexico. To avoid data congestion, the scores for individual
specimens are not shown.

N
ci

Mn

ci
,•
Sc

0
0
I

N
0
u
Q.
ci
I

N ,, '
ci
I
,,
'
"
,
.... ' Co
ci
I
• B.
...ci
I -0.6 -0.4 - 0 .2 -0.0 0 .2 0 .4 0 .6
PC01

Figure 2.10. R-mode bivariate plot on the first two principal components showing scaled-loading
factors for elements in central Mexican obsidian sources.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 47

Table 2.3. Eigenvalues and Percentage of Variance Explained by

Principal Components Calculated from the Variance-Covariance
Matrix of Concentration Data in Log Base-10 ppm for the
Eighteen Sources in Central Mexico

Principal Cumulative
Component Eigenvalue %Variance %Variance
1.5461 64.82 64.82
2 0.3392 14.22 79.04
3 0.3010 12.62 91.66
4 0.1235 5.18 96.84
5 0.0265 1.11 97.95
6 0.0173 0.72 98.67
7 0.0097 0.41 99.08
8 0.0073 0.30 99.38
9 0.0049 0.21 99.59
10 0.0029 0.12 99.71
11 0.0023 0.10 99.81
12 0.0018 O.o7 99.88
13 0.0011 0.05 99.93
14 0.0005 0.02 99.95
15 0.0004 0.02 99.96
16 0.0003 0.01 99.98
17 0.0002 O.ol 99.99
18 0.0001 0.01 99.99
19 0.0001 0.00 100.00
20 0.0000 0.00 100.00
21 0.0000 0.00 100.00

Two of the source groups (Santa Elena and Derrumbadas) have only five
analyzed specimens each. Therefore, in order to consider all of the central
Mexican sources simultaneously, probability calculations using Mahalanobis
distances are limited to a maximum of three variables. Approximately 92% of the
variance in the dataset listed in Table 2.3 is subsumed by the first three PCS.
Mahalanobis distances and F-value probabilities based on the first three PCS were
calculated to generate the posterior classification matrix presented in Table 2.4.
The classification table is organized with the obsidian source from listed vertically
and the source assigned to listed horizontally. As shown, the Santa Elena source
48 MICHAEL D. GlASCOCK et al.

00
N
PPochuco- l
....
~
N

Poehueo-3

0 Tuloncinoo
N

~ 0
ea. ~ Pochuco-2

Tepolzinoo
.9;
i: ~
I

Pored on Zocuoltipon
00
Zoro;ozo
t? Ucoreo-1
()Santo Eleno. . /
Ucoreo - 2
0 ,Altotonoo
Otumbaet ~
/
~
--==--
Cuodolupe
Victoria /o 0
-~
Molpais
Zinopecuoro

Pica de Orizoba Oerrumbodos

0~~~~~~--~~--~~--~~--~~~--~~--~~__J

0 2 4 6 8 10 12 14 16 18
Cs (ppm)

Figure 2.11. Bivariate plot of Cs versus Hf showing differentiation of the eighteen central Mexican
compositional groups. Probability ellipses at the 95% confidence level indicate the locations of the
subgroups.

and two of the Pachuca subsources are the only ones causing difficulty Santa
Elena has so few specimens that the Mahalanobis distance calculation extends its
probability hyperellipsoid to overlap with fringe samples from several sources.
This, of course, supports the earlier arguments that sources be thoroughly
characterized through analysis of a sufficient number of source samples. Until
more Santa Elena specimens are analyzed, we must critically examine all artifacts
assigned to Santa Elena. Fortunately, there is no archaeological evidence that
either Santa Elena or Derrumbadas sources were of any importance.
As mentioned earlier, the erroneous Pachuca classifications are caused by
the relative similarity of these subsources to one another on the first few PCS. In
order to eliminate errors in Pachuca assignments we can revert to inspection of
bivariate plots of Cs versus Hf or Cs versus Eu both of which easily differentiate
the Pachuca subsources. On the other hand, a procedure for robustly differenti-
ating all sources in central Mexico (after omission of Santa Elena and Derrum-
badas sources) can be developed by repeating our posterior classification
procedure using the first five PCS which describe about 98% of the total variance.
Table 2.5 presents the posterior classification results for central Mexican sources
when using the first five PCS. As shown, the classification is completely
successful as there are no misplaced specimens. Additional support for the
robustness of this procedure is found in Table 2.6 which presents the average
 VI
Table 2.4. Posterior Classification Matrix Based on the First Three PCs for Obsidian Source Specimens ~
from Central Mexican Sources-jackknife Method

Source n AV ZP PV GP DP pp LH OM MH TH PH ZH SHI SH2 SH3 UMI UM2 ZM

~
::l
('\
AV 14 14

ZP 31 29 2

PV 58 58
~
GP 23 23
~
DP 5
Q
pp 38 35
~
LH ~
OM 47 47 s
MH 19 19 ~
VI
TH 40 40
0
PH 10 9 ~
ZH 30 30 Q
SHI 129 125 4

SH2 II II

SH3 27 25
~
UMI 48 48 s
UM2 10 10
ZM 24 24
~
::l
Source abbreviations: AV = Altotonga, Ver.; ZP =Zaragoza, Pue.; PV = Pico de Orizaba, Ver.; GP =Guadalupe Victoria, Pue.; DP = Derrumbadas, ~
Pue.; PP = Paredon, Pue.; LH = Santa Elena, Hgo.; OM = Otumba, Mex.; MH = Malpais, Hgo.; TH = Tulancingo, Hgo.; PH = Tepalzingo, Hgo.; ZH
= Zacualtipan, Hgo.; SHI, SH2, & SH3 = Sierra de Pachuca, Hgo. subsources; UMI & UM2 = Ucareo, Mich. subsources; and ZM = Zinapecuaro,
Mich. ~
 ~

Table 2.5. Posterior Classification Matrix Based on the First Five PCs for Obsidian Source Specimens
from Central Mexican Sources-jackkni fe Method

n AV ZP PV GP pp OM MH TH PH ZH SHI SH2 SH3 UMI UM2 ZM

Source
AV 14 14

ZP 31 31

PV 58 58

GP 23 23

pp 38 38

OM 47 47

MH 19 19

TH 40 40

PH 10 10

ZH 30 30
....~
SHI 129 129

SH2 II II ~
~
27 27
SH3 !=I
48 48
UMI
10
UM2 10 ~
VI
24
ZM 24 8
l"l
~
t'l
....
~
 Vl
~

Table 2.6. Average Probabilities for Group Membership Based on the First Five PCs for Obsidian Sources Specimens
s::lMl
li
from Central Mexican Sources-jackknife Method

Source AV ZP PV GP PP OM MH TH PH ZH SH! SH2 SH3 UM! UM2 ZM

~
~

AV 54.292 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.00! 0.000 ~
ZP 0 000 50.29! 0.000 0.000 0.000 0.000 0.000 0 000 0.000 0.000 0.000 0.000 0.000 0.000 0.00! 0.000 g
PV 0.000 0.000 55.665 0.000 0 000 0.000 0 000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ~
GP 0.000 0.000 0.000 50.20! 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
pp 0.000 0.000 0.000 0.000 530!0 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.00! 0.000

OM 0.000 0.000 0.000 0.000 0.000 47.523 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 ~
~
MH 0.000 0.000 0.000 0.000 0.000 0.000 48.699 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Vl
TH 0.000 0.000 0.000 0.000 0.000 0.000 0.000 49.! 09 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0
PH 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 49.7!0 0.000 0.000 0.000 0.000 0.000 0.00! 0.000
~
Q
ZH 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 50.7!7 0.000 0.000 0.000 0.000 0.002 0.000

SHI 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 60.663 0.000 0.004 0.000 0.00! 0.000

SH2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 49.!59 0.000 0.000 0.00! 0.000 ~
SH3 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 49.373 0.000 0.003 0.000

UM! 0.000 0.000 0~ ~00 0~ MOO 0~ 0~ ~00 ~00 0~ 0~ ~00 ~~ 0~ 0.000

UM2 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 50.353 0.000 ~
ZM 0.000 0.000 o~ o~ ~oo o~ ~oo o~ ~oo o~ o~ o~ o~ o~ Mm 5!.523
::l
~

1.11
....
52 MICHAEL D. GlASCOCK et al.

probabilities for membership in each source group. (Ideally, the average prob-
ability of membership in one's own group will be -50%.) The results show that
when using five PCS, even in the worst case, the average probability of being
assigned to a wrong source group in central Mexico is less than 4 in 100,000.
As a test of our ability to use the latter procedure to reject specimens from
sources in other regions, the principal component scores for Guatemalan
source specimens were calculated. All of the Guatemalan source specimens
were then projected against the Mexican source groups and membership
probabilities calculated from Mahalanobis distances. Again, the procedure's
ability to reject wrong specimens were found to be an outstanding success. Two
Sansare specimens had probabilities of about 1 in 100,000 of belonging to the
Ucareo-2 subsource while all other Guatemalan source specimens were much
less probable of being associated with any of the eighteen central Mexican
sources.

Distinguishing between Individual Sources in Guatemala

Can we formulate a reliable procedure for differentiating between all sources
in Guatemala? What are its limitations? Does the procedure reject specimens from
Mexican sources?
The process used to examine data for the obsidian sources from Guate-
mala is analogous to that applied to the Mexican sources. Data for the twelve
Guatemalan source groups was extracted to yield a total of 13 7 specimens. Six
outlier specimens from five different sources were eliminated on the basis of a
preliminary screening of elemental means, ranges, and standard deviations. An
RQ-mode PCA was performed on the remaining 131 source specimens.
Table 2. ?lists the resulting eigenvalues, percent variance, and cumulative
variance described by successive components. Figures 2.12 and 2.13, respec-
tively, show the Q-mode and R-mode plots on components PC01 and PC02.
Confidence ellipses for several of the SMJ subsources and the San Bartolome
Milpas Altas source overlap one another in Figure 2.12. However, plots of
components PC03 and PC04 (not shown here) are successful in separating
these overlapping groups. In order of importance, we note that the elements
Co, Cs, Sb, and Mn make the largest contributions to the variance described
by the first two components in the Guatemalan source data .. Contrarily, the
REEs provide the least amount of source differentiation information. A bivari-
ate plot of Cs versus Co shown in Figure 2.14 successfully separates the
confidence ellipses for all twelve source groups.
Since the SMJ-4 subsource is characterized by only four specimens, we
are limited to use of the first two PCS (subsuming approximately 76% of the
variance) when calculating Mahalanobis distances and F-value probabilities for
all twelve sources. As shown in the Table 2.8, limiting our classification to the
first two PCS results in a large number of misclassifications, particularly into
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 53

Table 2. 7. Eigenvalues and Percentage of Variance Explained by

Principal Components Calculated from the Variance-Covariance
Matrix of Concentration Data in Log Base-10 ppm for the Twelve
Sources in Guatemala

Principal Cumulative
Component Eigenvalue %Variance %Variance

0.1351 55.55 55.55

2 0.0509 20.92 76.48

3 0.0416 17.09 93.57

4 0.0045 1.84 95.41

5 0.0032 1.31 96.72

6 0.0026 1.16 97.78

7 0.0021 0.87 98.64

8 0.0011 0.47 99.12

9 0.0005 0.23 99.34

10 0.0004 0.18 99.52

11 0.0004 0.16 99.68

12 0.0002 0.09 99.77

13 0.0002 0.07 99.84

14 0.0001 0.04 99.88

15 0.0001 0.04 99.91

16 0.0001 0.03 99.94

17 0.0000 0.02 99.96

18 0.0000 0.02 99.98

19 0.0000 0.01 99.99

20 0.0000 0.01 100.00

21 0.0000 0.01 100.00

the source at jalapa and the two San Martin Jilotepeque subsources with the
fewest specimens.
If the SMJ-4 subsource is omitted and three PCS (describing 94% of the
variance) are used the results are much more satisfactory. Tables 2.9 and 2.10
show the resulting classification matrix and average probabilities, respectively
 54 MICHAEL D. GLASCOCK et al .

...
0
ci
S'-IJ -5
h:tepeque@

N
@
0
Son Bartolome
ci MilpOll Altos

Lo;uno de Ayor-zo

0
C>
5'-IJ-6 (j
0
N
0
u
a. Sons ore ./'
S"J- 4

@
N
0
ci
I

...
0
J alapa ~
ci El Choyol
I
~
"'ci
0

I -0.06 - 0 .04 - 0 .02 0 .00 0 .02 0 .0 4 0 .06 0 .08

PC01

Figure 2.12. Q-mode bivariate plot on the first two principal components showing 95% confi-
dence ellipses for obsidian sources in Guatemala and Otumba regions.

ci
~ ~--~----~----~---,.---~----,-----~---,----,

...
0
ci
··- ....... ---
Co

..

0
0
0

N
0
...
0
u 0
a. I

"'ci
0

ci
I

<0

ci
I -o.3o -0.25 -o.2o - o . 15 -o. 10 - o .o5 -o.oo o .o5 o . 10
PC01

Figure 2.13. R-mode bivariate plot on the first two principal components showing scaled-loading
factors for elements in Guatemalan and Otumba obsidian sources.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 55

,. . ,Qj
"':
"'
0
,_;

Son Bartolome lr!Ailpos Altos

ea.
.e0 "'
u

"'0
lxtepeque

j j :-·
..
SMJ-J
·J-5 ~~tSMH Jolopo

~
0 SMJ-1
• SMJ- 2 (I Cnoyol
ct!l1P
' lo;uno de Ayorzo
0
02 3 4 5 6 7 8 9
Cs(ppm)

Figure 2.14. Bivariate plot of Cs versus Co showing differentiation of the twelve G uatemalan
compositional groups. Probability ellipses at the 95% confidence level indicate the locations of the
subgroups.

Table 2.8. Posterior Classification Matrix Based on the First Two PCs for Obsidian
Source Specimens from Guatemalan Sources-jackknife Method

Source 10 co JO sso SBM LAO SMJ1 SMJ2 SMJJ SMJ4 SMJ' SMJ6

10 12 12

C7

JO

sso 10 10

SBM 10

LAO 24 24

SMJI 28 23

SMJ2

SMJJ

SMJ4

SMJ'

SMJ6

Source abbreviations: IG = Ixtepeque; CG = El Chayal; JG = Jalapa; SSG = Sansare; SBM = San Bartolome Milpas
Altas; LAG = Laguna de Ayarza; SMJ I, SMJ2, SMB, SMJ4, SMJ5, & SMJ6 = San Martin Jilotepeqne subsources.
56 MICHAEL D. GLASCOCK et al.

Table 2.9. Posterior Classification Matrix Based on the First Three PCs for Obsidian
Source Specimens from Guatemalan Sources-jackknife Method

Source IG CO JO SSG SBM LAG SMJI SMJ2 SMJ3 SMJ5 SMJ6

10 12 1l

co
JG

SSG 10 10

SBM 10 10

LAG 24 24

SMJI 18 28

SMJl

SMJ3

SMJ5

SMJ6

None of the source specimens are assigned to a wrong source and, with the
exception of two SMJ entries in Table 2.10, the average membership prob-
abilities are less than 1 in 100, and the vast majority are far less than 1 in 1,000.
The smaller sizes of these Guatemalan source groups explains why these
probabilities are much greater than those we calculated for the central Mexican
sources. Even so, with probabilities of misclassification on the order of 1% or
less, we are generally satisfied.
To test the Guatemalan sourcing procedure's ability to reject specimens
from Mexican sources, principal component scores were calculated using the
three PC basis. The Mexican source specimens were then projected against the
Guatemalan sources and membership probabilities calculated from the Maha-

Table 2.10. Average Probabilities for Group Membership Based on the First Three
PCs for Obsidian Sources Specimens from Guatemalan Sources-jackknife Method

Source IG CG JG SSG SBM LAG SMJI SMJ2 SMJ3 SMJ5 SMJ6

IG 5l.l49 0.000 O.ot8 0.000 0.001 0.000 0.000 0.000 O.ot5 0.000 0.000
CG 0.000 49.819 0.011 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000

JG 0.000 0.057 52.040 0.000 0.000 0.000 0.000 0.003 0.100 0.000 0.000

SSG 0.000 0.001 0.002 52.246 0.000 0.000 0.000 0.001 0.001 0.000 0.000

SBM 0.000 0.000 0.254 0.000 49.860 0.000 0.000 0.042 1.828 0.000 0.000

LAG 0.000 0.000 0.003 0.000 0.000 45.909 0.000 0.001 0.000 0.000 0.000

SMJI 0.000 0.000 O.Q28 0.000 0.001 0.000 46.314 4.446 0.397 0.000 0.000

SMJ2 0.000 0.000 0.059 0.000 0.001 0.000 0.730 48.687 0.965 0.000 0.000

SMJ3 0.000 0.000 0;S86 0.000 0.299 0.000 0.000 0.835 49.891 0.000 0.000

SMJ5 0.000 0.000 0.120 0.000 0.000 0.000 0.000 0.014 0.012 50.750 0.000

SMJ6 0.000 0.000 0.014 0.000 0.000 o.ooo 0.000 0.003 0.000 0.000 48.879
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 57

lanobis distances. Again, the method is quite successful with specimens from
all but two sources having membership probabilities less than lin 1,000. The
sources at Otumba and Malpais have a few specimens with probabilities of
approximately lin 100 of being placed into the two Guatemalan source groups
with the fewest members Qalapa and SMJ-3). Therefore, we recommend use of
a probability cutoff limit of l% when using this procedure which would
virtually eliminate all possibility of misclassification.

Examining the Efficacy of an Abbreviated-NAA Procedure

Can we develop a sourcing method that relies on the elements available by an
abbreviated-NAA procedure? What are its limitations? As noted earlier, Pires-Fer-
riera (1973) used the short-lived elements Mn and Nato source artifacts with
success for most but not all sources in Mexico. Later, Stross et al. (1983)
employed a similar procedure on obsidian from Guatemala with five elements
being measured (Ba, Dy, K, Mn, and Na). Our short irradiation procedure which
measures Ba, Cl, Dy, K, Mn, and Na is similar to the method employed by Stross.
The procedure is described in detail elsewhere (Glascock et al. 1994). Although
it is anticipated the classification error rate will increase, the abbreviated
method offers the advantages of reduced costs and rapid analysis. Therefore,
we should test its applicability by performing a critical examination of the data
in our database. In particular, the examination should identify those sources
for which the method may be unsatisfactory.
For obsidian from Mexico, the variability of Ba, Dy, Mn, and Na concen-
trations between different sources suggests that these four elements might be
employed in an abbreviated procedure. To examine this possibility, data for the
569 Mexican source samples was extracted and log base-l 0 transformations
were made. A posterior classification matrix using these four elements was
generated as shown in Table 2.11. The number of misclassifications is small
and all occur for specimens within related regions (i.e., Orizaba specimens into
Guadalupe Victoria, Otumba specimens into Mal pais, Pachuca specimens from
one subsource into one another, and Ucareo specimens into Zinapecuaro).
More importantly, the average probabilities for these misclassification ranged
up to a maximum of 5%. Therefore, if we establish a cutoff of 5%, or more
conservatively 10%, for use of the abbreviated-NAA method, misclassifications
in central Mexico should be extremely rare. For example, Figures 2.15 and 2.16
show bivariate plots of Mn versus Na and Mn versus Dy for almost 700 artifacts
projected against the central Mexican sources to which they were sourced.
None of the obsidian sources not shown on these figures were linked to any of
these artifacts. Specimens with probabilities below our cutoff or with signifi-
cant probabilities in two or more Mexican sources can be resolved by submit-
ting them to the full-NAA and employing the more robust classification
procedures described earlier.
 ~

Table 2.11. Posterior Classification Matrix Based on Log Base-10 ppm Concentrations for Ba, Dy, Mn, and Na for Obsidian Specimens
from Central Mexican Sources-Jackknife Method

Source n AV ZP PV GP pp OM MH TH PH· ZH SHI SH2 SID UMI UM2 ZM

AV 14 14

ZP 31 31

PV 58 56 2

GP 23 23

pp 38 38

OM 47 46

MH 19 19

TH 40 40

PH 10 10

ZH 30 30 !;
SHI 129 128 I

SH2 11 11
~
~
SID 27 I 26
~
UMI 48 46 2

UM2 10 10 ~
Ill
ZM 24 24
8
(""\
~
Ill
...
~
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 59

.......

,
Guadalupe Vic toria
...
~
Pore don ~<~<,
Zorogozo

\:
"'...,
.
+
••
Ucoreo- 1
~ N
...;

~~-~
z"' leo de Ori zabo

•• +

"'r-i

":
N
~ Zocuoltipon
Otumbo

+
0
r-io 200 400 600 BOO 1000 1200 1400
Mn (ppm )

Figure 2.15. Bivariate plot of Mn versus Na showing artifacts projected against 95% confidence
ellipses for the eight important sources in central Mexico. To avoid data congestion, sources to
which artifacts were not assigned are not shown.

Zocuoltipon

~ Porodon

~ Zoro~ozo
~ + ~ Otu~<+ p;co do Orizobo

Ucoroo-1 (.7~
Cuodolupe Victoria
0~--~--~--L---~------~------~--~--~--~--~--~--~
0 200 400 600 BOO 1000 1200 1400
Mn (ppm)

Figure 2.16. Bivariate plot of Mn versus Dy showing artifacts projected against 95% confidence
ellipses for the eight important sources in central Mexico. Sources to which artifacts were not
assigned are not shown.
60 MICHAEL D. GLASCOCK et al.

Table 2.12. Posterior Classification Matrix Based on Log Base-10 ppm

Concentrations for Ba, Dy, Mn, and Na for Obsidian Specimens from
Guatamalan Sources-jackknife Method

Source IG CG JG SSG SBM LAG SMJI SMJ2 SMJJ SMJ5 SMJ6

IG 12 12

CG
JG

SSG 10 10

SBM 10 10

LAG 24 23

SMJI 28 26

SMJ2

SMJ3

SMJ5

SMJ6

A test of the same four elements in the Guatemalan sources results in the
classification matrix shown in Table 2.12. Misclassifications occur primarily
into the sources established by small numbers of specimens (i.e., El Chayal and
Jalapa) or from one San Martin jilotepeque subsource into another. The
probabilities for misclassification ranged up to 7% in the worst cases involving
the El Chayal and jalapa sources and 15% between the San Martinjilotepeque
subsources. As we have recommended for the Mexican sources, a conservative
probability cutoff of 10% should result in about 90% of all artifacts being
properly classified on the basis of data from an abbreviated analysis using
concentrations for Ba, Dy, Mn, and Na. Figure 2.17 shows a bivariate plot of
Mn versus Na for 500 artifacts from sites in Belize and Guatemala projected
against the three main Guatemalan sources to which they were sourced. As
expected, the number of artifacts requiring the full-NAA is quite small.
Finally, our work has shown that only about 10% of all artifacts in
Mexico and Guatemala are likely to require the full-NAA. Thus, the abbre-
viated-NAA method results in a significant savings in both time and cost will
result. These advantages make the abbreviated-NAA procedure very cost
competitive with XRE

CONCLUSION

In summary, the process we describe for obsidian sourcing studies should

identify methods which find that source specimens are properly assigned by
rejecting incorrect sources. This is accomplished by collecting and analyzing a
large number of specimens by measuring a large number of geochemical
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 61

..,...;
N
...;
..
...;

0
...;
l

(j)
S"'J-1

z"' en
N

co
<'<

...<'< ....
~ ~~--~~--~~--~~--~~~~--~-L--~~--~~
<'< 400
440 480 520 560 600 640 680 720
Mn (ppm)

Figure 2.17. Bivariate plot of Mn versus Na showing artifacts projected against 95% confidence
ellipses for the three important sources in Guatemala. Sources to which artifacts were not assigned
are not shown.

elements (correlated and uncorrelated) at high precision and using the data to
develop a powerful sourcing procedure. Once a satisfactory sourcing procedure
is identified, applications to artifacts on a step-by-step basis are possible: l) to
identify the specific region from which a specimen came by rejecting all other
regions, 2) to identify the individual sources by rejecting all other sources, and
3) if possible, to identify the specific subsource (quarry or outcrop) within the
source by rejecting all other subsources. The multivariate methods described
above offer a highly rigorous way to perform these operations.

ACKNOWLEDGMENTS

We acknowledge the support of three National Science Foundation grants

(BNS 79-15409 to Cobean, DBS 91-02016 to Glascock, and DBS 93-01152 to
Braswell) and a Fulbright scholarship to Braswell that supported various
aspects of this work. In no particular order, E.V Sayre, G. Harbottle, A.M.
Bieber, Jr., R.L. Bishop, and H . Neff are gratefully acknowledged for their
significant contributions in the advocacy of applying multivariate methods to
compositional studies over the past 25 years. The assistance of many other
colleagu es in Mexico, Guatemala, and the United States who provided artifacts
62 MICHAEL D. GLASCOCK et al.

for analysis are much appreciated. Finally, the encouragement and advice from
H. Neff who patiently reviewed this manuscript are much appreciated.

REFERENCES

Ahrens, L.H. 1954 The lognormal distribution of the elements. Geochimica et Cosmochimica Acta
5:49-73.
Asaro, F., Michel, H.V, Sidrys, R., and Stross, F. 1978 High-precision chemical characterization of
major obsidian sources in Guatemala. American Antiquity 43: 436-443.
Avila-Salinas, W 1975 Elementos trazas de algunas obsidianas Bolivianas, Centro de Investigacio-
nes Arqueologicas, publicacion No.4, La Paz, Bolivia, 16 pp.
Baxter, M.]. 1991 Principal component and correspondence analysis of glass compositions: an
empirical study. Archaeometry 33: 29-41.
Baxter, M.j. 1992 Archaeological uses of the biplot--a neglected technique? In Lock, G. and Moffett,
]., eds., Computer applications and quantitative methods in archaeology, 1991. British Archae-
ological Report International Series, S577, p. 141-148, Tempus Raparatum, Oxford.
Bettinger, R.L., Delacores, M.G., and jackson, R.]. 1984 Visual sourcing of central eastern California
obsidians. In R.E. Hughes, ed., Obsidian Studies in the Great Basin. Contributions of the
University of California Archaeological Research Facility 45: 63-78.
Bieber, A.M.Jr., Brooks, D.W, Harbottle, G., and Sayre, E.V 1976 Application of multivariate
techniques to analytical data on Aegean ceramics. Archaeometry 18: 59-74.
Bishop, R.L. and Neff, H. 1989 Compositional data analysis in archaeology. In R.O. Allen, ed.,
Archaeological Chemistry--Ill. American Chemical Society: Washington: 57-86.
Blackman, M.]. 1984 Provenance studies of Middle Eastern obsidian from sites in highland Iran.
In Lambert.] .B., ed., Archaeological Chemistry--lii Washington, American Chemical Society:
19-50.
Bowman, H.R., Asaro, F., and Perlman, I. 1973 On the uniformity of composition in obsidians and
evidence for magmatic mixing. journal of Geology 81:312-327.
Braswell, G.E. and Glascock, M.D. 1992 A new obsidian source in the highlands of Guatemala.
Ancient Mesoamerica 3: 47-49.
Burger, R.L., and Asaro, F. 1977 Trace element analysis of obsidian artifacts from the Andes: new
perspectives on pre-hispanic economic interaction in Peru and Bolivia. University of
California Lawrence Berkeley Laboratory preprint 6343: Berkeley, California.
Cann, ]. R. and Renfrew, C. 1964 The characterization of obsidian and its application to the
Mediterranean region. Proceedings of the Prehistoric Society 30: 111-133.
Cattell, R.B. 1952 Factor Analysis. Harper: New York, 462 pp.
Charlton, I. H., Grove, D.C., and Hopke, P.K. 1978 The Paredn, Mexico obsidian source and Early
Formative exchange. Science 201: 807-809.
Clark, ].E. 1981 Multi-faceted approach to the study of Mesoamerican obsidian trade: an example
from early Chiapas. Paper presented at the 46th annual meeting of the Society for American
Archaeology in San Diego, California.
Clark,j.E. 1988 The lithic artifacts of La Libertad, Chiapas, Mexico: an economic perspective. In
Papers of the New World Archaeological Foundation, No. 52., Provo, Utah.
Cobean, R.H., Coe, M.D., Perry, E.A.jr., Turekian, K.K., and Kharkar, D.P. 1971 Obsidian trade at
San Lorenzo Tenochtitlan, Mexico. Science 174: 141-146.
Cobean, R.H., Vogt, ].R., Glascock, M.D., and Stocker, T.R. 1991 High precision trace element
characterization of major Mesoamerican obsidian sources and further analyses of artifacts
from San Lorenzo Tenochtitlan, Mexico. Latin American Antiquity 2: 69-91.
Davis, ].C. 1986 Statistics and Data Analysis in Geology. New York, john Wiley and Sons.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 63

Duerden, P, Bird, ].R., Clayton, E., Cohen, D. D., and Leach, B.E 1984 Provenance studies of New
Zealand obsidian artifacts. Nuclear Instruments and Methods in Physics Research B3: 419-4 23.
Duranni, S., Khan, M.T., and Renfrew, C. 1971 Obsidian source identification by fission track
analysis. Nature 233: 242-245.
Elam,j .M. 1993 Post-Formative Obsidian Exchange in Oaxaca, Mexico. Ph.D. dissertation, University
of Missouri: Columbia.
Ericson, ].E. 1981 Exchange and Production Systems in Californian Prehistory: The Results of
Hydration Dating and Chemical Characterization of Obsidian Sources. British Archaeological
Reports, International Series 110.
Freter, A.C. 1992 Chronological research at Copan. Ancient Mesoamerica 3: 117-133.
Fuller, R.E. 1927 The mode of origin of the color of certain varicolored obsidians. journal ofGeolo~y
35: 570-573.
Gabriel, KR. 1971 The biplot-graphic display of matrices with application to principal components
analysis. Biometrika 58: 453-467.
Glascock, M.D. 1992 Characterization of archaeological ceramics at MURR by neutron activation
analysis and multivariate statistics. In Neff. H., ed., Chemical Characterization of Ceramic
Pastes in Archaeology. Madison, Prehistory Press.
Glascock, M.D. and Anderson M.P. 1993 Geological reference materials for standardization and
quality assurance of instrumental neutron activation analysis. journal ofRadioanalytical and
Nuclear Chemistry, Articles 174: 229-242.
Glascock, M.D., Elam, ].M., and Aoyama, K. 1990 Provenience analysis of obsidian artifacts from
the La Entrada Region, Honduras. In Pernicka, E. and Wagener, G.A., eds., Archaeometry
'90. Basel, Birkhauser Verlag: 365-374.
Glascock, M.D., Elam, ].M., and Cobean, R.H. 1988 Differentiation of obsidian sources in
Mesoamerica. In Farquhar, R.M., Hancock, R.G.V, and Pavlish, L.A., eds.,Archaeometry '88.
Toronto, Archaeometry Laboratory, University of Toronto: 245-251.
Glascock, M.D., Neff, H., Stryker, K.S., and johnson, T.N. 1994 Sourcing archaeological obsidian
by an abbreviated NAA procedure. journal ofRadioanalytical and Nuclear Chemistry, Articles
180:29-35.
Graham, C. C., Glascock, M.D., Carni, ].]., Vogt, ].R., and Spalding, T.G. 1982 Determination of
elements in National Bureau of Standards' geological standard reference materials by
neutron activation analysis. Analytical Chemistry 54: 1623-1627.
Harbottle, G. 1976 Activation analysis in archaeology. In Newton, G.W.A., ed., Radiochemistry.
London, The Chemical Society: 33-72.
_ _ 1982 Chemical characterization in archaeology. In Ericson,] .E. and Earle, T. K., eds., Contexts
for Prehistoric Exchange. New York, Academic Press: 13-51. .
Hughes, R.E. 1982 Age and exploitation of obsidian from the Medicine Lake Highland, California.
journal of Archaeological Science 9: 173-185.
_ _ 1984 Obsidian sourcing studies in the Great Basin: problems and prospects. In Hughes, R.E.
ed., Obsidian Studies in the Great Basin. Contributions of the University of California
Archaeological Research Facility No. 45: Berkeley, California: 1-19.
_ _ 1986 Diachronic Variability in Obsidian Procurement Patterns in Northeastern California and
Southcentral Oregon. Berkeley, University of California Press.
_ _ 1988 The Coso Volcanic Field reexamined: implications for obsidian sourcing and hydration
dating research. Geoarchaeology 3: 253-265.
Huntley, D.]. and Bailey, D.C. 1978 Obsidian source identification by thermoluminescence.
Archaeometry 20: 159-170.
Hurtado de Mendoza, L. and jester, W.A. 1978 Obsidian sources in Guatemala: a regional approach.
American Antiquity 43: 424-435.
Jackson, T.L, and Love, M.W. 1991 Blade running: middle preclassic obsidian exchange and the
introduction of prismatic blades at La Blanca, Guatemala. Ancient Mesoamerica 2: 4 7-59.
64 MICHAEL D. GLASCOCK et al.

Klecka, W.R. 1980 Discriminant analysis. Sage University Paper, Quantitative Analysis in the Social
Sciences, Series No. 07-019. Beverly Hills, California, Sage Publications.
Lachenbruch, P.A. 1975 Discriminant Analysis. New York, Hafner.
Leach, F. 1973 Obsidian from the Chatham Islands. New Zealand Archaeological Association
Newsletter 16: 104-106.
Leach, F. and Manly, B. 1982 Minimum Mahalanobis distance functions and lithic source charac-
terization by multi-element analysis. New Zealand journal of Archaeology 4: 77-109.
Leach, E and Warren, S.E. 1981 Neutron activation analysis of New Zealand and oceanic obsidians:
towards a simple screening technique. In Leach, E and Davidson,]., eds., Archaeological
Studies of Pacific Stone Resources. Oxford, BAR International Series 104, Oxford, England:
151-166.
Leach, B.E, Warren, S.E., and Fankhauser, B. 1978 Obsidian from the far north of New Zealand:
A method of sourcing based on natural radioactive emission. New Zealand journal of Science
21: 123-128.
Longworth, G. and Warren, S.E. 1979 The application of miissbauer spectroscopy to the charac-
terisation of western Mediterranean obsidian. journal of Archaeological Science 6: 179-193.
McDougall,J.M., Tarling, D.H., and Warren, S.E. 1983 The magnetic sourcing of obsidian samples
from Mediterranean and Near Eastern sources. journal of Archaeological Science 10:441-452.
Moholy-Nagy, H. and Nelson, F.W. 1990 New data on sources of obsidian artifacts from Tikal,
Guatemala. Ancient Mesoamerica 1: 71-80.
Neff, H. 1990 A series of routines written in GAUSS language (unpublished).
_ _ 1994 RQ-mode principal components analysis of ceramic compositional data. Archaeometry
36: 115-130.
Neff, H., Bishop, R.L., and Arnold, D.E. 1988 Reconstructing ceramic production from ceramic
compositional data: an example from Guatemala. journal of Field Archaeology 15: 339-348.
Neff, H., Bove, E]., johnson, T.N., and Arroyo, B. 1993 Fachimentos a traves de hidratacion de
obsidiana en !a costa sur de Guatemala. Apuntes Arqueologicos 2: 55-79.
Nelson, EW. 1985 Summary of the results of analysis of obsidian artifacts from the Maya lowlands.
Scanning Electron Microscopy II: 631-649.
Perlman, I. and Asaro, E 1969 Pottery analysis by neutron activation. Archaeometry 11: 21-52.
Picon, M., Carre, C., Cordoliani, M.L., Vichy, M., Hernandez, ].A., and Mignard, ].L. 1975
Compositions of the La Graufesenque, Banassac and Montans terra sigillata. Archaeometry
17: 191-199.
Pires-Perreira, ].W. 1973 Formative Mesoamerican Exchange Networks with Special Reference to the
Valley of Oaxaca. Memoirs of the Museum of Anthropology No. 7, Ann Arbor, University
of Michigan.
Reeves, R.D. and Armitage, G.C. 1973 Density measurements and chemical analysis in the identifi-
cation of New Zealand archaeological obsidians. New Zealand journal of Science 16: 561-5 72.
Sappington, R.L. 1981 The archaeology of the Lydie Gulch Site (10-AA-72): prehistoric occupation
in the Boise River Canyon, southwestern Idaho. University of Idaho Anthropological
Research Manuscript Series 66. Moscow, Idaho, University of Idaho.
_ _ 1984 Procurement without quarry production: examples from southwestern Idaho. In
Ericson, j.E. and Purdy, B.A., eds., Prehistoric Quarries and Lithic Production. Cambridge,
Cambridge Umversity Press: 23-34.
Sayre, E.V. 1975 Brookhaven procedures for statistical analyses of multivariate archaeometric data.
Brookhaven National Laboratory Report BNL-21693, New York.
Shackley, M.S. 1988 Sources of archaeological obsidian in the Southwest: an archaeological,
petrological, and geochemical study American Antiquity 53: 752-772.
Skinner, C. 1983 Obsidian Studies in Oregon: An Introduction to Obsidian and Investigation of Selected
Methods of Obsidian Characterization Utilizing Obsidian Collected at Prehistoric Quarry Sites
in Oregon. M.A. thesis, University of Oregon, Eugene.
SYSTEMATIC APPROACH TO OBSIDIAN SOURCE CHARACTERIZATION 65

Sneath, P.H.A. and Sokal, R.R. 1973 Numerical Taxonomy. San Francisco, W.H. Freeman and
Company
Stark, B.L., Heller, L., Glascock, M.D., Elam, j.M., and Neff, H. 1992 Obsidian-artifact source
analysis for the Mextequilla region, south-central Veracruz, Mexico. Latin American Antiq-
uity 3: 221-239.
Stevenson, C.M. and McCurry, M.O. 1990 Chemical characterization and hydration rate develop-
ment for New Mexican obsidian sources. Geoarchaeology 5: 149-170.
Stocker, T.R. and Co bean, R.H. 1984 Preliminary report on the obsidian mines at Pico de Orizaba,
Veracruz. In Ericson, j.E. and Purdy, B.A., eds., Prehistoric Quarries and Lithic Production,
Cambridge, Cambridge University Press: 83-95.
Stross, EH., Sheets, P., Asaro, E, and Michel, H.V. 1983 Precise characterization of Guatemalan
obsidian sources, and source determination of artifacts from Quirigu. American Antiquity
48: 323-346.
Vogt,j.R., C. C. Graham, M.D. Glascock, and R.H. Co bean 1982 A study of Mesoamerican obsidian
sources using activation analysis. journal ofRadioanalytical Chemistry 69: 271-289.
Ward, G .K. 1974 A systematic approach to the definition of sources of raw material. Archaeometry
16: 41-53.
Wheeler, M.E. and Clark, D.W. 1977 Elemental characterization of obsidian from the Koyukuk
River, Alaska, by atomic absorption spectrophotometry Archaeometry 19: 15-31.
Wilson, A.L. 1978 Elemental analysis of pottery in the study of its provenance: a review. .Journal
of Archaeological Science 5: 219-236.
 Chapter3

Mediterranean Islands and

Multiple Flows
The Sources and Exploitation of Sardinian
Obsidian

RoBERT H. TYKoT

ABSTRACT

Obsidian tools found at prehistoric sites in the Mediterranean are evidence of

a complex series of activities including procurement and transport of the raw
material from island sources, production and distribution of cores or finished
tools, and consumption and eventual disposal. Recent developments in the
study of the western Mediterranean sources include the localization and
characterization of five Monte Arci (Sardinia) sources used in antiquity; more
detailed survey and characterization of the Lipari, Pantelleria, and Palmarola
sources; and the provenance determination by ICP-mass spectrometry and
electron probe microanalysis of more than 700 tools from sites in Italy and
France doubles the provenance data previously available for reconstructing
western Mediterranean exchange systems and provides a framework for inter-
preting the specific cultural context in which their acquisition was embedded.
Furthermore, the analysis of samples from each stratigraphic level of multi-
component sites such as Grotta Filiestru in Sardinia and Basi in Corsica allows
for temporal control over three millennia. While differential use of the various

ROBERT H. TYKOT • Department of Anthropology, University of South Florida, Tampa,

Florida.

Archaeological Obsidian Studies, edited by Shackley.

Plenum Press, New York, 1998.

67
68 ROBERT H. TYKOT

obsidian sources has been recognized in peninsular Italy, it is apparent now

that distribution patterns vary both geographically and chronologically, and in
ways not explained by differences in accessibility or functional suitability
between sources.

INTRODUCTION

The volume Advances in Obsidian Glass Studies (Taylor 1976) included a

detailed assessment of the state of obsidian characterization studies in the
Mediterranean at that time (Dixon 1976). By then, the first successful applica-
tion of trace-element sourcing ( Cann and Renfrew 1964) had spawned a number
of more in-depth studies focusing on particular source regions (e.g. Hallam et al.
1976; Renfrew and Dixon 1976). In the western Mediterranean, the island
sources of Lipari (Li), Palmarola (PI), Pantelleria (Pa), and Sardinia (SA, SB, SC)
could be differentiated from one another, and from the eastern Mediterranean
sources of Melos (Sta Nychia and Demenegaki) and Giali (Fig. 3.1).
Dixon (1976: 289) identified four stages of increasing sophistication of
obsidian characterization studies, which remains a useful measure for archae-

I~

Figure 3.1. Obsidian sources in the Mediterranean region, with important sites mentioned in the
text.
MEDITERRANEAN ISLANDS AND MULTIPLE FLOWS 69

ologists interested in interpreting obsidian source data. At that time, prove-

nance studies in the western Mediterranean region had not yet reached the
routine stage where definitive results are produced, since the "source and
sample data sets [for Sardinia] contained more diversity than is consistent with
their relating to a single flow." Differentiating the Sardinian flows is important
because it is a very large island (24,000 km 2 ) and, unlike the other Mediterra-
nean source islands, was inhabited by an indigenous culture predating the
earliest obsidian exploitation.
Obsidian has been identified at some 1000 western Mediterranean sites
(Pollmann 1993), but none are demonstrably earlier than the Neolithic (ca.
6000-3000 BC). Obsidian is not even found in the pre-neolithic levels at
sites in Sardinia and Corsica, where sea voyages would not have been an
issue. It is thus apparent that obsidian served a social as well as functional
role, perhaps as an ethnic identifier for populations culturally diversifying
due to increased sedentism (Tykot 1996). This hypothesis is supported by
the closed nature of the western Mediterranean obsidian exchange system,
with only two pieces of Carpathian obsidian identified at north Adriatic sites
near Trieste (Randle et al. 1993), and the only sure pieces of Melos obsidian
from undatable contexts at Grotta del Leone near Pisa (Bigazzi et al. 1992b;
1986; Bigazzi and Radi 1981).
In the past twenty years, considerable effort has been expended to clarify
the Sardinian source situation, and to find an optimal range of techniques for
determining the provenance of archaeological artifacts. In this time, some 60
papers reporting analyses of geological and archaeological samples of western
Mediterranean obsidian have been written (references in Tykot 1995). Much
recent work has also been done in central Europe (Bigazzi et al. 1990; Williams
Thorpe et al. 1984), the eastern Mediterranean (Torrence 1986), and the Near
East (Keller and Seifried 1990), with the latest results presented in no less than
12 papers and posters at the 1994 International Archaeometry Symposium in
Ankara, Turkey (cf. Phillips 1992 and most recently Williams-Thorpe 1995 for
a review of European and Near Eastern obsidian provenance studies).
For the western Mediterranean, recent developments include: (1) the
localization and characterization of five Monte Arci (Sardinia) sources used in
antiquity (Tykot 1996; 1995; 1992; Herold 1986; Francaviglia 1984; Mackey
and Warren 1983); (2) more detailed survey and characterization of the Lipari
(Cortese et al. 1986; Pichler 1980), Pantelleria (Francaviglia 1988; Mahood and
Hildreth 1986) and Palmarola (Herold 1986) volcanic sources; (3) the ability
to differentiate all of the Mediterranean island sources, including 4 of 5
Sardinian subsources, using major/minor element composition; and (4) signifi-
cant numbers of chemical analyses of western Mediterranean archaeological
artifacts (Tykot 1996; 1995; Ammerman and Polglase 1996; 1993; Crisci et al.
1994; Randle et al. 1993; Bigazzi et al. 1992b; Ammerman et al. 1990; Fran-
caviglia 1988; Francaviglia and Piperno 1987; Crummett and Warren 1985;
70 ROBERT H. TYKOT

Michels et al. 1984; Williams Thorpe et al. 1984; 1979; Mackey and Warren
1983). These data permit not only the reconstruction of prehistoric exchange
networks but also more accurate obsidian hydration rate measurements for
refining cultural chronologies (Dyson et al. 1990; Michels 1987).

MULTIPLE METHODS

Archaeological interest in source tracing has fueled many experimental

efforts at finding analytical methods appropriate for obsidian artifacts. Analysis
of archaeological materials must always take into account the size of the artifact,
whether a sample may be removed, and any factors which might affect the
interpretation of analyses of ancient samples against a database of modern
geological specimens. In addition, the cost and time of analyses should not
dilute a provenance study. to the point where insufficient data are produced to
answer particular archaeological questions.
Neutron activation analysis and X-ray fluorescence are by far the most
common-and successful-geochemical techniques used to source obsidian,
relying on inter-source variation in trace element content being greater than
intra-source heterogeneity. Attempts at using Mossbauer spectroscopy (Aramu
et al. 1983; Longworth and Warren 1979), magnetic parameters (McDougall et
al. 1983) and electron spin resonance (Mello 1983) have proven unsuccessful
at differentiating all four island sources; refractive index, density, thermolumi-
nescence, and optical transmission cannot differentiate all of the Monte Arci
flows (Herold 1986). Strontium isotope ratios, in combination with rubidium
and strontium concentrations, appear effective (Gale 1981), but this approach
is too costly and time-consuming for the routine analysis of artifacts. Fission-
track dating has also been widely used to source western Mediterranean
obsidian artifacts on the basis of their differing geological ages (Bigazzi et al.
1992a; Arias et al. 1986; 1984; Bigazzi and Radi 1981; Arias-Radi et al. 1972);
the multiple Sardinian sources cannot be resolved, however, by this method
since they all appear to be about 5 million years old (see also the potassium-
argon determinations by Belluomini et al. 1970; Bigazzi et al. 1976; and di Paola
et al. 1975). As we will see, chronological and geographic differences in the
exploitation of the Sardinian sources existed, and raise important questions
about neolithic exchange mechanisms in northern Italy where multiple island
sources were used simultaneously.

MULTIPLE FLOWS: THE MONTE ARCI SOURCES

The presence of multiple obsidian flows in the Monte Arci region of

Sardinia was suggested long before Cann and Renfrew's ( 1964) analyses of four
MEDITERRANEAN ISLANDS AND MULTIPLE FLOWS 71

artifacts fell into two distinct groups. De la Marmora (1839-40) mentions

several places on Monte Arci where obsidian could be found in situ; both
translucent and opaque types of obsidian were explicitly recognized at the
archaeological site of Monte Urpinu (Loddo 1903: 47); and Puxeddu (1958)
identified 4 sources, ll collection centers, 74 workshops, and 157 additional
sites with obsidian scatters in his survey of the Monte Arci region. Three
chemical subgroups (SA, SB, SC) were later confirmed by Hallam's (1976;
Hallam et al. 1976) analysis of western Mediterranean obsidian artifacts for his
Master's thesis at Bradford University, although geological material from only
one source (SA) was analyzed. Four independent studies have since been done
to locate and characterize the different Sardinian sources represented in the
archaeological record.
Mackey did a detailed survey of obsidian sources on Monte Arci for her
doctoral dissertation in geology at the University of Nottingham, but in the end
her thesis was not accepted. The only available data come from a brief
conference paper in which three chemical subgroups, determined by X-ray
fluorescence analyses, are associated with specific source areas on Monte Arci
(SA = Conca Cannas, SB = Santa Maria Zuarbara, SC = Perdas Urias) although
in situ material was not located at Perdas Urias (Mackey and Warren 1983).
The main chemical subgroups thus correspond to the geographical source
locations identified by Puxeddu. The SB source was also found to consist of at
least two chemically differentiable outcrops (Mt. Sparau, Cucru Is Abis), both
represented among the 51 archaeological artifacts analyzed in Mackey and
Warren's ( 1983) study. Data for individual analyses are not given, however, and
the number of geological samples tested remains unknown.
Francaviglia ( 1984) analyzed a total of 172 geological specimens from 5
source locations in Sardinia, but gave little information about the deposits
themselves, and did not publish individual sample data either. Only one of the
SB outcrops was tested, and two of the chemical source groups (both found in
a detrital deposit at Cave della Ceca) represent obsidian nodules too small to
have been utilized for artifacts. Francaviglia's main contribution was the
identification of two SC subgroups, neither found in situ, and the demonstra-
tion that all of the main Mediterranean obsidian sources (SA, SB, SC, Li, Pa, PI,
Melos, Giali) could be chemically differentiated on the basis of major/minor
elements alone.
Herold (1986) conducted an extremely thorough survey and analysis of
the Sardinian (and Palmarola) sources for his doctoral dissertation in geological
science at the University of Karlsruhe, but these data also have not been
formally published. For Monte Arci, he identified 4 chemical subgroups among
numerous localities with small unworkable nodules, in addition to the 4 usable
sources identified by Mackey and Warren (1983).
Finally, this author (Tykot 1992; 1995; 1996) has also completed an
extensive program of survey and analysis of the obsidian sources in Sardinia,
72 ROBERT H. TYKOT

confirming the archaeological use off others of unworkable size or quality.

Importantly, the SC source was finally located in situ. In contrast to the
preceding studies, this research was designed to demonstrate the utility of
low-cost, major-element analysis for determining the provenance of hundreds
of artifacts from stratigraphically excavated archaeological sites. It was also
discovered that obsidian artifacts from the western Mediterranean could be
visually assigned a provenance with a high degree of accuracy. In a blind
test, visual determination of provenance was recorded prior to chemical
analysis for nearly 600 artifacts. No errors were made distinguishing between
Lipari, Sardinia, and Pantelleria obsidian (no Palmarola obsidian was present);
the only erroneous attributions were among the Sardinian sources, for which
68% were correctly identified to a specific Monte Arci source, 16% were
correctly narrowed down to one of two specific sources, and 16% were
incorrectly attributed (Tykot 1995). Since many of the incorrect attributions
were bi-directional between types SA and SB, the frequency of each source
represented in an entire lithic assemblage could be estimated with the same
degree of confidence associated with the chemical analysis of a selected
sample of artifacts.

ANALYSIS OF GEOLOGICAL AND ARCHAEOLOGICAL MATERIAL

A complete account of the geological survey has been given elsewhere

(Tykot 1992; 1995). Hundreds of hand samples were collected from the
localities shown in Fig. 3.2, visually examined, and a subset selected for
chemical analysis. At first, 186 geological specimens and 33 archaeological
artifacts were analyzed for 38 major and trace elements by inductively coupled
plasma-mass spectrometry to explore whether additional subgroups could be
chemically distinguished. For archaeologists, ICP-MS has several important
advantages over neutron activation and X-ray fluorescence analysis: 1) only a
tiny powdered sample is required, so the technique is minimally destructive to
valuable artifacts; 2) the large number of elements that can be accurately and
precisely analyzed is particularly important for characterization and prove-
nience studies; 3) isotope ratio measurements to 3 significant figures are
possible without extensive sample preparation; and 4) the combination of small
sample size and low per-sample cost allows assemblages of artifacts rather than
individual objects to be studied (Tykot and Young 1996). The number of
elements that can be accurately and precisely analyzed is unmatched by any
other single technique, since the problem of mass spectral overlap is minimal,
in contrast to the overlapping energy or wavelength spectra produced by other
techniques, including ICP-atomic emission spectroscopy. Like ICP-AES and
other solution-based methods, the obsidian sample must be pulverized and
dissolved by hydrofluoric acid digestion or by fusion with a lithium metaborate
MEDITERRANEAN ISLANDS AND MULTIPLE FLOWS 73

1 [ill Perlite & Obsidian

2 Q Rhyolite & Rhyodacite

3~ Andesite & Dacite

4 Trachyte & T rach. rh, olite

5 ~ Basalt

6~ ubmarine Basic La as

7 ~ Alluvial Sediments

st8J Marine Sediments

Figure 3.2. Geological map of the Monte Arci region of Sardinia, showing locations mentioned in
the text.

flux. No sample preparation at all is necessary if a la ser ablation device is used,

at least for artifacts small enough to fit into the sample chamber.
Cluster analysis of the lCP-MS obsidian data did not suggest the existence
of additional, useful subdivisions beyond those already established by Mackey
and Warren ( 1983) and Francaviglia (1984), while stepwise discriminant
analysis provides clear separation of the five g roups represented by both
geological and archaeological samples (Fig. 3.3) . A description of each ty pe
and its geological source follows :
74 ROBERT H. TYKOT

8
• SA
• •
r- ..
SBI

t oo:t
4
0 SB2

0
.~8
co
0
SCI
oO 0
0
~ S A
~A 0
QJO 0

,.
;~ 0 SC2
•
00 ... 0

oo

-8

-8 0 4 8 12
Disaiminant Function I
Figure 3.3. Plot of first two discriminant functions for JCP-MS analyses of Monte Arei (Sardinia)
obsidian.

1. SA. Type SA obsidian is very glassy, black in color but highly translu-
cent, with nodules up to 40 em found in situ within perlite beds near
Conca Cannas and Su Paris de Monte 8ingias. Individual microlite
crystals are visible in transmitted light, often with some flow orien-
tation.
2. S81. Type S81 is less glassy and black, but usually opaque; it is found
in situ, also in a perlite matrix, at the higher elevations of Punta Su
Zippiri and Monte Sparau North, and in the form of bombs up to 30cm
in length on the slopes of Cuccuru Porcufurau.
3. S82. Type S82 is often very glassy, but ranges from virtually transparent
to nearly opaque; it is sometimes characterized by the frequent pre-
sence of white phenocrysts up to 2mm in diameter. Microlites are
usually not present in the transparent variety. Type S82 occurs in large
blocks (occasionally up to 1 meter in length) within a perlite matrix
near Cucru Is Abis, Seddai, Conca S'Ollastu, and 8runcu Perda Crobina.
4. SCI. Type SCI is less glassy, black but frequently with well-defined
external gray bands, and totally opaque. Rare pieces have red streaks
or are partially transparent but tinted brown. This type can be found
in situ in a perlite seam along the ridge from Punta Pizzighinu to
Perdas Urias on the northeastern side of Monte Arci, in blocks up to
20cm in length.
MEDITERRANEAN ISlANDS AND MULTIPLE FLOWS 75

5. SC2. Type SC2 is visually indistinguishable from SCI, and differs

chemically only in its trace concentrations of strontium and a few
other elements. It has only been found in detrital deposits between
Perdas Urias and Santa Pinta, mixed with material of SCI type. For
these reasons, I consider the characterization of material as simply SC
sufficient for archaeological purposes, since it is extremely unlikely
that one type could have been selected over the other for making lithic
artifacts.
Since all useful Mediterranean source distinctions could be made on the
basis of major/minor element concentrations, electron probe microanalysis
using wavelength dispersive X-ray spectrometers was selected as the method
of choice for analyzing large numbers of archaeological artifacts since the
precision of the microprobe is superior to laser ablation ICP-MS, only a tiny
I-2mm sample needs to be removed, sample preparation is minimal, and the
per-sample cost is equally low-a fraction of the price of XRF or NAA (see
Merrick and Brown I984 for another example). Up to I8 samples were mounted
in one-inch epoxy disks, polished flat, and analyzed for II major/minor
elements using a beam diameter of 40 microns and counting times of IO to 80
seconds per element. Typically, three points on each sample were analyzed to
account for any heterogeneity. Table 3.I shows the average concentrations for
each source, for nearly 2000 analyses of over 700 specimens. Discriminant
analysis of major/minor element data provides highly confident attribution of
obsidian artifacts to the four main Sardinian sources (SA, SBI, SB2, SC) as well
as to all of the other Mediterranean island sources, including two subgroups
for Melos (Sta Nychia and Demenegaki) and three for Pantelleria (Balate dei
Turchi, Lago di Venere, and Gelkhamar).

Table 3.1. Average Major Element Concentrations of Monte Arci Obsidian Sources,
Based on Nearly 2000 Electron Probe Microanalyses
Source Si02 Al 20 3 Ti02 Fe20 3 MgO CaO Na20 K,O MnO P205 BaO Total
SA ave 74.72 13.40 0.09 1.25 0.08 0.59 3.44 5.26 0.08 0.06 0.02 99.00
sd 0.26 0.15 0.01 0.09 0.01 0.04 0.16 0.22 O.Ql 0.01 0.02
n 207 207 207 207 207 207 207 207 60 60 202
SBI ave 73.87 13.63 0.17 1.33 0.12 0.75 3.38 5.55 0.10 0.04 0.05 99.00
sd 0.35 0.10 O.o3 0.20 0.04 0.06 0.10 0.20 O.o2 O.Ql 0.02
n 26 26 26 26 26 26 26 26 21 21 26
SB2 ave 75.05 12.97 0.13 1.17 0.11 0.57 3.34 5.51 0.08 0.04 0.02 99.00
sd 0.33 0.15 O.o2 0.17 0.02 0.02 0.22 0.35 0.01 O.Ql 0.02
n 130 130 130 130 130 130 130 130 81 81 130
sc ave 72.71 13.92 0.27 1.53 0.21 0.88 3.30 5.90 0.14 0.03 0.11 99.00
sd 0.37 0.19 O.o3 0.21 0.07 0.10 0.20 0.30 O.Ql O.Ql O.o3
n 341 342 342 339 339 342 342 342 120 120 340

n =number of individual specimens tested.

76 ROBERT H. TYKOT

THE DISTRIBUTION OF SARDINIAN OBSIDIAN

Obsidian artifacts from more than 50 archaeological sites in Sardinia,

Corsica, and northern Italy have been chemically and/or visually analyzed by
this author, and add to the existing body of data (compiled in Tykot 1995,
appendices A and B) produced primarily by neutron activation at Bradford
(Hallam et al. 1976; Williams Thorpe et al. 1979; 1984; Mackey and Warren
1983), Milan (Ammerman et al. 1990; Ammerman and Polglase 1993; 1995),
Pavia (Bigazzi et al. 1992b; 1986) and Birmingham (Randle et al. 1993), and
elsewhere by atomic absorption spectroscopy (Michels et al. 1984), X-ray
fluorescence (Dyson et al. 1990; Crisci et a!. 1994), and fission-track dating
(Arias et al. 1986; 1984; Bigazzi and Radi 1982; Arias-Radi et al. 1972).
The most significant results are obtained through comparison of the
relative importance of each Sardinian source at different sites and in different
cultural phases, and in the type of artifacts and/or lithic reduction debris
represented. It has been suggested that in the Early Neolithic (ca. 6000-5000
BC), less-organized selection of source material would result in the presence
of obsidian flakes from multiple sources, and in its relatively inefficient use; by
the Late Neolithic (ca. 4000-3000 BC), procurement would have been better
organized, focusing on the higher quality SA obsidian, and featuring more
efficient reduction technology in the production of cores and blades (Hur-
combe and Phillips 1995). Such a shift is observed at Arene Candide, where
less-intensive, on-site production of tools from multiple obsidian sources (SA,
SB, SC, PI) in the Early Neolithic is followed by nearly exclusive importation
of blades of high-quality Lipari obsidian in the Late Neolithic (Ammerman and
Polglase 1993; 1996).
This pattern is not paralleled in Sardinia and Corsica. Type SB1 obsidian
was not used often, but nevertheless is present in archaeological assemblages.
Type SA, SB2, or SC may be the most frequent at any one site in a given
chronological period. Here, I focus on two of the most important stratigraphi-
cally excavated sites, Grotta Filiestru (Mara) in Sardinia (Trump 1983) and
Basi (Serra-di-Ferro) in Corsica (Bailloud 1969a; 1969b), from which 300
pieces of obsidian have been analyzed. At Grotta Filiestru, the use of clear,
glassy type SB2 obsidian from the western flanks of Monte Arci decreases
over 4 Neolithic cultural periods, while the use of opaque, less-glassy type
SC obsidian from the northeastern part of Monte Arci increases (Fig. 3.4);
type SA is never more than 20% of the assemblage. In contrast, type SA
accounts for an average of 40% of the assemblage at Basi, both SB varieties
are never important, and this pattern does not change for 8 stratigraphic
levels including the Early (Cardial) and Late (Basien) Neolithic (Fig. 3.5), a
span of approximately 2000 years. This finding is contrary to the suggestion,
based on a limited number of analyses (Hallam et al. 1976), that type SB was
the most frequently used obsidian source in Corsica. Type SB obsidian is also
MEDITERRANEAN ISLANDS AND MULTIPLE FLOWS 77

%
27 20 29 14
100

80

60

40

20

Cardial Filiestru Bonu lghinu Ozieri

Culture Phase
Figure 3.4. Neolithic obsidian use in 4 culture phases at Grotta Filiestru (Mara, Sardinia). Cardia!
and Filiestru are Early Neolithic, Bonu Ighinu is Middle Neolithic, and Ozieri is Late Neolithic.
Number at top of each bar is number of pieces analyzed.

%
100
• sc
0 SB2
• SBI
80 OSA

60

20

oLL~-L~~~L-~~~~~~~L-s~d~~~s~~~s~
b~~~s.~~
7 6.3 6 .2 6.1 5e e
Stratigraphic Level
Figure 3.5. Neolithic obsidian use in 9 stratigraphic levels at Basi (Serra-di-Ferro, Corsica). Levels
7 and 6 are Early Neolithic; level 5 is Late Neolithic. Number at top of each bar is number ofpieces
analyzed.
78 ROBERT H. TYKOT

%
27 72 26
100
II PI
• sc
DSB
80 Ei]SA

60

40

20

0
Filiestru Basi Arene Candide
Early Neolithic Sites
Figure 3.6. Early Neolithic obsidian use at Grotta Filiestru. Basi. and Arene Candide. Number at
top of each bar is number of pieces analyzed. Arene Candide data from Ammerman and Polglase
(1993; 1996).

infrequent at sites in southern Sardinia, and even at sites near the Monte
Arci source zone.
Comparison of obsidian source frequencies in the Early Neolithic be-
tween Grotta Filiestru, Basi, and Arene Candide (Liguria) suggest that different
obsidian use patterns existed in different geographic regions by contemporary
Cardial Impressed Ware culture groups (Fig. 3.6). The importance of type SB
obsidian at Arene Candide, but its low frequency at Basi, implies that obsidian
may not have always been dispersed through simple down-the-line transactions
from the source zone (see Tykot 1996). It is also possible that maritime contacts
between Sardinia and the mainland were not necessarily routed across the
shortest open-water crossings (from Sardinia to Corsica to Elba to Piombino
and then up the Ligurian coast).

CONCLUSIONS

As Ammerman and Polglase (1993; 1996) have already noted, a new

generation of theoretical models are necessary to explain prehistoric obsidian
exchange mechanisms. These models should take into account more than
MEDITERRANEAN ISLANDS AND MULTIPLE FLOWS 79

just the procurement and movement of raw materials, but also the transfor-
mation of the raw material into a finished product, how and why it was used
and re-used, and finally its discardment. For areas outside obsidian source
zones, these multiple stages probably represent activities by at least two
parties, and perhaps four or five. Technological studies of the sort carried
out by Hurcombe and Phillips (1995) and Polglase (1990; Ammerman et al.
1990; Ammerman and Polglase 1993) enlighten our understanding of where
obsidian reduction occurred (at quarry, workshop, or domestic sites), while
microscopic use-wear studies (Hurcombe 1992a; 1992b) are narrowing down
the functions these tools served. Finally, the combination of visual and
low-cost chemical analysis allows comprehensive sourcing of entire obsidian
assemblages and the effective statistical comparison of spatially and tempo-
rally dynamic obsidian source exploitation patterns. The integration of all
these data categories will ultimately provide insight into the social significance
of obsidian.

REFERENCES
Ammerman, A., Cesana, A., Polglase, C. and Terrani, M. 1990 Neutron activation analysis of
obsidian from two neolithic sites in Italy. journal of Archaeological Science 17(2): 209-220.
Ammerman, A. and Polglase, C. 1993 The exchange of obsidian at Neolithic sites in Italy. In Healy,
F and Scarre, C., eds., Trade and Exchange in European Prehistory. Oxbow Monograph 33.
Oxford, Oxbow Books: 101-107.
Ammerman, A. and Polglase, C. 1996 Analyses and descriptions of the obsidian collections from
Arene Candide. Quaternaria Nova.
Aramu, F, Maxia, V, Serci, S. and Uras, I. 1983 Mossbauer study of Mount Arci (Sardinia) obsidian.
Lettere al Nuovo Cimento 36(5): 102-104.
Arias, C., Bigazzi, G., Bonadonna, FP., Cipolloni, M., Hadler,j.C., Lattes, C.M.G. and Radi, G. 1984
Fission track dating in archeology. A useful application. In Parrini, P.L., ed., Scientific
Methodologies Applied to Works of Arts. Montedison Progetto Cultura: 151-159.
Arias, C., Bernardes, C., Bigazzi, G., Bonadonna, FP., Cesar, M.F, Hadler Neto,].C., Lattes, C.M.G.,
Oliveira, ].X., Osorio Araya, A.M. and Radi, G. 1986 Identifica~o da proveniencia de
manufaturados de obsidiana atraves da datac;;o com o metodo do tra~o de fisso. Ciencia e
Cultura 38(2): 285-308.
Arias-Radi, G., Bigazzi, G. and Bonadonna, FP. 1972le tracce di fissione. Un metodo per lo studio
delle vie di commercio dell'ossidiana. Origini. Preistoria e Protostoria delle Civilta Antic he 6:
155-169.
Bailloud, G. 1969a Fouilles de Basi (Serra-di-Ferro, Corse). Campagne de 1968. Corse Historique
9(33): 49-64.
Bailloud, G. 1969b Fouille d'un habitat neolithique et torreen a Basi (Corse). Bulletin de la Societe
Prehistoire Fran(aise 66: 367-384.
Belluomini, G., Discendenti, A., Malpieri, land Nicoletti, M. 1970 Studi sulle ossidiane italiane.
II. - Contenuto in 40 Ar radiogenico e possibilita di datazione. Periodico di Mineralogia 39:
469-479.
Bigazzi, G., Bonadonna, FP., Maccioni, L. and Pecorini, G. 1976 Research on Monte Arci (Sardinia)
subaerial volcanic complex using the fission-track method. Bolletino della Societa Geologica
Italiana 95: 1555-1570.
80 ROBERT H. TYKOT

Bigazzi, G., Marton, P., Norelli, P. and Rozloznik, L 1990 Fission track dating of Carpathian obsidians
and provenance identification. Nuclear Tracks and Radiation Measurements 17: 391-396.
Bigazzi, G., Meloni, S., Oddone, M. and Radi, G. 1986 Provenance studies of obsidian artifacts:
trace elements and data reduction. journal ofRadioanalytical and Nuclear Chemistry, Articles
98(2): 353-363.
Bigazzi, G., Meloni, S., Oddone, M. and Radi, G. 1992a Nuovi dati sulla diffusione dell'ossidiana
negli insediamenti preistorici italiani.In Herring, E., Whitehouse, R. and Wilkins,]., eds.,
Papers of the Fourth Conference of Italian Archaeology 3. New Developments in Italian
Archaeology Part 1. London, Accordia Research Centre: 9-18.
Bigazzi, G., Meloni, S., Oddone, M. and Radi, G. 1992b Study on the diffusion of Italian obsidian
in the neolithic settlements. Atti del VIII Convegno Nazionale sulla Attivita di Ricerca nei
Settori della Radiochimica e della Chimica Nucleare, delle Radiazioni e dei Radioelementi,
Torino, 16-19 Giugno 1992. CNR and Universitil degli Studi di Torino: 243-247.
Bigazzi, G. and Radi, G. 1981 Datazione con le tracce di fissione per l'identificazione della
provenienza dei manufatti di ossidiana. Rivista di Scienze Preistoriche 36(1-2): 223-250.
Cann, j.R. and Renfrew, A.C. 1964 The characterization of obsidian and its application to the
Mediterranean region. Proceedings of the Prehistoric Society 30: 111-133.
Cortese, M.; Frazzetta, G. and La Volpe, L 1986 Volcanic history of Lipari (Aeolian Islands, Italy)
during the last 10,000 years. journal of Volcanology and Geothermal Research 27: 117-133.
Crisci, G.M., Ricq-de Bouard, M., Lanzaframe, U. and de Francesco, A.M. 1994 Nouvelle methode
d'analyse et provenance de !'ensemble des obsidiennes neolithiques du Midi de Ia France.
Gallia Prehistoire 36: 299-309.
Crummett, ].G. and Warren, S.E. 1985 Appendix I. Chemical analysis of Calabrian obsidian. In
Ammerman, A., ed., The Acconia Survey: Neolithic Settlement and the Obsidian Trade.
Institute of Archaeology Occasional Publication 10. London, Institute of Archaeology:
107-114.
de la Marmora, A. 1839-40 Voyage en Sardaigne. Turin,]. Bocca.
di Paola, G .M., Puxeddu, M. and Santacroce, R. 1975 K-Ar ages of Monte Arci volcanic complex
(central-western Sardinia). Rendiconti della Societa Italiana di Mineralogia e Petrologia 31:
181-190.
Dixon, J .E. 1976 Obsidian characterization studies in the Mediterranean and Near East. In Taylor,
R.E., ed., Advances in Obsidian Glass Studies. Park Ridge, NJ, Noyes Press: 288-333.
Dyson, S.L; Gallin, L.; Klimkiewicz, M.; Rowland, R.j., Jr. and Stevenson, C.M. 1990 Notes on
some obsidian hydration dates in Sardinia. Quaderni della Soprintendenza Archeologica per
le Province di Cagliari e Oristano 7: 25-42.
Francaviglia, V. 1984 Characterization of Mediterranean obsidian sources by classical petrochemi-
cal methods. Preistoria Alpina 20: 311-332.
Francaviglia, V. 1988 Ancient obsidian sources on Pantelleria (italy). journal of Archaeological
Science 15: 109-122.
Francaviglia, V. and Piperno, M. 1987 La repartition et Ia provenance de l'obsidienne archeologique
de Ia Grona dell'Uzzo et de Monte Cofano (Sicile). Revue d'Archeomttrie 11: 31-39.
Gale, N.H. 1981 Mediterranean obsidian source characterisation by strontium isotope analysis.
Archaeometry 23(1): 41-51.
Hallam, B.R. 1976 Neutron Activation Analysis of Western Mediterranean Obsidians. Unpublished
Master's Thesis, Bradford University.
Hallam, B.R., Warren, S.E. and Renfrew, A.C. 1976 Obsidian in the western Mediterranean:
Characterisation by neutron activation analysis and optical emission spectroscopy. Proceed-
ings of the Prehistoric Society 42: 85-110.
Herold, G. 1986 Mineralogische, chemische und physikalische Untersuchungen an den Obsidianen
Sardiniens und Palmarolas. Grundlagen zur Rekonstruktion Prahistorischer Handelswege im
Mittelmeerraum. Unpublished PhD dissertation, Universitat (TH) Fridericiana Karlsruhe.
MEDITERRANEAN ISLANDS AND MULTIPLE FLOWS 81

Hurcombe, L.H. 1992a New contributions to the study of the function of Sardinian obsidian
artifacts. In Tykot, R.H. and Andrews, T.K., eds., Sardinia in the Mediterranean: A Footprint
in the Sea. Sheffield, Sheffield Academic Press: 83-97.
Hurcombe, L.H. 1992b Use Wear Analysis and Obsidian: Theory, Experiments and Results. Sheffield,
Sheffield Archaeological Monographs 4.
Hurcombe, L.H. and Phillips, P Chronology and artifact function at the Grotta Filiestru. Paper presented
at the international conference "Sardinian Stratigraphy and Mediterranean Chronology," Tufts
University, Medford, Massachusetts, March 17-19, 1995. Proceedings forthcoming.
Keller,]. and Seifried, C. 1990 The present status of obsidian source identification in Anatolia and
the Near East. In Albore Livadie, C. and Widemann, F, eds., Volcanologie et Archeologie.
Proceedings of the European Workshop of Ravello, November 19-27, 1987 and March 30-31,
1989. PACT 25(4): 57-87.
Loddu, R. 1903 Stazione neolitica del Monte Urpino ( Cagliari). Bullettino di Paletnologia Ita Iiana
29: 45-52.
Longworth, G. and Warren, S.E. 1979 Mossbauer spectroscopy and the characterisation of
obsidian. journal of Archaeological Science 6(2): 179-193.
Mackey, M. and Warren, S.E. 1983 The identification of obsidian sources in the Monte Arci region
of Sardinia. 1n Aspinall, A. and Warren S.E., eds., Proceedings of the 22nd Symposium on
Archaeometry, University of Bradford, Bradford, U.K. March 30th-April 3rd 1982. Bradford:
420-431.
Mahood, G.A. and Hildreth, W 1986 Geology of the peralkaline volcano at Pantelleria, Strait of
Sicily. Bulletin of Volcanology 48: 143-172.
McDougall,]., Tarling, D. and Warren, S.E. 1983 Magnetic sourcing of obsidian samples from the
Mediterranean. journal of Archaeological Science 10(5): 441-452.
Mello, E. 1983 1ndagini scientifiche per l'individuazione della provenienza dei manufatti di ossidiana.
In Tine, S., Passo di Corvo e Ia civiltil neolitica del Tavoliere. Genova, Sagep Editrice: 122-124.
Merrick, H.V and Brown, FH. 1984 Rapid chemical characterization of obsidian artifacts by
electron microprobe analysis. Archaeometry 26(2): 230-236.
Michels, J.W 1987 Obsidian hydration dating and a proposed chronnological scheme for the
Marghine region. In Michels, j.W. and Webster, G.S., eds., Studies in Nuragic Archaeology:
Village Excavations at Nuraghe Urpes and Nuraghe Toscano in West-Central Sardinia. BAR
International Series 373. Oxford, BAR: 119-125.
Michels,]., Atzeni, E., Tsong, 1.S.T. and Smith, G.A. 1984 Obsidian hydration dating in Sardinia.
In Balmuth, M.S. and Rowland, R.j., Jr., eds., Studies in Sardinian Archaeology. Ann Arbor,
University of Michigan: 83-114.
Phillips, P. 1992 Obsidian distribution and the European neolithic. In Tykot, R.H. and Andrews,
T.K., eds., Sardinia in the Mediterranean: A Footprint in the Sea. Sheffield, Sheffield Academic
Press: 71-82.
Pichler, H. 1980 The island of Lipari. Rendiconti della Societa Ita Iiana della Mineralogia e Petrologia
36(1): 415-440.
Polglase, C. 1990 Appendice l. In Bernabo Brea, M.; Ghiretti, A.; Polglase, C.; Visconti, V, 1 siti
neolitici lungo il torrente Cinghio (Parma). Preistoria Alpina 24: 140-155.
Pollmann, H.-0. 1993 Obsidian im nordwestmediterranen Raum. Seine Verbreitung und Nutzung im
Neolithikum und Aneolithikum. BAR International Series 585. Oxford, Tempus Reparatum.
Puxeddu, C. 1958 Giacimenti di ossidiana del Monte Arci in Sardegna e sua irradiazione. Studi
Sardi 14-15(1955-1957) Parte 1: 10-66.
Randle, K., Barfield, L.H. and Bagolini, B. 1993 Recent Italian obsidian analyses. journal of
Archaeological Science 20: 503-509.
Renfrew, C. and Dixon, ].E. 1976 Obsidian in western Asia: a review. In Sieveking, G. de G.,
Longworth, l.H. and Wilson, K.E., eds., Problems in Economic and Social Archaeology.
London, Duckworth: 137-150.
82 ROBERT H. TYKOT

Taylor, R.E., ed. 1976 Advances in Obsidian Glass Studies. Park Ridge, NJ, Noyes Press.
Torrence, R. 1986 Production and Exchange of Stone Tools: Prehistoric Obsidian in the Aegean.
Cambridge, Cambridge University Press.
Trump, D.H. 1983 La Grotta di Filiestru a Bonu Ighinu, Mara (55). Quaderni 13. Sassari, Dessi.
Tykot, R.H. 1992 The sources and distribution of Sardinian obsidian. In Tykot, R.H. and Andrews,
T.K., eds., Sardinia in the Mediterranean: A Footprint in the Sea. Sheffield, Sheffield Academic
Press: 57-70.
Tykot, R.H. 1995 Prehistoric Trade in the Western Mediterranean: The Sources and Distribution of
Sardinian Obsidian. PhD dissertation, Harvard University. Ann Arbor, University Micro-
films.
Tykot, R.H. 1996 Obsidian procurement and distribution in the central and western Mediterranean.
journal of Mediterranean Archaeology 9: 39-82.
Tykot, R.H. and Young, S.M.M. 1996 Archaeological applications of inductively coupled plasma-
mass spectrometry. In Orna, M.V., ed., Archaeological Chemistry V. Washington, DC,
American Chemical Society: 116-130.
Williams-Thorpe, 0. 1995 Review article. Obsidian in the Mediterranean and the Near East: a
provenancing success story. Archaeometry 37(2): 217-248.
Williams-Thorpe, 0., Warren, S.E. and Barfield, L.H. 1979 The distribution and sources of
archaeological obsidian from Northern Italy. Preistoria Alpina 15: 73-92.
Williams-Thorpe, 0., Warren, S.E. and Courtin,]. 1984 Sources of archaeological obsidian from
Southern France. journal of Archaeological Science 11(2): 135-146.
Williams-Thorpe, 0., Warren, S.E. and Nandris,]. 1984 The distribution and provenance of
archaeological obsidian in Central and Eastern Europe.]ournal of Archaeological Science 11:
183-212.
 Chapter4

Intrasource Chemical
Variability and Secondary
Depositional Processes
Lessons from the American Southwest

M. STEVEN SHACKLEY

ABSTRACT

For nearly 30 years, archaeological obsidian has been collected from geological
and archaeological contexts, analyzed by a spate of methods, and offered to the
scientific community as internally valid research. Virtually none of this work
has been gathered in explicitly scientific ways such that the results are reliable
and valid. In an attempt to provide a starting point for consideration, this
chapter discusses the field and laboratory methods used in the ongoing South-
west and Northwestern Mexico Archaeological Obsidian Project dealing with arid
environments, very old glass sources, relatively extreme secondary deposition,
and a great range of chemical variability common in this geological context.

INTRODUCTION

Two important factors affecting the understanding of the distribution of

archaeological obsidian include the potential for intrasource chemical variabil-

M. STEVEN SHACKLEY • Phoebe Hearst Museum of Anthropology and Department of Anthro-

pology, University of California, Berkeley, California.

Archaeological Obsidian Studies, edited by Shackley.

Plenum Press, New York, 1998.

83
84 M. STEVEN SHACKLEY

ity and the cultural effect of secondary depositional processes. Until quite
recently, these factors have essentially remained ignored, partly due to the
nascent nature of obsidian studies in both geology and archaeology and the
consequent effect-or lack thereof-on archaeological research orientation
(Glascock 1994; Hughes 1994; Hughes and Smith 1993; Shackley 1992, 1994).
Recent advances in instrumental analyses, many discussed in this volume, have
allowed us to deal with most of these processes in the geological record to great
advantage for archaeologists dealing with problem domains in production and
distribution of obsidian in the archaeological record. What does seem to be
lacking, however, is a critical examination of the field methods of potential
utility in the understanding of secondary depositional processes, as well as the
field strategies needed to adequately sample a source such that the chemical
variability of a single source can be derived (see Hughes 1994; Shackley 1992).
Fortuitously, two important archaeological sources of obsidian in the
western New Mexico and eastern Arizona region in the American Southwest
provide an excellent field laboratory to test a number of methods dealing with
these issues; the Cow Canyon source, the Mule Creek Obsidian Source Group,
and their erosion into the greater Gila River Stream Basin (Fig. 4.1). Intensive
field and laboratory research at Mule Creek for the past ten years iudicates a
source that is quite variable in its chemical composition and, along with Cow
Canyon, has eroded since at least the Pliocene into the Gila River system great
distances from the parent dome structures and immediate alluvial environment
(Shackley 1992). So, in essence, we can discuss both chemical variability and
secondary depositional issues from one locality. Furthermore, this particular
area happens to be located at the boundary of three important prehistoric
cultural groups in the late period that could have had access to primary or
secondary source material; Mogollon, Hohokam, and Salado (Fig. 4.1).

REGIONAL GEOLOGIC HISTORY

The Mogollon-Datil Province and the Mule Creek Area

The Mule Creek Source Region is one of the most geologically explored
archaeological sources of obsidian in the American Southwest (Brooks and
Ratte 1985; Ratte 1982; Ratte and Brooks 1983, 1989; Ratte and Hedlund 1981;
Rhodes and Smith 1972). Ratte has organized most of the research in the area
focusing on mapping and establishing the origin of the volcanics during the
Tertiary as originally described by Rhodes and Smith ( 1972). This region, which
is on the boundary between the Basin and Range complex to the west and
southwest, and the southeastern edge of the Colorado Plateau, exhibits a silicic
geology that is somewhat distinctive; from the decidedly peraluminous glass
of Cow Canyon with relatively high strontium values and the distinct chemical
CHEMICAL VARIABILITY AND DEPOSITIONAL PROCESSES 85

Figure 4.1. Location of selected archaeological sources of obsidian, cultural groups, and signifi-
cant drainage systems in the greater American Southwest. Sources = 1 Los Vidrios; 2 Antelope
Wells; 3 Cow Canyon; 4 Mule Creek; 5 Gwynn Canyon; 6 Mount Taylor; 7 Valles Caldera Region.
Not to exact scale.

variability of the Mule Creek glasses (Elston 1984; Ratte et al. 1984; Rhodes
and Smith 1972; Shackley 1988, 1990). The province has been named Mogol-
lon-Datil for its location and major floristic association (Elston 1965, 1976).
The region is, in part, characterized by pre-caldera andesites and later high-sil-
ica alkali rhyolites in association with caldera formation , subsequent collapse
and post-caldera volcanism. Most recently, fieldwork and chemical analyses by
Ratte and Brooks (1989) lead them to conclude that the Mule Creek Caldera
is actually just a graben, although the typical succession from intermediate to
silicic volcanism apparently holds.
The obsidian has been directly dated at the Antelope Creek locality
(locality lin Fig. 4.2) to 17.7±0.6 mya by K-Ar. A single obsidian marekenite
taken from the perlitic lava at the Antelope Creek locality discussed below was
used in the analysis. Unusual in geological descriptions, the obsidian proper
was discussed as an integral part of the regional geology.
86 M. STEVEN SHACKLEY

I
1109"
I

~· I • aEDHIU.
'
,..._
.
I
I
I "-
'
I

\
. . GWYNN CANYON
I (1!WE CANYON)

I
r

tO • » .. 10

r-1 I I I j a::::-..J
'-:j:::!::;:::¢~:::: o-. -
....... 10

Figure 4.2. Study area with selected archaeological sites (filled circles), obsidian sources (filled
triangles), modem communities (filled squares) , and modern drainage system. Mule Creek
Chemical Group in-situ localities= l Antelope Creek; 2Mule Mountains; 3 Mule Creek!N. Sawmill
Creek.

Rhyolite of Mule Creek (Miocene)

Aphyric, high-silica, alkali-rhyolite domal flows from the Harden Cienega
eruptive center along southwestern border of quadrangle [Wilson Mountain
1:24,000 Quad, New Mexico; Locality l herein) . Unit ob, commonly at the base
of the flows, consists of brown, pumiceous glass that grades upward into gray
to black perlitic obsidian and obsidian breccia. Extensive ledges of partly
hydrated, perlitic obsidian contain nonhydrated obsidian nodules (mare-
kenites) which, when released by weathering, become the Apache tears that
are widespread on the surface and within the Gila Conglomerate in this region.
Age shown in Correlation is from locality about l km south of tank in Antelope
Creek in Big Lue Mountains quadrangle adjacent to west edge of Wilson
CHEMICAL VARIABILITY AND DEPOSITIONAL PROCESSES 87

Mountain quadrangle. Thickness of flows is as much as 60 m and unit ob as

much as 25m (Ratte and Brooks 1989: map text).
This description adequately characterizes what is found at the other two
primary localities (Mule Mountains, and Mule Creek/North Sawmill Creek;
Figs. 4.3 and 4.4). Aphyric, artifact quality marekenites are remnant within
perlitic glass lava units. Nodules at all localities are up to 15 em in diameter
although most are under 10 em. The devitrified perlitic lava, quite friable,
erodes easily into the local alluvium. As discussed elsewhere, this is relatively
unique in Tertiary sources in the Southwest where most of the obsidian breccia
and perlitic lava is often completely eroded away leaving only the rhyolite
interior of the dome and a consequent inability to assign the surrounding
marekenites to a specific dome structure (Shackley 1988, 1994, 1995; see also
Bouska 1993; Hughes and Smith 1993). This glass, as well as the Cow Canyon
obsidian, has been found in late period Hohokam and Salado sites 250 to 300
km west in the Phoenix and Tonto Basins, and in significant quantities at Casa
Grande (Mitchell and Shackley 1995; Fig. 4.1 this chapter) .

Figure 4.3. Obsidian zone with remnant marekenites at Antelope Creek locality. Height of section
approximately four meters on the far right of section. Weathering has differentially exposed strata
containing only perlite or perlite and marekenites. Rock in alluvium in foreground is composed
of >95% marekenites that are eroding into Antelope Creek, into Harden Cienega Creek and then
into the San Francisco River to the north.
88 M. STEVEN SHACKLEY

Figure 4.4. Remnant rnarekenites in perlite at Antelope Creek locality. Pointed end of rock
hammer locates 5 ern diameter nodule in situ.

Cow Canyon Tertiary Source

The precise geological origin of this important archaeological source of
obsidian is essentially unknown (Shackley 1988, 1995). According to the rather
non-specific regional geological map (Wilson and Moore 1958) the remnant
dome structures are Tertiary and probably Miocene in age. Located on a very
large ridge composed of coalesced rhyolite domes that erode west into Eagle
Creek and finally into the Gila River, and east into the Blue River, San Francisco
River, and finally the Gila River system, any obsidian breccia or obsidian zones
have apparently completely eroded away, unlike the Mule Creek localities (Fig.
4.2). The nodules, up to 5 em in diameter, are present in the rhyolite regolith
at the top of the ridges at the head of Cow Canyon. Specific data on the source
are available in Shackley (1988, 1990, 1995). Again, the erosion of the glassy
obsidian zone on these old domes is not uncommon in the Southwest. It is
possible that the Cow Canyon glass is of pre-caldera origin and actually older
than the Mule Creek rhyolite to the south, although the lack of in-situ obsidian
is more likely an effect of the specific geology and/or weathering processes.
While there is some chemical variability in the Cow Canyon glass, it appears
CHEMICAL VARIABILIIT AND DEPOSITIONAL PROCESSES 89

to be less extreme than the Mule Creek case and will not be discussed further
here (Shackley 1995).

INTRASOURCE CHEMICAL VARIABILITY: THE MULE CREEK CASE

Perhaps nowhere are the problems of careful and complete field sampling
more important than when addressing the issue of intrasource variability
(Duerden et al. 1987; Ericson and Glascock 1992; Hughes 1988, 1994; Hughes
and Smith 1993; Shackley 1994, 1995; and discussed by many of the authors
in this volume). Perhaps nowhere is this more important than in Oceania where
the accurate assignment of a single piece of obsidian debitage may indicate
contact over thousands of kilometers of ocean (Green 1987; Summerhayes et
al., this volume). This problem was not much of an issue in the early years of
obsidian studies when only a few samples from each source were analyzed and
that deemed sufficient Uack 1971, 1976). This was arguably defended on the
grounds that silicic glasses are by definition homogeneous rocks quenched over
very short periods of time precluding the alteration that typically occurs in
crystalline volcanic rocks (Macdonald and Bailey 1973). Partly due to the
changes in silicic volcanic petrological theory and partly due to overwhelming
empirical evidence that many sources are quite variable chemically, recent
studies have admitted that more intensive and extensive sampling is necessary
to deal with the problem (Hughes 1994; Shackley 1994, 1995). And frankly, I
think an additional reason that these sampling issues have arisen is the more
intensive input of the archaeologists in archaeometric problem solving (Shack-
ley 1993a).

Source of Chemical Variability

Hughes and Smith (1993) recently discussed in some detail the structural
variables that create homogeneity and heterogeneity in silicic glass chemistry.
They cite magma mixing as one probable reason for this variability (Hughes
and Smith 1993:83-84). I recently found that wall-rock reaction with basal
limestone in southern Nevada caused extreme variability such that, without
knowledge of the source chemistry of the Tertiary source, an analysis of
obsidian from local archaeological contexts would invariably indicate two
distinct sources (Shackley 1994). Additionally, magma fractionation over time
can have a significant effect on glass chemistry from a single magma chamber
(Macdonald eta!. 198 7; Mahood and Hildreth 1983; Mahood and Stimac 1990).
If the chamber remains active for long enough, fractionation and large partition
coefficients can and generally will occur essentially resulting in linear changes
in trace element chemistry (Mahood and Hildreth 1983). This latter process
may be the case at Mule Creek, although all the other processes discussed above
90 M. STEVEN SHACKLEY

may be operative. A zircon fission-track date from the Mule Mountain rhyolite
yielded a age of 17.7±1.0 mya statistically identical to the Antelope Creek
nodule mentioned above (Ratte and Brooks 1983). Chronology, therefore is
probably not the culprit here, and some effect of magma mixing and/or
fractionation is responsible for the chemical variability.
Table 4.1 exhibits the mean and central tendency data for the Mule Creek
Source Group. Most obvious immediately is the virtually double rubidium
elemental levels for the Mule Creek/North Sawmill glass versus the other two
major groups. This was the chemistry originally characterized for Mule Creek
(Shackley 1988). Barium essentially separates the other two from each other
(see Figs. 4.5 and 4.6). Rubidium and zirconium generally separate these data
well enough, however, including the low zirconium river alluvium samples
(Fig. 4.7). An analysis by Ratte and Hedlund (1981) of rhyolite and obsidian
from the Antelope Creek locality is essentially identical to my analyses (Table
4.2). Instrumental methods used in the EDXRF analyses here are essentially
the same as those used by Davis et al. in this volume. For the precise methods
refer to Shackley (1994, 1995).
So, in the Mule Creek Source Region at least four distinct chemical groups
are evident, distinguished by Rb, Sr, Y, Zr, Nb, and Ba, concentration values,
especially Rb, Sr, Zr, and Ba, and are named after the localities where mare-

-
300

• • •• •
'• '
- ....
200

-
Ba
~
100

·'-"' -~ •• ·-ti

·•·......
0

I t$
••
...
400

300 Rb

•i!" ...
~~
I'~
•• -:
~-'
200

.;...
100

-
•• • •
.l
20

-~ Sr •
•

.......
-..~ • SOURCE GROUP

.
10
••
,~·
• Blue/SF River
0

4! .H
Alluvium

•~~
120
• Mule Mts

• •• •
100
80 •I Zr
• Mule Creek/

80 •• •• •• N. Sawmill Cr.

40 ~ Antelope Cr

Figure 4.5. Ba, Rb, Sr, Zr bivariate matrix plot of Mule Creek Chemical Groups.
CHEMICAL VARIABILITY AND DEPOSITIONAL PROCESSES 91

Table 4.1. Mean and Central Tendency Data

for the Mule Creek Chemical Groups,
Arizona and New Mexico"

Element Mean SD Minimum Maximum N

Antelope Creek Group (I)

Ti 679.27 109.11 470.91 849.27 14

Mn 352.17 39.97 285.92 425.05 14
Fe 9686.14 401.87 8880.69 10386.48 14
Rb 239.63 12.97 220.89 271.99 14
Sr 17.22 93.0 15.95 19.21 14
y 40.68 3.03 33.72 45.35 14
Zr 123.71 7.55 110.37 131.34 14
Nb 26.51 2.66 22.50 31.49 14
Ba 115.57 20.91 85.09 146.61 14

Mule Mountains Group (2)

Ti 720.45 105.29 614.11 992.83 12

Mn 439.02 43.33 357.05 520.26 12
Fe 8058.51 578.32 7446.45 9252.74 12
Rb 193.10 31.09 170.63 262.91 12
Sr 10.87 1.28 7.85 13.15 12
y 28.39 9.57 22.30 50.76 12
Zr 118.35 4.16 107.59 125.01 12
Nb 32.14 1.93 28.29 34.38 12
Ba 86.14 11.16 64.13 99.88 12
Mule Creek/North Sawmill Creek Group (3)

Ti 613.62 189.49 305.48 1276.39 34

Mn 575.36 41.04 503.21 671.58 34
Fe 8423.30 564.44 7602.84 11022.36 34
Rb 411.31 14.61 383.73 444.04 34
Sr 5.52 1.70 1.27 10.47 34
y 70.93 2.36 66.64 76.70 34
Zr 118.15 7.22 102.09 131.22 34
Nb 119.85 4.00 110.96 130.15 34
Ba 45.34 1.72 43.01 49.52 26

Blue/San Franciso River Alluvium at Clifton, AZ (4)

Ti 663.63 187.81 448.91 1023.15 7

Mn 490.15 144.74 303.37 726.46 7
Fe 7262.97 504.97 6444.17 8051.88 7
Rb 180.36 21.27 147.22 215.23 7
Sr 14.21 6.40 6.65 22.60 7
y 26.58 5.60 16.12 32.85 7
Zr 78.73 10.87 60.38 90.03 7
Nb 23.53 4.43 17.29 28.86 7
Ba 119.84 66.12 66.01 212.51 7
"Total Fe can be converted to total Fe 20 3 with a multiplier of
0.14297x!0-4 (see also Glascock !991). SD=one standard deviation.
92 M.STEVENSHACKLE¥

Mule Cr Group
160
120140 • Mule Mountains
80 100
60 • Mule Cr/
Zippm 20 40 Bappm
N. SawmiiiCr

" Antelope Creek

Figure 4.6. Rb, Zr, Ba three-dimensional plot of the three in situ Mule Creek Chemical Group
localities.

500.00
450.00
MIJe Cr/N Sawmill,.
400.00
350.00

300.00
.iii

,,
E Artelope
8::250.00
r~
'
..0 Cr

'"'
0:: 200.00 !
150.00 River AllN. ! MIJe Mts
100.00

50.00
0.00
8 0
0
0
0 8 8 8
0 0
N g g g g
Zrppm

Figure 4.7. Rb, Zr bivariate plot of all four Mule Creek Chemical Groups. Error bars represent
first standard error of population measurements along both axes.
CHEMICAL VARIABILITY AND DEPOSITIONAL PROCESSES 93

Table 4.2. Trace Element Analysis of Rhyolite

of Mule Creek by Ratte and Hedlund
(1981: Table 1 and 2)"

Sample No. Ti Mn Fe Rb Sr Zr

8 599 <!55 7333 227 5 97

9* 234 5 92
10 233 4 94
II 208 7 102
12 208 19 109

*Ti, Mn, and Fe not reported for samples 9--12.

"Ti, Fe, and Mn data calculated from the oxides. All measurements in
parts per million (ppm) Total Fe can be converted to total Fep 3 with
a multiplier of0.14297x 10-4 (see also Glascock 1991 ).

kenites have been found in perlitic lava: Antelope Creek; Mule Mountains; and
Mule Creek/North Sawmill Creek all in New Mexico (see Table 4.1; Figs. 4.5
and 4.6). A rather straightforward stepwise discriminant analysis (SPSS for
Windows, Release 6.0) using the Mahalanobis distance measure with Ba, Rb,
Sr, and Zr as elemental variables yields a good separation of the defined groups
(Table 4.3; Fig. 4.8).
During the 1994 field season, a fourth sub-group was discovered in the
San Francisco River alluvium near Clifton, Arizona and in older alluvium
between Arizona Highway 191 and Eagle Creek in eastern Arizona north of
Clifton (Fig. 4.2). While in situ nodules have not yet been found they are
certainly located somewhere west of Blue River and north and west of the San
Francisco River since none of this low zirconium sub-group was discovered in
alluvium upstream from the juncture of the Blue and San Francisco Rivers.
While the primary context of the source has not yet been discovered, it has
been recovered in archaeological contexts in the San Simon River Valley
southwest of this area (Shackley l993b; Fig. 4.2 here). The genetic relationship
is apparent in the trace element matrix plot (Fig. 4.5), and signifies the very
complex nature of the Mule Creek silicic geology, with subsequent depositional
mixing in the Gila Conglomerate.
Glass at other Tertiary sources in the Southwest, such as Sauceda Moun-
tains and Antelope Wells, also appear to exhibit more than one chemical mode,
although not as distinct as Mule Creek (LeTourneau 1994; Shackley 1988,
1990). The Mule Creek case is similar to Sauceda Mountains because the
chemical groups are not always spatially discrete and occur together in the
alluvium, in this case the extensive Gila Conglomerate which is mainly com-
posed of Mule Creek rhyolite and tuffs in the area where the marekenites do
occur (see Ratte and Brooks 1989).
94 M. STEVEN SHACKLEY

Table 4.3. Posterior Classification Matrix for Elemental Variables

Rb, Sr, Zr, and Ba from the Mule Creek Region*

Predicted Group Memhership"·b,c Total

SOURCE GROUP I 2 3 4

Original

Countb Antelope Creek (I) 21 0 0 0 21

Mule Mountains (2) 0 34 0 0 34

Mule Cr/N Sawmill (3) 0 0 15 0 15

Blue/SF Riv. Alluv. (4) 0 0 0 7 7

Percent Antelope Creek (I) 100 0 0 0 100.0

Mule Mountains (2) 0 100 0 0 100.0

Mule Cr/N Sawmill (3) 0 .01 100 0 100.0

Blue/SF Riv. Alluv. (4) 0 0 0 100 100.0

Cross validated'

Count Antelope Creek (I) 21 0 0 0 21

Mule Mountains (2) 0 34 0 0 34

Mule Cr/N Sawmill (3) 0 0 15 0 IS

Blue/SF Riv. Alluv. (4) 0 0 6 7

Cross validated0

percentage Antelope Creek (I) 100 0 0 0 !00.0

Mule Mountains (2) 0 100 0 0 100.0

Mule Cr/N Sawmill (3) 0 0 100 0 100.0

Blue/SF Riv. Alluv. (4) 0 0 14.3 85.7 100.0

*Cross-validated (jack knifed) analysis indicates that two source standards for the San Francisco
River alluvium (4) group are similar to the Mule Mountains (3) group (see Fig. 4.8).
•cross validation is done only for those cases in the analysis. In cross validation, each case is
classified by the functions derived from all cases other than that case.
hi 00% of original grouped cases correctly classified.
c98. 7% of cross validated grouped cases correctly classified.
CHEMICAL VARIABILITY AND DEPOSITIONAL PROCESSES 95

In the earlier study, I noted two "outliers" collected at Mule Creek with
significantly higher rubidium concentration values (Shackley 1988: 767).
These outliers now constitute the Antelope Creek chemical group. A single
collection day was simply not sufficient to deal with the potential variability
at this source.

SECONDARY DEPOSITIONAL EFFECTS AND

ARCHAEOLOGICAL PROBLEM DOMAINS

The role of secondary depositional processes in the procurement of

prehistoric raw materials, particularly obsidian, has not been explored to any
extent in North America (see Hughes and Smith 1993;Jackson 1989; Lavin and
Prothero 1992; Shackley 1988, 1992; Shelley 1993; Skinner 1983). Most of this
work has been focused on glacial transport during the late Pleistocene, and few
have attempted to integrate secondary transport processes into research do-
mains (Church 1995; Meltzer 1985; Wyckoff 1993). I find this particularly

0
SOURCE GROUP
N
c
0 0 Group Centroids
:0:0
u -2
c .,. Blue/SF River
:::l
u..
Alluvium (4)
-4
• Mule Mts (3)

-6 e Mule Cr/
N. Sawmill Cr (2)

-8 • Antelope Cr (1)
- 0 0 10 20

Function 1
Figure 4.8. Plot of the first two canonical discriminant scores for the four groups based on Ba,
Rb, S r,and Zr elemental variables. Group numbers correspond to those in Table 4.3.
96 M. STEVEN SHACKLEY

vexing since one generally assumes that raw materials are most often procured
from the nearest source as a null hypothesis and proceeds from there (Church
1995; Newman 1994; Renfrew 1977; Shackley 1987).
Continuing work in the study area has shown that all of the Mule Creek
Chemical Groups and the Cow Canyon glass are mixed within the Gila River
alluvium and may be present in proportions relative to the greater stream
dynamics in the region (Shackley 1992: Figure 1; Fig. 4.2 here). Recent
historical geological research in the region indicates that during the late
Pliocene and throughout the Pleistocene the stream flow from all the streams
eroding these Miocene obsidian sources was considerably greater than during
the present (Houser et al. 1985; Richter et al. 1983). At some times the entire
Safford and San Simon Valleys would be inundated with water (Fig. 4.2).
Indeed, during the Pliocene the Safford-San Simon Valleys were dominated by
a shallow lake and an extensive low-gradient alluvial fan from south and
southwest flowing streams including the early Gila River (Houser et al. 1985;
see also Burkham 1972; Huckleberry 1995). These sediments are now charac-
terized by white caliche-cemented terraces often called the 111 Ranch Forma-
tion (Houser et al. 1985; see Fig. 4.2 here). During the Plio-Pleistocene
boundary period, the Gila River effluent often flooded these valleys, and Eagle
Creek, eroding the Cow Canyon source, was much larger as it joined the Gila
River system. The effect of this rapid sedimentation is that at times a single
source may have contributed more or less to the streamflow at any given time.
What we see today is that a given section of these early sediments may be
completely dominated by a single source of nodules. This was recently the case
near the present 111 Ranch where collections of marekenites by Gay Kinkade,
analyzed at Berkeley, indicated that all 20 samples were exclusively from the
Cow Canyon source with no Mule Creek material in that particular portion of
the sediment. Interestingly, one of the nodules measured 54 mm in diameter,
as large as any at the primary source at Cow Canyon over 80 linear kilometers
north, suggesting very rapid transport (Shackley 1988, 1990; Fig. 4.2).
This creates one major implication for archaeological interpretation. If
an archaeological sample from a single site is dominated by any one of the Mule
Creek chemical groups or Cow Canyon glass, it may be that the knapper
procured nodules on one of these Plio-Pleistocene terraces rather than engaging
in an exchange relationship with a group at the source in the highlands. It may
be tempting to assume that, given the particular prehistoric cultural and
geological processes in this region, this is a unique case. Many of the Tertiary
sources in the Southwest, however, are transported through stream erosion
great distances (Shackley 1988, 1990). In the Great Basin where interior
drainage basins are the rule, this is apparently not the case (Shackley 1994).
The southern Southwest, while dominated by basin and range topography is
less dominated by interior drainage systems (Nations and Stump 1981;
Reynolds et al. 1986).
CHEMICAL VARIABILITY AND DEPOSITIONAL PROCESSES 97

The point here is that simply determining the primary origin of a given
obsidian source is only a small part of the research. Particularly in arid regions
with Tertiary sources of obsidian that are not subject to interior drainage
systems, a strategy must be generated that takes into account the ultimate
secondary distribution of the source material as well as the potential for
chemical variability.

A GENERAL SAMPLING STRATEGY FOR RECORDING

TERTIARY SOURCES IN ARID ENVIRONMENTS

As is evident from the discussion above, a strategy to deal with both

source chemical heterogeneity and secondary dispersion through erosion is
relatively simple. As noted by Hughes and Smith, because of the fundamental
axiom underlying all early obsidian provenance studies that obsidian sources
are by definition chemically homogeneous, many sources have been charac-
terized by only five samples or less (1993:82). My first sampling and analysis
at Mule Creek with 16 samples identified only two of the four chemical groups
(Shackley 1988). This was certainly not enough (see Skinner 1983).
Gone are the days when a collector, archaeologist, or geologist would
send an archaeometrist a few marekenites picked up on Highway 78 near the
Arizona-New Mexico line and the archaeometrist identified the new source. A
very basic minimum standard for field sampling and instrumental analysis
should include:
l. A thorough geological background search of the area just as an
archeologist would perform a record and literature search before
working in a new area. In the U.S., USGS, state, and university
geologists can usually be contacted and are an important source of
information. Similar agencies in other countries have corresponding
contacts. This scientific mutualism between archaeologists and geolo-
gists discussed by Hughes and Smith cannot be overemphasized, and
may simply save a lot of effort and improve our efforts to explain the
archaeological record (1993:90).
2. Map and describe samples taken along transects from the entire area
where artifact quality obsidian occurs. This may mean, as is the case
with Mule Creek, that many days or even weeks may be required
before the entire area is covered to the boundary point where obsidian
no longer occurs. Recording the location of the samples collected may
allow one to determine the locality of a primary source when it was
not located in the field by concentration of distinct chemical signa-
tures identified by instrumental analysis. (See also Glascock et al. this
volume for further comments on sampling).
98 M. STEVEN SHACKLEY

3. If the environment is appropriate, determine the limit of the secon-

dary distribution of the raw material. This cannot be emphasized
enough.
4. The proper number of samples to analyze may only be discernible
experimentally (lAOS 1994). Five is certainly not enough. Ten might
be, but only further sampling will tell. On Quaternary sources with
little secondary dispersion, I prefer to collect a minimum of 200
samples from transects across the dome(s), judgmentally select 10 to
20 based on size, color, fabric etc., and select an additional 10 to 20
probabilistically (simple random sample). For earlier Tertiary sources
I use the same sampling strategy from various stratum such as primary
lava, ash flow contexts, and alluvial samples along the entire drainage
where artifact quality obsidian occurs.
5. Analytical sampling is somewhat more straightforward (see Glascock
et al., this volume). I am convinced that any of the analytical methods
discussed in this volume which measure incompatible elements with
a fair degree of certainty and are calibrated to international standards
are adequate for source assignment (see Hughes 1988; lAOS 1994).
6. Report the data in a manner that is easy to interpret. An important
issue, rarely discussed, are the methods used to assign artifacts to
sources. A method favored by a number of researchers, particularly
those using NAA data yielding a relatively large number of elemental
and compound variables, are multivariate statistical analyses based
on Mahalanobis distance measures to indicate group (source) prove-
nance (Glascock 1992, 1994; Harbottle 1976). While this is conven-
ient, it is important to remember that if the number of cases is not
two to three times the number of variables (elements measured)
[p>3n] then the group assignments based on the sum of squares in
the cross-products matrix (SSCP) can often yield spurious results due
to zero cells in the matrix, and misclassifications can occur without
knowledge (Baxter 1994a, 1994b;johnson and Wichern 1982). More
importantly, of course, is that group assignments based on multivari-
ate statistical measurement do not necessarily represent groupings
based on what is occurring in the field. The best procedure in my
opinion is to combine multivariate analyses (i.e., cluster or discrim-
nant analysis), if one must use them, with graphic displays. If the
multivariate group assignments do not agree with that observed in
the graphic displays (bivariate or trivariate), then it would be advis-
able to carefully assign the artifacts to sources.
7. Finally, all source standard analyses should be reported regularly and
shared freely. Source standard data should be submitted to a central
clearinghouse such as Ronald L. Bishop at the Smithsonian Institu-
tions Conservation Analytical Laboratory (SARCAR), and/or Michael
CHEMICAL VARIABILITY AND DEPOSITIONAL PROCESSES 99

Glascock at the University of Missouris Research Reactor Facility

(MURR), from which researchers from all over the world will ulti-
mately be able to access the data through Internet (see Glascock et
al., this volume).

CONCLUSION

The chemical analysis of archaeological obsidian in the service of explain-

ing past human behavior has advanced tremendously in the last few decades.
Unfortunately, the field relationships that can influence the chemical variability
that hampers our ability to confidently assign artifacts to sources has lagged
behind these instrumental advances. No longer can the archaeometrist analyze
a few samples from a source sent by a collector and characterize a source. We
know now that many sources vary so considerably in their chemical composi-
tion, that a systematic collection strategy must be employed to deal with this
variability Furthermore, secondary depositional processes can have a great
effect on the validity of inferences of exchange, interaction and social processes
in some regions. Taken together, these two effects can seriously skew the
analytical results reported to an archaeologist, making for a very unstable
foundation for archaeological inference.
Our understanding of the chemistry, production, and distribution of
archaeological obsidian has increased tremendously. Our field collection and
recording strategies should follow these advances.

ACKNOWLEDGMENTS

Research for the Southwest Archaeological Obsidian Project was funded by

National Science Foundation Grant (DBS-9205506), and the Stahl Endowment
for Archaeological Research, University of California, Berkeley. The fieldwork
was aided by Eroica D. Shackley, Stephanie Kuluva, Gay Kinkade and Rudy
Benskin (BLM, Safford, AZ), Bruce Donaldson (Apache-Sitgreaves National
Forest, AZ), and Robert Schiowitz (Gila National Forest, NM). Fieldwork in
eastern Arizona and western New Mexico was facilitated by Special-Use Permit
2076-01 Apache-Sitgreaves and Gila National Forests. jim Ratte (US Geologi-
cal Survey, Denver) provided admirable consultation on the Mule Creek
problem and is probably the most knowledgeable researcher on the geology of
that region. Brenda Houser (US Geological Survey, Tucson) provided invaluable
information on the stream dynamics of the Safforcl/San Simon Basin based on
her research. Pat Gilman provided the impetus for dealing with the secondary
depositional issues in the San Simon Valley by inviting me to particpate in her
project. Suzanne Kerho assisted in the EDXRF lab. Special thanks to Tim Teague
IOO M. STEVEN SHACKLEY

(Department of Geology/Geophysics, University of California, Berkeley) for

constant aid in the EDXRF lab. Many others have aided in the research, and all
are integral parts of the whole, although I take full responsibility for the data
and conclusions here.

REFERENCES

Baxter, M.j. 1994a Stepwise discriminant analysis in archaeometry: a critique. journal of An:hae-
ological Science 21:659-666.
_ _ 1994b Exploratory Multivariate Analysis in An:haeology. Edinburgh, Edinburgh University
Press.
Bouska, V. 1993 Natural Glasses. New York, Ellis Horwood.
Brooks, WE., and Ratte,]. C. 1985 Geologic map of Bear Mountain Quadrangle, Grant County,
New Mexico. U.S. Geological Survey Miscellaneous Field Studies Map MF-1782.
Burkham, D.E. 1972 Channel Changes of the Gila River in Safford Valley, Arizona, 1846-1970. USGS
Professional Paper 655-G. U.S. Geological Survey.
Church, T. 1995 Comment on Neutron Activation Analysis of Stone from the Chadron Formation
and a Clovis Site on the Great Plains by Hoard et al. ( 1992).]oumal of An:haeological Science
22:1-5.
Duerden, P., Clayton, E., Bird, j.R., Ambrose, W, and Leach, E 1987 Artefact composition and
computation. In Ambrose, WR., and Mummery, ].M.], eds., Archaeometry: Further Aus-
tralasian Studies. Canberra, Australian National University: 232-238.
Elston, W E. 1965 Rhyolite ash-flow plateaus, ring-dike complexes, calderas, lopoliths, and Moon
craters. Annals of the New York Academy of Sciences 124:817-842.
_ _ 1976 Glossary of stratigraphic terms of the Mogollon-Datil volcanic province, New Mexico.
New Mexico Geological Society Special Publication 5:131-145.
_ _ 1984 Mid-Tertiary ash flow tuff cauldrons, southwestern New Mexico.]oumal of Geophysical
Research 89:8733-8750.
Ericson,j.E., and Glascock, M.D. 1992 Chemical characterization of obsidian flows and domes of
the Coso volcanic field, China Lake, California. Poster presented at the 28th International
Symposium on Archaeometry, Los Angeles, California.
Glascock, M. D. 1991 Tables for Neutron Activation Analysis (3rd Edition). Research Reactor Facility,
University of Missouri, Columbia.
_ _ 1992 Characterization of archaeological ceramics at MURR by neutron activation analysis
and multivariate statistics. In Neff, H. ed., Chemical Characterization of Ceramic Pastes in
Archaeology. Madison, Prehistory Press: 11-26.
_ _ 1994 New world obsidian: recent investigations. In D.A. Scott, and P. Myers eds., Ar-
chaeometry of Pre-Columbian Sites and Artifacts, Proceedings of the 28th International
Symposium on Archaeometry. Getty Conservation Institute, Los Angeles: 113-134.
Green, R.C. 1987 Obsidian results from the Lapita sites of the Reef/Santa Cruz Islands. In Ambrose,
WR., and Mummery,j.M.j, eds., Archaeometry: Further Australasian Studies. Canberra, The
Australian National University: 239-249.
Harbottle, G. 1982 Chemical characterization in archaeology. In Ericson,]. E. and Earle, T.K., eds.,
Contexts for Prehistoric Exchange. New York, Academic Press: 13-51
Houser, B.B., Richter, D.H., and Shafiqullah, M. 1985 Geologic map of the Safford quadrangle,
Graham County, Arizona. Miscellaneous Investigations Series Map 1-1617. U.S. Geological
Survey.
Huckleberry, G.A. 1995 Archaeological implications of Late-Holocene channel changes on the
Middle Gila River, Arizona. Geoan:haeology 10:159-182.
CHEMICAL VARIABILITY AND DEPOSITIONAL PROCESSES 101

Hughes, R. E. 1988 The Coso Volcanic Field reexamined: implications for obsidian sourcing and
hydration dating research. Geoarchaeology 3:253-265.
_ _ 1994 lntrasource chemical variability of artefact-quality obsidians from the Casa Diablo
Area, California. journal of Archaeological Science 21:263-271.
Hughes, R. E., and Smith, R.L. 1993 Archaeology, geology, and geochemistry in obsidian prove-
nance studies. In Stein, j.K. and Linse, A.R., eds., Scale on Archaeological and Geoscientific
Perspectives. Geological Society of America Special Paper 283: 79-91.
International Association for Obsidian Studies (lAOS) 1994 Field and laboratory standards in
obsidian geochemistry: SAA workshop sponsored by lAOS. International Association for
Obsidian Studies Bulletin 11: l.
jack, R.N. 1971 The source of obsidian artifacts in northern Arizona. Plateau 43:103-114.
_ _ 1976 Prehistoric obsidian in California: geochemical aspects. In Taylor, R.E., ed., Advances
in Obsidian Glass Studies: Archaeological and Geochemical Perspectives. Park Ridge, New
jersey, Noyes Press: 183-217.
jackson, T.L. 1989 Late prehistoric obsidian production and exchange in the North Coast Ranges.
California. In Hughes, R.E., ed., Current Directions in California Obsidian Studies. Berkeley:
Contributions of the University of California Archaeological Research Facility 48: 79-94.
johnson, R. A., and Wichern, D. W 1982 Applied Multivariate Statistical Analysis. Englewood Cliffs,
New jersey, Prentice-Hall.
Lavin, L., and Prothero, D.R. 1992 Prehistoric procurement of secondary sources: the case for
characterization. North American Archaeologist 13:97-113.
LeTourneau, I' D. 1994 Geologic investigations of the Antelope Wells obsidian source, southern
Animas Mountains, New Mexico. Paper presented at the 59th Annual Meeting of the Society
for American Archaeology, Anaheim, California.
Macdonald, R., and Bailey, O.K. 1973 The Chemistry of the Peralkaline Oversaturated Obsidians.
Geological Survey Professional Paper 440-N-l.
Macdonald, R., Davies, G.R., Bliss, C.M., Leat, P.T., Bailey, O.K., and Smith, R.L. 1987 Geochemistry
of high-silica peralkaline rhyolites, Naivasha, Kenya Rift Valley. journal of Petrology
28:979-1008.
Mahood, G., and Hildreth, W 1983 Large partition coefficients for trace elements in high-silica
rhyolites. Geochimica et Cosmochimica Acta 4 7:11-30.
Mahood, G., and Stimac, j.A. 1990 Trace-element partitioning in pantellerites and trachytes.
Geochimica et Cosmochimica Acta 54:2257-2276.
Meltzer, D. J. 1985 On stone procurement and settlement mobility in eastern fluted point groups.
North American Archaeologist 6:1-27.
Mitchell, D. R., and Shackley, M.S. 1995 Classic period Hohokam obsidian studies in southern
Arizona. journal of Field Archaeology 22:291-304.
Nations, D., and Stump, E. 1981 Geology of Arizona. Dubuque, 1owa, Kendall/Hunt Publishing
Company.
Newman, JR. 1994 The effects of distance on lithic material reduction technology. journal of Field
Archaeology 21:491-501.
Ratte,]. C. 1982 Geologic map of the Lower San Francisco Wilderness Study Area and contiguous
roadless area, Greenlee County, Arizona, and Catron and Grant Counties, New Mexico.
U.S. Geological Survey Miscellaneous Field Studies Map MF-1463-B.
Ratte,]. C., and Brooks, W E. 1983 Geologic map of the Mule Creek Quadrangle, Grant County,
New Mexico. U.S. Geological Survey Miscellaneous Studies Map MF-1666.
_ _ 1989 Geologic map of the Wilson Mountain Quadrangle, Catron and Grant Counties, New
Mexico. U.S. Geological Survey Geologic Quadrangle Map GQ-1611.
Ratte,]. C., and Hedlund, D.C. 1981 Geologic map of the Hells Hole Further Planning Area (RARE
II), Greenlee County, Arizona and Grant County, New Mexico. U.S. Geological Survey
Miscellaneous Field Studies Map MF-1344-A.
102 M. STEVEN SHACKLEY

Ratte,]. C., Marvin, R.F, and Naeser, C.W. 1984 Calderas and ash flow tuffs of the Mogollon
Mountains, southwestern New Mexico. journal of Geophysical Research 89:8713-8732.
Renfrew, C. 1977 Alternative models for exchange and spatial distribution. In Earle, T.K., and
Ericson, ].E., eds., Exchange Systems in Prehistory. New York, Academic Press: 71-90.
Reynolds, S.]., Welty,]. W., and Spencer,]. E. 1986 Volcanic history of Arizona. Arizona Bureau of
Geology and Mineral Technology Fieldnotes 16:1-5.
Rhodes, R. and Smith, E. 1972 Geology and tectonic setting of the Mule Creek Caldera, New
Mexico, USA. Bulletin Volcanologie 36:401-411.
Richter, D.H., Houser, B. B., and Damon, P.E. 1983 Geologic map of the Guthrie quadrangle, Graham
and Greenlee Counties, Arizona. Miscellaneous Investigations Series Map 1-455, U.S. Geo-
logical Survey.
Shackley, M. S. 1987 Comment on Tomato Springs: the identification of a jasper trade and
production center in Southern California. American Antiquity 52:616-623.
_ _ 1988 Sources of archaeological obsidian in the Southwest: an archaeological, petrological,
and geochemical study. American Antiquity 53:752-772.
_ _ 1990 Early hunter-gatherer procurement ranges in the Southwest: evidence from obsidian
geochemistry and lithic technology. Ph.D. dissertation, Arizona State University. Ann Arbor.
University Microfilms.
_ _ 1992 The Upper Gila River gravels as an archaeological obsidian source region: implications
for models of exchange and interaction. Geoarchaeology 7:315-326.
_ _ 1993a Gamma rays, x-rays, stone tools and the "sourcing" myth: are we missing the point?
Paper presented in the symposium Archaeological Science: Past Achievements/Future Direc-
tions, at the 58th Annual Meeting of the Society for American Archaeology, St. Louis.
_ _ l993b (ms) Obsidian procurement and distribution in prehistoric sites in the San Simon
Archaeological Project, Southeastern Arizona: an energy dispersive x-ray fluorescence
(edxrO study. Prepared for Pat Gilman, Department of Anthropology, University of Okla-
homa, Norman.
_ _ 1994 Intersource and intrasource geochemical variability in two newly discovered archae-
ological obsidian sources in the Southern Great Basin: Bristol Mountains, California and
Devil Peak, Nevada. journal of California and Great Basin Anthropology 16:118-129.
_ _ 1995 Sources of archaeological obsidian in the greater American Southwest: an update and
quantitative analysis. American Antiquity 60:531-551.
Shelley, P.H. 1993 A geoarchaeological approach to the analysis of secondary lithic deposits.
Geoarchaeology 8:59-72.
Skinner, C. 1983 (ms) Obsidian studies in Oregon: an introduction to obsidian and an investigation
of selected methods of obsidian characterization utilizing obsidian collected at prehistoric
quarry sites in Oregon. M.A. thesis, University of Oregon.
Wilson, E. D., and Moore, R. T. 1958 Geologic map of Graham and Greenlee Counties, Arizona.
Arizona Bureau of Mines, University of Arizona.
Wyckoff, D.G. 1993 Gravel sources of knappable alibates silicified dolomite. Geoarchaeology
8:35-58.
 Chapter 5

Characterization of
Archaeological Volcanic Glass
from Oceania
The Utility of Three Techniques

MARSHALL I. WEISLER AND DAVID A. CLAGUE

ABSTRACT

Distributional studies of archaeological volcanic glass play key roles in deter-

mining the scale, complexity, and duration of prehistoric social interaction
between island societies in Oceania. Volcanic glass distributions have shed light
on the spatial limits and antiquity of Melanesian exchange systems and on
strategies necessary to colonize and sustain communities on isolated landfalls.
The efficacy of three characterization techniques of Oceanic glasses are exam-
ined: macroscopic source assignment of a unique volcanic glass (ignimbrite)
from the Pitcairn Group, southeast Polynesia; non-destructive energy disper-
sive x-ray fluorescence (EDXRF) of Melanesian obsidian; and electron micro-
probe analysis of Hawaiian volcanic glass. Suggestions for facilitating volcanic
glass studies in Oceania are offered.

MARSHALL I. WEISLER • Department of Archaeology, University of Otago, Dunedin, New

Zealand. DAVID A. ClAGUE • Monterey Bay Aquarium Research Institute, Moss Landing,
California.

Archaeological Obsidian Studies, edited by Shackley.

Plenum Press, New York, 1998.

103
104 MARSHALL I. WEISLER AND DAVID A. CLAGUE

INTRODUCTION

Volcanic glass is an important commodity for archaeological distribu-

tional studies due to its widespread archaeological occurrence (in terms of
both time and space), limited geological provenance, and the ability to
accurately characterize artifacts and assign to source by chemical composi-
tional studies.
Since Roger Green's important exposition on the application of Pacific
obsidian research to archaeology more than three decades ago (1962), it is
now almost routine to geochemically analyze volcanic glass artifacts-by a
range of techniques-to address important questions on long-distance social
interaction or "trade and exchange" (Best 1987; Weisler 1990, 1993a), island
colonization strategies and mobility patterns (Summerhayes and Allen 1993),
differential access to and changing use of sources (Kirch et al. 1991; Sum-
merhayes et al. 1993), and technological and use-wear studies (Schousboe
et al. 1983; Sheppard 1993). Additionally, a series of influential articles during
the 1960s on Mediterranean and Near Eastern obsidian by Renfrew and
colleagues (Renfrew et al. 1965, 1968; Cann and Renfrew 1964) established
the world-wide importance of obsidian distributional research which further
encouraged investigations in those areas of the Pacific with obsidian sources
and artifacts. These include, for example, regions of Near Oceania (islands
east of the Solomons [Green 1991:495]), New Zealand, and Rapa Nui (Easter
Island).
In this paper, we review the efficacy of three characterization techniques
of Oceanic glasses. First, we examine the macroscopic identification of vol-
canic glass from the Pitcairn Group, southeast Polynesia, and suggest that
simple unaided visual identification of this unique rock is possible within
the context of Polynesia. Second, we demonstrate that the non-destructive
application of energy dispersive x-ray fluorescence (EDXRF) to obsidian from
the Bismarck region of Near Oceania is a powerful, quick, and inexpensive
technique for sorting large artifact assemblages prior to more detailed geo-
chemical analyses. Third, the utility of wave-length dispersive electron mi-
croprobe characterization of volcanic glass artifacts and source rock from
Hawaii is seen as a necessary technique for characterization of these routinely
collected small artifacts (see Fig. 5.1 for location of island groups mentioned
herein). Statistical and geological grouping techniques are explored and
suggestions are offered for increasing inter-laboratory comparability of de-
veloping databases.
Throughout this paper, we use the term volcanic glass as a general label
to indicate non-crystalline glasses. Obsidian is a subclass of volcanic glass and
includes a high-silica volcanic glass, essentially rhyolitic in composition.
Basaltic glass refers to Hawaiian glasses that are more or less basaltic in
composition.
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA 105

ANDESITE LINE
20'

:...,

ISLANDS

'.
:.":.-:: MafqJoHa-1 " ·

~...
.,--.
~~·_'·(·~· '-.

Ga-.- . ·. Pftcolm ...

POLYNESIA

,..,. 130' 110'

Figure 5.1. Map of the Pacific showing the location of the Andesite Line.

MACROSCOPIC SOURCE IDENTIFICATION OF VOLCANIC

GLASS ARTIFACTS FROM SOUTHEAST POLYNESIA

Of all the characterization techniques for Oceanic glasses, macroscopic

descriptions are normally the most unreliable for assigning artifacts to source.
This is especially true when describing and attributing source to fine-grained
basalt (Weisler 1993b: 68- 69). However, macroscopic attributes such as color,
luster, translucency, and opaqueness, are useful characteristics for assigning
artifacts to source when the sources and regional geology are well-known. This is
true for particular regions of California (Bettinger et al. 1984) and, perhaps,
unique deposits in New Zealand (Davidson 1972). Here, we make the case that
macroscopic identification of the major source of Pitcairn Island volcanic
glass-within the geological context of southeast P olynesia-is fast, inexpen -
sive, and reliable.

Regional Geology

For purposes of this discussion , we define southeast Polynesia as the

island groups of Mangareva and Pitcairn, separated by about 400 km, and the
distant landfall of Easter Island (Rapa Nui) situated nearly 1600 km to the east
106 MARSHALL I. WEISLER AND DAVID A. CLAGUE

>20· 100"

MAROUESAS IS.

... .<u-i ...

.. • :<r·, ·.~P l"u
COOK IS.·· SOCIETY IS. ·. .. ... / ~

Henderson

AUSTRAL IS.
/' ~0 / . Ducle
MANGAAE'A IS. \
Pllcairn EASTER IS.

Figure 5.2. Southeast Polynesia with Mangareva, the Pitcairn Group, Rapa Nui (Easter Island),
and the Tuamotus.

(Fig. 5.2). The closest archipelago to this region is the Tuamotus, a highly
dispersed group of low coral atolls, each lying atop a submerged volcanic base.
All these islands are of Oceanic Island Basalt origin and the Pitcairn-Mangareva
lineament is time-progressive with islands increasing in age toward the north-
west (Duncan and McDougall 1976), thus conforming to the "hot-spot" hy-
pothesis (Woodhead and McCulloch 1989). Geologic studies in southeast
Polynesia, especially during the past two decades, have added significantly to
our understanding of the age, petrology, geochemistry, and geomorphology of
these islands (Baker et al. 1974; Bandy 1937; Blake 1995; Carter 1967; Duncan
and McDougall1976; Spencer 1989; Spencer and Paulay 1989; Woodhead and
McCulloch 1989).
Volcanic glass has a limited distribution within the region of southeast
Polynesia. Despite two months of survey on all the main islands of Mangareva
(Weisler 1996), only thin selvages of volcanic glass-measuring only a few
millimeters in width-were observed at the margins of some dikes and their
manufacture into stone tools is unlikely. At the opposite extreme, Rapa Nui has
four major sources of obsidian (Beardsley et al. 1991) that were manufactured
into bifacial forms (mata'a) and simple flake tools, but, on present evidence,
have not been recovered from archaeological contexts off the island. Pitcairn
has the only other sources of volcanic glass in southeast Polynesia: thin, glassy
borders of some dikes found at the western end of the island at Tedside (Carter
1967); and the major source located at the southeast coast at Down Rope
(Weisler 1995, in press). Within the whole of Polynesia, this glass source is
unique as it is an ignimbrite formed by the deposition and consolidation of ash
flows. This welded tuff has characteristic black to grey "patches" in a glassy
groundmass (Fig. 5.3).
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA 107

Figure 5.3. Photos of thin-section (u pper) a nd ha

n d specimen (lower) of Down Rope glass. Scale:
4.5 em ; 1.5 microns.
108 MARSHALL I. WEISLER AND DAVID A. CLAGUE

Archaeological Context and the Analytical Assemblage

The colonization of the remote eastern Pacific islands marked the con-
cluding chapter in the greatest over-water migration in human history. At the
margin of southeast Polynesia lie several islands with evidence of prehistoric
habitation, yet were found to be abandoned at European contact. Limited in
sustainable resources, these ecologically marginal landfalls are typified by the
four islands in the Pitcairn Group. Archaeological research was conducted on
the elevated limestone island of Henderson to understand the role of inter-is-
land voyaging in sustaining small isolated populations in this region of south-
east Polynesia (Weisler 1994, 1995, 1996,in press). From the geographic scale,
frequency, and duration of volcanic artifacts found on Henderson, we could
infer the relative degree of isolation of the Henderson population.
An ideal context for conducting a provenance study of volcanic artifacts
is when the assemblage derives from a non-volcanic island such as a low coral
atoll or raised limestone island. On islands such as these, all volcanic artifacts
are imported (save, perhaps, those arriving in the roots of drift logs). Conse-
quently, the emphasis is on where did the volcanic artifacts originate and not
merely if they are imported. In this study, we selected volcanic glass artifacts
from stratified well-dated habitation sites on the raised limestone island of
Henderson island, one of four in the Pitcairn Group (Weisler 1994, 1995, in
press). The 14 artifacts had mean dimensions of: length 25.42±6.53 mm;
thickness 6.46±4.38mm; and weight 3.26±3.27gms.

Methods and Results

To determine whether the Down Rope ignimbrite could be identified
consistently from other siliceous rocks from Polynesia, we selected volcanic
glass from archaeological and geological sources from Samoa, Mangareva, the
Marquesas, Hawaii, and Rapa Nui for visual identification. While there is some
visual variability between non-Pitcairn, Polynesian volcanic glass, we were
100% successful in separating the Down Rope, Pitcairn rock from the other
Polynesian sources. This observation was tested by analyzing, by non-destruc-
tive energy dispersive x-ray fluorescence (EDXRF), l4 artifacts macroscopi-
cally identified as Down Rope glass, with 14 sources from throughout
Polynesia. (Complete details of this analysis are reported in Weisler 1993a:
216-217, 233-239). Fig. 5.4 is a bivariate plot ofEDXRF data illustrating the
correspondence between Henderson artifacts and the Down Rope source, thus
confirming our results obtained macroscopically.

Conclusions
We conclude that the Down Rope, Pitcairn volcanic glass can be iden-
tified macroscopically which has been confirmed by EDXRF analysis. We
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA 109

0 .5

•
Raoul

O•

03
....
N

• •
....... Samoa
>- Rapa Nul Tonga

•
02
New Zealand

0'

Samoa •
• New Zeala.nd

• Hawaii
.·~··-
Down Rope, P~c.oirn
Marquesas
•
Samoa

0 ' 0.2 o• o.o 0.8 0.7 1.0 80

Sr/Rb
Figure 5.4. Bivariate plot of Henderson artifacts, the Down Rope source, and other Polynesian
sources of volcanic glass.

suggest that the macroscopic identification and source assignment of Poly-

nesian volcanic glass is possible only when: ( l) the regional geology is
reasonably well-known; (2) the major archaeological sources of volcanic glass
have been documented; and (3) the artifacts to be sourced have unique
macroscopic characteristics. We stress that macroscopic determinations of
artifact source should be confirmed initially by the use of more powerful
geochemical techniques.

ENERGY DISPERSIVE X-RAY FLUORESCENCE (EDXRF)

ANALYSIS OF MELANESIAN OBSIDIAN

It is no coincidence that the exchange systems of the southwestern Pacific

are, to date, the best described ethnographically and archaeologically for the
whole of Oceania. Ethnographers found the ubiquity and diversity of Melane-
sian exchange systems fertile ground for plying their craft and have produced
a wealth of descriptive tomes (e.g., Davenport 1962, 1964; Harding 1967;
Malinowski 1922). Archaeologists were later to demonstrate the antiquity and
historical development of some of these systems (e.g., Allen 1977, 1984; Irwin
1983, 1985; Lilley 1988; Roades 1982) and long-distance exchange has been
archaeologically documented throughout the southwestern Pacific between
various La pita sites (ca. 3600 to 2000 BP). Summerhayes and Allen ( 1993; Allen
1984) sourcing obsidian, have also demonstrated that some form of interaction
took place during the late Pleistocene in the Bismarck Archipelago.
110 MARSHALL I. WEISLER AND DAVID A. CLAGUE

That obsidian artifacts are found in many sites throughout Melanesia and
over nearly 20,000 years of prehistory, attests to its great importance in
archaeological interaction studies. Furthermore, the oftentimes large quantities
of obsidian artifacts make sampling an important consideration prior to any
analytical techniques at the site or regional levels. Large sample sizes and
expensive, often time-consuming technical analyses have prompted, over the
years, experimentation with alternative kinds of techniques for characterizing
obsidian artifacts and assigning specimens to source.
In this section, we analyze obsidian artifacts from a Lapita site in the
Mussau Islands (Fig. 5.1) to demonstrate the efficacy of the energy dispersive
x-ray fluorescence (EDXRF) procedure-a fast, reliable, inexpensive, and
non-destructive technique.

Archaeological Context and the Analytical Assemblage

As part of the La pita Homeland Project (Allen 1984) and the Mussau
Project (Kirch 1987), Weisler conducted excavations at a deeply stratified
coastal rockshelter (EKQ, Epakapaka) on the island of Mussau, in 1986, where
more than 12,000 ceramic sherds and 745 obsidian flakes were recovered from
cultural deposits reaching 2.6m in depth (Kirch et al. 1991). Two square meters
were excavated near the center of the rockshelter on both sides of the drip line.
Eight layers were identified: four cultural overlying four non-cultural strata.
Obsidian artifacts were recovered from the surface and from 18 arbitrarily
defined levels spanning the entire cultural sequence.
As part of the Mussau Project, a sample of obsidian flakes from the EKQ
rockshelter were sampled by a previous investigator and subjected to heavy
liquid sorting by submersion into specific densities of sodium metatungstate
(see Green 1987 for analytical procedures; summary results reported in Kirch
et al. 1991). Density boundaries were established at 2.356 and 2.387 specific
gravity based on Green (1987), following the established density ranges for
Admiralty (Manus) and Willaumez (West New Britain) sources (Ambrose n.d.;
Kirch et al. 1991:156).
The EKQ obsidian was sorted into three groups based on heavy liquid
sorting: (1) Manus; (2) West New Britain; and (3) "mixed." We selected eight
obsidian flakes each from the artifacts sourced to a Manus and West New
Britain provenance and 18 specimens from the "mixed" group. These 34
obsidian flakes were from throughout the sequence of test pit l. The mean
specimen dimensions were: length 23.69±5.77mm; thickness 4.36±l.41mm;
and weight l.84±l.22gms. Selection was biased towards the larger specimens
that were within the optimum size requirements for the EDXRF equipment.
Our analysis, then, was directed towards checking the efficacy of the heavy
liquid technique and assigning the "mixed" specimens to a more specific
source.
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA lli

Methods

The EDXRF analysis was undertaken by Weisler at the Department of

Geology and Geophysics, University of California, Berkeley in December, 1992.
Elemental analyses were obtained using a Spectrace 440 energy dispersive x-ray
fluorescence spectrometer equipped with a Tracor Northetn X-ray (Spectrace)
TX 6100 analyzer with a dedicated IBM PC based microprocessor and Tracor
Northern reduction software. Oxides (Ti0 2 , MnO, and FeO) and trace elements
(Ni, Cu, Zn, Ga, Pb, Th, Rb, Sr, Y, Zr, and Nb) were analyzed with a Rh x-ray
tube operated at 30 keY, 0.20 rnA pulsed, with a O.l27mm Rh primary beam
filter under vacuum at 500 seconds livetime (sample time). Ba, La, Ce, Pr, Nd,
and Sm were analyzed using a 241 Am 100 mCi radioscope source used to excite
these heavier atomic elements. Oxide and trace element intensities were
converted to concentrations by a least squares calibration line calculated for
each element intensity of interest by analyzing nine international standards:
G-2, BCR-1, AGV-1, GSP-1, STM-1, BHV0-1, W-2, RGM-1, and NBS-278. The
U.S. Geological Survey standard BHV0-1 was used along with the analysis of
every 19 unknowns for monitoring machine accuracy and precision.
Having been subjected previously to heavy liquid sorting, the artifacts
were clean and free of surface dirt, but specimens were submerged in a distilled
water sonic bath for 15 minutes as part of a standardized procedure.

Results
A dendrogram was produced using Ward's method with 34 unknowns
and two source samples each from the Manus and West New Britain sources.
All analytical results with acceptable values for oxides and elements were used
as discriminators and included: Ba, La, Ce, Ti0 2 , MnO, FeO, Ni, Zn, Ga, Rb,
Sr, Y, Zr, Nb. Fig. 5.5 displays two groups (West New Britain and Manus source
assignments) formed by this statistical clustering method.
Table 5.1 presents the comparative results of the heavy liquid sorting and
the EDXRF analyses. Note that by using EDXRF all of the "mixed" group was
assigned to the Manus source, eight of the West New Britain sourced samples
were confirmed, seven of the West New Britain source designations were
correct, while one was clearly in error. This incorrect source assignment made
by heavy liquid sorting may have been due to the specimen containing
impurities or vesicles that effect specific gravity measurements (see Bandy
1937:1606).

Conclusions

The non-destructive energy dispersive x-ray fluorescence technique is a

powerful procedure for accurately sorting large groups of obsidian artifacts and
112 MARSHALL I. WEISLER AND DAVID A. CLAGUE

Dendrogram using Ward Method

Rescaled Distance Cluster Combine
C A S E 0 5 10 15 20 25
Label Seq
92-549 5 -+
92-554 10 -+
92-550 6 -+
92-562 18 -+
92-545* 1
92-551 7 -+
92-546* 2 -+
92-547 3 -+
92-548 4 -+
92-560 16 -+
92-577 32 -+
92-573 28 -+
92-575 30 -+-·
92-582
92-578
92-572
37
33
27
-+
-+
-+
I
92-552 8 -+
92-571
92-563
92-570
26
19
25
-+
-+
-+-·
I
92-574 29 -+
92-576 31 -+
92-580 35 -+
92-581 36 -+
92-556** 12 -+
92-566 21 -+
92-558 14 -+
92-555** 11 -+
92-565 20 -+
92-553 9 -+
92-559 15
92-567 22 -+
92-561 17 -+
92-583 38 -+
92-557 13 -+
92-569 24 -+
92-568 23 -+
92-579 34 -+
Figure 5.5. Dendrogram of EDXRF data using Ward's method with Mussau obsidian. Samples
marked by a single asterisk are from the Talasea, West New Britain source, while those noted by a
double asterisk are from the Lou, Manus Islands source.

source material. The technique is fast (up to 38 unknowns can be analyzed

during a 12 hour period), inexpensive, relatively easy to operate in reference
to other geochemical techniques, and requires minimal sample preparation for
non-destructive applications.
We believe that heavy liquid sorting of obsidian artifacts continues to be
an important analytical tool for sorting large assemblages without expensive
equipment and specialized training. However, the non-destructive EDXRF
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA 113

Table 5.1. Mussau Obsidian Source Assignment by

Heavy Liquid Sorting and Energy Dispersive X-ray
Fluorescence

Obsidian source Heavy liquid Energy dispersive

assignment sorting x-ray fluorescence

Manus 8 27

West New Britian 8 (I in error) 7

"Mixedn 18 0

Total 34 34

technique can be efficiently used in conjunction with heavy liquid sorting to

discriminate further problem or "mixed" specimens and samples that are not
good candidates for this technique. These include, for example, obsidian with
impurities or vesicles that affect specific gravity measurements.

ELECTRON MICROPROBE CHARACTERIZATION OF

HAWAIIAN VOLCANIC GLASS

During the 1950s and well into the following decade, Hawaiian volcanic
glass was thought of little scientific value and was routinely discarded on the
back dirt piles at most excavations. On rare occasions, volcanic glass artifacts
were collected, but merely tabulated in reports (Bonk 1954). The analysis of
Hawaiian volcanic glass artifacts came into prominence in the 1970s when it
was thought that the hydration-rim dating technique could be applied to this
material type (Barrera and Kirch 1973; Morgenstein and Riley 1974; Morgen-
stein and Rosendahl 1976); others have commented on the futility of this
approach (Olson 1983).
The most common subaerial geological sources of Hawaiian basaltic glass
are dikes. More than 1250 are found throughout the islands and are potential
sources (Weisler 1990: Table 1) because some dikes have narrow glassy chilled
margins up to 30 mm wide (Weisler 1990: Figure 4). Other volcanic glass
sources include the glassy tops and bottoms of young lava flows, trachyte cones
and domes (e.g., Pu'u Wa'awa'a), and redeposited nodules in gulches.
The production of volcanic glass flakes from small cores prised from the
glassy borders of dikes has been described (e.g., Schousboe et al. 1983), yet
archaeologists still have not presented a convincing argument detailing how
volcanic glass was used in prehistory (Barrera and Kirch 1973).
114 MARSHALL I. WEISLER AND DAVID A. CLAGUE

Despite the almost routine presence of volcanic glass at most Hawaiian

habitation sites, and its localized geological provenance, it is only recently that
the application of geochemical techniques have been used to address questions
of prehistoric social interaction (Weisler 1990). We describe the electron
microprobe analysis of volcanic glass artifacts from two archaeological projects
on Moloka'i, Hawaiian Islands. Because of the generally small size of volcanic
glass artifacts, the electron microprobe is especially well-suited to the geo-
chemical analysis of these specimens since only the glass itself is analyzed by
excluding phenocrysts and other inclusions.

Archaeological Context and the Analytical Assemblages

The role of inter-group communication in Hawaiian prehistory has never
been assessed, yet Hawaii is considered an exemplar of a highly stratified
Polynesian chiefdom. During late prehistory it is generally believed that each
traditional land unit (ahupua'a) was economically self-sufficient (Handy and
Pukui 1958; Earle 1978) and that exchange of resources was primarily between
productively specialized households within ahupua'a. Examination of the
distribution of fine-grained basalt and volcanic glass resources in relation to
political boundaries suggests, however, that some form of interaction or ex-
change must have occurred. Taking the island of Moloka'i as an example, we
analyzed volcanic glass source samples and artifacts from two different settle-
ment regions.
Archaeological research on the island of Moloka'i has been sporadic, yet
extensive with major island-wide surveys of religious sites (Stokes 1909),
coastal fishponds (Summers 1964), intensive settlement pattern surveys and
excavations in windward (Kirch and Kelly 1975) and leeward environments
(Weisler and Kirch 1985), and a recent summary of the radiocarbon chronology
of the island (Weisler 1989). Fifth in size of the major Hawaiian Islands,
Moloka'i displays a striking array of ecological contrasts with large well-wa-
tered windward valleys and broad expanses of dry leeward relatively undis-
sected topography. The oldest habitations, dating to ca. A.D. 300 to 600, have
been found in the ecologically-favored zones with ample rainfall, excellent
agricultural soils, and easy access to a broad range of marine resources. In
contrast, the leeward western regions were settled late in prehistory, beginning
around the 12th century A.D. (Weisler 1989), and habitations are generally
found at specific resource areas such as stone quarries, pockets of arable land,
and at sheltered embayments along the coast.
Our largest analytical assemblage of volcanic glass artifacts derives from
four rockshelters, a coastal midden, and a small surface scatter from the
Mo'omomi region of leeward west Moloka'i. Settlement patterns, subsistence
data, and artifact analysis suggest that these sites were occupied sporadically
to exploit seasonally abundant marine and terrestrial resources and to collect
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA 115

fine-grained basalt and volcanic glass in nearby quarries (Bonk 1954; Weisler
1990, 1991).
In 1952, William Bonk excavated a series of dry rockshelters on the
western region of Moloka'i mostly within the coastal dune strip (Bonk 1954).
These excavations were significant in that Bonk sampled more than 50% of
each rockshelter (measured within the drip line), and he saved surface finds
and volcanic glass samples from most excavation levels. Weisler returned to
several of Bonk's sites and excavated additional samples using finer screens to
better understand Bonk's collection biases (Weisler 1991).
The analytical assemblages consist of 50 Mo'omomi volcanic glass arti-
facts with mean dimensions of: length 17.41±4.58mm; thickness
7.53±4.14mm; and weight 2.38±2.66gms. In addition to the Mo'omomi arti-
facts, we analyzed two flakes from a volcanic glass working area near sources
at the summit of Maunaloa volcano above Mo'omomi, and two flakes from each
of two late prehistoric residential complexes along the south-central coast at
Kawela (Weisler and Kirch 1985).
During the 1980s, Weisler conducted a long-term geological survey of
west Moloka'i to record fine-grained basalt quarries and to locate sources of
volcanic glass (Weisler 1990). Samples from four sources of volcanic glass are
included in this study. In total, 55 artifacts from nine sites, and five geological
samples from four sources were analyzed.

The Volcanic Glass Sources of Moloka'i

The only known archaeological sources of volcanic glass on Moloka'i are
within the geological province known as the West Moloka'i Volcanics. The
sources are all dikes found exposed in gulches. Despite intensive survey along
the western rift zones of volcanic cones which can sometimes contain glass
deposits, no sources were located (Weisler 1990). Glassy rims still exist on
dikes in the deep valleys on the windward north shore of East Moloka'i. These
are admittedly more altered than those from West Moloka'i, but they still have
unaltered glass. However, the East and West Moloka'i lavas are chemically
distinct.
The West Moloka'i Volcanics consists of tholeiitic lavas erupted during
the shield-building stage and a small volume of alkalic lavas erupted during a
postshield stage. The alkalic lavas include small cinder and spatter cones and
the thick flows erupted from them; they are concentrated in the northwestern
part of the island. These lavas are characterized by low Si0 2 and high Al 20 3 ,
and Na 2 0, K20, Ti0 2 , and P 20 5 contents. The summit region consists entirely
of tholeiitic lava flows, including some that have unusually low MgO, CaO,
and Alp 3 and high Si0 2 , Ti0 2 , and Pp 5 contents. This lava is formed by
crystallization in the magma chamber and removal of olivine, plagioclase, and
augite from the more typical tholeiitic magma. All the thoeliitic lavas from West
116 MARSHALL I. WEISLER AND DAVID A. CLAGUE

Moloka'i have a distinctive mineralogy and compositions characterized by the

presence of rare orthopyroxene and high Si0 2 contents. They are most similar
in their chemistry to lavas from Mauna Loa Volcano on Hawaii.
Three sources, FDD, WFD-1, and WM0-11, are located in the summit
region of Maunaloa volcano exposed along gulch drainages (Weisler 1990:
Figure 3). These sources are all tholeiitic basalts and the FDD source is the
unusuallow-MgO, high Si0 2 lava noted above. The other two sources are more
typical tholeiitic basalt. These dikes average about 1.5m wide and measure up
to 60m in length. The glassy borders are ca. 10 to 30mm wide. From survey of
this summit area over a 10 year period, it was discovered that dike glass sources
are subject to repeated burial and exposure from seasonal floods that descend
rapidly through the steep terrain.
The fourth source is a large east-west trending dike that crosses Ka'a
Gulch just west of the Mo'omomi area. It is an alkalic basalt typical of the
postshield alkalic lavas. The dike appears as a linear dense mass of basalt
exposed ca. 1m above the eroded and steep slopes of the gulch. Glass can be
easily collected near the base of the dike, but is of poor quality.

Methods
The archaeological and geological source glasses were prepared as pol-
ished thin sections about 30 microns thick and analyzed using a 9-channel
SEMQ electron microprobe at the U.S. Geological Survey in Menlo Park,
California. Instrument parameters included 15 Kv accelerating voltage, 10 nA
beam current, and a roughly 5 micron beam diameter. The glasses were
analyzed against natural and synthetic standards and backgrounds were meas-
ured on high- and low-mean atomic number standards. Two natural glasses
(VG-2 and A-99, both distributed by the Smithsonian Institution) were ana-
lyzed repeatedly through the run to check for precision and accuracy. The data
were reduced on-line using a modified Bence-Albee correction procedure. The
microprobe set-up is such that Si, Fe, Mg, Ca, Na, and Ti are on fixed channels,
and AI, P, K, S, and Mn are analyzed on programmed scanners. The analytical
technique is non-destructive, although the preparation of the polished thin
sections destroys some sample. Glass chips as small as 50 microns can be
mounted and polished easily and even smaller specimens (down to about 20
microns) can be handled, but with more difficulty. Most of the samples used in
this study were 1mm across or larger.

Results
When faced with a large array of multivariate data, archaeologists often
resort to statistical clustering techniques to discover groups and find patterns
(see, for example, Fig. 5.5). Here, however, we take a geological approach to
discovering patterns in our data set. Illustrated in Fig. 5.6 are bivariate plots of
ARCHAEOLOGICAL VOLCANIC GlASS FROM OCEANIA Il7

oxide values of volcanic glass artifacts and source rocks based on data in Table
5.2. We have selected two plots: Ti0 2 vs MgO and CaO vs TiOr These plots
can be used to trace the geochemical changes that occur within a group of lavas
that are related by simple fractionation of the observed minerals olivine,

0 0 X1
4.5
0
c:P'o
........
~
ofl
..: 3.5
~
'-' 0 0J+ o jf A
N 0 ~ 3+0 + 8
0 + ""+
E= 0.
2.5

2\
1.5
7.0 8.0 9.0 10.0 11.0
CaO (wt.%)

0 Tholeiitic basalt-lavas
0 0
4.5 X1
0 0 0 Alkalic basalt-lavas
00 0 0 0
,....... 0
0 + Tholeiitic archaeological glasses
oQ:>
~ X
..: 3.5 0
Alkalic archaeological glasses
! 0 J+
q,O +
§ItA 0
s
Geologic glasses
ooo 3++ o~ H
E=
2.5 ~~
+
+2
1.5 +-------,.------.....----r- ----1
3.0 4.0 5.0 6.0 7.0
MgO (wt.%)

Figure 5.6. Plots of Ti0 2 vs MgO and Ti0 2 vs CaO. The tholeiitic and alkalic lava analyses are
unpublished whole-rock data of Clague and Weisler. The archaeological and geological glasses are
from Table 5.2. Some select groups are identified by the Group identification as listed in Table 5.2
and the text. Group A is the Ka'a Gulch alkalic basalt location; Groups B, H,andj are three distinct
groups from West Moloka'i; Group l is from Mauna Kea, Hawaii, and Groups 2 and 3 are probably
from the Ko'olau Range of O'ahu. Groups B, 2 and 3 consist of two points each.
 ........QO
Table 5.2. Microprobe Analyses of Geological and Archaeological Glasses from West Moloka'i"

Sample Group Si02 Al 203 FeO* MnO MgO CaO Na20 K 20 P20s Ti02 s Total

MOLOMOMI

G-KA'A A 49.5 15.0 11.7 0.16 6.41 9.83 3.42 0.98 0.56 3.32 0.015 101.0
WM21R-5 A 49.0 14.9 11.6 0.17 6.49 9.68 3.35 1.01 0.72 3.27 0.000 100.2
WM24-2 A 48.8 14.4 11.8 0.17 6.49 9.67 3.34 0.97 0.58 3.23 0.014 99.4
WM24-14 A 49.4 15.0 11.8 0.15 6.42 9.79 3.43 0.98 0.58 3.24 0.016 100.9
WM28-7 A 48.5 14.6 11.5 0.16 6.41 9.64 3.41 1.01 0.63 3.22 0.005 99.1
WM28-17 A 49.7 15.2 11.8 0.17 6.53 9.86 3.42 1.01 0.63 3.26 0.009 101.6
~
WM890-1 A 49.0 14.9 11.8 0.17 6.50 9.82 3.36 1.00 0.64 3.30 0.005 100.6 s:
WM24-1 B 52.0 14.1 12.2 0.20 5.59 9.40 2.90 0.77 0.53 3.04 0.011 100.8 ~
WM21-66B c 52.2 14.5 11.1 0.17 6.44 10.3 2.74 0.61 0.38 2.62 0.002 101.1 e-o
WM21-74 c 51.4 14.4 11.0 0.17 6.36 10.0 2.68 0.57 0.34 2.56 0.009 99.4
WM24-3 c 51.3 14.2 11.1 0.18 6.47 10.0 2.72 0.58 0.38 2.52
~
0.005 99.5 c:;
WM24-4 51.7 14.2 11.0 0.19 6.38 10.1 2.67 0.59 !'""
c 0.36 2.61 0.008 99.8
WM24-6 c 51.6 14.2 11.1 0.17 6.43 10.1 2.75 0.58 0.39 2.62 0.007 100.1 ~
~
WM24-11 c 51.9 14.6 11.0 0.17 6.41 10.1 2.73 0.59 0.38 2.57 0.009 100.5 :t:
WM28-3 c 52.1 14.7 11.0 0.15 6.44 10.3 2.70 0.59 0.39 2.58 0.010 100.9 ~
WM28-4 c 51.5 14.2 11.0 0.17 6.40 10.0 2.66 0.60 0.35 2.58 0.004 99.5
WM28-6 c 52.1 14.5 10.9 0.17 6.37 10.2 2.74 0.60 0.33 2.54 0.013 100.5
~
;S
WM28-8 c 52.0 14.4 10.9 0.16 6.44 10.3 2.72 0.61 0.35 2.58 0.008 100.5 ~
WM28-9 c 51.2 14.1 10.9 0.17 6.38 10.1 2.66 0.58 0.34 2.54 0.010 99.0 ;r..
WM28-10 c 52.0 14.2 10.8 0.15 6.35 10.3 2.76 0.60 0.36 2.63 0.007 100.2
WM28-ll c 52.1 14.5 10.7 0.16 6.46 10.3 2.71 0.60 0.33 2.60 0.004 100.4 ~
C'l
iii
 ~
~
Sample Group Si02 Al 20 3 FeO* MnO MgO CaO Na20 K 20 P20s Ti02 s Total
~
1"1

WM28-18 c 52.1 14.5 10.8 0.17 6.47 10.3 2.82 0.62 0.36 2.67 0.006 100.8 ~
WM28-21 c 52.0 14.2 11.1 0.17 6.45 10.1 2.71 0.60 0.37 2.57 0.010 100.2 ~
WM21-71 D 51.6 14.3 11.6 0.16 6.50 10.2 2.57 0.47 0.32 2.50 0.010 100.3
WM24-12 D 51.2 14.4 11.4 0.18 6.58 10.3 2.58 0.48 0.30 2.45 0.010 99.9
WM24-13 D 51.4 14.6 11.0 0.16 6.53 10.5 2.60 0.51 0.29 2.49 0.006 100.0
s6
WM21-66A E 50.9 14.1 10.8 0.18 6.70 10.3 2.52 0.56 0.36 2.51 0.003 99.0 t"'
WM21-69 E 51.0 14.1 10.9 0.19 6.69 10.2 2.57 0.53 0.36 2.44 0.013 98.9
WM21-72 E 50.9 14.1 10.9 0.17 6.77 10.3 2.55 0.52 0.34 2.50 0.008 99.0 ~
WM21-73 E 51.5 14.1 11.1 0.17 6.71 10.5 2.61 0.58 0.35 2.53 0.011 100.1 !"'\
WM21R-1 E 51.3 14.2 11.0 0.17 6.72 10.3 2.56 0.55 0.36 2.57 0.013 99.7
WM21R-3 E 51.2 14.1 11.0 0.17 6.60 10.3 2.55 0.54 0.39 2.38 0.010 99.3 ~
~
WM21R-4 E 51.2 14.6 10.9 0.17 6.53 10.5 2.55 0.54 0.34 2.41 0.008 99.7 ~

WM21R-6 E 51.7 14.3 11.0 0.16 6.72 10.4 2.60 0.55 0.35 2.48 0.010 100.3 ~
WM24-5 E 51.2 14.2 11.2 0.16 6.74 10.6 2.46 0.46 0.33 2.50 0.011 99.8
WM24-7 E 51.6 14.4 11.2 0.18 6.73 10.5 2.48 0.49 0.31 2.36 0.014 100.3 ~
WM24-8 E 51.4 14.4 11.0 0.18 6.79 10.3 2.51 0.54 0.28 2.50 0.004 99.9 0
WM24-10 E 52.0 14.4 10.9 0.16 6.82 10.6 2.53 0.50 0.31 2.34 0.006 100.5
WM26-1 E 51.1 14.2 10.8 0.18 6.67 10.3 2.55 0.53 0.34 2.40 0.005 99.1
WM26-2 E 51.6 14.3 11.1 0.16 6.79 10.3 2.61 0.53 0.34 2.51 0.009 100.2 ~
WM28-13 E 50.9 14.4 11.2 0.18 6.88 9.96 2.58 0.57 0.37 2.56 0.011 99.7
WM28-14 E 51.3 14.2 11.2 0.15 6.78 10.2 2.61 0.58 0.36 2.62 0.014 100.0
WM28-15 E 51.6 14.7 11.2 0.17 6.44 10.5 2.53 0.51 0.31 2.54 0.003 100.5
WM21R-2 F 51.6 14.6 10.8 0.16 6.52 10.7 2.62 0.52 0.35 2.42 0.005 100.3

(continued)

....
....
l:l
 Table 5.2. (Continued)
....
N
Q

Sample Group Si02 Al 203 FeO* MnO MgO CaO Na20 K20 P20S Ti02 s Total

WM21R-7 F 50.9 14.7 10.9 0.18 6.50 10.4 2.53 0.52 0.37 2.38 0.014 99.3
WM28-1 G 51.1 14.5 11.3 0.17 6.51 10.4 2.42 0.50 0.36 2.47 0.008 99.8
WM28-2 G 51.6 14.3 11.4 0.15 6.53 10.5 2.53 0.49 0.34 2.58 0.009 100.4
WM28-16 G 51.7 14.7 11.4 0.17 6.54 10.6 2.34 0.50 0.30 2.58 0.008 100.8
WM28-20 H 51.2 13.9 11.8 0.16 6.06 10.2 2.68 0.62 0.45 2.77 0.014 99.8
WM28-12 J 51.5 13.7 13.7 0.19 5.13 9.12 2.80 0.78 0.52 3.32 0.016 100.4

SUMMIT s
~
G-WM0-11 53.6 13.6 12.4 0.18 4.11 8.08 2.63 1.02 0.67 3.27 0.007 99.6 ~
G-FDD 51.4 14.5 11.2
r-o
0.16 6.70 10.3 2.24 0.51 0.30 2.55 0.011 99.8 r-o
G-WFD-1 E 52.1 14.3 11.2 0.16 6.69 10.6 2.50 0.51 0.38 2.50 0.002 !'""'
100.8
G-WFD-IA E 52.0 14.6 11.4 0.17 6.33 10.3 2.58 0.55 0.33 2.66 0.020 101.0
WM820-1 3 52.0 14.1 11.7 0.18 4.99 8.81 2.76 0.84 0.49 2.87 0.016 98.7
WM820-2 3 52.5 14.1 11.8 0.17 5.08 8.95 2.61 0.92 0.55 3.00 0.012 99.7
~r-o
t"i
~
KAWELA ~

~
WM40-2 B 51.2 13.9 12.1 0.18 5.87 9.83 2.62 0.55 0.40 2.86 0.016 99.6
WM40-1 1 48.5 13.1 15.4 0.24 4.20 8.74 3.21 1.53 0.96 4.57 0.068 100.6
~
WMIS0-1 2 52.0
::;
14.9 10.7 0.17 6.51 10.1 2.46 0.38 0.21 1.97 0.004 99.5 t::;j
WMIS0-2 2 52.5 15.0 10.6 0.17 6.70 10.0 2.59 0.52 0.28 2.11 0.014 100.5 ?>-

"Samples beginning with G are geological samples. Groups designated by letters are from West Molokai whereas those designated by numbers are
E
~
from other islands.
~
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA 121

plagioclase, and augite. The liquid (glass) becomes enriched in Ti0 2 and
depleted in MgO and CaO as olivine (MgO), plagioclase (CaO), and augite
(MgO and CaO) crystallize. These oxide plots are commonly used by geologists
to demonstrate that groups of samples are related to one another by the simple
removal of crystals. When the liquid reaches about 4% MgO, an iron-titanium
oxide also crystallizes and Ti0 2 then decreases with decreasing MgO. Many of
the ten identified artifact groups are distinguished by mineralogy as their glass
compositions are similar, and in some cases, overlapping. Nearly all the samples
plot in a tightly defined array that overlaps the array defined by unaltered lava
and glass samples from West Moloka'i, indicating that they are from West
Moloka'i and that they are geochemically similar and can be related to one
another by the simple removal of the observed crystals.
The majority of artifacts plot with the geological glass samples and lava
samples from West Moloka'i demonstrating local sources for most specimens.
Nine volcanic glass groups or sources derive from West Moloka'i:
• Group A. Ka'a Gulch source. An alkali basalt that contains olivine
phenocrysts and skeletal olivine crystals. Six artifacts.
• Group B. Unknown source. An aphyric tholeiitic basalt with low MgO
(5.7%). Two artifacts (one, 40-2, from Kawela).
• Group C. Unknown source. A tholeiitic basalt that is nearly aphyric
but contains rare phenocrysts of plagioclase and clinopyroxene. These
glasses average 6.42% MgO. Fifteen artifacts.
• Group D. Unknown source. A tholeiitic basalt that is nearly aphyric
but contains rare phenocrysts of plagioclase and olivine and has 6.54%
MgO and lower Nap, Kp, Pp 5 , and Ti0 2 than Group C. Three
artifacts.
• Group E. WFD-1 source. A tholeiitic basalt that contains abundant
microlites of clinopyroxene and rare skeletal olivine crystals. It con-
tains an average of 6. 71% MgO. Seventeen artifacts.
• Group F. Unknown source. A tholeiitic basalt that contains only
skeletal olivine crystals and has 6.51% MgO. Two artifacts.
• Group G. Unknown source. A tholeiitic basalt that contains abundant
clinopyroxene microlites and rare plagioclase phenocrysts. It has
6.53% MgO, but higher Ti0 2 than Group F. Three artifacts.
• Group H. Unknown source. A tholeiitic basalt petrographically similar
to GroupE, but contains much lower MgO (6.06 vs. 6.71%). One
artifact.
• Group ]. Unknown source. A tholeiitic basalt that has low MgO
(5.13%) and high FeO (13.7%). It is a fractionated lava that is compo-
sitionally between GroupE and the even more fractionated WM0-11,
which has lower FeO and Ti0 2 due to crystallization of Fe-Ti oxide
minerals. One artifact.
122 MARSHALL I. WEISLER AND DAVID A. CLAGUE

In summary, all 50 Mo'omomi artifacts from six habitation sites derive

from nine West Moloka'i sources. Some 39 artifacts (78%) are from three
sources. The known source in Ka'a Gulch (the closest to the sites) and the
WFD-1 source near the summit of Maunaloa account for 23 (47%) of all
artifacts. The Maunaloa source has the widest distribution among the
Mo'omomi sites.
The Kawela and Pu'u Nana artifacts (the latter located near the summit
of Maunaloa) produced the most unexpected results in that three sources, one
from Hawaii island and the other two apparently from O'ahu, are indicated in
addition to a West Moloka'i source for artifact 40-2 (see above).
• Group 1. Mauna Kea, Hawaii Island source. Kawela artifact 40-1 is a
hawaiite (differentiated alkalic basalt) that is chemically dissimilar
from those of West Moloka'i. In particular, it contains high FeO and
low Al 20 3 compared to those from West Moloka'i. Perhaps the best
clue as to its origin comes from the sulfur content of the glass which
is 680ppm (all glasses from West Moloka'i contain <200ppm). We
suspect that this glass erupted under water and that the hydrostatic
pressure kept it from degassing its sulfur. The likely place such a glass
occurs, and with the possibility of prehistoric exploitation, is the
subglacially erupted lavas that occur near the summit of Mauna Kea
volcano. These formed when Mauna Kea was capped by glaciers
during the Pleistocene. The Mauna Kea quarry is probably the source
of this artifact.
• Group 2. Ko'olau volcano, O'ahu. Kawela artifacts 150-1 and 150-2
are tholeiitic basalt characterized by low Ti0 2 , high Alp 3 , and by
somewhat low FeO. It is very similar in composition to glass rinds on
dikes in the Ko'olau Range on O'ahu (Clague and Hazlett 1989), and
unlike lavas from any other Hawaiian volcano. Two artifacts.
• Group 3. Ko'olau Volcano, O'ahu. Pu'u Nana artifacts 820-l and 820-2
are tholeiitic basalt characterized by low Ti0 2 (compare the Ti0 2
contents to that of Group], which has similar MgO content). They
appear to be differentiated from melts having compositions similar to
Group 2 and are also probably from the Ko'olau Range on O'ahu. It is
of interest that the Pu'u Nana site is very close to the WFD-1 source,
yet at least a few artifacts derive from a probable O'ahu source.

Conclusions
The unexpected results regarding the sources of volcanic glass artifacts
from Moloka'i can be explained in terms of the different settlement patterns
and periods of occupation within the regions of Mo'omomi and Kawela. The
initial use of the Mo'omomi region began during the Early Expansion Period,
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA I23

A.D. 1100 to 1400 when population levels expanded at an unprecedented rate

from their windward locales into the dry leeward regions (Weisler
1989: 126-127). Harvesting the seasonally abundant marine and terrestrial
resources and collecting fine-grained basalt and volcanic glass from nearby
sources is documented (Weisler 1991). At Mo'omomi, the settlement pattern
is dominated by rockshelters and a few coastal middens. The lack of con-
structed residential complexes or agricultural features further suggests short-
term use of the region. Mo'omomi then, can be viewed as a hinterland for
permanently settled areas of the island such as in the windward valleys.
Conversely, Kawela typifies a permanently occupied late prehistoric
settlement pattern with numerous residential complexes, agricultural sites,
coastal fishponds, and religious features (Weisler and Kirch 1985). Although
initial settlement of Kawela was at a coastal dune occupied as early as the 16th
century, ethnohistoric sources and radiocarbon dates suggest that the upland
community was in place during the last two centuries before European contact.
During this period in Hawaiian prehistory inter-island voyaging may have been
frequent for the exchange of commodities (such as fine-grained basalt and
volcanic glass) and transport of people for military conquest. Indeed, ethno-
historic sources mention a famous battle fought at Kawela with canoes from
Hawaii island lined along the shoreline (Kamakau 1961: 171).
The reason for the large number of volcanic glass sources for the
Mo'omomi sites remains to be addressed. While numerous dikes are often
found exposed along sea cliffs, most of the easily accessible dike sources of
volcanic glass are situated within gulches. We believe that erosional processes
have an important role to play in the discovery of dike glass and long-term
prehistoric use of these resources. Over a period of about 10 years, Weisler
visited the summit region of Maunaloa and walked the gulches in search of
volcanic glass sources. It was discovered that some dikes become exposed from
the seasonal torrential flooding of the gulches, while other dikes became
buried. This may account for why dike glass was collected prehistorically from
a particular source (FDD), but not another nearby locale (WM0-11). It also
suggests that many prehistorically exploited dikes may be obscured today by
recent alluvium. Therefore, the quest to identify all the specific sources within
a volcano may be optimistic, as out study has demonstrated. Consequently,
identification to the geographic scale of volcano (see Weisler l993b:63) may
only be possible in these circumstances.

SUMMARY AND CONCLUSIONS

We have examined the utility of three techniques for characterizing

volcanic glass artifacts and source samples from Oceania. The first study
examined the efficacy of macroscopic source identification of volcanic glass
124 MARSHALL I. WEISLER AND DAVID A. CLAGUE

(ignimbrite) artifacts from Henderson island, southeast Polynesia. We stress

that source assignment of Polynesian volcanic glass artifacts based on macro-
scopic attributes is only reliable when the regional geology is well-known, the
major archaeological sources of volcanic glass have been documented, and the
artifacts to be sourced have unique macroscopic attributes. In the example of
the Pitcairn island ignimbrite, its grey to black patchy appearance was a key to
identification.
The non-destructive energy dispersive x-ray fluorescence (EDXRF) study
of obsidian artifacts from the Mussau archipelago demonstrated its utility as a
rapid, accurate, and low cost technique for analyzing large assemblages.
Selecting obsidian flakes that were previously sorted by the heavy liquid
technique utilizing sodium metatungstate, we were able to further refine
unassigned artifacts as well as demonstrate that artifacts with impurities (such
as vesicles and phenocrysts) are not good specimens for sorting by heavy liquid.
The electron microprobe is especially advantageous when limited to small
artifacts of volcanic glass which commonly occur in Hawaiian archaeological
sites. Although sample preparation is destructive, pieces as small as 30 microns
thick and 30 to 50 microns in length are the smallest sizes that can be handled
and mounted. However, one mm glass chips are more than adequate for
geochemical analysis and are also large enough for petrographic descriptions.
Because more than 1,000 potential sources of volcanic glass are located on all
the major Hawaiian Islands, finding specific sources may not be possible in most
archaeological circumstances and, in fact, may not be necessary to address
important research questions. Our examination of 50 volcanic glass artifacts
from six sites from Mo'omomi confirmed ten sources from the west Moloka'i
volcanics, yet only 56% could be assigned to a specific source. Since most dikes
with accessible volcanic glass are found in areas of high erosion, we believe
that these sources will be subject to episodes of continuous burial and exposure,
as discussed previously. Consequently, artifact assignment to a particular
volcano may be our best approximation to a source.
Of particular importance was the identification of five volcanic glass
artifacts that originated from sources on O'ahu and Hawaii island. Building on
previous research (Weisler 1990), we have empirically demonstrated the first
inter-island movement of volcanic artifacts in the Hawaiian Islands. What is
considered a rather nondescript artifact class in formal terms may, indeed, hold
great value for addressing questions of prehistoric long-distance interaction.
In closing, we would like to offer some suggestions for advancing the
utility of volcanic glass studies in the Pacific region.
First, as Summerhayes et al. have demonstrated in this volume, we need
to acquire more source samples to better understand the variability of specific
obsidian sources found on the continental side of the Andesite Line. This may
not be possible with the huge number of potential volcanic glass deposits found
on Oceanic islands. We suggest that a knowledge of regional geology (that is,
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA 125

the general geochemical composition of specific volcanoes) may suffice for

facilitating assignment of volcanic glass artifacts to a specific volcano.
Second, as new analyses are reported, sample preparation, analytical
equipment used, operating conditions, and evaluation of accuracy and preci-
sion should be reported in full (Weisler 1993b:75). Accuracy can be demon-
strated by reference to known geological standards, whereas precision may be
evaluated by analyzing the same sample several times and comparing the
results. Minimal reporting standards should be adopted and sharing of source
standards should be encouraged.
We hope that the demonstration of these three techniques for the geo-
chemical analysis of archaeological volcanic glass from Oceania will encourage
further the systematic collection and study of this material class.

ACKNOWLEDGMENTS

Field investigations in the Pitcairn Group were supported by grants from

the Stahl Endowment and Robert H. Lowie Scholarship from the University of
California at Berkeley Participation in the Mussau Expedition was by invitation
by Patrick Kirch in 1986. The Mo'omomi, Moloka'i fieldwork was made
possible by a generous grant from the Nature Conservancy of Hawaii. Weisler's
geochemically analyses were supported by the Wenner-Gren Foundation for
Anthropological Research (grant #5376). Clague acknowledges the support of
the U. S. Geological Survey, Menlo Park and Hawaiian Volcanoes Observatory
This is publication number 16 of the Pitcairn Islands Scientific Expedition.

REFERENCES

Allen,]. 1977 Sea traffic, trade and expanding horizons. In Allen,]., Golson,]. and Jones, R. eds.,
Sunda and Sahul: Prehistoric Studies in Southeast Asia, Melanesia and Australia. London,
Academic Press: 387-417.
_ _ 19841n search of the La pita homeland. journal of Pacific History 19: 186-201.
Ambrose, W n.d. A note on obsidian density measurement for separating Melanesian obsidian
sources. Manuscript, Division of Archaeology and Natural History, Research School of
Pacific and Asian Studies, Canberra, The Australian National University.
Baker, PE., Buckley, F, and Holland, ].G. 1974 Petrology and geochemistry of Easter Island.
Contributions to Mineralogy and Petrology 44: 85-100.
Bandy, M.C. 1937 Geology and petrology of Easter Island. Bulletin of the Geological Society of
America: 48: 1589-1610.
Barrera, W. and Kirch, PV 1973 Basaltic-glass artifacts from Hawaii: their dating and prehistoric
uses. journal of the Polynesian Society 82: 176-187.
Beardsley, FS., Ayres, WS., and Goles, G.G. 1991 Characterization of Easter Island obsidian
sources. Indo Pacific Prehistory Association Bulletin 11: 179-187.
Best, S. 1987 Long distance obsidian travel and possible implications for the settlement of Fiji.
Archaeology in Oceania 22: 31-32.
126 MARSHALL I. WEISLER AND DAVID A. CLAGUE

Bettinger, R., Delacorte, M., and jackson, R. 1984 Visual sourcing of central eastern California
obsidians. In Hughes, R. ed., Obsidian Studies in the Great Basin, Contributions of the
University of California Archaeological Research Facility 45. Berkeley: 63-78.
Blake, S.G. 1995 Late Quaternary history of Henderson Island, Pitcairn Group. Biological]oumal
of the Linnean Society 56: 43-62.
Bonk, W.J. 1954 Archaeological Excavations on West Molokai. Unpublished M.A. thesis, Depart-
ment of Anthropology, University of Hawaii.
Cann, ]. R. and Renfrew, C. 1964 The characterization of obsidian and its application to the
Mediterranean region. Proceedings of the Prehistoric Society 30: 111-133.
Carter, R.M. 196 7 The Geology of Pitcairn Island, South Pacific Ocean. B. P. Bishop Museum Bulletin
231. Honolulu.
Clague, D.A. and Hazlett, R.W 1989 Geological Field Guide to the Hawaiian Islands. Field Trip
Guidebook T188/304 for the 28th International Geological Congress. Washington, DC,
American Geophysical Union.
Davenport, WH. 1962 Red feather money. Scientific American 206: 94-103.
_ _ 1964 Notes on Santa Cruz voyaging. journal of the Polynesian Society 73: 134-142.
Davidson, J.M. 1972 Archaeological investigations on Motutapu Island, New Zealand. Introduc-
tion to recent fieldwork and some recent results. Records of the Auckland Institute and
Museum 9: 1-14.
Duncan, R.A. and McDougall, I. I 976 Linear volcanism in French Polynesia. journal of Volcanology
and Geothermal Research 1: 197-227.
Earle, T.K. 1978 Economic and Social Organization of a Complex Chiefdom: The Halelea District,
Kaua'i, Hawaii. Museum of Anthropology, Anthropological Papers 63. Ann Arbor: Univer-
sity of Michigan.
Green, R.C. 1962 Obsidian: its application to archaeology. New Zealand Archaeological Association
Newsletter 5: 8-16.
_ _ 1987 Obsidian results from the Reef/Santa Cruz Islands. In Ambrose, W. and Mummery,].
M.]. eds., Archaeometry: Further Australasian Studies. Department of Prehistory, Research
School of Pacific Studies, Canberra, The Australian National University: 239-249.
_ _ 1991 Near and Remote Oceania- disestablishing "Melanesia" in culture history. In Pawley,
A., ed., Man and a Half: Essays in Pacific Anthropology and Ethnobiology in Honour of Ralph
Bulmer. Auckland: The Polynesian Society: 491-502.
Handy, E.S.C. and Pukui, M.K. 1958 The Polynesian Family System in Ka'u, Hawaii. Wellington:
The Polynesian Society.
Harding, T.G. 1967 Voyagers of the Vitiaz Strait: A Study of a New Guinea Trade System. Seattle:
University of Washington Press.
Irwin, G. 1983 Chieftainship, kula and trade in Massim prehistory. In Leach,]. Wand Leach, E.
eds, The Kula: New Perspectives on Massim Exchange. Cambridge: Cambridge University
Press: 29-72.
_ _ 1985 The Emergence ofMailu as a Central Place in Coastal Papuan Prehistory. Terra Australis
10. Canberra: Department of Prehistory, Research School of Pacific Studies, Australian
National University.
Kamakau, S.M. 1961 Ruling Chiefs of Hawaii. Honolulu, The Kamehameha Schools Press.
Kirch, P.V. 1987 Lapita and Oceanic cultural origins: excavations in the Mussau Islands, Bismarck
Archipelago. journal of Field Archaeology 14: 163-180.
Kirch, P.V., Hunt, T.L., Weisler, M., Butler, V., and Allen, M.S. 1991 Mussau Islands prehistory:
results of the 1985-86 excavations. In Allen,]. and Gosden, C., eds., Report of the Lapita
Homeland Project. Canberra, Department of Prehistory, Research School of Pacific Studies,
The Australian National University: 144-163.
Kirch, P.V. and Kelly, M. eds 1975 Prehistory and Ecology in a Windward Hawaiian Valley: Halawa
Valley, Molokai. Pacific Anthropological Records 24. Honolulu: B. P. Bishop Museum.
ARCHAEOLOGICAL VOLCANIC GLASS FROM OCEANIA 127

Lilley, I. 1988 Prehistoric exchange across the Vitiaz Strait, Papua New Guinea. Current Anthropol-
ogy 29: 513-516.
Malinowski, B. 1922 Argonauts of the Western Pacific. London: Routledge and Kegan Paul.
Morgenstein, M. and Riley, T. 1974 Hydration-rind dating of basaltic glass: a new method for
archaeological chronologies. Asian Perspectives 17: 145-159.
Morgenstein, M. and Rosendahl, P. 1976 Basaltic glass hydration dating in Hawaiian archaeology.
In Taylor, R. E. ed., Advances in Obsidian Glass Studies. Park Ridge, New jersey: Noyes Press:
141-164.
Olson, L. 1983 Hawaiian volcanic glass applied "dating" and "sourcing": archaeological context.
In Clark, j.T. and Kirch, P.V eds., Archaeological Investigations of the Mud/and-Waimea-
Kawaihae Road Corridor, Island of Hawaii: An Interdisciplinary Study of an Environmental
Transect. Departmental Report Series 83-l. Honolulu: B. P. Bishop Museum: 325-340.
Renfrew, C., Cann, JR., and Dixon,j.E. 1965 Obsidian in the Aegean. Annual of the British School
of Archaeology at Athens 60: 225-247.
Renfrew, C., Dixon, ].E., and Cann, j.R. 1968 Further analysis of Near Eastern obsidians.
Proceedings of the Prehistoric Society 34: 319-331.
Rhoads, j.W 1982 Prehistoric Papuan exchange systems: the Hiri and its antecedents. In Dutton,
T. ed., The Hiri in History: Further Aspects of Long-distance Motu Trade in Central Papua.
Pacific Research Monograph 8. Canberra, The Australian National University: 131-151.
Schousboe, R., Riford, M., and Kirch, P.V 1983 Volcanic-glass flaked stone artifacts. In Clark,j.T.
and Kirch, P.V, eds., Archaeological Investigations of the Mudlane-Waimea-Kawaihae Road
Corridor, Island of Hawai'i: An Interdisciplinary Study of an Environmental Transect. Depart-
mental Report Series 83-l. Honolulu, Department of Anthropology, B. P. Bishop Museum:
348-370.
Sheppard, P.R. 1993 Lapita lithics: trade/exchange and technology. A review from the Reef/Santa
Cruz. Archaeology in Oceania 28: 121-137.
Spencer, T. 1989 Tectonic and Environmental Histories in the Pitcairn Group, Palaeogene to Present:
Reconstructions and Speculations. Atoll Research Bulletin 322. Washington, DC, Smith-
sonian Institution.
Spencer, T. and Paulay, G. 1989 Geology and Geomorphology of Henderson Island. Atoll Research
Bulletin 323. Washington, DC, Smithsonian Institution.
Stokes, ].F. G. 1909 Heiaus of Molokai. Manuscript. Honolulu, Archives, B. P. Bishop Museum.
Summerhayes, G.R. and Allen,]. 1993 The transport of Mopir obsidian to late Pleistocene New
Ireland. Archaeology in Oceania 28: 144-148.
Summerhayes, G.R., Gosden, C., Fullagar, R., Specht,]., Torrence, R., Bird, R., Shahgholi, N., and
Katsaros, A. 1993 West New Britain obsidian: production and consumption patterns. In
Frankhauser, B. and Bird, j.R., eds, Archaeometry: Current Australasian Research. Depart-
ment of Prehistory, Research School of Pacific Studies, Canberra, The Australian National
University: 57-68.
Summers, C.C. 1964 Hawaiian Fishponds. Special Publication 52. Honolulu, B. P. Bishop Museum
Press.
Weisler, M.l. 1989 Chronometric dating and late Holocene prehistory in the Hawaiian Islands: a
critical review of radiocarbon dates from Moloka'i island. Radiocarbon 31: 121-145.
_ _ 1990 Sources and sourcing of volcanic glass in Hawai'i: implications for exchange studies.
Archaeology in Oceania 25: 16--23.
_ _ 1991 The Archaeology of a Hawaiian Dune System: the Nature Conservancy$ Mo'omomi Preserve.
Honolulu: The Nature Conservancy of Hawaii: 166 p.
_ _ 1993a Provenance studies of Polynesian basalt adze material: a review and suggestions for
improving regional data bases. Asian Perspectives 32: 61-83.
_ _ 1993b Long-distance Interaction in Prehistoric Polynesia: Three Case Studies. Unpublished
Ph.D. dissertation, Department of Anthropology, University of California, Berkeley.
128 MARSHALL 1. WEISLER AND DAVID A. CLAGVE

_ _ 1994 The settlement of marginal Polynesia: new evidence from Henderson Island. journal
of Field Archaeology 21: 83-102.
_ _ 1995 Henderson Island prehistory: colonization and extinction on a remote Polynesian
island. Biological journal of the Linnean Society 56: 377-404.
_ _ 1996 An archaeological survey of Mangareva: implications for regional settlement models
and interaction studies. Man and Culture in Oceania 12.
_ _in press Taking the mystery out of the Polynesian "mystery islands": a case study from
Mangareva and the Pitcairn Group. In Davidson,]., Irwin, G., Brown, D. and Pawley, A,
eds., Pacific Culture History: Essays in Honour of Roger Green.
Weisler, M.l. and Kirch, P.V. 1985 The structure of settlement space in a Polynesian chiefdom:
Kawela, Molokai, Hawaiian Islands. New Zealand journal of Archaeology 7: 129-158.
Woodhead, J.D. and McCulloch, M.T. 1989. Ancient seafloor signals in Pitcairn Island lavas and
evidence for large amplitude, small length-scale mantle heterogeneities. Earth and Planetary
Science Letters 94: 257-273.
 Chapter 6

Application of PIXE-PIGME to
Archaeological Analysis of
Changing Patterns of
Obsidian Use in West New
Britain, Papua New Guinea
G. R. SuMMERHAYES, J. R. BIRD, R. FuLLAGAR, C. GosDEN,

J. SPECHT, AND R. ToRRENCE

ABSTRACT
During the past twenty years, PIXE-PIGME analyses of the elemental compo-
sition of obsidian have made significant contributions to the study of resource
use and distribution in the western Pacific region. More recently, detailed
recording and systematic sampling of obsidian exposures combined with
enhancements in the characterization technique itself have led to much finer
discrimination between sources in New Britain. As a result of these advances,
over 1200 archaeological samples from a large suite of sites on New Britain and

G. R. SUMMERHAYES • School of Archaeology, LaTrobe University, Bundoora, Victoria, Aus-

tralia. J. R. BIRD • Australian Nuclear Science and Technology Organization, Menai, New
South Wales, Australia. R. FULLAGAR • Division of Anthropology, Australian Museum, Syd-
ney, Australia. C. GOSDEN • Pitt Rivers Museum, Oxford University, Oxford, United Kingdom.
J. SPECHT • Division of Anthropology, Australian Museum, Sydney, Australia. R. TORRENCE
• Division of Anthropology, Australian Museum, Sydney, Australia.
Archaeological Obsidian Studies, edited by Shackley.
Plenum Press, New York, 1998.
129
130 G. R. SVMMERHAYES et al.

other islands has been assigned to their sources. The archaeological research
and the extensive characterization study reported here have revealed major
changes in the exploitation of New Britain obsidian during the past 20,000
years. It is concluded that the advantages of PIXE-PIGME as a rapid, non-
destructive technique for sourcing obsidian merit its wider use in archaeology.

INTRODUCTION

Melanesia has witnessed a great deal of recent volcanic activity because of

its location at the intersection of several Pacific plates. Outcrops of obsidian
resulting from these events have been identified within four main areas (Fig.
6.1): (1) Admiralty Islands (Lou, Pam, and Manus); (2) New Britain (Willaumez
Peninsula, often referred in previous studies as Talasea, and Mopir); (3) D'Entre-
casteaux group (East and West Fergusson, Dobu, Sanaroa); (4) Banks Islands
(Vanua Lava and Gaua). Archaeological research has shown that obsidian was
transported considerable distances as early as 20,000 years ago (e.g. Allen et al.
1989; Summerhayes and Allen 1993) and has continued to be traded or other-
wise distributed away from the sources up to the present day. Obsidian artifacts
traced to Melanesian sources have been found in archaeological contexts spread
across the vast region spanning from Sabah in Malaysia (Bellwood and Koon
1989) to Fiji (Best 1987) and in both highland and lowland zones of Melanesia
(White 1996 offers a chronological summary of obsidian distributions).
Since the early 1960s, a wide battery of techniques has been used to
analyze the chemical composition of the various obsidian outcrops and to

Admirally 0 500 1000

Islands .____...__._~~ ...... km
f-·· <>~ew Ireland
~:\) •Nissan
New Britain ~~ SOLOMON
• ..,_ISLANDS
3.\ A "'~ Reef Islands
O'EntrecasteauP ..,.
Group ··santa Cruz Islands

4.: Banks I lands

tt"ivANUATU

NEW~ ~
...
•
.
AUSTRALIA CALEDON~.•

Figure 6.1. Melanesian obsidian sources and archaeological sites mentioned in the text.
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 131

assign artifacts to their original sources (e.g., Reeves and Ward 1976). By far
the greatest contribution to the archaeological analysis of obsidian distributions
has been made by the application of Proton Induced X-ray Emission-Proton
Induced Gamma-ray Emission (PIXE-PIGME), a highly sophisticated but
non-destructive and relatively quick technique. In this paper we describe how
recent advances in source sampling and in the PIXE-PIGME method have made
important contributions to research on obsidian use and exchange in the West
New Britain Province of Papua New Guinea. Much finer discrimination be-
tween subgroups within this geographical region is now possible. The assign-
ment of 1200 artifacts from archaeological sites in Papua New Guinea and the
Solomon Islands to their geographical region and in most cases to their
subgroup has led to important insights into the role of natural and social factors
in the changing patterns of prehistoric interaction in West New Britain. The
advantages of PIXE-PIGME as a fast, accurate sourcing technique argue for its
wider adoption outside the Pacific region.
In this paper we use the term geographic region in the same sense as Hughes
( 1986: 289) to mean a geographically defined unit of space within which natural
deposits of obsidian can be found. Within any geographic region there may be a
number of sources. A source is any natural context in which obsidian occurs.
Both primary contexts such as flows and secondary deposits such as stream beds
or beaches are considered to be potential sources of raw material. We adopt
Hughes (1986: 290) definition of source localities as "specific sampling loci
where obsidian specimens were collected." All the source localities found within
a geographic region are therefore considered to consist of a regional group.
Finally, source localities occurring within a single geographical region can be
combined into subgroups (Hughes' [1986: 290] "subgroupings" or "chemical
groups") on the basis of similarities in their chemical composition. Within a
single geographic region there may be multiple subgroups of obsidian sources.
Whereas previous characterization studies in Melanesia have aimed at assigning
artifacts to their regional group, our recent analyses of the Willaumez Peninsula
and Mopir geographic regions have shown that important results are obtained
when the focus of analysis is shifted to the level of subgroups.

PREVIOUS OBSIDIAN CHARACTERIZATION IN MELANESIA

The initial characterization of Melanesian obsidians employed emission

spectroscopy on samples from Lou Island, Willaumez Peninsula, East Fergus-
son Island and volcanic glass from Rabaul (Key 1968; 1969). At the same time,
artifacts from Watom Island were also shown to have a similar composition to
samples from source localities in the Willaumez Peninsula area, whereas flakes
from Collingwood Bay and the Trobriand Islands were found to resemble
samples from Fergusson Island. In a later study using the same technique,
132 G. R. SVMMERHAYES et al.

Ambrose and Green (1972) attributed obsidian artifacts from Ambitle Island
(off the east coast of New Ireland) and Gaua (Reef-Santa Cruz Group in the
southern Solomon Is.) to the Willaumez Peninsula regional group (but see
Ambrose 1973). These results were important because the obsidian artifacts
were found in association with pottery bearing Lapita-style decoration which
has been dated to the first millennium BC and has a very large spatial
distribution (Green 1979; Kirch and Hunt 1988).
X-ray fluorescence was used by Smith (1974; Smith et a!. 1977) to
separate outcrops on Fergusson Island and by Lowder and Carmichael (1970:
28) on one sample from the Willaumez Peninsula group. Smith eta!. (1977)
noted the limitations of sample size and sample preparation for this and other
methods in use at that time. In contrast to the early studies, which were
undertaken by geologists, the archaeologists were concerned with considering
the utility of different outcrops for human consumption (Specht 1981; Ambrose
1976). Consequently, their fieldwork focused on places where obsidian had
been procured by the local inhabitants in the recent past. Their accounts also
include information about how the obsidian was procured and traded. The
samples collected by Ambrose in association with Specht on the Willaumez
Peninsula were initially analyzed by Wall (1976) using neutron activation.
Source samples from Lou Island, Willaumez Peninsula, West and East Fergus-
son and the Banks Islands, as well as artifacts from Malo (Vanuatu), Ile des Pins
(New Caledonia), Ambitle, and the Moem Peninsula and Mt. Hagen on the
mainland of Papua New Guinea were included in this analysis (Ambrose 1976).
Unlike many other areas of the world, neutron activation was not widely
adopted for the analysis of Melanesian obsidians because an alternative method
which does not damage the artifact became available. The new analytical
technique reported by Coote eta!. (1972), Proton Induced Gamma-ray Emis-
sion (PIGME), proved useful for rapid non-destructive determination of three
elements (fluorine, sodium, and aluminum) in obsidian (Bird and Russell1976;
Ambrose et al. 1982). The results gave a good separation between the Wil-
laumez Peninsula, Admiralty, Fergusson and other regional groups, although
there was some overlap between the Willaumez Peninsula and the Banks
Islands. PIG ME was subsequently used to compare samples from all the known
Melanesian source localities and 700 archaeological artifacts (Bird et al. 1981a;
1981b). It was found that outcrops from eight areas on the Willaumez Peninsula
(labeled as Pilu, Voganakai, Dire, Kumeraki, Mt. Bao, Bitokara, Talasea Hospi-
tal, Talasea Administration) could be separated from the island sources on
Garua and Garala (Schaumann) Islands and that all the Willaumez Peninsula
source localities could be differentiated from artifacts found at Waisisi on the
Hoskins Peninsula. On this basis the existence of a second regional group
within New Britain was suspected. Subsequent fieldwork discovered the prob-
able source as within the geographic region of Mopir (Specht and Hollis 1982;
Fullagar eta!. 1991).
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 133

Artifacts from a wide range of sites in Melanesia were allocated to the

Admiralty, New Britain, Fergusson or Banks Islands regional groups using a
plot of two ratios (FIN a and AliNa). A notable exception was a group of artifacts
from sites along the coast of Papua that was clearly different from any of the
known sources but is nearest in composition to obsidian from West Fergusson.
The source for this material has yet to be discovered. Geographically contigu-
ous specimens were combined into 11 chemically distinguishable subgroups.
A Mahalanobis discriminant function analysis of the subgroups based on
concentrations of three elements was made and 97 per cent of the source
samples were reallocated to their correct subgroups (Bird et al. 1981a: 24).
Although PIGME was successful at discriminating among the four major
geographical regions in Melanesia, a more powerful technique was needed to
satisfactorily separate the source localities into subgroups. As Ambrose and
Duerden (1982: 85) noted, the three elements produced by PIGME could not
carry the burden of such close partitioning of subsources as the archaeologists
desired. Consequently, Proton Induced X-ray Emission (PIXE) which allows
simultaneous measurements of gamma-rays and X-rays (Duerden et al. 1979,
1980; Bird et al. 1983) was adopted. PIXE allows up to 20 or more elements to
be measured, and 12 (Si, K, Ca, Ti, Mn, Fe, Zn, Rb, Sr, Y, Zr, Nb) were found to
occur at useful concentrations in Melanesian obsidians (Bird et al. 1988; Duer-
den et al. 1986, 1987). Multivariate analyses using ratios of the elements from
the combined PIXE-PIGME results proved most useful in distinguishing sub-
groups (Bird et al. 1988). Provided the element concentrations are independent
of each other, the ratios are useful because they help to normalize variations in
measurement efficiencies between separate runs (cf. Torrence et al. 1992: 93).
From 1986 to 1988, some 1200 artifacts collected during the Lapita
Homeland Project of 1985 (Allen and Gosden 1991) were analyzed together
with repeat measurements on many source samples. The resulting databank
provided improved definition of the major Melanesian obsidian sources and an
indication that some source localities could be placed into subgroups, as well
as allocating most archaeological samples to their source region. Furthermore,
the use of differences in the relative density of obsidian from sources in the two
major geographic regions of New Britain and of the Admiralty Islands provided
an important first-order sampling technique in advance of PIXE-PIGME analy-
sis (Ambrose 1976, n.d.; Green 1987; Kirch et al. 1991:156-7). When our own
work began in 1988, then, significant progress had already been made.

DEMAND FOR FINER DISCRIMINATION OF SOURCES

Despite the success of the characterization program using PIXE-PIGME,

improvements were required in order to answer new archaeological questions
which arose from the previous results. For example, Green (1987) noted
134 G. R. SVMMERHAYES et al.

changes through time at the site of Watom in the quantities of obsidian from
the Willaumez Peninsula and Mopir, as well possible differences in the repre-
sentation of material from different localities within the Willaumez Peninsula
region. His analysis also pointed out the need for better information about the
occurrence of Mopir obsidian, but since so few source samples had been used
in the initial characterization studies, the ability to recognize material from
Mopir was on insecure ground. In addition, renewed fieldwork at Bitokara
Mission on the Willaumez Peninsula by Specht et al. (1988) proposed several
hypotheses concerning differential use of the mainland and island sources in
this region. For instance, they noted that sites with Lapita-style pottery tended
to be located near the obsidian sources on Garala and Garua Islands, but pottery
was absent at the mainland outcrops. Since waste by-products dating to a period
before the pottery were abundant at one mainland source, it seemed likely that
there may have been a major shift in the use of sources or even in control over
access to raw material in the later period. Our recent work has focused on
sources and sites in West New Britain as part of a larger program of research in
the province (e.g. Specht et al. 1991; Gosden 1989, 1991). It was important,
therefore, to find out which Willaumez Peninsula sources supplied sites outside
the area in different periods.
To satisfy the changing needs of archaeological research, a finer discrimi-
nation between the New Britain sources was sought. Two advances in method
were made to satisfy this requirement. Firstly, a much larger sample of obsidian
from more source localities was needed so that the nature and degree of
variability within and between individual occurrences could be described. This
was especially true for the Mopir sources, since no one had systematically
sampled in situ outcrops of obsidian because they are located in an uninhabited
and remote area. Secondly, although progress was made as a result of an increase
in source sample size (e.g., Fullagar et al. 1989; Torrence et al. 1992), several
changes were needed to increase the precision of the PIXE-PIGME analyses.
An important enhancement was the adaptation of measurement conditions to
suit the specific needs of particular archaeological problems.
We begin with a brief description of the PIXE-PIGME technique because
it is not widely used for obsidian analysis outside Australia. We then summarize
the two improvements for the characterization of Melanesian obsidian using
PIXE-PIGME, and consider some implications of the results of the analyses of
archaeological items.

PIXE-PIGME

Proton Induced X-ray Emission (PIXE) and Proton Induced GaMma-ray

Emission (PIG ME) are extremely valuable for archaeological analyses of obsid-
ian because they can be combined to provide a rapid, non-destructive analysis
of a very wide range of elements. The results are very similar to those from
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 135

X-ray fluorescence analysis, but include additional information on light ele-

ments which are obtained from the gamma-ray spectrum. Most importantly,
the minimal amount of sample preparation allows a much larger number of
specimens to be analyzed in a much shorter time.

Sample Preparation
Samples can be irradiated without any pretreatment, but two factors
which can affect the results must be controlled for: (1) the 2.5 MeV proton
beam only penetrates about 50 J.lm into the sample and most of the gamma-ray
and X-ray emission takes place within 10-20 :m. Surface dirt or weathering,
therefore, will influence the results and (2) low energy X-rays (e.g., 1.5 to 4
keY from Al to Ca) are significantly attenuated as they emerge from the
irradiated surface layer. The degree of attenuation varies with the angle of the
sample surface relative to beam direction.
The obsidian samples are cleaned by washing in freon or ethanol and
attached to the special mounting plates so that a relatively smooth and flat
portion of each sample faces the proton beam. The mounting plates are 14 x
40 mm in size and a raised portion at the center (with a 5 mm diameter hole)
has been designed to help in mounting irregular-shaped samples (Fig. 6.2a).
As the mounts are chromium plated, the results for chromium are not used in
the analysis in case any protons strike the plates.
When artifacts cannot be mounted so that a fresh, flat surface completely
covers the hole, either because they are too small or are very irregular in shape,
an alternative sample must be chosen. Artifacts with weathered surfaces can be
assigned to geographical region, but not to subgroup. If information about
subgroups is required, the weathered artifacts should be polished or broken to
expose a fresh surface.

Irradiation
PIXE-PIGME has been carried out for a number of years at the Australian
Nuclear Science and Technology Organisation (ANSTO) at Lucas Heights, New
South Wales (NSW), where obsidian are irradiated by a 2.5 MeV proton beam
from a 3 MeV Van de Graaf accelerator. The beam can be focused and collimated
with aperture diameters between <l mm and 10 mm. In addition, the beam
current can be easily adjusted from a few nanoamps to hundreds of microamps.
The choice of current involves a compromise between achieving sufficient
detector count rates and avoiding overheating of the sample. In the case of
obsidian analysis, an additional factor arises because the samples are electrical
insulators and they become charged during irradiation. This is overcome by
placing a heated carbon filament near the sample to emit electrons which
neutralize the positive charge of the proton beam, thus preventing the emission
of low energy X-rays which would otherwise swamp the X-ray detector.
136 G. R. SVMMERHAYES et al.

Figure 6.2. PIXE-PIGME analysis: a. (upper) mounted obsid_ian artifact. b. (lower) mounted
samples for placement into the vacuum chamber.

Sample irradiation takes place within a 20 em diameter vacuum chamber

(Fig. 6.3). Up to 60 samples, including control standards, can be mounted on
a 1 m long holder (Fig. 6.2b), and each is positioned in turn in front of the
proton beam by a computer-controlled stepping motor. The same computer
also controls data collection and filing so that one complete set of samples can
be analyzed without operator intervention.
An alternative method of irradiation, using the external beam, is used
when samples are too large to fit inside the chamber. In this case, a thin foil is
placed at the end of the accelerator beam line so that the protons pass through
the foil and 5 to 10 em of air before losing all their energy by ionization processes
(similar to XRF analysis) . This external beam method suffers from the fact that
proton energy loss and absorption of low energy X-rays occurs in the air and
reduces the number of elements that can be analyzed and the precision that can
be achieved. Since obsidian is not affected by being placed in a vacuum and
artifacts are usually relatively small, in-vacuum irradiation is preferable.
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 137

Figure 6.3. PIXEIPIGME analysis: Vacuum chamber for irradiation of obsidian samples.
138 G. R. SVMMERHAYES et al.

Detection
For gamma-ray detection, a lithium drifted germanium detector is located
just outside the sample chamber at 135 degrees to the direction of the incident
beam and at a distance of 155 mm from the sample. A typical pulse-height
spectrum from a 10 minute irradiation of obsidian is shown in Fig. 6.4a. In this
case, a number of narrow peaks arise from gamma-ray proton interactions with
fluorine, sodium and aluminum in the sample. A broad region of counts at high
energies is also from fluorine and can also be used for fluorine determination.
Small peaks from lithium, boron and magnesium are present but irradiation
times longer than 10 minutes are needed to determine these with satisfactory
precision.
For X-ray detection, a lithium drifted silicon detector is located inside the
sample chamber at 135 degrees to the beam direction (on the other side of the
beam) and at 90 mm from the sample. A special "pinhole" filter is placed in front
of the X-ray detector to limit the intensity of X-rays from major elements such

4t40 kt'V
No
x5 a. PIOME SPECTRUM 6129 k<V
F

.....
5618 k~tV
2 F

z
::>
0
u

0
0 1500

104 z ~ b. PIXE SPECTRUM

:1: CCl
1.1) '":' .:.::
1-
z 103 a:
u
~
.:.::
:::J ll
0 X ~
u :1: "' Q:

102 O:m ~ N
.,a:
"'" Q:
":>- ID
z
II
10'

10°
0

Figure 6.4. PIXE/PIGME spectra: a. (upper) A ypical

t pulse-height spectrum from a 10 minute
irradiation of obsidian. b. (lower) A typical X-ray spectrum from a PIXE analysis. The peaks
represent the values for different elements.
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 139

as silicon, potassium and calcium to make it easier to detect minor and trace
elements with higher atomic numbers. A typical X-ray spectrum is shown in
Figure 6.4b. Multiple peaks are observed for each element and special fitting
routines are needed to derive X-ray intensities and element concentrations.

Choice of Measurement Conditions

Standard conditions have been used for much of the obsidian work at
ANSTO since 1986: beam energy at 2.5 MeV; beam current at 100 nA; beam
diameter 1.2 mm; beam dose at 50 :C; measuring time of 10 minutes per sample.
Samples from Lou Island (Admiralty source group) and Mayor Island (New
Zealand) have been adopted as reference material to calibrate yield and effi-
ciencies. In this way, routine can be used to mount samples, collect and process
data and carry out cluster analysis with a minimum of effort and training.

Adjustments to Specific Requirements

In 1990, a new program of measurements was initiated with the aim of
trying to distinguish as many as possible of the source localities in New Britain.
The beam diameter was increased to 2 mm with an associated increase in
current to 300 nA, allowing for a trebling of the dose to 150 :C with the same
10 minute measuring time (Summerhayes and Hotchkis 1992; Summerhayes
et al. 1993). By increasing the dose in this way trace element concentrations
can be detected more accurately, to a finer discrimination between samples from
different outcrops.
An alternative approach to distinguish major obsidian sources is to reduce
the run time but increase the beam currents to obtain, for example, a dose of
20 :Cor more in one minute. In this way a much larger number of samples can
be run in a limited period of time. The important point is that measuring
conditions are flexible and can be adapted to make the best use of elemental
analysis to solve archaeological problems while maintaining comparability.

Unassigned Artifacts
Prior to 1990, sourcing studies using PIXE-PIGME commonly revealed
several discrete clusters that were composed solely of artifacts and whose
chemical compositions did not match any of the known subgroups but which
fell closest to the composition of the Willaumez Peninsula and Mopir regional
groups (e.g., Ambrose et al. 1981; Bird et al. 1988). These unassigned subgroups
were labeled TT, UU, and ZZ. One possibility is that these artifacts are of
obsidian from sources which have not yet been sampled. The outcrops may
have been covered over by recent volcanic activity (cf. Torrence et al. 1992). It
also seems likely that modification of the artifacts due to weathering or
140 G. R. SUMMERHAYES et al.

post-depositional effects such as abrasion, leaching or hydration has altered

the surface chemical composition, which is detected by PIXE-PIGME.
The second hypothesis was tested in 1993. Firstly, four artifacts which
had weathered surfaces were broken and then run again with a fresh surface
exposed to the beam and using the new measuring conditions. One was also
run a second time with a weathered surface exposed. All four cases, which had
previously been grouped as UU or listed as "unallocated", were now found to
cluster with the Kutau/Bao subgroup. The weathered surface was found to have
higher than expected amounts of Ca, explaining why the artifacts had pre-
viously formed a separate cluster.
In a second test, 15 artifacts with fresh surfaces which had been assigned
to UU or TT were carefully mounted and rerun. All except one sample were
found to cluster in the Kutau/Bao subgroup. Only one of three run from the ZZ
group could be reassigned to as Kutau/Bao. Further testing of the ZZ subgroup
is, therefore, warranted. All these artifacts are extremely small and require very
careful mounting. It seems likely that in previous runs they had not been
aligned so that a flat surface completely covered the hole. Low energy X-rays
from elements such as Si, Cl, K, and Ca are particularly likely to be affected by
the angle of the sample.
These tests highlight the importance of careful sampling and mounting for
the achievement of accurate results with very fine discrimination. In particular,
samples with weathered surfaces (having a characteristic gray, pitted appear-
ance) should either be polished over an area of 3 mm in diameter or broken and
then mounted with a fresh surface exposed. These limitations, however, are very
minor when compared to the higher costs in time and damage to artifacts of the
sample preparation required by some other characterization methods. Since
these sources of error can be controlled for, there is no sound reason for
preferring another analytical technique for obsidian characterization.

FIELDWORK AT THE WEST NEW BRITAIN SOURCES

At the initiation of our fieldwork in 1988, obsidian was known to occur

in both primary and secondary geological contexts in two areas of New Britain:
(1) near the center of the Willaumez Peninsula Qohnson et al. 1973; Ryburn
1975;johnson 1976; Specht 1981), and (2) at Mopir (Specht and Hollis 1982).
These volcanic features form part of the larger Bismarck arc Qohnson et al.
1973), and are situated over a Benioff zone which dips 70 degrees northward
below New Britain. Because of this zone, there are chemical differences in the
composition oflavas across the region Qohnson 1976: 108;johnson et al. 1973:
529; Lowder and Carmichael1973: 27). For this reason, obsidian outcropping
in different areas should have slightly different chemical makeups, despite their
geographical proximity.
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 141

Specht and a number of co-workers had previously undertaken fieldwork

on the Willaumez Peninsula aimed at identifying and sampling obsidian
sources known by local informants (Ambrose 1976; Specht 1981). Although
attempts to visit the Mopir sources had been made on several occasions, Specht
had only limited success in reaching this remote area and had been unable to
undertake systematic sampling (Specht and Hollis 1982). Nevertheless, these
initial efforts provided enough samples to allow satisfactory discrimination to
be carried out between the Willaumez Peninsula regional group, Mopir, and
the other Melanesian regional groups (e.g. Bird et al. 1981a, 1981b, 1988;
Duerden et al. 1987; Bird in prep).
Our fieldwork in 1988-90 had two aims. Firstly, we sought to locate as
many obsidian exposures as possible and to record a wide range of properties.
An impression of the scale of the sampling program can be gained by comparing
the number ofWillaumez Peninsula source localities studied in previous work
(Table 6.1) with a summary of the results of the recent fieldwork, which allowed
a threefold increase in the number of source localities analyzed and the number
of sample runs (Table 6.2). The nature of each locality was recorded in terms
of a suite of variables related to its potential for sustained low cost exploitation
as well as the gross physical characteristics and flaking properties of the
obsidian. A detailed description of the variables used and a summary of our
findings for the Willaumez Peninsula has been published (Torrence et al. 1992).
Obsidian was found in primary flows and in secondary contexts, such as in
stream beds and on beaches. The quality of the material in terms of size of
blocks, presence of inclusions or flow structures also varied greatly, as did the
scale of the exposures and the ease of extracting useable pieces. Despite
differences between source localities, a large majority of the exposures provided
obsidian that could have been used during most of prehistory In fact, one of
the most important conclusions of our study of obsidian exposures on the
Willaumez Peninsula was that exploitable obsidian was extremely common

Table 6.1. Willaumez Peninsula Obsidian Sources

Analyzed before 1988

Source Location Sample Localities Sample Runs

Gulu 3 II
Bao 4 14
Kutau!Bao 6 20
Hamilton I 8
Baki (Garala Island) 3 4

TOTAL 17 57
142 G. R. SUMMERHAYES et al.

Table 6.2. West New Britain Sources Analyzed 1989-90

(Figs. 6.5, 6.6)

Source Sampling Localities Samples

Locations Localities Analyzed Analyzed

Gulu 9 6 32
Kutau!Bao 15 14 46
Bao 7 7 15
Baki (Garala) 3 4 17
Baki (Garua) 15 9 20
Hamilton (Garua) 12 7 17
Mopir: Ulip 9 6 9
Mopir: Hitalong 3 I 5
Mopir: Yes 4 4 8
Mopir: Dovu I I 2
Mopir: East 4 4 5

TOTAL 81 63 176

throughout the area. Nearly everyone living in this region would have had
access to reasonably adequate sources of obsidian within a short walking
distance of their settlements. The abundance of obsidian, therefore, questioned
the importance of ownership of this resource as a factor in past social and
economic life (Torrence et al. 1992: 91), although access restrictions are known
to have been important in recent times (Specht 1981).
From the fieldwork, a series of predictions could be made about source
selection if minimization of effort were the primary organizing principle for
extraction. Within the Willaumez Peninsula regional group, the source locali-
ties associated with the Bao and Kutau volcanoes are the most extensive and
comprise the thickest flows. The source localities associated with Mt. Baki also
contain many large pieces which are easy to extract, but the overall extent of
the flows is smaller. The source localities assigned to the Gulu volcano can be
ranked as third best since they are also rather restricted in space and many of
the flows contain a high density of phenocrysts, which impede flaking. Finally;
although the flows on Mt. Hamilton are large, thick, and spatially extensive,
the obsidian fractures conchoidally on only rare occasions because the ratio of
phenocrysts to obsidian is very high (Torrence et al. 1992).
The Mopir source localities were found to consist of several thick flows.
Furthermore, a large supply of sizable blocks of easily-flaked obsidian can be
collected from stream beds with very little effort (Fullagar et al. 1991). On the
basis of the field survey, Mopir was assessed as having an extremely high
potential for exploitation.
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 143

The Kutau/Bao outcrops near the coast at Bitokara Mission and the nearby
village of Bamba can be placed on the same level of overall quality as the Mopir
flows, and there is little to choose between them on the basis of exploitability
alone. Differences in the degree to which they were used in the past, therefore,
must be ascribed to other causes (Fullagar et al. 1991; Torrence et al. 1996).
An analysis of changes in the quantity and quality of production is one useful
way to assess how sources were used in the past; preliminary results concerning
the selection of sources have already been obtained through analyses of
obsidian production at two major outcrops (Kutau/Bao and Baki) and at nearby
settlement sites (Specht et al. 1988; Torrence 1992; Fullagar 1992). An alter-
native method is to use patterns of consumption based on the source charac-
terization of assemblages at sites distant from the sources.
The second aim of the fieldwork at the obsidian sources was to obtain
multiple samples from each source locality to allow better fingerprinting of
each New Britain regional group. To control for intra-source variability, samples
were taken at regular intervals (usually 10 or 50 em wide) across flows or
spaced at regular distances in secondary contexts. The location of the recorded
source localities is shown in Figs. 6.5 and 6.6; a summary of the samples sizes
is provided in Table 6.2 (cf. Torrence et al. 1992; Summerhayes et al. 1993).

A Volcanic features associated

with obsidian sources

0 3 km

t 30

21

19

•
Mt Bao
•
MtKut.au
16.1 7
20 g IS
14

Figure 6.5. Willaumez Peninsula region: obsidian outcrops recorded in 1988-9.

144 G. R. SVMMERHAYES et al.

4
.& obsidian flow

N
0

Figure 6.6. Mopir region: recorded obsidian outcrops.

Although our coverage of both geographic regions is quite extensive, there is

room for additional fieldwork on the Willaumez Peninsula, particularly on the
slopes of Mt. Bao where only secondary deposits were observed and sampled.
Nevertheless, the consistency of the chemical composition of obsidian revealed
in the PIXE-PIGME analyses and the relative rarity of artifacts that cannot be
ascribed to a particular subgroup provide a good measure of confidence that
we have adequately sampled the range of chemical variability of the obsidian
in both regions.

IMPROVED DISCRIMINATION BETWEEN OBSIDIAN SOURCES

Using the greatly extended number of source samples obtained from the
Willaumez Peninsula (Torrence et al. 1992), PIXE-PIGME analyses were car-
ried out in 1989 using standard machine conditions (Fullagar et al. 1989;
Torrence et al. 1992). In 1990, the extended suite of samples from Mopir
(Fullagar et al. 1991) was added and from that time enhanced dosages have
been employed (Summerhayes and Hotchkis 1992; Summerhayes et al. 1993).
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 145

Previous discrimination between sources in the Willaumez Peninsula region

had been based on a sample of 17 source localities (Table 6.1). Only one outcrop
from Mt. Hamilton had been sampled and the spatial extent and intra-source
variability of all the outcrops could only be guessed. In the new study, a total
of 176 source sample runs was carried out using samples from 62 of the 81
obsidian source localities which were studied (Table 6.2). Using the higher dose
rates, it is now possible to routinely discriminate five groups of obsidian flows.
The compositional subgroups have been designated by the name of the volcano
or volcanic feature with which they appear to be associated (Figs. 6.5, 6.6, 6.7):
l. Gulu (incorporating Voganakai and Pilu); 2. Kutau/Bao (including sources
previously described a sTalasea); 3. Hamilton (Garua Island); 4. Baki (flows on
both Garua and Garala Islands) and 5. Mopir.
Except in a few cases, the clustering of the chemical composition matched
the actual grouping of outcrops in geographical space. Firstly, all the flows in
the vicinity ofVoganakai as well as obsidian in a pyroclastic flow on the e astern

FRL....,_O ' Garua Island

Talasla FAO
FAQ
FSZ
Fiu

e MOOcm Klllemenls •FRK

9"L_:.;:1~_.=;.2~:...._....:3fJ km

t
•FSS

•FHC
•FGT
FST
FNY .,
FOF. ·..,
FOH 0
FFS
FOJ Arawe Islands
FOL ~~
FO F1..E
FLF
FLQ

Figure 6.7. Site designations for West N ew Britain sites o n which PIXE-PIGME analyses under-
taken .
146 G. R. SUMMERHAYES et al.

side of Mt. Gulu were found to be chemically identical. Secondly, although

sources of obsidian are found in the vicinity of both the Kutau and Bao
volcanoes, the chemical composition of samples from these widely separated
localities overlaps significantly. From an archaeological point of view, this is
disappointing since the potential sources of obsidian are spread over a consid-
erable distance and comprise both inland and coastal locations (Fig. 6.5). In
contrast, the third major result is that the two flows on Garua Island were found
to be chemically distinct. Clearly outcrops on the nearby island of Garala as
well as those on the northeastern side of Garua Island belong to the same
subgroup. The group is named by the most prominent volcanic feature, Mt.
Baki. The very distinctive obsidian which outcrops on Mt. Hamilton belongs
to a different subgroup. Summarizing the Willaumez Peninsula obsidian
sources, one can usefully distinguish between the mainland subgroups of Gulu
and Kutau!Bao and the island subgroups of Baki and Hamilton (Fig. 6.8) .
Finally, the source localities in the Mopir region are very homogeneous
in terms of their chemical composition. The distinction made in the field
between well-spaced flows which were exposed in different watercourses (Fig.
6.6) appears not to be chemically significant. Presumably the obsidian in this
area emanated from a single volcanic event. The Mopir regional group can be
easily distinguished from all of the Willaumez subgroups.

CHANGING PATTERNS OF OBSIDIAN DISTRIBUTION

Given the ability to consistently discriminate between five subgroups

within New Britain, we now examine how and when obsidian was distributed

8 .-------------------------------------~----------~

sl
41
2
KUTAU

-4 MOPIR

-6
BAKI

-S L---------------------------------------------~
-10 -5 0 5 10

Figure 6.8. Correspondence analysis plot of West New Britain obsidian sources.
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 147

outside the regions associated with each subgroup. When the data on distribu-
tion is compared and contrasted with the assessment of the potential quality
of source localities based on the fieldwork (Torrence et al. 1992) and the
archaeological analyses of the history and intensity of use of specific localities
(Specht et al. 1988; Fullagar 1992; Torrence 1992), some very surprising
patterns of behavior are revealed. The results obtained indicate that the history
of obsidian procurement in New Britain during the past 20,000 years has been
shaped by a complex combination of environmental and social factors. Sea level
changes and volcanic activity must have been crucial factors behind the relative
changes in the popularity of the Mopir source in relation to the Willaumez
Peninsula obsidian. In contrast, social factors must also have played an impor-
tant role, particularly in the choice of obsidian sources on the Willaumez
Peninsula (Torrence et al. 1996).
Significant results of the 1990-1993 PIXE-PIGME analyses have been
summarized previously (Summerhayes and Hotchkis 1992; Summerhayes et al.
1993; Summerhayes and Allen 1993; Torrence and Summerhayes 1994); more
detailed results will also be reported by the excavators of the sites involved. A
total of 1200 archaeological samples from sites in Papua New Guinea (mostly
in West New Britain Province) and the Solomon Islands has been analyzed by
PIXE-PIGME during 1990-93. A summary of the sites and sample sizes is given
in Table 6.3. A breakdown of the results by chronological period is given in
Tables 6.4, 6.5, 6.6. In this paper we emphasize the results of the extensive
sampling of sites within West New Britain; the Pleistocene sites from New
Ireland are included to extend the time depth as far as possible (Fig. 6. 7).
When the results are broken down into chronological periods, as shown
in Tables 6.4, 6.5, and 6.6, a number of trends can be detected. The boundaries
between the periods for obsidian usage reflect what are currently thought to be
significant changes in the prehistory of the area, based on dated sites: ( l)
Pleistocene (ca. 20,000-10,000 bp); (2) Early-Middle Holocene (10,000-3,500
bp); (3) contemporary with Lapita-style pottery, and extending up to the final
use of pottery in most of the areas studied (3,500-1,500 bp); (4) recent (1,500
bp-present). The length of time represented by these divisions is somewhat
gross in terms of the prehistories of other areas of the world, but archaeological
research in Melanesia is still very much in an exploratory phase and there are
few well-dated archaeological contexts. Despite this limitation, four significant
changes in the use of obsidian from the identified subgroups have been
detected.
(l) Summerhayes and Allen (1993) report the results of PIXE-PIGME
analysis of nearly 100 samples from two Pleistocene cave sites in New Ireland,
Matenkupkum and Matenbek. Their results show that roughly three quarters
of the obsidian artifacts analyzed are derived from Mopir and the remainder
are nearly all from Kuatu/Bao, with a sprinkling of pieces from Hamilton and
Gulu. No obsidian can be ascribed to the Baki subgroup. A number of important
148 G. R. SVMMERHAYES et al.

inferences can be drawn from these results. Firstly, the presence of obsidian
throughout the archaeological sequence at Matenbek and the range of sources
represented suggest that obsidian was repeatedly introduced into the site,
suggesting that repeated sea crossings were made between the islands of New
Ireland and New Britain even at this early date. Clearly, the early island
colonizers were capable mariners, and Summerhayes and Allen (1993: 147)
have suggested that during the Pleistocene people were highly mobile. Obsid-

Table 6.3. PIXE-PIGME Analyses of Obsidian Artifacts 1990-1993 (Fig. 6.7)

SITE SAMPLE SIZE SITE SAMPLE SIZE

Willaumez Peninsula/Buvussi Kombe

FAO Garua Island 29 FCL Poi Island 3

FAP Garua Island 2 FPA Kautaga Island 5
FAQ Garua Island 20 FPC mainland 2
FSZ Garua Island 25 FPD Talagone Island 2
FEA Boduna Island 28 FPE Kalapiai Island 5
FRL Bitokara Mission 21 FPJ mainland 3
FRI Walindi Plantation 33
FRK Buvusi 3 Sub-total 20

Sub-total 161

Inland

FHC Misisil Cave 40

FGT Yombon 25
FSS Lamogai 13

Sub-total 78

Arawe Islands/south coast Kandrian/ south coast

FNY Pililo lsland-Paligmete 37 FFS Apugi Island 32

FOF Kumbun Island-Lolmo Cave 44 FFT Apugi Island 39
FOH Adwe Island-Makekurl 83 FLB Yimilo 9
FOJ Kumbun lsland-Apalo 80 FLE Awakuo 19
FOL Amulut 27 FLF Alanglongromo 40
FON Winkapiplo 53 FLQ Alanglong 47
FST Murien 13

Sub-total 437 Sub-total 186

Gasmata

Alolo Island 3
FXH Awrin Island 22

Sub-total 25
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 149

Table 6.3. (Continued)

SITE SAMPLE SIZE SITE SAMPLE SIZE

Morobe Province Watom Island

KBQ Sio 14 SAC 9

KU Malai Island 3 SAD 49
KLK Tuam Island 4 Recent 5

Sub-total 21 Sub-total 63

SW Manus (Admiralty Island) 22 Nissan

Sub-total 22 DFV, DBE, DGW 18

New Ireland Sub-total 18

EAB BalofCave 39
EFR Matenkupkum 43
EFS Matenbek 49

Sub-total 131

Reef Santa Cruz Islands

SE-RF-2 26
SE-RF-6 4
SE-SZ-8 8

Sub-total 38

TOTAL 1200

ian might have been obtained from one subgroup when people happened to be
in the area and was then used up gradually as they moved further away (cf.,
Torrence 1992). Down-the-line exchange might have also played a role in
supplying the New Ireland sites.
Irrespective of whether direct access or simple exchange took place, the
higher proportion of obsidian from the Mopir regional group, rather than the
Willaumez Peninsula, could be explained by the relative proximity of Mopir to
the New Ireland sites. In contrast, Torrence et al. (1996) have suggested that
sea-level changes may also have been an important factor, since it seems likely
that the Kutau!Bao sources were landlocked until post-glacial times, whereas
Mopir may have been more accessible by water transport at that time.
(2) The importance of proximity as a factor in the relative proportion of
obsidian from different sources present at sites is again highlighted by the
results from the Early-Middle Holocene samples. As represented in Table 6.4
150 G. R. SUMMERHAYES et al.

(cf. Summerhayes and Hotchkis 1992; Summerhayes et al. 1993), distance from
a source emerges as the primary factor in determining the relative amounts of
Kutau/Bao versus Mopir obsidian within an assemblage. For example, at Misisil
(FHC) and Lolmo (FOF) caves and at Yombon (FGT) there are roughly equal
proportions from each geographic region. In contrast, sites close to a geographi-
cal source region contain only the local obsidian: the sites on Garua Island are
dominated by Baki obsidian; Bitokara Mission (FRL) contains only Kutau/Bao
obsidian; Walindi (FRI) is also dominated by Kutau/Bao, but has one piece from
Baki; and the small sample from Buvusi Hill is composed solely of Mopir
obsidian.
(3) An exception to the rule of distance as determining factor is the period
from ca. 3,500-1,500 (Table 6.5). Within this phase Mopir obsidian is notice-
ably absent, except in the Kandrian region. The most likely explanation for this
absence or scarcity of Mopir obsidian is the occurrence at about 3,500 years

Table 6.4. Obsidian Characterization in West New Britain: Pre-3,500 bp

SITE MOPIR KUTAUIBAO HAMILTON BAKI GULU ADMIRALTY UNKNOWN

Pre-1!!,000 b(!

Inland

FHC 2 0 0 0 0 0

New ireland

EFR 30 7 0 0 0 J
EFS 37 9 2 0 0 0

JS00-10 000 h.(!.

Willaumez Peninsula

FAO 0 I 0 9 0 0 0
FAP 0 0 0 2 0 0 0
FAQ 0 I 0 9 0 0 0
FRI 0 7 0 I 0 0 0
FRL 0 II 0 0 0 0 0

Buvusi

FRK 0 0 0 0 0 0

Arawe Island

FOF 21 0 0 0 0

Kandrian

FLE 0 s 0 0 0 0 I
FLF J 4 0 0 0 0 0
FLQ 0 2 0 0 0 0
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 151

Table 6.5. Obsidian Chracterization in West New Britain: 3,500-1,500 bp

SITE MOPIR KUT AU/BAO HAMILTON BAKI GULU ADMIRALTY UNKNOWN

Willawnez Peninsula

FAO 0 5 0 5 0 0 0
FAQ 0 5 0 0 0 0 0
FEA 0 28 0 0 0 0 0
FRI 0 15 0 0 0 0 0
FRL 0 5 0 0 0 0 0
FSZ 0 25 0 0 0 0 0

Inland

FGT 3 13 0 0 0 0 0
FHC 0 I 0 0 0 0 0

Arawe Island

FNY 0 30 0 0
FOF 0 12 0 0 0 0 0
FOH 0 141 0 0 I I 3
FOJ 72 0 0 0 0 I I
FOL 0 6 0 0 0 0 0

Kandrian

FFS 7 II 0 0 0 0
FFT 3 18 0 0 0 0 6
FLB 3 2 0 0 0 0 0
FLF 5 17 0 0 0 0 0
FLQ 6 19 0 3 0 0

ago of a massive volcanic eruption at Mt. Witori, close to Mopir. The enormous
scale of this eruption must have had a significant impact over large areas of
West New Britain. Its thick tephra deposits have been found across the north
coast (Torrence et al. 1990; Specht et al. 1991) and inland as far as Yombon
(FGT) (Pavlides 1993). Given the huge deposits of pyroclastic flows in the
vicinity of Mopir (Torrence et al. 1990: 463), there would have been few local
survivors and access to the region must have been extremely difficult for some
time after the eruption. It seems likely that access to Mopir by a water route
was closed off. The continued presence of Mopir obsidian only in the Kandrian
area suggests that the south coast people acquired their supplies overland. At
some point within the period 3,500-1,500 bp the Mopir sources became
sufficiently accessible for Mopir obsidian to be extracted and transported to
Watom Island at the eastern end of New Britain (Green and Anson 1991).
Although the Willaumez Peninsula also suffered severe environmental
change due to the effect of the eruption, there appears to have been very little
152 G. R. SUMMERHAYES et al.

disruption in the procurement of obsidian from the Kutau/Bao subgroup. It is

also significant that this is the period when obsidian from the Admiralty Island
first appears on archaeological sites outside its local region (e.g., Table 6.5;
Ambrose 1976).
(4) During the most recent period, the pattern established during the
Early-Middle Holocene is repeated (Table 6.6). At this time sites on the north
coast contain mainly obsidian from the Kutau/Bao or Baki subgroups. The
Arawe sites, which were probably supplied along a trade route through FSS and
FST, also contain mostly Kutau/Bao obsidian, whereas the sites around Gasmata
have only Mopir obsidian. The Kandrian area has obsidian from both geo-
graphic regions and represents a meeting point between north-south trade
routes running across the island from each source region.
Throughout the long sequence the overall picture is highly colored by
proximity to sources as the major factor in determining access to obsidian.
There is, however, one notable exception that highlights the possible impor-

Table 6.6. Obsidian Charcterization in West New Britain: Post-1,500 bp

SITE MOPIR KUTAU/BAO HAMILTON BAKI GULU ADMIRALTY UNKNOWN

Willawnez Peninsula

FAO 0 0 0 9 0 0 0
FAQ 0 I 0 4 0 0 0
FRI 0 10 0 0 0 0 0
FRL 0 5 0 0 0 0 0

Inland

FHC 2 32 0 0 2 0 0
FGT 3 6 0 0 0 0 0
FSS 0 10 0 0 I 0 0

Arawe Island/south coast

FOF I 2 0 0 I 0 0
FOH 2 34 0 0 I 0 0
FOJ 0 5 0 0 I 0 0
FON I 45 0 0 5 0 2
FST I 9 0 0 2 0 I

Kandrian

FFS 2 10 0 0 0 0 I
FFT 5 4 0 0 0 0 3
FLB 0 4 0 0 0 0 0
Fl.E 4 6 I 0 0 0 0
FLF 4 7 0 0 0 0 0
FLQ 6 9 0 0 0 0 0
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA I 53

tance of social factors. The unusual pattern of consumption comes from sites
on Garua Island, the location of the Baki obsidian source (Tables 6.4, 6.5, 6.6;
cf., Torrence and Summerhayes 1994). Throughout the history of the island,
only the local source is utilized except in the period following the Witori
eruption (ca. 3,500 bp until about 1,500 years ago). Although the entire region
was covered by the Witori tephra and all obsidian sources were buried more or
less to the same extent, obsidian from the Kutau/Bao subgroup is the main
source used in the period subsequent to the Witori event. Yet one site on Garua
Island, FAO, is located directly on top of one of the Baki flows. Social factors
such as exchange and/or conflict seem the most plausible causes of this unusual
pattern of consumption. The choice of a distant rather than a local source of
raw material is highly suggestive of the operation of an exchange system in
which local communities chose to obtain ordinary commodities from some-
where else in order to maintain social links with other groups.
The history of use of the Baki source also raises questions about the
meaning of the dominance of the Kutau/Bao source in other time periods.
Although fieldwork at the source localities identified the Kutau/Bao subgroup as
the most extensive source of high quality obsidian, the Baki source was also
identified as an excellent source. Therefore, it is surprising that only one piece
ofBaki obsidian, from a pre-3,500 year old level at Walindi Planation (FRI), has
been found outside of Garua Island. Very small quantities from the much inferior
Hamilton and Gulu subgroups have been recovered at several sites, but the Baki
subgroup is strangely absent or extremely rare. These data imply that although
distance to a source region may be a key factor during the majority of prehistory,
the choice of which obsidian was exported outside the region may not have been
based solely on the obvious physical properties of the obsidian itself.

CONCLUSIONS

PIXE-PIGME analysis has a number of important properties which have

made it ideal for archaeological research on the prehistoric use and exchange
of obsidian in Melanesia. First and foremost, it is a non-destructive technique.
Only in cases where the surface has undergone extensive alteration do speci-
mens need to be polished or broken. Furthermore, the analyses can be carried
out in a very short period of time. For example, for each years analyses we have
achieved final results on hundreds of samples in less than two working weeks,
even taking into account the time involved in mounting samples and processing
statistical data. The capacity of the method to detect concentrations of a wide
range of elements has meant that extremely fine discrimination can be
achieved. In our case, the ability to recognize subgroups within the Willaumez
Peninsula regional group generated some intriguing archaeological results
which in turn raise new questions for additional archaeological research. The
154 G. R. SVMMERHAYES et al.

ascription of obsidian artifacts to the geographic region of origin also led to a

number of important results. Among the most significant has been the identi-
fication of obsidian from both geographic regions of New Britain in Pleistocene
contexts in New Ireland. This constitutes one of the earliest cases in the world
for the long-distance maritime movement of a raw material and was an
important precursor to the transport of New Britain obsidian to Sabah in the
west and to Fiji in the southeast around 3,000-2,000 years ago. Given the small
amount of fieldwork so far undertaken in Melanesia, additional important
discoveries will undoubtedly follow.
The research reported in this paper has attempted to answer a set of
questions about the nature of obsidian procurement and distribution that are
different from those of earlier work on Melanesian obsidians. To assess the
hypothesis that people living near obsidian sources attempted to control access
to the raw material, we needed to discriminate between source localities within
the larger Willaumez Peninsula region. In particular, the difference between
mainland and island sources was predicted to be important because of the
differential distribution of sites with lapita-style pottery. PIXE-PIGME has
suited our requirements admirably; it has been highly successful in discrimi-
nating among source localities from local areas within the wider region, and
has allowed rapid analyses of a large suite of archaeological samples. The
analyses undertaken during 1990-93 revealed changes through time in the
distribution of obsidian from different subgroups in which both environmental
and social factors appear to have played crucial roles. It is already clear,
however, that untangling the complex prehistory of causation will require a
more quantitative assessment of the results so far, as well as a much larger suite
of samples. The seemingly large number of samples analyzed to date converts
into very few obsidian pieces per hundred years over the 20,000 years of
obsidian use covered by the study.
Finally, it is important to stress that the enhancements to the use of
PIXE-PIGME would not have been sufficient on their own to produce the
results summarized above. The detailed program of fieldwork at the obsidian
sources provided the larger number of samples from numerous source localities
essential to adequately identify intra- and inter-source variability. Now that the
basic groundwork has been completed, we can move beyond the original
questions asked of the characterization program and start to examine the
complex interplay of social and environmental factors in the development of
exchange systems in this region and elsewhere.

ACKNOWLEDGMENTS

As always, we are grateful to the West New Britain Provincial Government,

Bitokara Mission, Garua Plantation, Walindi Plantation and Diving Resort, Mosa
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 155

Village, and many local people of West New Britain for their permission to carry
out research as well as their tremendous hospitality and assistance during the
course of our fieldwork. We especially acknowledge the considerable contribu-
tions of Wilfred Oltomo and Baiva Ivuyo (National Museum and Art Gallery, Port
Moresby), john Normu and john Namuno (West New Britain Cultural Centre),
Leo Metta (Talasea) and Neville Baker (Sydney) who participated in the field-
work. Our affiliating institutions, the Department of Anthropology and Sociol-
ogy at the University of Papua New Guinea and the National Museum and Art
Gallery provided hospitality and support and, with the Institute of Papua New
Guinea Studies, assisted with permits and visas. Funds for the fieldwork were
provided by the Australian Research Council, the Australia and Pacific Founda-
tion, the Australian Museum, and Sheffield University. Part of the work was
carried out while Torrence and Fullagar held Australian Research Council
Fellowships at the Australian Museum. PIXE-PIGME analyses were assisted by
a grant to Gosden and Summerhayes by the Australian Institute of Nuclear
Science and Engineering (AINSE). At ANSTO we especially thank Michael
Hotchkis and Graeme Bailey. Artwork is by Fiona Roberts.

REFERENCES

Allen, ]. and Gosden, C. eds. 1991 Report of the Lapita Homeland Project. Occasional Paper in
Prehistory 20. Canberra, Department of Prehistory, Research School of Pacific Studies,
Australian National University.
Allen,]., Gosden, C. and White, J.P. 1989 Human Pleistocene adaptations in the tropical island
Pacific: recent evidence from New Ireland, a Greater Australian outlier. Antiquity 63: 54&-561.
Ambrose, WR. 1973 3,000 years of trade. Australian Natural History 16: 370-373.
_ _ 1976 Obsidian and its prehistoric distribution in Melanesia. In Barnard, N., ed., Ancient
Chinese Bronzes and Southeast Asian Metal and other Archaeological Artefacts. Melbourne.
National Gallery of Victoria: 351-378.
Ambrose, WR. n.d. A note on obsidian density measurement for separating Melanesian obsidian
sources. Unpublished manuscript.
Ambrose, W R. and Duerden, P. 1982 PIXE analysis in the distribution and chronology of obsidian
use in the Admiralty Islands. In Ambrose, W and Duerden, P., eds., Archaeometry: an
Australasian Perspective. Canberra, Department of Prehistory, Research School of Pacific
Studies, Australian National University: 83-89.
Ambrose, W.R. and Green, R. 1972 First millennium B.C. transport of obsidian from New Britain
to the Solomon Islands. Nature 237: 31.
Ambrose, W.R., Bird, R. and Duerden, P. 1981 The impermanence of obsidian sources in Melanesia.
In Leach, F. and Davidson,]., eds., Archaeological Studies of Pacific Stone Resources. Oxford,
British Archaeological Reports, International Series 104: 1-9.
Bellwood, P. and Koon, P. 1989 'Lapita colonists leave boats unburned!' The question of Lapita
links with Island Southeast Asia. Antiquity 63: 613-622.
Best, S. 1987 Long distance obsidian travel and possible implications for the settlement of Fiji.
Archaeology in Oceania 22: 31-32.
Bird, R. in prep. Results of obsidian sourcing for Lapita Homeland Project. Canberra, Division of
Archaeology and Natural History, Research School of Pacific and Asian Studies, Australian
National University.
156 G. R. SVMMERHAYES et al.

Bird, R. and Russell, L.H. 1976 Applications of prompt nuclear analysis techniques to the study
of artefacts including Southwest Pacific obsidian. In Barnard, N., ed., Ancient Chinese
Bronzes and Southeast Asian Meta! and other Archaeological Artefacts. Melbourne, National
Gallery of Victoria: 317-336.
Bird, JR., Ambrose, WR., Russell, L.H. and Scott, M.D. 1981a The characterisation of Melanesian
obsidian sources and artefacts using proton induced gamma-ray (PIGME) emission technique.
AAEC/E510. Lucas Heights, Australian Atomic Energy Commission.
Bird, JR., Duerden, P., Ambrose, WR. and Leach, B.E 198lb Pacific obsidian catalogue. In Leach,
E and Davidson, J, eds., Archaeological Studies of Pac;ific Stone Resources. Oxford, British
Archaeological Reports, International Series 104: 31-34.
Bird, JR., Duerden, P. and Wilson, D.]. 1983 Ion Beam Techniques in Archaeology and the Arts.
Nuclear Science Applications 1: 357-516.
Bird, R., Ambrose, W, Shahgholi, N. and Kannemeyer, C. 1988 Melanesian obsidian. In Prescott,
JR., ed., Archaeometry: Australasian Studies. Adelaide, Department of Physical and Mathe-
matical Physics, Adelaide University: 107-114.
Coote, G.E., Whitehead, N.E. and McCallum, G.J 1972 A rapid method of obsidian charac-
terisation by inelastic scattering of protons. journal of Radioanalytical Chemistry 12: 491.
Duerden, P., Cohen, D.O., Clayton, E., Bird, J.R., Ambrose, WR. and Leach, B.E 1979 Elemental
Analysis of Thick Obsidian Samples by Proton Induced X-ray Emission Spectrometry.
Analytical Chemistry 51: 2350-2354.
Duerden, P., Bird, JR., Scott, M.D., Clayton, E., Russell, L.H., and Cohen, D. D. 1980 PIXE-PIGME
studies of artefacts. Nuclear Instruments and Methods 168: 447-452.
Duerden, P., Clayton, E., Bird, JR. and Cohen, D.O. 1986 Recent ion beam analysis studies on
archaeology and art. Nuclear Instruments and Methods in Physics Research Bl4: 50-57.
Duerden, P., Clayton, E., Bird, JR., Ambrose, W and Leach, E 1987 Obsidian composition
catalogue. In Ambrose, Wand Mummery,JM.j., eds., Archaeometry: Further Australasian
Studies. Canberra, Department of Prehistory, Research School of Pacific Studies, Australian
National University: 232-238.
Fullagar, R. 1992 Lithically Lapita: functional analysis of flaked stone assemblages from West New
Britain Province, Papua New Guinea. In Galipaud, JC., ed., Poterie Lapita et peuplement.
Noumea, New Caledonia, ORSTOM: 57-68.
Fullagar, R., Ambrose, WR., Bird, R., Specht,]., Torrence, R. and Baker, N. 1989 Stocktaking the
Rocks: Obsidian sources in West New Britain, Papua New Guinea. Proceedings of the Sixth
Australian Nuclear Techniques of Analysis. Lucas Heights (NSW), Australian Institute of
Nuclear Science and Engineering: 187-189.
Fullagar, R., Summerhayes, G.R., Ivuyo, B. and Specht, J 1991 Obsidian sources at Mopir, West
New Britain. Archaeology in Oceania 26: 110-189.
Gosden, C. 1989 Prehistoric social landscapes of the Arawe Islands, West New Britain Province,
Papua New Guinea. Archaeology in Oceania 24: 45-58.
_ _ 1991 Toward an understanding of the regional archaeological record from the Arawe Islands,
West New Britain, Papua New Guinea. In Allen,]. and Gosden, C., eds., Report of the Lapita
Homeland Project. Canberra, Department of Prehistory, Research School of Pacific Studies,
Australian National University, Occasional Papers in Prehistory 20: 205-216.
Green, R.C. 1979 Lapita. In Jennings, ].D., ed., The Prehistory of Polynesia. Cambridge, Harvard
University Press: 27-60.
_ _ 1987 Obsidian results from the Lapita sites of Reef-Santa Cruz Islands. In Ambrose, W.R.
and Mummery,].M.j., eds., Archaeometry: Further Australasian Studies. Canberra, Depart-
ment of Prehistory, Research School of Pacific Studies, Australian National University:
239-249.
Green, R.C. and Anson, D. 1991. The Reber-Rakival Lapita site on Watom. Implications of the
1985 excavations at the SAC and SDI localities. In Allen,]. and Gosden, C., eds., Report of
OBSIDIAN USE IN WEST NEW BRITAIN, PAPUA NEW GUINEA 157

the La pita Homeland Project. Canberra, Department of Prehistory, Research School of Pacific
Studies, Australian National University, Occasional Papers in Prehistory 20: 170-181.
Hughes, R.E. 1986 Variability in Obsidian Procurement Patterns in Northeastern California and
Southcentral Oregon. Berkeley, University of California Publications in Anthropology 17.
Johnson, R.W. 1976 Late Cainozoic volcanism and plate tectonics at the southern margin of the
Bismarck Sea, Papua New Guinea, in Johnson, R.W., ed., Volcanism in Australasia. Amster-
dam, Elsevier Scientific Publishing Company: 101-116.
Johnson, R.W, Mackenzie, D.E., Smith, I.E. and Taylor, G.A.M. 1973 Distribution and petrology
of late Cenozoic volcanoes in Papua New Guinea. In Coleman, P.]., ed., The Western Pacific:
Island Arcs Marginal Seas Geochemistry. Nedlands, Western Australia, University of Western
Australia Press: 523-534.
Key, C. 1968 Trace element identification of the source of obsidian in an archaeological site in New
Guinea. Nature 219: 360.
Key, C. 1969 The identification of New Guinea obsidian. Archaeology and Physical Anthropology of
Oceania 4: 47-55.
Kirch, P.V and Hunt, T.L., eds. 1988 Archaeology of the Lapita Cultural Complex: a critical review.
Seattle, Thomas Burke Memorial Washington State Museum, Research Report 5.
Kirch, P.V, Hunt, T.L., Weisler, M., Butler, V and Allen, M.S. 1991 Mussau Islands prehistory:
results of the 1985--86 excavations. In Allen,]. and Gosden, C., eds., Report of the Lapita
Homeland Project. Canberra, Department of Prehistory, Research School of Pacific Studies,
Australian National University, Occasional Papers in Prehistory: 144-163.
Lowder, G .G. and Carmichael, I. 1970 The Volcanoes and Caldera of Talasea, New Britain: Geology
and Petrology. Geological Society of America Bulletin 81: 17-38.
Pavlides, C. 1993 New archaeological research at Yombon, West New Britain, Papua New Guinea.
Archaeology in Oceania 28: 55-59.
Reeves, R.D. and Ward, G.K. 1976 Advances in obsidian glass studies: archaeological and geo-
chemical perspectives. In Taylor, R.E., ed., Advances in Obsidian Glass Studies. Princeton,
New Jersey, Noyes Press: 259-287.
Ryburn, R.]. 1975. Talasea-Gasmata, New Britain. 1:250,000 Geological Series- Explanatory Notes.
Canberra, Department of Mineral Resources, Geology and Geophysics, Australian Govern-
ment Publishing Service.
Smith, I.E.M. 1974 Obsidian sources in Papua New Guinea. Archaeology and Physical Anthropology
in Oceania 9: 18-25.
Smith, l.E.M., Ward, G.K. and Ambrose, WR. 1977 Geographic distribution and the characterisation
of volcanic glasses in Oceania. Archaeology and Physical Anthropology in Oceania 12: 173-201.
Specht,]. 1981 Obsidian sources at Talasea, West New Britain, Papua New Guinea. journal of the
Polynesian Society 90: 337-356.
Specht,]. and Hollis,]. 1982 A new obsidian source in West New Britain, Papua New Guinea.
Mankind 13(5): 424--427.
Specht,]., Fullagar, R., Torrence, R. and Baker, N. 1988 Prehistoric obsidian exchange in Melanesia:
A perspective from the Talasea sources. Australian Archaeology 27: 3-16.
Specht,]., Fullagar, R. and Torrence, R. 1991 What was the significance ofLapita pottery at Talasea?
Bulletin of the Indo Pacific Prehistory Association 11: 281-294.
Specht,]. and Fullagar, R. n.d. Preliminary report on archaeological research in West New Britain
Province, Papua New Guinea, July- August, 1988. Unpublished report to the Institute of
Papua New Guinea Studies.
Summerhayes, G.R. and Hotchkis, M. 1992 Recent advances in Melanesian obsidian sourcing:
results of the 1990 and 1991 PIXE-PIGME analyses. In Galipaud, ].C., ed., Poterie Lapita
et peuplement. Noumea, ORSTOM: 127-134.
Summerhayes, G.R. and Allen,]. 1993 The transport of Mopir obsidian to late Pleistocene New
Ireland. Archaeology in Oceania 28: 145-149.
158 G. R. SVMMERHAYES et al.

Summerhayes, G.R., Gosden, C., Fullagar, R., Specht,]., Torrence, R., Bird, R., Shahgholi, N. and
Katsaros, A. 1993 West New Britain Obsidian: Production and Consumption Patterns. In
Fankhauser, B. and Bird, J.R., eds., Archaeometry Studies in Australia 1991. Canberra,
Department of Prehistory, Research School of Pacific Studies, Australian National Univer-
sity: 57-68.
Torrence, R. 1992 What is Lapita about obsidian? A view from the Talasea sources. In Galipaud,
].C., ed., Poterie Lapita et peuplement. Noumea, ORSTOM: 111-126.
Torrence, R. and Summerhayes, G.R. 1994 How homogeneous is the Lapita Cultural Complex?
Questions raised by recent fieldwork on Garua Island, Papua New Guinea. Paper presented
at the 15th Indo-Pacific Prehistory Association Conference, Chiang Mai, Thailand.
Torrence, R., Specht,]. and Fullagar, R. 1990 Pompeiis of the Pacific. Australian Natural History
23: 457-463.
Torrence, R., Specht,]., Fullagar, R. and Bird, R. 1992 From the Pleistocene to Present: Obsidian
Sources in West New Britain, Papua New Guinea. Records of the Australian Museum
Supplement 15: 83-98.
Torrence, R., Specht,]., Fullagar, R. and Summerhayes, G. 1996 Which obsidian is worth it? A
view from the West New Britain sources. In Irwin, G., Davidson,]. and Pawley, A. and
Brown, D., eds., Oceanic Culture History: Essays in Honour of Roger Green. Wellington, New
Zealand journal of Archaeology Special Publication.
Wall, T. 1976 Use of the research MOATA and associated facilities for the source identification of
obsidian artifacts. In Barnard, N., ed., Ancient Chinese Bronzes and Southeast Asian Metal and
other Archaeological Artefacts. Melbourne, National Gallery of Victoria: 337-350.
White, J.P. 1996 Rocks in the head. In Irwin, G., Davidson,]., Pawley, A. and Brown, D., eds.,
Oceanic Culture History: Essays in Honour of Roger Green. Wellington, New Zealand journal
of Archaeology Special Publication.
 Chapter 7

Factors Affecting the

Energy-Dispersive X-Ray
Fluorescence (EDXRF)
Analysis of Archaeological
Obsidian
M. KATHLEEN DAVIS, THOMAS L. jACKSON,
M. STEVEN SHACKLEY, TIMOTHY TEAGUE, AND
JoACHIM H. HAMPEL

ABSTRACT

Non-destructive XRF analysis of obsidian artifacts presents unique challenges

for quantitative trace element characterization. These empirical tests of the
effects of sample size and surface configuration allow estimation of minimum
size requirements for different analyses. The results also show that errors

M. KATHLEEN DAVIS • Northwest Research, Corvallis, Oregon. THOMAS L. JACKSON •

Pacific Legacy, Aptos, California. M. STEVEN SHACKLEY • Phoebe Hearst Museum of Anthro-
pology and Department of Anthropology, University of California, Berkeley, California. TIMOTHY
TEAGUE • Department of Geology and Geophysics, University of California, Berkeley, California.
JOACHIM H. HAMPEl • Department of Geology and Geophysics, University of California,
Berkeley, California.

Archaeological Obsidian Studies, edited by Shackley.

Plenum Press, New York, 1998.
159
160 M. KATHLEEN DA VlS et al.

resulting from surface irregularities are usually insignificant relative to other

errors in the analysis.

INTRODUCTION

Non-destructive X-ray fluorescence (XRF) is now widely accepted by

archaeologists as a tool for identifying the geologic origins of obsidian artifact
raw material, and to a lesser extent, artifacts of basalt and other volcanic rock.
While the advantages of the non-destructive EDXRF approach are obvious for
artifact studies, analysis of unmodified obsidian rather than powdered samples
contributes to an already long list of analytical uncertainties inherent in XRF
analysis. Understanding the magnitude and source of these uncertainties is
crucial to accurate ascription of artifacts to source, particularly when charac-
terizing artifacts from closely related sources. Previous attempts to quantify
analytical uncertainties in non-destructive analysis have focused on errors
related to variation in artifact size and surface morphology (Bouey 1991;
jackson and Hampel 1992; Shackley and Hampel 1992; Davis 1994). These
studies suggest that artifact size is potentially the more important contributor,
although the actual magnitude of errors due to surface morphology remains
poorly understood.
Elaborating on these earlier studies, this paper attempts to quantify errors
associated with artifact size and surface morphology via two different experi-
ments. The first experiment is an analysis of unmodified samples of various
sizes made from a single obsidian core; the second involves multiple analyses
of flaked and powdered samples to measure errors related to artifact surface
variability. All samples are analyzed for the elements Ti, Mn, Fe, Zn, Pb, Rb, Sr,
Y, Zr, and Ba, using analytelscatter and peak ratio techniques to compensate
for errors associated with the analysis of unmodified obsidian. Since both
experiments involve multiple analyses of samples and comparisons to pow-
dered samples, it is also hoped that the data may be useful as a general indicator
of the accuracy and precision of the method.
In EDXRF analysis, samples are irradiated with photons, usually from an
X-ray tube or radioisotopic source. The primary X-ray beam excites electrons
in the sample, causing it to emit secondary X-rays, i.e., to fluoresce. The emitted
X-rays occur at energies specific to elements in the sample, appearing as peaks
over a given energy spectrum (Figs. 7.1 through 7.4). The height or intensity
of a given peak reflects, in part, the concentration of that element in the sample,
and may be converted to units of concentration, usually by comparison to a
regression line based on well-characterized standards (Bertin 1978;] enkins et
a!. 1981).
Traditional destructive XRF analysis of geologic samples requires rigorous
and time-consuming sample preparation. Samples are ground to a fine powder
 7000

Rh Compton Peak
6000

5000

Vl
4000
c::::>
0
0 Zr
3000

2000

1000

0
10 15 20
Energy (keV)

Figure 7.1. X-ray spectrum of a sample o funmodified Glass Mountain obsidian acquired at 30kV
showing K-a peaks for elements Rb, Sr, Y, Zr, and the Compton Scatter peak ofRh .

40 .-----------------------------------------
Fe Ka

30

"iii
_c
Vl"C

~ 20
§ ::::>
0
oo
c.
~

10
FeKJ3

o L_~~--~~~~~~~==~
0 5 10
Energy (keV)

Figure 7.2 . X-ray spectrum of a sample of unmodified Glass Mountain obsidian acquired at 1 2kV
showing peaks for the elements Ti, Mn , and Fe.
 4000
Ba Ka.

3000

Ba K ~
.!1
c:
0
:::::1 2000
()

1000

25 27 29 31 33 35 37 39
Energy (keV)

Figure 7.3. X-ray spectrum of a sample of unmodified Glass Mountain obsidian acquired at 50kV
showing the K-a. and K-~ peaks of Ba. The darkened bremsstrahlung region between 25 and 31
keY is used to correct for sample mass and thickness.

3000

Compton scatter Am
2500

2000

c Ba Ka.
Vl

:::::1 1500
0
()

1000

500

0
0 10 20 30 50 60
Energy (keV)

Figure 7.4. X-ray spectrum of unmodified Glass Mountain obsidian acquired using an 241Am
radioisotope source. Labeled are the K-a peak of Ba and the Compton scatter peak Am.
X-RAY FLUORESCENCE (EDXRF) ANALYSIS OF OBSIDIAN 163

and pressed into briquettes, or for analysis of major elements, prepared as fused
glass disks. This ensures (l) homogeneous sample composition; (2) a flat and
smooth sample surface to ensure optimal incident and emission angles of
primary and analyte X-rays; (3) particles that are uniform in size and small
enough so that analyte X-rays are not reabsorbed by the sample; and (4) sample
a
mass/thickness that is either infinite, or uniform from sample to sample enkins
1981). Infinite thickness is the point at which X-rays from the X-ray tube or
radioisotope source are completely absorbed by the sample. Samples less than
infinitely thick generate lower element intensities due to incomplete absorption
of primary X-rays by the sample, or put another way, through the loss of primary
X-rays through the back of the sample. Once infinite thickness is attained, the
intensity of a measured element (and therefore its concentration) will remain
constant regardless of further increase in sample thickness. To complicate
matters, infinite thickness is specific to any given element and sample matrix at
a given X-ray energy such that infinite thickness increases with the energy of the
incident X-rays. While artifact thickness is usually unimportant in our analysis
forTi, Mn and Fe, where infinite thickness at l2kV is around 20 microns, it is a
major problem in analysis for barium where infinite thickness at 50 kV is over 1
em. Thus, analytical tolerance for sample thickness varies greatly depending
upon the element in question and the excitation energy used.
While it is geiJerally assumed that obsidian, as a glass, satisfies the particle
size and homogeneity requirements, obsidian artifacts are rarely flat, smooth,
or free of contaminants and surface irregularities. Additional, and potentially
more severe, errors may be introduced when artifacts are smaller than the field
of view of the detector and/or less than infinitely thick. Further sources of error
relevant to both non-destructive and destructive techniques include variation
in the sample matrix (particularly for elements of Z=26 or less), background
subtraction and peak overlap, regression fitting error where applicable, and
counting uncertainty.
Archaeometrists have adopted a variety of techniques to correct for those
errors related to non-destructive analysis. One of the most ambitious attempts
thus far is the work of Giaque et al. (1993) at the Lawrence Berkeley National
Laboratory in Berkeley, California. Their technique utilizes the relative inten-
sities of the Ag K-a and K-13 Compton scattered radiations to calculate
mass/thickness and matrix absorption coefficients of obsidian artifacts. The
method, using a custom-made EDXRF system, is applicable to large and
small/thin samples in a variety of shapes and sizes. Other analysts, the authors
of this study included, have opted for simpler techniques such as the use of
analyte/scatter ratios, or in some cases, relative abundances or peak ratios
(Hampel1984; Hughes 1988; Bouey 1990; Shackley 1992).
The analyte/scatter technique, though less accurate than traditional
destructive techniques Qenkins et al. 1981) or the method developed by Giaque
et al. (1993) mentioned above, continues to be employed in obsidian artifact
164 M. KATHLEEN DAVIS et al.

characterization on a large scale. The technique is based on the general

principle that X-rays scattered by a given sample reflect the mass, morphology
and overall composition of that sample. By calculating element intensities as
ratios to a scatter region prior to their conversion to units of concentration, the
resulting concentration values will, in theory, be adjusted to account for the
particular mass and morphology of the sample. Element intensities are typically
calculated against the Compton scatter peak (Figures 7.1 and 7. 4), or a selected
area of the bremsstrahlung region (Figure 7.3), and converted to units of
concentration via comparison to regression lines derived from identical ana-
lyte/scatter ratios of reference standards.
The Compton scatter correction requires samples to be infinitely thick,
which is a limiting factor in some cases (e.g. the mid-Z analysis in this study).
But because it is related to the mass absorption coefficient of a sample (Franzini
1976; Hampel 1984), it may also be used to correct for matrix variation and
absorption effects. This is not true of the bremsstrahlung region, and when this
region is used matrix corrections must be undertaken separately.
Peak area ratios may also be used in lieu of absolute concentrations (e.g.
Fe/Mn and Fe/Ti ratios reported in this study). This technique works best when
the elements in question are close in atomic number, and has the considerable
advantage of eliminating errors imposed by regression fitting of reference
standards in addition to minimizing those errors related to r_he size and surface
of the artifact (Giaque et al. 1993). The drawback of this method is that peak
ratio values depend upon acquisition energy and peak extraction, and may be
only marginally (if at all) comparable between laboratories. For these particular
elements we have chosen the peak ratio technique in order to avoid matrix
effect errors associated with the large range of iron present in obsidians ( Giaque
et al. 1993; Hughes 1988).

METHODS

Size Experiment
To determine the limits of the above scatter and peak ratio techniques to
correct for artifact size, we analyzed 20 samples of varying thickness and
diameter prepared from a single core of obsidian from the Glass Mountain
source in northeastern California. Samples were prepared at the Department of
Geology and Geophysics, University of California at Berkeley, and are also the
subject of a paper by jackson and Hampel (1992). Samples were carefully cut
and ground to specified dimensions using a lapidary saw and abrasive, and are
smooth and polished on both target and opposite sides. The first eight samples
are round with a fixed diameter of 25 mm and thicknesses of 0.03, 0.2, 0.5, 1,
2, 3, 4 and 5 mm. By stacking various combinations of these eight samples,
X-RAY FLUORESCENCE (EDXRF) ANALYSIS OF OBSIDIAN 165

analyses were also conducted for sample thicknesses of 6, 7, 8, 9, 10 and l l

mm, and for the 1.2, 1.5, 1.7, and 2.5 mm thick samples reported in the mid-Z
analysis. The remaining twelve samples are square, measuring 3, 5, 10 or 15
mm on a side with thicknesses of 1, 2 and 3 mm for each.

Surface Variability Experiment

A second experiment, using a different set of obsidian samples, was
devised to measure errors resulting from the irregular surfaces of obsidian
artifacts. These samples, also prepared at the Department of Geology and
Geophysics at Berkeley, are the subject of a paper on surface effects by Shackley
and Hampel (1992). With a fixed diameter of 25 mm, the target side of each
sample has been flaked with a copper flaking tool to approximate the surface
of a typical flaked stone artifact. Ten such samples are analyzed here, five from
the Bodie Hills source located in eastern California, and five from the Napa
Valley source located in the North Coast Ranges of California. Each set of five
samples, designated here BH-1 through BH-5 for Bodie Hills, and NV-1 through
NV-5 for Napa Valley, was prepared from a single core of obsidian. The Bodie
Hills samples range from 5 to 7 mm thick, while the Napa Valley samples range
from 8 to l3 mm thick.
The flaked surface of each sample was analyzed in a random orientation
and then rotated 45, 90, 180, 270, and 315 degrees from the original orientation
for five subsequent analyses. These results were then compared to analyses of
pressed powder samples, one each of Bodie Hills and Napa Valley obsidian,
which were acquired in the same six orientations. The Bodie Hills and Napa
Valley powdered samples, were not prepared from the same core as the flaked
samples, and thus are not strictly comparable. However, our primary concern
here is the comparison of relative and not absolute errors. Bodie Hills and Napa
Valley powdered samples were chosen so that they would have peak counts
comparable to those of the flaked samples. Two additional powdered samples,
U.S. Geological Survey standard RGM-1 and Geological Survey of]apan stand-
ardjR-2, were also included in the analysis and analyzed in the six orientations
described above.

Analytical Conditions
Both sample sets analyzed in this study were subjected to the following
four procedures. Analyses for the elements Ti, Mn, Fe, Zn, Pb, Rb, Sr, Y, Zr, and
Ba were conducted on a Spectrace 5000 EDXRF spectrometer, at BioSystems
Analysis Inc. The Spectrace 5000 is equipped with a Si(Li) detector with a
resolution of 155 eV FWHM for 5.9 keY X-rays (at 1000 counts per second) in
an area 30 mm 2 . Signals from the spectrometer are amplified and filtered by a
time variant pulse processor and sent to a 100 MHz Wilkinson type analog-to-
166 M. KATHLEEN DAVIS et al.

digital converter. The X-ray tube employed is a bremsstrahlung type, with a Rh

target and 5 mil Be window. The tube is driven by a 50 kV, 1 rnA high voltage
power supply, providing a voltage range of 4 to 50 kV.
For analysis of the mid-Z elements (Zn, Pb, Rb, Sr, Y, and Zr) the Rh X-ray
tube is operated at 30 kV, 0.30 rnA (pulsed), with a 0.127 mm Pd filter.
Analytical lines used are: Zn (K-a), Pb (L-a), Rb (K-a), Sr (K-a), Y (K-a) and
Zr (K-a). Samples are scanned for 200 seconds live-time in an air path. Peak
intensities for the above elements are calculated as ratios to the Compton
scatter peak of rhodium, and converted to parts-per-million (ppm) by weight
using linear regressions derived from the analysis of 20 U.S. Geological Survey
(USGS), U.S. National Bureau of Standards (NBS) and Geological Survey of
Japan (GSJ) rock standards. The analyte/Compton scatter ratio is employed to
correct for variation in sample size, surface morphology, and sample matrix.
For analysis of the elements Ti (TiO), Mn (MnO), and Fe (Fe 20/), the
X-ray tube is operated at 12 kV, 0.27 rnA with an 0.127 mm aluminum filter.
Samples are scanned for 200 seconds live-time in a vacuum path. Element
values are reported in two ways: in units of concentration (parts per million
forTi and Mn, and weight percent for total Fe); and as Fe/Mn and Fe/Ti peak
ratios. Peak ratios are calculated as simple ratios of extracted peak intensities
(k-a and k-j3). Concentration values are calculated using linear regressions
derived from the analysis of 13 standards from the USGS, the NBS and the GSJ.
These values are reported to better evaluate the precision of each measurement.
It should be noted that the concentration values are not corrected against a
spectral reference, however, and are corrected for matrix effects only to the
extent that the concentration range of Fe is limited in the chosen standards. As
a result, large errors are evident among the smaller samples.
For analysis ofBa, the X-ray tube is operated at 50 kV, 0.25 rnA with a 0.63
mm Cu filter in the X-ray path. Samples are scanned for 200 seconds live-time
in an air path. Trace element intensities for Ba, extracted from a net fit of the k-a
peak, are calculated as ratios to the Bremsstrahlung region between 25.0 and
30.98 keY (Figure 3). Ppm values are generated using a polynomial regression
derived from the analysis of eight USGS and eight GSJ rock standards.
Samples were analyzed a second time for Ba using a 241 Am radioisotope
source on a Spectrace 440 at the XRF lab in the Department of Geology and
Geophysics at the University of California, Berkeley. The system is equipped
with a Si(Li) detector with a resolution of 142 eY FWHM@ 5.9 keY in an area
approximately 20 mm 2• Peak intensities for Ba were obtained by irradiating
specimens with the radioisotope source for 200 seconds livetime. Peak inten-
sities were calculated as ratios to the Compton scatter peak of Am, and
converted to ppm by weight using linear regressions derived from the analysis
of ten rock standards from the USGS, the NBS, and the GSJ.
With the exception of analysis for Ba on the Spectrace 5000 (see above),
peak intensities for elements in both systems are extracted using the Super ML
X-RAY FLUORESCENCE (EDXRF) ANALYSIS OF OBSIDIAN 167

data analysis routine, where peak intensities for both k-a and k-13 peaks are
extracted and corrected for overlap and background via comparison to stored
reference standards (McCarthy and Schamber 1981; Schamber 1977). Analyti-
cal results for selected reference standards are given in Table 7 .l.

RESUI.TS

Size Experiment
Analytical results and sample dimensions for the Glass Mountain samples
are reported in Table 7.2, and presented in Figs. 7.5 and 7.6. The accuracy of
these results may be evaluated against reported element values for RGM-1, a
USGS rhyolite standard also derived from the Glass Mountain obsidian flow
(Tatlock et al. 1976). Trace element values for RGM-1 reported by Govindaraju
(1989) are given in Table 7.1. In general, agreement between the larger Glass
Mountain specimens (i.e. samples thicker than 3 mm and with a diameter of
25 mm), and RGM-1 values is quite good for most elements, though some
systematic errors are evident. In the mid-Z analysis, element values for those
elements with adequate counting statistics (i.e., Rb, Sr, and Zr) are an average
of 5% higher than the values reported for RGM-l. We suspect that this error is

Table 7.1. Selected Trace Element Concentrations

of Two International Rock Standards"

Zn Pb Rb Sr y Zr Ti Mn Fe Ba Ba
(Am)

RGM-1
(Govindaraju 32 24 149 108 25 219 1601 279 1.86 807 807
1989)
(this study) 44 20 149 107 25 218 1673 288 2 794 785
± 4.9 2.5 2.8 6.3 1.6 5.1 81.3 29.7 0.1 9.2 12.8

NBS-278
(Govindaraju 55 16.4 127.5 63.5 41 295 1469 403 2.04 1140 1140
1989)
(this study) 53 16 128 65 40 284 1407 450 2 nm nm
± 5.1 2.4 2.7 6.3 1.6 5.2 81.3 29.8 0.1 nm nm

"±values represent a combined estimate of counting and fitting error uncertainties. All values are in ppm with
the exception of Fe (Fe 20/) which is in weight precent. RGM-1 is a U.S. Geologia! Survey rhyolite
(obsidian) standard and NBS-278 is a National Bureau of Standards obsidian standard. Values for barium
acquired with the 241 Am radioisotope source are labeled Ba(Am). nm =not measured.
168 M. KATHLEEN DAVIS et al.

Table 7.2. Trace Element Values for Unmodified (i.e., Unpowdered) Glass Mountain
Obsidian of Varying Diameter and Thickness"
Diameter Thh:knen y Fol Fol !Ia
Pb Rb Sr Zr Tt Mn Fe
(mm) (mm) Mn Tl (Am)
Control grouo
25 II 38 20 152 112 29 233 67 35 1576 281 1.87 781 533
25 10 41 21 156 Ill 29 233 69 36 1560 270 1.87 813 538
25 38 25 155 112 28 234 63 34 1620 297 1.87 794 539
25 38 21 !55 115 28 236 67 35 1605 279 1.87 824 553
25 43 23 161 112 25 229 68 34 1630 274 1.88 778 559
25 39 24 154 Ill 27 235 65 36 1545 286 1.86 798 564
25 34 20 158 112 27 230 71 35 1635 268 1.91 767 597
25 35 24 154 114 27 231 75 35 1613 256 1.93 814 640

38 22 156 112 28 233 68 35 1598 276 1.9 7% 565

SD 2.8 1.9 2.6 1.5 1.2 2.5 3.7 0.6 33.4 12.3 0.03 20.2 36.4
cv 7.4 86 1.7 1.4 4.4 1.1 5.5 1.8 2.1 4.5 1.4 2.5 6.4

Fixed diameter
25 3 47 23 158 115 27 232 73 34 1633 256 1.90 793 642
25 25 36 15 154 115 25 235 run run run om run
25 44 22 157 117 30 234 75 34 1624 247 1.87 815 670
25 1.7 45 20 161 116 26 230 run run
25 1.5 55 21 161 119 27 235 nm run nm nm
25 1.2 so 19 172 123 31 238 nm run nm nm run run
25 47 24 184 1311 28 2SO 74 35 1597 252 1.87 784 7311
25 0.5 61 32 207 145 29 270 71 34 1638 260 1.87 755 744
25 02 86 45 251 167 35 285 74 34 1630 254 1.89 784 770
25 0.03 185 60 240 159 29 256 65 30 1416 223 1.43 602 1177

Fixed thickness
25 47 23 !58 115 27 232 73 34 1633 256 1.90 793 642
15 36 20 155 115 27 234 75 36 1536 245 1.84 852 592
10 59 22 151 110 25 231 71 36 994 164 1.14 824 476
5 56 18 135 98 27 205 71 35 275 38 0.18 656 221
45 II 102 88 20 180 55 35 130 18 0.00 418 93

25 44 22 157 117 30 234 75 34 1624 247 1.87 815 670

15 51 19 151 114 31 228 76 36 1604 254 1.93 845 659
10 42 20 161 115 27 239 74 34 987 ISO 1.08 780 417
5 69 15 135 95 22 201 59 28 339 48 0.20 644 158
49 16 108 74 17 162 91 34 139 12 0.01 410 64

25 47 24 184 1311 28 2SO 74 35 1597 252 1.87 784 730

15 37 26 178 128 26 248 70 35 1399 235 1.65 769 477
10 47 17 175 124 27 245 71 34 953 152 1.04 805 360
5 65 26 133 91 21 194 75 29 332 38 0.20 568 Ill
3 41 14 100 76 :Z4 160 97 34 131 10 -4.00 274 34

"All element values are expressed in ppm, with the exception of Fe/Mn and FefTi peak ratios, and Fe which is
expressed in weight percent. Values for barium acquired with the 241 Am radioisotope source are labeled Ba(Am).
Values in the lower half of the table appear in bold type when they differ significantly (two sample t-test with a =
0.05) from the control group above. nm =not measured.

related to the calculation of element values for unmodified obsidian samples

against calibration lines derived from the analysis of powders, due to the lower
element/Compton ratio of powdered samples compared with that of unproc-
essed obsidian samples. Analyses of Ba with the radioisotope source are
exceptionally poor. In this case, only three of the twenty-six measurements
made fall within IO% of the 807 ppm reported by Govindaraju. It appears that
the analyte/Compton ratio is not correcting for sample size, or that some other
source of error is present.
X-RAY FLUORESCENCE (EDXRF) ANALYSIS OF OBSIDIAN 169

To evaluate the precision of the results, and thus determine the point at
which the trace element values are measurably affected by sample size, we
compare element values of the smaller specimens to values for those samples that
are well above estimated size limits. For this purpose we have chosen those
samples with a diameter of 25 mm and a minimum thickness of 4 mm as a control
group. The 25 mm diameter samples completely cover the sample slot so that
sample diameter or skewed placement of a sample should not factor into the
measurements. The minimum thickness of 4 mm in the control group is well
above infinite thickness for all elements except Ba, for which infinite thickness
is over a centimeter at this energy. This is not ideal. However, a test of all element
values in the control group revealed no systematic variation in element values
relative to thickness, except for Ba acquired with the radioisotope source. As
mentioned above, precision and accuracy are poor in this analysis for all samples.
For all other elements, significant deviation from the control group averages in
the smaller samples should indicate the point at which sample size has affected
the measurement. In Table 7.2, element values for the smaller samples appear in
bold type when they differ significantly (two-sample t-test at a= 0.05) from the
control sample above. To better emphasize general trends in the data, selected
element values are depicted as a function of sample thickness in Figures 7.5a-d,
and as a function of sample diameter in Figs. 7.6a--d.
At the physical level, observed element distortions are the result of two
conditions infinite thickness, and the field of view of the detector. When a
sample is too small to cover the area seen by the detector, the relative deficit in
X-rays from the sample results in element intensities that are lower than one
would expect for a given material. This effect is best illustrated in the Ti, Mn
and Fe concentration values plotted in Fig. 7 .6c, where no ratio correction is
employed. In this case element values fall drastically at the lO mm diameter.
As discussed previously, infinite thickness for a given element in an
analysis depends upon sample matrix and the excitation energy used. Ulti-
mately, however, tolerance for thickness and diameter will be determined by
the way in which element values are calculated. In this study, peak ratio values
(i.e. Fe/Mn and Feffi) show the greatest overall success. These values are
statistically indistinguishable from the control group down to a thickness of
0.2 mm for Fe/Ti and to a thickness of 0.03 mm for Fe/Mn, assuming a fixed
diameter of 25 mm. The difference in precision between the two ratios (see also
Figures 7 .Sd and 7 .6d) is likely due to the low concentration of Mn relative to
that of Ti in Glass Mountain obsidian (279 ppm for Mn vs. 1601 ppm forTi,
Govindaraju 1989). Similarly, and at a fixed thickness of 3 mm, values are
indistinguishable down to a diameter of 5 mm for Fe/Mn and 3 mm for Feffi,
though these limits rise to 10 mm for the 1 and 2 mm thick samples.
For analysis of Ba with the X-ray tube, ppm values are indistinguishable
from the control group down to a thickness of 0.2 mm. This is an excellent
result given that infinite thickness for Ba in this analysis is over 1 em. However,
 170 M. KATHLEEN DAVIS et al.

250

e2oo
12.
.!:!:
c
0
!c 150

Gl
u
c
0
(.)
100

50

01
0.03 0.2 0.5 1.2 1.5 1.7 2 2.5 3 4 5 6 7 8 9 10 11
Thickness (mm)
1200

1100

1000

e
12.
.!:!: 900
c
.2
!
c Gl 800
u
c
0
(.)

700

600

500
0.03 0.2 0.5 2 3 4 5 6 7 8 9 10 11
Thickness (mm)
Figure 7 .5. Trace element values for unmodified Glass Mountain obsidian as a function of sample
thickness. All concentrations reported in ppm, with the exception of Fe (Fe 20 3r) which is reported
in weight percent. All samples have a fixed diameter of 25mm. a) mid-Z elements; b) Ba acquired
at SOkV with an X-ray tube, and Ba acquired with a 2• 1Am radioisotope source.
X-RAY FLUORESCENCE (EDXRF) ANALYSIS OF OBSIDIAN 171

1800 3

Tl/ .... ..... --- ---

1500
rl
---~- ----- 2.5

e
1:1. .;e
Fe/
1:1.
';;' 1200 2 E
g
,:e
c
~

!
Ill tl
II.
j:: 900 1.5
Ji c

I
1: 600
.2
!
1:
tl
21
c
0
§
(.)
(.)
Mn
300 ~
0.5

0 0
0.03 0.2 0.5 2 3 4 5 6 7 8 9 10 11
Thickness (mm)

00~----------------------------------------------,

j60
...
~ 50~--------------------------------------------------------l

~~-------------------------------------------l
FeiTI

~~L__L_ _~~---L--~--L--L--~--L-_L--~--~~~

0.03 0.2 0.5 2 3 4 5 6 7 8 9 10 11

Thickness (mm)

Figure 7.5. (Continued). c) Ti, Mn and Fe. Ti and Mn are in ppm (lefty-axis) and Fe is in weight
percent (right y-axis); d) Feffi and Fe/Mn peak ratios.
 172 M. KATHLEEN DAVIS et al.

250

200
Zr
E
1:1.
.sc 150

.
0
i
cCll Rb
u 100
c Sr
0
(J

Zn
50
y
Pb:::
:
0
3 5 10 15 25
Diameter (mm)

1000

900

800
/ ---
--- ------
Ba wf X-ray tube -
E
1:1.
700
/ ______.
.sc 600

.2 / ~
!
/
500
cCll /
7
u 400
c
0 Ba wfAm source
(J
300

200
/
100 ../
0
3 5 10 15 25
Diameter (mm)

Figure 7.6. Trace element values for unmodified Glass Mountain obsidian as a function of sample
diameter. All concentrations reported in ppm, with the exception of Fe (Fep 3T) which is reported
in weight percent. Samples are square with a fixed thickness of 3mm. a) mid-Z elements; b) Ba
acquired at 50kV with an X-ray tube, and Ba acquired with a 241 Am radioisotope source.
X-RAY FLUORESCENCE (EDXRF) ANALYSIS OF OBSIDIAN 173

1800 3

1500 2.5

ec. -.!.
c. l:
~ 1200 2 Cl)
:::!:
'C
c
CIS
!
Ql
u..
i= 900 1.5 c
c 0
0
~ ~
'!:
'!: Ql
u
600
2!c c
0
0 (.)
(.)

300 0.5

0 0
3 5 10 15 25
Diameter (mm)

80

75

70
Fe/M/'
-
65

.. 60 /
/
0
:;
...0::
CIS
55
Ql
ll. 50

45

40

35

30
3
Fe/Ti

5 10 15
-
25
Diameter (mm)
Figure 7.6. (Continued). c) Ti, Mn and Fe. Ti and Mn are in ppm (lefty-axis) and Fe is in weight
percent (right y-axis); d) Fe/Ti and Fe/Mn peak ratios.
174 M. KATHLEEN DAVlS et al.

the same analysis seems particularly sensitive to sample diameter. Values for
two of the 15 mm diameter samples (the 2 and 3 mm thicknesses) are
significantly different from the control group, though the value for the 15 mm
by 1 mm thick sample is not. This points to one of the limitations of this data
set, which is that too few sample diameters are analyzed to fully describe the
relationship between diameter and element concentration and by extension,
effects of diameter and thickness combined. For now we will assume that the
allowable diameter in this analysis is somewhere between 15 and 25 mm.
Analytical results for Ba acquired with the Am source are poor for all
samples. Of 27 measurements, only three (the 0.2, 0.5, and 1 mm thick
samples) are within 100 ppm, or 12% of the 807 ppm reported for RGM-1
(Govindaraju, 1989). Since the analyte/Compton correction usually requires
samples of infinite thickness (Franzini et al., 1976) the large infinite thickness
of barium at this energy may be a factor, or perhaps some other source of error
is present.
Size tolerances for elements in the mid-Z analysis, as defined by compari-
son to the control group, are somewhat diverse. In general the results are more
conclusive, and probably more reliable, for those elements with large peaks so
this discussion will focus on Rb, Sr, and Zr. Zn and Ph are seldom used in
obsidian characterization, and the poor precision of these values reflects the
fact that this analysis is optimized for Rb, Sr, Y and Zr. If required, much better
results may be obtained for these elements by using a lower excitation energy.
For those samples with a fixed diameter of 25 mm, element values are
indistinguishable from the control group down to thicknesses of 2 mm for Rb,
to 2.5 mm for Sr, and to 1.2 mm for Zr. The lower thickness limit for Zr relative
to the other two elements is unexpected. Since the elements are close in atomic
number, they would be expected to behave similarly relative to thickness. More
likely this difference points to inadequate sampling in the control group. In
any case, substantial distortion in all three element values begins at about the
1.2 and 1 mm thicknesses (Fig. 7.5a). Results are more consistent for sample
diameter. At a fixed thickness of 3 mm trace element values for all three
elements are unaffected down to a diameter of 10 mm. For Sr and Zr this
threshold rises for the 2 mm thick samples and at 1 mm thick, values for all
diameters are measurably distorted.

Surface Experiment
Analytical results for the surface experiment are presented in Table 7.3.
By comparing the coefficients of variation ( CV) for six runs of the flaked
samples to six runs of powdered samples, we expected to see a difference in
precision that would reflect the error contributed by the flaked surfaces. What
we see instead is that for most elements, any potential difference in precision
between flaked and powdered specimens is obscured by the much larger errors
X-RAY FLUORESCENCE (EDXRF) ANALYSIS OF OBSIDIAN 175

contributed by peak counting uncertainty. In general, the distribution of counts

in an analyte peak follows the normal distribution, and a one s error is given
by the square root of counts in the peak Qenkins 1981). While other factors
such as peak-to-background ratio and element sensitivity contribute to the
overall precision of a measurement, the effects of counting uncertainty in this
analysis are apparent in Figs. 7. 7 and 7.8 where the CV for the six runs on each

Table 7.3. Trace Element Data for 10 Flaked Obsidian Samples (BH-l through
NV-5) and Two Pressed Powder Samples"
y Fe/ Fe/ Ba
Sample Zn Pb Rb Sr Zr Ti Mn Fe Ba
Ti Mn (Am)

mean 34 38 189 92 15 104 561 412 0.57 36 IS 431 331

min 27 35 186 90 12 101 529 403 0.56 33 IS 422 323
42 40 190 94 17 107 611 424 0.58 38 16 444 341
5.2 2.0 1.3 1.9 1.8 2.0 24.8 6.6 0.01 1.4 03 8 I 58
CV 15.1 5.3 0.7 21 12.2 2.0 4.4 1.6 1.46 3.9 1.7 1.9 1.8

mean 36 37 188 91 14 103 522 393 0.54 37 IS 445 345

min 30 33 183 90 II 99 493 386 0.50 JS IS 433 341
45 41 190 93 15 106 556 400 0.56 39 16 454 350
sd 4.9 2.8 2.8 1.1 1.4 2.7 23.6 5.1 0.01 1.4 OJ 8.6 3.1
13.8 7.4 1.5 1.2 10.4 2.6 4.5 1.3 2.26 37 19 1.9 0.9

mean 37 37 190 93 13 lOS 546 389 0.52 JS IS 447 354

min 28 36 186 91 12 101 538 377 0.51 33 IS 436 345
max 43 42 197 98 14 109 SSI 397 0.54 35 IS 468 365
sd 4.5 2.3 3.6 2.2 1.0 2.2 4.2 8.0 O.Ol 0.6 0.2 10.1 7.3
cv 12.4 6.1 1.9 2.3 7.8 2.1 0.8 2I 1.94 1.8 13 2.3 2.1

32 37 189 95 14 103 511 380 0.52 37 IS 443 352

min 29 33 184 92 12 101 487 373 0.51 36 IS 436 345
max 35 42 195 98 IS lOS 540 386 0.53 39 16 456 356
sd 2.2 3.1 3.6 1.9 0.9 1.3 18 7 4 9 0.01 1.2 0.1 6.5 4.7
6.9 8.3 1.9 2.0 6.4 1.2 3.7 1.3 1.70 3.3 0.9 1.5 1.3

mean 34 38 188 92 14 103 518 376 0.50 JS IS 441 342

min 29 36 179 91 12 97 502 364 0.49 34 IS 422 338
max 37 41 193 95 IS 107 537 394 0.52 37 16 465 356
sd 2.9 1.4 S.l 1.5 1.1 3.0 I 1.4 10.2 0.01 0.9 OS 13.8 6.6
cv 8.6 3.6 2.7 1.6 8.2 29 22 2.7 2.27 2.6 3.1 3.1 1.9
NV-I
mean 63 35 194 47 240 442 141 1.17 89 89 367 256
min 58 33 192 45 236 416 137 us 82 88 354 251
69 39 1% 10 51 243 465 145 1.19 94 93 376 260
sd 4.2 2.0 1.3 1.0 2.1 2.3 15.2 2.4 0.02 3.8 2.0 6.7 3.4
cv 6.6 5.6 0.7 12.5 4.5 1.0 3.5 1.7 1.35 4.3 2.2 1.8 1.3

mean 64 34 195 46 243 451 143 118 87 89 364 265

mm 58 32 193 40 239 433 129 116 83 83 349 258
68 36 197 49 247 469 ISO 119 92 100 372 276
sd 3.5 1.2 1.5 II 2.8 2.8 I 1.6 7.6 O.DI 2.8 59 8.5 62
cv 5.5 3.4 0.8 14.5 6.2 1.2 2.6 5.3 0.85 32 6.6 2.3 23

mean 63 37 193 46 247 453 145 1.17 86 87 369 273

min 56 32 190 43 243 433 131 114 82 84 362 256
max 78 42 198 8 49 251 476 ISO 1.19 90 97 381 287
sd 7.6 3.2 3.0 0.5 2.3 27 14.3 6.6 0.02 2.8 45 6.2 9.8
CV 12.0 8.9 1.5 6.5 s.o II 3.1 4.6 1.53 32 s1 1.7 3.6

63 35 194 45 244 468 148 119 84 87 367 268

min 56 32 190 7 43 241 448 128 1.17 81 80 359 264
max 65 38 198 10 47 250 485 159 1.23 87 100 372 273
sd 3.2 2.1 2.6 0.8 1.2 3.1 12 9 11.0 0.02 2.2 7.9 5.4 2.8
5.0 6.1 1.3 9.5 2.7 1.3 2.8 7.5 1.64 2.6 91 1.5 1.0

(continued)
176 M. KATHLEEN DAVIS et al.

Table 7.3. (Continued)

y Fe/ Fe/ Ba
Sample Zn Pb Rb Sr Zr Ti Mn Fe Ba
Ti Mn (Am)

~
mean 63 35 194 46 239 451 145 l.l9 88 88 365 265
min 56 33 191 45 236 416 133 l.l7 83 80 348 257
max 68 38 199 10 47 242 472 !57 1.26 96 96 380 272
sd 3.4 1.8 2.8 1.4 0.7 1.9 18.6 8.1 0.03 3.9 5.2 ll.O 5.2
CV 5.5 5.1 1.4 18.2 1.6 0.8 4.1 5.6 2.54 4.4 5.9 3.0 2.0
NY Powder
mean 59 34 189 45 240 526 167 1.42 88 89 385 375
min 54 30 184 42 237 496 !55 1.41 83 86 373 332
max 68 38 194 48 243 556 172 1.43 93 96 394 418
sd 4.9 2.7 3.3 l.O 2.3 2.1 20.2 6.5 0.01 3.6 3.8 6.2 39.1
cv 8.3 8.0 1.7 13.2 5.0 0.9 3.8 3.9 0.39 4.1 4.2 1.6 10.5
~
X 32 34 180 89 14 107 638 487 0.72 38 16 463 374
min 26 31 177 87 13 105 598 479 0.72 35 15 446 342
max 43 36 182 92 15 109 688 500 0.73 41 16 483 403
sd 5.2 1.7 1.6 1.5 0.7 1.2 27.9 6.6 0.00 1.9 0.2 11.9 27.2
16.4 4.9 0.9 1.7 5.1 1.2 4.4 1.4 0.41 4.9 1.6 2.6 7.3
RQM·I ~tandard
X 37 22 !51 103 25 221 1665 291 2.01 36 69 803 791
min 31 20 149 101 22 218 1638 285 2.00 35 67 794 739
max 44 25 154 107 26 224 1687 297 2.02 37 70 819 824
sd 3.9 1.7 1.4 1.8 1.2 2.1 15.0 4.2 0.01 0.5 l.l 9.0 26.8
cv 10.5 7.7 1.0 1.7 4.8 0.9 0.9 1.4 0.39 1.3 1.7 l.l 3.4
lli·~ Standard
X 38 22 !50 102 25 219 403 1000 0.80 70 nm nm
min 34 20 145 100 23 218 393 992 0.80 67 nm
max 42 24 !52 103 28 221 415 1007 0.81 72 nm nm
sd 3.0 1.4 2.5 1.3 1.5 1.4 7.4 4.6 0.00 1.5 0.1 nm nm
CV 7.9 6.2 1.7 1.3 5.8 0.6 1.8 0.5 0.28 2.2 0.7 nm mn

"RGM-1 and JR-2 reference standards are also included. Statistics for all samples represent six runs in six different
orientations. All element values are expressed in ppm with the exception of Fe/Mn and Fe/Ti peak ratios and Fe
which is expressed in weight percent. Values for the Ba acquired with the 241 Am radioisotope source are labeled
Ba(Am). nm =not measured.

sample are reported by element. Element concentrations, in addition to average

peak counts for each element, are given in Table 7.4.
In general, and within a given spectral region, error is larger for elements
where concentration and hence peak counts are low. This is reflected in both
the magnitude and range of the CV values for elements in all samples.
Coefficients of variation for Mn in the Napa Valley samples, for instance, range
from 1.8 to 7.8% where Mn concentration is an average of 167 ppm with a peak
area of approximately 2400 counts. In the Bodie Hills samples, where Mn
concentration averages 487 ppm with a peak area of approximately 8200
counts, CV ranges from 1.5 to 2.2%.
For the flaked vs. powdered sample comparison, while lower variability
was expected of powdered samples, coefficients of variation ( CV) for six runs
of the powdered Napa Valley and Bodie Hills samples are consistently at or
below the lowest CV values for the flaked samples only for Fe and Zr. These
elements are distinguished by large peaks and high peak-to-background ratios.
Only for Fe is the CV of the powdered sample actually lower than those of the
flaked samples. Furthermore the difference between the two is quite small a
difference of 1.05 units of CV for the Bodie Hills samples, and 0.46 units of CV
for the Napa Valley samples. (Figs. 7.7 and 7.8, Table 7.3). These data suggest
 20
18
...
32

16
1!1
14
•c 14
~

a
2 12 "'
)(
·;; 10 1:1
... 34
.2 8
>
0 ..
0 a
D
...
..
...
138 38
6

..
I
•"' ...
•
110 107 487 11 413

.. • :.. -~
0.72%
4
88 407
D
Jil ., D m
2
:
!;I

0
!I
y
• I II
I
Zn Pb Rb Sr Zr Ti Mn Fe FefTi Fe/Mn Ba Ba/Am
Element

Figure 7. 1. Relative errors acquired from six runs of Bodie Hills obsidian. Each point represents
the coefficient of variance of a given element in a single sample analyzed in six different
orientations. Squares represent CV for unmodified samples of obsidian with flaked surfaces (BH-1
through BH-5), while the triangle represents th e C Vof a pressed powder sample of Bodie Hills
obsidian. Numbers in bold above each element indicate the average concentration of tha t element.
All values are in ppm with the exception of Fe/Ti and Fe/Mn peak ratios, and Fe (Fe 2 0 3 1 ) which
is in weight percent. Values for Ba acquired with the 241 Am radioisotope source are labeled Ba/Am.

20
e
18

16

...
0

.. 1214
c
59
0
2
)(
34 19
·;; 10
... c
.2
> 8 ... c 187 c
0 45 c
c
6
8 8'"'
0 0
9 ... 0

- ...
5.2e c
4
189 ! c 1.42%
Iii
385 341
0
0
240 § IJ
c 0
0

•
0
2 c
0 0 0

~ i
L.J

~ 8
0
Zn Pb Rb Sr y Zr 11 Mn Fe FefTi Fe/Mn Ba Ba/Am
Element
Figure 7.8. Relative errors acquired from six runs o fNapa Valley obsidian. Each point represents
the coefficient of variance of a given element in a single sample analyzed in six different
orientations. Squares represent CV for unmodified samples of obsidian with flaked surfaces (NV- l
through NV-5) , while the triangle represents the CV of a pressed powder sample of Napa Valley
obsidian. Numbers in bold above each element indicate the average concentration of that elem ent.
All values are in ppm with the exception o f Fe/Ti and Fe/Mn peak ratios, and Fe (Fe20 31 ) which
is in weight percent. Values for Ba acquired with the 241 Am radioisotope source are labeled Ba!Am.
178 M. KATHLEEN DAVIS et al.

Table 7..... Concentration Values and Approximate Peak

Counts for Powdered Napa Valley and Bodie Hills Samples"

Obsidian source

Element Napa Valley Bodie Hills

Concentration Counts Concentration Counts

Zn 59 750 32 350
Pb 34 800 34 900
Rb 189 17,800 180 17,200
Sr 8 700 89 700
y 45 6,600 14 1,700
Zr 240 40,600 107 16,500
Ti 526 2,800 638 3,400
Mn 167 2,400 487 8,200
Fe 1.41 238,000 130,000
Ba 384 32,000 463 36,000
Ba(Am) 341 4,600 407 7,200

"All concentrations are in ppm, with the exception of Fe (Fe20 3T) which is in weight
percent. Values for barium acquired with the 241 Am radioisotope source are labeled
Ba(Am).

that in most cases errors related to the surface topography of obsidian artifacts
are negligible when compared to those contributed by counting uncertainty.

CONCLUSIONS

Two experiments were conducted to measure effects related to artifact

size and surface variability in the non-destructive analysis of obsidian. Based
on the analysis of 20 samples of Glass Mtn. obsidian, the following size limits
are observed. Thickness limits assume a fixed sample diameter of 25 mm, and
diameter limits assume a fixed thickness of 3 mm. For Fe/Mn and Fe!fi peak
ratios element values are unaffected down to a thickness of 0.2 mm and to a
diameter of 5 mm. Similarly, limits for Ba acquired with the X-ray tube are 0.2
mm for thickness and somewhere between 15 and 25 mm for diameter. Results
for Ba acquired with the Am radioisotope source are poor for all samples and
we are unable to explain the source of error at this time. For the mid-Z elements
the observed thickness limit ranges between 1.2 and 2.5 mm, and the diameter
limit is lO mm.
Archaeologists requiring the analysis of small artifacts will want to note
that the size limits reported here reflect the points to which no statistically
X-RAY FLUORESCENCE (EDXRF) ANALYSIS OF OBSIDIAN 179

measurable element distortions are observed. Such a level of precision is

rarely required for accurate source ascription of obsidian artifacts, and element
values for many of the smaller samples in this set reflect only minor error.
Furthermore, relative element proportions often remain intact, even for sig-
nificantly undersized samples (for example 0.5 mm thick by 8 mm in diame-
ter), and may be accurately characterized in some cases depending upon the
context of the analysis. It is up to the archaeologist and the analyst to
determine the analytical precision required to distinguish sources or chemical
groups in a given study area.
Results of the surface experiment, where samples of flaked Napa Valley
and Bodie Hills obsidian were analyzed in an attempt to measure error related
to surface irregularities, suggest that this error is nearly always negligible when
compared to counting uncertainty. The only instance where precision for
powdered samples ( CV over six runs of one sample) differs visibly from those
of flaked samples is for the element Fe, where the peak is large and peak-to-
background ratio is high (Figs. 7. 7 and 7.8, Table 7.3). In this case precision
for the powdered samples improved over those of the flaked samples by an
average of less than 1% CV for both the Bodie Hills and Napa Valley samples.
Regarding the applicability of these results to the analysis of artifacts, all
samples in this study were prepared from freshly broken and clean obsidian,
which is, needless to say, uncommon in the world of obsidian artifacts. Effects
due to soil contamination and weathering are not considered here. Secondly,
the flaked samples used in the surface experiment are relatively flat so that the
convexity often encountered in flaked stone tools is not represented. Finally,
these results are specific to the analytical conditions outlined herein, and may
or may not be comparable to results from other laboratories. For this reason it
is important that analytical methods and conditions be reported completely
with XRF results, so that comparability may be evaluated.

REFERENCES

Bertin, E. 1978 Introduction to X-ray Spectrometric Analysis. New York, Plenum Press.
Bouey, P. 1991 Recognizing the limits of archaeological applications of non-destructive energy-dis-
persive X-ray fluorescence analysis of obsidians. Materials Research Society Proceedings 185:
309-320.
Davis, M. K. 1994 Bremsstrahlung ratio technique applied to the non-destructive energy-dispersive
X-ray fluorescence analysis of obsidian. International Association for Obsidian Studies
Bulletin 11.
Franzini, M., L. L. and Saitta M. 1976 Determination of the X-ray fluorescence mass absorption
coefficient by measurement of the intensity of Ag Ka compton scattered radiation. X-ray
Spectrometry 5: 84-87.
Giaque, R. D., Asaro, F., and Stross, F. H. 1993 High precision non-destructive X-ray fluorescence
method applicable to establishing the provenance of obsidian artifacts. X-ray Spectrometry
22:44-53.
180 M. KATHLEEN DAVIS et al.

jenkins, R., Gould, R.W., and Gedcke, D. 1981 Quantitative X-ray Spectrometry. New York, Marcel
Dekker, Inc.
Govindaraju, K. 1989 1989 Compilation of working values and sample description for 272
geostandards. Geostandards Newsletter l3 (special issue).
Hampel,j.H. 1984 Technical considerations in X-ray fluorescence analysis of obsidian. In Hughes,
R. E. ed., Obsidian Studies in the Great Basin. Berkeley, Contributions of the University of
California Archaeological Research Facility: 21-25.
Hughes, R. E. 1984 Obsidian source studies in the Great Basin: Problems and Prospects. In Hughes,
R. E. ed., Obsidian Studies in the Great Basin. Berkeley: Contributions of the University of
California Archaeological Research Facility: 1-20.
_ _ 1988 The Coso Volcanic field reexamined: implications for obsidian sourcing and hydration
dating research. Geoarchaeology 3: 253-265.
jackson, T.L. and Hampel, ].H. 1992 Size Effects in the Energy-Dispersive X-ray Fluorescence
(EDXRF) Analysis of Archaeological Obsidian Artifacts. Presented at the 28th International
Symposium on Archaeometry, Los Angeles.
McCarthy,].]., and Schamber, EH. 1981 Least-squares fit with digital filter: a status report. In
Heinrich, K.Ej., Newbury, D.E., Myklebust, R.L. and Fiori, E. eds., Energy Dispersive X-ray
Spectrometry. Washington, DC, National Bureau of Standards Special Publication 604:
273-296.
Schamber, EH. 1977 A modification of the linear least-squares fitting method which provides
continuum suppression. In Dzubay, T.G., ed., X-ray Fluorescence Analysis of Environmental
Samples. Ann Arbor, Science Publishers: 241-257.
Shackley, M. S. 1988 Sources of archaeological obsidian in the Southwest: an archaeological,
petrological, and geochemical study. American Antiquity 53: 752-772.
_ _ 1992 The Upper Gila River gravels as an archaeological obsidian source region: implications
for models of exchange and interaction. Geoarchaeology 4: 315-326.
Shackley, M. Steven and Hampel,]. 1992 Surface effects in the energy dispersive X-ray fluorescence
(EDXRF) analysis of archaeological obsidian. Presented at the 28th International Sympo-
sium on Archaeometry, Los Angeles.
Tatlock, D. B., E]. Flanagan, HarryBastron, Sol Berman, and A. L. Sutton,Jr. 1976 Rhyolite, RGM-1,
from Glass Mountain, California. In E ]. Flanagan, ed., Descriptions and Analyses of Eight
New USGS Rock Standards. U.S. Geological Survey Professional Paper 850: 11-14.
 Chapter 8

Laboratory Obsidian
Hydration Rates
Theory, Method, and Application

CHRISTOPHER M. STEVENSON, jAMES J. MAZER, AND

BARRY E. ScHEETz

ABSTRACT

The development of laboratory hydration rates is considered to be the most

promising approach for the chronometric dating of obsidian artifacts. The
technical aspects of accelerated hydration, hydration rim measurement, and
the determination of effective hydration temperature and soil relative humidity
are reviewed. It is proposed that glass hydration is controlled primarily by the
amount of intrinsic water contained within the unhydrated obsidian and that
rates of hydration may be estimated once the concentration level is known. The
ability of the intrinsic water model to produce age determinations compatible
with other chronometric methods is examined with a case example from
Xaltocan, Mexico.

CHRISTOPHER M. STEVENSON • Archaeological Services Consultants. Columbus, Ohio.

JAMES j. MAZER • Chemical Technology Division, Argonne National Laboratory, Argonne,
Illinois. BARRY E. SCHEETZ • Materials Research Laboratory, Pennsylvania State University,
University Park, Pennsylvania.

Archaeological Obsidian Studies, edited by Shackley.

Plenum Press, New York, 1998.
181
182 CHRISTOPHER M. STEVENSON et al.

INTRODUCTION

With the increased use of laboratory methods for hydration rate devel-
opment in the early 1980s, the expectations of the archaeological community
were raised to the point where obsidian hydration dating (OHD) was seen as
having the potential to be a reliable dating method, and not just a technique
for the relative ordering of artifacts. With certain significant revisions to the
methodology, we believe that this optimism was well founded and that the
laboratory approach to hydration rate development holds the most promise for
future developments and successful applications.
The approach of developing rates at elevated temperatures was initially
applied to obsidians by Ambrose (1976), Friedman and Long (1976) and Ericson
(1977) but remained largely unevaluated with archaeological case examples.
The explicit presentation of the mathematics involved in the Arrhenius equation
(Friedman and Long 1976; Ericson 1977; Michels and Tsong 1980; Michels,
Tsong, and Smith 1983) has made it possible to acquire a understanding of the
parameters involved in glass hydration. However, the physical processes de-
scribed in the mathematical equations and the actual implementation of the
model that attempts to replicate the natural environment are difficult problems.
We see the current challenges within the OHD field as threefold: (1) to establish
the calibrations used in the prediction of hydration rates from composition, (2)
to accurately measure ambient field conditions of soil temperature and relative
humidity, and (3) to adequately model the ambient environment in the labora-
tory during the development of induced hydration rates.
Over the last few years the authors have undertaken a systematic evalu-
ation of many of the analytical steps involved in the dating process with the
objective of improving the method. We have proposed an improved method
for identifying and measuring the hydration rim, assessed the potential impact
of improper sample preparation, discussed the limits of optical resolution,
documented the presence of hydration rim dissolution for hydration rims
developed at elevated temperatures in deionized water, examined the effect of
reduced soil relative humidity on hydration rate, and documented the effect of
chemical composition (Stevenson et al. 1987; Scheetz and Stevenson 1988;
Stevenson et al. 1989a; Mazer et al. 1991, Mazer et al. 1992). In this chapter we
summarize these findings, identify problem areas, and discuss an application
of the dating method.

PRINCIPLES OF GLASS-WATER INTERACTION

In order to achieve an understanding of the obsidian hydration dating

method, a model that outlines the process of glass hydration under a certain
range of environmental conditions is necessary. The development of such a
LABORATORY OBSIDIAN HYDRATION RATES 183

model requires that the nature of the starting materials (e.g., glass and water)
be adequately known in order that the interaction between them can be
anticipated on the basis the properties of each substance. Unlike crystalline
materials, in which the arrangement of molecules is ordered, the structure of
glass is amorphous and cannot be characterized in a precise manner. This has
served as an effective deterrent to modeling glass-water interaction on a
quantitative basis, yet, the study of this interaction over the last decade has
resulted in a generalized model.
Pure silica glass is a three dimensional network composed of silica
tetrahedra that share oxygens with other tetrahedra. In natural glasses such as
obsidian, the occurrence of network modifiers such as sodium, calcium,
potassium, and ferrous iron depolmerize the silica network by breaking the
Si-0-Si bonds and create non-bridging oxygens (NBOs) with a negative charge.
These modifiers then attach themselves to the NBOs and reestablish the charge
balance (White 1988).
When a glass is exposed to liquid water two reactions occur. These
include the direct dissolution of the silica-alumina matrix and an ion exchange
reaction in which hydronium ions replace the mobile cations (e.g., sodium).
In static aqueous solutions, Hench et al. (1980) noted that cation exchange and
formation of a leached layer is the initial process that is followed by surface
dissolution of the hydration rim as the alkalinity of the solution increases by
the formation of silicic acid. With extended time the dissolution rate will
decrease as saturation of the media is achieved (White 1986).
Under natural (archaeological) conditions an obsidian fragment is con-
tained within a soil matrix under ambient conditions. In many situations the
predominant form of moisture is in the form of water vapor. Thus, a thin
molecular layer of atmospheric water is present on the surface of the glass. With
this small amount of moisture, the molecular water layer may be quickly
saturated by glass dissolution of the obsidian or from silica materials in the
surrounding soil matrix. Although saturation may be quickly reached the glass
is not in thermodynamic equilibrium with the aqueous layer and a continued
reaction of the water with glass results in the formation of a larger hydration layer
(White 1986). Therefore, despite the presence of two reaction processes, the
diffusion of atmospheric water into the glass is much faster than the surface
dissolution of obsidian (Stevenson and Scheetz 1989). This accounts for the
observed correlation of the thickness of the hydration layer with increasing time.

THE DEVELOPMENT OF HYDRATION RATES AT ELEVATED

TEMPERATURE

The initial strategy of hydration rate development involved the correla-

tion of rind widths with independent chronometric dates provided primarily
184 CHRISTOPHER M. STEVENSON et al.

by radiocarbon or dendrochronological assays (e.g., Ericson 1975, Findlow et

al. 1975). The limits of this method stem largely from the uncertainly of the
association between the obsidian fragment and the absolute date. This is
compounded by the differences in thermal history and soil relative humidity
(Rh) when hydration rims from multiple contexts within a large region consti-
tute the data set (Ericson 1988). The result of these less than certain associa-
tions and hydration histories has generated a significant amount of debate
about the rate equation (Meighan 1983; johnson 1969). From a materials
science perspective however, the ability of field studies to establish a rate of
water diffusion are premature given the lack of control over the starting
conditions of context temperature and relative humidity.
In an innovative approach to the rate development problem, Ambrose
(1976), Friedman and Long (1976) and Ericson (1977) initiated the "induced
hydration" approach. In the pioneering experiments of Friedman flakes of
obsidian were hydrated at one atmosphere in a reaction vessel raised to
temperatures between 95°C and 245°C. Under this experimental design, drops
of water were injected at regular intervals into the vessel in order to maintain
the environment at 100% relative humidity at one atmosphere of pressure
(Friedman: personal communication).
Several years later Michels et al. (1983) presented an alternate method of
rate development. Freshly fractured obsidian flakes were placed in 500 ml of
distilled deionized water and hydrated at temperatures ranging between 150°C
and 250°C for periods of up to 6 days. The resulting hydration rims are used
to calculate the preexponential and the activation energy. Using these con-
stants, the hydration rate at elevated temperature can be extrapolated to
ambient conditions of the archaeological site and an absolute date calculated.
In a reevaluation of this research design, Stevenson and Scheetz (1989), and
Stevenson et al. ( 1989a) added powdered silica to the aqueous media in order
to retard surface dissolution predicted by glass-water interaction studies docu-
mented by glass scientists.
Each of the experimental procedures used to develop laboratory hydra-
tion rates has a significant deficiency. The procedures of Friedman and Long
(1976) were designed so that the environment was continuously replaced with
steam. As a result, it can be anticipated that the continuous replenishment of
the surface layer of molecular water through moisture condensation may have
resulted in the excessive dissolution of glass matrix and has not allowed the
hydration layer to develop in a manner that reflects periodic replacement.
The experimental parameters proposed by Michels et al. (1980; 1983)
were also problematic. Under these conditions surface dissolution appeared to
exceed the development of the hydration rim at elevated temperature as the
replications of the experiment by Stevenson and Scheetz (1989) and Tremain
and Frederickson (1988) indicated. Reaction in a large volume of distilled
deionized water did not result in the saturation of the aqueous media. As a
LABORATORY OBSIDIAN HYDRATION RATES 185

result, obsidian flakes reacted under such conditions did not possess hydration
rims representative of the hydration history since the increasing alkalinity of
the solution resulted in various degrees of surface etching.
In a reconsideration of the experimental conditions suggested by Michels
et al. (1983), Stevenson et al. (l989a) added powdered amorphous silica to the
reaction solution. Preferential attack of the distilled deionized water on the
powdered silica led to rapid solution saturation thereby allowing water diffu-
sion to dominate. Although this approach is more compatible with current
theories concerning glass-water interaction, the results were also unevaluated
in any critical manner. Nevertheless, it is proposed that all previous experimen-
tal parameters for the development of hydration rates be revised and that a new
experimental procedure be adopted; one that more closely models the natural
hydration process under ambient conditions. Such a procedure is available and
entails the hydration of natural glass in a vapor environment maintained at
100% relative humidity.

COMPOSITIONAL DEPENDENCE OF GLASS HYDRATION

For two decades it has been proposed that differences in glass composi-
tion can profoundly effect hydration rate (Aiello 1969; Ericson et al. 1976;
Friedman and Long 1976). The awareness of this situation has made the
documentation of volcanic glass major, minor, and trace element composition
a routine procedure prior to obsidian hydration dating (Nelson 1984). The
operating assumption behind this procedure is that artifacts from the same
geological source, or individual flow within a volcanic field, are composition-
ally homogeneous and hydrate at the same rate when found in a similar
environmental setting.
However, just what compositional factor causes obsidians to hydrate
differently under equivalent conditions of temperature and relative humidity
has never been convincingly demonstrated. Quantitative models of obsidian
compositional dependence have included the Chemical Index (Friedman and
Long 1976), the S-value silicon oxygen ratio (Ericson et al. 1976), and the Zeta
Structural Factor (Ericson et al. 1976).
The empirically determined index of Friedman and Long (1976) with which
one estimates the rate of hydration from the index value at temperature T is:

Chemical Index= (Si0 2 - 45(Ca0 + MgO)- 20(Hp+)) (l)

where the values of each oxide and the total intrinsic water are in weight
percent. A second model proposed by Ericson et al. (1976) was based on earlier
work by Haider and Roberts (1970) who found a dependence of the diffusion
coefficient in simple glasses on the Al 20/Na 20 ratio. The Zeta Structural Factor
186 CHRISTOPHER M. STEVENSON et al.

extended this original study to further consider the effects of calcium and
potassium. It is calculated as follows:

(2)

using the mole percent of each oxide. Ericson et al. (1976) also suggested that
obsidian diffusion coefficients may be controlled by the general structure of
the glass and proposed using the silicon-oxygen bond ratio in the glass as a
measure to check this. The 5-value is calculated as follows:

Silicon (mole%)
5== (3)
Oxygen (mole%)

Over the last few years additional experiments were conducted by the
authors to identify the compositional factor(s) which controlled the hydration
rate. Similar to the study by Friedman and Long (1976), a suite of 13 distinct
obsidians and a tektite with different rates of hydration were analyzed for their
major and minor elements using X-ray fluorescence analysis (Table 8.1). The
intrinsic water content (H 20+) in the form of H 20 and OH- were also deter-
mined using infrared spectroscopic techniques (Newman et al. 1986). The

Table 8.1. Obsidian Source and Tektite Chemical Analyses

Source Si02 Al203 Na,O K,O Fe,O,' MnO MgO CaO TiO, OH' A(~day'") E(Jimol)

Red Hill 74.0 14.47 4.97 4.02 0.72 0.11 0.03 0.54 0.02 0.89 2.31 76300
Coso4·1 74.9 14.20 4.68 4.59 0.78 0.03 0.02 0.53 0.04 0.71 2.10 78500
Coso 1·1 74.2 14.49 5.04 4.36 1.03 0.03 0.03 0.55 0.04 0.22 1.62 83600
Superior 74.3 14.38 5.02 4.47 0.67 0.08 0.00 0.63 0.09 0.28 1.56 81589
Polvadera 74.5 14.31 4.36 4.71 0.57 0.06 0.05 0.49 0.07 0.24 1.48 81324
Obsidian R. 74.6 13.81 4.76 4.6 1.10 0.08 0.03 0.32 0.07 0.24 1.47 81785
Antelope 73.4 13.29 4.76 4.85 2.96 0.14 0.24 0.10 0.20 0.16 1.41 84100

Government 73.3 14.88 5.37 4.57 0.92 0.08 0.76 0.04 0.04 0.18 1.33 83902
Pachuca 74.5 14.31 4.37 4.61 0.57 0.06 0.50 0.50 0.07 0.12 1.23 84900

Mule Creek 73.6 14.34 4.62 4.45 0.97 0.05 0.06 0.65 0.06 0.10 1.18 85942

Gwynn 74.1 14.80 4.78 4.49 0.88 0.06 0.66 0.09 0.12 0.16 1.11 85000

lxtepeque 73.3 13.54 5.15 4.50 1.39 0.61 0.23 !.24 0.20 0.09 0.98 85290
Mayor Is. 72.1 12.53 5.28 4.40 4.59 0.!0 0.37 0.00 0.19 0.08 0.78 86259
lndocbinite 74.4 12.17 1.32 2.61 5.58 0.1! 1.85 1.52 0.76 0.026 0.13 91!00
LABORATORY OBSIDIAN HYDRATION RATES 187

hydration rate and activation energy constant for each obsidian were then
developed under conditions of 100% relative humidity at elevated temperature
and pressure for each glass.
The hydration rate at 160°C was regressed against each of the composi-
tional variables and it was found that the intrinsic water content (H 20+) was
the strongest predictor variable with a correlation coefficient of r=0.96. This
was a significantly better statistical descriptor than the Chemical Index model
(r=0.30), the Zeta Structural Factor (r=0.58), or the S-value model (r=0.58).
It was subsequently realized that OH- constituted the bulk of the
intrinsic water content (Silver et al. 1990). It had been observed that only at
concentration levels of approximately 0.3% and above did H 2 0 begin to
contribute to the total water content. Since many obsidians used in the
analysis contained water values below the 0.3% concentration level, it was
proposed that OH- was the dominant factor which controlled the hydration
process. Regression analyses were then rerun on the data set (Table 8.1) to
establish the calibration curves for the hydration rate at 160oC and the
activation energy (Figure 8.1). The correlation coefficients for these regres-
sions were also in excess of r=0.95.
With the identification of the controlling compositional parameter, the
question was immediately asked: Did the assumption of obsidian flow compo-
sitional uniformity hold with respect to the intrinsic water content? To address
this issue Stevenson et al. (l993a) examined the variation in OH- concentration
within the Coso volcanic field, Inyo County, California. Approximately 140
obsidian samples were selected from seventeen prehistoric quarries located
within primary flows and lag deposits. Infrared spectroscopy was used to
determine the total water values. For the volcanic field as a whole, the study
indicated that the water content varied between 0.31% and 2.34%.
Two implications of this study were immediately clear. First, if volcanic
flows were variable with respect to the controlling parameter of hydration then
it could no longer be assumed that the hydration rate developed from a
geological hand sample would be representative of the range of hydration rates
present within the flow material. Additional analyses at Diffusion Laboratory
have shown that the Coso case example is not unique. OH- values from the
Pachuca, Mexico source have been observed to vary between 0.08% and 0.13%,
a small yet influential range of values in the estimation of hydration rates.
Secondly, it is now apparent that obsidian hydration rates need to be deter-
mined for individual artifacts based upon their composition.
This situation is best demonstrated by a study recently completed by
Ackerly (Ackerly and Giese 1993) on materials excavated from Site NMSU 1565
(Table 8.2). Here, OH- concentrations within six artifacts ranged from 0.16%
to 0.58%. Similarly, the hydration rim widths were also variable and varied
between 1.91 pm and 4.80 pm. If a single hydration rate was to be used, ':he
range of absolute dates would be correspondingly large. However, hydration
188 CHRISTOPHER M. STEVENSON et al.

rates estimated from the composition of the obsidian produced dates which fall
within a 115 year age span (344 to 459 BP). Thus, simply controlling for OW
produces tight groupings of hydration dates. If a single hydration rate had been
applied to the rim measurements, the site would have been interpreted to

0
a ~ ~
.......; ....................; ....................:---······ ........
0

"'...
2
; : ;
..
..
.
-,...
t::!
't>
'•:-···················f·················
.
'E ~
z ..................t................~
< :

•
0 L--L--~-------L------L-----~
-2.0 -1.5 - 1.0 -0.5 0.0
Log OH-

100000

90000 ....t................... .\. .. .

0
E
...'
w
80000 ..................;....................; ...................;.. .
•
•

70000
- 2.0 - 1.5 -1.0 -0.5 0.0
Log OH-

Figure 8.1. (upper) Plot of the hydration rate at l60°C versus log of OW. (lower) Plot of th e
activation energy versus log of OH-.
LABORATORY OBSIDIAN HYDRATION RATES 189

represent a multicomponent occupation over many thousands of years rather

than predominantly a single component, late period site.

MEASUREMENT OF THE HYDRATION RIM

The diffusion of molecular water into the glass results in an increase of

the molar volume of the glass. As a result, the altered surface of the glass
acquires the property of double refraction (birefringence) when viewed in
transmission with polarized light. It is the accurate measurement of this
birefringent rind that is critical to the dating of archaeological contexts and
developing the experimental hydration rate constants.
In the optical measurement of the rind it has been the common practice
to identify ancllor confirm the presence of a birefringent layer under crossed
nichols (where the analyzer and the polarizer are set perpendicular to each
other). Under these low light conditions, establishing the boundaries of the
diffusion front may be problematical, thus the polarizer is often rotated to a
position of less than goo in order to allow the additional light to pass through
the thin section and improve the detail of the image. Using this method
thousands of measurements have been made.
The rotation of the polarizer to a position at less than goo to the analyzer
also serves to reduce the contrast between the hydration rim and the unaltered
glass. In order to improve this situation Stevenson et al. (lg87) have proposed
that the gypsum plate (first-order red plate) be used to confirm the presence
of hydration and to enhance contrast. Optical measurements with the gypsum
plate on the same thin sections read in uncrossed-nichols have resulted in the
definition of slightly wider hydration bands. The authors argue that these
determinations are truer representations of the depth of water since the method

Table 8.2. Obsidian Hydration Dates for Site NMSU 1565"

Lab No Provenience OR(wt%) A(~m'/day) E(J/mol) Rate Rim(~) Date S.D.b

93-67 4NJIE, Surface 0 33 2.95 80738 28.5 206 149BP IS

93-69 3N9E, O~!Ocm 0.26 2.51 81667 215 3.11 449BP 30

93-70 4Nl2E, 0-!0cm 0.16 1.70 83559 liS 199 344 BP 36

93-412 4N30E, 10-20cm 0 58 4 16 78540 53 I 4.80 433 BP 19

93-413 2N27E, 0-\0cm 029 271 81378 24 I 3.22 430 BP 27

93-414 7N26E, 20-JOcm 0 31 2 83 80981 26.5 3 49 459 BP 27

93-415 8N27E, 30-40cm 0.14 1.52 84080 96 191 380 BP 40

• EHT= 24 28"C, Rh=82% at 10 em below ground sutface

"Date tmcertainty calculated using a Q_l Jlffi rim measurement error
190 CHRISTOPHER M. STEVENSON et al.

is based on principles of interferometry and phase angle mismatch rather than

light contrast.
In order to further establish the boundary condition for the obsidian
dating process Scheetz and Stevenson (1988) also examined the justification
for the error factor associated with hydration rim measurement under the
resolution limits associated with many optical systems. It was their conclusion
that the resolution limits of the optical microscope imposed a error of 0.25 pm
when the filar micrometer is used in the measurement process. The image
splitting measurement instrument has the ability to reduce error well below
the 0.25 pm level (Dyson 1960) when the opposite sides of an image are
morphologically identical. In the case of hydration rind measurement the
opposing sides of the diffusion layer are close but not identical. Thus, the
measurement accuracy of this instrument is probably less than 0.25 pm and
approaches a value of 0.1pm. However, the exact quantitative limits of error
have not been established for the operating conditions used in obsidian
hydration dating.
Until recently, there has been little discussion over the comparability
of measurements made by researchers in the field. Beginning about 1984
there was a growing uncertainty in the discipline that promoted the initiation
of several blind tests to compare the results generated by different operators.
An comparison by jackson (1984) between Sonoma State University and
University of California, Davis demonstrated a good degree of correspon-
dence. A more extensive blind test by Green (1986) incorporating 10 indi-
viduals from the discipline reported a high degree of correspondence between
nine of the ten labs. A single laboratory returned readings where 9 of the 10
samples possessed hydration rims that were approximately 50% smaller than
the mean value developed from the 90 measurements made by other labora-
tories. An additional and more recent interlaboratory comparison that in-
cluded six operators documented a good agreement between laboratories
(Stevenson et al. 1989a).
However, measurement differences between laboratories still occur (Ber-
gland et al. 1992) and it is clear that non-optical measurement techniques need
to be developed. Kondo and Matsui (1992) have developed a non-destructive
method of hydration rim measurement using a Fourier-Transform Spectro-
photometer. In principle the approach is simple. A light source in the visible
light range is irradiated on the surface of the artifact and reflects off the surface
of the hydrated layer as well as at the interface between the hydration layer and
the unaltered obsidian. Because of the difference in refractive indices between
the hydration layer and bulk obsidian, the reflectance at the area of the diffusion
front is delayed. The interferometer detects the phase differences between the
reflected lights which can be converted to a depth value. Accuracy of the system
is reported to be less than 0.1 pm and good correspondence between this
method and optical methods have been documented.
LABORATORY OBSIDIAN HYDRATION RATES 191

MEASUREMENT OF EFFECTIVE HYDRATION TEMPERATURE

AND RELATIVE HUMIDITY

A number of researchers have repeatedly documented the fact that the

rate of obsidian hydration is an exponential function of temperature (Friedman
and Long 1976; Ericson 1975; Findlow et al. 1982). However, as Ericson (1988)
has noted, although it has been considered as a critical variable since the
development of the method (Friedman and Smith 1960), it has been practice
to diminish the significance of this critical variable in rate development using
archaeological data (Findlow 1977; Hurtado de Mendoza 1978; Aiello 1969).
With the development of laboratory methods that determine hydration rate
constants from high temperature experiments, and require extrapolation to
ambient conditions, the thermal history of the artifact is now an integral
component of the dating procedure.
The effect of temperature is significant due to the exponential depend-
ence on the diffusion of water in glass. Thus, accurate and precise soil
temperature determinations are required for archaeological contexts. The
complexities of such studies have been demonstrated by Riddings (1991) and
by Leach and Hamel (1984) in their measurement of soil temperatures in New
Mexico and New Zealand. In this latter analysis, significant differences in
effective hydration temperature (EHT) were obtained from monitoring instru-
ments that were separated by less than 100m. These and other studies indicate
that the soil temperature of a locality have the potential to be effected by
elevation, orientation to the sun, ground cover, and soil composition. It has
also been demonstrated that the temperature will decline with increasing depth
below the surface (Redfield 1965; Van Wilks and Derksen 1966; Labs and
Harrington 1982).
Estimation of the EHT for a locality has been routinely approximated
using air temperature data (Lee 1969) and on a lesser scale by the computer
simulation of soil temperature variation (Smith 1977; Stevenson et al. 1989a).
These procedures are first order approximations with definable limitations and
they have always been conceived as such by the persons that use them. Lee's
approach is constrained by the fact that it approximates a temperature for the
atmosphere and cannot account for temperature decreases that occur with
depth. Soil temperature simulations of temperatures at different depths within
the soil profile have attempted to overcome this limitation. However, they are
severely constrained by the difficulties associated with the estimation of the
model parameters (e.g., mean annual soil temperature, temperature amplitude,
soil thermal diffusivity). As a result of the modeling restrictions this latter
approach is in need of repeated evaluation.
The direct measurement of a yearly EHT with the thermal cell appears
to be the most accurate determination of soil temperature. As initially devel-
oped by Ambrose (1976), this instrument consists of a small sealed plastic
192 CHRISTOPHER M. STEVENSON et al.

sphere filled with a desiccant that is immersed in water. The difference in

the activity of water between the exterior (A=l.O) and the interior (A=O.O)
of the cell causes the diffusion of water through the wall to be a function of
temperature. At the end of an exposure period, normally one year, the weight
gain of the cell can be converted to an integrated water diffusion rate and
subsequently an EHT.
In a recent reevaluation of the traditional thermal cell design, Trembour
et al. (1988) have proposed that the dry desiccant used in the cell be replaced
with a saturated saline solution. The authors have observed that dry desiccants,
such as silica gel, change in their activity as water is absorbed. The saturated
solution of NaCl however, has a constant solubility over a wide range of
temperature and maintains a water activity of close to 0. 75 (Young 1967). They
also note that water diffuses through a polycarbonate cell containing a saturated
salt solution at a rate that is 75% slower than the Ambrose cell. As a result, this
new design permits the measurement of a higher EHT. The Ambrose cell has
an upper range of approximately 20°C and in arid or geothermal areas the
instrument may need to be replaced after six months. With a slower transmis-
sion of water the saline cell can measure mean exponential temperatures in
excess of 25°C with a error factor of 0.05°C.
Several ground temperature and soil relative humidity studies have
recently been completed using thermal cells, the results of which demonstrate
the necessity of careful environmental monitoring. A recent study by Wirth
Environmental Services (Cleland 1990) at the Coso geothermal field in the
southern California desert (Table 8.3) shows a temperature variation of 2°C
within 1 meter of the surface at Well Pad 33-7. Temperature cells placed in the
near surface soil regions (<25 em) on sub-tropical Easter Island (Stevenson et
al. 1993b) have demonstrated that even within relatively uniform weather
regions that there are temperature differences between open sites, rubble

Table 8.3. Soil Temperature, Relative Humidity, and Hydration Rate

Determinations for Well Pad 33-7, Coso Volcanic Field,
lnyo County, California

100% RhRate Rh Corrected Rate

Depth (em) %Rh EHT("C) (J.tm2/1000 years) (J.lm2/1 000 years)

5 68 23.57 16.13 11.29

25 90 21.49 12.40 8.69

50 96 21.99 13.22 11.77

100 97 23.53 16.05 14.45

LABORATORY OBSIDIAN HYDRATION RATES 193

structures, and caves that is slightly greater than 2°C. These apparently small
differences in temperature can have a significant impact upon the final age
determination.
Since the development of the dating technique it has been assumed that
sufficient ambient moisture is available in any environment for the continuous
development of the hydration rim as dictated by the chemical composition.
This assumption was supported by a set of experiments conducted by Ambrose
(1976). In this study the weight uptake (degree of hydration) of obsidian
powders were nearly equivalent at relative humidities greater than 50%. It also
been argued by Leach and Hamel (1984) that the %Rh of the soil will be close
to 95% even in arid environments. Thus with these two sets of supporting data,
the effect of a reduced or fluctuating Rh has not been considered a significant
environmental constraint.
Subsequent experimental results obtained from the hydration of manu-
factured glasses at varying Rh are contrary to these results. Bartholomew et al.
(1980) hydrated two high silica glasses under different vapor conditions and
reported that the rate of hydration appeared to be a linear function of Rh. A
similar experiment by Bates et al. (1988) on SRL-131 nuclear waste glass also
documented a near linear dependence. A non-linear relationship appears to
exist for obsidians (Friedman et al. 1994; Mazer et al. 1991) with the greatest
reduction in hydration rate occurring in the 90-100% Rh range (Figure 8.2).
In light of these new developments a careful measurement of the soil Rh is now
required.

15

1~
10

~
=
c:::
5

0 1 2 3 4 5

Time (Days 1fZ)

Figure 8.2. Effect of reduced relative humidity on the rate of hydration (open symbols represent
hydration at 175°C while filled symbols represent hydration at 160°C).
194 CHRISTOPHER M. STEVENSON et al.

Abrajano et al. (1989:30) offer an explanation for the dependence of the

rate of hydration on relative humidity:
The dependence of layer thickness growth on relative humidity can be exam-
ined qualitatively in the context of a water diffusion model. There are three
fundamental mass transfer steps for molecular water incorporation into glasses
[see Figure 8.3]. The first step is simple adsorption of water molecules onto the
glass surface. This initial step results in a layer of water forming on the glass
surface: at a given temperature, the layer thickness depends upon the partial
pressure (fugacity) of water or the relative humidity. The second step involves
absorption of water into the portion of the glass in contact with the absorbed
water. This establishes the concentration gradient of water between the glass
surface and the bulk glass thereby driving the final diffusional transport of
molecular water into the bulk glass. At a given temperature and total pressure,
the equilibrium concentration of water in a particular glass is determined solely
by PH 20. If this equilibrium solubility is reached very quickly in the outermost
portion of the glass, the partial pressure of water in the atmosphere can control
the rate of diffusion into the bulk glass by limiting the equilibrium water
concentration in the outermost portion of the glass (thereby controlling the
concentration gradient).

The soil Rh measurements completed by Wirth Environmental Services

(Cleland 1990) within the Coso geothermal field have demonstrated the
relationship with soil depth for this area (Table 8.3). Soil relative humidities
range from 68% at 5 em to 97% at 100 em. In conjunction with the differences
in soil temperature, the hydration rate for West Sugarloaf obsidian artifacts
(Stevenson and Scheetz 1989) within this soil profile may vary between 8.69
um 2/l000 years and 14.45 um2/1000 years. On Easter Island, the relative
humidity of the soil reaches 100% at 10 em below the surface. On the surface
however, the relative humidity was recorded to range between 82% and 92%
(Stevenson et al. 1992). This case study suggests that for Easter Island and
similar environments that obsidian flakes recovered from depths greater than

Glass

Dissolved Water Adsorbed Water Laver

~------------~'~/
~ : .

5
Bulk Glass < Diffusion S < Adsorption
' :8
< Absorption
Figure 8.3. Proposed model of natural h ydration.
LABORATORY OBSIDIAN HYDRATION RATES 195

10 em are preferred for dating since the Rh is a approximately constant and

does not require large corrections to the hydration rate.
The relative humidity of a soil can be determined by the burial of
additional temperature cells. Rh is defined as the 'ratio of the water vapor
pressure present to the pressure of saturated water at the same temperature'.
The temperature cell, immersed in its water jacket, provides a temperature
of hydration at 100% Rh or the vapor pressure of saturated water. A second
cell, without a water jacket, positioned next to the temperature cell will
absorb water at a rate dependent upon the availability of moisture. Since the
two cells experience the same temperature history the relative humidity may
be determined.

AN ARCHAEOLOGICAL CASE EXAMPLE

The ability of the calibration constants predicted from the OH- content
of the obsidian to accurately date a prehistoric site is evaluated with a data from
the Postclassic site of Xaltocan located on a small island within Lake Xaltocan
in the northern part of the Valley of Mexico. Test excavations by Elizabeth
Brumfiel were conducted at this residentiaVceremonial complex to define
patterns of Postclassic resource exploitation and to refine the ceramic chronol-
ogy for this section of the valley. A stratified random sampling program resulted
in the investigation of nineteen areas (Operations A to P, R toT). Obsidian blade
and flake fragments from ten dated contexts were available for analysis. Each
of these contexts is briefly described and interpreted in order to help under-
stand the cultural and non-cultural formation processes.

Operation D
Operation D was a 4 m 2 test unit located on the basal slope of a large
burial mound (Structure 54). Excavations revealed three depositional episodes:
(l) An overburden of recently redeposited material containing modern, Late
Postclassic and Middle Classic debris (0-90 em), (2) a primary midden deposit
with many large sherds (90-120 em), and (3) a low density midden of
Culhuacan Aztec I ceramics and a variety of calligraphic Aztec II bowls. At the
lowest level, Level 25 (230-240 em), two corncobs were dated (calibrated) to
AD 1229±70 (AD 1168-1282). High densities of Aztec I sherds and 7 sherd
discs suggested that the level may have been a ceramic workshop.

Operation G
Operation G was a test unit located on the edge of a low mound on
the north side of the island. Excavated to a depth of 2.45 m, the stratigraphy
196 CHRISTOPHER M. STEVENSON et al.

was interpreted to represent three depositional events. The first consisted of

loose overburden containing modern trash (0 to 50 em) while the second
event consisted of multiple thin strata of different soil types below 50 em,
along the south wall of the unit. The third depositional event was represented
by the remains of a room with walls of compact sand and clay paralleling
the west and south wall of the unit below 50 em. Two floors of black clay
were also identified. Associated with the latest floor in Level 15 (160 to 170
em, 120 to 130 em BGS) charcoal fragments from a probable hearth produced
a calibrated date of AD 989±80 (AD 890-1026). Aztec I ceramics were found
within this level and below.

Operation H
Operation H was located 12 m to the west of a large temple mound
(Structure 120) and was excavated to a depth of approximately 1.0 m. Below
several shallow layers of modern and prehistoric fill, two stone platforms were
encountered lying side by side and were left intact for the remainder of the
excavation. A homogeneous Aztec II ceramic assemblage was found in the fill
associated with these platforms. An exploratory trench into the fill between the
platforms uncovered a hearth which produced a (calibrated) date of AD 1328,
1333, 1395±70 (AD 1300-1416).

Operation I
Operation I was located on a large artificial platform mound near the
central part of the island. Recent disturbance in the platform 20 m north of
the excavation unit location revealed 11 superimposed plaster floors. The
test unit excavation revealed a destroyed clay wall and hard packed dirt floors
indicative of a house mound. At a depth of 1.9-2.0 m large quantities of
obsidian debitage were encountered and has been interpreted to represent a
workshop area. From this same context a charcoal sample was dated to AD
967±60 (calibrated) (AD 885-1005). Aztec I ceramics are present with the
workshop obsidian and below.

Operationj
Operation J was located 12 m south of Operation H and was excavated
to a depth of 1. 7 m. The stratigraphy of the unit was complex largely due to
the modern disturbance which extended to a depth of 95 em. Beneath this were
multiple lenses of ash and baked earth which rested upon a yellow brown sand
at approximately 120 em. Within this deeper deposit at a depth of 134 em,
charred material from a grey ash lens produced a corrected date of AD 1168±60
(AD 1041-1226). Aztec I ceramics were associated with this layer and below.
LABORATORY OBSIDIAN HYDRATION RATES 197

Operation P
Operation P was placed adjacent to Mound 94 and was excavated to a
depth of 1.2 m. Stratigraphically complex, the upper 70 em of the unit had been
disturbed by a modern trash pit which had penetrated through several sus-
pected floors. From deeper soil strata the remains of carbon, organic matter,
bone, and corn cobs were associated with Aztec II ceramics. A carbon sample
from this suspected midden context gave a calibrated date of AD 14 26±60 (AD
1402-1444).

Operation R
Operation R was located at the base of a mound (Structure 108) in a
surface concentration of ceramics. At a depth of 20-30 em a large refuse pit
was encountered. The pit extended to depths of up to a meter and contained
substantial quantities of Aztec III-IV black-on-orange ceramics, obsidian,
bone fragments, shell, and rock. A corrected date of AD 1410±70 (AD
1321-1439).

Hydration Rate Determination

Obsidian hydration rates for the Xaltocan artifacts were estimated from
the composition of the glass using the calibrations that relate the concentration
of OH- to the rate of hydration at high temperature and activation energy (Fig.
8.1). Infrared spectroscopic analysis was used to determine the OH- concen-
tration within each artifact. The estimated hydration rate constants were then
adjusted to reflect the ambient site conditions of temperature and relative
humidity. The adjusted hydration rates were then used to convert the hydration
rim widths to an absolute age before present.
A single obsidian blade fragment was selected from each radiocarbon
dated context for dating. The thickest and clearest specimens which were
free of inclusions or a cloudy appearance guided the selection process since
inclusions within the glass would interfere with the infrared analysis. A
parallel sided section approximately 1 mm thick was removed from the artifact
and each side polished to a 600 pm finish. A Bomem MB-120 Fourier
transformed infrared spectrometer was used to measure the hydroxyl con-
centration. The calibration constants provided by Newman et al. (1986) were
used to convert the oH- peak at 4500 cm- 1 to a weight percent value. Table
8.4 presents the results of the OH- analysis and show that concentration
levels range from 0.08 to 0.13 wt.%. These values were then used to estimate
the hydration rate at 160°C (A pm 2/day) and the activation energy (E ]/mol)
by application of the calibrations which relate composition to hydration rate
(Figure 8.1).
198 CHRISTOPHER M. STEVENSON et al.

Table 8 ...... Obsidian Hydration Rate Constants and Dates from Xaltocan, Mexico

Lab No. Provenience Y.OH. A(IUil2/day) E(Jimol) Width' S.D. EHT" %Rh Rate' Date S.D.

93-120 OPD,L.25 0.087 0.92 85945 2.34 0.07 21.3 100 4.42 711AD 108

93-127 OPG,L.I5 0.08 0.91 861% 2.28 0.05 21.3 100 4.24 724AD 110

93-136 OPH,L. 7 0.11 1.27 84774 1.90 0.05 21.7 95 6.72 1413AD 101

93-141 OPH,L.9 0.091 0.97 85760 2.52 0.05 21.6 95 4.46 526AD 115

93-145 OPI,L.I5 0.094 1.01 85621 2.13 0.05 21.3 100 5.07 1055AD 86

93-154 OPLL.34 0.102 1.11 85309 2.69 0.05 21.3 100 5.80 702AD 94

93-158 OP1,L.43 0.13 1.45 84280 3.00 0.05 21.3 100 8.76 912AD 70

93-166 OP J, L.13 0.114 1.24 84873 2.51 0.05 21.3 100 6.86 1032AD 75

93-176 OP P, L.IO 0.082 0.86 86156 1.43 0.05 21.6 95 3.89 1424AD 71

93-186 OPR,L.4 0.110 1.26 84836 2.55 0.06 21.9 95 6.77 990AD 77

'Width in microns
'EHT in"C
'Rate in 1'1'1'/IOOOyears

The high temperature hydration rate was then extrapolated to site ambi-
ent conditions. In 1990, pairs of saline based monitoring cells (Trembour et al.
1988) were placed at the site of Xaltocan at depths of 50 em, 100 em, and 150
em to record the conditions of temperature and relative humidity at different
depths. The results of the monitoring effort (Table 8.5) indicated that the EHT
was 21.9"C at 50 em and declined slightly with increasing depth (Fig. 8.4). The
relative humidity was 94% at the depth of 1.0 m and reached a 100% level by
1.5 m. These values were then used to adjust the 100% relative humidity high
temperature rates according to the depth of the artifact within the soil profile.
Artifacts found at depths of 1.3 m and below required no adjustment for relative
humidity. However, the hydration rate for artifacts found at shallower depths
in a 95% Rh environment was reduced by 10% (Mazer et al. 1991). Calendar
dates were then computed (Table 8.4).

Table 8.5. EHT and Relative Humidity Conditions

for Xaltocan, Mexico

Cell No. Depth Exposure EHT("C) %Rh

77/78 50 em 380 Days 21.95 95

75/76 IOOcm 380 Days 21.56 94

73/74 !50 em 380Days 21.30 100

LABORATORY OBSIDIAN HYDRATION RATES 199

22.0

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

......... ..... ...... ... ....... .......... .. ............................ ·-.. ............................. ..
---·-··
21.9 .. ..
. ...
21.8 ...... ·- ........................ .... .. .... . ................. ··:··· .................................. .. -- ......................... .
~- ~-

.. --.
.. ......... --- ............... . ............................... ..
--
-------
21.7 ..
.................................................................
..
C)
. .
-
. ............................. ............................. .
................................ .. ........................... ··!···
E'-' 21.6 .
~-

. . . -~-

......
!:OJ
21.5 ...
........................................................ .
----------
. .................................................................. .

21.4 .. .. . ...... .... .... .. ...............................................

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

21.3 .............. --- ........ ---~·

. ................................... .:.................................... .

21.2
0 50 100 150 200
Depth (m)
Figure 8.4. Xaltocan effective hydration temperature for varying soil depths.

Comparison of the Radiocarbon and OHD Results

The correspondence between the radiocarbon assays and the obsidian
hydration dates is presented in Figure 8.5. Here, we see that five of the 10 date
pairs agreed very well and had overlapping age ranges at one standard devia-
tion. In these cases, the mean values for the obsidian dates just slightly over-
or under-estimated the radiocarbon date means. Of the remaining five date
pairs, none had overlapping date ranges. In each case the obsidian dates were
substantially earlier than the radiocarbon dates. The high degree of correspon-
dence in half of the data set suggested that the temperature and relative
humidity calibrations were suitable. However, other intervening variables
appear to have affected the outcome for the remaining samples.
The radiocarbon evidence (Table 8.6) indicated that the site of Xaltocan
was occupied between approximately AD 900 and AD 1400 (Aztec I to Aztec
200 CHRISTOPHER M. STEVENSON et al.

Age A.D. Age AD

1600 1600
1400
+ +
"' 1400

.
1200 + 1200
1000 *
+ • 1000
800
*
.. .. 800
600 . 600
400 400
200 200
0 0
OP I OP I OP I OP G OP J OP D OP H OP H OP A OP P 0
Excavation Units
* Obsidian Dates + Radiocarbon Dates

Figure 8.5. Comparison of radiocarbon and obsidian age estimates on Pachuca obsidian at
Xaltocan, Mexico.

IV). Of the non-corresponding obsidian dates, the sample from the refuse pit
in OP R was much earlier than the radiocarbon and ceramic evidence suggested,
yet the age determination occurred within the age range for the site as a whole
and may have been present because of local recycling and/or redeposition of
the artifact. The remaining four samples however, have produced dates between

Table 8.6. Radiocarbon Dates from Xaltocan, Mexico

Operation Level Beta Lab No. Uncalibrated Date Calibrated Date

D 25 41910 AD 1130±70 1229 AD (1168-1282)

G 16 41911 AD 880±80 989 AD (890-1 026)

H 10 41912 AD 1350±70 1328,1333,1395 AD

(1300-1416)

15 41913 AD 840±60 967 AD (885-1005)

32 50303 AD 830±90 898, 906, 961 AD

(820-1014)

13 41914 AD 1060±60 1168 AD (1041-1226)

p 10 50315 AD 1440±60 1426 AD (1402-1444)

R 50316 AD 1410±70 1410 AD (1321-1439)

LABORATORY OBSIDIAN HYDRATION RATES 201

A.D. 526 and A.D. 724, a date range which was several centuries before the
earliest radiocarbon evidence.
The reason for this occurrence of early obsidian dates is not clear but is
possibly a result of obsidian scavenging and recycling from nearby sites.
Evidence for this type of behavior was found in Unit G (Levels 1, 3, 12, 15)
and Unit D (Levels 3 and 24). Here, Classic period figurings were found in
contexts with no Classic-period ceramics. It is also possible that the scavenging
of obsidian by the Xaltocamecas from earlier sites was practiced in the ninth
through eleventh centuries when trade networks were disrupted by the collapse
of Teotihuacan, or by the domination by Tula from the north (Brumfiel:
personal communication).

SUMMARY AND DISCUSSION

The goal of dating is to provide temporal control over the systemic

processes which have created the archaeological record and are observed in
the form of features such as pits, platforms, rooms, and stratigraphic layers
(Dean 1978:244). However, in the use of any chronometric dating method,
two forms of dating error can occur. They include calibration errors and errors
of convergence. In the latter case, dating errors arise from the fact that the
dated event is not related to the behavioral target event with which the dated
specimen was associated. Dean (1978) has discussed at length the reasons
for this divergence.
In this chapter, we have been concerned almost exclusively with the
problem of calibration. We began by arguing that under many archaeological
conditions the process of obsidian hydration is principally one of molecular
water diffusion rather than the commonly accepted model of alkali depletion
and replacement with hydronium ions. It was also argued that the best
approximation of the hydration process could be adequately modeled in the
laboratory under conditions of elevated temperature within a vapor environ-
ment maintained at 100% relative humidity. It was then proposed that the
hydration rate constants obtained for individual artifacts using these methods
can be used to calculate chronometric dates under measured temperature
and relative humidity conditions. The case example from Xaltocan supported
the ability of the laboratory methods to produce dates that converge with
other chronometric methods. However, a considerable amount of divergence
between the radiocarbon and obsidian dates was present and could not be
explained using the available evidence. Clearly, additional case examples will
be need to be completed in order to demonstrate the accuracy of the cali-
brations involved.
The research of the last few years conducted by the authors and others
has brought about an increased level of understanding and enhanced utility.
202 CHRISTOPHER M. STEVENSON et al.

We believe however, that the most pressing problems still remain within the
field of calibration and instrumentation. Accurate and precise calibrations
for the response of water diffusion to reduced levels of relative humidity is
still needed as is the routine implementation of a non-optical instrumentation
for the measurement of the hydration rim. These and other advances will
serve only to enhance the utility of the obsidian dating method in all future
applications.

ACKNOWLEDGMENTS

The authors would like to thank Maria Klimkiewicz for conducting the
X-ray fluorescence analysis of the obsidian source samples and Elizabeth
Brumfiel for providing us with samples from Xaltocan and use of her radiocar-
bon dating results. Tom jackson provided many useful comments that helped
to improve the quality of this manuscript.

REFERENCES

Abrajano, T., Bates,]., and Mazer,]. 1989 Aqueous corrosion of natural and nuclear waste glasses;
II. Mechanisms of vapor hydration of nuclear waste glasses. journal ofNon-Crystalline Solids
8:269-288.
Ackerly. N ., and R. Giese 1993 Status report at Tortugas Mountains, Dona Ana County, New Mexico.
Report to the Office of the President, The New Mexico State University. Las Cruces, NM.
Aiello, P. 1969 The chemical composition of rhyolitic obsidian and its effect on hydration rate:
some archaeological evidence. Masters Thesis, Department of Anthropology. UCLA.
Ambrose, W 1976 Intrinsic hydration rate dating of obsidian. In R. E. Taylor, ed., Advances in
Obsidian Glass Studies, Park Ridge, New jersey. Noyes Press: 81-105.
Bartholomew, R., Tick, P., and Stookey, S. 1980 Water/glass reactions at elevated temperatures and
pressures. journal of Non-Crystalline Solids 38 and 39:637-642.
Bates,]., Abrajano, T., Ebert, W, Mazer,]., artd Gerding, T. 1988 Experimental hydration studies
on natural and synthetic glasses. In E. Sayre, P. Vandiver, E. Drusick, and C. Stevenson, eds.,
Materials Issues in Art and Archaeology. Proceedings of the Materials Research Society:
237-244.
Bergland, E., McAlister,]., and Stevenson, C. 1993 An induced hydration rate for Obsidian Cliffs
glass. In P. W Baxter, ed., Contributions to the Archaeology of Oregon, 1990-1994. Association
of Oregon Archaeologists, Occasional Papers No. 5: 1-13.
Cleland,]. 1990 Induced hydration at Coso: Part Ill. Paper presented at the 24th Annual Meeting
of the Society for California Archaeology. Foster City. California.
Dean, ]. 1978 Independent dating in archaeological analysis. In Schiffer, M.B. ed., Advances in
Archaeological Method and Theory, Volume 1. New York, Academic Press: 223-255.
Dyson, J. 1960 Precise measurement by image-splitting. journal of the Optical Society of America
50: 754-757.
Ericson,]. 1975 New results in obsidian hydration dating. World Archaeology 7: 151-159.
_ _ 1977 Evaluation of prehistoric exchange systems: results of obsidian dating and tracing.
Ph.D. Dissertation, UCLA.
LABORATORY OBSIDIAN HYDRATION RATES 203

_ _ 1988 Obsidian hydration rate development. In E. Sayre, P. Vandiver, E. Drusick, and C.

Stevenson, eds., Materials Issues in Art and Archaeology. Materials Research Society Proceed-
ings, Volume 123: 215-224.
Ericson,]., Mackenzie,]., and Berger, R. 1976 Physics and chemistry of the hydration process in
obsidians, II: experiments and measurements. In Taylor, R.E., ed., Advances in Obsidian
Glass Studies. Park Ridge, New jersey, Noyes Press: 46-62.
Findlow, F, Bennett, V, Ericson,]., and De Atley, S. 1975 A new obsidian hydration rate for certain
obsidians in the American Southwest. American Antiquity 40:344-348.
Findlow, F, Martin, P., and Ericson,]. 1982 An examination of the effects of temperature variation
on the hydration characteristics of two California obsidians. North American Archaeologist
3: 37-49
Friedman, I. and Long, W. 1976 Hydration rate of obsidian. Science 159: 347-352.
Friedman, I. and Smith, R. 1960 A new dating method using obsidian: part 1, the development of
the method. American Antiquity 25: 476-493.
Friedman, I., Trembour, F, Smith, F and Smith, G. 19941s obsidian hydration dating affected by
relative humidity? Quaternary Research 41: 185-190.
Green,]. 1986 Obsidian hydration measurement: are we getting what we expect? Paper presented
at the Great Basin Archaeological Conference, Las Vegas, Nevada.
Hench, D., Clark, D., and Yen-Bower, E. 1980 Corrosion of glasses and glass ceramics. Nuclear and
Chemical Waste Managment l: 59-75.
Hurta to de Mendoza, L. and jester, \VA. 1978 Obsidian sources in Guatemala: a regional approach.
American Antiquity 43: 424-435.
jackson, R. 1984 Current problems in obsidian hydration analysis. In R. E. Hughes, ed., Obsidian
Studies in the Great Basin. Contributions of the University of California Archaeolog1cal
Research Facility. Berkeley: 103-116.
johnson, L. 1969 Obsidian hydration rates for the Klamath Basin of California. Science 165:
1354-1356.
Kondo, Y., and S. Matsui 1992 Application of obsidian hydration dating with the Hitachi Model
U-6000 microscopic fourier-transform spectrophometer- Dating of stone implements using
a new non-destructive technique. Hitachi Scientific Instrument News 35: 11-14.
Labs, K. and Harrington, K. 1982 A comparison of ground and above-ground climates for
identifying appropriate cooling strategies. Passive Solar journal 1: 4-11.
Leach, B.F, and Hamel, G.E. 1984 The influence of archaeological soil temperatures on obsidian
dating in New Zealand. New Zealand journal of Science 27: 399-408.
Lee, R. 1969 Chemical temperature integration. journal of Applied Meteorology 8: 423-430.
Mazer,]., Stevenson, C., Ebert, W., and Bates,]. 1991 The experimental hydration of obsidian as
a function of relative humidity and temperature. American Antiquity 56: 504-513.
Mazer,]., Bates,]., Stevenson, C., Bradley, C. 1992 Obsidians and tektites: natural analogues for
water diffusion in nuclear waste glass. Materials Research Society Symposium Proceedings.
Volume 257: 513-520.
Meighan, C. 1983 Obsidian dating in California: theory and practice. American Antiquity 48:600-609.
Michels,]., and Tsong, I.S.T. 1980 Obsidian hydration dating: a coming of age. In M.B. Schiffer,
ed., Advances in Archaeological Method and Theory, Volume 3. New York, Academic Press:
405-439.
Michels,]., Tsong, 1.S.T., and Smith, G. 1983 Experimentally derived hydration rates in obsidian
dating. Archaeometry 25: 107-117.
Nelson, F 1984 X-ray fluorescence analysis of some western North American obsidians. In Hughes,
R.E., ed., Obsidian Studies in the Great Basin. Contributions of the University of California
Archaeological Research Facility, University of California, Berkeley: 27-62.
Newman, S., Stolper, E., and Epstein, S. 1986 Measurement of water in rhyolitic glasses: calibration
of an infrared spectroscopic technique. American Mineralogist 71: 1527-1541.
204 CHRISTOPHER M. STEVENSON et al.

Redfield, A.C. 1965 Terrestrial heat flow through salt-marsh peat. Science 148: 1219-1220.
Riddings, R. 1991 Obsidian hydration dating: the effects of mean exponential ground temperature
and depth of artifact recovery. journal of Field Archaeology 18: 77-85.
Scheetz, B., and Stevenson, C. 1988 The role of resolution and sample preparation in hydration
rim meas urement; implications for experimentally determined hydration rates. American
Antiquity 53: 110-117.
Silver, L., Ihinger, P., and Stolper, E. 1990 The influence of bulk composition on the speciation of
water in silicate glasses. Contributions to Mineralogy and Petrology I04: I42-162.
Smith, T. I977 Obsidian hydration as an independent dating technique. Masters Thesis, Depart-
ment of Anthropology, University of Alaska: Fairbanks, Alaska.
Stevenson, C., Freeborn, W, and Scheetz, B. 1987 Obsidian hydration dating: an improved optical
technique for measuring the width of the hydration rim. Archaeometry 29: 120-123.
Stevenson, C., and Scheetz, B. 1989 Induced hydration rate development of obsidians from the
Coso volcanic field: a comparison of experimental procedures. In Hughes, R.E., ed., Current
Directions in California Obsidian Studies, No. 48. Contributions of the University of Califor-
nia Archaeological Research Facility, Berkeley: 23-30.
Stevenson, C., Carpenter,]., and Scheetz, B. 1989a Recent advances in the experimental determi-
nation and application of obsidian hydration rates. Archaeometry 31: 193-206
Stevenson, C., Dinsmore, D., and Scheetz, B. 1989b An inter-laboratory comparison of hydration
rim measurement~. International Association for Obsidian Studies Newsletter 1.
Stevenson, C., Knaus, E., Mazer, j., and Bates, j. 1993a Homogeneity of water content in obsidian
from the Coso volcanic field: Implications for obsidian hydration dating. Geoarchaeology
8:371-384.
Stevenson, C., Friedman, I., and Miles,]. 1993b The importance of soil temperature and relative
humidity in obsidian dating, with case examples from Easter Island. In Fischer, S.R. ed.,
Easter Island Studies: Contributions in Memory of William T. Mulloy. Oxbow Monograph 32,
Oxford: 96-102.
Tremaine, K., and Frederickson, D. 1988 Induced obsidian hydration experiments: an investigation
in relative dating. In Sayre, E., Vandiver, P., Drusick, E., and Stevenson, C. eds., Materials
Issues in Art and Archaeology. Proceedings of the Materials Research Society. Volume 123:
271-278.
Trembour, E, Smith, E, and Friedman, I. 1988 Diffusion cells for integrating temperature and
humidity over long periods of time. In Sayre, E., Vandiver, P., Drusick, E., and Stevenson,
C., eds., Materials Issues in Art and Archaeology. Proceedings of the Materials Research
Society. Volume I23: 245-252.
Van Wilk, W, and Derksen, W 1966 Sinusoidal soil temperature variation in a layered soil. In W
Van Wilk, ed, Physics of the Plant Environment, Amsterdam, North Holland Publishing Co.
White, W.B. 1986 Dissolution mechanisms of nuclear waste glass: a review. Advances in Ceramics,
Vol. 20: Nuclear Waste Managment II: 431-442.
_ _ 1988 Glass hydration mechanisms with application to obsidian hydration dating. In Sayre,
E., Vandiver, P., Drusick, E., and Stevenson, C., eds., Material Issues in Art and Archaeology.
Materials Research Society Symposium Proceedings. Volume 123: 225-236.
Young, j.F. 1967 Humidity control in the laboratory using salt solutions-a review. American
Laborat01y 17: 24I-245.
 Chapter 9

Obsidian Hydration Dating at

a Recent Age Obsidian Mining
Site in Papua, New Guinea
w. R. AMBROSE

ABSTRACT

The relatively high temperature of the humid tropics accelerates weathering

processes, including the hydration of obsidian, where the warm conditions can
promote a measurable hydration thickness in less than fifty years. On the other
hand, the aggressive weathering environment is capable of reducing the obsid-
ian surface to make standard measurements untenable. By targeting protected
fissures not exposed to surface weathering effects, a consistent series of
hydration readings has allowed age determinations of an obsidian mine site
where the radiocarbon dates are unreliable, or fall within the undateable
"modem" of the last 300 years. The results of this approach, including long
term laboratory-based hydration rate determinations, show that obsidian hy-
dration dating is a viable dating system free of dependence on other dating
methods to provide its time constants.

W. R. AMBROSE • Division of Archaeology and Natural History, Research School of Pacific and
Asian Studies, Australian National University, Canberra, Australia.
Archaeological Obsidian Studies, edited by Shackley.
Plenum Press, New York, 1998.

205
206 W. R. AMBROSE

INTRODUCTION

Obsidian hydration dating may have had a reputation as a doubtful

performer among the dating systems developed over the last few decades, but
a better appreciation of the fundamental importance of site temperature and
the chemistry of obsidian weathering as part of the hydration process have
greatly improved its prospects. From its promising beginnings and the system-
atic approach initiated by Friedman and his co-workers (Friedman and Smith
1960; Friedman et al. 1966), successive researchers have found the dates
derived from obsidian hydration to be wanting for various stated reasons. A
lack of correspondence with other chronometric systems, particularly radio-
carbon dating, is often seen in the literature of obsidian hydration dating. The
radiometric systems should be generally more reliable, being relatively free of
the host of environmental constraints that beset the chemical reaction rate
systems such as obsidian hydration dating and amino acid racemization dating.
Nevertheless, poor correspondence alone would not be sufficient to disregard
the obsidian dates in favor of another system in all cases. For example, the
radiocarbon time scale lacks credibility for determinations covering the last
300 years, mainly due to disturbance in the balance of atmospheric C0 2 by
fossil C0 2 from world-wide industrialization (Fig. 9.1).
Radiocarbon dates for these recent centuries are usually lumped within
a single "modern" result. In such circumstances, the use of obsidian hydration
dating could be expected to produce more reliable dates than radiocarbon
dating and evidence is presented here to support this view.
This research includes the use of hydration rates from both long term and
short term exposure of experimental obsidians, an estimate of site temperature
from short term monitoring, computer assisted measurement of hydration layer
thicknesses and a targeted selection of hydration measurements from concealed
fissures that are protected from obsidian surface weathering effects. The hydra-
tion dates are applied to a prehistoric obsidian mining location at the Umleang
site (GBJ), 2.5° south of the equator on Lou Island, in the Manus Province of
Papua New Guinea. Lou Island is formed from a series of coalesced volcanic
cones made up of rhyolite, pumice and tephra.
In one location, at the traditional remnant village of Umleang, at least
twenty vertical shafts, about 60-SOcm in diameter, were sunk through beds of
pumice and volcanic ash to locate imbedded blocks of obsidian up to 16m
below the ground surface. The shafts are still visible though most are backfilled
to a varying extent, with some in evidence only as dimples in the ground
surface, but with one shaft still accessible to 16m depth. At the present time,
the mining site is shaded by cocoa trees planted as a modern cash crop. The
Manus people have exploited other obsidians on the island for several thousand
years but without evidence for shaft mining as seen at the Umleang site. At
present, the island is volcanically dormant, the last obsidian producing erup-
OBSIDIAN HYDRATION DATING IN PAPUA, NEW GUINEA 207

500 ......
......
......
......
......
400 ......
......

_g-
Q)

~ 300
c:
0
.J:J

~ 200
0
"0

"'
a:
100

OL---~--~---L--~~--~--~---L--~----~~
500 400 300 200 100 0
Calender years BP
Figure 9.1. Radiocarbon age curve versus calendrical age for the last 500 years. The two enclosed
rectangles indicate periods of ambiguity in the age determination particularly prevalent in the last
three centuries (compiled from Stuiver and Reimer 1993). The equal age value between calendrical
and radiocarbon values is indicated by the dashed line.

tion occurring off-shore in 1953. The last use of the mines is not recorded but
was probably in the early part of this century.
The last major island-wide volcanic eruption was around 1600 years ago,
with the first construction of the mines an indeterminate time after this. The
deep shaft mines produced a considerable amount of sterile pumice upcast
which, when spread over the surrounding area and sealing earlier soil layers,
would potentially provide a good indicator for the chronology of mining.
Several deep shafts are still open in the area so that backfilling of older mine
shafts with pumice from ones dug later would not remove all the waste material
from the vicinity. The obsidian hydration study is an attempt at one locality to
date the start of this activity by examining the upcast age of mine number 9
(Fullagar and Torrence 1991) presently backfilled to within about two meters
of the surface.
Archaeological work was located 3m from the rim of the mine shaft where
a 5m by 1m trench, up to 1.25m deep was excavated. The deposits were
removed as 12 natural layers distinguishable on texture and appearance.
Obsidian was the main artifact material although a small amount of pottery was
also recovered.
208 W. R. AMBROSE

Experimental Hydration Rates

The difficulty of measuring a thin hydration layer from a recently made
obsidian artifact is offset in the humid tropics by the rapid growth of hydration
in high mean annual temperatures above about 25°C. With some obsidians a
measurable hydration thickness of more than 1 micrometer can be produced
in less than 50 years, so that it becomes possible to investigate the chronology
of archaeological sites that otherwise fall within the ambiguous last 300 years
range of radiocarbon dating. To take advantage of this it is first necessary to
determine the relationship of hydration rate to temperature.
As part of an experimental program to define the hydration rates of the
main Papua New Guinea obsidian sources, fresh flakes struck from obsidian
collected at the GBJ quarry site were placed in Pyrex beakers and exposed at
normal air pressure to saturated water vapor within stainless tanks immersed
in temperature controlled water baths. The rate constants for obsidians in the
study reported here have been calculated from these experimentally hydrated
flakes exposed at30°C and 40°C for between 10 to 15 years and others exposed
in a pressure vessel at 105 o Cfor up to 7 months. The short 5o C extrapolation,
to around 25 oC, should therefore allow a close matching of the experimentally
produced hydration rate to the temperature of obsidians in the normal terres-
trial conditions at the site.
Thin wafers, 0.5mm thick, were cut from the experimentally exposed
obsidians and the site artifacts, using a Lastec diamond wire saw. Small samples
between 1mm to 3mm wide were snapped from the main slide wafer for
chemical analysis of major elements by scanning electron microscope EDAX.
The main portions were then mounted and prepared as microscope slides for
hydration measurement in the normal way. Flakes are routinely measured for
their density using a modified system of Hughes and Oddy (1970), since at
other sites in Papua New Guinea there is evidence that density is correlated
with the hydration rate.
The hydration of obsidian progresses at a rate that is a function of the
square root of time. It follows therefore that hydration thickness is most accurate
in measuring time in the most recent period when hydration is proceeding most
rapidly. This should commend the system for dating recent sites, but unfortu-
nately the constraint of resolving power in optical microscopy limits the meas-
urement of very small thicknesses of hydration in the obsidian surface and
results in poor definition of the brief hydration increments that span the recent
period. The adoption of nuclear profiling (Lee et al. 1974) would be advanta-
geous in allowing laboratory based experiments to define the hydration rate
constants over a relatively short period at normal atmospheric temperatures.
One such case was demonstrated in this way for the GBJ obsidian after it was
exposed to water vapor at 40°C for 1.22 years (Duerden et al. 1982: 241), but as
the method is not readily available, optical microscopy remains the standard
OBSIDIAN HYDRATION DATING IN PAPUA, NEW GUINEA 209

measurement system. Another approach is to greatly accelerate the hydration by

exposing obsidian to high temperatures for brief periods measured in days. For
example, the exponential relationship of the hydration rate to temperature has
encouraged the use of elevated temperatures of 140 to 160°C in saturated water
vapor to hasten the growth of hydration in order to derive a hydration rate from
short term experiments, which can then be extrapolated to a longer term
hydration rate in normal temperature environments (Bates et al. 1988; Stevenson
and Scheetz 1989; Tremaine and Fredericksen 1988).

Obsidian Surface Erosion

It has been shown (Marshall1980) that the temperature dependence of the
process of aqueous silica dissolution alters between the high pressure, high
temperature conditions above l00°C, and lower normal atmospheric vapor
pressure temperatures. As a result it is standard practice now to expose experi-
mental obsidians in water vapor. However, it is yet to be demonstrated that the
phenomenon described by Marshall does not also effect the hydration process in
water vapor in a similar fashion. The extrapolation of experimental hydration
rates from water vapor at around 140°C to normal terrestrial temperatures
requires some attention to precision in maintaining the experimental conditions.
A small error in the measurement of hydration thickness, or in the high
experimental temperature, can translate into a significant deviation when the
rate constant is projected across some ll5°C to a normal site temperature.
While rapid dissolution of obsidian in aqueous solution may be alleviated
by conducting experimental hydration rate studies in water vapor, no such
measures are possible at archaeological sites. The identification of microbial
activity as an agent in removing the surface of basaltic glasses (Thorseth et al.
1992) is an important biological factor additional to the usual chemical erosion
of glass that occurs on either side of neutral pH (Casey et a!. 1992) and glass
surface dissolution by clay minerals through ion exchange mechanisms (Curti
et al. 1993). It seems likely therefore that surface erosion will commonly occur
at many archaeological sites where there is sufficient ground moisture, whether
in the tropics or cooler climates. This is consistent with the evidence that two
competing sets of reactions are operating in the weathering of obsidian, namely
hydration and dissolution (White 1988: 230).
The benefit of the rapid hydration rate in the tropics can be offset by the
aggressive nature of the weathering environment in moist soils in both open
sites and rock-shelters. This effect is apparent in many thin sections of obsidian
artifacts where surface pitting can be seen and in hand specimens where the
obsidian flakes have lost the gloss of freshly fractured surfaces. This latter effect
is the result of surface loss due to general silicate dissolution along with the
less obvious removal of migrating ions that exchange for the diffusing water in
the hydration layer. A strategy is necessary which avoids those surfaces where
210 W. R. AMBROSE

loss has occurred and where measurement would record only residual hydra-
tion. The conventional selection of clear obsidian flake surfaces is therefore not
suited for hydration measurements where surface dissolution is occurring.
Even when an obsidian artifact appears to have a glassy surface in hand
specimen, microscopic examination may often show evidence of surface loss.
Where the surface loss is equal to or faster than the hydration rate at the
outset, then no hydration can be present. Where this is not the case, while the
hydration is progressing as a function of the square root of time, the surface
loss is proceeding at a simple linear rate, until there is a stage when the
hydration increment diminishes to equal the surface loss. The time that must
pass for this to occur, will be governed by the relative rates of hydration and
surface dissolution. The effect of this asymptotic growth of hydration, where
it is not recognized as a feature produced by surface loss, would be to produce
hydration thicknesses on old obsidians that are too small and derived dates
which are too young. This means that if old obsidian artifacts affected by
unrecognized surface loss are coupled with radiocarbon dates to calculate
hydration rate constants, then the dates of younger artifacts will be given
erroneously old dates. The reverse will occur if radiocarbon dates for young
obsidians are used to provide the hydration rate constants for unknown age
older specimens affected by surface loss, which will appear then to be too
young. An additional complication arises from the effects of microenvironmen-
tal differences within the archaeological deposit that can differentially weather
a single obsidian flake.

Protected Fissure Hydration

When an obsidian flake is struck from a core the blow at the point of
percussion may also create a series of inconsequential fractures. These features
can be seen at the surface as cones which terminate in the body of the flake or
cracks that similarly penetrate the obsidian without detaching a flake (Fig. 9 .2).
In themselves these incompletely propagated flakes may offer no protection to
the inner surfaces from the adverse effects of weathering but they can show
whether the hydration is complete or reduced.
In thin section some of the hydrated flake scars that terminate within the
obsidian can be seen as two parallel bands in contact, but others may show a
characteristic open channel between the two bands, indicating that weathering
has removed the hydration in penetrating along the flake scar surfaces into the
body of the obsidian. It is the unweathered parallel bands that have been
targeted for measuring hydration thickness for the purposes of dating the GBJ
site. Where outer surfaces or weathered channels also bear hydration these have
been measured for comparative purposes. The rate of surface loss can be
calculated from paired readings of the inner protected surfaces and the outer
weathered ones.
OBSIDIAN HYDRATION DATING IN PAPUA, NEW GUINEA 211

Figure 9.2. Schematic representation of a cross section of flake development about the point of
percussion with (A) exposed surface, (B) partly detached flake with erosion forming an open cleft,
and (C) concealed crack. The concealed crack is protected from dissolution and provides the most
reliable hydration thickness reading.

In a study of obsidian from other sites on Lou Island (Ambrose 1993)

hydration rates were reported that were based on the conventional use of
radiocarbon dates and obsidians from one site to provide the hydration rate
constants for dating another site on the island. Both sites were under at least
Sm of siliceous volcanic ash; the obsidian flake surfaces would therefore be
expected to be well protected against silica dissolution. However, the obsidian
hydration rate constants calculated from conventional surface hydration
measurement and the radiocarbon dates of these two ash-sealed sites show
a major discrepancy when compared with the intrinsic rate from the experi-
mentally found rates, and the measurement of closed fissures on the obsidians
at the GBJ mine site. The experimentally derived obsidian hydration rate
constants applied to the closed fissure measurements of the GBJ mine site
samples show that the hydration velocity appears to be about twice as fast
as the radiocarbon based rate constant from the ash-sealed sites. This dis-
crepancy can be attributed to surface dissolution of obsidians from the
radiocarbon dated sites, despite their burial beneath meters of siliceous
volcanic ash. In general, the instances of questionable dates supplied from
obsidian hydration dating seem to err more often on the side of "too young"
than "too old". This may indicate a phenomenon of surface dissolution of
obsidian more widespread than the humid tropics.
212 W. R. AMBROSE

Hydration Measurement
Microscope measurement of the hydration depth is done on the standard
thin sections with a xlOO oil immersion objective giving a magnification at the
eyepiece of xlOOO. The microscope image is scanned through a CCD camera
connected to PC based digitizing equipment and manipulated with video
analysis software.
The procedure scans a strip several micrometers wide crossing the
hydration zone from its outer straight edge to beyond its inner boundary, to
produce an integrated gray scale profile (Fig. 9.3). Where the extent of straight
hydration edges allows, up to 20 integrated strip profiles are recorded. Each
profile is stored as a pixel series so that both the outer edge and inner boundary
of the hydration are registered by steep inflexions in the gray scale.

Q)

<13 0
0 Q
U) 0

>. "a
Q) 0
0
oo
01

>-
<13

..0
<(

000
oo
0
oo 0

0 10 20 30 40 50 60
Pixels
Figure 9.3. Typical averaged grey scale profiles across the band of external surface hydration, (A)
nll, and the double band of an internal crack (B) n6, showing differences in hydration width on
sample 4904. Each pixel represents .ll03!lm. The 21.8 pixels from A represents 2.41\lm, while
the half width pixel count of 29.3 on Bis 3.24\lm.
OBSIDIAN HYDRATION DATING IN PAPUA, NEW GUINEA 213

Gray scale enhancement has been used to increase the contrast at the
inflexion points and allows for easier reading of the pixel defined hydration
dimensions. A measure of the hydration width is found by pixel counting
between the inflexion points. Adisk file is made of the profiles of each obsidian
from which the mean and standard deviation of the hydration readings in
micrometers are derived from the pixel record. The pixel dimension is cali-
brated from the same digitized image arithmetic across a stage micrometer. This
procedure gives a calibrated value of (n=32) O.ll03±.001pm per pixel and set
the limit of resolution with this rig.
Previous work (Ambrose 1993) shows that computer based image analy-
sis of hydration thickness decreases the measurement error to about half
compared with operator based visual microscopy. Table 9.1 and Fig. 9.4 give
the results of hydration measurements on the set of obsidians from the GBJ site.

r- 06

2 - ~ !::,. !/!!:,.
•
!::,.
• •
4 6 ~

/':,. .... • •
6 r- •• M

•
Cll
>.. r- !::,.!::,.
cQ
~·
•
_J

8 tM

M6
•
10 6 6 /':,./':,.

12
r- 66
I
•
0 2 3 4 5 6
Micrometres
Figure 9.4. Mean hydration thicknesses in micrometres determined fro m the digitized image
scans. Open triangles indicate measurements made on hydration from the exposed outer surfaces
or the eroded fissures. Infilled squares are mean readings for hydration from protected inner crack
surfaces.
214 W. R. AMBROSE

The hydration thicknesses in micrometers are given with a standard deviation

based only on the measurement error of the pixel scans.

Site Temperature
Meteorological air temperature records from the Momote station, 40
km north on Manus Island, give an annual monthly mean of 27.2±0.rc,

Table 9.1. Hydration Thickness Measurements in Micrometers

of Obsidians Recovered from Archaeological Excavations
near Mine 9 at GBJ Site, Lou Island

Lab Closed Open Exposed

No. Layer fissure fissure swface

4888 I 0.97 ± 15 5.70 ± 25

4889 I 1.04 ± 10 1.31±19
4890 I 0.99± 12 1.03 ± 09
4891 2 3.66± 28 2.99±33
4891 2 2.65± 19
4892 2 0.77 ± 14 0.91 ± 17
4893 2 2.95±31
4894 3 2.29± II
4895 3 4.59±32
4896 3 2.60± 19
4897 4 2.36± 17 2.28±28
4898 4 2.21 ±23 1.47 ± 27
4899 4 2.48± 12
4900 5 3.50± 14 2.22 ± 19
4901 5 2.12±04 1.05± 14
4902 5 2.91 ± 14
4903 6 2.50± 14
4904 6 3.24±08 2.41 ± 23
4905 6 2.87 ± 16
4906 7 3.12± II 2.22± II
4907 7 3.35 ± 09 2.42 ± 21
4908 7 4.90± 14 3.01 ± 19
4909 8 3.70± 10 2.10± 19
4909 8 1.97 ± II
4911 8 1.87 ±II
4912 9 2.85 ± 12
4913 9 2.59± 19
4914 9 3.38± 16 2.48±26
4915 10 2.92±24
4916 10 2.72±20
4916 10 2.16±53
4917 10 2.48± 15
4918 II 3.96±08 2.71 ± 19 2.89±21
OBSIDIAN HYDRATION DATING IN PAPUA, NEW GUINEA 215

(McAlpine et al. 1983: 183). The site sub-surface temperature at 20 and 30cm
depth was measured over a short 56-day period using thermal cells. It would
have been preferable to adopt the normal long temperature cycle approach
(Ambrose 1984) where a full annual range could be monitored, but this was
not possible at the site. Converting the short period record into an estimate
of the annual temperature in the oceanic equatorial zone may be less haz-
ardous than a similar exercise in higher latitudes. The double peak of solar
radiation during the equinoxes at the equator around March 21 and September
21 and the narrow temperature range of small tropical oceanic islands reduce
the error inherent in a limited monitoring period of this sort. By using the
double peak annual curve at the equator and superimposing the period of
the monitoring on the relevant part of the annual cycle, it is possible to
approximate the annual mean temperature, as set out in Fig. 9.5, by using
equation (l):

m =COS [(365)(alb)] r + T (l)

where m = mean annual temperature; a = median day number for the monitor-
ing period; b =thermal cycle (i.e., 182.5 days); r =half total temperature range;
and T = cell temperature for monitoring period.
With an estimated annual ground temperature range of ±0.5 o Ccombined
with the monitored ground temperature over 56 days of 25.4°C, the approxi-
mate annual mean site temperature is calculated at 25.8oC. Air temperature
monitoring at the GBJ site, 2.0m above ground in a tree shaded position, using
daily maximum and minimum thermometer readings for a two week period
around the June solstice, gave a value of 25.7±2.rC. As the GBJ site is an

29.0

28.0

(..) 27.0 1----~-------

Ul

~
0>
~ 26.0

25.0

24.0 u....__ _L__....I....__....J....__....J....__....I....__....J....__....J....__....J....__ _,__...L.__ _L__~

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 9.5. The temperature of the site expressed as a ±.5 oc range about the mean of 25.8°C. The
median day for the monitoring period and the temperature value from which the approximate
mean temperature is derived are indicated by a cross.
216 W. R. AMBROSE

exposed landscape, it is possible that it received direct solar radiation for

periods when it was used for mining, gardening or habitation. On the other
hand, the high 3500 to 3800mm annual rainfall and evaporation would tend
to lower the ground temperature. The archaeological excavation was under-
taken in a low-lying swale that would have been a zone of soil accumulation
from mining upcast or rainwash. It could be expected that surface obsidian
would soon have been buried by these local factors, but the short term
temperature registered by the cell monitoring can only be an approximation to
the mean temperature of the site.
A better result could be expected from a longer term soil temperature
monitoring program at the site, but in the absence of this no attempt is made
to allow for differences in the depth and temperature lag status of the site. As
an error of approximately 10% in the age estimation will result from a 1 oc error
in temperature, this is a salutary reminder of the importance of accurate site
temperature measurement. The lack of precise temperature details for the GBJ
site should be noted when considering the hydration dates set out below. Only
the hydration reading error is included in the figures without any attempt to
evaluate the temperature based error.

CALCULATED OBSIDIAN CHRONOLOGY OF THE GBJ SITE

The value of k, the hydration rate constant, can be found from the
hydration thickness X and the time t of the experimentally exposed obsidians
by using the standard hydration equation X2 =kt. From the long term exposure
of the experimental obsidians at the three temperatures of 30°C, 40°C and
105°C, the respective k values are found to be 0.0358, 0.0723 and 14.3148.
The Arrhenius least squares regression equation of the reciprocal Kelvin
temperature, 1/K, over the logarithm of the rate constant, ln k, gives a slope of
-9312, intercept of 27.264 and a correlation, r, of -.999 from which the k value
of the approximate site temperature (25.8°C) is predicted as 0.02051. This can
be contrasted to the k value of 0.00945 from the radiocarbon based h~dration
rate for obsidians at sites beneath 5 meters of volcanic ash discussed above.
The age calculations of the obsidians listed in Table 9.2 and Figs. 9.6 and 9.7
are based on a k value of 0.02051. There is no reason to believe that the sealed
soil layers beneath the volcanic ash would have had a much cooler thermal
history than the GBJ site, or a drier microclimate, that could explain the
different hydration rate between the sites. The open porous nature of the
pumice and ash beds would allow a ready percolation of moisture so that high
moisture content would be a permanent feature of all the sites.
The results give indications, firstly, of major differences between the
age from hydration in concealed cracks and those from hydration at exposed
surfaces and, secondly, mixing above layer 5, although a larger sample size
OBSIDIAN HYDRATION DATING IN PAPUA, NEW GUINEA 217

Table 9.2. Hydration Dates from the GBJ (Umleang) Site"

No Layer Closed Open Exposed Radiocarbon

fissure fissure swface dates

4888 I 1944± 14 405 ± 139 1730±80 Uncal AD.

4889 I 1937± 10 1906 ± 24 1640(1670)0 Cal AD.
4890 I 1942± 12 1938 ± 9 (ANU-2019)
4891 2 1337 ± 100 1554 ± 96
4891 2 1647 ±51
4892 2 1961 ±II 1950± 15
4893 2 1565 ± 89
4894 3 1734 ± 25
4895 3 962 ± 141
4896 3 1660± so
4897 4 1718 ± 39 1736 ± 62
4898 4 1752 ±51 1885 ± 39
4899 4 1690± 29
4900 5 1392 ±46 1750±41 Modem
4901 5 1771 ±8 1936 ± 14 (ANU-2195)
4902 5 1577 ±40
4903 6 1685 ± 34
4904 6 1478± 25 1707 ±54
4905 6 1588 ± 45
4906 7 1515 ± 33 1750±24
4907 7 1443 ± 29 1704± 50
4908 7 819±67 1548±56
4909 8 1322±36 1775 ± 39 Modem
4909 8 1801 ±21 (ANU-2196)
4911 8 1819 ± 20
4912 9 1594 ± 33
4913 9 1663 ± 48
4914 9 1433 ±53 1690±63
4915 10 1574±68
4916 10 1629 ±53
4916 10 1762±112
4917 10 1690± 36
4918 II 1225 ± 31 1632± 50 1583 ±59
12 I 090±200 Uncal. AD
996(1210)1302 Cal. AD
(ANU-8254)

"The raw calculated values are presented but should be rounded to the nearest decade to indicate that single digit
precision is not implied.

would be necessary to draw firmer conclusions about site use. Flake number
4888 from layer 1 gives an exposed reading of 405±139 AD and a concealed
crack reading of 1944±14 AD which indicates recent flaking percussion of
an old piece that has not suffered major surface loss. A source for this piece
218 W. R. AMBROSE

...
2
•
4
• . 6.

• •
• • &

• ·~
8
•
10
•
• 6.6.
12 1 1
4Lo~o~~s--~~~~~L_L_~_L~--L-~~~~
-t 1

00 800 1000 1200 1400 1600 1800 2000

Years AD
Figure 9.6. Comparison of the ages AD for the GBJ site based on the readings of hydration from
the protected inner crack surfaces, shown as solid squares, and the outer exposed surfaces or
corroded fissure surfaces, indicated by open triangles. The four radiocarbon dates are shown as
crossed circles.

could be the cortex from a quarry block from deep below the ground surface
away from the influence of site weathering. The young reading on 4888 and
the other two youngest flakes ( 4889 for layer l and 4892 for layer 2), all fall
within the last 50 years and are later than the abandonment of obsidian
mining earlier this century.
Three upper layer flakes ( 4889,4890, 4892) have reversed readings, with
exposed surface hydration giving ages older than those from concealed cracks,
and indicate recently struck cleavages on relatively modern flaked material.
This could be the result of modern gardening in the mining area, perhaps
related to soil disturbance during plantings of cocoa, and this activity may also
explain the presence of an old flake (4891) in the recent deposits. Other
specimens at layers 3 and 7 (4895 , 4908) indicate that in the past older flakes
have been added to the general depositional trend of the site.
OBSIDIAN HYDRATION DATING IN PAPUA, NEW GUINEA 219

The hydration dates show a steady depositional sequence, with the

occasional addition of earlier material as noted, until layer 5, when there is a
larger introduction of derived obsidian (Fig. 9. 7). Whether this is an indication
of mining after around 200 years ago or soil disturbance from some other cause
remains to be determined.
The four indicated radiocarbon dates, with two as calibrated values
determined from the program of Stuiver and Reimer (1993), give a confused
picture of site history in indicating that a recent disturbance has affected layers
5 and 8 with the appearance of "modern" dates. The result for layer 5 may be
explained by referring to the radiocarbon curve of Fig. 9.1, where periods of
dating ambiguity are indicated for the last 500 years. The "modern" radiocar-
bon date for layer 8 is less easily interpreted as the hydration date of 1320±40

•
8

10

fOl
•
12 L-_L--L--L--L--L--~-L--~_i_ _J_~--~~
800 1000 1200 1400 1600 1800 2000
Years AD
Figure 9.7. Dates calculated from hydration readings of closed fissures , with the standard devia-
tion indicated, based on the pixel reading error. Other errors introduced by inaccuracy of the site
temperature determination are not included. The calibrated radiocarbon dates for the site are
indicated as crossed circles and show a good correspondence between the two dating systems for
the lowest layers ll and 12. The "modern" radiocarbon dates for layers 5 and 8 do not reflect the
true age.
220 W. R. AMBROSE

AD indicates the period is within the useful range of radiocarbon dating. A

fuller discussion of the GBJ site in the local archaeological context will be
addressed elsewhere, but it is clear that the hydration dating record provides a
reasonable chronological basis for further analysis of site history. Further
examination of the site obsidians is necessary, with a larger sample size, to
provide a clearer pattern of site development over time and to clarify when
mining began.

CONCLUSIONS

By utilizing long term experimental hydration rate determinations, con-

cealed cracks within the protected interior of the obsidian as the target for
hydration readings, computer assisted hydration measurement and limited
local site temperature monitoring, a more reliable age estimate for the GBJ site
has been achieved than could be expected from radiocarbon dating. The
evidence from paired readings of protected and exposed hydration surfaces
shows a systematic loss of the latter that seriously limits the value of conven-
tional hydration age determinations based on exposed surfaces. Previously
reported dates from obsidian hydration dating may need to be re-evaluated, not
only as specific dates, but also where obsidian and radiocarbon pairs have been
used to set general hydration rate constants.
In obsidian hydration dating, the effects of site weathering have often
been discounted or ignored. The results from the GBJ site on Lou Island show
that the evidence for the weathering status of obsidian artifacts is readily
observable in the comparison between surface and closed fissure hydration
layers, and needs to be assessed routinely in obsidian hydration dating projects.
The rapidity of hydration development in a tropical site has allowed the effects
of weathering on obsidians to be readily observed in the recent past, but the
same effects can be expected to occur in cooler climate sites over longer time
spans, whenever soil processes support the dissolution of silica. Another
analysis of hydration dates from a different Pleistocene age site on Manus shows
that very active surface dissolution has occurred to remove surface hydration
entirely, but by focusing attention on the closed fissures a good relationship of
site age with depth is shown back to around 12,000 BP (Ambrose 1994).

ACKNOWLEDGMENTS

The owners of the Umleang site were very helpful in providing access and
support during work there. Korup, Pakop Korup, Kavon Kekes and johnny
Savo in particular gave great assistance. Many other people at Rei village also
gave help from time to time for which I am very grateful. For help in the field
at the site I acknowledge the assistance of]. Kennedy, R. Fullager and]. Allen.
OBSIDIAN HYDRATION DATING IN PAPUA, NEW GUINEA 221

REFERENCES

Ambrose, W.R. 1984 Soil temperature monitoring at Lake Mungo, implications for racemisation
dating. Australian Archaeology 19:64-74.
_ _ 1993 Obsidian hydration dating, In Fankhauser B.L., and Bird j.R., eds., Archaeometry:
current Australasian research. Canberra, Prehistory Department, Australian National Uni-
versity: 79-84.
_ _ 1994 Obsidian hydration dating of a Pleistocene age site from the Manus Islands, Papua New
Guinea. Quaternary Geochronology 13(2): 137-142.
Bates, j.K., Abrajano, T.A., Jr., Ebert, WL., Mazer, J.j. and Gerding, T.j. 1988 Experimental
hydration studies of natural and synthetic glasses. In Sayre, E.V, Vandiver, P.B., Druzik].
and Stevenson C., eds., Materials issues in art and archaeology. Pittsburgh, Pennsylvania.
Materials Research Society Symposium Proceedings, 123: 237-244.
Casey, WH., Eggleston, C., Johnsson, P.A., Westrich, H.R. and Hochella, M.E Jr., 1992 Aqueous
surface chemistry and corrosion of minerals. Materials Research Society Bulletin,
17(5):23-29.
Curti, E., Godon, N. and Vernaz, E.Y. 1993 Enhancement of glass corrosion in the presence of clay
minerals: testing experimental results with an integrated glass dissolution model. In
Interrante C. G., and Pabalan R.T., eds., Scientific basis for nuclear waste management XVI.
Pittsburgh, Pennsylvania, Materials Research Society: 163-170.
Duerden, P., Cohen, D.D. and Ambrose, WR. 1992 The measurement of hydration profiles in
obsidian. In Ambrose W. and Duerden, Peter, eds.,Archaeometry: an Australasian Perspective.
Canberra, Prehistory Department, Australian National University: 236-242.
Friedman, I. and Smith, R.L. 1960 A new dating method using obsidian: part 1, the development
of the method. American Antiquity 25(4): 476-522.
Friedman, 1., Smith R.L. and Long, WD. 1966 Hydration of natural glass and the formation of
perlite. Geological Society of America Bulletin 25( 4 ): 323-328.
Fullagar, R. and Torrence, R. 1991 Obsidian exploitation at Umleang, Lou Island. In Allen]. and
Gosden C., eds., Report of the Lapita Homeland Project. Canberra, Prehistory Department,
Australian National University: 113-143.
Hughes, M.] and Oddy WW 1970 A reappraisal of the specific gravity method for the analysis of
gold alloys Archaeometry 12(1): 1-11.
Lee, R., Leich, D.A., Tombrello, T.A., Ericsonj.E. and Friedman I. 1974 Obsidian hydration profile
measurements using a nuclear reaction technique. Nature 250: 44-4 7.
Marshall, WL. 1980 Amorphous silica solubilities-l. Behaviour in aqueous sodium solutions;
25-300°C, 0-6 molal. Geochimica et Cosmochimica Acta 44: 907-913.
McAlpine,]. R., Keig, G. and Falls, R. 1983 Climate of Papua New Guinea. Canberra, Commonwealth
Scientific and Industrial Research Organisation and Australian National University Press.
Stevenson, C.M., and Scheetz, B.E. 1989 Induced hydration rate development of obsidians from
the Coso volcanic field: a comparison of experimental procedures. In Hughes R.E., ed.,
Current directions in California obsidian studies. Contributions of the California Archaeological
Research Facility. California, Berkeley, Dept. of Anthropology, University of California 48:
23-30.
Stuiver, M, and Reimer, P.j. 1993 Extended He database and revised CALIB radiocarbon calibration
program. Radiocarbon 35: 215-230.
Thorseth, I. H., Fumes, H. and Heldal, M. 1992 The importance of microbiological activity in the
alteration of basaltic glass. Geochimica et Cosmochimica Acta 56: 845-50.
Tremaine, K.]. and Fredericksen D.A. 1988 Induced obsidian hydration experiments: an investi-
gation into relative dating. In Sayre, E.V, Vandiver, P.B., Druzik]. and Stevenson C., eds.,
Materials issues in art and archaeology. Pittsburgh, Pennsylvania, Materials Research Society
Symposium Proceedings 123: 271-278.
222 W. R. AMBROSE

White, W. B., 1988 Glass hydration mechanisms with application to obsidian hydration dating. In
Sayre, E.V., Vandiver, P.B., Druzik,]. and Stevenson C., eds., Materials issues in art and
archaeology. Pittsburgh, Pennsylvania, Materials Research Society Symposium Proceedings
123: 225-236.
 Chapter 10

A 1990s Perspective on
Method and Theory in
Archaeological Volcanic Glass
Studies
R. c. GREEN

ABSTRACT

The commentary of this overview stems from reading seven papers in this
volume on sourcing volcanic glasses found in archaeologial sites in a range of
geographic regions throughout the world and two on using rhyolitic obsidian
for hydration rim dating. One perspective developed from this exercise is
historical and personal relating to the progress made in the last three decades
in the field. Another highlight some of the variation in termonology applied by
different analysts when speaking of sources and the material in them. A third
has to do with reasonably successful procedures and techniques now being
employed in both sourcing and independent age determinations from hydra-
tion rims. Finally, comments are directed at the need to connect these studies
with cultural analyses and interpretations which focus on the debris from the
source localities and whole site assemblages.

R. C. GREEN • Department of Anthropology, University of Auckland, Auckland, New Zealand.

Archaeological Obsidian Studies, edited by Shackley.
Plenum Press, New York, 1998.
223
224 R. C. GREEN

INTRODUCTION

The papers in this volume cover current aspects of the state of play in
volcanic glass studies in several parts of the world. Highlighted are advances
in strategies for sourcing and age determinations; much less is said about
parallel developments in the studies of debitage in or near ancient quarries or
source localities, and there is little discussion on production techniques for tool
making, qualities of the various volcanic glasses for different purposes, or their
use and function of objects in these materials within cultural systems. I will
return to these latter topics at the end, but accept that the thrust of my
commentary must focus on the first two subjects.

A PERSONAL BACKGROUND PERSPECTIVE

When, in 1958-59, I first seriously began to investigate the potential of

volcanic glass studies for archaeology in New Zealand, and subsequently for
the rest of Oceania, method and theory in the field was minimal. In the field
of sourcing, the initial stimulation was the New Mexican work of Boyer and
Robinson (1956) on the refractive index, and my original attempts began with
that technique, and with the aim of showing that not all New Zealand obsidian
came from a single source on Mayor Island (Duff 1956:231). Shortly thereafter
I learned of Friedman and Smith's (1960) work on a new method of dating
using obsidian, and then through Graham Clark of an about to be published
paper of Cann and Renfrew (1964) on more powerful methods (optical emis-
sion spectroscopy) of sourcing.
This lead to a series of papers (Green 1962, 1964, Green et al. 1967,
Ambrose and Green 1962) which explored the potential of these techniques
for some 12 source locations in New Zealand and then in the early 1970s
extended them to Oceania (Ambrose and Green 1972, Green 1991:202) where
source locations were still poorly known. By the late 1970s as a result of a
series of investigations the number of potential individual obsidian source
localities in New Zealand had risen to circa 42 different locations where
obsidian could be found naturally, and clustered together into 18 petrographi-
cally distinguishable sources on the basis of their trace element chemistry
(Ward 1973, 1974a, 1974b). By the early 1980's the techniques for sourcing
which had been explored ranged from colour in transmitted light and refrac-
tive index (Green 1964, Leach and Anderson 1978), through density to a
wide variety of techniques of elemental analysis (Reeves and Ward 1978:
Table 14.1, Leach and Fankhauser 1978, Leach and Warren 1981). Similar
developments in sourcing had taken place in Oceania, and many of the
numerous source localities in that region too had now been identified (Smith
et al. 1977, Bird eta!. 1978, 198la, 1981b).
I 9905 PERSPECTIVE ON VOLCANIC GlASS STUDIES 225

Major problems, however, had been encountered with obsidian hydration

dating in both New Zealand (Leach 1977, Reeves and Ward 1978:278, Leach
and Naylor 1981, Leach and Hamell984) and Oceania (Green 1991:202). The
same thing occurred with the use alteration rims on the basaltic glasses of
Hawaii as a means for age estimates (Graves and Ladefoged 1991). Thus,
despite a general protocol outlined by Ambrose (1976) for the development of
an intrinsic system of obsidian hydration dating system independent of any
radiocarbon age determinations, its potential remained unfulfilled.
So far as I can tell, most geographic regions of the world where volcanic
glass studies have been consistently pursued exhibit a similar history. In this
volume, Tykot provides a summary of current advances within the last 20 years
in the Western Mediterranean area, Glascock et al. do the same for two regions
in Mesoamerica, and Summerhayes et al. provide a similar outline of progress
in understanding the West New Britain region in the Bismarck Archipelago of
Oceania. In each case the number of individual source localities within each
region has multiplied dramatically, as often through the efforts of archaeolo-
gists, as through those of the geologists. Moreover, techniques for grouping
these source locations geochemically, allowing their fairly precise charac-
terization, have improved markedly, and with an ever increasing degree of
confidence in the acceptance of the results. In contrast, the uptake of obsidian
hydration dating has proved a more variable enterprise, with the result that
among dating methods, it was still in the mid-1980's, assigned to an "experi-
mental" as opposed to an operational procedure (Taylor 1987:318, Table 1).
The exception, perhaps, has been in North America and especially in California
(Meighan et al. 1974, Meighan and Vanderhoeven 1978, Meighan and Scalise
1988). Thus, as Ambrose (1993:79) has observed, while this means of dating
may in fact be responsible for a larger number of age estimations than any other
single method, "the zeal with which the method has been employed is not
always accompanied by widespread acceptance of the results that emerge."

VOLCANIC GLASS AS A CATEGORY

Given this background, what general issues does this volume of papers
raise for which various resolutions are offered, and what are some of the
problems that remain? First, given the title of the volume, it is apparent that
the general category under discussion is broadly volcanic glass, and not just
rhyolitic obsidian, its most prominent member over much of the world. For
instance, Weisler and Clague use volcanic glass as the general label, and
obsidian as one of its sub-classes in which there is a high silica content, and
basaltic glass as a another grouping for glassy materials with a more or less
basaltic composition. In a similar vein, one can compositionally distinguish
between peralkaline and rhyolitic obsidians based on the higher Fe content
226 R. C. GREEN

typical of the former as Glascock et al. indicate, as well as identify an often

culturally less useful water enriched form of glassy material derived from
geologically ancient rhyolitic obsidian and known as perlite in which cultur-
ally useful less hydrated nodules occur. Thus one can often distinguish,
sometimes visually as Weisler and Clague do for Pitcairn, or in combination
with density, between rhyolitic obsidians, trachytic cone and dome obsidians,
ignimbrite or welded tuff obsidians, marekenites, and a multitude of types
of glassy volcanic selvages which occur on dikes and glacially or water chilled
lava flows as in Hawaii. And if one employs a geological perspective, based
on composition, there are several types of obsidian ranging from rhyolite
through andesite and trachyte to phonolite, to all of which the cover term
volcanic glass applies.
In Remote Oceania at least, the volcanic glasses of sources in the Banks
Islands, Tonga, American Samoa, the Marquesas, and Hawaii are just as impor-
tant as are the rhyolitic obsidians of Near Oceania, Easter Island, and New
Zealand. Moreover from a cultural perspective, they were probably not distin-
guished; all were useful, used, and frequently exist under a single vernacular
taxon. Thus if one's perspective is wider than just the high-silicate obsidians,
then Glascock et al. need to take more care in their statements about the
impossibility of glass formation in Hawaii and with their failure to cite the
whole expanse of Oceania (and not just New Zealand) as another geographic
zone in which a range of glasses including rhyolitic obsidians are an important
human resource.
Finally, these geologically based distinctions among the volcanic glasses
used by humans are important to the strategies employed in their sourcing and
dating. For example, as Weisler and Clague suggest, the situation beyond the
continental side of the Andesite Line on Oceanic islands, with a huge number
of potential glass sources, may require a somewhat different approach to
sourcing. This certainly applies in Hawaii and Pitcairn, and they have outlined
some possible procedures to deal with those regions different from continental
areas. Equally important, in the use of alteration rims for age estimates, the
quite different amounts of intrinsic water in the various kinds of volcanic
glasses, as well as differences in Fe content, means that hydration dating in
rhyolitic obsidian will probably significantly differ from that of the surface
alteration of basaltic glass which Friedman et al. (1997) describe as being "more
akin to the rusting of iron".

SPATIAL AND TEMPORAL VARIATION IN "SOURCE"

TERMINOLOGY

In reading these papers, it becomes apparent that there is still a fair degree
of variation in the terminology employed by authors when describing different
1990s PERSPECTIVE ON VOLCANIC GlASS STUDIES 227

levels of "source" discrimination. At the lowest level of discrimination geolo-

gists often speak of very broad areas or source systems and at high levels of
resolution of source localities, with the distinctions of source subsystems and
locality complexes lying somewhere in between. In contrast Summerhayes et al.
are very explicit about what they mean by geographic regions with a number of
sources, and source localities as specific sampling loci where naturally occurring
obsidian specimens were collected. All source localities within a geographic
region are for them a regional group, within which similarities in chemical
composition make it possible to distinguish subgroups or "chemical groups".
Other authors, such as Glascock et al. also speak of geographic regions, but
they also talk of subregions, source areas and complex source areas, and
chemical or compositional subgroups for these, while Shackley speaks of four
distinct chemical groups for a named source region.
In an ascending hierarchy, the concepts of source locality, chemical or
compositional subgroups, regional source groups, both simple and complex,
and broad geographic regions seem to be some of the most productive discrimi-
nations employed by archaeologists. In contrast, when trying to understand the
geological and compositional relationships between chemical subgroups,
source systems and subsystems might be more applicable terminology.
Most workers, I believe, also find it essential to distinguish between
primary and!or outcrop sources and secondary or detrital and erosionally
distributed sources. Shackley highlights how important this second category
and its consideration may become in certain geographic regions with extreme
secondary deposition as in the southern Southwest U.S.A. But my experience
of source localities in New Zealand and the rest of Oceania leads me to believe
he is right in saying that determining the primary region of a given obsidian
source is only a small part of the research; a strategy must be generated that
takes into account its ultimate secondary distribution as well as the potential
for chemical variability that occurs within it. Humans as often used what nature
had handily provided as secondary products, as they did by quarrying volcanic
glasses from an outcrop. And for archaeologists this has a lot to do with the
outcome when they seek to understand those factors which have structured
ease of access, and the exploitation and production of portable cultural objects
made in these materials, as well as characterizing the size of the dot which
pinpoints its supposed origin. Sheppard (1996) in discussing the problem with
respect to chert, usefully refers to point sources with high concentrations of
the desired materials within a confined location and an abundance of source
localities within it, and a series of ever larger areas where both the spatial
concentrations of the desired material and the abundance of localities yielding
it markedly declines. The chapter by Summerhayes et al. for the Mopir and
Willaumez regional groups of West New Britain nicely outlines this situation
and again reinforces how important secondary contexts become when assessing
extraction processes in respect to availability. As Shackley suggests, they can
228 R. C. GREEN

have a great effect on the validity of inferences about exchange, interaction and
social processes in some regions.
Finally a number of papers (Summerhayes et al., Weisler and Clague;
Shackley) address the issue of the "impermanence" of volcanic glass sources
and especially ofits source localities (Ambrose et al. 1981). For some regions
of the Pacific as in Hawaii, American Samoa, the Willaumez Peninsula, Fergus-
son Island, and the Admiralties, the inability to identify a primary source is a
serious though not insurmountable issue. This also applies to the precise
geological origin of the Cow Canyon Tertiary source in the Southwestern
United States. At other times sources like Mopir in New Britain go in and out
of use and then in again, effected by volcanic activity and the nature of the
exchange relationship. It is not just that there are some geologically known
sources of volcanic glass which were never in use, or were not available for use,
or proved unsuitable for use, it is also that a number of useful chemical
subgroups known from site assemblages imply former sources that exhibit
various degrees of permanence through time which have effected their exploi-
tation by people in the past. Thus we may be able to locate their likely origin
only provisionally by assigning it to a regional group, rather than to a specific
source locality, or cluster of localities where it was actually obtained.

PROCEDURES AND TECHNIQUES FOR SOURCE

CHARACTERIZATION

The analytical methods discussed in this volume, as Shackley indicates,

have now become more than adequate for source assignment, as long as they
measure incompatible elements with a fair degree of certainty and are calibrated
to international standards. Four techniques stand out here among a wide
number of methods available. One is neutron activiation analysis (Glascock et
al.) which measures a greater number of elements with more sensitivity and
precision than x-ray fluorescence (XRF), and is only minimally destructive of
artifact samples. Another comprises various kinds of XRF analysis which can
either be non-destructive (Weisler and Clague, Davis et al.), although geolo-
gists and some archaeologists also use destructive processes with pellets to
increase precision and sensitivity. A third is the non-destructive PIXE-PIGME
technique currently largely employed in the Oceanic region (Summerhayes et
al.), and the last is a relatively more recent and less common technique of
ICP-MS requiring only a tiny powdered sample (Tykot) applied to the obsidian
sources in Sardinia. In this case the number of elements which can be accurately
and precisely analysed with little or no sample preparation makes it highly
attractive. It seems, after a long period of experimentation with a wide range
of techniques, sophisticated methods of employing a few of them to good effect
is being achieved.
19905 PERSPECTIVE ON VOLCANIC GLASS STUDIES 229

More importantly, all the archaeological players in this volume seem

concerned with moving from more expensive to less expensive strategies with
higher rates of throughput that will allow assemblages of artifacts rather than
individual objects, or other small samples of archaeological collections, to be
studied. As I have observed on several occasions (Green 1987:241, Green and
Bird 1989: footnote 2), even if sophisticated elemental analysis seems theoreti-
cally and procedurally possible, stratified sampling techniques, and a range of
methods from the visual, to density, to abbreviated forms of elemental analysis,
are going to have to be developed by archaeologists if they wish to achieve a
sufficient understanding of the range and frequency of source subgroups within
the large cultural assemblages they often recover. One is therefore encouraged
by statements like that of Tykot that it was discovered that obsidian artifacts
from Sardinia could be visually assigned a provenance with greater than 85%
accuracy, or that visual variability was 100% successful in separating the Down
Rope, Pitcairn ignimbrite volcanic glass from other Polynesian volcanic glasses,
and that in the Mussau group of the Admiralties rapidly established density
groupings plus EDXRF proved an inexpensive but powerful procedure for
accurately sorting large groups of obsidian artifacts and source materials using
minimal sample preparation plus a non-destructive method of examination
(Marshall and Clague). Finally, if one must employ more powerful and destruc-
tive methods, Glascock et al. have devised a systematic and rigorous series of
steps, which once a preliminary analysis has been done, and cut-off points
established, shows that only about 10% of all Mexico and Guatemala are likely
to require the full NAA treatment. Otherwise a time-saving abbreviated-NAA
procedure is now, they claim, sufficiently accurate and very cost effective when
compared to XRF, and thus applicable to processing a far greater number of
specimens.
Occasional unconvincing results have helped to identify a number of
potential sources of variation in PIXE-PIGME data-changes in the equipment
used, the use of different standards for calibration, the effects of roughness,
hydration and leaching and natural inhomogeneities within an obsidian flow
(Green and Bird 1989:95). Where they have been observed, they are often
controlled for by re-runs or multiple runs of the same specimen. Some of these
cases are discussed by Summerhayes et al. including one where the angle of the
presentation of a flat surface to the beam is also advanced as a possible factor
for spurious data, particularly in respect of unassigned artifacts. In a similar
fashion, Davis et al. discuss rather similar problems of physical factors, espe-
cially size, thickness, and surface configuration in respect to non-destructive
EDXRF analysis of obsidian. More importantly they conduct a number of
experiments to test the extent to which such factors become relevant and
conclude that in general they do not constitute a major source of error. Glascock
et al. cite the case of Chatham Island artifacts as another example yielding
ambiguous outcomes, this time due to XRF versus NAA techniques of data
230 R. C. GREEN

collection and the methods used in its statistical analysis. They have the
Chatham Island artifacts coming from Easter Island, 7000 km to the east, and
this in Pacific prehistory would indeed be a most dramatic outcome were it
true. More recent investigations using the PIXE-PIGME technique and yet
other statistical methods have shown it is not. They come, as the initial analysis
suggested, from Mayor Island in New Zealand (Leach et al. 1986).
Under the heading for this section one should also applaud the very useful
discussion by Glascock et al. of source discrimination and of attribution of
artifact specimen to source, usually by statistical means, and most frequently
some form of multivariate analysis. In my view it is an important summary of
the major problems which have been encountered so far in developing some
satisfactory protocols for this process. The issue is also raised by Shackley, who
warns of certain conditions that must be met when using Mahalanobis distance
measures, if misclassifications are to be avoided, and he suggests a preference
for combining any multivariate analysis with graphic displays. The major
concern I see here is the use of procedures which pick out and identify the
outlier specimens, and leave these as unknowns rather than assigning them to
groups or envelopes possessing a somewhat greater dispersion.

OTHER ASPECTS

Overall, if these seven papers on sourcing are any guide, then fairly
sophisticated paradigms for the sourcing of volcanic glass are now in hand, and
archaeologists are presently doing the intensive fieldwork on source locality
relationships necessary to understand from whence these glassy materials have
arrived in their sites. The next step, as Tykot observes, is the generation of
models and even more detailed studies which focus on procurement and
extraction processes as well as on the mechanisms for the movement of raw
materials. But we cannot stop there as we must also tackle the transformation
of raw material into a finished product, explain how and why it was used and
re-used, and finally relate this to its discard. Where we have taken all of these
steps not only for sourcing whole assemblages but analysing them technologi-
cally and distributionally within sites, as in the Reef/Santa Cruz example of the
Solomons, strong arguments have been developed that these exchange net-
works were structured more by social than simple economic and distance
considerations (Green 1987, Sheppard and Green 1991, Sheppard 1993).
Scattered through the papers in this volume as well, are similar sentiments
about social factors and social links as being of equal or greater importance
(Shackley, Tykot, Summerhayes et al.). I particularly appreciated the viewpoint
expressed by Tykot that "It is apparent that obsidian served a social as well as
functional role, perhaps as an ethnic identifier for populations culturally
diversifying due to increased sedentism". Aspects of that statement fit well with
19905 PERSPECTIVE ON VOLCAN1C GLASS STUDIES 231

one possible interpretation for the Reef/Santa Cruz case, except that "seden-
tism" would be replaced with "colonization" (see for example Kirch 1991:159).

PROBLEM OF AGE DETERMINATION THROUGH OBSIDIAN

HYDRATION MEASUREMENTS

Perhaps it is my location in Auckland where its Centre for Archaeological

Research has recently been actively engaged in exploring and upgrading this
technique and its application in New Zealand (Sutton and Sheppard 1994,
Stevenson et al. 1996), or my involvement with Wal Ambrose, and a keen
interest in his long-standing attempts to improve the whole intrinsic obsidian
dating process, but I am more than a little enthused by the papers of Stevenson
et al. and Ambrose. In my opinion they are an indication that the whole process
of obsidian hydration dating is currently moving from a mature experimental
phase to an operational method in which we can at last have some confidence.
The reason for this view is that a number of well-known problems with the
method seem to have been overcome, so that we can claim, as Ambrose does
"that obsidian hydration dating is a viable dating system free of dependence on
other dating methods to provide its time constants".
These are the issues for which I think these two papers signal important
developments. First, as Stevenson et al. say, it is now apparent that obsidian
hydration rates need to be determined for individual artifacts based on their
composition. The dominant factor is not the elemental variation in the
chemical composition of the sources and subsources to which each piece
needs to be traced (Glascock et al., this volume and Friedman et al. 1997)
as has frequently been assumed from previous research. Rather it is the
original or structural water content of each specimen or flake that matters,
especially in light of the knowledge that volcanic glass flows were in fact
variable with respect to this controlling variable. Stevenson et al. propose
here that OR is the dominant factor which controls the hydration process,
one which can be modelled as involving molecular water diffusion rather
than depletion and replacement of ions.
Next, as Ambrose observes, density has been shown in Papua New Guinea
sites to be correlated with the hydration rate. From this observation it has more
recently been possible to show that high density obsidians will have low
quantities of structural water and hydrate slowly while low density glasses will
have higher OR concentrations and faster hydration rates (Stevenson et al
1996, Figure 4). Thus relative or more precise density measurements, which
some have previously found useful in preliminary sorting and sampling strate-
gies when doing sourcing studies, now also become valuable in the estimation
of hydration rate ranges for obsidian chemical subgroups, and precise densities
become crucial for determining the rate to be applied to each individual piece.
232 R. C. GREEN

It also seems from Stevenson et al's work that relative humidity in a site
needs to be assessed to determine at what depth it becomes a constant that will
not effect the hydration rate.
Moving to Ambrose's paper, the call by Stevenson et al. for the develop-
ment of better instrumentation and perhaps non-optical measurement tech-
niques, has in part been answered. Ambrose (1994 and this volume) has found
that because of differential and unpredictable but aggressive weathering on the
external surfaces of flakes from two sites in the tropics, a phenomenon also
found in New Zealand (Stevenson et al. 1996), there is often a necessity to focus
on internal or concealed crack surfaces or protected fissures to obtain a
consistent series of hydration readings. This can then be combined with
computer assisted imaging technology providing an integrated hydration meas-
urement over a band of the optical image resulting in an average width value
for the measured area. In this way operator precision of the measurement is
greatly enhanced. The two procedures lead to very high quality hydration rim
measurements.
These measurements may then be interpreted as age estimates by apply-
ing long term experimental hydration rate determinations to yield acceptable
dates independent of other methods. Although Stevenson et al. (1996) have
successfully used induced or high temperature hydration rate determinations
to achieve acceptable dating results for specimens from the Mayor Island, Kaeo,
and Pungaere sources in New Zealand, their recommendations in this volume
is that a new experimental procedure be adopted, one that more closely models
the natural hydration process under ambient conditions. This viewpoint seems
in keeping with their own recent work and that of Ambrose. Still, it would
appear that at last methods for obtaining truly independent dates based on fairly
reliable rate constants for subsource groupings with narrow structural water
content ranges are now available.
Finally as Stevenson et al. describe, the determination of likely tempera-
tures at which the obsidian artifacts whose hydration rims are being measured
were stored, is being more closely monitored and modelled. At present the area
of EHT remains one where there is still quite a way to go; however, the problem
is continually being addressed to greater effect as overall site histories and their
taphonomy become better understood.
In summary, one might now predict that at last we will be able to
adequately and independently date the last 300 years of human history, where
radiocarbon dating fails us. While this will apply in a number of regions where
such information is vital, obsidian hydration dating, because it spans much
longer periods, is currently threatening to move into the ranks of a fully
operational, acceptable, and reasonably understood means of age estimation in
the many regions in which that material occurs. Overall obsidian studies are
now realizing the potential that they appeared to hold when some of us began
in the business some 30 years ago.
19905 PERSPECTIVE ON VOLCANIC GLASS STUDIES 233

REFERENCES

Ambrose, WR. 1976 Intrinsic hydration rate dating of obsidian. In R.E. Taylor, ed., Advances in
Obsidian Glass Studies, Archaeological and Geochemical Perspectives. Park Ridge, N.] ., Noyes
Press: 81-105.
Ambrose, WR. 1993 Obsidian hydration dating. In Fankhauser, B.L. and Bird, ].R., eds., Ar-
chaeometry: Current Australasian Research. Canberra, Research School of Pacific Studies,
Australian National University. Occasional Papers in Prehistory 22: 79--84.
Ambrose, WR. 1994 Obsidian hydration dating of a Pleistocene age site from the Manus Islands,
Papua New Guinea. Quaternary Geochronology (Quaternary Science Reviews) 13: 137--42.
Ambrose, W.R., Bird, ].R., and Duerden, P. 1981 The impermanence of obsidian sources in
Melanesia. In Leach, B.F and Davidson, ].M. eds, Archaeological Studies of Pacific Stone
Resources. Oxford, BAR International Series, 104: 1-19.
Ambrose, W.R. and Green, R.C. 1962 Obsidian dating: Preliminary results. New Zealand Archae-
ological Association Newsletter 5: 24 7-48.
Ambrose, W.R. and Green, R.C. 1972 First millennium B.C. transport of obsidian from New Britain
to the Solomon Islands. Nature 237: 31.
Bird, JR., Ambrose, W.R., Russell, L.H., and Scott, M.D. 1981a The Characterisation of Melanesian
Obsidian Sources and Artefacts Using Proton Induced Gamma-Ray Emission (PIGME) Tech-
nique. AAEG/E510. Lucas Heights, NSW, Australian Atomic Energy Commission.
Bird, ].R., Duerden, P., Ambrose, WR., and Leach, B. F. 1981b Pacific obsidian catalogue. In B. F.
Leach and ].M. Davidson (eds), Archaeological Studies of Pacific Stone Resources. Oxford,
BAR International Series 104: 31-43.
Bird, ].R., Russell, L.H., Scott, M.D. and Ambrose, WR. 1978 Obsidian characterisation with
elemental analysis by proton induced gamma-ray emission. Analytical Chemistry 50:
2082--84.
Boyer, W.W. and Robinson, P. 1956 Obsidian artifacts of Northwestern New Mexico and their
correlation with source material. El Palacio, 63: 333--45.
Cann, ].R. and Renfrew, A.C. 1964 The characterization of obsidian and its application to the
Mediterranean region. Proceedings of the Prehistoric Society 30: ll1-33.
Duff, R.S. 1956 The Moa-hunter Period of Maori Culture. 2nd edition. Wellington, Government
Printer.
Friedman, I. and Smith, R.L. 1960 A new dating method using obsidian. Part 1, the development
of the method. American Antiquity 25:4 76-93.
Friedman, 1., Trembour, F.W, and Hughes. R.E. 1997 Obsidian hydration dating. In Taylor, R. E.
and Aitken, M., eds., Chronometric Dating in Archaeology. New York, Plenum Press.
Graves, M.W. and Ladefoged, TN. 1991 The disparity between radiocarbon and volcanic glass
dates: New evidence from the island ofLana'i, Hawai'i. Archaeology in Oceania 26: 70-77.
Green, R.C., 1962 Obsidian, its application to archaeology. New Zealand Archaeological Association
Newsletter 5: 8-16.
_ _ 1964 Sources, ages and exploitation of New Zealand obsidian: An interim report. New
Zealand Archaeological Association Newsletter 7: 134-43.
_ _ 1987 Obsidian results from La pita sites of the Reef/Santa Cruz Islands. In W R. Ambrose and
Mummery, ].M.]., eds., Archaeometry: Further Studies in Australasia. Canberra. Research
School of Pacific Studies, Australian National University. Occasional Papers in Prehistory
14: 239--49.
_ _ 1991 A reappraisal of the dating for some Lapita sites in the Reef/Santa Cruz Group of the
Southwest Solomons. journal of the Polynesian Society 100: 197-207.
Green, R.C. and Bird, JR. 1989 Fergusson Island obsidian from the D'Entrecasteaux Group in a
Reef Islands Lapita site from the Santa Cruz group. New Zealand journal of Archaeology 11:
87-99.
234 R. C. GREEN

Green, R.C., Brooks, R.R. and Reeves, R.D. 1967 Characterisation of New Zealand obsidians by
emission spectroscopy. New Zealand journal of Science 10: 675-82.
Kirch, P.V., 1991 Prehistoric exchange in Western Melanesia. Annual Review of Anthropology 20:
141-65.
Leach, B.E 1977 New perspectives on dating obsidian artefacts from New Zealand. New Zealand
journal of Science 20: 123-38.
Leach, B.E, and Anderson A.]. 1978 The prehistoric sources of Palliser Bay obsidian. journal of
Archaeological Research 5: 301-307.
Leach, B.E, Anderson, A.]., Sutton, D.G., Bird, JR., Duerden, P., and Clayton, E. 1986 The origin
of prehistoric obsidian artefacts from the Chatham and Kermadec Islands. New Zealand
journal of Archaeology 8: 143-70.
Leach, B.E and Fankhauser, B. 1978 The characterization of New Zealand obsidian sources using
thermoluminescence. journal of the Royal Society of New Zealand 8: 331-4 2.
Leach, B.E and Hamel, G.E. 1984 The influence of archaeological soil temperatures on obsidian
dating in New Zealand. New Zealand journal of Science 27: 399-408.
Leach, B.E and Naylor, H. 1981 Dating New Zealand obsidians by resonant nuclear reactions. New
Zealand journal of Archaeology 3: 33-49.
Leach, B.E and Warren, S.E. 1981 Neutron activation analysis of New Zealand and Oceanic
obsidians: Towards a simple screening technique. In Leach, B.E andDavidson, j.M. eds.,
Archaeological Studies of Pacific Stone Resources. Oxford, BAR International Series 104:
151-66.
Meighan, C. E., Findlow, EE and De Atley, S.P. 1974 Obsidian dates 1: A compendium of the obsidian
hydration determinations made at the UCLA Obsidian Hydration Laboratory, Institute of
Archaeology Archaeological Survey. Archaeological Survey Monograph 3. Los Angeles,
University of California.
Meighan, C. E. and Scalise, J.L. 1988 Obsidian dates IV: A compendium of the obsidian hydration
determinations made at the UCLA Obsidian Hydration Laboratory, Institute of Archaeology
Archaeological Survey. Archaeological Survey Monograph 24. Los Angeles, University of
California.
Meighan, C.E. and Vanderhoeven, P.l. 1978 Obsidian dates II: A compendium of the obsidian
hydration determinations made at the UCLA Obsidian Hydration Laboratory, Institute of
Archaeology Archaeological Survey. Archaeological Survey Monograph 6. Los Angeles:
University of California.
Reeves, R.D. and Ward, G.K. 1976 Characterization studies of New Zealand obsidians: Towards a
regional prehistory. In Taylor, R.E., ed., Advances in Obsidian Glass Studies: Archaeological
and Geochemical Perspectives. Park Ridge, New Jersey, Noyes Press: 259-87.
Sheppard, PJ ., 1993 Lapita lithics: Trade/exchange and technology. A view from the Reefs/Santa
Cruz. Archaeology in Oceania 28: 121-37.
_ _ 1996 Hard rock: Archaeological implications of chert sourcing in Near and Remote Oceania.
In Janet, M., Davidson, M., Irwin, G.]., Leach, E, Pawley, A., and Brown, D., eds., Oceanic
Culture History: Essays in Honour of Roger Green. Dunedin: New Zealand Journal of
Archaeology Special Publication: 99-115.
Sheppard, P.J. and Green, R.C. 1991 Spatial analysis of the Nenumbo (SE-RF-2) Lapita site,
Solomon Islands. Archaeology in Oceania 26: 89-101.
Smith, l.E.M., Ward, G.K. and Ambrose, WR. 1977 Geographic distribution and the charac-
terization of volcanic glasses in Oceania. Archaeology and Physical Anthropology in Oceania
12: 173-20 l.
Stevenson, C.M., Sheppard, P.j., Sutton, D.G. and Ambrose, WR. 1996 Advances in the hydration
dating of New Zealand obsidian. journal of Archaeological Science 23: 233-24 2.
Sutton, D.G. and Sheppard, P.J. 1994 Dating New Zealand prehistory using obsidian hydration.
Archaeology in New Zealand 37: 272-81.
19905 PERSPECTIVE ON VOLCANIC GLASS STUDIES 235

Taylor, R.E., 1987 Dating techniques in archaeology and paleo-anthropology. Analytical Chemistry
59: 317A-31A.
Ward, G.K., 1973 Obsidian source localities in the North Island of New Zealand. New Zealand
Archaeological Newsletter 16: 85-103.
Ward, G. K., 197 4a. A paradigm for sourcing New Zealand archaeological obsidians. journal of the
Royal Society of New Zealand 4: 47-62.
_ _ 197 4b. A systematic approach to the definition of sources of raw material. Archaeometry 16:
41-53.
 About the Editor
M. Steven Shackley is Associate Research Archaeologist and Adjunct Associate
Professor in the Phoebe Hearst Museum of Anthropology and Department of
Anthropology, and Director of the Archaeological EDXRF Laboratory, Univer-
sity of California, Berkeley. He has been involved in and directed a number of
archaeometric and archaeological projects in western North America, particu-
larly the American Southwest and northern Mexico.

237
 Correspondence

W. R. Ambrose Robert H. Cobean

Division of Archaeology and Natural Subdirecci6n de Estudios
History Arqueol6gicos, INAH
Research School of Pacific and Asian Lie. Verdad No. 3, col Centro
Studies CP 06060, Mexico D.F
Australian National University
Canberra, ACT 0200, Australia M. Kathleen Davis
Northwest Research
Roger Bird 1414 NW Polk
Australian Nuclear Science and Corvallis, Oregon 97330
Technology Organisation
Private Bag Richard Fnllagar
Menai, NSW 2234, Australia Division of Anthropology
Australian Museum
Sydney, NSW 2000, Australia
Geoffrey E. Braswell
Department of Anthropology
Michael D. Glascock
State University of New York
Research Reactor Facility
Buffalo, NY 14222-1095 University of Missouri
Columbia, MO 65211
David Clague
Monterey Bay Aquarium Chris Gosden
Research Institute Pitt Rivers Museum
P. 0. Box 628 University of Oxford
7700 Sandholdt Road 64 Banbury Road
Moss Landing, CA 95039-0628 Oxford OX2 6PN, United Kingdom
239
240 CORRESPONDENCE

Roger C. Green jim Specht

Research Associates (Pacifica) Ltd. Division of Anthropology
P. 0. Box 60-054 Australian Museum
Titirangi Sydney, NSW 2000, Australia
Auckland 7, New Zealand
Christopher M. Stevenson
Joachim Hampel Archaeological Services Consultants
301 McCone Hall 4620 Indianola
Department of Geology and Columbus, OH 43214
Geophysics
University of California Glenn Summerhayes
Berkeley, CA 94720-4767 Department of Archaeology
La Trobe University
Thomas L. jackson Bundoora, Victoria 3083, Australia
Pacific Legacy
Timothy Teague
10096 Soquel Drive #3
Department of Geology and Geophysics
Aptos, CA 95003
301 McCone Hall
University of California
james j. Mazer
Berkeley, CA 94720-4767
Chemical Technology Division
Argonne National Laboratory Robin Torrence
Argonne, Illinois Division of Anthropology
Australian Museum
Barry E. Scheetz Sydney, NSW 2000, Australia
Materials Research Laboratory
Pennsylvania State University Robert H. Tykot
University Park, PA 19104 Department of Anthropology, SOC 107
University of South Florida
M. Steven Shackley Tampa, FL 33620-8100
Phoebe Hearst Museum of
Anthropology Marshall I. Weisler
103 Kroeber Hall Department of Archaeology
University of California University of Otago, P. 0. Box 56
Berkeley, CA 94720-3712 Dunedin, New Zealand
 Index
Abbreviated-NAA procedure, 57-60 Fission-track dating, 70, 76, 90
Ahupua'a (traditional Hawaiian land unit), 114 Fissure hydration, 210-211
Andesite Line (Pacific Basin), 105, 226 FORTRAN, 36
Anglo-American archaeology, 4
Apache tears, 86; see also Marekenites GAUSS, 36,37
Atomic absorption spectroscopy (AAS), 19 GIS (geographic information systems), 11
GPS (global positioning system), 22
Basi, Corsica, 67, 76 Grona del Leone, Italy, 69
Birefringence, 189 Grotta Filiestru, Sardinia, 76, 78
Bivariate analysis (plots), 27,98
Heavy liquid sorting, 110
CasaGrande (Hohokam), Arizona, 87 Hohokam (American Southwest), 84, 87
Cave della Ceca, Sardinia, 71 Hydration rim, 189
Chemical index of glass hydration, 185
Chemical variability, sources of in obsidian, ICP-MS (inductively coupled plasma-mass
89-90 spectrometry), 3, 5, 8, 9, 67, 72, 73,
Cluster analysis, 27-28, 73, 98 75, 228
Colors of obsidian, 18 Image analysis (of hydration layer), 213
Composition of obsidian, 18-19 "Impermanence" (of glass sources), 228
Compton scatter correction (in XRF), 164 INAA: see NAA
Induced hydration approach, 184
Dbase, 36 Inductively coupled plasma-mass spectrome-
Discriminant function analysis, 29-30, 93, 98 try: see ICP-MS
Infinite thickness (of EDXRF samples), 163,
Early Expansion Period (Hawai'i), 122-123 164
EDXRF, 6, 8, 9, 70, 71, 75, 90, 103, 104, 109, International Association for Obsidian Stud-
124,132,159-178,228,229 ies (lAOS), 3, 23
Effective hydration temperature (EHT), Intrinsic water (in obsidian hydration), 181
191-192 Isotope ratios, 70
Electron microprobe, 113, 116, 124
Electron spin resonance, 70 K-Ar (potassium-argon dates), 85
Energy dispersive x-ray fluorescence spec-
trometry, 6; see also EDXRF Laboratory hydration rates, 181
Euclidean distance, 27 Lapita sites (the Lapita Homeland Project),
Experimental hydration rates, 208 109, 133

241
242 INDEX

Macroscopic source identification (basaltic Sources of obsidian or basaltic glass (cont.)

glass), 105-109 Mediterranean (cont.)
Magma fractionation, 89 Palmarola, Italy, 68,69
Mahalanobis 0 2, 15, 27, 30-32, 93, 98, 230 Pantelleria, Italy, 68, 69
Marekenites, 86-88, 96 Sardinia, 68, 69
Mogollon (American Southwest), 84 Mesoamerica, 15
Mogollon-Datil Province (New Mexico), Altotonga, Mexico, 21, 34, 37
84-85 Derrumbadas, Mexico, 34, 47
Moloka'i, 114-116, 121 El Chayal, Guatemala, 35, 60
Monte Urpinu, Sardinia, 71 Guadalupe, Mexico, 34
Moss bauer spectroscopy, 70 Jalapa, Guatemala, 35, 53, 60
Laguna de Ayarza, Guatemala, 35
NAA, 3, 5, 6, 8, 9, 19-20, 75, 98, 229; see Malpais, Mexico, 34, 57
also Neutron activation analysis Mil pas Altas, Guatemala, 35
Neolithic (various phases), 71, 76, 77, 78 Otumba, Mexico, 21, 34, 57
Neutron activation analysis, 3, 70, 132; see Pachuca, Mexico: see Sierra de Pachuca
also NAA Paredon, Mexico, 21, 34
Non-optical measurement technique for hy- Pico de Orizaba-Guadalupe Victoria,
dration, 190 Mexico, 34, 35
San Bartolome, Guatemala, 35
"Obsidian bearing regions," 16 San Martinjilotepeque, Guatemala, 35,
111 Ranch Formation, southeastern Arizona, 36,39,41,53,60
96 Sansare, Guatemala, 35, 36
Optical emission spectroscopy, 19 Santa Elena, Mexico, 34, 47,48
Sierra de Pachuca, Mexico, 34, 35, 40,
PIXE/PIGME (proton induced X-ray emission 41,43,187
(PIXE) and proton induced gamma- Tapalzingo, Mexico, 34
ray emission (PIGME)), 3, 6, 8, 9, Tualancingo, Mexico, 34
129,134-140,153,228,229,230 Ucareo-Zinapecuaro, Mexico, 34, 43
Potassium-argon, 70; see also K-Ar Victoria, Mexico, 34
Principal components analysis (PCA), 28-29, 37 Zacualtipan, Mexico, 34
Proton induced X-ray emission (PIXE) and Zaragoza, Mexico, 34, 37
proton induced gamma-ray emis- Zinapecuaro, Mexico, 34
sion (PlGME): see PIXE/PIGME North America
Antelope Wells, New Mexico, 85, 93
Salado (American Southwest), 84, 87 Bodie Hills, California, 165, 179
San Simon river valley, Arizona, 93, 96 Coso volcanic field, California, 187
Society for American Archaeology (SAA), 3, 4 Cow Canyon, Arizona, 84, 85, 88, 96,
Society for Archaeological Science (SAS), 2, 4 228
Source localities (discussion oD, 227 Glass Mountain, Siskiyou County, Cali-
Sources of obsidian or basaltic glass fornia, 162, 164, 169, 178
Mediterranean Gwynn Canyon, New Mexico, 85
Giali, 68 Los Vidrios, Sonora, 85
Lipari, Italy, 68, 69 Mount Taylor (Grants), New Mexico, 85
Melos, 67, 68 Mule Creek (Source Group), New Mex-
Demenegaki, 68 ico,84,89-90,95,96,97
Sta Nychia, 68 Antelope Creek, 85, 86, 93, 95
Mount Arci, Sardinia, 67, 69, 71, 73-75 Mule Creek/North Sawmill Creek, 86,
Conca Cannas, 71 87,90,93
Perdas Urias, 71 Mule Mountains, 86, 87, 93
Santa Maria Zuarbara, 71 San Francisco River Alluvium, 93
INDEX 243

Sources of obsidian or basaltic glass (cont.) Sources of obsidian or basaltic glass (cont.)
North America (cont.) Oceania (cont.)
Napa Glass Mountain, Napa County, West New Britain and Willaumez,
California, 165, 179 Papua, New Guinea (cont.)
Sauceda Mountains, Arizona, 93 Kutau/Bao, 145, 152
Valles Caldera Qemez) Region, New Mopir, 130,139, I42, 144-151,228
Mexico, 85 Talasea, and Kutau/Bao, 130, 142,
Oceania I45, 146
Admiralty Islands, 130 Source standard library on World Wide Web
Lou Island, 130, 206-207 (URL), 12
Manus, 11~ 111, 130 Source systems (conceptual definition),
Pam, 130 227
Banks Islands, 130, 133 SPSS for Windows, 93
Gaua, I30 Structure (S-value) of glass hydration, 186
Vanua Lava, 130 Surface erosion (obsidian), 209
D'Entrecasteaux Group, 130 SYSTAT,36
Dobu, 130
East and West Fergusson, 130, 133 Umleang (GBJ) site, Lou Island, Papua, New
Sanaroa, 130 Guinea, 206
Mauna Loa, Hawai'i basaltic glass, 116
Moloka'i basaltic glass, 114, 115-116, Valley of Oaxaca, Mexico, 21
121-122 Visual (megascopic or macroscopic) charac-
Pitcairn Island Group ignimbrite, 103, terization, 18, 79, 105
106 Volcanic glass as non-silicic glass, 104
Down Rope ignimbrite, 106, 108, 229
Rapa Nui (Easter Island), 104, 106 WDXRF (wavelength dispersive x-ray fluores-
West New Britain and Willaumez, cence spectrometry), 9, 75
Papua, New Guinea, 110, 111, 130,
131, 133, 139, 140-146, 225 Xaltocan, Mexico, 195
Baki, 145, 152, 153 X-ray fluorescence: see EDXRF
Gulu, 145
Hamilton ( Garala), 14 2, I 45, 146 Zeta Structural Factor of glass hydration,
Hamilton (Garua), 142 185-186