Aldenderfer Maschner 1996 Anthropology, Space, and GIS

Anthropology, Space, and
Geographic Information Systems

Spatial Information Series
General Editors
M. F. Goodchild
P. A. Burrough
R. McDonnell
P. Switzer
Mark Aldenderfer
Herbert D. G. Maschner
New York Oxford

OXFORD UNIVERSITY PRESS
1996
Oxford University Press
Oxford New York
Athens Auckland Bangkok Bogota Bombay
Buenos Aires Calcutta Cape Town Dar es Salaam
Delhi Florence Hong Kong Istanbul Karachi
Kuala Lumpur Madras Madrid Melbourne
Mexico City Nairobi Paris Singapore
Taipei Tokyo Toronto
and associated companies in
Berlin Ibadan
Copyright © 1996 by Oxford University Press, Inc.

Published by Oxford University Press, Inc.
198 Madison Avenue, New York, New York 10016
Oxford is a registered trademark of Oxford University Press
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or means,
electronic, mechanical, photocopying, recording, or otherwise,
without the prior permission of Oxford University Press.
Library of Congress Cataloging-in-Publication Data

Anthropology, space and geographic information systems / [edited by]
Mark Aldenderfer, Herbert D. G. Maschner.
p cm. — (Spatial information systems)
Papers from a conference held in Santa Barbara, January, 1992.
Includes bibliographical references and index.
ISBN 0-19-508575-2 (cloth)
1. Anthropology—Data processing—Congresses. 2. Cartography—
Data processing—Congresses. 3. Geographic information systems—
Congresses. 4. Spatial analysis (Statistics)—Congresses.
I. Aldenderfer, Mark S. II. Maschner, Herbert D. G. III. Series.
GN346.5.A57 1996
301'.0285—dc20 95-32350
3 5 7 9 8 6 4 2
Printed in the United States of America

on acid-free paper
Preface
Although spatial thinking has long been a part of anthropological inquiry, it
has waxed and waned in its perceived utility and centrality to the field. Much
anthropological thought at the beginning of the twentieth century was con-
cerned with the concept of diffusion and the definition of culture areas on a
continental scale. Scholars—using material culture, kinship systems, house
form, and social institutions, for example—attempted to identify centers of
diffusion and thus used the notion of spatial proximity to explain similarities
and differences between cultures. Archaeologists of the era used similar con-
cepts to describe the distribution of material culture, and were concerned with
tracing the movement of peoples or charting the origins of agriculture and the
emergence of civilization. Anthropology and archaeology parted ways after
1940 in their thinking about the role of space in their fields.
Anthropologists rejected most diffusionist theories and in their place began
to develop new schools of thought and theories, few of which integrated space
and spatial thinking in a meaningful way into the research process. In one
sense, anthropology turned inward and sought to demonstrate the roles of
history, place, and locality as the primary means by which an understanding of
human cultural diversity could be appreciated. Space thus became passive and
sterile as an analytical concept. Archaeologists, however, maintained their in-
terest in space, and in the 1950s, with the emergence of settlement archaeol-
ogy, began to explore more sophisticated ways in which to employ spatial think-
ing and concepts. With the emergence of the "New Archaeology" in the 1960s
and its emphasis on explanation, quantitative thinking, and a scientific per-
spective on the past, archaeologists increasingly turned to other fields, notably
geography, for tools and ideas for spatial analysis. Geographical information
systems (GIS), as they became practical tools for spatial analysis in the early
1980s, were quickly seized upon by archaeologists, who immediately recog-
nized their potential.
Despite the inward turn of anthropological thought, significant subfields,
such as development and ecological anthropology, managed to retain an ap-
preciation of space and—independent of developments in archaeology—dis-
covered the value of GIS and how it could help them achieve their own re-
search goals. This became especially important for those scholars concerned
with finding ways to integrate the results of traditional anthropological in-
quiry, which tends to be small-scale and personal, with data obtained from
research into regional-scale phenomena, such as deforestation.
By the early 1990s, it became clear to many of us working with GIS that it
vi Preface
was time sample the field, identify the areas of strength in the use of GIS
within it, and demonstrate the value of GIS to our colleagues. To that end, a
conference was held at the University of California (UCSB) in Santa Barbara
in January 1992 entitled "The Anthropology of Human Behavior Through
Geographic Information and Analysis: An International Conference." A total
of 22 papers was presented. The papers in this volume are good representa-
tives of the variety of issues discussed in the conference. While it is clear that
much remains to be done to demonstrate the utility of GIS to a broader
anthropological audience, we feel these papers mark an important first
step toward that goal.
Many organizations and people were instrumental to the success of the
conference and the appearance of this volume. The National Center for Geo-
graphic Information and Analysis (NCGIA), directed by Mike Goodchild,
provided the most of the funding for the conference, and we are grateful for
his support and encouragement. Staff members of NCGIA who cheerfully
assisted us with the financial and other myriad details of running a complex
conference were Judith Parker, Sandi Glendenning, Yasmina Mhemedi, and
Carol Wasteneys; to each of them, we offer our thanks. We also owe a debt of
gratitude to the Social Science Computing Facility (SSCF) at UCSB. Joan
Murdoch, its director, allowed attendees of the conference access to E-mail
and other computing services; further, she helped us overcome a variety of
logistical and other problems. Other SSCF staff members who provided use-
ful assistance were Chris Arnold and Jeff Stein. The UCSB Department of
Anthropology also lent us a number of services: Dirk Brandts prepared con-
ference brochures, while Brian Billman, Pat Lambert, and Nelson Siefkin cheer-
fully acted as chauffeurs and assistants. Thanks also go to Stephanie Golledge
of the Department of Geography, who ran the slide projectors and supervised
the audiovisual equipment used in the conference.
This book could not have been assembled without the skills of Karen
Doehner, who had the patience to deal with questions about page layout and
graphic design from impatient and ignorant editors. Without her, in fact, we
would still be trying to churn this manuscript out on our word processors. She
was ably assisted at crucial moments by Dirk Brandts and John Kantner. We
would also like to express our gratitude to the authors of these papers, who have
exhibited remarkable patience in the light of the difficult genesis of this volume.
Santa Barbara, Calif. M. A.

January 1995
Madison, Wise. H. M.
January 1995
Contents
1
Introduction 3
Mark Aldenderfer
2
Land Degradation in the Peruvian Amazon: Applying GIS in Human Ecology
Research 19
William M. Loker
3
The Use of GIS to Measure Spatial Patterns of Ethnic Firms in the Los Ange-
les Garment Industry 44
Christopher G. Arnold and Richard P. Appelbaum
4
A Formal Justification for the Application of GIS to the Cultural Ecological
Analysis of Land-Use Intensification and Deforestation in the Amazon 55
Clifford A. Behrens
5
Integrating Socioeconomic and Geographic Information Systems: A Method-
ology for Rural Development and Agricultural Policy Design 78
Susan Stonich
6
Empirical and Methodological Problems in Developing a GIS Database for
Yanomano Tribesmen Located in Remote Areas 97
Ken McGwire, Napoleon A. Chagnon, and Charles Brewer Carias
7
A Time to Rend, A Time to Sew: New Perspectives on Northern Anasazi
Sociopolitical Development in Late Prehistory 107
Carla Van West and Timothy A. Kohler
8
Moving from Catchments to Cognition: Tentative Steps Toward a Larger
Archaeological Context for GIS 132
Vincent Gaffney, Zoran Stancic, and Helen Watson
viii Contents
9
An Analysis of Late-Horizon Settlement Patterns in the Teotihuacan-
Temascalapa Basins: A Location-Allocation and GIS-Based Approach 155
AmyJ. Ruggles and Richard L. Church
10
The Politics of Settlement Choice on the Northwest Coast: Cognition, GIS,
and Coastal Landscapes 175
11
The Role of GIS in the Management of Archaeological Data: An Example of
Application for the Spanish Administration 190
Conception Blasco Bosqued, Javier Baena Preysler, and Javier Expiago
12
The Role of GIS in the Interdisciplinary Investigations at Olorgesailie, Kenya,
a Pleistocene Archaeological Locality 202
Richard Potts, Tom forstad, and Daniel Cole
13
Danebury Revisited: A English Iron Age Hillfort in a Digital Landscape 214
Gary R. Lock and Trevor M. Harris
14
Geographic Information Systems and Spatial Analysis in the Social Sciences 241
Michael F. Goodchild
References 251
Contributors
Mark Aldenderfer Department of Anthropology
University of California-Santa Barbara
Richard P. Appelbaum Department of Sociology
Christopher G. Arnold Department of Anthropology
Javier Baena Preysler Servicio de Cartograffa
Universidad Autonoma de Madrid
Clifford A. Behrens Information Sciences Research
New Jersey
Concepcion BlascoBosqued Servicio de Cartograffa
Charles Brewer Carias Department of Anthropology
Napoleon A. Chagnon Department of Anthropology
Richard L. Church Department of Geography
Daniel Cole Department of Paleobiology
Smithsonian Institution, Washington D.C.
Javier Espiago Servicio de Cartografia
Vincent Gaffney Department of Archaeology
University of Reading
Michaeol F. Goodchild Department of Geography
Trevor M. Harris Department of Geography
University of West Virginia
Tom Jorstad Department of Paleobiology
Smithsonian Institution, Washington, D.C.
Gary R. Lock Institute of Archaeology
University of Oxford
William M. Loker Department of Sociology, Anthropology,
and Social Work
Mississippi State University
Herbert D. G. Maschner Department of Anthropology
University of Wisconsin-Madison
x Contributors
Ken McGwire Desert Research Institute
University of Nevada
Richard Potts Department of Paleobiology
Smithsonian Institution, Washington, D.C.
AmyJ. Ruggles Department of Geography
University of Iowa
Zoran Stancic Scientific Research Centre of the Slovene
Academy of Sciences and Art
Slovenija
Susan Stonich Department of Anthropology
Helen Watson Somerset
Great Britain
Carla Van West Statistical Research, Inc.
Tucson, Arizona
This page intentionally left blank
1
Introduction
Mark Aldenderfer
Although spatial thinking has long been a part of anthropological inquiry, it
has waxed and waned in its perceived utility and centrality to the field. Al-
though the papers in this volume attest to a vigorous tradition of spatial think-
ing in anthropology and further suggest that, for at least some branches of the
field, spatial thinking and analysis are truly central to their definition and mis-
sion, it is nevertheless clear that this has not always been the case. Further,
despite differences in historical trajectories of development between the two
major subfields of anthropology—cultural anthropology and archaeology1—
in terms of the way space has been used, it is also clear that the two subfields
share a number of common interests and themes that deserve discussion and
exploration. This exploration is not only interesting from a purely historical
perspective, but also has a very practical, down-to-earth dimension. The lit-
erature on the history of science is replete with cases of communication fail-
ures both within and between scientific disciplines. While in many cases this is
'merely annoying (different terms used to describe the same procedure, for
instance), there are occasions when these failures lead to the creation of a
highly idiosyncratic jargon used by small cliques of investigators, which clearly
offers the opportunity to inhibit scholarly communication. This, in turn, can
lead to redundancy of effort, failure to learn from the mistakes of others, and
wasted time and money. By providing a forum in which similarities and differ-
ences can be examined, the natural tendency of scientific disciplines to form
these cliques can be overcome.
I intend this paper to be such a forum for an exploration of the ways in
which geographic information systems (GIS) have been employed by anthro-
pologists and archaeologists as represented by the authors of the papers pre-
sented in this volume. I will briefly describe the GIS for those readers unfa-
miliar with it and then turn to a review of the history of spatial thinking and
the kinds of tools used to implement this thinking for each of the subdisci-
plines. Following this, I will turn to a discussion of the themes of the use of
GIS common to both. My hope is that the reader will not only gain a deeper
insight into the range of practice in the fields but also become aware of the
very significant points of contact they share. In this way, anthropologists and
others who wish to use the outcomes of anthropological research can avoid,
insofar as that is possible, the formation of cliques of users that can further
fragment an already very disparate field. I also hope to show that the use of
4 Anthropology, Space, and Geographic Information Systems
GIS in anthropology is currently very strong and that future prospects for its
continued development are bright.
A GIS is a sophisticated database management system designed for the acqui-

sition, manipulation, visualization, management, and display of spatially ref-
erenced (or geographic) data.2 The GIS has its origins in computer-assisted
mapping software developed during the 1970s, but it has evolved substantially
from these roots, particularly through its emphasis on expanded analytical ca-
pabilities, the capacity to accept a wide range of data types as input (i.e., satel-
lite imagery, standard aerial photographs, and digitized maps), and its ability
to generate new information through queries to a variety of sophisticated
databases. Although some GIS packages can be used on desktop comput-
ers, the full capabilities of a GIS are best exploited using work stations or
minicomputers.
In a GIS, data are represented as layers or themes, with each layer being a
specific natural, cultural, or derived variable, broadly defined, that describes
the environment within the context of the problem under study . The infor-
mation within each of these layers can be represented in either of two distinct
formats: rasters, in which data are aggregated into a grid of cells, or vectors, in
which data are represented by combinations of lines, points, and polygons.
Each scheme has strengths and weaknesses. Raster systems are used frequently
to represent environmental data layers, and they have had a long history of use
in GIS applications in which remotely sensed data have been important. Since
cells are aggregates, there is a loss of accuracy in the way in which the grid
describes the data layer in question; further, there may be problems with reso-
lution and description if an inappropriate grid size is used to represent a data
layer. However, raster systems are well suited to modeling, analysis, and dis-
play, since data layers can be easily overlain to discern patterning. Vector sys-
tems, in contrast, are valuable when accuracy in the representation of a data
layer is required. They are ideally suited for the production of high-quality
maps or certain data themes such as property or political boundaries, net-
works (streams, roads, etc.), and similar features.
Which of these two approaches will be used in a particular project depends
primarily upon the goals of the research. This acknowledges that the GIS,
therefore, is properly viewed a tool, albeit a very powerful one, for the analysis
of spatial data. This recognition in no way should be seen as identifying the
GIS as "just" a tool; instead, it provides a basis for using the tool in the most
efficacious manner. Marble (1990:14-17) has discussed this concept at length
and has argued that the GIS is precisely the tool those with spatial data need to
define new problems, open new research horizons, and integrate, in a way not
seen before in the social sciences, an informed spatial perspective.
Introduction 5
A Little History
Much anthropological thought at the end of the nineteenth and the beginning
of the twentieth century was concerned with the concept of diffusion and the
definition of culture areas on a continental scale. Scholars—using material
culture, kinship systems, house form, and social institutions, for example—
attempted to identify centers of diffusion of these and other traits and thus
used the concept of spatial proximity to explain similarities and differences
between cultures. In great part, the region in which such diffusions took place
was simply assumed to be isotropic, and the rate of diffusion was held to be
relatively unconstrained by the reality of physical space. In the United States,
diffusionist thought was associated with the culture area concept, which was
an attempt to map out the distribution of ethnic groups defined by language
or similarities of material culture in some geographicly defined space. Inter-
estingly, these regions could be scaled in size depending on the problem of
interest. The anthropologist Alfred Kroeber, in his book Cultural and Natural
Areas of Native North America (1939), listed large areas such as California, the
Eastern Woodlands, and the Great Plains. His goal was to look at variation in
Native American culture on a continental scale; thus, regions were large and
the data used to construct these regions were relatively coarse. At a smaller
scale, Gifford and Kroeber (1937) studied variation in material culture
within the "Porno culture area," a region within the larger "California"
culture area. Here, scale was measured in tens of miles, and very detailed
lists of overlapping sets of material culture were generated. The data, then,
were relatively fine-grained.
The tools used to manipulate these data were fairly simple and consisted of
maps, map overlays, and tables (labeled as culture element distribution lists)
that contained data on some cultural trait (hunting technology, clothing style,
etc.) cross-classified by the trait's presence or absence in the ethnic groups said
to inhabit the culture area under study. Data were tabulated by hand and plot-
ted on maps. While these data could have been portrayed as contour maps of
trait frequencies, described by fall-off curves in simple gravity models or some
other quantitative convention, substance was almost always deemed more im-
portant than a focus upon method.3
The European experience was focused upon the notion of the Kulturkriese,
or "culture circles," which were defined as large sets of traits spread across vast
geographic spaces. The concept had its origin in the German school of
anthropogeography, and its main emphasis was upon the distribution of mate-
rial traits such as the details of bow and arrow manufacture and other, simple
cultural traits (Harris 1968: 373, 382-383). The spatial extent of these circles
could be vast; in one instance, similarities in material culture were observed in
Melanesia, Indonesia, and West Africa and were thus included in a single circle
(Ratzel 1896). Refinements in this theory led to the development of circles of
smaller spatial scales. Schmidt (1939), for example, defined three levels of
circles—primary, secondary, and tertiary—and within each there were several
smaller-scale circles. The approach to developing these circles was very simi-
lar to how American anthropologists built their culture areas—cross-clas-
sified trait lists, usually based upon aspects of material culture, were con-
structed, and distributions of these traits were then plotted on maps and
affinities assessed.
Archaeologists of the era were engaged in similar enterprises. While much
archaeological work of the late nineteenth and early twentieth centuries fo-
cused upon the excavation of single sites deemed of major importance to the
cultural history of some region, there was nevertheless considerable interest
in delineating patterns of the diffusion of cultural traits, albeit in this instance
in prehistory. Perhaps the most famous of these diffusionist approaches was by
V. Gordon Childe, who wrote two very influential books, The Dawn of Euro-
pean Civilization (1925) and The Danube in Prehistory,(1929). Diffusionist thought
had always been very strong in archaeology, but in these books Childe very
carefully reviewed the empirical data for the spread of civilization from the
core area of the ancient Near East. As Renfrew and Bahn (1991: 32) note,
diffusionist ideas were part of a broader trend in archaeology that sought to
classify and consolidate knowledge of the past. What this meant in practice
was the definition of artifact types, primarily upon some stylistic grounds, fol-
lowed by an examination of their spatial distribution. In Europe, this led to
the definition of entities called "archaeological cultures" (Shennan 1974), and,
in North America, depending on the system used, they were labeled as foci,
traditions, or phases. Both the archaeological culture and the phase, two of the
most successful and durable products of this epoch of consolidation, have a
spatial and temporal definition in addition to their cultural content. That is,
they are found within regions within a limited temporal frame. In this sense,
they are not dissimilar from culture areas; in general, they tend to be defined
in precisely the same manner, but with explicit concern with time. Thus ar-
chaeological materials are recovered from a series of sites in a region, and
their contents are compared within discrete segments of time. The regional
distribution of artifact types as found at sites is identified, and thus a spatial
boundary for the phase is established. The data used to construct these enti-
ties are wholly concerned with some form of material culture. They are com-
bined into trait lists, and their spatial distribution is plotted onto maps and
map overlays. While criticized, this method of defining spatiotemporal cul-
tural boundaries is still used in this manner today.
During the 1930s and 1940s, however, anthropology as a discipline rejected
most diffusionist theories for their obvious explanatory inadequacies. It was
heavily criticized from more historically focused viewpoints of culture as well
as a reemergent emphasis on evolutionary explanations for observed patterns
of cultural similarity or difference. Some of these new schools, though, de-
spite potential improvements in explanatory potential, did not integrate space
and spatial thinking in a meaningful way into the research process. In one
sense, anthropology turned inward and sought to demonstrate the roles of
history, place, and locality as the primary means by which an understanding of
human cultural diversity could be appreciated. Space thus became passive and
sterile as an analytical concept. Anthropological concerns were typically small-
Introduction 7
scale, and most field methodologies emphasized the development of styles of
face-to-face interaction, such as participant observation. Samples of informants
were generally small, and the anthropologist could easily integrate these ma-
terials into a report, paper, or synthesis. The spatial scale was the village, or
perhaps a set of villages, in a small area. While there was implicit recognition
of the broader world or region, the "outside" was dealt with when necessary
but was not frequently a focus of research effort. Consequently, the analytical
tools used to deal with these limited spatial domains were likewise simple.
Some areas within anthropology, however, maintained their interest in spa-
tial phenomena; in great part, these developments have led to the adoption of
the GIS as a major tool in anthropological research. Schorr (1974: 166-168)
has described how the use of aerial photography helped to maintain a practical
interest in larger-scale spatial phenomena in anthropological research. While
archaeologists were the first to use aerial photography extensively following
its successful application to military intelligence in World War I (Schorr 1974:
163-165; Deuel 1969), ethnographers were slow to adopt it. Much of this
delay can be attributed to a genuine lack of a theoretical perspective that val-
ued spatial data. As I noted above, space per se was only a field of action, and
when recognized, it was only at a small scale. However, from the combined
stimuli of cultural geography, the sociological analysis of rural phenomena,
particularly agricultural systems, and the eventual emergence of a robust theory
of ecological (or materialist) anthropology, a consideration of space was rein-
tegrated into anthropological thought. As Schorr (1974: 167) relates, John
Rowe, who would eventually turn to archaeology, was the first American
anthropologist who turned to the systematic use of aerial photography as
an aid to ethnographic research. Rowe used it to get regional scale data on
settlement patterns, land tenure, house types, cultivation cycles, and other
data (Rowe 1953).
The importance of the development of a theory of ecological anthropology
and its influence upon the reemergence of spatial thinking in anthropology
and archaeology cannot be underestimated. Although much of the effort in
creating a robust ecological theory in anthropology was directed at establish-
ing causal linkages to demonstrate how humans adapt to their environments, a
key element of ecological anthropology was to adopt the ecosystem concept.
Ecosystems, of course, are spatially referenced, and clearly, human activity
within them varies at least in part as a consequence of spatial and temporal
variation in energy availability and how it flows through the system. Julian
Steward (1938) was responsible for the early development of an ecological
approach to anthropology through his meticulous reconstructions of settle-
ment patterns and environment of the Great Basin Shoshone in the American
West. As his approach was adopted and modified by others, the concept of
spatial variation in environmental potential and the way it affected human
utilization was carried into other regions and problems. For example, Conklin
(1957, 1967), one of the pioneers in the use of aerial photography as an aid to
the study of agricultural systems, particularly in the Philippines, was also in-
strumental in developing a variant of ecological anthropology.
Despite the reintroduction of space into the research process, the spatial
scale of research remained relatively small, and further, although aerial pho-
tography became more commonly used as an adjunct to field research, no
other significant spatially based methods or techniques were adopted by most
anthropologists working within this ecological paradigm. Maps and map over-
lays, as augmented by photography, remained valuable and were used prima-
rily as visual aids or iconic devices rather than as data themselves. Thus while
the settlement pattern within a region might be plotted as a map overlay on a
vegetation map, for example, aside from the demonstration of spatial propin-
quity, there was generally no more sophisticated use of spatially referenced
data in the research process. Many of the most successful case studies of eco-
logical anthropology have been described as "microlevel" studies (Moran 1979:
57) and are similar in size to the smallest-scale analyses done by anthropolo-
gists working from the culture area perspective.
Archaeologists of the era, however, moved beyond the efforts of their an-
thropological colleagues. Aerial photography was already well embedded in
the field during the 1930s and continues to play a major role today. Archae-
ologists wholeheartedly embraced the ecological approach and with it the eco-
system concept. Another innovation was the development of the settlement
pattern approach, an explicit concern with the identification of spatial vari-
ability in types of human settlements on the landscape. As defined by Willey
(1953: 1), settlement patterns were "the way in which man disposed himself
over the landscape on which he lived. It refers to dwellings, to their arrange-
ment, and to the nature and disposition of other buildings pertaining to com-
munity life." Note the early focus on settlement patterns was upon regional-
scale variability in the kinds of sites made by people and the degree to which
these distributions reflected social norms and social interactions. It was not
until the 1960s that this approach to settlement analysis was connected to the
ecological approach and therefore to broader issues that attempted to charac-
terize human behavior as an adaptation to ecological variability in the land-
scape. Regardless of theoretical perspectives, however, the direct observation
of spatial variation in these phenomena is an important innovation.
With the emergence of the "new archaeology" in the 1960s and its empha-
sis on explanation, quantitative thinking, and a scientific perspective on the
past, archaeologists increasingly turned to other fields, notably geography, for
tools and ideas for spatial analysis. A whole suite of spatial analytic methods
and techniques were borrowed wholesale, and they were quickly integrated
into archaeological field research. These included modern variants of the Von
Thunen model of agricultural land use, Weber's model of industrial location,
Christaller's central place model, Hagerstrand's model of innovation and its
diffusion, and gravity models of all kinds (Hodder and Orton 1976), and many
of these models were quickly connected to the ecological paradigm and settle-
ment pattern analysis. Graduate students in archaeology were directed to
courses in human geography. The outcome of this interest in geography was
that at least in archaeology, there was a very explicit understanding that spatial
data and spatial analytic techniques were fundamental to archaeological research.
Introduction 9
Another significant technical innovation that helped to better define re-
gional-scale studies was the appearance of remotely sensed data following the
launch of the LANDSAT 1 satellite in 1972. In many ways, the launch of this
satellite revolutionized anthropological perspectives on regional analysis. For
the first time, very large areas could be viewed in a number of productive ways
(i.e., different wavelengths of the spectrum), yet modest levels of detail of these
very large regions could be observed. Further, images could be updated re-
peatedly. For LANDSAT 1, for example, each region could be reexamined
every 18 days. The incorporation of a dynamic temporal element meant that
at least in principle, temporal and spatial variability in resource availability
could be studied simultaneously and thus were not strictly dependent upon
being the field to monitor this variability in person. Not surprisingly, archae-
ologists were first to take advantage of remotely sensed data, but ecological
anthropologists quickly followed suit (Ebert 1984).
By 1980, many archaeologists and a significant fraction of anthropologist
were deeply imbued with the importance of spatial data in their research ef-
forts. Despite the introduction of powerful new models of spatial behavior,
new methods for the acquisition of spatial data at very large scales, and useful
theoretical constructs that directed inquiry, there remained a significant gap
between the desire to work at larger spatial scales and the ability to do it in a
practical manner. The stage, then, was set for the introduction of GIS to an-
thropological and archaeological inquiry.
Recent Trends
In a recent overview of applications of GIS in archaeological research, Kvamme

(1989: 162) identifies five broad themes of its use: regional data management,
management of remotely sensed data, regional environmental analysis, simu-
lation, and locational modeling. Although these themes of use obviously over-
lap, they each have slightly different emphases that are useful to explore. While
these may be particularly apt descriptions of GIS use in archaeology, I believe
they are also appropriate descriptors of anthropological uses of GIS as well.
This is especially true given the recent convergence of interest in regional
scale studies in both anthropology and archaeology. In this section of the pa-
per, I will discuss how these themes of model use are reflected in the papers
presented in this volume and, further, I will explore some of the new theoreti-
cal and conceptual contexts that have emerged over the past fifteen years and
how they have affected the ways in which GIS is used today.
Regional Data Management
Within archaeology, the impetus for the development of this model use has
come from the expansion of interest in historic preservation in the United
States and, to a lesser extent, Europe. Through legislation at the federal, state,
and local levels, archaeological and historical sites are protected from destruc-
tion in varying degrees. Despite the ravages of the modern era, the numbers of
extant archaeological sites are truly staggering, and much historic preserva-
tion legislation mandates that governmental agencies maintain inventories of
archaeological sites and historic properties under their jurisdiction For ex-
ample, California has over 160,000 historic properties in its rolls, and New
Mexico has almost 100,000. Obviously, the GIS is ideally suited the task of
dealing with these spatially referenced data, and the paper by Blasco Bosqued,
Baena Preysler, and Espiago (Chapter 11) is a good example of how this can be
accomplished. While Kvamme (1989: 164) notes that most GIS applications
involving regional data management are not directly concerned with manage-
ment per se, there is a growing trend to build GIS applications specifically for
this purpose. Many states in the United States are currently developing GIS
for historic preservation, among them Nebraska, California, New Mexico, and
others, and major efforts toward this end are also found in Great Britain and
much of Europe (Larsen 1992).
The papers by Stonich (Chapter 5) and McGwire, Chagnon, and Brewer
Carias (Chapter 6) are excellent examples of how cultural anthropologists have
used GIS for regional data management. These studies, though, have an im-
petus different from that of this model use in archaeology. As the ecological
paradigm in anthropology developed, it became apparent that more attention
had to be paid to questions of how data from very disparate sources, such as
informant interview and remotely sensed images, could be effective integrated.
This became especially important for those scholars concerned with finding
ways to integrate the results of traditional anthropological inquiry, which tends
to be small-scale and personal, with data obtained from research into regional-
scale phenomena. An additional concern was how the ecosystem concept could
be effectively put into practice as an operational construct rather than as a
convenient concept (Morren 1991; Winterhalder and Evans 1991).
Yet another dimension of the need to manage regional-scale data within
anthropological inquiry comes from the field of development anthropology.
Development anthropology is generally concerned with finding ways to ame-
liorate or reduce the impact of rapid cultural change in so-called traditional
societies and to investigate the ways in which change is made manifest in these
societies when confronted with significant and persistent contact from more
complex groups. Research thus takes place within an "applied" or practical
context. Unlike many forms of anthropological inquiry, then, the outcome of
research may have a powerful influence on the lives of the people or groups
under study, for either good or ill. Consequently, those who practice in this
field of anthropological inquiry must be more diligent and thorough than their
purely academic brethren simply because at some level, people's lives depend
on what they do. Thus their tools must be first-rate and comprehensive.
One aspect of development anthropology that must be emphasized is that
the spatial scale of the entity affected by this rapid change is often consider-
ably larger than that dealt with in more traditional anthropological research
settings. Historically, anthropologists have tended to limit their inquiry to small-
scale cultural phenomena, such as families, households, villages, and, in some
cases, even individuals. The development anthropologist, however, is faced
Introduction 11
with change generally taking place across regions and larger geographic spaces.
Given this, it is clear that traditional forms of investigation, such as participant
observation, while providing useful insights, must nevertheless be supplemented
by other sorts of instruments, such as surveys and questionnaires.
As Stonich shows, an obvious role of GIS in this area is to provide a plat-
form through which regional-scale data can be better integrated into the re-
search process. In this case, the GIS is a tool useful for automation of research
and is thus a natural extension of the desire to use computer-assisted tech-
nologies whenever feasible. In effect, the GIS allows the researcher, along
with complementary data-gathering methods, to develop a more reliable re-
gional picture of variation or homogeneity of the phenomena under study.
These data might include satellite imagery, aerial photographs, and space-
borne photos. Combined with recent advances in global positioning system
(GPS) technology and ground-truth or validation studies, it is clear that through
the use of GIS, development anthropologists will be better able to understand
the region in a way heretofore not possible. The GIS also provides a common
platform for sharing data across different scientific disciplines. In her Hondu-
ran work, Stonich worked with agronomists and other natural scientists with
very different perspectives on data collection and field research. As she de-
scribes it, though, the GIS was instrumental in developing a broader basis for
cooperation among these scientists from different fields.
The data management needs faced by McGwire, Chagnon, and Brewer
Carfas are somewhat different. For almost thirty years, Chagnon has been
engaged in the study of the Yanomamo, a group of tribal peoples of southern
Venezuela. His work has involved extensive informant interview, demographic
reconstructions of population history, comparison of genetic data obtained
from spatially distant and proximate villages, village movements over a re-
gional scale, and much more. As they describe in their paper, it is clear that in
order to understand these data as well as to gain insight into warfare, political
alliance, and subsistence, many villages must be studied and compared. Again,
the GIS is ideal for managing these data, which are composed of a number of
distinct spatial scales. In recent years, their work has taken on greater urgency,
since the Yanomamo are threatened by land invasions and political turmoil.
Management of Remotely Sensed Data
As I pointed out above, the advent of remotely sensed data, particularly that
obtained through space-borne platforms, has been of enormous importance
to both cultural anthropology and archaeology. It has allowed the anthropo-
logical researcher to examine truly large-scale phenomena in way heretofore
impossible. The papers by Loker (Chapter 2), Behrens (Chapter 4), and to a
lesser extent the papers by McGwire et al. and by Stonich, are good examples
of this trend. One of the most important environmental issues of the modern
era is the combination of rapid population growth and environmental degra-
dation. Although clearly a global phenomenon, population growth and its ef-
fects on environment are best observed at a regional level. One region of the
world in which this topic has been hotly debated is in the Amazon basin. As
Loker rightly notes, while governmental policy may foster policies tolerant of
destructive land-use practices, land is destroyed by the land manager or farmer.
Thus a "global" problem has a very real local manifestation. The question for
research, then, is to identify how these land managers are destroying the land,
what kinds of land are under the most threat, and how productive steps can be
taken to slow or even halt these activities. One of Loker's main points is that
secondary data on land use, soil type, and crops planted—data traditionally
used by economists, agronomists, and anthropologists to chart changes in land
use through time—are generally inaccurate and thus almost always suspect.
Further, they are always dated and are likely to be of little use in such a dy-
namic, ever-changing situation. Remotely sensed data, in combination with
local-level studies for ground truth and depth, are an obvious solution to the
problem of data quality. To Loker, the GIS is the only platform capable of
handling these data in a timely and useful fashion; his paper is a very good
description of how such a research effort can be organized.
Behrens' paper provides a somewhat different perspective on the use of
remotely sensed data that is complementary to the approach used by Loker.
While Behrens is concerned with deforestation in the Amazon, his emphasis is
upon building a formal model of the process, one that shows how indigenous
groups in the region intensify their use of the land and how this leads to defor-
estation and other forms of land degradation. The remotely sensed data were
used to test this formal model, and the GIS was used to organize data over a
very large area of the Peruvian Amazon basin. As Behrens notes, his work is
firmly within an ecological approach to cultural phenomena.
Regional Environmental Analysis
This theme of model use is concerned with the examination of the way in
which some spatially distributed phenomenon is correlated with features of
the physical environment. In one sense, it is a clear outgrowth of the develop-
ment of the ecosystem concept as applied to understanding human behavior,
and the goal of this theme of model use is to determine that set of features
which appear to have the greatest influence on the character of human settle-
ment in some region. As Kvamme (1989: 168) notes, while there has been a
long tradition of this type of environmental modeling in archaeological re-
search, most of the results were not convincing until the advent of GIS, which
allowed the researcher to look at larger regions in a far more systematic man-
ner. The most successful uses of this model theme have come from archaeol-
ogy, and a number of papers in this volume deal with it, including those by
Van West and Kohler (Chapter 7); Gaffney, Stancic, and Watson (Chapter 8);
Maschner (Chapter 10); and Lock and Harris (Chapter 13). In each of these
papers, some aspect of the regional environment has been deemed of impor-
tance to understanding some aspect about the way in which the landscape was
used in past times. In the paper by Lock and Harris, for example, the authors
are concerned with understanding the place of Danebury, an English hillfort,
Introduction 13
in both the environmental and social landscapes of the past. Data used in their
modeling exercise include present-day hydrology, soils, and terrain elevation.
In a very different part of the world, Tebenkof Bay of the Alaskan coast,
Maschner collected a different suite of environmental data, including grade,
drainage, beach quality, vegetation, distance to fresh water, climatic exposure,
solar exposure, and resource patches (such as shellfish beds, salmon streams,
etc.). However, he only incorporated a subset of these in his GIS: cardinal
exposure, island size, climatic exposure, beach quality, slope, drainage, and
distance to fresh water. Maschner's goal is to model the determinants of settle-
ment placement, and while he rightly emphasizes the role environmental fac-
tors have to play in this process, he argues that environmental data alone are
not sufficient to explain why people placed their sites where they did. This is a
topic to which I will return to below.
The paper by Van West and Kohler is an interesting example of how envi-
ronmental data can be used creatively in the research process. The context of
their research is the Four Corners area of the American Southwest in a time
frame ranging from A.D. 900 to 1350. They are interested in the degree to
which environmental factors influenced food sharing in this region and how
this, in turn, was related to long-term cycles of population dispersion and ag-
gregation. Obviously, regional environmental data are of critical importance
to the exploration of this problem. Of the many different kinds of environ-
mental data to choose from, however, they selected only five categories of
information: soil depth and type, available water capacity, natural plant pro-
ductivity, and agricultural productivity. While they could have chosen many
others, they identified these variables as the most important to their study.
Their paper is a good example of how to approach the modeling process;
there must be good congruence between hypotheses posed, data collected,
and methods of analysis.
In each case, the authors rightly note the critical role played by the GIS in
their research. Although each model could have been constructed without a
GIS, none of them could have been used to investigate the problems posed in
anything like a systematic and thorough fashion without it. Maschner, for ex-
ample, is quite explicit in this belief when he argues that without the GIS, his
particular approach to the determinants of settlement choice could not have
been accomplished. Lock and Harris agree and emphasize the importance of
GIS in terms of allowing researchers to explore their data more fully than
had previously been possible.
Simulation
Simulation asks the question "What if...?" and, through the use of some sort
of model, explores the consequences of that model in a dynamic manner
(Aldenderfer 1990: 196-199). Simulation has had a long history of use in an-
thropology and archaeology, and it is no surprise that users of GIS within
these fields have found a productive way in which to combine the two meth-
odologies. While relatively few studies using simulation and GIS together have
been published, two papers in this volume—Van West and Kohler and
Behrens—demonstrate the potential of the approach. Van West and Kohler
develop a very sophisticated model of prehistoric agricultural activity, and use
the model to explore how human populations dependent upon sufficient lev-
els of agricultural productivity would have responded to extreme climatic vari-
ability. Among other things, they are able to predict the degree of cooperation
in food sharing using this model, and further, their results can provide deeper
insight into the process of population dispersal and aggregation seen through
the Four Corners region from A.D. 900 through 1350. The thrust of Behrens'
paper is similar, and although he does not provide a complete empirical test of
his model of agricultural land intensification in the Peruvian Amazon, he does
indicate the strong points of his model when the GIS is used to develop data
based on its implications. The role of GIS in both of these examples is clear;
without the GIS, neither of these models could have been examined in a
systematic manner.
Locational Modeling
One of the most important and obvious applications of the GIS to human
behavior is its use to predict the location of some aspect of human behavior on
the landscape. Depending on the goal of the modeling effort, this can be ei-
ther a very simple or very complex thing. As I have discussed in my historical
review, the idea of modeling and predicting spatial aspects of human behavior
has had a long history in both anthropology and archaeology, although the
degree to which it has been realized has varied with the skill of the modelers,
the quality of the model, and the kinds of methods available useful in making
the model operational. Moreover, the quality of theory has a very strong
influence on the success of any modeling effort, as I will discuss more
extensively below.
Within cultural anthropology, there has been relatively little use of locational
modeling, although there has been a great desire to find some way to imple-
ment it. Unlike their archaeological colleagues, however, cultural anthropolo-
gists have been little interested in predicting the locations of sites or other
forms of habitations on the landscape. Instead, their interest in locational
modeling is directed at unraveling spatial patterns in more complex behaviors.
For instance, Winterhalder and Evans (1991), in their study of agricultural
productivity on the eastern flanks of the Andes in southern Peru, looked pri-
marily at field distribution and patterns of exchange rather than the placement
of villages vis-a-vis some set of natural features. A similar approach in a very
different environmental and cultural context has been taken by Arnold and
Appelbaum (Chapter 3). In this paper, the authors are interested in modeling
the relationship between ethnicity and spatial location in the Los Angeles gar-
ment district. Theory from economics, geography, and sociology predicts that
the geographic concentration of businesses promotes their competitiveness
through a variety of mechanisms. A neglected aspect of this is the degree to
which these concentrations overlap in their distribution with other spatial
Introduction 15
phenomena, such as ethnic neighborhoods. Arnold and Appelbaum hypoth-

esize that ethnic groups demonstrating higher levels of form concentration
will have high levels of economic success, but they also ask the key question of
at precisely what spatial scale is this success manifest. Therefore, they must
"locate" ethnic variability at some spatial scale and relate it to economic suc-
cess. This approach to locational modeling is reflected in different degrees in
the papers by Loker, McGwire et al. and Behrens.
Most GIS applications within archaeology have been concerned with pre-
dictive modeling (Kvamme 1989; Allen et al. 1990). Specific modeling meth-
odologies used to explore the data generated through the use of the model
include log-linear modeling, multivariate methods, various forms of numeri-
cal simulation, and, of course, various spatial statistics. As Kvamme (1989) has
noted, however, there has been less emphasis on the use of spatial statistical
thinking than might be expected. The general consensus regarding the use of
the GIS for locational modeling in archaeology is that while it has been rela-
tively successful in its application to date, there is still the lingering feeling
that GIS could be far more important to the field than is currently recognized.
This feeling appears to have its origin in the way in which GIS was first ap-
plied as a locational modeling tool. Early models focused almost exclusively
on environmental parameters of land use (reflected in the regional environ-
mental analysis model theme) and, while these models had some limited suc-
cess, they clearly ignored other kinds of data that structure human use of the
landscape. The paper by Gaffney et al. (Chapter 8) expresses some of these
concerns in an eloquent manner. These investigators argue that the GIS has
not been as effective as a tool for archaeological research because it has been
tied, albeit unfairly, to theories of human behavior that emphasize functional
and economic determinism. Archaeologists have long been concerned with
how territories, boundaries, and regions may have been defined in the prehis-
toric past. Site catchment analysis, types of location modeling, and even pre-
dictive models have been employed to this end. There is considerable suspi-
cion, though, that many of the approaches and data types we have adopted to
define these constructs have actually imposed a structure on the past that is
not isomorphic with what may have existed in prehistory. While we must rec-
ognize that this problem can never be resolved with archaeological data (in
the sense that we can ever empirically verify a prehistoric mental construct
like "territory"), it may prove possible to identify material correlates of those
constructs with some success. To date, most of the GIS-based approaches that
have worked with this problem have used various combinations of economi-
cally related variables (arable land, soil type, etc.) and sought boundaries through
the creation of Theissen polygons or some similar spatial method (see, for
example, Savage 1990a, b). In most cases, the application of these methods has
provided plausible insight into possible territorial boundaries. Yet it is the case
that humans as a species often define their territories using other criteria and
that economic hinterlands only capture these boundaries imperfectly if at all.
The challenge, then, is to identify those variables accessible to the archaeolo-
gist that could inform us about past conceptions of territory. The writers fur-
ther suggest that if the GIS is ever to see widespread use in archaeology, it will
have to find ways to accommodate very different theoretical perspectives.
Lock and Harris (Chapter 13) agree with this prescription and make it clear
that the GIS, to be used effectively, must be employed within a theoretical
perspective. They echo the concerns of Gaffney, Stancic, and Watson, and
they explore the use of viewsheds as a means by which archaeologists can be-
gin to conceptualize new ways to perform locational modeling. Viewshed analy-
sis a common method in landscape architecture, and a number of archaeolo-
gists have begun to employ it in their research . Viewsheds are simply graphi-
cal means of displaying points of view from any location on a digitized land-
scape, and GIS is superbly capable of creating them. There remains the ques-
tion, however, of just what we look at from where. An early answer to this
question was simply to compute viewsheds from each major archaeological
site type, overlay the viewsheds, and determine which of them overlap, which
are invisible from one another, and which simply do not seem to correlate
with others. To be charitable, this can be called exploratory data analysis, and
under many circumstances, this may be the most profitable means by which
insight into the past can be obtained. Being less charitable, these efforts are
little more than spatial analogs of the uninhibited data dredging using
multivariate statistical methods that characterized archaeology during the
1970s. In effect, the calculation of viewsheds can be used in lieu of think-
ing about the problem.
It is easy to criticize, however, and more difficult to offer useful advice.
Mortuary sites have been shown to be excellent candidates for viewshed analy-
sis, as is rock art. Art is well-known to be intimately tied to religion, world
view, and, in many instances, ethnic differentiation. While there have been
numerous distributional studies of rock art, there have been few attempts to
look at the viewsheds of different types of art across a regional landscape. The
hard part, though, is developing some set of expectations of how this art ar-
ticulates with perceptions of that landscape. Given our extensive ethnographic
analysis and understanding of rock art of a number of different cultures, rea-
sonable progress toward this goal seems achievable.
Yet another type of data that could be used to develop viewsheds is the line
of site. This is similar to the viewshed, but it is somewhat more specific to
particular orientations and goals. It is well known that, in many societies, reli-
gious sites, shrines, and habitation sites are placed so as to be in view of natu-
ral, not cultural, features of the landscape. In the Andes, for example, moun-
tain peaks, both small and large, are sites of religious activity, and complex
networks of lines of site between these peaks are known from the ethnohistorical
record. While many of these peaks are known to have archaeological sites atop
them, many other have not been explored. Using GIS and line of site, it may
be possible to predict which peaks are likely to have sites and if so, of what
type. Would it not be easier just to get photos or maps of these locations and
create lines of site more cheaply? It might be easier, but traditional methods
are unlikely to provide the insight necessary to such an approach. Seeing might
not be believing, but this sort of virtual vision is often of considerable value. A
Introduction 17
real problem with line-of-site analysis is one that commonly plagues

archeoastronomy. From any sort of monumental architecture, it is often very
easy to create a line of site to virtually any natural feature, such as a star. Which,
if any, of these is one that would have been perceived in the past cannot be
easily determined. Line of site analysis, therefore, may be more useful in circum-
stances in which direct historical analogy suggests that certain natural features
have a clear articulation with some type of land use.
Other cognitive approaches to the landscape may prove to be useful as well,
but again, much depends on whether or not a coherent theoretical perspective
can be developed that can then be integrated with a set of methods. For in-
stance, neo-Darwinian approaches to cultural variation are beginning to show
that there may well be developmental differences in human landscape percep-
tion when viewed over evolutionary time (Maschner, Chapter 10; Kaplan 1992).
Whether or not any of these will be visible in the archaeological record re-
mains open to serious question, but it is clear that GIS can help to resolve and
explore these issues should sufficient data be generated.
All of this is not meant to imply that traditional approaches to locational
modeling cannot be extremely useful and successful. The papers by Ruggles
and Church (Chapter 9) and Potts, Jorstad, and Cole (Chapter 12) amply at-
test to this. In many ways, their analysis of Late Horizon settlement patterns
in the Basin of Mexico is a classic application of GIS to an archaeological
problem. What is of greatest interest in this paper is that while in this instance
it is concerned only with the Aztec example, the methodology could easily be
employed in the study of any imperial (or less complex) political system. Us-
ing this model, it is possible to develop comparative analyses of the "efficiency"
of ancient imperial states, and, by so doing, to gain deeper insight into how
these polities were organized. The paper by Potts and associates examines the
role of locational modeling from a very different perspective. The scale of
analysis in this instance is far smaller than the region, it involves a small area
surrounding already defined, very ancient archaeological sites. While the area
described in this paper—the Olorgesailie basin in southern Kenya—is much
larger than a single archaeological site, it is smaller than any of the regions
examined by the other contributors. This is unimportant, however, because
the goal of the modeling effort is the same—how to identify areas within
the basin that have a high probability of containing archaeological re-
sources of various time periods. The predictors in this case are a complex
mix of geological variables.
Conclusions
The GIS has a very bright future as a tool in anthropological and archaeologi-
cal research, and I think the papers in this volume have defined the ways in
which this development will take place. It is important to stress, however, that
as a tool, GIS and associated technologies are "theory-free," in that there is no
necessary isomorphism between a particular data type or category and the use
of GIS to solve or explore a problem. The GIS will be useful if the problem at
hand has a significant spatial dimension that has been carefully identified and
articulated by the researcher. Therefore, while it is unlikely that an anthro-
pologist interested in deconstructionism or postmodern approaches to culture
will find GIS useful, there is nothing intrinsic to GIS that makes this the case.
If space is somehow relevant to the postmodernist's problem, however, it is
quite probable that GIS could be used and used effectively.
Whatever the theory, space is an intrinsic property of life and society. If we
are interested in developing more reliable and robust quantitative idioms to
help us understand our "place" in space, the GIS has extraordinary advantages
to offer us, and I believe the next ten years will witness something of a rebirth
in the anthropological use of space, helped in great part by emerging GIS
systems and technologies.
Notes
1. In the United States, most archaeologists are trained in departments of anthropol-

ogy, and they consider themselves to be anthropologists first, then archaeologists.
However, since there are significant differences in the kinds of data obtained and em-
ployed by these modes of inquiry, it is useful to keep them distinct.
2. Good introductions to GIS can be found in Maguire, Goodchild, and Rhind (1991)
and Star and Estes (1990).
3. Driver and Kroeber (1932) did compute very large similarity matrices of trait list
data, but the technique was never widely adopted.
2
Land Degradation in the Peruvian Amazon:
Applying GIS in Human Ecology Research
William M. Loker
Land degradation, a reduction in the productive capacity of land, is a process
of increasing concern in the challenge to maintain and enhance global food
production. It is an especially critical problem in developing countries faced
with the need to increase food availability for growing populations. Billions of
dollars are invested in agricultural research and development aimed at increasing
the food supply. At the same time, land degradation threatens to reduce pro-
duction in large areas of agricultural land. While estimates of the magnitude
of the problem vary widely (see WCED 1987; WRI/IIED 1988; and Lai and
Stewart 1990 for recent reviews), there is a growing consensus that land deg-
radation is a serious and complex problem that merits increased attention from
both natural and social scientists.
A recent review of this topic by Blaikie and Brookfield (1987) highlights the
role of the social sciences in studying land degradation problems. According
to these authors, the term "land degradation" refers to a reduction in the ac-
tual or potential uses of land due to human activities (1987: 1). The costs of
land degradation ("the product of work on degraded lands is less than that on
the same land without degradation") make it a serious social problem for mil-
lions of farmers around the world and thus a priority for social science inquiry.
A central actor for understanding the causes and consequences of land deg-
radation is the land manager—most often the farmer—who makes the land-
use decisions for particular plots of land. Social science has a key role in un-
derstanding this process of decision making, including the social and ecologi-
cal contexts in which decisions are carried out. Anthropology's emphasis on
working with peasants, small farmers, and indigenous people holds out the
promise for important empirical and theoretical contributions in understand-
ing land degradation. A human ecology approach that focuses on the adaptive
strategies of individuals and groups and the environmental consequences of
these behaviors seems particularly well placed to contribute to this topic. Much
of the necessary research must be carried out "in the field," in close contact
with land managers, to observe the consequences and process of decision mak-
ing at first hand.
Recent studies of household decision making recognize the challenges in-
herent in modeling this process, including the need to understand
intrahousehold resource distribution and authority as well as the external con-
text that shapes the choices available at the farm level (see Gladwin 1980; Wilk
1989; Schminck 1984; Barlett 1980; von Braun and Pandaya-Lorch 1990).
The enormous data requirements for a comprehensive understanding of deci-
sion making have limited this type of study to microlevels—a household, vil-
lage, town, or other small group of people. Past efforts to measure land degra-
dation have likewise been restricted to microlevels: individual fields or experi-
mental plots. These microlevel data are extremely valuable in identifying the
factors involved and magnitudes of the impacts of agriculture in particular
environments. However, geographic extrapolation of results generated at the
microlevel to wider units of analysis is difficult or impossible.
Yet the formation of sound research and development strategies to address
land degradation frequently demands information from larger spatial scales.
Policy makers and development planners frequently need information gener-
ated at the regional level in order to make better decisions regarding the cu-
mulative effects of land degradation and the allocation of resources for ad-
dressing the problem. At present there are few reliable techniques for examin-
ing regional-level environmental impacts to guide policy and resource alloca-
tion decisions. This paper outlines a low-cost, objective means of analyzing
land-use patterns and variation in the natural resource base in order to iden-
tify specific geographic areas most in need of technical or policy interventions
aimed at ameliorating land degradation and its social effects.
Setting the Context: the Peruvian Amazon

The Peruvian Amazon is an area of about 75 million hectares (ha) located east
of the Andes mountains (Figure 2.1). For many years this selva area of Peru has
been perceived as having enormous agricultural potential (Belaunde 1959;
Hegen 1966). National governments have undertaken a series of initiatives
and invested millions of dollars in the development of infrastructure (mostly
penetration roads) in order to realize this potential. These policies have en-
couraged large-scale migration to the region, accompanied by expansion of
the agricultural frontier (Table 2.1). National statistics indicate the growing
importance of the selva in the agricultural economy (INE 1987)—the fruit, to
a certain extent, of investments made in the region (Table 2.2).
Table 2.1 Population growth in selected departments of the Peruvian Amazon.

Department Population Growth Rate
1949- 1961- 1972-
1949 1961 1972 1981 61 72 81
Loreto 152,457 272,933 375,007 482,289 2.8 2.9 2.8
Madre de Dios 4,950 14,890 21,304 33,007 5.4 3.3 5.0
San Martin 94,84? 161,763 224,427 319,751 2.6 3.0 4.0
Ucayali 16,154 64,161 120,501 163,208 6.8 5.9 3.4
PERU 6,207,967 9,906,746 13,538,208 17,005,210 2.2 2.9 2.5
Source: INEI 1990

Land Degradation in the Peruvian Amazon 21
21
3DcL:r*±s±
Figure 2.1 Map of Peru, indicating departments of the Peruvian Amazon.
In spite of the growing importance of the selva in the national economy,

serious questions persist regarding the long-term viability of agricultural settle-
ment of the region. Many observers see the growth of settlement and the pro-
cesses of land degradation that accompany agricultural expansion of the fron-
tier as undermining the long-term sustainability of agricultural production in
the region (Collins 1986; Bedoya 1986). Fundamental questions facing policy
makers and planners include the following: How serious is land degradation in
Table 2.2 Recional production of selected crops.
Crop Peru Coast Sierra Selva

Production/% Production/% Production/% Production/%
Rice 977,043 / 100 651,669/66 3,531/0.4 321,843/33
Maize 599,684/100 168,747 / 28 220,020/36 210,916/35
Beans 17,075 / 100 7,173/42 3,154/19 6,748/39
Yuca 644,259/100 1 9,246 / 3 20,686 / 3 604,327/94
Platano 835,160/100 35,517 / 4 76,167 / 9 723,476/87
Notes: Selva includes lowland and highland Amazonian regions (see footnote 3).
Source: INE/ENAHR 1986
22 Anthropology, Space, and Geographic Infonnatim Systems
the Peruvian Amazon? What areas are most affected? What sort of agricul-
tural activities are being carried out in these areas? What are the priorities for
agricultural research and development to address this problem?
Methodology
To understand the environmental consequences of agriculture at the regional

level, we need a method that can integrate two sets of data: (1) information on
characteristics of the natural resource base (such as soils, vegetation, etc.) and
(2) data on patterns of agricultural land use.
There is ample experience in the field of resource inventories used to guide
the planning and development process. The studies of ONERN (the National
Office of Natural Resource Evaluation) in Peru are excellent examples of this
type of indispensable work (ONERN 1962,1982,1986). It is also not unusual
to carry out socioeconomic surveys (diagndsticos socioeconomicos) in the course of
planning and implementing development projects. What has generally been
lacking are efforts that integrate these two approaches to understand the in-
teraction of human populations with their environments.
Both natural resources and human activities share the common character-
istic that they have a spatial dimension. Spatial analysis—supported by com-
puterized GIS—can provide a framework for the simultaneous consideration
of natural resources and human activities. Therein lies the appeal of a geo-
graphic analysis for gaining a better understanding human-environmental
interrelationships.
Given the rather daunting data requirements for a clear understanding of
household decision making and the vast area of the Peruvian Amazon, a de-
tailed study of the human ecology land use in the entire region would take an
army of researchers many years to complete. This is clearly impractical; policy
makers need reliable information quickly and at a reasonable cost to make
informed decisions. The approach taken in this study is to examine the out-
comes of land-use decisions, as reflected in statistics and observations of land
use, rather than the decision-making process itself, which remains something
of a "black box." Admittedly such a study will not be able to answer the question
of why certain land uses are chosen over others. However, if properly carried out
it should be able to answer important questions regarding where agriculturally
induced degradation is taking place and provide preliminary information on what
sort of activities are provoking degradation and how. This information can, in
turn, prioritize particular zones and shape hypotheses for further research.
Crop Distribution Data
The outcomes of land-use decisions are reflected in agricultural statistics on

the cultivated area of various crops. The agricultural statistics of Peru pre-
sented several difficult challenges (see Loker 1989 for details) and were prob-
ably less well organized than those of most Latin American countries. The last
agricultural census in Peru was carried out in 1973 and is clearly out of date
for use in a dynamic area like the Peruvian Amazon. Collecting reliable infor-
mation required travel to local statistical offices to get detailed information,
which was then cross-checked by field visits and interviews with key infor-
mants such as local agricultural researchers and farmers. These field visits also
permitted a first-hand look at the predominant farming systems in use to un-
derstand how these crops were grown. For example, rice is grown in three
very distinctive agronomic and ecological contexts: under irrigation, dry rice in
upland areas and along major rivers in seasonally exposed areas (regionally termed
barriales). These differences are not reflected in agricultural statistics; thus field
research was an indispensable component of the overall data-gathering process.
The five crops studied in detail are rice, maize, beans, yucca, and plantains.
Their distribution has been calculated and mapped in the departments of
Loreto, Madre de Dios, San Martin, and Ucayali— departments that cover
almost all the selva baja and part of the selva alta. The crops were chosen based
on three criteria: economic importance, available statistical information, and
the fact that, as annual or semiannual crops, they closely reflect recent land-
use decisions taken by farmers. An historical series (1980-88) of area culti-
vated, production, price and total value of each crop, listed by province, for
the departments studied was created (see Loker 1989). Table 2.3 presents the
data on area cultivated of the five crops studied in 1988 by department.
After collecting, analyzing, and adjusting the data on cultivated area and
production for the crops studied, the next step was to map the distribution of
crops. This mapping process demanded disaggregated data in order to pin-
point the location of cropping activities as finely as possible. Even the rela-
tively disaggregated data collected from local statistical offices often covered
relatively large areas in the selva. Several sources of ancillary information were
involved in determining more closely the distribution of cultivated areas.
Field visits were made to the four departments to discuss agricultural ac-
tivities with local researchers and observe local agricultural patterns. These
discussions were often carried out over maps of the region in order to locate
principal agricultural regions within the local areas. These visits also served to
discuss the predominant production systems in the region, including principal
crops, degree of input use, and predominant environmental problems associ-
ated with agricultural activities.
Table 2.3 Cultivated area of rice, maize, beans, yucca, and plantains in the departments
of Loreto, Madre de Dios, San Martin, Ucayali: 1988.
Area Cultivated (ha)

Department Rice Maize Beans Yuca Platano TOTAL
Loreto 17,096 22,299 2,317 8,530 10,350 60,592
Madre de Dios 5,429 2,932 213 640 499 9,713
San Martin 30,473 61,555 4,262 2,098 3,345 101,733
Ucayali 6,025 11,748 1,887 2,456 3,486 25,602
TOTAL 59,032 98,550 8,693 13,742 17,730 197,747
Source: Author's data derived from Ministry of Agriculture statistics

Mapping of crop distribution was also guided by demographic information
for the region, particularly distribution of the rural population. Population
data and projections were obtained from national statistics (INE-DGD 1985)
for the 1980-90 period. In mapping crop distributions, it was assumed that
agriculture occurs where people are, in particular where the rural population
is located, an assumption supported by limited statistical evidence on rural
employment patterns (INE/ORELORETO 1987; INE/OREMAD 1988).
The result was a "dot map" of the area under cultivation of the five crops
studied in the study area (see Figure 2.2). The crops were mapped in 100-ha
units. Placement of the symbols on the map was made as precisely as possible
given the limits of the data. Yet there is a certain amount of leeway as to where
a particular dot may be placed on the map. It is assumed that the current crop-
distribution map is sufficiently accurate given the 1:1 million scale of the study.
This study does not consider the extent or environmental impact of coca.
As is well known, illegal coca cultivation has expanded in recent years, acquir-
ing enormous economic importance in certain areas of the Peruvian Ama-
zon—especially the selva aha. Estimates of the actual area planted to coca vary
widely—from 100,000 to 500,000 ha. Coca was not included in the study for
several reasons: a preference to focus on legal activities, difficulty in gathering
Figure 2.2 Cultivated area of crops studied in the Peruvian Amazon.
I<3L3^"is!±
accurate information (Table 2.4 shows the varying estimates of coca cultiva-
tion in the Peruvian Amazon), and because significant areas of coca cultivation
lie outside the formal bounds of the study area [above 900 meters above sea
level (masl), and outside the Departments of Loreto, Madre de Dios, San
Martin, and Ucayali]. The environmental impact of coca cultivation and pro-
cessing has been discussed by Dourojeanni (in press).
Another important land use not considered in this study is pastures. Pas-
tures are not dealt with in detail due to the lack of reliable statistics on the area
in pastures in the Peruvian Amazon. Local statistical offices warned of the
unreliability of pasture and livestock figures. The small amount of data en-
countered in the course of researching this topic has been organized into Table
2.5, which includes information from some regions outside the study area.
Where relevant, these and other sources of information will be drawn on to
illustrate the role and importance of pastures in the Peruvian Amazon.
An alternative to the use of agricultural statistics would be the use of re-
mote sensing imagery to map land-use patterns. This strategy was not chosen
because of the expense of obtaining coverage for the entire Peruvian Amazon
and the difficulty, due to the prevalence of cloud cover over much of the area,
of obtaining even roughly contemporaneous images.
Natural Resource Distribution
The second set of information needed for the methodology proposed here is
data on the distribution of natural resources—in particular those elements of
the natural environment most relevant to agricultural production. This paper
uses an the agroecological study of the South American lowlands published by
CIAT and EMBRAPA entitled Land in Tropical America. In the case of Peru,
the basic information used in the CIAT-EMBRAPA study is derived from stud-
ies conducted by ONERN, complemented by other local sources. This infor-
mation is spatially limited and was extrapolated regionwide based on available
satellite and air photo imagery.
Land in Tropical America, covers 820 million ha at the 1:1 million scale, in-
Table 2.4 Estimates of the extent of coca cultivation, Peruvian Amazon.

Region Estimated Area Cultivated (ha)
San Martin 45,000 - 300,000
Huanuco 13,200 - 100,000
Cuzco (Quillabamba) 15,000 - 20,000
Junin-Pasco 5,000 - 15,000
Total Peru 100,000 - 500,000
Sources: PEAH-OSE, 1988

Ministerio de Agricultura, Region Agraria XIII (San Martin), n.d.
Que Hacer, No 59 (June-July 1989)
Personal Communications
Table 2.5 Pastures, livestock information, Peruvian Amazon.

Region Area in Livestock Population (head)
Pastures (has.)* Cattle Swine Poultry
Loreto 30,000 32,000 n.d. n.d.
Madre de Dies 10,276 20,355 7,533 200,352
Ucayali 38,120 21,953 47,840 21,000 1,144,000
San Martin 81,102 101,756 100,000 2,500,000
Aha Huallaga 10,238 25,138 28,598 20,908 426,633
Alto Mayo 426 5,300 13,747 n.d. n.d.
Baja Mayo-Huallaga
Central 40,000 71,670 n.d. n.d.
Palcazu 12,177 13,139 3,005 11,801
Satipo/Chanchamayo 14,343 13,000 7,485 n.d.
TOTAL 207,971 228,090 >159,920 >4,283,000
* First column refers to "Pastes naturales," second column to "Pastes
cultivados"; where only one figure is listed, source did not distinguish between
"naturales"and "cultivados."
Sources: Loreto: personal communication, Oficina Regional de Estadistica,
Ministerio de Agricultura
Madre de Dios: personal communication, Oficina Regional de Estadistica,
Ministerio de Agricultura
Ucayali: Ministerio de Agricultura, Region Agraria XXIII (Ucayali), 1987.
San Martin: ONERN-PNUMA (in press) and Ministerio de Agricultura,
Region Agraria XIII, mimeo
Alta Huallaga: PEAH-OSE 1987; Alto Mayo: Mimeo, INADE-APODESA
Bajo Mayo-Huallaga Central: ONERN-PNUMA, in press.
Palcazu: INADE-APODESA, 1988; Satipo/Chanchamayo: Cerron Rivera 1985
eluding Peruvian territory east of the Andes below 900 masl. The study de-
fines land systems, which are "an area or group of areas throughout which
there is a recurring pattern of climate, landscape and soils." (Cochrane et al.
1985: 2). Specifically, the variables studied include topography, hydrology,
vegetation, physical and chemical soil characteristics, temperature, rainfall,
potential evapotranspiration and other climatic factors. Fifty-nine land sys-
tems, covering a little over 61 million ha, fall wholly or partially within the
study area (see Figure 2.3). Table 2.6 lists the land systems, their areal extent,
and their approximate location in the study area. These land systems are far-
ther divided into land facets that describe the topographic and edaphic varia-
tion within the land systems.
The CIAT-EMBRAPA study provides basic information on natural resources
that most immediately affect agricultural potential of a given region. This in-
formation is stored in grids of cell size 5 cm by 4 cm, an area of about 9.25 by
7.4 km at this latitude. The resulting 1:1 million scale map provides a coarse-
grained view of resources, ignoring much of the spatial variability to which
agriculturalists genuinely adapt (see Moran 1990 for discussion of scale in
human ecology studies). But at present it is the most comprehensive study of
agricultural resources in the region and the scale seemed suitable for the re-
gional-level study contemplated here. In fact, one of the goals of this research
was to test the suitability of this database as a research and planning tool.
Figure 2.3 Land systems of the Peruvian Amazon.
Effects of Crops on Resources: Analytical Procedures
The next step in the analytical process was to relate the crop distribution to
the natural resource base. It was here that the GIS greatly facilitated the analysis.
The GIS is capable of taking various sets of spatially referenced data and trans-
forming them to create new data sets that can be translated into computer-
generated images.
The impact of agricultural activities is a combination of the sensitivity of
the land base to degradation (based on inherent agronomic characteristics)
and the nature of the cropping activities. Therefore it is necessary to classify
production systems in terms of their impact on the environment, given the
latter's susceptibility to degradation. Production systems in the region were
classified into three basic types: irrigated, riverine, and upland.
Irrigated systems are rice-based with little rotation of crops or fields, make
extensive use of chemical inputs (fertilizer and biocides), and are concentrated
28
Table 2.6 Area and location of land systems, Peruvian Amazon.
System Areas (has.) Location
220 9,229,437 Putumayo Plain

224 753,201 Putumayo River
228 584,991 Small scattered pockets in interfluves of Ucayali, both E & W sides
370 1,515,172 Large interfluve E side Ucayali, source of numerous rivers into Brazil
372 1,958,720 Interfluves around Rio Purus
373 803,441 Interfluves below 372 along Purus into Brazil
375 209,264 Rio Alto Purus into Brazil
400 141,076 Interfluve N of Madre de Dios (mostly Bolivia)
405 294,768 Rio Madre de Dios into Bolivia
408 1,298,356 Interfluves around Madre de Dios, Tambopata, Inambari
413 6,680 Interfluves upper Rio Madrid into Bolivia
803 369,468 Interfluve Morona/ Santiago
804 1,235,383 Interfluve Morona/ Santiago; Interfluve Maranon/Huallaga W of Tarapoto, Alto May
805 130,010 Rio Morona
806 4,938,683 Interfluve Pastaza/Morona, Pastaza/Tigre
807 1,602,718 Interfluve Tigre/Pastaza
808 2,669,975 Interfluve Napo/Tigre
809 4,147,596 Large area of interfluve above 808
810 2,774,281 Huallaga uplands, Interfluve middle Huallaga/Ucayali level of Juanjui; Interfluve
Huallaga/Aguaytia
811 1,038,554 Small pocket off Napo; numerous small pockets scattered near major rivers
812 862,378 Lower Rio Maranon
813 308,300 Lower Rio Napo
814 41,079 Tributary to Napo
815 390,775 Upper Napo/Curaray
816 123,233 Rio Tigre (tributary to Maranon) & Rio Corrientes
817 116,386 RioPastaza
818 355,504 Lower Maranon at confluence of Ucayali
819 259,723 Maranon above confluence w/Huallaga
820 191,096 Lower Huallaga near Yurimaguas
821 532,544 Lower Ucayali, N of 6 degrees S
822 741,138 Lower Ucayali, S of 6 degrees S to confluence of Pachitea
823 54,406 Small tributary, W side Ucayali
824 2,570,874 Large interfluve E side Ucayali
825 2,183,258 Interfluve Maranon/Huallaga, Huallaga/Ucayali near Yurimaguas
826 1,024,554 Scattered pockets along major rivers, E bank Ucayali, interfluve Ucayali/Aguaytia,
& scattered small pockets
827 1,735,187 Interfluve lower Maranon/Ucaya
828 1,720,457 Interfluve Ucayali/Tapiche also scattered along Ucayali & Ucayali/Huallaga interflv
Interfluve Ucayali/Pachitea/Aguaytia & scattered small pockets
829 319,495 Scattered pockets along major rivers
830 270,621 Terrace W bank Alto Ucayali near Atalya
831 784,380 Interfluve between small tributaries E side Ucayali
832 261,985 Rio de las Piedras/Tahuamanu
833 5,373,303 Interfluve E of Urubamba to Rio de las Piedras and Brazilian border
834 1,369,545 Pockets E of Ucayali and source of E tributaries, also interfluve Urubamba/Tambo
835 998,314 Andean foothills in south, headwaters of Inambari, Tambopata (shared w/Bolivia)
836 335,022 Urubamba, Tambo, small tributaries E Ucayali and Upper Rio Madre de Dios
837 27,106 Long narrow zone between Ucayali & Jurua
838 494,056 Pachitea, Alto Ucayali to 836, confluence Urubamba/Tambo
839 648,031 Interfluve Pachitea/Ucayali, Tambo/Urubamba
847 326,976 Rio Santiago plain; scattered pockets in 810, tributaries to middle Huallaga
848 68,029 Huallaga below Juanjui & small tributary to Ucayali
850 95,043 Huallaga above Juanjui to Tingo Maria
851 386,345 Large pocket E of Ucayali near confluence w/Pachitea
854 255,668 Pocket along Bolivian border between Iberia-Inapari
855 190,828 Tarapoto valley middle Huallaga & Bajo Mayo
in San Martin, particularly in the Alto Mayo region. In contrast, riverine sys-
tems are concentrated along the lower courses of major rivers (Maranon,
Ucayali, Napo) and are characterized by production of rice and beans in annu-
allyfloodedlands (barriaks) with maize, plantains, and yucca on levees (restin
with very limited production in upland areas away from the rivers. Finally
upland production systems are found along roads on nonalluvial soils and fo-
cus on the production of annual crops (upland rice, maize, beans or cowpea,
yucca, and plantains), frequently including a pasture component for dual-pur-
pose (milk and beef) cattle raising. The riverine and upland systems are char-
acterized by shifting cultivation with varying crop-fallow periods and very lim-
ited use of chemical inputs. Thus, these systems are very dependent on the
inherent local agroecological conditions.
The sensitivity of land systems can be assessed in terms of their erodibility
and their fertility. This responds to the two major forms of land degradation
prevalent in the Peruvian Amazon: deterioration of soil quality through physi-
cal processes (erosion) and soil biochemical degradation or nutrient loss through
leaching, harvest, etc., loss of organic matter, and exchangeable bases and re-
lated processes. The information on land-system characteristics was used to
generate erodibility and fertility scores for each land facet in the study area.
Figures 2.4 and 2.5 are computer-generated maps of the erodibility and
Figure 2.4 Erodibility of land systems, Peruvian Amazon.

fertility of lands in the Peruvian Amazon. Figure 2.4 displays land systems
with very high or high credibility. Note that the land systems with the highest
credibility are located along the western margin of the basin, in the Andean
foothills. This is the selva aha, which is also the area which has undergone the
most intensive colonization and agricultural exploitation in the Peruvian Ama-
zon. (Significant portions of the selva aha lie outside the area covered by this
study—i.e., the selva of Junin, Pasco, Cuzco, and Puno.) The other areas of
high to very high credibility are upland areas on the east side of the Ucayali
and the uplands surrounding the Rio de las Piedras in Madre de Dios.
Figure 2.5 is a map of soil fertility and classifies land systems in terms of the
percentage of their area that is of high to very high fertility or low to very low
fertility. Areas that are of average fertility are represented as white. Notice
that significant areas of the most fertile land are located along major river
courses. This conclusion is not surprising; alluvial lands are known to be gen-
erally more fertile than uplands. But it is worth noting, as it appears to vindi-
cate the method used to generate the fertility rankings.
Figure 2.5 Fertility of land systems, Peruvian Amazon.
Icl:*:;* i.s' i
Results
Crop Distribution by Land System
The first step in relating the impact of agricultural practices on natural re-
sources is to determine the distribution of cultivated area relative to land sys-
tems: in which land systems are agricultural activities concentrated? The geo-
graphic information system permits the cropping patterns to be overlaid on
the land systems and a count of the number of hectares under cultivation by
land system can be generated. A key assumption of this study is that, all other
things being equal, land degradation will tend to occur where cultivation is
concentrated in land systems of high credibility and/or low fertility.
Table 2.7 lists the distribution of cultivated area by land system. The culti-
vated area is broken down by three types of rice (irrigated, floodplain, upland)
plus the other four crops and the total cultivated area. Table 2.7 illustrates that
there is a range of cultivated area in the land systems.
The distribution of total cultivated land in the crops studied is presented in
Figure 2.6, a computer-generated orthographic projection of cultivated area
where the "peaks" represent area under cultivation. Note the concentration of
peaks in the western portion of the figure; this area corresponds to San Martin.
Table 2.8 lists the fifteen most heavily cultivated land systems, their loca-
tion, and the type of cropping activities present. The five most extensively
cultivated land systems (804, 810, 822, 847, and 855) account for about 60%
of the area cultivated in the study crops. Four out of five of these land systems
are located in San Martin. The ten most extensively cultivated land systems
(the five mentioned plus 812, 826, 820, 825, and 828) account for 76% of the
total cultivated area. Including an additional five land systems representing
Fgure 2.6 Orthographic projection of total cultivated area of study crops.

32
Table 2.7 Cultivated hectares by land system.
System IRRIHA BARRHA SECRHA TOTRHA MAIZHA FRIJHA YUCAHA PLATHA CROPSHj"
804 18600 0 2900 21500 19800 1200 1000 1200 44700
810 1200 0 1100 2300 24800 2000 700 1000 30800
822 0 2900 300 3200 9600 700 800 1900 16200
847 1400 0 100 1500 9500 100 200 600 11900
855 2400 0 500 2900 6300 700 300 200 10400
812 0 2700 400 3100 1500 100 1500 3300 9500
826 0 100 1200 1300 5400 100 900 600 8300
820 0 0 1900 1900 1900 200 800 800 5600
825 0 0 1000 1000 2400 300 700 900 5300
828 0 100 200 300 2300 100 400 900 4000
806 0 0 1000 1000 1900 0 300 600 3800
838 0 400 200 600 700 600 900 400 3200
850 900 0 600 1500 1200 200 0 300 3200
105 0 0 1700 1700 900 100 200 200 3100
808 0 300 500 800 900 300 500 500 3000
408 0 0 1600 1600 700 200 100 300 2900
818 0 800 200 1000 800 200 400 400 2800
827 0 400 100 500 500 0 500 800 2300
833 0 0 600 600 700 400 300 200 2200
821 0 800 0 800 600 0 300 300 2000
370 0 100 0 100 1000 100 300 200 1700
220 0 700 0 700 100 100 200 500 1600
836 0 0 300 300 900 100 200 0 1500
819 0 0 800 800 500 0 100 0 1400
824 0 0 400 400 500 0 300 200 1400
813 0 200 100 300 200 0 300 300 1100
816 0 0 200 200 300 100 100 400 1100
851 0 300 100 400 300 0 100 300 1100
830 0 0 100 100 400 100 200 200 1000
854 0 0 400 400 200 200 0 0 800
224 0 200 0 200 100 0 100 300 700
829 0 0 200 200 400 0 0 0 fiOO
831 0 0 0 0 0 200 200 200 600
835 0 0 200 200 100 300 0 0 600
832 0 0 500 500 0 0 0 0 500
834 0 0 0 0 100 200 100 0 400
839 0 0 0 0 200 200 0 0 400
848 100 0 0 100 300 0 0 0 400
807 0 0 0 0 0 0 200 100 300
375 0 0 100 100 0 0 100 0 200
811 0 0 0 0 0 100 100 0 200
814 0 0 0 0 0 0 100 100 200
823 0 0 0 0 200 0 0 0 200
373 0 0 0 0 100 0 0 0 100
805 0 0 100 100 0 0 0 0 100
815 0 0 0 0 0 0 0 100 100
837 0 0 100 100 0 0 0 0 100
228 0 0 0 0 0 0 0 0 0
350 0 0 0 0 0 0 0 0 0
372 0 0 0 0 0 0 0 0 0
400 0 0 0 0 0 0 0 0 0
403 0 0 0 0 0 0 0 0 0
413 0 0 0 0 0 0 0 0 0
803 0 0 0 0 0 0 0 0 0
809 0 0 0 0 0 0 0 0 0
817 0 0 0 0 0 0 0 0 0
841 0 0 0 0 0 0 0 0 0
844 0 0 0 0 0 0 0 0 0
852 0 0 0 0 0 0 0 0 0
IRRIHA = Hectares of irrigated rice MAIZHA = Hectares of maize
BARRHA = Hectares of barrial rice FRIJHA = Hectares of beans
SECRHA = Hectares of upland rice YUCAHA = Hectares of yuca
TOTRHA = Sum of IRRI+BARR+SECRHA PLATHA = Hectares of platanos
CROPSHA = Sum of five crops studied
Table 2.8 Location and agricultural activities in most heavily cultivated land systems.
Land Total ha
System Location (Department) Cultivated Comments
SOT Alto Mayo (San Martin) 44,700 18,600 ha in irrigated rice, much oi
remaining area upland, shifting
cultivation
810 Middle Huallaga, south of 30,800 Upland areas south of Tarapoto,
Tarapoto (San Martin) 1,200 ha under irrigation
822 Ueayali from Pachitea to 16,200 Riverine, significant areas in maiz<
Contamana including Pucallpa (9,600 ha), river rice (2,900 ha),
(Ueayali) platanos (1,900 ha).
847 Tributaries to Upper Huallaga, 11,900 Mostly upland cultivation of maizi
e.g., Sapasoa, Billabo (San (9,500 ha), 1,400 ha irrigated rice,
Martin) much coca and pastures (not
included in cultivated area total).
855 Tarapota Valley (San Martin) 10,400 Mostly upland cultivation of maizi
(6,300 ha) 2,400 ha of irrigated
rice.
812 Confluence of Ueayali, 9,500 Riverine, significant areas of river
Maranon to Brazil border, rice (2,700 ha), platanos (3,300 ha).
Iquitos (Loreto)
826 Pucallpa-Lima road, to km 80 8,300 Upland cultivation of maize (5,400
(Ueayali) ha) rice (1,200 ha), much pasture.
820 Huallaga below Shapaja to 5,600 Mostly upland rice and maize
Lagunas, Yurimaguas (San (about 1,900 ha in each).
Martin & Loreto)
825 Uplands near Yurimaguas 5,300 Mostly upland rice and maize
(Loreto) (1,000 ha in rice, 2,400 ha in
maize).
828 Pucallpa-Lima road km 80 to 4,000 Upland cultivation of maize (2,300
Aguaytia (Ueayali) ha), some platanos (900 ha), river
rice (100 ha) on Aguaytia River.
838/ Upper Ueayali from Pachitea 4,700 Riverine and upland mixed; rice
836 to Atalaya (Ueayali) (400 ha river, 500 ha upland),
maize (1,600 ha), beans (700 ha).
405/ Puerto Maldonado, Inapari 8,200 Riverine and upland mixed; rice
408/ road to Cuzco, also Rios de las (3,500 ha river, 2,000 ha upland),
833 Piedras, Madre de Dios and maize (2,700 ha), yuca and
Tambopata (Madre de Dios) platanos (800 ha each).
contiguous geographic areas of moderately heavy cultivation (838 and 836 on

the Upper Ueayali and 405,408, 833 in Madre de Dios), accounts for 82% of
the total cultivated area of the study crops. (See Figure 2.7 for the location of
the fifteen most heavily cultivated land systems.)
Thus fifteen land systems covering a total of 13,934,536 ha (22% of the
land in the study area) contain 157,900 ha of cultivated land in the study crops
(82% of the total area cultivated in the five study crops). The remaining 44
land systems covering 47,189,589 ha (78% of the land) have relatively small
areas under cultivation—a total of 35,800 ha in the study crops (18% of the
total cultivated area in the crops studied). Cropping activities are concentrated
on a relatively small proportion of the lands in the Peruvian selva—basically
those lands adjacent to major rivers and along roads, especially in San Martin.
This enables us to focus our attention on the areas of more concentrated crop-
ping activities and examine the environmental effects of agriculture in these
Figure 2.7 Location of fifteen most heavily cultivated land systems.
areas. We can assume that agriculturally induced land degradation is either

highly localized or absent in the 44 land systems with relatively little cultivation.
The question then becomes, of the extensively cultivated land systems, which
ones are most likely to be undergoing processes of agriculturally induced land
degradation due to some combination of low inherent fertility, high credibil-
ity, and inappropriate land use practices.
Impacts of Agriculture by Land System
We can begin to answer this question by looking at the data in Table 2.9 and
Figures 2.8 and 2.9. Table 2.9 lists the fifteen land systems where cultivation
activities are concentrated and their credibility and fertility ranks. This infor-
mation is listed by noting the percentage of each land system that is of very
high, high, average, low, or very low credibility and fertility. Figure 2.8 is a
graphic representation of the information in Table 2.9 generated by a GIS.
Table 2.9 Erodibility and fertility of agriculturally important land systems.
Land Erodibility (Percent Area) Fertility
System VH HI AV LO VL VH HI AV LO VL
804 0 92 8 0 0 0 0 100 0 0
810 94 0 6 0 0 0 0 100 0 0
822 0 0 100 0 0 28 0 72 0 0
847 0 0 100 0 0 0 0 100 0 0
855 25 50 25 0 0 50 0 50 0 0
812 0 0 50 50 0 0 0 100 0 0
826 0 0 100 0 0 0 0 68 32 0
820 0 0 100 0 0 0 0 100 0 0
825 0 92 8 0 0 0 0 0 100 0
828 0 90 10 0 0 0 0 100 0 0
836 0 0 100 0 0 0 100 0 0 0
838 0 0 100 0 0 0 40 60 0 0
405 0 0 30 70 0 0 0 70 30 0
408 0 0 80 20 0 0 0 80 20 0
833 90 0 10 0 0 0 0 10 90 0
The image overlays the information on credibility on the "peaks" of culti-

vated area. The peaks outlined in red hues indicate cultivation of land systems
with significant amounts of erodible land. San Martin is notable in this regard.
Figure 2.9 relates the information on area under cultivation to the fertility of
the land systems. Note that the San Martin region is relatively fertile land, as
are the cultivated alluvial lands (e.g., land system 822—Ucayali around
Pucallpa). Some fertility problems manifest themselves in the uplands west of
Figure 2.8 Overlay of credibility of land systems, cultivated area.

Percent Area in i>pri Systems with Erosion Hazard
Superimposed upon : Distribution of Crops Studied, Peruvian flmazon
IcfLarrLsi
Low Fertility fand Systems, Peruvian Amazon

Superimposed upon : Distribution of Crops Studied, Peruvian Amazon
-, - - — K M
TfHTK*-| SS!T-
Figure 2.9 Overlay of fertility of land systems, cultivated area.
Pucallpa (land system 826), the Yurimaguas uplands (land system 825) and in
Madre de Dios (land systems 405, 408, and 833).
The same information is conveyed in Table 2.9. Examining the first group
of five land systems, we can note that all of the land in these areas is of average
or above average fertility. However, the three of the four land systems located
in San Martin (804, 810, and 855) also have major portions of highly and very
highly credible land. Thus we can expect that these lands are susceptible to
degradation due to erosion. These land systems represent 4,218,492 total ha
and 85,900 ha of cultivated area in the crops studied. Two additional land
systems in the first group of five—847 and 822—appear to be less susceptible
to degradation than the three just discussed. Both land systems represent river
valleys: 847 the tributaries to the Huallaga and 822 the middle Ucayali.
The second tier of land systems also presents a mixed situation regarding
risk of degradation. The two alluvial land systems (812, Lower Ucayali-
Maranon around Iquitos, and 820, Lower Huallaga around Yurimaguas) are
relatively less likely to undergo agriculturally induced land degradation due to
their average to low credibility and average fertility-—fertility that is renewed
periodically through flooding. The upland land systems (826, km 15-80 on
the Pucallpa to Lima road; 825, uplands around Yurimaguas and between the
Ucayali and Huallaga; and 828, km 80-150 on the Pucallpa to Lima road)
appear more susceptible to degradation, though the underlying factors differ
among these areas. Land system 826 is not susceptible to erosion (average
credibility); however, a significant portion of its territory is of low fertility
(32%). Land systems 825 and 828 differ from 826 in that they are much more
susceptible to erosion, due both to topography and soil physical characteris-
tics. Land system 825 is also characterized by below-average fertility.
There are two other centers of cultivation that remain to be discussed—the
Upper Ucayali between Atalaya and the Pachitea (land systems 836 and 838),
and the Puerto Maldonado area (land systems 405, 408, and 833). Land sys-
tems 836 and 83 8 are both predominantly alluvial. They are not susceptible to
erosion and are of above-average fertility. This area seems to present little risk
for degradation. However, should cultivation expand outside of the valley
on the east bank of the Ucayali, there would be significant prospects for
land degradation.
The Puerto Maldonado area is a mixture of alluvial and upland environ-
ments in close proximity to one another. Land systems 405 and 408 contain
both types of land while 833 is predominantly (90%) upland. Land systems
405 and 408 are susceptible to degradation due to the significant portions of
low-fertility land. In reality the 70 to 80% of "average-fertility" land in these
land systems is below average, falling just outside the cutoff point of "low-
fertility" land, so the situation is more precarious than Table 2.9 indicates.
Land system 833 has a high potential for degradation due to its large amounts
of very highly credible land and its low fertility.
Discussion
Based on this analysis, four areas, comprising nine land systems, can be iden-
tified as "at risk" in terms of agriculturally induced land degradation. Six ex-
tensively cultivated areas, located on flatter and more fertile alluvial soils, are
predicted not to be undergoing a process of land degradation. The four
"at risk" areas are:
Area 1: the Alto Mayo-Tarapoto middle to upper Huallaga region (land
systems 804, 810, and 855)
Area 2: the area adjacent to the Pucallpa-Lima road from km 15-150
(land systems 826 and 828)
Area 3: the Yurimaguas uplands (land system 825)
Area 4: the Puerto Maldonado area, particularly north of Puerto
Maldonado along the Puerto Maldonado-Iberia road (land systems 405,
408, and 833).
This listing corresponds to the order of importance of these areas in terms
of agricultural production, the areal extent of land degradation problems, and
the number of people whose livelihoods are affected by land degradation.
Testing the Results: A Brief Review of Existing Research
A major question arising from this research is the degree to which it actually
succeeds in identifying areas undergoing land degradation. While it is not pos-
sible to answer this question definitively until field research designed specifically
to test the results of this research is carried out, a brief review of existing research
in these areas can begin to provide some insight into the utility of this method.
Area 1: This area encompasses most of the cultivated area of the Depart-
ment of San Martin. Agroecologically the region is atypical in that it has sub-
stantial areas of relatively fertile, nonacid (calcareous) soils even in uplands
away from major rivers. Rainfall is also quite variable in the region, ranging
from semiarid (850 mm per annum) to humid (>2000 mm). The region has
long been recognized as one of the most fertile in the Peruvian Amazon and
has been the object of numerous colonization and development initiatives.
Socioeconomically, the region is one of long-term settlement with relatively
high population density (ranging from 2.1 to 3 8.6 inhabitants/km2). In the late
1970s, road access to the region was improved with the construction of the
Carretera Marginal de la Selva, which facilitated a massive influx of settlers.
While much of this migration has been directed to the expansion of irrigated
rice on the relatively flat lands of the Alto Mayo region (the 8% of average-
erodibility lands in land system 804), there has also been a dramatic increase in
shifting cultivation of maize on the surrounding hillsides (see Foster Chaparro,
n.d., OIT/DGE 1984, for information on this process). Agricultural statistics
document a dramatic expansion of irrigated rice and maize cultivation in San
Martin; in the period from 1980-88 area in irrigated rice increased from about
4500 to over 20,000 ha, while the area in maize more than tripled (from 29,774
to 61,555 ha.) Maize and rice cultivation were stimulated by the provision of
low-interest credit, a guaranteed market, and price supports for maize (see
Cannock and Cuadra 1990 for a discussion of these issues). The local economy
has also been strongly influenced by the expansion of coca cultivation and, in
recent years, seriously affected by political and drug-related destabilization.
Recently, a multiyear study of land degradation in the region was carried
out by Peru's Office for Natural Resource Evaluation (ONERN) in connec-
tion with the United Nations Environmental Program (here referred to by its
Spanish acronym, PNUMA; see ONERN/PNUMA, in press). The study docu-
mented rapid deforestation in the region with the expansion of cultivation,
including areas deemed unsuited for cultivation by ONERN due to their steep
topography. Among the problems cited by the ONERN/PNUMA study are
excessive erosion, increased siltation of waterways, and invasion of substantial
areas by scrub vegetation, indicative of acidification and degradation of the
soil. Thus the method employed here independently corroborates the finding
of this field study, supporting its results. For our purposes, the ONERN/
PNUMA study provides empirical support for the method employed here to
detect and delimit land degradation.
Area 2: This area corresponds to the cultivated area along the Pucallpa to
Lima road, from 15 to 150 km from Pucallpa. Agroecologically, this area is
more representative of conditions in the selva baja of the Peruvian Amazon
and the Amazon Basin in general. Soils in the uplands are highly acid and
infertile. Topography is flat to undulating, with steeper slopes to the west as
one approaches the Andean foothills. Climate varies from subhumid (three
months with rainfall of less than 100 mm, total precipitation 1800 mm) to
perhumid (no months <100 mm of rain, total precipitation over 3000 mm).
Flat terrain is characterized by extensive areas of poor drainage occupied by
palm forest (known locally as aguajales—Mauritiaflexuosa). ). Well-drained up-
lands and uncultivated alluvial soils are characterized by species-rich tropical
forest (both true rain forest and semievergreen seasonal forest). Socioeconomi-
cally the area is also one of long-term colonization, having been opened up for
settlement with the construction of the Huanuco-Tingo Maria-Pucallpa sec-
tion of the highway in the 1940s. However, the real boom in settlement oc-
curred with the improvement of the highway in the 1960s, drawing massive
numbers of colonists in the 1960s and early 1970s, a period characterized by a
reform-minded government that introduced radical changes in the agricul-
tural sector. Settlement is confined to a narrow strip along either side of the
highway (varying from 2-10 km in width). Production systems are almost ex-
clusively shifting agriculture with annual crops and pastures. (Probably 70-
80% of the pastures—about 45,000 ha—found in Ucayali are located along
this road.) Farmers are highly market-oriented due to access provided by the
road and the presence of a large regional market in Pucallpa. The economic
mainstays of the region are forestry (extraction and milling) and commerce.
This region has been the site of extensive agricultural research on the part
of national and international agricultural research institutions. While research
has focused on a number of topics, one of the primary areas of interest has
been in the development of sustainable livestock production systems and the
recovery of degraded pastures (see CIAT 1988, IVITA 1989 for summaries of
this research). The long-term occupation of the margins of the road has led to
extensive deforestation and the growth of secondary forests and degraded pas-
tures on land that has been exhausted by slash-and-burn agriculture either by
itself or in conjunction with cattle production. Land degradation, mostly in
the form of chemical deterioration of the soil (nutrient loss) is thus recognized
as a major problem in Area 2. Again, this tends to confirm the results of the
methodology employed here.
Area 3: Agroecologically, Yurimaguas is similar to the Pucallpa region,

though soils are sandier and consequently better drained. Climate is more
uniform in terms of temperature and especially rainfall (about 2000 mm per
year and only one or two months <100 mm). Native vegetation is neotropical
rain forest. Socioeconomically, Yurimaguas is an old city by Amazonian stan-
dards, with a history of being an important transshipment point in the region.
It grew in size and importance with the advent of steamship transportation
and the rubber boom of the late 19th-early 20th centuries; Yurimaguas is the
headwaters for steamship travel on the Huallaga and an important transship-
ment point for goods headed down river to Iquitos and out of the region. Its
road connection to the rest of the country remains precarious, with a sea-
sonal dirt/mud road connecting Yurimaguas with Tarapoto and beyond.
The area is not dynamic in demographic or economic terms; population
growth is about 2.2% per year and cultivated area has only grown by about
13% in the last ten years.
The area has seen extensive agricultural research on the part of national
and international institutions (the North Carolina State University Tropical
Soils Program supported by US AID). The problem of land degradation in
the local area is widely recognized by agricultural scientists working in these
programs (NCSU 1986). The design of more appropriate agricultural systems
(using varying levels of inputs) is a major research thrust.
Area 4: Agroecologically, the Puerto Maldonado region (12 to 14° south

latitude) differs significantly from areas farther to the north due to somewhat
wider swings in temperature and more prolonged periods of cool tempera-
tures in the austral winter (June to September). The rainfall pattern is similar
to that of the Pucallpa region, with an even steeper gradient of increased rain-
fall in the Andean foothills to the west. (Rainfall in Puerto Maldonado aver-
ages 2345 mm with three months <100 mm; Quincemil in the Andean foot-
hills averages over 7000 mm of rain per annum.) Topography is generally roll-
ing to steeply sloping, with only limited areas of flat, poorly drained areas
away from the rivers and narrow river floodplains. Alluvial soils are limited in
extent and there is no seasonal cultivation of barriales due to their brief period
of exposure and unpredictable river water levels. Soils in the area are highly
variable, with uplands generally acid and infertile and better soils in alluvial
areas. The area is a patchwork of microenvironments caused by variation in
soils and topography. Native forest is extremely species-rich, with numerous
endemic species (Prance 1982). Socioeconomically, the region has been iso-
lated from developments in the upper Amazon, as its watershed drains south
and east into Brazil—away from Iquitos and Pucallpa, the major commercial
centers in the Peruvian Amazon. The region was economically important dur-
ing the rubber boom but languished thereafter. This isolation has diminished
with the penetration of the region by the Cuzco-Puerto Maldonado-Inapari
road in the 1960s. The road is only seasonally passable, particularly to the
north of Puerto Maldonado. Still, it has stimulated an influx of migrants from
the Cuzco region and the expansion of agriculture, including pastures, to meet
market demands. The area is unique in the Peruvian Amazon in the impor-
tance of gold mining to the economy; there is currently a gold rush occurring
in the region (similar to developments downriver in Brazil). The mining sec-
tor has both helped and hurt the agricultural economy of the region, provid-
ing a market for agricultural products but drawing labor away from the agri-
cultural sector, aggravating the chronic labor shortage that afflicts most of the
Amazonian region. Population is still very sparse in the region (density aver-
age 0.5 persons/km2), but the area is one of active colonization that will be fur-
ther stimulated if and when the road connection is improved. The Inapari-Puerto
Maldonado-Cuzco road is projected to link up with roads leading out of Rondonia
and eventually provide an all-weather road connection between Brazil and the
Pacific. This would be quite worrisome from an ecological point of view, given
this area's high propensity for degradation and the expected influx of settlers.
Agricultural research has been much less extensive in this region. A recent
study carried out by national environmental organizations has pointed to nu-
merous problems in the region, mostly connected with the mining sector
(FPCN/WWF 1989), though problems of land degradation are mentioned.
Existing research is inadequate to confirm the seriousness of land degra-
dation in this area.
The results of previous and ongoing research tend to support the current
study's identification of areas currently at risk of undergoing land degradation.
Those areas identified as undergoing land degradation have been confirmed
as problem areas by research carried out in each of the areas (with the possible
exception of Area 4). However, we must be alert to the possibility that the
method may have omitted significant areas of land degradation. There may be
significant areas of land degradation that were not identified as such by this
study. Therefore, any future research and development programs to counter-
act land degradation based on the results of this study should include some
sampling of areas outside those identified by this research to be sure that im-
portant areas of land degradation were not overlooked.
Conclusions
This paper has presented an objective, low-cost methodology for the detec-
tion and delimitation of areas susceptible to land degradation given current
land-use patterns in the Peruvian Amazon. This methodology was facilitated
by the availability of a low-cost GIS program capable of the simultaneous
manipulation and transformation of distinct sets of spatially referenced data.
In this the case, the GIS program handled data sets on the distribution of
natural resources and the distribution of agricultural activities in order to de-
lineate and map of areas at risk of land degradation in the Peruvian Amazon.
This process also allowed some preliminary hypotheses to be framed regard-
ing the nature of land degradation in each of the areas at risk. This method is
conceived as a first step in a multi-phase research program designed to address
problems of land degradation. While definitive testing of the research results
awaits further field research, the existing literature tends to Confirm the re-
sults of this study. Elsewhere I have suggested potential technological and policy
interventions to address the problems identified in this paper (Loker 1989).
The study encountered several problems in its execution. Data on agricul-
tural production derived from secondary sources are always somewhat sus-
pect; this problem is exacerbated by the weakness of the Peruvian institutions
charged with collecting agricultural statistics. Fieldwork and interviews with
local officials only partially resolved these problems. Agricultural data are spa-
tially referenced only at the district level, which frequently involves large, het-
erogeneous units in the Peruvian Amazon. Future efforts should aim at the
integration of remote sensing imagery—either from airborne and/or satellite
sources—to provide more accurate land-use data. The exclusion of coca and
pastures from the current study might also be remedied by using remote sens-
ing data. Also, this study has little to say regarding the degradation of land
under irrigated rice. Specialized studies of this agroecosystem are necessary to
determine whether land degradation is a problem in these systems and, if so,
to develop appropriate research strategies.
Despite its limitations, the study provides a good first approximation of the
environmental effects of agriculture in a very large region. With only a mod-
est investment of time and money, the method provides policy makers and
agricultural scientists with an overview of conditions in the region and focuses
attention on a reduced area for further research. The methodology described
makes extensive use of existing sources of information (a trait dear to the hearts
of most administrators and policy makers). However it also points up the ne-
cessity for strategic, well-focused fieldwork to resolve ambiguities in existing
data and provide crucial additional information on farming systems and crop
distribution not found in the secondary sources.
The GIS Difference
This study has demonstrated that GIS technology is an important tool to aid
anthropology and other disciplines in applied human ecology research. While
many of the analytical processes carried out in this study can, in theory, be
executed by hand using dot maps and overlay techniques, use of a computer-
based GIS presents several advantages. Unlike hand overlays, the GIS enables
the investigator to manipulate the scale and dimensions of the images being
analyzed to discover the best format for their presentation. For example, it
was possible to experiment with various degrees of vertical exaggeration in the
preparation of orthographic projections to discover which presented the data
most clearly. One of the key advantages of using the GIS is that it facilitates
the rapid preparation of maps for conveying related sets of information. Maps
can be prepared quickly and selected or discarded depending on their ability
to communicate critical information. For example, it was easy to represent
various combinations of crops (all crops, single crops, irrigated vs. nonirrigated
crops, upland vs. irrigated and riverine crops) in order to see what particular
crops and farming systems were most responsible for causing land degrada-
tion. Once baseline maps are created, the process of digitizing additional data
is often more rapid than representing the same data by hand. As with the use
of computers in statistics, the GIS opens up new analytical possibilities simply
because it performs numerical transformations so much more rapidly than can
be done by hand. This allows the investigator to explore multiple questions
that would be too cumbersome or time-consuming to explore by hand. As
with the impact of computers on statistical analyses, this increases the possibility
of the misuse of concepts and analytical techniques. But on balance, the fact that
the GIS gives researchers new analytical capabilities is a positive development.
In addition to asking what GIS can do for a particular research problem, it
is interesting to contemplate how different interpretive frameworks use the
information generated by a GIS. In the present case, the human ecology frame-
work examines the reciprocal relationship between human beings and the natu-
ral environment. Rather than restricting ourselves to questions regarding how
environment shapes human adaptation, this research asked how human activi-
ties are affecting the natural environment. To the extent that human activities
decrease the productive potential of the environment, they generate new prob-
lems of human adaptation and affect more inclusive biological and biogeochemi-
cal cycles. The GIS is a technology capable of implementing various tech-
niques of spatial analysis; these proved useful in answering the questions gen-
erated by this research. As research progresses to asking "why" questions of
land use and decision making, we may begin to stretch the capacity of current
GIS technologies and analytical methods—pushing the technology and meth-
ods in new directions in response to new research questions. It is here that
particularly close collaboration between disciplinary specialists—geographers,
computer scientists, and anthropologists—will be necessary.
Finally, GIS encourages researchers to "think spatially," something geog-
raphers do as a matter of course. Virtually any human attribute that is differ-
entially distributed across geographic space is amenable to analysis with a GIS—
including phenomena such as ethnicity and language as well as aspects of ma-
terial culture. As such, the GIS represents an important tool for conceptualiz-
ing problems, analyzing data, and presenting results effectively, with potential
applications to many problems in human ecology and anthropology in gen-
eral. As the GIS becomes better known and more widely adopted in anthro-
pology, it may lead to increasing convergence between anthropology and ge-
ography. Geographers and anthropologists often approach similar questions
from different perspectives; by combining their efforts, they may well gener-
ate significant new insights into human behavior.
Acknowledgments
The research for this paper was carried out while the author was a Visiting Research
Fellow at the International Food Policy Research Institute (IFPRI) in Washington,
D.C. and a Postdoctoral Fellow of the Centre Internacional de Agricultura Tropical
(CIAT) in Cali, Colombia. Funding for the research was provided by the Rockefeller
Foundation, CIAT, and IFPRI. Essential technical and intellectual support was pro-
vided by the Agroecological Studies Unit of CIAT, especially Drs. Simon Carter and
Peter Jones. The research greatly benefited from discussions with Drs. Filemon Torres,
Jose Toledo and Carlos Sere of CIAT and Drs. Stephen Vosti and Nurul Islam of IFPRI.
Thanks also to Keneth Reategui of CIAT and Jorge Vela of the Peruvian Institute
Nacional De Investigation Agraria y Agroindustrial (INIAA). I also thank numerous
other Peruvian colleagues who aided in data gathering and interpretation. Any
errors are my own.
3
The Use of GIS to Measure
Spatial Patterns of Ethnic Firms
in the Los Angeles Garment Industry
Christopher G. Arnold and Richard P. Appelbaum
The purpose of this chapter is to describe the methodology and present some
initial results in our efforts to understand the role of spatial organization in
ethnic economies of the Los Angeles garment industry. Our research goal was
to provide a spatial dimension to our database in order to test theories main-
taining that space is a critical component of economic transactions. To this
end, we created analytical grids of various resolutions using three GIS func-
tions: address matching, polygon generation, and identity overlay. The grids
were then transformed into matrices and tested for spatial autocorrelations
using Statistical Analysis System (SAS) routines proposed by Griffith (1992).
The following section presents brief arguments for the importance of space
from the perspectives of industrial district and ethnic economy theories. The
third section describes the research site and data sources. The fourth section
outlines the analytical concerns of spatial autocorrelation and the modifiable
areal unit problem (MAUP), which leads to a discussion of the relevant GIS
operations in the fifth section. The sixth section presents spatial autocorrelation
measures that show a direct relationship between spatial concentration and
economic success in the Los Angeles garment district. The concluding section
summarizes the importance of GIS as an analytical tool.
Theoretical Issues
Industrial Districts and Ethnic Economies
Both industrial district and ethnic economy perspectives are based on assump-
tions about the importance of space. Business economists and geographers
show that arrangements within spatially concentrated, tightly integrated in-
dustrial districts are critical to globally competitive industries, rapid informa-
tion flow, lowered transaction costs, and increased control over production,
permitting quick and flexible responses to changing market demands (Scott
and Mattingly 1989; Piore and Sable 1884; Storper and Walker 1992; Storper
and Christopherson 1987). In particular, Porter (1990)—a Harvard Business
Study of Ethnic Firms in Garment Industry 45
School economist and member of former President Reagan's Council on Com-
petitiveness—emphasizes that geographic concentration increases local compe-
tition as well as fostering such "emotional factors" as trust, pride, and
bragging rights.
Spatial concentration, while less prominent in discussions of ethnic econo-
mies, remains an underlying factor in providing a venue for the exchange of
cultural capital. To avoid marginalization from the primary economy, immi-
grants establish a secondary economy that serves the existing ethnic commu-
nity and permits the proliferation of social networks and business practices
(Sanders and Nee 1987; Fortes and Bach 1985; Light and Bonacich 1988).
New immigrants, lacking information and local language skills, easily enter
this secondary labor market, where they can take advantage of human capital
provided by a commonality of ethnicity and language.
Hypotheses
While each of these theoretical perspectives is clear on the point that space is
important, they are less precise on what spatial configurations might yield the
best economic advantage. Hypothetically, we expect firm clustering to corre-
spond to economic success. This being the case, we also expect that those
ethnic groups demonstrating higher levels of firm concentration will have high
levels of economic success. But how far (or near) should an entrepreneur travel
to capitalize on spatial patterns? Should firms be located within the same build-
ing, across the street, or within the same county? Since these theoretical dis-
cussions implicitly rely on space, it was natural for us to turn to the analytical
capabilities available through GIS.
Research Site and Data Sources
The Los Angeles Garment Industry
The Los Angeles garment industry provides a unique opportunity to study the
interaction of industrial districts and ethnic economies. Capital and labor are
linked through a set of hierarchical relationships between ethnically diverse
groupings of bankers, landowners, wholesalers, retailers, factors, jobbers, manu-
facturers, contractors, subcontractors, and laborers. Appelbaum and Bonacich
(1993) characterize the industry as a line of progressively smaller piranha, each
biting at the one below. The manufacturer-contractor link is critical to the
garment industry because of its role in the organization of labor and absorp-
tion of labor risk. The link serves as a linchpin between primary and secondary
economies, where world-class designers meet third-world laborers. Manufac-
turers design apparel and arrange the purchase, preprocessing, and delivery of
raw materials to contractors, who then provide the sewing labor to assemble
the final product. "Cash and handshake" arrangements give manufacturers
immunity from the harsh labor conditions often imposed by contractors. These
arrangements provide ideal opportunities for ethnic economies while at the
same time encouraging the type of social and spatial arrangements described
by Porter and others.
The Data
The data set combines four sources to provide a comprehensive coverage of

5158 contractors, retailers, wholesalers, and manufacturers in Los Angeles County:
1. The California Department for Labor Standards Enforcement (CDL)
is responsible for licensing garment contractors and provides a list of
existing firms for each year, including the names, addresses, and tele-
phone numbers for firms contracting labor. Due to the high number of
firms operating illegally and the high turnover rate, Bonacich estimates
that the CDL represents only about two-thirds of the existing contractors
(Bonacich and Hanneman 1991).
2. Dun and Bradstreet (D&B) provides credit reports of companies do-
ing at least $1 million in sales per year, primarily manufacturers; infor-
mation includes number of years doing business, types of clothing pro-
duced, chief executive officer, and parent-child companies.
3. Fabric Marketing Research of New York City produces a business
profile for investors, that includes sales and the types of clothing produced.
4. We were also able to survey 185 of the 210 companies reporting at
least $10 million in sales. Questions included ethnicity, percentage of
imports and exports, type of clothing, number of contractors, and plans
once the North American Free Trade Agreement (NAFTA) took effect.
Initial ethnic coding was based on Bonacich and Hanneman (1991). This
was refined by native speakers, who identified names as being from the coun-
try of their birth, as well as by expert informants within the industry, who
identified the ethnicity of people with whom they were acquainted. Surveyed
respondents' ethnicities were self-identified. Ethnic categories are Armenian,
European, Chinese, Japanese, Jewish, Korean, and Vietnamese. Not included
in this current research are ethnically ambiguous categories: "Other," which
includes Middle East, Greek, Persian, and ethnically unidentifiable names;
"Asian-Other," which includes Thai, Philippine, Indian, as well as unidentifi-
able Asian names; and "Unknown," where no name was listed.
Based on the combined data source, we were able to locate and ethnically
identify 3314 garment firms in Los Angeles County, including: 141 Armenian,
315 Chinese, 463 European, 10Japanese, 3 78 Jewish, 748 Korean, 934 Latino,
and 325 Vietnamese. For 899 of these firms, we also had sales information for
32 Armenian, 77 Chinese, 227 European, 6 Japanese, 233 Jewish, 157 Korean,
143 Latino, and 24 Vietnamese. Because of the diverse nature of reported sales
for manufacturing (selling a product) and contracting (selling labor) the sales
data were further divided into manufacturers and contractors (See Table 3.1).
The portion of the data set used for address matching is essentially a 100%
Table 3.1 Los Angeles garment firms ethnicity by yearly sales in millions of dollars
(Arnold 1993).
MANUFACTURERS
N SUM MEAN MAX MIN STD
Ethnicity Firms Sales Sales Sales Sales Dev
Armenian 12 55.44 4.62 20.00 0.25 5.95
Chinese 26 245.38 9.44 100.00 0.15 20.99
European 128 2013.43 15.73 137.00 0.06 24.88
Japanese 2 26.00 13.00 17.00 9.00 5.66
Jewish 165 3218.00 19.50 250.00 0.10 36.40
Korean 35 366.41 10.47 102.00 0.28 21.03
Latino 28 74.29 2.65 14.00 0.03 3.16
Vietnamese 4 18.72 4.68 12.00 0.22 5.56
CONTRACTORS
Armenian 20 29.51 1.48 7.00 0.06 1.85
Chinese 51 54.32 1.07 4.60 0.11 0.93
European 99 170.79 1.73 25.00 0.00 3.07
Japanese 4 2.92 0.73 2.20 0.10 0.99
Jewish 68 167.12 2.46 34.20 0.04 4.72
Korean 122 191.32 1.57 15.00 0.04 1.99
Latino .115 118.80 1.03 5.00 0.04 1.07
Vietnamese 20 18.92 0.95 3.00 0.13 0.73
COMBINED CONTRACTORS AND MANUFACTURERS

Armenian 32 84.95 2.65 20.00 0.06 4.13
Chinese 77 299.70 . 3.89 100.00 0.11 12.71
European 227 2184.22 9.62 137.00 0.00 20.01
Japanese 6 28.92 4.82 17.00 0.10 6.87
Jewish 233 3385.42 14.53 250.00 0.04 31.67
Korean 157 557.73 3.55 102.00 0.04 10.64
Latino 143 193.09 1.35 14.00 0.03 1.80
Vietnamese 24 37.64 1.57 12.00 0.13 2.55
sample of identifiable firms in Los Angeles County, yet it is subject to error

contained within the extant data sources. As cited above, these sources tend to
exclude small, illegal shops. Even more "high-end" bias exists for those firms
recording sales information, since the D & B and Mass indices are concerned
with success and stability. However, since this particular research concerns levels
of economic success, we do not believe that this bias negates our conclusions.
Analytical Issues
Spatial Autocorrelation
Our previous research, using dot maps, demonstrated heuristic differences in

spatial arrangments between ethnic garment firms (Applebaum and Arnold,
forthcoming). More recently, Appelbaum and Bonacich (1993) and Arnold
(1993) have demonstrated economic differentials between ethnic groups in
the garment industry; that is, Jewish and European manufacturers have the
largest average sales volumes, followed by Japanese, Koreans, Chinese, Viet-
namese, Armenians, and Latinos (See Table 3.1).
Our current research concerns more empirical measures of space, which
we can use in comparing spatial arrangements with income levels of the vari-
ous ethnicities within Los Angeles County. One such measure is spatial
autocorrelation. A thorough explanation of the topic is beyond the scope of
this chapter, and excellent discussions have already been provided by Cliff and
Ord (1973) and Griffith (1987). Essentially, spatial autocorrelation measures
how the values of a particular variable are distributed on a planar map (Griffith
1987: 9-10). Two measures of spatial autocorrelation are the Geary ratio and
the Moran coefficient. The Geary ratio is based on a paired comparison of
neighboring map values and is inversely related to clustering. The bounds of
the ratio normally fall between 0 and 2. The closer the ratio is to 0, the more
similar values tend to cluster. Conversely, Geary ratios nearing 2 are an indi-
cation of the clustering of unlike values. Ratios close to 1 result when there is
no discernible pattern.
The Moran coefficient is based on the covariation of neighboring map val-
ues and is directly related to clustering. The coefficient normally ranges from
-1 to 1. Moran coefficients approaching 1 indicate the clustering of like val-
ues, while coefficients approaching -1 indicate the clustering of dissimilar val-
ues. A coefficient of 0 corresponds to no clustering.1
Spatial Aggregation
One of the most perplexing issues concerning spatial analysis is the modifiable
areal unit problem (MAUP), which has been discussed at some length by
Openshaw(1978,1983,1991; Openshaw and Taylor 1979), who demonstrates
that spatial units can be manipulated to produce nearly any result. Simply put:
"A correlation at the aggregate or ecological level need not imply a similar
correlation at the individual level" (Goodchild 1984: 51). For example, the
1990 U.S. Census reports that 40% of Los Angeles County residents clas-
sified themselves as Hispanic. Yet, by increasing resolution on a specific
location within a given neighborhood (for example, comparing a small
area within Bel Aire Estates, which is largely Anglo, with one in Mac Arthur
Park, which is largely Hispanic) we could conceivably produce area-based
demographics showing anywhere from a near 0% to a near 100% "His-
panic" population in either case.
To farther complicate the MAUP, most common spatial units—such as coun-
ties, census boundaries, zip codes, and voting districts—are generated without
regard to what data those units might contain. The boundaries of Los Angeles
County have not changed since the 1800s. Census spatial units are based on
how much territory bureau workers can cover in a day. Zip codes promote
efficient mail delivery. Voting districts are, at best, a compromise between
political parties.
The aggregation issue posed by the MAUP parallels the problem of know-
ing how much space is important to industrial districts and ethnic economies.
Just as we previously asked whether firms should be located within the same
building or county, how can we objectively determine how much of an area to
include in spatial autocorrelation measures? It was our intention to use the
spatial manipulation capabilities of GIS to produce spatial autocorrelation
measures at different levels of resolution. Using more than one spatial unit
would allow us to look for central tendencies in spatial patterns, which
Goodchild (1984) and Openshaw (1977, 1978, 1983,) suggest as being a best-
case solution to the MAUP. Thus, we used address matching to provide point-
referenced data, rather than using preexisting areal units such as zip codes.
Once the point coverage was built, using the data as guide, we generated a
series of grids representing multiple levels of aggregation.
GIS Methodology
Address Matching
The first step was to locate the firms on a map of Los Angeles. After using
ARC/INFO to build a street coverage of Los Angeles County from the 1990
U.S. Census TIGER files (Topographically Integrated Geographic Encoding
and Referencing system), we used the ADDRESSMATCH operation to lo-
cate and build a point based coverage of all Los Angeles County garment firms.2
The Census Bureau codes streets into segments with a node at either end;
each node contains the address range of that segment. ADDRESSMATCH
finds the correct segment and estimates the firm's position on the segment
based on the number range. For example, to locate a firm at 175 Spring Street
on a map of Los Angeles County, the address matching routine would first
locate the Spring Street segment that had nodes numbered below and above
175. If the segment had a low-numbered node of 100 and a high-numbered
node of 2 00, the address 175 would be located three-quarters of the way to the
200 node. Using the "reject processing option," we were able to match
95% of the businesses; the remaining 5% consisted of firms with incom-
plete or incorrect addresses.
Polygon Generation
The next step in determining Geary and Moran statistics was to arrange the
data into matrix form. One method is to place a grid over the coverage and
sum the point values within each cell. Using the point coverage as a guide, we
created three 10 by 10 grids (rectangular Arc/Info polygons containing 100
cells) using the GENERATE command with the FISHNET option. We se-
lected this resolution because of the computational advantages afforded by
Griffith's (1992) spatial autocorrelation algorithm for a square matrix. The
first grid, DISTRICT (cell perimeter=542), was fitted over the official gar-
ment district. The second grid, GREATER (cell perimeter=1140), was se-
lected via the "looks like a lot of points" method and more closely represents
what our ethnographic research indicates as being an area of intensive gar-
ment manufacturing activity (the "expanded garment district"). The third grid,
COUNTY (cell perimeter=32,460), was fitted over the entire point coverage.
Identity Overlay
To calculate the number of points within each grid cell, we began with the
IDENTITY overlay procedure for each polygon. This function overlays the
point coverage (firms) with the polygon coverage (grid), creating a new point
coverage. The new point coverage includes a cell ID number for each point,
indicating which cell contains the point. The cell IDs can then be summed
within Info to determine the number of points within each cell. However, we
needed to extract the data from Arc/Info for further processing in SAS. Using
Info's REPORT function, we created an ascii file which included the Point
ID, Cell ID, and ethnicity. The file was then loaded into SAS and put into the
square matrix format described by Griffith (1992).
Analysis
Given the theoretical insight of industrial district and ethnic economy per-
spectives, we expected that spatial clustering would be directly related to eco-
nomic success. For the purpose of this chapter, we have included density maps
and Geary's ratios only for Jewish, Korean, Latino, and Vietnamese firms (see
Arnold 1993 for a more comprehensive analysis). Jewish and European firms
are considered part of the primary economy because they tend to control more
of the high-end processes and report much higher sales than other ethnic cat-
egories. Korean and Latino firms represent the largest number of businesses
outside the primary economy. Table 3.1 shows the economic hierarchy of the
four ethnic groups measured in mean annual sales. Jewish firms have the high-
est, with 14.53 million (19.5 million for manufacturers and 2.46 million for
contractors); European are second, with 9.62 million (15.73 million for manu-
facturers and 1.73 million for contractors); Korean firms are third, with 3.55
million (10.47 for manufacturers and 1.57 for contractors); followed by Latino
firms, with 1.35 million (2.65 million for manufacturers and 1.03 million
for contractors).
The following density maps (Figures 3.1 to 3.3) demonstrate that Openshaw
was correct in asserting the problems of the modifiable areal unit. "N" is the
number of ethnic firms at the COUNTY, EXPANDED DISTRICT, or DIS-
Jewish Garment Firms European Garment Firms
N-378 Max Z-233 GR-1.02 p.. 13 N-463 Max Z-181 GR-0.87 p-.OO
Korean Garment Firms Latino Garment Firms
N-738 Max Z=592 GR=1.04 p- 19 H-934 Max Z-367 GR-0.72 p-.OO
Figure 3.1 "County" density maps.
TRICT resolution, "Max Z" is the number of firms lying within the most
populated cell, "GR" is the Geary ratio, and "p" is the probability associated
with GR. For each level of analysis, different results were obtained. At the
COUNTY level, Latino firms have the lowest GR, with 0.72, followed by
European firms, at 0.87. The ratios for Jewish firms, 1.02, and Korean firms,
1.04, have high ^values (0.13 and 0.19, respectively), indicating that neither of
the ratios is significant to a 0.1 tolerance (Fig. 3.1).
At the EXPANDED DISTRICT level, Jewish firms have the lowest Geary
N=119 Max Z=23 GR.0.47 p..00 N=92 Max Z»17 GR=0.50 p«.00
N-449 Max Z-69 GR-0.62 p.. 00 N-144 Max Z.31 QR-0.73 p = . 0 0
Figure 3.2 "Expanded district" density maps.
ratio, 0.47, indicating the highest clustering, followed by European firms, with
a ratio of 0.50. Korean firms scored 0.62, while Latino firms had a ratio of
0.73. All of the/) values at the EXPANDED DISTRICT level are signifi-
cant to 0.00 (Fig. 3.2).
At the DISTRICT level, only Latino firms, with a ratio of 0.82, are statis-
tically significant to a 0.02 tolerance. The remaining Geary ratios are 0.84 for
Korean firms, 0.89 for European firms, and 0.97 for Jewish firms (Fig. 3.3).
The theoretical problem of "how much space is significant" can now be
N-97 Max Z=18 GR-0.97 p-.OS N=75 Max Z-12 GR-0.89 p..01
N-387 Max Z-36 GR.0.84 p..01 H=105 Max Z=23 GR.0.82 p..00
Figure 3.3 "District" density maps.
seen methodologically in measures of spatial autocorrelation. Clearly, our ex-

pectations were met at the EXPANDED DISTRICT level of resolution with
a direct correlation between aggregation and income. However, at the other
two levels (COUNTY and DISTRICT), probability values indicate that the
measurements are unreliable. This is primarily an artifact of the spatial
autocorrelation measures used. Since these measures are based on the aggre-
gation of similar values they are affected by spatial divisions resulting in vary-
ing contiguous values. Although not as yet attempted, it is probable that GIS
smoothing functions will yield more statistically significant results at all levels
of resolution. It is telling that the one level which was inductively selected
(EXPANDED DISTRICT) not only proved to be the most statistically sig-
nificant but also yielded the expected results.
Conclusion
The functionality of GIS is often classified into three levels. The first level
involves using GIS to do simple things we have always done. The second
level involves doing complex things we seldom or never do. The third level
involves using GIS to do brand new things that revolutionizes our thinking,
and create new hypotheses (Aldenderfer 1992). Our own use of GIS fits into
the second level's "never" category. If we had enough time, money, and re-
search assistants, it would have been possible to match over three thousand
addresses and then created grids by hand; but without GIS, we would not
have attempted the research.
In our exploration of the Los Angeles garment industry, we have been able
to use GIS to generate the matrices suggested by Griffith (1987, 1992). Per-
haps more importantly, we ameliorated the problems of aggregation suggested
by Openshaw (1983). Our methodology also indicates anthropological uses of
GIS, which approach the third level of GIS functionality. By mapping the
actual location of human activities rather than relying on arbitrary spatial units
such as neighborhoods, anthropologists working in urban settings can better
understand the relationship of space and their research subjects. This new
understanding has the potential to redefine, if not "revolutionize" our thinking.
It is important to emphasize that GIS posed its own limitations and was an
intermediate step for our research goals. Since the focus of GIS development
is on commercial uses, it falls short in its research applications. This makes the
role of the National Center for Geographic Systems Information and Analysis
(NCGIA) even more critical in bringing together commercial technology and
academic applications.
Notes
1. Currently, we are using spatial autocorrelation to compare and contrast spatial pat-
terns. However, there are further implications for statistical models. Positive and sta-
tistically significant spatial autocorrelation would invalidate standard regression mod-
els because an assumption of independence between cases could not be made.
2. The TIGER files for Los Angeles County are over 100 megabytes and Arc/Info
BUILD and CLEAN operations require even larger amounts of disk space. Before we
obtained our current multi-gigabyte Unix platform, we were able to use these intensive
processes by subsetting Census street ranges prior to building coverages.
4
A Formal Justification for the Application
of GIS to the Cultural Ecological Analysis
of Land-Use Intensification and
Deforestation in the Amazon
Clifford A. Behrens
What is the process by which indigenous Amazonian people intensify their

utilization of tropical forest resources, and what are the roles of population
demography, settlement patterns, and resource degradation in this process?
These are the central problems of this chapter. Over the last fifty years, eco-
logically oriented anthropologists have focused on these questions because of
their significance for explaining the socioecological variability found among
Amazonian Indians.
A common theme in many attempts to account for socioecological variabil-
ity in the Amazon is that large, sedentary populations necessitate increasing
levels of social integration. Therefore, some explanations for this variability
have sought factors that limit population density, such as the local availability
of arable soils and protein-rich faunal foods. Simple single-factor frameworks
have been criticized, yielding slightly more complex kinds of explanation, some
based on evolutionary ecology and decision theory. Nevertheless, none of these
approaches has successfully managed to relate population growth, village for-
mation, resource degradation, and intensification of land use together in a
single formalism that derives its first principles from a comparative analysis of
the ethnographic literature. As a result, culture has not been assigned the cen-
tral role it deserves in any theory purporting to characterize the process of
land use intensification among indigenous Amazonians.
This paper will review the ethnographic literature on the Amazon to (1)
establish an empirical basis for the ingredients required to formulate cultural
ecological theories of land-use intensification among indigenous Amazonians
and (2) propose a developmental sequence based on increasing sedentism, in-
tensification of land utilization, and growing market demand for production.
Thus, this paper attempts to integrate seemingly disparate ideas from the past
and present, each with some "ring of truth," in the kind of mathematical frame-
work advocated but never really achieved by Steward.1
The resulting paradigm converges on one very much resembling "land-
scape ecology," but with greater emphasis on the role of culture and human
decision making in a generative process. The need for detailed land-use data
on a regional scale implicates the application of new technologies, such as remote
sensing and geographical information systems, to test the proposed theories.
Intensification of Land Use in the Amazon

The ethnographic literature suggests that there are many reasons why indig-
enous Amazonian populations abandon their traditional dispersed settlements
for a more sedentary existence and, subsequently, intensify their use of land.
Forced settlement (Bodley 1982; Davis 1977) or territorial circumscription
resulting from warfare (Carneiro 1961; Chagnon 1973) and outside encroach-
ment, often fueled by speculation (Collins 1986; Hecht 1984; Hegen 1966;
Moran 1981; Schmink and Wood 1984; Schumann and Partridge 1989; Shoe-
maker 1981), may be factors in some cases. At times extensive deforestation
accompanies village formation or colonization as part of a strategy to lay claims
on land in anticipation of future title disputes (Rudel 1983). With pacification
and some preservation of tribal lands, it seems that nowadays permanent vil-
lages are formed more often, for what are perceived by natives as opportuni-
ties to obtain formal education, improved health care, and regular access to
commercial goods (Behrens 1989, 1992; Descola 1981; Yost 1981). The fol-
lowing three stages characterize the transformation of subsistence-oriented
extended family homesteads, typical of the Amazon in the past, to more con-
temporary villages who participate in a regional market economy (cf. Hames
andVickers 1983).
Stage I
Traditionally, most indigenous peoples in the Amazon lived in extended fam-

ily homesteads scattered along major rivers and their tributaries. Economies
were subsistence-based with shifting cultivation supplying calories principally
in the form ofManihot esculenta (manioc) and varieties ofmusa (plantains and
bananas), while fishing and hunting provided animal proteins to the diet. Typi-
cally, quality land and wild food resources were located close to home in the
surrounding forest, and when gardens or wild resource procurement sites were
no longer productive, homesteads were relocated elsewhere. A spatial depic-
tion of this stage is given in Figure 4.1.
This landscape consists of numerous extended family homesteads that are
dispersed in space (represented by filled circles) alongside a major water source.
Here, homesteads are shown as small rectangles, each enclosed by a filled circle.
A family exploits resources in plots of land nearby, represented by several circles
surrounding each homestead. When the effort required to produce or procure
resources on this land exceeds a tolerable level, the family (or part of it) moves
to one of a number of other potential sites, represented by small unfilled
circles in the figure.
As mestizos and whites entered a region, indigenous people often adopted
Using GISfor Cultural/Ecological Land-Use Analysis 57
Figure 4.1 Land used for subsistence production by dispersed extended family
homesteads.
a more sedentary lifestyle. This occurred for several reasons: natives wanted
access to more efficient tools and ready-made goods (or the cash to purchase
them), and they desired schools and health clinics that would improve their
quality of life and help them compete better in a larger literate society. Forces
for change often came from within as well as from without. Influential head-
men and local schoolteachers, in collaboration with missionaries or powerful
mestizo patrons (who often had their own motives), brought kinsmen together
and convinced them of the "advantages" to be realized with village life (e.g.,
Behrens 1992; Yost 1981).
Stage II
Once a village is founded, the household economy may continue to thrive

primarily at a subsistence level, but increased interaction with outsiders fur-
ther accelerates change by increasing the demand for Western goods and by
expanding natives' awareness of employment opportunities outside of their
village. With sedentism, nearby land and wild resources such as fish and game
are gradually depleted, even though more productive regions of tropical forest
exist beyond the distance from their houses that fishers and hunters may regu-
larly be willing to travel (Baksh 1985; Hames 1987). Spatially, this stage is
represented by Figure 4.2.
In this landscape, all homesteads have aggregated to form a village, now
shown as a single filled circle containing multiple rectangles. Large planar
areas near the confluence of rivers or their tributaries are typically selected for
village settlements. Families temporarily exploit plots of land within an exten-
sive territory surrounding the village, but they also reserve other areas for
future needs. For example, a family may cultivate a swidden for several years,
shown as one pattern of small circle, while it maintains some fallow land for
later use, represented by another pattern of small circle. Though population
density within the village's territory may increase, the total population density
for the entire region does not necessarily change from the previous stage.
Moreover, with sedentism, families gradually seek new areas (e.g., arable land
or wild food procurement sites) more distant from the village, thus increasing
their travel time to workplaces (Behrens 1989). Some may actually build houses
on these sites and live there for substantial periods during the year.
The desire for school materials, medicines, more efficient tools, ready-made
clothing, and other goods creates a growing demand for cash. At first, villag-
ers often sell forest products or small domesticated animals to outsiders, but
Figure 4.2 Land used for subsistence production by a rural village.

Using GISfor Cultural/Ecological Land- Use Analysis 59
later many decide to exchange their labor for cash. However, wage labor is not
always an attractive option for gaining cash, as it typically forces men (and
sometimes women) to live away from their villages for long periods.
Stage III
Natives often learn to produce cash crops or care for cattle while working for
outside employers, and they eventually apply this new knowledge to generate
cash on their own. However, cash cropping and cattle ranching are labor-
intensive, so less time is allocated to subsistence activities that take people
long distances from their village—as for wild food procurement (Baksh 1985;
Behrens 1986a, 1992; Macdonald 1981). Consequently, domesticated animals
gradually replace fish and game as sources of dietary protein.
Initially, farmers may seek more distant land to raise their level of agricul-
tural production and meet external demands of the regional market. But higher
travel costs and the need to recruit scarce labor in the village during times of
planting and harvesting lead to a greater intensification of land use. The latter
occurs as farmers shorten fallows or bring what was once considered "mar-
ginal" land near the village into production (Behrens 1989). In a similar man-
ner, cattle ranching entails more labor input and greater social organization to
create and maintain pasture, manage the herd, and distribute or sell its prod-
ucts. And, as with agricultural intensification, pasture land is often recycled
after short fallow periods. Fences are built or gardens moved to more distant
locations to protect them from menacing livestock (Behrens 1986a, 1989).
The amount of land a farmer allocates to cash cropping or pasture is deter-
mined not only by his family's subsistence needs but also by the farmer's evalu-
ation of future market demand for his product. Propitious increases in market
prices and greater interaction with clients or intermediaries who purchase prod-
ucts in the village and transport them to the marketplace can influence a farmer's
perception of market demand and hence his use of land.
At some point, cash croppers and cattle ranchers may spend their earnings
to purchase pack animals or more modern transportation, such as motorized
canoes and trucks. These are used to facilitate travel between their houses and
gardens or pastures, deliver products to markets, and make occasional trips to
distant wild food procurement sites. Improvements in transportation also open
the possibility for some to either extend their use of land by expanding distant
clearings or by establishing satellite settlements in primary forest, which they
may occupy much, or all of the year (this is considered below in greater detail).
A spatial model for this stage is shown in Figure 4.3.
The landscape depicted above is similar in appearance to the one shown in
Figure 4.2. However, cash-producing uses of land, shown by a third pat-
tern of small circle, are intermingled with fallow land and plots exploited
for subsistence.
Admittedly, these are three somewhat idealized stages, but they generally
represent the process of village formation and changes in land use frequently
reported for indigenous Amazonians. The isolation of factors that either fa-
Figure 4.3 Land used for subsistence and market production by a rural village.
cilitate or interrupt this sequence have occupied the attention of cultural ecolo-
gists for the last fifty years.
Limiting Factors to Social Complexity

Much previous ecological research has sought to explain the apparent lack of
social complexity in the Amazon by "limiting" environmental factors thought
responsible for relatively low population densities. Others have focused on
cultural factors by attempting to determine whether Amazonian peoples con-
serve their resources or intentionally manage sustainable yields of them. This
research points to the need for theories that integrate better environmental
and cultural approaches in a paradigm that has potential for studying the process
of agricultural intensification without population pressure, typical of the Amazon.
Environmental Factors
Early ecological research on Amazonian people grew out of comparative analysis

made possible by the wealth of ethnographic information published in the
Handbook of South American Indians (Steward 1940). Steward concluded that
social evolution in the Amazon was constrained by resource zonation. Meggers
(1954) followed this with the suggestion that levels of social complexity are
limited by the agricultural potential of soils in different regions of the Ama-
zon. After examining the productive potential of Kuikuru soils, Carneiro (1961)
concluded that at least some Amazonian populations had the potential to ex-
ceed their current sizes without degrading their environments. But since his
study, more detailed research on indigenous plant-soil relationships in the
Amazon indicates that significant local and regional variability in soil quality
exists and soils are, indeed, degraded with continued use (Behrens 1989; Gross
1983; Moran 1991).
Others have argued that faunal foods as either sources of dietary protein
(Denevan 1966; Gross 1975; Harris 1984; Lathrap 1968; Meggers 1971; Ross
1978) or fat (Baksh 1985; Johnson and Baksh 1987), not necessarily the pro-
duction of highly caloric cultigens, is limiting on aboriginal settlement size
throughout Amazonia. These claims, too, have been challenged by their op-
ponents (Beckerman 1979; Chagnon and Hames 1979; Lizot 1977), who have
shown that, in many Amazonian groups, the amount of protein per capita
consumed often equals or exceeds amounts consumed in modern Western so-
cieties. It does seem, however, that faunal resources are rapidly depleted in the
vicinity of large sedentary settlements. Moreover, the disappearance of wild
foods is readily recognized by natives and a matter that concerns them deeply
(Baksh 1985; Behrens 1986b; Hames 1987).
Not unlike the demographic focus of previous cultural ecologists, more
recent research has embraced Boserup's idea that population growth fuels ag-
ricultural intensification and that this, in turn, entails greater social and politi-
cal integration (cf. Boserup 1965; Johnson and Earle 1987; Lele and Stone
1989; Vasey 1979). Nevertheless, while relatively low population densities have
been reported throughout much of Amazonia, many indigenous communities
are rapidly intensifying their production for regional cash markets (Bodley
1982,1985). These findings indicate that factors other than merely demography
and population pressure are responsible for land-use intensification and social
variability among contemporary indigenous Amazonians. This seems particu-
larly true for groups that are seeking to increase their participation in
regional cash markets.
Cultural Factors
Others have investigated the role of culture in resource utilization and land-
use practices by examining indigenous management principles. This approach
takes the position that the environmental knowledge and decision-making
principles upon which natives base their behavior (e.g., what they conceptual-
ize as the costs and benefits of resource exploitation along with the calculus
they employ for evaluating these), also have a significant cultural basis.
For example, Siskind (1973) interpreted long-time settlement of the tropi-
cal forest without exhausting game resources as a "conservation-oriented" ad-
aptation of tropical forest cultures. Stocks (1983) suggests that the Cocamilla
manage their lake fishery both at the conscious level, supported by an eco-
nomic rationale, and at the unconscious level, supported by folklore and magi-
cal beliefs. Hames (1987), attempting to discern a conservation ethic among
the Yanomamo, found that he could not distinguish between the predictions
of conservationism and short-term maximization. Irvine (1987) claims that

the San Jose Runa reveal a concern for resource sustainability in their use of
fallows for hunting, their attempts to protect garden land from encroachment
by obtaining titles, and their management of fishing territories by traditional
corporate group control. As one might expect, others question the existence
of any "conservation ethic" among Amazonian Indians who, to them, seem
ambivalent (or sometimes even destructive) in their exploitation of the tropi-
cal forest environment (e.g., Henley 1982).
Based on his work with the Machiguenga, Johnson (1989) has attempted to
expose the demographic basis for resource conservation among indigenous
cultures. He concluded that, when population densities are low, natives do not
need to manage their habitat. He argues that Amazonian Indians will only
begin to invest their labor in resource management when population pressure
on resources is great, access to critical resources by individuals and their de-
scendants must be secured, or when through warfare (or other encroachment)
people's territory becomes "circumscribed" (cf. Carneiro 1961; Chagnon 1973).
Need for an Integrative Approach
Many of the points made by previous researchers are well taken and supported
empirically. But each by itself provides a rather incomplete characterization of
land-use intensification in the Amazon. The review above suggests that envi-
ronmental and cultural factors are related in a complex manner. Thus, there
exists the need to unify these points of view in a single framework so as to
better understand the cultural ecological process by which intensification of
land use occurs in the Amazon (see Blaikie and Brookfield 1987).
The debate over limiting factors in Amazonia has encouraged researchers
to closely examine the relationship between time allocation, agricultural pro-
duction, and level of participation in cash markets. Quantitative studies of
indigenous activity patterns and diet have caused some to conclude that people's
work strategies are influenced more by the overall range of activities in which
they spend time and the various needs these satisfy, not just as a response to a
single, absolute need such as protein (Behrens 1986a; Johnson and Behrens
1982; Saffirio and Scaglion 1982; Werner et al. 1979). Thus, this research has
pointed to the need for multifactorial explanations, greater quantification, and
comparative analyses of agricultural intensification on a regional scale (cf. Gross
1990;Moran 1990; Sponsel 1986).
Moreover, research on indigenous systems of conservation has implicated
native cultural knowledge and demonstrated the need for an approach that
stresses the interactive process by which humans adapt to their environment
(Behrens 1986a; Ellen 1978; Orlove 1980; Vayda and McCay 1974). Though
opinions may differ as to when a "conservation ethic" becomes part of indig-
enous culture, it seems most would agree that management principles become
increasingly important as local land resources become stressed. The position
taken in the next section of this paper attaches particular importance to the
reasons why natives organize into larger social units, their evaluation of new
economic opportunities, and the manner in which they make strategic deci-
sions affecting their use of land.
Unifying Theories
A formal theory relating village settlement, population density, and agricul-

tural intensification has been proposed (Stryker 1976). With some modifica-
tions and ethnographic elaboration, Stryker's formulation seems a useful con-
ceptual framework for representing the intensification of land use that accom-
panies increasing sedentism and production for external markets, typical among
shifting agriculturalists in Amazonia. Separate theories are developed for each
of the three developmental stages presented earlier.
Main Ingredients of the Theories
To unify the many ideas offered to account for land-use intensification in the
Amazon, a theory should integrate concepts of resource degradation, popula-
tion growth and increasing sedentism, and eventually production for market.
Moreover, these key concepts should be cast in a framework that is sufficiently
general in outline but, ideally, one that has already found use in cultural ecol-
ogy. The formalism developed here accomplishes this by wedding the theory
of time allocation to central place theory (Gross 1984; Johnson and Behrens
1989; Kellerman 1983; King 1984; Plattner 1989; von Thunen 1826). The
path diagram in Figure 4.4 characterizes the process of land-use intensifica-
Figure 4.4 Path diagram showing ingredients of theories.

tion in the Amazon from the perspective of an individual producer and cap-
tures his or her decisions to allocate more time to either labor or travel; where
c = total cost incurred in production
/ = labor cost incurred in production
t = travel cost incurred in production
a = land put into subsistence production per capita
q = current level of land degradation
n = population size of settlement
s = number of potential settlement sites
k = land put into market production per buyer (assumed constant)
b = number of buyers per capita
m = land put into market production per capita
xo = amount of production desired for consumption (assumed constant)
x - amount of production
yo = amount of cash desired (assumed constant)
y = amount of cash gained from sales of x
p = market price paid for a unit off x (assumed constant)
The signs of arrows in the diagram indicate the effect one variable has on
another. Thus a plus indicates that an increase in a variable is associated with
an increase in another, while a minus shows that an increase in a variable is
associated with a decrease in another. For example, an increase in travel time
is shown to accompany an increase in settlement population size, but a de-
crease in settlement population size is associated with an increase in the num-
ber of sites where a population may settle after fissioning. Bounded by certain
conditions, described next, conclusions are derived from each theory for settle-
ment members who attempt to meet their desire for food and cash in a way
that minimizes their total production cost.
Central to the proposed theories is the assumption that, due to travel time
between houses and resource production sites, land is not a costless produc-
tion input even when more is available (cf. Chisholm 1962; De Lisle 1978;
Found 1970; Stryker 1976; but also see Stevens and Lee 1979). It is further
assumed that only land surrounding the settlement is exploited and that the
production potential of this land is homogeneous in quality. On this idealized
landscape, there exist no natural or physical barriers to movement or location,
and travel is equally easy in all directions. Thus, the extent of land use is theo-
rized as being equidistant from the center of a settlement, so that the total area
of land exploited is enclosed within a circle.2 If A represents the area of this
circle and R its radius, then A/p = R2. Furthermore, if A= an, then (an)/p = R2 or,
put another way, R = g(an)I/2. SSince travel time is assumed to be proportional
to R, then t = g'(an)l/2 ((cf. Stryker 1976). When subsistence production is aug-
mented by production for market, then A = n(a+kb). If we let m = kb, then^ =
n(a+m). Therefore, when land is used for both subsistence and market pro-
duction, t = g"[n(a+m)]I/2.
65
Added to the spatial constraints above is the assertion that all settlement
members employ the same production technology and both production costs
and output are shared equally among them. Consistent with these assump-
tions and the path diagram in Figure 4.4, signs of the first- and second-order
derivatives are given as t, t ,t,l', m',t,t ,x.x,x,x.x.>0 and n'.t.t
X,. X . X < 0.
IP ai? mm
Theory I:
Subsistence Production in Dispersed, Extended Family Homesteads
Let the total cost (in time) that an individual incurs in subsistence production
be a function of labor cost plus traveling time. Labor cost is expressed as a
function of land degradation. Traveling cost depends on the amount of area
per capita utilized by homestead members for subsistence production, and the
population size of the homestead. Homestead size, in turn, is given as a func-
tion of the number of other sites available for settlement (i.e., homestead size
can be reduced by some of its members leaving to establish another home-
stead elsewhere). Hence, the magnitude of s in the theory captures the force of
attraction that a settlement has for its inhabitants—in other words, a
population's degree of sedentism. This verbal formulation for an individual's
cost function can be expressed formally with the following:
c= I +1; where /= l(q) and t= t[an(s)]
In a similar manner, a production equation for each homestead member

can be written:
x = x(l,a)
The previous equation states that one's total resource output is a function
of the amount of labor applied and one's share of land in production.
In addition to previous assumptions, analysis begins by also assuming that
homestead population size is fixed. (Below, we will evaluate the effects of re-
laxing this assumption.) The cost and production functions can be combined
in a Lagrangean equation and solved for its optimum (Chiang 1974).
Minimize z
At equilibrium the following marginal condition is satisfied:
The solution above expresses the conclusion that the marginal product of
land is positive and equals the marginal product of labor multiplied by the
marginal rate at which travel time may be substituted for land. This result
expresses formally the assumption made earlier that the utilization of new land
is not without its costs. Moreover, current land quality (q) need not be limit-
ing, though relative land degradation affects choices to relocate as one must
expend more labor to produce the same amount of output.
This same theory can be used to show the effects of shirting settlement
location on land degradation, labor cost and per capita land use. These con-
clusions are derived by totally differentiating the first-order equilibrium con-
ditions, and then applying Cramer's rule (Chiang 1974; Kuznar 1991). While
noting that the determinant of the Jacobian is reasonably assumed to be nega-
tive,3 the operations above yield these results:
and by the Chain Rule,
Moreover, from the previous formulation we can assess the effect of increas-
ing the number of sites on total land exploited (A) by a settlement:
This effect depends on the relative values of the first term, which is negative,
and the second term, which is positive. However, a reasonable interpretation
of this result is that fissioning might occur when average land holdings be-
come too small. In this scenario, average land use would stay the same, making
the second term of the expression zero and its overall sign would become nega-
tive. Indeed, this is the result one would expect among extensive swidden
agriculturalists.
The analysis demonstrates that, when other settlement sites are maintained,
land degradation and labor allocations can be reduced, while per capita amount
of land either remains the same or increases. Thus, the effects of growing popu-
lation can be offset by moving the settlement (or part of it) to another location.
The land-use pattern generated by this theory is exactly the one shown in
Figure 4.1. Here the vegetation is spatially heterogeneous and consists prima-
rily of forest punctuated by numerous small settlements, each enclosed by
areas of land in different stages of succession. Some of these areas are newly
cleared, while others exhibit forest regrowth. The salient features of this land-
scape are its many small settlements and a low ratio of utilized to fallow land in
the area surrounding each. Over the long term, the amount of tropical defor-
estation associated with this land-use pattern is minimal, as forest regrowth
invades clearings once these are permanently abandoned or a settlement relo-
cates to another site.
Theory II:
Subsistence Production in a Rural Village
In this theory, an individual's cost and production functions are the same as
those above but with a single important difference. Once homesteads aggre-
gate in a village, it becomes less desirable for families to relocate their houses
in distant areas beyond the land exploited by others that surrounds the village.
Sedentism, a result of shared social, economic and political interests (i.e., a
cultural factor), is expressed formally by reducing the number of potential
settlement sites to one (presumably a much larger area). Hence, this theory
gives individual traveling cost only as a function of the per capita amount of
area used for subsistence production by village members and village popula-
tion size, as follows:
Using the procedure of totally differentiating the first-order conditions and

then applying Cramer's rule, the results derived from this theory are:
and by the Chain Rule,
At equilibrium, the negative effect of population increase on average land use

stands in contrast to the previous theory where it had little effect. Here, popu-
lation is sedentary and there exist no other sites of equivalent "importance" to
settlers. Hence, an increase in village population will always decrease average
land holdings, revealed formally in the derivation above.
The change in the total amount of land utilized with respect to increases in
village population size is given as:
The value of this derivative depends on the magnitude of the first term, which
is positive, and the second term, which is negative. Thus, the change in total
village land use accompanying an increase in village population depends, to a
large degree, on current average land use and population size.
In summary, this second theory's conclusions are that, with sedentism, a
growth in population leads to an increase in land degradation and labor allo-
cations. These increases are associated with a decrease in the per capita amount
of land used, even though there is more available land surrounding the village.
But in this case, individuals (and their families) will not shift settlement to
another area when land becomes less productive, because to do so would de-
prive them of desired goods and services. Therefore, levels of production can
be sustained only through increasingly intensive uses of land, not by moving
elsewhere, as was allowed in the previous theory.
Spatially, the pattern of land use generated from this theory is represented
by Figure 4.2. The landscape is characterized by one (or merely a few) large
village settlements. The total amount of land exploited by a village far exceeds
the amount utilized by a dispersed homestead. Each individual's share of land
is smaller and used more intensively. This is shown in Figure 4.2 by the many
small plots surrounding the village and the higher ratio of utilized to fallow
land. On this landscape, deforestation is more severe and continuous in the
vicinity of each village.
Theory III:
Market Production in a Rural Village
A mathematical argument can also be made to show that the desire for cash,
together with an increase in market demand for production, can lead to rapid
intensification much the same as an increase in village population size. In this
theory, an individual's cost function expresses two reasons for exploiting more
land: the production of food and cash. Together with the consumption re-
quirements of the village population, the amount of land exploited is also de-
termined by demand from outsiders (e.g., buyers, clients, or agricultural banks).
Therefore, travel cost is given as a function of village population size along
with the per capita amount of land used for subsistence and market pur-
poses, as follows:
t = t[n(a+m)],
where m = m(b)
Furthermore, in this theory the production function specifies a target in-

come and a market price for placing a cash value on output:
This constraint indicates that the only production sold to market is any
remaining after subsistence needs are met. Notice that when total production
is less than or equals subsistence needs, then this production function simpli-
fies to the same one specified in the previous theory. Hence, this theory does
not allow for full commercialization (or total market dependency) as individu-
als still produce, rather than purchase, the food that they consume.
Derivation of conclusions begins by fixing village population size and fur-
ther assuming that subsistence needs are met with current amounts of subsis-
tence land. Consequently, per capita land devoted to subsistence production is
also held constant, so that only market production land is used to generate
cash. In this manner, attention is directed to the overall effect of increasing
market demand and per capita land used for cash generation. As before, we
minimize the cost function subject to the production function employing the
technique of Lagrangean multipliers:
Minimize z
At equilibrium the following marginal condition is satisfied, similar to the

previous two theories:
An analysis of the effect that either changing village population size or in-
crease market demand for production has on labor input and the amount of
market land exploited per capita also yields results similar to those obtained
from the previous theory. Growth in village population leads to greater land
degradation. But an increase in market demand tends to stimulate village popu-
lation growth. This growth could result from improvements in health care
made possible through the purchase of medicine and payments to physicians
or the arrival of new settlers attracted by opportunities to acquire cash and
purchase Western goods (cf. Pingali and Binswanger 1987). These conclu-
sions are derived by applying the same mathematical techniques as before:
and since
then dn/db > 0
We can also determine the manner in which labor and the amount of sub-
sistence land exploited by each person change with an increase in market de-
mand for output. In a situation where land is taken from subsistence to pro-
duce for market, am< 0. Then, by the Chain Rule:
da/db =ff,,,™'< 0
and
Earlier it was assumed that villagers' subsistence needs were met and that
therefore the amount of land put into subsistence production was fixed. How-
ever, the conclusion just derived shows that an increase in outside demand for
production can encourage the conversion of land from subsistence to market
uses. Moreover, growing market demand for production yields an effect simi-
lar to that of an increase in village population size—namely, increased labor costs.
The effect on total land utilized by increasing either village population size
or the number of those who purchase village production is expressed as:
As in the previous theory, the effect of village population increase on total

land use is ambiguous and depends to a considerable degree on the relative
values of the first two terms, which are positive, and the last two terms, which
are negative. However, an increase in the number of market buyers may have
the effect of increasing total land usage, assuming that more of these purchas-
ers will stimulate an increase in the average amount of land that farmers put
into production for market.
The signs of these derivatives indicate that, even in the absence of village
population growth, production for external markets will also lead to in-
creasing labor inputs while at times decreasing the per capita amount of
subsistence land cultivated. Thus, an increase in market demand for out-
put only exacerbates the effect of sedentism and village population
growth—i.e., intensification of land use.
The ecological effects of this process on the theorized landscape are illus-
trated in Figure 4.3. This spatial representation is similar to the one generated
by the previous theory with two noteworthy exceptions. First, plots surround-
ing the village show two uses of land, production for subsistence and for
market. Second, with market participation, the ratio of utilized to fallow
land decreases even further, indicated by the smaller number of dot-filled
circles surrounding the village.
These three theories seem to possess many desirable properties. They are
built from concepts that, over the years, anthropologists have identified as
being important for understanding indigenous land use patterns in the Ama-
zon. Furthermore, these theories "behave" in a manner that is consistent with
what we already know ethnographically about the complex relationship be-
tween population growth, village formation, and intensification of land use.
Nonetheless, these conclusions beg testing with new data.
Testing the Proposed Theories

The landscapes generated by the proposed theories are essentially hypotheses
that can be tested with data on indigenous land use. However, rigorous tests of
these theories require detailed ground-cover data on a regional scale. The
immense volume of data that must be processed implicates the application of
new methodological tools, principally satellite remote sensing and the analysis
of satellite imagery with a GIS.
Anthropologists are making increasing use of remote sensing data in their
research (Behrens and Sever 1991; Conant 1990; Green and Sussman 1990;
Reining 1979; Wilkie and Finn 1988). Remote sensing is the science and art of
acquiring information about an object or area from a distance through instru-
ments that are sensitive to various bands of the electromagnetic spectrum
(Lillesand and Kiefer 1987). Space-borne platforms, such as Landsat and SPOT
earth observation satellites, acquire images by measuring amounts of radia-
tion reflected by ground objects, each with different spectral properties or
"signatures." Once these signatures are discovered statistically, using digital
processing techniques, images can be classified to reveal the distribution of
information classes within a region (Richards 1986; Schowengerdt 1983).
As a preliminary study, a Thematic Mapper (TM) subscene of the Lower
Pisqui River region of eastern Peru was analyzed. This area is inhabited by
Shipibo Indians and some mestizos, who are primarily subsistence farmers but
also produce cash crops, raise cattle, and cut timber for sale in regional mar-
kets (Behrens 1986a, 1989, 1992). This image consists of 512 x 512 pixels,
covering an area of about 15x15 km. The data for bands 1, 4, 5, and 7 were
extracted from tape and processed with ELAS, Earth Resources Laboratory
Applications Software (NASA 1987).4
To make a land-use/land-cover classification for this region, an unsuper-
vised classification was performed with the cluster analysis and maximum like-
lihood modules contained in the ELAS library. Cluster analysis produced six-
teen spectral classes and these were interpreted in two ways (Behrens 1991).
First, all pixels for a spectral class were highlighted to determine their location
in the image. This step yielded the following tentative groupings of spectral
classes into land-use/land-cover categories: Cropland (classes 4, 6,10,11,15),
Pasture (class 14), Forest Land (classes 1, 3, 5, 7, 9, 13), Water (class 8), Wet-
land (classes 2 and 12), and Barren Land (class 16) (cf. Anderson et al. 1976). A
second step plotted spectral class centroids within the feature space created by
combinations of the TM bands used for classification. By examining these
plots, it was possible to interpret some of the more enigmatic classes, gener-
ated statistically, through their association with better-known classes with which
they clustered. In addition, since each TM band actually detects different lev-
els of water absorption and biomass in objects on the ground, the location of
spectral classes in plots informs one about physical and biological properties
shared by classes within each cluster, further aiding the assignment of spectral
classes to land-use/land-cover categories.
With the methods described above, the Lower Pisqui TM subscene in Fig-
ure 4.5 was classified and painted to show the distribution of land use/land
cover types in the vicinity of Shipibo settlements. While this image represents
only a preliminary attempt at classification, it seems to provide a potentially
accurate picture of land use-patterns and landscape heterogeneity. With bet-
ter ground-truth data, it should be possible to distinguish gardens with differ-
ent crops and, perhaps, fallow land in different stages of forest succession.
The information content of satellite images such as this can be exploited
even more with a GIS. A GIS consists of data layers that are georeferenced—
i.e., rectified to the same map projection (Burrough 1986; Jensen 1986). Once
a classified satellite image is integrated in a GIS, it is possible to use its compu-
tational power to estimate variables of interest. For example, it is known that
TM data have a 30-m spatial resolution. So the GIS merely counts the num-
ber of grid cells in a data layer that share a particular land-use/land-cover
classification. Then, this sum is multiplied by 900 m2 to calculate the total
amount of land area for the class.
To demonstrate the potential of integrating classified remote sensing data
Figure 4.5 Land use/land cover classification of lower Pisqui Thematic Mapper subscene.
with a GIS, the classified Lower Pisqui subscene in Figure 4.5 was imported
into the Geographical Resource Analysis and Support System (GRASS), a public
domain GIS developed by the Construction Engineering Research Labora-
tory (CERL) of the U.S. Army Corps of Engineers (USAGE 1991). Once a
GRASS data layer, each pixel in the classified TM image becomes a single cell
in a raster or grid. An extensive set of commands exists in GRASS to display,
overlay and statistically process data layers. For this study, GRASS was used to
produce areal estimates of land use/land cover categories in the Lower Pisqui
TM subscene. These are listed in Table 4.1.
Approximations such as these are useful for testing hypotheses implicated
by the proposed theories. Measures of intensification can be formed from the
ratio of cultivated to fallow land and levels of deforestation compared for dif-
ferent settlements. When combined with attribute data such as settlement
population sizes, sizes of herds, time allocations to work activities, and each
settlement's distance from its nearest marketplace, the effect of changes in
these variables on land-use patterns can also be examined quantitatively.
Discussion
Though not addressed formally, the theories above anticipate still other changes
in land use patterns as local economies become further commercialized. In
addition, this study underscores the importance of formalization in cultural
ecology and the central position given to cultural knowledge in theory build-
Table 4.1 Estimated areas for each land use/land cover category in the lower Pisqui
Thematic Mapper subscene.
Category Hectares % of Total Area
Cropland 1,871 7.93
Pasture 26 0.11
Forest Land 18,068 76.58
Water 739 3.13
Wetland 2,664 11.29
Barren Land 225 0.95
Total 23,593 100.00
ing. A review of related literature reveals that, in many ways, the interests in
regional land-use patterns among cultural ecologists converge on those of land-
scape ecologists. These points are discussed below in more detail.
Other Theoretical Implications
The proposed theories have highlighted the significance of spatial factors and
travel time for understanding intensification of land use among indigenous
Amazonians. These theories assumed that all producers travel by foot from
their houses to locations surrounding their settlements. However, the ethno-
graphic literature suggests that further deforestation often occurs when cash
is used to purchase improved modes of transportation, e.g., motorized canoes
or pack animals (Behrens 1989). This effect is anticipated theoretically by
decreasing the travel time, and hence the total cost, of putting new land
into production.
Small satellite settlements may form when either the transportation costs
or land degradation associated with living in a central village become intoler-
able (cf. McCall 1985; Siddle 1970). Such settlements, often established on
sites to which families can claim ancestral ties, may become permanent with
improved transportation to the parent village or when producers have rela-
tives in the village to care for school-age children. Over time, a satellite settle-
ment may actually grow large enough to support (or seek support for) some of
the same facilities and services found in the village—e.g., a school or health
clinic. Gradually, land between these two villages is also cleared, producing
even more widespread deforestation.
As the third theory shows, outside buyers—e.g., itinerant merchants/ped-
dlers—can also stimulate land-use intensification by increasing demand for
production. By purchasing output in a village, they reduce the travel costs to
producers, but usually at a reduced price paid to producers. Once a village
amasses sufficient political clout, it may pressure government agencies to build
a road linking the village to regional markets. Even without colonization by out-
siders, this would most likely have the effect of increasing the amount of land put
into production by either increasing the number of outside buyers or by raising
the price paid to producers for the goods diey deliver to market themselves.
Role of "Culture" in Theory Building
The theories developed in this paper are based on the idea that human adapta-
tion is the interaction individuals have with their environment as part of a
decision-making process aimed at providing solutions to resource exploitation
problems they perceive. Culture is accorded an important role in providing
useful conceptual categories and "folk" principles for making decisions about
labor allocations and resource consumption (Chibnik 1981; Johnson 1982;
Keesing 1974). Here, the focus is on the native decision maker who, while
drawing on accumulated ecological knowledge encoded in culture, evaluates
perceived opportunities and applies available technology within the constraints
set by the local social and physical environment.
As mentioned earlier, a test of any of the proposed theories requires a model,
here meant an ethnographic interpretation of the theory, using terms and con-
cepts understood by natives (Behrens 1986a). Cultural principles governing
time allocation and land management are indispensable for model building.
Activity classes, production objectives, and land types, along with appropriate
uses for each, are informed by indigenous culture. For example, while it may
be acceptable to fell trees when clearing land for a new garden, the felling of
trees to create pasture or to sell for timber may represent behavior not as
highly regarded by natives. Similarly, the value that people place on money
and its use to purchase other things, such as food, also varies across cultures.
The number of potential sites where a population may settle, or just the
mere conceptualization of their existence, is a simple measure of sedentism.
But the value placed on village life is really a cultural ideal that expresses the
importance of services and social interaction to people who live together. Be-
cause the theories revealed that sedentism often contributes to increasingly
intensive uses of land, they highlight the need to discover those values holding
a village together. Concomitantly, forces that counter greater social integra-
tion in a village, such as growing political intrigue and social tension or in-
creasing number of human/animal conflicts (Rappaport 1968), should also
be explored fully.
Convergence with Landscape Ecology
When cultural ecologists analyze regional land-use patterns, their interests

overlap those of landscape ecologists in many ways. While landscape ecology
tends to focus on the manner in which spatial patterns affect ecological pro-
cess, it also recognizes that "the landscape (like many ecological systems) rep-
resents an interface between social and environmental processes" (Turner 1989:
Using GISfor Cultural/Ecological Land-Use Analysis 75
189). Both disciplines attempt to measure and account for spatial heterogene-
ity resulting from human disturbance in land forms. Hence, it is not surprising
that landscape ecologists have expressed the need for many of the same
research tools, such as remote sensing data and GIS, as cultural ecologists
(Turner and Gardner 1990).
Shared interest among cultural and landscape ecologists in the analysis of
environmental mosaics also entails concern for spatial and temporal scale (Ur-
ban et al. 1987). As the proposed theories showed, the amount of heterogene-
ity in land forms changes through time as increasingly intensive uses of land,
involving more regular clearing of primary and secondary forests, can lead to
less biodiversity. Moreover, certain local-level processes, such as tropical de-
forestation, are repeated in other regions, sometimes yielding environmental
changes on a global scale (Botkin et al. 1989).
It should be noted, however, that cultural ecology differs from landscape
ecology in its emphasis on culture and its contribution to the formation of
landscapes. The theories developed in this paper are based on individual deci-
sion makers who apply their cultural knowledge in the management of time,
land, and other resources. Thus, indigenous culture is also given a role in
delineating spatial and temporal scale for individuals, and in providing the
many schemes they use to classify different components of the landscape. Rather
than approaching the analysis of landscape heterogeneity naively, this orienta-
tion tends to anchor processes of disturbance on a landscape in areas of human
settlement, where they originate.
Advantages of Formalization
The formal theories developed in this paper helped to organize and make
explicit what have often been presented as competing or disparate ideas and
showed the complex manner in which they are related. Formal theories such
as these can guide the identification and field measurement of those variables
required to test a theory and its derivative models. This study generated
three land-use patterns from first principles discovered through an ex-
amination of the ethnographic literature. Conclusions derived for each of
these theories reconciled different views about the causes for land-use in-
tensification in Amazonia.
The importance of deriving first principles for a theory from the ethno-
graphic record cannot be overstated. The fact that the properties and conclu-
sions of the proposed theories have a familiarity about them provides a firm
footing for future elaboration (in much the same way that the Lotka-Volterra
equations provide a formal basis for evolutionary ecology, or the Hardy-
Weinberg model gives population genetics its foundation). As an example, if
comfortable with the theories above, one may explore logically the conse-
quences of adding new means of transportation such as motor boats or mules,
or improvements to land such as fertilizer, or even changes in the sexual divi-
sion of labor so that more garden work is performed by women after men
take up cattle production.
Conclusions
In this paper some factors, that had been offered alone in previous frameworks
to explain land use intensification in Amazonia, were shown to be inextricably
related in complex ways. Formal theories were developed to support the claim
that perceived changes in rates of return for time invested is a better reason for
changing land use patterns than the realization that some absolute minimum
level of "limiting" resource has been exceeded. By incorporating travel time as
a key concept, these new formulations weighed the total cost of intensification
against the perceived benefits of sedentism and production for markets.
These theories demonstrated how population growth, increasing sedentism,
and greater market participation can lead to increasingly intensive uses of land.
Shifting settlement was shown to be one means of adjusting a population's
demand for resources to land degradation. However, once people aggregate in
a village, population growth and market participation can force labor increases
and decreases in per capita land utilization—i.e., more intensive uses of land.
Hence, population growth and market participation tend to increase the length
of time a plot is utilized, yielding more widespread deforestation and less
biodiversity. By stressing the effects of travel costs on land utilization, the theo-
ries provide some possible reasons for landscape heterogeneity, even with the
low population densities typical of Amazonia. But the theories presented in
this paper only represent a simple attempt to formalize existing ideas about a
very complex cultural ecological process. Yet with the growing availability of
new data and more sophisticated analytical tools, such as GIS and digital im-
age processing of remote sensing data, it should be possible to refine these
theories and advance our understanding of land use intensification among
indigenous Amazonians.
Acknowledgments
This research was supported by NSF grants BNS 88-12823 and BNS 89-11090.1 wish
to thank Thomas Sever of NASA's Science and Technology Laboratory at the Stennis
Space Center, Mississippi, for providing the satellite imagery used in this study. Assis-
tance with image processing was given by Richard Sellers, Don Powell, and Fred Mayer
of Lockheed Engineering and Sciences, who also are located at the Stennis Space Center.
I am also grateful for comments made by Duran Bell on an earlier version of this paper.
Notes
1. Steward (1955: 90) was quite explicit on this point: "If cultural studies have a scien-
tific rather than descriptive purpose, cross-cultural comparisons should lead to causal
formulae which state that certain synchronic or diachronic conditions for factors pre-
suppose certain other conditions or factors. If features A, B, and C occur only when
features D, E, and F occur, some fundamental relationship between all six features may
have identity or near-identity of both form and function..."(emphasis added).
2. The conclusions of this theory should be affected little by choice of geometry pro-
vided thata somewhat regular form is assumed. With an additional simplifying assump-
tion, the area of many polygons can be expressed as a function of a single parameter,
such as a radius or side. For example, one might assume that utilized land is enclosed by
a rectangle, an assumption that seems reasonable for many riverine settlements (e.g.,
Behrens 1986a). If it were further assumed that there was a tradeoff (K) between ex-
tending a clearing farther along the river (base b of the rectangle) or deeper into the
forest (height kb of the rectangle), then this area could also be represented as A = b(kb)
= kb2 = an, and sott - gb"2 = g' (an)1/2&s dderived in the text.
3. For the existence of a minimum, the Lagrangean must be concave upward at the
optimum. Therefore, second-order conditions require that the determinant of the Ja-
cobian matrix (J) of endogenous variables be negative (Chiang 1974; Kuznar 1991).
4. The Multi Spectral Scanner (MSS) and Thematic Mapper (TM) are the two sensors
on board Landsat satellites 4 and 5. The Pisqui image used in this study consists of
parts (path 7, rows 65 and 66) from two Landsat 5 TM scenes (Y50987142 81X0 and
Y5098714284X0) acquired on November 13, 1986. Bands 2 and 3 were not processed
because they exhibited striping—distortion caused by a malfunction in one of the sen-
sors—nor was band 6 examined because of its different spatial resolution.
5
Integrating Socioeconomic and
Geographic Information Systems:
A Methodology for Rural Development
and Agricultural Policy Design
Susan Stonich
Understanding the factors related to destructive ecological processes in the

tropics has expanded significantly in the last decade. Much has been learned
about heterogeneity in geomorphology, soils, hydrology, and climate and about
associated vulnerability to ecological damage. Research on cropping systems
has divulged both the suitability and the liability in swidden agricultural prac-
tices and has led to recommendations involving alternative cropping and
agroforestry complexes (Altieri 1987). At the same time, there has been a grow-
ing awareness that a more comprehensive knowledge of tropical ecology and
enlarged technological and/or agricultural options will not necessarily affect a
sustainable ecology (Altieri and Hecht 1990; Redclift 1984, 1987). Research
on peasant economies in Latin America and elsewhere has demonstrated the
existence of a highly differentiated peasantry, the vast majority of whom are
landless or land-poor and who are more dependent on income earned from
off-farm than from on-farm sources (Collins 1986; Deere and Wasserstrom
1981; Stonich 1991b). Such studies have demonstrated that systemic inter-
connections among family and corporate farmers with landholdings of all sizes
promote environmental destruction (Stonich 1989); have established the ex-
istence of labor scarcity rather than labor surpluses in many peasant commu-
nities and the related environmental consequences (Brush 1977,1987; Collins
1987,1988; Posner and MacPherson 1982; Stonich 1993); and have called for
rural and agricultural development policy that takes into account a socially
differentiated peasantry and diversified rural poverty (de Janvry and Sadoulet
1989). It is increasingly evident that ecological destruction cannot be fath-
omed apart from the demographic, institutional, and social factors that influ-
ence the agricultural practices and other natural resource management deci-
sions of agricultural producers.
This paper describes a multidisciplinary methodology designed to examine
the interactions among demographic trends, social processes, agricultural pro-
duction decisions, and ecological decline in southern Honduras, a region char-
Integrating Socioeconomic and Geographic Information Systems 19
acterized by widespread and worsening human impoverishment and environ-

mental degradation (Figure 5.1). The methodology integrated the research
efforts and databases compiled by anthropologists from the University of Ken-
tucky using a farming systems approach, who were part of the socioeconomic
component of the International Sorghum Millet Project (INTSORMIL) with
potentially complementary research conducted by the natural and agricultural
scientists working as part of the Comprehensive Resource Inventory and Evalu-
ation System Project (CRIES) at Michigan State University. Very broadly, the
social science research used a political economy approach to examine the rela-
tionships between household production and consumption and to raise ques-
tions concerning how the expansion of export-oriented capitalist agriculture
in the region affected land tenure and related aspects of agrarian structure that
influenced land-use decisions. Research findings are available in a number of
publications and technical reports (see, e.g., DeWalt and DeWalt 1982; DeWalt
and Stonich 1985; Stonich 1986, 1989, 1991a-c, 1992a, b, 1993). In brief,
analysis revealed that agricultural change, and related changes in land distri-
bution and crop allocation patterns, resulted in mutually reinforcing conse-
quences: the exacerbation of underdevelopment and poverty and the adoption
of survival strategies that are incompatible with sustainable land use. These
therefore, lead to ecological deterioration. The focus of this paper is not on
the results of the research but rather on the methodology. The methodology
represents an example of the integration of two multidisciplinary projects: one
an agricultural development project dominated by agricultural scientists (plant
breeders, plant pathologists, etc.) and the other primarily a natural/agricul-
tural resources project commanded by natural resource scientists. Both projects
Figure 5.1 Map of Honduras and southern Honduras.

were significantly enhanced by the inclusion of GIS methods. Significantly,
however, the integration of the projects was perceived, organized, and carried
out by anthropologists.
The Overall Research Methodology
The general goal of the methodology was to provide a systemic framework in

which to articulate the relevant factors and levels of analysis involved in the
social and ecological processes related to the expansion of capitalist agricul-
ture (Figure 5.2). The ultimate objective was to integrate a representative sample
of microlevel studies (of individuals, households, and communities) into a hi-
erarchical structure composed of multiple, relevant, and increasingly
macrolevels of analysis. A further aim was to add the dimension of history by
positioning the traditional microlevel foci of anthropology—individuals and
the community—at the convergence of local and world history. The underly-
ing assumption that variability existed in terms of all factors and at all levels
required that sampling strategies be designed so as to be capable of collecting
data representative of that heterogeneity.
In this article, it is possible only to sketch the research methodology de-
signed to accomplish these numerous objectives. (For more thorough treat-
ment of the theoretical, methodological, and applied issues involved, see Stonich
Figure 5.2 Relationship among various sources of data, relevant factors, and
levels of analysis.
tfl
Integrating Socioeconom.ic and Geographic Information Systems 81
1992b.) The fundamental task was to identify the factors (defined and
operationalized in terms of sets of variables) that could be used to link the
relevant social and ecological processes through the appropriate levels of analy-
sis. Of the several factors that could have been chosen, patterns in land use
appeared the most germane and an analysis of changes in land-utilization pat-
terns emerged as the most promising means of synthesizing the social and
ecological processes occurring in the region. Employing land-utilization pat-
terns as a window for integrating these processes was particularly useful for a
number of reasons. First, the spread of commercial agriculture in such con-
texts fundamentally alters the ecology in areas in which it occurs. An examina-
tion of changes in land-utilization patterns—the changes in the ways in which
human populations exploit land resources—specifically focuses on these eco-
logical transformations, highlighting the interaction of biophysical and social
processes. Second, land-use changes could be examined at many levels, from
the study of the land-use decisions of one farmer to superhousehold and com-
munity patterns of land use, and from those to regional, national, and ulti-
mately world-wide configurations. Finally, such an analysis could aid in the
evaluation of determinants influencing the evolution of land-utilization pat-
terns through identifying the producers, the beneficiaries, and the losers of a
given pattern of production and the changes in consumption patterns that
accompany such transformations.
The choice of specific levels of analysis (and appropriate scales) from which
data were gathered was based on theoretical considerations but was influenced
as well by the conceptualization and construction of the levels of analysis re-
flected in the data available from national and international sources. For ex-
ample, little relevant data (social and/or ecological) had been collected by such
agencies and/or were available at the levels of the individual householders or
particular communities. Rather, such agencies provided information at the
levels of the municipality (similar to counties in the United States), the de-
partment (similar to states in the United States), and the region ("the South"
is treated as a separate region of the country as defined by national and inter-
national agencies). As much as possible, both qualitative and quantitative data
were collected from such agencies in the forms of written materials (e.g., pub-
lished and unpublished documents, statistics, maps, census data, and so on)
and interviews with administrators and officials. Considerable care was given
to the collection of data from such sources both early in the research process
and continuously throughout the study, and the contacts made during the ini-
tial search and interviewing process proved invaluable. Given the macrolevel
focus of data from such sources, the core of the anthropological research was
the study of individuals and communities—the traditional loci of anthropo-
logical research (Ortner 1984). The objective was to obtain data from a sample
of communities that were representative of the variation in social/environ-
mental interactions that existed in the region. After a preliminary study of
written documents, numerous interviews with officials, and a short period of
initial field research throughout the south (one month), a sample of research
communities were chosen to represent the agroecological variation present
within the region. Using a combination of ethnographic (including both quali-
tative and quantitative techniques) and survey research methods, primary data
were collected from samples of male and female householders in these com-
munities. Variables were defined and operationalized so as to be analogous to
the more macrolevel data available from the published and unpublished re-
ports of national and international agencies. In summary, the research meth-
odology involved the integration of data on land utilization patterns from the
level of the male/female householder, through the community, municipality,
region, and nation. Minimally, for each level, data were obtained on (1) the
ways land was classified (in terms of biophysical characteristics, quality, use,
and so on; (2) the relative allocation of land (e.g., cropland versus pasture); (3)
the agricultural and other practices that involved the use of land (e.g., farming
and cropping systems and fuel-wood collection); and (4) land tenure (the dis-
tribution, size, and ownership/rentership of land).
Because research goals included understanding trends over time, an attempt
was made to collect data that spanned a number of years. At the levels of the
nation, region, department, and municipality, government publications were
most useful. Agricultural censuses from 1952, 1965, and 1974 were consulted
and then supplemented by data from more recent regional development, natural
resource, and agricultural reports, as well as by data from remote sensing de-
vices (aerial photos and satellite imagery). At the local level, additional inter-
views designed to measure changing use of land resources over the same thirty
year period were administered to farmers. These were augmented by moni-
toring selected parcels of land for a period of one year.
Documented changes in land-utilization patterns were tied to social factors
influencing the evolution of such patterns and to their ecological outcomes
through an integrated master database. This database was used first as input to
the geographic and agroeconomic information system developed by CRIES
(1984) and later disaggregated into files managed by dBASEIII+, although
other database managers could have been used. The master database consisted
of physical, ecological, and agricultural information contained in digitized land
use, topographic, soil, and land parcel maps; tabular ecological data collected
from parcel monitoring; as well as primary data collected through various ag-
ricultural, socioeconomic, and nutritional household surveys (Table 5.1). The
result was a geographically referenced database keyed to the level of the house-
hold but capable of being disaggregated to the levels of the male and female
householder or concatenated to the levels of the community, municipality, or
region. In addition, the identification variables (of individuals, households,
communities, etc.) in the master database were keyed to those in the com-
puter-coded ethnographic field notes, making it possible to coordinate searches
of field notes with analyses of the more "quantified" information included in
the master database.
Integrating Socioeamomic and Geographic Information Systems 83
Table 5.1 Sources of data compiled by INSORMIL for Pespire.
DATA SOURCE - RESOLUTION/SCALE

E E
DATE PRIMARY USE
E
REMOTE SENSING:
Landsat (MSS) - 79 m. x 79 m. 1978 Land cover/use
Aerial Photos - 1:40,000 1946 Land cover/use
1973 Land cover/use
1980 Land cover/use
Orthophotos - 1:10,000 1982 Land cover/use
MAPS:
Topographic- 1:50,000 1965 Agro-ecological
1:20,000 1982 Agro-ecological
Soil 1:20,000 1982 Agro-ecological
Cadastral 1:10,000 1982 Land tenure
LAND PARCEL MONITORING:
Soil samples 1983-5 Agro-ecological
Erosion measure 1983-5 Agro-ecological
Agricultural production 1981-5 Agricultural
SURVEY:
Household surveys - male householders 1981-90 Agricultural
Economic
Ecological
Social
Household surveys - female householders 1981-90 Nutritional
Demographic
Economic
Social
Ecological
COMMUNITY ETHNOGRAPHIES: 1981-90 Demographic
Social
Economic
Ecological
Nutritional
The Integration of Image Processing and GIS into the Methodology
Collaboration between INTSORMIL and CRIES

INTSORMIL is one of the commodity-focused Coordinated Research Sup-
port Programs (CRSP) funded by the United States Agency for International
Development (USAID) in response to the so-called New Directions Man-
date. It was one of the first CRSP efforts to be funded because of the impor-
tance of sorghum and millet as food grains in some of the most impoverished
nations of the world. The overall goal of INTSORMIL is to increase the pro-
duction and utilization of these grains. Although most of the funding for
INTSORMIL supports the research of agricultural scientists, a small propor-
tion of funds was allocated to support social science research. The responsibil-
ity of the team of University of Kentucky anthropologists was to use the meth-
odologies and field research techniques of anthropologists to understand the
socioeconomic constraints on the production, distribution, and utilization of
sorghum, which is an important human food in southern Honduras. Focusing
on the farm household and using a combination of ethnographic and survey
research methods, data were collected on the farming, nutritional, ecological,
and economic systems from a sample of nine communities in various
agroecological zones.
The CRIES project was initiated in 1976 with joint funding and participa-
tion by USAID, the U.S. Department of Agriculture (USDA), and Michigan
State University. CRIES used a multidisciplinary approach in assisting devel-
oping nations to assess their natural resources and related agricultural produc-
tion potential with an aim toward integrated rural development. The project's
scope was directed toward the realization of a country's natural resources pro-
duction potential within the framework of relevant regional, national, and in-
ternational constraints. CRIES had chosen southern Honduras as a pilot area
in which to develop and test their system (CRIES 1984).
Data-acquisition methods used by CRIES focused on the systematic, na-
tionwide integration of physical resource attributes, agronomic and economic
characteristics, and political or administrative regions into a spatially defined
database for policy evaluation and design. Current land cover and land-use
information was derived from available remote sensing data obtained by air-
and/or space-borne sensors (e.g., aerial photos, Landsat imagery). Estimates
of erosion, crop yields, area planted, total production, associated costs and
other agronomic and economic factors pertinent to the evaluation of natural
resource and agricultural production potential in defined agroecological zones
were derived from secondary sources and through limited field surveys.
CRIES developed both geographic and agroeconomic information man-
agement and analysis systems (Schultink 1987; Schultink and Amaral 1987).
The agroeconomic information system's analytical routines (linear program-
ming, goal programming, and basic statistical and economic models) are linked
to the geographically referenced databases created and managed by the geo-
graphic information system.
Although admirably able to describe and analyze regional level patterns, a
major shortcoming of applying the macrolevel approach of CRIES to a region
such as southern Honduras is that it tends to mask the considerable social,
economic, agricultural, and ecological variation that exists. In so doing such
an approach does little to help understand the processes that are occurring
more at the microlevel. It is precisely this kind of microlevel variation that the
socioeconomic component of INTSORMIL was designed to measure. On
the other hand, even though INTSORMIL research communities had been
carefully chosen to sample relevant regional variation, the precise repre-
sentativeness of research results based on nine communities was difficult
to know with any precision.
Thus, as both projects evolved, it became increasingly apparent that the
two projects had complementary perspectives as well as databases, the integra-
tion of which could have important methodological and applied policy signifi-
cance. Both projects defined many of the same variables as relevant and were
collecting similar kinds of data (see Figure 5.3). The major difference between
the two projects was the level of analysis on which research was focused. CRIES
concentrated on the compilation of secondary data sources and the subse-
quent analysis of data more at macrolevels (i.e., geopolitical units larger than
the village), while INTSORMIL focused on the collection of primary data more
at microlevels (i.e., male and female householders, farm households, and villages).
After preliminary collaborative efforts with CRIES personnel in Honduras
and Costa Rica, it was concluded that an integrated methodology was feasible
and would represent an innovative multidisciplinary approach combining the
methodologies most often utilized by agricultural and natural resource scien-
tists with those usually employed by social scientists. This initial collaboration
resulted in slight reorientations in variable definitions and data collection that
would allow for more thorough subsequent integration and analysis.
Figure 5.3 Data requirements of both INSORMIL and CRIES.
NATURAL RESOURCE An inventory of land resources identified

CONSTRAINTS qualitatively and measured by hectares
available for agricultural purposes.
An inventory of general land use categories and

specifications of cropping patterns.
PRODUCTION Current agricultural systems - management

CONSTRAINTS techniques by crop and agroecological
zones.
Yields and production costs for each management

technique, crop, and agroecological zone.
Production goals and yields.
SOCIOCULTURAL AND Labor availability (seasonal and total)

SOCIOECONOMIC
Availability of credit
water
fertilizer
Transportation costs
Using GIS to Integrate Satellite Imagery and Other Sources of Data
In summary, the methodology included the following steps:
1. Deriving a land-use/cover map for the Department of Choluteca from

Landsat satellite imagery
At the level of the department of Choluteca, CRIES (1984) determined an

initial land-use/cover classification scheme by visually interpreting LANDSAT
photographs (Figure 5.4).1 The categories were based on differentiation in
tonal, textural, and pattern variations found in the Landsat imagery and dur-
ing supporting fieldwork. The Choluteca inventory was used as a component
in the CRIES-Geographic Information System and was combined with physi-
Figure 5.4 Map of land cover/use derived from Landsat satellite data, Choluteca
Department, southern Honduras.
Integrating Socioecmomic and Geographic Information Systems 87
cal resource attributes (e.g., soils, climate, agroecological information) in or-

der to assess current and potential agricultural production at the level of the
department. (For details on the mapping procedures and final classification
scheme, see CRIES 1984: 59-74).
2. Determining a land-use classification scheme at the level of the mu-

nicipality using digital data from the Landsat multispectral scanner
The first task in integrating information from more microlevels was to refine
the land-use classification scheme at the level of the municipality. To do this,
the municipality of Pespire, in which five INTSORMIL research sites were
located, was chosen. This classification was done on the basis of Landsat digi-
tal imagery augmented by aerial and orthographic photos, using the ERDAS
Image Processing System and associated photographic systems available at
the CRIES center at Michigan State University. Determining land-use was an
interactive process involving numerical taxonomic techniques, topographic
maps, a preliminary land-use classification scheme determined by analyses of
aerial and ortho photos (Figure 5.5), and ground truth supplied by photos and
by persons who had spent considerable amounts of time in the area. Once a
refined classification scheme was established and different land-use patterns
were determined, the scheme and patterns were merged into a master data-
base according to the requirements of the CRIES-GIS.
3. Establishing a hierarchically, nested sample of sites: research area 1

and research area 2
To move to the subsequent (increasingly more micro) level of analysis, the

geographic boundaries of two research areas within the municipality of Pespire
were determined. Both areas were of equal size (20 km2) and defined on the
basis of UTMs (Figure 5.6). Research area 1 was located in the foothills/high-
land region of the municipality and included three INTSORMIL research
communities. Research area 2 was in a zone of higher elevation and encom-
passed two research communities.
Land-use patterns within each research area were calculated using the clas-
sification scheme that had been previously determined at the level of the mu-
nicipality. Topographic maps furnished by the Honduran Institute of Geogra-
phy and soil maps provided by the Honduran Ministry of Natural Resources
were digitized for both areas. As shown by Figures 5.7 and 5.8 considerable
diversity between the two areas was revealed through this effort. In terms of
elevation, the majority of area 1 (63%) was located below 200 m while the
major part of area 2 (54%) was found between 200 and 400 m with the re-
mainder at higher elevations (for the significance of this variation on potential
cropping/agricultural systems, see Stonich 1986). At the same time, the use of
the 1:20,000 soil maps revealed variability in soil types hidden by the macroscale
maps of 1:250,000 used by CRIES for their analysis at the departmental level.
Next, property boundary maps supplied by the Honduran Cadastral office
Figure 5.5 Examples of topographic map and aerial photo used in analysis
(Perspire, Research Area 1).
Figure 5.6 Hierarchy of geographic areas used in analysis (Honduras, southern
Honduras, Choluteca Department, Pespire, Area 1, Area 2).
Figure 5.7 Elevations: Pespire Area 1 and Pespire Area 2.
were digitized and aggregated according to the size of land parcels (Figure
5.9). Finally, all digitized maps were integrated into the CRIES-GIS, which
made it was possible to generate color and line printer maps as well as to
statistically analyze land use and distribution patterns within the research
areas and then relate these to elevation, soils, and so on. Moreover by
moving back to higher levels of analysis—the municipality and the de-
partment—it was possible to compare patterns within the two research
areas with those previously determined for the municipality of Pespire
and the department of Choluteca. Figures 5.10 and 5.11 present two ex-
amples of what was done. Figure 5.10 compares patterns of land use among
the municipality (Pespire), research area 1, and research area 2. It shows a
fairly consistent pattern in land use for the municipality as a whole as well
as for the two research areas: 53 to 61% of land is in cultivation, 13 to
Figure 5.8 Soil types: Pespire Area 1 and Pespire Area 2.
18% is in fallow, 7 to 9% is in pasture, 5 to 6% is covered by deciduous

forest, 1% by pine forest, and 5 to 7% in scrub forest. Figure 5.11, on the
other hand, reveals greatly dissimilar patterns in the relative distribution
of farmland according to farms in different farm-size categories in the
department of Choluteca, the municipality of Pespire, research area 1,
and research area 2; with inequality in the distribution of land decreasing
as one moves from the level of the department, the municipality, research
area 1, and research area 2. For example, while 53% of the farmland
throughout Choluteca and 35% of the farmland in Pespire is in farms of
100 ha or more, only 7% of the farmland in research area 1 is found in
that category and no farms of that size are found in research area 2. In
contrast 77% of all farmland in research area 2 and 58% of all farmland in
research area 1 are on farms of 20 ha or less while such farms make up
92 Antbropology, Space, and Geograpbic Information Systems
Figure 5.9 Landholdings by size of parcels: Pespire Area 1 and Pespire Area 2.
only 32% of the farmland throughout Pespire and 25% throughout the
entire department of Choluteca.
4. Creating a master database using CRIES-GIS and CRIES-AIS
A master database was developed that was used first as input to the CRIES
geographic and agroeconomic information systems (CRIES-GIS and CRIES-
AIS) and later disaggregated and managed by dBASEIII+. The master data-
base consisted of physical, ecological, and agricultural information contained
in the land-use, topographic, soil, and land parcel maps; tabular ecological
data collected from parcel monitoring; as well as data from the agricultural,
economic, and nutritional household surveys. The result was a geographically
referenced database to which information from subsequent survey data or other
sources could easily be added. Subsequently, the database was disaggregated
Figure 5.10 Percentage distribution of land in various land use/cover categories:

Municipality of Pespire, Area 1, and Area 2.
into a hierarchical system of files integrated by dBASEIII+. The latest file

integrated into the system was one containing data on the migration behavior
and characteristics of the children of householders.
The ability to incorporate information at the levels of the household and
community provided the means to determine relative heterogeneity within
those levels and then compare that variability to higher levels of analysis. As
shown in Figure 5.12, in which values for research area 2 come from digitized
cadastral maps and values for village 1 and village 2 (both located within re-
search area 2) are based on the analysis of household surveys, the distribution
of landholdings in village 1 and village 2 are quite distinct—from each other as
well as from the overall pattern within area 2. As revealed, farmland is most
equitably distributed (albeit with a significantly lower mean) in village 1, where
the majority of farmland (62%) is found on farms that are 5 ha in size or less
and where no households have access to farms of more than 20 ha. On the
other hand, 66% of all the farmland owned by families in village 1 is owned by
households with access to farms of 20 to 100 ha in size. The significant degree
of heterogeneity in household access to land between communities revealed
by the survey data is obscured if one considers only the aggregate information
obtained from digitized cadastral data for area 2. At the same time the diver-
sity between the two villages would make any overall estimate of patterns in
land distribution within area 2 quite problematic. This situation demonstrates
the great usefulness in employing complementary approaches: one based on
the geographically bounded area and the other on household surveys and
community ethnography.
Figure 5.11 Percentage distribution of farmland by farm size categories: Choluteca
Department, Municipality of Pespire, Area 1 and 2.
Figure 5.12 Percentage distribution of farmland by farm size categories: Pespire

Area 2, Village 1, and Village 2.
Conclusions
The last decade saw an increasing recognition worldwide of the complex,

multilevel, and multidimensional interrelationships among demographic, so-
cial, economic, and environmental processes. This emerging awareness pointed
to the need for interdisciplinary research efforts aimed at halting human im-
poverishment and environmental destruction in the developing world and
opened the door to enhanced social science participation in research related to
global environmental change (Redclift 1987). This widespread acknowledg-
ment occurred as many anthropologists attempted to expand their traditional
research focus to examine how relatively small groups are integrated into larger
regional, national, and international systems (Ortner 1984). These develop-
ments within anthropology have raised important theoretical issues regarding
the general processes involved in sociocultural change and in the nature of
integration of systems as well as methodological concerns regarding appropri-
ate levels of analysis (and appropriate scales). Recognizing the augmented ef-
forts of anthropologists to understand the relationships between relatively
small-scale and larger-scale processes, DeWalt and Pelto (1985: 187) stress
the importance of developing appropriate methodologies to guide research .
According to them, two of the most important methodological tasks that are
demanded are the clear depiction of systems of postulated relationships among
units and levels of analysis and precise definitions of the relevant factors that
are used as the specific linkages that articulate those levels of analysis (DeWalt
and Pelto 1985: 187). Within these guidelines, the methodology developed
for the research reported in this article attempted to integrate "the environ-
ment" into anthropological research and to incorporate the more traditional
micro-level focus of anthropologists with the more macrolevel locus of agri-
cultural and natural resource scientists. The resulting methodology provided
the means to examine the economic and agricultural decisions of peasants—to
consider peasants as actors and agents of social and ecological change—rather
than merely as passive recipients who are acted upon by the state or other
powerful forces (Stonich 1992b). The use of the CRIES-GIS and CRIES-AIS
afforded the specific means with which this goal was accomplished. Using a
multistage approach, data management using these systems permitted
georeferencing multidata sets but also statistical analysis based on the correla-
tion of spatial distributions.
Recent advancements within and outside of anthropology provide the po-
tential for the augmented participation of anthropologists within the develop-
ment process, especially in terms of establishing the complex multilevel inter-
relationships between social processes and environmental destruction. Realiz-
ing this potential will enhance the contributions of anthropology to wide-
spread concerns regarding global environmental change. Models from cul-
tural and human ecology represent the efforts of a significant segment of a
profession long concerned with constructing theoretical perspectives and meth-
odological procedures forintegrating the natural environment into anthropo-
logical research (Moran 1986). "Human systems ecology" (Bennett 1976,1985)
and "political ecology" (Blaikie and Brookfield 1987) now are emerging as
possible useful research approaches with which to examine the human dimen-
sions of environmental destruction, especially in the developing world (Stonich
1993). But these represent only two of many possible perspectives and ap-
proaches that may evolve in the future.
The methodology described in this article illustrates one attempt to sys-
temically and systematically integrate the natural environment into anthropo-
logical research. The use of GIS—which combines satellite data with other
sources of demographic, social, and economic information—provides the means
to systematically move through various levels of analysis in ways that were
never feasible before. In so doing, however, this research raises several signifi-
cant and more comprehensive issues for future research. Among the most
important of these are considerations regarding choices of specific levels of
analysis and appropriate scales of sampling to answer particular questions about
ecological changes and their human dimension. Another vital concern has to
do with training ourselves and our students in order that we may as fully as
possible communicate with colleagues in other vital disciplines and use the
rapidly emerging technologies in the best ways possible.
Acknowledgements
The research for this article was supported by the International Sorghum/Millet Col-
laborative Research Support Program (INTSORMIL) through contract #AID/DSAN-
G-0149, by a Fulbright Senior Scholar Award through their Central America Repub-
lics Research Program, by a University of California Academic Senate Research Grant,
and by two University of Kentucky Summer Fellowships.
Note
1. The Landsat imagery interpreted for this project utilized diazo processing for con-
trast stretching and color composite production. The technique consisted of printing
Landsat 185 mm x 185 mm positive black-and-white transparencies onto color diazo
film. For details see CRIES 1984.
6
Empirical and Methodological Problems
in Developing a GIS Database for Yanomamo
Tribesmen Located in Remote Areas
Ken McGwire, Napoleon A. Chagnon,
and Charles Brewer Carias
For almost thirty years Chagnon has been studying the settlement patterns of
a large cluster of remote Yanomamo communities in southern Venezuela, docu-
menting population growth, mortality patterns, fissioning, dispersal, and pio-
neering of adjacent virgin areas of tropical forest. Approximately fifteen vil-
lages, with a current (1992) population of about 2000 individuals, have been
studied. During a period of some 150 years, members of these communities
have cleared and subsequently abandoned approximately five hundred sites
whose geographical locations are very poorly known. In 1990 and 1991, Charles
Brewer Carias, a Venezuelan naturalist, joined Chagnon in this research ef-
fort. Recent field research has resulted in geographic and demographic data
suggesting that long-term warfare patterns may be contests over the appar-
ently more desirable lowland areas, where economic activities are less costly in
terms of energy and resources are more abundant or easier to obtain. Periodic
village movements, provoked by hostilities with neighbors, require that rela-
tively large lowland areas must be controlled so that groups can move around
within them and maintain maximum distance from enemy groups. To do this,
lowland villages must grow large and politically bellicose. When they fission,
usually at a size of about 150 to 200 people, some of the resulting smaller
groups are driven out and take refuge in more rugged but economically less
productive highland terrain, where they adopt a less bellicose political stance
toward their neighbors. Rates of mortality due to warfare, frequencies of
abduction of women from neighbors, and other sociodemographic at-
tributes distinguish highland from lowland communities in the overall area
(Chagnon 1992).
Geographic information systems are considered effective methods for or-
ganizing and analyzing the variety of spatial information required to test such
hypotheses of relationships between environment and social processes. A GIS-
based approach would allow maps of parameters relating to resource distribu-
tion and environmental characteristics to be compared to a rich and growing
record of field observations. Analysis based on GIS would support data man-
agement requirements by allowing accurate identification and positioning of
cultural and environmental features within a consistent map base. Such capa-
bilities for managing spatially referenced data will be required for both the
historical and ongoing monitoring of populations, especially epidemiological
patterns and associated sociodemographic data. The consistent organization
of data within a GIS would greatly enhance the ability to analyze relationships
such as co-occurrence and proximity within acquired data. This would allow
accurate estimation of the resource base and carrying capacities within the
region, modeling of site suitability, and assessment of relationships between
environmental and cultural characteristics. The GIS would also allow the cre-
ation of valuable derived information products that represent a synthesis of
multiple factors. Such products might include calculation of travel times for
migration of individuals, innovation, and disease. Certain large-scale environ-
mental phenomena, such as hydrologic regimes and vegetation productivity,
might be estimated as well. Finally, the graphic orientation of a GIS database
allows effective presentation of social and demographic information. This ca-
pability goes beyond the creation of traditional map products to include three-
dimensional terrain views and potential animation of dispersion in popula-
tions, innovation, and disease patterns through time.
The GIS allows for more effective storage, analysis, and presentation of
spatial data than is possible with hard copy products. It provides a framework
for processing spatial data that exceeds the capabilities of relational database
technologies as well. However, the power of GIS-based approaches requires
consideration of numerous factors. The most fundamental of these is the avail-
ability and cost of spatially referenced data. The cost of data acquisition and
entry is often estimated at as much as 90% of the overall cost of creating a GIS
for developed areas, where much data of good quality may already exist. Data
products that are brought into the system will have variable degrees of accu-
racy and precision in both spatial positioning and measured attributes. This
error must be understood and accounted for in all subsequent operations us-
ing that data, as errors tend to accumulate throughout an analysis. Cultural
and environmental features will also be dynamic at varying time scales. Changes
may be gradual or sporadic, making estimation of the effective lifetime of a
data set difficult to estimate. Temporal transitions in mapped features will af-
fect both of the aforementioned cost and accuracy issues.
In order to incorporate data from a number of disparate sources into a GIS
environment, the accuracy and precision of available data must be considered
in relation to the desired spatial analysis and information products. The accu-
racy and precision of cartographic products in both positional accuracy and
thematic content must be compatible between data sources and of sufficient
quality that the sum of errors does not jeopardize the validity of results. Inac-
curacy in data sources will jeopardize analyses by introducing unwanted asso-
ciations between variables. In map-based analysis, this concern goes beyond
simple misclassification or blunders in data entry because even identical, le-
gitimate boundaries from two map products will never overlay exactly. Preci-
sion, as indicated by map scale, will also place fundamental limits on the qual-
Developing a GIS Database for Yanomamo Tribesmen 99
ity of information that can be obtained from the database. Precision of the
taxonomic divisions in mapped objects will set a conceptual limit on what in-
formation can be extracted. The precision of map scale will determine the
degree of line generalization that occurs along boundaries and the amount of
internal heterogeneity allowed within mapped objects. The geometric prob-
lems associated with merging data sets of differing scales may result in errors
that could greatly reduce the informational value expected from original map
products. A summary of the most important considerations regarding accu-
racy/precision problem includes the following (Goodchild and Gopal, 1989):
1. The precision of GIS processing is effectively infinite.

2. All spatial data are of limited accuracy.
3. The precision of GIS processing exceeds the accuracy of the data.
4. In conventional map analysis, precision is usually adapted to accuracy.
5. The ability to change scale and combine data from various sources
and scales in a GIS means that precision is usually not adapted to accuracy.
Existing Data Types and Characteristics
In order to relate social characteristics to terrain, elevation, and other geo-

graphical attributes, data collection has begun to accurately locate current and
historical occupation sites and relate these to map data. Chagnon and Brewer
Carias have used helicopters and global positioning system (GPS) instruments
to locate latitude and longitude coordinates of Yanomamo sites, such as cur-
rent and long abandoned villages and gardens. Using GPS receivers, one can
receive navigational information from a series of satellites operated by the
U.S. Department of Defense. A single portable receiver is capable of provid-
ing the coordinates of a point in the field to within ±15 m. The database that
has been created from field observations includes locations of features such as
villages, garden clearings, prominent mountains, elevations of many sites, and
river confluences. These GPS observations allow socio-demographic data to
be related to environmental phenomena and provide a baseline for tracking
settlement fissioning and migration through time.
Practically all maps for the region were derived from a mosaic of side-look-
ing airborne radar (SLAR) imagery created by the International Aero Service
Division of Goodyear Aerospace in 1971. This type of radar mosaic can be
quite useful for regional assessment of natural resources. The SLAR imagery
is generated by continuous scanning of radar response obtained at an oblique
angle to one side of the aircraft platform. Because of the low platform altitude,
oblique angle of acquisition, and continuous scanning technique, the geom-
etry of SLAR imagery is greatly affected by both changes in surface elevation
and aircraft attitude. Topographic relief causes foreshortening, layover, and
shadowing in imagery. Foreshortening will cause higher elevations in the im-
agery to be shifted toward the location of the flight line. "Layover" describes a
pathological case of foreshortening which may be encountered in steep ter-
rain. Shadows in radar imagery are areas that cannot be imaged by a given
flight line because of obstruction by topographic relief.
Strips of overlapping SLAR imagery gathered from numerous flight lines
were combined to create the base map for the region. Such mosaics attempt to
match features along adjacent strips while adhering to the locations of known
geodetic control points. Such points of known geographic coordinates are of
limited number in the region. Though the positional accuracy of the com-
bined SLAR products may be good for certain obvious and culturally relevant
features, such as river confluences, geometric distortion within the sparse
ground-control network may be high. This is especially true in areas that are
of different elevation from selected control points or that have high local re-
lief. Estimates of planimetric accuracy of SLAR image products were reported
to be ±2 min of latitude/longitude (MARNR-ORSTROM 1988). This trans-
lates to a distance of approximately 3.5 to 4 km on the ground. Because of
these effects on geometric fidelity and image data quality, radar imagery is not
generally a good candidate as a base map for GIS. The primary reason that
SLAR imagery was used rather than aerial photography in the 1971 mapping
effort was that radar imaging systems are not limited by the persistent cloud
cover of the humid tropics.
In comparisons between field-measured GPS coordinates and the SLAR
mosaic maps, regular discrepancies were found in the reported position of
geomorphic features. This observed degree of error is unacceptable for the
desired spatial analysis of social processes relevant to the study of Yanomamo
cultural ecology. Such error would allow confusion regarding the environ-
ment associated with specific village and garden sites. This level of accuracy
cannot support investigations that focus on the aggressive exclusion of weaker
villages from more desirable areas, as villages located near transitional zones
may be associated with improper environmental characteristics. In addition,
these geometric distortions may affect the estimation of the total resource
base that is available in a particular region.
An example in the discrepancy between features measured by GPS and
mapped in SLAR imagery is presented in Figure 6.1. A portion of the Upper
Siapa River drainage basin that has been digitized from SLAR-based maps is
presented. Locations of field-measured GPS readings are plotted for com-
parison. Discrepancies are often as large as 1 km. In this case the error is
systematic shift northward, perhaps indicating that local elevation is different
than that of the closest geodetic control points. Areas of greater slope that
separate drainages or which are nearer to headwaters are likely to be more
affected by topographically induced distortions than those points indi-
cated in Figure 6.1.
A difficulty arises in that almost all environmental information for the re-
gion, such as the commonly cited Atlas del Inventario de Tierras del Territorio
Federal Amazonas by the Ministerio delAmbientey los Recursos Naturales Renovables
(MARNR, Venezuela) and the Office de la Reserche Scientifique det Technique
Outer Mer (ORSTROM, France), were primarily derived from this SLAR base
Figure 6.1 Positional error in SLAR mosaic for part of the Upper Siapa basin.
map. Though a second atlas and some Landsat MSS images were used in the
compilation of the MARNR-ORSTROM atlas, the SLAR mosaic was used as
both the cartographic base and for interpreting land-cover characteristics. In
addition to known problems with geometric accuracy, the SLAR mosaic was
compiled at a scale of 1:250,000. All maps derived from this imagery are lim-
ited by this initial scale of observation. The level of detail that can be derived
from this scale may be difficult to relate to actual site conditions at the scale
affecting a single village. For example a typical village and garden complex
might extend over a linear distance of 250 m. This translates to 1 mm on the
MARNR-ORSTROM atlas maps. Minimum mapping units in map compila-
tion generally cannot be much smaller than 2 mm across, or else the width of
lines being drawn on the map and the ability to label mapped units becomes
quite problematic. Visual inspection of the MARNR-ORSTROM atlas actu-
ally suggests a larger minimum mapping unit than 2 mm, as would be appro-
priate given the characteristics of the image base. As a result, internal hetero-
geneity which would not be represented within mapped units, may be capable
of supporting entire villages.
The potential for geometrically correcting maps derived from the SLAR
mosaic using field-measured GPS coordinates was examined. The process of
distorting the SLAR-based data to match known coordinates is referred to as
geometric rectification or rubber sheeting. This type of correction may be accom-
plished by applying a single polynomial transformation to the SLAR data based
on a least squares fit between control points which are identified in both the
map and GPS coordinate systems. Another possible approach develops a tri-
angulated irregular network from the control points and applies a locally de-
veloped transformation within each triangulated facet. The global polynomial
approach is solved numerically and allows an estimate of the residual geomet-
ric error. The triangulated network approach is solved analytically and allows
localized distortions to be fitted independently.
An initial investigation suggests that because of the complex distortions in
the SLAR mosaic, a simple rectification using either of these two methods
probably would not improve overall accuracy significantly. The frequency and
irregularity of distortions due to topography is too complex to fit with the
mathematical surface of a polynomial transformation. The triangulated net-
work approach would be more suitable for dealing with this complex distor-
tion. However, the existing network of GPS control points is too limited to
describe the geometric distortion within the region as well. The geometry of
SLAR-based products in the vicinity of GPS measurements, primarily along
river courses and on mountaintops, could definitely be improved with either
method. However, as of this time, there is insufficient control to significantly
improve the geometry so that GIS analysis could confidently be used through-
out the region. An additional problem is that features that appear to be promi-
nent when measured in the field are difficult to identify confidently on the
1:250,000 scale maps.
Remote Sensing Alternatives

The accuracy and precision of existing map products appears to be of insuffi-
cient quality to support the desired analyses. Since the creation of the 1971
SLAR mosaic, several generations of space-borne imaging sensor systems have
been developed for which numerous applications have been tested. These sys-
tems have been shown to provide valuable information on both natural and
cultural features. Sensor systems are being evaluated by the authors to assess
their potential contribution to both detection of cultural activities and more
refined estimation of environmental characteristics.
A number of sensor systems may provide coverage of the study area. The
return frequency of space-borne platforms over large regions is much higher
than is possible using aircraft. This frequency of observation can allow re-
gional coverage by visible and infrared imagery despite the high probability of
cloud cover on any given overpass. Though smoke from fires in areas where
forest is being cleared can seriously degrade image quality in visible wave-
lengths, the only burning of jungle in this region is limited to the small gar-
dens cleared by the Yanomamo. If necessary, imaging systems could be used
that acquire data in those infrared wavelengths not greatly affected by smoke.
Imagery acquired in the visible and infrared wavelengths has much better geo-
metric fidelity than radar data and provides a more readily interpreted image
product. Because of the reduced significance of topographic variation relative
to platform altitude, space-based remote sensing systems are also much less
distorted by topographic relief than imagery from aircraft platforms.
Satellite image data for the area may be relatively accessible. In addition to
satellite data acquisition in the United States, the Institute de Pesquisas Espaciais
in Sao Jose dos Campos, Brazil, has a large archive of image data from its
ground receiving station for Multispectral Scanner (MSS) and Thematic Map-
per (TM) data from the Landsat satellites. Table 6.1 lists some spaceborne
sensors that may provide useful data of the area. Space-based imagery that has
good geometric fidelity might be used to refine boundaries originally delin-
eated on the SLAR mosaics. The GIS systems are currently capable of dis-
playing digitized map boundaries directly on digital image backdrops. This
facilitates the correction and ongoing maintenance of cartographic databases.
Boundaries from the MARNR-ORSTROM atlas could be overlain on imag-
ery from sensors with much higher spatial resolution, greater geometric fidel-
ity, and increased surface information. Human analysts could then validate or
correct each mapped boundary based on visual inspection. This approach would
allow the most effective correction of SLAR-based map products. However,
any additional effort to increase the precision of mapped features will encoun-
ter difficulties in dealing with the coarse scale at which original maps were
compiled. The internal heterogeneity within mapped units that is ignored at
1:250,000 scale may dominate images at higher spatial resolutions. Efforts to
capture this new detail in the GIS must be thorough and may require indi-
viduals with knowledge of the specific features and processes being mapped.
Such an effort to supplement the atlas maps with finer-scale detail will be
expensive and time-consuming. An effort that simply redefines initial bound-
aries based on higher-resolution imagery but that does not also deal with the
original internal heterogeneity of mapped features will be prone to error and
overconfidence in derived information products.
Data from space-borne imaging platforms may also support the described
anthropological study by providing a new source of information on both settle-
ment locations and environmental characteristics. Modern space-borne re-
mote sensing systems may provide substantially more detail than the SLAR
mosaics for the region. A number of human activities can be detected by space-
borne sensors. The clearing of forested areas associated with the establish-
Table 6.1 Sensor characteristics.Figure 6.2: The effect of resolution on the detection
of clearings.
Period of Spatial Overpass Type of
Sensor Acquisition Resolution Frequency Measurement
Landsat MSS (USA) 1972 - present 80m 16 days visible, infrared
Landsat TM (USA) 1982 - present 30m 16 days visible, infrared
AVHRR (USA) 1982 - present 1000m 0.5 days visible, infrared
SPOT-XS (FRANCE) 1986 - present 20m 26 days visible, infrared
SPOT-PAN (FRANCE) 1986 - present 10m 26 days panchromatic
MOS-1 QAPAN) 1991 - present SOm 17 days visible, infrared
SIR-A (USA) 1981 40m 2 images radar
SIR-B (USA) 1984 40m 2 images radar
ERS-1 (EC) 1991 - present 30m 3 days radar
ment of villages and gardens can be detected in the visible and near infrared
wavelengths by a reduction in the amount of absorbed photosynthetically ac-
tive solar radiation. Radar systems may detect such clearings by the reduction
in land surface roughness. The ability to detect these clearings will be a func-
tion of both the spatial resolving capacity of the sensor and the degree to which
vegetation has been removed. The ability to detect certain features will be
dependent on the date of image acquisition. For example, fully developed in-
tercropping systems will be more difficult to discriminate from forest canopy
than recent burns. Figure 6.2 simulates the potential measurement of a village
and garden clearing at various resolutions. The sizes of features are to scale
with known village characteristics. Based on this casual simulation, it seems
that Thematic Mapper imagery at a 3 0-m spatial resolution would be the coars-
est data that might be applied to the problem. Though AVHRR imagery from
the NOAATiros-N series of satellites has very coarse spatial resolution (1000
m), the location and frequency of fires associated with clearing of forested areas
are detectable in these data because of the large amount of energy released.
Remotely sensed data may provide further information on site quality that
relates to the hypothesis of relationships between social behavior and resource
availability. Image data may be classified either by manual or automated tech-
niques to indicate areas with similar environmental characteristics. This clas-
Figure 6.2 The effect of resolution on the detection of clearings.

sification of image data may provide maps that identify more or less attractive
environments for settlements. For example, it may be possible to delineate
areas that are poorly drained and subject to prolonged inundation by the char-
acteristic spectral response of species growing in such an environment. There
is also a large body of literature documenting the ability of sensors that mea-
sure wavelengths in the visible and near infrared to provide quantitative infor-
mation regarding the quantity and productivity of vegetation (Tucker 1979).
This information may be valuable in identifying regional productivity when
the resource potential of an area is being estimated. Such an approach would
be limited by the reduced sensitivity of sensor response to secondary and ter-
tiary vegetation canopies. A more direct application of this approach would be
to assess the relative vigor of existing garden plots in various regions.
Field observations suggest that one of the major discriminants of site qual-
ity is the difference between highland and lowland sites. Information to dis-
criminate between these two geomorphic classes will be based primarily on
elevation, slope, and hydrology. It is possible that a much more accurate and
precise map of elevations could be generated using satellite imagery. Sensors
made by the SPOT Image Corporation are pointable, allowing stereo imag-
ery to be acquired. Methods have been developed that may produce elevation
data comparable to 1:50,000 scale maps with 20-m contour intervals from this
imagery (Theodossiou and Dowman 1990). Automated methods have been
developed to extract hydrological parameters from such digital elevation data
(Band and Robinson 1986). Remotely sensed image data can have great utility
in the study of remote regions. Opportunities to apply GIS and remote sens-
ing techniques are continuing to evolve. Several new space-borne sensor sys-
tems are to be launched in the 1990s. Among these are Landsat Thematic
Mapper 6, which will have a 15-m spatial resolution panchromatic channel;
SPOT 4, with an added smoke-penetrating 20-m resolution infrared channel;
and SPOT NG, which is planned to have 5-m resolution and true stereo im-
aging. The upcoming Earth Observing System (EOS) being developed for the
earth science community may also assist in regional environmental character-
ization. At the time of this writing, policies are being discussed in the U.S.
Congress to make data of this type more accessible to the educational and
research communities.
Conclusion
The GIS-based analysis of the relationships between social process and envi-
ronmental characteristics for the Yanomamo poses a challenging situation.
Existing data, which are useful for the assessment of natural resources for the
region in general, are of very questionable quality for use in analyses at the
scale that affects tribal behavior. The existing information does have value,
but it would be necessary to correct and supplement these data in order to
utilize the analytical capabilities of GIS technology. Modern remote sensing
capabilities could provide the basis for such an effort.
Effective methods of data collection and analysis such as these must be de-
veloped in order to understand the interactions that will continue to acceler-
ate between the two radically different cultures of the region. This under-
standing will be of immense value in anticipating the impacts of acculturation,
implementing effective health programs for the Yanomamo, and developing
insight into the human condition.
7
A Time to Rend, A Time to Sew:
New Perspectives on Northern Anasazi
Sociopolitical Development
in Later Prehistory
Carla Van West and Timothy A. Kohler
Slightly before A.D. 1300, the Four Corners area of the North American South-
west was abandoned by prehistoric agriculturists. By that time, populations
had undergone three major cycles of aggregation into large settlements, first
constructing relatively large "public" facilities and then redispersing. The
reasons for the final abandonment of this area, as well as for the earlier col-
lapse of the Chacoan-related system of the mid-11 OOs, are classic areas of
archaeological inquiry. Recently, the earliest cycle of village formation and
dispersal, in the A.D. 800s, has come under increased scrutiny as well (Orcutt
et al. 1990; Wilshusen 1991).
In this paper we reexamine these phenomena by posing a simple but fun-
damental question: Under what conditions will farmers find it in their own
best interest to share the food they produce? Whatever the particular fea-
tures of these cycles of aggregation and dispersion, we suggest that periods of
increasing complexity in the fabric of sociopolitical organization—-which in-
volve the growth of settlements, elaboration of social roles and networks, and
heightened cooperation in building, hunting, and exchange—are constructed
on top of reliable systems of food sharing beyond that expected among close
kin. Such resource pooling has the effect of reducing the impact of variability
in agricultural production in an area where great unpredictability surrounds
the growing of food. Our thinking about how to approach these systems of
food sharing has been influenced by recent analyses of sharing among
hunter-foragers (e.g., Kaplan and Hill 1985; Smith 1988) and by current
discussions of risk and uncertainty in behavioral ecology and
microeconomics (Clark 1990; Stephens 1990).
Background
Biotic Setting
This study focuses on an area in southwestern Colorado about 3 5 km north of

the New Mexico border and immediately east of the Utah state line (Figure
7.1). Notable landmarks include the northward bend of the Dolores River on
the northeast, the escarpment of the Mesa Verde in the southeast, and the
commanding presence of a volcanic laccolith—Sleeping Ute Mountain—on
the south. In the lower portion of the study area, McElmo Creek collects wa-
ter from intermittent washes to the north that drain the mesa tops and broad
highlands in the upper portion of the study area. With the exception of Ute
Mountain (3011m) and an uplifted section known as the McElmo Dome (2182
m) in the south central section, the highest elevations in the study area are in the
northeast toward the San Juan Range of the Rocky Mountains, and the lowest
elevations (1500 m) are in the southwest along McElmo Creek at the Utah line.
Figure 7.1 The study area in southwest Colorado (edges marked by latitude/longitude
ticks) (after Van West 1994:2).
Northern Anasazi Sociopolitical Development 109
The study area is a cold, middle latitude, semiarid steppe, where potential
atmospheric evaporation exceeds the usual amounts of available precipitation
(Trewartha 1954). Most of the precipitation occurs as snow from December
through February; a smaller amount is deposited by patchy thunderstorms
between July and early September. The Dolores River is the only natural per-
manent surface water; other streams are seasonal. Groundwater is available as
seeps and springs at the contact between water-bearing sandstones and imper-
vious shales. These contacts outcrop in canyon head and canyon wall locations
throughout the study area.
Most of the study area supports the Upper Sonoran Biotic Community
(Lowe 1964), including the "Great Sage Plain" (Newberry 1876). Today, after
significant land clearance for modern agriculture, pinyon pine and juniper
woodlands remain mostly on the more shallow or rocky soils of the mesas and
highlands. In areas not cleared for fields, sagebrush and other deciduous shrubs
are found on deep, relatively level soils at middle elevations. Woodlands of
oak, Ponderosa pine, and occasionally Douglas fir thrive in the highest, wet-
test elevations, whereas the lowest, most arid elevations support Great Basin
Desert scrub grasses and shrubs.
The middle elevations of the study area are fertile and productive for dry-
land agriculture; here the growing season is long enough to mature a crop,
and soil depth and moisture are adequate to sustain plant growth. Today this
area is known for its dry-farmed pinto beans, alfalfa, and winter wheat and its
well preserved archaeological heritage. The Mesa Verde, with its famous Anasazi
ruins, rises just southeast of the study area. Recently it has been acknowledged
that the "beanfield country," of which the study area is a significant part, was
more densely settled and probably more typical of the Anasazi occupation of
the Mesa Verde region than even the Mesa Verde itself (Rohn 1989).
The research area is ideally suited for a variety of detailed archaeological
studies of these prehistoric farming peoples. It is situated in an area that fa-
vors preservation by virtue of aridity and modest impact by contemporary
populations. It has been the object of archaeological investigations for over
100 years, and the record of occupation is fairly well known (see reviews in
Brew 1946; Herold 1961; Nickens and Hull 1982; Eddy et al. 1984; and
Gleichman and Gleichman 1991). Because agriculture remains important to
the local economy, local weather station data, soil surveys, and documenta-
tion of historic crop yields—all essential for modeling prehistoric agricul-
tural productivity—are readily available.
Cultural Context
The Anasazi occupation of the study area began in approximately A.D. 500 and
persisted more or less continuously until almost A.D. 1300 (see Tables 7.1 and
7.2 for examples of population trends in various locations within the study
area). This occupation followed a much earlier but very sparse use, focused
first on big-game hunting, and later on (5500/6000 B.C. to 1 A.D.) foraging of
wild foods by Archaic populations. Elsewhere in the northern San Juan area,
Table 7.1 Population density estimates from surveys in the study area (using a 20-
year use-life estimate).
Dolores Archaeological Project a
600-720 720-800 800-840 840-880 880-920 920-980 980-1025 1025-1100 1100-1175 1175-1250
0.4 4.3 5.0 34.4 20.7 2.2 1.2 0.7 0.7 0.0
Mockingbird Mesa3
600-720 720-800 800-840 840-880 880-920 920-980 980-1025 1025-1100 1100-1175 1175-1250
0.0 7.0 0.5 2.0 0.0 2.0 28.5 5.0 15.5 53.0
Mockingbird Mesab
Basketmaker III early Pueblo I late Pueblo I early Pueblo II late Pueblo II early Pueblo III late Pueblo III
2.0 1.4 0.6 4.1 4.5 7.3 9.6
Upper Sand Canyon Survey0

930-980 980-1060 1060-1150 1150-1300a 1 1150-13001)
2.8 6.6 10.0 10.9 | 17.1
Lower Sand Canyon Surveyd

600-725 700-900 900-1100 1100-1300
1.3 0.0 0.8 7.0
Note 'Taken directly from Schlanger (1985:199).
Computed from Fetterman and Honeycutt (1987).
c
Computed from Adler (1992); earlier BMIII and PI occupation noted but not used in this study.
Computed from Gleichman and Gleichman (1989:36).
the first Anasazi populations, known as Basketmakers for one of their distinc-
tive crafts, were raising some maize and squash, hunting with atlatls, and using
dry rock shelters for storage, shelter, and burial as early as 1200 B.C. Subse-
quent Basketmaker (ca. A.D. 450 to 750) populations in and around the study
area regularly grew maize, beans, and squash; raised turkeys; built shallow
pithouses; made ceramic containers and bows and arrows; and sometimes lived
in loosely clustered but widely separated communities. Such people were the
first farmers in the study area. Their settlements are broadly dispersed across
a wide variety of elevations and topographic settings and usually occur near
good arable soils.
In the subsequent Pueblo I period (A.D. 750 to 900) Anasazi populations
built slab-outlined surface rooms of post-and-adobe construction and experi-
mented with simple masonry techniques, often building their suites of surface
rooms and deep pit structures in aggregated sites and settlement groups. The
distribution of Pueblo I communities is more restricted, with most known
local populations living in large villages in the northeastern portion of the
study area along the Dolores River. Explanations for this pronounced aggre-
gation have focused on the appropriateness of these higher elevations for suc-
cessful and predictable agriculture during a generally dry period (Petersen
1988: 123-124). Similarly, the breakup of the Pueblo I villages in the late 800s
or early 900s is often explained as movement to lower elevations in response
to the effect of climate change on agricultural productivity.
In this paper we focus on the Pueblo II (A.D. 900 to 1150) and Pueblo III
(A.D. 1150 to 1300) periods. In the early portions of the Pueblo II period, the
population lived in small, dispersed residences. By the beginning of the 1000s,
population size seems to be steadily increasing, either from in situ growth or
immigration, and clusters of apparently contemporaneous settlements appeared
Table 7.2 Population estimates from Sand Canyon Locality and Mockingbird Mesa.
Sand Canyon Locality 930-980 980-1060 1060-1150 1150-1300a 1150-1300b
Number ot components 20 75 96 107 109
Mean habitation size (n of rooms) 6 6 8 13 20
Mean momentary n of habitations 8 19 21 14 15
Mean n of people per habitation 9 9 12 19.5 30
Mean momentary population per yeair 72 169 256 278 436
2
Mean momentary pop density (km ) 2.8 6.6 10.0 10.9 17.1
Number of settlement clusters 0 3 4 3 3
Number of sites per cluster 0 13 (91 rooms) 15 (122 rooms) 2 1(2 7 5 rooms) 22 (275-n-)
Mean size per cluster (km2) 2.6 2.8 2.4 2.4
Mockingbird Mesa BMIII early PI late PI earlvPII latePII earlvPIII late PHI
Number of components 62 11 8 47 37 47 49
Mean habitation size (n of HH) 1.3 1.5 1.3 1.4 2.0 2.5 3.2
Mean momentary n of habitations 5.0 2.9 1.6 9.4 7.4 9.4 9.8
Mean n of people per habitation 6.5 7.7 6.3 7.0 9.8 12.5 15.8
Mean momentary population per yeair 32 23 10 66 72 118 155
Mean momentary pop density (km2) 2 1.4 0.6 4.1 4.5 7.3 9.6
Number of settlement clusters 6 1 1 4 7 6 6
Number of sites per cluster 5.2 6 5 5 4.1 6 5.8
Mean distance between sites (m) 239±137 620+520 600+300 2521175 357±189 344+205 323±236
Note: All values are rounded to the nearest tenth of whole number for the final table, not in the actual calculations.
All estimates based on a 20-year habitation use life.
Sand Canyon estimates assume a constant 1.5 people per room and a survey area of 25.5 km2. Estimates for 1150-1300a
are made without including the large towns of Sand Canyon Pueblo and Goodman Point. Estimates for 1150-1300b
do include these large towns (computed from Adler 1992, Adler and Varien 1991).
Mockingbird Mesa estimates (computed from Fetterman and Honeycutt 1987) assume a constant 5 people per household
and a survey area of 16.1 km2.
once again in the eleventh century. The Pueblo II period in general is notable
for new techniques of masonry construction, the appearance of multistory struc-
tures, the emergence of the architectural form known as the kiva from its pit-
house antecedents, the appearance of corrugated pottery, the development of
water-control features to impound potable water in reservoirs and capture soil
and slow runoff, and the florescence and demise of the "Chaco Regional Sys-
tem." Named for the great towns and associated ceremonial and integrative
architecture in Chaco Canyon some 165 km southeast of the study area, great
houses, roads, great kivas, and other Chacoan attributes appear in the study
area by at least A.D. 1045, when the first great house was constructed at Wallace
Ruin. In many cases, these local "Chacoan outliers" were established in exist-
ing communities and were apparently linked with each other or with Chaco
through some as yet poorly understood system of political, social, and eco-
nomic ties (Lekson 1991).
The subsequent Pueblo III period began in a time of environmental stress
and cultural reorganization following the collapse of the Chacoan system. The
distribution of local populations in the mid-1100s—a time of few cutting dates—
is not well understood. Likely it was a time of population mobility and disper-
sion. There was a renewal of building activity toward the end of the 1100s,
accompanying population growth at both the site and community levels. No-
table characteristics of Pueblo III, particularly of the post-12 00 period, in-
clude pecking and smoothing masonry into regular ashlar blocks, adoption of
a carbon-based paint to decorate ceramics, refinements in kiva construction,
addition of new architectural forms such as towers, and bi- and triwalled struc-
tures, enclosure of some habitation sites with walls that create interior plazas,
proliferation of water-control features, and movement from mesa tops to mesa
edge, canyon rim, and rock shelter locations close to springs. The 1200s wit-
nessed the construction of the impressive cliff dwellings in the canyons of
Mesa Verde National Park. Especially after about A.D. 1225, there was a strong
trend toward aggregation into a few very large towns throughout the region.
During a period of pronounced dryness in the 1270s and 1280s, however, all
of these towns were abandoned, and surviving populations emigrated south-
ward to form groups considered to be among the ancestors of modern Pueblo
Indians. Drought and other climate change resulting in failure to produce
adequate food have been used for many years as the prevalent explanation for
this apparently sudden, widespread depopulation of the study area and the
surrounding Northern San Juan Region (Douglass 1929; Euler et al. 1979;
Hack 1942; Haury 1934; among others).
Modeling Prehistoric Agricultural Productivity

The raw data used in this research are estimates of total maize yield for each
year from A.D. 901 to 13 00 within the study area. These values were developed
in an earlier study (Van West 1994; see also Van West 1991; Van West and
Lipe 1992) that reconstructed the effects of climatic fluctuation on potential
agricultural productivity in the 1816 km2 research area. One goal of that ear-
lier study was to examine whether the droughts of the late 1200s were suffi-
ciently severe and widespread to disrupt the agricultural economy of the Mesa
Verde Anasazi and precipitate the final abandonment of the region.
The approach taken was as follows (Figure 7.2): A study area was defined
by a set of 12 contiguous 7.5-min quadrangle sheets for which 7.5-min Digital
Elevation Models were also available. A grid of 200 x 200 m (4-ha) cells orga-
nized as 200 rows by 227 columns, registered to Universal Transverse Mercatur
(UTM) values, was superimposed on this space. Soils data provided by the
Soil Conservation Service were recorded for each of the 45,400 cells. Infor-
mation on soil depth, available water capacity, natural plant productivity, and
agricultural productivity was also recorded for the 98 soil types in the study
area. A total of 36,759 cells representing some 1470 km2 had complete soil
information. The 98 soils were reduced to eleven classes based on the quantity
of water held in the soil layers. Five weather stations at various elevations in
and around the study area, with sufficiently long records of monthly tempera-
ture and precipitation, were used to calculate Palmer Drought Severity Indi-
ces (PDSIs; Palmer 1965) for these eleven soil classes for the length of their
instrumented recording periods.1 (The differing elevations of these stations
allowed calculation of the effect of elevation on climatic variation for soils at
different elevations.) The PDSIs are temporally sensitive measures that inte-
grate the effects of temperature and precipitation on soil moisture and take
into consideration water already stored in or depleted from a soil profile. They
were calculated for every month of each year for the instrumented historic
period. Values for the month of June, which reconstruct soil moisture condi-
Data Preliminary Preliminary Final
Collection Analyses and Products Analyses and Final Modeling Products
Paleoenvironmenlal Data ^.PDSI Analysis_ Long-Term PDSI

(Tree-Ring Chronologies) Reconstructions
Elevational Data Elevational Data PDSI T Annual Annual Productivity Annual Long-Term
(DEMs) Plane (s) 1 Reconstruction • PDSI Reconstructions £ Population • Population
Data Plane Reconstructions Estimates Estimates
Soils Data
(Types, Locations,
Depth and Water-holding • Soils Data Plane(s) -
Capacity)
Soil Productivity Data Crop Yield and

(Agricurtural and Natural Natural Plant
Plant Productivity) Productivity
Studies ^
Agricultural
Productivity and
PDSI Calibration
Studies
Ethnographic Data
(Agrwultural Practices,
Consumptkm and Per Capita
Demand for Yield)
Figure 7.2 The conceptual model for reconstructing prehistoric agricultural

productivity. The GIS-managed portions of the study are enclosed by the rectangle
(after Van West 1994:4).
tions as they would have existed every year on July 1, were correlated with
tree-ring data for the same set of historic years. These tree-ring data, which
are used as a proxy for prehistoric climatic variability, were in the form of
eigenvector amplitudes, created from a principal components analysis per-
formed on seven expanded southwestern tree-ring chronologies, and repre-
sented the period from A.D. 901 to 1970 (Rose et al. 1982).2 This correlation of
instrumented PDSI values and tree-ring data produced a calibration, or trans-
fer function, that was applied to the full length of the tree-ring series, includ-
ing the pre-instrurnented period prior to 1912.
In this way, fifty-five long-term reconstructions of PDSI were produced,
one for each combination of elevational stratum and soil moisture group. These
were used to assign PDSI values to each 4-ha cell in the study area for each
year in the 1070-year tree-ring record and to reconstruct the state of farmland
at the driest point of the year. Thereafter, each reconstructed annual PDSI
value and its effect on a specific soil type was reexpressed in terms of potential
bean and maize yield, based on another series of regression analyses that de-
termined the historical relationship of these two crops to soil moisture condi-
tions. These productivity values were stored, analyzed, and displayed using
GIS technology. A mainframe raster system, VICAR-IBIS, and a PC raster
system, EPPL7, were used to process the data. VICAR was used to log in,
verify, correct, and mosaic the DEMs, whereas EPPL7 was used to store and
process all the subsequent data through the stage where annual production
maps were created (Figure 7.3). Maps displaying estimates of potential agri-
cultural productivity were created for each year in the A.D. 901 to 13 00 period.
The annual maps depict spatial variation in the potential agricultural land-
AGRICULTURAL
PRODUCTIVITY
A.D. 902
Figure 7.3 An example of a GIS-produced map (EPPL7) depicting agricultural

productivity in the 1816 km2 study area of southwestern Colorado for the year AD 902
(from Van West 1994:122).
scape. (These were later animated to simulate the changing patterns of pro-
ductivity through time.) The annual productivity estimates were used to cal-
culate how many people could be supported, given assumptions about which
soils of what level of productivity would be sought out, what levels of food
storage were desired, and what levels of demand were being made on the avail-
able maize supply. Any cell that produced a value under a predefined threshold
was not allowed to contribute to the total potential yield for that year, thereby
simulating Anasazi selection of only the better-yielding soils. The annual esti-
mates of total productivity used in the present study reflect this modification
to potential gross yield.
Three different estimates of the maximum population sustainable in any
year were calculated, depending on whether populations were assumed to be
seeking one, two, or three years of maize in storage after harvest. Carrying-
capacity estimates for various periods of time (the entire 400-year period;
archaeologically visible periods of say, 50 years; or occupation episodes as small
as 15 to 20 years) were also generated. These estimates were of three kinds
(Hassan 1981): a maximum carrying capacity reflecting the mean agricultural
production of each period; a critical carrying-capacity zone determined by the
worst year in a period; and an optimal carrying-capacity zone reflecting yields
of 20 to 60% of the mean for a given period.
Employing GIS technology, these data were further analyzed at three spa-
tial scales: first, for the entire study area, then at a smaller scale using two
intensively surveyed archaeological localities within the study area, and finally
at a yet smaller scale using the catchments surrounding eight tree-ring-dated
sites also within the study area. Each scale was appropriate to different ques-
tions. For the entire study area, it was found during the lowest years in the
400-year record, the maximum sustainable population annually storing the
equivalent of two years of maize was approximately 31,000 (21 people/km2).
Despite periods of dryness and significant spatial variation in yield, this esti-
mate suggests that a sizable population could have been supported at all times
within the 400-year period if either mobility or access to the productive places
in the study area were not restricted or if food distribution systems could sup-
port populations at some remove from productive locales. Otherwise, there
clearly would have been times when specific populations restricted to farming in
certain places would have been unable to meet their need for maize. This conclu-
sion is amply borne out by studies at the scale of localities within the region.
Two block survey areas, one on Mockingbird Mesa and the other in the
vicinity of Sand Canyon Pueblo (see Figure 7.1), provided population esti-
mates that were compared with estimates of theoretical population limits gen-
erated by the reconstruction. Data from the upland and very attractive Sand
Canyon locality suggest that populations there probably never exceeded what
the local fields could support, whereas in the nearby but lower area around
Mockingbird Mesa, population more than likely exceeded the productive lim-
its of the area at some times within the A.D. 980 to 1025 period and again
within the A.D. 1175 to 1250 period. Clearly there were places and times that
were more productive and predictable than others.
The site catchment studies suggested that Anasazi populations were aware
of the differential productivity of their environment and that they selected for
those locations that would reduce the risk of low yields. Site establishment
and growth, particularly for small residential hamlets, occurred in periods of
local high production. Conversely, site abandonment or cessation of building
occurred in periods of relatively high fluctuations in yield.
One important conclusion of this earlier study was that climatic factors
affecting agricultural productivity were probably not causing shortfalls in crop
yield at the scale of the 1816-km2 study area sufficient to force its abandon-
ment in the late 1200s. If prehistoric populations could have dispersed to fa-
vored locations in the late thirteenth century, there would have been sufficient
productive land to grow food for many thousands of people. Apparently other
limiting factors or social considerations came into play in the decision to aban-
don a region that had been home for at least 800 years.
Pooling Resources: Why and When?

We now reexamine these production questions using theoretical perspectives
provided by microeconomics (an equivalent analysis, in the language of be-
havioral ecology, could be made by interpreting our utilities as having fitness
consequences). In particular, we are interested in whether considering the
Anasazi to be risk-sensitive consumers can help us understand (1) why they
began to live in large aggregated settlements when they did, (2) why certain
cooperative behaviors such as the construction of public architecture emerged
when they did, and finally (3) why this area was abandoned in the late 1200s,
when absolute production levels were apparently adequate to support the
resident population.
What Does It Mean to Be "Risk-Sensitive"?
Consider the function graphed in Figure 7.4a. It represents one possible rela-
tionship between the yield or output of a foraging or production system as
measured in units such as bushels or kilocalories (on the x axis) and the value
or utility of that production to the producer (on the y axis).3 If the relationship
were a straight line of appropriate slope instead of a curve, then 20 bushels of
corn would be twice as valuable to a producer as 10 bushels. There is, how-
ever, reason to expect that this relationship is often nonlinear and in many
cases may be of the general sigmoid form graphed here (see reviews in Krebs
and Kacelnik 1991; Smith 1988). In such a relationship, the value of marginal
production accelerates more quickly than production itself when relatively
few units are being produced but more slowly when many units are being
produced. Thus, 20 bushels of corn may be either more—or less—than twice
as valuable as 10 bushels, depending on how the inflections on the utility func-
tion interact with these production levels on the x axis. Very small harvests,
especially if they are common, are scarcely more valuable than no harvest at
all, since they threaten the viability of the farming way of life. For this reason
the utility function is not a simple, concave-downward, decreasing, marginal
utility function. However, beyond a certain point, maize, like most other goods,
Figure 7.4a Good year economics. Units of production on the x axis; units of value to
the producer of a certain level of harvest on the y axis, a and b represent approximate
variance limits, either for different plots in the same year, or different years in a series
of generally good years. The average value realized after pooling is marked [v(x)]
and [v(a) + v(/3)]/2 represents the average value realized by not pooling. Following
Smith (1988) with minor changes in notation.
should respond to the law of diminishing marginal utility; therefore, in its
upper portions, our total utility function is concave downward.
We may observe that when the mean production falls within the diminish-
ing marginal value segment of the curve (as in Figure 7.4a), the mean value of
having chosen a "risky gain," simplified here as [v(a) + v(/3)]/2 is lower
than the mean value of having chosen a "risk-free gain" [v(x)] • (In this study,
we define "risk" as the random variation in the outcome of some decision
rather than the probability of falling below some acceptable threshold of in-
come.) Therefore, when agricultural supply was generally high relative to de-
mand, the Anasazi farmer would optimize the value of production by partici-
pating in a food-sharing system, here represented by the mean production
value [v(x)J, rather than acting independently and averaging the value of
randomly varying harvests, here represented by [v(a) + v(j3)]/2 .This is so
because in good years, the value of the group mean production (or the
"postpooled" mean) is greater than the value of the individually averaged pro-
duction (or the "nonpooled" mean), because v(x) > [v(a) = v(ft)]/ 2. In
other words, the household should avoid risk in a string of generally good
years by sharing food resources. (This is analogous to buying insurance and is
attractive because it trades cheap maize during good years for highly valued
maize in downside years.) Conversely, if the mean household production level
is low, as in Figure 7.4b, the Anasazi farmer would optimize the value of his
production by avoiding pooling systems and should accept the risk of not shar-
ing because the mean value of the variation in yield from either year-to-year
or field-to-field is higher than the value of the portion that he and his kin
would receive from the group under conditions of food sharing. This is be-
cause, as in Figure 7.4b, v(x) < [v(a) = v(f})]/2. Thus, in bad years, the
Figure 7.4b Bad year economics. In contrast to good years, in bad years [v(x)J (the
expected value of pooling) is less than [v((X) + v(/3)]/2 (the expected value of not
pooling). Notation as in Figure 4a.
household should be risk-seeking in its production and consumption strate-
gies. Here gamblers will do better than will insurers because the value of the
maize they will harvest in an uncommonly good year, when averaged with
their own previous and later production, is greater than the value of the maize
they could expect to receive through the depressed exchange network.
Two major dimensions to variability in agricultural production can be rec-
ognized. First, holding space constant, there is variability from year to year
due to climatic fluctuation (and other factors); second, within any year, there
is spatial variability in the productivity of various agricultural plots. A variety
of behaviors have been linked to attempts to avoid these risks. Three impor-
tant strategies are storage (to buffer year-to-year variability), pooling food
among producers,4 and dispersing fields in a variety of topographic situations
(the latter two buffer spatial variability within any year). Since stored food
may also be subject to pooling, and pooled resources may not be immediately
consumed, it might be expected that these analytically distinct responses to
temporal and spatial variability would in practice be correlated. Based on abun-
dant archaeological data for this time and area as well as on specific ethno-
graphic analogy, we assume that storage was always important and that it, too,
responded to production variability, but detailing that relationship is beyond
the scope of this paper. We do risk confusing food sharing with dispersion of
fields as responses to the same problem, especially since some of our measures
of food sharing (such as aggregation) automatically entrain dispersion of fields
on the level of the village (Kohler 1989).
In this paper we discuss behaviors connected with pooling (especially food
sharing, although pooling of labor and of information were probably impor-
tant to Anasazi adaptations and likely are entangled with food sharing). Al-
though the model we examine is simple, some of its predictions are not intu-
itively obvious. In periods characterized by relatively high mean production,
behaviors involving pooling of harvests ought to be attractive for the reasons
discussed above. If these same periods are also subject to relatively high year-
to-year fluctuation, the difference between the value of the mean postpooling
consumption rate v(x) ) and the value of the mean nonpooling consumption
rate [v(a) + v(fi)]/2 is accentuated (compare Figures 7.5a and b). There-
fore, it follows that periods with high mean production coupled with high
annual fluctuation in yield ought to be especially favorable for the develop-
ment of pooling behaviors. Finally, and for the same reasons, periods with
high mean production cooccurring with periods of high spatial variability in
yield should also tend to favor risk-averse (i.e., pooling) behaviors.
On the other hand, in periods when mean production is low, sharing ought
to be unattractive. Such periods would favor defection from any ongoing sys-
tem of sharing. This is because the value (on the y axis) of the mean risky (i.e.,
nonpooling) consumption is higher than the value of the mean risk-free (i.e.,
postpooling) consumption. [Sebastian (1991: 111) reaches a similar conclu-
sion by reference to the ethnographic record.] Moreover—and this point seems
to go against traditional archaeological intuition—high temporal variability
or high spatial variability in periods of low mean production will exaggerate
Figure 7.5a A series of years with a relatively high mean production and relatively
high variance results in relatively large differences between the expected value of
sharing versus not sharing. Notation as in Figure 7.4. In good years, increasing
production variance (through time or space) increases the value of pooling when
average production is held constant.
the difference between the values of the mean risky and the mean risk-free
consumption rates. In periods of low mean production, then, the relative value
of "defection" from a system of sharing [to use Axelrod's (1984) terminology]
is greatest when temporal or spatial variability is highest [compare Figures
7.6a and b; see Hegmon (1989: 93) for a related point, expressed in the cur-
rency of risk reduction rather than utility maximization].
Figure 7.5b A series of years with the same mean but relatively low variance results
in less difference between the expected value of sharing versus not sharing. Notation
as in Figure 7.4.
Figure 7.6a A series of years with relatively low mean production and relatively high
variance results in relatively large differences between the expected value os sharing
versus not sharing. Notation as in Figure 7.4. In bad years, increasing production
variance (through time or space) increases the value of not pooling when average
production is held constant.
Summary of the Model
Pooling of food—perhaps the primitive foundation of cooperative behav-

ior in general—is a form of risk-averse (variance-reducing) behavior. We ex-
pect systems favoring the pooling of food to develop in periods characterized
by high mean productivity, high fluctuation in productivity from year to year,
and great differences in productivity across space. We further expect that dur-
ing periods characterized by low mean productivity but high temporal or spa-
tial variability, pooling would not be in the best interests of the producers and
Figure 7.6b A series of years with the same mean but relatively low variance results
in less difference between the expected value of sharing versus not sharing. Notation
as in Figure 7.4.
•
is expected to break down, if present, or to not develop. Finally, periods with

intermediate yields are neutral in terms of the model; that is, no clear expecta-
tions can be formed as to whether insuring or gambling would be more desir-
able. In the next two sections, we discuss how we define periods that meet
these conditions and how we attempt to recognize "pooling behavior" (and its
demise) in the archaeological record.
Operational Considerations
Identifying Periods of Interest
We gave priority to high mean production in identifying periods in which

pooling should develop. The shape of the utility function suggests that the
expected value of pooling will exceed the expected value of not pooling only
when the mean production is relatively high. Although we can never be sure of
the exact shape of the utility function, if the general shape assumed is correct,
the expected value of pooling will increase along with the mean production. It
was necessary to identify periods with given characteristics that were as long
as possible. There are two reasons for this, one of which is strictly operational:
in order to have some chance of being able to recognize periods in the ar-
chaeological record, they have to be relatively long. The other is that decision
makers must have some idea of the relative payoff for pooling versus defection
in order to be able to make decisions on that basis. This might be impossible
on the basis of just a few years with given conditions.
To identify periods with relatively high mean production, we first smoothed
the 400-year sequence of total maize production for the study area using a
4253H filter (Velleman and Hoaglin 1981) [see Van West (1994: 133) for a
histogram of the unsmoothed estimates]. This smoothed series of production
values is displayed in Figure 7.7. A look at this series allowed us to identify five
Figure 7.7 Smoothed annual estimates of total maize productivity in kilograms. Periods
identifiable as favorable or unfavorable for pooling are identified by bars above and
below the series, respectively.
relatively long periods (from 24 to 50 years in length) of relatively high aver-
age production, identified by solid bars above the graph. The same technique
was used to identify five periods of relatively stable low means (identified by
bars below the graph). We allowed these periods—expected to be unfavorable
for pooling—to be somewhat shorter (they ranged in length from 10 to 50
years) than the periods expected to be favorable for pooling. In part, this was
out of simple necessity, since many of the unfavorable periods were rather
short; we also assumed that it might take less time to destroy a cooperative
system than to erect it. This left 117 years (A.D. 901 to 1005 and 1289 to 1300)
that we could not characterize as relatively stable periods of either high or low
means and that we considered to be neutral in terms of our model.
The next step was to compute measures of the temporal and spatial varia-
tion within each period. Under the model, pooling will be most attractive in
periods with relatively high means that also exhibit high temporal and spatial
variability in production. We first measured temporal variability as the stan-
dard deviation of the annual production for each period, but this measure was
positively correlated with the mean annual production for each period (and
the coefficient of variation was negatively correlated with mean annual pro-
duction). We therefore regressed the standard deviations around the means
on the production means for each period and worked with the residuals, which
were free from any linear association with the mean (Figure 7.8). The same
Figure 7.8 The relationship between mean maize productivity (in kg) and the standard
deviations (in kg) around those means for the ten periods identified in Figure 7.7 and
Table 7.3. The residuals from this regression are used as the measure of relative
temporal variability in Table 7.3.
operation was necessary to construct a measure of spatial variability within
each year that is unassociated with the total annual regional productivity, since
there is a strong tendency to have high variability in years of high production
(Figure 7.9). These annual residuals, a measure of relative spatial variation,
were averaged within each defined period to form a measure of average rela-
tive spatial variability in each period.
For each period, the mean production and these relative measures of temporal
and spatial variation are displayed in Table 7.3. In the first column of this table,
the periods are ranked according to our best estimate of the overall attractiveness
of pooling (in the case of periods a to e) or the overall attractiveness of "defec-
tion" (for periods g to k). For the periods in which we expect pooling to develop,
this ranking is achieved by first ranking the scores in columns 2, 4, and 5 from
high to low (with a rank of 1 assigned to high positive scores). Then each period
is assigned an overall rank by taking the median of these three ranks. Thus, the
ranks for period d are 2, 2, arid 3.5, yielding a median of 2, the highest rank for
any period. We used ranks, rather than standardizing these three values as z scores
and taking their mean, because of some disjunction between what we would like
to measure (the achieved production per household, given some particular dis-
tribution of population and fields in any given year) and what we are actually
measuring (an estimate of potential production across the entire landscape).
The periods in which the expected value of defection is greater than that of
Figure 7.9 The relationship between the total maize productivity for the study area
(in kg) and the annual standard deviation around the per-hectare mean productivity (in
kg) for each year in the 400-year sequence. The residuals from this regression are used
as the measure of relative spatial variability in Table 7.3.
sharing were ranked according to the same logic. Sharers are most disadvan-
taged when production is low (the lowest is assigned a rank of 5 in Table 7.3,
column 2), and relative temporal variation (the highest is assigned a rank of 5
in column 4 of the same table) and relative spatial variation (the highest is
assigned a rank of 5 in column 5) are high. The median of these three ranks,
then, could be used to assign to each period an overall attractiveness for shar-
ing behavior. For example, period k with ranks of 5, 3, and 5 receives a median
rank of 5, identifying it as the least favorable period for pooling in the 400-
year record. These are the years between A.D. 1272 and 1288.
Identifying Pooling Behavior in the Archaeological Record
To examine the predictions of this model we must be able to monitor the

growth and demise of cooperative behaviors in the northern Anasazi South-
west, with emphasis on those that might involve food sharing. The vast litera-
ture on integration, reciprocity, redistribution, political complexity, and ag-
gregation is all germane in attempting to identify what these behaviors might
be. Our tactic in choosing measures that ought to be involved with food shar-
ing was to choose as many indices as possible in the full realization that none
Table 7.3 Periods with differential advantages for pooling, ordered by median
rank for mean total maize productivity, relative temporal variation, and mean
relative spatial variation.
Standard Dev.
Mean Annual for Annual Relative
Maize Maize Temporal Mean Relative
Productivity Productivity Variation Spatial Variation
Period* in Years during period during period during period during period0
A.D. (median rank) (x 100) (rank) (x 100) (rank) (rank) Value of Pooling
[d] 1118-1211(2) 70,797 (2) 14,893 225 (2) 1.33(3.5) strongly positive
[a] 1006-1029(3) 68,118(3) 13,693 -249 (3) 2.79(2) 1
[e] 1100-1129(3.5) 71,356(1) 13,702 -1120(4) 1.33(3.5) 1
[h] 1049-1088 (4) 68,109(4) 16,000 2071 (1) -1.99(5) 1
[e] 1222-1271 (5) 66,418 (5) 11,772 -1692 (5) 2.80(1) weakly ^positive
[f] all 117 years not
included in a 64,666 14,284 1301 -0.53 (6) approx. neutral
favorable or
unfavorable period
[h] 1089-1099(1) 59,607 (1) 10,961 -631(2) -5.42 (1) weakly negative
[i] 1130-1179(2) 59,433 (2) 12,130 585 (4) -2.40 (2) 1
[gl 1030-1048 (3) 59,046 (3) 10,687 -751(1) -1.81 (3) 1
[j] 12 12-122 1(4) 58,333 (4) 11,866 623 (5) -1.59(4) 1
[k] 1272-1288 (5) 58,033 (5) 10,797 -363 (3) 2.82 (5) strongly negative
901-1300 64,925 13,937 n/a 0.00
a
the letters in brackets front of each period identify the periods in Figures 7.7 and 7.8,
t> these are the residuals from the model [Standard Deviation around total during period] = -4792 + 0.27 x
[Mean Annual Productivity during period]. High positive residuals identify periods with more temporal
variability than would be expected, based on the average relationship between the mean productivity
and the standard deviations for each period.
c
these are the averages, for each period, of the residuals from the model [Standard Deviation around
per-pixel mean productivity for each year in the 400-year sequence] = 167 + 0.0000019 x [Total
Productivity for each year in the 400-year sequence]. High positive residuals identify periods with more
spatial variability than would be expected, based on the average relationship between the mean productiv
and the standard deviations for each year in the 400-year sequence.
may be a pure measure of the dimension. We consider all to be at least weakly
involved with the dimension of interest.
The measures selected are listed in the first column of Table 7.4. More discus-
sion as to how these facets of the archaeological record might be involved with
food sharing can be found for aggregation at the site level in, for example, Glassow
(1977: 206); for growth and aggregation at the community level in Orcutt et al.
(1990) and Sebastian (1991); for great kivas and triwall structures in Plog (1974:
127); for reservoirs in Haase (1985); and so forth. Ford (1972) provides a general
perspective on the importance of the movement of food in contemporary Tewa
ritual and society. Other measures of increased interaction, such as higher
intraregional rates of exchange of regionally produced ceramic and lithic materi-
als, should also be expected in periods in which regional production potentials
favored development of pooling. Unfortunately, we know of no studies that de-
scribe the volumes of any intraregional flows of materials with enough temporal
precision to be useful in testing the present model.
Table 7.4 lists only one "positive" test implication for the periods in which
sharing is expected to break down—the dissolution of aggregated sites. Of course,
we also expect no evidence of behaviors connected with sharing in these periods.
The Model Meets the Evidence

The temporal precision of our expectations exceeds the temporal precision of
a good deal of the archaeological record. For this reason the tree-ring-dated
sites in the study area (Table 7.5) are especially valuable for testing the expec-
tations. In Table 7.6 we add to the group of tree-ring-dated sites another group
Table 7.4 Initial test of the pooling model.

High Positive < Predicted Strength of Pooling Advantage > High Negativ
1180- 1006- 1100- 1049- 1222- 1089- 1130- 1030- 1212- 1272-
Neutral
Manifestation of Pooling Behavior 1211 1029 1129 1088 1271 1099 1179 1048 1221 1288
Expectations for Periods in which
Pooling is Expected:
Growth/aggregation at site level
(e.g., more rooms indicate more + + ++ •f +++ p 3 + ? 3
-
residents)
Growth/aggregation at community
level (e.g., more sites indicate more + + ++ + ++ 3 3 3 ? 3
-
members)
Great Kivas 2 ?
?
7
?
1 5 3 ? ? ? ? 0?
Reservoirs + + ++ 3 ? 4- ? ? -?
Great Houses 3 3 10 1 3 ? 3 -f ? ? 0
Roads 1 0 S ? 1 3 ? + ? ? 0
Enclosing Walls/Interior Plazas 3 0 1 0 6 3 3 3 ? + 0?
Triwalled and Biwalled Structures 0 0 1? 0 5 3 ? ? ? ? 0?
Foundation of hamlets associated
with the establishment/growth of + ++ ++ + +++ ? ? - ? 3 -
aggregates
Expectations for Periods in which
Defection is Expected:
Breakup of aggregates 3 3 3 3 3 +a 3 •f 3 3 +
a
This refers to the breakup of the mid-9th-century Pueblo I villages (as in the Dolores Archaeological Project Area) which
may take place slightly before A.D. 900. The "neutral" period includes all of the tenth century (to A.D. 1005) and the
years from 1289-1300 that follow the local collapse of the great towns.
Table 7.5 List of tree-ring dated sites in the study area.
Site Name Site No. Reference Dates (A.D.) '
DCA Site 5MT8371 Dykeman 1986 935-950
Norton House 5MT8839 Fuller 1987; Kuckelman and Morris 1029-1048
10RS
lyoo
Aulston Pueblo 5MT2433 Kane 1975; Morris 1986 1030-1050
Wallace Ruin/Ida Jean SMT6970; Bradley 1974, 1984, 1988a; 1045-1124
Ruin
Ruin 5MT4126 Brisbin and Brisbin 1973
Lowry Pueblo SMT1566 Martin 1936; White and Breternitz 1086-1120
1976
Escalante Ruin 5MT2149 Hallasi 1979 1124-1138
Mustoe Site 5MT3834 Gould 1982 1173-1231
Knobby Knee Stockade 5MT2S25 Morris 1991 1193-1201+
Lillian's Site 5MT3936 Varien 1990 1211-1214+
Roy's Site 5MT3930 Varien 1990 1213-1223+
Saddlehorn 5MT262 Kleidon 1991 1237-1256+
Sand Canyon Pueblo 5MT765 Adams 1985, 1986; Bradley 1986,
1987, 1988b, 1992; Kleidon and
Bradley 1989
Castle Rock Pueblo 5MT1825 Kleidon and Lightfoot 1991 1252-1277+
Troy's Tower 5MT1825 Varien 1990, 1991 1265-1271+
Lester's Site 5MT10246 Kuckelman 1991 1270-1271+
Note: tree-ring dates are given as a range where the first date is a cutting date believed to
represent a construction episode and the final date is the latest date from the structure
or site, even if it is a non-cutting date.
of sites for which probable peaks of occupation can reasonably be derived from
ceramic materials and tabulate those items of public architecture from this
larger set of sites that we wish to use as indices of increased sharing of re-
sources. Finally, in Table 7.4, these data are tabulated against the periods iden-
tified under our model (in Table 7.3) as either rewarding cooperative food
sharing or rewarding defection.
Results
The general pattern of the record is strongly in the directions anticipated by

the model. The period between 1100 and 1129, in which the expected value
of cooperative behaviors is high, coincides with the local manifestation of the
Chacoan system. The "terminal" aggregation in the mid-1200s at canyon
head sites such as Sand Canyon is also correctly predicted. The breakup of
the Chacoan system between 1130 and 1179 and the final abandonment of
the region in the 1270s or 1280s are likewise found in those periods in which
we predict defection.
Some of the apparent weaknesses of the model may reflect weaknesses in
our ability to precisely date the archaeological record. We suggest that many
of the "+" signs in the 1130 to 1179 period in fact pertain to sites actually
belonging to the immediately preceding periods of more favorable condi-
tions. Three of the periods in which we predict defection (1030 to 1048,
Table 7.6 Public architecture in the study area.

Great Great Bi- or tri- Enclosing
Site Date3 Houses Roads Reservoirs' Kivas walls walls/plazas
Wallace 1045-1125(1125) +b + +
1224-1275 (?)
Reservoir 1050-1 125 (?) +
Casa Negra 1060-1150(1125) + + +
SMT4700 1075-1125(1100) ?
Yellowjacket 1075-1150(1125) + + <- +
1175-1250(1225)
Yucca House 1080-1150(1125) + +
1225-1275 (1250)
Ansell Hall 1080-1150(1125) + +
Lowry 1080-1150(1125) + t +
Haynie 1100-1135(1125) +
Bass 1100-1150(1125) +
Albert Porter 1100-1150(1125) + ?
Emerson /Sun dial 1100-1150(1125) +
5MT10581 PII +
Ida Jean 1125-1150(1125) +
Pigge 1175-1225(1200) +
5MT10581 1175-1225(1200) +
N. Lowry 1100-1250(1225) +
Mustoe/Gd. Ft. ComniPIII +
Bear tooth PHI +
SMT751 PHI +
5MT726 PHI +
Mud Springs 1200-1250(1225) + +
Goodman Point 1200-1250 (1240) + +
Mitchell Springs 1200-1275 (1225) + 4-
5MT4421 1225-1275 (1250) +
Hibbets PHI +
Little Cow 1225-1275(12(50) +
Morley Kidder 1225-1275 (1260) +
Easter 1250-1275 (1260) ? +
Castle Rock 1250-1275 (1260) +
Cotton wood 1250-1280(1260) ?
Sand Canyon 1252-1277(1260) + + + +
Woods Mesa 1250-1280(1260) +
Pedro Point 1250-1300(1260) +
a
Range of years, e.g., 1100-1150, indicates core occupation range for the major occupation. Date in
parentheses (e.g., 1125) indicates estimated peak date of occupation for the major component of the site.
b
Here and elsewhere, for multiple-component sites, no attempt is implied to apportion features to components,
f From Haase (1985), otherwise all from Varien et al. (1990), as revised by Varien 1991, personal communication.
1089 to 1099, and 1212 to 1221) are simply too short to identify with
confidence in the record.
However, with the exception of our identification of 1272 to 1288 as the
period most likely to result in defection, the relative strength of the predicted
effect is not strongly correlated with the strength of the effect actually seen in
the archaeological record. For example, the behaviors we consider associated
with pooling appear more strongly in the 1222 to 1271 period than they do
between 1180 and 1211, when the model predicted the effect to be strongest.
Likewise, the effect for the 1100 to 1129 period, predicted by the model to be
only moderate, appears to be very strong. It is tempting to consider these as
random failings due to measurement error, but, in fact, we suspect that these
deviations are systematic and subject to further explanation.
Discussion
The model we have constructed and tested predicts how an individual produc-
ing and consuming hotisehold ought to behave in order to best utilize produc-
tion. This is a simple model, and it is not surprising that it fails to explain every
facet of the record; if anything, it is surprising that it explains so much. Never-
theless, its failings highlight important factors that must be built into future
versions of the model. We now discuss the probable effects of some of the
most important complications of reality not accounted for in the model.
Perhaps the most important shortcoming is the model's omission of the
direct or indirect effects of population size. Population is not constant in size
or in spatial distribution throughout the period (see Tables 7.1 and 7.2). The
probable (smoothed) trajectory is for slow growth, starting from a small base
in the 900s, with the rate of increase accelerating in the early 1100s and then
decelerating in the mid-HOOs. Finally, in the late 1100s or early 1200s,
population may again increase rapidly until shortly before the final aban-
donment of the region.
Inspection of Table 7.4 suggests that the simple model examined here un-
derestimates the attractiveness of defection when population levels are low
[Kohler and Van West (1992) discuss this aspect of the analysis in more detail].
For example, it may be that the "neutral years"—which include all the A.D.
900s—are essentially times of defection. Further, we may also have underesti-
mated the attractiveness of sharing when population is high. Significantly, the
effects of large sedentary populations on the biotic landscape are now begin-
ning to be understood (Floyd and Kohler 1990; Kohler 1992b; Kohler and
Matthews 1988; Speth and Scott 1989). Depletion of wild-food in conjunc-
tion with high population levels increased Anasazi commitment to and depen-
dence upon agriculture. Under such conditions the landscape begins to fill up
with marks of ownership (Adler 1994; Kohler 1992a), particularly where agri-
culture is most reliable and productive. In periods of both low population and
low production, defection into an open landscape retaining abundant wild re-
sources is more attractive than it would be in periods of low production coupled
with high population. We suggest, then, that population levels changed the
shape of the utility function to make sharing of crops more attractive when
population is high and less attractive when population is low. The dilemma of
the 1270s and 1280s was that the biotic and social characteristics of the land-
scape necessitated a sharing adaptation at the same time as climatic character-
istics were making such cooperative behaviors increasingly unattractive. The
abandonment of the study area may be due more to the impossibility of dis-
persing—and of making a living as a farmer in an isolated farmstead—in the
landscape that had been created over the last several hundred years than it was
to the absolute production levels that could have been achieved.
Our simple model predicts that cooperative behaviors such as aggregation
can develop in the absence of any increase in population and at any population
level. This prediction may accommodate the circumstances of aggregation in
the Zuni and Taos areas, where some observers (Crown et al. 1990; Kintigh
1985) claim that village formation may have taken place in the absence of
significant population growth. If our interpretation of the weaknesses of the
simple model is correct, however, the productive characteristics that favor
pooling would have to be correspondingly more exaggerated to have the same
effect in a sparsely populated region as in a densely populated region (see Plog
et al. 1988). Thus, we surmise that population history and population size are
key contexts within which the utility model must operate.
A second important area in which this model is almost certainly too simple
is in its ignorance of extraregional productive characteristics. Presumably at
least part of the population history of our study area can be explained by its
attractiveness relative to other regions, in the same was as some population
movements within the study area have been explained (Schlanger 1985). If so,
this will affect the shape of the utility function within the study area and in
turn change the value of the productive outcome (or, in game-theoretic terms,
the nature of the payoff matrix for cooperating versus defecting).
Finally, to segregate some portion of the Southwest from the whole for
analysis is to assume that developments in each subregion are independent.
Clearly our study area is influenced by the development of the Chacoan sys-
tem outside of its borders and probably by other external events and systems
that are not so obvious. Our "underprediction" of the value of pooling for the
Chacoan period may be due in part to this effect. This study area is one of
convenience, primarily justified on the basis of the high-quality data that makes
it possible to formulate and test the present model in a preliminary fashion.
Summary and Conclusions

In this paper we define a method for predicting the development of coopera-
tive behaviors among small-scale horticulturists. The model behind the method
focuses on sharing of food. Foodsharing is not only one of the most funda-
mental forms of human cooperation but is symbolically implicated in more
elaborate cooperative endeavors and may form the foundation for their elabo-
ration. Our goal in this paper has not been to define the links between food-
sharing and the various cooperative behaviors that leave more visible residues
in the archaeological record. Nevertheless, examination of the performance of
the model suggests that its utility goes well beyond the prediction of foodsharing
to the prediction of a large set of cooperative behaviors.
The model also has important implications for how we think about spa-
tial and temporal variability in food production. Within the last fifteen
years, archaeologists have acknowledged the importance of production
risk in the formation of tribal social networks and in enlarging the scope
and scale of sociopolitical organizations in general (Johnson and Earle
1987). In one influential analysis (several others of similar emphasis could
be cited), it was expected that
sustained increases in the intensity of integration will occur asconcomitants
of sustained increases in region-wide uncertainty or risks arising from envi-
ronmental change. The more sustained the environmental change, the more
permanent will be the effect on societal organization (Braun and Plog 1982:
508; emphasis in original).
The model put forward here, on the contrary, predicts that the effect of
variance (risk) in production on cooperative systems involving food sharing
will depend on whether these risks are situated in a context of relatively high
or relatively low production. Cooperative behaviors are most valuable in the
context of high production cooccurring with high temporal and spatial vari-
ability. Such behaviors are least valuable in circumstances of low production
cooccurring with high spatial and temporal variability. Of course, our method
requires knowing something about the shape of the utility function before any
predictions can be made about the relative value of cooperation versus defec-
tion. We have assumed that years of production substantially below the mean
(or customary level) are low enough so that they will be located in an increas-
ing marginal value segment of the curve, and that years well above the mean
are within a decreasing marginal value segment of the curve. These assump-
tions require continued scrutiny. Provisionally, the success of the predictions
they allow tends to confirm their general accuracy.
Finally, we consider it an advantage that this method focuses on decisions
at the level of the individual or the household. This "methodological indi-
vidualism," to use by Smith's (1988) term, allows us to go beyond group-level
functionalism to consider the motives of the actors at an effective level. This
level of analysis allows us to predict, for example, that when a town such as
Sand Canyon Pueblo was abandoned in the late 1270s, it was not as part of a
group-level decision to move elsewhere together, but rather it was a matter of
individual households, or slightly larger corporate groups, going their sepa-
rate ways, leaving behind a system in which continued participation was not
in their own best interests.
Acknowledgments
We would like to thank Mark Varien and Angela Schwab of Crow Canyon Archaeo-
logical Center for access to unpublished data and manuscripts, Larry Hammack and
Jim Kleidon for recent tree-ring data from Knobby Knee Stockade and Sand Canyon
Pueblo, and William D. Lipe of Washington State University and Crow Canyon for
discussion of some aspects of this paper in its early stages. Ian Thompson provided
thoughtful editorial assistance. Useful comments were also received from Linda Cordell,
Michelle Hegmon, Keith Kintigh, and Eric Alden Smith. The Wenner-Gren Founda-
tion for Anthropological Research through grant 4799 helped support the research
that produced the annual estimates of prehistoric agricultural productivity on which
the present analysis is based.
Notes
1. These stations were Bluff, Utah; and Cortez, Ignacio, Mesa Verde and Ft. Lewis,
Colorado; their records began as early as 1912 and continued annually through
1983. They ranged in elevation from a low of 1315 m at Bluff to a high of 2317 m
at Ft. Lewis.
2. These chronologies were derived from the vicinity of Santa Fe, the Jemez Moun-
tains, Gobernador, Cibola, Chama Valley, and Cebolleta Mesa, New Mexico; and
Mesa Verde, Colorado.
3. This and the following curves are based on theory of consumer behavior. We have
graphed the total utilities rather than the marginal utilities. See Awh (1976: 48-66), or
any other introduction to microeconomics, for a general discussion of such curves.
4. We use the terms "pooling" and "sharing" as general synonyms to encompass both
the restricted sharing and the more general pooling of resources distinguished by
Hegmon (1989), since it is not our goal here to address the questions of what kinds of
sharing may be taking place through what mechanisms. The number of sharing house-
holds (approximating the number of producers in the present case) necessary to achieve
dramatic reductions in income variance is related to the degree of correlation in pro-
duction among households and may in practice be fairly small (Hegmon 1989;
Winterhalder 1986).
8
Moving from Catchments to Cognition:
Tentative Steps Toward a Larger
Archaeological Context for GIS
Vincent Gaffney, Zoran Stancic, and Helen Watson
Geographical informations systems are being used increasingly frequently

within archaeological applications. Given the nature of much archaeological
data, there can be little doubt that this technology probably represents one of
the most flexible and comprehensible tools for the analysis of spatial data
presently available. However, there are causes for concern relating to the ar-
chaeological context of GIS. This paper suggests that the nature of most GIS
is such that they are most readily applied to data that are most conveniently
stored in map format and that this may ultimately be restrictive to the natural
development of archaeological analysis. In particular it is suggested that the
use of GIS modules may lead to the unwitting exposition of an environmen-
tally or functionally deterministic viewpoint of a type that has largely been
rejected by most archaeologists. The need to develop cognitive models is
emphasized and it is suggested that GIS has an important role to play in the
development of such approaches. Particular emphasis is placed on the abil-
ity of GIS to incorporate the whole environment within archaeological
models and to transform abstract spatial information in order to place it
within a cultural domain.
Two case studies are presented to support these suggestions. The first in-
volves the re-analysis of a GIS study of late prehistoric settlement and burial
data on the island of Hvar by the authors. It is suggested that the original
interpretation of these data can be greatly improved through a more thought-
ful consideration of the belief systems operating within these communities.
The second case study involves prehistoric rock art and other ritual monu-
ments in mid-Argyll in southern Scotland. The GIS-generated viewshed data
are used to explore the cognitive context of the monuments within the land-
scape and to explore the perceived relationship between monuments.
The GIS clearly has a lot to offer archaeology. However, there is a need to
ensure that we use the technology on the terms of archaeology rather than
simply transfer the techniques for which GIS is most commonly used into an
archaeological context.
A Larger Archaeological Context for GIS 133
Tools and Theories: The Archaeological Context of GIS
The ability of GIS to manipulate and analyze map data is proving to be in-
creasingly attractive to archaeologists (Kvamme 1989). The allure of the tech-
nique is not hard to fathom. Archaeological data are often most conveniently
stored in map format and the frequent recourse to mapped displays to explain
and interpret archaeological material makes GIS one of the more flexible and
comprehensible analytical tools available to the archaeologist. Moreover, as
more archaeologists use GIS they begin to understand that the technique al-
lows us to handle vast amotmts of complex data in a manner that was previ-
ously unimaginable. However, in establishing such a profitable relationship,
there are certain dangers for the unwary. At the simplest level, we should note
the alluring visual products of GIS and suggest that there is a danger that
archaeologists may be distracted by such outputs to the extent that esthetics
may dominate interpretation. The ultimate aim of research or the quality of
the data for such analysis may be forgotten. One suspects that most practitio-
ners have been guilty at some point of the surreptitious admiration of GIS
imagery for its own sake (Gaffney and Stancic 1991b). These are, however, the
pitfalls of a novel technique. We can be sure that familiarity will breed a healthy
degree of complacency over time.
More significant than this is the manner in which the most attractive quali-
ties of GIS technology may mold archaeological thought and practice in a less
than desirable manner. Here there is more cause for concern. If the strength
of GIS lies in its ability to analyze mapped data, the archaeological potential of
GIS is related to the types of mapped data available or thought suitable for
such purposes. Topographic, environmental, and palaeo-environmental data
have to date proved particularly amenable to archaeological GIS analysis. Such
variables can be conveniently mapped and measured and have been assessed
against archaeological data to provide information pertinent to a variety of
problems or models (Allen et al. 1990). However, the aptitude of GIS to ana-
lyze such data and the relative sophistication of the results do not legitimate
such pursuits archaeologically. Indeed, there are good reasons to suggest that
the application of GIS techniques in such a way could ultimately prove to be
restrictive to the general development of archaeological thought. In its least
harmful form, the indiscriminate use of GIS solely in conjunction with mapped
physical data may result in the slick but repetitious confirmation of otherwise
obvious relationships. In the worst case it might involve the unwitting exposi-
tion of an environmentally or functionally determinist analytical viewpoint of
a type that has largely been rejected by the archaeological community. There
is a need, therefore, to explore the archaeological context for GIS applications.
Initial inspection suggests that the context of GIS applications within ar-
chaeological research has, on the whole, been related to the physical analysis
of landscape settlement traces. This is a well-established analytical approach
(Clarke 1968). Following trends largely instigated within geography, archae-
ology has held a consensus since the 1960s that the archaeological landscape is
a structured phenomenon. The spatial structure within archaeological depos-
its or landscapes was held as interpretable, and the quantitative revolution

within archaeology during the 1970s was largely built upon this credo. How-
ever, the quantifiable data central to such analysis during the 1960s and 1970s
largely resulted in an emphasis on abstract pattern location and economic analy-
sis. The works of the Cambridge paleoeconomists (Higgs 1972) and publica-
tions including Hodder and Orton's (1976) Spatial Analysis in Archaeology sug-
gest themselves as good examples of such trends. Edmund Leach's comments
on the "new archaeology" at the Explanation of Culture Change conference in
1971 emphasize the concern for such developments:
I appreciate that the new archaeology, being functionalist and behaviorist, is

practical, down to earth and scientific, so that its practitioners tend to be rather
disdainful of symbolic, non-rational human activity. It is consistent with this that
practically every paper has been concerned with problems of economic subsis-
tence, settlement patterns, demography. Religious rituals have scarcely been men-
tioned. Archaeologists who concentrate their attention exclusively on the kitchen
aspects of the garbage pit are certainly missing a lot [Leach 1973, 768-769].
Of course, such a situation was not simply a product of quantitative devel-

opment. The material nature of the archaeological record lent itself to such
forms of analysis and the limits of archaeological inference touted by some
archaeological pioneers; indeed, Edmund Leach institutionalized this perspec-
tive to some extent (Hawkes 1954). The Marxist perspective of some influen-
tial archaeologists earlier diis century, most notably Gordon Childe, also provided
intellectual support for the primacy of economic analysis in the study of early societies.
However, the acknowledgment of the relative sterility of an abstract or purely
quantitative approach to the study of human landscapes occurred early within
geography (Taffe 1974), and an emphasis on the need to incorporate a cultural
perspective within analytical approaches rapidly followed (Butzer 1977). Ar-
chaeology has followed this trend with the acknowledgment that "there is al-
ways a cultural geography" (Binford 1982: 7). Culture and belief systems are
increasingly interpreted as being able to order the physical environment within
absolute limits (Drennan 1976; Fletcher 1977). The need to develop "cognized
models" that incorporate the belief systems and perceptions of past societies
has become an imperative (Renfrew 1982).
In attempting such models, we are encouraged by Renfrew's (1982:11) sug-
gestion that "if people's actions are systematically patterned by their beliefs,
the patterning (if not their beliefs, as such) can become embodied in the ar-
chaeological record." The suggestion that cognitive information on the way
communities perceive and interpret their environment should be patterned
indicates that such qualities will be measurable and potentially mappable. To-
day, work on the symbolic properties of material culture, within both archae-
ology and anthropology, allows some degree of faith in the use of archaeologi-
cal data for such purposes (Douglas 1966; Hodder 1982a, b). The analysis of
portable objects, pottery in particular, has in some instances already proceeded
in such a manner (Bradley 1991: 78).
Despite this, some difficulties remain. It is clear that most studies still view
artifacts or the relationships between artifacts as the primary channels of sym-
bolic communication. Yet this is only a partial solution. Communities and cog-
nitive systems operate at a number of levels, and the need to order the larger
environment presents situations in which perception and belief can transcend
material culture and incorporate the natural. Within more mobile popula-
tions such perception may be of particular importance. Archaeology, with its
reliance upon crafted objects, has not necessarily been well placed to investi-
gate such situations. From sacred groves to "high places," natural features
may have an imbued value without needing any concomitant material associa-
tion, while the relationship of paths, tracks and natural markers in the envi-
ronment suggests that natural and cultural order are not only hard to distin-
guish (Renfrew 1982: 21) but that the distinction may be artificial and mis-
leading. There is a danger, therefore, in accepting the works of humanity alone
as an adequate symbolic metaphor. Bradley (1991: 77), drawing on the work of
Ingold (1986), has put the point well in suggesting that in some instances the
restrictive dominance of cultural remains in archaeology has led to an analysis
of "land rather than landscape....of a terrain in which cultural features have
taken the place of those natural elements that map the world of mobile peoples."
This observation is particularly clear when we consider the context of spe-
cific types of monuments. Prehistoric barrows, for instance, may originally
have commanded a greater visual impact on the landscape by virtue of their
striking appearance, "the neon of their time" (Evans 1985: 84) or placed to
dominate natural routes or viewpoints (Babic 1984: 38-39; Cace 1985). How-
ever, the difficulties of incorporating such observations coherently into the
analytical process are considerable. Evans (1985: 83) has suggested that most
traditional forms of mapping are unsuitable for such investigations. They in-
volve the reduction of place and space to location and distance, modern ab-
stractions that lose their cultural and cognitive context.
However, most GIS have the capability to overcome some of these difficul-
ties. According to the availability of data, the GIS provides access to all the
landscape irrespective of the presence or absence of archaeological sites. Con-
sequently, we are not only presented with the ability to investigate the positive
and negative characteristics of archaeological distributions but can also, if we
wish, explore the larger relationship of archaeological and natural space. There
is a more important point. Although earlier discussion has tended to accredit
GIS with a powerful ability to analyze space, it is also correct to say that GIS
may manipulate space according to variable, imposed values. We can assign
value to space, which can then be interactively analyzed with other forms of
mapped data. The simplest example of such a situation is the use of GIS to
explore access and movement via least-cost modules. In using such modules,
ease of movement across a surface becomes a measurable quantity. However,
the use of such techniques in the context of a variety of landscapes or by ad-
justment to reflect different transportation systems—pantechnicons or
packhorses—suggests that distance can be viewed through GIS as a relative,
not a constant, variable. The GIS is not, under such circumstances, to be con-
sidered as an objective observer of patterns implicit within spatial data; rather,

it is a tool to create spatial relationships according to values we regard as im-
portant. Although many people might consider such a judgment to be an indict-
ment, in the context of exploring values or belief systems through spatially struc-
tured phenomena, natural and archaeological, this is a strength to be explored.
Some of this potential can be illustrated through two studies carried out by
the authors. The first relates to a group of settlement and stone cairn data
from the island of Hvar in Dalmatia. Although already published in some de-
tail, the publications have limitations in that they were mainly concerned with
demonstrating the technical potential of GIS to a European archaeological
audience (Gaffney and Stancic 199la, 1991b). Consequently, only a relatively
small number of economic models and databases were utilized and the ar-
chaeological context of the data was not fully explored. It is therefore worth
reconsidering the results within a more explicitly archaeological context in
order to explore the implications for further GIS research. The second study
is intimately connected with GIS approaches to cognitive analysis and is an
interim report on a study of prehistoric rock art carried out to complement
recent analysis and fieldwork in mid-Argyll, southern Scotland.
Archaeology on Hvar: Sins of Omission by GIS Novices

The island of Hvar lies off the coast of central Dalmatia, Croatia (Figure 8.1).
Its peculiar elongated shape (it is about 68 km long and nowhere more than 15
km wide) results from its origin as an anticlinal peak, the strike of which runs
almost due east-west. The formation of the anticline has provided the island
with a central mountainous spine which peaks at 625 m. In the east of the
island, it forms a beveled upland plain containing a series of intermittent fer-
tile basins. The north central section of the island is dominated by a secondary
trough that forms the Stari Grad plain, the largest fertile area on the island.
However, other important geological deposits include the quaternary deposits
occurring within the Stari Grad plain and the intermittent outcrops of fertile
Eocene flysch that occur along the south coast of the island, especially around
the town of Hvar. Both of these are relatively fertile and well-watered zones.
The period under consideration here is the Bronze Age/Early Iron Age ca.
2200 to 400 B.C. Evidence for human activity is limited and is primarily linked
to the appearance of a series of cist burials under tumuli, which have been
compared with graves from the Cetina culture. This is a culture group defined
almost entirely on the basis of grave assemblages excavated around the source
of the river Cetina on the Croatian mainland opposite Hvar (Marovic 1963,
1976,1985; Marovic and Covic 1983). During the later Bronze Age, inhumation
burial under a mound continued in the area. This is in contrast to the appear-
ance of Urnfield burial traditions in the north of Croatia. However, there is
increasing evidence for permanent settlement in the form of a series of de-
fended enclosures. These vary in size and probably in_function. Some are very
large and, if interpreted as settlements, could have contained a sizable popula-
tion. Others are smaller and are often represented by little more than a small
Figure 8.1 Location map of the Island of Hvar.
rampart sealing off a mountainous spur. These could never have held more
than a couple of huts. The smallest enclosures on Hvar are circular and ap-
proximately 30 m in diameter. They often occur on elevated positions with
extensive views and are best interpreted as something like watchtowers rather
than settlements. Other habitation evidence is rare, although a series of pot-
tery scatters may represent the remains of unenclosed settlement (Figure 8.2).
The qualitative evidence for Iron Age occupation on Hvar is essentially an
extension of that of the late Bronze Age. Defended sites are the principal settle-
ment form, while burials continue to occur within tumuli. However, there are
reasons to suggest that at least one large site on the island, Hvar Castle, was
incorporated within a series of long-distance exchange networks. The finds of
eighth century B.C. Apulian pottery from this site indicate contact with Italy,
while the discovery of amber within nearby tumuli excavated during the nine-
teenth century indicate further-flung trade contacts (Petric M. 1986; Petric
N. 1979, 1980;). The overall context of the archaeological material on Hvar
suggests an affiliation with the Liburni. This is a historically attested tribal
grouping associated with the Creation coastline between Istria and the river
Krka which classical sources suggest may once have been in possession of the
central Dalmatian islands (Wilkes 1969: 8).
The initial published analysis of the Bronze/Iron Age data was carried out
primarily as a technical exercise in order to illustrate how GIS could be used
Figure 8.2 The principal Bronze Age and Iron Age sites on Hvar.
to establish the relationship between the principal prehistoric sites on the is-
land and their economic basis. In order to achieve this, an investigation of the
economic catchments of the largest defended sites on the island was under-
taken on the assumption that these sites represented the apex of the settlement
hierarchy. Cost surface catchments were constructed for each hillfort using a
timed and measured journey across the Stari Grad plain as a calibration factor.
The details of this process are published elsewhere and need not be repeated
here. However, the results suggest that these sites might be interpreted as the
central places of small prehistoric communities and that they were situated in
order to control large expanses of fertile land (Figure 8.3).
A second supporting analysis was carried out to correlate the distribution
of stone tumuli on the island with fertile soils. The rationale behind this lay in
the observation that the stone cairns frequently associated with burial could
also be interpreted as agricultural features. Excavation of such tumuli on the
Dalmation karst often provides no evidence for burial (Chapman et al. 1987).
The nature of the karst is such that stone clearance is a normal by-product of
arable agriculture (Gams 1987). As such, the positive correlation of the pre-
Figure 8.3 Soils within major hillfort catchments.

historic cairn distribution with fertile soils illustrated by the analysis was inter-
preted as further evidence for the agricultural basis of the prehistoric commu-
nities on Hvar. This supported the interpretation of the economic rationale
behind the siting of the major enclosures and cairns on the island. The distri-
bution of the fertile soil, enclosures, and cairns on the island also suggested
that further settlements could be expected in areas containing fertile soil and
cairns but no major enclosures (Figure 8.4).
Within the terms of the original publication, the results as summarized
above were satisfactory. They demonstrated the utility of GIS analytical tech-
niques and, at a technical level, they suggested that cost surface catchments
were a more useful guide to land associated with past settlements than
catchments produced in a more traditional manner. More pertinent, however,
were the results of the interactive analysis of the catchments with the environ-
mental data, the results of which represented a significant contribution to our
knowledge of the prehistoric communities in Hvar and Dalmatia. In particu-
lar, the emphasis on the role of arable agriculture within such communities
was a welcome counterbalance to the frequent assertion, deriving from later
classical sources, that pastoralism was the dominant economic mode during
these periods (Zaninovic 1977).
The isolation of a hierarchy of settlements topped by a series of larger en-
closures sited to dominate strategic portions of fertile land is also a useful
observation, although it may be suggested that such a result is descriptive rather
than explanatory. However, it is possible to rationalize the emergence of such
hierarchies from an economic viewpoint. The karst is a fragile environment. It
is vulnerable to serious soil erosion when its delicate equilibrium is disturbed
by agriculture (Shiel and Chapman 1988). Recent work in northern Dalmatia
suggests that significant erosion, almost certainly the result of the onset of
agriculture, was under way at least from the late Bronze Age onward (Chapman
et al 1987; Shiel and Chapman 1988). Although there are no paleoenvironmental
data available for Hvar itself, we can assume that it also suffered from serious
erosion. The single earthen tumulus located on the island during the recent
survey is situated on land that today is eroded to such an extent that there is no
significant soil cover remaining. Within such a context, it is no great step to
Figure 8.4 Settlements and cairn distribution in the eastern part of Hvar.
suggest that the threat to critical agricultural resources posed by erosion was
met by an increased emphasis on the possession of land, and that this is re-
flected in the construction of defensive enclosures sited to dominate fertile
areas. The emergence of such settlements was then associated with elites whose
power was based upon such access.
Presented in this manner, the results represent an essentially environmen-
tally deterministic approach to the data, a straightforward relationship be-
tween settlement, economy, and society. Although pleasingly supported by
GIS analysis, this interpretation is not satisfactory. In suggesting that pressure
on land led to the emergence of a hierarchical society, we are side-stepping the
problem of isolating the belief systems that permitted such social divisions to
develop and be maintained. An alternative explanation can be suggested using
the evidence included in the original GIS analysis and other information that
we have on belief systems within the prehistoric communities on Hvar.
Key to such an approach is the relationship between agriculture and the
dead. It has already been noted that stone cairns, although normally inter-
preted as funereal monuments, frequently provide no trace of burial and have
been interpreted as agricultural clearance cairns. However, it is significant that
there is no clear distinction between cairns of apparently different functions.
Cairns with burials frequently occur within fertile agricultural zones, while
mounds with no trace of any burial also occur within groups of cairns that
clearly are not related to agriculture and are more overtly "ritual." The most
striking example of the latter instance on Hvar is the barrow "cemetery" at
Vira (Figure 8.5). Here, twenty-two mounds are grouped around a small pen-
insula. Although the area may have maintained some agricultural potential in
the past, the surrounding land is highly eroded today. The placement of the
mounds in this group has no agricultural rationale. Indeed, it can be shown
that they are sited in order to be intervisible with a single small mound on the
central Vira peninsula. Despite this, excavation on the mounds since the nine-
teenth century has indicated that the majority of the cairns contained no evi-
Figure 8.5 Intervisibility analysis of the barrow cemetery at Vira.

dence of grave construction. Several of the mounds that did provide evidence
for grave cists were otherwise empty, with no trace of a body being recovered
(Petric 1979: 70). Such finds have often prompted the suggestion that the
mounds were "cenotaphs," monuments for those whose manner of death at
sea or in battle prevented the burial of a body (Zaninovic 1978). The fre-
quency of anomalous mounds on Hvar and in Dalmatia in general, either
as graves devoid of bodies or empty mounds in situations where funereal
evidence could have been expected, suggests that in some cases the mound
itself was important and the physical presence of a body might be of sec-
ondary consequence.
A second related observation is concerned with the evidence of ritual asso-
ciated with these mounds. Although most published emphasis is placed on the
evidence for formal grave goods when present, the most frequent finds during
excavation are simple pottery shards, which may be incorporated within the
body of the cairns or deposited within grave cists. This practice occurs through-
out the Bronze and Iron Ages (Marovic 1985; Novak 1959). The quantities of
shards within mounds may vary from a few individual pieces to large numbers
representing more than one hundred vessels (Batovic and Kukoc 1987). Shells,
gravel, and scraped-up earth are also noted as occasional deposits within cists.
In many cases the deposits suggest domestic refuse. One mound on the main-
land, a cairn without burial evidence, incorporated fragments of pottery and
house daub (Milosevic 1986). The incorporation of such material within
mounds suggests the presence of rituals concerned with cleansing and purity
associated with monuments related to the dead. The close relationship be-
tween death and purification and the maintenance of fertility through rituals
associated with these concepts is often stressed in anthropological studies
(Hodder 1982 a). In the case of Hvar, the ambiguous relationship of cairns
relating to burial and agricultural clearance, both spatially and in respect to
associated rites, makes analogy with such situations particularly attractive.
Treated in isolation, the linkage between mounds, burials, agricultural land,
and fertility suggests a concern with the legitimation of the control of valuable
agricultural resources through the ancestors and the dead. The existence of
dummy tumuli and empty graves emphasizes how important such contacts
may have been considered. There is, however, a broader context for such an
interpretation. Nearly all the "top ranking" enclosures on the island are also
associated with stone cairns of exceptional size. The precise situation of these
cairns varies. At the hillfort of Vela Glava, a massive cairn measuring 36 m in
length and 4.2 m in height is situated within the enclosure. The mound at
Purkin Kuk, the largest on the island at approximately 43 m in length and 5 m
high, lies within the probable territory of the Gracisce hillfort and high on a
hill overlooking the Stari Grad plain. The enclosure at Mosevcica is com-
posed of three ramparts cutting across a spur. The central rampart incorpo-
rates a large tumulus about 5 m in height. At Likovic there are a pair of large
enclosures, only one of which is interpreted as a settlement. The second
nonsettlement enclosure, Grcka Gomila, contains a massively enlarged mound,
6 m in height within its defenses. Despite the enlargement of the rampart of
Grcka Gomila, its construction close to a hill of approximately the same height
suggests that defense was not the principal function of the enclosure. The
only exception to this pattern is the site at Hvar Castle. It, however, has been
almost totally destroyed as a result of the medieval and postmedieval settle-
ment of the town of Hvar, but it is significant that the site is linked with
the Vira barrow cemetery described above, an exceptional ritual monu-
ment on the island.
Such mounds are not uncommon elsewhere in Dalmatia and western Bosnia.
A number of interpretations have been mooted for their function from towers
to temples for Sylvanic cults (Benac 1986). Excavations have taken place on
these mounds on Hvar. Unfortunately, the size of the monuments and the
small scale of past work has been such that the results cannot be taken as de-
finitive (Novak 1959; Zaninovic 1978). However, it is interesting that no buri-
als have been recorded in any of these mounds so far, although finds of domes-
tic refuse are frequent. In the case of Vela Glava in particular, the situation of
the mound on the peak of a prominent hill and the quantity of material lying
on its surface suggests, despite the probable existence of a settlement on the
lower slopes of the hill, ritual deposition rather than simple refuse disposal as
the source of the archaeological material.
Consequently, although there is a degree of variation between the situa-
tions of individual monuments, the cairns large and small form a coherent
group. Morphology and associated rituals link cairns and settlements. This
suggests a common concern with legitimation and the maintenance of fertil-
ity. However, the scale of the larger mounds sets them apart and their siting
suggests that some elements of the rituals associated with them may have had
a different emphasis. With the exception of the mound at Purkin Kuk, all the
mounds, while highly visible landscape monuments, are isolated in some way,
either within an enclosing wall or associated with some sort of excluding bar-
rier. This is also true for the Vira cemetery which combines the virtues of
general visibility of the central peninsular with the separation from the penin-
sula by two mounds that may originally have formed a barrier or a gateway to
the central mound. Although the mound at Purkin Kuk does not follow this
pattern so clearly, it does occupy an isolated hilltop position with exceptional
visual connections to virtually the whole of the largest fertile plain on the
island. The peculiar position of these monuments suggests that access to asso-
ciated rituals may not have been freely available and may have involved the
physical, but not necessarily the visual, exclusion of a portion of the population.
The possibility that there may have been restrictions in access to ritual on
Hvar is an important one. The role of ritual and religion as an arcane means
of placating, making obeisance, or communicating requests to higher authori-
ties is quite clear. In the social sphere, ritual also gains importance in the
relationship that individuals or groups are perceived to have with the super-
natural. The divine right of kings to rule is an explicit example of such a
relationship, demonstrating the legitimation of power religious convention
may confer on individuals. Equally important, however, is the power that
results from restricted access to ritual and the right to interpret divine in-
structions or carry out rituals that are seen as critical for the survival of society.
Within the context of the data from Hvar, the evidence for the origin of
rites associated with burials during the Early Bronze Age in Dalmatia suggests
that there was a perceived concern with the maintenance of fertility from at
least this time. The continuation of such practices and construction of large
monuments and perhaps settlements, intimately linked with similar rituals
during the Late Bronze Age, suggests that power within these communities
was being achieved by parts of the community through control of rituals main-
taining land fertility rather than simply through coercive control of the land
itself. There can be little doubt that with the onset of noticeable erosion dur-
ing the Late Bronze and Iron Ages, groups with a perceived preferential access
to ritual power associated with the maintenance of fertility would achieve sta-
tus. Such a situation would be enhanced by the increased isolation of ritual
within monuments which physically excluded the remainder of the popula-
tion. The role of ritual and religion in maintaining social conventions and
emphasizing the status of individuals and groups under the guise of divine
authority has a logic within societies with weakly developed political struc-
tures (Drennan 1976: 346). Such a context seems most appropriate to the
evidence from Hvar.
This paper is not the place to develop this argument further, although the
literature suggests that an extended analysis of the Croatian or central Dalma-
tian Late Bronze Age and Iron Age following these lines would be rewarding
(Benac 1986). What is important in the context of the reanalysis of the Hvar
data is the need to avoid the restrictions of a narrow economic viewpoint,
which might be encouraged by the use of GIS, and to emphasize its potential
within the context of an explicit theoretical perspective integrating the belief
systems that engender and maintain social relations. In short, there is a need
to be theory-led, if not theory-laden (Renfrew 1982: 2).
From Environmental Determinism to Cognition:

Paying a Suitable Penance
The second part of this paper relates to the study of prehistoric rock art and
ritual monuments in Argyll in southwestern Scotland (Figure 8.6). This is a
useful contrast with the data from Hvar, in that we are treating objects that
unequivocally carry a culturally embedded message both by virtue of the sty-
listic and symbolic information inherent within their designs and in their rela-
tionship to the larger physical landscape in which they operate. The area un-
der investigation is that of Kilmartin which includes the estuary of the Crinnan,
an important area of lowland north of Kintyre (Figure 8.7). The Crinnan es-
tuary faces west to the island of Jura and Ireland. It is a significant porterage
intersecting with a series of important routes to the north and west. Apart
from the agricultural potential that such an area of lowland might have of-
fered to prehistoric communities, the area is also significant in that it con-
tains an impressive concentration of ritual monuments relating to a number
of prehistoric periods (RCAHMS 1988).
Figure 8.6 Location map of the Kilmartin study area.
During the Early Neolithic period, the primary evidence for human occu-
pation is a series of burials within chambered cairns of various types. Settle-
ments of the period are rare and none have been found in the immediate area
of study. Interpretation of the evidence has therefore been limited to the burial
data and the tentative suggestions that the cairns held a symbolic role and
indicated the legitimation of a particular group's rights to land. Certainly the
cairns seem to have been used over a long period of time and appear to have
been opened and sealed intermittently for burials or other rites. The contin-
ued significance of these monuments within the landscape is indicated by the
insertion of later cists within the mounds (RCAHMS 1988: 6).
A series of later monuments, including an impressive variety of standing
stones and stone circles, is also concentrated within the study area, although
those associated with the timber and later stone circle at Temple Wood are of
particular prominence (RCAHMS 1988: 11). During the later Neolithic, at
least one henge monument was constructed at Ballymeanoch (RCAHMS 1988:
10, 52). Cist burials, frequently associated with cairns and barrows, appears to
have been introduced into the area along with Beaker cultural material and
were constructed throughout the Bronze Age. There is little evidence for settle-
ment until the Iron Age, when a series of settlement types including large
hillforts and Duns occur. Specific mention should be made of the fort at
Dunadd, which stands on a distinctive isolated hill in the center of the Crinnan
Figure 8.7 Kilmartin study area..
plain. The find of a carved Neolithic stone ball in the fort suggests that this
peculiar natural feature achieved some degree of inferred importance from an
early period (Bradley, personal communication; RCAHMS 1988: 156). There
are also persistent finds indicating the special status of the site over a long
period. These include rock carvings and exotic imported pottery. During the
early historic period, the fort became the capitol of the Dal Riata, a tribe of the
Scotti (RCAHMS 1988: 33-34, 149-159).
The prehistoric rock art associated with this area is one of the largest and
most varied groups in the British Isles (Morris 1989: 52). Although the major-
ity of the artwork is composed of simple cup or cup-and-ring marks on natural
flat rock faces, more complex designs are also found. These include spirals,
starred circles, and various grooves and enclosures. The cup-and-ring marks
themselves display significant variation in their composition, association, size,
and numbers of rings (Bradley 1991; RCAHMS 1988: 10). Rock art is notori-
ously difficult to date, and it is possible that the carvings were produced over
millennia. However, as Bradley (1991: 79) has noted; "the ultimate inspiration
for this art style...comes from stone-built tombs in Brittany and Ireland," and
there are elements of the Argyll material that link it with the Boyne and Irish
Passage Grave art (Bradley 1991, Fig. 14; Morris 1989: 52; RCAHMS 1988:
120). The westward orientation of the study area and other evidence for contact
with Ireland provide a suitable context for such observations (RCAHMS 1988:6).
It is significant, however, that carved rock panels share motifs with carved stones
within standing monuments. This suggests some degree of contemporaneity.
When the carving started in the Kilmartin area remains uncertain, but carved
representations of bronze axes on monuments including the cist graves at
Nether Largie north must be late in the sequence (RCAHMS 1988: 68).
The situation of rock art panels in the British Isles is variable, although a
recent summary by Bradley (1991) has noted the tendency of principal groups
of rock art to be situated both in the vicinity of major monuments and on land
interfacing with lowland and highland zones. The integrative quality of the
art is also emphasized in the frequent relationship that it demonstrates with
dominating viewpoints and important routes. The arcane nature of rock art
has laid itself open to a wide variety of interpretations, not all of which sustain
rational examination (Jones 1991: 69-70; Meaden 1991: 189-194). Bradley
has wisely rejected any attempt at formal interpretation of the rock art and
chosen to emphasize the need to treat the art as a channel of information of
greater or lesser complexity, suggesting that complexity would be expected
where larger and more varied groups might expect to meet and where com-
munication between such groups would be of particular importance. The gen-
eral relationship of art with public monuments, routes, and intermediate land
zones suggests that this is a profitable line of inquiry. Bradley's more detailed
analysis of the Northumberland and Kilmartin rock art has, to a large extent,
substantiated this assertion. In the Kilmartin area in particular, there is a rela-
tionship between the complexity of rock art and points of entry to the lowland
area that contains the major public monuments, prompting Bradley to suggest
that the symbolic message of the art may have been directed at those entering
the estuary (Bradley 1991: 99).
The complementary study to Bradley's stylistic analysis presented here at-
tempts to use GIS in a manner that allows some quantification of the percep-
tion of the monuments in the Kilmartin area. The study incorporates the analy-
sis of viewshed data constructed for all the nonsettlement sites listed by the
Royal Commission for the Ancient and Historical Monuments of Scotland for
the Kilmartin area from the Neolithic to the Bronze Age, a total of 76 sites
(RCAHMS 1988). The decision to use all the pre-Iron age sites partly springs
from the desire to contrast patterns for different classes of site and earlier and
later monuments. However, it was also noted that the explicit inclusion of
earlier sites into the later cognitive system, most amply demonstrated by the
insertion of the beaker cist grave at Nether Largie south chambered cairn
(RCAHMS 1988: 51) and the incorporation of reused rock art within later cist
graves, strongly suggests that the data should be treated as a continuum. Fol-
lowing from this, the monuments were broken down into five basic groups for
analytical purposes: chambered cairns (6); individual or groups of decorated
natural rock faces (28); standing stones, alignments, and circles (15); the henge
at Ballymeanoch; and cists, cairns, burials and barrows (26). The distribution
of sites is given in Figure 8.8.
The analysis of these data involved the calculation of the viewshed for each
monument—that is the area that can be seen from an individual site. This
information represents the area within which that monument is likely to com-
municate visual information. Monument viewsheds can overlap, producing
zones in which an observer might be aware of many monuments, all of which
Figure 8.8 Archaeological sites in the Kilmartin study area..
may carry information. Presumably, the increasing density of such informa-

tion can be interpreted in some circumstances as a measure of the importance
of a particular area. In the context of this work, it is perhaps better to emphasize
the ability of such a procedure to provide a mappable, spatially variable index of
perception, which incorporates groups of monuments and plots their visual rela-
tionship with the surrounding landscape. Analysis of these data should give an
insight into the cognitive landscape within which the monuments operated.
An example of the process is given in Figure 8.9. Here, the viewsheds for
the Ballymeanoch Henge and a pair of standing stones at Dunamuck are illus-
Figure 8.9 Example of cumulative viewsheds.
trated, along with the cumulative viewshed for the two sites. Within the cu-
mulative viewshed the areas that are visible to only one monument have a
value of one and the areas that are intervisible have a value of two. The sum of
this process for the individual monument groups can be seen in Figure 8.10.
Analysis of these viewsheds provides a number of interesting patterns. Fig-
ure 8.11 illustrates the mean intervisibility values of monument groups. Cham-
bered cairns emerge with the lowest mean value for intervisibility. Rock art also
has a relatively low value of intervisibility, while there is an increasing level of
intervisibility through the groups of standing stones, cists and burials, with
the Ballymeanoch henge emerging as the most visible single monument type.
These data can be broken down further to measure the way in which
these monuments interact with each other. Table 8.1 illustrates data on
intervisibility between monument types. This emphasizes how very little
the chambered cairns and rock art interrelate with any group of sites. There
is an increasing association between the standing stones and the henge
Figure 8.10 Cumulative viewshed data for individual monument types.
monument, while there is a clear positive relationship between the henge

and the later cists and barrows.
Several significant points are suggested by these results. The first is that
chambered cairns, as a group, are not integral with the other monuments,
even though individual cairns provide evidence for their incorporation into
later cognitive systems. It is also, perhaps, surprising that the rock art, despite
the relatively large number of such sites, interacts at a very low level with
other monuments. However, the henge is clearly a focal point for the standing
stones and cists. The strong relatedness of the henge and cists indicates that
Plotted within one standard deviation
Figure 8.11 Mean monument intervisibility.
many of the later burial monuments were deliberately sited to be in visual

contact with the henge, emphasizing its special position within the area.
Some insight into the processes leading to this situation can be inferred
from the nature of the land contained within monument group viewsheds.
Figure 8.12 illustrates the mean proportions of lowland within the viewsheds
of each monument group. This shows that although the viewsheds of cham-
bered cairns and rock art may visually incorporate considerable amounts of
lowland, there is a trend toward an increasing emphasis on the lowland within
the remaining three monument groups, with the Ballymeanoch henge, again,
emerging as the monument type with the greatest mean amounts of land, but
the cists and burials also frequently contain very large amounts of lowland.
Table 8.1 Index of intervisibility between sites (Mean number of visible sites divided
by the total number of sites of each type.)
Monuments Viewed
Chambered Rock art Standing Henge Cists etc.
Cairns Stones
Viewpoints
Chambered Cairns 0.19 0.05 0.07 0.00 0.11
RockArt 0.06 0.12 0.11 0.11 0.10
Standing Stones 0.07 0.11 0.15 0.33 0.20
Henge 0.00 0.07 0.33 — 0.54
Cists etc. 0.10 0.09 0.16 0.54 0.31
Figure 8.12 Lowland within cumulative viewsheds.
In part, this is a chronological pattern. The Early Neolithic communities

that built the chambered cairns may well have been concerned with the regu-
lar exploitation of a wide variety of environments. This is reflected in the varia-
tion displayed by the monuments of the period. This may contrast with the
later communities associated with the cist graves, which were probably more
closely involved in agricultural pursuits. The importance of the Crinnan low-
lands to such communities is reflected in the siting of contemporary monu-
ments and visual association with the lowland zone. How are we to incorpo-
rate the other monuments within this scheme? It has already been emphasized
that the presence of shared motifs suggests some degree of contemporaneity
between rock art panels and monuments, including the Temple Wood stone
circles. Therefore, we should not seek some simple temporal movement to-
ward the lowland as an explanation for patterning.
What may be more significant is to emphasize the contrasting nature of the
decorated rocks and other major public monuments, including the standing
stones, circles, and henge. Bradley (1991) has emphasized the tendency for the
most complex monuments, including the henge, stone circle, decorated standing
stones, and the most complex rock art forms in the Kilmartin area to be asso-
ciated with the valley entrances to the Crinnan estuary. This may well contrast
with the majority of rock art panels whose message is more pervasive and per-
haps subtle. Rock art panels occur in a wider variety of settings but essentially
integrate upland and lowland. The artwork itself, though a visual phenom-
enon, is considerably less visible than the other public monuments. The rock
faces upon which such carvings occur are frequently almost horizontal (Morris
1989: 48) or occasionally even in situations which may obstruct views (Bradley
personal communication). Figure 8.11 illustrates that intervisibility is rela-
tively low for this monument group. The information these monuments im-
parted seems to have been very closely associated with the landscape in a more
intimate sense, a point firmly emphasized by the use of natural surfaces for
decoration.The intergrative position of the monuments at the junction of the
highland and lowland also emphasizes this role but suggests some degree of
ambiguity in their relationship between land zones. Braithwaite (1982: 81), in
her discussion of Sudanese ethnographic data, has emphasized the role of deco-
ration and ritual to express, communicate, authorize, and guide action on
boundaries and other ambiguous areas of social interaction. It seems reason-
able to view much of the rock art in the Kilmartin area in this light. The large
and complex public monuments may be communicating to diverse groups ar-
riving in the area. The less complex rock art panels may be communicating
information on areas of ambiguity across the landscape, perhaps to those who
utilize these diverse economic zones.
There are clearly a series of potentially conflicting symbolic structures op-
erating within the monuments of the Kilmartin area. These are important and
will provide further insights as analysis proceeds. However, it is worth stress-
ing the emergence of a coherent pattern from this disorder over time. By pro-
ducing a cumulative overlay of all monuments for the area, (Figure 8.13), it is
clear that specific areas achieved prominence as defined by their potential per-
Figure 8.13 Filtered cumulative viewsheds of the Kilmartin study area.

ception and illustrated through the viewshed data. We can discriminate to a
certain extent between natural and archaeological features. Within Figure 8.13,
the cumulative viewshed image for all monuments has been filtered to dis-
criminate between those areas with viewshed values equal to or less than the
maximum value of the archaeological monuments and those that exceed this
value and must relate to the natural, nonarchaeological features. The bluff to
the south of the Crinnan canal, although not the highest part of the survey
area, assumes particular significance within the latter group. It must have al-
ways been a prominent feature in the cognitive map of the inhabitants or visi-
tors of the estuary. The valley entering the estuary to the north at Nether
Largie (Figure 8.14) clearly forms the ritual cognitive focus of the area center-
ing on some of the most important and imposing monuments in the valley.
Around this spot are the massive Nether Largie chambered cairn, the Temple
wood stone circles and the Ballymeanoch henge (RCAHMS 1988:48-52, IBS-
HI). All three of these major monuments have later cist burials inserted,
thus integrating elements of the whole sequence into a single cognitive
system, while the cists and burials clustering around these monuments
include several that incorporate earlier carvings, further emphasizing the
very real perceived relationship between these temporally separate monu-
ments (RCAHMS 1988: 68-70).
GIS: Archaeological Panacea or Pandora's Box?
It does not do to labor a point, but it is hoped that the above examples clarify
some of the observations made in the introduction to this paper. Clearly, ar-
chaeology has prospered by shedding some of the shackles of functional and
economic determinism. We would be well advised to ensure that these do not
return in modified and seductive new wrappings. The reanalysis of the Hvar
Figure 8.14 Ritual monuments around Nether Largie.

data suggests that GIS, if not used thoughtfully, is open to such criticism.
However, archaeology abounds in such cautionary tales, and it is more pro-
ductive to emphasize the new opportunities that GIS offers. The interim analy-
sis of the Argyll ritual monument data clearly illustrates the potential of GIS
for the study of large-scale cognitive phenomena and its ability to utilize the
full landscape for such purposes. This is new and we are only just beginning to
explore the possibilities on offer. In this respect, it is worth emphasizing that
the Argyll data were analyzed solely through the application of standard GIS
modules. While that may have been convenient for this exercise, the limita-
tions of modules that were primarily designed for the analysis of physical space
will eventually prove restrictive. The future for innovative work in GIS will
therefore lie in the development of more sophisticated mathematical modules
explicitly for archaeological purposes and within the context of GIS technolo-
gies. Archaeology has a reputation as a somewhat eclectic discipline, and there
is no doubt that in the past this has proved a strength. In this case, however, we
have the opportunity to treat GIS as a "core" technology that can be devel-
oped according to our own agenda. Used carefully, archaeology can only ben-
efit from such a situation.
Acknowledgments
We would particularly like to thank Professor Richard Bradley of Reading University,

Great Britain, Bozidar Slapsak, and all other members of the Hvar Project for provid-
ing access to important unpublished data and for discussing the results of the analysis.
Thanks also go to the National Institute for Geographical Information Systems in the
United States and to the University of Ljubljana for assistance with a traveling grant.
We would like to thank everyone who has commented upon the results presented in
this paper. All errors are, of course, our own.
9
An Analysis of Late-Horizon Settlement Patterns
in the Teotihuacan-Temascalapa Basins:
A Location-Allocation and GIS-Based Approach
Amy J. Ruggles and Richard L. Church
The general interest of linking GIS capabilities and location-allocation (L-A)

techniques to investigate certain spatial problems should be evident. The tech-
niques and the technology are often complementary. A GIS can provide, man-
age, and display data that L-A models require; in turn, L-A models can en-
hance GIS analytic capabilities. This combination of information manage-
ment and analysis should have wide appeal. The technique and technology
may be especially wellmatched when one considers many of the special re-
quirements of archaeological applications of L-A models. We intend to inves-
tigate and illustrate the value of such a combined approach though the ex-
ample of a regional settlement analysis of the Late Horizon Basin of Mexico.
Quantitative Settlement Modeling in Archaeology

GIS-Based Modeling
Geographic information systems are increasingly common in archaeology.

Their ability to manage, store, manipulate, and present spatial data is of real
value, since the spatial relationship between objects is often an archaeological
artifact in its own right. Space is central to both archaeological data (Spaulding
1960; Savage 1990a) and theory (Green 1990). Although GIS may not always
offer intrinsically new and different manipulations or analyses of the data, they
can make certain techniques easier to apply.
There is a wide spectrum of GIS-based modeling applications in archaeol-
ogy (Allen 1990; Savage 1990a). The anchors of this spectrum range from the
use of GIS in the public sector in cultural resource management settings to
more research-oriented applications. The strongest development of GIS-based
archaeological modeling is probably in the former context. Models developed
here are predominantly what Warren (1990) identifies as "inductive" predic-
tive models where patterns in the empirical observations are recognized, usu-
ally using statistical methods or probability models. This type of application is
usually identified with "site location" modeling (Savage 1990a). As defined,
these models do not predict the probable locations of individual sites but rather
calculate the probability that a geographic area will contain a site, given its
environmental characteristics (Carmichael 1990: 218). The primary role of
GIS in many of these applications is to manage and integrate spatial informa-
tion and feed it to some exterior model.
The other area of GIS-based modeling involves research applications. These
models attempt to incorporate sociocultural factors into the analysis and are
often implicitly or explicitly linked with landscape archaeology (Green 1990).
Landscape archaeology encompasses a range of geographic topics, including
subsistence studies, settlement and site-type analysis, and geographic location
theory (Savage 1990b). Although some definitions of research-oriented GIS
applications appear specifically to rule out any archaeological research into
site locations,1 this distinction may be made primarily to distinguish this class
of applications from the predictive class where regional "site location" models
are a near-universal focus. One common element of these applications is that
to date they have often worked within the confines of GIS abilities. Applica-
tions of standard techniques have been imaginative, but in some cases the li-
brary has proved limiting.2 There are exceptions to this tendency, such as Farley
et al. (1990) and Williams et al. (1990) who integrate GIS into a tool kit of
techniques that include GIS, remote sensing (RS), exploratory data analysis
(EDA) and database management systems (DBMS).
Optimization and Location-Allocation Models
Applications of operations research techniques are not unknown in archaeol-

ogy. Many of these applications involve applying some form of quantitative
linear programming (LP) model (Keene 1979, 1981; Reidhead 1979, 1980).
Less common are applications of LP or integer programming (IP) models
with an explicitly spatial focus (Dickson 1980), or of the class of IP models
known as L-A models (Bell et al. 1988; Bell and Church 1985; Church and
Bell 1988). These spatial models provide a means to construct tailor-made
idealized configurations for an archaeological system. Their optimizing
function(s) can be chosen in light of the specific cultural system, context, and
problem under investigation, while the expression of those goals can take place
in a nonisotropic "real-world" setting. These characteristics can be contrasted
to classic central place theory, where an implicit maximization function is as-
sumed universally applicable and the resulting settlement pattern is expressed
in an isotopic environment.
The use of optimizing models in archaeological contexts has not been with-
out criticism (Evans and Gould 1982; Jochim 1983; Keene 1983; Lewarch
1978). Most objections to the use of optimization models are concerned with
the fundamental behavioral implications of optimization, contrasting the pro-
cess of "optimal" decisionmaking to actual human behavior. These concerns
cannot easily be addressed. The degree to which strict optimality can reason-
ably be expected in the human world is a critical point and the assumption that
humans are perfectly rational, economic decisionmakers is difficult to justify.
GIS-Based Analysis of Late-Horizon Settlement Patterns 157
Optimizing LP models should certainly not be attempted on the justification
that they replicate the process of real-world decisionmaking. However, this
does not mean that human behavior cannot be economically and rationally
optimizing1 under certain circumstances. If a case can be made for a push to-
ward or desire for the efficient use of some resource existing in the system
under study,4 then an optimizing approach is probably appropriate and may
provide insights not otherwise available. We believe that by requiring evi-
dence that optimizing behavior can be expected and by more rigorously test-
ing model results, many of these criticisms can be addressed.
Many of the objections to the use of optimizing models are not uniquely
archaeological in nature but arise because of the human context of archaeo-
logical modeling, a context that is shared with contemporary planning.
Because of the human context of their models, archaeologists should prob-
ably follow the lead of planners in searching for robust optimal solutions
and investigating alternative close-to-optimal solutions. They would like-
wise benefit from comparing the results of significantly different model
formulations and hypotheses, including very simple alternatives (Jochim
1983). Archaeologists should also be cautious when interpreting the re-
sults of optimizing models and may choose to use model results primarily
to explore the problem context (Keene 1981) or explicitly to identify unex-
pected behavior (similar to Altschul 1990).
Another class of objections to archaeological applications of geographic
location and optimization models is founded upon the different nature, qual-
ity, and quantity of geographic and archaeological data (Evans and Gould 1982;
Lewarch 1978). Geographers and planners benefit from more readily avail-
able data, while archaeological data are often derived from models themselves
and are more heavily dependent on assumptions and theory. Sensitivity test-
ing is of great importance in any application of these models, but it is usually
mentioned only in early archaeological applications of LP models (Keene 1979;
Reidhead 1979), and then only to the degree appropriate to contemporary
planning applications. Because of the nature of archaeological data, archaeo-
logical sensitivity testing must be more radical. This more radical testing should
involve testing model sensitivity not only to parameters but also to entirely
different estimates of input data.
The Combination of GIS and L-A Modeling
There are a few obvious ways in which one can organize a quantitative analysis
of an archaeological (or other) settlement system (Figure 9.1). Track I is de-
fined as working entirely within the capabilities of a GIS: the method followed
by many research applications of GIS in archaeology. Track II is defined as
working entirely exterior to the GIS: a category that subsumes most modeling
in archaeology and most archaeological applications of L-A to date. Finally,
track III is defined as using GIS capabilities while incorporating outside tech-
niques: examples of this mode can be found in many planning applications of
GIS in archaeology and a few research applications as well (Farley et al. 1990;
Figure 9.1 Different general methods of incorporating a GIS into spatial analysis.
Williams et al. 1990). The distinction between the defined tracks is not hard
and fast, as it depends upon the state of GIS development. We will take a track
III approach in the our research for the following reasons.
The data requirements of L-A modeling are considerable. Furthermore,
the L-A model solutions of most interest to archaeology, robust solutions that
avoid a strict focus on single-point optimality, make even greater demands
upon the quality and volume of data required. Archaeological data sets are
hard-pressed to provide appropriate information; GIS, on the other hand, al-
low for the easy manipulation, creation, and presentation of spatial data. In
their current implementation, GIS have limited modeling and analytic capa-
bilities but can provide spatial, nonisotropic information to an exterior model.
The ability of a GIS to create spatial data, and to present and analyze the L-A
configurations, once developed, will be valuable but not critical to this effort.
The single most critical capability GIS can offer is easy access to spatial data in
a standard format. This will allow L-A models to be run multiple times under
various assumptions and data conditions that will allow the user to better search
and test for robust model solutions. The GIS not only facilitates this type of
analysis but, to a large degree, makes this type of analysis feasible.
The linkage of L-A models and GIS techniques is not proposed as a pana-
cea for all spatial analysis, let alone for all archaeological spatial analysis. The
suitability of the L-A model to the archaeological context and the adequacy
of the archaeological data to the task should always be of concern. Still, L-A
models when used and interpreted appropriately and supported by the capa-
bilities of GIS technology, provide a valuable technique for addressing a class
of archaeological problems.
A Problem-Driven Description of the Late Horizon Cultural System
Several conditions must exist before there is a strong case for using an opti-
mizing Operations Research (OR) approach to model a cultural system. There
must be some reasonable and not incidental goal either implicit or explicit to
the functioning of the system. This goal need not be singular, although single
and simple goals are easier to model. There must be constraints upon the
system that inhibit achieving this goal. All of these goals and constraints must
be capable not only of being reasonably approximated by linear mathematical
equations but also of being "measured" by some common (shared) quantita-
tive measure. There should also be some demonstrable push toward optimiza-
tion or efficiency. Optimizing behavior is more likely to occur if the system
being modeled is of fundamental importance and if the constraints or demands
upon that system are severe. The Late Horizon (A.D. 13 50 to 1519 or "Aztec")
regional settlement system appears to be a promising candidate for analysis
using a L-A approach.
This quantitative analysis is possible because a baseline survey data set (Sand-
ers et al. 1979) exists with fundamental information concerning the settlement
pattern, demographics, and agricultural potential of the Basin of Mexico. There
is a wealth of ethnographic and historic data for the Basin. Fundamental to
our work is the existence of investigations specifically into the agricultural
potential and settlement characteristics of the Temascalapa-Teotihuacan re-
gion (Evans 1980ab) and earlier quantitative spatial analyses of settlement in
this region (Bell et al. 1988; Evans and Gould 1982; Gorenflo and Gale 1986).
The Basin of Mexico is the geographic and political core of the Aztec em-
pire (Figure 9.2). The Basin is part of a broad and largely volcanic upland
(approximately 2240 m above sea level) zone. Until the nineteenth century,
the Basin supported between three and six shallow Pleistocene relic lakes. The
Basin is a semi-arid environment; precipitation and temperature are highly
seasonal and variable. Variations in the magnitude and timing of either pre-
cipitation or frost is a critical factor for agriculture. This element of risk is one
reason to expect the efficient use of agricultural resources.
Following the work of Bell, and colleagues (1988), our research focuses
upon the Temascalapa-Teotihuacan region. This region of approximately 900
km2 in the northeast Basin of Mexico is reasonably coherent. It is physically
separated from the rest of the Basin by the Patlachique Range to the south, the
lakes Texcoco and Xaltocan to the west and an arid region of low population
to the northwest. This region fell entirely under the Acolhua administration
Figure 9.2 Late-Horizon settlement in the study area of the Temascalapa-

Teotihuacan basins.
(Bell et al. 1988: 167), which defines the less physiographically distinct bor-
ders. Most of our data concerning this region comes from the Temascalapa
and Teotihuacan survey regions (Sanders et al. 1979).
If anything, the general risks and limits on agriculture that apply to the
Basin of Mexico are more severe in the Temascalapa-Teotihuacan region. There
is less precipitation, and the dangers posed by variations in frost and precipita-
tion are increased because of elevational differences. The Temascalapa-
Teotihuacan area is probably better suited for the cultivation of maguey and
nopal than maize (Bell et al. 1988:169). However, it is unlikely that maize was
not grown in the area. The classic Mesoamerican diet was a maize and beans
combination, and there is empirical evidence that maize tended to be grown
to the limits of its range (Sanders et al. 1979:233). Because of this and because
maize benefits from the most abundant data on cultivation and yield, we will
adopt it as our standard crop for the estimation of agricultural production
following precedents established by other subsistence models of the Basin (Bell
et al. 1988; Parsons 1976).
Simple demographic stresses upon the resource base during this period are
another reason to expect the efficient exploitation of agricultural resources.
The Late Horizon Basin of Mexico may have supported as many as 1 million
people (Sanders et al. 1979: 162) and may have been operating at 88% of its
carrying capacity (Sanders et al. 1979: 378).
The Aztec Empire was a conquest state ruled by the "triple alliance" of
Tenochtitlan, Texcoco, and Tlacopan, although in reality the empire was pri-
marily controlled from Tenochtitlan. It has been suggested that the Aztec
Empire had an unusual and/or increasing reliance on tribute (Fagan 1984;
Parsons 1976). Tribute could take the form of both subsistence and high-sta-
tus goods or services (labor, military service, etc.) (Davies 1980: 112) and was
delivered at fixed intervals, although extraordinary demands could be made
for certain occasions. Collection of tribute from a large hinterland may have
been partly an adaptation to subsistence stresses. However, the Aztec system
of tribute collection may have imposed its own costs upon the subsistence
system already subject to simple physical and sociodemographic demands. The
Aztec system of tribute collection was complex. Subordinate states did not
always pay tribute to their closest regional center, which is to say that the
official hinterlands of regional centers were interdigitated. Some argue this
was done to increase political stability by depriving any regional center of a
coherent spatial base for revolt (Evans and Gould 1982: 287). Additionally,
Aztec subordinate states could owe tribute simultaneously to different centers
at the same or different levels of the administrative hierarchy (Hodge 1984).
The Aztec maintained two distinct types of regional administrative centers:
city-state capitals where the traditional local lineages continued to rule and
confederation or regional government outposts administered by representa-
tives of the triple alliance (Evans and Gould 1982: 287). The distinction be-
tween these types appears based upon a difference of administration rather
than one of function. There were either eleven or twelve regional administra-
tive centers in the Temascalapa-Teotihuacan region; eleven according to Evans
(1980a) or twelve if we follow Bell and colleagues (1988), who modeled only
city-state capitals and included an additional site in their analysis. Of these
twelve centers, six were definitely city-state capitals and five were regional
government outposts.
Beyond the indirect evidence of optimizing tendencies such as collecting
tribute from a larger area, there is archaeological evidence of a trend toward
the efficient use of agricultural potential in the expansion, intensification, and
control of agricultural production during the Late Horizon (Sanders et al.
1979: 176-177). The present research is based on the belief that a similar
trend toward efficiency can be expected in the choice of settlement location.
The location of settlements could be determined by a desire to maximize gross
agricultural production, or it could be determined by the desire to minimize
the cost of transporting foodstuffs to their final destinations, which in effect
maximizes the amount of subsistence foodstuffs available to the society.
The pattern of interaction between places is thus important, as it has an
impact on the efficiency and perhaps basic viability of the cultural system. In
the present study, we will focus specifically upon the political pattern of inter-
action. Although it is not clear whether trade or tribute was the more funda-
mental force behind the movement of subsistence goods across the Basin (Evans
1980b; Smith 1979, 1980 ), the volume and regularity of subsistence goods
being demanded as tribute is a good indication of their probable importance.
Both Bray (1980) and Santley (1986) identify the pattern of interaction
among the Aztec as a "dendritic marketing pattern," where the movement of
goods, etc., is centripedal. Such a tribute system is said to be efficient at mov-
ing primary goods (often crops) out of an area "by means of collection points
of intermediate rank" (Bray 1980: 6). Santley (1986) sees this structure as be-
ing one method to buffer a state-level system in an semiarid environment against
subsistence risk and as an adaptation to stresses posed by a highly developed
politicoeconomic superstructure (Santley 1986: 234-241).
To date quantitative L-A based analyses of the settlement pattern in this
region have found little evidence of the expected efficiency in settlement loca-
tion (Bell et al. 1988; Gorenflo and Gale 1986). During the development of
our model, we will comment briefly upon the research of Bell and colleagues
(1988) as it is the most similar to our own. Their data set is the same 344
known Late Horizon settlements in the Temascalapa-Teotihuacan region. Both
studies use L-A models to locate regional administrative centers. Finally, both
models use similar measures of the basic integrating components: an estimate of
surplus agricultural production (in maize) and an estimate involving population.
A Location-Allocation Model Approach
The Maximum Coverage Location Problem
Bell and associates (1988) chose to use the maximum coverage location prob-
lem, or "Maxcover" (Church and ReVelle 1974), as their basic model formula-
tion. Maxcover locates a given number of centers (facilities) on a network so
that they will serve or "cover" as many users as possible within a prespecified
time or distance standard. The nonhierarchical formulation of Maxcover they
applied does not take into account the economics of getting the tribute and
agricultural production to where it was needed, nor does it account for the
actual hierarchical importance and function of the facilities being located. Bell
et al. (1988) modeled the Temascalapa-Teotihuacan region as a bounded, in-
dependent region and not a sub-system within a larger entity. There are some
additional implementation and interpretation problems that are not necessar-
ily a general criticism of the Maxcover model. Bell and colleagues (1988) sited
an incomplete set of administrative centers in the region, locating only the
seven city-state capitals. They also did not attempt an extensive sensitivity
6
analysis of model results to assumptions inherent in their data sets. Finally,
they interpreted their L-A model results as the optimal configuration and level
of performance for the empirical system and measured the empirical system
efficiency and performance against this baseline. This use of model results can
be questioned on the basis of whether the Maxcover model adequately cap-
tured the behavior and goals of the Aztec system.
One objective of this research was to investigate the feasibility and demon-
strate the probable value of incorporating capabilities of GIS and L-A tech-
niques to address a class of archaeological problems. The more immediate,
problem-specific objective was to address as many of the weaknesses of Bell
and coworkers' (1988) modeling the Late Horizon Temascalapa-Teotihuacan
system as possible. This would be in part through the application of a more
appropriate L-A model, a more rigorous program of sensitivity analysis, and a
more cautious interpretation of results.
For this L-A analysis, we selected the supporting P-median problem (Weaver
and Church 1987), as it will explicitly consider the flow of surplus agricultural
production and/or people through the environment. Our application involves
the siting of the full twelve facilities, taking into account their intermediate
hierarchical position and links to a larger exterior system. We applied the L-A
model to different estimates of agricultural productivity and road networks in
order to investigate model sensitivity to the assumptions inherent in these
data sets. Finally, the facility configurations produced by this model are treated
as idealized and oversimplified hypotheses concerning the sociocultural sys-
tem. Comparisons involving the "efficiency" of the various modeled and em-
pirical configurations are made primarily to judge if the model-generated con-
figurations appear to replicate the empirical pattern of facilities and not to
definitively measure the efficiency of the empirical system. Evaluating the ef-
ficiency of the Late Horizon system is valid only if the locational model is an
acceptable hypothesis concerning the functioning of the system and (1) the
data sets the model used are well established or (2) the model results are rela-
tively robust to variations in these data sets.
The Supporting Facility Location Model
The supporting P-median problem, (Weaver and Church 1987) locatesp sup-
porting facilities on a transportation network in a system that already has q (q
> 1) central facilities. A supporting facility is one that serves passengers or
shipments before and/or after service at a central facility. This model assumes
that travel between a supporting facility and a central facility has a cost that is
some fraction a, (0 <a < 1), of the direct travel cost between these locations.
The objective of the supporting P-median problem is to locate supporting
facilities in order to minimize the transport costs between sources and the
final destination (central facility). Supporting facilities will be used only if it is
more efficient to do so. This will occur only when the travel cost from the
village or hamlet to the central facility via the supporting facility is less than
the direct travel cost from the hamlet to the central facility. For the purposes
of this application, the supporting facilities to be located will be the provincial
administrative centers. The central facility for this region will be Texcoco for
all centers. Thus tribute flow moves from hamlets to administrative centers
(supporting facilities) to Texcoco (the single central facility). The system cost
of tribute movement represents the total movement from areas of agricultural
productivity to Texcoco both directly and via administrative centers. This rep-
resents a dendritic system of tribute flow.
There are several possible interpretations of a in the present context. When
the measure of demand being served is surplus agricultural production, a could
represent a change in the cost or difficulty of transporting goods between the
supporting center and the final destination or the different amounts of surplus
being shipped on to the final destination (or some combination of the above
factors). If the demand being served is people, a might represent some mea-
sure of interaction between sites and administrative centers. The interaction
between administrative centers might be reduced in either case because there
is relatively less interaction between these higher-level centers than with local
sites or because interaction with these higher centers is easier because of es-
tablished and regular channels of transportation or communication.
Bell and associates (1988) solved a multiobjectiveMaxcover model simulta-
neously upon their measures of agricultural surplus and population. The cur-
rent research will run parallel single-objective models on the alternative sur-
rogate measures. This single-objective approach is undoubtedly a simplifica-
tion of the complex, multiobjective problem-space of real-world settlement
location decisions. However, if single-objective models working independently
upon agricultural surplus and population produce similar patterns, this tells us
something about the nature of the problem space and potentially robust na-
ture of the configuration. The more complex multiobjective model may not
be required to capture the heart of the empirical locational decision.
For similar reasons, we also employed the basic P-median model in
addition to the Supporting P-median model, although the latter better
captures the hierarchical nature and function of the facilities being lo-
cated. The P-median model locates p facilities on a transportation net-
work so as to minimize the total system costs of transporting goods from
the hamlets to the facilities (administrative centers). It may be that the
differences between the hierarchical and nonhierarchical P-median con-
figurations are effectively insignificant. The P-median model is of some
interest in itself as it produces more of a solar configuration of interaction
and facilities, whereas the supporting P-median model produces a den-
dritic pattern of interaction and facilities.
Importance of the Database

There are two general types of input information critical to L-A modeling: (1)
some measure of supply or demand associated with nodes and (2) the network
that connects those nodes. These define the environment the model optimizes
within and thus the resulting solutions. These solutions are, in fact, "optimal"
CIS-Based Analysis of Late-Horizon Settlement Patterns 165
only to the degree that the measures of supply/demand and network accessi-
bility hold true (once we accept the model as appropriate). As no doubt will
prove common in archaeological modeling, these last two critical elements for
L-A modeling must be estimated for the Late Horizon Temascalapa-
Teotihuacan system (Bell et al. 1988: 167). In the present application, the
major role of the GIS will be in creating, supporting this information and
making it available.
Measures of Supply or Demand
For this research, we investigate the sensitivity of model results to the esti-
mate of supply or demand using three estimates of surplus agricultural pro-
duction. Using agricultural surplus production as a surrogate measure for avail-
able agricultural tribute is common (Bell et al. 1988; Parsons et al. 1982; Sand-
ers and Santley 1980). Such a surrogate is intuitively reasonable although not
perfect. Conceptually, determining surplus agricultural production is straight-
forward: simply subtract local consumption from local production.
One estimate of surplus agricultural production we used was that of Bell
and coworkers (1988). Theoretically, they estimated the agricultural surplus
production at a site as follows: the gross yield associated with a site was com-
puted by multiplying the estimated number of food producers (in production
units) per site, the estimated productivity levels of the land available to it, and
the approximate amount of land that each production unit would have farmed.
The local residents' subsistence needs were then computed by multiplying the
local population by 160 kg/year/person (Sanders and Santley 1980). Surplus
agricultural production was the gross agricultural yield at a site minus the
local subsistence demand. Practically, what Bell and colleagues (1988) did was
to adjust some measure of estimated surplus production by the site-type
(Church, personal communication, 1991), retroactively adjusting for the elite
population (Bell et al. 1988).
The overall structure of our alternative estimates of agricultural surplus
production defined as AMB—from Agricultural productivity, Modified: Best
first rule—is the same as that of Belland associates (1988): subtract local con-
sumption from local production. Many of the component estimates and as-
sumptions are shared. The differences between our estimate of agricultural
surplus (AMB) and that of Bell and colleagues (1988) are concentrated in how
land and agricultural productivity were associated with individual sites:
1. AMB assigns Evans's (1980a) productivity land classes to zones through-
out the Temascalapa-Teotihuacan region consistently, whereas in Bell
and colleagues the estimates of productive land class in Temascalapa were
of a different quality than those in Teotihuacan.
2. AMB is organized around the explicit construction of discrete hinter-
lands about sites—the assignment of hinterlands is explicit, whereas in
Bell and colleagues the assignment of both productivity class and amount
of land to a site is unclear.
3. Furthermore, AMB explicitly assigns lands of varying productive
quality to a single site, whereas in Bell and colleagues it was unclear
how this was done.
4. Finally, both land and labor can be limiting factors in AMB's cal-
culation of gross agricultural yield. This was not the case in Bell and
colleagues.
Modified Thiessen polygons were the first-cut estimate of how much total
land would be available for the exclusive use of any site. Any unbounded
Thiessen polygons were cut off at the survey limits set by Sanders and co-
workers (1979) to account for possible competition from exterior, unmodelled
sites. Evans's (1980a) land classes had previously been assigned to the environ-
mental zones from Sanders and coworkers (1979) and sites identified as hav-
ing chinampas in Bell and associates (1988) had their entire territories con-
verted to the chinampa productivity class (Figure 9.2). The amount of land of
varying quality that each site had available for its exclusive use was then esti-
mated. The total amount of labor these different quality zones could absorb
was calculated as was the productive labor available to the individual sites us-
ing the appropriate methods from Bell and coworkers (1988).
The labor available to a site was allocated to the available land according to
a set of rules that took into account the productivity of the land, the amount of
land each production unit could cultivate, and the finite availability of both
land and labor. The rule we chose was to first allocate some labor to land
immediately adjacent to the site no matter what land class it was. The remain-
ing labor was allocated to land according to a most-productive-land-first rule
until either the land or labor available to a site ran out (see Figure 9.3). The
estimate of gross productivity associated with a site was calculated on the basis
of these allocations. The space-filling Thiessen polygons defined a site's po-
tentially available land and allowed land to be a limiting factor in production.
The allocation of labor to areas of different productivity defined what land the
site actually cultivated, and this allowed labor to be a limiting factor in pro-
duction as well as make explicit the allocation of land of different quality to a
single site. The rest of the calculations involving local consumption and surplus
production were performed exactly as in the work of Bell and colleagues (1988).
A third surrogate measure of available tribute—or at least another estimate
that should be highly correlated with "productivity"—was agriculturally pro-
ductive labor [derived from Sanders et al.'s 1979 single basic value (SBV) esti-
mate of population]. The results of the L-A models run on this measure may
not be easily interpretable in themselves, nor will comparing the L-A results
run on this measure and those run on the alternative measures be unambigu-
ous. However, as this measure is a component in the construction of the ear-
lier ones, it can serve as a check on model stability and may prove valuable in
understanding the differences between the other estimates.
Figure 9.3 Flowchart showing the method of land and labor allocation during the
construction of the AMB estimate of total agricultural production.
Network Approximations
The definition of arcs and the creation of a network is another of the basic
requirements of L-A modeling. A network is critical to L-A models because,
as it determines accessibility, it directly influences measures of efficiency and
optimality. Modeled system solutions can be embedded in network assump-
tions. To use exactly the empirical network (even if we knew it completely)
would slightly bias the L-A model toward choosing empirical sites as optimal
(assuming we correctly specified the process). By the same token, any arbi-
trarily defined network also contains its biases toward site location, although
they will not be specifically for or against empirical sites.
As no doubt will prove common in archaeological applications, little is known
of the exact configuration of the network in the Temascalapa-Teotihuacan area
(Belletal. 1988: 170). Gonzalez's (1973) reconstruction of the Aztec road net-
work does not extend into our area and only involves major, masonry links,
while Santley's (1986) analysis of the Aztec road network is conceptual. One
obvious method to attempt to reduce the model biases imposed by the net-
work would be by overdefining the network by adding "too many arcs," even
though each arch is essentially an added hypothesis (Bell et al. 1988). This
more continuously connected network would be less constraining for the
modeled solutions. Such an approach would be more an attempt to provide a
large and reasonable set of possible links rather than an attempt to exactly
replicate the empirical network per se. At the same time, the biases created by
using some kind of discrete network should not be ignored, especially when a
discrete road network was known to exist (Bell et al. 1988: 170). The question
of how connected a network estimate should be constructed is not clear. It is
also not clear how sensitive the L-A model solutions are to different networks
and network connectivities. It is this last question that we will attempt to ad-
dress in our sensitivity analysis.
The Delaunay triangulation link series (DTLS) (Boots 1986) is a somewhat
regular ordering of different definitions of geographic connectivity. Boots sug-
gests that the DTLS could be used to identify an approximate trade-off curve
between user and system costs. Beyond providing a series of networks of vary-
ing connectivity, another interest of this set of network estimates in the present
situation is that they all are constructed from the DTLS—and the DTLS is
easily available in many GIS as it is the geometric dual of the Thiessen poly-
gon diagram and is often the underlying structure of a triangular irregular
network (TIN). The alternative networks in the DTLS is as follows (in order of
increasing connectivity—see Boots (1986) for a full description of these networks):
the minimal spanning tree, (MST); the relative neighborhood graph, (RNG);
the Gabriel graph, (GG); and the Delaunay tessellation (DEL).7 The RNG was
found to be the same as the MST in our case and was dropped from further
analysis. This left the MST, GG and DTLS as our set of alternative networks.
One problem with the above series of network approximations is that they
are completely isotropic and geometric; they do not take into account the
topography of the region. We chose not to calculate distances upon a topo-
graphic model of the region because the digitizing necessary would have been
substantial and the resulting topographic model not that good because of the
coarse resolution of our available base maps. We did, however, try to account
for topography heuristically by modifying the GG network (chosen because
of its intermediate connectivity) by eliminating those links that were unfea-
sible or implausible because of topography and those that were redundant and
did not result in a great savings in total distance.
CIS-Based Analysis of Late-Horizon Settlement Patterns 169
Additionally, the DTLS series does not take into account the tendency to
more directly link large centers. A final network was therefore constructed
from three minimal spanning trees, the standard MST, another that mini-
mized the number of times contour lines were traversed by any link (to ac-
count for topography), and a final tree that minimized the distance between
larger centers (population size based on their agriculturally productive popu-
lation). This network also required some trimming of implausible or infea-
sible links (POPNET).
Therefore, we have a total of five alternative networks, three constructed
directly from the DTLS and two heuristically modified networks. All of these
are ultimately based on the DTLS. Note that while these last two networks
may be more realistic in that they either take into account topography or more
directly link high population centers, they are not necessarily realistic in terms
of exactly replicating the empirical Aztec network, nor are they meant to. Our
network are, in order of increasing connectivity:
MST < GG < MGG < POPNET < DEL
These networks had to be connected to Texcoco in order to run the sup-

porting P-median model. A point-estimate of Texcoco's location was chosen
from the digitized SPS maps of the Basin of Mexico and that node linked to a
set of reasonable sites in Teotihuacan. The straight-line distance to Texcoco
was known for a few of the empirical centers (Evans 1980a). In general, the
straight-line distances from our estimate of Texcoco's location were similar
(within a kilometer or so) to the distances calculated by Evans.
Comparison of Patterns of Administrative Centers and Conclusions
In total we have proposed two alternative L-A models to the Maxcover model
used by Bell et al. 1988 , the simple P-median model and the supporting P-
median model. We expected the latter to better capture the hierarchical na-
ture of the settlement system and to better approximate the hypothesized den-
dritic interaction of the Aztec system (Bray 1980; Santley 1986). Since the
appropriate value of a (or reduction in travel cost between intermediate and
higher-order administrative centers) is unknown, we solved the supporting P-
median model using a values of 0.2 5, 0.50 and 0.75. We also solved the model
using an a value of 0.00; this is equivalent to solving the P-median problem. In
order to assess the nature, relative performance and robustness of the result-
ing L-A derived "optimal" solutions to uncertainties in the input data, we have
developed three alternative productivity estimates and five network approxi-
mations and solved each model upon all possible combinations of productivity
estimate and network approximation.
Our comparison of modeled and empirical settlement patterns is based upon
the relative performance of configurations as measured by the total weighted
distance of the least-cost solution for transporting surplus production from
the areas of production to its "final" destination which could be either a local
regional center or Texcoco. Surplus production could be transported directly
to Texcoco in the latter case or via a regional administrative center in order to
take advantage of the decreased transportation costs represented by a. This
aggregate transportation measure can be computed for any pattern of facilities
by applying a consistent criterion to assign hamlets to regional centers. De-
spite the implicit spatial nature of this measure (as it creates regional assign-
ments of hamlets to centers), this measure evaluates alternative configurations
of facilities in terms of their aspatial levels of performance; very different spa-
tial patterns may attain similar levels of performance.
We compared the performance of three alternative configuration types:
the empirical Aztec settlement pattern and the idealized patterns created by
the P-median and supporting P-median models. The performance of the em-
pirical and P-median configurations was computed under all values of a so
that all configurations could be evaluated under conditions where tribute ship-
ments were made to the final destination of Texcoco. This measure was cal-
culated for all combinations of network and surplus estimates. We did not
need to recalculate the performance levels of the supporting P-median con-
figurations at these different levels of a because they were simply the models'
objective function.
The most basic comparison is of the modeled and empirical configurations'
relative performance under a common network approximation and produc-
tivity estimate (e.g., comparing all the modeled and empirical configurations
using the AMB productivity estimate and the MGG network). This compari-
son can be seen in Figure 9.4, for all combinations of a, productivity estimate,
and network approximation. In these diagrams we have adopted the support-
ing P-median performance as a "standard" level so that comparisons can be
more readily made between productivity estimates and a conditions with dif-
ferent levels of total possible weighted distance. The supporting P-median
level of performance was chosen strictly as a matter of convenience. At times
during the following discussion the Aztecs may be described as "improving"
their performance vis-a-vis optimality, but this is strictly because of the
structure of the diagrams.
The various combinations of network approximation and a level are repre-
sented along the x axis of these diagrams. The alternatives are organized ac-
cording to network approximation (from most to least connected approxima-
tion) and within these divisions according to a value (in ascending order). There
is a different diagram for each productivity estimate. The performance of the
supporting P-median configurations are equivalent to the x axis because they
are the baseline for numerical comparison, but they are nonetheless indicated
by the line of rilled triangles. The performance of the other configurations
vis-a-vis this baseline are measured on they axis (in units described below).
The Aztec configuration performance is represented by the line with the filled
square and that of the simple P-median by the line with the open diamond.
Two different measures of difference from the baseline measure (support-
ing P-median optimality) are used in the following evaluation: the difference
measured directly in weighted distance (on the left); and the difference mea-
Figure 9.4 Comparison of configuration types in terms of performance.
sured proportionally (on the right). The method of solving supporting P-me-
dian problems and computing hamlet allocations to regional centers results in
higher weighted distance values at higher a. There can therefore be an appar-
ent "improvement" (or decrease in apparent difference from the supporting
P-median baseline) in these measures at higher a due solely to this. This prob-
lem is more evident using the percentage difference from optimality measure.
Although the simple difference in weighted distance measure shares this prob-
lem (a unit difference has different meanings under the different a conditions),
the comparison of these two measures should give a better indication of whether
the apparent improvement is "real" or not.
The difference from optimality measured directly in weighted distance shows

that the absolute difference between Aztec system performance and optimal
supporting P-median performance decreases as a increases. This is to say that
as more surplus is shipped on to a final destination or, as the same amount of
surplus is shipped on to that final destination but with lesser savings due to
bulk shipment, the performance of the two configurations becomes more simi-
lar. This decrease is despite the increasing magnitude of the general weighted
distance values. On the other hand, the simple P-median configurations differ
more and more from the supporting P-median standard when evaluated un-
der the conditions of higher a. Although the simple P-median solution and the
Aztec configuration are very different when compared directly against each
other, they attain similar levels of performance under conditions of high a.
The difference between models measured in percentage difference from
optimality shows the same pattern of greater similarity between the perfor-
mance of the empirical Aztec system and that of the supporting P-median
configurations as a increases. Again, the greatest apparent difference is be-
tween the simple P-median configuration performance and Aztec configura-
tion. The general similarity of the Aztec configuration to the supporting P-
median model solutions is again marked by greater similarity at higher a. Eya
= 0.75 the Aztecs are performing within 10% of optimality, a heuristic stan-
dard often used in planning for determining solutions that do "good enough."
The P-median configurations again show greater differences in relative per-
formance levels vis-a-vis the supporting P-median model at higher a. By a =
0.75, the apparent difference between the simple P-median configuration and
the Aztecs' performance is practically nonexistent.
This general pattern of behavior appears very consistent across networks,
not only in form but also usually in value or magnitude (although this last is
not true for the MST network which has generally higher values). According
to these network estimates and performance measures, network connectivity
does not appear to alter model performance substantially except at very low
connectivity. Although there is some variation between productivity estimates
and networks, this behavior is remarkably consistent and robust. The general
pattern of decreasing percentage optimality and decreasing absolute differ-
ence in optimality indicates that this improvement in performance is real
and not a function of the numbers (supporting P-median distance matri-
ces) involved. Based upon these results, our conclusion would be one or
both of the following:
1. The Aztec configuration of facilities is more consistent with the

ssupporting P-median (dendritic) configuration than the simple P-me-
dian (more solarlike) configuration of sites. The supporting P-median
model is best able to replicate the performance of the Aztec configura-
tion when there exists either a great deal of tribute to be shipped through
supporting centers to some exterior final destination (Texcoco) or when
the savings due to bulk shipping are not great (but goods are still shipped
to that final destination).
2. Because this behavior appears robust across estimates of productivity
and network, we may be able to assert that the Aztec system of adminis-
trative centers modeled appears to be attaining relatively close to opti-
mum performance under conditions where they existed to service a great
deal of surplus tribute that was to be shipped to an exterior final destina-
tion or under conditions where the savings due to bulk shipment to that
final destination were not great.
The Aztec configuration appears to be a robust and fairly efficient pattern

for the collection and transport of substantial amounts of surplus and for the
administration of population when travel between the hamlets, and interme-
diate and superordinate administrative centers is incorporated. This apparent
efficiency of the Aztec system is a new finding, one not found in earlier L-A
based models of the Late Horizon Basin of Mexico (Bell et al. 1988; Gorenflo
and Gale 1986). Our comparison of the relative levels of model and empirical
performance provides additional physical support in favor of the many theo-
retical claims of a dendritic pattern of Aztec settlement (Bray 1980; Santley
1986). Santley (1986) found evidence of a dendritic pattern expressed in the
(largely reconstructed) physical road network, whereas our evidence is in the
more assured locations of regional centers and is less dependent upon a single
reconstruction of the Aztec environment and road network.
We have attempted to illustrate the interest of a combined GIS and L-A
approach and to present an application to a regional settlement analysis of the
Late Horizon Basin of Mexico. The combination of the L-A technique and
GIS technology seems well suited when one considers certain shortcomings
of most GIS based analysis and many of the special requirements of archaeo-
logical applications of L-A models. The data requirements of L-A modeling
in general are considerable and the robust model solutions of interest to ar-
chaeology make even greater demands upon the quality and volume of data
required. Archaeological data sets are hard pressed to provide appropriate in-
formation. GIS, on the other hand, allow for the easy manipulation, creation,
and presentation of spatial data, although they currently have limited model-
ing and analytic capabilities. Combining the two as we have done will allow
one to search and test for robust model solutions. The GIS not only facilitates
this type of L-A analysis and data support but, to a large degree, makes this
type of analysis feasible.
Notes
1. "The model-building approach in GIS as defined here focuses on the development

of models for long-term processes, rather than for the purpose of predicting or under-
standing site locations" (Allen, 1990: 197-199, italics added for emphasis.)
2. For example, Savage (1990b) develops an approach to prehistoric catchment basins
that attempts to take into account topography and ease of travel in order to take his
analysis "out of the isotropic plain" (Savage 1990b: 352) but he resorts to a standard
isotropic Thiessen polygon function to estimate catchment areas.
3. Defined more as progress toward an ideal goal rather than the attainment of that goal.
4. A case made either implicitly on the basis of standard archaeological or anthropo-
logical theory about the cultural system or explicitly (and more satisfactorily) upon the
archaeological and historic record.
5. A not entirely satisfying but commonly accepted generic term for all members of
the triple alliance.
6. This last is important, as the network assumptions behind the BCG88 road network
have the real potential to bias model results.
7. The Delaunay network is not the continuously connected network—not all nodes
are joined by a single arc.
10
The Politics of Settlement Choice
on the Northwest Coast:
Cognition, GIS, and Coastal Landscapes
The reasons why evolutionary ecology and, more specifically, optimal forag-
ing theory, do not work in many archaeological situations are varied. Most
importantly however, is our lack of understanding of basic human decision-
making processes in societies intermediate between bands and states. From
evolutionary ecology, we can predict some foraging behavior and thus explain
some of the settlement behavior of foraging societies (Mithen 1991; Smith
1991). In states and empires, we can use modern microeconomic theory to
predict settlement, trade, and political organization. However, we have very
little understanding of how to predict behavior in societies that fall between
these two extremes.
One of the basic assumptions of modern economic, geographical, and cul-
tural ecological studies is that humans are energy maximizers. Ecologists view
this ability to be economically efficient as a product of our evolutionary his-
tory of being adaptive (Jochim 1981; Krebs and Davies 1991; Smith and
Winterhalder 1992; Stephens and Krebs 1986; Winterhalder and Smith 1981).
Support for this assumption is clearly seen in studies of small, mobile foraging
societies where individuals and kin-based groups tend to maximize their eco-
nomic return with subsistence and settlement behaviors that most would agree
are adaptive in that particular context (Jochim 1981; Mithen 1991; Smith 1991).
For sedentary communities with more complex political organizations (tribes
and simple chiefdoms), however, this is not the case, and this discrepancy is
seen archaeologically in settlement and subsistence strategies that do not con-
form to predictions derived from optimal foraging theory.
Thus, an underlying assumption in ecological studies is that models of sub-
sistence economizing behavior and studies of subsistence efficiency will work
well for hunters and gatherers (Keene 1981; Winterhalder and Smith 1981) or
small-scale horticulturalists (Keegan 1986), but will decrease in their explana-
tory power with increasing social and political complexity. Although this has
not been specifically tested, the fact that optimal foraging theory is less effec-
tive in explaining behavior in agricultural and sedentary hunter-and-gatherer
societies (Maschner 1992) and is not usually applied to chiefdoms and states at
all supports this contention. I contend that the reason for this discrepancy in
the performance of optimal foraging theory is that the psychological adapta-
tions for cost-effective foraging that evolved in and work so well for small,
kin-based, mobile groups, do not work so well in multi-kin-based sedentary
societies. With this change in the social environment, foraging efficiency is
sacrificed for political and social efficiency. This emphasis on the social and
political rather than the economic results in a number of changes. Although
there is no necessary reduction in adaptive tracking efficiency, the overall dif-
ferences in foraging decision making between bands and more complex orga-
nizations indicates that there is a more multifarious decision process in multi-
kin group constructions.
If we create predictive models of archaeological site location based purely
on environmental characteristics, we find that they often fit the observed pat-
tern quite well when applied to mobile foragers. Yet when such predictive
models are attempted for more complex political forms, they are often aban-
doned because the prehistoric inhabitants do not seem to follow the environ-
mental rules established by models of subsistence or general economic effi-
ciency, as has been described by E. A. Smith (1991) and others. This is because
humans, based on their evolutionary history, are not adapted to live in multi-
kin-group sedentary communities. In such communities, social conditions
necessitate that economic efficiency be traded for social and political efficiency
and conflict-minimizing strategies thus assume a more important role than
subsistence-maximizing strategies. This does not mean that archaeologists
should not build the economic predictive model anyway, but simply that they
should not conceptualize these models as an end product. Rather, these mod-
els should be used as screens on which to project the actual settlement pattern
and anticipated behaviors. The deviation from the predicted model will in-
form on areas in which the prehistoric peoples were sacrificing economic effi-
ciency for political and social ends.
This paper will demonstrate that most of the problems archaeologists have
with geographical and ecological models stems from our preconceived expec-
tation that non-industrial societies will behave adaptively, expectations that
are formulated without any understanding of the conditions that regulate adap-
tive decision making. I would, in fact, argue that we have almost no under-
standing of the individual decision-making process— when, for example, eco-
nomic efficiency will be traded off for other environmental conditions or how
spatial perception and cognitive mapping abilities affect this process. Answers
to these questions will be found in the relatively new discipline of evolutionary
psychology, an offshoot of cognitive psychology. Combining evolutionary
models with spatial analysis derived from a geographic information system
(CIS) and building of the work of Carr (1984), Hodder and Orton (1976), and
Kvamme (1983,1985), I will demonstrate that spatial models, combined with
extensive field research and modern evolutionary theory, leads to interpreta-
tions of prehistoric human behavior not previously possible. This new ap-
proach will be applied herein to settlement data collected by the author on
Kuiu Island, southeast Alaska (Figure 10.1).
Cognition, GIS, and Coastal Landscape, in
Figure 10.1 The northern Northwest Coast.
Evolutionary Psychology, Cognition, and the Landscape

Evolutionary psychology is a new discipline aimed at searching out and study-
ing human psychological mechanisms that regulate and structure human be-
havior. Evolutionary psychology does not have as its goal the study of cultural
traits that increase fitness, nor is it guided by the assumption that human be-
havior is inherently "adaptive." According to Tooby and Cosmides, two of its
foremost proponents; "die promise of the evolutionary perspective lies in-
stead in its power to assist in the discovery, inventory, and analysis of innate
psychological mechanisms.... By directly regulating human behavior and learn-
ing, these mechanisms directly govern cultural dynamics..." (1989: 30; see also
Barkow 1989; Blurton-Jones 1976; Cosmides and Tooby 1987; Daly and Wil-
son 1988; Tooby and DeVore 1986). They go on to state that "Natural selec-
tion defines information processing problems the organism must be able to
solve (in a given adaptive context) and cognitive psychology now has methods
available to assist in discovering what algorithms exist in the psyche to solve
these problems" (Tooby and Cosmides 1989: 30).
Tooby and Cosmides (1989: 3 0) conclude their justification by arguing that
"Application of these methods will allow the exploration of design features of
innate human psychological mechanisms, including the design features of those
learning mechanisms that create and maintain cultural phenomena" (Tooby
and Cosmides 1989: 30).
A considerable body of psychological research has revealed that even the
simplest cognitive tasks require "innate" abilities in their solution (Chomsky
1975, 1980; Fodor 1983, 1985). Given the complex nature of the tasks hu-
mans perform as a part of daily survival, the innate psychological structures
of the human brain can be expected to be very complex in design and organi-
zation (Tooby and Cosmides 1989: 34). It is reasonable to assume, and
paleontologically verifiable, that complex physical structures, such as the eye
and hand, developed over extremely long periods of time. The complex na-
ture of the human psyche similarly suggests a long period of development
and, like the hand, probably assumed approximately modern form long ago
as a consequence of selection on a Pleistocene landscape (Tooby and Cosmides
1989: 34). Of particular importance were the development of "general solu-
tions," such as cost-benefit algorithms—mechanisms applicable to varying
local conditions (Tooby and Devore 1987). Ultimately, the psychological
mechanisms underlying cultural behavior came to be so powerful that they
created culture change as a historical rather than evolutionary process that, as
early as the Neolithic, began changing human social organization at a rate far
beyond natural selection's ability to track (Tooby and Cosmides 1989: 35).
Therefore, we have no a priori reason to assume that any modern behaviors
are adaptive in their modern context (Barkow 1989; Symons 1989; Tooby and
Cosmides 1989) or that "modern cultural dynamics will necessarily return
cultures to adaptive trajectories if perturbed away" (Tooby and Cosmides 1989:
35). Although there has been enough continuity in many cultural systems to
argue that many components of culture do function to modify behavior in
adaptive ways, there are no reasons to assume that this is generally so (Tooby
and Cosmides 1989: 35).
In essence, the human brain and the components of the human psyche are
products of a long evolutionary history of solving problems encountered on
the Pleistocene landscape. From an archaeological perspective, we can assume
that certain conditions—such as small group size, kin-based group composi-
tion, a high degree of mobility, and foraging—were present for over 3 million
years. With the rise of the Neolithic (for example), when humans moved for
the first time into formations that contained multiple kin groups, sedentism,
and often a non-foraging economy, culture change escalated to a rate greater
Cognition, GIS, and Coastal Landscapes 179
than selection's ability to keep pace. We are thus organized around a Pleistocene
psychology and an urban sociology; there is therefore no reason to assume that humans,
in a modern or recent cultural context, behave adaptively. This has enormous im-
pact on the investigation of the human decision-making process and thus our
ability to create accurate models of settlement behavior. Given this theoretical
perspective, I will demonstrate that there is reasonably strong evidence to in-
dicate that in the transition from lineage- or kin-based villages to multilineage
or non-kin-based villages on the Northwest Coast, site locations changed from
locales that were more optimal for resource exploitation to locations that were
optimal for defense, a proxy measure of political optimization.
A Northwest Coast Example

Background
My research has been conducted in Tebenkof Bay, located on the west side of
Kuiu Island in southeast Alaska, on the Northwest Coast of North America.
Historically it was one of the main territories occupied by the Kuiuquaan
Tlingit. The Tlingit are a group of politically complex hunters and gatherers
that lived in large, sedentary villages; they had part-time craft specialists, insti-
tutionalized slavery, a complex system of warfare for revenge and territory,
and a subsistence strategy that was largely based on the yearly salmon run.
Despite a large and diverse ethnographic and ethnohistoric literature, very
little is known about prehistoric Tlingit settlement and subsistence, particu-
larly in relation to the decision process that governed habitation site selection.
This is the problem I chose to address by integrating geographic information
systems, evolutionary theory, and field survey.
According to historic records on the Tlingit, a number of criteria that
were critical to the selection of habitation site locations (de Laguna 1990; de
Laguna et al. 1964; Emmons 1991; Krause 1970; Langdon 1979; Niblack
1970; Oberg 1973). These included southerly exposures, sheltered coves,
beaches of pebbles and small rocks so as not to damage the cedar canoes,
level and well-drained terrain, and defensibility. Proximity to food re-
sources was also important.
Three seasons of archaeological survey and an intensive excavation pro-
gram revealed over 150 archaeological sites along 350 km of shoreline in
Tebenkof Bay. One of the primary goals of the survey was the construction of
an archaeological model of site location that would account for the variability
expressed in the locations of habitation sites. Models of site location have not
been a major emphasis in Northwest Coast archaeology. Hobler's study of the
relationship between site location and salmon harvesting areas met with only
limited success (1982). Acheson identified a number of geographic and envi-
ronmental conditions that seemed to correlate with site location in the Kungit
Haida territory of the Queen Charlotte Islands. He found that sites were of-
ten located on the exposed outer coast, but in locations that were not exposed
to the primary surf (1991). He also argued that village sites were distant from
salmon streams and had very poor freshwater sources (1991). C. Rabich-
Campbell's study of site location in southeast Alaska (1980) is the first attempt
to quantify the environmental determinants of site location. Using univariate
tests, she found a preference for southward exposures, sand or pebble beaches,
ice-free zones, protection from southeasterly storms, and fresh water (1980:
17), yet she was unable to account for the considerable variability in the data
presented for her sample of twenty four house-depression villages.
Of the sites identified on the Tebenkof Bay Project, ninety four were
identified as villages or temporary habitations. Habitation sites were iden-
tified on the basis of either house depressions or large accumulations of
shell middens (Figure 10.2). Parameters were recorded for each site on
the survey, with the data being collected either in the field or through the
project GIS and aerial photographs. These variables included cardinal
exposure, island size, climatic exposure, beach quality, slope, drainage,
distance to water, and the cultural variable, site area.
Figure 10.2 Site distributions in Tebenkof Bay, Kuiu Island.

Data Collection
The field survey consisted of surface reconnaissance combined with subsur-

face testing with augers, soil probes, and shovels. Site area was determined
based on the systematic probing for subsurface shell-midden deposits. Data
were collected on grade, drainage, beach quality, vegetation, distance to fresh
water, climatic exposure, solar exposure, resources (shellfish beds, salmon
streams, berry patches, etc.), and cultural features. A GIS was constructed on
a SUN Unix work station running Arc/Info. Elevation data were digitized
from 1:63360 USGS topographic maps; shorelines were digitized from 1:63 3 60
color infrared aerial photographs, and beach quality, for areas without archaeo-
logical sites, was recorded from 1:10,000 black-and-white aerial photographs.
Vegetation zones and some other landscape features were already in Arc/Info
format and were provided by the Stikine area of the Tongass National Forest.
The combination of field data and the GIS allowed us to not only check and
evaluate the field recording system but also to gather data on a number of
nonsite locations to be used as random points in the following analyses.
The categories used for each variable are important to this discussion. Expo-
sure was recorded as simply south-facing (west, southwest, south, southeast) or
north-facing (east, northeast, north, northwest). Exposure was recorded in the
field but was also checked using the project GIS. Island size was done in the GIS
and categorized as large (a site on Kuiu Island itself), medium (a site on one of the
larger, 0.5 to 3 km2 islands in the bay), and small (really an islet or island < 0.5
km2). Island size data were collected because many of the small islands do not
have a fresh water source, and I was interested in the characteristics of site loca-
tion in the absence of potable water. Climatic exposure was categorized as com-
pletely sheltered, seasonally exposed to storms and/or wind, and usually exposed
to storms and/or wind. Climatic exposure data were created in the GIS using
prevailing wind data and the seasonal direction of storms. Beach quality was iden-
tified as four categories: sand and pebbles, small rocks and gravel, large rocks and
boulders, and cliffs and breakwaters. Beach quality data were collected in the
field and checked with aerial photographs. Drainage was determined both from
field observations and remotely sensed images and was recorded as well-drained
or poorly drained. Slope was reconstructed using the GIS and recorded as less
than or greater than 5% slope. Distance to fresh water was a metric measure-
ment in the field and categorized for this study as < 10 m from the stream or
spring, between 10 and 100 m from the source, and > 100 m or no water. Water
distance could not be recorded from either images or the GIS because of the
numerous small springs that are not identifiable in either format. Site area was
also recorded in the field and was used in this study as < 400 m 2 ,400 to 1000 m2'
and > 1000 m2. As the only purely archaeological variable in the study, area al-
lowed me to determine if the intensity of occupation of a particular location (site
area as a proxy measure of intensity of occupation) could be used in determining
site location. From the beginning it was recognized that there should be a strong
relationship between island size and distance to fresh water, because many of the
smaller islands tend not to have sources of fresh water.
Analysis
In a recent paper modeling site location variability on the Northwest Coast

using the Tebenkof Bay data, Jeffrey Stein and I demonstrated that there were
clear patterns to the kinds of environmental characteristics chosen for pre-
historic habitations (Maschner and Stein 1995). Through logistic regression
analysis, we built a model of site location that explained over 90% of the
variability in the environmental data based on beach quality, drainage, aspect,
island size, grade, and climatic exposure. Thus, we found that habitation sites
were generally located on pebble or gravel beaches for boat landings, that
most sites faced south, that the site landscape was well drained, and that the
sites were reasonably sheltered from intermittent storms (Maschner and Stein
1995). This was done as follows.
We began with a known population of sites (n = 94 with one rejected as
an outlier) and a randomly derived sample of nonsites (n = 100 with one
rejected as an outlier). Next a logistic regression model, following Kvamme
(1985) and Warren (1990), was created for the prediction of site locations.
The most useful model constructed included the variables cardinal expo-
sure, climatic exposure, grade, drainage, beach quality, and island size. As
we stated in our original presentation, for model-validation purposes, 75
of the 194 observations were randomly excluded before any of the models
were fit. Then a training sample of 119 observations was used to create an
appropriate predictive model and the 75 randomly excluded values acted
as an "independent" data set with which to validate the model. Although
the number 75 was arbitrarily chosen, we wanted to ensure that the vali-
dation sample was large enough to fit a logistical model with nine param-
eters (Maschner and Stein 1995).
Using cardinal exposure (Xt), beach (X2), climatic exposure (2Q, island
type (X4), grade (Xy), and drainage (X6) as the six covariate effects, the follow-
ing logit model was fit to the training sample observations:
where the probability that a site is present, given the covariate vector x, is
denoted asp(x) ) (Maschner and Stein 1995).
The data fitted to this model exhibited a highly significant goodness-of-fit
statistic of G2 = 74.046 on 6 d.f. with a corresponding/) value of < 0.0001.
Because of the strong association between level, well-drained terrain and ar-
chaeological site locations, we found the overall fit and predictive capability
of this six-covariate model to be optimal for site location prediction. The
training sample fitted to this model also exhibited a highly significant good-
ness-of-fit statistic of G 2 = 114.43 on 8 d.f. (p value < 0.0001). Table 10.1 lists
the classification table which was obtained after fitting the validation sample
to the six-covariate model (using a cut point probability of 0.50). Both the
sensitivity (92%) and specificity (82%) are high, yielding an overall classifica-
tion rate of 87% (Maschner and Stein 1995).
Table 10.1 Classification table for model with six covariates [adapted from Maschner
and Stein (1995, Table 5)].
Predicted Percent
Observed Site Present Site Absent Correct
Site Present 34 3 91.89 I
Site Absent 7 31 81.58 1
Overall 86.67
Pairwise interactions between all main effects were tested and found to
be statistically insignificant. Although we expected a significant interac-
tion between grade and drainage to be present, it did not prove to be the
case. With an underlying goal of parsimony in mind, we were comfortable
with the inclusion of only main effects. Assuming the model is indeed of the
correct functional form, it is evident that sites are most probable on level,
well-drained terrain on smaller islands with at least some form of shelter from
the storms. Sites are also likely to be found near sand and pebble beaches as
opposed to shorelines with rocks and cliffs. It is clear from these parameter
estimates that the environmental characteristics of sites and random points
are different (Maschner and Stein 1995).
We then used log-linear analysis to model the specific elements of our
environmental variables that combine to influence site location. Since we were
not using a nonsite test sample for log-linear modeling, and because grade
and drainage were univariate in the sense that all sites were on level, well-
drained surfaces, we used only the variables climatic exposure, island size,
cardinal exposure, and beach quality. To these we added distance to fresh
water, which was recorded in the field for every site, and site area; this is the
only archaeological variable in the model and is used here as a proxy measure
of intensity of occupation. From a simple perusal of the data using only four
of the variables, it can be clearly seen that there are certain environmental
characteristics of site location that seem to cooccur (Table 10.2).
Table 10.2 Cross classification of a sample of 93 archaeological site locations

according to (1) island size (2) beach quality (3) cardinal exposure and (4) climatic
exposure [adapted from Maschner (1992: 228, Table 7.1) and Maschner and Stein
(1995, Table 6)].
Sheltered from Storms Seasonally Exposed to Storms

Southern Northern Southern Northern
Island Size Beach Quality Exposure Exposure Exposure Exposure
Large Sand/Gravel 23 8 8 0
Rocks/Boulders 0 2 0 1
Medium Sand/Gravel 14 6 2 1
Small Sand/Gravel 9 3 6 1
The log-linear model, denoted by [12 3] [15] [4] [6] (where 1 = island size, 2 =
beach quality, 3 = solar exposure, 4 = climatic exposure, 5 = distance to water,
6 = site area, plus their interactions), is expressed as:
assuming the usual ANOVA-like constraints.

The model's close fit is confirmed by the goodness of fit statistics G2 =
54.73, df = 83, P value = 0.993 and c2 = 57.62, df = 83, P-value = 0.985. This
demonstrates, as we argued in the original study (Maschner and Stein 1995),
that there is a strong relationship between certain environmental variables
and the distribution of archaeological sites in Tebenkof Bay, southeast Alaska.
The log-linear analysis shows a strong association between cardinal exposure,
beach quality, and island size. With decreasing island size, the prehistoric de-
cision makers increasingly traded off favorable characteristics in order to pur-
sue whatever benefits they hoped to gain from landing at those locations. Cli-
matic exposure was shown to be completely independent of the other three
geographic variables but equally important in determining the location of a
site. Fresh-water access was also important but not mandatory, as demonstrated
by the number of sites on small islands that, at least today, have no fresh water
sources. Sites on large and medium-size islands, however, exhibited a large
negative relationship to limited water accessibility, as was expected.
Site area is a useful indicator of the occupational intensity at particular sites.
Upon examination of this model, we found that site area is completely inde-
pendent of the other five variables and their interactions. This implies that the
intensity of occupation by the prehistoric inhabitants was not conditioned by
any particular one of these factors. The parameters that did influence duration
of occupation at these sites (whatever they are) are not covered by the data.
Thus, we clearly demonstrated that there are specific characteristics to site
location that were being selected by the prehistoric and early historic decision
makers in Tebenkof Bay over the last 5000 years. These included south-facing
pebble beaches, shelter from storms and surf, and level, dry landscapes. We
also demonstrated that beach quality, cardinal exposure, and island size are
structurally related, but site area, as a proxy measure of site occupational in-
tensity, is independent of these factors. The relationships between site loca-
tion and the purely environmental variables is so strong as to make the predic-
tion of site locations feasible in this area.
Social and Political Factors in Site Location
Having identified a powerful relationship between certain environmental char-

acteristics and the locations of archaeological sites in Tebenkof Bay, I will
now focus on analyses that I performed using these models as a screen onto
which I could project variability within the data. It is hoped that once the
environmental models are secure, political and social variability within the
models can be investigated.
The first pattern concerns a core-periphery effect that is seen in the distri-
bution of the habitation sites. A simple perusal of Figure 10.2 makes it obvious
that the shell-midden-based habitation sites tend to be densest in the center of
the bay, becoming scattered with distance. At the heads of the bays and along
the arms of the channels, shell-midden sites are rare. In these areas, fish camps,
weir complexes, or very temporary camps make up the site complement.
There are several possible reasons for this distribution. Based on the distri-
bution of winter ice zones, this pattern could be purely functional. Ice forms
periodically between December and February at heads of the long coves and
along the deep channels in Tebenkof Bay. Although it generally does not last
for more than a few weeks, we might expect that habitation in these areas would
be problematic because water travel would be difficult during cold periods.
Seasonal differences in resource extraction could also result in this pat-
terning. As stated above, the only sites in the deep coves are fish camps
and weir complexes associated with the harvesting of salmon. The Tlingit
did not commonly harvest shellfish in the summer (Moss 1989), resulting
in the lack of shell middens in these areas, a defining characteristic of
habitation sites for this study.
But when the sites are broken down further, the implications of these pat-
terns become more clear. Habitation sites were classified according to settle-
ment type following a system developed by Acheson (1991). Permanent vil-
lages were characterized by either the presence of house depressions, or by the
presence of shell middens greater than 1000 m2. This reclassification made it
apparent that the more substantial sites are all located in the central part of the
bay, in areas with fairly equal access to most resources—perhaps all within a
half-day canoe trip.
A temporal breakdown of the village site distribution provides further in-
sight. We can look at settlement variability among the largest sites, primarily
the differences between house-depression villages, all dating after A.D. 500,
and the large shell middens, all probably dating between 1500 B.C. and A.D.
500 (based on radiocarbon determinations on three of the seven). Figure 10.3
shows the distributions of these two site types divided, with some presump-
tion (because not all are dated) into Middle and Late Phases. It is obvious that
the house-depression villages are primarily concentrated in the north-central
part of the bay, while the temporary villages are located in the center of the bay.
These changes occur within the parameters of the logistic regression and
log-linear models described above. All villages are located on level, well-drained
landscapes that face somewhere between west and south and have beaches
composed of small pebbles and gravel for landing boats safely. The differences
lie in climatic exposure and access to resources, as well as a number of qualita-
tive factors that were recorded in the field but which, due to the scale of the
data, were not measurable in the GIS. The Middle-Phase villages are gener-
ally found on highly convoluted shorelines with excellent intertidal resources.
Figure 10.3 Distributions of village sites and defensive sites.
Being in the middle of the bay situates them with equal access to both salmon
streams and the resources of the outer waters—particularly bottomfish and
sea mammals. The house-depression village sites of the Late Phase are more
often exposed to wind and surf, being situated on long, straight shorelines.
Although they are generally closer to the primary salmon harvesting streams,
they have much poorer intertidal resources.
Why would we expect to see a shift from resource-dense and energy-opti-
mal locations to more exposed shorelines? As can be seen in Figure 10.3, there
is a general relationship between the distributions of Late-Phase villages and
that of defensive sites suggesting that defensibility was a more important issue
in the Late Phase. The defensiveness of the village locations is difficult to
evaluate, yet a proxy measure of security is possible. In a GIS it is possible to
compute the viewshed of any point. This was done to compute the total area
of surface water visible from a particular village. I would argue from the eth-
nographic data that villages located in order to be defensive would require a
commanding view of the bay from the village itself. This would provide a sort
of early warning system to prepare for attacks, since due to the forest-choked
hinterland, threats would most likely come over water.
Table 10.3 shows the differences in surface-water visibility between the
Middle-Phase villages and the Late-Phase house-depression villages. Although
the standard deviations are large and the samples small, house depression vil-
lages have nearly three times the surface water viewshed of their non-house-
depression counterparts. This would make them much more defensible but at
the same time more exposed to wind and surf.
Fundamental to understanding this change is the occupation sequence of
these two lands of sites. From the data at hand, it appears that most of the
Middle-Phase villages were occupied simultaneously, while the Late-Phase
house-depression villages were occupied sequentially (Maschner 1992). That
is, there is a transition from a number of simultaneously occupied small vil-
lages to one large village that was periodically relocated.
If we look at the size of the villages in relation to the number of houses
and then compare this ratio with the size of the Middle-Phase villages, an
interesting pattern emerges. Of the seven villages with house depressions
in the bay, there are a total of twenty two recognizable houses. These
houses have a total of 51,480 m2 of site area or approximately 2340 m2 per
house. The Middle-Phase villages total 15,363 m2 of site area for a mean
of 2194 m2. Table 10.4 presents a £-test demonstrating clearly that these
are not significantly different.
We know from historic accounts that large villages were traditionally or-
ganized with one lineage per house (Emmons 1991). If, as I suspect, this was
true prehistorically, then Middle-Phase (non-house-depression) villages were
probably single-lineage residences and large, Late Phase villages were amal-
gamations of these small, Middle-Phase communities. This pattern implies a
fundamental transition in settlement from the Middle to the Late Phase. The
village settlement pattern shifts from one based on the distribution of key
resources to one based on defensibility. This includes the transition to more
defensive locations, the creation of larger corporate entities, and the estab-
lishment of defensive constructions. In evolutionary terms, this is seen as a
transition from an economically maximizing settlement pattern to a pattern
of political maximization.
Table 10.3 T-test of surface water area visible from village locations [adapted from
Maschner (1992:286, Table 7.27)].
Period Count Mean (m2) Std. Dev. (m2)
Late Phase Villages 7 9233053.30 380221.36
Middle Phase Villages 7 3855025.40 321375.09
Separate Variances t=2.86 df=11.7 probability=0.01
Pooled Variances t=2.86 df=12 probability=0.01
Table 10.4 T-test of site area in temporary villages to the ratio of site area to houses in
house depression villages [adapted from Maschner (1992:288, Table 7.28)].
Period Count Mean (m2) Std. Dev. (m2)
Late Phase Villages 7 2340.0/house 640.6
Middle Phase Villages 7 2194.0/site 1314.9
Separate Variances t=.277 df=11.7 probability=0.787
Pooled Variances t=.265 df=12 probability=0.787
Conclusion
In this paper I have attempted to integrate evolutionary theory, field survey,
GIS, and multivariate statistics to explain prehistoric settlement and settle-
ment change in Tebenkof Bay, southeast Alaska. Through this study and pre-
vious work (Maschner 1992; Maschner and Stein 1995) it was demonstrated
that the basic environmental characteristics of beach quality, climatic expo-
sure, cardinal exposure, and island size explain most of the variability in ar-
chaeological habitation site location. This is not surprising, since these were
among the main considerations for site location among the historic Tlingit. It
was further shown that these variables do nothing to explain site area, used
here as a proxy measure of intensity of occupation.
These environmentally based patterns, founded purely on observable fea-
tures of the landscape, were shown to be useful as a screen onto which settle-
ment patterns that might be more socially or politically significant could be
projected. This patterning can be seen quite clearly in the transition from
small, probably semisedentary, lineage-based communities to large, sedentary,
multi-kin-group amalgamations in the Middle to Late-Phase transition. Here
I found that in the transition from single-lineage to multilineage settlements,
the characteristics of site location shifted from very convoluted coastlines with
excellent shellfish-gathering areas and equidistant access to most of the re-
source zones in the bay, to locations with long, straight shorelines providing
excellent open-water views but also greater exposure to surf and storms and
poor shellfish-harvesting areas. Further, palisaded forts and other defensive
constructions are added, which is a powerful implication that foraging effi-
ciency and economic diversity are being sacrificed for political efficiency.
This is a condition that can be predicted from evolutionary psychology.
Humans are not, given their evolutionary history, adapted to living in multi-
kin-group conglomerates. On the Northwest Coast, it appears that multi-kin-
group conglomerates have been present only for the last 1500 years, clearly
not enough time for natural selection to have any impact on the human phe-
notype. Thus, when people are forced to live in larger amalgamations, much
greater emphasis is placed on interpersonal interactions, including feasting
activities (Chagnon 1983; Hayden 1990), alliance through trade and exchange
of valuables and luxury items (Arnold 1987), and warfare (Chagnon 1988; Lam-
bert 1993; Lambert and Maschner 1993; Maschner 1992, 1993), conditions
that will probably be present in all midlevel societies between bands and
states (Gregg 1991). In these cases predictive, maximizing, and other op-
timal models should be seen as screens on which we can project the actual
data. The degree of deviation from the predicted model will be the the
intensity of effort being put toward political and social efficiency by the
prehistoric decision makers.
Acknowledgements
I would like to thank Jeff Stein, Patricia Lambert, Mark Aldenderfer, and Michael
Goodchild for comments on early versions of this paper. Jeff Stein and I did most of
the statistical analyses together and his work is credited for the discussion of the logis-
tic regression procedure in this paper. Funding for the Tebenkof Bay Archaeological
Project was provided to Maschner through a Wenner-Gren Foundation for Anthropo-
logical Research Predoctoral Grant, a National Science Foundation Doctoral Improve-
ment Grant, the Stikine Area of the Tbngass National Forest-Petersburg, the Univer-
sity of California-Santa Barbara, the Alaska State Historic Preservation Office, and
Sigma-Xi Grants-in-Aid of Research. I would like to thank Mark McCallum of the
Tongass National Forest (USDA) for making my research possible. The statistical analy-
ses and GIS construction were done in the Social Science Computing Facility (SSCF)
at the University of California-Santa Barbara and a special thanks is given to Joan
Murdoch and Mark Schildhauer, codirectors.
11
The Role of GIS in the
Management of Archaeological Data:
An Example of Application for the
Spanish Administration
Concepcion Blasco Bosqued,
Javier Baena Presyler, and Javier Espiago
The Archaeological Maps of Spain

The idea of collecting all archaeological findings and sites of a particular re-
gion is at least as old as the first archaeological studies and perhaps prior to
the scientific development of the discipline of archaeology itself. However,
the first archaeological maps (cartas arqueologicas) had very far different objec-
tives than those of today.
The first example of this sort dates from 1818. In Spain, the Law of Ar-
chaeological Excavations promotes the elaboration of an exhaustive inventory
since 1941, when the first archaeological map of the province of Soria, com-
piled by Mr. Bias Taracena, appeared, and in 1945, M. Almagro Basch, to-
gether with Jose Colominas and Jose Serra Rafols, led the effort to compile
the archaeological map of the province of Barcelona and four other provinces.
These six archaeological maps of provinces followed the criteria adopted by
other European countries like Italy, Switzerland, and Yugoslavia. In each map,
the names of all counties were organized alphabetically; within counties, the
names of sites and other archaeological findings, either industrial or artistic
sites, appeared in chronological order from the Paleolithic to the end of the
Hispanic-Visigothic period (eighth century). The scales of these maps varied
from 1:50,000 to 1:400,000.
Changes produced by the new political map of Spain during the 1980s,
together with the transfer of responsibilities, transformed the "official" ar-
chaeological maps into archaeological maps for each "autonomic" government
of the different provinces. These inventories had a fundamental goal: an ex-
haustive knowledge of the patrimony for its preservation and its study. They
include a profound and methodical process and contain not only all existing
literature, but also a complete survey of the field to locate whatever possible,
including medieval, modern or contemporary testimonies. The volume of data
GIS in the Management of Archaeological Data 191
is so great that a computerized system is necessary, both for the creation of a
database and for the need to have a precise cartography that helps to preserve
archaeological remains. At the moment, the autonomic governments initiated
are in the stage of data collection for these comprehensive maps. In any case,
besides these "official" series, other works covering smaller regions exist. These
types of studies proliferated in the 1970s as a consequence of the development
of spatial archaeology and new orientations in the discipline. As far as Madrid
is concerned, there are few works; the only "archaeological map" was com-
piled by Mr. Fernandez Galiano and published in 1976. In this map, unlike
those previously mentioned, sites are ordered by cultural horizons indepen-
dently of the counties to which they belong. There are some antecedents in the
scientific literature of Madrid, where sites of a particular horizon appear together
with a cartography. The lack of systematic surveys in this region has made the
creation of the archaeological map a priority for the regional authorities.
Our team works nowadays on the elaboration of a complete archaeological
inventory of the Community of Madrid (8028 km2), which includes the met-
ropolitan area of the capital, an important network of freeways, and the great
infrastructure needed for its millions of inhabitants. This work was initiated in
1985 to review the impact of public construction in the area and to plan ahead
on the unavoidable destruction provoked by the expansion of a large city. To
elaborate this map, we planned a systematic process implying a survey of the
totality of the area; this survey is organized by natural regions, taking as unit
of reference the county lines and giving priority to the surrounding areas of
the metropolis where the impact on the environment and on the patrimony is
greater. For this purpose, teams of archaeologists are hired to survey the fields
and directly assess the situation of remains. After surveying fifty counties, re-
sults are very uneven due to both the complexity of some of these communi-
ties (some have been extensively urbanized) and the differing past occupations
of them. This unevenness is manifest in the results; some counties offer three
archaeological points and others almost a hundred. Data collected so far have
been published (Velasco et al. 1987: 189-195).
These data help the administration in the decision-making process and pro-
vide an important database for the development of archaeological research. In
particular, this project is used for the following:
1. To determine high concentration of archaeological remains and de-

clare them areas of cultural interest. These places will require a report by
an archaeologist in case of construction or other alteration and, if remains
exist, an excavation of the place must be carried out previous to any work.
2. To rescue every single historical or/and archaeological remain in dan-
ger of disappearance prior to any public or private construction.
3. To inform the administration every time the territory is going to be
subjected to an evaluation or before the initiation of an urban project;
this will guarantee the preservation of remains; otherwise, an excavation
must take place prior to their destruction.
4. To inform private enterprises of the existence of archaeological re-

mains to be preserved or excavated; the expense of those actions has to
be included in their budgets.
5. To provide the scientific community with a database that facilitates re-
search with no need to further survey or study some areas already researched.
System Configuration
There are several factors involved in the choice of a specific configuration to
develop a GIS applied to archaeology. We have considered the following:
The requirements for the process of collection and storage of cartographic

data in a context of distributed GIS (externally connected network)
The requirements of the application, organized in a context of multiple
uses, and the diffusion of the information
The possibilities of expansion of the system already applied
The technical features, price and maintenance of commercial soft-
ware and hardware
Other factors not mentioned here, which offer more particular features,
were also considered; for instance, those factors related to previous experi-
ences of the team in charge, or certain aspects of the administrative organiza-
tion of the final receptive agency about to receive the application.
The collection of data is based on the organization of a distributed GIS. First,
in order to use the existing digitized data (in our case, exclusively cartographic
data), it is necessary to know how to transmit them and how to read them.
Second, we need to have tools to define the structure of the information and
evaluate the quality of the data so as to correct possible errors and defects. It is
necessary to count on the usual capability of the GIS to capture data and store
them in such a way that we can establish topographic relations as well as a
precise location of the sites. Finally, both the collection and the storage of data
also affect the choice of the peripheral hardware for data display.
These considerations affect the selection of the software, graphic pe-
ripherals, storage peripherals, and communications system and network.
They imply a series of works for data reception, data capture, and data
treatment/actualization. ..
As far as the analysis of the application is concerned, the most striking fea-
tures in the selection process have been the need to organize the access to the
data depending on the different levels of confidentiality and the capacity to
alter the information. They affect both the operating system and the proce-
dures to be used with different GIS software. When an application is estab-
lished in a multiuser context, the information must be protected by limiting
the access to files and directories through the issuing of specific conditions to
read and modify data depending on type of users. The UNIX systems give
permission to reading, writing, and executing; they allow three types of users
(owner, group, and others) together with the possible combinations of the
nine. They also permit checking of the system's use and provide a statistical
history of that usage. In this case, the PCs were exclusively auxiliary elements.
A RISC station with large hard disk and EtherNet or Token-Ring connec-
tivity can serve several graphic terminals at almost the same cost as the equal
number of 486 PCs (with which communications would be slower and more
unsafe and without the close to mainframe possibilities of station software).
Furthermore, in the Spanish market, RISC stations tend to be less expensive
than PCs. For instance, the price of a IBM RT PC (RISC) computer with 16
Mb RAM and 200 Mb hard drive was 2.5 times higher than that of an IBM
RISC station, like the RS/6000 320H (500 Mb disk and 32 Mb RAM) in 1988.
Prices for the PCs have not declined as much. The 286or386PCs have dropped
in price, but 486 PCs cost three times as much as RISCs (RISC 1991 and
RISC 1988) three years ago, and they have less capacity. Possibilities for ex-
pansion of the system were clearer in a RISC context than in a PC context.
Once the UNIX system had been chosen instead of RISCs machines,
the selection of a base software was reduced to the few introduced prod-
ucts whose maintenance is guaranteed. Operative requirements for that
type of software were as follows:
The usual procedures for data collection and for vectorial topological
structuralization already mentioned.
Operators of topological relationships and graphic expression that allow
the calculation of inclusion, adjacency, intersection, union or superposi-
tion, buffers, Thiessen-Voronoi polygons, zooms, areas of visibility. An-
other desirable requirement was the capability of performing an analysis
based on the calculation of digital terrain models which is of the most
fruitful techniques in spatial archaeology.
The configuration finally selected by the administration uses commercial

software ARC/INFO on an IBM 530 RISC Sys/6000 with two disk units of
857 Mb each, 64 Mb in memory, and connections to graphic peripherals (A-0
Tablet, plotter, etc.). Through an EtherNet network they are connected to
Xstations and Ps/2 devices (Figure 11.1).
Collection of Cartographic Data

Digitized cartographic information mostly comes from the Office of Territo-
rial Planning (Oficina de Planeamiento Territorial, or OPT), the agency in
charge of regional planning. Apart from other considerations, it was necessary
to use it instead of direct cartographic sources. The preservation of archaeo-
logical patrimony directly relates to urban construction and decisions by the
OPT based on their own cartographic information.
Collected information presents 1:25,000 scale in the UTM projection sys-
194 Anthropology, Space, and Geographiclnfort».ation Systems
InterNet/BITNet X.25
hosts
IBM 3090 PAD
VAX/6000
workstation plotter
EtherNet IBM RS/6000
TCP/IP
XWindows digitizer
XStation
workstation DAT
laser
printer
workstation streamer
colour
printer
scanner
Several plat-
forms linked - PC
via EtherNet digitizer
which allows
simultaneous PC
interactive plotter
graphical •*-
access all-in-all modem
plain old
telephone system
Figure 11.1 Configuration of network.
tern and is filed in five DXF files without structured topology (spaghetti struc-
ture). The treatment consisted of several works of cleaning and cartographic
correction, topological restructuralization, and in the introduction of punc-
tual entities to serve as nexuses to the alpha numbers of sites.
The resulting cartographic information of the system contains the following:
Administrative limits with polygon topology

Reference grid with polygon topology
Hydrographic network in which different codes indicate different
elements (main rivers, affluents, channels, and so on). It is organized
with line topology
Network of roads and rails with line topology
Areas of population and construction with polygon topology
Toponyms and geodesic vertices at three levels and without internal
topological relation
Location of sites and areas of protection
This information is sufficient to carry out the productive exploitation of

the system in the administrative context. However, more detailed data collec-
tion at a 1:5000 scale began in 1992. Also, as a product of direct collection or
research conducted by other university groups, we will add other pieces of
information that will clarify the archeological situation. These are the following:
Elevation; it will allow the elaboration of DTMs

Lithology/tectonic/geomorphology, tied to the analysis of human-
environment interaction
Edaphology/paleo-vegetation and paleofauna reconstructions
Information from remote sensing or, in general, information in
raster format
Alphanumeric Information
The design of fields and their contents was one of the steps prior to the treat-
ment of alphanumeric information. Technicians of the Council of Culture
(Consejeria de Cultura), the future users; university archaeologists, and spe-
cialized GIS personnel from the Cartographic Service (Servicio de Cartograffa)
participated in these tasks.
The administration already had a great deal of the alphanumeric informa-
tion to be used in their own documents, therefore our work focused on the
computerization of the information. For the same reason, we differentiated
the treatment of cartographic information from alpha-numeric information.
Archaeological information was based on a type of survey that took into
consideration only archaeological aspects and presented a hierarchical struc-
ture hardly adequate for a relational context. In this model, data collection in
each site presented open fields that required modifications in form and struc-
ture if they were to be integrated into a computerized system.
Creation of the Database
First, a series of criteria based on importance and grouping of existing fields in

the documentation applied by the administration had to be established. Sec-
ond, a structure was designed for each of the fields composing the database
based upon previous experience during the survey of different counties within
the community. Finally, we created different codes for any of the possible at-
tributes of each site to accelerate the processing of data.
The alphanumeric information contained in the index cards of site inven-
tory is vast and very heterogeneous from site to site. Some sites have been
196 Anthropology, Space, and Geographic Infm~m,atim Systems
excavated several times and have generated immense amounts of informa-
tion; others have been surveyed or visited, and yet others are known but
have never been located.
A database containing the "most complex case" would be excessive for the
"most common case," and that is why we consider the use of different levels of
information. The first level should include basic identification data collection
and location of sites to which has been added a series of fields, with additional
information such as cultural classification of a site: typology, status of preser-
vation, legal situation, and so on.
Description of Fields of the Database
We present here some of the fields included in the information used by the
administration. They express the current status of the alphanumeric informa-
tion loaded in the system.
1. Number. The code number is the exclusive identifier of the site

throughout the system. The two first digits correspond to the county
code and the rest indicate the order assigned to each site within the county.
2. Name. It refers to the toponym or the name by which the site is
known in the literature.
3. Cultural classification. Sites are ascribed to a period of relative chro-
nology based provisionally on surface-collected materials. Each type
consists of two digits and includes the span from Lower Paleolithic to
historical times. Through an extended definition of this field, the possi-
bility of several cultural occupations of the same site is considered.
4. Typology of site. This field classifies each site depending upon the
possible activities carried out in it. Since it is based on surface collection,
this classification is provisional. Three digit codes have been established
to reflect the way archaeological remains can appear in a site. The first
digit refers to the general cultural ascription; the next two refer to the
type of site. For instance, 101:1 equals Paleolithic; 01 means undetermined.
5. Location. Topographic and morphological characterization of a site
using two-digit codes.
6. Axisjy. UTM coordinate of centroid of site area.
7. Axis x. UTM coordinate of centroid of site area.
8. Extension. Site-area surface in square meters.
9. Vegetation. Actual vegetation on the site. This field refers to a de-
scription of the surroundings and identifies through two-digit codes the
vegetation series present in the context of a site.
10. Land use. Current land use.
11. Status of preservation. How the site has been preserved since last visit.
12. Land classification. Legal situation of site. Two-digit codes reflect
the legislation enacted by the Community of Madrid.
13. Declaration. Legal situation of site as well.
14. Works done. Type of work carried out in a site. Distinction between
simply visited or surveyed sites, basically unknown and adding little in-
formation, and excavated sites offering a great deal of information.
15. Reference of materials. Brief reference to features of collected materials.
16. Author of index card. Identification of person who carried out field-
work and is responsible for the information contained in the card; it does
not refer to whoever enters data from the index card in the computer. This
information is very useful to improve information and eliminate bias.
Code Dictionary
In systems of closed fields, the contents of tables have to correspond to a thesau-

rus. The possible contents of each field form a sublist called a field dictionary.
Loading of Alphanumeric Data
Archaeological data from sites were loaded on DBIII, a program compatible

with versions of PC ARC INFO. Once data had been collected, it was easy to
generate a coverage with the fields of site number as future identifier of the
element, and the two site axes. The rest of the fields noted before were related
to this graphic coverage in a way that, finally, we created an interrelated carto-
graphic and alphanumeric database of sites for all the territory proposed
in our project.
Through asynchronic transmission, the coverage was collected within the
work space previously created in a work station. This support was essential for
the use of the volume of information in this project.
Development of the Application for Administrative Management

After we counted on all necessary coverages to respond to all requirements of
this project, we initiated the development of the final application for the user.
A menu-based AML language was used, particularly, pulldown and sidebar
menus. As required by the user, two parts are distinguished in the application.
First, the user selects the geographic context in which to work, either on car-
tographic sheets of 1:50,000 scale, or in particular municipalities.
Once the context has been selected, a working menu offers the follow-
ing operations:
1. Drawing. This option permits the activation or deactivation of the dif-

ferent coverages of information to make them graphic. A zoom is available.
2. Selection of sites. This facilitates the search of sites on simple or mul-
tiple conditions. There is an option to list on the screen the database
associated to each of the sites which fulfill those conditions.
3. Buffers. It permits the creation of areas of influence (buffers) from any
of the coverages included in the system. The user works by elaborating
areas either from road networks or from the sites themselves.
4. Creating BICS. It creates areas of archaeological protection. These
are defined by the administration based on the concentration of existing
sites on the surface. Through the calculation of the coordinates of the
points, the program offers the possibility to design these areas from the
program itself.
5. Queries. It allows quick queries in a graphic form over any of the
elements on the screen. The query shows a window in which all the
database fields of the element in question appear.
6. Graphic output. This option allows the elaboration of reports as well
as the creation of maps of a chosen area.
7. Site update. This permits the modification or addition of informa-
tion on the sites.
Applications forResearch Purposes
Apart from the creation of a product that serves the interests of the regional
administration of Madrid, we have initiated several lines of research. These
relate to our own interests in the discipline of archaeology and serve to open
new dimensions in the field. The first of these attempts correspond to a pre-
Roman site located in the mountainous area near the main water pipe in Madrid.
Castro del Ponton de la Oliva is located in the municipality of Patones de
Abajo (Madrid), on the right margin of the Lozoya River before it joins with
the Jarama River. On the south face of the Mount of Guadarrama, a few meters
from the province of Guadalajara, the site could have belonged to "the chain
of fortified towns that defended the access to the Castillian Plateau" (Cuadrado
1991: 191). The only access to this place is through the south, where a
wall was built.
Recovered materials (coins, particularly) in the excavation by Cuadrado al-
low him to date this town between 105 B.C. and A.D. 423 (Cuadrado 1991:
232). Chronology could be extended to the light of the last excavations up to
the seventh and eighth centuries B.C. The long occupation of this site could
have been a consequence of, among other things, its strategic location, near
one of the only areas of access in the eastern side from the Central Mountain-
ous System (Port of Somosierra) to the North Plateau, and "on the valleys and
plains that form and develop in the confluence of the Lozoya and Jarama Riv-
ers; these are at the same time obligatory areas to pass from the northern and
eastern mountains to the southern plains" (Mufioz 1981: 58). The Jarama River
was also a natural route toward Complutum, the main Roman city of this re-
gion. The site as well as Complutum, belonged to the Conventus
Caesaraugustanus, located on westernmost side of the valleys of Guadarrama,
Guadalix, and Jarama (Balil 1987: 137).
No doubt such a long occupation was due to other factors besides the stra-
tegic position for the control of routes, such as a careful selection based upon
the array of possibilities offered by the environment. To get a close look at
environmental resources and characteristics, the application of GIS allowed
us to combine and to manage all kinds of data.
Unfortunately, excavations have been so far very limited and there is a lack
of detailed analysis of the materials recovered in different areas of the fortified
structure (castro). We do not know if the settlement was continuously occu-
pied or if it was abandoned; we also lack data on the extension that was occu-
pied in different stages, which would allow calculations of population or culti-
vable land, among other data. Thus, in this first stage, we have only consid-
ered some general aspects such as topography, vegetation and weather, poten-
tial use of land, lithological characteristics and existence of mines, Roman roads,
shepherd roads (canadas), and natural passes. We have also outlined the areas
of visibility and Thiessen polygons including mines and possibilities of land use.
In order to include a wide framework, we have digitized approximately 59
km to the west and 37 km to the east of the fortified structure, an area which
includes the mountains and land of the province of Segovia located in the
north; in this way we have inferred possible contacts with the towns in the
Valley of the Duero River through the natural passes in this eastern area of the
Central Mountainous System. We have also incorporated 59 km to the south,
because the natural route along the Jarama River provided an easy communi-
cation route toward the Tajo River. Finally, only 23 km of the eastern area
have been digitized, since the available data of the easternmost part of
Guadalajara are scarce and few known sites relate to Ponton de la Oliva; this
problem takes substantial proportions if we consider the scarcity of cartographic
data on shepherd roads.
The first important factor is its extraordinary visibility 60 km to the south
controlling the valley of the Jarama River, which added a double advantage: on
the one hand, it allowed the control of one of the most important natural
passes between the two sides of the Central Mountainous System which sepa-
rates the Central Plateau of the Iberian Peninsula into two halves; on the other,
this visually dominated area is agriculturally the most productive land.
Secondly, there is a great proximity to areas of surfacing metals, specifically
tin and copper, which have been specially important in Western protohistory
and, in particular, protohistory of the Iberian Peninsula. These surfacing met-
als, easy to exploit in ancient times, appeared in a 10-km buffer around the
site. The area also has granite and metamorphic rocks used to make tools such
as axes, adzes, grinding stones and manos, etc. Unlike the cultivable lands, this
mining area is not visible from the site though we can see the different exits
from the area of exploitation for its commercialization.
The selection of the place had to do with the suitability and the rentability
of surrounding land, because it is an area with variable lithological compo-
nents; there are areas for cattle ranching without possibility of cultivation,
together with flat terraces with high freatic level which were covered by scrub
land. These were used at the time of occupation for hunting and other re-
sources and also some agricultural activity additional to those generated by
commerce and mining.
The interest in this study comes not only from the importance of the site
at Ponton de la Oliva but also from the mining and lithological wealth of the
site; this attracted other people settled in the region, particularly in the low
areas of the rivers, where great agricultural and cattle resources made large
human occupations possible.
For the last few years, our research project has focused on the low areas of
the rivers of Madrid. We aim at a global vision of the human occupation of the
area, its ways of life and resource exploitation from the Paleolithic to pre-
Roman times. Works of excavation and survey are complemented with labora-
tory analyses and geographic applications based on GIS.
A second research project is the group of the Paleolithic which, using a
GIS, tries to study the settlement in the medium and final areas of the
Manzanares River, part of which runs through the city of Madrid.
Archaeological work has so far demonstrated the presence of a very old
settlement corresponding to the Lower Paleolithic (300,000 to 120,000 B.P.).
This research is trying to establish the chronological sequence of different
riverine terraces as departing point to date different sites of the valley. Related
to this study and through the work with the digital elevation model obtained
from the 1:50,000 topography, we try to reconstruct the fluvial model during
Paleolithic times.
Researchers of recent prehistory attempt to reconstruct the paleogeogra-
phy in order to know the environment: it implies a climatic, vegetal, and fau-
nal reconstruction based on pollen analyses and paleozoology. These data (hy-
drology and vegetation) will be combined with archaeological information
(identified by chronological horizons and cultural phases).
Because GIS help to determine the land type and land suitability as well as
the geological characteristics and the distances between synchronic occupa-
tions, it is possible to determine which could have been the priority activities
and the needs of with other sites located in different areas. On the other
hand, we try to observe changes in settlement patterns through time and look
for what caused them.
Our major hopes in the use of GIS are based upon the reconstruction of
ancient geography, the situation of ancient hygrography (river basins and river
flows). We also try to demonstrate, taking into consideration level and depth
of the rivers, etc., if they were channels of communication and transportation.
This last aspect is very important, considering that many agricultural com-
munities of the Bronze Age (1600 to 1100 B.C.) used to transport great quan-
tities of granite in blocks of over 60 kg, to elaborate domestic and industrial
artifacts, particularly grinding stones, from a distance of 70 km or more. Such
effort would be considerably reduced if they had used rafts (balsas) pulled by oxen.
On the other hand, this reconstruction of the paleolandscape allows us to
check natural passes for transhumancy of cattle or mine prospectors as well as
to have an approximate reconstruction of the area of catchment of resources,
taking into consideration the recovered materials in the light of geological
and vegetation maps.
At the present moment we are still introducing the location of sites and
studying the land suitability in the environment surrounding some of the ex-
cavated sites. We are also creating a database of materials recovered in those
sites and locating their areas of closer origin. From here, we try to generate
routes, control of visibility, Thiessen polygons, etc. The area under study, the
subject of detailed cartography, has a radius of 20 km; however, we plan to
include the total area of catchment in order to obtain a complete view of lifestyle
and activities of the pastoral and agricultural communities of the region from
recent prehistory to Roman times.
Acknowledgments
A joint effort of a team of technicians and researchers at the SCUAM (Servicio de

Cartografia de la Universidad Autonoma de Madrid) has made this work possible. Ap-
plied research and teaching of GIS are some of the tasks of the SCUAM. The follow-
ing members of SCUAM have equally contributed to this work: Carlos Almonacid
helped with cartographic questions and software; Marco Leber y Alberto del Rio con-
tributed to matters of software and hardware; finally, Virginia Recuero and Emma
Arnaiz added their expertise in archaeology by helping in the processing of informa-
tion. Soledad Vieitez, Ph.D. Candidate in Anthropology at the University of Califor-
nia in Santa Barbara, translated this paper into English.
12
The Role of GIS in the
Interdisciplinary Investigations
at Olorgesailie, Kenya,
a Pleistocene Archaeological Locality
Richard Potts, Tom Jorstad, and Daniel Cole
A geographic information system is an ideal tool for use in interdisciplinary

studies because it provides automated means of linking and relating different
spatial databases. In this paper we discuss GIS applications to ongoing ar-
chaeological and paleoecological studies at Olorgesailie, an early hominid
archaeological locality in the rift valley of southern Kenya (Figure 12.1) and
one of the most noted Acheulian handaxe sites worldwide (Isaac 1977). The
questions being asked in early hominid archaeology require thinking beyond
individual artifacts and site excavations to broader spatial scales within well-
defined time intervals (or chronostratigraphic units) (Blumenschine and Masao
1991; Potts 1991). The sedimentary exposures at Olorgesailie permitthe small-
est spatial scale of individual artifacts and fossils to be integrated with re-
gional-scale studies. Since many of the GIS applications are still in initial
form, the purpose here is largely to illustrate the conceptual framework by
which GIS integrates the analysis of spatial data at varying geographic scales
in the Olorgesailie basin.
Geologic Background
Covering over 4000 km in length, the African Rift System trends southward
from the Afar Triangle in the Red Sea region to south of the Zambezi River in
Zambia. The numerous continental rift basins that make up the rift system
have a complex structural and volcanic history. For most of its length, the
African Rift traverses Ethiopia, Kenya, and Tanzania. The rift is divisible into
eastern and western portions, which merge into a broad faulted region in north-
ern Tanzania (Baker et al. 1972). Between the eastern and western rifts, occu-
pying portions of Uganda, Tanzania, and northern Kenya, is an uplifted pla-
teau 1000 to 1200 m in elevation.
Uplifted, elongated domal structures located in Ethiopia and Kenya form
the structural base from which the East African Rift System has developed.
GIS in Interdisciplinary Investigations in Kenya 203
igure 12.1 The Gregory Rift Valley in southern Kenya. Locations of Olorgesailie,
ake Magadi, and major volcanic cones are noted.
The rocks that make up this shield complex are Precambrian gneisses, quartz-
ites, and schists. In addition to intrusions by dikes and plutons, these basement
rocks have been altered by partial melting and metamorphism.
Significant though episodic uplift of the Kenyan dome and its flanks during
the late Cretaceous and middle and late Tertiary contributed to the develop-
ment of a graben structure (Baker 1986; Baker et al. 1972). Basalts, phonolites,
alkali trachytes, rhyolites, and ignimbrites are present in the rift and have been
the focus of numerous chemical and petrological studies (King and Chapman
1972; Williams 1965, 1969, 1970, 1972). These volcanic rocks also play an
important role in the geochronology (Baker etal. 1971; Deino and Potts 1990;
King 1970) and structural interpretation of the rift (Baker and Wohlenberg
1971). As the rift evolved, central volcanoes, fault blocks comprising horsts,
grabens, and stepfault platforms became characteristic features. The grid
faulting in the rift has undergone crustal extension on the order of 10%
(Karson and Curtis 1989).
Around 12 million years ago the structure of the rift in southern Kenya was
developing. Volcanic activity in the form of fissure eruptions deposited large
amounts of phonolites and trachytes along the fractured uplift (Logatchev et
al. 1972). Emanating from a network of fissures and fractures, total volume for
these deposits has been estimated to range from 25,000 to 30,000 km3
(Logatchev et al. 1972; Williams 1970).
The nature of the volcanic activity changed from fissure eruptions to cen-
tral volcanoes in the early to middle Pliocene (10 to 5 million years ago). Nu-
merous volcanoes developed along the floor and margins of the rift and con-
tained variable but alkaline compositions. Mt. Olorgesailie began to develop
around 2.5 million years ago and was active for nearly 500,000 years (Baker et
al. 1971). Also active during this period but to the east of the Gregory Rift was
Mt. Kenya. Slightly younger in age were Shira, Mawenzi and Kibo, the three
peaks that form Mt. Kilimanjaro. Mt. Kenya (5199 m) and Mt. Kilimanjaro
(5855 m) are the two highest mountains in Africa and attest to the magnitude
of eruptive events when these volcanoes were active. The faulting that oc-
curred previously in the eastern branch of the rift system was partially masked
by deposits emplaced by central volcanoes during this time.
During the Upper Pliocene and Lower Pleistocene, the flood or Plateau
Trachyte series continued to be an important event in the Southern Kenya
rift. Throughout the formation of the rift, volcanic activity also occurred out-
side the rift margins, as witnessed by flood deposits well beyond the rift valley.
The changing chemistry between fissure and central volcano deposits pro-
vide data on the evolution and correlation of volcanism and faulting. Although
variable in composition, the volcanics in the southern Kenya rift range from
alkaline to transitional basalts and trachytes (King 1970; Williams 1972).
Magma chambers and plutonic igneous bodies underlying the rift floor were
the source of these volcanic rocks (Karson and Curtis 1989).
As part of the eastern branch of the rift system, the Gregory Rift is a fault
bounded graben that is 60 to 80 km wide and 450 km in length. It extends
from the Magadi-Natron basin in the south to the Baringo and Suguta
grabens in the north (Baker et al. 1972). Within the central and southern
portions of the Gregory Rift a series of dense gridfaults with small dis-
placements are present. These subparallel north-south-trending faults cre-
ate a series of stepfault platforms giving an en echelon structural layout
(Baker 1958; Baker et al. 1978; McCall 1967; Walsh 1969). The numerous
lake basins present have in part been created by back tilting of the footwall
blocks (Frostick and Reid 1989). The interfault basins are bounded by fault
zones called accommodation zones and are generally 50 km or less in length
(Bosworth 1985; Rosendahl et al. 1986). Some of these locales became re-
gions in which diatomaceous and carbonaceous deposits occur (Isaac 1977;
Cohen and Thouin 1987). Investigations on the character and nature of
changes in African Rift lakes have been undertaken (Butzer et al. 1972; Frostick
and Reid 1986; Gasse et al. 1980; Grove et al. 1975).
The Olorgesailie basin consists of lacustrine and fluvial deposits exposed in
the southern extent by the drainage of the Ol Keju Nyiro River; the unex-
posed northern area of the basin is divided into the Oltepesi (eastern) and
Legemunge (western) Plains. The Ol Keju Nyiro cuts the southwest portion
of the Legemunge Plain, where it enters the Koora Graben along the western
flank of Mt. Olorgesailie and continues southward for over 40 km. The Kwenia
Plain is situated along the eastern margin of the mountain; it extends 12 km
and terminates in a swampy area (Baker 1958).
Along the northern flank of Mt. Olorgesailie the ephemeral Ol Keju Nyiro
River has eroded through sections of volcanic bedrock. Here the Ol Keju Nyiro
and Oltepesi basalt flows (1.4 to 1.6 million years ago) overlie Mt. Olorgesailie
volcanics. These deposits may be the result of vent activity associated with
Olorgesailie volcanic activity (Baker 1958). These deposits, which dip 10 to
15° west-northwest, can be differentiated into a series of flows and vary from
weathered basalts to lava breccias. Overlying these volcanics are the Plateau Tra-
chyte series (0.9 to 1.3 million years ago [Baker et al. 1971]). In the central part of
the southern Kenya rift, this trachyte series is composed of up to eleven distinct
flows totaling 150 m of quartz trachyte deposits (Baker and Mitchell 1976).
Archaeological Background
Investigations by Gregory and Hobley established the foundation for inter-

preting the structural as well as the cultural history of the Gregory Rift (Gre-
gory 1896). While on safari and camping at the northern flank of Mt.
Olorgesailie along the Ol Keju Nyiro River, they noted the presence of Pale-
olithic artifacts on white diatomaceous deposits (Gregory 1921). This associa-
tion of archeological materials within the rift began a long tradition of inves-
tigating the evolution of early hominids in East Africa.
Since the discovery of the Olorgesailie prehistoric site by Louis and Mary
Leakey in 1942 and subsequent investigations by the Leakeys and by Glynn
Isaac during the 1960s, Olorgesailie has been recognized as an important site
in hominid evolution in the Gregory Rift (Figure 12.1). Beginning in 1985, a
Smithsonian Institution project in collaboration with the National Museums
of Kenya has been conducting fieldwork at Olorgesailie. The focus of current
research has differed from that of earlier work by the Leakeys and by Isaac.
Their efforts concentrated on artifact recovery (Leakey 1952) and artifact as-
semblage and functional interpretation (Isaac 1977). In addition, their efforts
conducted within the site museum boundary looked at archaeological assem-
blages primarily within a fluvial facies. These coarser-grained deposits consist
of silts, sands, gravels, and pebble conglomerates. The association of these
artifacts in these deposits does not provide optimum conditions for interpret-
ing hominid activity areas due to the greater chance of reworking and mixing.
The current work expands upon the previous research in using a
paleolandscape approach (Potts 1989a, 1991; Potts and Behrensmeyer n.d.).
By collectively integrating interdisciplinary fields such as paleoecology,
geoarchaeology, and geochemistry with archaeological efforts, a more com-
plete reconstruction of the site and surrounding region is possible. Compared
with traditional methods that examine smaller areas and isolated artifact asso-
ciations, the paleolandscape approach provides a larger environmental per-
spective. Questions regarding hominid land use, ecological overlap with car-
nivores and spatial distribution of hominid activities can be put into broad
geological and paleoenvironmental perspectives (Potts 1989b).
The geological interpretation of the Olorgesailie deposits has been to date
based on surficial exposures and erosional outcrops and not on borehole or
coring data. The main exposures and localities for the archaeological excava-
tions occur on the northern side ofMt. Olorgesailie, especially along the south-
ern portion of the Legemunge Plain. Thick deposits are also visible within the
gorge cut by the Ol Keju Nyiro River along the Oltepesi Plain east and south-
east of the site museum.
Various names have been used to describe the deposits at Olorgesailie, in-
cluding the Kamasian Beds (Shackleton 1955), Olorgesailie Lake Beds (Baker
1958), Legemunge Beds (Baker and Mitchell 1976) and Olorgesailie Forma-
tion (Isaac 1978). Shackleton's fieldworkin 1944 and 1946 in conjunction with
Isaac's investigations from 1961 to 1965 led to the division of the Olorgesailie
deposits into fourteen members (Isaac 1978; Shackleton 1978). This recogni-
tion of geologic members varies from the "landscape surface" stratigraphic
system used by Leakey (1952) and Posnansky (1959). Subsequent fieldwork
and geochemical studies have confirmed Shackleton's assignment of members
and mapping of the formation (Bye et al. 1987).
Thicknesses for the Olorgesailie Formation average more than 80 m, al-
though most sections are incomplete or compressed. These sediments are
unconformable with the underlying faulted and eroded volcanic deposits, which
have notable relief. Generally, as grid faulting proceeded, sediment traps de-
veloped and created lakes or seasonal water bodies within the basin. This,
combined with continuous and differential deformation of the deposits, built
a complex history of sedimentation. The numerous composite sections pro-
duced by Isaac (1977) illustrate that, within the basin, faulting and deformation
were contemporaneous with deposition of part of the Olorgesailie Formation.
The sediments are made up of diatomites, volcanic sands and gravels, silts,
diatomaceous silts and clay. Strata within this series represent lake margin and
fluvial fioodplain paleosols, which contain paleontological and archaeological
information. Episodes of regional volcanism in the form of airfall ash and
volcaniclastic deposits are also present.
Intercalation of ashes and pumiceous strata with fossil-rich deposits and
lake beds has provided a means for extremely fine calibration of hominid ac-
tivities, faunas, and climatic changes within the Olorgesailie basin by single
crystal laser fusion '"•AR-ÂR dating (Table 12.1). The oldest dated ash in Mem-
ber 1 is 992+39 ka, and the youngest age estimate is a pumice dated 49±ka, above
the upper boundary of the Olorgesailie Formation (Deino and Potts 1990).
The focus of the Smithsonian excavations has been on Member 1 deposits
Table 12.1 Olorgesailie age determinations.
Post Olorgesailie Formation 49± 6
Upper Olorgesailie Formation 215±17
Member 14 493+ 1
Member 12 601± 3
Member 10 662± 4
Member 9 747± 6
Member 8 <1000
Member 5 974± 7
Member 4 992±39
Member 1 2660±30
Ages determined by single-crystal 40Ar/39Ar dating of

pumice, ash, sandstone, and lavas (Deino and Potts
1990). Readings are in thousands of years plus or mi-
nus one standard deviation.
(Figure 12.2). This basal unit of the Olorgesailie Formation is visible in out-
crops north of the Ol Keju Nyiro River as well as in discontinuous areas within
the site museum. Archaeological excavations since 1986 have concentrated on
an artifact- and fossil-rich paleosol above the 992,000-year-old ash layer.
Figure 12.2 Edited portion of Olorgesailie map from Shackleton (1978), highlighting
the Member 1 outcrops. Box west of the site museum boundary represents area illustrated
in Figure 12.3.
GIS and Other Mapping Applications
Research objectives in the Olorgesailie basin have differed conceptually from
most previous archaeological studies of early and mid-Pleistocene hominids,
though they follow directly from theoretical considerations (Isaac 1980; Binford
1982) and applications to later Stone Age materials (Foley 1981). The first
part of the research agenda involves the paleolandscape approach. Excavation
has been directed along an extensively exposed fossil-bearing horizon in a low-
energy environmental context. This stratum represents a root-marked paleosol
in the upper part of Member 1 of the Olorgesailie Formation (ca. 990 ka), and
is exposed over several kilometers. Excavation sampling of this paleosol aims
to evaluate how stone artifacts and other evidence of hominid activities vary
with topography, vegetation, evidence of hyena and other carnivore activities,
death sites of animals, and other paleoecological data. In contrast with the
typically small excavations of previous archaeological studies of early homi-
nids, the time-averaged paleolandscape is the unit of analysis in the current
research at Olorgesailie.
The second objective arises from the surprising degree of time depth evi-
dent at Olorgesailie on a basinwide scale. Previous geochronological studies
had suggested that the Olorgesailie deposits might represent a time interval
about 100 ky long during the middle Pleistocene. As noted earlier, ÂR-ÂR
analyses have shown that the Olorgesailie deposits span nearly 1 million years.
Stone artifacts, fossil animals, and paleoenvironmental information (e.g. lake
characteristics from diatoms and vegetation from soil carbonates) collected
throughout the basin will aid in documenting the interface between hominid
activities and ecological history in the region over this extensive time period.
These two developments in the research at Olorgesailie, the paleolandscape
approach and diachronic studies across the entire sedimentary basin, require a
system for handling and interpreting significant amounts of spatial informa-
tion. A two-part system was initiated to collect and manage data. The first part
is a multifaceted program that involves the use of a Geodetic Total Station to
conduct geological mapping and spatial positioning of archaeological and geo-
logical samples. A coordinate grid system was established in order to integrate
and unify all archaeological excavations. With this information, all previous
excavations as well as future sites are combined into one system. With current
applications, this method provides very accurate three-dimensional
proveniencing of all artifactual materials, sediment samples, and strati-
graphic information.
The second part of the system utilizes the GIS to map and interpret data.
The GIS was used to create a regional base map with the aid of existing geo-
logical maps (Shackleton 1978) and geological fieldwork. By selectively digi-
tizing information within the study area and layering pertinent geological data
such as faults and rivers a customized map was generated. Figure 12.2 illus-
trates the extent of the Olorgesailie Formation members and archaeological
localities. The boundaries of sedimentary exposure in the western half of the
Olorgesailie basin are shown, with divisions into lower members (1 to 9) and
upper members (10 to 14) of the Olorgesailie Formation. The area north of
these exposures, near the Magadi Road, represents the uneroded Legemunge
Plain. Contour intervals are every 20 ft, as originally mapped by Shackleton
(with discontinuous contours inside the Site Museum boundaries).
This artifact-rich paleosol in upper Member 1 is bracketed by two marker
units: an underlying airfall ash deposited in diatomite lake beds; and an over-
lying carbonate horizon, representing a broad mud-cracked layer in the lake
beds deposited later. Elevations of these three units (ash, paleosol, and car-
bonate) have been measured at many points over the lateral extent of the
paleosol. These measurements, instigated by project geologist A. K.
Behrensmeyer, have been taken so that the relative elevation of the paleosol
compared with these two other horizontal marker horizons can be determined;
this provides a means to correct for geologic dips and faults and thus to recon-
struct the original relative topography of the landsurface (Potts and
Behrensmeyer, n.d.). Figure 12.3 displays a small, archaeologically rich por-
tion of the outcrop. These maps are based on surface elevation measurements
and depict two important features of the outcrop: Hyena Hill is one of the
richest areas of archaeological and paleontological finds, and includes excava-
tion Site 102; and the Elephant (Site 15) locality represents a relatively flat
area easy to access by excavation in comparison with Hyena Hill.
The GIS, which facilitates the mapping and analysis of each archaeological
excavation in the region, permits easy visualization of the relationship between
the spatial distributions of artifacts and fossil bones and larger scale features of
the outcrop. Since three-dimensional data have been taken on all stone arti-
facts and fossil bones found in excavation, maps representing the distributions
of these materials can be created. Combined with an extensive program of
sampling for the purpose of microhabitat interpretation, the paleosol maps
facilitate an examination of the relationship between hominid and other ani-
mal activities and paleoenvironmental variation.
Figure 12.3 Partial topography of the Olorgesailie region including Hyena Hill and
Elephant (Site 15) localities. Symbols indicate surveyed points along the paleosol
exposure. One of several transit locations is noted.
Site 15 contains a skeleton of the extinct species Elephas recki in association
with stone artifacts (Potts 1989a). The intimate and nearly exclusive spatial
association between the elephant bones and stone artifacts (Figure 12.4) pro-
vides important information about taphonomic context, leading to the inter-
pretation that this elephant carcass attracted toolmaking hominids to this place
on the ancient landscape. Layering of data for specific bone type, artifact type,
and depth within the paleosol is possible. So with the aid of GIS applications,
interpretations of hominid activity areas on a paleolandscape are refined.
Site 102 is a second example relating important fossil finds to mappable
features of the paleolandscape. This site represents a series of burrows into the
diatomite strata directly beneath the paleosol and is filled with a mix of paleosol
and diatomite. The ancient soil collapsed on one of the burrow chambers en-
trapping several hyenas (Crocuta crocuta) nearly 992 ka (Potts 1989a). Preserved
in a series of complicated depressions, a dense concentration of bones was
Figure 12.4 Partial plan of Elephant (Site 15) locality, displaying the distribution of
stone artifacts and bones. Triangles represent 1m intervals.
G/5 in Interdisciplinary Investigations in Kenya 211
carefully excavated and mapped. Using GIS, data can be entered regarding
den dimensions and areal density of fossil bones; then data can be used to
examine spatial relationships between carnivore and nearby hominid activi-
ties and their possible overlap.
In an effort to acquire easier access to the fossil- and artifact-rich area sur-
rounding Site 102, a bulldozer operation was undertaken at Hyena Hill. This
activity removed up to seven m. of overburden and provided access to the
paleosol over an area of 1800 m2.
A conductivity survey was undertaken in this bulldozed area to assist in
formulating an excavation strategy. Soil conductivity is electrolytic in nature
and depends on the presence of ions in moisture-filled pores (Keller and
Frischlcnecht 1966; Scollar et al. 1990). Mapping using this method is a
noninvasive means of measuring how well an electric current flows through a
soil. In principle, a conductivity probe induces a current into the soil and records
the magnitude of the magnetic field that is generated. Terrain conductivity is
dependent on soil moisture as well as porosity, temperature, salinity and clay
content (Hesse et al. 1986; McNeill 1980). Taken collectively, these factors—
along with matrix type, topographic setting, and season of the year—affect
conductivity. Although experimental in nature, given the age and kind of de-
posits at Olorgesailie, prospecting using electrical methods has been used in a
variety of settings (Frohlick and Lancaster 1986; Tite and Mullins 1970). The
areas of test excavation were selected by viewing variations in the conductiv-
ity values. Areas of relatively abrupt conductivity change and of little change
have been excavated.
As a result of this survey and subsequent excavations, it will be possible to
use GIS to relate two highly disparate databases: spatial positions (and areal
densities) of excavated objects and electrical conductivity. Figure 12.5 is a map
that expresses variation in electrical conductivity over the extent of Hyena
Hill. Information obtained from excavation enables study of the relationship
between conductivity data and the concentration of large and small animal
fossils, stone artifacts, and paleosol features. The conductivity data and exca-
vation sampling of Hyena Hill reflect a spatial scale intermediate between that
of a typical excavation site and that of the paleolandscape. The GIS system
will permit excavation of these multiple spatial scales in assessing how homi-
nid activities relate to habitat variation across the paleolandscape.
Significance
The GIS framework outlined here is significant to the Olorgesailie research
in several ways. In the paleolandscape approach, excavations many tens to thou-
sands of meters apart must be tied into one spatial system rather than assessed
separately. This effort reflects the general shift of early hominid archaeologi-
cal research away from analysis of small sites in isolation to much broader
landscape and regional contexts (Potts 1991). The visual and analytical linking
of small excavations enables the long erosional transect through the paleosol
to be used as a unit of analysis. It does this by permitting researchers to define
Figure 12.5 Conductivity of Hyena Hill after levelling by bulldozer. Contours

are expressed in units of conductivity, millisiemens/meter, where one Siemen is
equal to one mho.
foci of hominid activities and other site formation processes in relationship to

the overall in situ distribution of artifacts, fossils, and environmental evidence
throughout the paleosol. It is on this basis that the paleolandscape approach
differs from previous approaches in early hominid archaeology, and GIS pro-
vides critical spatial support for this shift in thinking and data collection.
Depending on the purpose, archaeological data can be plotted by true el-
evation, stratigraphic position, or artifact type. Since the Olorgesailie strati-
graphic record spans the past 1 million years, stone artifacts may be related
spatially to fossil animal bones and geologic samples that yield specific
paleoenvironmental data representing this vast time scale. Also, GIS permits
the layering of information following each field season. Thus, the editing op-
tions available within GIS are well suited to the kind of long-term,
multidisciplinary research fundamental to early hominid studies.
As archaeological field research on early hominids goes beyond simple arti-
fact recovery to encompass ecological approaches, spatial relationships between
objects and their context become essential to know and to analyze. Using GIS
in the multidisciplinary effort at Olorgesailie effectively integrates and man-
ages large amounts of data derived from archaeological, geological, and pale-
ontological studies. It thus facilitates the reconstruction of ecological condi-
tions during a lengthy period of hominid and faunal evolution.
Acknowledgments
The Olorgesailie research is conducted in collaboration with the National Museums of

Kenya, Nairobi. We thank Drs. M. Isahakia and M. G. Leakey for permission and
assistance. R. P. gratefully acknowledges funding for field research from the Smithsonian
Scholarly Studies Program, the National Museum of Natural History, and Robert and
Joan Donner. D. C. acknowledges support from the Smithsonian Research Opportu-
nities Fund. We thank A. K. Behrensmeyer, R. Chapman, and J. Clark for their superb
contributions to this research and to the original manuscript. This is a publication of
the Smithsonian's Human Origins Program.
13
Danebury Revisited:
An English Iron Age Hillfort
in a Digital Landscape
Gary R. Lock and Trevor M. Harris
The doyen of British field archaeology, O. G. S. Crawford, noted that "The

surface of England is a palimpsest, a document that has been written on and
erased over and over again" (Crawford 1953: 51). Many centuries of detailed
observation and recording of the English landscape have resulted in a wealth
of archaeological data, covering many thousands of years of human habita-
tion. The need to record and decipher these extensive field data has led to the
adoption of methods and techniques developed in other disciplines, including
that of geography. Geographic information systems are the latest tools to be
adopted in the quest for effective methods of field recording and archaeologi-
cal analysis (for introductory GIS texts, see Aronoff 1989; Burrough 1986;
Star and Estes 1990; Tbmlin 1990). The applications of GIS in archaeology
can be differentiated according to scale and type, although relatively few ma-
ture applications currently exist. Studies that have been undertaken range from
intrasite to intersite analyses, and from research-driven applications to inven-
torying and cultural resource management (see for example Allen et al. 1990;
Gaffney and Stancic 1991, 1992; Harris and Lock 1992; Larsen 1992; Lock
and Harris 1991). This regional study, based on the Iron Age hillfort of
Danebury in England represents a contribution to this growing literature and
to the development of GIS use in archaeological analysis.
This paper has two main aims. First, it seeks to identify and examine the
archaeology of the Danebury region within the context of existing archaeo-
logical theory and to refine and add to these interpretations where applicable.
Second, the paper seeks to undertake this analysis within a GIS environment.
Our goal here is to illustrate how GIS can contribute to archaeological analy-
sis, shed new light on existing knowledge, and enhance our understanding of
the prehistoric use of the landscape. In landscape archaeology, there are sev-
eral well-established themes that are strengthened and augmented by the data
handling and analytical capabilities of GIS. This paper elaborates and devel-
ops these themes in the context of the ongoing archaeological study of the
Danebury hillfort region.
An English Iron Age Hillfort in a Digital Landscape 215
GIS and Landscape Archaeology
Maps are the bare bones upon which the details of field archaeology are hung.
Since 1791, the Ordnance Survey (OS) in Britain has provided much of that
map skeleton; recently in digital form. Before the appointment of Crawford
in 1920 as the first OS archaeological officer, the recording of archaeological
sites was undertaken by interested individuals and societies. Typical of these
individuals was Sir Richard Colt Hoare, who published detailed archaeologi-
cal site observations in Ancient Wiltshire (1812) and contributed greatly to
the archaeological content of the first OS 1:63,360 map series of that county
in 1811. From these early endeavors the systematic recording, inventorying,
and mapping of archaeological sites and site information by archaeologists
has become a fundamental component of archaeology. Much of this record
exists in innumerable excavation studies, voluminous reports, and site records
stored in county and national archaeological archives. The variety and nature
of the record media has restricted the availability and utility of these archaeo-
logical data, especially for regional intersite studies. The constraints caused
by the inherent difficulties of incorporating and using disparate archaeologi-
cal information in landscape archaeology is the first of our themes. The spa-
tial data handling and mapping capabilities of GIS will not only have a
major impact on the way in which archaeological data is recorded but also
greatly facilitate the dissemination of valuable archaeological data to the
archaeological community.
A second major theme in landscape archaeology is concerned with evaluat-
ing archaeological evidence in the context of the surrounding landscape in
which it is found. The focus of early field archaeologists was almost exclu-
sively on observing and recording individual archaeological sites. A few far-
sighted individuals, such as John Aubrey in the seventeenth century and Wil-
liam Stukely in the eighteenth century, did go beyond this role and attempt
the reconstruction of past natural environments of areas surrounding the re-
corded monuments (Ashbee 1972). Aubrey, for example, commented on the
preferential siting of barrows on false crests rather than on the highest points
of the landscape and explained the clustering of barrows by reference to line-
of-sight visibility with Stonehenge. Stukely, referring to the cutting of bar-
rows by Roman roads, developed the theme of chronological relationships.
Concern for the meticulous measurement and recording of sites, chronologi-
cal relationships, and the relationship of field monuments to contextual cul-
tural and natural landscapes marked these studies as significant developments
for the age. It was Crawford, though, who began the major shift toward the
interpretation of archaeological sites within their environmental context. His
use of distribution maps for spatial analysis demonstrated the interpretive po-
tential of overlaying archaeological distributions on environmental background
maps [developed by Fox (193 2)]. Distribution maps remain among of the most
widely used spatial analytical tools in archaeology. Reconstructing the evolv-
ing relationships that existed between former cultural and physical landscapes
forms the basis of modern landscape archaeology. Study of the contextuality
of archaeology and landscape is particularly well suited to examination and

exploration in a GIS environment. A major strength of GIS lies in its ability
to relate, integrate, and analyze a variety of geographical data within a single
computing environment.
A third broader strand linking the development of landscape archaeology
to GIS lies in the shared methodologies and techniques that exist between the
disciplines of geography and archaeology. The borrowing of spatial statistics
and spatial analysis from geography (Gamble 1987; Goudie 1987), and paral-
lel disciplinary developments in landscape analysis (Vita-Finzi and Higgs 1970),
locational analysis, central place theory (Grant 1986), perception studies, cul-
tural geography (Rowntree et al. 1989), human-environment relationships,
and human ecology (Butzer 1982, 1989) have all contributed to the develop-
ment and enrichment of shared methodologies and techniques between the
two disciplines. The GIS represents an important continuation of this trend.
The ability to pursue existing modes of archaeological analysis, such as distri-
bution mapping or spatial statistical analysis, and to develop new analytical
techniques within GIS is an important bridge linking traditional archaeologi-
cal to recent technological advances. These developments will continue to
strengthen the links between geography and archaeology in the pursuit of
cultural interpretations of the visible landscape.
The appeal of GIS technology to archaeologists is the ability to pursue
traditional landscape archaeology themes, as outlined above, within a single
integrative digital environment. The traditional map-based approach to land-
scape archaeology; the quantification of spatial phenomena; and contextual
analytical approaches can be pursued within a powerful spatioarchaeological
GIS environment. The dynamic social and cultural use of a single landscape
can thus be explored (or, as in this study, reexplored), tested, interpreted, or
reinterpreted in the context of existing archaeological theory. The GIS also
provides new techniques for archaeological analysis. The approach taken in
this study is for GIS analysis to be archaeological theory-driven such that ex-
isting theories relating to the Danebury region are examined and tested and
new observations, insights, and interpretations of the archaeological record
developed where applicable. Part of the challenge of implementing GIS func-
tionality within archaeology is to explore traditional modes of analysis while
shaping new approaches to the evolving methodological and technological
base of landscape archaeology. In the use of GIS for archaeological analysis
and interpretation, we draw comparison to the approach of Tukey in explor-
ing numerical data (Tukey 1977). The emphasis on visualization, manipula-
tion and reexpression; the teasing out of themes, patterns, and processes by
exploring, reworking, and reinterpreting the data; are common themes be-
tween exploratory data analysis and GIS. This approach contributes substan-
tially to the analytical and interpretive power of GIS in archaeological studies.
Equally, GIS brings about its own issues and complexities relating to the spa-
tial and temporal components of archaeological data, particularly in a complex
landscape such as that around Danebury. As we indicate in this study, these
fundamentals cannot be ignored without deleterious effects on study results.
An English Iron Age Hillfon in a Digital Landscape 217
Danebury Revisited
After some twenty years of survey and excavation, Danebury hillfort repre-
sents one of the most intensively studied prehistoric sites in western Europe
(Figure 13.1). More than one third of the enclosed 5 ha has been excavated
and a wealth of data and information about the area has been collected (Cunliffe
1984; Cunliffe and Poole 1991). As a result, Danebury and its environs repre-
sents one of the most important landscapes contributing to our understanding
of later prehistoric society. The area has a rich prehistoric chronology repre-
sented by field monuments dating from the Neolithic to the Romano-British
period; the fourth millennium B.C. to the first century A.D. Archaeological evi-
dence suggests that Danebury was occupied throughout the first five centuries
B.C., supporting a population of several hundred people at any one time. A
variety of activities and crafts are known to have been practiced on the site,
and biological remains suggest a mixed farming regime based on crops and
domesticated animals. The massive timber and earthen ramparts and impos-
ing entrance were redesigned and rebuilt several times during the life of the
hillfort; this had perhaps as much to do with prestige and status as with defense.
As a result of this detailed attention, Danebury has been the focus of much
theoretical work concerning Iron Age sociopolitical organization in southern
England (e.g., Haselgrove 1986). Cunliffe has generated ideas about the rise
to dominance of developed hillforts in Wessex, such as Danebury, at the ex-
pense of other hillforts. He also hypothesizes their role as redistribution cen-
ters for local and long-distance trade (Cunliffe 1984). Excessive storage capac-
ity for grain and theoretical models based on a client-king hierarchy of large
centralized hillforts surrounded by smaller subordinate farmsteads support these
ideas. A gradual shift in British excavation policy over the last two decades
away from large-scale research excavations suggests that Danebury and its
environs will be unique for many years to come in terms of the amount of
archaeological information and knowledge available for any one hillfort.
Reconstructing the Archaeological Record
This analysis is based upon Palmer's (1984) pioneering study, which collated
much of the considerable disparate archaeological data pertaining to Danebury
and its environs into a single publication. Palmer's study brought together
data from the long sequence of excavations in the region, over a century of
fieldwork, data from cultural archival repositories, and thousands of aerial
photographs. Many new sites were identified and plotted by Palmer from crop
marks on aerial photographs, and these included important settlements, field
systems and linear ditch features previously unidentified from ground survey.
The two Sites and Monuments Records (SMRs) of Wiltshire and of Hamp-
shire, and the National Archaeological Record of the RCHM(E) [see Harris
and Lock (1990) for the role of these bodies in the cultural resource manage-
ment of the United Kingdom] also contributed substantially to the database.
The resultant data were presented by Palmer as a composite map covering
2
450 km of the Danebury region (see Figure 13.1) at a scale of 1:25,000, with
the archaeology classified into four archaeological time periods.
For the purpose of this study, additional data were obtained from the
Wiltshire SMR and added to Palmer's to cover a "missing" 50 km2 section to
the northwest of the study area. This addition produces a rectangular study
area of 500 km2 and significantly reduces the impact of edge effects on the
analysis. Detailed aerial photographic data, however, were not available for
the additional area; thus some variation in data quality exists within the study
area database. Archaeological sites and features were digitized from Palmer's
Figure 13.1 The location of Danebury in Hampshire, England.

maps using a finer period classification. Present day hydrology, soils, and ter-
rain elevation data were also encoded (Figure 13.2 and subsequent figures).
Using present-day environmental data to reconstruct the landscape of some
6000 years ago invokes a number of assumptions regarding environmental
change. Environmental information is used in two ways; first, the modern
data are extrapolated backward with the aid of, albeit limited,
archaeoenvironmental information such as that of colluvium formation. Sec-
ond, assumptions are made concerning the extent of environmental change
during this time, as, for example, in the rather intractable areas of clay-with-
flints deposits remaining wooded. Indications of such environmental change
within the Danebury region are discussed in more detail below. "Negative
areas," where the archaeological evidence is known to have been obliterated
or obscured, are also recorded (Figure 13.3). Despite the richness of the ar-
chaeological record, these areas are extensive and reflect the imprint of subse-
quent building and road development, disturbed ground through quarrying,
and the presence of wooded areas where both earthworks and aerial photo-
graphic evidence are obscured. Such "missing" pieces in the archaeological
jigsaw puzzle are significant and must be taken into account in any analysis.
In many respects the demands of GIS bring the issues associated with the
spatial and temporal accuracy of the prehistoric archaeological record into
sharper focus. The two main conceptual axes that underpin archaeology, the
Figure 13.2 Soils and hydrology of the Danebury area.

Figure 13.3 Negative areas.
temporal and spatial dimensions, invoke gross assumptions about the record-
ing of the archaeological record. Methodologies for dealing with time as a
continuum are as yet insufficiently developed and have traditionally enforced
the use of a categorical approach to data recording and analysis (Castleford
1992). The construction of appropriate temporal classifications on the basis of
existing knowledge is notoriously difficult. Extensive excavation can alleviate
this problem to some extent, but a temporal classification often into periods of
several centuries is the norm. The implied assumption of historical
contemporaneity between sites attributed to any one period is problematic.
While a refined chronology is a worthy goal in such circumstances, it is virtu-
ally impossible in an extensive regional study of this kind.
The spatial domain is equally problematic. Determining the spatial extent
of a "site" has been a long-standing methodological problem faced by archae-
ologists. The very nature of GIS seeks to locate objects in precise coordinate
space, yet the archaeological evidence does not always lend itself to such im-
plied accuracy. How to define the boundary of a prehistoric settlement, or to
what extent a linear ditch feature implies human influence over the surround-
ing area, are not easily ascertained or recorded. Recourse has invariably been
made to traditional categorical approaches to site recording and analysis. The
need to determine fuzzy space within a fuzzy temporal domain is admirably
demonstrated within the archaeological context and promises one fruitful av-
enue of GIS and archaeological research.
The issue of spatial accuracy is a matter of additional concern in encoding
and interpreting the archaeological record within a GIS. The archaeological
data recorded by Palmer were published at a scale of 1:25,000. In the case of
most archaeological evidence, this scale is sufficient only to represent the spa-
tial extent of an archaeological site in its more generalized form. The detailed
plans that accompany most excavation reports cannot and do not warrant be-
ing incorporated in a study undertaken at this scale. Thus archaeological fea-
tures recorded at this scale tend to be representative or symbolic spatial depic-
tions of the known archaeological record rather than "true" cartographic de-
scriptions. For example, a series of crop marks that together make up a prehis-
toric field system can be extremely complex in form. The field systems are
recorded in this study have been reduced to polygons delimiting only the esti-
mated external shape of the feature, with no attempt to record the detailed
spatial description of the individual fields or boundaries which made up that
area, even though that information is known to exist. For the purpose of this
regional analysis of Danebury a more accurate spatial representation of the
archaeology at a scale smaller than 1:2 5,000 was deemed unnecessary. Under
differing circumstances—for example, for intrasite analyses—a more detailed
spatial record may be a vital prerequisite. As in any study, the results of this
analysis must be evaluated against the inadequacies of data quality and ar-
chaeological classification.
Reconstructing the Environmental Record
Today, a thin mantle of poor-quality soil covers the undulating chalk downland
of the Danebury region, interspersed with occasional patches of clay-with-
flints mostly on the hilltops (Figure 13.2). The valley bottoms are covered by
colluvial and alluvial deposits to varying depths. Today the region is not prime
agricultural land and supports some cereal production, but it is predominantly
given over to sheep and cattle rearing on the downs and valleys, respectively.
There is evidence, however, that this land was not always of marginal quality.
At the beginning of the Neolithic period, the region was covered with loessic
deposits (Catt 1978; Robinson 1984), providing fertile, light, easily worked
soils suited to prehistoric ards and attractive to early farming communities.
The density of prehistoric archaeology in the area in itself suggests an agri-
culturally productive landscape. At the beginning of the Neolithic period,
much of the land was covered with woods which were gradually cleared for
agricultural purposes (Smith 1984). By the Iron Age, the landscape was a
mosaic of mixed farmland, with good downland turf for sheep grazing, arable
fields, patches of managed woodland for timber and pig pannage, and good
valley-bottom grazing for cattle.
It is generally accepted that woodland clearances during the prehistoric
period led to soil erosion and the depletion and removal of the loess cover.
Continued depletion of the soil resource and bedrock denudation has pro-
duced the poor-quality chalk-derived soils of today. The existence of exten-
sive colluvial deposits in the valley bottoms is attributed to this process of soil
depletion. In some valleys up to 2 m of colluvium, containing much loessic
material, has formed. Crude estimates of soil erosion from valley sides vary
from 2 to nearly 20 cm, and the lynchets surrounding Celtic fields could be
one early prehistoric response to containing this erosion. Detailed excava-
tions of colluvial deposits in the chalk valleys suggests that extensive soil ero-
sion began during the Neolithic period and intensified during the Late Bronze
and Iron Ages (Bell 1983).
The study area is crossed in the east by the Test and Anton rivers and their
tributaries, the Wallop Brook and Pillhill Brook, and in the west by the Bourne
and Avon rivers (Figure 13.2). All four rivers show evidence of meandering
and course changes within their shallow, flat bottomed valleys, especially the
River Test, with its floodplain of nearly 1 km in width in places. Alluvial and
colluvial deposits in numerous dry valleys indicate previous running water,
especially evident along the valley of the Pillhill Brook, though such deposits
could be the result of occasional floods. The level of the water table during
prehistoric times could have a bearing on the hydrology of the area, but this is
notoriously difficult to establish. Evidence from the nearby Wilsford Shaft
excavation, however, indicates that the Late Bronze Age water table was not
very different from that of today (Ashbee et al. 1989).
Establishing erosional and depositional rates to model the prehistoric to-
pography is problematic and fraught with issues of accuracy and substantia-
tion. Although estimates for specific locations have been calculated, these can-
not be extrapolated over the entire area. Because of this insufficiency of evi-
dence to support meaningful models of the prehistoric environment of the
entire study area, the nature of prehistoric environmental conditions must be
borne in mind throughout the following analysis. Our understanding of the
archaeology of the period, especially the earlier periods of the Neolithic and
Bronze Age, is necessarily influenced by the long-term impact of erosive and
depositional forces on the landscape. Similarly, interpretation of some six mil-
lennia of archaeology must be undertaken against the backcloth of varying re-
lationships between humans and nature and a dynamic physical environment.
Disaggregating the Archaeological Landscape
The analysis that follows adheres to a traditional approach of examining the

archaeology by sequential archaeological periods. For each chronological pe-
riod, the major archaeological features are described and explained within ex-
isting theory and interpretation. The GIS is used to rework the archaeological
data to support, amplify, amend or add to existing ideas about Danebury and
its environs. The archaeology is portrayed in a series of computer-generated
plots. The inherent difficulties of displaying a complex archaeology under-
lines the utility of GIS to disaggregate, recombine, and display archaeological
phenomena in a variety of forms to give clarity and aid interpretation. How-
ever, while visualization is central to GIS, the benefits are heavily constrained
when complex color computer-generated images are converted to monochrome
for publishing purposes. Technological advances in computer graphics have
outstripped cost effective reproduction in traditional book form and new me-
dia technology is necessary to fully take advantage of developments in scien-
tific visualization and enable dissemination of results to a wider audience.
Despite these limitations this study seeks to add to the sum understanding of
prehistoric societies in this region and to demonstrate, through example, the
extensive capabilities of GIS for archaeological analysis.
An english Iron Age Hifffort in a Digital223
The Neolithic Period
Compared to later periods, there is a paucity of data for the Neolithic period.
Furthermore, the Neolithic represents some two to three thousand years of
the archaeological record, and the problems of chronology become manifest.
Very little evidence for settlement exists for this period generally, and the
emphasis of the archaeological record in the Danebury region is on the funerary
monuments of long barrows as well as on individual finds. It has been sug-
gested that, among other functions, Neolithic long barrows could represent
an outward expression of the territoriality of a social group over an area. Burial
was not the sole function of these large monuments, for while some were used
for multiple burials and associated complex ritual activity, others contain no
burials at all. Theories about long barrows have, therefore, included their pos-
sible role as symbolic monuments demarcating the social and political terri-
tory of a local group of people. The theme of visible territorial markers is well
established in Neolithic landscape studies, and central to this idea is the place-
ment of the monument in the landscape to obtain maximum visibility over an
area. Traditionally, defining this territoriality has been explored through
Thiessen polygons (Figure 13.4), which identify areas proximal to each of the
long barrows. If the long barrows were uniformly dispersed across the land-
scape, then the polygons would equally appear uniform in size and shape. Such
is not the case in Figure 13.4, for several long barrows appear to cluster and
many barrows tend to be located toward one corner of the polygon. In the
context of the topography this is significant. It suggests that long barrows
could have been located so as to visually dominate specific segments of the
surrounding territory, which dropped away beneath it, and the figure shows
this for several individual barrows. A natural extension of these ideas is to
equate each long barrow with a designated "territory" over which members of
the local group exerted control. The visual symbolism of a long barrow thus
Figure 13.4 Neolithic long barrows and Thiessen polygons.

provided tangible expression of territoriality to any person who entered the area.
Existing manual spatial analyses of these barrows have tended to stop at this
point. The GIS, however, permits the extension of these theories to explore
related or alternative hypotheses. Given the suggested linkage between visual
dominance and territoriality, it was hypothesized that the viewshed of each
long barrow should correlate reasonably closely with the patterns generated
by the Thiessen polygons. This was not the case, however, with barrow
viewsheds being very irregular in both shape and size (Figures 13.5 and 13.6).
In some instances, the viewsheds dominated only small sectors of the land-
scape; in others, they covered broad swathes of the surrounding countryside
with little apparent relationship between the patterns created by the viewsheds
and the Thiessen polygons. In at least two cases, the viewsheds exhibit definite
rim effects (Figure 13.5), suggesting a classic use as territorial marker to alert
people crossing the surrounding ridge tops.
Perhaps of greatest interest arising from the GIS-generated viewsheds is
the discovery that the long barrows do not fall within each other's line-of-
sight (Figures 13.5 and 13.6). In almost every case, each barrow appears to
have been located so as to exclude other barrows from its viewshed, even in
those areas where the barrows are in close proximity to one another. Using a
conservative offset of only 2 m to represent the height of a long barrow when
newly created, the viewsheds are remarkable for the way that the barrows have
been located so as to utilize the topography to shield each barrow one from
the other. A sensitivity analysis with a barrow offset height of 4 m showed
almost no variation in the patterns so discerned. The consistent
nonintervisibility between barrows appears unlikely to have occurred through
chance but was probably the result of very careful siting procedures under-
taken by native peoples. Such findings suggest that other factors, over and
above the desire to visually dominate a part of the landscape, were at play.
Many of the viewsheds are small and the visual extent of territoriality would
Figure 13.5 Neolithic long barrows with two 'rim' effect viewsheds.
Figure 13.6 Selective Neolithic long barrow viewsheds showing non-intervisibility.
have been minimal. Furthermore, for each cluster of long barrows, one or two
small locations do exist where the barrow viewsheds overlap. The implication
is that within discrete parts of the landscape, long barrows have been deliber-
ately and sensitively sited so as to avoid intervisibility with each other while
permitting one or two common viewpoints of the barrows to exist.
This situation introduces some interesting perspectives related to Appleton's
(1975) theory of prospect and refuge in the landscape. The traditional archaeo-
logical view has been that long barrows provided strong symbols of prospect
within a prospect-dominant landscape. In many respects, the current analysis
indicates a more subtle situation to have existed in which the barrows were
symbols of both refuge and prospect; that they were both symbols of territori-
ality while "hiding" from other long barrows within the landscape. Caution is
clearly required in drawing conclusions from scanty archaeological evidence,
although in this instance there would appear to be a more complex form of
territorialiry being exhibited in this region than was previously envisaged. Rim
effects, sectoral dominance, nonintervisibility, and mutualintervisibility sug-
gest that complex locational decisions underlay the decisions to site individual
barrows. It would appear that viewshed analysis is a more robust and use-
ful measure for identifying territoriality as a basis for sociopolitical units
than Thiessen polygons.
The Bronze Age

It is customary to divide the British Bronze Age into at least two parts, reflect-
ing differences in the archaeological evidence. The Early Bronze Age is as-
signed to approximately the early second millennium B.C., during which indi-
vidual burials beneath round barrows form the most prolific monument type
(Figure 13.7), representing an important change in burial rite. In contrast to
the communal burial of the Neolithic period, the advent of round barrows
226 Antbropology,ormation Systems
Figure 13.7 Early Bronze Age round barrows, ring ditches, and oval barrows.
indicates individual wealth and an increase in the importance of social rank in

Early Bronze Age society. The wide distribution and large number of round
barrows shown in Figure 13.7, together with their ploughed-out equivalent,
the ring ditches, indicates greater organized use of the landscape and increased
clearance of the original woodland. The location of round barrows is clearly
linked to the topography, with many being placed along ridge tops and on
prominent locations within the landscape. The clustering of barrows into
cemeteries is also apparent.
Around the middle of the second millennium, a fundamental change oc-
curs in the archaeological record with the emphasis switching from funerary
and ceremonial monuments to settlement sites and evidence for more formal
land organization, such as field systems and linear ditches (Figure 13.8). Lin-
ear ditches are generally thought to have had a dual function, representing
land-management boundaries and routeways across the landscape. The prox-
imity of many of these linear ditches to ridge tops in the landscape, evident in
Figure 13.8, would seem to support both these contentions. The field evi-
dence for the Later Bronze Age is similarly concentrated on the high ground.
The figure also shows the way in which many of the barrows are apparently
located around the edges of the new land divisions, suggesting that the build-
ers showed some respect for earlier burial practices. The enclosed settlements
of this period contain one or more round houses and subsidiary buildings
enclosed within a bank and ditch and are probably the farmsteads of extended
family groups. Taken together, the greater, more formal organization of the
landscape during the Late Bronze Age suggests that a change occurred in
Bronze Age society to move away from the conspicuous display of personal
wealth and prestige within funerary practices toward that of land ownership
and organizational power. In addition, it appears that large areas of the ear-
lier woodland cover had been removed by this period.
Figure 13.8 Early Bronze Age barrows and Late Bronze Age field systems, linear
ditches, and enclosed settlements.
The Iron Age
By the beginning of the British Iron Age in the seventh century B.C., further
changes in the archaeological record occur involving substantial settlement
expansion in the form of hillforts and smaller, more numerous, farmsteads
(Figure 13.9). A direct comparison of the archaeology of the Late Bronze Age
(Figure 13.8 without the round barrows) and of the Iron Age (Figures 13.9
and 13.10) suggests that a sudden, dramatic increase in the area of organized
landscape took place. Evidence of a large-scale movement of human activity
from the central highlands into other parts of the landscape is apparent, but
the implication of a sudden and dramatic expansion of human activity is highly
Figure 13.9 Iron Age hillforts and settlements.

misleading. The evolution of landscape change is here subsumed within a coarse
categorical temporal framework, and extreme caution must be exercised. As
will become apparent, a refined temporal classification is crucial in order to
understand the relative development of hillforts and settlements within the
Danebury region, and such a finding must also reflect upon other archaeo-
logical periods in this study.
Fortunately, extensive field excavation and typological studies have enabled
a much better archaeological chronology to be determined for the Iron Age.
Palmer's chronology of the Iron Age has been disaggregated within the GIS
into finer temporal periods comprising "very early" (seventh century B.C.),
"early" (sixth to fourth centuries B.C.), "middle" (third to second centuries
B.C.), and "late" (first century B.C.). Additional categories comprise "continu-
ous" occupation and "of unknown date." Woolbury hillfort, for example, can-
not be assigned to any specific phase of the Iron Age and falls within the later
category. Linear ditches and field systems are also difficult to date and have,
by necessity, been assigned to the "continuous" category (Figure 13.10).
Several themes have emerged from studies on the later prehistoric land-
scape of southern England. First, traditional locational analysis and statistical
analysis have sought to interpret the spatial relationships between hillforts
and other types of contemporaneous settlement in terms of the politically
and economically defined hillfort territories dominating smaller, second-or-
der settlements. Only in a few instances has any kind of contextual environ-
mental information been included in these analyses. A second theme com-
prises a detailed economic analysis of territoriality based on site catchment
analysis (SCA) and examines land quality and other aspects of the immediate
environment to establish the economic viability of hillforts and farmsteads
within a hypothesized sphere of influence. A third, more recent theme repre-
sents a movement away from the traditional spatio-environmental emphasis
toward the concepts of social theory. These studies seek to integrate knowl-
Figure 13.10 Iron Age hillforts, settlements, linear ditches and field systems.
edge of the social and political structure of Late Iron Age society into a spa-
tial and environmental framework.
Early analyses of Iron Age hillforts in England concentrated on the precur-
sor to the first of these themes and sought to understand hillfort phenomena
through an examination of simplified distribution maps. Dyer's (1961) study
moved beyond this stage to suggest that hillfort territories were bounded by a
combination of natural linear features and man-made boundaries and that in
general each territory, between 5 and 9 sq mi in area, contained a spring, ar-
able land, hill pasture, and woodland. Subsequent analyses focused heavily on
the application of formal locational models to hillfort distributions in south-
ern England. These studies have advanced concepts of an Iron Age settlement
hierarchy in which hillforts functioned as higher-order central places sur-
rounded by lower-order farmsteads. Various socioeconomic-military interpre-
tations have been suggested to account for these territorial patterns. Clarke's
(1968) study provided the first published research to link Raggett's work on
locational analysis to Iron Age hillfort distributions. By imposing rigid hex-
agonal lattices over the Iron Age landscape, Clarke suggested that varying
sizes of hillfort territory existed for different tribal groupings and used devia-
tions from a hypothesized pattern as the starting point for a socioeconomic
interpretation. Clarke's later study (1972a) again pioneered the use of locational
models. Here a third-order settlement pattern was developed according to
Christaller's "administrative principle" in which lower-order farmsteads owed
allegiance to higher-order hillforts which maintained that allegiance by eco-
nomic, social, political, and military means.
Subsequent studies have continued to mine central place theory to explain
the placement of hillforts in the Iron Age landscape (Collis 1977). Clarke's
studies were quickly refined and adapted to take account of hillfort chronol-
ogy and size. Cunliffe (1971), for example, noted that the apparently uniform
distribution of hillforts on the Sussex Downs was significantly altered when
evidence for Late Iron Age occupation was taken into consideration. Indi-
vidual hillfort territories of between 25 to 40 sq mi in area were hypoth-
esized, again influenced by river boundaries. In Wiltshire hillfort "spheres of
influence" or "socioeconomic units" were identified, containing one or more
smaller farmsteads. Hogg (1971) suggested a variant of this approach based
on weighted Thiessen polygons. By a weighting schema based on the en-
closed area within each hillfort, Hogg produced an alternative territorial dis-
tribution for sites south of the River Thames. Using these territories, it was
possible to comment on population densities and to loosely correlate terri-
tory size and underlying geology.
Despite the evolution and development of these ideas, a major criticism of
this earlier work has been the gross oversimplification of complex prehistoric
landscapes into a single series of central points as a basis for constructing
Thiessen territories. The difficulty of integrating contextual data, both envi-
ronmental and archaeological, has severely limited the benefits and potential
of the research undertaken. A disaggregation of the Danebury archaeology as
undertaken here provides a valuable if simple insight into the Iron Age. Fig-
Figure 13.11 Very early and early Iron Age hillforts and settlements.
ures 13.11 to 13.13 represent the disaggregation of the Iron Age into the four
period categories defined earlier. Only the hillfort of Danebury is included on
all three figures on archaeological evidence. Balksbury was in use in the very
Early Iron Age (Figure 13.11), though it was soon replaced by nearby Bury
Hill. The latter—together with Quarley, Figsbury, Danebury, and possibly
the undated Woolbury—made up the Early Iron Age hillforts in the region
(Figure 13.11). Noticeably, the early-period farmstead sites predominantly clus-
tered around Figsbury and Quarley hillforts, with very few around Danebury.
Danebury and possibly Woolbury made up the middle-period hillforts (Fig-
ure 13.12), during which many new settlements appeared around Danebury,
and these subsequently continued in use into the late period. In the Late Iron
Age (Figure 13.13), Bury Flill is reestablished and—along with Danebury,
Figure 13.12 Middle Iron Age hillforts and settlements.

Figure 13.13 Late Iron Age hillforts and settlements.
Woolbury, and the new hillfort at Boscombe Down West—formed the surviv-
ing hillforts at the close of the Iron Age.
Based on the disaggregated Iron Age archaeology, environmental features,
terrain, and the hypothesized relationships between hillforts, settlements, lin-
ear features, and field systems, the sequence of hillfort and farmstead occupance
in the Danebury region can perhaps be viewed in terms of the evolving domi-
nance of Danebury over the surrounding region. The Early Iron Age period
appears to have been one of competition between the five hillforts that had
arisen equispaced across the study area. In line with the earlier studies, there is
certainly visual evidence for a hierarchy in which low-order farmsteads devel-
oped social, economic, and military ties with the nearest higher-order hillfort.
Danebury would appear to exhibit a relationship with the surrounding field
systems and linear ditches, suggesting that the territory roughly delimited by
the River Test in the east and the ridge to the west of the Wallop Brook could
be economically and socially related to the hillfort (Figure 13.14). Similar claims
could be made with respect to the other hillforts and their surrounding areas.
The acknowledged problems of boundary effects, however, and the influence
of hillforts lying beyond the edges of the study area could have a bearing on
these findings, though marginally so.
The Middle Iron Age period is characterized by the succession and domi-
nance of Danebury hillfort over all the surrounding hillforts that were aban-
doned during this time. The rise to dominance of Danebury is accompanied
by the establishment of many new farmsteads around the hillfort. It is hypoth-
esized that Danebury exerted social, military, and economic influence over
these lower-order centers. While Danebury was dominant during the Middle
Iron Age, the outlying areas experienced a reemergence of hillforts during the
late period of the Iron Age. Cunliffe (1991) contends that the establishment,
abandonment, and reestablishment of hillforts and farmsteads in the Danebury
region reflects a dynamic process of regional hillfort dominance and the influ-
Figure 13.14 Early Iron Age hillfbrts, settlements, linear features, field systems, and
Thiessen Polygons (based on the hillforts).
ence of ever greater spheres of influence. Grant (1986) in his territorial recon-
struction of hillforts has suggested a similar hypothesis in which hillfort domi-
nance was a function of site size, which in turn was a surrogate for population
size. During the Middle and Late Iron Age, Danebury formed the peak of a
central-place hierarchy in which the surrounding hillforts and settlements were
at lower levels in the system. In addition to these regional central-place func-
tions, long-distance trade placed Danebury within a much larger network of
hillfort trading centers within southern England. It can be suggested that the
reemergence of hillforts during the late period of the Iron Age represented a
recolonization of the immediate region and a visible extension of the Danebury
power base reaching out to dominate and exploit its outlying territory.
Changing Scale: Territoriality and Iron Age Hillforts
Within the power politics of hillfort development, the relationships that ex-
isted between individual hillforts and their surrounding hinterland provides
an important theoretical theme within later prehistoric landscape archaeol-
ogy. The role of hillforts as higher-order central places serving lower-order
farmsteads has already been discussed. A change in scale, however, from the
supraregional to the microlevel can be revealing of other relationships existing
between hillforts and their surrounding farmsteads and field systems. The nature
of human-environment relationships and the extent of economic potential are
also crucial elements in understanding Iron Age society and are more readily
revealed at the microscale. The early period hillfort of Figsbury, and Danebury
in the middle period, are examined in detail to explore many of these points.
The Early Iron Age hillfort of Figsbury was positioned to attain maximum
visual dominance over the adjacent valley, and surrounding farmsteads (Fig-
ure 13.15). The extent to which maximum defensive protection was waived in
Figure 13.15 Figsbury hillfort, settlements, linear features, field systems, soils,
and hydrology.
order to obtain maximum visibility over the surrounding area is not known.
Visual dominance over the local populace has clear symbolic and practical uses
in exerting and maintaining power over an area, especially in a feudal society.
The viewshed of Figsbury hillfort clearly dominates the adjacent valley system
and farmsteads, yet it is severely lacking in visibility from other points of the
compass (Figure 13.16). All-round visibility from the hillfort to meet defen-
sive needs would appear to have been deferred in favor of maximizing visibil-
ity over the valley and farmsteads. The Early Iron Age hillfort of Quarley is
similar in its setting to Figsbury in that it maximizes visual dominance of the
adjacent farms and field systems at the cost of all-round defensive capability.
Danebury hillfort similarly emphasizes the apparently crucial importance of
visual dominance over the surrounding farmsteads (Figure 13.17). Here the
Figure 13.16 Figsbury hillfort and viewshed over surrounding farmsteads.

Figure 13.17 Danebury hillfort viewshed and surrounding farmsteads.
correlation between viewshed and the distribution of farms, field features, and
linear ditches is remarkable. Danebury's viewshed emphasizes the hillfort's
prospect looking predominantly toward the west, though farms immediately
to the east and south come under its coverage. Noticeably, the valley of the
River Test to the east does not fall within the defensive viewshed of Danebury.
A cost surface for the Danebury area based on a function of distance and slope
from the hillfort indicates that not only was visibility a critical factor in the
relationship between hillfort and surrounding districts but also that accessibil-
ity between hillfort, farms, and field systems was a key factor in the develop-
ment of the area (Figure 13.18).
Site catchment analysis (SCA) (Ellison and Harriss 1972; Gaffney and Stancic
1991, 1992; Jarman et al. 1982; Limp 1990; Vita-Finzi and Higgs 1970) has
Figure 13.18 Danebury least effort distance surface.

provided a valuable insight into prehistoric landscape and settlement patterns
through a consideration of the economic viability of land and site suitability.
The precise determination of site catchment size has been heavily influenced
by the work of Chisholm (1968) into the distances involved in the working
regime of farming communities. Ellison and Harriss used a 2-km radius
catchment to represent an area of movement beyond which net returns dimin-
ish to such an extent as to adversely affect farming practice and economic
viability. The nature of these catchment areas tends to vary according to land-
use potential and farming technology, thus it differs for sites of different peri-
ods. Late Bronze Age Wiltshire, for example, showed a preference for downland
locations, whereas sites in the Iron Age showed a preference for a combina-
tion of both downland and high-quality arable land (Ellison and Harriss 1972).
Though there was a bias in the latter study toward downland sites because
hillforts were not included, the variation between the Late Bronze Age and
Iron Age has been interpreted as representing a shift toward an infield-out-
field agricultural system in which intensive arable farming was practiced on
the infield and extensive grazing on the more distant downland.
To examine the ideas of site catchment and the spatial relationship between
farmsteads, circular buffers have been placed around each settlement surround-
ing Figsbury and Danebury at distances of 400 and 1000 m radius (Figures
13.19 and 13.20). The buffer sizes are based on distances used in previous
SCA applications. The visual effect is quite revealing, for while there is very
little overlap between the 400-m buffers, the 1000-m buffers do coalesce into
an amalgamation of farmstead sites. The use of buffers would appear to be a
more useful approach than Thiessen polygons in this case (Figure 13.21). The
uneven spatial distribution of farmsteads precludes the imposition of a single
buffer size to differentiate farmstead catchment sizes. However, the 400-m
and 1000-m buffers provide a very useful range within which to evaluate the
spatial patterning of farmsteads in the Danebury microregion. In line with
Figure 13,19 Early Iron Age Figsbury settlements buffered to 400m and 1000m radii.
Figure 13.20 Middle Iron Age Danebury settlements buffered to 400 m and
1000m radii.
earlier ideas, there is a clear suggestion of an infield system under the control
of individual settlements. Accessibility would be an important factor influenc-
ing the level of energy input; for fields close to a farmstead, this could be
anticipated to be high. The 1000-m buffer indicates the overlapping spheres
of farmstead activity and suggests some sharing of communal grazing resources
that are less energy-intensive. Hingley's (1984) work on the importance of
intensive agricultural production as a means of control and exploitation of
territory in Iron Age society is of particular interest in the context of the pat-
terns shown here. He suggested that "property" could be divided into two
forms, comprising actual land division, and control over territory, which re-
sulted from social organization. At the smallest social grouping of the indi-
Figure 13.21 Middle Iron Age Danebury settlements with Thiessen Polygons.
vidual farmstead, "domesticated" property represents land close to the farm-
stead, which is used for animal and crop production. Further away from farm-
steads is the "wild" property that belongs to the larger social group and re-
quires very little energy input compared to the local resources. This commu-
nity land is controlled, exploited, and defended by the broader social group,
which is based on the nearby hillfort. In terms of the relationship between
energy resources and land organization, this appears to be an early precursor
to the later-medieval infield-outfield system.
Comparison of these buffered areas around the farmsteads with the field
systems and linear features surrounding Figsbury and Danebury is equally re-
vealing (Figures 13.15,13.19,13.20,13.22). The linear features appear to pro-
vide extensive interconnectivity between the farmsteads and field systems. The
role of these linear ditches as routeways has been alluded to previously. There
is clear suggestion from these plots that the linear ditches acted as drove roads,
providing access both between farmsteads and to the field systems. Underly-
ing the social organization of the Iron Age landscape is the suggestion that
social rank in Iron Age society was partly a consequence of the ownership and
control of cattle and, therefore, of good pasture. One interpretation of the
linear ditch system so prevalent within the Iron Age landscape is that drove
roads for the movement of cattle connected different areas of pasture and
were an important mechanism for sustaining power, control, wealth, and so-
cial rank in Iron Age society.
It is interesting to note also how the 1000-m buffers are bounded in part by
the River Bourne to the north of Figsbury and the Wallop Brook to the south-
west of Danebury. Iron Age settlements reflect an inherent competition for
resources which resulted in generally regularly spaced settlement patterns (Bra-
dley 1978; Hudson 1969). Natural boundaries, however, did contribute to de-
fining communal resource areas, and the 1000-m buffer may be an accurate
assessment of the spatial extent of farmstead activity and influence. The exist-
Figure 13.22 Middle Iron Age Danebury settlements, field systems, and linear features.
ence of dry valley systems with soil evidence for previous watercourses pro-
vides additional explanation for the location of several adjacent farmsteads
now located at some distance from present-day water sources. The cumulative
evidence from linear ditches, field systems, farmsteads, and hillforts suggests a
well-developed and organized social and economic landscape with defined social
groupings interacting from territories defined by a central hillfort.
There is a very real tendency for archaeological studies using GIS to over-
emphasize the role of the physical environment in influencing human activ-
ity. The return of environmental determinism can already be seen in recent
conference contributions which link GIS and archaeology. This can be un-
derstood, though it is no more acceptable, given the ready availability of physi-
cal environmental information such as topography, slope, aspect, surface wa-
ter, and the inherent difficulties of interpreting social, economic and cultural
influences from the archaeological evidence. The role of complex socioeco-
nomic factors in influencing the Iron Age landscape can only be inferred from
the archaeological record (Gent and Dean 1986; Steponaitis 1978). Tribute
flow between low-order farmstead sites and hillforts, for example, in con-
junction with land productivity, could provide one construct in explaining
settlement patterns surrounding hillforts and the hillfort catchment size. Simi-
larly, anthropological concepts of ethnic groupings and core areas suggests
the importance of cultural interpretations of material evidence (Blackmore et
al. 1979). Hingley's (1984) work, for example, used settlement evidence to
interpret the social relations of production and focused on the differences
between enclosed settlements of the uplands and the denser distribution of
open settlements on valley terraces. The identification of different scales of
perception and control of space at the settlement and communal levels is a
concept relevant to the present study.
The Romano-British Period

While the gradual Romanization of England from 43 A.D. allowed for some
local chieftains to retain much of their existing power base, hillforts as cen-
ters of military, economic, and political power were obvious targets for the
Roman military and economic machine. There is some similarity between
the field systems, linear features, and settlement of the Late Iron Age and
that of the Romano-British period, though there was clearly a contraction in
the extent of landscape use (Figures 13.10 and 13.2 3). It is noticeable that the
roads of the Romano-British period cut across the landscape, paying scant
attention to the earlier social and economic units based on the hillfort cen-
ters. The roads indicate the increasing importance of long-distance move-
ment, especially for military purposes, the developing money economy; and
an economic and political system based on newly constructed Roman town-
ships. In bypassing the Iron Age hillforts the Roman roads exemplify the pass-
ing of one epoch and the advent of another.
Figure 13.23 Romano-British settlement, roads, linear features, and field systems.
Conclusion
This paper has focused on the analytical capability of GIS to support regional
archaeological investigation within the context of developing archaeological
theory. The development of the Danebury landscape has been tracked from
the Neolithic to Romano-British period. The emphasis has been on identify-
ing and explaining the socioeconomic relationships that existed between Iron
Age hillforts and farmsteads. The interpretation of prehistoric relict features
necessarily imposes substantial constraints on interpretation of their true mean-
ing. The transformations within a society are inferred by a cultural interpreta-
tion of landscape archaeology. A hierarchical social organization of Celtic so-
ciety as noted by the classical writers (Crumley 1974) can be seen reflected in
the settlement hierarchy of hillfort and surrounding farmsteads in a client-
patron relationship. This symbiotic relationship enabled a diverse economy to
flourish with hillforts acting as redistribution centers for both locally pro-
duced goods and imported exotic items (Cunliffe 1984). Through an interpre-
tation of the political, social, and economic significance of archaeological fea-
tures in the context of changing human-environment relations, the Danebury
landscape is recreated as a complex dynamic. Danebury contrasts with other
areas because of the richness of the archaeological record. Evidence for a con-
tinuous human presence on the landscape is everywhere, and the powerful
spatial data-handling capabilities of GIS are demonstrated for archaeological
analysis. This study clearly bears witness to the prehistoric landscape as a pal-
impsest of increasing complexity and of evolving social organization.
We contend that for archaeology to benefit from the new technology of
GIS, analyses must be theory-driven rather than hopeful, hi-tech "fishing trips."
Using the exploratory power of GIS, it is possible to examine existing archaeo-
logical theory and to reinterpret the archaeological evidence in the pursuit of
greater understanding and the development of theory. The impression of GIS
as a sterile, atheoretical technology, as feared by some in geography, may not
enter into the archaeological debate on the use of GIS but should be borne in
mind nonetheless (Openshaw 1991; Taylor and Overton 1991).
Despite the many demonstrable advantages of GIS functionality for ar-
chaeological analysis, there are real limitations because of the quality of ar-
chaeological data and the current inability of GIS to handle imprecise spatial
information and dynamic temporal data. Defining the spatial and temporal
nature of an archaeological feature lies at the core of archaeology's main ana-
lytical axes, those of space and time. There is considerable fuzziness in defin-
ing archaeological site boundaries, especially at the regional level. There are
also current limitations in incorporating information that is either poorly de-
fined spatially or has no spatial description at all. For example, social and eco-
nomic ideas of political, religious, and ritual influences, all of which were prob-
ably important factors acting on the visible landscape in prehistoric times, are
important considerations. Similarly the dependence on multilayer GIS cover-
ages to handle the temporal nature of data reflects current limitations in ar-
chaeology itself. The categorical classification of time by archaeologists and
within GIS is unsatisfactory. Temporal GIS, and the ability to deal with tem-
poral fuzziness would greatly enhance the potential of GIS for archaeological
analysis. There are obvious concerns, therefore, regarding how archaeologi-
cal data and ideas can be expressed in a system that demands spatial and tem-
poral exactitude. Similarly, how human perception and cognitive space is
handled within a GIS are important considerations for archaeologists. In much
the same way as an understanding of aboriginal dreamscapes contributes to an
understanding of aboriginal culture and its organization of the landscape, so
archaeologists will need to consider how GIS can be used in modeling prehis-
toric cognitive landscapes. It is within these areas of research rather than in
the repetition of existing applications that the full potential of GIS in archaeo-
logical analysis is to be found.
Acknowledgments
The authors are indebted to Joel Halverson for technical assistance. Gary Lock would
like to thank the organizers of the Santa Barbara conference for funds towards expenses.
14
and Spatial Analysis in the Social Sciences
Michael F. Goodchild
Although geographers have used them for centuries, the idea that they might
be part of the formal or informal process of data analysis and scientific infer-
ence is still novel in many areas of the social sciences. In economics, for ex-
ample, the dominant paradigm emphasizes search for theories and principles
that apply uniformly within human society or its economies, and if geographic
areas such as nation-states are used as units of analysis, their role is more or
less that of statistical samples, each equally representative of a hypothetical
population of all possible nation-states. Within this paradigm (and I do not
wish to suggest that it is the only paradigm within economics or that it is
necessarily inappropriate), a list of nation-states, ordered alphabetically, is more
precise and useful for analysis than a map, since maps tend to lump data values
into coarse classes of color or shade and make it difficult to identify the at-
tributes of small nation-states like Luxembourg or St. Lucia. Tables, on the
other hand, present each nation state's value clearly and precisely.
The chapters in this book represent a striking departure from this ortho-
doxy, since each describes research that depends in some way on a belief in the
importance of spatial context or in looking at data from a spatial perspective.
This might mean merely looking at a map, despite the disadvantages already
noted. Or it might mean direct reference to such primitive spatial concepts as
adjacency, proximity, direction, or coincidence in space. We might choose to
call this examination of data in spatial context GIS or spatial analysis, and the
methods used might be mathematically sophisticated, computationally inten-
sive, or simply intuitive—the chapters in this book cover a wide range of op-
tions. Taken together, however, it is clear that they represent a striking depar-
ture from the aspatial ways of thinking described earlier. It seems appropriate
in this concluding chapter to attempt a summary of the main features and
assumptions of spatial analysis and its limitations and impediments. One of
the aims of GIS is to make spatial analysis easier, more flexible, and more
powerful. This chapter reviews some of these efforts, particularly those that
are likely to be of use to social science.
For the purposes of this discussion, spatial analysis is defined as "a set of
techniques whose results are dependent on the locations of the objects of analy-
sis" (Goodchild 1987). This excludes the conventional application of many
basic statistical techniques such as regression or the calculation of a mean,
since one can freely relocate the objects of analysis without affecting their
attributes. I assume that the definition includes descriptive and intuitive tech-
niques, such as simple map display and the analysis carried out by the eye-
brain when it receives a visual stimulus, as well as the more sophisticated ap-
plications of spatial statistics and operations research.
The first part of the chapter discusses the concepts that underlie spatial
analysis and its implementation in GIS. The second part reviews widely rec-
ognized impediments to spatial analysis and some current research efforts to
remove them. Although GIS is often seen as a tool for support of analysis and
modeling in any science that concerns itself with distribution and differentia-
tion over the surface of the earth, the emphasis in this chapter is strictly on
social science applications.
Concepts of Spatial Analysis
Space as an Organizing Frame
Perhaps the simplest function of a map is as a way of organizing data. Just as a

person's home street address is often used as a form of identification, so locat-
ing an artifact on a plan of a dig provides a way of coping with large amounts
of information that might seem impossibly confusing if presented in the form
of a table. In and of itself, spatial organization of data seems very unlikely to
lead to any kind of insight. But once data is organized spatially (as described,
for example, in Chapter 11), it becomes easy to display, and to gain insight
through more sophisticated concepts of spatial analysis.
Spatial Anomalies
Other things being equal, we might expect geographic space to be uniform

and geographic distributions to be similarly uniform in space. Random pro-
cesses distribute objects over space with uniform density, as when raindrops
fall on an area of dry pavement. But the human eye is remarkably efficient at
scanning otherwise random patterns and detecting anomalies such as the clus-
ters, or areas of anomalously high density, that often occur in the distribution
of cases of a disease like cancer. Spatial anomalies can take the form of abnor-
mally high or low densities or clusters of objects having similar values.
In practice, the human ability to detect pattern can be almost too powerful.
People have spun elaborate stories around imagined patterns in the configura-
tion of stars in the night sky and searched fruitlessly for explanations for apparent
spatial clusters of certain diseases. So in addition to providing the ability to dis-
play data in ways that invite the eye to find patterns, a GIS might also usefully
help the observer by providing access to objective tests of pattern based on sound
statistical principles. In this way, GIS can provide tools that improve on or assist
the observer's skills of human intuition. For an example of the use of GIS to
search for clusters in patterns of disease, see Openshaw and coworkers (1987).
GIS and Spatial Analysis in the Social Sciences 243
Spatial Coincidence
Insight comes not so much from the observation of an anomaly as from the
recognition that it coincides spatially with some other factor and the implica-
tion that this factor is somehow involved in causing the anomaly. As soon as a
cancer cluster is observed on a map, the observer begins a rapid and instinctive
process of searching his or her experience for factors that might be more or
less unique to that place. Is there not a high-voltage transformer in that area,
or a factory that emits a large amount of unpleasant smoke? Each of us carries
vast stores of such information about familiar neighborhoods, and we are pow-
erfully equipped to retrieve and associate it given a suitable display of data.
The map itself may offer clues if the disease locations have been plotted on top
of a standard city street map. A GIS raises this potential to an even higher
plane by automating the process of assembling and displaying the data and by
allowing the user to overlay layers of information for areas that may be totally
unfamiliar. At the same time, it may make it too easy for the observer to ignore
the fact that spatial coincidence is not necessarily an indication of cause and
thus to become too easily diverted by spurious geographic associations.
Spatial Proximity
The causal factor for a cancer cluster might be spatially coincident with the
cluster, or it might occur some distance away if we are willing to assume that
the cluster and the cause are somehow linked over space. They might be linked
through the atmospheric movement of pollution if the cluster is downwind of
the source of pollution. Because disease occurrences are typically recorded at
place of residence, the linkage between pattern of disease and causal factor
might be the journey to work, with a cluster spread over the dormitory neigh-
borhoods near a particular contaminated workplace.
One of the best-known examples of this type of spatial thinking is the work
of Dr. James Snow on a cholera outbreak in the Soho district of London in the
1800s (Gilbert 1958). By making a map, Dr. Snow was able to show that deaths
in the outbreak clustered around a particular public water pump, thus imply-
ing that contaminated water from the pump was the source of the infection. In
this case the link between cluster and cause was the collection of water; Dr.
Snow was able to show conclusively that the outbreak was cpnfined to those
who habitually drew water from that pump and approximated by those who
lived closer to that pump than to any other.
Spatial Dependence
Closely related to the concept of spatial proximity is spatial dependence, or

the tendency for "things that are close together in space to be more alike than
more distant things" (Tobler, 1970). The presence of spatial dependence is not
in itself of great interest, since virtually all geographically distributed phe-
nomena display it to some degree and it is impossible to imagine a world in
which spatial dependence is absent; the slightest movement in any direction
would produce a complete change of scene. But some phenomena display
greater spatial dependence than others, and the distance over which depen-
dence is observed varies also. It is sometimes interesting to compare phenom-
ena according to their patterns of spatial dependence or to observe how spatial
dependence changes through time.
On the one hand, spatial dependence is an interesting property of geo-
graphically distributed phenomena that can provide useful insights, as Arnold
and Appelbaum show in Chapter 3. On the other hand, many statistical tests
assume that observations are drawn randomly and independently, and this as-
sumption is clearly invalid for much geographic data. In much of the literature
of spatial analysis, we find spatial dependence being viewed as a problem to be
treated and removed rather than as a useful source of insight, since its pres-
ence causes such problems for conventional statistical reasoning.
Consider the following example. Two points are antipodal if a line between
them passes exactly through the center of the earth. The antipodal point of
Santa Barbara, California, at longitude 120° west and latitude 34° north, is a
point at latitude 34° south and 60° east, located south of Mauritius in the In-
dian Ocean. Roughly one-third of the earth's surface is land and a smaller
proportion is antipodal land, in the sense that pairs of antipodal points are
both on land. If land is randomly distributed over the earth's surface, one would
expect this proportion to be roughly one third times one-third, or one-ninth.
In fact, the proportion is very much smaller, the only substantial areas of an-
tipodal land being in southern South America and northern China. This is
surprising until one realizes that the distribution of land displays a very high
degree of spatial dependence, since it is caused by the movement of very large
continental masses. Spatial dependence and its effects are important issues in
reasoning about spatial patterns and their meaning.
Spatial Heterogeneity
Anselin (1989) argues that two properties make spatial data "special"—spatial
dependence and spatial heterogeneity. Both to some degree question the as-
sumptions of traditional statistical inference. Because of spatial dependence,
the number of truly independent observations in a given sample taken over
geographical space is normally lower than one might expect—spatial depen-
dence lowers the effective number of degrees of freedom in a test. Spatial
heterogeneity, on the other hand, invalidates the assumption that all cases in a
sample are drawn randomly from the same population.
Consider, for instance, a sample of early settlement sites in a small region
of South America. It is likely that one or more gradients exist within the
region—for example, the settlements might be dispersed over an area sloping
upward away from the shore of a lake. The study is confined to the area
covered by a single air photo and includes an examination of every known
site within the area. These are treated as a sample and subjected to stan-
dard statistical tests.
Unfortunately, there is an inherent conflict in this example between actual
practice and the assumptions of the test. Rather than a sample drawn ran-
domly from a population, we have analyzed all of the cases that occur within a
defined area, and the area is differentiated by a gradient. It is not clear what
population is being described when inferences are drawn from analysis of the
sample. Is it the set of all early South American settlements, the ones in a
region of which this study area is typical, or the settlements that might have
occurred had history repeated itself? If it is not possible to identify the popu-
lation, then what is the point of statistical inference?
In practice, spatial heterogeneity is most problematic when one considers
the effects of the choice of study-area boundary on the results of the analysis.
If the air photo had been positioned further upslope, the mix of settlements in
the sample would have been different and the conclusions would have been
affected. Thus, one consequence of spatial heterogeneity—the tendency for
conditions, particularly the parameters of models, to vary geographically—is
an unwanted dependence of the results of investigation on the definition of
the study area. In traditional statistical thinking, the choice of study area merely
defines one random sample from a homogeneous population rather than an-
other. For geographical phenomena, however, it is necessary to think much
more carefully about the effects of the choice of study area and about the
nature of the population that is the subject of statistical inference.
Issues in Spatial Analysis
Spatial analysis, or more generally the examination of data in their spatial con-
text, seems to offer the potential for powerful new insights into the processes that
occur on the geographical landscape, and there is ample evidence of this in the
pages of this book. In recent years, the growth in sophistication and widespread
availability of GIS has led to a revival of interest in many of the methods of spatial
analysis that can be found in a rich literature dating back over many decades
(Goodchild et al. 1992; Heywood 1990). In this climate of enthusiasm, it is easy
to forget that spatial analysis was the subject of a series of powerful critiques,
beginning in the 1970s (Gregory 1978; Johnston 1987), and that these extended
to spatial thinking in general. The previous two sections have already identified
ways in which spatial thinking may be more complex or problematic than first
appearances would suggest. This section reviews some of the other impediments
to spatial analysis, and this leads to the subsequent section on recent efforts to
remove or reduce these impediments dtrough the use of GIS.
Form and Process
Geographical information is primarily cross-sectional—that is, it represents a

"snapshot" at a point in time. Mechanisms of cause and effect, on the other hand,
are strongly temporal, and it is consequently difficult or impossible to deduce
cause and effect from patterns observed in geographical data. Spatial analysis is
primarily analysis of form, whereas understanding requires analysis of process.
There are two simple but distinct ways of producing a cluster of points in
space. One is by contagion—the presence of one case or carrier makes other
cases in the vicinity more likely. Alternatively, a cluster can arise because the
density or likelihood of occurrence varies geographically in response to some
causal factor. Although a number of tests for clustering in a spatial point pat-
tern have been devised (Ripley 1981), none of them is capable of distinguish-
ing between contagion and varying density as the causal mechanism. On the
other hand, it is easy to resolve them if data is available on the temporal
sequence of occurrence.
Although the clustering of cholera cases around the Broad Street pump
strongly suggested water from the pump as the cause of the disease, many
other mechanisms could explain the observed geographic pattern equally well.
Transmission by human contact would also lead to geographic clustering, as
would a causal factor in the environment. It was only when Dr. Snow put a
lock on the pump and the outbreak came to a rapid end that the drinking
water hypothesis was truly confirmed. In such cases, it is probably better to
see spatial analysis as a source of possible hypotheses about cause rather than
as a means of confirmation.
The Ecological Fallacy
The geographical landscape is almost unbelievably complex; the more closely

one looks at most phenomena, the more detail one sees, apparently ad infinitum.
In mapping, this is expressed as the relationship between scale and generaliza-
tion—a map at a scale of 1:24,000 shows more phenomena and more detail
than a map at a scale of 1:100,000. Fractals (Mandelbrot 1982) provide per-
haps the only theoretical framework for modeling the effects of scale and gen-
eralization on the natural landscape.
A similar need for generalization occurs in dealing with social data. Al-
though data are often collected on individuals, for reasons of practical neces-
sity as well as protection of privacy it is more often distributed, mapped, and
analyzed on the basis of reporting zones such as precincts, wards, census tracts,
counties, or nation-states. Choropleth maps present the characteristics of each
reporting zone as if they applied uniformly to the zone's entire area, inviting
the map reader to assume that the zone is more homogeneous than it really is.
Because much data are available in that form, much spatial analysis in social
science uses information aggregated by reporting zone. For example, one might
compare the level of unemployment in each county to the percentage of people
in the county who have not completed high school. Robinson (1950) was among
the first to draw attention to the fact that a positive correlation between these
two aggregate measures is not necessarily evidence that individuals without
high school graduation are more likely to be unemployed. The correlation
indicates that unemployed and people without high school education are likely
to be found in the same areas, but it does not indicate that they are necessarily
the same people. It is an ecological fallacy to make a false conclusion about
individuals from an observation about spatially aggregated data.
GIS and Spatial Analysis in the Social Sciences 24 7
Modifiable Areal Units
Closely related to the ecological fallacy is what is often called the modifiable
areal unit problem (MAUP). While the results of aggregate analysis can be
falsely imputed to individuals or lower levels of aggregation, they are also
strongly dependent on the particular choice of reporting zones, such that it
may be possible to manipulate the outcome of the analysis by manipulating
the zones. Openshaw (1983) was the first to draw attention to this problem in
spatial analysis, and Openshaw and Taylor (1979) provide a striking example.
Using data on the percent over 65 and percent registered Republican voters in
Iowa counties, they were able to produce virtually any result from perfect posi-
tive to perfect negative correlation by manipulating the reporting zone bound-
aries. Fotheringham and Wong (1991) have made a series of detailed simula-
tions of the effects of zone boundaries on a variety of forms of spatial analysis.
Recent Developments in GIS
Because of the valuable applications of GIS in the management of utility com-

pany operations, land records, and natural resources, the growth of interest in
it during the past two decades has been driven at least as much by its use for
information management and inventory as its capabilities for spatial analysis.
Maguire (1991) argues that spatial analysis is only one of three traditions of
GIS. But considering only its scientific applications, it is often argued that
GIS is a set of computerized tools for implementing the techniques of spatial
analysis, just as statistical packages are tools for implementing statistical analysis.
The capabilities of GIS for spatial analysis have grown substantially over
the past few years. Versions 6.1 and 7.0 of ARC/Info (developed and distrib-
uted by Environmental Systems Research Institute, Redlands, CA) added sig-
nificant new functions for spatial analysis, particularly in the types of network
analysis illustrated by Ruggles and Church in their chapter, the measurement
of spatial autocorrelation, and the modeling of spatial interactions. IDRISI
(developed and distributed by The IDRISI Project, Clark University, Worces-
ter, MA) has added new capabilities for analysis in support of decisionmaking
and the management of error and uncertainty in spatial data. Other GIS with
significant spatial analysis capabilities include TransCAD and GisPlus (Cali-
per Corporation, Newton, MA), and Maplnfo (Maplnfo Corporation, Troy, NY).
The next sections review efforts that are in place or under way to address
the impediments to spatial analysis identified earlier.
Spatial Heterogeneity
Traditional uses of statistical methods have emphasized the power of tests to

confirm or deny hypotheses by following rigorous procedures. In the early
days of computing, many of these procedures were implemented on main-
frames, taking over much of the tedium of mechanical calculations. Beginning
in the 1970s, however, with the introduction of interactive computing, new
methods of analysis based more on exploration than confirmation began to
emerge. Interactive techniques, which now dominate computing, are ideally
suited to dealing with problems of spatial heterogeneity. Rather than analyz-
ing all data within a study area, a user of an interactive system can explore the
characteristics of data using user-defined windows. In the example of South
American settlement discussed earlier, an interactive user could "brush" the
map with a window, determining the characteristics of the settlements lying
within the window's current position. Results then become a function of the
(user-controlled) window position rather than a function of the (uncontrolled)
study area boundary. Monmonier (1989) was among the first to draw atten-
tion to the potential of geographical brushing.
A related development is the use of logically linked windows in packages
designed for advanced spatial analysis, such as REGARD (Haslett et al. 1991),
or the work of MacDougall (1992). Here the user is able to display several
representations of data simultaneously, such as a map and a histogram or a
map and a scatterplot, in separate windows on the screen. Moreover, the win-
dows are linked, so that cases highlighted in one window are automatically
highlighted in all other windows. For example, pointing to an outlying case
on a scatterplot causes its position to be highlighted on the linked map,
and "lassoing" a group of cases on a map causes them to be highlighted on
the linked histogram.
Spatial Dependence
Several GIS, including IDRISI and ARC/Info, now offer the ability to com-
pute indices of spatial dependence. The Geary and Moran indices of spatial
autocorrelation are perhaps the most widely used, although the variogram,
which computes spatial dependence as a function of distance, may provide
more insight for suitable kinds of data (Goodchild 1986). Unfortunately, al-
though it has been relatively easy to add the ability to compute such indices,
it has proven much more difficult to provide the user with help in interpreta-
tion or insight into the consequences of spatial dependence. A wide range of
models of spatially dependent data have been proposed, as well as modifica-
tions to statistical tests when spatial dependence is present, but these remain
largely inaccessible to the nonspecialist user. Anselin's SpaceStat (1992) in-
cludes many of these techniques, but it must be linked to a GIS through a
suitable data-transfer mechanism.
Animation
As noted earlier, the time dimension to data is often essential if insight is

to be gained into causal mechanisms. Maps, on the other hand, are static,
as are many geographical data. Recently, efforts have been made to intro-
duce the time dimension into spatial databases (Langran 1992), and to
develop GIS that allow animation (Buttenfield et al. 1991). An animated
display of the temporal sequence of deaths in the Snow cholera data, for
example, could have led to much richer insight into the transmission
mechanism and would presumably have eliminated the possibility of di-
rect transmission between individuals. Unfortunately, many existing GIS
use designs that have been optimized for the display of static data and give
very poor performance in animations.
Reporting Zone Effects
The literature on the ecological fallacy and MAUP gives conspicuously little
help in finding solutions to either of these problems. No simple techniques
can be applied to aggregate data to determine whether a fallacious inference
is being made, and there are no methods for removing the effects of zone
boundaries on the results of analysis. In general, the less aggregated the data
the lower the chance of an ecological fallacy. Similarly, lower levels of aggre-
gation are observed to produce a smaller range of results when reporting
zones are manipulated.
The GIS can offer two forms of assistance in helping to reduce the impact
of reporting zone effects. First, the power of modern computers means that it
is almost always practical to process, analyze, and display data at the lowest
possible levels of aggregation. In the past, the labor-intensive manual meth-
ods used to handle and map data meant that aggregation was often essential.
In today's computer-rich research environment, volume of data has virtually
no effect on cost. Martin and Bracken (1991) provide some striking illustra-
tions of the insights gained by processing U.K. census data at the lowest avail-
able level of aggregation.
Second, while they cannot be removed, reporting zone effects can at least
be explored by using the capability of GIS to repeatedly reaggregate and re-
analyze data. Although this can be done only at levels of aggregation above the
lowest available, the results may still be useful in indicating the importance of
reporting zone effects in a given analysis.
Conclusion
In this concluding chapter I have tried to identify the consistent themes of

spatial analysis that run through many of the earlier chapters of this book.
The GIS can bring useful concepts and techniques to bear on data and can
lead to interesting insights. On the other hand, while the eye-brain combina-
tion is an extraordinarily powerful processor of spatial information, it is easily
misled, particularly when it is aided by statistical techniques whose assump-
tions do not match the nature of geographical data. As Anselin argues, spatial
data are special, and their power to yield insights must be balanced by an
awareness of the issues identified in this chapter.
Many techniques of spatial analysis are already implemented and available
in packages like ARC/Info or IDRISI. Newer techniques, and particularly
those which exploit the capabilities of interactive computing for exploration
of data and animation, are available to date only in specialized research pack-
ages. But many development projects are under way, many of them in con-
nection with NCGIA's Research Initiative 14 on GIS and Spatial Analysis
(Fotheringham and Rogerson 1992), and it is likely that the availability of
interesting and powerful techniques will improve rapidly in the next few years.
References
Abasolo, J.A.
1974 Carta arqueologica de la provincia de Burgos. I Partidos judidales de Belorado
y Miranda de Ebro. Studia Arqueologica No. 33, Valladolid.
Abasolo, J.A. and I. Ruiz Velez
1977 Carta arqueologica de la provincia de Burgos. Partido Judicial de Burgos. Burgos.
Acheson, S.R.
1991 In the wake of the ya'aats' xaatgaay ("iron people"): A study of changing
settlement strategies among the Kunghit Haida. Unpublished Ph.D. dissertation.
Oxford, University of Oxford.
Adams, E.G.
1985 Annual Report of Test Excavations at 5MT165, Sand Canyon Pueblo and Ar-
chaeological Survey in T36N, R18W, Sections 12 and 24, and T36N, Rl 6W, Sections 29
and 3. Crow Canyon Archaeological Center. Submitted to Bureau of Land Man-
agement, San Juan Resource Area Office, Durngo, Colorado.
Adams, E.G.
1986 Report to the National Geographic Society on excavations at Sand Canyon
Pueblo, an Anasazi ceremonial center in Southwestern Colorado. Ms. on file, Crow
Canyon Archaeological Center, Cortez, Colorado.
Adler, M.A.
1994 Population aggregation and the Anasazi social landscape: A view from the
Four Corners. In The Ancient Southwestern Community: Models and Methods for the
Study of Prehistoric Social Organization, edited by W.H. Wills and R.D. Leonard,
pp. 85-101. University of New Mexico Press, Albuquerque.
Adler, M.A.
1992 The upland survey. In The Sand Canyon Archaeological Project: A Progress
Report, edited by W.D. Lipe, pp. 11-23. Occasional Paper 2. Crow Canyon Ar-
chaeological Center, Cortez, Colorado.
Adler, M.A. and M.D. Varien
1991 The changing face of the community in the Mesa Verde region, A.D. 1000-
1300. Paper presented at the Third Anasazi Symposium, October, 1991, Mesa
Verde National Park, Mesa Verde, Colorado.
Aldenderfer, M.
1990 The analytical engine: Computer simulation and archaeological research.
In Archaeological Method and Theory, vol. 3, Schiffer, M. (ed.), pp. 195-248. Uni-
versity of Arizona Press, Tucson.
Aldenderfer, M.
1992 The state of the art: GIS and anthropological research. Anthropology Neivs-
/etter33(5):14.
Allen, K.M., S.W. Green, and E.B.W. Zubrow (editors)
1990 Interpreting Space: GIS and Archaeology. Taylor and Francis, London.
Allen, K.M.
1990 Manipulating space: A commentary on GIS applications. In Interpreting Space:
CIS and Archaeology, edited by K. Allen, S. Green, and E. Zubrow, pp. 197-200.
Taylor and Francis, London.
Almagro, M., J. de Serra Rafols, andj. Colominas
1945 Carta arqueologica de Espana. Barcelona. Madrid.
Altieri, M. (with contributions by R. Norgaard and others)
1987 Agroecology: The Scientific Basis of Alternative Agriculture. Westview, Boulder,
Colorado.
Altieri, M. and S. Hecht (editors)
1990 Agroecology and Small Farm Development. CRC Press, Boca Raton, Florida.
Altschul,J.
1990 Red flag models: The use of modeling in management contexts. In Interpret-
ing Space: GIS and Archaeology, edited by K. Allen, S. Green, and E. Zubrow, pp.
226-238. Taylor and Francis, London.
Anderson, J.R., E.E. Hardy, J.T. Roach, and R.E. Witmer
1976 A Land Use and Land Cover Classification System, for Use with Remote Sensor Data.
USGS Professional Paper 964. U.S. Government Printing Office, Washington, D.C.
Anselin, L.
1989 What Is Special about Spatial Data? Alternative Perspectives on Spatial Data
Analysis. Technical Paper 89-4. National Center for Geographic Information and
Analysis, Santa Barbara, California.
Anselin, L.
1992 SpaceStat. Software Series S-92-1. National Center for Geographic Infor-
mation and Analysis, Santa Barbara, California.
Appelbaum, R.P. and C.G. Arnold
forthcoming) Space and the global economy: How the forces of dispersal and
concentration are reshaping the contemporary Los Angeles garment industry. In
Geographic Information Systems: A Handbook for the Social Sciences, edited by C. Earle,
L. Hochberg, and D. Miller. Basil Blackwell, New York.
Appelbaum, R.P. and E. Bonacich
1993 A tale of two cities: The garment industry in Los Angeles. Report to the
Haynes Foundation, Los Angeles, California.
Appleton, J.
1975 The Experience of Landscape. John Wiley & Sons, London.
Aramburu, C., E. Bedoya, andj. Recharte
1982 Colonizacion en la Amazonia. Centre de Investigacion y Promocion
Amazonico, Lima.
Arnold, C.G.
1993 Space and the ethnic economy: A case study of the Los Angeles garment
industry. Unpublished Ph.D. dissertation in progress, University of California,
Santa Barbara.
Arnold, J.E.
1987 Craft Specialization in the Prehistoric Channel Islands, California. University of
California Press, Berkeley.
Aronoff, S.
1989 Geographic Information Systems: A Management Perspective. WDL Publi-
cations, Ottawa.
References 253
Ashbee, P.
1972 Field archaeology: Its origins and development. In Archaeology and Land-
scape, edited by PJ. Fowler, pp. 38-74. John Baker, London.
Ashbee, P., M. Bell, and E. Proudfoot
1989 The Wilsford Shaft: Excavations 1960-62. English Heritage Archaeological
Report No. 11. HBMC, London.
Atrian, P., C. Escriche, J. Vicente, and A.I. Herce
1980 Carta arqueologica de Espana. Teruel. Teruel.
Awh, R.Y.
1976 Microeconomics: Theory and Applications. John Wiley & Sons, Santa Barbara,
California.
Axelrod, R.
1984 The Evolution of Cooperation. Basic Books, New York.
Babic, I.
1984 Prostor Izmedju Trogira i Splita. Muzej Grada "Irogira, Trogir.
Baker, B.H.
1958 Geology of the Magadi Area. Geological Survey of Kenya, Report No. 42.
Government Printer, Nairobi.
Baker, B.H.
1986 Tectonics and volcanism of the southern Kenya Rift Valley and its influence
on rift sedimentation. In Sedimentation in the African Rifts, edited by L.E. Frostick,
R.W. Renaut, I. Reid, and J.J. Tiercelin, pp. 45-57. Geological Society of London
Special Publication No. 25. Blackwell Scientific Publications, Boston.
Baker, B.H., R. Crossley, and G.G. Coles
1978 Tectonic and magmatic evolution of the southern part of the Kenya Rift
Valley. In Petrology and Geochemistry of Continental Rifts, edited by E.R. Neuman
and I.B. Ramberg, pp. 29-50. Reidel Publishing, Boston.
Baker, B.H. and J.G. Mitchell
1976 Volcanic stratigraphy and geochronology of the Kedong-Olorgesailie area
and the evolution of the Southern Kenya Rift Valley. Journal of the Geological Soci-
ety, London 132:467^184.
Baker, B.H., L.A.J. Williams, J.S. Miller, and FJ. Fitch
1971 Sequence and geochronology of the Kenya Rift Volcanics. Tectonophysics
11:191-215.
Baker, B.H. andj. Wohlenburg
1971 Structure and evolution of the Kenya Rift Valley. Nature 229:538-542.
Baker, B.H., L.A.J. Williams, T.A. Miller, and P.A. Mohr
1972 Rift System of Africa. Special Paper 136. Geological Society of America ,
Boulder, Colorado.
Baksh, M.
1984 Cultual ecology and change of the Machiguenga Indians of the Peruvian
Amazon. Unpublished Ph.D dissertation, University of California, Los Angeles.
Baksh, M.G.
1985 Faunal food as a "limiting factor" on Amazonian cultural behavior: A
Machiguenga example. Research in Economic Anthropology 7:145-175.
Balil, A.
1987 La romanizacion. In 130 anos de Arqueologia madrilena, pp. 135-165.
Band, L. and V Robinson
1986 Automated construction of a hydrologic information system from digital
elevation data. Proceedings of the Geographic Information Systems for Environ-
mental Protection Workshop. United States Environmental Protection Agency
Environmental Monitoring Systems Laboratory (EPA/EMSL), Las Vegas, Nevada.
BarkowJ.H.
1989 The elastic between genes and culture. Ethnology and Sociobiology
10:111-129.
Barlett, P..F.
1980 Agricultural Decision Making: Anthropological Contributions to Rural De-
velopment. Academic Press, New York.
Batovic, S. and S. Kukoc
1986 Podvrsje/Matkov Brig. Arheoloski Pregled (1985) 27:61-63.
Beckerman, S.
1979 The abundance of protein in Amazonia: A reply to Gross. American An-
thropologist 81:533-560.
Bedoya, E.
1986 Intensification y degradation en los sistemas agricolas de la selva alta: El
case del Alto Huallaga. In Estrategias Produaivasy Recursos Naturales en la Amazonia,
edited by E. Bedoya, J. Collins, and M. Painter. Document 9. Centre de
Investigation y Promotion Amazonico, Lima.
Behrens, C.A.
1986a The cultural ecology of dietary change accompanying changing activity
patterns among the Shipibo. Human Ecology 14:367-396.
Behrens, C.A.
1986b Shipibo food categorization and preference: Relationships between indig-
enous and Western dietary concepts. American Anthropologist 88:647-658.
Behrens, C.A.
1989 The scientific basis for Shipibo soil classification and land use: Changes in
soil-plant associations with cash cropping. American Anthropologist 91:83-100.
Behrens, C.A.
1991 Applications of satellite image processing to the analysis of Amazonian cul-
tural ecology. In Applications of Space-Age Technology in Anthropology —Conference
Proceedings, edited by C. Behrens and T. Sever, pp. 9-33. NASA Science and Tech-
nology Laboratory, Stennis Space Center, Mississippi.
Behrens, C.A.
1992 Labor specialization and the formation of cash markets for food in a Shipibo
subsistence economy. Human Ecology 20:435-462.
Behrens, C.A. and T. Sever (editors)
1991 Applications oj Space-Age Technology in Anthropology — Conference Proceedings.
NASA Science and Technology Laboratory, Stennis Space Center, Mississippi.-
Belaunde T.F.
1959 La Conquista del Peru para losPeruanos. Editorial Minerva, Lima.
Bell, M.
1983 Valley sediments as evidence of prehistoric land-use on the South Downs.
Proceedings of the Prehistoric Society 49:119-150.
Bell, T.L. and R.L. Church
1985 Location-allocation modeling in archaeological settlement pattern research:
Some preliminary applications. World Archaeology 19:354-371.
References 255
Bell, T.L., R.L. Church, and L. Gorenflo
1988 Late Horizon regional efficiency in the Northeastern Basin of Mexico: A
location-allocation perspective. Journal of Anthropological Archaeology 7:163-202.
Benac, A.
1986 Utvrdjena praistorijska naselja u zapadnom dijeluJugoslavije. MaterialiXXII
(Odbrambeni sistemi u praistoriji i antici na tlu Jugoslavije), Novi Sad: 22-34.
Bennett, J.W.
1976 The Ecological Transition: Cultural Anthropology and Human Adaptation,
Pergamon, New York.
Bennett, J.W.
1985 The micro-macro nexus: Typology, process, and systems. In Micro and Macro
Levels of Analysis in Anthropology: Issues in Theory and Research, edited by B. DeWalt
and P. Pelto. Westview Press, Boulder, Colorado.
Binford, L.R.
1982 The archaeology of place. Journal of Anthropological Archaeology 3:5-31.
Blackmore, C., M. Braithwaite, and I. Hodder
1979 Social and cultural patterning in the Late Iron Age in Southern England. In
Space, Hierarchy and Society. Interdisciplinary Studies in Social Area Analysis, edited
by B.C. Burnham and J. Kingsbury, pp. 93-112. International Series 59. British
Archaeological Reports, Oxford.
Blaikie, P. and H. Brookfield
1987 Land Degradation and Society. Meuthen, London.
Blumenschine, RJ. and F.T. Masao
1991 Living sites at Olduvai Gorge? Preliminary landscape archaeology results in
the basal Bed II lake margin zone. Journal of Human Evolution 21:451-462.
Blurton-Jones, N.G.
1976 Growing points in human ethology: Another link between ethology and the
social sciences? In Growing Points in Ethology, edited by P.P.G. Bateson and R.A.
Hinde, pp. 427-450. Cambridge University Press, Cambridge.
BodleyJ.H.
1982 Victims of Progress. Cummings, Menlo Park.
BodleyJ.H.
1985 Anthropology and Contemporary Hum.an Problems. Cummings, Menlo Park.
Bonacich, E. and P. Hanneman
1991 A statistical portrait of the Los Angeles garment industry. University of
California-Riverside, Department of Sociology. Riverside, California.
Boots, B.N.
1986 Voronoi (Thiessen) Polygons. Concepts and Techniques in Modern Geography
(CATMOG) No. 45. GeoBooks, Norwich, U.K.
Borobio, M.J.
1985 Carta arqueologica de Soria. Campo de Gomara. Soria.
Boserup, E.
1965 The Conditions of Agricultural Growth: The Economics of Agrarian Change Un-
der Population Pressure. Aldine, Chicago.
Bosworth, W.
1985 Geometry of propagating continental rifts. Nature 316:625-627.
Botkin, D., M. Caswell, J. Estes, and A. Orio
1989 Changing the Global Environment: Perspectives on Hum.an Involvement. Aca-
demic Press, New York.
Bradley, B.A.
1991 Rock art and the perception of landscape. Cambridge Archaeological Journal
1:77-101.
Bradley, B.A.
1974 Preliminary report of excavations at Wallace Ruin, 1969-1974. Southwestern
Lore 40(3 and 4):63-71.
Bradley, B.A.
1984 The Wallace Ruin and the Chaco phenomenon. In Insights into the Ancient
Ones. 2nd ed., edited byJ.H. Berger and E.F. Berger, pp. 122-127. Interdiscipli-
nary Supplemental Educational Programs, Cortez, Colorado.
Bradley, B.A.
1986 1985 Annual Report of Test Excavations at Sand Canyon Pueblo (5MT165).
Crow Canyon Archaeological Center. Submitted to Bureau of Land Management,
San Juan Area Office, Durango, Colorado.
Bradley, B.A.
1987 Annual Report of Excavations at Sand Canyon Pueblo (5MTJ65), Montezuma
County, Colorado, 1986 Field Season. Crow Canyon Archaeological Center. Submit-
ted to Bureau of Land Management, San Juan Area Office, Durango, Colorado.
Bradley, B.A.
1988a Wallace Ruin interim report. Southwestern Lore 54:8-33.
Bradley, B.A.
1988b Annual Report on the Excavations at Sand Canyon Pueblo, 1987 Field Season.
Crow Canyon Archaeological Center. Submitted to Bureau of Land Management,
San Juan Area Office, Durango, Colorado.
Bradley, B.A.
1992 Excavations at Sand Canyon Pueblo. In The Sand Canyon Archaeological Project:
A Progress Report, edited by W.D. Lipe, pp.79-99. Occasional Paper No. 2. Crow
Bradley, R.
1978 The Prehistoric Settlement of Britain. Routledge & Kegan Paul, London.
Braithwaite, M.
1982 Decoration as a ritual symbol: A theoretical proposal and an ethnographic
study in southern Sudan. In Symbolic and Structural Archaeology, edited by I. Hodder,
pp. 80-88. Cambridge University Press, Cambridge.
Braun, D.P. and S. Plog
1982 Evolution of "tribal" social networks: Theory and prehistoric North Ameri-
can evidence. American Antiquity 47:504-525.
Bray, W.
1980 Landscape with figures: Settlement patterns, locational models and politics
in Mesoamerica. Paper presented at the Burg Wartenstein Symposium no. 86.,
Prehistoric Settlement Pattern Studies: Retrospect and Prospect. August 16th-
24th, 1980.
Brew, J.O.
1946 The Archaeology of Alkali Ridge, Southeastern Utah. Papers of the Peabody
Museum of American Archaeology and Ethnology 21. Harvard University, Cam-
bridge, Massachusetts.
Brisbin, J.M. and CJ. Brisbin
1973 Contribution to the sites being excavated in the north McElmo drainage
area of Southwestern Colorado. Ms. on file, Crow Canyon Archaeological Cen-
ter, Cortez, Colorado.
References 257
Brush, S.
1977 The myth of the idle peasant. In Peasant Livelihood, edited by R. Halperin
and J. Dow, pp. 60-78. W.W. Norton, New York.
Brush, S.
1987 Who are traditional farmers? In Household Economies and Their Transforma-
tions, edited by M.D. Machlachlan. University Press of America, Lanham, Maryland.
Burrough, P.A.
1986 Principles of Geographical Information Systems for Land Resources Assessment.
Monographs on Soil and Resources Survey. No. 12. Clarendon Press, Oxford.
Buttenfield, B.P., C.R. Weber, M. MacLennan, and J.D. Elliott
1991 Bibliography on Animation of Spatial Data: A Guide to the Literature. Technical
Paper 91-22. National Center for Geographic Information and Analysis, Santa
Barbara, California.
Butzer, K.W.
1977 Cultural perspectives on geographical space. In Dimensions of Human Geog-
raphy; Essays on Some Familiar and Neglected Themes, edited by K.W. Butzer, pp. 1-
14. University of Chicago Press, Chicago.
Butzer, K.W.
1982 Archaeology as Human Ecology: Method and Theory for a Contextual Approach.
Cambridge University Press, Cambridge.
Butzer, K.W.
1989 Cultural ecology. In Geography in America, edited by G.L. Gaile and C J.
Willmott, pp. 192-208. Merrill Publishing Co., Columbus, Ohio.
Butzer, K.W, G.L. Isaac, J.L. Richardson, and C. Washbourne-Kamau
1972 Radiocarbon dating of East African lake levels. Science 175:1069-1076.
Bye, B.A., EH. Brown, T.E., Cerling, and I. McDougall
1987 Increased age estimate for the Lower Paleolithic hominid site at Olorgesailie,
Kenya. Nature 329:237-239.
Cace, S.
1985 Nekropola u prostoru zajednica. Materiali XX (Sahranjivanje pokojnika sa
aspekta ekonomskih i drustvenih kretanja u preistoriji i antici). Beograd, 65-74.
Cannock, G. and V Cuadra
1990 Politicas de ajuste economico y production agricola en la selva. Debate
Agrario 9:43-67.
Carmichael, D.
1990 GIS predictive modeling of prehistoric site distributions in central Mon-
tana. In Interpreting Space: GIS and Archaeology, edited by K. Allen, S. Green, and
E. Zubrow, pp. 216-225. Taylor and Francis, London.
Carneiro, R.L.
1961 Slash-and-burn cultivation among the Kuikuru and its implications for cul-
tural development in the Amazon Basin. Anthropologica Supplements 2:47-67.
Carr, C.
1984 The nature of organization of intrasite archaeological records and spatial
analytic approaches to their investigation. Advances in Archaeological Method and
Theory 7: 103-222.
Castleford,J.C.A.
1992 Archaeology, GIS, and the time dimension: An overview. In Computer Appli-
cations and Quantitative Methods in Archaeology, edited by G.R. Lock and J. Moffett,
pp. 95-106. International Series S577. British Archaeological Reports, Oxford.
CattJ.A.
1978 The contribution of loess to soils in lowland Britain. In The Effect of Man on
the Landscape: the Lowland Zone, edited by S. Limbrey and J.G. Evans, pp. 12-20.
CBA Research Report 21. Council of British Archaeology, London.
Cerron Rivera, J.
1985 Diagnostic/) de Desatrollo Econdmico de la Selva Central, Provincias de Chanchamayo
y Satipo. Institute Nacional de Desarrollo, La Merced, Peru.
Chagnon, N.
1973 The cultural-ecology of shifting (pioneering) cultivation among the
Yanomamo Indians. In Peoples and Cultures of Native South America, edited by D.
Gross, pp. 126-142. Natural History Press, New York.
Chagnon, N.
1983 Yanomamo: The Fierce People. Holt, Reinhart, and Winston, New York.
Chagnon, N.
1988 Yanomamo: The Fierce People. Holt, Reinhart, and Winston, New York.
Chagnon, N.
1992 Yanom.am'6. (4th ed.). Harcourt, Brace, Jovanovich, Fort Worth, Texas.
Chagnon, N. and R. Hames
1979 Protein deficiency and tribal warfare in Amazonia: New data. Science
203:910-913.
Chapman, J.C., R. Shiel, and S. Batovic
1987 Settlement patterns and land use in neothermal Dalmatia, 1983 A. Journal of
FieldAnhaeology 14:124-146.
Chiang, A.C.
1974 Fundamental Methods of Mathematical Economics. McGraw-Hill, New York.
Chibnik, M.
1981 The evolution of cultural rules. Journal of Anthropological Research 37:256-268.
Childe, V.
1925 The Dawn of European Civilization. London.
Childe, V
1929 The Danube in Prehistory. London..
Chisholm, M.
1968 Rural Settlement and Land Use. Hutchinson, London.
Chomsky, N.
1975 Reflections on Language. Random House, New York.
Chomsky, N.
1980 Rules and Representations. Columbia University Press, New York.
Church, R.L. and C. ReVelle
1974 The maximal location covering problem. Papers of the Regional Science As-
sociation 32:101-118.
Church, R.L. and T.L. Bell
1988 An analysis of ancient Egyptian settlement patterns using location alloca-
tion covering models. Annals of the Association of American Geographers 78:701-714.
CIAT (Centre Internacional de Agricultura Tropical)
1988 Annual Report, Tropical Pastures Program.. Working Document No. 44, 1988.
CIAT, Call.
CIAT (Centre Internacional de Agricultura Tropical)
1985 Land in Tropical America. Cali, Colombia
References 259
Clark, C.
1990 Uncertainty in economics. In Risk and Uncertainty in Tribal and Peasant
Economies, edited by E. Cashdan, pp. 47-63. Westview Press, Boulder.
Clarke, D.L.
1968 Analytical Archaeology. Methuen, London.
Clarke, D.L.
1972a A provisional model of an Iron Age society and its settlement system. In
Models in Archaeology, edited by D.L. Clarke, pp. 801-869. Methuen, London.
Clarke, D.L. (editor)
1972b Models in Archaeology. Methuen, London.
Cliff, A. andj. Ord
1973 Spatial Autocorrelation. Pion, London.
Cochrane, T.T., L.G. Sanchez, L.G. de Azevedo, J.A. Porras, and C.L. Garver
1985 Land in Tropical America. 3 vols. Ciat-Embrapa. Ciat, Cali.
Cohen, A. and C. Thouin
1987 Carbonate sedimentation in Lake Tanganyika. Geology 15:414-418.
Collins, J.L.
1986 Smallholder settlement of tropical South America: The social causes of eco-
logical destruction. Human Organization 45:1-10.
Collins, J.
1987 Labor scarcity and ecological change. In Lands at Risk in the Third World,
edited by P. Little andM. Horowitz, pp. 19-37. Westview Press, Boulder, Colorado.
Collins, J.
1988 Unseasonal Migration: The Effects of Rural Labor Scarcity in Peru. Princeton
University Press, Princeton, New Jersey.
Collis, J.
1977 An approach to the Iron Age. In The Iron Age in Britain—A Review, edited by
J. Collis, pp. 1-7. Department of Prehistory and Archaeology, University of
Sheffield, Sheffield.
Collis, J.
1981 A theoretical study of hillforts. In Hill Fort Studies, edited by. G. Guilbert,
pp. 66-76. Leicester University Press, Leicester.
Conant, EP.
1990 1990 and beyond: Satellite remote sensing and ecological anthropology. In
The Ecosystem. Approach in Anthropology, edited by E. Moran, pp. 3 5 7-388. Univer-
sity of Michigan Press, Ann Arbor.
Conklin, H.
1957 Hanunoo Agriculture. FAO, Rome.
Conklin, H.
1967 Some aspects of ethnografic research in Ifugao. Transactions of the New York
Academy of Sciences 30:99-121.
Cosmides, L. and J. Tooby
1987 From evolution to behavior: Evolutionary psychology as the missing link. In
The Latest on the Best: Essays on Evolution and Optimality, edited by J. Dupre, pp.
277-306. The MIT Press, Cambridge, Massachusetts.
Crawford, O.G.S.
1953 Archaeology in the Field. Phoenix House Ltd., London.
CRIES (Comprehensive Resource Inventory and Evaluation System Project)
1984 Resource Assessment of the Choluteca Department. Michigan State University
and the United States Department of Agriculture, East Lansing, Michigan.
Crown, P.L., J.D. Orcutt, and T.A. Kohler
1990 Pueblo cultures in transition: The northern Rio Grande. Paper presented at
the Crow Canyon Archaeological Center conference "Pueblo Cultures in Transi-
tion: A.D. 1150-1350 in the American Southwest," Cortez, Colorado.
Crumley, C.L.
1974 Celtic Social Structure: The Generation of Archaeologicatty Testable Hypotheses
from Literary Evidence. Museum of Anthropology Anthropological Papers No. 54.
University of Michigan, Ann Arbor.
Cuadrado, E.
1991 El Castro de la Dehesa de La Oliva. Arqueologia, Paleontologia, y Etnografia,
No. 2:189-255. Comunidad de Madrid.
Cunliffe, B.W.
1971 Some aspects of hill-forts and their cultural environments. In The Iron Age
and its Hill Forts, edited by D. Hill and M. Jesson, pp. 53-69. Monograph Series
No. 1. University of Southampton, Southampton.
Cunliffe, B.W.
1984 Danebury. An Iron Age Hillfort in Hampshire, vols. 1 and 2. Research Report
52. Council for British Archaeology, London.
Cunliffe, B.W.
1991 Iron Age Communities in Britain, 3rd ed. Routledge, London.
Cunliffe, B.W. and C. Poole
1991 Danebury. An Iron Age Hillfort in Hampshire, vols. 4 and 5. Research Report
No. 73. Council for British Archaeology, London.
Daly, M. and M. Wilson
1988 Homicide. Aldine de Gruyter, New York.
Davies, N.
1980 The Aztecs: A History. University of Oklahoma Press, Norman. Originally
published 1973, Macmillan Ltd., London.
Davis, S.H.
1977 Victims of the Miracle: Development and the Indians of Brazil. Cambridge Uni-
versity Press, New York.
Deere, C. and R. Wasserstrom
1981 Ingreso familiar y trabajo no agricola entre los pequefios productores de
America Latina y El Caribe. In Agricultura de Ladera en America Tropical Informe
TecnicoNo. 11, edited by A.R. Novoa andJ.L. Posner. CATIE, Turrialba, Costa Rica.
Deino, A. and R. Potts
1990 Single crystal 40Ar/39Ar dating of the Olorgesailie Formation, Southern
Kenya Rift. Journal of Geophysical Research 95:8453-8470.
de Janvry, A. and E. Sadoulet
1989 Investment strategies to combat rural poverty: A proposal for Latin America.
World Development 17:1203-1221.
de Laguna, EK.
1990 Tlingit. InNorthivest Coast, edited by W. Suttles, pp. 203-228. Handbook of
North American Indians, vol. 7. Smithsonian Press, Washington, D.C.
de Laguna, F.K., F.A. Riddell, D.F. McGeein, K.S. Lane, J.A. Freed, and C. Osborne
1964 Archaeology of the Yakutat Bay Area, Alaska. Bureau of American Ethnology
Bulletin 192. Smithsonian Institution, Washington, D.C.
De Lisle, D.D.G.
1978 Effects of distance internal to the farm: A challenging subject for North Ameri-
can geographers. Professional Geographer 30:278-288.
References 261
Denevan, W.M.
1966 A cultural ecological view of former aboriginal settlement in the Amazon
Basin. The Professional Geographer 18:346-351.
Descola, P.
1981 From scattered to nucleated settlement: A process of socioeconomic change
among the Achuar. In Cultural Transformations and Ethnicity in Modern Ecuador,
edited by N. Whitten, pp. 614-646. University of Illinois Press, Urbana.
Deuel, L.
1969 Flights into Yesterday: The Story of Aerial Archaeology. St. Martin's, New York.
DeWalt, B. and K.M. DeWalt
1982 Socioeconomic Constraints in the Production, Distribution, and Consumption of
Sorghum in Southern Honduras. INTSORMIL, Farming Systems Research in South-
ern Honduras, Report No. 1. University of Kentucky Department of Anthro-
pology, Lexington.
DeWalt, B. and PJ. Pelto (editors)
1985 Micro and Macro Levels of Analysis in Anthropology: Issues inTheory and Re-
search. Westview Press, Boulder, Colorado.
DeWalt, B. and S. Stonich
1985 Farming systems research in Southern Honduras. In Fighting Hunger with
Research, edited byJ.F. Winn. University of Nebraska, Lincoln.
Dickson, D.B.
1980 Ancient agriculture and population at Tikal, Guatemala: An application of
linear programming to the simulation of an archaeological problem. American
Antiquity 45:697-712.
Dominguez, A., M.A. Magallon, and P. Casado
1984 Carta arqueologica de Espana. Huesca. Huesca.
Douglas, M.
1966 Purity and Danger. Praeger, New York.
Douglass, A.E.
1929 The secret of the Southwest solved by talkative tree-rings. National Geo-
graphic Magazine 54:737-770.
Dourojeanni, M.
1990 Amazonia: iQue Hacer? Centro de Estudios Teologicos de la Amazonia,
Iquitos, Peru.
Drennan, R.D.
1976 Religion and social evolution in formative Mesoamerica. In The Early
Mesoamerican Village, edited by K. Flannery, pp. 345-368. Academic Press,
New York.
Driver, H. and A. Kroeber
1932 Quantitative expression of cultural relationships. University of California
Publications in American Archaeology and Ethnology 31:211-256.
DyerJ.F.
1961 Dray's Ditches, Bedfordshire, and Early Iron Age territorial boundaries in
the Eastern Chilterns. Antiquaries Journal XLL32-43.
Dykeman, D.D.
1986 Excavations at 5MT8371, an Isolated Pueblo II Pithouse Structure in Montezuma
County, Colorado. Studies in Archaeology No. 2. Division of Conservation Archae-
ology, San Juan County Museum Association, Farmington, New Mexico.
Ebert,J.
1984 Remote sensing applications in archaeology. Advances in Archaeological Method
and Theory 7. Academic Press, New York.
Eddy, F.W, A.E. Kane, and PR. Nickens
1984 Southwest Colorado Prehistoric Context: Archaeological Background and Research
Directions. Office of Archaeology and Historic Preservation, Colorado Historical
Society, Denver.
Ellen, R.F.
1978 Problems and progress in the ethnographic analysis of small scale human
ecosystems. Man 13:290-303.
Ellison, A. and J. Harriss
1972 Settlement and land use in the prehistory and early history of southern
England: A study based on locational models. In Models in Archaeology, edited by
D.L. Clarke, pp. 911-962. Methuen, London.
Emmons, G.
1991 The Tlingit Indians. University of Washington Press, Seattle.
Euler, R.C., G.J. Gumerman, F. Plog, R.H. Hevly, and T.N.V Karlstrom
1979 The Colorado Plateaus: Cultural dynamics and paleoenvironment. Science
205:1089-1101.
Evans, C.
1985 Tradition and the cultural landscape: An archaeology of place. Archaeological
Review from. Cambridge 4:80-94.
Evans, S.T.
1980a A settlement system analysis of the Teotihuacan region, Mexico, A.D. 13 50-
1520. Unpublished Ph.D. dissertation, Department of Anthropology, The Penn-
sylvania State University, University Park.
Evans, S.T.
1980b Spatial analysis of basin ofMexico settlement: Problems with the use of the
central place model. American Antiquity 45:866-875.
Evans, S. and P. Gould
1982 Settlement models in archaeology. Journal of Anthropological Archaeology
1:275-304.
Pagan, B.M.
1984 The Aztecs. W. H. Freeman and Company, New York.
Farley, J.A., F.W. Limp, andj. Lockhart
1990 The archaeologist's workbench: Integrating GIS, remote sensing, EDA and
database management. In Interpreting Space: GIS and Archaeology, edited by K. Allen,
S. Green, and E. Zubrow, pp. 141-164. Taylor and Francis, London.
Fernandez-Galiano, D.
1976 Carta Arqueologica de Alcald de Henaresy su Partido. Alcala de Henares.
Fetterman, J. and L. Honeycutt
1987 The Mockingbird Mesa Stirvey. Cultural Resource Series No. 22. Colorado
State Office, Bureau of Land Management, Denver.
Fletcher, R.
1977 Settlement studies (micro and semi-micro). In Spatial Archaeology, edited by
D.L. Clarke, pp. 47-162. Academic Press, London.
Floyd, M.L. and T.A. Kohler
1990 Current productivity and prehistoric use of Pinion (Pintis edulis, Pinaceae) in
the Dolores Project Area, Southwestern Colorado. Economic Botany 44 :141-156.
References 263
FodorJ.A.
1983 The Modularity of Mind: An Essay on Faculty Psychology. MIT Press, Cam-
bridge, Massachusetts.
Fodor, J.A.
1985 Precis of modularity of mind. The Behavioral and Brain Sciences 8:1-42.
Foley, R.
1981 Off-site Archaeology and Human Adaptation in Eastern Africa. International
Series 97. British Archaeological Reports, Oxford.
Ford, R.I.
1972 An ecological perspective on the Eastern Pueblos. In New Perspectives on the
Pueblos, edited by A. Ortiz, pp. 1-17. School of American Research Book. Univer-
sity of New Mexico Press, Albuquerque.
Foster Chaparro, P.E.
n.d. Analisis del proceso de desarrollo socioeconomico de una microregion: Valle
del Alto Mayo, 1970-85. Unpublished Masters Thesis, Universidad del Pacifico, Lima.
Fotheringham, A.S. and D.W.S. Wong
1991 The modifiable areal unit problem in multivariate statistical analysis. Envi-
ronment and Planning A23:1025-l 044.
Fotheringham, A.S. and P. Rogerson
1992 GIS and Spatial Analysis: Initiative 14 Specialist Meeting Report. Technical
Paper 92-11. National Center for Geographic Information and Analysis, Santa
Barbara, California.
Found, W.C.
1970 Towards a general theory relating distance between farm and home to agri-
cultural production. Geographical Analysis 2:165-176.
Fowler, P.J. (editor)
1972 Archaeology and the Landscape. John Baker, London.
Fox, C.
1932 The Personality of Britain. National Museum of Wales, Cardiff.
FPCN/WWF (Fundacion Peruana para la Conservation de la Naturaleza and World
Wildlife Fund)
1989 El Departamento de Madre de Dios: Informe Preliminar. Ms. on file,
FPCN/WWF.
Frohlick, B. and WJ. Lancaster
1986 Electromagnetic surveying in current Middle Eastern archaeology: Appli-
cation and evaluation. Geophysics 51:1414-1425.
Frostick, L.E. and I. Reid
1986 Evolution and sedimentary character of lake deltas fed by ephemeral rivers
in the Turkana basin, northern Kenya. In Sedimentation in the African Rifts, edited
by L.E. Frostick, R.W. Renaut, I. Reid, and JJ. Tiercelin, pp. 113-125. Geologi-
cal Society of London Special Publication No. 25. Blackwell Scientific Publi-
cations, Boston.
Frostick, C. and I. Reid
1989 Is structure the main control of river drainage and sedimentation in rifts?
Journal of African Earth Sciences 8:165-182.
Fuller, S.L.
1987 Cultural Resource Inventories for the Dolores Project: The Dove Creek Canal
Distribution System. andDaiuson Draw Reservoir. Four Corners Archaeological Project
Report No. 7. Complete Archaeological Service Associates. Submitted to Bureau
of Reclamation, Upper Colorado Region, Salt Lake City.
Gaffney, V. and Z. Stancic
199la Predicting the Past—GIS and Archaeology. Germany.
Gaffney, V. and Z. Stancic
1991b Predicting the past: GIS and archaeology. GIS 4: 427-32.
Gaffney, V and Z. Stancic
1992 Diodorus Siculus and the Island of Hvar, Dalmatia: Testing the text with
GIS. In Computer Applications and Quantitative Methods in Archaeology 1991, edited
by G.R. LockandJ.Moffett, pp. 113-125. BAR International Series S577.Tempus
Reparatum, British Archaeological Reports, Oxford.
Gaffney, V, Z. Stancic, J. Farley, R. Harris, F. Limp, J. Lockhart, and I. WilliamsForth
1991 Geographical information systems, territorial analysis, and prehistoric agri-
culture on the island of Hvar, Dalmatia. Paper presented at the Xllth Congress of
the International Union of Prehistoric and Protohistoric Sciences, Bratislava, 1991.
Gamble, C.
1987 Archaeology, geography and time. Progress in Human Geography 11:227-246.
Gams, I.
1987 Adaptation of the karst land for agrarian use in Mediterranean problems of
research and conservation (a survey). Endine 3:65-70.
Gasse, F., P. Rognon, and F.A. Street
1980 Quaternary history of the Afar and Ethiopian Rift Lakes. In The Sahara and the
Nile, edited byM.A.J. Williams and H. Faure, pp. 361-400. Balkema, Rotterdam.
Gent, H. and C. Dean
1986 Catchment analysis and settlement hierarchy: A case study from pre-Roman
Britain. In Central Places, Archaeology and History, edited by E. Grant, pp. 27-36.
Department of Archaeology and Anthropology, University of Sheffield, Sheffield.
Gifford, E. and A. Kroeber
1937 Culture element distributions: EV, Porno. University of California Publica-
tions in American Archaeology and Ethnology 37:117-254.
Gilbert, E.W.
1958 Pioneer maps of health and disease in England. Geographical Journal
124:172-183.
Gladwin, C.H.
1980 A theory of real life choices: Applications to agricultural decisions. In Agri-
cultural Decision Making: Anthropological Contributions to Rural Development, edited
by P. Barlett, pp. 45-85. Academic Press, New York.
Glassow, M.A.
1977 Population aggregation and systemic change: Examples from the American
Southwest. In Explanation of Prehistoric Change, edited byJ.N. Hill, pp. 185-230.
School of American Research Book. University of New Mexico Press, Albuquerque.
Gleichman, C.L. and P.J. Gleichman
1989 An Archaeological Inventory of Lower Sand Canyon, Montezuma County, Colo-
rado. Native Cultural Services, Boulder, Colorado. Submitted to Bureau of Land
Management, San Juan Resource Area Office, Durango.
Gleichman, C.L. and P.J. Gleichman
1991 Context Document: Great Pueblo Period of the McElmo drainage unit,
A.D. 1075-1300. Ms. on file, Native Cultural Services, Boulder, Colorado.
Gonzalez A.L.
1973 Piano Reconstructive de la Region de Tenochtitlan. Colaboro en elDihujo del Piano
Manuel Najera Zamora. Instituto Nacional de Antropologia e Historia, Mexico.
References 265
Goodchild, M.
1984 Geocoding and geosampling. In Spatial Statistics and Models, edited by G.L.
Gaile and C.J. Willmott, pp. 33-53. D. Reidel, Boston.
Goodchild, M.F.
1986 Spatial Autocorrelation. Geo Books, Norwich.
Goodchild, M.F.
1987 Towards an enumeration and classification of GIS functions. Proceedings,
International Geographic Information Systems (IGIS) Symposium: The Research Agenda
11:67-77. National Aeronautics and Space Administration, Washington, D.C.
Goodchild, M.F.
1989 Accuracy of Spatial Databases. Taylor and Francis, New York.
Goodchild, M.F. and S. Gopal
1989 The Accuracy of Spatial Data Bases. Taylor and Francis, London.
Goodchild, M.F., R.P. Haining, S. Wise, and 12 others
1992 Integrating GIS and spatial data analysis: Problems and possibilities. Inter-
national Journal of Geographical Information Systems 6:407-423.
Gorenflo, L. and N. Gale
1986 Population and productivity in the Teotihuacan Valley: Changing patterns
of spatial association in prehispanic Central Mexico. Journal of Anthropological
Archaeology 5:199-228.
Goudie, A.S.
1987 Geography and archaeology: The growth of a relationship. In Landscape
and Culture, edited byJ.M. Wagstaffe, pp. 11-25. Blackwell, London.
Gould, R.R.
1982 The Mustoe Site: The application of neutron activation analysis in the in-
terpretation of a multi-component archaeological site. Unpublished Ph.D. disser-
tation, Department of Anthropology, University of Texas, Austin.
Grant, E.
1986a Hill-forts, central places and territories. In Central Places, Archaeology
and History, edited by E. Grant, pp. 13-26. Department of Archaeology and Pre-
history, University of Sheffield, Sheffield.
Grant, E. (editor)
1986b Central Places, Archaeology and History. Department of Archaeology and
Prehistory, University of Sheffield, Sheffield.
Green, G.M. and R.W. Sussman
1990 Deforestation history of the eastern rain forests of Madagascar from satel-
lite images. Science 248:212-215.
Green, S.
1990 Approaching archaeological space: An introduction to the volume. In Inter-
preting Space: GIS and Archaeology, edited by K. Allen, S. Green, and E. Zubrow,
pp. 3-8. Taylor and Francis, London.
Gregg, S.A. (editor)
1991 Between Bands and States. Center for Archaeological Investigations, Occa-
sional Paper No. 9. Center for Archaeological Investigations, Carbondale.
Gregory, D.
1978 Ideology, Science, and Human Geography. Hutchinson, London.
Gregory, J.W.
1896 The Great Rift Valley. John Murray, London.
Gregory, J.W.
1921 Rift Valleys and Geology of East Africa. Seeley Service, London.
Griffith, D.A.
1987 Spatial Autocorrelation: A Primer. Association of American Geographers,
Washington, D.C.
Griffith, D.A.
1992 Spatial regression analysis on the PC: Spatial statistics using SAS. Ms.
on file, Department of Geography and Interdisciplinary Statistics Program,
Syracuse University.
Gross, D.
1975 Protein capture and cultural development in the Amazon Basin. American
Anthropologist 77:526-549.
Gross, D.
1983 Village movement in relation to resources in Amazonia. In Adaptive Re-
sponses of Native Amazonians, edited by R. Hames and W Vickers, pp. 429-449.
Academic Press, New York.
Gross, D.
1984 Time allocation: A tool for the study of cultural behavior. Annual Review of
Anthropology 13:519-558.
Gross, D.
1990 Ecosystems and methodological problems in ecological anthropology. In
The Ecosystem Approach in Anthropology, edited by E. Moran, pp. 3 09-319. Univer-
sity of Michigan Press, Ann Arbor.
Grove, A.T., F.A. Street, and A.S. Goudie
1975 Former lake levels and climate change in the Rift Valley of Southern Ethio-
pia. Journal of Geography 141:177-202.
Haase, W.R., IV
1985 Domestic water conservation among the northern San Juan Anasazi.
Southwestern Lore 51:15-27.
HackJ.T.
1942 The Changing Physical Environment of the Hopi Indians of Arizona. Papers of
the Peabody Museum of American Archaeology and Ethnology 35(1). Harvard
University, Cambridge, Massachussetts.
HallasiJ.A.
1979 Archaeological excavation at the Escalante Site, Dolores, Colorado, 1975
and 1976. In The Archaeology and Stabilization of the Dom.inguez and Escalante Ruins,
pp. 197-425. Cultural Resource Series No. 7. Colorado State Office, Bureau of
Land Management, Denver.
Hames, R.
1987 Game conservation or efficient hunting? In The Question of the Commons:
The Culture and Ecology of Communal Resources, edited by B. McCay and J. Acheson,
pp. 92-107. University of Arizona Press, Tuscon.
Hames, R. and W. Vickers (editors)
1983 Adaptive Responses of Native Amazonians. Academic Press, New York.
Harris, M.
1968 The Rise of Anthropological Theory. Crowell, New York.
Harris, M.
1984 Animal capture and Yanomamo warfare: Retrospect and new evidence. Jour-
nal of Anthropological Research 40:183-201.
Harris, T.M. and G.R. Lock
1990 The diffusion of a new technology: A perspective on the adoption of geo-
graphic information systems within UK archaeology. In Interpreting Space, edited
by K. Allen, S. Green, and E. Zubrow, pp. 33-53. Taylor and Francis, London.
References 267
Harris, T.M. and G.R. Lock
1992 Toward a regional GIS site information retrieval system: The Oxfordshire
Sites and Monuments Record (SMR) prototype. In Sites and Monuments: Na-
tional Archaeological Records, edited by C.U. Larsen, pp. 185-199. The National
Museum of Denmark.
Haselgrove, C.
1986 Central places in British Iron Age studies: A review and some problems. In
Central Places, Archaeology and History, edited by E. Grant, pp. 3-12. Department
of Archaeology and Prehistory, University of Sheffield, Sheffield.
Haslett, J., R. Bradley, P. Craig, and A. Unwin
1991 Dynamic graphics for exploring spatial data with applications to locating
global and local anomalies. American Statistician 45:234-242.
Hassan, F.A.
1981 Demographic Archaeology. Academic Press, New York.
Haury, E.W.
1934 Climate and human history. Tree-Ring Bulletin 1:13-15. Museum of North-
ern Arizona, Flagstaff.
Hawkes, C.
1954 Archaeological theory and method: Some suggestions from the old world.
American Anthropologist 56:155-168.
Hayden, B.
1990 Nimrods, piscators, pluckers, and planters—The emergence of food pro-
duction. Journal of Anthropological Archaeology 9:31-69.
Hecht, S.B.
1984 Cattle ranching in Amazonia: Political and ecological considerations. In
Frontier Expansion inAm.azonia, edited by M. Schminkand C. H. Wood, pp. 366-
398. University of Florida Press, Gainesville.
Hegen, E.E.
1966 Highways into the Upper Amazon Basin: Pioneer Lands in Southern Colombia,
Ecuador and Northern Peru. University of Florida Press, Gainsville.
Hegmon, M.M.
1989 Risk reduction and variation in agricultural economies: A computer simula-
tion of Hopi agriculture. Research in Economic Anthropology 11:89-121.
Henley, P.
1982 The Panare: Tradition and Change on the Amazon Frontier. Yale University
Press, New Haven, Connecticut.
Herold,].
1961 Prehistoric Settlement and Physical Environment in the Mesa Verde Area.
University of Utah Anthropological Papers No. 53. University of Utah Press,
Salt Lake City.
Hesse, A., A. Jolivet, and A. Tabbagh
1986 New prospects in shallow depth electrical surveying for archaeological and
pedological applications. Geophysics 51:584-594.
Heywood, I.
1990 Geographic information systems in the social sciences—Introduction. En-
vironment and Planning A 22:849-852.
Hietala, H.J. (editor)
1984 Intrasite Spatial Analysis in Archaeology. Cambridge University Press, Cambridge.
Higgs, I.
972 Papers in Economic Prehistory. Cambridge University Press, Cambridge.
Hill, D. and M. Jesson (editors)
1971 The Iron Age and its Hill-forts. Monograph Series No. 1. University of
Southampton, Southampton.
Hingley, R.
1984 Towards social analysis in archaeology: Celtic society in the Iron Age of the
Upper Thames valley (4000 B.C.). In Aspects of the Iron Age in Central Southern
England, edited by B. Cunliffe and D. Miles, pp. 72-88. Monograph No 2. Oxford
University Committee for Archaeology, Oxford.
Hobler, P.M.
1983 Settlement location determinants: An explanation of some northwest coast
data. In The Evolution of Maritime Culture on the Northeast and Northwest Coast of
America, edited by R.J. Nash, pp. 65-76. Department of Archaeology Publication
11. Simon Eraser University, Burnaby.
Hodder, I.
1982a Symbols in Action. Cambridge University Press, Cambridge.
Hodder, I.
1982b Symbolk and Structural Archaeology. Cambridge University Press, Cambridge.
Hodder, I. and C. Orton
1976 Spatial Analysis in Archaeology. Cambridge University Press, Cambridge.
Hosmer, D.W., Jr. and S. Lemeshow
1989 Applied Logistic Regression. John Wiley & Sons, Inc., New York.
Hodge, M.G.
1984 Aztec City-States, edited byj. Marcus. Memoirs of the Museum of Anthro-
pology, No. 18, Studies in Latin American Ethnology & Archaeology, Vol. 3. Uni-
versity of Michigan, Ann Arbor.
Hogg,A.H.A.
1971 Some applications of surface fieldwork. In The Iron Age and its Hill-Forts,
edited by D. Hill and M. Jesson, pp. 105-125. Monograph Series No. 1. Univer-
sity of Southampton, Southampton.
Hudson, J.C.
1969 A location theory for rural settlement. Annals of the Association of American
Geographers 59:365-381.
INADE (Institute Nacional de Desarrollo)
1988 Cuenca del Rio Palcazu: Sistema de Information Geogrdfico. INADE-
APODESA, Lima.
INE (Insitituto Nacional de Estadistica)
1986 Encuesta Nacional de Hogares Rurales (ENAHR): Resultados Definitives. INE, Lima.
INE-DGD (Direction General de Demografia)
1985 Peru: Proyecciones de Poblacion par Anos Calendarios, Segun Departamentos,
Provinciasy Distritos, Periodo: 1980-1990. Boletfn Especial No. 9. INE, Lima.
INE-ORELORETO
1987 Compendia Estadistico, Loreto, 1985, 1987. Oficina Regional de Estadistica
de Loreto. INE, Lima.
INE-OREMAD
1988 Compendia Estadistico, Madre de Dios, 1985, 1988. Oficina Regional de
Estadistica de Madre de Dios. INE, Lima.
INEI (Instituto Nacional de Estadistica e Informatica)
1990 Peru: Compendia Estadistico, 1989-1990. 2 vols. INEI, Lima.
Ingold, T.
1986 The Appropriation of Nature. Manchester University Press, Manchester.
References 269
Irons, W.
1979 Cultural and biological success. In Evolutionary Biology and Human Social
Behavior: An Anthropological Perspective, edited by N. Chagnon and W. Irons, pp.
257-272. Duxbury Press, North Scituate, Massachussetts.
Irons, W.
1990 Let's make our perspective broader rather than narrower: A comment on
Turke's "Which Humans Behave Adaptively, and Why Does It Matter?" and on
the so-called DA-DP Debate. Ethology and Sociobiology 11:361-374.
Irvine, D.
1987 Resource management by the Runa Indians of the Ecuadorian Amazon.
Unpublished Ph.D. Dissertation, Department of Anthropology, Stanford Univer-
sity, Palo Alto, California.
Isaac, G.L.
1977 Olorgesailie: Archaeological Studies of a Middle Pleistocene Lake Basin in Kenya.
University of Chicago Press, Chicago.
Isaac, G.L.
1978 The Olorgesailie Formation: Stratigraphy, tectonics and die palaeogeographic
context of the Middle Pleistocene archaeological sites. In Geological Background of
Fossil Man, edited by W. W. Bishop, pp. 173-206. Scottish Academic Press, Edinburgh.
Isaac, G.L.
1980 Casting the net wide: A review of archaeological evidence for early hominid
land-use and ecological relations. In Current Argument on Early Man, edited by
L.K. Konigsson, pp. 226-251. Pergamon Press, Oxford.
IVTTA (Institute Veterinario de Investigaciones Tropicales y de Altura)
1989 Sistemas de Produccion Amazonicos. Universidad Nacional Mayor de San
Marcos, Lima.
Jarman, M.R., G.N. Bailey, and H.N. Jarman
1982 Early European Agriculture. Cambridge University Press, Cambridge.
Jensen, J.R.
1986 Introductory Digital Image Processing: A Remote Sensing Perspective. Prentice-
Hall, Englewood Cliffs, New Jersey.
Jochim, M.
1981 Strategies for Survival, Cultural Behavior in an Ecological Context. Academic
Press, New York.
Jochim, M.
1983 Optimization models in context. In Archaeological Hammers and Theories,
edited byj. A. Moore and S.A. Keene, pp. 157-172. Academic Press, New York.
Johnson, A.
1982 Reductionism in cultural ecology: The Amazon case. Current Anthropology
23:413-428.
Johnson, A.
1989 How the Machiguenga manage resources: Conservation or exploitation
of nature? In Resource Management in Amazonia: Indigenous and Folk Strategies,
edited by D. Posey and W. Balee, pp. 213-222. New York Botanical Gardens,
Bronx, New York.
Johnson, A.W. and M. Baksh
1987 Ecological and structural influences on the proportions of wild foods in the
diets of two Machiguenga communities. In Food and Evolution: Toward a Theory of
Human Food Habits, edited by M. Harris and E. Ross, pp. 387-405. Temple Uni-
versity Press, Philadelphia.
Johnson, A. and C.A. Behrens
1982 Nutritional criteria in Machiguenga food production decisions: A linear
programming analysis. Human Ecology 10:167-189.
Johnson, A. and C.A. Behrens
1989 Time allocation research and aspects of method in cross-cultural compari-
son. Journal of Quantitative Anthropology 1:313-334.
Johnson, A. and T.K. Earle
1987 The Evolution of Human Societies: From Foraging Group to Agrarian State.
Stanford University Press, Stanford, California.
Johnston, R.J.
1987 Geography and Geographers: Anglo-American Human Geography since 1945.
Arnold, London.
Jones, K.
1991 The Ancient British Goddess. Ariadne, Glastonbury.
Kane, A.E.
1975 Archaeological resources in the Great Cut-Dove Creek area, Dolores Project:
Report of the 1974 Field Season. Ms. on file, Bureau of Land Management, Anasazi
Heritage Center, Dolores, Colorado.
Kaplan, H. and K. Hill
1985 Food sharing among Ache foragers: Tests of explanatory hypotheses. Cur-
rent Anthropology 26:223-246.
Kaplan, S.
1992 Environmental preference in a knowledge-seeking, knowledge-using or-
ganism. In The Adapted Mind, edited byj. Barkow, L. Cosmides, andj. Tooby, pp.
581-598. Oxford University Press, Oxford.
Karson, J.A. and PC. Curtis
1989 Tectonic and magmatic processes in the eastern branch of the East African
Rift and implications for magmatically active continental rifts. Journal of African
Earth Sciences 8:431-453.
Keegan, W.H.
1986 The optimal foraging analysis of horticultural production.. American Anthro-
pologist 28:92-107.
Keene, A.S.
1979 Economic optimization models and the study of hunter-gatherer subsis-
tence setdement systems. In Transformations: Mathematical Approaches to Culture Change,
edited by C. Renfrew and K. Cooke, pp. 369-404. Academic Press, New York.
Keene, A.
1981 Prehistoric Foraging in a Temperate Forest: A Linear Programming Model. Aca-
Keene, A. S.
1983 Biology, behavior, and borrowing: A critical examination of optimal forag-
ing theory in archaeology. In Archaeological Hammers and Theories, edited by J.A.
Moore and A.S. Keene, pp. 137-155. Academic Press, New York.
Keesing, R.M.
1974 Theories of culture. Annual Review of Anthropology 3:73-97.
Keller, G.V and EC. Frischknecht
1966 Electrical Methods in Geophysical Prospecting. Pergamon Press, New York.
Kellerman, A.
1983 Economic and spatial aspects of von Thunen's factor intensity theory. Envi-
ronment and Planning (A) 15:1521-1530.
References 271
King, B.C.
1970 Vulcanicity and rift tectonics in East Africa. In African Magmatism and Tec-
tonics, edited by T.N. Clifford andl.G. Gass, pp. 263-283. Oliver and Boyd, Edinburgh.
King. B.C.
1978 Structural and volcanic evolution of the Gregory Rift Valley. In Geological
Background to Fossil Man, edited by W.W. Bishop, pp. 29-54. Scottish Academic
Press, Edinburgh.
King, B.C. and G.R. Chapman
1972 Volcanism of the Kenya rift valley. Philosophical Transactions of the Royal Soci-
ety of London. Series A 271:185-208.
King, LJ.
1984 Central Place Theory. Sage Publications, Newbury Park, California.
Kintigh, K.
1985 Settlement, Subsistence, and Society in Late Zuni Prehistory. Anthropological
Papers of the University of Arizona 44. University of Arizona Press, Tuscon.
Kleidon, J.
1991 Saddlehorn Hamlet (5MT262). In 1990 Sand Canyon Project Testing Pro-
gram: Preliminary Report on the Excavations at Saddlehorn (5MT262), Mad Dog Tower
(5MT181), Castle Rock Pueblo (5MT1825), Lester's Site (5MT10246), Lookout House
(5MT10459), and Cougar Cub Alcove (5MT1690), edited by K.A. Kukelman, J.
Kleidon, M.D. Varien, and R.R. Lightfoot, pp. 20-41. Crow Canyon Archaeo-
logical Center, Cortez, Colorado.
Kleidon, J. and B. Bradley
1989 Annual Report of the 1988 Excavations at Sand Canyon Pueblo (5MT765). Crow
Kleidon, J. and R.R. Lightfoot
1991 Castle Rock Pueblo (5MT1825). In 1990 Sand Canyon Project Site Testing
Program: Preliminary Report on the Excavations at Saddlehorn (5MT262), Mad Dog
Tower (5MT181), Castle Rock Pueblo (5MT1825), Lester's Site (5MT10246), Lookout
House (5MTW459), and Cougar Cub Alcove (5MT1690), edited by K.A. Kukelman,
J. Kleidon, M.D. Varien, and R.R. Lightfoot, pp. 62-84. Crow Canyon Archaeo-
Kohler, T.A.
1989 Introduction. In Bandelier Archaeological Excavation Project: Research Design
and Summer 1988 Sampling, edited by T.A. Kohler, pp. 1-12. Reports of Investiga-
tions No. 61. Department of Anthropology, Washington State University, Pullman.
Kohler, T.A.
1992a Fieldhouses, villages, and the tragedy of the commons in the early north-
ern Anasazi Southwest. American Antiquity 57:617-635.
Kohler, T.A.
1992b Prehistoric human impact on the environment in the upland North Ameri-
can southwest. Population and Environment 13:255-268.
Kohler, T.A. and M.H. Matthews
1988 Long-term Anasazi land use and forest reduction: A case study from south-
west Colorado. American Antiquity 53:537-564.
Kohler, T.A. and S.C. Parker
1986 Predictive models for archaeological resource location. In Advances in Ar-
chaeological Method and Theory, vol. 9, edited by M.B. Schiffer, pp. 397-452. Aca-
Kohler, T.A. and C.R. Van West
1992 The calculus of self-interest in the development of cooperation: Sociopolitical
development and risk among the Northern Anasazi. Paper presented at the Santa
Fe Institute Workshop "Resource Stress, Economic Uncertainty, and Human
Response in the Prehistoric Southwest," Santa Fe.
Krause, A.
1970 The Tlingit Indians. University of Washington Press, Seattle.
Krebs, J.R. and N.B. Davies (editors)
1991 Behavioral Ecology: An Evolutionary Approach. 3rd ed. Blackwell Scientific
Publications, Oxford.
Krebs, J.R. and A. Kacelnik
1991 Decision making. In Behavioral Ecology: An Evolutionary Approach. 3rd ed.,
edited by J.R. Krebs and N.B. Davies, pp. 105-136. Blackwell Scientific Pub-
lications, Oxford.
Kroeber, A.
1939 Cultural and Natural Areas of Native North America. University of California
Publications in Anthropology, vol. 38.
Kuckelman, K.A.
1991 Lester's Site (5MT10246). In 1990 Sand Canyon Project Site Testing
Program-.Preliminary Report on the Excavations at Saddlehom (5MT262), Mad Dog
Tower (5MT181), Castle Rock Pueblo (5Mtl825), Lester's Site (5MT10246), Lookout
House (5MT10459), and Cougar Cub Alcove (5MT1690), edited byK.A. Kuckelman,
J. Kleidon, M.D. Varien, and R.R. Lightfoot, pp. 85-144. Crow Canyon Archaeo-
Kuckelman, K.A. and J.N. Morris
1988 Archaeological Investigations on South Canal. 2 vols. Four Corners Archaeo-
logical Project Report No. 11. Complete Archaeological Service Associates. Sub-
mitted to Bureau of Reclamation, Upper Colorado Region, Salt Lake City.
Kuznar, L.A.
1991 Mathematical models of pastoral production and herd composition in tradi-
tional Andean herds. Journal of Quantitative Anthropology 3:1-17.
Kvamme, K.L.
1983 Computer processing techniques for regional modeling of archaeological
locations. Advances in Computer Archaeology 1:26-52.
Kvamme, K.L.
1985 Determining empirical relationships between the natural environment and
prehistoric site locations: A hunter-gatherer example. In For Concordance in Ar-
chaeological Analysis: Bridging Data Structure, Quantitative Technique, and Theory,
edited by C.Carr, pp. 208-238. Westport Publishers, Kansas City, Missouri.
Kvamme, K.L.
1989 Geographic information systems in regional archaeological research and
data management. In Archaeological Method and Theory, vol. 1, edited by M.A.
Schiffer, pp.139-204. University of Arizona Press, Tuscon.
Lai, R. and B. Stewart
1990 Soil Degradation. Advances in Agronomy Series 11. Springer-Verlag, New York.
Lambert, P.M.
1993 Patterns of violence in prehistoric hunter-gatherer societies of coastal Cali-
fornia. Paper presented in "Crime and Punishment in Prehistory" Symposium,
Annual Meeting of the American Anthropological Association 1993, Washington.
References 273
Lambert, P.M. and H. Maschner
1993 Hunter-gatherer warfare on the west coast of North America. Paper pre-
sented at the 58th Annual Meeting of the Society for American Archaeology, St. Louis.
Langdon, S.
1979 Comparative Tlingit and Haida adaptation to the west coast of the Prince of
Wales Archipelago. Ethnology 18:101-119.
Langran, G.
1992 Time in Geographic Information Systems. Taylor and Francis, London.
Larsen, C.U. (editor)
1992 Sites and Monuments: National Archaeological Records. The National Museum
of Denmark, DKC, Copenhagen.
Lathrap, D.
1968 The "hunting" economies of the tropical forest zone of South America: An
attempt at historical perspective. In Man the Hunter, edited by R.B. Lee and I.
DeVore, pp. 23-29. Aldine Publishing Company, Chicago.
Leach, E.
1973 Concluding Address. In The Explanation of Culture Change: Models in Prehis-
tory, edited by C. Renfrew, pp. 761-771. Duckworth, London.
Leakey, L.S.B.
1952 The Olorgesailie prehistoric site. In Proceedings of the First Pan-African Con-
gress on Prehistory, 1947, edited by L.S.B. Leakey and S. Cole, p. 209. Philosophi-
cal Library, New York.
Leibman.J.C.
1976 Some simple-minded observations on the role of optimization in public
systems decision-making. Interfaces 6:102-108.
Leith, C.J., K.A. Schneider, and C. Carr
1976 Geophysical investigation of archaeological sites. Bulletin of the International
Association of Engineering Geology 14:123-128.
Lekson, S.H.
1991 Settlement pattern and the Chaco region. In Chaco and Hohokam: Prehistoric
Regional Systems in the American Southwest, edited by PL. Crown and WJ. Judge,
pp. 31-55. School of American Research Press, Santa Fe, New Mexico.
Lele, U. and S. Stone
1989 Population Pressure, the Environment, and Agricultural Intensification: Variations on
the Boserup Hypothesis. MADIA Discussion Paper #4. World Bank, Washington, D.C.
Lewarch, D. E.
1978 Locational models and the study of complex societies: A dilemma in data
requirements and research. Western Canadian Journal ofAnthropology 8:75-88.
Light, I. and E. Bonacich
1988 Immigrant Entrepreneurs. University of California Press, Berkeley.
Lillesand, T.M. and R.W. Kiefer
1987 Remote Sensing and Image Interpretation. 2nd ed. Wiley, New York.
Limp, W.F.
1990 Continuous cost movement models. In Applications of Space-Age Technology in
Anthropology, edited by C.A. Behrens and T.L. Sever, pp. 237-250. NASAJohn C.
Stennis Space Center, Mississippi.
Lizot, J.
1977 Population, resources, and warfare among the Yanomami.Maw 12:497-517.
Lock, G.R. and T.M. Harris
1991 Integrating spatial information in computerized SMRs: Meeting archaeo-
logical requirements in the 1990s. In Computer Applications and Quantitative Meth-
ods in Archaeology, edited byK. Lockyearand S. Rahtz, pp. 165-173. International
Series 565. British Archaeological Reports, Oxford.
Lock, G.R. and T.M. Harris
1992 Visualizing spatial data: The importance of Geographic Information Sys-
tems. In Archaeology and the Information Age: A Global Perspective, edited by P. Reilly
and S.P.Q. Rhatz, pp. 81-96. Routledge, London.
Lock, G.R. and J. Moffett (editors)
1992 Computer Applications and Quantitative Methods in Archaeology. International
Series S577. British Archaeological Reports, Oxford.
Logatchev, N.A., W Beloussov, and E.E. Milanovsky
1972 East African rift development. Tectonophysics 15:71-81.
Logather, N.A. and Y.A. Zorin
1987 Evidence and causes of the two-stage development of the Baikal Rift.
Tectonophysics 143:225-234.
Loker, W.
1989 Environment and agriculture in the Peruvian Amazon: A methodological
experiment. Paper prepared for the CIAT-IFPRI Symposium: "Rural Develop-
ment in the Peruvian Amazon: Ecological, Socioeconomic and Technological
Factors." Lima, Peru.
Lowe, C.H.
1964 Arizona's Natural Environment: Landscapes and Habitats. The University of
Arizona Press, Tuscon.
Macdonald, T.
1981 Indigenous response to an expanding frontier: Jungle Quichua economic
conversion to cattle ranching. In Cultural Transformations and Ethnicity in Modern
Ecuador, edited by N. Whitten, pp. 3 5 6-3 8 3.University of Illinois Press, Urbana.
MacDougall, E.B.
1992 Exploratory analysis, dynamic statistical visualization, and geographic in-
formation systems. Cartography and Geographic Information Systems 19:237-246.
Maguire, D.J.
1991 An overview and definition of GIS. In GeographicalInformation Systems: Prin-
ciples and Applications, vol. 1, edited by DJ. Maguire, M.F. Goodchild, and D.W.
Rhind, pp. 9-20. Longman, London.
Maguire, D.J., M. Goodchild, and D. Rhind
1991 Geographical Information Systems: Principles and Applications. John Wiley &
Sons: New York.
Maluquer de Motes, J.
1956 Carta arqueologica de Espana. Salam,anca. Salamanca.
Mandelbrot, B.B.
1982 The Fractal Geometry of Nature. Freeman, San Francisco.
Marble, D.
1990 The potential methodological impact of geographic information systems on
the social sciences. In Interpreting Space: GIS and Archaeology, edited by K. Allen, S.
Green, and E. Zubrow, pp. 9-21. Taylor and Francis, London.
MARNR-ORSTROM
1988 Atlas del Inventario de Tierras del Territorio Federal Am.azonas. Caracas.
References 275
Marovic, I.
1963 Iskopovanja kamenih gomila oko vrela rijeke Cetine g. 1953, 1954 i 1958.
Vjesnik za Arheologiju i Historiju Dalmatinsku 56:5-80.
Marovic, I.
1976 Rezultati dosadasnjih istrazivanja gomila oko vrela rijeke Cetine u god. 1953,
1954, 1958, 1966 i 1968. MaterialiXII. Zadar 55-73.
Marovic, I.
1985 Iskopovanja kamenih gomila u Bogomolju na otoku Hvaru. Vjesnik za
Arheologiju i Historiju Dalmatinsku 78:5-33.
Marovic, I. and B. Covic
1983 Cetinska kultura. In Praistorija Jugoslav enskih zemalja, vol. IV, edited by A.
Benac, pp. 191-241. Sarajevo.
Martin, D., and I. Bracken
1991 Techniques for modelling population-related raster databases. Environment
and Planning (A) 2 3:1069-107 5.
Martin, P.S.
1936 Lowry Ruin in South-western Colorado. Anthropological Series 23(1). Field
Museum of Natural History, Chicago.
Martinez, H.
1983 Los Estudios Acerca de la Migration y Ocupacion Selvdtica Peruana. Amazonia
Peruana 9. Lima.
Maschner, H.D.G.
1992 The origins of hunter and gatherer sedentism and political complexity: A
case study from the northern northwest coast. Unpublished Ph.D. dissertation,
University of California, Santa Barbara.
Maschner, H.D.G.
1993 The origins of northwest coast warfare. Paper presented in "Crime and
Punishment in Prehistory" Symposium. Annual Meeting of the American Anthro-
pological Association 1993, Washington.
Maschner, H.D.G. and J. Stein
1995 Multivariate approaches to site location on the northwest coast of North
America. Antiquity 69: 61-73.
McCall, G.J.H.
1967 Geology of the Nakuru-Thomson 's Falls-Lake Hannington Area. Kenya Geological
Survey Report No. 78. Geological Survey of Kenya. Government Printers, Nairobi.
McCall, M.K.
1985 The significance of distance constraints in peasant farming systems with
special reference to sub-Saharan Africa. Applied Geography 5:325-345.
McNeillJ.D.
1980 Electrical Conductivity of Soils and Rocks. Technical Note TN-5. Geonics
Ltd., Ontario.
Meaden, G.T.
1991 The Goddess of the Stones: The Language of the Megaliths. Souvenir Press, London.
Meggers, B.J.
1954 Environmental limitations on the development of culture. American Anthro-
pologist 56:801-824.
Meggers, B.J.
1971 Amazonia: Man and Culture in a Counterfeit Paradise. Aldine, Chicago.
Milosevic, A.
1986 Bisko kod Sinja. Arheoloski Pregled (1985) 26:60-61.
Ministerio de Agriculture—Oficina Regional de Estadistica
1987 Ucayali: Information Estadistica Bdsica del Sector Agropemario. Region Agraria
XXIII, Pucallpa, Peru.
Ministerio de Agricultura—Region XIII (San Martin)
1988 Plan de desarrollo rural integrado del Departamento de San Martin,
Documento de Trabajo (mimeo). Peru, Tarapoto.
Mithen, S.
1991 Thoughtful Foragers. Cambridge University Press, Cambridge.
Monmonier, M.
1989 Geographical brushing: Enhancing exploratory analysis of the scatterplot
matrix. Geographical Analysis 21:81-84.
Moran, E.F.
1979 Ecological Anthropology. Duxbury Press, North Scituate, Massachusetts.
Moran, E.F.
1981 Developing the Amazon. Indiana University Press, Bloomington.
Moran, E.F.
1986 Anthropological approaches to the study of human impacts. In Natural Re-
sources and People: Conceptual Issues in Interdisciplinary Research, edited by K.
Dahlberg andj. Bennett, pp. 107-128. Westview Press, Boulder, Colorado.
Moran, E.F.
1990 Levels of analysis and analytical level shifting: Examples from Amazonian
ecosystem research. In The Ecosystem, Approach in Anthropology, edited by E. Moran,
pp. 279-308. University of Michigan Press, Ann Arbor.
Moran, E.F.
1991 Human adaptive strategies in Amazonian blackwater ecosystems. American
Anthropologist 93:361-382.
Morren, G.
1991 New technology and regional studies in human ecology: A Papua New Guinea
example. In Applications of Space-Age Technology in Anthropology, edited by C. Behrens
and T. Sever, pp. 137-166. NASA Science and Technology Laboratory, John C.
Stennis Space Center, Mississippi.
Morris, J.N.
1986 Monitoring and Excavation atAulston Pueblo (Site 5MT2433), a Pueblo II Habi-
tation Site. Four Corners Archaeological Project Report No. 6. Complete Archaeo-
logical Service Associates. Submitted to Bureau of Reclamation, Upper Colorado
Region, Salt Lake.
Morris, J.N.
1991 Archaeological Excavations on the Hovemveep Laterals. Four Corners Archaeo-
logical Project Report No. 16. Complete Archaeological Service Associates. Sub-
mitted to Bureau of Reclamation, Upper Colorado Region, Salt Lake.
Morris, R.
1989 The prehistoric rock art of Great Britain: A survey of all sites bearing motifs
more complex than simple cup-marks. Proceedings of the Prehistoric Society 55: 45-88.
Moss, M.
1989 Archaeology and cultural ecology of the prehistoric Angoon Tlingit. Un-
published Ph.D. dissertation, University of California, Santa Barbara.
Mufio/,, G.
1981 El castro de la Dehesa de La Oliva. In II Jornadas de Estudio sobre la Provincia
de Madrid, pp. 57-62. Madrid.
References 277
National Aeronautics and Space Administration
1987 ELAS: Earth Resources Laboratory Applications Software, vol. II: User
Reference. NASA, Earth Resources Laboratory, Mississippi.
NCSU (North Carolina State University)
1986 Tropical Soils Program. Annual Report 1984-85. NCSU, Raleigh, North
Carolina.
Newberry, J.S.
1876 Geological report. In Report of the Expiating Expedition from Santa Fe, New
Mexico to the Junction of the Grand and Green Rivers of the West in 1859, edited by
J.N. Macomb. U.S. Government Printing Office, Washington, D.C.
Niblack, A.
1970 The Coast Indians of Southern Alaska and Northern British Columbia. Reprinted.
Johnson Reprint Company, New York. Originally published in 1888, Annual Re-
port of the United States National Museum, pp. 225-386. National Museum,
Washington, D.C.
Nickens, P.R. and D.A. Hull
1982 Archaeological resources of Southwestern Colorado: An overview of the
Bureau of Land Management's San Juan Resource Area. In Archaeological Resources
of Southwestern Colorado, edited by S. Eininger, PJ. Gleichman, D.A. Hull, P.R.
Nickens, A.R. Reed, and D.D. Scott, pp. 1-308. Cultural Resource Series No. 22.
Colorado State Office, Bureau of Land Management, Denver.
Novak, G.
1959 Prethistorijske gomile na Paklenin otocima. Arheoloski Radovi i Rasprave
1:237-244.
Oberg, K.
1973 The Social Economy of the Tlingit Indians. University of Washington Press, Seattle.
OIT/DGE (Organization Internacional del Trabajo y Direction General del Empleo)
1984 Migraciones Laborales: Colonization, Poblacion y Empleo en el Alto Mayo.
Ministerio de Trabajo y Promotion Social, Lima.
ONERN (Oficina Nacional de Recursos Naturales)
1962 Evaluation e Integration del Potential Economico y Social de la Zona de Tingo
Maria-Tocache (Huallaga Central). ONERN, Lima.
ONERN
1982 Mapa de Capacidad de Uso Mayor de las Tierras del Peril. ONERN, Lima.
ONERN
1986 Perfil Ambiental del Peru. ONERN-USAID, Lima.
ONERN and PNUMA (United Nations Environment Program)
In press) Vigilanda Ecologica de los Procesos de Degradation de las Tierras y Desertification
en El Peru (mimeo).
Openshaw, S.
1977 A geographical solution to scale and aggregation problems in region-build-
ing, partitioning, and spatial modeling. Institute of British Geographers Transactions
2 (NS):459-472.
Openshaw, S.
1978 An empirical study of some zone-design criteria. Environment and P/<zw-
Kmg-A10:781-794.
Openshaw, S.
1983 The Modifiable Areal Unit Problem.. Geo Books, Norwich, England.
Openshaw, S.
1991 A view on the GIS crisis in Geography, or using GIS to put Humpty-Dumpty
back together again. Environment and Planning, A23:621-628.
Openshaw, S. and P.J. Taylor
1979 A million or so correlation coefficients: Three experiments on the modifi-
able areal unit problem. In Statistical Methods in the Spatial Sciences, edited by N.
Wrigley, pp. 127-144. Pion, London.
Openshaw, S., M. Charlton, C. Wymer, and A.W. Craft
1987 A Mark 1 geographical analysis machine for the automated analysis of point
data sets. International Journal of Geographical Information Systems 1:335-358.
Orcutt, J.D., E. Blinman, and T.A. Kohler
1990 Explanation of population aggregation in the Mesa Verde region prior to
A.D. 900. In Perspectives on South-western Prehistory, edited by P.E. Minnis and C.L.
Redman, pp. 196-212. Westview Press, Boulder.
Orlove, B.S.
1980 Ecological anthropology. Annual Review of Anthropology 9:23 5-273.
Ortner, S.B
1984 Theory in anthropology since the sixties. Comparative Studies in Society and
History 26:126-166.
OSE (Ministerio de Agricultura-Oficina Sectorial de Estadistica)
1988 Boletin Estadistico del Sector Agi-ario, 1975-1987. OSE, Lima.
Palmer, R.
1984 Danebury, an Inn Age Hillfort in Hampshire. An Aerial Photographic Interpre-
tation of its Environs. Supplementary Series 6. Royal Commission on Historical
Monuments, London.
Palmer, W.C.
1965 Meteorological Drought. Research Paper No. 45. U.S. Department of Com-
merce, Office of Climatology, U.S. Weather Bureau, Washington, D.C.
Palol, P. de and F. Wattenberg
1974 Carta arqueologica de Espana. Valladolid. Valladolid.
Parsons, J.R.
1976 The role of chinampa agriculture in the food supply of Aztec Tenochtitlan.
In Cultural Change and Continuity: Essays in Honor of James Bennett Griffin, edited
by C.E. Cleland, pp. 233-257. Academic Press, New York.
Parsons, J.R., E. Brumfiel, M.H. Parsons, and DJ. Wilson
1982 Prehispanic Settlement Patterns in the Southern Valley of Mexico, the Chalco-
Xochimilco Regifm. Memoirs of the Museum of Anthropology No. 14. University
of Michigan, Ann Arbor.
Pascual, A.C.
1991 Carta arqueologica. Soria. Zona Centra. Soria.
PEAH-OSE (Proyect Especial Alta Huallaga-Oficina Secretarial de Estadistica)
1988 Boletin Agricola, 1987. PEAK, Lima.
Perez de Barradas, J.
1921-1922 Yacimientos paleoh'ticos de los Valles de Manzanares y del Jarama
(Madrid). Memorias de la Junta Superior de ExcavacimesyAntiguedades, No. 50. Madrid.
Petersen, K.L.
1988 Climate and the Dolores River Anasazi. University of Utah Anthropological
Papers 113. University of Utah Press, Salt Lake City.
Petric, M.
1986 Izvjestaj o arheolosko-konzervatorskim radovima na rusevini Gazarovic u
Hvaru._Periodimi Izvjestaj Centra za Zastitu Kuhurne Bastine Kom.une Hvarske 156:7-17.
Petric, N.
1979 Hvarski tumuli. Vjesnik za Arheologiju i Historiju Dalmatinsku 72/3 -.67-78.
References 279
Petric, N.
1980 Prilozi paznavanju apulske geometrijske keramike na istocnom Jadranu.
Diadora 9:107-200.
Pingali, PL. and H.P. Binswanger
1987 Population density and agricultural intensification: A study of the evolution
of technologies in tropical agriculture. In Population Growth and Economic Develop-
ment: Issues and Evidence, edited by G. Johnson and R. Lee, pp. 2 7-51. University
of Wisconsin Press, Madison.
Piore, MJ. and C.F. Sabel
1984 The Second Industrial Divide: Possibilitiesfor Prosperity. Basic Books, New York.
Plattner, S.
1989 Markets and marketplaces. In Economic Anthropology edited by S. Plattner,
pp. 171-208. Stanford University Press, Stanford.
Plog,F.T.
1974 The Study of Prehistoric Change. Academic Press, New York.
Plog, F.T., GJ. Gumerman, R.C. Euler, J.S. Dean, R.H. Hevly, and T.VN. Karlstrom
1988 Anasazi adaptive strategies: The model, predictions and results. In TheAnasazi
in a Changing Environment, edited by GJ. Gumerman, pp. 230-276. Academic
Press, New York.
Porter, M.E.
1990 The Competitive Advantage of Nations. Free Press, New York.
Portes, A. and R. Bach
1985 Latin Journey: Cuban and Mexican Immigrants in the United States. University
of California Press, Berkeley.
Posnansky, M.
1959 The Hope Fountain Site at Olorgesailie, Kenya Colony. South African Ar-
chaeological Bulletin 16:83-89.
Posner, J.L. and M.F. MacPherson
1982 Agriculture on the steep slopes of tropical America: The current situation
and prospects. World Development 10:341-354.
Potts, R.
1989a Olorgesailie: New excavations and findings in Early and Middle Pleis-
tocene contexts, southern Kenya rift valley. Journal of Human Evolution 18:477-484.
Potts, R.
1989b Ecological context and explanations of hominid evolution. Ossa 14:99-112.
Potts, R.
1991 Why the Oldowan? Plio-Pleistocene tool making and the transport of re-
sources. Journal of Anthropological Research 47:153-176.
Potts, R. and Behrensmeyer, A.K.
n.d. Paleolandscape variation in early Pleistocene hominid activities. Ms. in pos-
session of authors.
Prance, G.
1982 Forest refuges: Evidence from woody angiosperms. In Biological Diversifica-
tion in the Humid Tropics, edited by G. Prance, pp. 137-157. Columbia University
Press, New York.
Que Hacer
1989 La coca en su laberinto. In Que Hacer, Revista Bimestral del Centre de
Estudios y Promotion del Desarrollo (DESCO). June-July 1989. Lima.
Rabich-Campbell, J.C.
1980 Sedentary seasonal settlements in southeast Alaska: An emerging design. Pa-
per presented at the 7th Annual Meeting of the Alaska Anthropological Association.
Rappaport, R.A.
1968 Pigs for the Ancestors. Yale University Press, New Haven.
Ratzel, F.
1896 The History of Mankind. Macmillan, London.
RCAHMS (Royal Commission on the Ancient and Historic Monuments of Scotland)
1988 Argyll, vol. 6. HMSO, Edinburgh.
Recharte, J.
1982 Prosperidad y pobreza en la agricultura de la Ceja de Selva: el valle de
Chanchamayo. In Colonization en la Amazonia, edited by C. Aramburii, E. Bedoya
and J. Recharte. Centre de Investigation y Promotion Amazonico, Lima.
Redclift, M.
1984 Development and Environmental Crisis. Metheun, New York.
Redclift, M.
1987 Sustainable Development: Exploring the Contradictions. Metheun, New York.
Reid, I. and L.E. Frostick
1986 Slope processes, sediment derivation, and landform evolution in a Rift Val-
ley basin, northern Kenya. In Sedimentation in the African Rifts, edited by L.E.
Frostick, R.W. Renaut, I. Reid, andJ.J. Tiercelin, pp. 99-111. Geological Society
of London Special Publication No. 25. Blackwell Scientific Publications, Boston.
Reidhead, V A.
1979 Linear programming models in anthropology. Annual Review of Anthro-
pology 8:543-578.
Reidhead, V. A.
1980 The economy of subsistence change: A test of an optimization model. In
Modeling Change in Prehistoric Subsistence Economies, edited by T.K. Earle and A.
Christensen, pp. 141-186. Academic Press, New York.
Reining, P.
1979 Challenging Desertification in West Africa: Insights from LANDSAT into Carry-
ing Capacity, Cultivation, and Settlement Sites in Upper Volta and Niger. Center for
International Studies, Ohio University, Athens.
Renfrew, C.
1973 The Explanation of Culture Change: Models in Prehistory. Duckworth, London.
Renfrew, C.
1982 Explanation revisited. In Theory and Explanation in Archaeology, edited by C.
Renfrew, M. Rowlands, and B. Segraves, pp.5-24. Academic Press, New York
Renfrew, C. and P. Bahn
1991 Archaeology: Theories, Metliods, and Practice. Thames and Hudson, London.
Renfrew, A.C. and E.V. Level
1979 Exploring dominance: Predicting polities from centers. In Transformations:
Mathematical Approaches to Culture Change, edited by C. Renfrew and K.L.
Cooke, pp. 145-167. Academic Press, New York.
Revilla, M.L.
1985 Carta arqueologica de Soria. Tierra de Almazan. Soria.
Richards, J.A.
1986 Remote Sensing Digital Image Analysis: An Introduction. Springer-
Verlag, New York.
References 281
Ripley, B.D.
1981 Spatial Statistics. John Wiley & Sons, New York.
Robinson, M.
1984 Landscape and environment of central southern Britain in the Iron Age. In
Danebury, an Iron Age Hillfort in Hampshire, edited by B.W. Cunliffe and D.
Miles, pp. 1-11. Research Report No. 73. Council for British Archaeology, London.
Robinson, W.S.
1950 Ecological correlations and the behavior of individuals. American Sociological
Review 15:351-357.
Rohn, A.H.
1989 Northern San Juan prehistory. In Dynamics of Southwest Prehistory, edited by
L.S. Cordell and GJ. Gumerman, pp. 149-177. Smithsonian Institution Press,
Washington, D.C.
Rose, M.R., WJ. Robinson, and J.S. Dean
1982 Dendroclimatk Reconstruction for the Southeast Colorado Plateaus. Final Report
to Dolores Archaeological Project and Division of Chaco Research. Laboratory of
Tree Ring Research, University of Arizona, Tucson.
Rosendahl, B.R., DJ. Reynolds, P.M. Lorber, C.F. Burgess, J. McGill, D. Scott, JJ.
Lambiase, and SJ. Derksen
1986 Structural expressions of rifting: Lessons from Lake Tanganyika, Africa. In
Sedimentation in the African Rifts, edited by L.E. Frostick, R.W. Renaut, I. Reid,
and JJ. Tiercelin, pp. 29^43. Geological Society of London Special Publication
25. Blackwell Scientific Publications, Boston.
Ross, E.B.
1978 Food taboos, diet and hunting strategy: The adaptation to animals in Ama-
zonian cultural ecology. Current Anthropology 19:1-36.
Rowe, J.
1953 Technical aids in anthropology: A historical survey. In Anthropology Today,
edited by A. Kroeber, pp. 895-940. University of Chicago Press, Chicago.
Rowntree, L.B., K.E. Foote, and M. Domosh
1989 Cultural ecology. In Geography in America, edited by G.L. Gaile and CJ.
Willmott, pp. 209-217. Merrill Publishing Co., Columbus, Ohio.
Rudel, T.K.
1983 Roads, speculators, and colonization in the Ecuadorian Amazon. Human
Ecology 11:3 85-403.
Saffirio, G. and R. Scaglion
1982 Hunting efficiency in acculturated and unacculturated Yanomama villages.
Journal of Anthropological Research 23:315-327.
Sanchez Meseguer, J. et ales
1983 El Neolitico y la Edad del Bronce en la Region de Madrid. Arqueologia y
Paleologia No. 3, Madrid.
Sanders, J.M. and V. Nee
1987 Limits of ethnic solidarity in the enclave economy. American Sociological
Review 52:745-773.
Sanders, W.T. and R.S. Santley
1980 A tale of three cities: Energetics and urbanization in prehispanic central
Mexico. Paper prepared for the Burg Wartenstein Symposium no. 86, Prehistoric
Settlement Pattern Studies: Retrospect and Prospect, August 16-24th, 1980.
Sanders, W.T.
1976a The natural environment of the basin of Mexico. In The Valley of Mexico:
Studies in Prehispanic Ecology and Society, edited by E.R. Wolf, pp. 39-68. Univer-
sity of New Mexico Press, Albuquerque.
Sanders, W.T.
1976b The agricultural history of the basin of Mexico. In The Valley of Mexico:
Studies in Prehispanic Ecology and Society, edited by E.R. Wolf, pp. 101-160. Uni-
versity of New Mexico Press, Albuquerque.
Sanders, W.T, J.R. Parsons, and R.S. Santley
1979 The Basin of Mexico: Ecological Processes in the Evohition of a Civilization. Aca-
Santley, R.
1986 Prehispanic roadways, transport network geometry, and Aztec politico-eco-
nomic organization in the basin of Mexico. Research in Economic Anthropology,
Supplement 2 :Economic Aspects of Prehispanic Highland Mexico, edited by B.L.
Issaac, pp. 223-244. JAI Press, London.
SAS Institute Inc.
1989 SAS/STAT User's Guide, Version 6, 4th ed., vol. 1. SAS Institute Inc., Gary,
North Carolina.
Savage, S.
1990a GIS in archaeological research. In Interpreting Space: GIS and Archaeology,
edited by K. Allen, S. Green, and E. Zubrow, pp. 22-32. Taylor and Francis, London.
Savage, S.
1990b Modeling the late archaic landscape. In Interpreting Space: GIS and Ar-
chaeology, edited by K. Allen, S. Green, and E. Zubrow, pp. 330-355. Taylor
and Francis, London.
Schlanger, S.H.
1985 Prehistoric population dynamics in the Dolores Area, Southwestern Colorado.
Ph.D. dissertation, Washington State University. University Microfilms, Ann Arbor.
Schmidt, W.
193 9 The Cultural Historical Method of Ethnology. Fortuny, New York.
Schminck, M.
1984 Household economic strategies: Review and research agenda. Latin Ameri-
can Research Review 19:87-102.
Schmink, M. and C.H. Wood (editors)
1984 Frontier Expansion in Amazonia. University of Florida Press, Gainesville.
Schorr, T.
1974 A bibliography, with historical sketch. In Aerial Photogivphy in Anthropological
Field Research, edited by E. Vogt, pp. 163-188. Harvard University Press, Cambridge.
Schowengerdt, R.A.
1983 Techniques for Image Processing and Classification in Rem.ote Sensing. Academic
Press, Orlando.
Schultink, G.
1987 The CRIES Geographic Information System: An appropriate technology
interface between thematic cartography and quantitative land evaluation for de-
velopment planning and policy analysis. Geomatics Applied to Municipal Manage-
ment, edited by G. Rochon and J. Cornut, pp. 235-262. L'Association, Montreal.
Schultink, G. and N.C. Amaral
1987 Assessment of agro-ecosystems production potential in developing coun-
tries: The CRIES agro-economic information yield model. Soil Survey and Land
Evaluation 7:187'-197.
References 283
Schumann, D.A. and W.L. Partridge (editors)
1989 The Human Ecology of Tropical Land Settlement in Latin America. Westview
Press, Boulder, Colorado.
Scollar, I., A. Tabbagh, A. Hesse, and I. Herzog
1990 Archaeological Prospecting and Remote Sensing. Cambridge University Press,
New York.
Scott, A.J. and D.J. Mattingly
1989 The aircraft and parts industry in southern California: Continuity and change
from the Inter-war years to the 1990s. Economic Geography 65:48-71.
Sebastian, L.
1991 Sociopolitical complexity and the Chaco system. In Chaco and Hohokam:
Prehistoric Regional Systems in the American Southwest, edited by P.L. Crown and
W.J. Judge, pp. 109-134. School of American Research Press, Santa Fe.
Shackleton, R.M.
1955 Pleistocene movements in the Gregory Rift Valley. Geologische Rundschau
43:257-263.
Shackleton, R.M.
1978 A geological map of the Olorgesailie area. In Geological Background to Fossil
Man, edited by W.W. Bishop, map insert, Scottish Academic Press, Edinburgh.
Shennan, S.
1974 Archaeological "cultures:" An empirical investigation. In The Spatial Orga-
nization of Culture, edited by I. Hodder, pp. 113-140. University of Pittsburgh
Press, Pittsburgh.
Shiel, R. andJ.C. Chapman
1988 The extent of change in theagricultural landscape of Dalmatia, Yugoslavia,
as a result of 7,000 years of land management. In Recent Advances in Yugoslav Ar-
chaeology, edited byj.C. Chapman,]. Bintliff, V Gaffney, and B. Slapsak, pp. 31-
44. International Series 431. British Archaeological Reports, London.
Shoemaker, R.
1981 The Peasants of El Dorado: Conflict and Contradiction in a Peruvian Frontier
Settlement. Cornell University Press, Ithaca.
Siddle, D.J.
1970 Location theory and the subsistence economy: The spacing of rural settle-
ments in Sierra Leone. Journal of Tropical Geography 31:79-90.
Siskind, J.
1973 To Hunt in the Morning. Oxford University Press, London.
Smith, E.A.
1988 Risk and uncertainty in the 'original affluent society': Evolutionary ecol-
ogy of resource-sharing and land tenure. In Hunters and Gatherers 1: History,
Evolution, and Change, edited by T. Ingold, D. Riches, and J. Woodburn, pp.
222-251. Berg, Oxford.
Smith, E.A.
1991 Inujjuam.iut Foraging Strategies. Aldine de Gruyter, New York.
Smith, E.A. and B. Winterhalder (editors)
1992 Evolutionary Ecology and Human Behavior. Aldine de Gruyter, New York.
Smith, M.E.
1979 The Aztec marketing system and settlement pattern in the valley of Mexico:
A central place analysis. American Antiquity 44:110-125.
Smith, M.E.
1980 The role of the marketing system in Aztec society and economy: Reply to
Evens. American Antiquity 45:876-883.
Smith, M.E.
1986 The role of social stratification in the Aztec Empire: A view from the prov-
inces. American Anthropologist 88:70-91.
Smith, R.W.
1984 The ecology of Neolithic farming systems as exemplified by the Avebury
region of Wiltshire. Proceedings of the Prehistoric Society 50:99-120.
Spaulding, A.C.
1960 The dimensions of archaeology. In Essays on the Science of Culture: In Honor of
Leslie A. White, edited by G. Dole and R. Carneiro, pp. 437-456. Thomas Y.
Crowell, New York.
SpethJ.D. and S.L.Scott
1989 Horticulture and large-mammal hunting: The role of resource depletion
and the constraints of time and labor. In Farmers as Hunters: the Implications of
Sedentism, edited by S. Kent, pp. 71-79. Cambridge University Press, Cambridge.
Sponsel, L.E.
1986 Amazon ecology and adaptation. Annual Review of Anthropology 15:67-97.
SPSS Inc.
1990 SPSS/PC+ Advanced Statistics 4.0. SPSS Inc., Chicago.
Star,J. andj. Estes
1990 Geographic Information Systems: An Introduction. Prentice Hall, Englewood
Cliffs, New Jersey.
Stephens, D.W.
1990 Risk and incomplete information in behavioral ecology. In Risk and Uncer-
tainty in Tribal and Peasant Economies, edited by E. Cashdan, pp. 19-46. Westview
Press, Boulder, Colorado.
Stephens, D.W. and J.R. Krebs
1986 Foraging Theory. Princeton University Press, Princeton.
Steponaitis, V.P.
1978 Location theory and complex chiefdoms: A Mississippian example. In
Mississippian Settlement Patterns, edited by B.D. Smith, pp. 417-454. Academic
Press, London.
Stevens, R. and Y. Lee
1979 A spatial analysis of agricultural intensity in a Basotho village of southern
Africa. Professional Geographer 31:177-183.
Steward, J.H.
1938 Basin-Plateau Aboriginal Sociopolitical Groups. Bureau of American Ethnol-
ogy, Bulletin 120. Washington, D.C.
Steward, J.H.
1940 South American cultures: An interpretative summary. In Handbook of South
American Indians, vol. 5, edited by J. Steward, pp. 669-772. Smithsonian Institu-
tion, Washington, D.C.
Steward, J.H.
1955 Theory of Culture Change. University of Illinois Press, Urbana.
Stocks, A.
1983 Cocamilla fishing: Patch modification and environmental buffering in the
Amazonian varzea. In Adaptive Responses of Native Amazonians, edited by R. Hames
and W. Vickers, pp. 239-267. Academic Press, New York.
Stonich, S.C.
1986 Development and destruction: interrelated ecological, socioeconomic, and
nutritional change in Southern Honduras. Unpublished Ph.D. dissertation. Uni-
versity of Kentucky.
References 285
Stonich, S.C.
1989 The dynamics of social processes and environmental destruction: A Central
American case study. Population and Development Review 15:269-296.
Stonich, S.C.
199 la The political economy of food security in Honduras. In Harvest of Want:
Food Security in CentralAmerica and Mexico, edited by S. Whiteford and A. Ferguson,
pp. 45-74. Westview Press, Boulder, Colorado.
Stonich, S.C.
1991b Rural families and income from migration: Honduran households in the
world economy. Journal of Latin American Studies 23:131—161.
Stonich, S.C.
1991c The promotion of nontraditional exports in Honduras: Issues of equity,
environment, and natural resource management. Development and Change 22:725-755.
Stonich, S.C.
1992a Struggling with Honduran poverty: The environmental consequences of
natural resource based development and rural transformation. World Development
20:385-399.
Stonich, S.C.
1992b Society and land degradation: Issues in theory, method, and practice. In
Anthropological Research: Process and Application, edited by J. Poggie, B. Dressier,
and B. Dewalt, pp. 137-158. SUMY Press, Binghamton, New York.
Stonich, S.C.
1993 1 Am Destroying the Land: The Political Ecology of Poverty and Environmental
Destruction in Honduras. Westview Press, Boulder, Colorado.
Storper, M. and S. Christopherson
1987 Flexible specialization and regional industrial agglomeration: The case of
the U.S. motion picture industry. Annals of the Association of American Geographers
77:104-117.
Storper, M. and R. Walker
1992 The Capitalist Imperative: Territory, Technology, and Industrial Growth.
Blackwell, Cambridge.
Stryker, D.
1976 Population density, agricultural technique, and land utilization in a village
economy. The American Economic Review 66:347-358.
Symons, D.
1989 A critique of Darwinian anthropology. Ethology and Sociobiology 10:131-144.
Taffe, EJ.
1974 The spatial view in context. Annals of the Association ofAm.erican Geogra-
phers 64:1-16.
Taracena, B.
1941 Carta arqueologica de Espana. Soria. Madrid.
Taylor, P. J. and M. Overton
1991 Further thoughts on Geography and GIS. Environment and Planning A
23:1087-1094.
Tite, M.S. and C. Mullins
1970 Electromagnetic prospecting on archaeological sites using a soil conductiv-
ity meter. Archaeometry 12:97-104.
Theodossiou, E. and I. Dowman
1990 The heighting accuracy of SPOT. Photogrammetric Engineering and Remote
Sensing 56:1643-1649.
Tobler, W.R.
1970 A computer movie simulating urban growth in the Detroit region. Economic
Geography (supplement) 46:234-240.
Tomlin, C.D.
1990 Geographic Information Systems and Cartographic Modeling. Prentice Hall,
Englewood Cliffs, New Jersey.
Tooby, J. and L. Cosmidies
1989 Evolutionary psychology and the generation of culture, part I: Theoretical
considerations. Ethology and Sociohiology 10:29-49.
Tooby, J. and I. DeVore
1987 The reconstruction of hominid behavioral evolution through strategic mod-
eling. In The Evolution of Human Behavior: Primate Models, edited by W.G. Kinzey,
pp. 183-237. State University of New York Press, Albany.
Trewartha, G.T.
1954 An Introduction to Climate. McGraw-Hill, New York.
Tucker, C.
1979 Red and photographic infrared linear combinations for monitoring vegeta-
tion. Remote Sensing of Environment 8:127-150.
TukeyJ.W.
1977 Exploratory Data Analysis. Addison Wesley, Reading, Massachussetts.
Turner, M.G.
1989 Landscape ecology: The effect of pattern on process. Annual Review of Eco-
logical Systems 20:171-197.
Turner, M. and R. Gardner (editors)
1990 Quantitative Methods in Landscape Ecology: The Analysis and Interpretation of
Landscape Heterogeneity. Springer-Verlag, New York.
USAGE U. S. Army Corps of Engineers)
1991 GRASS 4.0. Construction Engineering Research Laboratory (CERL),
Champaign, Illinois.
Urban, D.L., R.V O'Neill, and H.H. Shugart, Jr.
1987 Landscape ecology. Biosdence 37:119-127.
Van West, C.R.
1994 Modeling Prehistoric Agricultural Productivity in Southwestern Colorado: A GIS
Approach. Reports of Investigations 67. Department of Anthropology, Washing-
ton State University, Pullman, and Crow Canyon Archaeological Center,
Cortez, Colorado.
Van West, C.R.
1991 Reconstructing prehistoric climatic variability and agricultural production
in southwestern Colorado, A.D. 901-1300: A GIS approach. Paper presented at
the Third Anasazi Symposiums, Mesa Verde National Park.
Van West, C.R. and W.D. Lipe
1992 Modeling prehistoric climate and agriculture in southwestern Colorado. In The
Sand Canyon Archaeological Project: A Progress Report, edited by W.D. Lipe, pp. 105—
119. Occasional Paper 2. Crow Canyon Archaeological Center, Cortez, Colorado.
Varien, M.D.
1990 1988 Small Site Testing: Preliminary Descriptive Report on the Excavations at
Lillian's Site (5MT3936), Roy's Ruin (5MT3930), Shortened Site (5MT3918), and
Troy's Tomer (5MT395). Crow Canyon Archaeological Center, Cortez, Colorado.
References 287
Varien, M.D.
1991 1989 Sand Canyon Project Testing Program.-. Preliminary Report of the Excava-
tions at Troy's Tower (5MT3951), Catharine's Site (5MT3961), and Stanton's Site
(5MT10508), Crow Canyon Archaeological Center, Cortez, Colorado.
Varien, M., W.D. Lipe, B. Bradley, M. Adler, and S. Thomson
1990 Mesa Verde region settlement, A.D. 1100 to 1300. Ms. on file, Crow Canyon
Archaeological Center, Cortez, Colorado. Paper presented at the Crow Canyon
Archaeological Center conference "Pueblos Cultures in Transition: A.D. 1150-
1350 in the American Southwest".
Vasey, D.E.
1979 Population and agricultural intensity in the humid tropics. Human Ecol-
ogy 7:269-283.
Vayda, A.P. and B.J. McCay
1974 New directions in ecology and ecological anthropology. Annual Review of
Anthropology 3:293-306.
Velasco, E, P. Mena, and A. Mendez
1987 Excavaciones de urgencia y carta arqueologica. In 130 Anos de Arqueologia
madrilena. pp. 189-195. Madrid.
Velleman, P.E and D.C. Hoaglin
1981 Applications, Basics, and Computing of Exploratory Data Analysis. Duxbury
Press, Belmont.
Vita-Finzi, C. and E.S. Higgs
1970 Prehistoric economy in the Mount Carmel area of Palestine: Site catchment
analysis. Proceedings of the Prehistoric Society 36:1-37.
von Braun, J. and R. Pandaya-Lorch
1991 Income Sources of Malnourished People in Rural Areas: Microlevel Information
and Policy Implications. Working Papers on Commercialization of Agriculture Num-
ber 5. International Food Policy Research Institute, Washington, D.C.
von Thiinen, J.H.
1826 Der Isolierte Staat in Beziehung aufLandtsirtschaft und Nationalokonom,ie. (Ham-
burg) 2nd. ed. reprinted 1966, edited by P. Hall and translated by C.M. Wartenbert.
Pergamon Press, Oxford. Originally published 1826, Gustav Fischer, Stuttgart.
Vosti, S.A. and W. Loker
1989 Some health and environmental aspects of agricultural settlement in the
Western Amazon. Paper presented at the International Food Policy Research In-
stitute-Committee on Agricultural Sustainability-World Bank Symposium on En-
vironmental Aspects of Agricultural Development, Washington, D.C.
WalshJ.
1969 Geology of the Eldama Ravine-KabametArea. Survey Report No. 83. Ministry
of Natural Resources, Geological Survey of Kenya, Nairobi.
Warren, R.E.
1990 Predictive modeling in archaeology: A primer. In Interpreting Space: GIS and
Archaeology, edited byK. Allen, S. Green, and E. Zubrow, pp. 90-111. Taylor and
Francis, London.
WCED (World Commission on Environment and Development)
1987 Our Common Future. Oxford University Press, Oxford.
Weaver, J.R. and R.L. Church
1987 The formal and computational relationship of the supporting median prob-
lem to the P-Median problem. Transportation Research 21B:323-329.
Werner, D., N. Flowers, M. Ritter, and D. Gross
1979 Subsistence productivity and hunting effort in native South America. Hu-
man Ecology 7-3Q3-115.
White, A.S. and D.A. Breternitz
1976 Stabilization ofLcnury Ruins. Cultural Resources Series No. 1. Colorado State
Office, Bureau of Land Management, Denver.
Wilkie, D. S. andJ.T. Finn
1988 A spatial model of land use and forest regeneration in the Ituri forest of
northeastern Zaire. Ecological Modeling 41:307-323.
WilkesJ.
1969 Dalmatia. Routledge & Kegan Paul, London.
Wilk, R.
1989 Decision making and resource flows within the household: Beyond the black
box. In The Household Economy: Reconsidering the Domestic Mode of Production, edited
by R. Wilk, pp. 23-52. Westview Press, Boulder, Colorado.
Willey, G.
1953 Prehistoric Settlement Patterns in the Viru Valley, Peru. Bureau of American
Ethnology, Bulletin 155. Washington, D.C.
Williams, I., WE Limp, and EL. Briuer
1990 Using geographic information systems and exploratory data analysis for site
classification and analysis. In Interpreting Space:GIS and Archaeology, edited by K.
Allen, S. Green, and E. Zubrow, pp. 239-273. Taylor and Francis, London.
Williams, L.A.J.
1965 Petrology of volcanic rocks associated with the rift system in Kenya. In
East African Rift System.: UMC-UNESCO Seminar, vol. II, pp. 33-39. Univer-
sity College, Nairobi.
Williams, L.A.J.
1969 Geochemistry and petrogenesis of the Kilimanjaro volcanic rocks of the
Amboseli area, Kenya. Bulletin ofVolcanology 33:862-888.
Williams, L.A.J.
1970 The volcanics of the Gregory Rift Valley, East Africa. Bulletin of Volcanology
34:439-465.
Williams, L.A.J.
1972 The Kenya rift volcanics: A note on volumes and chemical compositions.
Tectonophysics 15:83-96.
Wilshusen, R.H.
1991 Early villages in the American Southwest: Cross-cultural and archaeo-
logical perspectives. Ph.D. Dissertation, University of Colorado. University
Microfilms, Ann Arbor.
Winterhalder, B.
1986 Diet choice, risk, and food sharing in a stochastic environment. Journal of
Anthropological Archaeology 5:3 69-3 92.
Winterhalder, B. and E.A. Smith (editors)
1981 Hunter-Gatherer Foraging Strategies. University of Chicago Press, Chicago.
Winterhalder, B. and T. Evans
1991 Preliminary GIS analysis of the agricultural landscape of Cuyo Cuyo, De-
partment of Puno, Peru. In Applications ofSpace-Age Technology in Anthropology, C.
Behrens and T. Sever, (eds.), pp. 195-226. NASA Science and Technology Labo-
ratory, John C. Stennis Space Center, Mississippi.
Referencess 289
WRI/IIED (World Resources Institute/International Institute for Environment and
Development)
1988 World Resources, 1988-89. Basic Books, New York.
Wrigley, N.
1977 Probability Surface Mapping. Concepts and Techniques in Modern Geogra-
phy 16. Geo Abstracts, University of East Anglia, Norwich.
Yost, J A
1981 Twenty years of contact: The mechanisms of change in Wao ("Auca") cul-
ture. In Cultural Transformations and Ethnicity in Modem Ecuador, edited by N.
Whitten, pp. 677-704. University of Illinois Press, Urbana.
Zaninovic, M.
1977 The economy of Roman Dalmatia. In Aufstieg und neidergang der Romanischen
Welt, Geschite und Kultur Roma in Spiegel der neuren Forschung, vol. 2, edited by H.
Temporini and W. Haas, pp. 767-809. De Gruyter, New York.
Zaninovic, M.
1978 Novi prilozi arheoloskoj topografiji otoka Hvara. Novija i neobjavljena
istrazivanja u Dalmaciji, Vodice. Hrvatsko Arheolosko Drustvo 49-62.
This page intentionally left blank
Index
Aerial photography, 4 Central place model, 8, 63. See also Model

in archaeology, 8 Chacoan, 111
in ethnography, 7 Chain rule, 66, 67, 69
Address matching, 49 Choluteca, 89 (Figure 5.6)
Agricultural productivity, Modified: Best Cluster analysis, 71. See also Analysis
first rule (AMB), 165,167 (Figure 9.3), Comprehensive Resource Inventory and
171 (Figure 9.4) Evaluation System Project (CRIES),
Analysis 79, 84, 85 (Table 5.3)
cluster, 71 Conductivity, 211, 212 (Figure 12.5)
location-allocation (L-A),163 Cramer's rule, 66, 67
logistic regression, 182, 185 Crop distribution, 23,31,32 (Table 2.7),
log-linear, 184 33 (Table 2.8)
settlement, 8 Culture areas, 5. See also Diffusion
site catchment, 15, 228, 234, 235
spatial, 22 Danebury, 214, 217, 218 (Figure 13.1)
definition, 242, 243 Delaunay
viewsheds 16, 225 tesselation (DEL), 168. See also Net-
Anasazi, 109 works
Anthropology triangular link series (DTLS), 168,174.
development, 10 See also Networks
ecological, 7, 8 Dendritic marketing pattern, 162
Archaeological cultures Diffusion, 5, 6. See also Culture areas,
definition, 6 Kulturkreise
Anasazi, 108 Digitization, 4, 16
Aztec Empire, 161 Disaggregation, 229-231 (Figures 13.11,
Basketmakers, 110 13.12,13.13)
Chacoan, 111 Dot map, 24, 48. See also Crop distribu-
Bronze/Iron age, Britain, 136,225-232 tion
Neolithic age, Britain, 144, 223
Late Horizon, Mexico, 159
Northwest Coast, North America, 179 Earth Observing System (EOS), 105
Argyll, 143 Earth Resources Laboratory Application
AVHRR imagery, 104. See also Imagery Software (ELAS), 71
Aztec Empire, 161 Ecological degradation, 55, 79
definition, 19
measure of, 20
Basketmakers, 110 ERDAS image processing, 87. See also
Bronze/Iron age, Britain, 136, 225-232 Imagery
Facility location model, 163. See also Kilmartin, 143, 144 (Figure 8.6), 145
Model (Figure 8.7)
Fall-off curves, 5 sites of, 147 (Figure 8.8)
Foraging Kuiu Island, 176, 177 (Figure 10.1), 179
optimal, 175, 176 Kulturkreise, 5. See also Diffusion
cost-effective, 176
Lagrangean equation, 65, 77

Gabriel graph (GG), 168, See also Net- Late Horizon, Mexico, 159
work Least-cost module, 135
Geary ratio, 48, 50-52 Linear programming (LP) model, 156,
Geodetic Total Station, 208 157. See also Model
Geographic Information System (GIS) Locational modeling, 9, 14-17. See also
definition, 4, 43 Modeling
application in research, 9, 15, 96, 133, Location-Allocation (L-A)
173,214,215 analysis, 163. See also Analysis
-based modeling, 155. See also Model- model, 155, 157, 162, 164. See also
ing Model
data-gathering methods, 11,71,216 Logistic regression
functionality of, 54, 97, 98, 240 analysis, 182, 185. See also Analysis'
regional data management model, 182. See also Model
in archaeology, 190 Log-linear
in ethnography, 78, 97 analysis, 184. See also Analysis
Geographical Resource Analysis and Sup- modeling, 15, 184. See also Modeling
port System (GRASS), 72
Geometric rectification, 101. See also
Rubber sheeting MARNR-ORSTROM atlas, 100, 101,
Global positioning system (GPS), 99 103
accuracy of, 100, 102 Maxcover, 162, 168. See also Model
Gravity model, 5, 8. See also Model Minimal spanning tree (MST), 168,172.
Gregory Rift, 204 See also Network
Model
central place, 8, 63
Hvar, island, 136, 137 (Figure 8.1) facility location, 163
sites on, 138 (Figure 8.2) gravity, 5, 8
integer programming, 156
linear programming (LP), 156, 157
Identity overlay, 50 location-allocation (L-A), 155, 157,
Imagery 162, 164
AVHRR, 104 logistic regression, 182
SLAR, 99, 100 maxcover, 162, 168
ERDAS, 87 optimization, 156, 157
Integer programming model, 156. See also P-median, 163, 164, 168, 172
Model spatial, 176
Intensification, land use, 55, 62, 72 time allocation, 63
International Sorghum Millet Project Van Thunen, 8, 63
(INTSORMIL), 79, 83, 85 (Table 5.3) Modeling
Intel-visibility, 148, 150 (Figure 8.11, CIS-based, 155
Table 8.1), 225 (Figure 13.6) locational, 9, 14-17
log-linear, 15, 184
Index 293
Modifiable areal unit problem (MAUP), RISC station, 193
48, 49, 247, 249 Rubber sheeting, 101. See also Geomet-
Moran coefficient, 48, 49 ric rectification
Multispectral scanner (MSS), 103
Multivariate methods, 15
Satellite
LANDSAT 1, 9, 70, 86 (Figure 5.4)
Negative areas, 219, 220 (Figure 13.3) NOAATiros-N, 104
Neolithic age, Britain, 144, 223 SPOT, 70, 105
Network, 167, 168 Settlement pattern
Delaunay tesselation (DEL), 168 definition, 8
Delaunay triangular link series (DTLS) analysis, 8. See also Analysis
168, 174 Side-looking airborne radar (SLAR) im
Gabriel graph (GG), 168 agery, 99, 100. See also Imagery
minimal spanning tree (MST), 168, Simulation, 13, 14
172 Single basic value (SBV), 166, 171 (Fig-
relative neighborhood graph (RNG), ure 9.4)
168 Spatial
triangular irregular network (TIN), aggregation, 48
168 analysis, 22. See also Analysis
Northwest Coast, North America, 179 definition, 242, 243
anomalies, 242
autocorrelation, 48
Office of Territorial Planning )OPT), 193
coincidence, 243
Olorgesailie, 202, 203 (Figure 12.1), 205
dependence, 243, 244, 248
Operations Research (OR), 159
heterogeneity, 244, 247, 248
Optimization model, 156, 157. See also
model, 176. See also Model
Model
proximity, 243
Optimizing, 159
Statistic
goodness-of-fit, 182, 184
Palmer Drought Severity Indices Statistical analysis system (SAS), 44, 50
(PDSIs), 112, 113 (Table 7.2) SUN Unix workstation, 181
P-medianmodel, 163, 164, 168, 172. See
also Model
Tebenkof Bay, 179
Pespire, 89, (Figure 5.6)
sites of, 180 (Figure 10.2)
Polygon generation, 49, 50
Temascalpa-Teotihuacan, 159, 160 (Fig-
Pueblo I, 110
ure 9.2)
Pueblo II, 110
Texcoco, 164
Pueblo III, 111
Thematic mapper (TM) 71, 72 (Figure
4.5), 73 (Table 4.1), 77, 103, 104
Rasters Thiessen polygons, 166,167 Figure 9.3),
definition, 4 168, 223 (Figure 13.4), 225, 232 (Fig-
system, 113 ure 13.14), 235, 236 (Figure 13.21)
Regional Time allocation model, 63. See also Model
data management, 9, 10 Topographically integrated geographic
environmental analysis, 12 encoding and referencing system (TI-
Relative neighborhood graph (RNG), GER) files, 49
168. See also Network Triangular irregular network (TIN), 168.
Resource pooling, 115, 118, 120, 121 See also Network
(Figure 7.7), 125 (Table 7.4), 131 T-test, 187 (Table 10.3), 188 (Table 10.4)
UNIX system, 192, 193
Van Thunen model, 8,63. See also Model

Vectors
definition, 4
use of, 113, 182
Viewshed, 16, 146, 148 (Figure 8.9), 149
(Figure 8.10), 152 (Figure 8.13), 153,
224 (Figure 13.5), 233 (Figure 13.16),
234 (Figure 13.17)
Yanomamo, 11, 97

Aldenderfer Maschner 1996 Anthropology, Space, and GIS

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Aldenderfer Maschner 1996 Anthropology, Space, and GIS

Încărcat de

Drepturi de autor:

Formate disponibile

Anthropology, Space, and

Geographic Information Systems

New York Oxford

Copyright © 1996 by Oxford University Press, Inc.

Library of Congress Cataloging-in-Publication Data

Printed in the United States of America

Santa Barbara, Calif. M. A.

Geographic Information Systems

A GIS is a sophisticated database management system designed for the acqui-

In a recent overview of applications of GIS in archaeological research, Kvamme

Regional Data Management

Management of Remotely Sensed Data

Regional Environmental Analysis

phenomena, such as ethnic neighborhoods. Arnold and Appelbaum hypoth-

real problem with line-of-site analysis is one that commonly plagues

1. In the United States, most archaeologists are trained in departments of anthropol-

Setting the Context: the Peruvian Amazon

Table 2.1 Population growth in selected departments of the Peruvian Amazon.

PERU 6,207,967 9,906,746 13,538,208 17,005,210 2.2 2.9 2.5

Source: INEI 1990

Figure 2.1 Map of Peru, indicating departments of the Peruvian Amazon.

In spite of the growing importance of the selva in the national economy,

Table 2.2 Recional production of selected crops.

Crop Peru Coast Sierra Selva

To understand the environmental consequences of agriculture at the regional

Crop Distribution Data

The outcomes of land-use decisions are reflected in agricultural statistics on

Area Cultivated (ha)

Source: Author's data derived from Ministry of Agriculture statistics

Figure 2.2 Cultivated area of crops studied in the Peruvian Amazon.

Natural Resource Distribution

Table 2.4 Estimates of the extent of coca cultivation, Peruvian Amazon.

Sources: PEAH-OSE, 1988

Table 2.5 Pastures, livestock information, Peruvian Amazon.

Figure 2.3 Land systems of the Peruvian Amazon.

Effects of Crops on Resources: Analytical Procedures

220 9,229,437 Putumayo Plain

Figure 2.4 Erodibility of land systems, Peruvian Amazon.

Figure 2.5 Fertility of land systems, Peruvian Amazon.

Crop Distribution by Land System

Fgure 2.6 Orthographic projection of total cultivated area of study crops.

contiguous geographic areas of moderately heavy cultivation (838 and 836 on

Figure 2.7 Location of fifteen most heavily cultivated land systems.

areas. We can assume that agriculturally induced land degradation is either

Impacts of Agriculture by Land System

The image overlays the information on credibility on the "peaks" of culti-

Figure 2.8 Overlay of credibility of land systems, cultivated area.

Low Fertility fand Systems, Peruvian Amazon

Figure 2.9 Overlay of fertility of land systems, cultivated area.

Testing the Results: A Brief Review of Existing Research

Area 3: Agroecologically, Yurimaguas is similar to the Pucallpa region,

Area 4: Agroecologically, the Puerto Maldonado region (12 to 14° south

The GIS Difference

Industrial Districts and Ethnic Economies

Research Site and Data Sources

The Los Angeles Garment Industry

The data set combines four sources to provide a comprehensive coverage of

COMBINED CONTRACTORS AND MANUFACTURERS

sample of identifiable firms in Los Angeles County, yet it is subject to error

Our previous research, using dot maps, demonstrated heuristic differences in

Korean Garment Firms Latino Garment Firms

N-738 Max Z=592 GR=1.04 p- 19 H-934 Max Z-367 GR-0.72 p-.OO

Figure 3.1 "County" density maps.