Monitoring regional market systems in prehistory:
Models, methods, and metrics
Leah D. Minc *
Radiation Center, Oregon State University, Corvallis, OR 97331, USA
Received 23 August 2004; revision received 20 September 2005
Available online 2 December 2005
Abstract
While market systems have long been recognized as key factors in the development and integration of many complex
societies, market exchange has been largely ignored in recent studies of archaic states. The underlying causes appear threefold:
(1) we overlook the rich conceptual framework developed within economic geography that identifies key differences
among regional market systems; (2) we underestimate the degree to which these differences constrain or create distinct
opportunities for other productive activities; and/or (3) we lack appropriate measures for monitoring organizational differences
in the archaeological record. To understand the role of markets within the ancient economy, measures are needed
to model and monitor its regional organization. The method proposed here focuses on the distribution of goods which
circulated through the market system. By examining how different types of market systems move goods through the regional
system in different ways, distinct patterns of commodity distribution are predicted. Metrics for monitoring exchange
based on artifact assemblage similarity are then proposed. As a case study, the analysis applies these insights to characterizing
the regional market system in the Aztec heartland, an area where the organization of exchange and the degree
of political interference remain topics of active debate.
Ó 2005 Elsevier Inc. All rights reserved.
Keywords: Market systems; Regional exchange; Aztec empire
Introduction
Although market systems have long been recognized
as a key factor in the development and integration
of complex societies (e.g., Berdan, 1989;
Brumfiel and Earle, 1987; Flannery, 1972; Morrison
and Sinopoli, 1992; Sanders and Price, 1968; Sinopoli,
1994), market exchange has received relatively
little attention in recent studies of archaic state
economies (Smith, 2004, pp. 83–84; Stein, 1998).
The slight is especially noticeable when contrasted
with the degree of attention focused on agrarian
and craft production (Costin, 1991, 2001; Denevan,
2001; Sinopoli, 2003; Trigger, 2003, pp. 358–373;
Whitmore and Turner, 2001), households and
domestic consumption (DÕAltroy and Hastorf,
2001; Feinman and Nicholas, 2000; Santley and
Hirth, 1993), and mechanisms of elite finance
(Earle, 2001, 2002; Feinman and Nicholas, 2004;
Yoffee, 1995).
The shift away from market studies is both
puzzling and troubling. In many complex societies,
Journal of Anthropological Archaeology 25 (2006) 82–116
www.elsevier.com/locate/jaa
0278-4165/$ - see front matter Ó 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jaa.2005.09.003
*
Fax: +1 541 737 0480.
E-mail address: mincleah@engr.orst.edu.
market exchange is inextricably linked to
production and consumption, and is critical for
the effective articulation and coordination of these
economic processes. Typically, market systems are
seen as supporting increasingly specialized production
in both agrarian and craft sectors by
mobilizing resources directly from producers,
and enabling households to provision themselves
with needed items. Further, market systems coordinate
these activities both spatially and temporally,
by concentrating exchange in a predictable
time and location that allows participants to
schedule.
Yet the relationship between market exchange
and production is not simply a supporting role. It
is widely recognized by economic anthropologists
and cultural geographers that different market systems
create very different contexts for production,
exchange, and consumption according to the structure
and scale of their regional organization (Hodges,
1988; Larson, 1985; Plattner, 1985, 1989a,b;
Smith, 1974, 1976a,b,c,d, 1977). Solar, dendritic,
and interlocking network systems, for example,
generate distinctive conditions for market participation
that affect the degree of reliance on (and
investment in) market exchange as a viable economic
strategy. In addition, these systems represent
different levels of political involvement or interference
in the arena of exchange activities on the part
of ruling elites.
To understand the complex role of markets
within the ancient economy, measures are needed
not only to detect the existence of market
exchange, but to monitor key dimensions of its
regional organization. Prior approaches to modeling
regional market system organization have
focused heavily on the infrastructure of market
systems, including the spatial distribution of economic
central places or market centers, and networks
of roads and transportation routes that
developed in response to market forces (Santley,
1986, 1991; Skinner, 1964, 1977; Smith, 1979;
Trombold, 1991). Aside from concerns over the
applicability of commercial models such as Central
Place Theory (CPT) to pre-industrial societies (e.g.,
Evans, 1980), archaeologists face the additional
problem of identifying economic nodes and networks
from settlement pattern data. Settlement
systems are determined by multiple factors in addition
to economic concerns, including ecological,
geographic, political, religious, and historical circumstances,
such that settlement size and location
are not responsive indicators of market forces
alone.1
The approach taken here, in contrast, focuses on
the distribution of goods which circulated through
the market system, and is grounded on observations
from economic geographers that different market
systems move goods through the regional system
in different ways. Accordingly, this study begins
by examining the major types of regional exchange
systems and their organizational features, drawing
from the insights of previous studies in economic
anthropology (Feinman, 1980; Fry, 1979, 1980;
Fry and Cox, 1974; Hodges, 1988; Plattner, 1985;
Renfrew, 1975, 1977; Santley, 1986, 1991; Smith,
1974, 1976d). Second, by evaluating how these organizational
features affect commodity flows, distinct
patterns of commodity distribution associated with
each market system type are identified.
The resulting models of regional exchange are
potentially amenable to testing using a range of data
types, including chemical, typological, and stylistic
analyses of goods that circulated through the market
system. Metrics for monitoring exchange based
on artifact assemblage similarity are proposed here,
in that these data are a common product of archaeological
surveys and provide the necessary regional
perspective. Finally, as a case study, the analysis
applies these insights to the task of characterizing
the regional market system in the Aztec heartland,
an area where the organization of exchange and
the degree of political interference remain topics of
active debate.
Models of market system organization
In a regional perspective, market systems consist
of a series of (1) market centers that function as foci
of economic exchange, and (2) the market zone or
service area provisioned by a given market center.
As economic central places, market centers bring
together producers and consumers on a regular
1
Central Place Theory and related models derived from
network analysis (Berry, 1967; Christaller, 1966; Skinner, 1964)
are based on assumptions that may not hold in pre-industrial,
non-commercial societies (Smith 1976a 1:8,12; Hassig, 1991, p.
20). For example, CPT holds that if the market principle is
dominant then microeconomic forces will generate a predictable
spatial patterning in the distribution and hierarchical arrangement
of economic central places (King, 1984); conversely, if this
predictable pattern is observed to be present, then the underlying
market principle is assumed to have been operative. For a more
detailed critique, see reviews by Evans (1980) and Hirth (1998).
L.D. Minc / Journal of Anthropological Archaeology 25 (2006) 82–116 83
basis (either daily or periodic) for exchanges involving
a broad range of goods and services. The area
provisioned by a market center depends on the
range of goods available at the market center and
their demand function (Plattner, 1989b). Lower-order
centers provide basic subsistence and household
items of relatively low unit cost; since such goods
are widely available, they are not typically transported
great distances to market nor distributed
far from the market centers, resulting in a relatively
small service zone. In contrast, higher-order centers
additionally offer commodities requiring greater
labor input or energy investment, with correspondingly
higher unit cost and lower demand. These
goods travel greater distances from producers to
point of redistribution in the market center and
attract consumers from a broader area.
The organization and integration of multiple
market centers within a regional market system
can be assessed along three primary dimensions:
(1) scale of inclusiveness (areal extent of exchange
interactions or system integration), (2) network
(the amount of horizontal commodity flows
between market centers of the same size), and (3)
hierarchy (the amount of vertical commodity flows
between market centers at different levels of the settlement
hierarchy) (Smith 1976d, pp. 314–315). A
fourth dimension, political congruence, assesses the
spatial organization of market systems relative to
features of political geography, including administrative
centers and territorial boundaries.
Economic geographers suggest that the dimensions
of scale, network, and hierarchy can be integrated
to characterize the distinct organizational
features of four different market system models, as
described below (Fig. 1; Table 1). While these ideal
types do not claim to cover all possible cases, they
do provide a well-tested starting point for evaluating
ancient, historic, and modern market systems
(Hodges, 1988, p. 25).
Solar market systems, also known as simple centralized
market exchange systems, are characterized
by both poorly developed hierarchy and network,
and relatively small scale. As a system of central
places, a solar market system consists of a localized,
low-level market hierarchy (usually involving an
administrative center serviced by several small rural
market places) that is poorly articulated with other
low-level hierarchies in the same region (Hodges,
1988, p. 73; Smith, 1974, p. 176). A key feature of
solar systems is that the center provides both political
and economic functions for dependent communities
within its hinterland. As a result, the extent of
political control is spatially congruent with the
sphere of economic influence.
Under a system of enforced dependence on the
primary center for both economic goods and administrative
services, rural market participants (as
either producers or consumers) are unable to exercise
a choice in market destinations—they must
patronize the center to which they are politically
dependent (Smith, 1974, pp. 176–177). Goods move
from production sources into the central marketplace
where they are redistributed throughout the
area served by the centralized market. If a number
of producers contribute to this system, their products
will have nearly identical, overlapping patterns
of spatial distribution, resulting in a high degree of
internal homogeneity within the area served by the
central market.
At the regional level, restricted consumer and
retailer movements result in poor articulation
between solar market systems, marked by sharp discontinuities
in commodity flows across system
boundaries. The resulting pattern is one of discontinuous,
bounded market territories. Linkages
between these territories occur only as elite-controlled
trade between administrative centers, indicated
by the movement of foreign commodities or
prestige goods between primary central places.
Under a solar market system, then, the regional
spatial patterns of commodity flow should be characterized
by (1) bounded, discontinuous market territories
that (2) are coterminous with political
administrative units. If commodity exchange was
centralized at the level of the local polity, then the
boundaries of market systems would be apparent
as sharp discontinuities in commodity distribution,
corresponding to polity borders. Household assemblages
within communities participating in the same
market system will appear relatively homogeneous
in composition but distinct from those served by
neighboring market systems.
Non-centralized or overlapping market exchange
systems (also called network exchange systems) are
also characterized by poorly developed hierarchy,
but the higher degree of horizontal exchanges
results in a higher degree of market network at the
regional level than found under solar market systems
(Bohannan and Bohannan, 1968; Fry, 1979;
Smith, 1974, pp. 179–180). In non-centralized market
exchange systems, political authority is decentralized
and relatively weak, and administrative
boundaries do not constrain the movements of
84 L.D. Minc / Journal of Anthropological Archaeology 25 (2006) 82–116
goods or people. Producers and consumers have
access to several independent market centers, and
goods move readily between adjacent market zones.
Based on the distribution of goods, no sharp boundaries
are visible between market zones; rather, a gradient
in similarity of assemblages extends along a
line between exchange centers, as the frequency of
each product declines with distance from its original
source. However, political instability precludes the
development of a market hierarchy and the system
Fig. 1. The primary dimensions of scale, network, and hierarchy generate distinct organizational features for four different regional
market systems.
Table 1
Dimensions of market system variation
Dimension of variation Regional market systems
Solar Non-centralized Overlapping Dendritic Complex Interlocking
Scale Small, local Small, relatively local Large, regional Large, Regional
Network Poorly developed Well developed Poorly developed Well developed
Hierarchy Poorly developed Poorly developed Well developed Well developed
Political Congruence Coterminous with
local polity
Not constrained by
political boundaries
Coterminous with control
of primate center
Coterminous with
regional polity
L.D. Minc / Journal of Anthropological Archaeology 25 (2006) 82–116 85
is characterized by poor regional articulation; as a
result, economic interaction between communities
declines rapidly with distance. Neighboring communities
will thus share a large number of commodities,
while the assemblages of more distant
communities will be more dissimilar.
These characteristics suggest several key aspects
of regional spatial patterning generated by non-centralized
or network market systems: (1) the structure
is characterized by a series of fluid, overlapping
market territories; and (2) the network is spatially
unbounded by local political or administrative
boundaries. If commodities were distributed
through a non-centralized market system in which
several independent exchange systems interacted,
then individual goods or artifact types would likely
form overlapping distributions and they would be
represented in increasingly smaller proportions with
distance from their source. Individual market territories
would emerge as areas with similar assemblages,
but the degree of similarity between
adjacent market territories would be high and differences
across these boundaries relatively weak.
Dendritic market systems, in contrast, are characterized
by a well-developed market hierarchy, in
conjunction with a poorly developed market network.
Within the regional system, political and economic
control emanates from a single primate
center, which connects to dependent communities
through a linear chain of exchange interactions
(Appleby, 1976; Johnson, 1970; Kelley, 1976; Smith,
1974, pp. 177–179). In the resulting branch-like
structure, goods flow vertically up and down the
market hierarchy, while horizontal connections
among communities at the same level of the hierarchy
are minimal.
Dendritic market structures depend on a strong
centralized authority in which ‘‘political and economic
elites are one and the same, and political
decisions are made to safeguard elite economic
interests’’ (Santley, 1991, p. 199). This imbalance
of power precludes the development of a competitive
market system, as elites are able to manipulate
market commerce to their own ends. Under a strategy
of commodity oligopoly, for example, administrative
elites control the availability of goods and
services needed by the rural populace to force them
to produce for the urban market; monopolistic controls
over access to trading partners, transportation,
and currency can be equally effective in directing
and controlling the flow of goods (Hodges, 1988,
pp. 42–52; Smith, 1974, p. 179).
Although primarily associated with mercantile
or colonial powers (e.g., Smith, 1976d; Larson,
1985; Vance, 1970), several authors note that such
market imperfections were probably widespread,
and the model has considerable utility in highlighting
the effects of strong vertical market integration
in prehistoric and early historic economies
as well (e.g., Allen, 1992; French, 1964, pp. 120–
122; Hodges, 1988, pp. 42–52; Santley, 1991;
Smith 1974, p. 179; 1976a, p. 36). In particular,
while the dendritic system supports the primate
center, this marketing structure suppresses economic
development within the hinterland (Plattner,
1989b, p. 203). Rural goods do not flow
with regularity across rural markets, and rural
producers cannot depend on those markets for
subsistence goods (Smith 1976a, pp. 34–35). The
incomplete development of market network reduces
hinterland dependence on the market system
either as an outlet for products or as a source
of desired goods (Smith 1974, pp. 177–178;
1976d, pp. 319–320).
A key characteristic of dendritic market systems,
then, is the differential integration of communities
into the regional system based on their
distance from the primate center (Smith, 1974, p.
177). If the regional economy is dendritic in structure,
the geographic pattern would be that of distinct
zones characterized by different levels of
central market participation, including (1) a wellintegrated
core zone around the primate center
displaying high market participation, and (2) a
poorly integrated rural periphery with low market
participation. The degree of central market system
participation will be reflected in the degree of
access to higher-order goods produced in (or
imported into) the primate center. As market participation
declines with distance from the primate
center, we can expect to see a decline in the abundance
of centrally produced goods. Within the
periphery, locally produced subsistence and utilitarian
goods will predominate. Reduced lateral
commodity flows within the periphery will result
in localized differences in style and assemblage
composition.
Complex, interlocking market systems or hierarchically
integrated market networks are characterized
by both well-developed market hierarchy and
network. In such a market system, goods are channeled
through a series of local and regional centers,
serving overlapping market regions of various spatial
scales:
86 L.D. Minc / Journal of Anthropological Archaeology 25 (2006) 82–116
‘‘In interlocking systems each market center is
linked to several higher-level centers as well as
to several lower-level centers. This creates a network
with several levels, several links between
levels, and hierarchically organized service to all
places in the system. Goods flow to and from
other systems and regions but are also exchanged
within the local system at each level. Trade areas
are overlapping and economic regions unbounded.
Hence, supply and demand or price information
is communicated across broad areas to
ensure coordination of specialization. Under
these circumstances the rural consumer can stay
put and still enjoy product diversity in his marketplace;
he can also depend on a broad market
for the goods he produces. This allows specialization
within the realm of food production, so that
rural areas become as market dependent if not as
diversified as urban centers’’ (Smith 1976d, p.
320).
A primary feature of interlocking market systems
is the high degree of regional economic integration
(Plattner, 1989b, p. 203). Because goods move both
vertically and horizontally between centers, distribution
patterns are geographically widespread and
create a higher degree of similarity in consumer
goods throughout the regional system. Local variations
in the availability of goods from a given source
may persist, however, generating minor differences
between local market systems.
If the regional economy was organized as a complex
interlocking market system, the distribution of
artifact types should reflect a pattern of widespread,
relatively uniform distribution throughout the
region served by the market system. Minimal
intra-regional variation in type frequencies would
occur, although pockets of locally produced types
might appear. However, because trade areas are
overlapping and unbounded, no sharp internal
boundaries in distribution would be apparent. Similarly,
because rural areas are well integrated into
the market system, no sharp core-periphery pattern
in market participation is expected.
Archaeological correlates of regional market systems
The preceding models of regional market system
organization present distinct expectations for commodity
distribution patterns against which the
archaeological case can be compared. Fundamental
to such an analysis, however, is the empirical identification
of areas participating in the same market
system, and the delineation of market zones—the
archaeological manifestation of market territories—based
on patterns of consumption. Once market
zones have been delineated, the organization of
the regional market system can be evaluated in
terms of the size and number of distinct market
zones, their spatial organization, the horizontal
and vertical relationships among market zones,
and their distribution relative to features of political
geography.
Identifying market zones and boundaries
In archaeological terms, market zones are defined
here as areas serviced by the same market center(s)
within which communities have access to the same
array of goods. One direct measure of whether communities
visit the same market center is the degree
to which they obtain goods from the same producers,
as determined through compositional analyses
of artifacts and raw materials (Arnold et al., 1991;
Bishop and Blackman, 2002; Bishop and Neff,
1989; Glascock and Neff, 2003; Neff, 1992; Neff
and Glascock, 1995). Trace-element studies have
been particularly fruitful in modeling the extent
and mechanisms of long-distance exchange in
high-status elite goods (e.g., Glascock, 2002; Renfrew,
1975, 1977). At the intra-regional level, however,
the success of trace-element studies depends
on the spatial scale of geochemical variability in
raw (source) materials and our ability to differentiate
the products of local producers (Bishop and
Blackman, 2002). Even when variability in source
materials supports a fine-grained spatial analysis,
the cost of such analyses has generally limited their
application in the assessment of regional market
systems, where a large number of samples and a
regional sampling strategy are required to map
quantitative differences in the volume of exchange
and identify market boundaries. As demonstrated
below, however, compositional analyses play a critical
role in verifying key aspects of market system
organization as delineated from less direct measures
of exchange.
An alternative means for defining market zones
focuses on the degree of assemblage similarity
among sites within the study area based on typological
and stylistic grounds. This tactic follows up on
the ‘‘distributional approach’’ formalized by Hirth
(1998), which demonstrated that market participation
to provision households results in a high level
L.D. Minc / Journal of Anthropological Archaeology 25 (2006) 82–116 87
of homogeneity in the distribution of circulating
commodities. Briefly, Hirth argues that market
places provide equal access to commodities for all
consumers, independent of social status (1998, p.
458). Since consumers participate directly in market
exchange, and all households (whether low-status or
high-status) have access to the same types of
resources in the marketplace, the result is an
increase in the homogeneity of material culture
assemblages at the community level. In the archaeological
record, ‘‘the homogenizing effects of trade’’
(Creamer, 1998) among households offers a clear
indicator of marketplace exchange.
Beyond identifying the existence of market
exchange, however, the distributional approach
offers a powerful tool for examining the spatial
organization of exchange at a regional level.2
Archaeologists have long utilized stylistic similarity
(particularly that of ceramics) to assess the relative
degree of economic interaction among different
communities, arguing that greater exchange of
goods, technology, and artistic norms results in
greater uniformity (Fry, 1979, 1980; Fry and Cox,
1974; Hodge and Minc, 1990; Pires-Ferreira, 1976;
Plog, 1976, 1978). Where markets provide the primary
mechanism for exchange and commodity distribution,
HirthÕs analyses suggest that the degree to
which communities share similar artifact assemblages
can be used to assess the degree to which they
attend the same market centers. Because they have
access to the same array of goods, communities
attending the same market center(s) will have highly
similar assemblages of basic craft goods; in contrast,
those participating in completely different exchange
networks will have largely dissimilar assemblages. It
is therefore possible to identify and map the spatial
distribution of communities participating in the
same market zone from areas which display compositionally
similar material culture assemblages. Conversely,
boundaries between market zones will be
apparent as discontinuities in assemblage similarity
reflecting a decline in exchange activities.
Several factors may distort the presumed relationship
between assemblage similarity and economic
interaction, however. Primary among these
are differences in site function and status that can
alter the mix of different functional and prestige
artifact classes found at a site. In order to minimize
variability in assemblage composition owing to such
functional or wealth differences, the delineation of
market zones can focus on stylistic variability within
a single functional artifact type or class. The implementation
of the distributional method is illustrated
below based on stylistic variability within the class
of decorated ceramic serving vessels.
The identification of market zones utilizes the
methodology entitled ‘‘unconstrained clustering for
the analysis of spatial data’’ developed by Whallon
(1984). The goal of unconstrained cluster analysis is
the identification of spatial clusters or areas with
similar artifact assemblages that are not constrained
as to their size, shape, density, composition, or patterns
of artifact covariation by the very quantitative
methods employed to identify them. This methodology
was originally designed to assist intra-site analyses
in the identification of activity areas within
occupation floors. It is, however, a general
approach rather than a specific technique (Whallon,
1984, p. 244), and as such is appropriate for the
analysis of regional spatial data as well.
The approach requires, first, that we have data
on the distributions of a number of different artifact
types over a contiguous spatial area. Second, it
requires that information on artifact type distributions
within this area be sufficiently detailed to characterize
the underlying distributional patterns.
Ceramic collections generated by regional site surveys
potentially satisfy both these requirements.
Surveys attempt to provide relatively complete
information on settlement distributions for a large
contiguous area. The surface collections generated
through these surveys theoretically represent samples
drawn from the underlying ‘‘true’’ ceramic distribution
patterns. Individual samples or
collections, however, are subject to local noise or
distortion of the regional pattern, due to differences
in surface collecting conditions and factors affecting
artifact preservation and visibility. One major goal
of unconstrained cluster analysis, therefore, is to
even out the local perturbations in artifact recovery
to expose the larger underlying patterns of
distribution.
Input data consist of information on both the
absolute and relative densities of artifact types, as
these potentially provide complementary types of
information. On one hand, relative measures (e.g.,
percent of ceramic assemblage) reflect potential similarities
and differences in consumer behavior that
2
Several reviewers (Hassig, 1998; Hicks, 1998) of HirthÕs
approach suggested that it had wider utility, e.g., ‘‘The judicious
elaboration of his approach holds promise for going beyond the
detection of markets to the tentative identification of market
types and systems’’ (Hassig, 1998, p. 467).
88 L.D. Minc / Journal of Anthropological Archaeology 25 (2006) 82–116
are independent of differences in assemblage size.
This is an advantage where assemblage size may
well vary according to differences in settlement
density or surface survey conditions. Assuming, as
Hirth (1998) suggests, that market participation
promotes equal access and that consumers (on average)
obtained goods proportionate to their availability
in the market place, then communities
participating in same market zone will be marked
by assemblages with highly similar proportions of
ceramic types.
On the other hand, absolute densities (e.g., differences
in the total number of sherds per unit area)
potentially reflect differences in access to a particular
source, as a function of distance. Typically, we
can associate source areas for a given ceramic type
with high availability and high consumption (high
absolute densities) and assume that accessibility
declines with distance from a source, such that marginal
areas will be characterized by low densities of
that type. Absolute densities are also critical in identifying
areas with low overall artifact densities within
which percentage data may be unreliable or
misleading. Thus, through combining both perspectives,
we gain insights into both the relative degree
of interaction as well as the volume of interaction
between different areas. To the extent that the
results of the two analyses agree, we can be assured
that the patterning within the spatial data is relatively
robust.
Unconstrained cluster analysis proceeds through
a number of steps; at each step specific decisions
must be made to tailor the approach to the problem
at hand. Briefly, the basic sequence involves:
(1) Creation of a regular data grid from irregularly
spaced data points for each artifact type included
in the analysis. A regular grid of points is
established over the study area, and the value
of each grid point (e.g., ceramic type density)
is estimated from neighboring data points
(e.g., sites or collection areas), usually as a
weighted average of the values of points within
a certain distance of the grid point. Generally,
the weights applied are the inverse square
of the distances from the central point.
(2) Smoothing the data within each grid. In a second
optional step, the density grid is smoothed
using a moving template of surrounding grid
points, and each grid point is replaced with
the average of itself plus its closest neighbors.
Grid smoothing does involve some degree of
generalizing from the data, but Whallon
(1984, p. 245) argues that this is necessary:
‘‘We are interested in distributional pattern,
and pattern is a characteristic of the data as
a whole rather than of the array of individual
item locations. That is, pattern is a generalization
from the data.’’ In this context, grid
smoothing is recommended in that it reduces
noise and enhances the underlying pattern by
averaging out random fluctuations.
(3) Clustering of grid points based on density values
for multiple artifact types. Cluster analysis is
utilized to summarize the information contained
in the multiple density contour maps
and to reveal the joint patterning of the different
artifact types or stylistic variants within
the study area (Whallon, 1984, p. 245). Input
data for the cluster analysis consists of type
densities as calculated for each grid point. It
should be emphasized here that what is being
clustered are grid points, with clusters of grid
points presumably representing areas with
similar assemblage composition. As a result,
the focus is on identifying areas with similar
ceramics, rather than on identifying groups
of sites with similar ceramics. An explicitly
hierarchical algorithm, such as the minimum
variance or WardÕs method that joins clusters
by minimizing intra-cluster variance (Aldenderfer
and Blashfield, 1984), is useful in revealing
the degree of relatedness among areas
within the study area. The clustering is based
on Euclidean distances between grid points
as calculated from the densities of the different
artifact types or stylistic variants at those
points. The appropriate cluster solution is
identified using the scree method, based on
major changes in error sum of squares (SSE).
According to this criterion, ‘‘clustering proceeds
. . . until a series of marked jumps in
the error sum of squares are produced by the
fusion of relatively dissimilar groups. These
sudden jumps in the clustering criterion indicate
significant increases in the heterogeneity
of the groups being defined’’ (Whallon, 1984,
p. 253). The strategy here is to examine cluster
solutions just prior to major increases in the
SSE, as these divisions represent relatively distinct
groups.
(4) Cluster mapping. Once the appropriate cluster
solution has been identified, grid points
belonging to each cluster are plotted on a
L.D. Minc / Journal of Anthropological Archaeology 25 (2006) 82–116 89
map of the study area, and these groups are
examined for spatial integrity and geographic
interpretability. It is expected that grid points
belonging to the same cluster will also cluster
spatially, and that as a group they represent
an area with a distinctive ceramic assemblage.
(5) Cluster interpretation. If clusters pass these
preliminary tests, the analysis proceeds to
cluster interpretation using descriptive statistics
to quantify differences in assemblage composition
between clusters or areas. Cluster
interpretation must also consider the spatial
configuration of clusters, their size, and their
location relative to features of political geography,
including political centers and polity
boundaries.
Assessing relationships among market zones
Once market zones have been delimited, the relationships
within and among the zones are examined
to characterize the regional market system structure.
Metrics for the key organizational attributes
of scale, network, hierarchy, and political congruence
are suggested below.
Scale
The scale of market zones can be evaluated
directly from the number of market zones and their
relative spatial extent. Of primary interest here is the
characterization of the regional system as comprising
one or a few large market zones as opposed to
many, smaller market zones. The size of market
zones relative to travel time and distance is also of
interest, as indicating whether the service zone
corresponds to a local, regional, or supra-regional
market center.
Network
The degree of horizontal integration among market
zones reflects the extent of trade and exchange
between adjacent market zones in goods of the same
order, and can be assessed from the overall similarity
among their ceramic assemblages. For example,
the Brainerd-Robinson agreement coefficient
(Cowgill, 1990) measures similarity between pairs
of market zones, when assemblage composition
data are expressed as percentages of various ceramic
types and variants. The coefficient can range from 0
(when pairs of market zones share no types in common)
to 200 (when pairs of market zones share all
types in common and in the same proportions).
As a measure of network, a high coefficient indicates
a high degree of interaction among market zones.
Conversely, a low coefficient suggests a lower degree
of interaction; a sharp decrease in the agreement
coefficient represents a boundary in economic interaction.
In addition, the degree of horizontal integration
can be assessed qualitatively from the spatial
congruence between market zones and individual
type distributions, that is, whether type distributions
appear to conform to market zone boundaries
(low network) or cross over market zone boundaries
(higher network).
Hierarchy
The degree of vertical integration within the
regional market system can be evaluated from the
dual perspective of differential access to a greater
diversity of goods and to better quality goods. In a
hierarchically organized system, higher-order markets
offer higher-order goods (including higher cost,
higher status items), plus all lower-order goods (such
as basic necessities) (King, 1984, p. 32; Smith, 1985).
As a result, higher-order markets are more diverse
than lower-order markets, and carry a greater proportion
of higher-order goods. In the presence of a
well-developed market hierarchy, market zones surrounding
the highest-order market centers are therefore
expected to have a greater diversity and greater
volume of high quality ceramics. In contrast, in a
non-hierarchically organized system, we can expect
that all market zones will be more equally diverse
and/or have roughly equal access to higher quality
goods.
A large number of indices have been developed to
measure assemblage richness and/or diversity (Ludwig
and Reynolds, 1988, pp. 85–95; Pielou, 1975);
most do not adequately account for the sample-size
effect; that is, that the larger the assemblage is, the
more artifact classes it should have, simply as a function
of sample size (Rhode, 1988). In contrast, the socalled
‘‘regression approach’’ (Grayson, 1984; Jones
et al., 1983), focuses on the systematic relationship
(generally linear or curvilinear) between assemblage
richness and assemblage size, and examines differences
in the rate of acquisition. The region with the steeper
slope addsartifact classes at a faster rate, indicating
access to a more diverse array of goods and, in this
context, proximity to higher-order markets.
With respect to higher quality goods, differential
access can be calculated by first using the production-step
index to rank the different types according
to the level of labor input or number of distinct steps
90 L.D. Minc / Journal of Anthropological Archaeology 25 (2006) 82–116
involved in their production (Feinman, 1980; Feinman
et al., 1981; Garraty, 2000). The distribution (relative
to market zone boundaries) of ceramics
belonging to different ‘‘labor input’’ classes can be
assessed using a simple v2
test for homogeneity, based
on type counts at sites assigned to market zones. Standardized
v2
residuals represent departures from the
expected frequencies in standard deviations (Reynolds,
1977, 1984). By examining the sign and
strength of the standardized v2
residuals, market
zones having greater access to more labor-intensive
ceramics can be identified.
Political congruence
The spatial organization of market zones relative
to features of political geography, including administrative
centers and territorial boundaries, provides
a potential measure of political controls over market
system participation. For prehistoric societies
where the distribution of administrative centers is
known from archaeological surveys, the extent of
market zones can be compared with reconstructed
political territories (e.g., Thiessen polygons) or
markers of ethnicity (such as representations of conquest
on stelae, etc.). Where administrative centers
and their dependent territories have been mapped
from documentary evidence such as tribute lists,
the degree of political congruence can be assessed
directly from the degree of spatial conformity
between ethnohistoric and archaeological data.
The proposed measures are summarized in Table
2. Clearly, all measures are relative scales, such that
assessment is best approached on a comparative
basis.
Case study: the Aztec empire and market system
reorganization
We now turn to an examination of regional patterns
of ceramic distribution as a basis for reconstructing
the major organizational features of the
Aztec market system through which these craft
goods and a host of other commodities circulated.
Market system organization is examined relative
to two distinct phases of political centralization:
before and after consolidation of the Aztec empire.
The purpose of this exercise is to illustrate the methodology
developed above as applied to the Basin of
Mexico; in the interests of brevity, it is not possible
to synthesize the extensive literature on the larger
Aztec economy, except as it directly relates to market
system organization.
Models for Early Aztec market systems
The Early Aztec or pre-imperial period in the
Basin of Mexico (traditionally dated to ca. 1150–
1350 AD)3
is known from the ethnohistoric documents
as a time of extreme political decentralization
and instability. At that time, the Basin was divided
Table 2
Archaeological measures of regional market systems
Dimension Definition Potential Measures
Delineation of market zones Areas serviced by the same market center(s)
that have access to the same array of goods
Spatially unconstrained cluster analysis to identify
market zones from areas with similar artifact
assemblages
Scale Relative size of market territories. Number and areal extent of distinct market zones
Network Degree of horizontal interaction or overlap
between market territories of the same order
a. Measures of similarity/ dissimilarity between
adjacent market zones
b. Degree of spatial congruence between artifact type
distributions and market zones
Hierarchy Degree of vertical interaction between market
territories of a different order
a. Assemblage richness/diversity
b. Differential access to higher quality goods
Political Congruence Spatial organization of market territories
relative to administrative centers and
territorial boundaries
Distribution of market zones relative to administrative
territories as reconstructed from:
a. Archaeological measures (CPT, Thiessen polygons,
etc.); or
b. Ethnohistoric documentation of territorial extent.
3
The absolute chronology for the pre-imperial period is
currently under revision (Cowgill, 1996; Nichols and Charlton,
1996; Parsons et al., 1996), in part based on radiocarbon dates
suggesting a much earlier inception than traditionally held. In
addition, while finer grain ceramic chronologies are available for
parts of the Basin, the complex variation in ceramic sequences
continues to frustrate attempts to refine the regional picture
beyond the traditional Early Aztec and Late Aztec employed
here.
L.D. Minc / Journal of Anthropological Archaeology 25 (2006) 82–116 91