31
Spatial and Statistical Inference of
Late Bronze Age Polities
in the Southern Levant
Stephen H. Savage
Department of Anthropology
Arizona State University
Box 872402
Tempe, AZ 85287–2402
stephen.savage@asu.edu
Steven E. Falconer
Department of Anthropology
Arizona State University
Box 872402
Tempe, AZ 85287–2402
falconer@asu.edu
introduction
A
rchaeological settlement patterns afford particularly
accessible and regionally comprehensive
avenues for interpreting societal
change according to a variety of themes, many of
them overlapping. Most political reconstructions
derived from pre- or protohistoric settlement patterns
infer the political status of whole societies.
This macroscopic approach is exemplified by the
definition of “states” as societies with specialized
central political control, which are manifested by
demonstrably hierarchical settlement systems (e.g.,
following Wright and Johnson 1975; Wright 1986).
According to the same logic, “pre-state” societies
(often subsumed under the terms “chiefdom” or
“complex chiefdom”) are conceptualized as antecedent
societal forms that lack some aspects of specialized
political control and are reflected by less
hierarchical settlement patterns (e.g., Earle 1987;
1991; Wright 1984). Even when heeding calls for
conceptual flexibility (e.g., Rothman 1994), this approach
implicitly assumes a linear evolutionary progression
from chiefdom to state (following Carneiro
1981; see critique in Yoffee 1993), which may engender
correspondingly constrained, linear interpretations
of prehistoric political trajectories.
In the case of the southern Levant (roughly the
region of modern Israel, the Palestinian Authority,
and western Jordan), these considerations are especially
pertinent because Bronze Age society, while
demonstrably complex as indicated most obviously
The Late Bronze Age (ca. 1500–1200 b.c.) marks the earliest opportunity to apply
substantial historical archives to the inference of spatially defined polities in the southern
Levant. A series of analyses of the Amarna Letters suggests numerous, small, bellicose
“city-states” differing considerably in political prominence and demographic
composition. We propose quantitative methods for analyzing archaeological settlement
data to explore the spatial configuration of Late Bronze Age polities and their varying
hierarchical structures. This approach provides an independent test of the historical
method, which identifies capital cities and assumes the adherence of surrounding communities,
by discerning polities from constellations of settlements, large and small,
amid the abundantly available regional survey data for the southern Levant. We infer
a political landscape that corresponds well with many aspects of historical reconstruction
and propose new ideas on the configuration and structure of Late Bronze Age
polities. In particular, the readily apparent balkanization of the southern Levant is
founded on significant structural variation between settlement and polities on the
Coastal Plain, the Central Hills, and the Jordan Rift. These results carry connotations
for the study of earlier and subsequent political dynamics. Our methods and inferences
are readily applicable to other cases of emergent political complexity in the southern
Levant and elsewhere, particularly those lacking historical documentation.
32 SAVAGE AND FALCONER BASOR 330
by the rise of early cities, does not fall neatly into
a standard anthropological category of “chiefdom”
or “state.” For example, Levy’s (1995) synthesis of
chiefdom-style organization during the Chalcolithic
period (ca. 4500–3500 b.c.) points out that statelevel
society did not arise in the Levant until the late
Iron Age, approximately two and one-half millennia
later. Traditional interpretations of Levantine sociopolitical
organization during the intervening Bronze
Age generally appeal to the less rigorously defined
concept of “city-state,” which may be characterized
as a compact, independent political unit consisting
of a capital and its hinterland, with correspondingly
limited political cohesion and power (see discussions
in Griffeth and Thomas 1981; Maisels 1990:
10–13; Charlton and Nichols 1997). Yoffee (1997)
has pointed out that the city-state concept seems so
intuitively obvious that it often eludes rigorous or
consistent definition. Not surprisingly, the inference
of city-state configurations on archaeological landscapes
also commonly lacks methodological rigor.
The challenge of tailoring analytically provocative
and archaeologically accessible lines of
sociopolitical interpretation may be addressed by
exploring the organization and interaction of political
units within societies, rather than attempting
to characterize these societies as whole political
entities. This approach avoids unwarranted assumptions
of political cohesion, as may be entailed when
pigeonholing an emergent complex society settled
over a large region as “chiefdom-” or “state-level.”
Renfrew’s idealized “early state module” (ESM)
provides one vehicle for conceptualizing the spatially
defined political units that might have comprised
many early complex societies. This concept
arises from the proposition that “in most, perhaps
all, early civilizations there function a number of
autonomous central places which, initially at least,
are not brought within a single unified jurisdiction”
(Renfrew 1975: 13). Among many potential functions,
each center administered a territorially defined
sociopolitical unit, or “polity” (Renfrew 1986: 1–2).
Analyses of the interactions between polities of
equivalent scale and status (i.e., “peer polity interaction”)
avoid the pitfalls of essentially taxonomic
descriptions of complex societies as “state” or “nonstate”
in favor of focusing on the dynamics between
ESMs and changes in those dynamics over time.
This approach normally entails the archaeological
or historical identification of political centers
and the subsequent extrapolation of their surrounding
ESMs, which may be depicted “notionally” as
Thiessen polygons1 (e.g., Renfrew 1975: fig. 2; 1986:
fig. 1.1). However, this entrée may prove less satisfactory
for societies lacking large central places or
in which large places were not necessarily “central.”
Primary attention to political centers also tends to
downplay the spatial implications of the remaining
communities in regional settlement systems. All these
considerations pertain to the study of early complex
society in the southern Levant, where Bronze Age
settlement trajectories feature persistent rural communities
and more transitory towns and cities (Falconer
1994; Falconer and Savage 1995).
The Bronze Age is epitomized traditionally as
the first era of urbanized society and “city-state” formation
in the southern Levant (Dever 1987; Richard
1987; Leonard 1989; Bunimovitz 1995; Gophna
1995; Ilan 1995), which may be associated culturally
with the Canaanites of the Old Testament and
other ancient texts (Matthiae 1981: 187; Bunimovitz
1995: 320; Ilan 1995: 297). These developments are
signaled by the advent, growth, and periodic collapse
of fortified towns and cities, although the sizes
of Levantine cities are modest compared with those
of neighboring regions (Falconer 1994). Bronze Age
diplomatic records, recovered primarily from Egypt,
also imply the nature of regional and local political
relations involving the cities of Canaan.
Levantine urbanization departs intriguingly from
prototypical patterns seen on grander scales elsewhere
in the Near East (e.g., Mesopotamia; Falconer
and Savage 1995). Further elucidation of these
differences requires the dissection of localized settlement
patterns within the subregions of the southern
Levant. We propose an archaeologically based,
quantitatively rigorous approach to the spatial analysis
of emergent polities and political trajectories.
In this pursuit, the interpretation of archaeological
settlement patterns need not be limited to corroborating
or refuting textual indications of political relations.
Instead, we treat texts and settlement patterns
1
Thiessen polygons divide a landscape into regions by choosing
prominent sites as centers, sketching connecting lines between
a center and its nearest neighbors, and then drawing
boundaries that bisect these connecting lines. The boundary lines
are then extended to define closed polygonal territories around
each center. This method invariably centers polities around a
major settlement, reifies an assumption of firm boundaries, and
overlooks surrounding networks of smaller communities.
2003 POLITIES IN THE SOUTHERN LEVANT 33
as independent aspects of the archaeological material
record, each requiring circumspect interpretation,
and each providing intriguing commentary on
the other.
documentary evidence of
late bronze age polities
A variety of ancient texts allude to the highly
fragmented political structure of Canaanite society,
usually as it impinges on foreign relations with
Syria and, especially, Egypt. The most robust and
systematic accounts of local and international Levantine
political relations appear in the “Amarna
Letters,” a diplomatic archive that includes numerous
communiqués between Canaanite rulers and the
pharaohs Amenhotep III and Amenhotep IV (Akhenaten).
These letters, written in Akkadian cuneiform
over about 25 years in the mid-14th century
b.c., form part of the royal archive brought by
Akhenaten from the former capital, Thebes, to his
new capital Akhetaten (Amarna) approximately 190
miles south of Cairo (Moran 1992: xiii; Naªaman
1992: 276). Subsequent pharaohs abandoned Akhetaten
and, beginning in 1887, local villagers began
unearthing tablets from Amarna’s relatively shallow
remains. Whole and fragmentary tablets now constitute
approximately 382 texts, which museums and
individuals have acquired from the antiquities market
and made available for analysis (Artzi 1988;
Moran 1992: xiii–xviii).
These texts voice the largely acrimonious interactions
of local Levantine polities and their petty
leaders, usually referred to as “mayors,” or less
commonly as “rulers,” “kings,” or “princes” (Moran
1992: xxxii; Naªaman 1992: 178). Although prior
Egyptian intervention in the southern Levant was
minimal (Ahituv 1978: 105; Weinstein 1981: 10),
Egypt established garrisons and administrative outposts
during the Late Bronze Age to secure tax
revenue, conscript labor, and political allegiance
(Naªaman 1981: 177; Weinstein 1981: 12–17). With
increased Egyptian military and political investment
by the end of the Late Bronze Age, Canaan apparently
assumed a loose colonial role within the imperial
economy (Naªaman 1981: 184; Weinstein 1981:
17–18). Six or seven of the recovered Amarna Letters
were sent from Pharaoh to Canaanite rulers, and
their blunt language conveys the unequivocal material
demands of the crown.
The much more numerous letters from Canaanite
mayors to Pharaoh begin with a “prostration formula”
that the author usually claims to perform
“seven times and seven times,” often “(both) on the
belly and on the back” (Moran 1992: xxvi–xxxiii).
These authors adopt a recurrent theme in proclaiming
their undying allegiance to Pharaoh and the
treacherous deceit of their neighbors. In so doing
they refer to their “cities” and other territorial holdings,
as well as to alliances between towns and internecine
quarrels over material goods or the favor
of the crown. Two unmistakable political subtexts
pervade this correspondence: very unequal power
relations between Canaan and Egypt, and serious
political fragmentation and volatility within Canaan.
Successive descriptive analyses of the Amarna
Letters offer historical appreciations of the Canaanite
political landscape. Helck (1962) attempted the
first Amarna-period political map of the entire eastern
Mediterranean seaboard by plotting the location
of towns that appear prominently in the Amarna
Letters. He then sketched territorial divisions from
his reading of the texts, segregating 23 Canaanite
polities in the southern Levant as we define it here
(Helck 1962: 191).
Subsequently, Naªaman (1988; 1992) used the
Amarna Letters, supplemented with Egyptian topographical
lists, royal inscriptions, biblical inferences,
and selected archaeological data, to identify 25–32
Canaanite “city-states” and autonomous towns lying
west of the Jordan River, along with six Egyptian
garrison towns whose local rulers were deposed and
replaced by the Egyptian crown. Naªaman proposes
a political landscape in which city-states of “higher
rank” influenced and sometimes dominated those of
“lesser rank” (1988: 18–19). Only Hazor and Shechem,
which apparently subordinated smaller nearby
cities, and possibly Gezer, which controlled coastal
and inland trade routes, were the heads of “territorial
kingdoms.” Each remaining city-state merely incorporated
a “capital” city surrounded by villages and
hamlets (1988: 18). He also suggests that the Galilee
and parts of the Levantine hill country were “barely
inhabited,” representing “a kind of no-man’s land”
occupied by “nomadic groups” (1988: 19).
This largely descriptive analysis inspired Bunimovitz
to infer his own matrix of polities, each “roughly
of the same size range, well below 1000 km2,” with
a radius less than 20 km (1995: 326; following
Naªaman 1988; 1992). He depicts the territorial
holdings around 18 capital cities as Thiessen poly-
34 SAVAGE AND FALCONER BASOR 330
gons (1995: fig. 6). This approach gives predominant
weight to historically selected political centers
between which equidistant territorial boundaries are
assumed (following, e.g., Haggett, Cliff, and Frey
1977: 436–39). Bunimovitz likens the political configuration
of the Late Bronze Age to a scaled-down
version of Renfrew’s ESM pattern for early civilizations
(1975: 12–18). This interpretation fills the
political map of the southern Levant, implicitly denying
Naªaman’s notion of settlement voids in the
hill country. However, he does note that the modest
scale of Levantine polities reflects a significant recession
in the number and size of cities between the
Middle and Late Bronze Ages (1995: 326; see also
Gonen 1984; Falconer 1994). This diminished scale
enabled lesser towns and city-states to enjoy political
autonomy, even when in close proximity to major
city-states.
Building on this prior scholarship, Finkelstein
(1996; see also Strange 2000) presents the most strident
reconstruction of the Late Bronze Age political
landscape based overwhelmingly on evidence from
the Amarna Letters, augmented by regional settlement
data. Spurning Naªaman’s no-man’s land, he
draws a political map “of the 14th century b.c.e.”
whereby “all of Canaan was divided between directly
adjacent polities with shared boundaries” (Finkelstein
1996: 225–26). This study identifies Late Bronze Age
political centers, then draws rough polygons “to establish
the theoretical, schematic boundaries between
the territories” (Finkelstein 1996: 226, maps 1–2). By
excluding 15 of Naªaman’s centers (because they are
judged arbitrarily to be too small, too close to other
centers, or without obvious hinterlands), Finkelstein
proposes a network of 14–17 city-states. Paralleling
Naªaman and Bunimovitz, Finkelstein argues for significant
variation between polities in territorial extent,
resident population, and labor force. For example, he
suggests that the highland city-states of Jerusalem,
Shechem, and Hazor, with their vast territorial holdings,
rugged topography, and limited settled population,
differed “in every possible respect” from the
smaller lowland city-states (1996: 243). Echoing Bunimovitz’s
commentary, he notes that this Late Bronze
Age configuration reflects a regional “demographic
crisis,” featuring a shortage of human labor in many
polities (1996: 243–45).
In sum, these historically based studies portray
the Amarna Age southern Levant as fragmented into
a relatively large number of modestly sized polities,
generally referred to as “city-states.” These entities,
which vary significantly in size, include larger hill
country polities (e.g., those around Shechem and
Jerusalem), with less dense, less sedentary populations,
and smaller lowland polities with more aggregated
agrarian towns. There is a general consensus
that this pattern reflects a decline in the number and
size of towns, as well as regional population, over
the Middle Bronze/Late Bronze Age transition. Although
the term “city-state” is used axiomatically,
there is no suggestion that Late Bronze Age society
involved state-level administration on a local or
regional scale.
The Amarna Letters provide the best, although
clearly less than exhaustive, documentation of Canaanite
political structure (e.g., see discussions in
Naªaman 1988: 17–21; 1992: 276–78). We compare
the evidence from this correspondence, written over
the course of about one generation, to settlement
data that span roughly three centuries (ca. 1500–
1200 b.c.). In doing so we do not claim to delineate
Canaanite polities at any specific point in time. Our
more modest goal is to propose and apply a quantitatively
based method with which we might approximate
the general structure and configuration of the
Levantine political landscape in the Late Bronze Age.
We compare historical reconstructions of Canaanite
polities with our independent settlement pattern analysis
to assess the degree to which the two approaches
approximate the same political landscape, thereby
providing a particularly robust portrait of Late Bronze
Age political structure.
quantitative inference
and analysis of polities
The southern Levant may be divided into a series
of longitudinal physiographic zones, including a low
elevation Coastal Plain, the Central Hills, and Galilee
(where maximum elevations exceed 1000 m asl),
and the Jordan Rift (which drops to 400 m bsl at
the Dead Sea) (fig. 1). The Jordan Rift stretches
north to include the Huleh Basin and communicates
with the Coastal Plain via the Jezreel Valley. Our
spatial analysis explores the Late Bronze Age political
landscape using site location and size data,
gleaned from a variety of published surveys and
from the records of the Israel Antiquities Authority,
that provide continuous coverage from the Mediterranean
coast to the eastern side of the Jordan
Valley (Ibrahim, Sauer, and Yassine 1976; 1988;
2003 POLITIES IN THE SOUTHERN LEVANT 35
Broshi and Gophna 1984; 1986; Gophna and Portugali
1988; Joffe 1991; 1993; Finkelstein and Gophna
1993). We use k-means and rank-size analyses to
explore these data as they reflect the political landscape
of Late Bronze Age Canaan. Our application
of k-means analysis considers site locations and
identifies clusters of settlements that, it may be argued,
represent approximations of spatially defined
polities. Rank-size analysis examines relative site
sizes to generate a quantified portrait of settlement
hierarchies. When used in combination, they provide
a quantitative means for integrating archaeological
and textual data to elucidate Late Bronze
Age and other ancient political landscapes.
There are several reasons why the spatial patterning
of settlements in early complex societies may
reflect geographically defined political units. A
fundamental tenet of locational analysis holds that
settlements may cluster around important geographical
features, local resources, or prominent centers
(Hodder and Orton 1976: 85). This notion is pivotal
for archaeological analyses based on distance-decay
relationships or gravity models. Simple economic
considerations, such as the cost of moving produce
in agricultural societies, suggest that people tend to
interact more frequently with the members of nearby
communities than with those of more distant settlements.
This observation is so fundamental to spatial
analysis that it very often goes unstated, becoming
a sort of “first postulate.” The transition to agriculture
and sedentary life engendered increasingly
“nodal” settlement systems. “Relatively speaking,
spatial order now became structured around points
and localities in place of vast distended ranges, and
around daily patterns of movements in place of annual
or seasonal migrations. . . . In short, increased
social interaction and social integration was used to
compensate for reduced spatial mobility” (Dodgshon
1987: 86–87). As this nodality and interaction
increased, new societal forms developed, based increasingly
on spatial organization. “Spatial propinquity
bred social propinquity . . .” (Dodgshon 1987:
86). That is, people living in nodes on the landscape
that were closer to each other tended to interact
more frequently. Settlement clustering undoubtedly
was molded by a variety of social and economic factors,
among which the spatial definition of polities
must have figured prominently. We should not presume
that these clusters constitute a straightforward
political map. Accordingly, we do not propose borders
between polities, even in the “notional” sense
implied by Thiessen polygons. We postulate simply
that these clusters of sites may be used to interpret
the spatial implications of Late Bronze Age political
organization.
K-Means Cluster Analysis
We infer site clusters using k-means analysis, a
form of divisive cluster analysis that is particularly
amenable to the analysis of spatial coordinates. Our
k-means analysis uses Palestine Grid coordinates for
474 Late Bronze Age site locations in the southern
Fig. 1. Aggregate survey coverage and physiographic regions
within the southern Levant considered in this study.
36 SAVAGE AND FALCONER BASOR 330
Levant, rendering each east site coordinate as an
x-axis value and each north coordinate as a y-axis
value. Each analysis begins with all site locations in
a single cluster, then divides the locations into two
clusters, then into three, and so on, until each observation
forms a cluster by itself or until a userrequested
maximum number of clusters is reached.
Each site location is included in the cluster with the
nearest center, based on a Euclidean distance calculation.
Through several passes, k-means analysis
minimizes the variance (i.e., the distance between
observations and cluster centers) at each clustering
level. As a result, it tends to form circular clusters,
and, accordingly, has been called “pure spatial clustering”
in archaeological literature (e.g., Kintigh and
Ammerman 1982).
Clearly, the variance for a data set is greatest
when all its observations are in one cluster and
reaches zero when each observation constitutes its
own cluster. For most analyses, the optimal clustering
solution lies somewhere in between. The analyst
must choose which clustering level or “solution” is
most appropriate for the data under consideration.
We use the program KMEANS (Kintigh 1994) to
generate a series of clustering solutions for each
data set, beginning with one cluster and continuing
to the program’s maximum of 30. KMEANS calculates
a statistic (Sqrt[SSE/N]) for each solution that
reflects the relative “tightness” of clustering within
each of its constituent clusters. This calculation
begins by summing the squared distances between
each site and its respective cluster center. This value
(SSE) is divided by the total number of sites under
analysis (N) to determine an average squared distance
for each solution. We calculate the square root
of SSE/N at each clustering level to measure the
“unsquared” average distance between sites and
their cluster centers, which is expressed in kilometers.
Absolute values for this statistic, Sqrt(SSE/N),
may be compared usefully between solutions. More
importantly, a comparison of Sqrt(SSE/N) for observed
data versus Sqrt(SSE/N) for randomized data
provides the basis for determining an optimal clustering
solution for each k-means analysis (see below).
KMEANS identifies the constituent sites for each
cluster and calculates a cluster center and root mean
square (RMS).2 The center of a k-means cluster is,
2
K-means cluster analysis starts with an observed distribution
of points and determines their most likely patterning of clustering.
more often than not, simply a point in space around
which a group of sites clusters. The RMS indicates
the average distance of all sites in a cluster
from their cluster center and provides the radius for
each cluster circle. Smaller RMS circles indicate
tighter clustering of constituent sites. Since all sites
are considered equally (i.e., there is no weighting of
sites according to size or presumed function), a cluster’s
“center of gravity” is pulled toward the greater
number of sites in that cluster (fig. 2).
KMEANS addresses the issue of whether the data
are clustered meaningfully by comparing the observed
data to a series of random runs. The program
stores the observed x-axis (east) and y-axis (north)
coordinates in arrays. For each random run, a y-axis
value from the array is randomly assigned to an
x-axis value from the array. Once all the x-values
have been assigned y-values, the program generates
a full range of k-means cluster solutions, each with
its own Sqrt(SSE/N) statistic. This process is repeated
100 times (the program’s maximum). The
summary statistics for the observed data are then
compared with those for the random runs at each
clustering level to determine which solution is optimal
(see application below).
Fig. 2. Spatial representation of a k-means cluster, summarizing
several basic concepts of k-means clustering.
We use the program KMEANS to generate a series of clustering
solutions (1-cluster, 2-cluster, 3-cluster, . . . , 30-cluster) for Late
Bronze Age site locations. In each clustering solution, every site
is assigned to the closest cluster center, forming circular clusters.
Each cluster can be depicted on a map by drawing a large circle
(the RMS circle), whose radius is equal to the average distance
of the cluster’s members from the center of the cluster. Figure 2
depicts several salient concepts underlying k-means analysis.
2003 POLITIES IN THE SOUTHERN LEVANT 37
Late Bronze Age Settlement Clusters
An optimal clustering solution for Late Bronze
Age settlement is determined from three particularly
important statistics. First, the %SSE value is calculated
at each cluster level. This statistic expresses
the percentage of the SSE in the observed data set
at each cluster level compared with the SSE in the
one-cluster solution. In effect, this value measures
the amount of residual variance in the data set at
each cluster level; it is 100 percent at the one-cluster
level, and would be zero if each site were in a cluster
by itself. Clearly, lower values of %SSE are preferred,
and values over 5 percent probably should
be avoided. Our analysis of Late Bronze Age settlement
data generated clustering solutions ranging
between 2 and 30 clusters. We remove the twothrough
ten-cluster solutions from consideration,
due to residual variance greater than 5 percent, as
indicated by the %SSE values (table 1).
Second, we subtract the Sqrt(SSE/N) for the observed
data from the mean Sqrt(SSE/N) from the
100 random runs for each cluster level to calculate
the “Sqrt(SSE/N) difference.” Since the Sqrt(SSE/N)
indicates the average distance between site locations
and their cluster centers, this value compares the
clustering of randomized locations with the clustering
of observed data. The difference is plotted at
each cluster level to show peaks (greater distances
between observed and random clustering) that suggest
preferred cluster levels. Peaks for the Late
Bronze Age data narrow the range of preferable
solutions to 11, 14, 18, and 24 clusters (fig. 3).
Third, we calculate a z-score for the SSE at each
cluster level by subtracting the mean SSE for the
random runs from the SSE for the observed data,
dividing the result by the standard deviation of the
random runs. The resulting z-score is always negative
in our analyses since the observed clustering is
always tighter than the random clustering. Z-scores
are normally distributed, and their associated probability
values may be read from statistical tables (e.g.,
Shennan 1988: 338–39). Lower (in this case, more
negative) z-scores indicate decreasing probabilities
that the observed clustering reflects a random distribution
of sites. Hence, we prefer clustering solutions
with lower z-scores, especially those scores
that differ most from z-scores for adjacent solutions.
The 24-cluster solution emerges as the optimal solution,
based on a z-score that is lowest among these
solutions and most different from scores for adjacent
solutions (in this case, the 23- and 25-cluster solutions)
(table 1).
This solution proposes a number of clusters comparable
to the number of polities expected by Helck
(1962), at the lower end of the range suggested by
Naªaman (1988; 1992), and greater than suggested
by Bunimovitz (1995) and Finkelstein (1996). Our
solution contrasts with those of Bunimovitz and
Finkelstein due to our inclusion of East Jordan Valley
Survey data and Finkelstein’s exclusion of many
potential political centers. More importantly, the
detailed results of this solution may be compared
Table 1. Residual Variance Values (%SSE)
and Z-Scores for Late Bronze Age
K-Means Clustering Solutions
Clusters % SSE Z-score
1 100.00 0.00
2 27.67 -1.25
3 20.02 -1.73
4 15.14 -2.89
5 11.28 -2.77
6 9.53 -2.28
7 8.41 -2.02
8 7.21 -2.09
9 6.17 -2.22
10 5.13 -2.52
11 4.30 -2.76
12 3.77 -2.85
13 3.25 -3.04
14 2.81 -3.25
15 2.60 -3.32
16 2.39 -3.45
17 2.18 -3.51
18 2.01 -3.54
19 1.92 -3.48
20 1.80 -3.44
21 1.68 -3.47
22 1.58 -3.57
23 1.50 -3.56
24 1.39 -3.66
25 1.35 -3.62
26 1.30 -3.68
27 1.25 -3.69
28 1.21 -3.66
29 1.18 -3.57
30 1.13 -3.57
Note: The preferred solution is shown to be at 24 clusters.
38 SAVAGE AND FALCONER BASOR 330
with historical reconstructions as each method
sheds light on the configuration of Late Bronze Age
polities.
The Configuration of Late Bronze Age Polities
Most generally, there is a remarkable correspondence
of k-means settlement clusters with the
geographical configuration of polities inferred historically
from the Amarna Letters (e.g., as plotted by
Bunimovitz 1995: fig. 6) (fig. 4). This agreement is
particularly strong for the Coastal Plain, the Jezreel
Valley, and the Central Hills. In a smaller number
of areas, incongruities may be interpreted in light
of Amarna-period politics, Levantine geography, or
the details of archaeological and historical methods
of inference. For example, our inclusion of data
from the East Jordan Valley Survey extends our
database slightly east of the regions discussed by
Naªaman, Bunimovitz, and Finkelstein. As a result,
two k-means clusters, one east of Shechem and the
other just north of the Dead Sea, are added to their
reconstructions. When the East Jordan Valley sites
are excluded from our k-means analysis, these clusters
are omitted accordingly.
Other specific questions of congruence involve
historical contention. For instance, Helck (1962)
proposes two polities in the Huleh Basin headed by
Hazor and Samhuna, which parallel the two k-means
clusters shown by our analysis, while the other
authors suggest a single municipality centered at
Hazor. On the southern Coastal Plain, Naªaman
(1988) and Finkelstein (1996) do not include GathCarmel
among their city-states, whereas our analysis
concurs with Bunimovitz in suggesting a polity
in this area. Other historical evidence suggests that
Bunimovitz’s Gezer polity should be subdivided, as
suggested by three k-means clusters in that polygon.
Beck and Kochavi note that “Late Bronze Age
Aphek, which is mentioned in the time of Thutmose
III and Amenhotep II as a Canaanite citystate,
probably became an Egyptian fortress during
the time of Ramesses II, as did nearby Jaffa [Yafo]”
Fig. 3. Values for the “Sqrt(SSE/N) difference” [observed
Sqrt(SSE/N)—mean random Sqrt(SSE/N)] for Late Bronze
Age k-means clustering solutions, showing preferred solutions
at 11, 14, 18, and 24 clusters.
Fig. 4. A comparison of Late Bronze Age city-state capitals
and territories (shown as Thiessen polygons) inferred historically
by Bunimovitz (1995: fig. 6) with settlement clusters
produced by k-means cluster analysis in this study.
Cluster circle sizes correspond to RMS radii; smaller circles
indicate tighter clustering.
2003 POLITIES IN THE SOUTHERN LEVANT 39
(1993: 68). These two centers figured prominently in
Egyptian grain exports from Palestine, as attested by
a letter from Ugarit found in the Late Bronze Age
Governor’s residence at Aphek (see Owen 1981;
Singer 1983).
The lack of glaring discrepancies suggests that
both methods are indeed approximating a common
political landscape, including a combination of welldefined
probable polities and a much smaller handful
of more debatable entities. In light of previous active
discussion of the relative cohesion and dynamic
interplay between these polities, we use rank-size
analysis to consider structural variations between
polities and their likely implications for Levantine
political dynamics.
Rank-Size Analysis
Rank-size analysis provides a quantitative means
of assessing the political and economic integration
of modern and premodern complex societies. The
“rank-size rule” states that in mature, well-integrated
settlement hierarchies, the size of any nth-ranked
place may be predicted by dividing the size of the
largest place by n, such that the rank and population
of these communities describe a straight-line
log-normal distribution when plotted logarithmically
(Zipf 1949; see discussions in Falconer and Savage
1995; Savage 1997). Such log-normal distributions
“appear to be typical of larger countries with a long
tradition of urbanism, which are politically and economically
complex” (Berry 1961: 582). Since archaeological
settlement hierarchies rarely adhere to
the rank-size rule, their interpretation generally derives
from the manner and degree to which rank-size
distributions depart from log-normal (e.g., Johnson
1972; 1977; 1980; Blanton 1976; Crumley 1976; Adams
1981; Paynter 1982; Wright 1986; Gophna and
Portugali 1988; Finkelstein and Gophna 1993; Bunimovitz
1995; Falconer and Savage 1995; Liu 1996;
Harrison 1997; Savage 1997). Site distributions with
slopes steeper than predicted by the rank-size rule are
referred to as “primate,” and those with shallowerthan-predicted
slopes are “convex,” while compound
curves with a primate upper element and a convex
lower component may be labeled “primo-convex”
(see Falconer and Savage 1995: fig. 1). These departures
are molded especially by the “closure” of the
settlement system under study (i.e., the degree to
which interactions are bounded within that system)
and the “integration” of communities within a system
(i.e., the relative frequency of interaction between
them) (Vapnarsky 1968; Johnson 1980).
Archaeological interpretations often rely simply
on judgmental appraisals of the shapes of rank-size
distributions. This approach avoids the issue of how
far a distribution must depart from log-normal to
be considered “primate” or “convex.” Some studies
use the Kolomogorov-Smirnov one-tailed goodnessof-fit
test (the “K- test”) to determine whether an
observed rank-size plot differs significantly from a
log-normal curve predicted by the rank-size rule
(see discussion in Sokal and Rohlf 1981).3 Traditionally,
a value (the “K- statistic”) is calculated as
the maximum deviation between the observed set of
site sizes and a log-normal distribution (Thomas
1986: 322–37; Shennan 1988: 53–61). The K- statistic
is compared with predetermined threshold values
to determine whether the observed departure from
log-normal is statistically significant (Thomas 1986:
504–6).
Traditional applications of the Kolomogorov test,
however, assume that the data under analysis represent
complete, clearly bounded settlement systems.
While this may be true for studies of modern
nations, rank-size analysis of prehistoric settlement
must accommodate the effects of sampling on archaeological
data. Therefore, we apply the K- test
using Monte Carlo simulation to estimate the probability
that an observed departure could be equaled or
exceeded by drawing sites at random from a lognormal
population (see Falconer and Savage 1995;
Savage 1997). Unlike traditional applications of the
K- test, this method provides a quantified basis for
judging whether any given rank-size curve departs
sufficiently from log-normal to merit interpretation
as “primate” or “convex.” This method provides an
archaeologically appropriate means for the structural
3
We use the Ktest
to assess the null hypothesis that any
given rank-size distribution simply reflects a random departure
from log-normal. The test calculates the absolute maximum deviation
between an observed rank-size distribution and the distribution
that is expected according to the rank-size rule. The
Kstatistic
expresses the proportion of observed cases that are
“out of place,” in the sense that they are smaller or larger than the
expected according to the rank-size rule. The value of the statistic
ranges, therefore, between 0.0 (indicating strict adherence to the
rank-size rule) and 1.0 (reflecting pervasive departure from the
rank-size rule). The Kstatistic
may approach, but will never
reach, 1.0 because the size of the largest site provides the starting
point for an expected log-normal distribution and, therefore,
always conforms to the rank-size rule.
40 SAVAGE AND FALCONER BASOR 330
assessment of settlement clusters and the polities
they may approximate.
Structural Variation of Late Bronze Age
Settlement Clusters
Fundamental contrasts between regional systems
(including early states) and localized city-states lie
not only in their geographical extent, but in their
political and economic structure. Coherent regional
systems incorporate communities varying in size
and function over large areas using a hierarchy of
administrative centers (e.g., Maisels 1990: 10–13;
Trigger 1993: 8–9). In contrast, city-states are characterized
by territorial compactness and tight integration
based on the dominance of a capital over its
subordinate hinterland (Griffeth and Thomas 1981;
Maisels 1990: 10–13; Trigger 1993: 8–14; Charlton
and Nichols 1997).
Well-bounded and integrated regional systems
should be reflected in log-normal distributions. However,
rank-size analysis of Late Bronze Age settlement
data (using 369 sites of known size) produces
convex curves for the southern Levant generally and
for the subregions of the Coastal Plain and Central
Hills, and a log-normal distribution only for the Jordan
Valley (table 2). These results most likely indicate
a pattern of regional and subregional “pooling”
of multiple localized systems, with the possibility of
a more bounded and/or integrated settlement system
along the Jordan Rift.
In contrast, tightly integrated city-states with
prominent capitals should produce primate distributions
reflecting the local dominance of these centers.
In addition, documentary analyses of Late Bronze
Age polities concur that we should expect basic
differences in political and economic structure between
subregions (most notably between the less
sedentary Hill Country and the more densely populated
Coastal Plain and Jordan Valley) and between
individual polities of varying political consolidation
and power.
Rank-size analyses of individual k-means settlement
clusters show marked contrasts in the structure
of polities within and between the subregions of
the southern Levant (fig. 5; table 3). The most pronounced
pattern shows consistently convex distributions
for the Hill Country clusters along the Central
Hills and the Jezreel Valley. Among these 10 clusters
only Cluster 19 has a primo-convex curve, with
a slightly primate upper portion to accompany its
convex lower component. In contrast, all of the lognormal
and primate distributions correspond with
clusters along the Coastal Plain and the Jordan Rift.
This pattern seems to corroborate Naªaman’s and
Finkelstein’s suggestion of a basic structural dichotomy
between the Levantine uplands and lowlands.
The convex curves for the upland clusters suggest
polities marked by varying combinations of poorly
defined boundaries, minimal settlement integration,
and the lack of a prominent capital. These characteristics
accord well with Finkelstein’s portrait of
large polities, including those around Jerusalem and
Shechem, populated with only a limited sedentary
population (1997: 243). The lone deviation from
this pattern stems from the large size of Gezer (in
Cluster 19), which is seen as the capital of a leading
Canaanite city-state by Naªaman (1988: 18) and
Finkelstein (1996: 233).
The clusters of the Coastal Plain include an array
of different rank-size distributions denoting varying
degrees of integration and prominence among their
Table 2. Results of Late Bronze Age Regional and Subregional Rank-Size Analyses*
Region
Number of
sitesa
Sample
proportionb
Number of sites
in populationc
Observed Kvalue
Probabilityd Curve shape
Southern Levant 369 .75 492 .495 <.001 Convex
Southern Coastal Plain 105 .75 140 .371 <.001 Convex
Central Hills 108 .75 144 .407 <.001 Convex
Jordan Valley 60 .75 80 .233 <.543 L-normal
* Cf. Falconer and Savage 1995: table 4.
a
Includes only sites of known size.
b
See discussion in Falconer and Savage 1995: 41–44.
c
Number of sites in population = number of observed sites x (1/sample percent).
d
Probability of drawing a Kvalue
greater than or equal to the observed value at random from a log-normal population, based on
1,000 random runs for each row.
LONG
2003 POLITIES IN THE SOUTHERN LEVANT 41
capitals. The coastal clusters that include larger possible
capitals (Acco, Gath-Padalla, Achshaf, Aphek,
Jaffa) produced log-normal, primo-convex, or primate
curves, suggesting the varying prominence of
those cities, while clusters with smaller potential
centers (Ashkelon, Yurza) revealed convex curves.
This pattern brings to mind Naªaman’s suggestion
of “higher rank” polities influencing or dominating
those of “lesser rank” (1988: 19) and suggests the
possibility of particularly acute political inequity and
volatility along the coastal lowlands.
The Jordan Rift is marked by commonly primate
rank-size distributions (for four out of five clusters)
that suggest a series of well-integrated city-state
polities, headed by clear capital communities. Clusters
1 and 11, with capitals possibly at Rehov (or
Table 3. Results of Rank-Size Analysis of 24 Late Bronze Age Settlement Clusters
Identified by K-Means Cluster Analysis*
Subregion Clusterd
No. of sites in
K-means
clustere
No. of sites in
rank-size curvef
Observed Kvalue
Probabilityg
Curve shape
Coastal Plaina
2 28 21 .333 <.001 Convex
5 31 19 .368 <.001 P-convexh
8 24 21 .286 <.084 L-normal
9 11 10 .600 <.001 Primate
14 17 15 .333 <.001 Primate
17 7 6 .500 <.001 Convex
21 22 17 .235 <.453 L-normal
22 7 5 .400 <.001 Primate
Hill Countryb
3 16 13 .385 <.001 Convex
4 37 31 .290 <.032 Convex
7 18 9 .500 <.001 Convex
10 30 29 .310 <.001 Convex
15 25 23 .522 <.001 Convex
16 11 8 .375 <.001 Convex
18 36 34 .324 <.001 Convex
19 20 16 .375 <.001 P-convexh
20 6 4 .500 <.001 Convex
24 9 5 .400 <.001 Convex
Jordan Riftc
1 46 43 .442 <.001 Primate
6 19 4 .500 <.001 Primate
11 28 24 .208 <.320 L-normal
12 16 7 .429 <.001 Primate
13 4 0 - - -
23 6 5 .600 <.001 Primate
* Note that small sample sizes for some clusters may make their interpretation less conclusive.
a
Includes Northern and Southern Coastal Plain
b
Includes Central Hills, Jezreel Valley, Upper and Lower Galilee
c
Includes Jordan Valley and Huleh Basin
d
Cluster numbers are assigned randomly by KMEANS
e
Includes all Late Bronze Age sites
f
Includes only Late Bronze Age sites of known size
g
Probability of drawing a Kvalue
greater than or equal to the observed value at random from a log-normal population, based
on 1,000 random runs for each row
h
Primo-convex curve
42 SAVAGE AND FALCONER BASOR 330
Pella) and Anaharath, respectively, represent heavily
populated polities with modestly sized centers. To
the north lie two clusters with fewer hinterland villages,
but pronounced primate curves indicative of
the prominence of their larger leading cities. Cluster
12 is particularly noteworthy because the primacy
and precociously large size of Hazor reinforce its
likely importance as a crucial node of political and
economic ties with the nearby Damascus Basin.
This study portrays the southern Levant with
three distinct political patterns as revealed by quantitative
analysis of regional and local settlement patterns.
The Coastal Plain, the setting for the region’s
largest political and economic centers, conventionally
seen as the hearth of Canaanite civilization,
emerges as a hodge-podge of polities with highly
variable structures and their attendant political
connotations. The Jordan Rift, normally seen as a
smaller-scale backwater off the Mediterranean littoral,
features settlement patterns most consistent
with a series of highly integrated peer polities or
city-states, and subregional political coherence. In
contrast to both of these lowland areas, the settlement
clusters of the Hill Country are more dispersed,
with consistent evidence of less settlement integration.
When considered structurally, these results suggest
three fundamentally different bases for political
development in a region normally viewed as a single,
albeit fractious, social and cultural entity during
the Late Bronze Age. These distinctions help illuminate
the foundations of the particularly volatile
political dynamics of the southern Levant.
conclusions
The Late Bronze Age of the southern Levant
has received considerable attention as the nexus between
materially based analyses of prehistoric societies
and textually based analyses of the subsequent
historical era. Among textual sources, the Amarna
Letters provide the earliest archival source to support
robust inferences of Levantine political dynamics.
Historical appreciations of the Amarna Age,
which tend to use archaeological data primarily in
the service of the written word, now may be supplemented
with independent quantitative analyses of
settlement data toward a fuller appreciation of the
Late Bronze Age political landscape. A major goal
of this study is to provide a quantitative means of
applying the abundantly available settlement data of
the southern Levant to explore local and regional
political dynamics more fully and fundamentally.
Our study corroborates the venerable interpretation
of a fractious and volatile countryside balkanized
into a relatively large number of modestly sized polities
that varied considerably in extent and composition.
In particular, we note considerable disparity in
polity size throughout the region, and most notably
in structure when comparing highland and lowland
polities and subregions. We suggest that these structural
disparities in particular lie at the heart of Late
Bronze Age political contentiousness. Given the general
congruence of our settlement analysis with preFig.
5. Geographical distribution of k-means settlement clusters
and their structures as interpreted from rank-size analyses
in this study. Cluster numbers are assigned arbitrarily by
the program KMEANS. Cluster 13 contained sites of uncertain
size and was not analyzed (see table 3).
2003 POLITIES IN THE SOUTHERN LEVANT 43
vious historical interpretations, the methods we propose
hold promise for inferring longer-term patterns
in the deeper past (e.g., during the emergence of
Bronze Age urbanism). Moreover, many of the patterns
we infer (most notably the dichotomy between
highland and lowland political configurations and
structure) hold considerable significance for the historical
periods that follow (e.g., regarding the Iron
Age emergence of Israelite ethnicity and statehood).
Adams, R. McC.
1981 Heartland of Cities: Surveys of Ancient Settlement
and Land Use on the Central Floodplain of
the Euphrates. Chicago: University of Chicago.
Ahituv, S.
1978 Economic Factors in the Egyptian Conquest of
Canaan. Israel Exploration Journal 28: 93–105.
Artzi, P.
1988 The Present State of the Amarna Documents.
Pp. 3–16 in Proceedings of the Ninth World
Congress of Jewish Studies, Bible Studies and
Ancient Near East, Jerusalem 1985, ed. M.
Goshen-Gottstein. Jerusalem: Magnes.
Beck, P., and Kochavi, M.
1993 Aphek (in Sharon). Pp. 62–72 in The New Encyclopedia
of Archaeological Excavations in
the Holy Land, Vol. 1, ed. E. Stern. Jerusalem:
Carta.
Berry, B. J. L.
1961 City Size Distributions and Economic Development.
Economic Development and Cultural
Change 9: 573–87.
Blanton, R. E.
1976 The Role of Symbiosis in Adaptation and
Sociocultural Change in the Valley of Mexico.
Pp. 181–201 in The Valley of Mexico: Studies
in Pre-Hispanic Ecology and Society, ed E. R.
Wolf. Albuquerque, NM: University of New
Mexico.
Broshi, M., and Gophna, R.
1984 The Settlements and Population of Palestine
during the Early Bronze Age II–III. Bulletin
of the American Schools of Oriental Research
253: 41–53.
1986 Middle Bronze Age II Palestine: Its Settlement
and Population. Bulletin of the American
Schools of Oriental Research 261: 73–90.
Bunimovitz, S.
1995 On the Edge of Empires—Late Bronze Age
(1500–1200 bce). Pp. 320–31 in The Archaeology
of Society in the Holy Land, ed. T. E.
Levy. New York: Facts on File.
Carneiro, R.
1981 The Chiefdom: Precursor of the State. Pp. 37–
79 in The Transition to Statehood in the New
World, eds. G. D. Jones and R. R. Kautz. Cambridge:
Cambridge University.
Charlton, T. H., and Nichols, D. L.
1997 The City-State Concept: Development and
Applications. Pp. 1–14 in The Archaeology of
City-States, Cross-Cultural Approaches, eds.
D. L. Nichols and T. H. Charlton. Washington,
DC: Smithsonian Institution.
Crumley, C. L.
1976 Toward a Locational Definition of State Systems
of Settlement. American Anthropologist
78: 59–73.
Dever, W. G.
1987 The Middle Bronze Age: The Zenith of the
Urban Canaanite Era. Biblical Archaeologist
50: 148–77.
Dodgshon, R. A.
1987 The European Past: Social Evolution and
Spatial Order. London: Macmillan.
Earle, T. K.
1987 Chiefdoms in Archaeological and Ethnohistorical
Perspective. Annual Review of Anthropology
16: 279–308.
1991 The Evolution of Chiefdoms. Pp. 1–15 in
Chiefdoms: Power, Economy, and Ideology, ed.
T. K. Earle. New York: Cambridge University.
Falconer, S. E.
1994 The Development and Decline of Bronze
Age Civilisation in the Southern Levant: A
Reassessment of Urbanism and Ruralism. Pp.
305–33 in Development and Decline in the
Mediterranean Bronze Age, eds. C. Mathers
and S. Stoddart. Sheffield Archaeological Monographs
8. Sheffield: Collis.
Falconer, S. E., and Savage, S. H.
1995 Heartlands and Hinterlands: Alternative Trajectories
of Early Urbanization in Mesopotamia
and the Southern Levant. American Antiquity
60: 37–58.
references
44 SAVAGE AND FALCONER BASOR 330
Finkelstein, I.
1996 The Territorial-Political System of Canaan in
the Late Bronze Age. Ugarit-Forschungen 28:
221–55.
Finkelstein, I., and Gophna, R.
1993 Settlement, Demographic, and Economic Patterns
in the Highlands of Palestine in the Chalcolithic
and Early Bronze Periods and the
Beginning of Urbanism. Bulletin of the American
Schools of Oriental Research 289: 1–22.
Gonen, R.
1984 Urban Canaan in the Late Bronze Period.
Bulletin of the American Schools of Oriental
Research 253: 61–73.
Gophna, R.
1995 Early Bronze Age Canaan: Some Spatial and
Demographic Observations. Pp. 269–80 in The
Archaeology of Society in the Holy Land, ed.
T. E. Levy. New York: Facts on File.
Gophna, R., and Portugali, J.
1988 Settlement and Demographic Processes in Israel’s
Coastal Plain from the Chalcolithic to the
Middle Bronze Age. Bulletin of the American
Schools of Oriental Research 269: 11–28.
Griffeth, R., and Thomas, C. G.
1981 Introduction. Pp. xiii–xx in The City-State in
Five Cultures, eds. R. Griffeth and C. G. Thomas.
Santa Barbara, CA: ABC-Clio.
Haggett, P.; Cliff, A. D.; and Frey, A. E.
1977 Locational Analysis in Human Geography. New
York: Wiley and Sons.
Harrison, T. P.
1997 Shifting Patterns of Settlement in the Highlands
of Central Jordan during the Early
Bronze Age. Bulletin of the American Schools
of Oriental Research 306: 1–37.
Helck, W.
1962 Die Beziehungen Ägyptens zu Vorderasien
im 3. und 2. Jahrtausend v. Chr. Wiesbaden:
Harrassowitz.
Hodder, I., and Orton, C.
1976 Spatial Analysis in Archaeology. Cambridge:
Cambridge University.
Ibrahim, M. M.; Sauer, J.; and Yassine, K.
1976 The East Jordan Valley Survey, 1975. Bulletin
of the American Schools of Oriental Research
222: 41–66.
1988 The East Jordan Valley Survey, 1976 (Part Two).
Pp. 189–207 in Archaeology of Jordan: Essays
and Reports, ed. K. Yassine. Amman: Department
of Archaeology, University of Jordan.
Ilan, D.
1995 The Dawn of Internationalism—The Middle
Bronze Age. Pp. 297–319 in The Archaeology
of Society in the Holy Land, ed. T. E. Levy.
New York: Facts on File.
Joffe, A. H.
1991 Early Bronze I and the Evolution of Social
Complexity in the Southern Levant. Journal of
Mediterranean Archaeology 4: 3–58.
1993 Settlement and Society in the Early Bronze I &
II Southern Levant: Complementarity and Contradiction
in a Small-Scale Complex Society.
Monographs in Mediterranean Archaeology 4.
Sheffield: Sheffield Academic.
Johnson, G. A.
1972 A Test of the Utility of Central Place Theory in
Archaeology. Pp. 769–85 in Man, Settlement,
and Urbanism, eds. P. J. Ucko, R. Tringham,
and G. W. Dimbleby. London: Duckworth.
1977 Aspects of Regional Analysis in Archaeology.
Annual Review of Anthropology 6: 479–508.
1980 Rank-Size Convexity and System Integration:
A View from Archaeology. Economic Geography
56: 234–47.
Kintigh, K. W.
1994 Tools for Quantitative Archaeology. Tempe,
AZ: Arizona State University.
Kintigh, K. W., and Ammerman, A. J.
1982 Heuristic Approaches to Spatial Analysis in Archaeology.
American Antiquity 47: 31–63.
Leonard, A., Jr.
1989 The Late Bronze Age. Biblical Archaeologist
52: 4–39.
Levy, T. E.
1995 Cult, Metallurgy and Rank Societies—Chalcolithic
Period (ca. 4500–3500 bce). Pp. 226–44
in The Archaeology of Society in the Holy
Land. New York: Facts on File.
Liu, L.
1996 Settlement Patterns, Chiefdom Variability, and
the Development of Early States in North
China. Journal of Anthropological Archaeology
15: 237–88.
Maisels, C. K.
1990 The Emergence of Civilization: From Hunting
and Gathering to Agriculture, Cities, and the
State in the Near East. London: Routledge.
Matthiae, P.
1981 Ebla: An Empire Rediscovered. Garden City,
NY: Doubleday.
Moran, W. L.
1992 The Amarna Letters. Baltimore: Johns Hopkins
University.
Naªaman, N.
1981 Economic Aspects of the Egyptian Occupation
of Canaan. Israel Exploration Journal 31:
172–85.
2003 POLITIES IN THE SOUTHERN LEVANT 45
1988 Historical-Geographical Aspects of the Amarna
Tablets. Pp. 17–26 in Proceedings of the Ninth
World Congress of Jewish Studies, Bible Studies
and Ancient Near East, Jerusalem 1985, ed.
M. Goshen-Gottstein. Jerusalem: Magnes.
1992 Canaanite Jerusalem and Its Central Hill Country
Neighbours in the Second Millennium b.c.e.
Ugarit-Forschungen 24: 275–91.
Owen, D.
1981 An Akkadian Letter from Ugarit at Tel Aphek.
Tel Aviv 8: 1–17.
Paynter, R.
1982 Models of Spatial Inequality: Settlement Patterns
in Historical Archeology. New York:
Academic.
Renfrew, C.
1975 Trade as Action at a Distance: Questions of
Integration and Communication. Pp. 3–59 in
Ancient Civilization and Trade, eds. J. A.
Sabloff and C. C. Lamberg-Karlovsky. Albuquerque,
NM: University of New Mexico.
1986 Introduction: Peer Polity Interaction and Sociopolitical
Change. Pp. 1–18 in Peer Polity
Interaction and Socio-political Change, eds.
C. Renfrew and J. F. Cherry. Cambridge: Cambridge
University.
Richard, S.
1987 The Early Bronze Age: The Rise and Collapse
of Urbanism. Biblical Archaeologist 50: 22–
43.
Rothman, M. S.
1994 Sealing as a Control Mechanism in Prehistory:
Tepe Gawra XI, X and VIII. Pp. 103–20
in Chiefdoms and Early States in the Near
East: The Organizational Dynamics of Complexity,
eds. G. Stein and M. S. Rothman.
Monographs in World Archaeology 18. Madison,
WI: Prehistory.
Savage, S. H.
1997 Assessing Departures from Log-Normality in
the Rank-Size Rule. Journal of Archaeological
Science 24: 233–44.
Shennan, S.
1988 Quantifying Archaeology. Edinburgh: Edinburgh
University.
Singer, I.
1983 Takuhlinu and Haya: Two Governors in the
Ugarit Letter from Tel Aphek. Tel Aviv 10: 3–25.
Sokal, R. R., and Rohlf, F. J.
1981 Biometry: The Principles and Practice of Statistics
in Biological Research. San Francisco:
Freeman.
Strange, J.
2000 The Palestinian City-States of the Bronze
Age. Pp. 67–76 in A Comparative Study of
Thirty City-State Cultures, An Investigation
Conducted by the Copenhagen Polis Centre,
ed. M. H. Hansen. Historisk-filosofiske Skrifter
21. Copenhagen: Kongelige Danske Videnskabernes
Selskab.
Thomas, D. H.
1986 Refiguring Anthropology: First Principles of
Probability and Statistics. Prospect Heights,
IL: Waveland.
Trigger, B. G.
1993 Early Civilizations: Ancient Egypt in Context.
Cairo: American University in Cairo.
Vapnarsky, C. A.
1968 On Rank-Size Distributions of Cities: An Ecological
Approach. Economic Development and
Culture Change 17: 584–95.
Weinstein, J. M.
1981 The Egyptian Empire in Palestine: A Reassessment.
Bulletin of the American Schools of
Oriental Research 241: 1–28.
Wright, H. T.
1984 Prestate Political Formations. Pp. 41–77 in On
the Evolution of Complex Societies: Essays in
Honor of Harry Hoijer, 1982, eds. T. K. Earle.
Los Angeles: Undena.
1986 The Evolution of Civilizations. Pp. 323–65
in American Archaeology Past and Future:
A Celebration of the Society for American
Archaeology, 1935–1985, eds. D. J. Meltzer,
D. D. Fowler, and J. A. Sabloff. Washington,
DC: Smithsonian Institution.
Wright, H. T., and Johnson, G. A.
1975 Population, Exchange and Early State Formation
in Southwestern Iran. American Anthropologist
77: 267–89.
Yoffee, N.
1993 Too Many Chiefs? Or, Safe Texts for the 90s.
Pp. 60–78 in Archaeological Theory: Who Sets
the Agenda? eds. N. Yoffee and A. Sherratt.
Cambridge: Cambridge University.
1997 The Obvious and the Chimerical: City-States
in Archaeological Perspective. Pp. 255–63 in
The Archaeology of City-States, Cross-Cultural
Approaches, eds. D. L. Nichols and T. H. Charlton.
Washington, DC: Smithsonian Institution.
Zipf, G. K.
1949 Human Behavior and the Principle of Least
Effort: An Introduction to Human Ecology.
Cambridge, MA: Addison-Wesley.