Society for American Archaeology
Heartlands and Hinterlands: Alternative Trajectories of Early Urbanization in Mesopotamia
and the Southern Levant
Author(s): Steven E. Falconer and Stephen H. Savage
Source: American Antiquity, Vol. 60, No. 1 (Jan., 1995), pp. 37-58
Published by: Society for American Archaeology
Stable URL: http://www.jstor.org/stable/282075 .
Accessed: 14/09/2011 13:44
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
Society for American Archaeology is collaborating with JSTOR to digitize, preserve and extend access to
American Antiquity.
http://www.jstor.org
HEARTLANDS
AND HINTERLANDS:
ALTERNATIVE TRAJECTORIES OF EARLY URBANIZATION
AND THE SOUTHERN LEVANT
MESOPOTAMIA
IN
Steven E. Falconer and Stephen H. Savage
Comparativerank-sizeanalyses revealhighly variablecoursesof urbanizationin ancient Mesopotamiaand the
southernLevantduringthefourth throughearly second millenniaB.C. Whiletraditionalrank-sizemethodsdo not
considerthe effectsof archaeologicalsampling, we proposea revisedapproachbased on Monte Carlo simulation,
whichincorporatessite-recoveryrates and demonstratesthe advantagesof 'full-coverage"survey.We highlightthe
rapiddevelopmentof urbanprimacyin southernMesopotamia'sheartland(Adams1981) and the morestatic rural
integrationof the Diyala hinterland(Adams 1965). In contrast,BronzeAge urbanizationin the southernLevant
describesa mosaic of urbanand rural systemsfollowing independenttrajectories.We call for greaterattentionto
small sites, whichoftendefinethe shape of rank-sizedistributions.Ourapproachilluminatesmodestcases of urbanization in terms of structure,ratherthan simply of reducedscale, and avoidsa tendencyto categorizesuch cases as
derivative.
Los andlisiscomparativosdel rangode tamaniode asentamientosrevelanuna gran variabilidaden el rumbohacia
la urbanizacionen la Mesopotamiaantiguay en el sur de Levantea travesdel cuartomilenio hasta principiosdel
segundo milenio A.C. Los metodos de rango-tamanoempleadostradicionalmenteno consideranlos efectos del
muestreoarqueologico.Por ello proponemosuna perspectivadistinta de aquellos,basada en la simulacionMonte
Carlo la cual incorporaestimacionesde los sitios recuperadosy demuestralas ventajasde los reconocimientosde
total."Distinguimosla primaciadel rdpidodesarrollourbanoen el nucleode la Mesopotamia
superficiede "cobertura
surena(Adams1981) de la integracionruralmds estdticaen la periferiadel Diyala (Adams1965). En contraste,la
urbanizacionen la Edad de Bronce en el sur de Levantepresenta un mosaico de sistemas ruralesy urbanosque
siguen trayectoriasindependientes.Ponemosmayoratencionen los sitiospequenos,los cualesconfrecuenciadefinen
la forma de las distribucionesde rango-tamano.Nuestroperspectivailustra casos modestosde urbanizacionen
terminosde estructuraen lugarde una simple escala reducida,y evita la tendenciade categorizarestos casos como
derivativos.
based on the "increasingly substantial proportion of the population of a settlement system [that] came either to live in a central
place or to be involved in a variety of ways
in the activities of a central place" (Clarke
1979:436). A substantial literature focuses on
the diverse forms and functions of pre-industrial cities (e.g., Sanders and Webster 1988;
Wheatley 1971; Adams 1966). However, we
argue that urban studies are most compelling
he title of Robert Adams's book Heartland of Cities succinctly captures a theme
that unifies many of the most influential analthe evoyses of early civilizations:
lution of urbanism. Urbanized societies featured city centers that were differentiated
from, but integrated with, their rural peripheries (e.g., Redman 1978:215-216).
The
closely related process of "urbanization" gave
rise to urban economic and political primacy,
T
Steven E. Falconer * Department of Anthropology, Arizona State University, Tempe, AZ 85287-2402
Stephen H. Savage * Department of Anthropology, University of Wisconsin, Milwaukee, WI 53201
American Antiquity, 60(1), 1995, pp. 37-58.
Copyright ? 1995 by the Society for American Archaeology
37
38
AMERICANANTIQUITY
when they comprehend entire networks of
sedentary settlement and, in so doing, distinguish different trajectories whereby cities,
towns, villages, and hamlets became incorporated or disarticulated as regional systems
coalesced or broke down (e.g., Adams 1981).
Nonsedentary populations tend to leave more
ephemeral archaeological signatures that often elude effective recovery and dating by
extensive regional surveys. Regrettably, this
aspect of society can only be addressed tangentially (e.g., regarding Early Bronze IV pastoralism in the Levant) with the data we assemble here.
This study offers a comparative perspective on the initial urbanization of lowland
Mesopotamia and the southern Levant, two
regions characterized by the early appearance
of cities and broad regional survey coverage.
Using revised methods of rank-size analysis
based on Monte Carlo simulation, we highlight a variety of urbanized settlement systems in the "heartlands" and "hinterlands"
of both regions. Our approach not only highlights fundamental distinctions between the
courses of Mesopotamian and Levantine urbanization, but also striking chronological and
geographical variation within each region.
Our results reveal further that the ordering
of rural sites often serves to distinguish the
rank-size distribution of one settlement system from another. Thus, when attempting to
capture urbanized systems as whole entities,
small places, as well as central ones, may serve
as key defining elements.
The diversity between and within Mesopotamian and Levantine settlement systems
calls for a renewed body of interpretive paradigms with which the full panorama of preindustrial urbanism may be explored. As a
case in point, the spectacular growth of metropolitan Uruk in the fourth and third millennia B.C. may provide the classic textbook
example of urban nucleation, but it does not
necessarily prefigure all courses of urbanization in Mesopotamia, let alone other regions
of the Near East. Our approach is particularly
valuable for interpreting less ostentatious expressions of urbanism in terms of structure
[Vol. 60, No. 1, 1995
and development, rather than simply reduced scale.
We begin with a discussion of rank-size
analysis, followed by a proposal for how it
might be adapted to the special nature of archaeological sampling and survey data. Our
rank-size analyses corroborate Adams's
(1981) portrait of southern Mesopotamia as
a centrally important urban "heartland" in
which the earliest cities, striking for their impressive size, are accompanied by the decimation of surrounding villages as they grow.
In contrast, the Diyala Plain constituted a
Mesopotamian "hinterland," distinct and well
removed from the Uruk heartland, that was
characterized by a dwindling array of small
towns and cities situated amid proliferating
rural settlement. Interestingly, Bronze Age
settlement in the southern Levant emerges as
neither an urban heartland nor a rural hinterland. Rather, a patchwork of urban and
rural systems followed variable trajectories
at different times and in different subregions.
Urbanism was primarily a coastal phenomenon apparently superimposed on a resilient
network of rural settlement. Systems of small
towns and villages in the Levantine hill country and Jordan Valley followed their own
courses of development that generally cannot
be attributed to the influences of waxing and
waning coastal urbanism.
These alternative expressions of urbanization heighten our appreciation of early urban
diversity and signal a need for renewed attention to the significance of small communities in stratified settlement systems. Analytical methods, such as those applied here,
that are tailored for archaeology and accommodate a full spectrum of settlement types
will inevitably enhance our insight on urbanism, and other issues of social complexity, in
southwestern Asia and elsewhere.
Rank-Size Analysis
Auerbach (1913) originally observed that the
cities of modern industrial nations, when
ranked according to their populations, are
distributed such that the largest city has twice
Falconer and Savage]
EARLYURBANIZATION
INMESOPOTAMIA
ANDTHESOUTHERN
LEVANT
39
1000
1000
10
Rank
100
x\
,N
----__-------I
.W7
.,.
oo
10
1
1
b
10
10
100
Rank
Figure 1. Examples of primate and convex rank-size distributions (a), and a primo-convex distribution created by
combining the two distributions in a (b).
the population of the second-ranked city,
three times the population of the third-ranked,
and so on. Following this "rank-size rule,"
the size of any nth-ranked place is predicted
by dividing the size of the largest place by n,
and the rank and population of cities describe
a log-normal distribution when plotted logarithmically (Haggett 1971:101; see Figure
1A).
Applying the Principle of Least Effort, Zipf
(1949) invoked "Economic Man" to explain
the interaction of two opposite courses of
economic action that create the rank-size rule.
The "Force of Diversification" encourages a
large number of "small, widely scattered and
largely autarchical communities" located near
raw material sources, while the "Force of
Unification" moves raw materials to a very
limited number of massive centers of production and consumption (1949:352).
Conventional applications in archaeology
assume that the forces of diversification and
unification are equal, which provides an expedient resolution to the dilemma of assigning values to Zipfs two forces. However, the
assumption also renders the formula used by
archaeologists a special case of the general
equation that is open to critical discussion
(Kowalewski 1982; Richardson 1973; Dziewonski 1972; Moore 1959).
Interpreting Rank-Size Curves
Log-normal distributions in accordance with
the rank-size rule "appear to be typical of
larger countries with a long tradition of urbanization, which are politically and economically complex" (Berry 1961:582). Archaeologists infer that such distributions
signify regional systems in which cities are
40
AMERICANANTIQUITY
well integrated with their subordinate communities (e.g., Adams 1981:72-74; Johnson
1980). Observation of "log-normality" in
western industrialized nations (e.g., Vining
1955:Figure 1) inspired Auerbach's original
formulation, but pre-industrial settlement
patterns tend not to conform to the values
expected under the special case of the ranksize rule applied in archaeology. Therefore,
most archaeological inferences are derived
from the manner and degree to which ranksize distributions depart from log-normal.
These departures may be classified into "primate," "convex," and "primo-convex" forms
(Johnson 1977; Paynter 1983), which potentially indicate various expressions of strong
or weak integration between large and small
communities.
Primate Distributions
Primate distributions are generated by settlement systems that contain fewer intermediate and large places than predicted by
the rank-size rule, or in which the first-ranked
place is considerably larger than expected
(Figure 1A). Typical primate patterns, which
are somewhat concave (see Johnson 1977),
may indicate an extraordinary centralization
of political or economic functions, as exemplified by several long-lived Mesoamerican systems in which "the primate center
provided unique services having to do with
maintenance of a regional boundary for a set
of local subsystems. The primate center's
special activities often involved a combination of high order sacred ceremonialism,
macroregional elite exchange, foreign diplomacy, and war" (Kowalewski 1982:65).
Primate distributions may also result from
the constraints of settlement system boundaries. Johnson, following Smith (1976) and
Blanton (1976), notes that the primate condition is associated frequently with centers or
peripheries of former colonial empires, "in
systems which are sufficiently bounded so as
to inhibit the development of more than one
highest order central place" (1977:496-497).
Johnson also cautions that "problems in sys-
[Vol. 60, No. 1, 1995
tem boundary definition. . . may produce essentially artificial primate distributions in
both archaeological and modern data sets"
(1977:498). This may be particularly true
when the entire extent of a settlement system
has not been identified. Therefore, archaeological interpretations should consider the
possibility that there may be "a role for the
primate city that extends beyond its regional
hinterland" (Skinner 1977:238).
Convex Distributions
A convex distribution contains more intermediate and large places than predicted by
the rank-size rule (Figure 1A). "In these cases
large settlements are smaller or small settlements are larger than expected" (Johnson
1977:497). In contrast to many primate systems, a convex distribution indicates relatively little integration of political and economic services among communities in a
settlement system, particularly less "vertical" integration between large cities and
smaller rural communities (Johnson 1980;
Paynter 1982). For example, data from the
Levantine Central Hills reveal pronounced
rank-size convexity that implies minimal articulation between dispersed Bronze Age villages (see discussion below).
Johnson (1980) suggests further that as they
become increasingly integrated, settlement
systems will shift from convex to log-normal
to primate distributions. Adams notes this
sequence in the combined Warka and Nippur-Adab survey data, and our analyses suggest another example in the upper end of the
rank-size curves for the Levantine Coastal
Plain (see discussion below). Alternatively, a
convex pattern may result from pooling two
or more adjacent settlement systems, or from
the exclusion of a primate center from its
subordinate settlement system (Johnson
1980; Paynter 1983).
In yet another twist to rank-size interpretation, some convex distributions may reflect
central place economic organization (Johnson 1977; Crumley 1976). Central Place Theory (e.g., Christaller 1933) predicts that plac-
Falconer and Savage]
EARLYURBANIZATION
INMESOPOTAMIA
ANDTHESOUTHERN
LEVANT
es of equivalent economic function will be
equivalent in size, resulting in a stepwise
ranking, rather than the more continuous distribution predicted by the rank-size rule. Such
a stair-step distribution is inherently convex,
especially when a system has multiple highest-order central places.
Primo-convex Distributions
Primo-convex distributions incorporate elements of rank-size primacy in their uppersize range and convexity in their lower range.
For example, the primo-convex curve in Figure 1B was created by combining the primate
and convex curves in Figure 1A. Primo-convex distributions may represent a special expression of pooling: the superimposition of a
centralized or colonially derived system (expressed in a primate upper curve) on a lowerlevel system that may be loosely integrated
or have an element of central place organization (reflected in a convex lower curve).
This possibility is particularly intriguing, since
it suggests the simultaneous operation of two
distinct settlement systems in a single region.
Toward an Accommodation of Rank-Size
Analysis and Archaeological Data
Although rank-size methods are used commonly in settlement pattern analysis, archaeological interpretations often rely simply on
judgmental appraisals of the shapes of ranksize distributions. This approach sidesteps the
issue of how far a distribution must depart
from log-normal to merit interpretation as
primate or convex. Attempts at greater statistical rigor use the Kolmogorov-Smimov
test (hereafter
one-tailed-goodness-of-fit
called the "K- test") to determine whether an
observed rank-size plot is significantly different from an expected one (Sokal and Rohlf
1969; Paynter 1982, 1983). The K- test is a
distribution-free test of the null hypothesis
that a sample is drawn from a particular population. "The Kolmogorov-Smirnov test rejects the null hypothesis if there are differences in the central tendency, range, or shape
41
of the sample and population distributions,
thus making it a very general test of nonidentity" (Paynter 1982:156). As it is used in
rank-size analysis, the K- test measures the
maximum deviation between cumulative
distributions of observed and expected site
sizes (Shennan 1990:55-61; Thomas 1986:
322-337). The maximum deviation (the Kstatistic) is compared to a predetermined value for a given alpha level in a statistical table
(e.g., Thomas 1986:504-506). If the deviation exceeds this value, the observed distribution is considered to be significantly different from the expected one.
Problems with Rank-Size Analysis of
Archaeological Data
There are several fundamental considerations that affect the applicability of rank-size
analysis to archaeological data. First, the Ktest may not be appropriate for archaeological analyses for at least four reasons:
1. The K- test assumes that the individual
observed and expected values are independent of each other. However, none of the
expected values are selected independently,
since they follow directly from the size of the
largest observed site as prescribed by the ranksize rule.
2. Although the test is designed for continuous frequency distributions, the expected
values assume a step-wise distribution, again
in accordance with the rank-size rule.
3. The observed frequency distributions are
assumed to be drawn from larger populations
with replacement. However, they are clearly
not drawn with replacement, since we count
each site only once.
4. Both frequency distributions are assumed to have no upper bound. This is true
of the observed data set, but the expected
distribution is bounded at its upper end by
the size of the largest observed site.
These characteristics of archaeological
rank-size data call traditional archaeological
applications of the K- test into serious question. While the procedure for calculating the
maximum deviation between observed and
42
AMERICANANTIQUITY
expected distributions remains acceptable, we
argue that appropriate measures of statistical
confidence for archaeological data must be
derived empirically.
The calculation of appropriate confidence
levels follows directly from a second major
consideration underlying archaeological ranksize analyses: archaeological site distributions are samples drawn from larger populations according to uncertain sample proportions that can only be estimated. Previous
tests of significance do not account for this.
Since the shape of any observed distribution
is affected by its sample proportion (based on
survey coverage and intensity), quantitative
analysis must incorporate estimated site recovery rates to produce meaningful results.
Thus, the issue of sampling cannot be separated from the issue of statistical confidence.
Additional issues stem from difficulties inherent in sensing archaeological sites and estimating their sizes. Archaeological site sizes
may be over- or under-estimated, particularly at deeply stratified sites or those subject to
alluviation. The area of a stratified site often
reflects the extent of its largest occupation,
which may inflate our estimates of smaller
habitations in other periods. In contrast, alluvial or colluvial blanketing of site fringes
in some regions (e.g., lower Mesopotamia and
the Jordan Valley) may cause underestimation of occupations at or below modern surface levels.
Further, the effects of such systematic errors on the analysis of any observed ranksize distribution will be particularly acute toward its lower end, since this is precisely where
the majority of sites in an expected log-normal distribution are located and where the
likelihood is greatest that sites in the target
population have been missed or obliterated.
Applying the K- Test Through Monte Carlo
Simulation
Given the limitations of the traditional Ktest, our analyses utilize the RankSize computer program (Savage 1993)1, which applies
[Vol. 60, No. 1, 1995
Monte Carlo simulation methods and accommodates the special characteristics of archaeological data and sampling. This program analyzes a set of observed site sizes based
on an estimate of the sample proportion it
represents. The simulation first sorts the observed data and calculates a K- statistic (that
is, the maximum deviation between the expected and observed values), keeping it for
later reference. The program then creates a
simulated population based on the rank-size
rule starting from the largest site size in the
observed data. The number of sites in this
simulated population is determined by multiplying the number of observed sites by the
reciprocal of the sample fraction. Thus, if a
survey yields a sample of 100 sites that are
assumed to represent 90 percent of the population, that population contains 100 / .90 =
111 sites. The largest site in the population
equals the largest site in the sample. The remaining 110 sites are generated according to
the rank-size rule. Clearly, this method is only
as accurate as the estimate of the sample proportion, but it forces the analyst to confront
the sampling issue explicitly.
The simulated population becomes the basis for a long series of random computer runs.
Each run draws a sample of sites from the
simulated population equal to the number of
observed sites. (The user may always include
the largest observed site size or allow it to
vary; all of our analyses include it.) The simulation then calculates a K- statistic that measures the maximum deviation between this
sample and its own expected, log-normal distribution. The program repeats this procedure over a large number of runs, stores the
K- statistics produced by each run, and sorts
them in ascending order. Finally, the simulation compares the value of the original Kstatistic for the observed data to the range of
K- values from the random runs to estimate
the probability that a K- value greater than
or equal to the observed value will be obtained by random samples drawn from a lognormal population. For example, if only 3
percent of the simulated K- values are as large
or larger than the observed K- value, we may
Falconer and Savage]
EARLYURBANIZATION
INMESOPOTAMIA
ANDTHESOUTHERN
LEVANT
infer that it is unlikely that the observed sample is drawn from a log-normal population.
Whereas a traditional test categorizes an
observed K- statistic as significant or nonsignificant at a predetermined alpha level (typically .05), we prefer to estimate the probability that an observed distribution could have
been drawn from a log-normal population.
The lower the probability, the greater our
confidence that the original settlement system may be interpreted as differing from lognormal (primate, convex, or a combination
of the two). The analyst may deem some
probability estimates as highly diagnostic and
others as more equivocal. Rather than a simple pass-fail result, this approach provides a
variety of possible outcomes and interpretive
avenues (see Cowgill 1977).
The potential for misestimating observed
site areas may be accommodated by randomly inflating or diminishing the sizes of sites
picked by the RankSize program for simulated distributions. This might be done within a preset percentage range, for a predetermined proportion of sites, for example. To
compensate for the likelihood of missing more
sites at the lower end of an observed distribution, it may be advantageous to incorporate a "sliding probability" of site recovery
ranging from near certainty for extremely large
cities to very modest levels for diminutive
villages and hamlets. The addition of these
two routines to the RankSize procedure would
raise a variety of questions regarding how
these potential sources of sample bias should
be measured and simulated. While we do not
attempt to address these concerns in detail
here, experimental results indicate that the
introduction of these two sources of variability into the RankSize program tends to
result in modestly higher probability figures.
Therefore, the probabilities discussed below
may tend to slightly exaggerate the departure
of observed rank-size distributions from lognormal.
We apply RankSize simulations to site size
data, revealing a variety of trajectories whereby early urbanized settlement systems arose
in Mesopotamia and the southern Levant.
43
Each analysis poses two questions: 1) How
low is the probability that the observed distribution could be drawn from a log-normal
population? 2) In light of this probability and
the observed rank-size plot, what is the shape
of the observed distribution (and the likely
shape of the original population)? In each case,
we assume that the largest site reported was
the largest settlement in the original system
under study. We introduce the settlement data
from each region with an assessment of survey methods and site recovery rates from
which we estimate an average sample proportion. The number of settlements in each
original population is calculated by multiplying the number of observed sites by the
reciprocal of the sample proportion. When a
probability value is very low, we argue that
our sample represents a noteworthy departure from the log-normal distribution predicted by the rank-size rule, and we address
the archaeological implications of that departure.
Interestingly, the settlement patterns analyzed in this study generate very few equivocal probability figures. The majority of observed distributions in both Mesopotamia and
the southern Levant have probabilities of
<.01 of simply representing samples of a
larger log-normal population. The remaining
cases, with two Mesopotamian exceptions,
provide substantially larger values (ranging
between .24 and .93) more clearly indicative
of effective adherence to the rank-size rule.
Our approach begins to accommodate the
uncertainty inherent in archaeological field
and analytical procedures and illustrates the
benefits of larger sample proportions. Simulated K- statistics rarely exceed observed
values when the observed data represent a
high sample proportion, rather than a low
one. For example, when analyzed at sample
proportions ranging between .05 and .95,
Early Bronze III settlement data from the
Levantine Coastal Plain show that site recovery rates must be greater than 75 percent
to generate an extremely low probability that
an observed sample was drawn from a lognormal population (Figure 2). This obser-
44
AMERICANANTIQUITY
[Vol. 60, No. 1, 1995
0
IM/m,
X
,.6
C
.
Cis
Ie
0).2 ~
A
n-~~~~I/:
5 10 15 20 25 30 35 40 50 55 60 65 70 75 in8085 90 95
Sample Percent
vation reinforces the call for full coverage
survey (e.g., Fish and Kowalewski 1990) and
careful consideration of the factors that determine site survivorship and recovery.
Mesopotamian Survey Chronology and
Methods
Our analyses of urbanization in Mesopotamia draw data from the Warka, Nippur-Adab,
and Diyala surveys conducted by Adams
(1981; 1965; Adams and Nissen 1972) (Figure 3). All three survey areas lie beyond the
frontiers of modern cultivation and present
only the "low, featureless relief' of sparse
vegetation, "long disused canal levees, and
the rubble strewn mounds of former settlements" (Adams 1981:xvii). The Warka and
Nippur-Adab surveys encompass the heartland of early Mesopotamian city and state
development: the central floodplain of the
Euphrates River. The Diyala Survey assesses
early canal systems and agrarian settlement
in the lower Diyala River drainage, a region
that constituted a rather detached hinterland
of the urban developments farther to the
south. We concentrate on the explosion of
Mesopotamian urbanism centered on Uruk
during the Early-Middle Uruk, Late Uruk,
and Early Dynastic I periods (following Ad-
Figure 2. Probability estimates
at increasing sample proportions
for Early Bronze III data from
the Levantine Coastal Plain.
ams 1981; see Table 1). We omit data from
the enigmatic phenomenon known as "Jemdet Nasr," which may have been a regional
ceramic style or an archaeologically elusive
time period (Adams 1981:81; Finkbeiner and
Rollig 1986).2
Large-scale archaeological reconnaissance
in Mesopotamia has relied on visits to sites
identified from maps or aerial photographs,
as well as vehicular reconnaissance along parallel transects (usually 0.5-1.0 km apart) or
across river levees (e.g., Adams 1965:120;
1981:38-39; Wright 1981:298). Perhaps the
major natural impediment to site recovery in
Mesopotamia is alluviation, which may reduce the apparent size of a site, or obscure it
altogether, particularly in cases of small sites
from earlier periods (e.g., Stronach 1961; Adams 1975). As a formal test of the efficiency
of jeep reconnaissance, Adams (1981:40-42)
restudied 13 sq km north and east of ancient
Nippur in which nine sites were identified
originally. More intensive coverage revealed
three additional sites and required modification of the descriptions or dating of four
others. These results suggest that settlement
inventories for the Nippur-Adab Survey and
others like it may be deficient by one-quarter
to "as much as one-third" (Adams 1981:42).
This deficiency tends to underestimate site
Falconer and Savage]
INMESOPOTAMIA
EARLYURBANIZATION
ANDTHESOUTHERN
LEVANT
45
Figure 3. Mesopotamian survey
areas. Approximate coverages:
Diyala Survey = 8,000 km2; Nippur-Adab Survey = 3,450 km2;
Warka Survey = 2,800 km2.
Adapted from Adams (1965:119,
1981:42), Adams and Nissen
(1972:4).
counts at the lower end of rank-size curves.
On the basis of Adams's systematic restudy,
a rarity in Near Eastern survey archaeology,
we estimate an average recovery rate of 70
percent, which we apply as the sample proportion for our analyses of Mesopotamian
settlement.3
Hyperurbanization in the Mesopotamian
Heartland
The Uruk and Early Dynastic I periods reveal
a pattern of population growth in a dwindling
number of communities in the combined
regions of the Warka and Nippur-Adab surveys. Ancient Uruk achieved primate status
through a process of urban "agglomeration"
in which it absorbed much of the rural population immediately surrounding it, thereby
swelling to 400 ha and "no less than 40,000
to 50,000" inhabitants by Early Dynastic I
(Adams 1981:85; Adams and Nissen 1972:
19-21). The rank-size plots in Figure 4 follow
Johnson's (1980) prescription for the development of such a primate system.4 To begin
the sequence, our simulation of Early-Middle Uruk settlement generates a probability
of only .03 that the observed distribution represents a random departure from log-normal
(Table 2). The Early-Middle Uruk rank-size
plot shows this distribution to be convex.
The Late Uruk rank-size distribution,
though also convex, has a much higher probability (p = .71) of representing a random
departure from log-normal. Therefore, this
settlement pattern is more justifiably interpreted as adhering to the rank-size rule, rather than deviating substantially from it. In this
case, very different probability estimates result from rather similar data sets primarily
46
AMERICANANTIQUITY
[Vol. 60, No. 1, 1995
Table 1. Mesopotamian Chronology.
Ending Date
Beginning Date
Period
ca.
ca.
ca.
ca.
Early Dynastic I
Jemdet Nasr
Late Uruk
Early-Middle Uruk
2900
3100
3500
4000
ca.
ca.
ca.
ca.
B.C.
B.C.
B.C.
B.C.
2600
2900
3100
3500
B.C.
B.C.
B.C.
B.C.
Note: Based on historical correlations and recalibrated radiocarbon dates (see Porada et al. 1992: Figures 3, 4;
Adams 1981:81-82).
due to an increase in the size of the largest
site (Warka) in the Late Uruk Period.
Data from the subsequent Early Dynastic
I Period generate a very low probability (p <
.01) that the primate rank-size curve in Figure 4 is a product of chance. Indeed, we estimate that by Early Dynastic I more than 60
percent of the sedentary populations around
Warka and Nippur lived in a total of only
five communities, each larger than 40 ha
(Falconer 1987:Figure 5; Adams 1981:Table
7).
Although our rank-size plots do not provide highly convex patterns suggestive of
pooling, separate analyses of northern and
southern settlement "enclaves" (following
Adams 1981:70-90) reveal two distinct, but
related routes by which urbanization developed more locally. In the Early-Middle Uruk
Period, the Nippur-Adab enclave to the north
was more heavily populated, particularly with
village farmers (Adams 1981:70; Falconer
1987:Figure 7). The growth of Uruk in the
subsequent Late Uruk Period (from 70 to 100
ha) was accompanied by an apparent shift in
rural settlement from the north to the south.
Adams suggests that "if we take into account
the artificial limitations of the survey area
... literally tens of thousands of small villagers appear to have abandoned their homes
and moved southward" (1981:70). This transition is manifested in rank-size curves that
develop in very different manners.
The Nippur-Adab data produce probabil-
Table 2. Summary of Simulation Runs for Mesopotamian Survey Regions.
Region
Period
Number
of Sites
Largest Number Sample in Popof Sites Percent lationb
Sitea
Observed
K-Value
Probabilityc
Curve
Shape
<.01
.71
Primate
L-norm
Warka/Nippur-Adab
Early Dynastic I
Late Uruk
Early-Middle
Uruk
400.0
100.0
98
156
70
70
140
223
.643
.288
70.0
203
70
290
.335
.03
Convex
Nippur-Adab
Early Dynastic I
Late Uruk
Early-Middle
Uruk
50.0
50.0
34
49
70
70
49
70
.441
.408
<.01
<.01
Convex
Convex
50.0
144
70
206
.389
<.01
Convex
Warka
Early Dynastic I
Late Uruk
Early-Middle
Uruk
400.0
100.0
64
107
70
70
91
153
.688
.336
<.01
.04
Primate
Primate
70.0
59
70
84
.492
<.01
Primate
Diyala
Early Dynastic I
Early-Late Uruk
36.0
36.0
37
21
70
70
53
30
.243
.286
.93
.68
L-norm
L-norm
a Size in hectares.
b
Number of sites in population = number of observed sites x (1/sample percent).
c Probability of drawing a K- value greater than or equal to the observed value at random from a log-normal
population, based on 1,000 random runs for each row.
Falconer and Savage]
INMESOPOTAMIA
ANDTHESOUTHERN
LEVANT
EARLYURBANIZATION
47
_
N
(D
v:5-
CT
t^
Rank
Rank
Figure 4. Rank-size distributions for Early-Middle
Uruk and Early Dynastic I settlement within the combined areas of the Warka and Nippur-Adab surveys. Data
from Adams (1981:Table 7).
Figure 5. Separate Early Dynastic I rank-size distributions for the Warka and Nippur-Adab survey areas.
Data from Adams (1981:Table 7).
ity estimates consistently less than .01, with
rising K- statistics, and rank-size distributions that denote increasing convexity (see
Adams 1981:Figure 17). This convexity becomes accentuated as the system loses smaller settlements. The Warka Survey data likewise produce very low probability values and
rank-size plots that start as distinctly primate, then approach log-normal in the Late
Uruk Period due to the addition of mediumand small-sized settlements (Adams 1981:
Figure 17). In Early Dynastic I, while the
Nippur-Adab enclave continued to lose villages, the city of Uruk quadrupled in size,
providing a rank-size distribution that once
again is resoundingly primate (Figure 5).
In general, these trends portray a society
that became highly urbanized by relocating
and reducing its rural population. In this case
the process had two related facets. A growing
southern populace "agglomerated" in the
quintessentially primate center of Uruk, while
settlement to the north became increasingly
urbanized only in a residual sense, as the hierarchy of towns and villages below Nippur
and Adab withered. Interestingly, both patterns strongly imply decreased agrarian productivity and contradict the expectation that
urban authorities will encourage rural farming communities to ensure a strong agricultural base and their own self-preservation
(Adams 1981:88). Only late in the third millennium B.C. (i.e., the Early Dynastic II
through Isin-Larsa periods), do cities encourage an abundance of surrounding villages in a process of "ruralization" (Yoffee
1986; Robinson 1972) more in keeping with
this expectation (Adams 1981:130-170; Falconer 1987:122-123). As we shall see, this
process of simultaneous urban aggregation
and rural depopulation describes only one,
counterintuitive and potentially maladaptive, expression of urbanization that does not
typify the development of early urbanism in
other time periods or regions of southwestern
Asia.
Small-scale Urbanism in the Diyala
Hinterland
The long-term trajectory for settlement on
the Diyala Plain, which stands as a counterpoint to that of the Warka and Nippur-Adab
regions, starts with the appearance of relatively modest cities in the Uruk Period and
charts their decline and disappearance by ca.
1000 B.C. (Falconer 1987:128-133; Adams
1965:36-42). The Diyala's sparse regional
population became overwhelmingly and increasingly rural over these millennia. By Early Dynastic I, 10 towns and cities measured
10 ha or larger (including Tell Asmar and
Khafajah; see Figure 3). More impressively,
AMERICANANTIQUITY
48
---Uruk observed
------ EDI observed
(both)
Expected
g
.N
c
100
Rank
Figure 6. Rank-size distributions for settlement within
the Diyala Survey area. Data from Adams (1981:Figures
8, 10; 1965:Table 10, Figure 2, Appendix C, map sections
1B-4B).
the frequency of villages (less than or equal
to 4 ha) jumped from 71 to 90 percent (Falconer 1987:Figure 12). These data would suggest a pattern of "ruralization" like that
around Uruk following the Early Dynastic
Period. However, despite apparently convex
rank-size plots (Figure 6), very high probability estimates reveal that the observed Uruk
and Early Dynastic I settlement distributions
represent merely chance departures from lognormal (Table 2). This approximation of lognormal contrasts fundamentally with the developing primacy of Uruk's settlement system. Adams observes that the emergence of
fortified towns in the Diyala found "little apparent reflection in the disposition of the remaining, smaller settlements over the countryside" (1965:38). Like their much larger
counterparts to the south, the Diyala's towns
and cities emerged amid an array of potentially subordinate villages, but apparently
failed to exert a comparable molding influence on surrounding settlement patterns.
While metropolitan behemoths came to
dominate the Mesopotamian heartland, settlement in the Diyala hinterland emerged as
an alternative expression of urbanism based
on a log-normal integration of small cities,
towns, and proliferating villages.
Lower Mesopotamia and the Diyala do not
exhaust the range of early urban expressions
[Vol. 60, No. 1, 1995
in the Near East. However, they do bear witness to two classic patterns by which urbanism might first appear: through the nucleation of a predominating primate city or the
coalescence of a stratified, but less centralized, settlement system on a much smaller
scale. By virtue of their early appearance and
massive scale (at least for Uruk) these patterns serve as touchstones for the interpretation of cities and regional settlement elsewhere in the Near East. The southern Levant
provides a striking case in point in which
conventional models of urbanism hinge explicitly or implicitly on reference to earlier,
larger-scale events elsewhere in southwestern
Asia, notably Mesopotamia or Syria.
Conventional Approaches to Levantine
Urbanism
A substantial and growing literature explores
the rise of Levantine cities and towns using
a variety of data and interpretive angles, but
sharing a set of common themes. With the
exception of a handful of recent works (e.g.,
Joffe 1991a; Finkelstein and Gophna 1993;
Gophna and Portugali 1988), urbanism is
treated as a regionwide phenomenon, with
little discussion of its variable manifestations
in the southern Levant's constituent settlement zones (Esse 1989; Dever 1987; Richard
1987). Secondly, urbanism is construed as
structurally equivalent to that found elsewhere; it simply occurred later, and on a reduced scale. For example, Kempinski (1978)
attributes the appearance of Early Bronze Age
cities to a process of nucleation analogous to
that around Uruk. Third, many authors propose that antecedent forms of Bronze Age
material culture, and the technology that produced it, are found elsewhere, notably in Egypt
and Syria (Ilan and Sebbane 1989; Kempinski 1989). Therefore, this material evidence
and an associated tradition of urbanism must
have spread into the southern Levant, presumably from earlier, exotic sources (Dever
1987; Gerstenblith 1983). Jointly, these suppositions hold that the first urbanized societies in the southern Levant were secondary,
derivative expressions of early urbanism. The
Falconer and Savage]
EARLYURBANIZATION
INMESOPOTAMIA
ANDTHESOUTHERN
LEVANT
49
southern Levant is described as a "backwater" in comparison to neighboring regions,
which nonetheless featured "truly urban" society during the Bronze Age (Dever 1976;
Kenyon 1973).
Our study reverses the conventional interpretive process applied to the southern Levant. Instead of emphasizing foreign relations and the effects of cities on the entire
southern Levant, we break the region into
constituent units, from which analytical results are assembled to reinterpret the mosaic
of Bronze Age urbanism.
Levantine Survey Chronology and Methods
Settlement data from the southern Levant
derive from a variety of surveys in Israel north
of the Negev Desert, the West Bank of the
Jordan River, and the Jordan Valley (e.g., see
Joffe 199 lb:Table 1, Figures 2-4). In contrast
to Mesopotamia, this region features significant variation in topographic relief, modem
vegetation, and human habitation. The
southern Levant can be divided longitudinally into a series of geomorphic zones (Figure 7; see Horowitz 1979:11-19). A coastal
plain up to 50 km wide extends in from the
Mediterranean shoreline. The landscape then
rises to form a north-south chain of hills
reaching 300 m asl. Major valleys separate
the northern hills of the Galilee from the central hills that extend through the West Bank.
The eastern slope of this hilly backbone
plummets rapidly to the Jordan Valley, where
the valley floor lies between 100 and 350 m
below sea level. Relatively sparse Mediterranean maquis and oak and pistachio remnant forests populate the Levantine hill country, while only vestiges of natural vegetation
exist amid the modem settlement and agriculture of the Levantine Coastal Plain and
Jordan Valley (Gophna et al. 1986; al-Eisawi
1985).
The success of early Levantine urbanism
oscillated dramatically through the Early and
Middle Bronze ages, which covered the third
and early second millennia B.C. (Table 3).
Early Bronze I provided a formative, villagelevel prelude to the fortified towns and cities
^^7
Negev Desert
Figure 7. Geomorphic/settlement zones in the southern
Levant. Hatched areas indicate survey subregions considered in this study. Approximate coverages: Total
southern Levant = 15,000 km2; Coastal Plain = 2,650
km2; Central Hills = 7,500 km2; Jordan Valley = 1,700
km2. Compiled from Broshi and Gophna (1984:Table 11),
Ibrahim et al. (1976:Figure 1, 1988:192-193).
that emerged in Early Bronze II and III (Joffe
1991 a). This urban florescence paralleled the
rise of the Egyptian state during Dynasties IVI (Kemp 1983:71-116; Ben-Tor 1991; Kantor 1992:17-21; Stager 1992:40-41). Subsequently, Egypt entered the First Intermediate
Period, during which centralized authority
collapsed and regional economic ties were attenuated (Ward 1971; Stager 1992:41). The
southern Levant simultaneously experienced
wholesale abandonment of cities in favor of
village life and nonsedentary pastoralism
during Early Bronze IV (Dever 1980; 1989)5.
AMERICANANTIQUITY
50
Table 3. Levantine Chronology.
Period
Middle Bronze
IIB-C
Middle Bronze
IIA
Early Bronze
IV
Early Bronze
III
Early Bronze
II
Early Bronze I
Beginning Date
Ending Date
ca. 1800 B.C.
ca. 1500 B.C.
ca. 2000 B.C.
ca. 1800 B.C.
ca. 2300 B.C.
ca. 2000 B.C.
ca. 2700 B.C.
ca. 2300 B.C.
ca. 3100 B.C.
ca. 3500 B.C.
ca. 2700 B.C.
ca. 3100 B.C.
Note: Based on historic correlations and recalibrated
radiocarbon dates (see Falconer 1993, Stager 1992:401, Joffe 1991b:Table 3, Richard 1987).
Unfortunately, published settlement data for
this striking period of collapse are available
only from the coastal plain.
The duration of the Middle Bronze Age
approximated that of Egypt's Middle Kingdom (Weinstein 1975). While Egypt reestablished state-level government and far-flung
economic influence, Levantine cities reemerged suddenly ca. 2000 B.C. in Middle
Bronze IIA. During Middle Bronze IIB and
C the region's cities and general population
grew to sizes unsurpassed until the Roman
and Byzantine periods (i.e., first-seventh centuries A.D.; Broshi 1979). The archaeological
distinctions between Middle Bronze IIB and
C are subtle and not universally recognized
(Falconer 1987:180-181), and settlement data
commonly are analyzed jointly, as we do here.
This study draws on data compiled by several authors for specific periods (Joffe 1991 b;
Broshi and Gophna 1984; 1986) or regions
(Finkelstein and Gophna 1993; Gophna and
Portugali 1988; Ibrahim et al. 1975, 1988) in
the southern Levant. Since this literature cites
a wide variety of local and regional surveys,
it offers few discussions of survey methods.
Levantine surveys typically collate data from
earlier surveys and maps, while new fieldwork combines purposive vehicular reconnaissance, intensive pedestrian tactics, and
questioning of local inhabitants (Joffe 1991 b:
34-44; Ibrahim et al. 1976:44). Most authors
presume virtually complete survey coverage
[Vol. 60, No. 1, 1995
(e.g., Finkelstein and Gophna 1993:2) and
judge the likelihood of finding additional
larger sites to be "practically nil" (Broshi and
Gophna 1986:88; 1984:50). Hence, as in
Mesopotamia, the vast majority of undetected sites are likely to have been small.
Despite the lack of an explicit study of survey efficiency like Adams's, we may approximate site recovery rates based on several
considerations. First, since Levantine surveys encompass areas of substantial modern
habitation and agriculture, these factors undoubtedly obscure small sites to a greater extent than in Mesopotamia. On the other hand,
Levantine alluviation, which does affect some
sites (e.g., Braun 1985; Rosen 1986:45), is
limited to the intermittent erosion of hill
slopes (Beaumont 1985), rather than the continuous massive silt transport of the Tigris
and Euphrates rivers. While natural factors
may affect site visibility more commonly in
Mesopotamia, cultural influences may have
a greater impact in the Levant. In sum, we
estimate Levantine site recovery rates on the
same order as those for Mesopotamia, but
nudge them slightly higher because of the predominance of intensive localized surveys that
rely more heavily on pedestrian tactics (Joffe
1991b:34-44), rather than macroscopic vehicular coverage.6 These considerations suggest an average site recovery rate on the order
of 75 percent, only slightly lower than that
guessed by Broshi and Gophna (1984:41,
1986:73, 88).
Rank-Size Analyses of Levantine Urbanism
In light of the preeminence normally ascribed
to fortified Bronze Age cities in the southern
Levant (e.g., Dever 1987; Richard 1987), we
might anticipate accordingly primate ranksize settlement distributions. When combined for the entire region, Bronze Age survey data do produce a series of probabilities
below .01 (Table 4). However, these values
reflect convex departures from log-normal, as
exemplified for the high points of Levantine
urbanism in Early Bronze III and Middle
Bronze IIB-C (Figure 8). This consistent pattern implies a series of poorly integrated re-
Falconer and Savage]
51
INMESOPOTAMIA
EARLYURBANIZATION
ANDTHESOUTHERN
LEVANT
gionwide systems or the pooling of many more
localized settlement networks. As we shall
see, these two interpretations are not mutually exclusive.
First, let us consider the possibility of
pooled systems before returning to arguments for integration and disintegration below. Finkelstein and Gophna (1988) note that
the Bronze Age ushered in the economic
"conquest" of the central hill country. Unlike
the coastal plain and Jordan Valley, which
were ideal for lowland agriculture, the central
hills were best suited for large-scale fruit
growing and summer pasturage. Thus, a corollary of Bronze Age urbanism was an emerging potential for significant economic specialization and pastoral interplay between
Table 4.
6
Nd
Rank
Figure 8. Rank-size distributions for Early Bronze III
and Middle Bronze II B-C settlement in the southern
Levant. Data from Joffe (1991b), Finkelstein and Gophna
(1993), Gophna and Portugali (1988), Broshi and Gophna
(1984; 1986), Ibrahim et al. (1976, 1988).
Summary of Simulation Runs for Southern Levant.
Sample
Percent
Number
of Sites
in
Populationc
Observed
KValue
Probabilityd
Curve
Shape
Largest
Sitea
Number
of Sitesb
Southern Levant Middle Bronze IIB-C
Middle Bronze IIA
Early Bronze III
Early Bronze II
Early Bronze I
80.0
80.0
30.0
40.0
40.0
219
96
202
151
243
75
75
75
75
75
292
128
269
201
324
.479
.573
.406
.377
.370
<.01
<.01
<.01
<.01
<.01
Convex
Convex
Convex
Convex
Convex
Coastal Plain
Middle Bronze IIB-C
Middle Bronze IIA
Early Bronze IV
Early Bronze III
Early Bronze II
Early Bronze I
64.0
65.0
5.0
25.0
25.0
25.0
59
51
16
12
24
42
75
75
75
75
75
75
79
68
24
16
32
56
.610
.745
.564
.250
.500
.595
<.01
<.01
<.01
.25
<.01
<.01
P-cnvxe
P-cnvxe
D-cnvxr
L-norm
Convex
Convex
Central Hills
Middle Bronze IIB-C
Middle Bronze IIA
Early Bronze III
Early Bronze II
Early Bronze I
15.0
12.0
11.0
15.5
15.0
91
9
49
43
81
75
75
75
75
75
121
12
65
57
108
.385
.444
.347
.465
.469
<.01
<.01
<.01
<.01
<.01
Convex
Convex
Convex
Convex
Convex
Jordan Valley
Middle Bronze IIB-C
Middle Bronze IIA
Early Bronze III
Early Bronze II
Early Bronze I
7.0
7.0
25.0
20.0
20.0
26
13
36
32
64
75
75
75
75
75
35
17
48
43
85
.308
.231
.444
.219
.219
.24
.29
<.01
.69
.78
L-norm
L-norm
Convex
L-norm
L-norm
Region
a
Period
Size in hectares.
Includes only sites with unequivocal dates of occupation and known size. Some counts are underestimated (e.g.,
MBIIA in the Central Hills).
c Number of sites in population = number of observed sites x (1/sample percent).
d
Probability of drawing a K- value greater than or equal to the observed value at random from a log-normal
population, based on 1,000 random runs for each row.
e P-cnvx = Primo-convex curve.
f D-cnvx = "Double convex" curve.
b
[Vol. 60, No. 1, 1995
AMERICANANTIQUITY
52
100 _.................
----
--
-_
EB I observed
EB ml observed
Expected (EB I & HI)
10CZ
._
c/
1_^^-~10
S U~:
100
Rank
Rank
Figure 9. Rank-size distributions for Early Bronze I
and Early Bronze III settlement on the Coastal Plain.
Data from Joffe (1991b), Gophna and Portugali (1988),
Broshi and Gophna (1984).
Figure 10. Rank-size distributions for Early Bronze IV
and Middle Bronze II B-C settlement on the Coastal
Plain. Data from Gophna and Portugali (1988), Broshi
and Gophna (1986).
highlands and lowlands (Finkelstein and
Gophna 1993). Therefore, it is not surprising
that Bronze Age society followed multiple
courses of development, which may be distinguished according to geomorphic zones of
settlement. We present analyses of settlement
in the coastal plain, the central hills, and the
Jordan Valley, which reveal these variable
trajectories, and collectively constitute the
early urbanization of the southern Levant.7
ized by stair steps (Figure 9). These unusual
distributions bear a striking resemblance to
primo-convex curves, but since their upper
portions are convex rather than concave, we
refer to their form as "double convex."
Through the Early Bronze sequence, the
lower curves become increasingly truncated.
The relatively high probability estimate for
Early Bronze III (p = .25) reflects a particularly drastic curtailment of the rural settlement that defined the lower curves of Early
Bronze I and II. This trend presages the subsequent abandonment of coastal sedentism
in Early Bronze IV. Coastal population growth
resumed on a grander scale in the Middle
Bronze Age (Gophna and Portugali 1988).
Very low probabilities (p<.0O) again reflect
compound rank-size distributions that depart substantially from log-normal (Figure
10). Unlike those of the Early Bronze Age,
these curves are primo-convex, and reflect
slightly increased numbers of medium- and
small-sized settlements.
Following the examples and general reasoning summarized earlier, we propose that
these primo-convex and "double convex"
rank-size distributions signify the superimposition of multiple, contemporaneous settlement systems. Coastal settlement during
the Early Bronze IV Period is particularly
noteworthy for interpreting these curves. A
The Coastal Plain
If there was a "heartland" of Levantine urbanism, it was the Mediterranean coastal
plain, which contained most Bronze Age
communities 10 ha and larger (Falconer
1994). Among the best-known sites in the
coastal plain are Tell Dor, Aphek, Tell Gezer,
Tell el-'Areini, and Ashkelon (see Figure 7).
During the Early Bronze Age, a growing
coastal population became increasingly concentrated in larger towns and cities as rural
settlement dwindled (Gophna and Portugali
1993). Coastal settlement data produce probabilities below .01 for Early Bronze I and II,
and a more equivocal estimate of.25 for Early Bronze III. Unlike any pattern found in
Mesopotamia, these data describe slightly
convex upper rank-size curves that join more
pronounced lower convex curves character-
Falconer and Savage]
EARLYURBANIZATION
INMESOPOTAMIA
ANDTHESOUTHERN
LEVANT
modest array of very small towns (two measure 5 ha each) and diminutive villages (none
larger than 1 ha) comprised a "decapitated"
settlement system that followed the abandonment of larger towns and cities. This convex distribution has a very low probability
(p<.01) of being drawn from log-normal
population. Since the Early Bronze IV ranksize distribution results from the elimination
of an upper settlement curve, this convex
curve is comparable to the lower curves of
Early Bronze I-III. This pattern provides circumstantial evidence that each compound
curve of Early Bronze I-III may indeed be
composed of an upper distribution of towns
and cities superimposed on a relatively discrete lower curve of villages and hamlets.
In keeping with this interpretation, our data
suggest that an Early Bronze Age network of
loosely knit towns and cities was established
amid a dwindling and increasingly disarticulated system of small villages. Statuary excavated from coastal sites and texts from
Egypt suggest that this patterning reflects the
presence of Egyptian commercial missions in
coastal towns (Ahituv 1978; Na'aman 1981).
While also "compound," the Middle Bronze
rank-size plots differ in three respects: the
more concave upper curves suggest more centralized integration of towns and cities, the
more extended lower limbs reflect more pronounced ruralism, and the upper and lower
curves remain more closely articulated.
We must conclude that despite containing
most of the Levant's cities, the coastal plain
was neither a heartland of urban nucleation,
nor a hinterland of lower level urban-rural
integration. Early Bronze coastal cities remained relatively static in size and number
(Joffe 199 lb:Table 23), and their lack of ranksize primacy suggests that they failed to exert
the influence of centers like Uruk, even on a
reduced scale. During the Middle Bronze Age,
both urban and rural populations grew, but
only slightly (Falconer 1994). An element of
Middle Bronze Age rank-size primacy, which
became more pronounced in MB IIB-C, suggests that MB cities could have had a molding
influence, but exercised it only to a limited
53
0)
rC
10
Rank
Figure 11. Rank-size distributions for Early Bronze II
and Middle Bronze II B-C settlement in the Central Hills.
Data from Finkelstein and Gophna (1993), Joffe (1991b),
Broshi and Gophna (1984; 1986).
extent. This portrait of coastal urbanism may
be clarified with reference to the Levantine
"hinterlands" of the central hills and Jordan
Valley.
The Central Hills
With the emergence of highland/lowland economic interaction (see above), we might predict that coastal settlement systems may have
incorporated rural communities in the adjoining central hills. Indeed, Early and Middle Bronze settlement data produce consistently convex rank-size distributions (p < .01;
Figure 11), as we would expect if villages in
the hills supplemented coastal ruralism. Other analyses verify that hill country villages
became more abundant between Early Bronze
I and II (Joffe 1991 b:Tables 20, 21), and proliferated dramatically in the Middle Bronze
Age (Broshi and Gophna 1986; Falconer
1994). However, these communities declined
drastically during the first urban climax of
Early Bronze III as part of a general population drop in the hill country (Finkelstein
and Gophna 1993). So, during Early Bronze
I and II and the Middle Bronze Age, some
settlement in the central hills may have compensated for the otherwise under-represented
ruralism of the coastal plain. In contrast, Early Bronze III represents a nadir in rural set-
AMERICANANTIQUITY
54
[Vol. 60, No. 1, 1995
Gophna 1986; Falconer 1994). Settlement
data produce relatively high probabilities that
the observed convex distributions are simply
samples drawn from a log-normal population. It appears that after the non-urban interlude of Early Bronze IV, local inhabitants
dispersed into a renewed hinterland network
of towns and villages, which was again roughly analogous to that of the Diyala.
a
.N
u
Summary of Levantine Urbanism
Rank
Figure 12. Rank-size distributions for Early Bronze I,
Early Bronze III, and Middle Bronze II B-C settlement
in the Jordan Valley. Data from Joffe (1991b), Broshi
and Gophna (1984; 1986), Ibrahim et al. (1976, 1988).
tlement contemporaneous with that of the
coastal plain.
The Jordan Valley
Among the settlement zones within the
southern Levant, only the Jordan Valley displays patterns of development that resemble
any of those found in Mesopotamia. Much
as we saw for the Diyala, very high Early
Bronze I and II probability estimates suggest
a similar pattern of close adherence to lognormal and low-level hinterland integration
(Figure 12). However, unlike the Diyala, the
Jordan Valley experienced a drop in population and settlement frequency (Joffe 1991 b:
Table 18). While valleywide population leveled off in Early Bronze III, the settlement
system became significantly less integrated as
suggested by its convex rank-size distribution
(p<.01). In this case, rather similar Early
Bronze II and III rank-size curves generate
very different simulation results, primarily
because the lower end of the Early Bronze III
curve consists of smaller sites, which imply
a shrinking population of rural farmers.
During the Middle Bronze Age, sedentary
settlement redeveloped in the Jordan Valley
on a more modest scale. The valley's population grew slightly, while the number of settlements roughly doubled (Broshi and
Early Bronze Age urbanization in the Levant
is striking, because while urban communities
grew, primarily along the Mediterranean
coast, regional population declined as smaller
settlements dwindled in number and size in
all three subregions (Joffe 1991 b:Tables 8, 11,
12, Figures 9, 10). Far from anticipating this
waning sedentary population, most conventional interpretations view the Early Bronze
Age as a period of pronounced population
growth (e.g., Amiran 1970; Richard 1987).
Instead, these patterns hint that rural populations may have adopted strategies of "resilience" (Adams 1978) based on increased
pastoralism that peaked during the urban collapse of Early Bronze IV. This interpretation
finds circumstantial support in the persistent
compound nature of coastal rank-size curves
and the declining ruralism of the central hills
and Jordan Valley.
In contrast, Middle Bronze Age urbanization fits conventional expectations much better. Survey data suggest population growth,
both in cities and the general countryside.
While compound rank-size curves still suggest that cities and towns were superimposed
on coastal ruralism, urban-rural integration
became enhanced and may have extended
into the hill country. In the Jordan Valley,
settlement returned to a distinct pattern of
hinterland development similar to that of the
Diyala, but on a further reduced scale.
Interestingly, the advent of cities in the third
millennium B.C. and their rejuvenation in
the early second millennium B.C. followed
different courses of development. Further,
when dissected geographically, settlement
data from both periods reveal distinct, some-
Falconer and Savage]
EARLYURBANIZATION
INMESOPOTAMIA
ANDTHESOUTHERN
LEVANT
times divergent settlement trajectories within
the coastal plain, central hills, and Jordan
Valley, a composite pattern perhaps most
characteristic of Levantine urbanism. While
the southern Levant must have been affected
by the actions of foreign cities and states,
notably those of Egypt, its own expression of
urbanism was not simply derivative. Rather,
the earliest urbanism in the southern Levant
described an intriguing patchwork, in which
the largest cities were superimposed on a
much broader network of resilient towns and
villages that followed their own courses of
development.
Conclusions
Orthodox approaches to Near Eastern urbanism treat the rise of Near Eastern cities
as a uniform phenomenon with core and
marginal expressions. Adams recalls that
Benno Landsberger, a noted Assyriologist,
criticized his choice of the Diyala Basin for
initial survey reconnaissance because this
work defined a derivative "dialect" of early
urbanism "before the paradigm of the heartland [was] known" (Adams 1981 :xviii). Most
treatments of the southern Levant follow a
similar logic in which the rise of Bronze Age
cities is interpreted as a local vernacular expression of Near Eastern urbanization. This
view holds that Levantine urbanism simply
followed and echoed that of Syria and Mesopotamia, but on a smaller scale. While Mesopotamia is justifiably renowned for its heartland of very early, highly nucleated cities, our
rank-size analyses also elucidate a distinct
expression of hinterland development in the
Diyala based on more modest, but consistently integrated town and village life. In contrast, we demonstrate that Early Levantine
urbanism rarely adheres to either of these
general patterns, but represents a distinct geographic and chronological mosaic that defies
simple categorization. We conclude that Near
Eastern urbanization did not include so much
a Mesopotamian core and its dialectical offshoots, as a polyglot array of alternative forms
of city life and its relations with the countryside.
55
Ultimately, our study implies that cities do
not, in any uniform sense, epitomize all "urbanized" societies. A fuller comprehension
of urbanization as a highly variable phenomenon requires that we expand our comprehension of non-urban components in stratified settlement systems. We suggest that small
communities are most important in this regard, not simply as a supporting foundation
for urbanism. Instead, configurations of rural
settlement often define the overall contours
of rank-size distributions and, in so doing,
may reveal peculiar courses of rural development that contribute to the variety of trajectories for early urbanization in Mesopotamia, the southern Levant, and elsewhere in
southwestern Asia.
Acknowledgments. We thank George Cowgill, Charles
Redman, Norman Yoffee, and especially Keith Kintigh
for their abundant commentary on preliminary versions
of this study. The mathematical assumptions underlying
the K- test were brought to our attention through lengthy
discussion with Dennis Young, Department of Mathematics, Arizona State University. Carole Crumley, Keith
Kintigh, Kenneth Kvamme, and Barbara Stark made
helpful suggestions during the development of the simulation program. Moawiyah Ibrahim, James Sauer, and
Khair Yassine kindly provided access to the field notes
and ceramic collections of the East Jordan Valley Survey.
Falconer derived site size estimates from these sources
with the support of a National Endowment for the Humanities Travel to Collections Grant.
References Cited
Adams, R. McC.
1965 Land Behind Baghdad: A History of Settlement
on the Diyala Plains. University of Chicago Press,
Chicago.
1966 The Evolution of Urban Society: Early Mesopotamia and Prehispanic Mexico. University of Chicago Press, Chicago.
1975 An Early Prehistoric Site in the Warka Region.
Sumer 31:11-16.
1978 Strategies of Maximization, Stability, and Resilience in Mesopotamian Society, Settlement, and
Agriculture. Proceedings of the American Philosophical Society 122:329-335.
1981 Heartland of Cities: Surveys of Ancient Settlement and Land Use on the Central Floodplain of the
Euphrates. University of Chicago Press, Chicago.
Adams, R. McC., and H. Nissen
1972 The Uruk Countryside: The Natural Setting of
56
AMERICANANTIQUITY
Urban Societies. University of Chicago Press, Chicago.
Ahituv, S.
1978 Economic Factors in the Egyptian Conquest of
Canaan. Israel Exploration Journal 28:93-105.
Amiran, R.
1970 The Beginnings of Urbanization in Canaan. In
Near Eastern Archaeology in the Twentieth Century,
edited by J. A. Sanders, pp. 83-100. Doubleday,
Garden City, New York.
Auerbach, F.
1913 Das Gesertz der Bevolkerungskontration. Petermanns Mitteilungen 1913:74.
Aurenche, O., J. Evin, and F. Hours (editors)
1987 Chronologies in the Near East: Relative Chronologies and Absolute Chronology 16,000-4,000 B.P.
BAR International Series 379, British Archaeological Reports, Oxford.
Beaumont, P.
1985 Man-induced Erosion in Northern Jordan. In
Studies in the History and Archaeology of Jordan,
vol. 2, edited by A. Hadidi, pp. 291-296. Jordanian
Department of Antiquities, Amman.
Ben-Tor, A.
1991 New Light on the Relations Between Egypt and
Southern Palestine During the Early Bronze Age.
Bulletin of the American Schools of Oriental Research 281:3-10.
Berry, B. J. L.
1961 City Size Distributions and Economic Development. Economic Development and Culture Change
9:573-588.
Blanton, R. E.
1976 The Role of Symbiosis in Adaptation and Sociocultural Change in the Valley of Mexico. In The
Valley of Mexico: Studies in Pre-Hispanic Ecology
and Sociology, edited by E. R. Wolf, pp. 181-201.
University of New Mexico, Albuquerque.
Braun, E. L.
1985 En-Shadud.: Salvage Excavations at a Farming
Community in the Jezreel Valley, Israel. BAR International Series 249, British Archaeological Reports, Oxford.
Broshi, M.
1979 The Population of Western Palestine in the Roman-Byzantine Period. Bulletin of the American
Schools of Oriental Research 236:1-10.
Broshi, M., and R. Gophna
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 Settlements
and Population. Bulletin of the American Schools of
Oriental Research 261:73-90.
Christaller, W.
1933 Die zentralen Orte in Suddeutschland: eine
ikonomisch-geographische Untersuchung uber die
Gesetzmiissigkeit der Verbreitung und Entwicklung
der Siedlungen mit stidtischen Funktionen. Jena.
[Vol. 60, No. 1, 1995
Clarke, D. L.
1979 Towns in the Development of Early Civilization. In Analytical Archaeologist. Collected Papers
of David L. Clarke, edited by N. Hammond, et al.,
pp. 435-443. Academic Press, New York.
Cowgill. G.
1977 The Trouble with Significance Tests and What
We Can Do About It. American Antiquity 42:350368.
1990 Towards Refining Concepts of Full-Coverage
Survey. In The Archaeology of Regions. A Case for
Full Coverage, edited by S. Fish and S. Kowalewski,
pp. 249-259. Smithsonian Institution, Washington,
D.C.
Crumley, C. L.
1976 Toward a Locational Definition of State Systems of Settlement. American Anthropologist 78:5973.
Dever, W. G.
1980 New Vistas on the Middle Bronze I Horizon
in Syria-Palestine. Bulletin of the American Schools
of Oriental Research 237:35-64.
1987 Palestine in the Middle Bronze Age: The Zenith of the Urban Canaanite Era. Biblical Archaeologist 50(3): 149-177.
1989 The Collapse of the Urban Early Bronze Age
in Palestine-Toward a Systemic Analysis. In L'urbanisation de la Palestine a l'age du Bronze ancien,
edited by P. de Miroschedji, pp. 215-246. BAR International Series 527, British Archaeological Reports, Oxford.
Dziewonski, K.
1972 General Theory of Rank-Size Distributions in
Regional Settlement Systems: Reappraisal and Reformulation of the Rank-Size Rule. Papers of the
Regional Science Association 29:75-86.
al-Eisawi, D. M.
1985 Vegetation in Jordan. In Studies in the History
andArchaeology of Jordan, vol. 2, edited by A. Hadidi, pp. 45-57. Jordanian Department of Antiquities, Amman.
Esse, D. L.
1989 Secondary State Formation and Collapse in
Early Bronze Age Palestine. In L'urbanisation de la
Palestine a l'age du Bronze ancien, edited by P. de
Miroschedji, pp. 81-96. Bar International Series 527,
British Archaeological Reports, Oxford.
Falconer, S. E.
1987 Heartland of Villages.:Reconsidering Early Urbanism in the Southern Levant. Ph.D. dissertation,
University of Arizona. University Microfilms, Ann
Arbor.
1994 The Development and Decline of Bronze Age
Civilization in the Southern Levant: A Reassessment of Urbanism and Ruralism. In Development
and Decline in the Mediterranean Bronze Age, pp.
305-333, edited by C. Mathers and S. Stoddart.
Sheffield Academic Press, Sheffield.
Falconer and Savage]
EARLYURBANIZATION
INMESOPOTAMIA
ANDTHESOUTHERN
LEVANT
Finkbeiner, U., and W. Rollig (editors)
1986 Jamdet Nasr: Period or Regional Style? Beiheft
zum Tubinger Atlas des Vorderen Orients, B62.
Wiesbaden.
Finkelstein, I., and R. Gophna
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.
Fish, S., and S. Kowalewski (editors)
1990 The Archaeology of Regions: A Case for Full
Coverage. Smithsonian Institution, Washington,
D.C.
Gerstenblith, P.
1983 The Levant at the Beginning of the Middle
Bronze. American Schools of Oriental Research
Dissertation Series, No. 5. American Schools of Oriental Research, Winona Lake, Indiana.
Gophna, R., and J. Portugali
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 269:11-8.
Gophna, R., N. Liphschitz, and S. Lev-Yadun
1986 Man's Impact on the Natural Vegetation of the
Central Coastal Plain of Israel During the Chalcolithic Period and the Bronze Age. TelAviv 13/14:7184.
Haggett, P.
1971 Locational Analysis in Human Geography. St.
Martin's, New York.
Horowitz, A.
1979 The Quaternary of Israel. Academic Press, New
York.
Ibrahim, M., J. Sauer, and K. Yassine
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). In Archaeology of Jordan: Essays and Reports,
edited by K. Yassine, pp. 189-207. University of
Jordan, Amman.
Joffe, A. H.
1991a Early Bronze I and the Evolution of Social
Complexity in the Southern Levant. Journal of Mediterranean Archaeology 4(1):3-58.
199 lb Settlement and Society in Early Bronze I and
II Canaan. Ph.D. dissertation, University of Arizona. University Microfilms, Ann Arbor.
Johnson, G. A.
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(3):234-247.
Kantor, H. J.
1992 The Relative Chronology of Egypt and Its Foreign Correlations Before the First Intermediate Pe-
57
riod. In Chronologies in Old WorldArchaeology, 3rd
ed., edited by Robert W. Ehrich, pp. 3-21. University of Chicago Press, Chicago.
Kemp, B. J.
1983 Old Kingdom, Middle Kingdom and Second
Intermediate Period, c. 2686-1552 B.C. In Ancient
Egypt: A Social History, edited by B. G. Trigger, B.
J. Kemp, D. O'Connor, and A. B. Lloyd, pp. 71182. Cambridge University Press, Cambridge.
Kempinski, A.
1978 The Rise of an Urban Culture. Israel Ethnographic Society Studies, no. 4. Israel Ethnographic
Society, Jerusalem.
1989 Urbanization and Metallurgy in Southern Canaan, in L'urbanisation de la Palestine a l'age du
Bronze ancien, edited by P. de Miroschedji, pp. 163168. BAR International Series 527, British Archaeological Reports, Oxford.
Kenyon, K.
1973 Palestine in the Middle Bronze Age, in Cambridge Ancient History, 3rd ed., vol. 2, part 1, pp.
77-116. Cambridge University Press, Cambridge.
Kowalewski, S.
1982 The Evolution of Primate Regional Systems.
Comparative Urban Research 9(1):60-8.
Moore, T.
1959 A Note on City Size Distributions. Economic
Development and Cultural Change 6:465-466.
Na'aman, N.
1981 Economic Aspects of the Egyptian Occupation
of Canaan. Israel Exploration Journal 31(3-4): 172185.
Paynter, R. W.
1982 Models of Spatial Inequality: Settlement Patterns in HistoricalArcheology. Academic Press, New
York.
1983 Expanding the Scope of Settlement Analysis.
In Archaeological Hammers and Theories, edited by
J. Moore and A. Keene, pp. 233-75. Academic Press,
New York.
Porada, E., D. P. Hansen, S. Dunham, and S. H. Babcock
1992 The Chronology of Mesopotamia, ca. 70001600 B.C. In Chronologies in Old World Archaeology, 3rd ed., edited by Robert W. Ehrich, pp. 77121. University of Chicago Press, Chicago.
Redman, C. L.
1978 The Rise of Civilization. W. H. Freeman, San
Francisco.
Richard, S.
1987 The Early Bronze Age: The Rise and Collapse
of Urbanism. Biblical Archaeologist 50:22-43.
Richardson, H. W.
1973 Theory of the Distribution of City Sizes: Review and Prospects. Regional Studies 7:239-251.
Robinson, D. J.
1972 Changing Settlement Patterns in Colonial Hispanic America. In Man, Settlement and Urbanism,
edited by P. J. Ucko, R. Tringham, and D. W. Dim-
58
AMERICANANTIQUITY
bleby, pp. 931-943. Schenkman, Cambridge, Massachusetts.
Rosen, A. M.
1986 Cities of Clay: The Geoarchaeology of Tells.
University of Chicago Press, Chicago.
Sanders, W. T., and D. Webster
1988 The Mesoamerican Urban Tradition. American Antiquity 90:521-546.
Savage, S. H.
1993 A Rank-Size Simulation Study of Settlement
on the Levantine Coastal Plain during the Chalcolithic, Early, and Middle Bronze Ages. Manuscript
on file, Department of Anthropology, Arizona State
University, Tempe.
Schiffer, M. B.
1987 Formation Processes of the Archaeological Record. University of New Mexico Press, Albuquerque.
Shennan, S.
1990 QuantifyingArchaeology. Academic Press, New
York.
Skinner, G. W. (editor)
1977 The City in Late Imperial China. Stanford University Press, Stanford.
Smith, C. A.
1976 Regional Economic Systems Linking Geographical Models and Socioeconomic Problems. In
Regional Economic Systems, vol. 1, edited by C.
Smith, pp. 3-68. Academic Press, New York.
Sokal, R. R., and F. J. Rohlf
1969 Biometry. Freeman Press, San Francisco.
Stager, L. E.
1992 The Periodization of Palestine from Neolithic
through Early Bronze Times. In Chronologies in Old
World Archaeology, 3rd ed., edited by Robert W.
Ehrich, pp. 22-41. University of Chicago Press, Chicago.
Stronach, D.
1961 The Excavations at Ras al 'Amiyah. Iraq 23:
95-137.
Thomas, D. H.
1986 Refiguring Anthropology: First Principles of
Probability and Statistics. Waveland, Prospect
Heights, Illinois.
Vining, R.
1955 A description of certain aspects of an economic
system. Economic Development and Cultural Change
3:147-195.
Ward, W. A.
1971 Egypt and the East Mediterranean World,22001900 B.C.; Studies in Egyptian Foreign Relations
During the First Intermediate Period. American
University of Beirut, Beirut.
Weinstein, J. M.
1975 Egyptian Relations with Palestine in the Middle Kingdom. Bulletin of the American Schools of
Oriental Research 217:1-16.
Wheatley, P.
1971 The Pivot of the Four Quarters: A Preliminary
[Vol. 60, No. 1, 1995
Inquiry into the Origins and Character of the Ancient
Chinese City. Aldine, Chicago.
Wright, G. A.
1973 Bristlecone Pine Calibrations of Radiocarbon
Dates: Some Examples from the Near East. American Journal of Archaeology 77:197-201.
Wright, H. T.
1981 The Southern Margins of Sumer: Archaeological Survey of the Area of Eridu and Ur. In Heartland of Cities, by R. McC. Adams, pp. 295-345.
University of Chicago Press, Chicago.
Yoffee, N.
1986 The Process of Ruralization in Social Evolutionary Theory. Paper presented at the 51 st Annual
Meeting of the Society for American Archaeology.
Zipf, G. K.
1949 Human Behavior and the Principle of Least Effort: An Introduction to Human Ecology. AddisonWesley, Reading, Massachusetts
Notes
' The RankSize program was written by Savage in Turbo
Pascal 6.0 (Borland International 1990) for IBM PC or
compatible computers.
2 These
periods lie at the crossroads of prehistoric radiocarbon-based chronologies and historic chronologies.
Radiocarbon recalibrations tend to push these horizons
earlier, often by several centuries (e.g., see Aurenche,
Evin, and Hours 1987).
3 Most archaeological sites in Mesopotamia and the
southern Levant are multiphase mounded tells in which
later deposits obscure the habitation areas of earlier strata. Since site size often can only be estimated according
to the area covered by the largest occupation, some estimates are inflated. However, most survey reports offer
period-by-period size estimates whenever possible, primarily for larger sites.
4 For the sake of graphic clarity only selected curves are
presented in our rank-size figures (Figures 4-6, 8-12).
However, simulation results for all rank-size analyses
are included in Tables 2 and 4.
5This period is referred to as "Intermediate Bronze" or
"Intermediate EB-MB" in some literature (see discussion in Falconer 1993).
6
Schiffer (1987:340-353) provides a broader discussion
of the many factors that influence site recovery rates by
archaeological surveys.
7 These zones correspond to geographic zones 4, 6, 7,
and 9 in Joffe 1991 b, and Broshi and Gophna 1984 and
1986.
Received September 28, 1993; accepted August 1, 1994