v. January 17,1997

To be archived until April 15, 1997

The purpose of this research note is to explore the possibility of extending back in time Tertius Chandler's (1987) census of world cities and to produce a list of the world's earliest cities , for the third and fourth millennia BC. That is a basically descriptive task, but it would also take us to the start of what Gordon V. Childe was the first to first call the "Urban Revolution" that launched not only the first network of cities but also civilization as we know it.

Let us adopt as our general baseline the work of Tertius Chandler (1987) that has now become a basic reference for students of the evolution of the world system (see i.a. Wilkinson 1992­3, Chase­Dunn and Willard 1993; Bosworth 1995). That compendium of major cities of the past 4000 years is notable not only for the novelty of the attempt at world­wide coverage of urbanism, and over such an extended period, but also for the time spent, and care taken, to collect population counts for most of the entries into that list. But while the census covers most of the historically extant cities, it does not fully explore the very earliest ones , in Mesopotamia and surrounding areas, and is not really useful for illuminating the inception of the world system of cities.

The point in time at which Chandler starts his list is ­2250, and for that year he shows eight entries, arranged in order of population size starting with Memphis and Akkad, and he also produces two other short lists, for ­2000, and for ­1800 (which last is our own cut­off date). The full data is presented in Appendix 1. We note that actual population data are sparse, but relative size is implied in the ranking on the list. We note too that only five of the close to 30 estimates in his lists ¹ are actually supported by references; the remainder are rank­ordered by the author in relation to these.

The only other work we know to make available a comprehensive population estimate for the early cities is Colin McEvedy's PENGUIN ATLAS OF ANCIENT HISTORY (1967:26­7) that shows, for ­2250, the same year as Chandler's, twelve cities that he estimates "would have had populations in the 10 ­ 15,000 range". His map highlights a total of 13, in Mesopotamia, Egypt, and in the Indus River Valley (cf. Appendix 1). There is some convergence between the two lists, but it is far from close.

Our own task is to collect fresh information for the first one thousand years of world cities, and to take another look at Chandler and McEvedy earliest data.

Criteria for inclusion

Let us define as a city a community with a significant degree of division of labor that makes it part of a network of cities. That would distinguish it, for instance from a settlement of farmers, such as Jericho, with a population believed to have been, ca. ­7000, in the region of 1­2,000. albeit protected, famously, by a wall, but not operating in a system of cities.

To qualify for a list continuous with Chandler's for this, initial period of urban formation (after about ­3500) let us consider all cities that fall within the range of 10,000 to 100,000. That seems to be the size of population within which a systemic division of labor might take firm hold. In this ancient era (­3500 to ­1200; cf. Modelski and Thompson 1996), the upper limit of city populations seems to have been about 100,000, the size attained on Chandler's list only once, by Avaris, in ­1600; McEvedy's first reaching of the 100,000 mark occurs in ­825, at Niniveh.

The lower limit of 10,000 is about the same size criterion as McEvedy's and only slightly lower than that implied in Chandler's listings for ­2250, 2000 and 1800. That would make inclusion warranted for settlements at least one order of magnitude larger than those comparable to Jericho. We aim to include all cities that meet the criterion of 10,000 but might not, of course, be able to do so due to data limitations.

The cities for this survey are thus selected on the basis of the criterion of population size. This is the onset of urbanization, and cities are small, and their populations hard to estimate. The written record of Sumer, as well as of Egypt, does supply us with the names and some descriptions, particularly so in the case of Sumerian literature, so strikingly proud of its cities, but that record does not include counts of populations; censuses are non­existent, even though the practice of counting e.g. armies and battle casualties does seem to be taking hold by the end of the period.

To determine the populations of cities known either from written records, or from archaeological research, so that they might be considered for inclusion in this list, two approaches might be pursued, the use of other authors' estimates, and making inferences from the area of settlement by introducing a population density factor.

One obvious source is the extant archaeological literature. Writing about Mesopotamia, Henri Frankfort (1948:222) agreed that the total population of the city states can be computed only in a general way, but that in the Early Dynastic period (ca.2900­2400) it would seem "to vary from, say, 10,000 to 20,000 people" ². Ruth Whitehouse (1977:48) took this thought one step further and maintained that in Sumer, at that time, there were "probably never more than 20 city­states" and proceeded to give a straightforward list of fourteen that she judges to have been the most important cities in that period (see Appendix 1); she then added: "we should perhaps not be far wrong if we thought in terms of populations of 10,000 to 20,000 for most of these cities, with perhaps 50,000 at Uruk" (ib.). McEvedy's population estimates, in the 10,000 to 15,000 range (same Table 1) were in the same range, if a shade more conservative.

Other such estimates are hard to pin down, though they do appear in print from time to time, and a selection is shown in Table 1. One important estimate concerns Uruk that on the basis of the work of Richard Adams (1967, 1981) features as the largest city of the very early period. Another useful contribution is provided by figures for the two major Indus Valley cities, Mohenjo­daro and Harappa, that are thought to have attained, at their peak, the 20,000 to 40,000 range (Allchin 1975:341; Hamblin 1977:145; Whitehouse 1983). But it is not always clear how the authors arrived at their estimates.

Population density

Two elements go into a population estimate: the archaeologists' site assessment (be it the area of urban settlement in general, or an estimate, or actual count, of houses), and a population density factor, be it "macro", for the entire urban site, or "micro", per house ratio (ranging say from eight persons per house at III Dynasty Ur (Woolley 1954:193) to an "ideal" five or "actual" 3.5 for Adams (1981:144,349­50). The micro­estimate requires a reliable house count, and that is not really available for most of the sites. On the other hand, a macro­factor is liable to error, such as ignoring local conditions.

Table 1 that lists several authors' city populations that include both site assessment, and population figures for the third and fourth millennium and it demonstrates that the population density factor implied in these estimates has ranged widely in recent literature. While the lowest is 100­125 inhabitants per hectare (in the published proposal for Uruk), the mean for all the estimates is around 350 inhabitants per urban hectare, and some proposals are significantly more expansive.

Mean of 13 density estimates: about 350

E.D. + Early Dynastic period ca.2900­2400.

As illustrated in Table 1, the literature on urban population shows little unanimity in that regard. Chandler (1987:7) explains in his section on "methods and sources" that "we have become accustomed to standard ratios of 100 people per hectare ... in the Orient and much of Europe, but around 75 per ha. for new walls and up to 200 per ha. just before a new wall is built, and even higher densities in special geographic circumstances" ³. But Yigael Yadin, writing about warfare engaging "cities of the ancient Middle East", most of which were walled, believes it to be a "reasonable assumption" to maintain that in such cities "there were roughly 240 inhabitants to an urban acre" and that yields a high density estimate of nearly 600 to a hectare (at 0.405 ha to an acre). He also proposes that the proportion of fighters among the inhabitants averaged 25 per cent.

McEvedy (1967:44) observes that "the average walled city of classical times" covered 20­60 hectares and that a density of about 250 inhabitants to the hectare is a reasonable assumption though probably a slightly lower figure is indicated for Western Europe and a slightly higher figure for the Near East". We need to remember though that McEvedy is referring here to "classical" times, after about ­1200. Does that indicate that classical towns had one half the densities of Yagin's ancient world? Maybe not, because in walled settlements (as in Yadin's and McEvedy's estimates) the pressure for compact arrangements and the need to man the ramparts must have been a strong or stronger force for higher density even in the earlier period.

In his basic study of Mesopotamian urbanization, Robert McCormick Adams (1981:69,349) proposes a "reasonable and perhaps conservative" standard: "125 persons per hectare of actual site area, or about 100 persons per hectare if calculated only from the measurement of maximum length and width". Early Dynastic Uruk is known to have been enclosed by an early double wall with a circumference of about 9.5 km. We also know that in addition to temple precincts and palaces, and one and some two­story houses, some parts of the city included orchards and clay pits ⁴. "Covering approximately 400 hectares ... Uruk appears likely on this basis to have had a population of certainly no less than 40,000 to 50,000" (ib.:85); this is the only actual population estimate to appear in a massive survey covering some three millennia of urban growth.

More recently, Hans Nisssen (1993:56), who worked with Adams on the Uruk and its country­side, has offered a higher figure: for late Uruk, at the end of the fourth millennium, he proposes 25­50,000 (density factor of 100­200 people/ha), and for the Early Dynastic I period, ­2990­­2800, a settled area of "slightly less than 600 hectares" (with population therefore presumably in the range of 60­120,000, much higher than earlier figures). He describes the 100­200 density standard as "one we have come to accept".

J. N. Postgate (1991:79­80) agrees that in the present state of knowledge, the best approach toward estimating population of Mesopotamian cities is from the archaeological evidence, but does not see it as a simple matter because density and homogeneity of occupation are likely to vary widely from city to city. He regards the macro­density factor method as unsatisfactory, because the range (that he gives as 100­400) is too wide to be useful, and because its basis is uncertain. He proposes instead to follow the micro­method of building our knowledge up from the ground level of individual houses and their uses. Given that this procedure is exceedingly time­consuming, and could never be applied to all extant sites, his approach does not take us very far in a global survey. In his imaginatively documented volume Postgate gives (ib.:27) a list of 36 ancient Sumerian cities, a "fairly uniform class of major population centers distributed widely across the Southern plain" but he leaves his readers without a sense of what populations of these area might have been.
That means that the decision must be the analyst's own. The safest bet would be to go with 100 people/hectare, but that standard (as Adams was well aware) appears overly conservative and at variance with some of the fragmentary data we have on other matters, such as size of armies, casualty counts, and employment. Moreover, while that figure might well be applicable to all settlements, including villages, it seems too low for the larger and largest cities we are interested in, cities in which ease of communication, hence density must be at a premium. A number of other analysts seem to incline toward a higher figure. That is why the standard adopted in the present study (and in Table 2) for converting size estimates to population counts will be 200 per hectare ⁵. It means that a site of about 50 hectares would yield our minimum population of 10,000. But since the site data is also shown, in Appendix 2, the reader might wish to make his own count from the raw data here assembled.

The rank­size rule

But that is not all. Population estimates derived from archaeological surveys allow only rough­and­ready figures indicating orders of general magnitude. Might there not a method for narrowing down those broad­gauged guesses? Why not introduce the rank­size rule?

Systems of cities tend to be characterized by a rank­size relationship, first noticed by George Zipf, now also known in more general terms as the power law, that states that a city's population is inversely proportional to its rank. When all cities of such a system are arranged in order of size (that is, are rank­ordered), the largest will be twice as large as the next largest, the third will be one third of the largest, and so on. Systems of cities that follow that rule will describe, on double­log graph paper, a straight line, downward sloping at an angle of 45 degrees.

Richard Adams (1981:72­5,84­5) used this general rule for plotting the distribution of human settlements in Southern Mesopotamia, and found it to be approaching log­normality (that is, coming close to forming a straight line) in the entire area under his investigation in Early Dynastic I period (ca. ­2800); due to data limitations, he did not attempt to show it for Early Dynastic II­III (ca.­2500). Using the Chandler data, Christopher Chase­Dunn and Alice Willard (1993) applied this rule to measuring the degree of concentration in a wide range of urban systems.

On this occasion the rank­size rule will be employed not for testing the degree to which a given system of cities approaches a log­normal, or "harmonious", distribution (or else inclines either toward urban primacy, hence also dependency on the one hand, or dispersion, on the other); rather the rank­size principle will be employed to improve the estimate of, or in fact predict, the population of individual cities.

What might be a predicted distribution of city sizes for Early Dynastic I Sumer? We assume at this point that Sumer did, in that period, form a recognizable system of cities, one whose distribution would in fact tend toward log­normality, that is a full demonstration of the rank­size rule. We also assume that the population of the largest city, Uruk, was at that point 80,000 (400 ha at 200 people/ha), and that the rank order of the other cities would be: Kish, Umma, Shuruppak, Bad­tibira, Nippur, Adab, and Suheri. Their population might then be read­off as shown in Table 2 in the first two sections of the column for ­2800. Figure 1 shows the predicted distribution for ­2500, Early Dynastic III, this time on the stronger assumption that the potential for the largest city approaches 100,000, but that neither of the two presumably largest cities approaches that potential.

(Figure 1 about here)

What might be the grounds for regarding the Sumerian distributions for ­2800 and ­2500 as log­normal? In the first place, we know that Adams has shown all human settlements in his study to be nearing log­normality in ­2800, and we suppose that by ­2500, that trend had not reversed yet, as it probably did in the Akkadian period, by ­2300. Log­normality means a movement away from primacy or dependency and toward lesser inequality, but not one of total equality. From the reading of the historical record, too, we would expect the size distribution to be log­normal rather than "uniform", one in which two or three of the largest city states would, by means of coalitions, contend for leadership.

Such a picture of an orderly distribution of city sizes might be a corrective to the "flattened" image presented by the scholars cited earlier, including Franfort (1948), Whitehouse (1977), Postgate (1991:26), and Roux (1995, in Appendix 1).


A provisional list

Table 2: Early world cities (above), presents a provisional listing of the world's major cities for the early ancient period ­3700 to ­1800. It is grounded in three sets of sources, first, in the archaeologist's site estimates (reported in Appendix 2), second, in the application of the rank­size rule to the Sumerian portion of our listings for ­2800 and ­2500, and third, in available documentary evidence such as king lists, inscriptions or historical accounts.

Let us now discuss our data in more detail, justifying each of these listings in Table 2 in turn, column by column. On the evidence at hand, no cities qualify on our minimum criterion prior to about ­3700 to ­3600. In the Ubaid period (to ­4000) Eridu, reputedly Sumer's first city, is thought to have had "no less" than 4,000 inhabitants (Mellowan 1970:331, see also Table 1). But in the early Uruk period Eridu expanded greatly, reaching, according to Wright (1981:325), "at most" 6,200­10,000 people on close to 50 hectares. For the Early Middle Uruk period Adams (1981:64,71,114,348) also shows Uruk at 70 hectares, and Larak, in the Nippur­Adab environs, at 50 hectares; here we appear to have s cluster of two­three cities that satisfy our criterion, and form the nucleus, (or embryo?) of an emerging system of cities.

The column for ­3400 (Late Uruk period) shows basically one major Sumerian entry, one whose settled city site, of some 100 hectares, indicates a population that fully qualifies on our basic criterion, and that is Uruk. We have one other firm entry, for Larak (cf. Adams 1981:348, 114, site 1306), one of the five "antediluvian" cities. We mark with a "c" the four other "antediluvian" cities that have come down in Sumerian tradition as the first sources of civilization, even though site reports on Bad­tibira and Shuruppak do not support a full listing ⁶.

In addition to Eridu (whose status after the late Uruk period is uncertain, cf. Annotations), the other "antediluvian cult centers" are Bad­tibira, Larak, Sippar, and Shuruppak. Hallo (1970) argues that they should be viewed as contemporaneous, and not as existing in a sequence. They feature prominently in the extant sources of epic literature: in the myth of the Deluge (Pritchard 1959:29), and in the Sumerian King List (Jacobsen 1939:71­77) that names them as the five antediluvian centers that successively claimed "kingship" (that is leadership". ⁷ The Epic of Gilgamesh (Pritchard 1959:49 ff) that opens with the foundation of Uruk, mentions only three other cities: Eridu, Nippur, and Shuruppak. The antediluvian cult centers may not have qualified on our population criteria but they seem to have functioned as a differentiated system of city­states, among members of which "leadership" may or may not have been transferred, and specialized economic functions might have been allocated, as e.g. for Bad­tibira, the city of metal­workers. The memory of their foundational activities seems to have been treasured for much longer than a millennium. Being described as "cult centers" also confirms that the original nucleus of these centers must have been the temple precinct.

The dating of the Sumerian Epic Deluge that demarcates the end of the antediluvian period remains an unresolved problem. Evidence of a great flood at Ur, on the Euphrates, found by Leonard Wooley, points to a date "not very much later than 3500 BC". On the other hand, the hero of the flood story, Ziusudra, apparently reigned as king of Shuruppak ca.­2900, and so did Gilgamesh, in Uruk, somewhat later. Nevertheless, evidence of a flood faced by both Kish and Shuruppak may be found in the strata separating the Ubaid and Jamdet Nasr periods on those sites (Mallowan 1970). Might we have here two instances of a five­hundred year flood, one at ­3400, and the other at ­2900?

The column for ­3100 (Jamdet Nasr period) shows once again the five antediluvian cult centers and has two other additional entries for Sumer; Uruk, as a major city in the 250 ha range (cf. Nissen 1993 for a full account), as well as Suheri (site 242), and Umma (site 230, in Adams 1981:101).

A geographical list from Uruk, dating from about ­3000, presents as its first entries the following cities, in that order: Ur, Nippur, Larsa, Uruk, Kes, and Zabalam (Green 1977). Of these, only Uruk meets our population standard (based on site measurements) , but the list does suggest not only spatial awareness but the existence of a system of interlinked cities.

These Sumerian cities functioned as the "heartland of cities" in the world system at large. Outside that core, urban settlements that might be viewed as "enclaves" or "outposts" of an "Uruk world system" did exist but mostly failed to match our criteria: in the Susiana plain to the east, Susa might have reached 25 hectares in extent, and Chogha Mish, 18 ha; in the Syrian­Mesopotamian Plains and Highlands, the largest settlement, at Habuba/Tel Qannas, with possibly up to 40 hectares, might have had a population range of 6­8,000 (Algaze 1993: Chs 2 and 3, ­esp.p.58). The only exceptions appear to be Anshan, capital of Elam and a trading center in southeastern Iran, and Tell Brak, in northern Mesopotamia, with a temple precinct not unlike that of Uruk.


Egypt: civilization without cities?

This account has been so far of the cities of Summer and of its immediate neighborhood. But what about Egypt, that is in many ways comparable to Mesopotamia, and that underwent political unification just about this time, at ­3100?

Some students of the ancient era have been known to argue that, unlike Mesopotamia, Egypt lacked anything that could be regarded as cities in modern terms. That great country did have temples, palaces, and cemeteries, often of monumental proportions, as early as the fourth and third millennia but its capitals seem to have lacked remarkable size, and have left little evidence either of intellectual life, or of commercial activity in particular (given the crown's monopoly of power, and of foreign trade). As John A. Wilson put it: "For nearly three thousand years, until the founding of Alexandria, Ancient Egypt was a civilization without a single major city" (in Kraeling and Adams 1960:135).

Manfred Bietak (1979,1991) rejects this rather stark view of the Egyptian experience and pleads that urban archaeology of that area is still only in an elementary stage. Fascinated by a multitude of beautiful objects, much in demand by the museums of the world, and hampered by the difficulty of digging in the overcrowded valley and delta of the Nile, archaeologists neglected the study of the ancient city. He argues nevertheless that Egypt was "an urbanized society from the beginning of the Old Kingdom", and even pre­dynastic (pre­3100) Egypt had several centers of political and economic power, both in Upper Egypt as at Hierakonpolis, or in the Delta, such as Buto.

The political unification of the country ca.­3100, argues Bietak (1987:31), led to the creation of urban settlements of a planned character, square and enclosed by a wall, but of limited size. "It appears that from the beginning of the first dynasty onward, the population was gathered in new sites founded by the crown. That way it was easier to control, and could be deployed over the whole country according to the wishes of the crown." These royal foundations seem to have anticipated the coming of the great construction projects, barrages, and canals that required whole armies of workers. The conditions for the emergence of cities were all in place "even if these agglomerations did not cover more than about 1.8. to 2 ha each". In other words, these early Egyptian planned settlements, presumably including Memphis founded about this time, did not meet our threshold criterion.

The first two sections of the column for ­2800 (Early Dynastic I) are based on Figure 1:Predicted population size for Sumer cities. As previously explained, Uruk is taken to be the largest city of the system (with a population of 80 thousand); the other cities are rank­ordered in relation to it, and their populations deduced from the rank­size model for log­normal distribution. It is assumed that Sumer, at this point of time, formed a coherent city­system. Note that this model allows only eight cities to exceed 10,000 in population.

The column for ­2500 (Early Dynastic III) employs the same model, except that it assumes that largest city to be potentially as large as 100,000. In fact though, Uruk, while still the largest city, is thought to have declined in population (Eannatum of Lagash claims to have conquered Uruk ca.­ 2450) and is taken to approximate the size of Umma or Lagash. The larger potential allows ten cities to exceed 10,000, but it also means that a number of cities customarily listed for the Early Dynastic period, such as Eridu, Bad­tibira, Sippar or Akshak (see Appendix 1), do not make our list.

Outside Sumer, we add some others for newer cities: Ebla, in Northern Syria, (whose population about ­2450 is estimated by Pettinato (1981:131) to have been "40,000 at most") and Mari, a commercial and competitor of Ebla in Northern Mesopotamia, that must have been comparable in size. In Iranian Baluchistan we might add Shakhr­I Sokhta, a recently excavated city of some 20,000 on more than 100 hectares and part of the then trading system (Whitehouse 1977:150).

The first city in Egypt making an appearance on this list as likely to meet our criteria is Memphis, then the capital of the country, but we do not have a site estimate for it. ⁸ This is the time of the building of the Great Pyramids in the vicinity of that location. In Bietak's (1979:129­130) view, these projects lent a strong impetus toward the development of urban life.

Our list also includes Mohenjo Daro at this early date. Recent research (reported in Paropla 1994:6­25) has shown this Indus Valley city to have had reached its mature stage at about ­2500, with a population that is put at some 40,000. It is thought that the laying of the foundation of the citadel would have required the work of some 15­20,000 workers for several years. Mohenjo­daro was part of a regular network of settlements, with major cities at an average distance of some 250 km. In addition to Mohenjo­daro and Harappa, both now rated at 40,000 inhabitants each, three other cities of "almost the same size" have been discovered more recently, but are still unexcavated ⁹.

The column for ­2300 stands for the Akkadian period that began with the conquests of Sargon ca.­3350. Of the 23 world cities we name for ­ 2300, the Akkadian rulers controlled fifteen, hence a good portion of the "civilized" world, but they did so at some cost. Only seven out of 12 Sumerian cities from ­2500 reappear, survivors of what must have been some devastating wars. Sargon defeated Lugal­zagesi of Uruk, and razed the ramparts of his city, as well as the walls of Ur, Lagash, and Umma. But the system of cities has also been spreading out. In addition to Tell Brak, new important centers have now appeared in the Syrian­Mesopotamian area. Other than Mari, and Ebla/Tell Mardikh, rated at 56 ha, we have three other archaeological sites in the 100 ha range: Qatna, in the Orontes basin, 100 ha; Tell Chuera (between the Balikh and the western Khabur), 100 ha, and Tell Taya, in the Mosul area, estimated at 70­160 ha (Algaze 1993:108). The Harappan cities are now at their peak.

Namazga­depe, near to­day's Ashkhabad, in Turkmenistan, a center of potteries and metallurgical trades, measures 70 ha (Whitehouse 1977:156). On our list, that is the city nearest to China, along the path of what later was to become the Silk Road.

The Third Dynasty of Ur (Ur III) (2112­2004) controlled most of Sumer, but its reach was not as long as that of Akkad. Ur­nammu, the founder, is remembered for his ziggurat at Ur. That city gave the name to the dynasty, but because of its peripheral position in the south, may not have been the central administrative focus of its rule. Leonard Woolley estimated its population at 34,000 within the walls, with a much larger number outside its walls. However, the city was sacked and ravaged by the Elamites in ­2004, and for that reason its population at ­2000 remains uncertain.

The column for ­1800 shows the close of the Isin­Larsa period. Isin, Larsa and Umma are now the major towns of Sumer but they soon fall prey to Babylon (1792); by ­1600 the cities of Sumer have faded ¹⁰. Mohenjo­daro has been abandoned by ­1800, and the remaining urban center now appears to be Egypt.

But here we do have the first entry for China. A site at Ehr­li­t'on, of some 350 hectares, and with two palatial buildings, shows radio­carbon datings going back from ­1800 to ­2100. We place it in the list for ­1800, as a possible capital of the Hsia Dynasty (2205­1766) even though debate continues whether that might not be an early Shang site (­1700 to ­1100) (Chang l986:309).

Table 2 makes it possible to put together a list of the world's largest cities at those dates, and these are shown in Table 3. As contrasted with Chandler's tabulations, they demonstrate the priority of Mesopotamian urban centers in the fourth and third millennia, but in the second millennium they affirm the importance of Egypt centers as the then oases of order in a world in turmoil (11).

Source: ­3400­1800: Table 3; for 1600,1400: Chandler 1987: 460,523ff.

In the fourth and third millennium, a system of cities emerged that covered the Fertile Crescent and the Indus Valley, and by ­1800, apparently extended to China. In Europe, cities appear on the Chandler list for the first time in ­1600, with Knossos on Crete in particular, followed in ­1360 by Mycene (30,000) in mainland Greece. In Meso­America, the Valley of Mexico, and the Yucutan peninsula see the first towns rise a thousand years later, in mid­first millennium BC: on Chandler's list of large cities, Cuicuilco (Mexico) is shown at ­300, and Izapac, of the Mayans, at +100. Tikal, the Mayan's largest city may have peaked at about 40,000. In Africa south of the Sahara, cities first develop in Ethiopia (Axum) at about 300, and in West Africa, at +1000.

This essay in collecting systematic information on the earliest of cities seems not unpromising. All the estimates most by their nature remain highly speculative and tentative but they do provide a basis for further discussion.

I wish to thank Richard McCormick Adams, Christopher Chase­Dunn, and William R, Thompson for their comments on an earlier version of this paper.

1. Tertius Chandler (1987:93) gives the following sources for his estimates: ­2250: EBLA (30,000): Howard LeFay, NATIONAL GEOGRAPHIC, December 1978; ­2000: UR (65,000): Woolley 1955, 34,000 within walls; Memphis (60,000): Lichtheim ANCIENT LITERATURE OF EGYPT,1973, vol.I:104, 10,000 untaxed men (soldiers? priests?); SUSA (25,000) Strabo GEOGRAPHY 1917/32:15:3,2:728, 120 stades around (use 1/3 only); ­1800: MOHENJO­DARO (20,000) M. Wheeler INDUS CIVILIZATION 1968:26.

2. Henri Frankfort (1948:396) offers the following individual population estimates for the Early Dynastic period: Lagash (that is Girsu) 19,000; Umma 16,000; Eshnunna 9,000, Kafajah (Tullub) 12,000; partly in the light of these figures, and assuming urban density of 200 per acre, Braidwood and Reed (1957:29­30) put the population of Sumer ca. ­2500, at half a million.

3. Chandler (1987:7) adds: "Chinese cities tend to have an especially low density because of the Chinese refusal to sleep below anyone, so their houses are ... nearly all of just 1 story. Hence, inland Chinese cities had a density of only about 75 per hectare, and even in seaports or the imperial capital the density hardly exceeded 100".

4. Harvey Weiss (1986:95) advocates a "correction factor", to 60 per cent of an "otherwise reasonable estimate of 100 person per ha of mounded settlement", to accommodate the argument that "only 40 to 60 per cent of a built­up city mound was comprised of residential structures".

5. With the exception of the entry for Ehr­li­t'on (in ­1800), presumably an imperial capital for which, in the light of Chandler's comments (see note 3), a factor of 100 people/ha will be applied in Table 3.

6. We owe that listing i.a. to the "antediluvian" section of the Sumerian King List (that was compiled ca. ­2000; see Jacobsen (1939:55­6), and to other epics (Hallo 1970). Jacobson notes the peculiarities of that section, and in particular its origin, that is likely to be Eridu, and its independence from the rest of the list, being probably a later addition. In the King List proper, the transitions between "kingships" held by the cities occur by force of arms ("city A was smitten with weapons"), in the ante­diluvian section the formula for the five cities is "I (the author) drop the city". All five cities are attested in the historical and archaeological record: Eridu from Ubaid onward, Larak and Shuruppak from ­3700, Bad­tibira from ­3100, but no report on earliest Sippar. A notable omission from the antediluvian section is Uruk whose site is largest and goes back to the Ubaid period. The entire Sumerian King List, most likely is, as Michalowski (1883) has argued, a propaganda piece on behalf of the Isin Dynasty that followed Ur III, and need to be viewed with caution.

7. The term "kingship" in the context of the Sumerian King List might be better rendered, especially for the "antediluvian" period, as "leadership". The archaeological evidence for "kingship" , such as royal palaces, is confined to the Early Dynastic period, and begins with Kish, and so is the use of the term "lugal" for king, or "war leader" (that dates from about ­2700).

By contrast, "ensi" (steward, manager) is older than "lugal", and emerges, in the written sources, from the temple ca. ­3000 (Maisels 1990:169­172).

8. In Bietak's (1979:98) view, "statistical definitions concerning population densities are probably not relevant to Ancient Egypt".

9. These are Judeiro­Daro, in Sind, Lurewala Ther in Haryana, and Ganweriwala Ther, on the dried­out course of Hakka River, in Haryana (Parpola 1994:6).

10. On Chandler's list for ­1600, the only Mesopotamian cities are Babylon (60,000), and Nineveh (less than 25,000). In Adams' Table 14 (1981:172­3) for the Old Babylonian period Umma rates 200+ ha, and Adab, Zabalam, Site 1389 (Umm al­Khezi), and Jidr, 40­200 ha.

11. Chandler's list of the world's largest cities (1987:523­7) starts with Memphis, from ­3100 onwards, and goes on to Akkad, from ­2240, Lagash, ­2075, Ur, ­2030, Thebes, ­1980, Babylon, ­1770, and Avaris, ­1670.

Adams, Richard McCormick (1966) THE EVOLUTION OF URBAN SOCIETY, Chicago: University of Chicago Press.
­­­­­ (1981) HEARTLAND OF CITIES, Chicago: University of Chicago Press;

­­­­­ and H.J. Nissen (1971) THE URUK COUNTRYSIDE, Chicago: Chicago University Press.

Algaze, Guillermo (1993) THE URUK WORLD SYSTEM: THE DYNAMICS OF EXPANSION OF EARLY MESOPOTAMIAN CIVILIZATION, Chicago: University of Chicago Press.

Allchin, Frank Raymond (1975) "History of the Indian Subcontinent: The Indus Civilization" ENCYCLOPEDIA BRITANNICA, Vol.9, pp.338­344.

Barrow, Finn and Manisha Shodhan (1977) NOTES ON THE URBAN HISTORY OF INDIA, Copenhagen: School of Architecture.

Bietak, Manfred (1979) "Urban archaeology and the town problem" AT PP. 97­144 OF in EGYPTOLOGY AND THE SOCIAL SCIENCES, Cairo.
­­­­­(1986) "La naissance de la notion de ville dans l'Egypte Ancienne, un acte politique?" at pp. 29­36 of SOCIETES URBAINES EN EGYPTE ET AU SOUDAN, Cahiers de recherches de L'Institut de Papyrologie et d"Egyptologie de Lille No.8, Lille: Presses Universitaires de Lille.
­­­­­ (1991) "Un project d'archeologie urbaine a Tell El­Dab'a" at pp. 115­125 of CITES DISPARUES, ed. Michael Baud, Serie "Monde", NS No.55, Paris: Autrement, September.

Bosworth, Andrew (1995) THE MARCH OF CITIES, Ph.D. dissertation, University of Washington.

Braidwood, Robert J. & Charles A. Reed (1957) "Achievment and early consequences of food production" at pp. 19­33 of COLD SPRING HARBOR SYMPOSIA ON QUANTITATIVE BIOLOGY, Vol.22.

Chandler, Tertius (1987) FOUR THOUSAND YEARS OF URBAN GROWTH: AN HISTORICAL CENSUS; Lewiston: St. Gavid's;

Chang, Kwang­chih (1988) ARCHAEOLOGY OF ANCIENT CHINA, New Haven: Yale University Press;

Chase­Dunn, Christopher and Alice Willard (1993) "Systems of Cities and World Systems: Settlement Size Hierarchies and Cycles of Political Centralization 2000 BC to 1988 AD" Annual Convention of the International Studies Association, Acapulco.

Darlington, C.D. (1969) THE EVOLUTION OF MAN AND SOCIETY, New York: Simon and Schuster;

Falkenstein, A. (1954) "La cite­temple summerienne" CAHIERS D'HISTOIRE MONDIALE, I:784­814.

Frankfort, Henri (1948) KINGSHIP AND THE GODS Chicago: Chicago University Press.

Gabriel, Richard (1991) FROM SUMER TO ROME; THE MILITARY CAPABILITIES OF ANCIENT ARMIES, New York: Greenwood.

Green, M.W. (1977) "A note on an archaic period geographical list from Warka" JOURNAL OF NEAR EASTERN Studies, 293­4.

Hallo, W.W. (1970) "Antediluvian cities" JOURNAL OF CUNEIFORM STUDIES, Vol.33:57­67;

Hamblin, Jane Dora (1977) THE FIRST CITIES, New York: Time­Life;

Huot, Jean­Louis, Jean­Paul Thalman and Dominique Valbelle (1990) NAISSANCE DES CITES, Paris: Nathan.

Jacobsen, Thorkild (1939) THE SUMERIAN KING LIST, Chicago:

The Oriental Institute;

­­­­­ (1957) "Early political development in Mesopotamia" ZEITSCHRIFT FUER ASSYRIOLOGIE, 91­140.

McEvedy, Colin (1967) PENGUIN ATLAS OF ANCIENT HISTORY, New York: Penguin Books;

Maisels, Charles Keith (1990) THE EMERGENCE OF CIVILIZATION; London: Routledge;

Mallowan, Max (1970) "The development of cities from Al­U'baid to the end of Uruk 5" in THE CAMBRIDGE ANCIENT HISTORY, Third edition, Vol.1, Pt.1, 327­462.

­­­­­ (1968) "The early dynastic period in Mesopotamia" CAMBRIDGE ANCIENT HISTORY, rev.ed., Vol.I & II.

Michalowski, Piotr (1983) "History as Charter: Some observations on the Sumerian King List" JOURNAL, AMERICAN ORIENTAL SOCIETY, 237­248.

Modelski, George and William R. Thompson (1996) "Evolutionary pulsations in the world economy". Paper presented to the International Studies Association annual convention, San Diego, April;

Nissen, Hans. J. (1993) "The context of the emergence of writing in Mesopotamia and Iran" in John Curtis ed. EARLY MESOPOTAMIA AND IRAN: CONTACT AND CONFLICT 3500­1600 BC, London: British Museum

Parpola, Aska (1994) DECIPHERING THE INDUS SCRIPT, Cambridge: Cambridge University Press.

Pettinato, Giovanni (1981) THE ARCHIVES OF EBLA: AN EMPIRE INSCRIBED IN CLAY; Garden City NY: Doubleday.

Postgate, J.N. (1992) EARLY MESOPOTAMIA: SOCIETY AND ECONOMY AT THE DAWN OF HISTORY, London: Routledge.

Pritchard, James B. (1959) THE ANCIENT NEAR EAST: AN ANTHOLOGY OF TEXTS AND PICTURES, Princeton: Princeton University Press.

Roaf, Michael (1991) ATLAS DE LA MESOPOTAMIE, Paris: Brepols.

Roux, Georges (1995) LA MESOPOTAMIE, (rev.ed.) Paris: Editions du Seuil.

Weiss, Harvey (1986) THE ORIGIN OF CITIES IN DRY­FARMING SYRIA AND MESOPOTAMIA IN THE FOURTH MILLENNIUM BC. Guildford CN: Four Quarters.
­­­­­ et al. (1993) The genesis and collapse of third millennium Mesopotamian civilization SCIENCE 261:995­1004.

Whitehouse, Ruth (1977) THE FIRST CITIES, Oxford: Phaidon
­­­­­ ed. (1983) MACMILLAN DICTIONARY OF ARCHAEOLOGY, London.

­­­­­ and John Wilkins (1987) L'AUBE DES CIVILIZATIONS, Paris: Bordas.

Wilkinson, David (1992­3) "Cities, Civilizations, and Oikumenes" COMPARATIVE CIVILIZATIONS REVIEW, Nos 27,28, Fall and Spring, pp.51­87 and 41­72.

Wilson, J.A. (1960) "Egypt through the New Kingdom: Civilization without cities" at pp. 124­136 of CITY INVINCIBLE, ed. C.H. Kaeling and R. Mc Adams, Chicago: Chicago University Press.

Woolley, Leonard (1954) EXCAVATIONS AT UR, London: Ernest Benn.

Wright, Henry T. (1981) "The Southern Margins of Sumer" at pp. 295­345 of Richard McC. Adams HEARTLAND OF CITIES, Chicago: Chicago University Press.

Yadin, Yigael (1963) THE ART OF WARFARE IN BIBLICAL LANDS, IN THE LIGHT OF ARCHAEOLOGICAL STUDY, Vol.I, New York: McGraw Hill.

Sources: Whitehouse 1977:48; Roux 1995:156; McEvedy 1967:26­7; Chandler 1987:468 (cities rank­ordered by population size).

Sources: For Sumerian cities: Adams 1981, esp. Tables 7 & 14, and Wright in Adams 1981; Northern Mesopotamia and beyond: Weiss 1993, Algaze 1993, Mallowan 1968; Roaf 1991;

* Roux 1995:492; ** Nissen 1993:56.

Unless otherwise noted, based on Encyclopedia Britannica 15th edition (1975, 1994) (EB), and the Macmillan Dictionary of Archaeology (Whitehouse 1983). Arranged in the order in which they appear in Table 3.

ERIDU: Founded on sand dunes probably in the 5th millennium BC. Mallowan (1970:231) describes it, in the Ubaid period, as an "unusually large city" of an area of approx. 20­25 acres, with a population of "not less than 4000 souls". Wright (1981:324­5) reports 12 ha settled area in Late Ubaid, and 4+40 ha in Early Uruk. The first of the five antediluvian cities cited in the Myth of the Deluge (Pritchard 1958:29), also in the Sumerian King List, as the first to exercise "kingship" (Jacobsen 1939:71). "Eridu was for all practical purposes abandoned after the Ubaid period" (Jacobsen 1957:98); after the Early Uruk period (Wright 1981:25). Whitehouse (1977:48) calls it a major Early Dynastic city. Massive Early Dynastic II palace (100 m in each direction) partially excavated there (Adams 1966:142).

BAD­TIBIRA Second of the five antediluvian cities cited in the Sumerian Myth of the Deluge (Pritchard 1958:29); in the Sumerian King List, as second to exercise "kingship" (Jacobsen 1939:71­2).

Known as the "fortress (or canal) of metal­workers" (Hallo 1970:65). Estimated area ­3100­2500 is ca.25 ha but with a large margin of uncertainty (Adams 21981: 105).

LARAK Third of the five antediluvian cities cited in the Sumerian Myth of the Deluge (Pritchard 1958:29); also in the same order in the Sumerian King List as exercising "kingship"(Jacobsen 1939:75). Important site No.1306, at al­Hayyad, probably on the left bank of the old Tigris, totally abandoned from ED III on (Adams 1981:348).

SIPPAR At early junction of Euphrates and Tigris. The fourth of five antediluvian cities in the Myth of the Deluge (Pritchard 1958:29); in the Sumerian King List, the fourth to exercise "kingship" before the Deluge (Jacobsen 1939:75). "City of bronze" (Hallo 1970:65). "Ancient, long­lived, but fairly modest town" (Adams 1981); never a dynastic capital but important religious and trading center. Studied by R. Harris (in Maisels 1990:182) for the period ­1894­1595; sacked by Elamite king in ­1174 (EB IX:235).

SHURUPPAK Remains found from late Ubaid period to 3rd dynasty of Ur (2112­2004); particularly important remains of Early Dynastic period. The last of five antediluvian cities cited in the Myth of the Deluge (Pritchard 1958:29); the last to exercise "kingship" before the Flood (Jacobsen 1939:77). On the ancient course of the Euphrates, scene of the Deluge in the Epic of Gilgamesh (Tablet XI, Pritchard 1958:66). Ca. 2600, had 600­700 soldiers on a full­time basis, for a population of 30­35,000 (Gabriel 1991:5).

ADAB Important up to the reign of Ur­Nammu, ca.­2000; Lugal­anne­mundu in the king list, ca. ­2400. (EB I:72). "City of some importance" in EDI (Adams 1981:88); Adab region is largest urban concentration in the Akkadian era.

AKSHAK In the Diyala valley; ca. ­2500 conquered by King Eannatum of Lagash. (EB I:179).

KISH Seat of first postdiluvian "kingship" (Sumerian King List, Jacobsen 1939:77); first dynasty of Kish (2750­2660); excavated E.D. palace indicates early power of kings. Site estimated at 84 ha in ­2500 but possibly "substantially larger" (Adams 1981:88) King Mesilim arbitrated boundary dispute between Lagash and Umma ca. ­2550. Defeated by Gilgamesh of Uruk ca. 2660 (EB V:837); seized by Hamazi ca. ­2500, , and six years later by the King of Akshak (Roux 1995).

URUK "The largest single center" throughout the Uruk period (Adams 1981:71). "In the Uruk period the city of Uruk is probably the largest settlement in the entire world" (Postgate 1991:112). According to the Sumerian King List, postdiluvian "kingship" shifted from Kish to Uruk (Jacobsen 1939:85). Must have had 25,000 ­ 50,000 people at the end of the 4th millennium" on 230 ha of settled area , and 20 ha of non­domestic use; slightly less than 600 ha around ­2900­2800 (Nissen 1993:56). Earliest writing emerged here; in one text the invention of writing is attributed to Enmerkar, Lord of Uruk (Roaf 1991:84).

Adams (1981:172,162) marks Akkadian Uruk as showing no evidence of occupation in Table on p.172, but as 200+ political capital in map on p.162.

UR Founded in the 4th millennium; early occupation ended by flood. Remarkable lst Dynasty royal tombs excavated from 26th century. Third dynasty great ziggurat built ca. ­2100 (EB 18:1021­2). Walls enclosed 60 ha. Leonard Woolley, estimates the city's area, at ­2800, at 220 acres (ca.90 ha), and a population of 34,000, and at ­2000 (Greater Ur), at 1300 acres (ca.525 ha), and a population of 250,000; estimate not accepted by Wright 1981. Sacked by Elamites ­2004.

LAGASH Founded in the Ubaid period. Abandoned in the I and II Early Dynastic period (Jacobsen 1957:98). "Probably the largest in area of the third­millennium city states" (Maisels 1990:172 quoting Maekawa 1973). In Early dynastic period rulers called themselves "Lugal" (kings); Steele of Vultures celebrates victory of King Eannatum over Umma. Most brilliant period ca. ­2l25, in the Gutian period (EB V:989). One Ur III textile workshop on the southern edge of the city employed 6,000 workers, mostly women and children (Postgate 1991:115).

NIPPUR Sanctuary of the head of family of gods, Enlil; a center of religious learning and of political order but never a capital (Adams 1966:129). Reached maximum extent ca. ­2500, and was fortified; eastern scribal quarter source of numerous tablets.

UMMA ED I site estimate has "large margin of uncertainty" (Adams 1981). Upstream from Lagash, interferes with its water supplies. Before ­2300, Rimus of Akkad defeats Umma and Ki­Dingir, reports 8900 men killed, 3000 captive and 3000 massacred (Jacobsen in Adams 1981:xiv); Rimus defeats Umma and Ur, with 8040 killed, and 5460 captives (Postgate 1991:76).

ISIN Seat of independent dynasty after 2020, continues into Middle Babylonian period (ca.­1000).

KESH (Tell al­Willaya) on the Tigris, "must have been a major city in the EDI/II period" (Adams 1981:88,172.

LARSA Emerges in Early Dynastic period; most prosperous when independent dynasty launched by Naplanum ca. ­2025, competing with Isin, Assur, and Eshnunna (EB VI:54). Canals blocked by Babylon ca. ­1800 (Adams 1981;134).

AKAD (Agade) Built by Sargon as capital city (after ­2350); destroyed at the end of the dynasty (ca.2150). According to inscriptions "5,400 warriors ate daily bread" before Sargon, presumably at Akad (Postgate 1991:41).

ASSUR (Ashur) Site (of about 60 ha, Frankfort 1948:396) originally occupied ca.­2500 as trading settlement; became religious capital of Assyrians; oldest excavated ziggurat dates from ca.­1800. Inner city protected by walls 4 km long (EB I:581­2).

MARI Wealthy and prosperous city. Excavations uncover remains from ­3100 to 700. Competes with Ebla ca. ­2500; exceptionally prosperous ca. ­1800; destroyed by Babylon's Hammurabi. (EB VI:616).

EBLA (Tell Mardikh) A North Syrian commercial transit and wool exporting center whose royal archive was discovered in 1975. Flourished ca. ­2500; when Greater Ebla controlled some 250 towns and villages (Pettinato 1981:136) but apparently lacked a standing army. Destroyed by fire, probably by Naram­Sin of Akkad ca.­2240. About ­2000, rebuilt by Amorites, with limited prosperity until finally destroyed in the great upheavals ca.1650­1600. Algaze (1993:108) mentions Ebla as having a site of 56 ha.

TELL BRAK in the Habur Valley, Northern Mesopotamia. Late Uruk ruins, inc. temple cover an area of 110 ha (Roaf 1991); a palace built by Nara­Sin , ca. 2250, destroyed by the Gutians (Roux 1995:187); reconstructed by Ur­Nammu of Ur III, ca. 2100. Close to center of the Mittani empire (ca.­1400).

ANSHAN (Tall­i Malyan) city of Elam; at onset of third millennium a wall protected an area of 200 ha, of which 1/4 was permanently occupied; had workshops for processing materials, inc. lapis lazuli from North Afghanistan. (Roaf 1991)

SHAKHR­I SOKHTA A third millennium valley town in Iranian Baluchistan, south of Zabol, with Proto­Elamite script, notable for stone working, alabaster, and lapis lazuli trade. In 2nd millennium, river Helmand changed its course and the city disappeared.

MOHENJO­DARO and HARAPPA: Allchin (1975:341) cites two independent (but undated) population estimates, and we add a third, in Whitehouse (1983):

H.T. Lambrick W.A. Fairservis Whitehouse

for Mohenjo­daro 35,000 41,250 40,000

Harappa app.35,000 23,500 20­25,000

Waddington (1969:89), citing Piggott, gives Mohenjo­daro 600 acres (243 ha) in ­2000. Whitehouse (1983) estimates the site at 100+ ha, and that of Harappa at +43 ha

ABYDOS Prominent sacred city near Thinnis, Upper Egypt, houses tombs of 1st dynasties; from 5th dynasty onward center of the cult of Osiris (EB I:41).

HELIOPOLIS Seat of the great temple of Re, second only to Amon at Thebes. Peak importance during 5th dynasty (ca.­2400) when Re became object of state cult. (EB IV:1001).

HERACLEOPOLIS Capital of the 9th and 10th dynasties in middle Egypt (ca.­2150). (EB IV:1036).

HIERAKONPOLIS (Nekhen, nr. Aswan) major importance ­3400­2700, took part in wars of unification. Narmer palette found there; continued as religious/historical center.

MEMPHIS Founded by Menes ca.­3100; capital of Egypt during the Old Kingdom (ca.­2686­2160). According to Bietak (1977) the site measures 2.5 x 1.5 km; Valbelle (1990) puts it at "close to 200 ha". The most extensive pyramid town may be found at Giza (about 18 km from Memphis), site of three great pyramids and the Sphinx that date from about ­2600.

COPTOS (Qift) Nearby gold and porphyry mines worked in 1st and 2nd dynasties. Cult of goddesses Min and Isis (EB VIII:333).

THEBES (Luxor) In the Middle Kingdom 11th dynasty (­2133­1991) was the capital of Egypt; earliest monuments date from that period. Greatest prosperity after about ­1560, peaked about ­1400. In the New Kingdom (­1780­1085), Thebes was "at least" 8 km.sq. (Bietak 1977).

EHR­LI­T'ON south of the Yellow River, in central Honan. Oldest Chinese city known so far, first discovered in 1957, on a site of some 350 ha, with two palatial buildings (Chang 1986:309).