GaWC Research Bulletin 464

GaWC logo
  Gateways into GaWC


City Generics: Ancient-Mesopotamian and Modern-Global City Networks Compared

P.J. Taylor*


City generics is a research choice to explore continuities of urban process across time and space. It directly addresses the portability of models between different societal contexts. A nuanced approach respects such differences. This is tested and illustrated through the application of the interlocking network model to extremely different social structures – the contemporary world-economy and the ancient Mesopotamian world-empire. In both cases data is collected and analysed to reveal a city network at the centre of societal reproduction; advanced producer services are enabling of corporate globalization through cities; advanced deity services are enabling of spiritual-political change through cities. Both analyses produce credible geographical results so that similarities and differences in results can be discussed as generic inter-city relations.


“Down every road there’s always one more city”

- Merle Haggard (

Introduction: The Choice Array in Analysing Cities

Amin and Thrift’s (2017) recent notable contribution to urban studies is entitled Seeing like a city, a form of wording for entree into social inquiry previously adopted by Scott (1998) in his Seeing like a state. I query their adaptation from state to city. For Scott, the state is a political institution with an executive decision making capacity that operates through simplification of social complexity for policy implementation. The city can be viewed as an administrative area with a local government but its executive functions are so much less than ‘the kind of problem a city is’ – organized complexity (Jacobs 1961). Cities interpenetrate each other in much more complex and sophisticated relations than in international relations, so much so that inter-city relations incorporate complementarities in myriad ways. Cities are the epitome of social complexity; to conceive of a city ‘seeing’ is quite limiting. The way I will approach the urban in this paper is as ‘being like a city amongst other cities’.

This is a well-worn notion. It appears in many guises: from Harris and Ullman (1945), through Berry (1965) and including Amin and Thrift (1992) in an earlier work, Bathhelt et al (2004) and as world city network (Taylor 2004) which is derived in part from Jacobs’ (1969) urban theory of economic development. The latter is particularly noteworthy because she integrates agglomeration processes within cities with connections between cities. This combination can be viewed as a single process of ‘promiscuous agglomerations’ at the core of the economic dynamisms of cities (Taylor 2017). In this context, focus on the external relations of cities in numerous ‘GaWC’ (Globalization and World Cities research network) publications has elicited criticism concerning the portability of urban models (Taylor and Derudder 2016). In this paper I will address this question by applying inter-city modelling in two very different social worlds.

Hoyler and Harrison (2017) have reviewed the recent debates on the veracity of urban theory derived from global/world cities research. Criticisms are both empirical and ideological. The basic postcolonial critique is that ideas derived from analyses of cities in the global North have been transplanted into the global South so that cities in the latter are judged by criteria alien to them. Not only are these two ‘global’ worlds very different, viewing them from just this one vantage point merely reflects global power differentials by representing the less powerful as inadequate versions of the more powerful. I do not want to debate further these profound differences but rather I wish to set the debate within a framework of reasonable research choices. I write in the spirit of Hoyler and Harrison’s (2017) plea, with regard to world/global cities, that ‘we should not throw out the baby with the bathwater in the pursuit of new urban theories’. 

There are two fundamental choices scholars make when they study cities. First, every city is constituted as a cacophony of cultural, social, political and economic processes interwoven by myriad urban agents that are ceaselessly changing the city. Scholars intervene in this complex whole by choosing a combination of process and agency to investigate in detail. In the GaWC approach the process investigated has been the advanced producer servicing of global capital and the agents are the financial, professional and creative firms that provide this service. This builds upon Sassen’s (1991/2001) original description of global city formation in New York, London and Tokyo and extends it to hundreds of other cities worldwide. This is one process amongst the multiple upon multiple of processes that constitute Sassen’s three cities and all the other cities in the GaWC extension into a world city network. Sassen chose this process and her three cities because they represented a particularly acute urban change in leading cities at the time of her study. Hence the label, and book title, Global City for, as she saw it, a new type of city evolving in response to economic globalization. But it remains a process, albeit, for her, one to be found in only certain cities. Extended beyond the small coterie Sassen thought of as global cities, this same process, broadly servicing global capital, can be found in all cities across the world: corporate globalization is not something a city can opt out of. But this does not mean cities across the world are thereby being labelled global or world cities. The adjective ‘world’ in world city network refers to the overall framing of the city network - a network in a corporate capitalist world – not to the individual cities in the network. Thus because the process Sassen analyses is a dominant process in her three cities, labelling them ‘global cities’ is meaningful, but for their country cousins Cincinnati, Nottingham and Nagoya labelling them global or world is misleading; certainly they are each part of the world city network (servicing global capital occurs there) but this is a much lesser part of their changing being than it is for New York, London and Tokyo.

Second, as Jacobs (1969, 35) tells us ‘a city always seems to have implied a group of cities’ and therefore there is a choice to be made as to which and how many cities to study. This can be viewed as a choice array ranging from focus on a single city to studying all cities in the group. Every city is unique and there is a large fiction literature that invokes the particular nature of a book’s city context. In social science urban research most studies are also at this end of the scale. Cities can also be studied singly but with critical reference to its external relations (e.g. Cronon 1991). Comparative analyses search out similarities and differences between a small number of cities to draw wider conclusions than possible for just one city. There are also literatures on regional or epochal types of cities that are deemed to have distinct features in common such as ‘American City’, ‘Asian City’, etc. or ‘Medieval City’, Classical City, etc. and mixtures of the two as in ‘Colonial City’ or “Industrial City”. The grossest example of this is the distinction between the ‘Modern City’ and the ‘Pre-modern City’, which can involve both space and time differentiations. At the far end of this choice array are studies that attempt to define the nature of cities in general. The very fact that we have the word and concept of ‘city’ in multiple languages across time and space means that there is a sense of cities in general being identifiable. It is a generic approach to cities that I employ in this paper.

The study of city generics has to be hugely dependent on abstraction, the peeling back of layers of meaning in attempts to find ‘essential’ aspects of cities. This is inevitably challenging and challenged ranging from Amin and Thrift’s (2017) ‘seeing’ and ‘thinking’ city to Batty’s (2013) ‘new science of cities’. The latter involves theoretical modelling generating quantitative empirics that can be generally applied to cities across time and space. This is the approach I employ but with a particular nuance that respects basic societal differences.

To explain my deviation from typical quantitative abstraction I will use the common research example of the rank size distribution of cities whereby the population sizes of cities are arrayed as ratio measures (numbers of people) against ordinal measures (ranking these numbers) to elicit an inter-city relation implying a sorting process. In practice this has produced a research path that contrasts the ‘rank-size rule’ of complex sorting with ‘primate city’ situations where the potential sorting has been delayed or distorted (Berry 1961). In a large-scale historical study of this process Chase Dunn has carried out such research on a wide range of regions from 2000BC (Chase-Dunn and Willard 1993; Chase-Dunn and Hall 1995) using a ‘primacy index’ whereby the rank-size rule hierarchy scores zero and values above this indicate increasing single city primacy. In fact the 60 results to AD1400 show 15 cases with positive scores (degrees of primacy) but 45 cases of negative scores, which simply do not feature in the rank-size versus primacy way of thinking (Taylor 2013, 199). What these negative scores show is a more network pattern of cities eschewing both rank-size and primate hierarchical possibilities. This probably reflects most regions including several political units each with their favoured ‘imperial capitals’ distorting any demographic sorting process in the opposite direction to a singular imperial ‘primate’ distortion. At the very least this example suggests the abstract modelling should take heed of the different political contexts within which cities are developing. But perhaps this lessen should go further. In this paper I will show how abstract modelling can be adapted to apply to different ways in which cities function in very dissimilar societies. Such a nuanced abstraction approach is explored through applying the interlocking network model to both cities in the society for which it was devised, corporate globalization, and to cities in a totally different society, the city-deity relations of ancient Mesopotamia. 

The argument is presented in four parts. First, the interlocking network model is presented in general terms as a process of multi-city practices by dominant city agencies. The latter are network builders, they interlock cities  through their work. Second, the model is applied to cities in corporate globalization; a new analysis of advanced producer service firms (‘the interlockers’) is presented that uses composite data on offices for 2000 and 2016 to describe the modern global city network. Third, the model is applied in a totally new context, a world-empire structure in which cities are equated with deities; a unique analysis of associations of deity cults (‘the interlockers’) that uses data on temple names to describe the ancient Mesopotamian city network. The fourth part compares the two sets of modelling results to indicate the communalities and dissimilarities evident in this city generics exercise. A short conclusion invites debate on the utility of this approach for urban studies.  

Inter-City Relations: The Interlocking Network Model

The traditional way of treating the external relations of cities is through central place theory (Christaller 1933/1966). His model of interlocking hinterlands has provided a basic hierarchical ordering of cities that was later developed as national urban systems, which at one time constituted a major research focus in urban studies (Bourne and Simmons, 1978). This model proved to be very portable and has been widely used beyond its modern ‘western’ origins (e.g. Skinner 1964; Renfrew 1975; Hodder and Orton 1976). But this common city generics is not the only way cities relate to one another. As Jacobs (1969, 35) tells it: ‘A city does not grow by trading only with a rural hinterland’. As a complement to central place theory’s focus on local links and hierarchies there is a central flow theory that focuses on non-local inter-city relations, horizontal links rather than hierarchical ones (Taylor et al 2010). This is based upon an interlocking network model; it is the generics of this model that is being assessed in this paper.

Most social networks have just two levels of operation, the network level where the net is formed (e.g. a juvenile gang) and the unit level of the network makers, the agents as nodes in the network (e.g. gang members). But in an interlocking network there are three levels due to the separation of agent and node. Thus there is an overall network level, a nodal level through which the network operates, and a sub-nodal level of agents, the network-makers who create the network through the nodes (Knoke and Kuklinski 1982). In the initial world city network analysis of contemporary inter-city relations this consisted of the network level of world-economy, the nodal level of cities and the sub-nodal level of advanced producer service firms with multiple offices across the cities. Thus the network of cities is created by firms working in cities; firms are said to interlock the cities through their everyday servicing work across their various city offices. To give a simple example: an Australian property company is buying up Brazilian land using German finance; its legal needs will likely be met by a London law firm with branches in Sydney, Sao Paulo and Frankfurt that can ensure contracts are compatible with Australian, Brazilian and German law, and are brought together as English common law, the international standard. Thus the work done is shared between London, Frankfurt, Sydney and Sao Paulo offices through flows of instruction, information and knowledge; thus the four cities are interlocked. The amalgamation of myriad such work projects creates an interlocking model that is the world city network. In this modelling the key item of interest is not the city per se, but rather city-dyads as measured by flows between them. For instance in the legal example the office in London is the key agent in the network making and therefore we might expect the London-Frankfurt, London-Sydney and London-Sao Paulo dyads to be more important than the Frankfurt-Sydney, Frankfurt-Sao Paulo and Sydney-Sao Paulo dyads. It is the purpose of interlocking network analysis to estimate these inter-city flows and thereby describe the world city network.

The interlocking network model was developed using advanced producer servicing because it was part of a research programme extending Sassen’s (1991/2004) global city formation to a worldwide network process (Taylor 2001). However, the process can encompass the everyday work of other agents who operate through multiple city locations. Thus the model has been used to describe city networks produced by media firms, architecture/engineering firms, UN agencies, diplomatic missions, NGOs, and maritime producer service firms (Taylor 2005; Derudder and Taylor 2017). It follows that the model can be stated generally, like central place theory, in terms of cities as service centres, but not for servicing their hinterlands rather for providing inter-city servicing. As such, the model is specified as follows. Let there be m service providers and n cities; for each provider, j, we know the importance of each city, i, in its office network denoted as its service value vij . This produces an m x n service values matrix: there are m columns each showing one service provider’s location strategy; and n rows each showing one city’s service mix.

The focus is on city-dyads: the basic relational link for each pair of cities, a and b, in terms of provider j is given by

            rab.j = vaj . vbj                                                                                        (1)

This product of the importance of the provider in the two cities is an estimate of the potential flows between the two cities through the provider’s work. From this we can derive estimates of the overall potential flows between the two cities by summing over all providers

            rab  = Σrab.j                                                                                           (2)

From equation (2) a new inter-city relational matrix, n x n, can be constructed, which is a numerical representation of the network.  Each city, a, is a member of n – 1 city-dyads so that the sum for all its dyads is a measure of the city’s gross connectivity within the matrix

            Ca = Σrai           (ai)                                                                          (3)

This gross connectivity measure is an indication of a city’s degree of integration within the network.

These three equations provide the fundamental building blocks of an interlocking network model.

An Interlocking Network Analysis of the Contemporary Global City Network

Corporate globalization is intrinsically an urban world-system. The global city network is a key functional constituent of this contemporary world-economy structure. The network formation is a response to demand for an enabling spatial framework through which ceaseless capitalist accumulation develops. Its leading enabling institutions are financial, professional and creative service firms that navigate the world of business for their clients across the world-system. They operate through ‘global cities’ as described by Sassen (1991/2004); it is the practice of these institutions by their multi-city connections that comprise the global city network. 

It is this process of servicing global capital by advanced producer service providers that has been described as the ‘world city network’ in a series of cross-sectional studies since 2000. Data are collected in a given year to create a service value matrix and the city network is produced for that year (e.g. Taylor (2004) for 2000 and Taylor and Derudder (2016) for 2012). In the analysis presented here I combine data over time to create a conglomerate city network for 2000-2016 that I term the global city network, broadly indicative of corporate globalization, to distinguish it from the specific single-year network analyses. The purpose of this change in methodology is to aid comparison with the Mesopotamian city network analysis that is perforce described over a broad time-span. In addition the size of the service values matrix is greatly reduced to bring it closer to the other analysis.

The new 2000-2016 service values matrix was created as follows. The connectivity rankings of the leading cities for 2000 and 2016 were combined and the top 35 cities selected (to compare with 34 cities found in the Mesopotamian data analysis). Selecting the advanced producer service firms was less straightforward. Over the sixteen years there were many changes in firms due to amalgamations and some firms dropping out of the analysis between 2000 and 2016. In the event there were 41 firms where continuity could be found, albeit with some name changes through amalgamations. The new list consists of 11 firms in financial services, 10 in law, 8 in accountancy, 7 in advertising, and 5 in management consultancy. The resulting service values matrix of 35 cities x 41 firms was created by summing the original service values for 2000 and 2016. These initial service values ranged from 5 to 0 (5 for the city housing a firm’s headquarters, 4 to 1 indicating the importance of a city to the firm’s office network, and 0 indicating no presence of a firm in a city). Therefore in the new combined matrix the values range from 10 to 0, from a headquarters remaining in a city through to no presence continuing in a city.

In an interlocking city network analysis the initial significant result is the measure of gross network connectivity indicating the degree of a city’s integration into the network (equation (3) above). The results for the 2000-2016 matrix are shown in Table 1 where the top 20 cities are listed. This shows familiar features found in all previous cross-sectional analyses, notably the dominance of London and New York, followed by the quartet of Hong Kong, Singapore, Tokyo and Paris. The three main ‘globalization arenas’ are fairly evenly represented: Western Europe with 7 cities, Pacific Asia and North America with 5 cities each. The other three cities in Eastern Europe (Moscow), Australia (Sydney) and Latin America (Sao Paulo) are perhaps adjuncts to the three main arenas leaving South Asia, the Middle East and Africa outside this top 20. These results suggest a quite hierarchical global city network strongly centred on ‘NYLON’.

Table 1 Global City Network: Gross Network Connectivity (GNC)

This network structure is explored further in Table 2 that includes all 35 cities. A cluster link analysis has divided the global city network into two sub-nets that affirm the hierarchical nature of these results: sub-net I includes all top 20 cities from Table 1 except for Toronto, and with San Francisco and Bangkok added from the North America and Pacific Asia globalization arenas respectively. Thus sub-net II is made up of the lesser-connected cities in the analysis. But this is not a replication of the traditional core-periphery description of the capitalist world-economy: as well as Toronto, sub-net II includes 4 cities each from western Europe and Pacific Asia. Rather what we have uncovered here is the distinction between intensive globalization and extensive globalization previously identified in some cross-sectional analyses (for 2008 in Taylor et al (2011) and for 2012 in Taylor and Derudder (2016)). The differences are in the type of servicing as shown in Table 3. This shows the contribution of the different service sectors to each sub-net wherein financial services and management consultancy are roughly equally represented as percentages in each sub-net. However legal services are the most biased with more double the percentage in sub-net I compared to sub-net II. This concentration of advanced legal services is indicative of intensive globalization. In contrast both accountancy and advertising are disproportionally represented in sub-net II. In the case of accountancy this service includes very large firms that are ubiquitous across hundreds of cities worldwide. Advertising remains state-bound to some degree for cultural and national reasons; in this period these firms have important offices in all the main media centres (notably for national TV) of medium-sized states. This is clearly reflected in sub-net II and is indicative of extensive globalization.

Table 2 Global City Network: Cluster Link Analysis

Table 3 Advanced producer service sector contributions to clusters

The visualization of this network in Figure 1 confirms the network departure from the core-periphery spatial organization: sub-net II is encompassed by sub-net I. Intensive globalization and extensive globalization are not spatially distinct but rather show variation in integration into the global city network both within and without the main globalization arenas. The global network space shows the importance on inter-city links by width of line and it can be seen that links between the encompassing major cities are far more important than between the interior encompassed cities. These latter are not ‘off the map’ as critically described by Robinson (2002) but might be described coincidently both on the inside and at the edge of the network. This inside-edge location within a strongly hierarchical pattern is conformed in Table 4 where the main city-dyad connectivities from the figure are listed. The top twelve connected dyads have London-New York (NYLON, the only dyad with a name) definitely number one as expected with these two cities’ links to the quartet below them in Table 1 taking up the next eight positions before Frankfurt and Chicago make an entrance. The dominance of London and New York controls this dyad analysis, one or other making up one part of most dyads in the top 50. The nine exceptions led by Hong Kong-Singapore ranked 28th are also listed on Table 4. But it is still the same eight cities found in the top 12 dyads. It would take a very large table to include inner-edge city-dyads: the highest ranked such city-dyads are Toronto-Zurich ranked 416th and Jakarta-Kuala Lumpur ranked 450th.

Figure 1 Visualization of the global city network

City Codes: C1 Amsterdam; C2 Bangkok; C3 Beijing; C4 Brussels; C5 Buenos Aires; C6 Chicago; C7 Dublin; C8 Frankfurt; C9 Hong Kong; C10 Istanbul:  C11 Jakarta; C12 Johannesburg; C13 Kuala Lumpur; C14 London; C15 Los Angeles; C16 Madrid; C17 Melbourne; C18 Mexico City; C19 Milan; C20 Moscow; C21 Mumbai; C22 New York; C23 Paris; C24 San Francisco; C25 Sao Paulo; C26 Seoul; C27 Shanghai; C28 Singapore; C29 Stockholm; C30 Sydney; C31 Taipei; C32 Tokyo; C33 Toronto; C34 Washington; C35 Zurich

Table 4 Global City Network: City Dyad Connectivity (CDC)

Finally, it is important in interpreting these results to appreciate that London and New York are exceptional cities in the fullest sense of that term. The other cities in sub-net I are also ‘global cities’ in Sassen’s terms and are also extremely rare worldwide. Therefore none of these cities dominant in the corporate servicing process can be sensibly used as exemplary cases, city proto-types to which other cities might aspire. The city ‘norm’ in corporate globalization is the inside-edge city as found in sub-net II; they are only a minority in this analysis because of restriction to the size of the service values matrix used for this analysis.

An Interlocking Network Analysis of the Ancient Mesopotamian City Network

Ancient Mesopotamia was intrinsically an urban world-system. The Mesopotamian city network was a key functional constituent of this original world-empire structure. The network formation was a response to demand for an enabling spatial framework through which legitimate continuity of political authority developed. Its leading institutions were deity cults that through their temples and shrines navigate a spiritual world for their adherents across the world-system. They operated through ‘gods resident in cities’ as described by Leick (2001); it is the practice of these institutions by their multi-city connections that comprise the Mesopotamian city network. 

It can be debated as to whether the very first cities are to be found in Mesopotamia but, what is not in doubt, is that this is where a network of relatively large cities developed to create a deeply integrated social world, commonly understood to be the first ‘civilization’. As such its economic processes have been studied as a world-system, a commercial world of production, trade and import shifting (Algaze 2005a, 2005b). It is the related innovative surge, most notably economic stocktaking leading to the invention of writing and thereby leaving an historical record, which warrants its civilization appellation. Although these commercial practices are very important for understanding this society they do not constitute capitalism: commerce is generic; capitalism is specific (Sanderson 1995). The latter’s specificity is that commerce pervades all aspects of society as ceaseless capital accumulation; ultimately the ‘bottom line’ account rules, as in today’s corporate globalization. In ancient Mesopotamia commerce provides the material basis of the society but it does not define it. Here it is a specific cultural world of deities that is pervasive. Nevertheless cities are equally central in both world-systems.

The Mesopotamian origin myth is an urban story: gods come down to Earth as cities, the first is the god Enki at Eridu in Sumer where the first building is a temple, the god’s house. Humans are then created to provide for the needs of god and temple (Leick 2001, 2). But Eridu is not alone, it is one of five early cities - the others being Larak, Shuruppak, Sippar and Badtibira - creating a first network of cities. Jacobson (1971) emphasizes the cooperation and mutuality between the cities postulating a federated ‘league of cities’, further developed as the ‘Hexopolis of Sumer’, wherein Shuruppak is joined by Adab, Lagash, Nippur, Umma and Uruk  (Leick 2001, 76). But the fact that cities are gods means that the economic decline of a city does not thereby diminish its societal importance. Thus although Eridu never becomes a large city – Modelski ‘s (2003, 22) population estimates peak at 10,000 – the city ‘maintained the myth of primacy to the subsequent ages of Mesopotamian history’ (Leick 2001, 17). This illustrates the cultural pervasiveness within, as integral to, these city networks.

It may be that ‘Mesopotamian cities are always sacred’ (Leick 2001, 2) but this is not a static world; there are vital political changes that directly affect the cities and their networks. This means that the relationship between city and god is never straightforward (p. 146). Like any ideological system of thought, the Mesopotamian belief system ‘provided a basis for political manipulation’ (p. 147). As Thomas (2010, 143) tells it ‘what is apparent when reading various religious myths ... is the extent to which the realm of gods reflects earthly social structures … the relative status of god and city rose and fell in concert’. For instance, initially ‘there was no indication that any one city deity was fundamentally more important than another’ but as a political hierarchy emerged between cities, this was reflected in the gods: Enlil of Nippur was the first to emerge as supreme god (Leick 2001, 147-8). This was an effect of the conversion of federated commercial cities into competitive political city-states and the consequent warfare (Taylor 2013, 124-9). Deities are immortal and therefore, cities being gods meant that they could not be destroyed; this necessary respect for the fate of opponents meant that construction of a city-empire encompasses other Sumerian cities by warfare was problematic. Thus where one city became militarily dominant its imperium was relative benevolent using the title “King of Kish”; Kish was a city to the north, and crucially outside Sumer. The first true city-empire was created by Sargon of Akkad, and decisively, Akkad was a new city, for whom gods were newly required, and again was outside Sumer. Without claims to be an ancient sacred site – there were no gods tied to this place -  one of Sargon’s successors declared himself to be ‘god of Akkad’ seemingly endorsed by the old gods – Ishtar of Eanna, Enlil of Nippur, Dagan of Tuttel, Ninhursagar of Kish, Enki of Eridu, Sin of Ur, Shamash of Sippar and Nergal of Kutha are mentioned (p. 100) – but the later crisis (demise of the empire) of Akkad is subsequently explained as Enlil’s disapproval and the gods withdraw from the city (Leick 2001, 105).

I have presented this detail to show the importance of gods but also their variety across numerous locations. The gods were related to one another in pantheons as families and, in addition, gods had specific functions (fertility, weather, war, etc.). They were worshiped in a cult-like manner resulting in shrines and temples dedicated to particular deities’ functions. An important part of this process is the practise of syncretrism, whereby new gods took on the functions and esteem of old gods, an assimilation that retained respect for the old god while confirming the special sacredness of the new god and city. This is associated with the centre of power moving away from Sumerian origins of the gods: Enki becomes Ea (Leick 1991, 37) and Ianna becomes Istar (p. 86) to be newly associated with Akkad (p. 96). In contrast Babylon’s god Marduk is said to have usurped Enlil’s position of primacy but his cult had limited reach (p. 115). It is these geographical presences across multiple cities that can be used to describe the inter-city relations that constitute a Mesopotamian network of cities.

The data I use for this empirical exercise is from George’s (1993) collection and editing of cuneiform tablets on which are inscribed lists of temples. He has created a gazetteer of the ceremonial names of temples to which he has added indexes of divine names and geographical names. I have identified 902 cross-references of gods with cities from these indexes to produce a list 52 deities arrayed across 34 cities. A corner of the resulting 52 x 34 matrix showing frequencies of co-listings of cities with gods is displayed in Table 5. This is a service values matrix: the more listings the more divine services provided in a city. In the corner displayed it can be seen that none of the five listed gods were mentioned in a ceremonial temple name alongside the cities of Adab, Akkad and Bad-tibira (i.e. no service provided in these cities). But Assur and Babylon get lots of mentions in association with these gods. The highest frequency of co-listings relates Assur to Adad plus Iskur (doubly named because this is a case syncretism); they are mentioned together on temple names six times. But Babylon has the most co-listings overall in Table 5 being mentioned together with both Anu and Assaku five times. How can these numbers be interpreted? If, as Leick (2001, 146) tells us: ‘All Mesopotamian cities were thought to depend entirely on the benevolence and active protection of the city gods’ these frequencies indicate which gods are promoting the wellbeing of a city and to what degree. Thus we can say that Adad plus Iskur provides a very strong divine service for the city of Assur; more generally each god can be described as advanced divine service provider. Therefore the matrix, of which Table 5 is a part, is directly equivalent to the service values matrix used as input to the interlocking network model to describe the modern global city network in the previous section; on the ground think services provided in temples analogous to services provided in skyscrapers. Each row of the matrix shows the geographical distribution of a deity’s service of protection and prosperity; every column shows the specific mix of deity services in a city. And this spiritual urban work involved professional services deriving from organized training and education; scribes are the first archetypal city profession involving many specializations including law, languages, composition, rhetoric, mathematics as well as writing, with the most proficient becoming teachers in their turn (p. 163). Nippur in particular was renowned as a great learning centre (p. 159-2); scribes in Akkad have even been referred to as ‘spin doctors’ for their efforts in linking their rulers to gods (p. 88). For the purposes of this analysis, advanced divine services and advanced producer services are treated as analogous between two networks of cities in very different worlds.

Table 5 Frequencies of associations between five cities and five deities in temple names

Table 6 shows the top 20 Mesopotamian cities in terms of gross network connectivity. There are three particular features that stand out. First, there is a wide geographical scope from Sumerian cities in the south to upper Mesopotamian cities (Assur, Nineveh, and Mari) in the north. Second, the spiritual perseverance of the early Sumer cities is attested by the presence of Eridu and four of the Hexopolis (Uruk, Nippur, Lagash and Umma). Third, the politics of empire formation is strongly present as imperial capitals (Akkad, Ur, Isin and Babylon) become increasingly important as the size of the empires rise (Taagepera 1978); Babylon ranked number one is the first great imperial capital city. These results are eminently credible, certainly the top five cities are predictable; Zabalam ranked a clear sixth is perhaps more interesting since it features much less in the literature than many cities ranked below it.

Table 6 Mesopotamian City Network: Gross Network Connectivity (GNC)

The cluster analysis of the links that create the connectivity is shown in Table 7 where two sub-nets are revealed. Sub-net I is the prime cluster, its includes the grand tradition of Mesopotamian culture of gods successively operating through the top seven cities listed from Eridu to Babylon and encompassing the three languages of Sumerian, Akkadian and finally Babylonian.  With sub-net II we find Zabalam coming to the fore again as the leading city of this cluster. This is a surprise; I cannot find any suggestion of this division of Mesopotamian deity culture; this requires careful, cautious, scrutiny.

Table 7 Mesopotamian City Network: Cluster Link Analysis

Sub-net II largely consists of less important Sumerian cities but also includes long distance trading links in the north, Nineveh and Mari, the east, Susa, and the south, Tilmun, the latter two both geographically outside Mesopotamia. The advanced divine services provided are mainly through the major god Inanna and a minor god Ningirsu. The geographical distribution of these two gods’ services are shown in Table 8. Inanna has homes in many major cities – Leick (2001) records Inanna of Assur (p. 197), Inanna of Nineveh (p. 194), Inanna of Uruk (p. 146) and reports she is also very important to Ur (p. 132), Nippur (p. 157) and Babylon (p. 248). This is reflected in the first column of Table 8 (a) where the frequency of Inanna services prominently includes all these cities of the grand tradition. But if we focus on the second column, where Inanna services are given as a percentage of all deity services available in a city, a very different picture arises: here we see subset II. These are cities that in relative terms specialise in Inanna services, in five cases constituting half or more services in a city due to this one god, culminating with Zabalam where she has a 90% near monopoly of deity services. Notice, however, that Girsu, Sirara and Sippar do not feature in Table 8 (a); they are brought into the sub-net via the god Ningursa (literally god of Girsu) through the link to Lagas, effectively Girsu’s ‘twin city’ (Table 8 (b)). This interpretation is confirmed when we turn to the ranking of the most connected city-dyads in the two sub-nets in Table 9. For sub-net I the top dyad links the first two cities ranked in Table 5, Babylon-Uruk, with other leading cities prominently featured at lower ranks (Assur, Nippur, Ur) in a dense interlocking pattern. This is not the pattern for the other smaller sub-net. In sub-net II the leading dyad is Girsu-Lagash, after which Zabalam dominates. This is a much simpler specialist net structure based on Zabalam as leading city and Lagas the strategic link city.

Table 8 Unravelling Sub-net B

Table 9 Mesopotamian City Network: City Dyad Connectivity (CDC) by Clusters

Finally, this Mesopotamian city network based upon deity servicing can be portrayed visually (Figure 2). This diagram locates cities in relation to the strength of their relations to other cities. In this case cities are shown by size of their gross network connectivity and the two sub-nets are identified. What we find is a core zone of leading cities representing the Grand Tradition sub-net including minor cities below, and the Specialist sub-net located separately above.

Figure 2 Mesopotamian City Network: Network Visualization

City Codes: C1 Adab; C2 Akkad; C3 Assur; C4 Babylon; C5 Bad-tibira; C6 Borsippa; C7 Dur-Kurigalzu; C8 Eridu; C9 Esunna; C10 Girsu:  C11 Hursag-kalamma; C12 Ilip/Urum; C13 Isin; C14 Kalah; C15 Karkara; C16 Kes; C17 Kis; C18 Kutha; C19 Lagas; C20 Larak; C21 Larsa; C22 Marad; C23 Mari; C24 Nineveh; C25 Nippur; C26 Sippar; C27 Sirara (Nina); C28 Susa; C29 Tilmun; C30 Umma; C31 Ur; C32 Uru-ku; C33 Uruk; C34 Zabalam

Finally, the credibility of these unfamiliar quantitative analyses needs to be addressed. Clearly the modelling is based on far less information than for the modern city network (compare differences in sizes of connectivity scores between Table 1 and Table 5). Nevertheless the number of city-deity associations in the service values matrix is 617, a more than respectable quantity for the type a simple analyses conducted. However there is the additional question of potential bias in the data – how representative is it of city-deity relations? Whereas for the modern network analysis cities and service providers were selected; in this case the data used was what was available in George’s (1993) gazetteer of temple names. This in turn reflects tablet collection and publication; results seem credible for Table 5 where results are as reasonably expected but with surprises – Table 2? – it may be more an outcome of serendipity that is not representative. But I will proceed with the discussion on the assumption that this interlocking network analysis has produced moderately sound results to make comparison with the other city network analysis worthwhile.

Contrasts and Comparisons

This paper has purposively dealt with extremely different worlds to illustrate its city generics argument; in Wallerstein’s (1976) terms contrasting world-system structures, a world-empire and a world-economy. Reverting to more traditional language, by terming both cases civilizations cities are brought to the fore: civilization and city have a shared etymology; civilizations are societies having cities. Linking this to Jacobs’ (1969) previously noted observation that cities are always being related ‘in groups’, city networks can be interpreted as the necessary organizational hubs of civilizations where their formation and development are enabled. Thus the global city network and the Mesopotamian city network are the dynamic geographical frames upon which their respective societal formations cohere. The contrast is in their coherence, materialist against spiritual.

The first point to make in comparing the two interlocking network analyses is that the service values matrix of the global network is very much denser than the Mesopotamian matrix. For instance, zeros are quite rare in the fist matrix (i.e. firms with no presence in a city) but are quite common in the second matrix (i.e. deities with no presence in a city). This should be treated as differences in data availability - far more data points for the global network – rather than suggesting the Mesopotamian city network was itself less cohesive to that society. But the analyses do suggest some interesting similarities and differences:

  • Both city networks exhibit strong hierarchical tendencies. Table 10 shows the top twenty cities in each city network in relative terms (as percentages of the connectivity of the first ranked city). This shows quite different profiles. The global city network exhibits a clear duopoly at the top (NYLON) and although Uruk is a distinct second rank, Babylon is firmly separate as dominant city in Mesopotamia. But the main difference is in the rates of decline in connectivity, which is far steeper for Mesopotamia: all global network cities in Table 10 have at least 50% connectivity, but this is only true of five Mesopotamian cities. It seems that advanced deity services generate stronger hierarchical tendencies than advanced producer services. This reflects their different provenances: as previously noted deity services are tied into political changes where imperial capitals can ‘collect’ gods to reflect their temporary prowess; Babylon as the first great imperial capital epitomises this political process.

  • Both city networks consist of two link clusters, one of which represents the main dimension of the network in the sense of including the leading cities in terms of connectivity. In the case of the global network this is intensive globalization with the strongest producer service provisions; in Mesopotamia this is the grand tradition of major cities featuring all the leading deities.

  • The second link clusters are very different. In the case of the global city network it is the familiar extensive globalization wherein the broader producer services dominate less well-connected cities. In contrast in Mesopotamia the second link cluster consists of a narrow specialist provision of deity services. However neither seems to represent an independent process but they relate to the first link clusters in different ways. Extensive globalization is completely encompassed indicating that corporate globalization is all pervasive throughout the city network. The Mesopotamian specialist deity services includes well-connected cities (Zabalam ranked 6th and Nineveh ranked 9th in Table 6) and appears separate in the network space (Figure 2) but neither indicates an independent process given that its specialism (Inanna services) is actually pervasive across both link clusters (Table 8): it is an off-shoot of the main tradition.

Table 10 Comparing Gross Network Connectivities

Despite their world-economy and world-empire structures being vastly different, the prime enabling processes of their respective societal development - advanced producer services and advanced deity services respectively – show comparable city network organization as a generic outcome of civilization.

Concluding Comments

I began this paper with a discussion of the choice array open to urban studies scholars resulting in a range of generalities in research findings. The framing of the argument as an array implies that there is not a simple right or wrong way for researchers to choose. Different approaches generate different types of knowledge that will be useful for some scholarly purposes but not for others. Every approach has its particular advantages but also is subject to serious pitfalls. For instance single city studies provide depth of understanding but suffer from concern for how representative is that understanding. City generics promotes a broad understanding of cities but it is particularly susceptible to queries about the portability of urban models.

The key step in tackling portability is to respect variation in urbanizations so that differences are incorporated into the modelling and not simply ignored. This is what has been attempted above. The city genetics has not been about simply applying a model devised in one context and then imposing it on another. Rather data customised for the specification of the model has been used in both cases – collecting information on firms servicing reproduction of global capital and on temples servicing reproduction of spiritual/ political empires. In both cases a description of city network formation has provided a geographical framing of how critical work in cities enables societal development.


Algaze, G. (2005a) The Sumerian take-off. Structure and Dynamics 1 (1), Article 2 (http://repositories.cdlib.or/imbs/socdyn/sideas/vol1/iss/art).
Algaze, G. (2005b) The Uruk World-System. University of Chicago Press, Chicago.

Amin, A. and N. Thrift (1992) Neo-Marshalliann nodes in global networks. International Journal of Urban and Regional Research, 16, 571-81.

Amin, A. and N. Thrift (2017) Seeing like a city. Polity, Cambridge, UK.

Bathelt, H., A. Malmberg and P. Maskell (2004) Clusters and knowledge; local buzz, global pipelines and the process of knowledge creation. Progress in Human Geography 28, 31-56.

Batty, M. (2013) The new science of cities. MIT Press, Cambridge, MA.

Berry, B. J. L. (1961) City size distributions and economic development. Economic Development and Cultural Change 9, 573-88.

Berry, B. J. L. (1965) Cities as systems in systems of cities. Papers of the Regional Science Association 13, 147-63.

Bourne, L. S. and J. W. Simmons (eds) (1978) Systems of Cities. Oxford University Press, New York.

Chase-Dunn, C. and T. D. Hall (1995) Cross world-system comparisons. In S. K. Sanderson (ed.), Civilizations and world-systems. AltaMira, Walnut Creek, CA.

Chase-Dunn, C. and A. Willard (1993) Systems of cities and world-systems: settlement size hierarchies and cycles of political centralization, 2000BC – 1988 AD. (

Christaller, W. (1933/1966) Central places in southern Germany. Prentice Hall, Englewood Cliffs, NJ.

Cronon, W. (1991) Nature’s metropolis: Chicago and the Great West. W. W. Norton, New York.

George, A. R. (1993) House most high: the temples of ancient Mesopotamia. Eisenbrauns, Winona Lake, IN.

Harris, N. and E. L. Ullman (1945) The nature of cities. Annals of the American Academy of Political and Social Science 242, 7-17.

Hodder, I. and C. Orton (1976) Spatial analysis in archaeology. Cambridge University Press, Cambridge, UK.

Hoyler, M. and J. Harrison (2017) Global cities research and urban theory making. Environment and Planning A 49, 2853-8

Jacobs, J. (1961) The Death and Life of Great American Cities. Random House, New York.

Jacobs, J. (1969) The economy of cities. Vintage, New York.

Jacobsen, T. (1970) Towards the image of Tammuz. Harvard University Press, Cambridge, MA.

Knoke, S. and J. H. Kuklinski (1982) Network analysis. Sage, Beverley Hills, CA.

Leick, G. (1991) A dictionary of ancient near eastern mythology. Routledge, London.

Leick, G. (2001) Mesopotamia: the invention of the city. Penguin, London.

Modelski, G. (2003) World cities: -3000 to 2000. Faros 2000, Washington, DC.

Renfrew, C. (1975) Trade as action at a distance: questions of integration and communication. In J. A. Sabloff and C. C. Lamber-Karloversusky (eds) Ancient civilization and trade. University of New Mexico Press: Albuquerque, NM.

Robinson, J. (2002) Global and world cities: a view from off the map. International Journal of Urban and Regional Research. 26, 531-54.

Sanderson, S. K. (1995) Expanding world commercialization: the link between world-systems and civilizations. In S. K. Sanderson (ed.), Civilizations and world-systems. AltaMira, Walnut Creek, CA.

Sassen, S. (1991/2001) Global City: New York, London, Tokyo. Princeton University Press, Princeton, NJ.

Scott, J. C. (1998) Seeing like a state. Yale University Press, New Haven, CT.

Skinner, G. W. (1964/1965) Marketing and social structure in rural China. Association of Asian Studies Reprint Series, 1. Association of Asian Studies, Ann Arbor, MI.

Taagepera, R. (1978) Size and duration of empires: systematics of size. Social Science Research 7, 108-127.

Taylor, P. J. (2001) Specification of the world city network. Geographical Analysis 33, 181-94.

Taylor, P. J. (2004) World city network: a global urban analysis. Routledge, London.

Taylor, P. J. (2005) Leading world cities: empirical evaluations of urban nodes in multiple networks. Urban Studies 42, 1593-1608.

Taylor, P. J. (2013) Extraordinary cities: millennia of moral syndromes, world-systems and city/state relations. Edward Elgar: Cheltenham, UK.

Taylor, P. J. (2017) Promiscuous agglomerations: towards integrating urban agglomerations with urban networks. GaWC Research Bulletin 457 (

Taylor, P. J. and B. Derudder (2016) World city network: a global urban analysis. Routledge, London (second edition).

Taylor, P. J. and B. Derudder (2017) Central flow theory: comparative connectivities in the world city network. Regional Studies

Taylor, P. J., M. Hoyler and R. Verbruggen (2010) External urban relation process: introducing central flow theory to complement social place theory. Urban Studies 47, 2803-18.

Taylor, P. J., M. Hoyler, K. Pain and S. Vinceguerra (2014) Extensive and intensive globalizations: explicating the low connectivity puzzle of US cities using a city-dyad analysis. Journal of Urban Affairs 36, 876-90.

Thomas, A. R. (2010) The evolution of the ancient city. Lexington, Lanham, MD.

Wallerstein, I. (1976) A world-systems perspective on the social sciences. British Journal of Sociology 27, 345-54.



* Peter J. Taylor, email:


Edited and posted on the web on 3rd February 2018