The structure of Upper Mesopotamian cities: Insight from fluxgate gradiometer survey at Kazane Höyük, southeastern Turkey

This paper presents the results of fluxgate gradiometer survey of the Bronze Age city at the site of Kazane Höyük, southeastern Turkey. We undertook this work to test the applicability of magnetometry to the study of the organization of urban space at this site within the context of urbanization in Upper Mesopotamia. Gradiometry collection covered a total of 37 520 m2 in five parts of the site. Results from each area were mixed but the most revealing data, from Area 1, show a roughly 2 ha area in the outer town that contains monumental, elite and administrative architecture as well as a main street. Low negative values indicate that most identified architecture is built with limestone foundations, and high positive values reveal that some of the buildings burned before their collapse. These interpretations are supported by excavations that reveal much about the use of the identified spaces and features. Although the structure of Area 1 is rectilinear, evidence for strict rules of city planning is lacking. Instead, the third millennium city at Kazane has a structure seen at other Upper Mesopotamian cities: dense, semi‐orthogonal architecture built along well‐maintained avenues. Combined with previous research, it is clear that Kazane contained multiple elite or administrative areas, which may indicate a degree of power‐sharing or heterarchy in the development and management of this city. Copyright © 2010 John Wiley & Sons, Ltd.


Introduction
Recent research shows that the first cities in Upper Mesopotamia developed during the fourth millennium BC (Frangipane, 2002;Gibson et al., 2002;Oates et al., 2007). The mid-third millennium marked a second period of urbanism in which cities spread widely across the entire region, as part of numerous small, urbanized states ( Figure 1) (Akkermans and Schwartz, 2003, pp. 232-287;Stein, 2004;Wattenmaker, 2009). Larger cities ranged in size from 20 ha to over 100 ha, their growth possibly limited by the productivity of dry-farming (Wilkinson, 1994). The structure of these cities reveals how they developed and the socio-political organization of their residents.
Although our knowledge is limited to a few geophysics plans and excavated exposures of varying extent, current data indicate that third millennium Upper Mesopotamian cities lacked a rigid master plan such as the grid system at Teotihuacán (Cowgill, 2007). Instead these cities shared characteristics that adhered to general planning principles that emphasized multiple nuclei of social activities, such as neighbourhoods. The two basic city shapes were oblong and round, enclosed by a city wall (Stone, 1995). Among an expansive lower town containing residences, craft production facilities and other kinds of spaces, these cities had a tall mound comprised of stratified remains from pre-urban periods. This mound served as a citadel in urban periods, and in some cities, such as Chuera, it formed a large upper city in excess of 25 ha (Meyer, 2006).
The radiating main streets and concentric connecting streets at round cities, including Tell Chuera (Meyer, 2006), Al Rawda (Castel et al., 2005) and Beydar (Lebeau, 1997(Lebeau, , 2006Lebeau and Suleiman, 2003), provide easier access to the city centre and the resultant wedge-shaped Archaeological Prospection Archaeol. Prospect. 17, 73-88 (2010) Published online 17 March 2010 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/arp.375 sectors. In contrast, winding main streets at oblong cities such as Titriş (Nishimura, 2008: figure 33) obscure lines of sight and may impede transit for visitors. Regardless of street pattern or city shape, these cities often have dead-ends or cul-de-sacs that create defensible space within neighbourhoods (Newman, 1973;Creekmore, 2008, p. 345). Streets also contain drainage systems that are a vital part of urban infrastructure. Main streets that connect sectors of the city may form the backbone of what MacDonald (1986, pp. 30-31) describes in Roman cities as armature (Creekmore, 2008, pp. 343-345). Thus, these cities are not highly planned but contain many aspects of city planning and express ideas about the social production and construction of space (Low, 2000, pp. 127-128).
This paper explores Upper Mesopotamian city space based on the results of magnetometry and associated excavations conducted from 2003 to 2005 at the site of Kazane Höyü k, southeastern Turkey. Settlement patterns indicate that during the mid-third millennium BC, Kazane was the 100 ha primate capital of a small state in the Harran Plain (Yardımcı, 2004;Creekmore, 2008). Kazane consists of an oblong area enclosed by a city wall, with an 8-12 ha, 20 m high multiperiod mound located in the northwest portion of the settlement (Figure 2). Surface survey and excavations led by Patricia Wattenmaker since 1992 identified third millennium architecture and artefacts in various parts of the site (Wattenmaker, 1997;Wattenmaker and Mısır, 1994). The three most substantial exposures of third millennium contexts all contain monumental architecture. In the first instance, excavations revealed a wall 50 m long in the centre of the site, just east of the tell, that is possibly the fortified perimeter of a palace (see Figure 8B: PW) (Wattenmaker, 1997, figure 5). In the second case, excavations in the southern part of the  Titriş;2,Harran;3,Banat;4,Sweyhat;5,6,Ebla;7,8,Bi'a;9,Chuera;10,Beydar;11,Mozan;12,Brak;13,Leilan;14,Hamoukar; 15,Taya. city uncovered a monumental tomb that is probably part of an elite house with wall foundations over 1 m wide, built from large limestone boulders (see figure 5, grid C5; Creekmore, 2008, figure 2.7). In the third case, excavations in the southern end of the city revealed yet another structure with a substantial limestone foundation associated with a pavement containing storage and weaving tools, and a poorly preserved, simple structure that may be the home of the weavers (Wattenmaker, 1997, figure 6). Prior to the application of magnetometry the wider context of these structures or the character of other parts of the third millennium city was largely unknown. We tested magnetometry in five different parts of the site with a view towards assessing the usefulness of this method at the site, and with the goal of identifying structures, streets and their articulation in various places within the city ( Figure 2).

Research goals
Geophysics reveals large segments of urban structure, providing a means for spatial analysis at a scale beyond the limits of practical excavation (Aspinall et al., 2008, 149-155). In cases where the characteristics of buried remains are especially suitable for geophysics, it is possible to construct detailed plans to study architecture, streets and the spatial organization of a settlement, as demonstrated by Benech's (2007) application of Hillier and Hanson's (1984) methods of spatial analysis to residential blocks at Doura-Europos, Syria. In another example, Branting used geophysics data as the basis for a GIS-T study of pedestrian activities at the city of Kerkenes Dag (Branting, 2004;Summers, 2009). We should be mindful that the partial, or in some cases complete, urban plans provided by geophysics are representations of a slice of time, often the last period or phase of a city, although uneven development may collapse multiple periods into a single plane. A long-term goal of this project is to use multiple geophysics methods and targeted excavation to reconstruct the life history of Kazane and discern the dynamics of its growth and decline. At Kazane our use of magnetometry was driven by the following research questions.
(i) What is the context of the monumental architecture found in excavations? Are these structures isolated buildings set among non-monumental structures or are they part of extended areas of elite or administrative facilities? (ii) What is the nature of city planning at Kazane? Do streets, buildings, or other features show a unified master plan?
(iii) What is the density of settlement at Kazane? Does Kazane have a crowded urban landscape as at contemporary cities such as Titriş Hö yü k or Tell Chuera? Can we identify open or low-density areas? (iv) What materials are used in the architecture at Kazane? Limestone foundations should be readily apparent in magnetometry data, and especially wide or monumental walls should mark elite or administrative structures. (v) Burning should be evident in magnetometry data.
Can we identify pyrotechnic features related to craft production, or building fires that preserve organic remains related to foodstuffs and craft production?

Methods
At Kazane we employed fluxgate magnetometry because this method is relatively quick compared with methods that require contact with the ground surface. Also, prior excavations revealed that the limestone wall foundations and terra rossa soil at Kazane are very similar to other sites, such as Titriş Hö yü k, where fluxgate magnetometry effectively identified buried architecture and features (Algaze et al., 1995, pp. 32-37;Matney and Algaze, 1995, pp. 36-39;Rosen, 1997;Creekmore, 2008). Although it would have been preferable to use a more recent instrument design with more memory and features, such as a Geoscan FM-256 gradiometer, or a caesium magnetometer with higher sensitivity, due to cost we used a locally available Geoscan FM-36 gradiometer. An FM-36 can reliably detect buried features up to about 1 m below the ground surface, although features with very strong magnetism may be detectable to 3 m depending on their context (Clark, 2000, pp. 78-80). At Kazane we collected data in 20 m T 20 m grids at eight samples per metre with 1 m spacing between zig-zag traverses. Collecting sixteen samples per metre by shrinking the collection area or employing 0.5 m traverses may have reduced aliasing and improved the visibility of thin walls or small features. Unfortunately higher resolution sampling would have been extremely time-consuming and would have inhibited our ability to test different parts of the site. We processed collected data with Geoplot 3.0 according to the scheme outlined by Somers et al. (2003, pp. 71-73) for the Geoscan Instruments FM-36. Standard procedures include clipping, interpolation, and zero mean traverse/grid to correct for errors introduced by zigzag traverses and uneven grid edges. The size of the collection area, the research objective, prior archaeological knowledge and ground conditions for each area are summarized in Table 1.

Results and interpretation
The interpretations discussed here derive from data collected over known features, comparative magnetometry data from other sites, factors such as thermoremnant magnetism, and subsequent test trenches. In Area 1 (Figure 2) we first collected magnetometry data over architecture and features excavated and reburied in 2002. This allowed us to compare known types of architectural material, in this case burned mudbricks and limestone foundations, with anomalies in the magnetogram. In addition, comparative magnetometry data from sites with similar kinds of architecture and features provided interpretive examples for the Kazane data. Especially comparable data from other Early Bronze Age cities include the work in Turkey at Titriş Hö yü k (Algaze et al., , 2001Matney and Algaze, 1995), and in Syria at Al-Rawda (Gondet and Castel, 2004;Castel et al., 2005), Sweyhat (Peregrine, 1996;Peregrine et al., 1997), and Chuera (Meyer, 2006, Abb. 2;Buthmann et al., 2001;Pruss, 2000Pruss, , p. 1432.
To draw the interpretations in Figures 4,5 and 8-11 we examined the magnetograms at different scales, with different palettes, along the path of different processing procedures, and with different clipped or extracted data ranges. We looked for patterns or magnetic values indicative of architecture, heated or burned contexts, and other features. We indicate the relative confidence of our interpretations by distinguishing between 'walls' and 'possible walls.' Note that the thickness of the interpretive lines does not correspond to the expected thickness of the walls, except for the excavated walls shown in Figure 5; instead these lines mark the midpoint of the linear anomaly. Despite the care taken during interpretation, any feature that has not been confirmed by excavation is subject to misinterpretation. Magnetic values discussed below are based on raw, pre-processed or unclipped data.

Area 1
The results from Area 1 reveal rectilinear architecture with a northwest-to-southeast orientation ( Figure 3). We tentatively identify at least 11 Building Units that may comprise integrated structures, as well as isolated pieces of architecture that are probably just the visible portion of larger units (Figures 4 and 5). These Building Units range in size from less than 50 m 2 to over 1000 m 2 . Gradiometry data indicate that many of these structures have walls a metre or more wide, an estimation confirmed by test trenches (see Figure 7: A-F). All of the structures have walls showing low magnetic values ranging from À1 nT to À20 nT. On the basis of known structures previously excavated and reburied in grid C5 (Figure 3), these values indicate the use of limestone foundations.
A street, appearing as a slightly positive linear value (0 nT to 5 nT) between two negative linear values (À1 nT to À10 nT), runs northwest-to-southeast along the western edge of the area from grid A6 to grid D2 (Figure 3; see E in Figure 7). Plough furrow-derived readings perpendicular to this street may mask cross streets. For example, plough furrows paralleling the field border of the collection area mask a potential southwest-to-northeast street intersecting the known street in grid B6 and running to at least grid C8 (Figure 3). In addition, largely 'empty' areas, for example grids F4 and E5 (Figure 3), contain hints of structures that appear as faint white lines. These 'structures' are probably buried more deeply than

Test trenches in Area 1
Our initial trenches were long, narrow exposures designed to test specific features. We subsequently expanded some trenches to create larger exposures and obtain artefact samples from various contexts ( Figure 6). The trenches revealed many structures oriented with respect to each other and to the street on the western edge of the area. Analysis of artefacts, radiocarbon dates and architecture from the trenches shows that these buildings date to the period 2550-2250 BC, except for Building Unit 11 (Operation 8, Figure 7: B) and modifications to Building Unit 3 (Operation 6), which date to the early second millennium (Creekmore, 2008).
The trenches confirm that the most prominent structures in Area 1 are monumental in either their dimensions or the thickness of their walls (Figure 7: A, C, D, F). All excavated structures were built of mudbrick (sun-dried bricks made of mud and straw) set upon limestone foundations, although the mudbrick superstructure only survived in Building Units 5 and 7. An exception is Building Unit 6, which excavations in Operation 7 reveal has intact walls over 2 m wide, built from relatively small field stones instead of large blocks (Figure 7: F). Notably, large negative values, such as those found in grids D6 and C7, correspond to limestone walls close to the surface whereas smaller negative values, such as those in the eastern half of grid D7 and in grid F7, are more deeply buried, sometimes up to 1 m below the ground surface.
The existence of thin-walled, non-monumental structures seen in the gradiometry data was also confirmed through excavation. The thin walls, each around 0.50 m wide, were revealed to be the southern room of Building Unit 4 (an abutting addition to the northern rooms), early second millennium domestic reuse of Building Unit 3 ( Figure 5: the thin, east-west wall), an early second millennium domestic context in Building Unit 11 (both contexts dated by common ceramic types; Figure 7: B), and two walls bordering the street exposed in Operation 5 (Figure 7: E). The walls on either side of the street, located about 1 m below the ground surface, serve as curbs but may also be the common wall of non-monumental structures bordering the street. The contrast between the slightly positive values of the street, which consists of compacted pebbles, pottery, bones and other garbage, and the negative values of the limestone walls on its margins, enhances the visibility of these features. In addition to the correlation between limestone walls and negative linear anomalies, we noted a correlation between burned contexts and areas of strong positive values ranging from 12 nT to 90 nT. Excavations show that strong positive values in grids F5, D6, C7 and E7 correspond to burned mudbrick, and in the case of Building Unit 4 in Operation 2, a room full of large storage jars and collapsed, burned bricks (Figure 7: D). Operation 1.1 (narrow trench connecting Operations 1 and 6) revealed that the wide, positive, elbow-shaped anomaly in grid C7 consists of burned, collapsed mudbricks (5 nT to 15 nT) that fell from the wall immediately to the west. The high positive reading in grid F5 (13 nT to 87 nT) was revealed by the trench Operation 3 to be a highly burned building in which the collapsed bricks -from the ground surface up to several metres deep -were pulverized by the heat of the fire. Burned bricks were also found within and east of Building Unit 4 in grid D6, and in the southern half of Building Unit 1. Areas of especially high magnetic values (12 nT to 40 nT) in Building Unit 5 correspond to highly burned bricks, although burned grain was also noted (Figure 7: G).
The results of the test trenches suggest that unexcavated patterns of high positive magnetic values paralleling negative linear anomalies, or limestone wall foundations, are collapsed, burned bricks. These and other high positive readings are highlighted in Figure 3. Widely scattered burning across Building Units 4 and 5 into the southern part of Unit 2 and into Unit 1 may indicate a spreading fire that started in the The distribution of burned bricks may reflect the nature of the construction materials or the relative flammability of the contents of these buildings. Notably, the andiron hearth excavated in grid D5 ( Figure 5) appears to match a dipolar anomaly (À18 nT to 7 nT), and the high positive values at the northern end of the room correspond to burned bricks (15 nT to 29 nT).

Interpretation of Area 1
The combination of the gradiometry and test trench data indicates that most of the identified structures in Area 1 are probably elite or institutional buildings marked by monumental wall foundations or large dimensions. The artefacts recovered from Building Unit 8 suggest that its southern room was the storage space for a temple. This room contained many plain and fenestrated stands, small bowls and drinking cups, a few unusual vessels and several grindstone pieces (Creekmore, 2008, pp. 165-167, figure 5.18). It is possible that the northern room is the sanctuary of a small temple, but we do not have enough evidence to validate this suggestion. In another possible templerelated context, we found several sacred or votive statue eye-inlays just outside Building Unit 7 (Creekmore, 2008, figure 5.14).
Large storage jars packed into Building Unit 4 identify it as a storage facility, while burned barley, rodent bones and well-plastered interior walls identify Building Unit 5 as a granary (Creekmore, 2008, figures 5.8 and 5.10). Cylinder-seal impressed clay sealings found in both buildings attest to the administration of goods. These sealings were for doors, jars, baskets and other containers. The long, narrow plan of Building Unit 10, located between Building Units 4 and 5, may indicate that it too is a storage facility. Finally, the large size of Building Unit 6, combined with its long rooms bordered by exceptionally wide walls, may mark it as a massive storage facility or an official building. Patricia Wattenmaker's excavation of a monumental tomb in Building Unit 1 marks this as an elite house. This structure may extend into Building Unit 2, where

Area 2
Despite modern disturbances (Table 1) some rectilinear features are visible in the Area 2 data (Figure 8), for example in grid B4, but none comprise unambiguous architecture. We suspect that many nearsurface stone walls in the southern part of the area, in the vicinity of the structure excavated in 1996, were removed by farmers because we witnessed stone robbing during our fieldwork. The strong dipole values throughout the area north of the fence, concentrated in grids C4, A5, B5, D5 and A6-C6, derive from modern metal debris in the plough zone, hearths or kilns, or recent trash fires on the surface. The values for these anomalies generally range from À40 nT to 60 nT, although some reach 100 nT. Similar dipole effects due to iron objects are noted by Kvamme (2003, pp. 444-445), and experiments show that even shortlived fires at relatively low temperatures can significantly increase the magnetism of the area around the burning (Morinaga et al., 1999;Linford and Canti, 2001). During data collection we noted evidence of two recent surface fires in grids C4 and C5. The C5 fire did not show up in collected data but the fire in C4 corresponds to a cluster of dipolar anomalies. The circular pattern of the groups of dipole anomalies in the northern half of Area 2 is difficult to explain but may derive from regular dumping or burning of modern garbage in a particular area. Alternatively, given their size, at least some of these dipole signals derive from nested or rebuilt hearths and kilns, possibly arranged around a room in a bakery. If the large wall previously excavated in this area belongs to a large official building then on the basis of comparable structures we would expect to find multi-oven bakeries, which service the facility, in or near the complex (see e.g. the bakery at Tell Brak; Emberling and McDonald, 2003).

Area 3
As in Area 2, small dipole values in grids B3, C2 and B4 (Figure 9), may derive from modern metal objects in the plough zone. In addition, we identified parts of two to four structures and two possible pyrotechnic installations. Two square structures, each about 100 to 150 m 2 , are oriented northwest to southwest and appear largely in grids A2 and B4. Two additional structure-like anomalies appear in grid B3, and the western edge of grid A1. All of these structures have higher magnetic values (2 nT to 5 nT) than the structures from Area 1, suggesting that they are built from mixed materials rather than strictly limestone. Finally, two dipole anomalies in grids B3 and B4 may mark small kilns. Test trench Operation 9 cut across what we expected to be the inside corner of two walls, possibly connected by a limestone pavement. Instead, the area of negative magnetism (À5 nT to À3 nT) corresponded to a collapse or pile of limestones of uncertain date, and we did not find any walls (Figure 9: B, grid B4). Due to field conditions we were unable to expand the test trench or investigate spots of high magnetism in the area.

Area 4
Striping from deep plough furrows makes it difficult to interpret linear features in Area 4 ( Figure 10). There may be a faint negative linear feature running northwest to southeast across grid A1. Based on test-trench results in Area 1, this faint negative feature, with values ranging from À2 nT to À7 nT, could indicate a wide limestone foundation about 1 m below the surface. Another potential negative linear feature runs north to south in grid D1-E1. Although these

Area 5
The most prominent features in Area 5 are the line of relatively large dipole values running southwest to northeast from grid A2 to C4 (Figure 11), and the other large dipole anomalies in grids B2, C3 and D3 (values vary but range from À50 nT to 50 nT). The strongest of these anomalies may be hearths or kilns in an outdoor setting or located within structures built from material, such as mudbrick, that is not as perceptible to the magnetometer. Although magnetic values are dependent upon a variety of factors, the values of these anomalies are similar to a kiln uncovered under similar field conditions at Ziyaret Tepe using an FM-256 fluxgate gradiometer (Matney and Donkin, 2006, p. 16). The values of the Ziyaret kiln range from À29.85 nT to þ41.80 nT (Tim Matney, personal communication). If the installations are outdoors then they may mark this space as an industrial area for ceramic or other pyrotechnic industries. In contrast, if the pyrotechnic features are within structures then their spacing, about 10 m to 20 m apart, may mark high density, small-scale housing.
Most of the rectilinear anomalies in Area 5 closely follow plough furrows and irrigation channels, making it likely that these are soil marks rather than buried structures. The most promising possible structure appears in the northwest corner of grid B3, continuing into grid B4, where we identified a rectangular feature with at least one subdivision. This feature has a magnetic value similar to the linear features in Area 3 (2 nT to 6 nT), again indicating that it was built from mixed materials rather than solely limestone. We tested this feature with Operation 10, which identified several phases of early second millennium architecture, dated by common ceramic types, reaching up 2 m below the ground surface ( Figure 11: B, grid B3; Creekmore, 2008, figures 5.32-5.34). Although these walls were indeed thin (less than 0.50 m), as the gradiometry indicated, and in some cases were oriented as we expected, none of these features corresponded precisely to the magnetic data and the closest matches were found too deep to be reasonably associated with the magnetic anomalies. Notably, two small (less than 1 m diameter), nested, thin-walled clay bread ovens found just over 1 m below the surface in the southern end of Operation 10 do not appear in the gradiometry data. This suggests that if the strong dipole signals across Area 5 are indeed pyrotechnic features, then they are probably buried less than 1 m deep or they were more highly fired than simple bread ovens, making their magnetism visible despite their greater depth.

Conclusions: city space at Kazane
The gradiometry work described here yielded very good results in Area 1 and suggestive results in the other areas. In less successful areas a caesium magnetometer may provide better results due to its greater sensitivity, while closer spacing of traverses may identify small or thin features not seen in the current data. Resistivity has the potential to locate features in places, such as Area 2, where metal materials interfered with the magnetic signal. An additional caesium magnetometer or resistivity survey may also provide a view of features beyond the optimal 1 m depth range of the FM-36. Finally, ground-penetrating radar would be especially useful for determining the degree of continuity in city structure over time by comparing time (depth) slices from deeply stratified areas within the city, such as Areas 1 or 5 (e.g. Casana et al., 2008, figure 8).
Although we failed to locate any indisputable pyrotechnic installations, isolated dipolar anomalies across Area 5 may mark a concentration of pyrotechnic industries. In Area 1, patterns of burning provide clues to building materials and the nature of the destruction, intentional or not, of the structures in this area. Although Kazane probably did not collapse at the end of the third millennium, a time in which cities across Upper Mesopotamia experienced contraction, abandonment or political reorientation (Akkermans and Schwartz, 2003, pp. 282-287), the finding of the remnants of a simple early second millennium structure built within and over the monumental walls of Building Unit 3 suggests that there was a significant reorganization of space in Area 1 during the transition from the third to the second millennium.
In Area 1 we were able to contextualize a previously known monumental structure, showing it to be just one of many such buildings spread over a large area. These buildings are densely packed with only narrow passages between them, and in some cases they share common walls. In contrast, the southeastern part of Area 1 seems relatively empty of architecture, perhaps   (Orthmann, 1995, Plan 13). Regarding city planning, the buildings and a street in Area 1 indicate that, like other Upper Mesopotamian cities, Kazane's structure is organized in a semiorthogonal manner for reasons of convenience rather than centralized planning (see Smith, 2007, p. 13). Clearly visible structures in Area 1 have gradiometry values consistent with limestone foundations, while structures in Area 3 and 5 depart from this pattern, although in the latter cases the interpreted features could not be confirmed by test trenches. The preference for such monumentality in the construction of storage facilities, housing, cultic and other structures may reinforce not only the power or prestige of the private households or institutions that made use of these structures, but of the city as a whole. Limestone is readily available in the low mountains bordering the plain, but the labour-intensive transport and use of large stone foundations marks a clear choice to build impressive structures.
The identification of a large area of monumental, administrative architecture in the outer town at Kazane is especially significant because such structures are generally found in the core or the citadel of third millennium cities, where they were protected in an often elevated, centralized and walled location, as at Tell Chuera, Tell Beydar, Tell Mozan, Tell Brak and Tell es-Sweyhat. Although these findings are due in part to the tendency to focus excavations in central areas or citadels, they project a sense that these cities grew around a single elite and administrative core. There are few excavated examples of clearly elite, administrative, or public buildings in the outer town of third millennium Upper Mesopotamian cities, but magnetometry shows some relatively larger structures in the lower town at Tell Chuera (Meyer, 2006, p. 184), and Titriş Hö yü k (Nishimura, 2008, p. 150). In addition, at Tell Beydar a large granary has been excavated in an area of the upper city that is some distance from the palace-temple complex at the core of the site (Sténuit, 2003).
The data from Area 1 at Kazane confirm that elite and administrative contexts were not confined to the core or the tell/citadel. This suggests that there were multiple nuclei of growth in the city and echoes the dispersal of small temples at Tell Chuera and Tell Al-Rawda, which may indicate that religious practice and power in these cities were multisited and multicentred. At Kazane, Building Unit 8 in Area 1 may be an example of a small temple nestled in the dense urban fabric. Thus, Kazane may best fit the multiple nuclei model of urban development, in which several nuclei develop within an emerging city (Harris and Ulman, 1945, figure 5). Alternatively, it is possible that the city grew through the expansion of core features along sectors bracketed by major streets. Viewing urban growth as the merging of multiple nuclei or the expansion of sectors may explain the characteristic oblong shape of many Upper Mesopotamian cities. Perhaps the city walls were built to encompass dispersed initial nuclei and developing neighbourhoods that later merged to form the densely built spaces we find during the apogee of a city's development. Dispersed nuclei also suggest some degree of power-sharing in both the production and management of urban space, perhaps a kind of heterarchal governing structure as suggested by scholars of Mesopotamian cities (Crumley, 1995;Ur, 2004;Fleming, 2004;Stein, 2004;Stone, 2007;Creekmore, 2008;Wattenmaker, 2009). To investigate these complex issues of urbanization, we should expand the use of geophysical techniques to reveal major portions of city structure.