The practice model has not previously been developed with regard to urban layout, although some oF Wilson's (1982) ideas developed at Takwa do move in this direction. Rather, practice has been invoked during discussion of institutions within the town, most notably the Swahili stonehouse. The wholesale introduction of domestic architecture in coral and lime construction occurs with the towns of the fourteenth to fifteenth centuries, although the development of this architectural traDition is seen earlier at sites like Shanga and in the construction of mosques from the eleventh century. The stonehouse has been seen as the quintessential expression of SwaHili identity (Allen 1974, 1979; Donley 1982, 1987; Donley-Reid 1990) and becomes widespread during this explosion of town building in the fourteenth century. At this time, towns also began to assume a different character as an increasing number of small community mosques supplemented the main congregational mosque, and town walls were buIlt at some sites.
Stonehouses have therefore been studied as miniature worlds wIth the practices and rituals of the domestic spaces seen as key aspects of the construction of Swahili identity (Fleisher and LaViolette 2007). The importance attributed to these structures has also led to the suggestion that it was the requirements of the houses that structured the urban spaces (for a similar argument regarding Upper Mesopotamia, see Creekmore, Chapter 2 in this volume). The concomitant assumption is that the site layout had no overarching plan: Garlake (1966), who conducted the largest survey of coastal architecture to date, described the towns as having little in the way of urban planning, at least in the sense of regular streets, or an over-
All pattern evident in the layout. He argues that this was owing to the houses:
Strong preferences for orientation [to the north or, less often, east] made the planning of towns difficult, for no house will have entrances to any street bordering its western or southern sides. Blocks of building surrounded by streets are therefore scarcely practical, and street plans are as a result irregular or, more correctly, non-existent. (Garlake 1966:89)
Yet, Garlake also laid the seeds of an important approach to the study of Swahili urban space, when he emphasized the Swahili house as the most important feature of the urban environment, defining the way that the townscape developed:
One of the most interesting features of the town buildings is the communal approach to, and cooperation in, planning. All adjoining houses invariably share a single common party wall (there is a single exception in Songo Mnara). Moreover, in almost every case, where houses adjoin, the plans interlock rather than simply abut one another, making for compactness and economy of building. This is evidence of a far greater degree of cooperation in planning and construction than that which would be found if just one owner allowed his neighbour to build against the irregular line of his outside wall. It entailed complete cooperation and joint planning from the start, followed by simultaneous building. (Garlake 1966:89)
He (1966:90) suggests that this is most likely attributable to "a close degree of kinship, or very firm family or tribal ties," recalling Ghaidan's notion of involvement in the town plan of Lamu. This idea was developed by Allen (1979:6) who links the apparently chaotic patterning of Swahili urban spaces to the lifecycles of the houses themselves and the importance attached to spatial prescriptions over the ways the houses were used. Garlake emphasized the orientation of the houses, of which 75-80 percent face north, and cited environmental reasons, claiming that this direction favored ventilation. Allen develops instead the social aspects of the houses' design, and particularly the importance of ritual in the orientation and layout.
Allen's (1979) analysis of the Swahili stonehouse draws on a combination of ethnographic data (from Lamu) and the evidence of stonehouses in the archaeological record from the fourteenth
Century onward. He feels that the continuity of the stone-building tradition and similarities in form of the houses justify this extrapolation, a justification later taken up by Donley-Reid (1990). Allen's interpretation of the houses is twofold: as the settings for economic activity and as important sites for domestic ritual and self-identiication. He suggests that both would have had an effect on the developing house plan, and hence the cumulative urban space. First, Allen considers the demands of finding space for family members anD for the conduct of trade. Trade seems, by the fourteenth to fifteenth centuries, to have been conducted in the context oF the houses and through a system of hospitality and the use of a trusted patron. Ibn Battuta's eye-witness description of fourteenth-century Mogadishu helps us understand how this might have worked:
SWAHILI URBAN SPACES Of THE EASTERN AFRICAN COAST
When a ship comes into port, it is boarded from sanbuq, that is to say, little boats. Each sanbuq carries a crowd of young men, each carrying a covered dish, containing food. Each one of them presents his dish to a merchant on board, and calls out: "This man is my guest." And his fellows do the same. Not one of the merchants disembarks except to go to the house of his host among the young men. . . when a merchant has settled in his host's house, the latter sells for him what he has brought and makes his purchases for him. Buying anything from a merchant below its market price or selling him anything except in his host's presence is disapproved of by the people of Mogadishu. (Freeman-Grenville 1962:27-28)
The houses are seen to cater for this through the provision of porches for the conduct of business, and an adjoining guest room where visiting merchants might stay. At the same time, the privacy and purity of the househoLd itself - and particularly of the women within it - would be ensured by the design of the house around an "intimacy gradienT" (Ghaidan 1971), with successive rooms of increasing levels of privacy. Over time, with what appears to have been increasing proscriptions on female appearances in public life, this could have led to the development of the house blocks seen in eighteenth-century towns such as Lamu, in which women were able to move between houses without having to venture outdoors. An earlier form of this was simply the idea that family houses would exist in compound arrangements, or courtyard blocks.
Ethnographically known SwaHili houses were fundamentally associated with women, who might only rarely leave the house or block in which they lived. Matrilocal marriage patterns meant that over the course of generations, houses would need to be extended and divided to cater for incoming husbands and the growing family. This fluidity of house size, and evolving architecture, necessitated against a rigid town plan and meant that inhabitants would prioritize the spaces around houses, and that house blocks or compound plans would develop organically. In addition, the interior spaces of the houses were strictly controlled by the rituals associated with marriage, birth, and death, with the family's links with, and understanding of, the spaces of the houses being renewed through these practices. In particular, this has been explored by Donley-Reid, who drew on a speciic ritual in which a newborn baby is formally introduced to the interior spaces as a means of elaborating the spatiality of Swahili social structures and the ways that the house is both the medium and the expression of these.
These interpretations of the houses also rely heavily on ethnographic data from twentieth-century Lamu, and it is necessary to take into account the diversity of houses in the fourteenth to if-teenth centuries, as well as the importance of change over time that might reflect changing meanings relating to the houses (Fleisher and LaViolette 2007). Rather than simply assuming an unchanging conceptual map for the conined spaces, though, these approaches do provide an important dynamic approach to space. The emphasis on practice and ritual as linked to the lifecycle of the houses could be extended to think through the delimited exterior spaces of the urban layout on a larger scale. A focus on practice, and the lifecycle of the house through the generations of its inhabitants, offers a more flexible and useful mechanism for understanding the layout of Swahili sites, and incorporating the diversity observed among the different town plans. It also brings some of the insights of the regional studies of Swahili urbanism into the town, thinking through the ways that urbanism was created through mapping interactions, rather than simply documenting its form.
At sites such as Jumba la Mtwana on the Kenyan coasT (Figure 4.9), which do not seem to conform to the centralized pattern at all, the practice-based model offers a heuristic tool that allows us an entry point into an otherwise anomalous layout. In fact, the scattered town plan of Jumba, which it has been suggested may have offered space
For individuaL farmsteads or gardens (Sassoon 1981), could also have been a tactic within a new settlement for families building houses that they knew might need expansion over time. Such a consideration would tend to favor a dispersed site pattern. Likewise, one of the standing buildings at Jumba - the one known as the "House of the Many Doors" - offers a direct glimpse of the process of development in microcosm, as the building was adapted numerous times during its period of occupation, resulting in the many in-filled doors that give it its name. With each change, the building was further extended and subdivided - once so extensively that it seems to have been almost demolished and rebuilt at a higher level - until it ended its days as a series of "apartments," which Sassoon (1981) suggests may even have housed visiting merchants, as they appear to be smaLl housing units.
SWAHILI URBAN SPACES Of THE EASTERN AFRICAN COAST
At Songo Mnara, in Tanzania, the numerous stone houses arranged around the site without any obvious street pattern or regular plan also speak to this type of spatial structuring (Figure 4.10). The orientation of the houses is clear, and what regular patterning does exist is formed by the house blocks, which have been built up around courtyards and through houses that shared some of the partition walls. Even the town wall, which enclosed the site to the landward side, is developed in parts through the joining together of several houses: a pattern that is mirrored at other walled sites. Songo Mnara also speaks to one of the advantages of exploring Swahili urban space in this way, as the site contains a structure known as the "palace" (Garlake 1966), as well as several other huge domestic structures. Each can actually be seen as compound structures, with numerous room blocks leading onto central areas, and sharing adjoining walls. These structures are very different from the palace of Husuni Kubwa aT Kilwa, which is a unique example of a planned layout associated with a particular sultan. If a correlation of space and social structure based on a single elite ruler is assumed, the "palace" at Songo Mnara must be seen as analogous, despite the presence of numerous other contenders for that title, ignoring the very real differences between it and Husuni Kubwa. Likewise, palaces have been idenTified at Tumbatu (Pearce 1920:402) and at Gedi (Kirkman 1963). A practice-based approach to the urban layout that instead explores these structures as built up through the functions and lifecycles of the inhabitants allows a more nuanced discussion of the differences between these sites and thus a greater understanding of the way the urban configurations would have been created.
Importantly, an approach to function, practice, and activity wIthin the towns also allows for a consideration of the delimited spaces, built up through the positioning of the structures, but nonetheless implicated in the activities and needs of daily life. Delimited spaces within Swahili towns commonly include areas within and against town walls, courtyards, areas surrounding central congregational mosques, especially near entrances and outside the north-facing mihrab or prayer niche, central open spaces delimited by domestic architecture, and less frequently by streets and alleys. Although few archaeologists have developed methodologies adequate to investigate these spaces, most supporT Wilson (1982:207), who argues that open spaces in Swahili towns "played an important part in the social, political and economic life of the community." Suggestions as to the use of these spaces, which are present in most Swahili towns, include: open-air meeting places (Garlake 2002:181), market areas, protected space for future town growth (Kusimba 1993:122; 1996:711), gardens and/or orchards (Garlake 2002; Kusimba 1993; see also Stark, Chapter 11 In this volume, for Mesoamerican examples); areas of impermanent architecture, and areas of industrial production (Garlake 2002; Gensheimer 1997:328-339; Kusimba 1993:122). Historic and ethnographic data support some of these hypotheses: Early sixteenth-century Portuguese accounts indicate that the sultan oF Kilwa was crowned in an open space adjacent to the palace, while specially designated open spaces in Comorian cities, called fumboni, were contexts for social gatherings, weddings, feasts, and game playing (Gensheimer 1997:334). Archaeologists have also explored certain aspects of these spaces; on Pemba Island, Tanzania (Fleisher and LaViolette 1999; LaViolette and Fleisher 2009) and Gedi, Kenya (Koplin and LaViolette 2008), excavations have demonstrated that some "open areas" were actually dense with earth-and-thatch houses; aT Shanga in Kenya, evidence of possible trade kiosks has been located in the central open space oF the town (Horton 1996).
World History
