Archaeological Investigations at Songo Mnara, Tanzania:

Urban Space, Social Memory and Materiality on the 15th- and 16th-century Southern Swahili Coast

Preliminary Report

Submitted to the Department of Antiquities, Republic of Tanzania

2010

Jeffrey Fleisher, Ph.D.

Department of Anthropology Rice University

Houston, Texas, USA

Stephanie Wynne-Jones, Ph.D. Department of Archaeology and Anthropology

University of Bristol Bristol, United Kingdom

2

Cover image: Pierced copper Kilwa coin, Hasan ibn Sulaiman type, from SM010 (SF#10122, context 10016).

3

Table of Contents

Acknowledgements………………………………………………………………………………5 Introduction………………………………………………………………………………………7 Research Problem………………………………………………………………………………..7 Methods and Research Procedures Conducted…………………………………………………..9 Excavation Results……………………………………………………………………………...11 Trenches within domestic architecture: House 44……………………………………………...11 Trenches within domestic architecture: House 23……………………………………………...27 Trenches outside domestic architecture………………………………………………………...36 Trenches in open areas and around coral-built features………………………………………...41 Conclusion……………………………………………………………………………………...59 References………………………………………………………………………………………62 Appendix 1: Excavation Form………………………………………………………………….63 Appendix 2: Intrasite Geoarchaeology at Songo Mnara, Tanzania (Dr. Federica Sulas)………65 . Appendix 3: Songo Mnara Intertidal Survey (Mr. Jack Stoetzel)……………………………...67 Appendix 4: Conservation work on the supporting wall of the Necropolis……………………68

Appendix 5: Archiving and storage: Songo Mnara 2009…………………………….………...70 Plates……………………………………………………………………………………………71

4

5

Acknowledgements Research at Songo Mnara was funded primarily by Rice University, through the Social Sciences Research Institute and the Archaeological Field School of the Department of Anthro-pology. The British Institute in Eastern Africa and The Leverhulme Trust provided supplemen-tary funding.

Fieldwork was conducted in collaboration with the Antiquities Department of Tanzania, and we are very grateful to the Director, Donatius Kamamba, for his assistance. We would also like to acknowledge the help of our Antiquities representative, Mohammed Chidoli, who was invaluable to the success of the 2009 fieldwork. In Kilwa Masoko we were also helped by Idi-phonse, Bwanga and Hassan.

Topographic and buildings survey of the site was completed by Mr Benson Kimeu of the British Institute in Eastern Africa. Dr Kate Welham and Mr Harry Manley of Bournemouth University were responsible for the geophysical survey, and the ongoing GIS work in mapping the site. Geoarchaeological work was completed by Dr Federica Sulas of the University of Cambridge; the latter also generously donated laboratory space and time for her analyses.

Excavations were conducted by Abidemi Babatunde Babalola, Brian Clark, Jack Stoet-zel, Kylie Klein, Sarah Nouri, Helen Horn-Mitchem, Elizabeth Johnson, Dominic Pistor and Noelle Tankard. We were assisted by Masoudi Mohammed, Yusuf Hamadi, Sulieman Moham-med, Ahmed Issa, Mohammed Ali Bakari, Ali Mohammed, Assi Abdallah, Sulieman Abdallah, Abdallah Omari, Ahmad Awari, Makame, Yaheya Mohammed, Mahibu Saidi, Issa Hassan, Hassan Mwinyi, Ahmed Abdallah, Zurafa Mwinyi, Rehena Abdallah, Yusufu Masud, and An-drea Seligman. Our camp was organised and run by Bw. Polo, Mama Sofia, Rukia, Fatouma, and Nankunda.

7

Introduction This report details research carried out at the site of Songo Mnara in southern Tanzania

from June 13 to July 17, 2009, under COSTECH permit number RCA2009/46. This project focused on the use of space in daily practice at this ancient urban center, investigating public (centrally-located religious architecture, cemeteries, and open areas) and private (stone houses)spaces. In comparing public and private, this project aimed to delineate particular material practices related to the constitution of social identities through domestic activity and through ritual/commemorative practices associated with centrally-located cemeteries and open areas.

Songo Mnara, nestled in the Kilwa archipelago on the southern coast of Tanzania (Figure 1), is a logical place to begin addressing these theoretical concerns. Despite its promi-nence and excellent preservation, most of the work completed there has been focused on re-cording the extensive architecture (Dorman 1938; Garlake 1966); only test excavations have been conducted at the site (Chittick 1961; Mathew 1959; Pradines 2005; Pradines and Blanch-

ard 2005). Most research in the region has been focused on the more famous and well-known site of Kilwa Kisiwani (Chittick 1974; Sutton 1998; Fleisher 2004; Wynne-Jones 2007). Songo Mnara is dominated by the well-preserved remains of more than 40 large domestic room-blocks, five mosques, and numerous tombs (Figure 2). Room blocks wrap around and enclose an open, central area of the site where tombs, a walled cemetery and a small mosque are located. The site was occupied for a maximum of 200 years, between the late 14th and early 16th centuries. This rela-tively short lifespan (in comparison with the 800-year occupation of Kilwa) means that the stratigraphy is simple, with little vertical complexity, making it an ideal site at which to explore horizontal com-plexity in the form of spatial differentia-tion. Research Problem Archaeological investigations at Songo Mnara focused on a number of re-search problems relating to life in the town itself and the relationship of the town to other coastal settlements. At the most general level, this project has helped document and describe the culture history of the settlement and how it relates to nearby sites, like Kilwa Kisiwani. Within Figure 1: East African Coast

8

the site, the project explored traces of material practices related to stone houses and examined whether they conform with ethnographically- and ethnoarchaeologically-known examples. As Donley-Reid (1990) has shown, the performance of social roles in these private places, or 'structuring structures,' was critical to the constitution and reaffirmation of Swahili social reali-ties. Although these insights have been shown to apply to 17th- to 20th-century Swahili houses in Lamu, few studies have investigated whether similar practices existed in earlier centuries (cf. Fleisher and LaViolette 2007) or in different parts of the eastern African coast. The spatial or-ganization of southern stonetowns (those along the Tanzania coast and south) suggests that a distinctive, more hierarchical social organization dominated these urban centers (LaViolette and Fleisher 2005). Thus, insights from 19th- and 20th-century northern towns such as Lamu may

not be directly analogous. Investigation of chronological and geographical variation is required if we are to understand the ways that private spaces were used and experienced locally in Swa-hili towns. Songo Mnara, with its extensive stone house architecture, will provide a solid southern example with which to compare data from northern towns.

More generally, the investigation of Songo Mnara allows town-level comparisons to be made with northern examples. Because southern towns tend to be constructed primarily in earth and thatch, there has been a real difficulty in comparing town plans from north to south. The research at Songo Mnara seeks to document thoroughly the town plan, which can be used in comparison with those to the north. Are the spatial arrangements of the stone architecture

Figure 2: Map of the standing buildings at Songo Mnara (after Garlake 1966: Plan 43)

9

that are believed to structure northern towns (Horton 1994) applicable to southern ones like Songo Mnara? This project has begun to address this question.

Finally, this project is the first to seriously investigate the central spaces of a Swahili town, with the hope of recovering what types of activities were occurring in those places. Were open, central areas simply places for the dead and thus relatively “empty spaces” except for the graves and tombs? If this was their main use, were they places where public rituals occurred, either through personal, practice or more group-oriented events? Swahili tombs seem to hint at these practices, including offering niches and incorporating symbols of authority, such as im-ported ceramics and sumptuary goods (Fleisher and LaViolette 2007). Modern ethnographies also speak of ritual practice at tombs as an ongoing part of Swahili memorialization practices (Wilson 1979). However, open spaces might also be locations of more prosaic economic activ-ity, such as craft production and/or marketplaces (Fleisher 2010). By approaching open spaces, then, the project begins to address the spatial practices of both the private and public sphere, through a multipronged approach to the material record. Methods and Research Procedures Conducted The methodological approach to archaeological research at Songo Mnara during the 2009 season was aimed at three goals: to build a foundational dataset by thoroughly mapping and describing the site, assessing previous efforts; to evaluate the future archaeological promise of the site; and finally, to address the research questions posed above. All of these efforts in-cluded traditional archaeological excavations and site surveying, as well as a set of new scien-tific approaches aimed at examining less-commonly investigated parts of Swahili sites (such as open areas) including geophysical surveys, soil chemistry studies, microstratigraphic analysis, environmental coring, and flotation. Excavations and Artifact Processing Trenches were dug in natural layers, with trowels, shovels, and hoes. All deposits were sieved, using a 2mm mesh. This proved the most effective way of recovering small artifacts, like beads, fish bones, and coins. Artifacts designated ‘special finds’, such as imports, coins, beads, metals and any other rare or unusual objects, were recorded in situ and their locations mapped.

The project adopted a modified version of the Museum of London Archaeology Service single-context recording system (MoLAS 1994), recording all contexts and cuts on context re-cording sheets (Appendix 1). Each new layer or feature was given a separate context number; context numbers were recorded consecutively within each trench (i.e., context 10003 is the third context recorded in Trench 10). Additional recording sheets documented stone architecture, and environmental samples. Scale plans and profiles were drawn of all layers; contexts were also photographed. In general, trenches were dug to sterile soil, unless otherwise noted below.

All artifacts were bagged in the field, labeled with contextual information, and then re-turned to the field lab, where they were washed, sorted, and preliminarily recorded. Some arti-fact classes, like imported pottery, were analyzed and catalogued in the field lab. Other materi-als, like beads and local pottery, were exported for further study and more detailed recording. In the field laboratory, local pottery was sorted into diagnostic and undiagnostic sherds, all arti-facts were counted and weighed, and then undiagnostic sherds—undecorated bodies only—were discarded, placed into excavated trenches prior to backfilling. All stored artifacts were

10

placed in archival quality plastic bags, as well as bug proof cotton bags; these were organized by trench in wooden boxes and stored in the storeroom at Songo Mnara. Topographic Survey Mr. Benson Kimeu, of the British Institute in Eastern Africa, was hired to conduct to-pographic surveys of the site. He completed a close, 2m-interval survey of the entire site, pro-ducing the first topographic map for the site. Using a Total Station, he was able to assess the reliability of previous site maps created by Garlake (1966) and Pradines (2005). In general, we found that these site maps misrepresented the position and shape of standing architecture in subtle but important ways. We are currently working to try and rectify these detailed maps with their real-world locations. Other survey work included the detailed recording of burials and tombs within the central part of the site as well as in the cemetery adjacent to the southwestern structure, which we refer to as the ‘Necropolis’ (Pradines calls it ‘La Pagode’, 2005). Previous maps of the site have simply noted that graves exist in these areas, but did not map their precise locations. All of these data have been compiled into a Geographic Information System, which brings together data from the topographic survey, excavation trenches, and efforts to accurately record burials at the site. Geophysical Surveys The first two weeks of the project were devoted to intensive geophysical surveys of the open areas at the site. Geophysical surveys offer unique opportunities to examine below sur-face anomalies based on difference in magnetism and moisture content, providing data that can be translated into a graphical representation to explore high probability areas for excavation. These surveys were conducted by Dr. Kate Welham and Mr. Harry Manley, of Bournemouth University. Three different survey readings were taken, including conductivity and magnetic susceptibility (with an EM38) and magnetometry (with a fluxgate gradiometer). These surveys covered more than 22,000 square meters of the site, in 20 x 20 meter survey units. Although formal results of these surveys are still being processed, in field results allowed the project to target areas of high probability during the field season. Trenches SM005, SM007, and SM011 were situated based on these surveys and will be detailed below. The final report on the 2009 geophysical survey will follow. Geoarchaeology (see Appendix 2) Another effort to begin understanding more ephemeral aspects of the archaeological re-cord was the application of geoarchaeology, including soil chemistry, phytoliths and micromor-phological sampling. This work was coordinated by Dr. Federica Sulas of the University of Cambridge. This work has the potential to delineate human activities within open areas and domestic spaces that may have only left chemical traces in the soil. Phytoliths will supplement and expand the paleobotanical record discovered through flotation (see below). Soil chemistry and phytolith samples (20g of soil) were taken along transects through open areas in the north-ern and central part of the site, and in many of the contexts in the excavation units. Micromor-phological samples were taken in areas of high human traffic, in House 44 and trenches SM006, SM007, SM013 and SM015. These samples, 25 x 25 cm blocks of soil taken from the profile of excavations, will allow for analysis of microstratigraphy within these trenches. Currently, all these materials are being processed at the McBurney Geoarchaeology Laboratory at the Univer-sity of Cambridge.

11

Mangrove study (see Appendix 3) Songo Mnara today is surrounded by stands of dense mangrove to the north and west. Mr. Jack Stoetzel, of the University of Virginia, conducted research in the mangrove swamps in an effort to determine how these natural resources have changed through time, and how human activity on the island may have impacted their growth and development. He conducted a coring survey along transects adjacent to, and perpendicular with the coastal line. Soil from these cores will be analyzed for pollen and wood charcoal, to determine if it will be possible to use these types of procedure for environmental reconstruction. This is the first such effort on the east African coast. Flotation In order to recover palaeobotanical remains, the project systematically took 2-5 liter soil samples from excavations and subjected them to flotation. This procedure, conducted by Mr. Dominic Pistor of Simon Fraser University, allows for the recovery of charred plant remains as well as other macroscopic artifacts. The samples from this work are currently being processed and analyzed by Dr. Sarah Walshaw of Simon Fraser University. Excavation Results What follows is a summary of each excavation trench dug in the 2009 season. In all, fifteen trenches were excavated across the site. Eight trenches were located within domestic architecture including six in House 44 (SM001, SM003, SM004, SM008, SM009, SM010) and two in House 23 (SM014, SM015). Two trenches were located just outside these structures (SM002 along the north wall of House 44, SM013 along the north wall of House 23). Three trenches were excavated in open areas (SM005, SM007, SM011), located according to anoma-lies from geophysical surveys. Finally, two trenches were situated around coral built features; SM012 surrounded a tomb and SM006 around a well. Trenches within domestic architecture: House 44

House 44 was identified for excavation for two reasons. First, it is a self-contained and relatively small structure (Figure 3). The single room-block represented by this house is much simpler than the palatial structures to the south of the site, and House 44 in particular seemed like a compact complex of rooms that would be simple to interpret, not complicated by issues of recognizing access points and axes of movement. Second, House 44 was representative of a set of individual houses in the northern part of the site. These smaller units seem to be of a differ-ent order to the complexes and palaces to the south of the central open area. Excavation was intended to cover both simple and complex, northern and southern, types of house. As per our larger strategy, House 44 was excavated in its entirety, as far as this was possible in the time available. This was an important part of our approach to the houses, which aims to reconstruct activities across the internal spaces, and the excavations here proved the usefulness of this strat-egy. The house was therefore excavated by room, with each room being given a different unit number (see Figure 3); the only exception to this was SM001, a sondage excavated during the first week at the site. The room in which SM001 was placed was later excavated in its entirety as SM003.

12

cont

ext

Loca

l pot

tery

, cou

nt

Loca

l pot

tery

, wt.

in g

r.

Impo

rted

pot

tery

, cou

nt

Impo

rt p

otte

ry, w

t. in

gr.

Gla

ss, c

ount

Gla

ss, w

t. in

gr.

quar

tz, c

ount

quar

tz, w

t. in

gr.

shel

l, w

t. in

gr.

grou

nd s

tone

1001 2 35 99 1003 34 241 1 0.9 2 3.6 406 1 1004 2 27 65 1006 28 125 63 1007 136 1275 2 32.9 167 1008 8 Total 202 1704 2 32.9 1 0.9 2 3.6 808 1

Table 1: Artifacts from SM001, by context

Figure 3: Plan of House 44, with trenches excavated in 2009

13

SM001 SM001 was a 1 x 1m test unit,

excavated in the southwestern room of House 44. The aim of this unit was simply to assess stratigraphy inside the house, before clearing the room more fully. An exploratory unit of this kind is essential to continuing large-scale exposures, demonstrating the depth of likely rubble fill, archaeo-logical layers and floor deposits (Table 1). In fact, SM001 encountered a very simple stratigraphy (Figure 4) which was mirrored throughout the house. A layer of disturbed earth and coral rubble covered the surface, ex-tending to a depth of approximately 23cm. This layer (#1001) was exca-vated quickly with jembes and shov-els, as the ground surface inside the house was much higher than the sur-rounding area and we assumed there would be a thick layer of rubble. In fact, #1001 immediately overlay a plaster floor (#1002). The cultural materials recovered from #1001 were therefore those that lay immediately on the floor of the room; this insight meant that the rest of the room (SM003) was excavated much more slowly, with attention to the thin layer of ceramics and other remains on top of the floor. Unit SM001 was excavated through the floor, which was unique for House 44: subse-quent excavations stopped at the level of the plaster. This showed a thick layer of sandy fill be-neath the plaster (#1003/1004), which seems to have been brought in to support the raised floor level. The sand was largely sterile, supporting this assumption; several contexts were recorded but these were cuts and fills within a basic homogeneity. At a depth of approximately 54cm, we encountered a darker sandy layer, with much more cultural material (#1007). This seems to be an activity layer that pre-dates the construction of the house, although similarities among the ceramics indicate that it is not very much older. This context corresponds with the lower levels in the back room of the house (SM010), discussed below. SM003 The entire south-west room of House 44 was then excavated, expanding on the results of SM001 (Table 2). This room covered an area of approximately 4 x 2.25m. As discussed above, the rubble from the top layer was removed much more slowly across this room, and several de-posits were recorded lying on top of the plaster floor. These seem to have been fairly coherent deposits, with ceramics and other materials left in certain areas when the house was abandoned. These ceramic concentrations were often associated with burning, which was assumed to be

Figure 4: SM001, east profile

14

linked to the preparation of food. Fish bones and blackened ceramics were found in these areas, and samples of the associated sediments were taken for flotation. Context #3001 was the rubble-filled layer that overlaid all rooms in the house, cleared fairly quickly using jembes and shovels. As we came down onto context #3002, we began to excavate more slowly with trowels and hand shovels. This sediment appeared quite similar to the topsoil, but the amount of coral declined substantially. #3002 surrounded and enclosed the series of archaeological deposits recognized in this room. In particular, concentrations of ce-ramics were mapped and recorded to the west of the northern door (#3003) and in the southeast-ern corner of the room (#3004). Next to the doorway a richer, crumbly, deposit was noted (#3007/3008) but this may have been partly formed by the degradation of the lime plaster floor in these areas (Figure 5). Still, there was a higher number of ceramics near the door. Excava-tions were halted at the plaster floor (#3009). SM004 The central room of the house (an area of approximately 8.5 by 2.5m) was then exca-vated, and showed a similar pattern (Table 3). The rubble layers at the top were particularly thick in this room, as there were several areas where the walls had fallen inwards. Contexts #4001 and #4002 were made up almost entirely of coral and plaster, and the subsequent layers were also extremely rubble-filled (#4003 and #4004). Beneath this rubble, the pattern was simi-lar to SM003, with a layer of slightly cleaner fill (#4005) above the plaster floor. Within that fill, several discrete contexts were recorded (Figure 6), which were all areas of concreted ce-ramics, in a matrix of denser dark sediment; this latter seems to have been created by the pres-ence of ash in these spots, as well as decomposing faunal and botanical remains. A particularly rich ceramic scatter was – again – located to the western side of the northern door. Here a larger area of burned material (#4008) contained a compact mix of ceramics and other materials. The entire area was extremely dark, with a lot of charcoal and organic matter. Samples were taken for geochemical and micromorphological analysis, which has confirmed the presence of burn-ing here, suggestive of non-food plant material. Excavations were halted at the plaster floor, recorded as context #4006. A series of contexts were recorded on the floor (#4007-4016).

cont

ext

Loca

l pot

tery

, cou

nt

Loca

l pot

tery

, wt.

in g

r.

bead

s, c

ount

Impo

rted

pot

tery

, cou

nt

Impo

rt p

otte

ry, w

t. in

gr.

spin

dle

who

rls, t

erra

cotta

spin

dle

who

rls, s

tone

shel

l, w

t. in

gr.

grou

nd s

tone

3001 1 3002 64 776 1 2 9.6 1 58 1 3003 13 316 3004 37 591 1 1.4 3007 12 115 3008 8 74

TOTAL 134 1827 1 3 11 1 1 58 1 Table 2: Artifacts from SM003, by context

15

Figure 5: SM003, Plan 3002

16

Figure 6: SM004, Plan 4002

17

SM008 SM008 was the entrance room of the house, an area of approximately 8.5 x 2.5m. The rubble-filled topsoil that covered it was very uneven, and the floor itself had been depressed in several areas, giving an uneven bottom surface. Near the door, deposits containing ceramics that had apparently once underlain the floor had been pushed up through the plaster, and the plaster reduced to dust. This end of the room, next to the door, also had almost no rubble cover-ing it, so the deposits had not been protected from trampling and weathering. The whole room therefore needed to be excavated carefully to account for these factors, and in places, this meant that shallow deposits were damaged slightly before it became clear that they were there. The area around the door, in which ar-chaeological materials were evident on the sur-face, and which had very little covering, was named context #8001, and the thick layer of

cont

ext

Loca

l pot

tery

, cou

nt

Loca

l pot

tery

, wt.

in g

r.

bead

s, c

ount

copp

er c

oins

, cou

nt

Impo

rted

pot

tery

, cou

nt

Impo

rt p

otte

ry, w

t. in

gr.

iron,

cou

nt

iron,

wt.

in g

r.

copp

er, c

ount

copp

er, w

t. in

gr.

quar

tz, c

ount

quar

tz, w

t. in

gr.

shel

l, w

t. in

gr.

4001 8 85 1 1 1.5 138 4003 11 416 68 4005 144 1719 14 1 1 1.8 332 4008 5 413 4009 52 529 1 1 9.4 1 1.1 18 4010 14 178 4011 1 3 4012 9 202 4013 13 233 4014 8 81 4016 16 318

TOTAL 281 4177 16 1 1 1.8 1 9.4 1 1.1 1 1.5 556

Table 3: Artifacts from SM004, by context

Figure 7: SM008, fingo and footscraper by doorway

18

rubble that covered the western end of the room was context #8002. #8001 was excavated with a trowel, revealing ceramics and bones in the area around the door, as well as a coral footscraper, inset into the floor to the side of the entry (Figure 7). In contrast, #8002 was removed using jembes and shovels, until we came down onto a single con-text along the length of the room - #8003. This sandy, silty sedi-ment contained several other contexts, which were concentra-tions of ceramics and artifacts deposited in certain areas. Of particular note are context #8005, immediately underneath the doorway, which was a concentrated area of pottery and bones, tentatively identified as a fingo offering beneath the floor, and #8006, a concentrated, ashy deposit against the north wall of the room (Figure 8). SM008 was a productive unit (Table 4), which could not be excavated in its entirety due to lack of time. Rather than excavating in a hurry, it was decided that the room would be back-filled, and excavations would be completed next season. SM009 SM009 was the southeastern room in House 44; this room was not fully excavated due to lack of time, and work was confined mainly to rubble clearance. Only a single context was recorded, #9001 (Table 4), which was the rubble-filled topsoil and mixed sediment that covered the floor. SM010 The back room of House 44 was excavated as a single unit, approximately 8.5 x 2.25m in size (Table 5). The entirety of the room had been covered by the internal wall, which had fallen southwards and was lying flat above the archaeological deposits. This had the effect of sealing the deposits, and also meant that – once the wall had been removed – there was very little coral rubble and disturbance in the layers beneath. It was interesting that no trace of any plaster ceiling was encountered in this room, suggesting that this back space may have had a makuti roof. Such an arrangement is also suggested for other buildings at the site, due to the sloping angle of the side walls in the back room. In House 44, the side and back walls did not survive up to full height and so could not be used for corroboration. The back wall was later suggested by a rubble fall in the archaeological stratigraphy (#10009), the circumscribed nature of which further suggested that it may not ever have stood to full height. Nevertheless, it is sus-pected that this room would have had a more open, makuti, roof: an interpretation supported by the identification of this room with cooking activities.

cont

ext

Loca

l pot

tery

, cou

nt

Loca

l pot

tery

, wt.

in g

r.

bead

s, c

ount

Impo

rted

pot

tery

, cou

nt

Impo

rt po

ttery

, wt.

in g

r.

Dau

b, w

t. in

gr.

spin

dle

who

rls, t

erra

cotta

shel

l, w

t. in

gr.

grou

nd s

tone

8001 37 642 73 8002 20 679 71 8003 189 787 4 289 8005 5 49 16 8006 141 4438 1 32.4 1.6 1 121 1

TOTAL 392 6595 4 1 32.4 1.6 1 570 1

9001 7 66.0 Table 4: Artifacts from SM008 and SM009, by context

19

Figu

re 8

: SM

008,

Pla

n 80

01

20

Figure 9: Plan of SM010, show

ing the top midden – contexts 10002, 10003, 10004 and 10005

21

Figu

re 1

0: P

lan

of S

M01

0, s

how

ing

ston

e fe

atur

es 1

0023

and

100

25

22

context spit

Local pottery, count

Local pottery, wt. in gr.

beads, count

copper coins, count

Imported pottery, count

Import pottery, wt. in gr.

Glass, count

Glass, wt. in gr.

Daub, wt. in gr.

spindle whorls, terracotta

spindle whorls, stone

jiko frag

ground sherd, count

lamp frag, count

ochre, count

iron, count

iron, wt. in gr.

copper, count

copper, wt. in gr.

quartz, count

quartz, wt. in gr.

shell, wt. in gr.

ground stone

1000

1A

3

62.0

10

001

B

48

748.

0

26

1000

1C

21

4 25

82.0

404. 6

1

1000

1D

32

.8

1000

1

60

744.

0

205

1

1000

2

226 5

1511

4. 526

1

9 14

6. 9 1

0. 6 32

6

1

1

1

1.3

1 0.

13

8.7

345 1

3

1000

2C

18

0 6

1000

3

684

7222

.018

1

2 1. 0

516. 7

2

1 7.

28

76.8

36

6 7 1

1000

4

251

2200

.07

1

4.2

435 8

1

1000

5

36

913.

01

94

2

10

006

0

0.0

0 1

2.8

1

47

10

007

15

16

2.0

3 1

1 46

.3

110

008

1

1000

9

44

298.

0

1

40

.5

104

210

010

18

5 14

75.0

10

012

87

72

6.0

3

37

8

1001

3

158

1264

.03

42

4. 1

10

014

29

2 24

43.4

7

1 1.

5 1

0.32

9.

1

3.1

1

Tabl

e 5:

Art

ifact

s fr

om S

M01

0, b

y co

ntex

ts (c

ontin

ued

belo

w)

23

context

spit

Local pottery, count

Local pottery, wt. in gr.

beads, count

copper coins, count

Imported pottery, count

Import pottery, wt. in gr.

Glass, count

Glass, wt. in gr.

Daub, wt. in gr.

spindle whorls, terracotta

spindle whorls, stone

jiko frag

ground sherd, count

lamp frag, count

ochre, count

iron, count

iron, wt. in gr.

copper, count

copper, wt. in gr.

quartz, count

quartz, wt. in gr.

shell, wt. in gr.

ground stone

1001

590

822.

03

120

510

016

741

8411

.040

52

46.7

10.

293

11

21

228

.121

901

1001

735

828

72.0

192

40.6

866

1188

110

018

175

1633

.01

10.

815

8.2

150

610

019

1411

8.0

5510

020

1872

1382

9.0

169

104

892

0.9

99.5

122

17.

12

0.7

963

.768

733

1002

134

838

37.0

51

25.5

31.

031

.93

13.

11

1002

222

116

48.0

31

9.3

10.

21

112

.61

0.4

957

110

024

346

3277

.013

53

35.6

10.

110

73

19.1

19.

312

802

1002

50

0.0

149

1002

655

334

07.0

318

230

.51

1.9

21

0.3

834

1002

729

029

22.0

231

162

.42

2.4

11

4.5

1800

1002

89

34.0

1002

945

671.

01

237

1003

027

288.

01

4010

033

256

2156

.04

15

17.7

563

310

034

1510

035

2317

2284

4.0

367

1324

02

0.3

41

19.5

120

6351

5.3

110

036

10.

610

037

257

2922

.06

10.

21

251

26.1

880

1

TOTA

L12

261

1076

44.9

751

3642

779.

319

9.55

1080

8.9

262

32

21

1186

.311

72.8

8871

0.6

2527

225

24

The stratigraphy of this room was complex; there was no plaster floor to provide a de-finitive identification of the living surface. Instead, an earth floor seems to have overlain the sandy fill, which was excavated down beyond the foundations of the walls. It was therefore pos-sible to identify a pre-construction layer of activity, which – although separated by layers of construction debris - contained ceramics and cultural materials of a similar type to those seen in the occupation of the house. Immediately beneath the fallen wall we excavated context #10001, which was a mixed, rubble-filled sediment made up of loose silty sands. This thin layer quickly came down onto richer sediments with more organic content, and dense artifactual content. These contexts were named #10002 in the eastern end and #10003 in the western end, reflecting a darker, richer ma-trix to the east. This was associated with buried midden deposits, and the distinction between #10002 and #10003 became less clear as we proceeded down. Within #10002 we excavated and recorded a dense midden, full of shells, ceramics and other artifacts. This midden was also overlain by a layer of burned daub, suggestive of some kind of previous structure or room divi-sion. The daub (#10004) never resolved itself into a distinguishable pattern, however, and was instead mixed in with the more general midden deposit to the western side of the north door (#10005) within which several discrete contexts were identified, such as an area of burning, named #10006. These all sat within context #10002, which continued down around the various midden deposits (Figure 9).

It was not possible to identify a floor layer, although this was hinted at by areas of red earth encountered beneath the midden #10005. These areas, #10012 and #10014, were both around the doorway and were immediately overlain by the midden and by rubble fall from the door. The layer was not seen elsewhere in the room, but it could have been preserved by the deposits lying on top.

This possible floor layer separated the top midden #10005 from another, dense, midden that seemed to extend down into the fill of the floor. This was a dense, dark, organic-rich de-posit, with large quantities of shell debris along with charcoal, bones and infrequent chunks of daub. #10014 was also identified as an extension of this scatter, which was much looser having been broken up beneath the rubble falling from the door.

Beneath this wealth of debris, the excavations came down onto a cleaner, more homoge-neous silty sand (#10016) across the entire room. This context replaces the previous designa-tions #10002 and #10003, as the distinction between east and west had become unclear. The sediments were, however, very similar. #10016 was interpreted as the fill that would have un-derlain, or served as, a floor. Although traces of the dense midden continued into this layer (#10017/#10018) it appeared relatively clean. Within this fill, as in SM001, several areas of darker soil, apparently representing small pits and fills (e.g. #10019) were identified and then the entire unit came down onto a clean fill layer (#10020) which it was assumed would be ster-ile. This layer was excavated by shovel.

Within and beneath this fill layer, however, we continued to identify contexts that corre-sponded to past activities. Of particular note are a pair of stone features that emerged in the western end of SM010 (Figure 10). One (#10023) seems to have been the footprint of a small wall that divided off this end of the room, although only a single row of stones remained. The second (#10025) was a circle of stones which upon excavation turned out to be a latrine, built in the corbelled shape known from Swahili sites elsewhere. This feature, which would have been beneath ground level in the room, had been infilled, and was excavated separately (fill context #10033).

25

Figu

re 1

1: P

lan

of S

M01

0, s

how

ing

low

er m

idde

n (c

onte

xts

1003

5, 1

0034

, 100

37)

26

In addition, excavation beneath the floor continued to reveal midden deposits (Figure 11), with one particularly dense lower midden, #10035, apparently associated with an earlier occupation or the construction of the house. An exceptionally-rich deposit, this midden con-tained ceramics, shell, bones, and beads in abundance. One special find – the shell of a sea tur-tle – was excavated as a separate context within this midden (#10036).

In a corresponding layer, representing the pre-construction levels of the house, a scatter of porites coral chips was recorded within the sandy fill (context #10034). Although this was dense and distinctive, it was not possible to excavate separately, due to the sandy loose nature of the sediments that surrounded it. It seems likely that this is the remains of carving porites blocks for the doors and finer features of the house, and would have been left at the time of con-struction.

Subsoil (#10039) was only reached in two places in SM010: at the bottom of the latrine, and in the central area of the room. Here, we had chosen to excavate a 1m trench through the centre of the midden deposits, since time would not permit full excavation of the room. Here, subsoil was reached at a depth of approximately 1 m, beneath context #10035. Summary of House 44

Across the house, excavations have demonstrated that it is possible to recover areas of activity, represented by distinct deposits. These seem to have been left in the rooms when the houses were abandoned, and careful excavation is able to recover spatial information from these layers. The initial results also give an indication that particular areas might have been used for specialized purposes. The deposits to the side of the doors suggests the patterning of discard and storage practices in particular spaces, which is an aspect to be further explored in the future. The general spread of activity also suggests much more diversification of activity than has been assumed by idealized models of the Swahili house, with cooking, food processing etc. indicated at every level.

Artifacts are not discussed in detail here, although they are catalogued by context throughout (Tables 1-5). The assemblage from the house included many ceramics, although only a few imported ceramics of specific types associated with the fifteenth century. There was no obvious concentration of imports in particular areas and, instead, they were found among the locally-produced wares. Excavations in House 44 also recovered large quantities of beads and Kilwa-type coins. SM010 produced the largest collection of coins found in the 2009 season, among the midden deposits and floor fills. Clearly, finding so many small artifacts is aided by a very careful excavation technique and – in particular – the sieving of sediments using a 2mm mesh. In the central room (SM004), many beads were found when the plaster floor was brushed for photographs, showing both that they are difficult to identify with the naked eye, and that they were slipping through the sandy sediments to come to rest on the plaster surface. A final interesting feature of the artifact assemblage was the presence of spindle whorls throughout the house, as well as a concentration of these artifacts in SM010. These finds suggest that cotton production might have been occurring inside the house, possibly associating it with women’s labor. Such an interpretation would also begin to allow us to integrate the house into the eco-nomic life of the town, if products intended for trade were coming from such household-level production.

Finally, the excavations throw some light onto the architecture and construction of the houses. As Garlake (1966, p. 18) claimed, the walls of House 44 had no real foundations, in-stead they were simply built on a sandy level, directly onto the ground. The foundation trenches

27

seen along the walls are shallow, and were then in-filled with the same sand used as packing beneath the floors. Finds in SM010 also suggest the porites coral used for doorways and orna-mental features was finished in situ, with the fine carving around the doors and doorsteps likely achieved once the coral was in position. This complements Chittick’s (1974, p.39-40) observa-tion that at Kilwa Kisiwani lime was burned for plaster on the site of houses under construction, suggesting that the full range of work on the materials was accomplished on site. No lime-burning pits were identified in House 44, but it should be noted that not all rooms were exca-vated down to pre-construction layers, in order to preserve the plaster floors. Trenches within domestic architecture: House 23 House 23, at the southwestern corner of the site, was selected for excavation as a representative of the grander, more ornamental architecture found on the southern side of Songo Mnara. The house has many of the features of the grandest architecture at the site, including a monumental stairway leading down to the Southern Open Area, a stepped internal courtyard, and many orna-mental wall niches in the interior spaces (Figures 12 and 13). The house is a relatively small example of these grand structures, however, and this recommended it to us for excavation. Due to a lack of time, House 23 could only be sampled, and the courtyard and a single interior room were investigated (Figure 14). For this reason, the range of activity is less well understood than for House 44; returning for fuller excavations of this structure will be a priority for future sea-

Figure 12: Courtyard in House 23, facing west

28

Figure 13: Niche in House 23; Trench SM015 in front

29

sons. Nevertheless, it is possible to ascertain differences from House 44 in the use of space and the interior features of the house. SM014 SM014 was a 4 x 1m unit excavated in the courtyard of House 23. It was positioned to cut through the steps that are a feature of these sunken courtyards, sampling the top and the base for evidence of activity. Initially begun as a 3 x 1m trench, the length was extended during excavation as the extent of the steps became clearer. The trench was excavated from the top down, in an attempt to clear off the steps and differentiate between the deposits above and below them. This was never entirely achieved, as the steps were very broken and disturbed, by the large quantities of rubble that had fallen into this courtyard since its abandonment. Thus, the stones that lined the steps were broken and dis-rupted, making it difficult to distinguish between the deposits above and beneath. There was very little guidance from the archaeological stratigraphy, as there was little in the way of arti-facts or occupation debris. It would appear that this space was kept relatively clean, or that this part of the courtyard was not a major locus for debris or storage. Few contexts were recorded (Figure 15). The majority of the sediments fell into context #14001, which was the rubble-filled, mixed, surface layer. The stones for the steps were en-countered almost immediately beneath this (#14003). Although care was taken to excavate the deposits immediately above (#14002) and below (#14004) the highest steps separately, there was little difference in the sediments recorded. Artifacts were very few. Only two shallow steps were identified, both faced with stone blocks, apparently set in a lime plaster, now extremely degraded and broken. The floor in the centre of the courtyard was excavated down to hard, pink, coral bedrock, found very close to the surface on this southern side of the site. It seems that the courtyard would originally have had either this surface, or a think covering of packed earth. No plaster floor was found in the center. A single burned context (#14008) was identified through excavation, apparently repre-senting a burned-out stump of a small tree. Although the burned roots extended beneath the bot-tom step, it is impossible to know if this tree was inside the courtyard when it was occu-pied, or if it grew and was burned after abandonment. In sum, the excavation of SM014 provided an impor-tant glimpse into the architec-ture and use of the courtyard, although it seems to have dem-onstrated that this area was kept quite clean. Finds were extremely few, with a single spindle whorl and bead stand-ing out in an assemblage that otherwise included only small amounts of ceramics and shell (Table 6).

cont

ext

Loca

l pot

tery

, cou

nt

Loca

l pot

tery

, wt.

in g

r.

bead

s, c

ount

Impo

rted

po

ttery

, cou

nt

Impo

rt p

otte

ry, w

t. in

gr.

Dau

b, w

t. in

gr.

spin

dle

who

rls,

terr

acot

ta

quar

tz, c

ount

quar

tz, w

t. in

gr.

shel

l, w

t. in

gr.

14001 12 80 1 3 16.9 1714002 6 45 11.0 2 4.6 1114004 33 302 1 4.2 2.1 19614005 53 392 1 2.4 1 2614006 21 100 9.0 12014007 14 69 14.7 77

TOTAL 139 988 1 1 4.2 39.2 2 5 21.5 447 Table 6: Artifacts from SM014, by context

30

Figure 14: House 23 plan, w

ith Trenches excavated in 2009

31

Figu

re 1

5: P

rofil

e of

SM

014,

sho

win

g co

ntex

ts re

cord

ed

32

context

Local pottery, count

Local pottery, wt. in gr.

beads, count

copper coins, count

Imported pottery, count

Import pottery, wt. in gr.

G lass, count

G lass, w t. in gr.

Daub, wt. in gr.

spindle whorls, terracotta

spindle whorls, stone

jiko frag

ground sherd, count

iron, count

iron, wt. in gr.

copper, count

copper, wt. in gr.

quartz, count

quartz, wt. in gr.

shell, wt. in gr.

ground stone

15001 12

228

15002 231

1877 11

1

1

0.1

2

1 8.6

477

15003 224

1520 8

3

45.72

0.4

1 1.3

1071 1

15004 84

954 1

167

15005 38

268

15006 141

1032 5

1

9.2

1 5.7

126 1

15007 296

1755 12

7

36.72

0.2

1

5 20.9

358

15008

5

1

0.3

1 3.8

348 1

15009 291

2427 21

1

0.3

3

6 43.8

1620

15010 673

6029 106

3

8.11

0.3

11

11

4 6.6

1259 1

15011

1

15012 10

36 2

129

15013 55

251 16

1.8

2

14.2

14.2

5 16.7

124

15015 14

341 3

67.3

2 5.4

21

15016 23

90 4

1

0.8

31

15017

8

TOTAL 2092

16808 202

1 15

100.58

1.5569.1

101

11

14.2

14.2

26 112.8

5731 4

Table 7: A

rtifacts from SM

015, by context

33

SM015 SM015 was a 2 x 2m unit excavated in the central room of House 23. This room con-tains two porites wall niches, to either side of a central southern doorway. SM015 was located beneath one of these niches, to the west of the room. The southern, western, and northern walls of the trench were room walls, allowing access into the corners and the space beneath the niche. This room was filled with a thick layer of coral rubble. This layer, #15001, was exca-vated using jembes and shovels, and was not sieved; it continued for 90cm (Figure 16). It seems that the depth of rubble in this room was due to the falling in of an upper storey, meaning that the entire house is deeply buried. Immediately beneath the rubble, we encountered a bright red, packed earth deposit (#15002). This red earth characterized the stratigraphy of the trench, and it seems that it may have been the floor material in this room. No plaster floor was encountered, although an ex-tremely dense, concreted layer was met within the red earth (#15003) which may have been the floor. A micromorphological block was taken from the trench profile at the transition with this context, to assess the nature of the floor and check for possible traces of a pulverized plaster layer.

Figure 16: Profile of SM015, showing contexts recorded

34

Despite its position in a prominent area of a decorative room, SM015 found a dense ac-cumulation of debris and burning, apparently associated with food consumption. This was con-centrated in the center of the trench (#15004) but extended as a layer across most of the area of the trench (Figure 17). The central concentration seems to have been inside a pit, dug directly into the packed earth floor. The general spread of debris was incredibly rich in artifacts (Table 7), with ceramics, slag, beads, a single coin and quantities of fish bones. There was also a par-ticular concentration of spindle whorls. An apparently separate and more diffuse midden layer (Figure 18), containing a lot of charcoal and ash, was found beneath the main packed earth floor levels (#15010) and overlay

Figure 17: Plan of SM015, showing context 15004

35

the sandy fill that seems to have been brought in to raise up the floor (#15013). Builders’ trenches along the walls were visible in this layer, as they were in-filled with a darker, artifact-rich fill (#15011). As elsewhere, the sandy fill overlay a surface with traces of pre-construction activity. In SM015 this included baked earth features, and a concentration of iron slag. The density did not suggest intensive iron-working on this spot, but there had certainly been some kind of activity area here that was covered by the sand brought in.

Figure 18: Plan of midden layer in SM015

36

This trench is hard to interpret due to its size. The features identified often extended be-yond the walls of the trench, and it was also not clear where the exact floor level was. Neverthe-less, there are several interesting features. A pit with evidence of burning and a high density of ceramic and faunal debris in the centre, below the niche, is perhaps surprising, as this area might have been expected to be kept clear of such domestic debris. The high quantities of spin-dle whorls indicate the spinning of thread in even these ornamental spaces. The packed earth floor is surprising, in this room of elegant architectural detail, and contrasts with the plastered floors of House 44, ostensibly an architecturally-simpler house. Clearly, further excavation is required to allow us to interpret the activities represented here. Trenches outside domestic architecture

Two trenches were excavated just outside houses in an effort to understand the stratigra-phy outside of structures, and to gauge the integrity of archaeological deposits. This is part of the larger effort to examine more public and open spaces. Although the soils were quite differ-ent in these trenches—that of SM002 much looser and sandier than in SM013, the nature of the stratigraphy was similar. In each, a cultural layer with pottery scatters and other artifacts was easily identifiable, sealed beneath rubble from architectural collapses and deposits that built up after the site was abandoned in the 16th century. These trenches demonstrate the rich potential that larger exposure excavations could have in delineating activities occurring outside of houses. SM002 Trench SM002 was a 2 x 2m trench located along the northern wall of House 44, outside the structure, excavated by Fleisher, Helen Horn-Mitchem, and Babatunde Babalola (Table 8). This trench was situated on the exterior wall thinking that it was associated with an exterior doorway (Figure 3). Although the doorway was located on the eastern side of the building, this trench still contained some interesting deposits (Figure 19).

Topsoil and grass was removed as context #2001, and contained much rubble and loose fill. Directly below was a dense concentration of coral rag, recorded as a masonry context, #2002; this is likely the remains of wall fall. Just below and surrounding #2002 was a fine sand silty loam context, #2003, removed in two spits: spit 1 contained coral rubble along with mixed modern and 15th century artifacts (copper Kilwa coin, spindle whorl, beads) while spit 2 con-tained less coral rag. Context #2004, a scatter of local potsherds, was revealed below #2003, located in the northwest corner of the trench. At this level, a small feature (#2005) was re-vealed extending into the western profile, semicircular in shape; this context was excavated and its cut (#2008) recorded. Context #2003 also sealed #2006, a loose sand layer which contained a number of special finds. Finally, #2007, a loose sandy soil, was revealed below #2006, and represents the top of subsoil, although artifacts in small frequencies were found in it, including pottery, bone and glass. Because this was the first trench of the season to reach subsoil, a 1 x 1m test unit was excavated in the northeast corner to make sure that #2007 was actually the top of subsoil; fill from this test pit was recorded as #2008.

37

Tabl

e 8:

Art

ifact

s fr

om S

M00

2, b

y co

ntex

t

context

spit

Local pottery, count

Local pottery, wt. in gr.

beads, count

copper coins, count

Imported pottery, count

Import pottery, wt. in gr.

Glass, count

Glass, wt. in gr.

spindle whorls, terracotta

spindle whorls, stone

jiko frag

copper, count

copper, wt. in gr.

quartz, count

quartz, wt. in gr.

shell, wt. in gr.

ground stone

2001

211

4.3

20

03

1 15

1 30

49

4 1

1 13

.82

43.2

31

1

20

193

120

03

2 27

8 16

58

16

1

613

20

04

51

44

5

2006

116

680

8 1

10.

1

1 .8

57

20

07

30

77

2

1

0.2

72

20

09

104

TO

TAL

62

6 59

09

30

2 1

13.6

615

83

11

1 .8

1

20

1039

1

38

SM013 This was a 2 x 2m trench at the base of monumental steps leading from House 23, exca-

vated by Elizabeth Johnson (Table 9). The trench was located directly at the base of the stair-case, centered on them (Figure 14).

Excavation of the trench began with hoes, removing topsoil and rubble, as the top level was dominated by coral rag from the collapsing stairway directly to the south (#13001). Imme-diately below this fill, we located an additional step. This masonry feature included thick flat stones (#13003) resting on a coral rag footer (#13005); the footer continued to the west, match-ing the width of the overall staircase. The trench was widened in this direction an additional 20cm, in order to expose the full step/footer. The fill from on top of the step was recorded separately (#13002). Once the rubble-filled context was removed, a much cleaner, dark gray-ish-brown fill layer was revealed and excavated (#13004). This fill context was excavated in two spits, one down to the level marked by the bottom of the coral rag footer, the other below.

Figure 19: Profile of SM002

39

Tabl

e 9:

Art

ifact

s fr

om S

M01

3, b

y co

ntex

t

context

spit

Local pottery, count

Local pottery, wt. in gr.

beads, count

copper coins, count

Imported pottery, count

Import pottery, wt. in gr.

Glass, count

Glass, wt. in gr.

Daub, wt. in gr.

spindle whorls, terracotta

copper, count

copper, wt. in gr.

quartz, count

quartz, wt. in gr.

shell, wt. in gr.

1300

1

123

735

3

2 6.

9

6.

6

1 0.

9 13

0 13

004

1 72

6 28

7451

13

43

1.0

6.8

2

70

46

.2

1104

13

004

2 63

5 35

1637

6

9.8

10.

3

18

42

1300

6

1259

13

049

13

007

12

20

1651

010

11

5 49

.32

1.3

42.5

1

5

47.9

10

141

1300

8

18

313

1

411

TOTA

L

3981

36

997

193

2 16

70

5

2.3

55.9

3 1

.3

76

95

1362

8

40

spit 2 of #13004 sealed a scatter of local pottery that covered the entire trench, sitting in a slightly compact dark brown silty sand loam (#13007). This surface likely represents the active cultural surface associated with the monumental staircase, resting just 20cm lower than the top

Figure 20: Plan of SM013, showing cultural layer

41

of the footer (Figure 20). When the pottery scatter and associated fill were removed, large frag-ments of coral rag were exposed; this is likely natural coral rag outcrop.

As time was short, and natural coral was encountered in the western part of the trench, excavations proceeded only in the eastern half. A final fill layer, #13008, was removed with a jembe; this layer contained only a few artifacts, including local pottery and shell. This layer was sealed by #13007 above, and sealed sterile sandy subsoil below. Trenches in open areas and around coral-built features Trenches situated in open spaces at the site (SM007, SM011, SM005), and around open air features such as tombs (SM012) and wells (SM006) were aimed at exploring the types of activities that could be discerned in these parts of the site (Figure 21). This is exploratory work, as there have been few excavations at Swahili sites that focused specifically on open areas. In general, the findings from these excavations supported those from trenches just outside of struc-tures: well-sealed and undisturbed cultural surfaces were preserved, with rich and variable arti-fact assemblages. Besides SM010 in House 44, the highest density of copper coins were found in these trenches with eight coins located in SM011 and seven in SM012. Additionally, these areas often contained well-defined and stratified pottery scatters that allowed for the definition of cultural surfaces even in the absence of significant soil changes. These trenches demonstrate a number of important findings: first, that there may have been earth and thatch houses within the coral built town; second, that artifact patterning is associated with active outdoor spaces, specifically those around the well (SM006) and tomb (SM012); and third, that industrial activi-ties can be isolated within the open areas, as indicated by iron smithing debris in SM005. SM005 Trench SM005 was a 2 x 3m trench in the area of an anomaly located through the geo-physical surveys (Table 10). Data from the magnetometry survey indicated that this area con-tained some of the highest magnetic readings at the site. The magnetometry readings were borne out by the excavations, with finds of iron scale, droplets and slag, from the uppermost contexts. These data suggest strongly the presence of iron smithing in the vicinity.

This trench was excavated over the course of a few weeks; it was begun by Kylie Klein on June 21st and once iron slag was encountered, the trench was halted for two weeks, so we could pursue other research goals. Helen Horn-Mitchem reopened the trench on July 6th and excavated it to subsoil.

The contexts in this trench were quite homogenous, represented by gray sandy silty soils from topsoil to subsoil (Figure 22). Contexts #5001, #5002, #5003, and #5004 contained simi-lar soils, but were separated due to higher artifact densities encountered in #5003 and #5004, including local pottery, bone, iron slag, and beads. Context #5004, excavated in two spits, con-tained a number of special finds, including two copper coins, grindstone fragments, a spindle whorl, and beads. This context sealed #5005 and #5007; the former was a fill lighter in color than stratigraphically higher contexts, but also sandy and silty, the latter was similar and con-tained frequent iron scale. Both of these contexts had many fewer artifacts than previous levels and were shallow contexts marking the transition from cultural layers to the subsoil, context #5008. Context #5009 was a darker, loose sandy silt fill, sealed by #5005 and sealing subsoil, #5008, containing one copper Kilwa coin. An isolated patch of dark fill at the center of the trench, #5010, contained yellow glass beads, pottery, bone and iron scale. This seems to be a patch of the darker fills that were pressed into subsoil (#5008); the cut was recorded as #5011.

42

Figure 21: Trenches excavated in central open area in 2009

43

Figu

re 2

2: P

rofil

e of

SM

005

44

SM006 Trench SM006 was a 2 x 3.5m trench bisecting the area around a circular coral-built well (Figure 23, Table 11). The trench was designed to examine the deposits surrounding the outside of the well, both in an effort to examine this open space and in the hopes of recovering useful geoarchaeological data through phytolith and micromorphological analysis. A small balk was retained against the southwestern exterior wall of the well, from where a micromorph sample was extracted after the trench was excavated. An uppermost context, #6001, contained grasses and other surface vegetation. This con-text also included the surface fill of a contemporary path that runs through the northern part of the trench, skirting the well; this area was more compacted than the rest of the trench, which generally contained a loose fine silt sandy loam. The top deposit sealed context #6002 which extended over the entire trench; this context was a slightly compacted, light brown silty sand loam. This context contained a high density of vessel glass compared with other contexts. Af-ter approximately 5cm of this context was removed, deposits in the northern part of the unit ap-peared darker and somewhat looser than those above it; this deposit was designated context #6003, and was restricted to the northern half of the trench. In the southern half, #6002 was removed to reveal a pottery scatter along the southern wall (Figure 23). At first this context was restricted to a small area in the southeast corner, but then extended to the southwest as well. This context contained many glass beads, and a few sherds of Islamic monochrome pottery. Once contexts #6003 and #6004 were removed, a dark brown loose silt sandy loam was re-vealed, #6005. This context covered the entire trench, and was approximately 5-10 cm deep. Artifact quantities decreases significantly in this context, with smaller amounts of local pottery and special finds. Context #6005 sealed a deposit that was a transition to subsoil; deposits were mottled, including loose dark brown fill, patches of lighter yellow sand, and a few areas of red

cont

ext

spit

Loca

l pot

tery

, cou

nt

Loca

l pot

tery

, wt.

in g

r.

bead

s, c

ount

copp

er c

oins

, cou

nt

Impo

rted

pot

tery

, cou

nt

Impo

rt p

otte

ry, w

t. in

gr.

Gla

ss, c

ount

Gla

ss, w

t. in

gr.

Dau

b, w

t. in

gr.

spin

dle

who

rls, s

tone

tuye

re, c

ount

shel

l, w

t. in

gr.

grou

nd s

tone

5001 191 326 12 1 1.4 1 0.2 72 5002 114 378 5 18.3 2 251 15003 3 32.5 83 5004 1 312 2248 1 2 1 5.9 39 1 886 15004 2 31.7 15005 42 224 1 18.4 5007 26 212 34 5008 98 553 6.2 553 5009 71 504 1 1 1 1.1 21.6 125 5010 29 199 2 155

TOTAL 883 4644 25 3 3 8.4 1 0.2 167.7 1 2 2159 3 Table 10: Artifacts from SM005, by context

45

Tabl

e 11

: A

rtifa

cts

from

SM

006,

by

cont

ext

context

Local pottery, count

Local pottery, wt. in gr.

beads, count

copper coins, count

Imported pottery, count

Import pottery, wt. in gr.

Glass, count

Glass, wt. in gr.

Daub, wt. in gr.

spindle whorls, terracotta

ground sherd, count

lamp frag, count

iron, count

iron, wt. in gr.

copper, count

copper, wt. in gr.

quartz, count

quartz, wt. in gr.

shell, wt. in gr.

ground stone

6001

69

17

8 3

1

0.1

10.

2 42

.2

60

02

564

2506

16

4 11

2.6

31.

6 12

5.3

1

1.8

1338

6003

11

84

8357

28

7 37

.1

10.

2 22

2.5

1

1

3.4

890

60

04

718

6989

21

1

6 53

.7

10.

9 14

1

1 3.

1

11

17

1 60

05

217

1554

1

2

13.6

1

0.2

1

11.

8 1

1.7

259

60

06

126

1009

5

1

0.6

433

60

07

7 29

1

0.

9

21

60

07B

1

1

3.2

TOTA

L 28

85

2062

2 76

1

20

217.

1 8

3.7

404.

9 1

1 1

3 9.

7 1

1.8

23.

540

58

1

46

Figure 23: Plan of SM006, showing pottery scatter (#6004)

47

Figu

re 2

4: P

rofil

e of

SM

006

48

soil. Also exposed in this context were irregular coral rag fragments along the exterior surface of the well, indicating that the deposit was cutting into the soils that had been intact when the well was constructed. Darker patches of soil against the exterior of well could have been from well construction. It also became clear that part of the well head had been exposed above the surface, since plastered surfaces became clear as contexts #6002/3/4 were excavated. A final context, #6007, was excavated at the bottom of the trench. This included a series of small dark brown circular stains cut into subsoil, possibly post holes but more likely root stains, based on their shallow and irregular shapes when excavated. Overall, the deposits in this trench were surprisingly rich in artifacts, especially vessel glass, glass beads, bone, and local and imported pottery. A midden, represented by pottery con-centrations in #6003 and #6004 may represent the living surface once associated with the well when it was in active use in the 15th and 16th centuries (Figure 24). SM007 Trench SM007 was a 1 x 4m trench across an anomaly located through the geophysical surveys, excavated by Babatunde Babalola (Table 12). The trench was positioned to cut from an area of high magnetic readings to one with much lower, standard readings. Overall, excavations revealed three main fill layers: #7001, #7002, #7003 (Figure 25). Context #7001 comprises the uppermost layer across the trench, but then continues deeper in the northern part of the trench; it is a slightly loose gray brown sand silty loam. After about 20cm of this fill was excavated across the trench, context #7002, a more compact, reddish brown fill, was revealed in the southern two-thirds of the trench. The transition between these contexts was important in defining the nature of the magnetic anomalies, yet defining that tran-sition was difficult as the soils shaded imperceptibly from one to the other. Excavations thus focused on either end of the trench, in order to determine what differences may exist between the fills. Context #7002, primarily in the southern part of the trench, contained local pottery, bone, glass and shell beads and copper coins. Context #7003 is directly below #7002, and sealed by it. In terms of compaction, color and composition, it is the same as #7002, but con-tained a higher density of local pottery and other materials; these artifacts were scattered throughout the fill, with no defined accumulations. In the northern third of the trench, #7001 continued unabated, with no evidence of the reddish brown soils of #7002 or #7003 present. However, approximately 40cm below the ground surface there was a pottery scatter set within #7001 fill; this scatter was recorded as two stratigraphically-distinct contexts, #7005 and #7006, but probably represents a dense concentration of pottery with a depth of 50-60 cm. These scat-ters and associated fill contained local pottery, glass and shell beads. These pottery scatters were found roughly at the same depth as context #7003, and therefore artifact densities were higher across the trench at this level, even though the soils were quite different. A loose sandy soil with many shells and cowries—subsoil—was found below #7001 in the north and #7003 in the south. It is difficult to interpret the cultural activities represented by the contexts in this trench. It may be that the increased artifact density found at approximately 40cm represents a cultural level associated with open air activities, the general use of the site. However, the differences in soil types, and the artifact scatters that seemed to mark the transition between them, may sug-gest a more specific activity or use. It is possible that this represents a melted earth house, with artifacts scatters marking the areas just outside the walls. However, in order to evaluate this proposition, a much larger exposure will need to be excavated, and careful attention paid to arti-

49

Tabl

e 12

: Art

ifact

s fr

om S

M00

7, b

y co

ntex

t

context

spit

Local pottery, count

Local pottery, wt. in gr.

beads, count

copper coins, count

Imported pottery, count

Import pottery, wt. in gr.

Glass, count

Glass, wt. in gr.

spindle whorls, terracotta

ochre, count

iron, count

iron, wt. in gr.

quartz, count

quartz, wt. in gr.

shell, wt. in gr.

ground stone

7001

1

355

1198

27

2

2.8

2 0.

3

1 22

8.7

1 1.

4 14

5 1

7001

2

441

2285

10

36

9

7002

B

16

1 44

8 20

1

50

7002

127

183

101

1 0.

4

20

7003

191

35.

6

1

32

7003

B

29

0 22

83

6

124

9.3

1 0.

1

45

6

7004

B

19

94

2

1

0.1

7004

C

35

23

2 10

70

04

10

34

8

1

0.6

39

70

05

34

39

9

70

06

2 33

37

6

1 70

07

20

67

3

1

0.1

68

TO

TAL

15

235

7599

11

53

725

8.3

6 0.

91

1 22

8.7

1 1.

4 11

79

2

50

Figure 25: Profile of SM007

51

fact densities and slight soil color differences. In many ways, this trench represents the promise of excavations in open areas, but also the challenges that will accompany their interpretation. SM011

Trench SM011 was a trench located in the area of an anomaly located through geophysi-cal survey, excavated by Kylie Klein. This trench was positioned in an area of higher magnetic readings, and meant as a test of one of the areas of red soil evident throughout the site. The trench was initially begun as 1 x 1m in size, but once a relatively high artifact concentration was encountered, after about 10 cm of excavation, the unit was expanded to 2 x 2m in size.

This trench seems to contain a primary occupation episode, represented mainly by #11001, a slightly compact, fine silt sandy deposit that covered the entire trench; set within this fill was a series of artifact concentrations (#11002, #11003, #11004 – Figure 26). Context #11001 was excavated in two spits: spit 1 was excavated until the artifact concentrations were revealed, at approximately 20cm below the surface, and spit 2 represented the fill below these concentrations. These artifact concentrations included two pottery scatters (#11002 and #11003) and an area with coral rag and stone concentrated together (#11004). The fill in and around these features included (in addition to local pottery) bone, copper coins (n=6), and a large number of glass and shell beads. Context #11001, spit 2 contained similar artifacts, in-cluding spindle whorls, imported pottery, and two more copper coins.

Another concentration of pottery and coral rag was sealed by #11001, spit 2, located in the southwest corner of the trench. This concentration, stratigraphically below #11004, seemed to be distinct from it. This context contained large fragments of local pottery, as well as a star-shaped terracotta oil lamp similar to one excavated at Kilwa. Also sealed below was #11005, a less compact, dark grayish brown fill in the northeast corner of the trench, and #11007, a dark red circular area of compact sandy loam (Figure 27). Within #11007 were small fragments of what appeared to be burned sand. This feature was sectioned and removed (the cut of which was recorded as context #11008). With all of these isolated deposits removed, the remaining fill, a fine silty sand, was subsoil (#11009).

Based on the types and number of artifacts located in this trench (Table 13), it is possi-ble that it represents the remains of a melted earth and thatch house. However, no daub was located in the excavations, nor architectural evidence such as post holes. However, the concen-tration of coral rag in particular places, along with the wealth of glass and shell beads as well as coins, does suggest that this may have been a structure with an earth floor. SM012

SM012 is a 3 x 3m trench surrounding Tomb 12 excavated by Sarah Nouri (Figure 28, Table 14). This trench was situated around this stepped tomb in order to investigate whether there was archaeological evidence of activities associated with this tomb, including ritual offer-ings. This tomb seemed promising since there was ceramics scattered on the tomb, likely from more recent offerings (Figure 29); additionally, there were two niches on the eastern and west-ern ends of the tomb which likely served as repositories for offerings. The surface ceramics were collected separately, as #12001. The dense root mat was removed as #12002; this context was very loose silty sand and, because the tomb was located under a tree, had much organic ma-terial. There was also some coral rubble in this context, likely collapse from the tomb itself. In this context a carved porites headstone fragment (SF#12036) was located with Arabic script on it (indecipherable). This fragment may have come from the tomb, but it was not clear from

52

context

spit

Local pottery, count

Local pottery, wt. in gr.

beads, count

copper coins, count

Imported pottery, count

Import pottery, wt. in gr.

Glass, count

Glass, wt. in gr.

Daub, wt. in gr.

jiko frag

lamp frag, count

iron, count

iron, wt. in gr.

copper, count

copper, wt. in gr.

quartz, count

quartz, wt. in gr.

ground stone

11001 1

387 2589

193 6

1 1

20.8

0.31

14

14.51

11001 2

707 3774

127 2

15.7

10

130.7 3

2.5

3

11002

34 839

11003

57 988

1 9.8

11005

51 438

4

3.1

7

87.6

11006

1 59

1

1

11007

4 25

6

1

1.9

11009

20 107

17

0.8

2

9.3

TOTAL

1261

8819 347

8 2

10.82

0.819.9

2 1

19227.6

3 2.5

15 16.4

4

Table 13: Artifacts from

SM011, by context

53

Figure 26: Plan of SM011, showing contexts 11001, 11002, 11003, and 11004

54

Figure 27: Plan of SM011, showing contexts 11005 and 11007 in corner

55

Tabl

e 14

: Art

ifact

s fr

om S

M01

2, b

y co

ntex

ts

context

spit

Local pottery, count

Local pottery, wt. in gr.

beads, count

copper coins, count

Imported pottery, count

Import pottery, wt. in gr.

Glass, count

Glass, wt. in gr.

Daub, wt. in gr.

spindle whorls, terracotta

ground sherd, count

ochre, count

iron, count

iron, wt. in gr.

copper, count

copper, wt. in gr.

quartz, count

quartz, wt. in gr.

shell, wt. in gr.

ground stone

1200

1

35

1170

8

12

002

61

16

05

1

1.2

1

4.4

450

112

003

1 70

41

4 3

2

2

2 2.

3 29

7 1

1200

3 2

182

798

5

3 7.

2

12

004

1 78

88

7 7

1

0.9

0.8

1 1.

6 41

2

1200

4 2

386

2518

7

1

1 1.

1 10

00

12

005

1 10

5 66

8 7

11

1.1

1

3

9.4

265

12

005

2 52

7 25

17

21

26

17.6

1

112

006

10

22

7

12

007

18

3 89

7 8

3

22.8

53

7 1

1200

8

172

1481

3

2

5.4

13.

5

1

1 6.

9 18

68

112

009

40

3 24

30

7 1

1 1.

9

2 18

.2

1229

1201

0

229

1295

7

2

1

0.5

358

12

011

67

93

0

1

0.9

1

51.1

16

82

12

012

30

0 27

0 10

1 0.

6

3

1.7

1507

TOTA

L

2808

18

107

86

721

60

.4

24

21

1 1

1 4.

4 3

1.7

11

90.6

96

13

5

56

where it fell. Once the root mat was cleared, the trench was excavated in three zones: one each to the north, south and east. This was done to provide some control for the artifacts after it be-came clear that the deposits were somewhat homogenous in the trench. The soil throughout the trench was a loose, grey brown silty sand. In each of these zones, the deposits were excavated as follows: a first context, excavated in two spits, sealed by the root mat (#12002); beneath this was a context that marked the transition to subsoil; and then a subsoil context (context numbers summarized in Table 15).

Within the trench, one pottery concentration on the eastern side of the tomb was re-

corded (#12006). This concentration was located in the main fill deposit, #12004. These exca-vations also revealed two more headstones (G29 and G30, Figure 30), to the north and south of the trench, buried below 30-40cm of fill. This implies that many other headstones/footstones remain buried in the central open area of the site. One of the most surprising aspects of this unit was the richness of the artifact assem-blage associated with the deposits (Table 14). Based on the depth of the base of the tomb, it is clear that most of the deposits in the trench were deposited after the tomb had been constructed; there is no evidence of any earlier occupation levels below the tomb itself. A number of special finds may be related to ritual activities carried out near the tomb: rounded quartz pebbles pos-

Figure 28: Tomb with trench SM012 surrounding it, facing southwest

Location South East North Main context

12003, spits 1 and 2

12004, spits 1 and 2

12005, spits 1 and 2

Subsoil transition

12007 12008 12009

Subsoil 12010 12011 12012 Table 15: Context numbers in SM012

57

Figu

re 2

9: P

lan

of S

M01

2, a

t gro

unds

urfa

ce, w

ith p

otte

ry s

catte

r (12

001)

58

Figure 30: Plan of SM012, w

ith headstones adjacent to tomb

59

sibly left as offerings on the tomb, a practice that continues to occur at the necropolis grave-yard; seven copper coins; a significant quantity of imported pottery. Conclusion The 2009 field season at Songo Mnara successfully addressed our research questions on domestic and open spaces and provides a firm foundation for future research on these issues. As previous researchers at the site have noted, the material assemblage from Songo Mnara is surprisingly limited given the monumentality of its architecture and its proximity to the power-ful center of Kilwa Kisiwani. In short, the architectural wealth of the site does not seem to ex-tend fully to wealth in material culture. For example, the 2009 excavations demonstrated a sur-prising dearth of imported pottery at the site; overall the imported sherd ratio (ratio of imported to local sherds) was less than 1% (.5%) far less than ratios from other sites which often range from 2 to 6%. The small numbers in the archaeological deposits also contrast with the signifi-cant number of imported ceramics that were built into structures at the site. For example, in the palace structure, an intact barrel-vaulted room still contains 121 small glazed imported bowls (Garlake 1966, p. 38); additionally, Burton (1872, p. 359) notes that the mosque at the Necropo-lis (which he calls the “Nabhani” mosque) once contained Persian painted tiles surrounding the mihrab. Other artifacts were found in similarly small quantities, including glass vessel frag-ments, and objects of iron and copper.

Two artifact classes, however, were found to be much richer than previously noted, coins and beads. Copper coins of Kilwa type were found across the site (63 in total), and in relatively high densities in a number of trenches including SM010, SM011, and SM012. Based on an initial analysis of these coins, they represent at least eight different types, related to indi-vidual sultans at Kilwa (Fleisher and Wynne-Jones n.d.). The distribution of these coins across the site, in houses, open areas, and around the tombs suggests that they were in common use, and offer the possibility of providing the first contextual understanding of coins at a Swahili settlement. Similarly, close to 2,000 glass and shell beads were recovered; they were common in most trenches, often related to floor deposits. These beads are currently being analyzed by Ms. Marilee Wood. The recovery of coins and beads at Songo Mnara, where few of these ma-terials had been previously recorded, is due in large part to the recovery techniques employed, including the sieving of all deposits with 2mm mesh.

For the purposes of this project, it was significant that excavations showed the integrity of the deposits at Songo Mnara. Although they are somewhat thin compared to other coastal sites, the strata identified represent a host of primary contexts, with horizontal variability evi-dent across the areas investigated, inside and outside the houses. This is amply demonstrated by the variation evident; rather than a scatter of undifferentiated remains, excavations recovered definite activity areas and artifact associations.

For example, the deposits excavated in Houses 23 and 44 represent a range of activities that seem to be relegated exclusively to the houses and their associated spaces. These include the spinning of thread, evidenced by concentrations of spindle whorls in the household deposits, and cooking, based on the presence of terracotta jiko fragments and food debris. The focused distribution of these materials shows that debris from household activities remains in context in the households themselves. Differential distributions of other finds, like copper coins and im-ported pottery, higher frequencies of which were found in house deposits, suggests that these were more commonly used in these contexts. House 44, in particular, demonstrated the future potential of excavating household remains, with many instances of ceramic scatters on house

60

floors, evidence of burn patterns that may correlate with household activities, and possible ritual deposits of ceramic vessels in the doorways and corners of the structure. We are also encour-aged by the variability of the deposits within and between houses. In House 44, for example, although most rooms had plaster floors, the back room (SM010) evidently had a sandy earthen floor and may have been partially open to the outside. This unusual room also had a distinctive material assemblage, rich in both productive household activities and special finds like coins. The presence of this type of variability within houses will allow us to begin exploring the way interior house space was used in diverse ways. Comparison of deposits from House 44 and 23 show very different patterns, with no plaster floors observed in House 23, and radically differ-ent types of fills used for floors and occupation surfaces. This was contrary to expectations, given that we expected House 23 to offer a more refined architectural space based on its loca-tion and the grandness of the structure. Finally, the location of pre-architectural deposits under both houses offers the possibility of examining the development of the town from its earliest phases. Although these pre-house deposits were not significantly earlier than those related to the houses themselves, they do provide a sense of the settlement before large-scale architecture was constructed. Testing and excavations in the open areas were equally productive and instructive. Geo-physical surveys have provided a range of anomalies to be investigated further. During the 2009 season, we tested three of these anomalies (SM005, SM007, SM011). Clear evidence of iron smithing was found in SM005, a somewhat surprising find in such a central location at the site. Further work in the area will provide a sense of the scale of this production, and its rela-tionship to other deposits in the open area, as well as the burial area that it skirts. Trenches SM011 and SM007 both exposed what appear to be intact deposits related to the 15th and 16th century ground surface. SM011 might be the remains of an impermanent structure, as the evi-dence from it indicates domestic activities. However, much larger exposures will be needed to fully understand the nature of this area. Although the deposits in SM007 are ambiguous, they point to the possibility that intact deposits related to open areas can be isolated and described. The two trenches associated with coral-built features in open areas provide a sense of the rich possibilities that such excavations hold. Trench SM006, which bisected the exterior of a coral rag lined well, exposed a rich assemblage of material associated with the space just adja-cent to the well. It was perhaps surprising to find such a large number and wide variety of cul-tural materials in this space; imported and local pottery, beads, and glass fragments were com-mon in these deposits, and indicate that this may have been as much a social space as it was a functional one. Similarly intriguing are the deposits found in SM012, the space surrounding a coral rag tomb in the central cemetery. Finds from this location might be indicative of ritual practices associated with the tomb; significant quantities of local and imported pottery, copper coins, and small quartz pebbles might be evidence for offerings left next to or on the tomb it-self. The deposits all seem to post date the construction of the tomb structure, and therefore are not likely midden deposits. Further study of the local ceramic forms will be an important way of determining whether these vessels were for these special uses. Notably, as excavations be-gan, the surface deposits contained what appeared to be recent ceramic vessels, indicating con-temporary offerings related to this tomb; similarly, some of the headstones in the Necropolis cemetery had coins and quartz pebbles left as offerings on them, the latter practice described by Burton (1872:359) in the 19th century. These historical and contemporary practices might pro-vide important analogs for ancient ones.

61

In sum, the 2009 field season provides a firm foundation from which to continue re-search at the site: new topographic and architectural renderings of the site are currently being produced; geophysical surveys are providing a rich understanding of the subsurface deposits in open and other areas; geoarchaeological research is offering clues to more ephemeral practices; and the archaeological excavations provide an initial sense of the rich material assemblages re-lated to different parts of the site. In future seasons we aim to continue this research, working to build a more comprehensive understanding of life in this ancient Swahili town, and the way that material practices and the use of space were central to its inhabitants and its history.

62

References Burton, R. 1872. Zanzibar: City, Island, and Coast. London: Tinsley Brothers. Dorman, M.H. 1938. The Kilwa Civilization and the Kilwa Ruins, Tanganyika Notes and Re-

cords 6: 61-71. Chittick, H.N. 1974. Kilwa: An Islamic Trading City on the East African Coast. Nairobi: British

Institute in Eastern Africa. Chittick, H.N. 1961. Excavations at Songo Mnara, Annual Report of the Department of Antiqui-

ties, pp. 4-6. Dar es Salaam: Government Printer. Donley-Reid, L. 1990. A Structuring Structure: The Swahili House, in S. Kent (ed.) Domestic

Architecture and the Use of Space, pp. 114-126. Cambridge: Cambridge University Press. Fleisher J. 2010. Housing the Market: Swahili Merchants and Regional Marketing on the East

African Swahili Coast, 7th-16th Centuries AD, in C. Garraty and B. Stark (eds.), Archaeologi-cal Approaches to Market Exchange in Pre-Capitalist Societies. Boulder: University Press of Colorado.

Fleisher, J. 2004. Behind the Sultan of Kilwa's 'Rebellious Conduct': Local Perspectives on an International East African Town, in A. Reid and P. Lane (eds.) African Historical Archae-ologies, pp. 91-124. New York: Kluwer Academic/Plenum Publishers.

Fleisher J. and A. LaViolette 2007. The Changing Power of Swahili Houses, Fourteenth to Nineteenth Centuries A.D., in R.A. Beck (ed.) The Durable House: House Society Models in Archaeology, pp.175-197. Occasional Paper 35. Carbondale: Center for Archaeological In-vestigations, Southern Illinois University.

Fleisher, J. and S. Wynne-Jones (under review) Kilwa-type coins from Songo Mnara, Tanzania: New Finds and Chronological Implications, Numismatic Chronicle.

Garlake, P. 1966. The Early Islamic Architecture of the East African Coast. Nairobi and Ox-ford: British Institute in Eastern Africa.

Horton, M.C. 1994. Swahili Architecture, Space and Social Structure, in M. Parker-Pearson and C. Richards (eds.) Architecture and Order, pp.147-169. London: Routledge.

Kirkman, J. 1964. Men and Monuments on the East African Coast. London: Lutterworth Press. LaViolette, A. and J. Fleisher 2005. The Archaeology of Sub-Saharan Urbanism: Cities and

Their Countrysides, in A. Stahl (ed.) African Archaeology: A Critical Introduction, pp.327-352. Oxford: Blackwell Publishing.

Mathew, G. 1959. Songo Mnara, Tanganyika Notes and Records 53: 154-60. Museum of London Archaeology Service 1994. Archaeological Site Manual (Third Edition).

London: Museum of London. Pradines, S. 2005. Archéologie et préservation du patrimoine: le projet franco-tanzanien de

Kilwa, 2002-2005, Nyame Akuma 63: 20-6. Pradines, S. and P. Blanchard 2005. Kilwa al-Mulûk. Premier bilan des travaux de conserva-

tion-restauration et des fouilles archéologiques dans la baie de Kilwa, Tanzanie, Annales Islamologiques 39: 25-80.

Sutton, J.E.G. 1998. Kilwa: A History of the Ancient Swahili Town with a Guide to the Monu-ments of Kilwa Kisiwani and Adjacent Islands. Azania 33: 113-69.

Wilson, T.H. 1979. Swahili Funerary Architecture of the North Kenya Coast, in J. de V. Allen and T.H. Wilson (eds.) Swahili Houses and Tombs of the Coast of Kenya, pp.33-46. London: Art and Archaeology Research Papers.

Wynne-Jones, S. 2007. Creating Urban Communities at Kilwa Kisiwani, Tanzania, AD800–1300. Antiquity 81: 368–380.

63

Appendix 1 Context Recording Sheet: Side 1

64

Context Recording Sheet: Side 2

65

Appendix 2 Intrasite Geoarchaeology at Songo Mnara, Tanzania Federica Sulas

Intra-site geoarchaeological investigations were undertaken as part of the Rice Univer-sity/University of Bristol archaeological project at Songo Mnara, Tanzania. The geoarchae-ological study aims to provide essential data for addressing the use of space inside buildings and in open areas through the study of soils/sediments (chemical analysis, soil micromorphol-ogy and phytolith analysis). The techniques have successfully been applied in European archae-ology (see Goldberg and Macphail 2006 and references therein) and studies on African contexts are now increasing (e.g. Shahack-Gross 2004; Mallol et al. 2007; French et al. 2009; Sulas et al. 2009).

The main aims during the two weeks of fieldwork undertaken in June 2009 were to as-sess the overall conditions of the site and to record and sample open areas and excavation units for subsequent laboratory analyses. Four main deposit-types were targeted:

1) Open areas in between buildings (Northern Open Area, Southern Open Area, Town

Wall); 2) Domestic structures (House 44, House 23) 3) Graveyard 4) Activity areas and high-resistivity spots

A total of 138 soil samples were collected across the site and the majority consist of loose soil/sediment samples and 9 oriented blocks for soil micromorphology.

Open Areas - A grid of 5 x 10m square was superimposed over two open areas located respectively to the N (Northern Open Area) and to the S (Southern Open Area) of House 44. The soil cover was recorded at 5m intervals over four E-W transects and samples were collected from approximately 10cm below the ground surface. In addition, three surface samples were collected along the main town wall. A total of 79 samples was collected from the open areas. The field data will be used to map the distribution of different soil cover across the site and laboratory analyses are expected to inform on the nature of the different soil types encoun-tered and on the type of activities that were taking place in the open areas (e.g. animal/human trampling, burning, etc). In addition, phytolith analysis will be undertaken on selected samples to acquire relevant data on the vegetation cover and the plants present at the site.

Domestic structures - Systematic soil sampling was performed in parallel with the exca-vation of House 44 and House 23. A total of 55 samples were collected from room fills, floors, thresholds and hearths. Micromorphological analysis will concentrate on the study of floor de-posits, including plasters, and hearths. Combined with chemical analyses, the micromorphologi-cal study is aimed at elucidating the use of domestic space. Phytolith analyses will be per-formed to investigate the plant types present indoors and potentially associated with food proc-essing.

Graveyard - A grid of 2 x 2m square was built on the graveyard and 8 soil samples were collected at 2m intervals and approximately 10cm depth. Soil analyses will be performed to de-tect any chemical signature related to the burials.

66

Activity areas and high-resistivity spots - The non-residential structures investigated in-clude a well (SM006) and other deposits possibly associated with the use of the site. A total of 19 soil samples were collected. Resistivity survey detected a series of anomalies across the site and a few of these were investigated by test excavations; soil samples were collected to acquire relevant information on the chemical characteristics of the deposits. A portion of the well (SM006) was excavated and soil samples were taken from key cultural deposits. The study of micromorphology samples is aimed at defining the microstratigraphy of these deposits and identifying evidence associated with the use of the well. References French, C., F. Sulas, M. Madella 2009b. New geoarchaeological investigations of the valley systems in the Aksum area of northern Ethiopia, Catena 78(3): 218-233. Goldberg, P., R.I. Macphail 2006. Practical and theoretical geoarchaeology. Oxford: Black-well. Mallol, C., F.W. Marlowe, B.M. Wood, C.C. Porter 2007. Earth, wind and fires: ethnoarchae-ological signals of Hadza fires, Journal of Archaeological Science 34(12): 2035-2052. Shahack-Gross, R., F. Marshall, K. Reyan, S. Weiner 2004. Reconstruction of spatial organiza-tion in abandoned Maasai settlements: implications for site structure in the Pastoral Neolithic of East Africa, Journal of Archaeological Science 31: 1395-1411. Sulas, F., M. Madella, C. French 2009. State formation and water resource management in the Horn of Africa: the Aksumite Kingdom of the northern Ethiopian highlands, World Archae-ology 41(1): 1-15

67

Appendix 3 Songo Mnara Intertidal Survey Jack Stoetzel

The intertidal survey conducted on Songo Mnara is the first step in an analytical process

that aims to reconstruct the relationship between settled populations and their coastal landscape. The theoretical impetus for this research approaches a landscape as the tangible result of inter-action between cultural and ecological processes. As a result, a contemporary landscape con-tains a chronology of such cultural and environmental interaction. The socio-environmental chronology targeted in this research intends to identify Swahili influences on erosional proc-esses. Soil cores were extracted in order to provide geomorphological evidence of hypothesized erosional influence. A coring instrument specifically designed to extract soil from super-saturated environments was used to obtain samples.

The survey universe covered approximately 150m of coastline (oriented on the north-south access) directly west of extant stone architecture (UTM location of survey extended be-tween E0560300, N9000680 and E0560505, N9000815). Nine east-west transects dissected the survey universe. Each transect began at the outer edge of mangrove forests and extend seaward to the ultimate low water mark. A core was extracted every ten meters along each transect; in total 36 were collected.

Cores were analyzed immediately upon collection and curated for in-lab analyses. Field analysis was primarily interested in recording stratigraphy based primarilyon color change stan-dardized by a Munsell color chart. Other factors involved with the delineation of stratigraphy included the presence/absence of calcium carbonate organisms, roots, and quartz. Informal field analysis of collected cores included tentatively indicates that the seaward extent of mangrove forests has varied by as many as 20m. The grain-size, pollen type and distribution, and diatomic communities will all be analyzed in-lab in order to assay a variety of nuanced chronologic changes within the intertidal ecosystems sampled. Ecological changes will be correlated with socio-cultural changes identified by terrestrial survey conducted by J. Fleisher and S. Wynne-Jones.

The results of this research will form part of the doctoral thesis of J. Stoetzel, University of Virginia, and will be presented in future site publications.

68

Appendix 4 Conservation work on the supporting wall of the Necropolis Just after our arrival on Songo Mnara (10th June 2009) heavy rains caused the collapse of a retaining wall at one of the mosque structures. This mosque has a particularly interesting mihrab design and an extensive and unparalleled associated cemetery. This structure has been called “La Pagode” by Pradines (2005) and the Nabhani Mosque by Burton (1872) and Kirk-man (1964). The cemetery contains dozens of burials with head and foot stones, and local oral traditions mark this as a resting place for prominent religious persons. This structure is not well understood, but was planned by Garlake in the 1960s, who suggested it was a rare coastal mosque type on architectural grounds. We believe that it also speaks to local Islamic traditions and burial practices in a unique fashion.

It is positioned to the south west of the site, atop an elevated coral bluff. The cemetery is enclosed by a retaining wall (see Figure 2 in main report), which support the earth and the graves it contains, preventing them from sliding into the mangrove. It was this wall that col-lapsed during the rains, leaving the earthen bank unsupported, with human remains beginning to show in the exposed profile (Figure 31). It was immediately clear that the cemetery could not be left in this state. It was impossi-ble to walk in the cemetery for fear of dislodging the entire complex, and there was a risk that the earth could begin to slide at any time. A heavy rain would have caused the loss of all buri-

Figure 31: Collapsed wall and exposed cemetery deposits at Necropolis

69

als, which would have rapidly been washed into the mangrove, undermining the mosque struc-ture and destroying the burial ground. With the help of Bw. Chidoli, Idiphonse and Mr Kamamba of the Department of Antiq-uities, an application was submitted to the UNESCO Internal Assistance Emergency Fund. We received advice on submission from Dr Fumiko Ohinata of UNESCO, Nairobi, who was im-mensely helpful, and from Ms Valerie Goulet of the UNESCO office in Dar es Salaam. As it was unclear how long the process of receiving this money would take, and due to the im-mediate threat to the wall and the structures it supported, we also spoke with the local commu-nity about the possibility of constructing a temporary solution. The Ruins Committee of the nearby village was extremely accommodating, and over 40 local residents volunteered to assist without payment. They spent two days using the collapsed coral blocks to rebuild the support-ing wall, with Valerie Goulet overseeing the work. We supplied food and water for the workers, as a small recompense. Due to the generosity of the local residents, a strong and potentially long-lasting wall was reconstructed, which would hold the deposits in place until such time as the UNESCO funds are made available. We heard before leaving Songo Mnara that the funds had been granted, and we look forward to seeing the more permanent structure that will replace our tem-porary solution. The villagers who gave so generously of their time hope that they will be given work when the new wall is built; Bw. Chidoli has a record of their names. For now, the quick response and generosity of all parties, from local residents to UNESCO’s committee, have saved this unique monument.

70

Appendix 5 Archiving and storage: Songo Mnara 2009 Ceramics

Most locally-produced ceramics from the site were taken to Rice University, Houston, for analysis. The cataloguing is ongoing; a full list will be made available to the Antiquities Unit, Dar es Salaam, as soon as it is available. All ceramics will be returned to Tanzania when analysis is complete, and storage arrangements will be agreed with the Antiquities Officers.

The imported ceramics from Songo Mnara were bagged and labelled by context, and stored in wooden boxes in the site house on the island. Analysis was completed in the field, and the numbers are presented fully in the trench descriptions above. The imported ceramics will be moved to more appropriate storage along with the locally-produced ceramics, when returned. Beads

As reported, the beads from Songo Mnara are currently being analysed in the USA by Ms Marilee Wood. A full report on the beads will be made available to the Antiquities Unit when it is finished. The beads themselves will be returned at this time. Coins

The coins found during 2009 excavations are catalogued and will be published in a forthcoming edition of the Numismatic Chronicle (as cited above). The coins themselves are being stored in acid-free coin slides, as per current museum practice. They will be returned to Dar es Salaam as soon as the project is complete. Plant macrofossils

The results of flotation on soil samples from Songo Mnara are under analysis by Dr Sarah Walshaw at Simon Fraser University, Vancouver. A copy of the report will be made available once it is complete. The samples will remain at the laboratory facilities in Vancouver. Phytoliths and micromorphology slides from the 2009 excavations are being analysed by Dr Federica Sulas at the Pitt-Rivers Laboratory, University of Cambridge. A copy of the re-port will be made available once it is complete. Again, the samples will remain at the laboratory facilities once analysis is complete. Bones The bones from Songo Mnara are being studied by Erendira Quintana of the University of Bristol. The results of this analysis will form part of her doctoral thesis. A copy of this, as well as the report on the Songo Mnara assemblage, will be made available to the Antiquities Unit in Tanzania once complete. The bones will be returned for storage along with other finds from the site. Reports

A copy of this report will be deposited with the Tanzanian Commision for Science and Technology (COSTECH), the British Institute in Eastern Africa, Nairobi, and made available via the Songo Mnara project website (http://www.songomnara.rice.edu/). All subsequent reports and artefact information will be made available in the same places.

71

Plates

View of palace northern façade

72

Decorative niche, palace courtyard

73

Reconstructed arches in palace courtyard, showing remains of second storey

74

Mihrab of the Great Mosque

75

House 44: before (above) and after (below) excavation

76

Excavations in progress: SM011 (above); SM015 (below)

77

Monumental staircase, House 21

The Necropolis and attached graveyard

78

View across southern open area, with central enclosed graveyard, facing south

View across northern open area, from House 44, facing northeast

79

Courtyard of the Palace

View of the Great Mosque from north, House 15 in background