Primitive Islam and Architecture in East Africa

Author(s): Mark Horton

Source: Muqarnas, Vol. 8, K. A. C. Creswell and His Legacy (1991), pp. 103-116
Published by: Brill
Stable URL: http://www.jstor.org/stable/1523158
Accessed: 15-01-2016 10:35 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/1523158?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Brill is collaborating with JSTOR to digitize, preserve and extend access to Muqarnas.

http://www.jstor.org

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
 MARK HORTON

PRIMITIVE ISLAM AND ARCHITECTURE IN EAST AFRICA

K. A. C. CreswellnevervisitedEast Africa;indeedifhe THE ORIGINS OF ISLAM AND

had he wouldhave seenverylittlethatwouldhave illu- THE SHANGA EVIDENCE
minatedhis understandingof early Islamic architec-
ture.In the1930'sarchaeologicalinvestigations had not The earlyIslamic historyof the East Africancoast is
begun, and most of the monuments were covered by poorlydocumented.Traditional chroniclesand oral
thickand impenetrablebush.But withinthelast thirty historyprovidea narrativeforthearrivalofIslam, but
yearsmuchhas beendone to cleartheprincipalmonu- thesewereonlywrittendownin thenineteenth century
mentsand makeplans and elevations.'Archaeological and are generallyheld to be unreliable.4 The onlyeye-
excavationsprovidea chronology, extendingback into witnessdescriptionsby Arabicwritersare thoseofal-
theearlyIslamic periodand detailedevidencefortim- Mascudi and Ibn Battuta. Al-Mas'udi5 describesa
ber as well as stonestructures.The resultis an impor- coast largelypagan in 916,withone Muslimroyalfam-
tantgroupofbuildingsthatilluminateourunderstand- ily residingin Kanbalu. Ibn Battuta6 describes,from
ingoftheuse ofIslamicarchitecture beyondthecenters his visitin 1331,a devoutand whollyIslamic society,
ofscholarshipand craftsmanship ofthecentralIslamic extendingfromMogadishu to Kilwa. Other descrip-
lands. tionscan be showntobe based on hearsayevidenceand
Creswell'scommentsthat"Arabia, at theriseofIs- oftenwithlittlebasis in fact.Idrisi,forexample,does
lam does notappear tohave possessedanything worthy notmentionany Muslimcommunities, butdatedAra-
ofthenameofarchitecture"'is a viewthatcouldbetter bic inscriptions survivefromsomefifty yearsbeforehe
apply to East Africa,south of Ethiopia. Indigenous was writing.7
stonestructures thatareknownfromtheregion,suchas The mostrecentarchaeologicalworkundertakenby
thosefromtheZimbabweplateauand in thehighveldof the BritishInstitutein EasternAfricahas gone some
southernAfrica,belonggenerallyto thelaterIronAge, way to establishthechronology ofIslam in theregion.
that is, afterA.D. 1000; archaeologicalinvestigations An eight-year program ofexcavationsat Shanga,in the
have as yetfailedtoidentify an Africantraditionin sub- Lamu archipelago(fig.1), has beenfollowedby exten-
stantialtimberarchitecture. sive surveysand excavationsin Zanzibar and Pemba
A commonviewofthegenesisofIslamicarchitecture and a reassessment ofpreviousworkat Manda, Mafia,
is that,as it spreadintocentersofcivilizationand city and Kilwa. The broadconclusionfromthisworkis that
life,so it adoptedand modifiedtheexistingarchitectu- old ideas ofArab "colonization"oftheAfricancoastfor
ral formsoftheseregions.One exceptiontothisthesisis the purposeof tradehave been largelyreplacedby a
thecoastofEast Africa,wheretherewas no sucharchi- broadprocessofconversionofindigenouscoastalcom-
tecturaltradition.Thereforeone mightexpectto find munitiesto Islam throughcontactwiththe monsoon
traces of "primitiveIslam" as definedby Creswell tradingsystemofthewesternIndian Ocean.8
transmitted directlyfromthe Arabian peninsulaand Excavationsat thesiteofShanga in theLamu archi-
unaffected bycontactwithcomplexsocieties.Whilethe pelago between1980 and 1988 revealeda sequence in
phase of primitiveIslam may have lasted onlya very thecenterofthesitewithwhichwe can begintoidentify
shorttimeinArabia,itwas able tocontinuein themore these processesof Islamization.Shanga was a minor
remoteareas oftheSwahilicoast.3Thus thepossibility Swahilitradingcenter,abandonedin theearlyfifteenth
of makingsignificantarchaeologicaldiscoveriesis so century,but whichenjoyedan earlyimportance,with
muchthegreaterthere.Duringexcavationsat Shanga stratified occupationlevelsup to fivemetersin depth,
in theLamu Archipelago,preciselysuch remainswere extendingback toca. 750,datedon thebasis ofimport-
discovered,whichare ofinterestto our understanding ed potteryand radiocarbondating.
ofthedevelopmentofIslamic architecture. The primaryoccupation lies directlyupon white

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
104 MARK HORTON

N1

ate L t suniga

rwa0Manda 0 0

K0ipungani kl
O LAMU ARCHIPELAGO

S 9th -:10th-century sites

O O Later sites

Ungwana 5 0 10 20
o0

Fig.1. LocationmapofearlysitesintheLamuarchipelago.

beach sand. The potteryassemblageis 96 percentlocal tranceson its westand east sides. The enclosurewas
potteryoftheTana tradition, butthereare a fewsherds preciselycardinalnorth-southinorientation,and in the
of importedSasanian Islamic and unglazed storage geometriccenterwas thewell. Later reconstruction of
jars, as well as Chinese stonewares,suggestingthat thiswell damaged its earlyform,but a concretedsur-
fromthe outsetShanga was in contactwiththe mon- faceofsand leadingdownto thewell,formedbythere-
soon tradingsystems.The house structures are oftim- peated sloppingof water onto the surface,indicated
berpost-holeconstruction and circularin plan. thatitwas eitheran open hollowor shuttered withtim-
The layoutofthisearlysettlement is verycharacter- ber in a temporary manner.Five metersto theeast of
istic.Excavationsrevealeda seriesofshallowgulliescut the well was a large tree,whose burnt-outtreestump
intothesand,whichhad beenredugon manyoccasions was excavated,and muchironslagwas foundaroundit.
and whichapparentlycontaineda fence.This fencecan Elsewherewithintheenclosuretherewas littletraceof
be shownto have encloseda rectangulararea, approxi- habitation,only shortgulliesand slots fortemporary
mately 100 metersby 80 meters,withexcavateden- structures,oftenassociatedwithcraftactivitysuch as

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
 PRIMITIVE ISLAM AND ARCHITECTURE IN EAST AFRICA 105

shell bead manufactureand the preparationof shell Africankayais a local manifestation of a widersettle-
cones. Immediatelyoutside the enclosurewere found ment plan found within both Bantu-speakingand
arcs ofpost holes suggestingroundhuts,bread ovens, Cushitic-speaking communities oftheregion,and may
and a muchhigherproportion ofanimaland fishbones. wellhave extendedup totheRed Sea coastofEthiopia.
The enclosureappearstorepresent demarcatedareas While no directlinkcan be postulated,earlyMuslims
ofspecializedactivity, withdefinedentranceswhosepo- that reached the Africancoast may well have found
sitionis retainedthroughout thenextsixhundredyears formswithwhichtheywerealreadyfamiliarin Arabia.
ofthesettlement. The importanceofthewellin thecen-
ter,thelargetreenearbywithitsironslag, but no fur- ARRIVAL OF ISLAM
naces,suggestan area ofritual.The cardinalalignment
of theenclosuremustalso be significant, and astrono- At Shanga,Islamicpracticecan be identified almostdi-
micalalignmentsare notunknownamongvariousearly rectlyabove thehorizonofwhitesand withitstrodden
pastoralgroupsin theregion.9 debrisof craftactivity.Direct evidencecomes froma
The ethnographicand ethnohistorical parallels for numberofburialscut intothesand and laid out in the
theseritualenclosuresare wellrecordedin East Africa. conventionalway thatcontinuesto thepresentday -
The bestknownare thekayaoftheMijikendawhooccu- that is, lyingeast-weston the side, with thehead in-
py thecoastal stripinlandfromtheSwahili.'oThe typ- clinednorthwardsto Mecca. They are foundin levels
ical kayais a doublerectangularenclosurewithmassive dated fromaround800.
gatessetwithintheforest.The gatesare associatedwith More controversialare the timberstructuresthat
specificclans,and buriedbelow thegate is a fingopot, were built directlyover the burnt-outtreestump,in
said to containthemagiccarriedfromShungwaya,the thecenterof thesite.They each lie directlybelow the
traditionalhomelandoftheMijikenda.The centralrit- prayerhall ofthelaterFridaymosqueand thus,on the
ual area has no domesticoccupation,onlyburialsand groundsof continuity, would appear to be mosques
themoro"whereeldersmeet.Each mijior tribehas its themselves. Each is laid out,usingthedhiraof518 mm.,
ownkayaand is modeledon theoriginalarchetypalkaya and theyhave a consistent qibla, albeitsome 500 away
of Shungwaya,the homeland,accordingto tradition. fromthe truedirectionof Mecca. What is interesting
Indeed theMijikendaand theSwahilispeakcloselyre- about thisorientationis thatit does notfollowthatof
latedlanguages,and theMijikendaappear torepresent thecentralenclosure,again suggestingthatthetimber
thenon-Islamicizedcomponentofa commonsociety.'2 structures, or mosques,are not primaryto it. Support
The Swahili also builtsimilarenclosures,although thatthedeviantqibla linewas perceivedas thecorrect
withinan Islamic context.The plan ofTakwa, builtin linecomesfromtheburialswhichlineup withthetim-
theeighteenth century,clearlyshowshow themosque bermosquesand notthecentralenclosure.
was setwithina largerenclosurewithentrancegates.In A total of seven timbermosques and one stone
theComores,wheremanytraditionalSwahilipractices mosque (figs.3-5) werefoundin theexcavationsbelow
have continued,a centralcommunalarea knownas the the prayerhall of theFridaymosque whichitselfwas
fumboni is foundin a numberofvillages,again withclan constructed around 1000.The earlieststructure liesdi-
and moietyassociationswithgateways.Thereare even rectlyabove theburnt-out treestumpand was builton-
the remainsofafumboni in thecenterof the Shangani to thehorizonoftroddenwhitesand and ironslag.The
quarter in Zanzibar Old Town. Drawings of nine- followingis a summaryofthesequence:
teenth-century Mogadishu show an open enclosureat
thepointwherethetwomoietiesmeetbythemosqueof MosqueA. This structure,rectangularin plan, was
Fakhral-Din (fig.2). Fingo potshave also been found markedby tracesofa thinwall trenchsupportedby an
underdoorwaysin a numberofSwahilisettlements."3 externalrowofpostholes.The entrancewas somewhat
Of coursethisEast Africanpracticeis reminiscent of eroded,withtracesofan emplacement fora timberstep.
descriptions ofpre-IslamicMecca, withitswalledsanc- This mosquehad twokinyokae made
floors, ofa greenor-
tuaryofKa'ba, containingwithinit thesacredwellof ganicmudcollectedfromthemangroves, and twophas-
theZemzemand settlement without.The suggestions of es of a post hole wall, suggestinga rebuilding.It was
Askumite,and thusAfrican,connectionsin thearchi- aligned 3100 with internaldimensionsof 1.64m. by
tectureoftheKa'ba thatweremade byCreswellmight 2.59m. (almost precisely9 cubits by 5 cubits using
extendto the formof the sanctuaryitself.'4The east 518mm.).

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
106 MARK HORTON

Fig. 2. The centralenclosurein Mogadishu,locatedbetweenthetwomoietiesofShanganiand Hamar Weyne.In thetopcentercan be seenone

ofthemoietygatesintotheenclosure.To theleftis the13th-centurymosqueofFakhral-Din. (FromG. Revoil,Voyage
chezlesBenadirs
..., 1882.)

MosqueB. A clearlydefinedoblongarea ofwellrounded outside.The plan is also a slightdeparture;it is a two-

beach pebbles directlyoverlaymosque A. There were celledstructure, withoffsetdoorwaysat thesouthend,a
apparentlyno associated post holes and thus no evi- largernorthern room,and a smallersouthernroom.In
dencefora superstructure. werevery
Its measurements thecenterofthewestwall is a pad ofkinyokae, possibly
approximately 6.4 m. by 4.0 m. thebase ofa stair.The doorjambs to theentrancewere
markedby square holes,suggestinga substantialdoor
Mosque C. This was a well-definedrectangular struc- frame. UnlikeE, thismosquedid nothavea centralpost
ture, with a wall trench and post-holerows along its hole. Its internaldimensionswere 7.24m. by 3.61m.
a
side, kinyokae floor,and a central
posthole.This struc- (14 cubits by 8 cubits,using518mm.).
turewas a littlelargerthanmosquesA and B and dis-
placed a littleto the east, witha verysimilarqibla of MosqueE. This was a rectangularbuildingthatwas a
3080. The internaldimensionsof 5.10m. by 2.85 m. completerebuildingof mosque D. The wall trenches
givesa size of 10 by 5.5 cubitsusing518mm. werealmostexactlycoincident,withthoseofmosqueD
cut away by thetrenchformosque E. There weretwo
MosqueD. This structurefollowedalmostexactlythe phases to thisbuildingrepresented by twofloorsusing
same linesas mosqueC, witha qibla of3090, butagain kinyokae, two cutsforthecentralpost,and a foreshort-
displacedto theeast and slightlylarger.It had a well- eningbythedemolitionofthesouthernroomin thesec-
definedkinyokae floorand a wall trench,withthe post ond phase. The wallsweremade up ofthinsticks,up to
holes actuallyset intothewall trench,ratherthanthe 40 mm.in diameterand thencoveredin kinyokae to pro-

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
 PRIMITIVE ISLAM AND ARCHITECTURE IN EAST AFRICA 107

MOSQUE A (phase 1) MOSQUE A (phase 2) MOSQUE D MOSQUE E (phase 1)

-- -
O -..-
-- ... -
io0 t" o:7 '9 i

q ) I
I(
stumpii
. . .I
i!_
" II 'I -
'00I 0 * I

. I i
IiO
1iI
I
LO
o 0 C) __I -- -
0

CL9
MOSQUE B MOSQUE C MOSQUE E (phase 2) MOSQUE F

-- - - - -- - - - -L I I - - -- - -- - - - . . . . ..

9,
I

bpebble
surfacete

0 5 t

4 0
~0

mosquesA, B, andC from Fig.4. Planofthefoundations

oftimber
Fig.3. Planofthefoundations oftimber
mosquesD, E, andF.
Shanga.The postholeswhichhavebeenshownsolidprovidedevi-
dence fortheexact positionofthe timberpost,or "ghost."

duce an even surface.The central post hole was theonly were 7.25 m. by 4.14 m. (14 cubits by 8 cubits using
substantial post associated with this building, suggest- 518 mm.).
ing that it had a flat,ratherthan a gabled, roof.The di-
mensions of thisstructurewere 6.21 m. by 3.61 m. or ex- Mosque G. A wooden building was definedby a number
actly 12 cubits by 8 cubits, using 518 mm. of large post pits, 0.80 m. in diameter and up to 1.0 m.
deep. As the posts had been robbed out, theirexact po-
Mosque F. This rectangular building was a new struc- sition could not be discovered. The floor level of this
ture,on a new qibla line of 3230. It too was divided into building had also been destroyed. However, the basic
a larger northernand a smaller southernroom, entered plan was of a rectangular building that was undivided
on axis from the south, but with an offsetinner door- internally.Two smaller posts suggested a porch or en-
way. It was constructed with a wall trench that con- trance on the east side. The qibla was 329', and the ap-
tained posts, as well as an external row of supporting proximate external dimensions were 9.7 m. by 5.4 m.,
posts. A large post in the centerof the northwall might that is 18 by 10 cubits, using 540 mm. and not the
be fora mihrab, as there is no comparable post on the 518 mm. of the previous structures.
southernwall required forstructuralsupport. Only one
plaster floorwas found,but the patternofre-cutsofpost MosqueH. The firststone mosque was built directlyover
holes suggests that there were at least two phases of the post holes of mosque G. It was constructedofneatly
building and replacement. The internal dimensions shaped poritescoral (that is, undersea coral quarried by

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
108 MARK HORTON

MOSQUE G MOSQUE H MosqueJ.Mosque J is theFridayMosque; it stillsur-

- ---- i
-------- vives(fig.6). It was builtin stone,whichwas stillin use
----_- whenthetownwas abandonedin ca. 1425.The mosque
had threeplasterfloorlevelsrestingupon a platform of
whitesand, 1.5m. deep. The mosque was enteredby
stepsin thecenterof thewestside. The east and west
wallshad an arcadeofflat-headed doorwayswithsmall
rectangular windowsabove. On theeast side,therewas
a verandaofthesame heightas theinternalfloor.The
mihrabthatwas presentin thefinalphase was inserted
throughthe northwall. The earlierarrangement was
destroyed,butitwas apparentlya muchsmallerniche,
possiblysetwithinthethickness ofthenorthwall. The
MOSQUE J externalwalls had side pilasters,whichwerebuiltup-
wardsto supporta thatchedroof.Internallytheprayer
hall was divided into a large northernand a small
o o
southernroom,but thisdividingwall was demolished
in thesecondphase and replacedbycolumns.Manyre-
used stoneswereincludedin its walls. The qibla was
342 .

0 0 DATING

In theabsenceofinscriptional evidence,datingis either

byradiocarbonor byassociationwithlocal orimported
pottery.The potterycomprisesthree basic assem-
blages:
ThefillofMosqueJ. The floorwas builtovera platform of
whitesand,whichcontaineda largenumberofsherdsof
Fig.5. Planofthefoundations mosqueG andofthetwo
oftimber
all ofwhichwererecovered,throughtotalsiev-
stonemosquesH (all phases)andmosqueJ(primary
phase). pottery,
ing.SherdsincludedSasanian Islamic,whiteglaze,and
lusterpottery,the laterformsof theearliestlocal pot-
diversand shapedwhilestillwet),bondedbymudwith tery,but not a singlesherdof sgraffiato pottery.This
a whiteplasterface. In plan, thebuildingwas divided groupcan be closelydated to around1000,providinga
intothreeparts- a square prayerhall,a narrowsouth construction date formosqueJ.
room,and a southerncourtyardbeyond.A westernan- Constructionlevelsassociated
withMosqueH and G. These
nexwas added duringthelifeofthebuilding.In thecen- containfewersherds,similarto thoseat Mosque Jbut
terofthenorthwall weretheremainsofa salientthat withno lusterpotteryand withwhiteglaze,and earlier
presumablycontainedthe mihrab.On the west side formsoflocal pottery, datingto 850-900.
weresteps,probablythebase ofa staircaseminaret.No LevelsassociatedwithMosquesA-F. Therewas verylittle
floorssurvived,but a layerof whitesand sealing the variationin thecontentoftheselayers,and pottery was
constructionlevel suggestedthat therewas a raised generallyrare.The veryearliesttypesoflocal pottery
floor,probably capped in plaster. The span of the wereassociatedwithSasanian Islamic,eggshellwares,
prayerhallswas toowide and musthave beensupport- unglazedstoragejars, and singlesherdsofChangsha
ed by a centralcolumn,but thistoo was robbedout. stoneware,Dusun and black stoneware.White glaze
Constructionalevidenceshowedthattherewerethree was entirelyabsent.Date rangeis 750-850.
phases ofbuilding.Internaldimensionsof5.10m. and A sequence ofcharcoalsampleswas recoveredfrom
wall thicknessof0.54 m. show thata cubitof540 mm. the stratifiedmosques providingdirect radiocarbon
was used, witha qibla of3290. dates; thedatingwas undertakenby Dr. R. Switsurof
the Godwin Laboratory,Cambridge University.Six

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
 PRIMITIVE ISLAM AND ARCHITECTURE IN EAST AFRICA 109

Fig. 6. Excavationswithintheprayerhall ofmosqueJ,showingtheplan ofmosqueH, and thepostholesofmosquesE (phase 2) F and G. The

threeprincipalqibla linesare visible.

Table1. RadiocarbonDates fromShanga Mosques

Sample ArchaeologicalContext Estimated Radiocarbon Calibrateddate range,cal AD

Reference Date Age BP 68% 95%

SA1982 Fill oftreestump,clearedto build 750 1235 ? 35 710 to 745 685 to 880
mosque A 760 820
835 855
SA1647 Post hole in wall ofmosque C 790 1180 ? 40 790 to 890 720 to 735
765 965
SA1512 Centralpost hole to mosque E, 825 1170 ? 45 785 to 895 720 to 735
associatedwithfirstofthreephases 925 935 765 965
SA0659 Post hole ofmosque F 850 1100 ? 50 890 to 980 815 to 840
850 1020
SA0616 Spread associatedwithearlyuse of 900 1060 ? 45 900 to 915 895 to 1025
mosque H 950 1020
SA0590 Spread below floorofmosqueJ,above 1000 985 ? 35 998 to 1042 980 to 1062
robbingofmosque H 1095 1115 1070 1125
1105 1150 1135 1160

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
110 MARKHORTON

sampleswerechosento representthespan ofthenine

mosquesexcavated.Samples werefromsecurelylocat-
ed featuressuch as post holesor contextssealed below OO 0
floorlevels.Table 1 showstheresultsobtained,together
i i ,
withourestimatesofthedate arrivedat throughceram- oo

ic analysisand generalevidenceofsequence.The sixra- o

diocarbonages obtainedfromthesesamplescorrespond 4 5

correctlyto theirstratigraphicorder.The twoearlyra-

diocarbon dates provide an indicationthat the first
1%
mosque is unlikelyto date fromthe seventhcentury,
even at 95 percentprobabilitylevels.The earliestpos-
sible date for the tree stump would be 685 and for
-LI,
mosque C 720; at 68 percentthesedates are 710 and
785. A date formosqueA would,on calibrateddates,lie
withintherange 750-850. The archaeologicaland ce-
*
ramic evidencewould point to the earlierend of this i~
range.A possibledatingschemewouldbe: mosqueA to
thesecondhalfoftheeighthcentury,withdecadal re-
buildingsoftheflimsy structuresA, B, C, tosay ca. 800
and generationalrebuildingofthemoresubstantialD, Fig.7. Planofthetimber hallat Shangaexcavatedin 1988,withthe
E, F and G to ca. 900; the stonemosque H in use ca. foundations oftheporites building shownabove.Thepreciseposition
ofthetimber
posts(shown as solid)couldbedefined
within
thelarger
900-1000,replacedby mosqueJ ca. 1000.
postpits.
SECULAR BUILDINGS
westbya shortflight ofsteps.On one side ofthesesteps
Contemporary withand verynearto theseearlytimber was a basin presumablyforwashingfeetbeforeenter-
mosques,on the westside of thewell,anothertimber ing.Insidetheplasterfloorswereraisedon a platform of
structure, whichcouldbe describedas a hall,was found whitesand, and as the buildingwas heavilyrobbed,
and excavated(fig.7). The postholeswereoftwotypes, therewas no directevidenceforthepositioning ofdoor-
an outerrowofsmallpostsand an innergridofmassive ways. There were threerows of rooms,the firstun-
postpitsup to 2 m. in depthand up to 1.6m. deep. The divided,possiblyan internalcourtyard, theseconddi-
ghostsofthepoststhemselvessuggestedtimbersup to videdintothree,and thethirdintotwo.At theeastend
40 cm. in diameter.In plan, the post holes formeda was another,muchshorterflight ofstepstoa rearspace.
grid,fiveby fourofequally spaced posts,withtheex- There was no evidenceofdomesticoccupationon the
ceptionof two centralposts,suggestingthe structure floors,or indeedin theexternalcourtyards.Some me-
had a symmetrical internalcourtyardand a porticoon tersto the west was a second, much smaller,porites
each side thatrannorth-south (fig.8). The alignmentof building,of two rooms,again withraisedfloorlevels.
thisstructure was thatofthecentralenclosureand not These mayhave been stores.
ofthemosquesto theeast. Its scale suggestsa ceremo-
nial function,and it was replaced,significantly, by a ARCHITECTURAL ORIGINS
stonebuildingon thesame alignment.
This new buildingwas constructedof poritescoral Findingparallelsforthesebuildingsis a taskbesetwith
bonded in mud with plaster facing.Its positioning, not least because such ephemeraltimber
difficulty,
slightlyto one side of the timberhall, shows thatthe structureshave neverbeforebeen excavatedeitherin
builderswere aware of the locationof the post pits. East Africaor in Arabia. However some suggestions
Otherwisethe two structures werealmostidenticalin can be made,althoughI realizetheyare based uponve-
size and shape, and thestonebuildingmustbe viewed ryslenderevidence.
as a replacementin stoneofthetimberhall below.Dat-
ingevidencesuggeststhatthistookplace around900. Thetimber The reconstruction
mosques. of the elevation
The poritesbuilding(fig.9) was enteredfromthe fromthepatternofpost holes suggeststhatthe build-

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
 PRIMITIVE ISLAM AND ARCHITECTURE IN EAST AFRICA 111

Fig. 8. The timberhall withtheprincipalpostpitsexcavatedout.

ings were flat-roofedratherthan gabled. It is note- weresimplywallsofwovenwattleattachedtoregularly

worthythattheside postholesare exactlythesamesize spaced timberuprights, whichwereforkedat thetopto
as the cornerposts. Mosques D and E, forexample, support a timberjoist that held the flat or wagon-
have no structuralsupportingtimberson thecorneror shaped roofof wattleand daub. By havinga gridof
on thecentralaxis, whichinitiallyled me to conclude posts,largeareas could be roofedin thisway,themaxi-
thattheywerenotroofedat all. The evidencefora cen- mumspan foreach wattlebeingabout 1.5 to 2.0 m. Re-
tralpost,however,indicatesthatsomeformofroofwas constructionsofthesetembehousesin theVillage Mu-
placed over thestructure,but it is verydifficult
to see seum in Dar es Salaam, providea post-holeplan very
how a gable roofwas supportedwithno majorpostsat reminiscent oftheShanga structures.
thenorthand southendstosupportthetimber.I think, However,such flat-roofed structures are clearlyun-
therefore,thattheevidencepointsto flat-roofed struc- suitable forthe East Africancoast, where rainfallis
turesofverysimpleconstruction. much heavier than in the Masai plains. Flat-roofed
There are parallelsfortheflat-roofed housesin East stonehouses are widespread,but thesehave guttering
Africa,knownas "tembes,"foundin theMasai plains and run-off channels,whichof coursecannotbe used
ofTanzania in thearea ofLake Manyara.'5Sometimes withwattle-and-daub roofs.But thereis someevidence
theywere dug into the ground,but more commonly thateven domesticmud houses of thecoast wereflat-

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
112 MARKHORTON

Fig. 9. The poritesbuildingat Shanga duringexcavation.The floorswereraisedon a platform

ofwhitesand thatwas oftenhigherthan the
robbedwall foundations.The steps,witha smallbasin at thefoot,are visiblebottomright.The wellis locatedin the top leftcorner.

roofedduringthe eighthand ninthcenturies.Neville timetheclappingofhandswillbe takenas partofthere-

Chittickpointedout thatthepatternofburningfound ligiousceremony'.""Now we apparentlyhave archae-
on daub fragments fromManda werein keepingwith ologicalevidenceforthedeliberatelayingofpebblesin
theirpositionon the roof,ratherthan on the walls.'6 earlymosques.
Similardaub has also been foundat Shanga. It seems The southernroomspresenta problemin interpreta-
most likelythat the flat-roofedbuildingtypewas in- tion. Some of the structureshave a southernroom;
troducedfroman aridarea,eithertheinterior ofAfrica, othersdo not,so it was clearlynotessential.The man-
or theArabianpeninsula.Clearlythelatteris mostlike- nerin whichitsdoorwaywas staggeredsuggeststhatit
ly,giventhe Islamic contextforthesebuildings. was therepartlytorestrict intotheprayingar-
visibility
A feature of particular interestin the Shanga ea. It has no washingfacilities,
and thewellwas located
mosques is the use of roundedpebbles on thefloorof at thenorthwest cornerofthemosque,notclose to the
mosque B, a practicefollowedin the firstmosque of entrance.Butitstillmaybe thatthissouthernroomwas
Amrin Fustatin 641. It was also adopted at Basra in used forablutions,introducedin thetimeofmosqueD.
665; about that mosque Creswell commentedthat The absenceofa mihrab,or indeedanyformofqibla
"whenpeopleprayed,theirhandsbecamecoveredwith marker,in thecenterofthenorthwallindicatesthatthe
dust, which theyused to removeby clapping. This mihrabwas notwidespreadevenbythemideighthcen-
caused Ziyad to say, 'I am afraidthatin thecourseof tury.In theShanga sequence,thefirstmihrabmaybe

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
 PRIMITIVE ISLAM AND ARCHITECTURE IN EAST AFRICA 113

withmosqueF, wherethereappearstobe a largeaxially centralcolumndoes notremain,butwas probablypre-

located post thatwas not answeredby anotherat the sentbecause theroofspan is toogreatto have beenun-
southend. This large post may have been partlyhol- supported.Anotherfeatureat Shangais thebase ofa set
lowedout to providea shallowrecessedniche. ofstepson theeastside,probablyfora staircaseminaret
(fig.10), again foundalso in theSirafand Samarraex-
Thestone Turningto thestonearchitecture,
mosques. the amples.'9
evidenceis muchclearer.The earlieststonemosque at AlthoughtheseparallelsbetweenShanga and theSi-
Shangafitscloselyintoa groupofsmallmosquesknown rafmosquesare quiteprecise,theyshouldnotnecessar-
fromSamarra and Siraf,'8whichare also dated to the ilyimplythedirectimportation ofSirafiarchitecture
to
tenthcentury.These structuresdo not contain the the East Africancoast. The earliertimbermosques at
courtyardsthatare characteristicofthecongregational Shanga have no courtyard,and theseparatesouthern
mosques of theperiod,but theyhave a tripartite divi- roomssuggestsomelocal continuity froma woodentoa
sion along theqibla axis oftheprayerhall, back room, coral architecture.The use ofporitescoral,a Red Sea
and courtyard,in exactlythe same way as at Shanga. technique,points to Yemeni connectionsand stone
They have square prayerhalls dividedby a transverse structureswithsimilarplan are knownfromthe Ye-
arcade. As the floorlevel does not surviveat Shanga men,20 although more recent in date. Very littleis
(and was probablyraised on a platformof sand), the knownabout thesmallmosquesoftheAbbasidand late

Fig. 10.The recessedminbarat Mbui Maji, Tanzania. It has beenclaimedthatthisarrangement

ofmihraband minbaris a survivalofearlyIs-
lamicpracticein East Africa,but it probablydates to the 18thor early19thcentury.

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
114 MARK HORTON

Umayyadperiod;theymay have been ofa fairlystan- againstthisinterpretation.Burialsstratified belowthis

dard plan throughout theIslamicworldas is suggested structuregivea clear indicationoftheperceivedqibla
by the early-to mid-eighth-century mosque at Qasr line at thisdate ofaround3100, whereasthisstructure
Muqatil in Iraq.2' faces 2670, virtuallydue west. This is too greata di-
The final mosque at Shanga, dated around 1000, vergenceto accept,evenin remoteEast Africa.
seemsto be a local developmentderivedfromitsprede- The traditionalderivationof courtyardmosques
cessors.The prayerhall is rectangular, in a proportion fromtheplan ofdomestichouses,in particularthatof
of 2:3, ratherthansquare; it is dividedby tworowsof Muhammad,is particularlysignificant.27 Reconstruc-
columns.Again thereis a southernback room. The tionofthehouseoftheProphetin Medina givesa rect-
floorwas raised 1.5m. above the externalground,so angular building,with an open courtyardhaving a
thatentrywas via stepsfroma washingcourtyardon southernporticoofthreebays,similarin somewaysto
thewestside. Both theeast and westwalls have an ar- thetimberhall at Shanga. The porticowas supported
cade of threeflat-topped doors.No traceoftheearlier on palm trunks, whichmaywellhave beenthematerial
mihrabarrangement survives,althoughtherecessmust used at Shanga, as thepostdiametersare too wide for
have been set intothewall above floorlevel.A partic- mangrovepoles.Such courtyard housesmayhave been
ularfeatureofthemosquewas thatithas pilasterstrips widespreadin the Arabian Peninsulaand copied by
on theoutsidewalls,whichendin shortpillarsthatsup- Muhammad forhis own house. A quasi-domesticor
porteda thatchedroof,whichwas presumablygabled. ceremonialfunction fortheShangabuildingis possible,
Some of thefeaturesofthismosque are reminiscent of perhapsassociatedwiththerulinggroupwho adopted
the earlier timberand stone structuresat Shanga; thisbuildingformfromArabia.
others,such as thepilasters,are newfeatures.
The stonemosque at Shanga is ofparticularimpor- Theporites The case fora ceremonialfunction
building.
tance as it providestheprototypefora groupofearly forthetimberhall is strengthened whenwe considerthe
mosquesin East Africa,whichare remarkably uniform. stonebuildingwhichreplacedit. It surviveson a gran-
This groupincludesMbui,22Kaole west,23 Kizimkazi,24 derscale and represents theadoptionofstonebuilding.
Kisimani Mafia I, II, and III,25 and Sanje ya Kati.26All The techniquethatuses poritescoral appears to have
these structuresdate to the eleventhor early twelfth been introducedfromelsewhere,and the Red Sea
century. Theyoftenhavesidepilasterswithprayerhalls would be themostlikelycandidate,giventhedistribu-
in a proportionof 3:4. On theeast and westwalls are tionofsimilarcoralsin thisarea and thecontinueduse
paireddoors,witha centrallyplaced southerndoor;the ofporitescoral in buildingto thisday.28
roofis supportedon four,or sometimestwo,circularco- As thebuildingonlysurvivesto floorlevel,itseleva-
lumns,whichoftenwerein timber.OnlyKisimaniMa- tion is difficultto reconstruct.The thicknessof the
fiaIII has a southernroom.Otherwisetheprayerhallis walls,especiallytheinternalwalls,and thequantityof
freestanding witha raised floorlevel,supportedon a rubblesuggestedthatit may not have been a single-
platform ofwhitesand. The wallsare constructed ofpo- storiedbuilding;it mayhave risento severalfloors-
ritescoral. theentrythrougha flight ofstepssuggestsa grandfront
elevation,whichwouldnothave fitin easilywitha low,
Thetimber hall.The postholesofthetimberhallsupport- squat building.I suggestthatthiswas a towerbuilding
ed majorstructuraltimbers,and are toowidelyspaced severalstoriesin height.Examples of such a building
tohavesupporteda tembe-style roof.The outerwall be- stylecan be cited on both sides of the Red Sea, from
yond these post holes does not survive,but may have wheretheuse ofporitescoral seemsalso to have been
been ofdaub construction. A suggestedreconstruction derived.29The towerpalaces of the kingsof Askum,
is ofa square building,withan internalcourtyard and a whichcontinuedto be builtat least untilthe seventh
portico on the west side. The arcade would have been century,providea particularly closeparallel.3o
In Asku-
entirelyoftimber,ofwhichno detailshave survived. mitearchitecture stresswas giventomonumental stair-
Similaritiesbetweenthis buildingand early court- wayentrances, whichfindsan echointheShangabuild-
yardmosquescan be suggested,albeiton a verymuch ing. At nearbyManda a similarporitesbuildingwas
smallerscale. Bothhave theporticoon one side witha foundwithan entrancestairoffoursteps."'The Asku-
centralcourtyard. Could thisstructure be a smallearly mite palaces are not that much largerthan the East
courtyardmosque? Evidence for its orientationis Africaexamples- thatat Ta~akha Mariam was only

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
 PRIMITIVE ISLAM AND ARCHITECTURE IN EAST AFRICA 115

Monumental
stone building
,onument--.
stonebuildings 1

L well ggver.av grea " -

--'

street

house Mosque

0 10 20 30

metres

Fig. 11. The complexofmonumentalbuildingsin thecenterofShanga,aroundtheyear900.

21 m. by25 m. comparedto 12.5m. by 10.5m. at Shan- theearlyUmayyadperiod.

ga and 16.2m. by 12.2m. at Manda. Creswelland otherswho have followedhimhave re-
The openingup oftradingconnectionsbetweenEast lied upondocumentary sourcesto reconstructtheearly
Africaand theRed Sea in thetenthcenturyprovidesa formsofIslamicarchitecture. Perhapsthetimehas now
contextforthe transmission of thisarchitectural style come forarchaeologyto be deployedso that we can
southwards.The poritesbuildingat Shanga is one ofa morecriticallyassess thesehistoricalsources.
wholegroupofsimilarstructures thatfilledthecentral
enclosure;theirwooden fenceswere replacedwithstone British inEastern
Institute Africa
ones. At thesame date thefinalwoodenmosqueG was London,England
also replacedin stone.The centerof Shanga, perhaps
coveringup to 8,000m2seems to have been filledby a
complex of monumentalbuildings with a modest NOTES
mosquein thecenter(fig.11). Againparallelsto Asku- 1. P. S. Garlake,TheEarlyIslamic
Architecture EastAfrican
mitepalace complexescan be suggested,such as sev- ofthe Coast,
BritishInstitutein Eastern Africa,Memoir no. 2 (Nairobi,
enth-century Dongur,32wherea multiplicity of build- 1966),providesthemostcomprehensive corpusofplansand ele-
ingsand courtyardsfilledan area of3,000m2. vations.
These examplesshowthatEast Africais a fruitful ar- 2. K. A. C. Creswell,A ShortAccount ed.
ofEarlyMuslimArchitecture,
ea in whichto understandtheoriginsofIslamic archi- JamesAllan (London: Scolar Press,1989),p. 3.
3. Ibn Battutafirstuses thetermSwahiliCoast, and travelwriters
tecture,bothbecause thesitesare availableforexcava- untilthenineteenth centuryrefertotheAfricanMusliminhabit-
tion, as the earliest centersof occupation are now antsas theSwahili.The SwahiliCoast extendsfromMogadishu
abandoned,and becausein coraland lateritesoils,stra- in thenorthto Mozambique,and includesoffshore islandssuch
as theComores,Zanzibar,Pemba,Mafia,and northern
tigraphyis well preserved,so we are able to document Mada-
thetransition fromwood to stoneand identify themost gascar.
4. Only thefragments oftheKilwa chroniclethatsurvivein a Por-
of
ephemeral buildings. While our dates do not yetex- tugueseprtciswritten downbyDe Barros(Da Asia,Dec. 1,book
tendback intotheseventhcentury, further excavations VII) predate the nineteenthcentury.On the unreliability of
maypush back theoriginsofsitessuch as Shanga into thesetraditions, theclaimedUmayyadexpeditions
particularly

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions
116 MARKHORTON

to the Swahili, see M. Tolmacheva, "They Came fromDa- gregational MosqueandOther oftheNinthtoTwelfth
Mosques Centuries,
mascus in Syria,"InternationalJournal ofAfrican HistoricalStudies The BritishInstituteof Persian Studies, Siraf III (London,
12.2 (1979): 259-69. n.d.).
5. Murujal-Dhahab,1: 112; 3: 31. 19. J. Schacht,"FurtherNoteson theStaircaseMinaret,"ArsOrien-
6. Les Voyagesd'IbnBatoutah 2: 179-96.G. S. P. Freeman-Grenville, talis4 (1961): 137-41,arguesthatthestaircaseminaretis a relic
The East AfricanCoast,SelectDocuments (London: Rex Collins, of earlyIslamic practiceon the East AfricanCoast. Its occur-
1975),containstranslations ofthemainsourcesthatrefer toEast renceat Shangaplacesitsuse thereas earlyas thetenthcentury,
AfricaduringtheearlyIslamic period. butstaircaseminaretscontinuedtobe usedin theGulfformany
7. Idrisistates,in theseventhsectionofthefirstclimeoftheKitdb centuries,as D. Whitehouseshows,"StaircaseMinaretson the
Rujar,thatBarawa (Somalia) was thelast ofthelands ofthein- PersianGulf,"Iran10 (1972): 155-58.Schacht'sargumentabout
fidels,whohave no religiouscreed,buttakestandingstonesand recessedminbars'beinganotherUmayyadrelicsurvivalin East
anointthemwithfishoil and bow downbeforethem.But in the Africais lessconvincing, withno knownexamplespredatingthe
Fridaymosqueat Barawa, thereis an inscription datedA.H.498 late eighteenth century;seeJ. Schacht,"An UnknownType of
(1104-5). Minbar and Its Historical Significance,"Ars Orientalis5,2
8. M. C. Horton,"Early MuslimTradingSettlements on theEast (1957): 149-73. Garlake, EarlyIslamicArchitecture, pp. 74-75,
AfricanCoast," Antiquaries Journal67, 2 (1987): 290-323; idem, G. R. Smith,"A Recessed Mimbar in the Mosque at Simam-
"Asiatic Colonizationof the East AfricanCoast: The Manda baya," Azania 8 (1973): 154-56; and H. N. Chittick,"The
Evidence"Journal oftheRoyalAsiaticSociety,pt. 2 (1986): 201-13. Mosque at Mbuamaji and theNabahani," Azania4 (1969): 59-
9. R. Soper,"Archaeo-Astronomical Cushites:Some Comments," 60, show thatSchacht'sdates are oftentoo early.
Azania 17 (1982): 145-62,pointsout thatsome of thedetailed 20. R. B. Lewcock,"Architectural ConnectionsbetweenAfricaand
claimsforastronomicalobservationby Cushiticgroupscannot PartsoftheIndian Ocean Littoral,"ArtandArchaeology Research
be sustained. Papers9 (1976): 15.
10. A. Werner,"The Bantu Coast TribesoftheEast AfricaProtec- 21. Creswell,Short Accountpp. 221-22.
torate,"JournaloftheRoyalAnthropological Association45 (1915): 22. Unpublishedfieldnotes.
326-54; T. Spear, TheKayaComplex (Nairobi:East AfricanLiter- 23. Garlake,EarlyIslamicArchitecture,fig.3.
atureBureau, 1978); H. Muturo,"An ArchaeologicalStudyof 24. Ibid,fig.50; H. N. Chittick,"PreliminaryReporton theExcava-
the MijikendaKaya Settlements on HinterlandKenya Coast," tionsat KizimkaziDembani, Zanzibar," Tanganyika Antiquities
Ph. D. diss., UniversityofCalifornia,Los Angeles,1987,espe- Report for1960(Dar es Salaam, 1962), pp. 17-19.
ciallyplan on p. 170. 25. H. N. Chittick, KisimaniMafia: Excavationsatan IslamicSettlement
11. Lit. "Cattle byre,"a possibleclue tothepastoralistoriginsofthe ontheEastAfrican Coast,AntiquitiesDivisionOccasional Paper
Mijikenda. no. 1 (Dar es Salaam, 1961),and "Reporton theExcavationsat
12. D. Nurseand T. Spear, TheSwahili(Philadelphia:University of Kisimani Mafia and Kua." TanzaniaAntiquities Reportfor1964
PennsylvaniaPress,1985). (Dar es Salaam, 1966),pp. 15-16.
13. T. Wilson,"Spatial Analysisand Settlement on theEast African 26. H. N. Chittick,"Kilwa, A Preliminary Report,"Azania1 (1966):
Coast," Paideuma28 (1982), fig.2; M. C. Horton,"The Early 3.
SettlementoftheNorthernSwahiliCoast," Ph. D. diss.,Cam- 27. Creswell,Short Account,p. 4.
bridgeUniversity,1984;C. Revoil,"Voyage chez les Benadirs, 28. J. P. Greenlaw, The CoralBuildingsof Suakin(London: Oriel
les Qomaliset les Bayouns,en 1883,"Le TourdeMonde56 (1888): Press, 1976); S. M. Head, "Corals and Coral ReefsoftheRed
385-414. Sea," in A. J. Edwardsand S. M. Head, eds. KeyEnvironments,
14. Creswell,Short Account, p. 3. RedSea (Oxford:PergamonPress,1987),pp. 130-33.
15. S. Denyer,African Traditional
Architecture(London: Heinemann, 29. F. Varanda,ArtofBuilding inYemen(London:Artand Archaeol-
1978),pp. 62-63. ogyResearchPapers, 1981),pp. 81-99.
16. N. Chittick,Manda,Excavations atanIslandPortontheKenyaCoast, 30. Y. M. Kobishchanov,Axum(UniversityPark,Pa.: Pennsylvania
BritishInstitutein Eastern Africa,Memoir no. 9 (Nairobi, State UniversityPress,1979),p. 141.
1984),p. 18. 31. Chittick,Manda,pp. 41-44.
17. Creswell,ShortAccount, p. 9. 32. F. Anfray, "L'archeologied'Axoumen 1972,"Paideuma 8 (1972):
18. Creswell,ibid,p. 414; D. Whitehouse,"The SmallerMosques at 60-76.
Siraf:A. Footnote,"Iran12 (1984): 166-68;and idem,TheCon-

This content downloaded from 129.96.252.188 on Fri, 15 Jan 2016 10:35:29 UTC
All use subject to JSTOR Terms and Conditions