. 8
( 11)


suggested being in the ¬rst millennium ad. Bananas are most widely grown
in the Lake Victoria Basin (Schoenbrun 1993, 1998) and in parts of western
Africa. The recent demonstration (p. 202 above) that they were present in
Cameroon in the mid-¬rst millennium bc now suggests that their antiquity
further east may be greater than was previously supposed. We do not yet
know to what extent people of the Chifumbaze complex grew these and
other vegetatively propagated crops, but they are likely to have been signif-
icant, particularly in the better-watered areas.
The evidence for domestic livestock is somewhat more comprehensive,
although of such uneven distribution that it is probably dangerous to gener-
alise. Sheep and/or goats, together with cattle, are attested on sites covering
virtually the whole time-span of the Chifumbaze complex: cattle seem to
have become much more numerous in southeastern Africa from around the
seventh or eighth century ad. The breed or breeds of these cattle is still a
matter for speculation (Grigson 1996, 2000); the extent to which they were
humped has not been ascertained and the possibility that cattle of Indian
origin were introduced through the Swahili coastal settlements remains
to be evaluated. Domestic fowl (MacDonald 1992; MacDonald and Edwards
1993) appear in eastern and southeastern Africa during the mid-¬rst mil-
lennium ad, in Chifumbaze-complex contexts. Whether they derived from
more northerly parts of Africa or more directly from their Asian homeland
remains unknown.
Throughout this period, the hunting of wild animals retained consider-
able economic importance. Iron arrowheads and spearpoints (which need
not, of course, have been used exclusively for hunting) have been recovered
from many sites, and those from which bones of wild animals have been
excavated are appreciably more numerous than those that have yielded
remains of domestic stock. At Kalundu, in southern Zambia, the faunal
remains from successive layers show the gradual replacement of wild by
domestic species (Fagan 1967). The wild animals represented include wilde-
beest and buffalo, as well as many of the smaller antelope. Fish bones are
rarely preserved, but have been recorded in Malawi, notably at Nkope.
The appearance and lay-out of houses and settlements represents an
aspect of the Chifumbaze complex that has only rarely been addressed by
the research so far undertaken. Elucidation of these matters requires far
more extensive excavation than has generally proved practicable, but will
undoubtedly be a focus for future investigation. Many sites have yielded
fragments of clay house-walls bearing impressions of vertical stakes and
interwoven withies; together with traces of clay ¬‚oors and of burnt thatch,
these indicate that the general type of house erected in Chifumbaze-complex
villages was not dissimilar to those used in many rural areas of eastern and
southern Africa during recent times. Whether these houses were round or
Iron-using peoples before ad 1000 261

rectangular is, however, rarely apparent; we generally have little informa-
tion about their internal arrangements. In recent times such features have
shown signi¬cant variation between different areas, and clear correlations
may be seen with the general life-style, economy and socio-political system
of the inhabitants (Huffman 1989, 1993, 2001). People in southeastern Africa
whose lives centre around their herds of cattle, for example, often live in
villages where houses, usually round, are arranged around a central corral.
In southern parts of D. R. Congo, on the other hand, the predominantly
agricultural people live in rectangular houses set in straight lines beside a
street. The extent to which these distinctive patterns may be traced back in
the archaeological sequence is clearly of considerable signi¬cance. In fact,
there is growing evidence that the former pattern, that of the southern
cattle-herders, had originated by late in the ¬rst millennium ad, when it is
attested at Toutswe and related sites in northeastern Botswana and adjacent
parts of South Africa. It is, however, important to recognise that cultural con-
tinuity cannot be assumed: practices and their rationale (cf. Huffman 1998)
can both change, and do so independently.
By the tenth century a clear distinction may be seen between Chifumbaze
communities in the central and western savannas and those of southeast-
ern Africa. In the former region, metal clearly formed a medium for the
accumulation and exchange of wealth, a function served in the southeast
by cattle (Vansina 1984; D. W. Phillipson 1989a). This important distinction
is one which rose to even greater prominence in subsequent centuries.

The contribution of Bantu linguistic studies
In order that maximum use may be made of the evidence afforded by linguis-
tic studies, it is necessary to consider the possibility of a correlation between
the archaeological picture described above and the processes of the disper-
sal of the Bantu languages. In the eastern and southern savanna the geo-
graphical distribution of the Chifumbaze complex is almost identical with
the area occupied in more recent times by people speaking languages of the
closely interrelated Bantu type. This observation lends support to the widely
held belief that this complex represents the archaeological manifestation
of the initial eastward and southward dispersal of Bantu-speakers beyond
the equatorial forest. The view that the people responsible for the Chifum-
baze complex were the ¬rst local speakers of Bantu languages is widely
accepted (Ehret and Posnansky 1982; Ehret 1998; Holden 2002; Bellwood
and Renfrew 2003) but has not gone undisputed (see, most recently, Robert-
son and Bradley 2000), particularly by those who have paid relatively little
attention to linguistic studies. The Bantu languages, which are today spoken
by upwards of 250 million people spread over an area of nearly 9 million
262 afric an archaeolog y

square kilometres, show a remarkable degree of inter-comprehensibility;
there can be no reasonable doubt that they have attained their present
wide distribution as a result of dispersal from a localised ancestral lan-
guage within the comparatively recent past -- certainly within the last 3000
or 4000 years. Linguists are virtually unanimous in the belief that this ances-
tral Bantu language was spoken close to the northwestern border of the
present Bantu-speaking area -- in what is now Cameroon and eastern Nigeria
(cf. Dalby 1975). It is likewise widely agreed that the modern Bantu
languages, considered together, may be divided into at least two major
groups, spoken respectively in the western and eastern parts of Bantu
Africa (Vansina 1984, 1995; but see Nurse 1982; Holden 2002). The mod-
ern boundary between the two groups closely follows the eastern edge
of the equatorial forest and the western branch of the Rift Valley from
Lake Albert southwards to the southern end of Lake Tanganyika, but is
less clearly de¬ned in the savanna further to the south where there seems
to have been much movement and interaction (Bastin et al. 1983; Vansina
It has recently been emphasised that a direct linguistic comparison
between western and eastern Bantu may be misleading (Ehret 2001): while
the latter is indeed a clear entity, the former is far more heterogeneous,
effectively de¬nable only as ˜the rest™ or ˜non-eastern Bantu™. This view, here
accepted, had in fact been foreseen by Vansina (1984, 1990), who had empha-
sised the distinction between the Bantu languages of the northwest and the
equatorial forest on the one hand and, on the other, those of the southwest-
ern savanna which are much less sharply distinguished from eastern Bantu
than are their forest counterparts.
The earliest dispersal of Bantu-speakers appears to have been that of the
western group in the equatorial forest (Vansina 1990; Bastin et al. 1999),
with lexicostatistical studies indicating several stages successively to Bioko,
Gabon, northern and central D. R. Congo (Fig. 132). In much of the savanna to
the south of the forest, the af¬nities of the recent languages are less clearly
understood; those in more westerly regions, while clearly related to those of
the forest, suggest that much interaction and linguistic development took
place following contact with speakers of eastern Bantu.
The eastern Bantu languages show much less diversity than do those of
the western group. This implies that their differentiation and expansion
may have been comparatively recent, but it also makes the elucidation of
these processes correspondingly dif¬cult (Nurse 1982). The problem is com-
pounded by the long history of linguistic borrowing that must have taken
place between populations living in close proximity and interaction with
one another (Bennett 1983). Studies of loanwords from non-Bantu sources
have, however, suggested that most of the modern Bantu languages of
Iron-using peoples before ad 1000 263


5 Zaire

3 6



Fig. 132: The
distribution of Old ˜Western™ Bantu languages,
with stages of expansion
Bantu languages
Eastern Bantu languages
0 1000 km
(modi¬ed from
Bantu languages showing
Vansina 1984, 0 500 miles western and eastern features

eastern and southern Africa may be derived from a form that was spoken
in the area north of Lake Tanganyika (Ehret 1973).
A large majority of the known sites of the Chifumbaze complex are in
the area where eastern Bantu languages are now spoken. While this may
be due to the fact that the more westerly regions have seen substantially
264 afric an archaeolog y

less archaeological research, the possibility remains that the Chifumbaze
complex may prove to be the archaeological signature speci¬cally of eastern
Bantu and its in¬‚uences.
It is now appropriate to compare the conclusions of historical linguis-
tic studies with the picture that emerges from archaeological research.
The methodological problems inherent to such an exercise have been set
out in chapter 1 (pp. 6--9): here it must be stressed once again that the
two approaches have totally different aims, emphases and data-bases, so no
absolute correlation between their conclusions should be either assumed or
expected. The comparison is nevertheless a perfectly valid exercise in histor-
ical reconstruction since each study may supplement or expand the results
of the other. An account of this stage in African prehistory, presented in
primarily linguistic terms, has been published by Ehret (1998).
The ¬rst point which emerges is that, as noted above, it is speci¬cally with
the eastern Bantu languages, including those of the southern savanna inter-
action zone, that the Chifumbaze complex shares its distribution. Archaeo-
logical evidence that the Urewe sites represent the earliest manifestation of
the Chifumbaze complex is mirrored by the linguists™ suggestion that the
modern eastern Bantu dialects are derived from a language that was for-
merly spoken in the area north of Lake Tanganyika. There is, however, no
reason to argue that a forest origin is inherently unlikely or to belittle the
contribution of indigenous East African elements to the development of the
Urewe group and the Chifumbaze complex as a whole.
The economic bases of life in the equatorial forest are markedly different
from those of the East African savannas. The herding of domestic animals
other than goats is effectively impossible in the former area. Likewise, agri-
culture in the forest is based upon vegetatively propagated crops, whilst in
the savanna cereals are of major importance. It is hardly surprising to ¬nd
these differences re¬‚ected linguistically. Some words for ˜goat™ in both east-
ern Bantu and its antecedents can be traced to a common ancestral form.
Many of the terms used in eastern Bantu languages to designate cattle and
sheep, on the other hand, were borrowed from non-Bantu sources that are
also represented in more northerly areas, and the same is true for words
connected speci¬cally with the cultivation or processing of cereals (Ehret
1973). The concurrence in the southern savanna of farming practices derived
from the forest zone as well as from more easterly regions is re¬‚ected in
the linguistic evidence in that area for interaction between Bantu-speakers
of the eastern savannas and those whose origins lay in the forests to the
north. We may conclude that there is strong circumstantial evidence that
the beginning of farming in central, eastern and southern Africa was con-
nected with the dispersal of people who spoke Bantu languages (Ehret and
Iron-using peoples before ad 1000 265

Posnansky 1982; D. W. Phillipson 2003a). In very broad terms, there is some
similarity between the regional divisions of the early farming cultures and
the linguistically de¬ned stages of Bantu expansion.
We have stressed above how the Chifumbaze complex represents a major
archaeological discontinuity in almost all parts of its distribution area. Its
settlements, economic practices and technology were all very signi¬cantly
different from those which had gone before, and were all known during
earlier times in more northerly parts of Africa. There can be little doubt that
the dispersal of the Chifumbaze complex was largely due to the expansion of
a new population element. The apparent link with a linguistic dispersal adds
strength to the argument, but no one-to-one correlation should be assumed:
there is no reason to suggest that all farmers in these regions spoke Bantu
languages, or that all Bantu-speakers were necessarily farmers. It is, however,
highly likely that the beginnings both of farming and of metal-working in
these regions were due at least in part to the activities of people who spoke
Bantu languages. On that basis, we may now turn to consider how these
processes may have taken place.

Mode of dispersal
From the data which have been summarised above, there can be little doubt
that the Chifumbaze complex was introduced into sub-equatorial Africa as
a result of a substantial and rapid movement of population. The entire
culture represented on sites of this complex can be shown to be foreign to
the areas in which it occurs, and most of its constituent features may be
traced to a source or sources in the northern savanna. The large number of
available radiocarbon dates accords with this view, and linguistic evidence
also lends it a considerable degree of support. More problematic is seeking
to understand how this process may have occurred (cf. Collett 1982).
Despite the novelty of the metal-using farmers™ life-style, economy and
technology, it is clear that these were not rapidly or totally accepted by the
indigenous populations. In several areas, indeed, there is plentiful evidence
both from archaeology and from oral traditions for the survival of people
who continued to practise the old microlithic technology long after the
appearance of metallurgy. In some parts of south-central Africa this con-
tinued until only two or three centuries ago, although the degree of this
survival obviously varied according to the intensity of farming settlement
in the various areas. The best-studied of these very late microlithic indus-
tries are those of northern and eastern Zambia. Here, detailed analyses
of industrial successions covering the last three millennia, such as those
from Nakapapula, Thandwe and Makwe, indicate no signi¬cant discernible
266 afric an archaeolog y

typological changes such as might be expected had there been any major
change in the hunter-gatherer economy and life-style of the stone-tool-
makers as a result of cultural contact with their farming contemporaries
(D. W. Phillipson 1976, 1977a). That some form of contact did in fact take
place between the indigenes and the immigrant food-producers is indicated
by the presence in nearly all the microlithic assemblages of this period of
varying numbers of sherds of characteristic Chifumbaze pottery (Musonda
1987). The persistence of the mode-5 industries, showing only gradual typo-
logical development following trends that were already apparent before the
¬rst arrival of farmers, suggests that such contact between the two groups
was usually minimal. By contrast, in areas where farming settlement at
an early date was relatively dense, there seems to have been fairly rapid
displacement of the hunter-gatherer populations as, for example, in much
of southern Zambia. One plausible reconstruction of the interactions which
may have taken place between hunter-gatherers and farmers is that of a tem-
porary client relationship, such as has been recorded in recent times both
in southern Africa and further to the north (Barnard 1992; Mercader et al.
2000). This analogy also suggests that, in some areas, relations between the
two population groups were characterised by aloofness or mutual avoidance
(Musonda 1987). It must be remembered that virtually all our evidence for
interaction comes from the territory of the Chifumbaze complex™s eastern
facies; we have as yet no reliable means of knowing whether the processes
were similar in more westerly areas.
The ¬nal absorption, conquest or displacement of the stone-tool-using
hunting peoples in south-central Africa may be attributed to the expanding
population of the early second millennium, which was marked by increased
emphasis on the herding of domestic animals, notably cattle. With the pas-
sage of time, continued expansion would have brought the cattle-keeping
farmers and the hunters into intensi¬ed competition. Except where stable
symbiosis was established, the ultimate result was apparently the disappear-
ance of the stone-tool-making traditions in all but those few areas which
were unsuitable for farming settlement. An example of the former situa-
tion is provided by the pygmies of the eastern equatorial forest (d™Hertefelt
1965; Turnbull 1965; Mercader et al. 2000). The early history of these
small-statured hunting people is completely unknown, as is the language or
languages which they originally spoke. Today, most pygmy groups are
involved in a client relationship with their agricultural neighbours whose
language, whether Bantu, Sudanic or Ubangian, they have adopted. What is
now abundantly clear is that the process was not one of simple conquest.
The arrival of the farming peoples in the territories of the indigenous
hunter-gatherers of southern Africa also appears to have been recorded
Iron-using peoples before ad 1000 267

Fig. 133: Rock
painting at
Silozwane Cave,

by the latter in rock paintings. The representation of a group of people
at Silozwane Cave in the Matopo Hills of Zimbabwe is believed to show
grain-grinding and other activities of agriculturalists; the people them-
selves are quite different in appearance from the hunters represented in the
earlier paintings (Fig. 133). Elsewhere, fat-tailed sheep are clearly recognis-
able (Summers 1959). Later paintings and occasional engravings, consisting
mainly of geometric and schematic designs (Fig. 134), are best attributed
to the farmers themselves. In some areas, including Malawi and eastern
Zambia, they may be linked with art forms associated with religious prac-
tices that have continued into recent times (D. W. Phillipson 1972b, 1976;
Gutierrez 1996). Similar instances have been discerned in South Africa
(Prins and Hall 1994), where there is evidence that much of the non-
representational art was the work of herding peoples (B. W. Smith and
Ouzman in press).
Although there can be little doubt that the inception of farming in Africa
south of the equator was brought about through movement of population, it
is not easy to suggest what may have been its causes and motivation. While
the number of people involved at any one time may not have been very
large, it was evidently suf¬cient to sustain the migrants™ distinctive life-style
and customs. It is most probable that entire family groups were involved in
the movements. While population pressure in the original Bantu homeland
in what is now Cameroon and eastern Nigeria may have been one of sev-
eral causal factors, it would not explain why they moved over such a vast
268 afric an archaeolog y

Fig. 134:
Schematic rock
painting at
Sakwe, Zambia

and sparsely inhabited area so rapidly -- the eastern facies appears to have
expanded southwards at an average rate of some 15 kilometres a year: about
350 kilometres per generation. The advantages conferred by the knowledge
of metallurgy, stock rearing and, probably to a lesser extent, agriculture
would greatly have facilitated this expansion (Vogel 1987), but there are
likely to have been other factors also of which we are as yet unaware. It is
useful to speculate on the reasons why the ¬rst-millennium-ad expansion of
the Chifumbaze complex, which proceeded with such remarkable rapidity
as far to the south as the Vaal River and KwaZulu-Natal, was arrested so
abruptly just beyond the Kei River. Although the site distribution remains
very imperfectly known, especially in the west, it is nevertheless reasonably
clear that its southern limit broadly coincides with the northern edge of the
southwest African zone of desert vegetation and that of the long-grass veld
of South Africa™s Free State Province. On the eastern side of the continent
the southernmost extent of the Chifumbaze complex is restricted to the
west by the uKhahlamba/Drakensberg and to the south by the Cape winter-
rainfall zone. Recent societies have demonstrated that this frontier presents
no barrier to herding communities; it is none the less an effective south-
ern limit for the cultivation of traditional African cereals and other food
crops. It seems reasonable to suppose, therefore, that agricultural potential
limited the initial expansion of farming people into southern Africa. It was
not until the development of the more pastorally oriented societies of the
early second millennium ad that further southward expansion took place,
Iron-using peoples before ad 1000 269

resulting in the settlement by Bantu-speaking peoples of such marginal
lands as those of the Free State Province and parts of Namibia.

Madagascar and the Comoro Islands
The huge island of Madagascar and the small Comoro archipelago require
consideration, albeit brie¬‚y. Human settlement of these islands has an antiq-
uity of less than two thousand years (Dewar and Wright 1993; Dewar 1996;
Allibert and V©rin 1996; Rakotoarisoa 1998). Throughout this time, they have
maintained varying degrees of contact with the African mainland, and this
is re¬‚ected in certain aspects of their culture and language. Today, the pop-
ulation of the Comoro Islands is essentially Swahili, although there is evi-
dence -- primarily but not exclusively linguistic -- that this situation was
preceded by one when connexions with Madagascar were dominant. The
Malagasy culture, on the other hand, has little in common with those of
the mainland and the language is Austronesian, being apparently derived
most directly from the speech of Borneo, although it includes numerous
loanwords and other evidence for contact with speakers of African Bantu
languages. There is historical evidence for the presence of Bantu-speaking
enclaves on the west and northwest coasts; it is likely that at least some
elements of the Austronesian population settled for a while on the East
African coast before moving to Madagascar, although no clear archaeological
indication of this has yet been recognised (cf. Blench 1996). The connexion
between these processes and the advent to East Africa of cultural elements
originating across the Indian Ocean remains to be demonstrated.
The oldest evidence for the human settlement of Madagascar is essen-
tially circumstantial, there being no de¬nite archaeological traces which are
securely dated before the late ¬rst millennium ad. However, major changes
in vegetation and in the animal population suggest that there may have been
a human presence, perhaps intermittent and restricted to coastal regions,
since around the beginning of the Christian era. The ¬rst urban development
on Madagascar took place at Mahilaka on the northwest coast around the
twelfth century. Between the fourteenth and the sixteenth centuries, there
is evidence for the growth of social hierarchies, followed by the formation
of states.

Stone-tool-using herders of southwestern Africa
Beyond the area that was settled by the farmers of the Chifumbaze complex,
principally in Namibia and in the Northern and Western Cape Provinces
of South Africa, use of lithic technology continued throughout the ¬rst
millennium ad and even into more recent centuries (Klein 1986; Kinahan
270 afric an archaeolog y

Snake Rock




ge Limerock
ran Doornfontein
R. O Blinkklipkop Dikbosch

Spoegrivier Seacow

Eland™s Bay
Tortoise Cave
Fig. 135:
Principal sites in Boomplaas Scott™s
southwestern Cave
Africa which
have yielded Blombos
Hawston Nelson
Land over 1000 metres Die Kelders Bay Cave
remains of
0 300 km Byneskranskop
domestic sheep
in precolonial 0 200 miles

1991; Brink and Webley 1996; Webley 1997; Bousman 1998). There can be
little doubt that the people responsible for these stone industries were
KhoiSan-speakers, ancestors of the more recent Khoi (˜Hottentots™) and San
(˜Bushmen™) of southern and southwestern Africa. At an early date some of
these stone-tool-using people (Fig. 135) obtained access to domestic sheep
and -- although this is less certain -- cattle. Bones of domestic animals make
their appearance on sites in Namibia and the Western Cape which are dated
to the ¬rst or second century ad. Pottery ¬rst occurs in the local archaeo-
logical sequence at approximately the same time, but there remains some
Iron-using peoples before ad 1000 271

Fig. 136: Cape
coastal pottery
(after Rudner

uncertainty as to whether the two innovations were intimately linked. The
pottery shows no particular similarity to wares in adjacent regions from
which it could have been derived, the earliest material being usually thin
and well made; later vessels, especially in coastal regions, comprised pointed-
based jars and beakers (Fig. 136; Rudner 1968; Schweitzer 1979). The latter
wares seem also to have been used by communities who did not have access
to domestic stock and who continued their traditional reliance on gathering
wild vegetable foods, collecting shell¬sh, hunting and ¬shing (A. B. Smith
et al. 1991). There has been much controversy over the nature of the cul-
tural processes which these archaeological manifestations represent: it has
frequently been assumed that the distinction between hunting and herd-
ing KhoiSan-speaking populations recorded by European visitors from the
¬fteenth century onwards could be projected back to the beginning of the
Christian era when their ancestors had entered southwestern Africa along
routes which might, it was sometimes proposed, be demonstrated archae-
ologically. A related assumption was that some hunting peoples may have
obtained varying numbers of sheep (perhaps as prey rather than as livestock)
from neighbouring herding populations whose remains have not yet been
recognised archaeologically.
Recent research makes such interpretations seem increasingly unlikely.
A complicated pattern of local seasonal resource-exploitation has been
revealed, but does not include any clear distinction between hunters and
herders except in parts of the Northern Cape where variation in lithic
technology has been attributed to such a division (Parsons 2003). At least
in some areas, such as the west coast north of the Cape of Good Hope
(Fig. 137), this seasonal pattern seems to have involved movement of popu-
lation between, on the one hand, the inland regions where wild vegetable
foods were plentiful during the summer and where the diet could then
readily be supplemented by the meat of small animals and, on the other
hand, the caves and open sites near the coast where shell¬sh were exploited
272 afric an archaeolog y

Fig. 137: The
early herder site
of Kasteelberg,
Western Cape

during the winter (Parkington and Hall 1987). This appears not only to have
been the main regimen of life during the centuries which preceded the
advent of domestic sheep, but also to have continued into relatively recent
centuries in several areas without signi¬cant discernible modi¬cation (Kina-
han 1996). It is now considered likely that some level of sheep-herding was
adopted by some of these seasonally migrating groups, and that herders
per se did not form a distinct population element (cf. A. B. Smith 1990,
1998b; Sadr 2003).
There are no local wild ancestors for domestic sheep or cattle in the rel-
atively well-studied South African faunal assemblages of the last millennia
bc, so there can be no reasonable doubt that the ¬rst domestic animals in
southernmost Africa were introduced into the region from more northerly
latitudes. Links with the stone-tool-using herder peoples of East Africa were
at one time considered but have not proved easy to demonstrate. Recent
research has located a potential source in much greater proximity.
In the Matopo Hills and neighbouring parts of southwestern Zimbabwe,
pottery of the idiosyncratic type known as Bambata ware occurs in the
upper levels of several rockshelter deposits (K. R. Robinson 1966a): it appears
to have been the ¬rst pottery produced in this particular area and to be
associated not with the working of iron but with the production of mode-5
lithic artefacts. At Bambata Cave it is also associated with bones of domestic
sheep and dated around the second century bc (N. J. Walker 1983). Although
typologically distinct, at least in its earlier occurrences, Bambata ware shows
Iron-using peoples before ad 1000 273

some af¬nity with pottery of the Chifumbaze complex, although its ¬ne,
thin fabric is particularly distinctive. Analogous material is now reported
from several sites in north-central Botswana and in Namibia (Jacobsen 1984;
Denbow 1986; Huffman 1994; Reid et al. 1998). It was noted above that the
earliest pottery in the Western Cape, as at Die Kelders (Schweitzer 1979),
may also be distinguished from the later wares by its exceptionally thin,
¬ne fabric.
The presence of domestic animals in southwestern Zimbabwe during the
last centuries bc is no longer surprising in view of the evidence cited above
(p. 256) for the very early establishment of the Chifumbaze complex in the
upper Zambezi Valley. If, as is now indicated, this region was penetrated in
about the third century bc by pioneers of the main dispersal of farming
economy into southern Africa, we can understand how, shortly afterwards,
certain technological and economic traits were adopted from this source by
people living further to the south, in what is now Botswana and southwest-
ern Zimbabwe (D. W. Phillipson 1989b). The relevant traits were, primarily,
the herding of sheep and (possibly later) cattle and the manufacture of
pottery. Metal-working and cultivation, if practised in the upper Zambezi
Valley at this time, were probably not adopted further to the south until
signi¬cantly later. The potential for cultivation was, in any event, severely
restricted by the arid climatic conditions of southwestern Africa. It was from
these central regions of southern Africa that herding and pottery manufac-
ture probably spread to the southwestern coastal regions by very early in the
Christian era (Sadr 1998; A. B. Smith 1998a). This conclusion is concordant
with linguistic arguments (Ehret 1982; Elphick 1985) which also suggest that
domestic stock was acquired by KhoiSan-speakers in northern Botswana or
a neighbouring area, and thence transmitted southwards. Thus it was that
herding was introduced into southwesternmost Africa some two centuries
before the farmers of the Chifumbaze complex penetrated the coastlands of
The second millennium AD in sub-Saharan Africa
The last 1000 years
In many parts of Africa the last 1000 years comprise a period for which
archaeology, although still of considerable importance, is by no means our
only source of information. Linguistic reconstructions which, when taken
together with the results of archaeology, have been valuable for illuminating
earlier times, now become less speculative. For the more recent periods the
oral historical traditions of many African societies preserve a great deal of
valuable information, even though their interpretation is exposed to many
pitfalls. In some areas of the continent, written records are also available.
For much of northern Africa, indeed, the last 1000 years fall fully within the
period of written history, and for that reason this chapter is concerned only
with the regions lying to the south of the Sahara. Here, some areas, such as
those of the sudanic kingdoms and parts of the East African coast, were in
more-or-less regular contact with people from literate communities in whose
records much useful historical detail has been preserved. These writings
may be used in conjunction with information from other sources in the
building up of a composite picture of the period™s events and developments.
Elsewhere, we have no signi¬cant written records pre-dating contact with
the European traders and colonisers who gradually established control over
most of Africa between the ¬fteenth and the nineteenth centuries. With
this process the subject-matter of this book comes to an ill-de¬ned end; and
the study of the African past enters the ¬eld of conventional history (Iliffe
To a large extent interest in the archaeology of sub-Saharan Africa after
ad 1000 has, until recently, focussed on evidence for the progress of long-
distance trade and the development of states (cf. Garlake 1978a; Connah
2001). In many areas the two processes may have been to some extent con-
current and dependent on one another. Trade has been seen as both a
stimulus and a mechanism for state-formation through the opportunities
it provided for individuals or groups to monopolise or control the distri-
bution of wealth. However, as will be shown below, it is in several areas
becoming increasingly apparent that essentially local resource exploitation
and exchange systems often provided the necessary basis. Political centrali-
sation made it possible for the products of a region to be gathered together
for organised redistribution. These factors have had a signi¬cant effect on
The second millennium ad in sub-Saharan Africa 275

our sources of historical information. State centres or capitals provide many
attractions for the archaeologist; such sites, because of their size, wealth
or monumental architecture, may be readily recognised, leading to a ten-
dency to ignore less conspicuous sites that were inhabited by peasants or
subject communities. Thus trade and political power have tended to be over-
emphasised in archaeological studies. Likewise, as was shown in chapter 1,
oral traditions often serve to record and to support the position of ruling
groups or dynasties, which may therefore tend to be stressed at the expense
of other aspects of the past. One result of this bias in our available sources is
that we tend to know far more about the history of those peoples who devel-
oped centralised state-systems than of those who did not, and more about
activities relating to long-distance trade than about those connected with
domestic economy. One of the strengths of archaeology as a data-source for
African history and later prehistory is its ability to throw light on aspects of
past societies which are not stressed by studies rooted in other disciplines.
Archaeology is of particular relevance also to the study of the development
of acephalous societies.

West Africa
Some account was given in chapter 7 of the early stages of urbanism and
socio-political centralisation in the savanna country of West Africa before ad
1000, with particular reference to Jenne-Jeno and to the kingdoms of Ghana
and Kanem. Later centuries saw the further development and proliferation
of such kingdoms as well as the establishment of comparable institutions
further to the south. For the sudanic kingdoms, much of our information
now comes from written Arabic sources, which are in some areas beginning
to be supplemented by the testimony of archaeology (Levtzion 1973, 1977;
Levtzion and Pouwels 2000; Connah 2001). Indeed, throughout the period
with which this chapter is concerned, the widespread adoption of Islam has
been a major factor in African history, archaeology and culture, especially in
the northern half of the continent and along the Indian Ocean coast (Adahl
and Sahlstr¨m 1995; Insoll 2003).
To the second half of the eleventh century belongs al-Bakri™s description
of Ghana. He commented on the near-divine status of the ruler, who was
succeeded, not by his son, but by his sister™s son; on his death he was buried
beneath a large earth mound, accompanied by the bodies of his retainers. A
large armed force enabled the king to maintain control over many tributary
chiefdoms. The capital of ancient Ghana was divided into two areas. One,
surrounding the royal residence, was built in the local African style with
predominantly round houses of mud; while the other, inhabited mainly by
276 afric an archaeolog y

Muslim traders and other migrants from the north, was built of stone and
included several mosques (Levtzion and Hopkins 1981).
These observations may be paralleled by evidence from archaeology. Inter-
ments under tumuli are known from a wide area of the West African
savanna from Nigeria to Senegal; several examples have been excavated
(Desplagnes 1903, 1951; Joire 1955; Saliège et al. 1980; Gallay et al. 1982; S. K.
and R. J. McIntosh 1993; Connah 2001). Some of these burials, notably those
at Rao in Senegal which yielded ¬ne gold ornaments of the fourteenth cen-
tury (Joire 1943), clearly belonged to very powerful and/or rich individuals,
but mound-burial was by no means restricted to this class. What the excava-
tions do show is that traditional burial customs continued to be practised
long after the introduction of Islam. The apparent slowness by which the
new religion was adopted is not otherwise re¬‚ected in the results of archae-
ological research, largely because so much attention has been focussed on
sites mentioned in Islamic records. At Kumbi Saleh (see p. 241, above), a
wide central avenue and market area are lined by the remains of two-storey
stone buildings; the site™s population has been estimated at between 15,000
and 20,000 people and the artefacts indicate the North African connexions
and long-distance trade of the inhabitants (Mauny 1978). It is noteworthy,
however, that excavation seems to have been restricted to the traders™ area,
and that the presumed royal quarter has not been investigated. The bias
in our archaeological information is thus not surprising. To the north, a
broadly contemporary settlement at Tegdaoust in south-central Mauritania
has been more thoroughly excavated (Robert 1970; Robert-Chaleix 1989;
Fig. 138).
The trade which contributed so signi¬cantly to the prosperity of ancient
Ghana was conducted chie¬‚y in gold and presumably slaves from the south,
in salt from the Sahara and in copper and a variety of manufactured
goods from the north. These items are only very selectively preserved in the
archaeological record: one particularly informative discovery is that of an
abandoned caravan at Majabat al-Koubra in the Mauritanian Sahara (Monod
1969). Copper-alloy bars and cowrie shells were the principal objects carried,
presumably southwards, at a date now con¬dently placed in the twelfth cen-
tury. Trade in these commodities was controlled and taxed by the rulers of
ancient Ghana, although none was actually produced within the territory
that was subject to their jurisdiction, the source of gold being around Bam-
buk on the upper Senegal far to the southwest. (Exploitation of the Akan
gold deposits in modern Ghana probably did not begin until later.) Ancient
Ghana was in a strong position to exploit its intermediary position between
the suppliers and the trans-Saharan traders (Brett 1983). The majority of
the people of the kingdom of Ghana were probably of northern Mande
stock. Oral traditions recorded in Timbuktu during the sixteenth century
The second millennium ad in sub-Saharan Africa 277

Fig. 138: Plan of
houses at
Tegdaoust (after
suggest that its founding rulers may have been of Saharan Berber Sanhaja
(ancestral Tuareg) origin, although by the time of al-Bakri™s account in the
mid-eleventh century this dynasty had apparently been replaced by one with
more southerly af¬nities.
By the eleventh century several of the Berber peoples of the western Sahara
had converted to Islam. They then united as the Almoravids, a militant
group which expanded rapidly both to the north and to the south, coming
in the process into con¬‚ict with the kingdom of Ghana. The stability of the
kingdom was severely weakened and it never fully reverted to its former
prosperity (R. J. McIntosh 1998). More southerly Mande groups now rose to
prominence, and by early in the thirteenth century the empire of Mali came
into being as the effective successor to Ghana. Mali had a richer agricultural
base than its predecessor and exercised more direct control over the gold-
¬elds. By the fourteenth century its rulers held sway, from their capital on
278 afric an archaeolog y

Maximum extent of Mali c.14th century
Maximum extent of Songhai c. 15th century
Maximum extent of Kanem c. 14th century


Rao Kumbi Saleh



Oyo Ile
Fig. 139: West Benin
African sites and 0 1000 km
0 500 miles
discussed in
chapter 8

the upper Niger, over an extensive territory (Fig. 139), stretching from the
southern Sahara to the northern edge of the forest.
Downstream along the Niger the Songhai had, by the end of the ¬rst
millennium ad, formed their own riverine kingdom, with its capital at Gao,
east of the still-¬‚ourishing Jenne (Insoll 1996, 1997, 2000; Insoll and Shaw
1997). They also had strong trade links with the Mediterranean world, the
most striking evidence for which is a series of inscribed twelfth-century
tombstones near Gao (Flight 1975), which appear to have been made to
order in Spain and transported thence across the Sahara by camel. It is
noteworthy how the West African savanna™s centres of greatest prosperity
arose in places where goods were transferred from one mode of transport
to another, as from the camels used to cross the Sahara to the donkeys and,
increasingly, horses of the savanna or the canoes of the Niger River.
Before the end of the thirteenth century the Mande had established con-
trol over the Songhai kingdom, greatly increasing the power and prosperity
of Mali. Not long afterwards, Musa, ruler of Mali, went on a pilgrimage
to Mecca; the wealth and size of his entourage created a lasting impres-
sion and served to consolidate Mali™s position in the Islamic world (Sutton
1997). The works of North African writers now become major sources for
our knowledge of the sudanic kingdoms; unfortunately, neither they nor
the archaeology provide adequate information about the trans-Saharan slave
trade which undoubtedly took place over a long period, although its scale
may sometimes have been exaggerated. The power of Mali was broken in the
late ¬fteenth century by the Songhai sultan, Ali, who extended his rule east-
wards through Hausaland and northwards to Aïr. Mali was by then restricted
The second millennium ad in sub-Saharan Africa 279

to a small area west of the upper Niger. The enlarged Songhai empire was
short-lived, being destroyed at the end of the sixteenth century by a force
from Morocco.
To the east, the rulers of Kanem were converted to Islam late in the
eleventh century. For some hundreds of years, Kanem did not receive the
attentions of trans-Saharan contacts on the same scale as did ancient Ghana
and Mali; its exports, although valuable, lacked the lure of the gold of
Bambuk. Eventually, after a long period of internecine strife among the
rulers, the former Bornu province southwest of Lake Chad became the cen-
tre of a new empire, with Kanem reduced to tributary status. Troops of
cavalry were important in maintaining the power of the state. By the six-
teenth century the main trans-Saharan trade had shifted eastwards: the
Hausa states and Bornu were established as major markets where the prod-
ucts both of the sudan and of more southerly regions were exchanged for
goods from North Africa and beyond. The direct successors to these states,
and their urban centres such as Hamdallahi (Gallay et al. 1990), played an
important part in the more recent history of the area into the twentieth cen-
tury. Future archaeological research may be expected to throw much light
on their chronology and on the economic processes which underlay their
To the south, in the West African forests and on their northern fringes,
a contrasting situation prevailed, for here the events of the period between
ad 1000 and the arrival of the ¬rst European voyagers at the end of the
¬fteenth century are known only from archaeology, supplemented to some
extent with the oral traditions preserved by more recent peoples. Only very
indirectly are events of these southerly regions recorded in the Arabic writ-
ings which have proved such useful sources for the history of the sudanic
kingdoms. Unfortunately, archaeological investigation of the second millen-
nium ad in the West African forest regions is still in its infancy, and has so
far been largely restricted to parts of modern Ghana and Nigeria. Research
has been concentrated at sites which have yielded important works of art in
terracotta or copper alloy, with the result that little comprehensive informa-
tion is yet available relating to the growth of states or the development of
the urbanism which has been such a prominent characteristic of the Yoruba
and neighbouring peoples in recent times.
In what is now Ghana, a series of small states arose by the fourteenth
century in an area between the Akan gold¬eld and the entrepˆts of the
middle Niger (Posnansky 1973; S. K. and R. J. McIntosh 1980). The best
known of these states was that centred on Begho near the Volta River, just
north of the forest margin. Excavations at Begho have demonstrated both
the relevance of recent oral traditions concerning the early history of the
town, and also the cultural continuity which has prevailed between the ¬rst
280 afric an archaeolog y

Fig. 140: Begho:
terracotta ¬gure
of a trumpet-
blower, and part
of an ivory
trumpet (after
Posnansky 1976)

inhabitants and the modern Brong (Fig. 140). The separate quarters of Begho
that were inhabited by artisans and by visiting traders may still be identi¬ed
(Posnansky 1973, 1976). Imported materials con¬rm the close connexions
with Jenne. Begho was ¬nally eclipsed by the rise of the Asante kingdom,
probably at the beginning of the eighteenth century (Anquandah 1982);
before this the ¬‚ow of gold northwards had been signi¬cantly reduced,
because the European traders who had arrived on the coast were also trad-
ing in gold. Knowledge of the subsequent history of the Ghanaian forest
area is, for the most part, derived from the oral traditions of the Asante
and their neighbours and from the records of European visitors to the coast
(DeCorse 1992, 2001).
In southern Nigeria, processes of urbanisation can be traced back to about
the beginning of the second millennium, but have been the subject of
remarkably little speci¬c archaeological research. Such work is, of course,
hindered by dense forest, intensive cultivation and the growth of modern
cities like Ife and Benin. One of the few early urban centres not obliterated by
modern building is Oyo Ile (Soper and Darling 1980; Soper 1993), although
disappointingly little archaeological information has yet been made avail-
able about its growth and development.
The second millennium ad in sub-Saharan Africa 281

Fig. 141:
Terracotta heads
from Ife

Probably the most signi¬cant archaeological discoveries of early second-
millennium date in this region have been made at Ife (Willett 1967; Eyo
1974; Garlake 1974, 1977, 2002). The site had evidently been occupied in
earlier times, but little is yet known about its initial phases. By about the
eleventh century more intensive settlement is indicated, marked archaeo-
logically by pavements made of large numbers of potsherds set on edge.
These pavements evidently formed open-air courtyards that were used for
domestic purposes, although in one case there were also the remains of
an altar, likewise decorated with sherds. Of the associated buildings, which
were made of sun-dried mud, few traces have survived. The greatest atten-
tion in archaeological work at Ife has been devoted to a remarkable series
of highly realistic ¬gures in terracotta (Fig. 141) or copper alloy, which are
thought by most art historians to represent a tradition derived -- albeit per-
haps indirectly -- from that of the much earlier ˜Nok culture™ described in
chapter 7. These ¬gures apparently had a religious signi¬cance and were on
occasion kept in shrines or on altars, which formed part of domestic houses.
They occur in several different areas of Ife, including those away from the
royal palace quarter, so it appears that they were not a prerogative of the
ruler. Although the archaeological arguments are tenuous, they suggest that
the Ife ¬gures were created for a variety of ritual situations, which may be
paralleled in more recent Yoruba traditional practices.
282 afric an archaeolog y

Fig. 142: Part of
the linear
earthworks in
the region of
Benin (after
Darling 1984)

Whilst Ife may be seen as heir to ancient urban and artistic traditions,
it probably owed much of its prosperity to its location, which enabled it
to participate in exchange between the products of the forest and luxuries
derived from or beyond the savanna. The copper used in its metalwork was
almost certainly derived from the east or the north, presumably by way
of the Niger Valley; connexions with the latter are attested by the pres-
ence there, as at Tada, of copper-alloy castings in an Ife-related style (Eyo
and Willett 1980). Glass beads were imported in large numbers and were
The second millennium ad in sub-Saharan Africa 283

Fig. 143: Benin
˜bronzes™: left -- a
head in early
style; right --
fragment of a
plaque depicting
a Portuguese

then melted down and re-worked. Virtually no information has survived con-
cerning the domestic economy of early Ife, but it is reasonable to assume
that yam cultivation, then as now, contributed signi¬cantly to its food
The ˜lost-wax bronzes™ -- they are actually made of brass -- of Ife are few
in number in comparison with the terracottas. Brass-working was carried
out on a far greater scale at Benin, where archaeological excavation shows
that there was, from at least the thirteenth century, a town where the
human sacri¬ces for which the place subsequently became notorious were
conducted (Connah 1975). Archaeological survey in the environs of Benin
has revealed the presence of a vast network of linear earthworks (Darling
1984; Fig. 142). It is suggested that the earliest of these date from the end of
the ¬rst millennium ad, and that by the twelfth century a territory several
thousands of square kilometres in extent had been delineated and subdi-
vided, including the city zones of Benin and Udo. The growth of Benin
itself may also be illustrated by the successive extensions to its surround-
ing walls and earthworks. The earliest Benin ˜bronzes™, which probably date
to the ¬fteenth century, show some stylistic connexions with those of Ife
(Fig. 143). At Benin, brass-working was carried out for the ruler, the Oba; and
it is relevant to note that the Obas of Benin originally had strong ties with
Ife. The history of Benin is fairly well known both from oral traditions and
284 afric an archaeolog y

from written sources, for the city was in contact with Portuguese traders
from the sixteenth century (Dark 1973; Connah 1975; see also Willett 1967).
Although most of the recorded history of Benin falls outside the period of
time covered by this book, archaeology is now beginning to illustrate the
growth of the city and set it in a comparative context.
In conclusion, it must be admitted that archaeology has so far contributed
rather little to our understanding of events in West Africa during the earlier
centuries of the second millennium ad. The concentration on sites that are
already known from written sources has had a stultifying effect on archaeo-
logical research, as has the emphasis on investigating artistic traditions,
especially those based on the working of terracotta and copper alloy. Since
wooden sculpture, which is of paramount importance in the art history
of more recent periods, hardly ever survives for longer than two or three
centuries, even this aspect of research has been noticeably incomplete.
Despite these problems, some important points do emerge from the archaeo-
logical investigations outlined above. Long-distance trade seems to have been
a sequel, rather than a causative factor, to the development of states and
cities both in the West African savanna and in the forest. Prosperity in both
zones originated in the exploitation of local resources. The emphasis of some
earlier writers on foreign elements, whether religious or material, should be
attributed to their focus and methodology. This revised viewpoint offers a
much-needed stimulus to further integrated research in West African archae-
Further to the east, in the equatorial forests of the Congo Basin and in
the savanna country to their north, the recent archaeology remains almost
completely unknown. The information about past events that we can derive
from other sources suggests that in the ¬rst half of the second millennium
ad there continued to be small-scale settlement of Bantu-speaking peasant
communities in the forest and of Ubangian-speakers on its northern fringes,
some of the latter still expanding eastwards, following the pattern that had
been established in earlier times (David 1980). The end result of this process
was the establishment in what is now the extreme southwestern Sudan
and adjacent parts of D. R. Congo and the Central African Republic of the
Ubangian-speaking Zande as rulers of a cluster of peoples of varied Bantu,
Sudanic and Nilotic antecedents.

Ethiopia, the southern Sudan and adjacent regions
Elsewhere in the southern Sudan, in the Bahr el Ghazal territory of the
Nilotic-speakers, peasant farming continued, with emphasis on the herd-
ing of cattle. The most signi¬cant economic development of this time, as
we see at the Wun Rok mound-sites early in the second millennium ad,
The second millennium ad in sub-Saharan Africa 285

was the replacement of the earlier humpless cattle by humped ones, which
may have been obtained from Arabic-speaking peoples further to the north
(David et al. 1981). The pottery, decorated with twisted-cord roulettes as in
earlier times, continued with no signi¬cant change. Indeed, the general pic-
ture which emerges from the limited research so far undertaken is one of
basic continuity from the ¬rst millennium ad into recent times. The ¬rst
half of the second millennium was, however, a period of major southward
expansion by Nilotic-speakers into East Africa, as will be further discussed
below. In parts of the southeastern Sudan, as in much of northern Kenya
and, probably, Uganda, iron remained extremely rare; use of stone tools
continued into the last few centuries (Fig. 144).
In highland Ethiopia, the history of this period is marked by the south-
ward shift in the centre of Christian culture from Aksum into the Lasta
and Shoa regions, the spread of Islamic culture in the east and southeast,
and the continued obscurity of events in the south. In the seventh century,
the kingdom of Aksum was in decline and its capital had probably been
moved to a new location. By this time, the inhabitants of the mountainous
Lasta area east of Lake Tana had adopted much of the earlier Christian cul-
ture. Here, around the beginning of the twelfth century, if not before, the
Zagwe dynasty arose and established its political authority over the area.
Like the written records, the surviving sites of this period are exclusively
ecclesiastical. Pride of place goes to the rock-cut churches at Lalibela, the
Zagwe capital (Fig. 145). Hewn on both the exterior and the interior from
solid rock, these churches show several architectural features that may be
traced back to Aksumite buildings (Gerster 1970). Indeed, the degree of con-
tinuity between ancient Aksum and the Christian civilisation of highland
Ethiopia has, until recently, been underestimated (D. W. Phillipson 2003b).
Numerous less elaborate rock-cut churches are widespread in Tigray. Towards
the end of the thirteenth century the Semitic-speaking Amhara of Shoa,
south of the Blue Nile, replaced the Zagwe ruling dynasty and the cen-
tre of political power shifted still further southwards (Sergew 1972); this
pattern of authority survived, albeit in modi¬ed form, into the twentieth
After the decline and eventual eclipse of Christian Nubia from the late
thirteenth century onwards, their co-religionists in Ethiopia became largely
cut off from the outside world. Islamic culture, by the beginning of the
second millennium ad, was reaching large areas of eastern Ethiopia from
the port of Zeila near the mouth of the Red Sea. So far, our knowledge
comes primarily from written records, but there are also extensive sites
of ruined towns which await archaeological investigation (Wilding 1980).
To the south, the later archaeology of southern Ethiopia remains almost
totally unknown apart from the presence of varied megalithic monuments,
286 afric an archaeolog y



Wun Rok

Gedi Malindi


Ivuna Kilwa

Bwana Mkubwa

Twickenham Road
Ingombe Ilede Ruanga
Naletale Taba Zikamambo
Khami Sofala
Nthabazingwe Zimbabwe
Bambandyanalo Schroda
{ Mapungubwe

Fig. 144:


sites of the
0 1000 km
millennium ad
0 500 miles
in eastern and
southern Africa
The second millennium ad in sub-Saharan Africa 287

Fig. 145: The
rock-cut church
of Abba Libanos,

some at least of which date within the last one thousand years (Azaïs and
Chambard 1931; Joussaume 1974, 1995; Anfray 1982). In northern Kenya and
much of Somalia stone tools probably continued in use until the last few
centuries: archaeological traces of nomadic pastoralist populations such as
the present Eastern Cushitic-speaking peoples of this area are notoriously
sparse and dif¬cult to interpret. The surviving oral traditions (H. S. Lewis
1966; Turton 1975) mostly relate to a southward expansion of the Oromo at
the expense of the Rendille and Somali, a process which has continued into
recent times.
In the better-watered areas of Kenya™s western highlands and parts of
Uganda, settled cattle-herders are attested from the beginning of the sec-
ond millennium. They built stone-walled semi-subterranean stock enclo-
sures and, at least in later times, practised irrigation agriculture (Sutton
1973, 1998c). There are good reasons for believing that these people were
Nilotic-speakers ancestral to the modern Kalenjin. Several later penetrations
of Nilotic-speakers southwards into East Africa are attested (Oliver 1977).
One incursion brought the Maasai into the Rift Valley country which had
been occupied by stone-tool-using herders during the last millennium bc.
Another introduced the Luo to their present territory on the northeastern
shore of Lake Victoria. The extensive stone-built terraces and irrigation works
at Engaruka in northern Tanzania (Sutton 1998a) may also be attributed,
albeit tentatively, to an early Nilotic-speaking population. To the south lay
the territory of the Bantu-speaking peoples, discussed below.
288 afric an archaeolog y

The east coast of Africa
Mention was made in chapter 7 of the written evidence, contained in The
Periplus of the Erythraean Sea, for trading voyages along the East African coast
as early as the ¬rst centuries of the Christian era. Only slight archaeological
evidence for such trade has yet been found (Chami 1999). It is not until the
early eighth century ad that we have clear traces of the coastal settlements
that were frequented by the Indian Ocean traders, who penetrated as far to
the south as the Maputo area of Mozambique (Sinclair 1982; Horton 1996b).
Indeed, we have virtually no information about any type of settlement on the
coast prior to the eighth century. Research has until recently been concen-
trated on the remains of stone buildings, and traces of more ephemeral occu-
pation of mud-and-thatch structures, perhaps extending back into earlier
times, have received comparatively little attention (Fleisher and Laviolette
1999). The distribution along the coast of Kwale-derived pottery in the early
centuries of the Christian era has been noted above (pp. 252--4). Late in
the ¬rst millennium a similar continuum is apparent for the pottery of
the so-called Tana tradition which is found from the Lamu Archipelago of
Kenya southwards to Vilanculos Bay in Mozambique, as well as on Zanzibar,
Pemba, the Comoro Islands and perhaps in northern Madagascar (Fig. 146).
Tana pottery (Horton 1987; see also Chami 1998) has af¬nities inland, but
it cannot yet be linked more precisely with any particular population or
area. Its extensive coastal distribution does, however, demonstrate a degree
of cultural unity along the Indian Ocean seaboard which was independent
of overseas traders and their activities.
Excavations at Shanga in the Lamu Archipelago (Horton 1996b) indicate
that there was probably a brief period of occupation before the introduc-
tion of imported pottery from the Arabian Gulf area provides evidence for
overseas trade, probably around the beginning of the eighth century. A very
small wooden building of this time has been interpreted as a mosque, sug-
gesting that only a few of the site™s inhabitants were Muslims. There are
good reasons to believe that conversion of local people to Islam did not
often take place until around the tenth or eleventh centuries, when the
Shanga mosque, still on the same site, was substantially larger and built
of coral. By this time overseas trade had expanded considerably and tiny
silver coins were in use, their Arabic inscriptions clearly demonstrating the
Islamic faith of the rulers.
Much of our information about the trading cities of the East African
coast comes from excavations on two off-shore islands: Manda in the Lamu
Archipelago of Kenya and Kilwa in Tanzania, some 350 kilometres to the
south of Dar es Salaam (Chittick 1974, 1984). At the more northerly set-
tlement, Manda, a massive sea-wall protected stone-built houses with plas-
tered walls. Imported pottery and glass were in frequent use, much of it
The second millennium ad in sub-Saharan Africa 289

Fig. 146:
Principal sites on
the East African
coast, and the
distribution of
pottery (after
Horton 1987)

apparently brought from the eastern shore of the Arabian Gulf, where the
port of Siraf is known to have ¬‚ourished at this time. The contemporary occu-
pation of Kilwa shows weaker foreign in¬‚uences: most of the houses were of
local wood-and-mud construction and imported goods were less numerous.
It may be concluded that settlements on the East African coast re-established
290 afric an archaeolog y

overseas trading contacts in about the eighth century. Settlements of for-
eigners were initially few, and the majority of the coast™s inhabitants were
of indigenous African stock. In later centuries these elements progressively
combined to form the modern urban Swahili culture, but it must be empha-
sised that many non-African facets of this culture appear to be comparatively
recent additions (Allen 1974, 1993). A similar picture emerges from studies
of the Swahili language which is basically, grammatically and structurally
Bantu, although with a signi¬cant number of loanwords from Arabic and
other Indian Ocean sources, many of which have been added only within
the past two centuries (Nurse and Spear 1985).
The chief items of export trade from eastern Africa were ivory, horn and
skins (Wright 1993). Slaves must also have been carried away; the evidence
comes mainly from written accounts of slave communities in lands beside
the Arabian Gulf, notably from the ninth century (Trimingham 1975). In
the south, as will be shown below, gold from the Zimbabwe mines was of
paramount importance. Beads, pottery, glass, cloth and other luxury man-
ufactures were the principal imports, together with much of the skills and
learning which contributed to the coastal culture at this time. Some of
the beads appear to have been produced at Fustat, Cairo (Saitowitz et al.
1996). Iron, which the Periplus indicates had been imported in earlier times,
was now produced locally, as is shown both by archaeological discoveries at
Manda and by contemporary Arabic writings. From the tenth-century record
of al-Mas™udi we get further information: he refers to the indigenous coastal
people as ˜Zenj™ and implies that, even in the towns, few of them were
yet Muslims. They used domestic cattle as beasts of burden, and cultivated
millet and bananas (the latter introduced from Indonesia). Ivory, obtained
inland by Zenj hunters and collectors, was brought by them to the ports and
then shipped to Oman and on to India and China. It was by similar means
that the coastal traders obtained gold from what they called ˜the land of
Sofala and the Waqwaq™ (Freeman-Grenville 1962).
During the eleventh and twelfth centuries the coastal settlements
increased in prosperity. This was particularly true in the case of towns on
the Benadir coast of the modern Somalia, notably Mogadishu (Brobeng 1995;
Dualeh 1996), and also in the south, as at Kilwa (Sutton 1998b). In the latter
place, the so-called Shirazi dynasty of rulers was established, and the town™s
importance grew rapidly; elaborate stone buildings were erected, pottery
and glass were imported, and coins were issued. The greatest prosperity of
the East African coast came in the late thirteenth and fourteenth centuries,
which saw the foundation of many new towns, such as Gedi (Fig. 147) in
Kenya, and the erection of the ¬nest stone buildings at Kilwa. In 1331 the
coast was visited by ibn Battuta, who has left a vivid eye-witness description.
By this time the rulers of Kilwa controlled the coast as far to the south as
The second millennium ad in sub-Saharan Africa 291

Fig. 147: Ruins of
a mosque at Gedi

Sofala, near the modern Beira on the coast of Mozambique. As will be shown
below, the trade in Zimbabwean gold reached its peak in the period around
ad 1400, Sofala being its main point of export (Chittick 1977).
It is now generally accepted that this coastal civilisation, which ¬‚ourished
from at least the early eighth century until the beginning of the ¬fteenth,
was ancestral to that of the Swahili (Chami 1998; C. Kusimba 1999; Horton
and Middleton 2000). As such, it is rightly seen as an essentially African,
outward-looking society, the overseas connexions of which have in the past
been stressed at the expense of those which linked it to the African hinter-
land (Abungu 1998; cf also Mutoro 1998).
Europe knew little about the civilisation of the East African coastal towns
until the Portuguese rounded the Cape of Good Hope, reaching Sofala in
1497. To guard their sea-route to India, they rapidly established forts at both
Sofala and Kilwa. The gold-rich interior then attracted their attention, and
control of the coastal trade, with many of the settlements on which that
trade depended, passed into European hands. Both the trade and the urban
settlements of the East African coast have survived into post-colonial times.

Bantu-speakers north of the Zambezi
In the interior of East Africa, the areas that had been settled by the peo-
ple responsible for the Chifumbaze complex, described in chapter 7, saw
accelerated cultural change around 1000 years ago. In some more southerly
292 afric an archaeolog y

Bantu-speaking regions, this development is fairly securely dated to about
the eleventh century, but in East Africa neither the archaeological sequence
nor the chronology is properly understood. It is commonly believed that in
much of Bantu Africa the second millennium ad was marked by increased
cattle-herding (Oliver 1982), and the contribution of Nilotic-speaking peoples
in East Africa has been emphasised. At the same time, it is important not to
underestimate the very real cultural unity of the last two thousand years in
the Bantu-speaking regions and the many lines of continuity which extend
from early in the ¬rst millennium ad into recent times (D. W. Phillipson
In much of southern Uganda, the archaeological occurrences of the sec-
ond millennium ad feature pottery decorated by means of a roulette made
of cord (Soper 1985; Desmedt 1991). The af¬nities of this pottery are with
more northerly areas, where it is particularly associated with speakers of
Nilotic languages. It is attested in Rwanda from about the ninth century,
but its period of manufacture in Uganda cannot yet be shown to have begun
at such an early date (Sinclair 1991). It has been widely believed that this
development was in some way linked with the establishment of the major
kingdoms of the interlacustrine region, including those of Buganda, Bun-
yoro and Ankole (Robertshaw 1994). Although Bantu-speaking, at least in
more recent times, these kingdoms preserve traditions which attribute their
foundation to a group called the Bachwezi, who appear to have been cattle-
herders, perhaps from Nilotic-speaking areas to the north.
These traditions need not be taken at face value, and there is much con-
troversy concerning the date, identity and activities of the Bachwezi. If they
were indeed a distinct group, the most likely interpretation of the avail-
able evidence is that they were the alien founders of ruling dynasties who
came to dominate sections of a pre-existing Bantu-speaking population. It
should be noted that control of long-distance trade was apparently not the
major stimulus for the formation of the interlacustrine states (Oliver 1977;
Connah 2001). On the contrary, local resources -- notably herds of cattle --
provided their economic basis, as is emphasised by some of their rulers™
regalia (Sassoon 1983). While the populations of these states made and
used roulette-decorated pottery, there is no reason to link the two develop-
ments. The widespread appearance of this pottery in Uganda should occa-
sion no surprise in view of the readiness with which it has been adopted
by Bantu-speaking peoples in more easterly and southerly parts of East
Africa. Although this process is still continuing today, it is one for which
the reasons are not at all clear, and it provides an excellent example of
a major change in artefact style that is not accompanied by any signi¬-
cant shift in population or, indeed, by any apparent economic or practical
The second millennium ad in sub-Saharan Africa 293

Fig. 148: Salt
crystallising on
the surface of a
Kibiro, 1990

The archaeology of the interlacustrine kingdoms has been investigated
at the capital sites of Bigo, Bweyorere and Ntusi (Posnansky 1968, 1969;
Reid and Robertshaw 1987; Reid 1996). Extensive dams and earthworks --
those at Bigo total more than 10 kilometres in length -- indicate the scale of
organisation that was achieved. Remains of large circular houses resemble
the royal residences of more recent times. Salt was evidently an important
commodity at this time in East Africa, as elsewhere, and major workings
have been investigated at Kibiro (Fig. 148) on Lake Albert and at Uvinza
in western Tanzania, as well as at Ivuna further south (Hiernaux and
Maquet 1968; Fagan and Yellen 1968; Sutton and Roberts 1968; Connah
Elsewhere in East Africa our knowledge of this period is even less compre-
hensive. In parts of the eastern Kenya highlands pottery derived from the
Kwale tradition may have continued to be made as late as the thirteenth
or fourteenth centuries, before being replaced by a style, still practised by
the modern Kamba and Gikuyu, which may have originated in the Chyulu
area of southern Kenya. In both Kenya and Tanzania there is evidence, early
in the present millennium, for peasant farmers owning domestic cattle and
sheep/goats and cultivating sorghum. Except in the immediate hinterland,
glass beads and other items imported from the coast remained rare until
recent times (Odner 1971; Soper 1976, 1979).
294 afric an archaeolog y

Fig. 149:
tradition pottery
(after D. W.
Phillipson 1977a)

A similar situation appears to have prevailed in most areas of the modern
Zambia and Malawi. Here, however, as in a wide area further to the south,
there was a series of pronounced and apparently sudden changes in pottery
styles around the eleventh century ad. The main exception to this is in the
Upper Zambezi region of western Zambia, which seems to belong to a west-
ern culture area, extending into Angola, where there is evidence for greater
continuity in pottery style from the ¬rst millennium ad into more recent
times (D. W. Phillipson 1974). In eastern and much of northern and cen-
tral Zambia the main pottery style of the second millennium ad has been
called the Luangwa tradition (Fig. 149). Its appearance around the eleventh
century represents a sharp break with its predecessors of both the eastern
and the western facies of the Chifumbaze complex, and it has continued
with relatively little modi¬cation ever since. The contrast is well seen at
the site known as Twickenham Road in Lusaka (D. W. Phillipson 1970). The
antecedents of the Luangwa-tradition pottery remain unknown but may lie
in the direction of Katanga in southeastern D. R. Congo (cf. Huffman 1989).
Today, Luangwa-tradition pottery is traditionally made by women. In more
westerly areas, vessels of the Lungwebungu tradition, apparently derived
from the Chifumbaze complex, are made by men. This suggests that the
contrast between the pottery of the Chifumbaze complex and that of the
Luangwa tradition may have been due to the establishment of communi-
ties amongst whom potting was undertaken by women. Luangwa-tradition
pottery is today made by people of many different societies, including the
The second millennium ad in sub-Saharan Africa 295

Bemba, Chewa and Nsenga. It was clearly established long before these soci-
eties became differentiated from each other following the arrival of chie¬‚y
dynasties (which claim an origin in what is now D. R. Congo) in about the
¬fteenth and sixteenth centuries (D. W. Phillipson 1974).
By the fourteenth/¬fteenth century copper mining, which had begun on
a small scale about one thousand years earlier, became much intensi¬ed,
resulting in enormous workings such as those at Kipushi, Bwana Mkubwa
and Kansanshi in the Zambia/Katanga Copperbelt area. Cross-shaped ingots
were cast in closely standardised sizes and widely traded; they probably
served as a form of currency (Bisson 1975, 2000; de Maret 1981). It is prob-
ably not coincidence that the development of copper mining and trading
is indicated at the same general time as the local rise of centralised states,
and that the chie¬‚y dynasties of surrounding areas trace their origin to
southeastern D. R. Congo, which saw the greatest development of the cop-
per trade. As shown in chapter 7, the local use of metal as an embodiment of
wealth may be traced back to the ¬rst millennium ad. A further feature of
later metal-working in this area was the tall natural-draught iron-smelting
furnace ¬red without the use of bellows (D. W. Phillipson 1968b; Sutton
The southern Congo Basin region appears to have contributed signi¬cantly
to the cultural development of a very large part of central Africa during the
second millennium ad. It is thus particularly unfortunate that its archae-
ology for this period remains virtually unknown (Ervedosa 1980). A con-
siderable amount of research has been undertaken on the oral traditions
relating to the kingdoms which ¬‚ourished in this area, especially along
the southern and eastern margins of the forest. Further evidence has been
obtained through historical linguistic studies, but it is not easy to correlate
these results with those obtained by archaeologists in neighbouring areas
(Vansina 1966, 1984, 1990).
Archaeologically, the only well-documented late sequence in the whole
of this vast area is that based on excavations around Lake Kisale near the
southeastern corner of the forest (de Maret 1977, 1985b, 1992). The great
cemetery of Sanga was noted in chapter 7: although the sequence began
late in the ¬rst millennium, its main period of use is dated by radiocarbon
to the eleventh and twelfth centuries. Already by this Classic Kisalian period
considerable metal-based wealth had been accumulated; ¬‚ange-welded iron
gongs, of a type which -- at least in later times -- served as symbols of kingship
(Vansina 1969), were found in some of the richer graves (Fig. 150). Later use
of the cemetery, attributed to the Kabambian phase, continued at least into
the sixteenth century. There is no good reason for regarding the Kabambian
as other than a descendant of the Kisalian, although there are important
differences between them, most pertinently in the much greater frequency
296 afric an archaeolog y

Fig. 150: A
Kisalian grave at

of copper cross-ingots in the Kabambian graves. The most recent graves at
Sanga are attributed to the Luba and probably date within the last two
Three important points emerge from this discussion. One is that, as sug-
gested above, there was a greater degree of continuity from the Chifumbaze
complex into later times in some more westerly areas than there was fur-
ther east. The second is that certain cultural items which became prevalent
in eastern areas during the second millennium ad may have had a greater
antiquity in the west. This view is supported by oral traditions which, as we
have seen, derive the ruling dynasties of many states in Zambia, Malawi and
adjacent regions from a Congo Basin origin. One interpretation of these tra-
ditions places the rise of the savanna kingdoms of southern D. R. Congo and
northern Angola at least as far back as the thirteenth century (J. C. Miller
1976), and this is not contradicted by the scanty archaeological evidence.
Lastly, although the region clearly saw the interplay of several cultural tra-
ditions, and although its natural resources and economic potential were
diverse, particular and widespread emphasis appears to have been placed
on the working of metal and its use not only for everyday utilitarian objects


. 8
( 11)