Implications of new AMS dates for the Khami Period in the Mapungubwe Landscape

FUNDING: University of the Witwatersrand Research Incentive Scheme (HUFF013) After the abandonment of Mapungubwe, the Limpopo Valley was reoccupied first by Sotho people, making Icon pottery, and then by Kalanga speakers making Khami pottery. The senior Kalanga chief, in this case Twamamba, was based at Machemma about 60 km to the south, while several petty chiefs administered various portions of the valley itself. Because of fluctuating rainfall, the occupations of both Sotho and Kalanga people occurred in pulses during higher rainfall periods. New AMS dates place one site in the Icon Period, eight sites in Pulse 1 (AD 1400–1480) and eight sites or components in Pulse 2 (AD 1520–1590). Kalanga people occupied the best agricultural land near the Limpopo floodplains and Sotho people lived on the plateau to the south. The two groups thus shared the landscape, but not the resources equally. The ceramic record documents this unequal interaction. This interaction, facilitated by male and female initiation schools on the ethnic boundary, helped to create Venda as a language and macro-cultural entity.


•
Initiation schools in the Limpopo Valley provide a model for interaction in the rest of Venda.
By AD 1320, Mapungubwe people had abandoned the Limpopo Valley and the landscape remained unoccupied by agriculturalists for some 80 years until early Sotho people took up residence. 1,2 With a probable origin in East Africa 3 , Icon pottery marks the first appearance of Sotho speakers in southern Africa. Shortly afterwards, Kalanga speakers (Western Shona) who made Khami pottery moved south from Zimbabwe and reoccupied the valley: it was their ancestors who had lived at Mapungubwe.

The Khami Period
Khami chiefdoms can be ranked according to a hierarchy in which each chiefly level was the apex of a pyramid of lower courts. 4 Between AD 1400 and 1450, Khami 5 near Bulawayo ( Figure 1) became the large capital of an independent Torwa state in an area known as Butua 6 . States with such large capitals encompassed some 90 000 km 2 and multiple smaller chiefdoms. In the early 15th century, Khami expanded its control into Botswana 7 , southern Zimbabwe 8 and across the Limpopo. The senior chief for the Limpopo settlements was based at Machemma 9 , about 60 km to the south of the Limpopo. A paramount chief at this time did not exist in South Africa nor in Botswana, and so, the provincial capital must have been in Zimbabwe.
Because Shona and Venda people use natural features to mark political boundaries 10 , the Shashi and Limpopo rivers most likely separated the southwest (Botswana) and southern (South Africa) districts from those in Zimbabwe. Interaction between early Sotho and Kalanga appears to have occurred only in the southern district and further southeast. Kalanga people withdrew during the dry period around AD 1500 and then returned to the Limpopo, and to Machemma, when rainfall improved. 10

Sotho and Kalanga interaction in the Mapungubwe Landscape
Interaction between Sotho and Kalanga peoples is of special interest because it led to the creation of Venda as a language and macro-cultural entity. Venda people are archaeologically and anthropologically important because they have continued the essence of the pre-colonial Zimbabwe Culture -class distinction and sacred leadershipdistinguishing them further from Sotho-and Nguni-speaking communities. 11 The Tshivenda language is known to be a unique blend of Shona (both Kalanga and Karanga) and Sotho. 12 Previous research near the Soutpansberg shows that the creation of Tshivenda parallels the creation of Letaba (Venda) ceramics from a blend of Icon (Sotho) and Khami (Kalanga) pottery. 13 It is not possible to use ceramic data alone to identify which, if any, present-day Sotho and Kalanga dialects were spoken 600 years ago. Broad correspondences between Icon and Sotho, and Khami and Kalanga, are nevertheless sufficient for our Limpopo research.

New dates
We have recently processed dung samples from 16 new sites or components related to Venda origins. They were collected from below the surface level inside central cattle kraals: eight in Pulse 1 and eight in Pulse 2 (Table 1). We use the southern hemisphere data set SHCal20 at one sigma accuracy, following Calib 8.10 (available online at http://calib. org/calib/calib.html). We use one sigma errors rather than two because each dung sample derives from a short, discrete dating event limited by our tight ceramic and baobab sequences. Our sequence confirms that recorded by Loubser 13

Initiation schools
Loubser's 13 research suggested that the Soutpansberg was the boundary between Shona speakers to the north and Sotho speakers to the south. Our survey shows that the initial interaction took place in more confined localities, such as the Mapungubwe Landscape. The channel for this interaction was probably through initiation schools. Such schools as domba inculcate various aspects of culture, including history, worldview and proper moral behaviour [15][16][17][18] , and both royal and commoner youths attend. It is therefore significant that the remains of these schools stretch along the boundary between Khami and Icon settlements (e.g. Faure AD 2 and Kilsyth AD 268) and nowhere else in the Mapungubwe area. They are not located in the centre as researchers once thought. 10 The new dates show that these schools date to both Pulse 1 and Pulse 2, when Icon changed to Tavhatshena. In addition, male circumcision sites occur on the eastern boundary associated with the Khami-Period palace called Haddon. 9,19 Both types of schools were therefore present in the valley at the same time, and over the years, hundreds of youth would have attended both. These schools provide a model for interpreting Tshivenda origins in the rest of Venda.

End of Khami
In AD 1644, the Khami capital was destroyed by a combined force of Portuguese and a disgruntled faction of the Torwa state 20 , and for the next 40 years, no single leader controlled the former kingdom. A few chiefs controlled some areas in Botswana 21 , but none appears to have been based in in the Limpopo Valley, perhaps because this Interregnum Phase was too dry. The third step in the evolution of Tshivenda took place at this time, when Butua lacked strong leadership.  (Figure 3). 13 The Tshivhula claim that their territory once extended from Machemma to the Limpopo. 26 Because each chiefdom level is two to three times the size of the level below, Machemma would have controlled about 10 000 km 2 , corresponding to the area claimed by the Tshivhula. It is therefore reasonable to conclude that the Khami Period in the Mapungubwe Landscape marks the arrival of the Tshivhula dynasty, or at least the Twamamba. 10 Originally, they would have spoken Kalanga, but by the 17th century, they had become Venda and independent of the Khami state.