When science alone is not enough: Radiocarbon timescales,history,ethnography and elite settlements in southern Africa

Introduction: A question of different scales

In his review of the politics of African archaeology, Shepherd (2002) observed that, when compared to the West, African archaeology neither sets nor directs research agendas. Instead, it follows the directions in western and global archaeology. Ironically, the African historical and anthropological record has been a major source of analogies for making interpretations in global archaeology (Hodder, 1982). According to Shepherd (2002), this contradiction is an artefact of the origins of archaeology as a discipline in eighteenth- and nineteenth-century Europe. With the success of the colonial project, the methodologies and thinking frameworks of the discipline were exported to the new colonies (Robertshaw, 1990). Today, despite most of Africa being politically independent, African archaeologists have broadly failed to tap into indigenous practices to gestate an African archaeology inspired by local philosophy and knowledge (see Depelchin, 2005).

This article dips into indigenous African philosophies and the latest developments in the application of Bayesian modelling (Bronk Ramsey, 2009) to develop an alternative way of looking at settlement and political archaeology in the Iron Age of southern Africa. The article focuses on indigenous conceptions of leadership which are primarily based on the concept of rotational succession (collateral or by primogeniture) (Beach, 1994; Holleman, 1952; Sinclair et al., 1993; Weinrich, 1971). There exists a rich body of Shona proverbs and idioms which emphasise the fact that ‘leadership is all about rotation’ – or ‘ushe madzoro hunoravanwa’ in Shona (Hamutyinei, 1987; Pongweni, 1989). More importantly for archaeology, when a new ruler ascended to the throne, he ruled from his pre-existing settlement and did not move into that of the predecessor, which was often located in a separate district (Beach, 1994; Mudenge, 1988; Pikirayi, 1993). Therefore, succession politics resulted in the creation of numerous elite centres with different occupational durations (Sinclair et al., 1993).

This contradicts the dominant ways of reading political meaning from southern African elite settlements, which often assume that the largest settlements were the capitals, with the smaller ones forming nodes of decreasing importance from provincial to district and local centres (see Huffman, 2000, 2007; Sinclair, 1987). For example, because of its size, Great Zimbabwe is depicted as the capital of a very big state whose influence extended from the Kalahari to the Indian Ocean. The consequence of this for our understanding of the past is that if one big centre collapses, it is replaced by another one – Mapungubwe is replaced by Great Zimbabwe and Khami, respectively. The principle of rotational succession advanced here indicates that centres of power shifted from time to time, changing the status of settlements (Beach, 1994). Thus size alone is not adequate to explain spatial and temporal relationships between the many elite settlements that are found in southern Africa. Drawing from primary and secondary sources and Bayesian analyses of radiocarbon dates, this model of rotational leadership is applied initially to the historical and anthropological records before being tested on the archaeological record.

Archaeology relies today on chronometric methods for establishing sequences, and particularly on radiocarbon dating (Buck, 2001). Since its discovery, a number of problems with radiocarbon dating, chief among them calibration and contamination, were noted and resolved (Bronk Ramsey, 2009). By the 1980s in southern Africa, radiocarbon dating had played an instrumental role in developing the chronology of the 2000-year-long Iron Age period (Huffman, 2007; Mitchell, 2002; Phillipson, 2003; Vogel, 1999). The radiocarbon timescale made it possible to explore correlations between radiocarbon dates and changing settlement size to reconstruct political hierarchies in the region (Chirikure and Pikirayi, 2008; Huffman, 1986).

The main challenge for archaeology is that, despite having the techniques of absolute dating in its arsenal, it deals with missing, fragmentary, layered and dispersed evidence. Archaeological sites are palimpsests of events that happened at different points in humanity’s long history but with short- and long-term durations (Bailey, 2007; Lucas, 2005). However, although advances were made in producing very tight chronologies using Bayesian modelling (Buck, 2001), radiocarbon dating still provides aggregate chronologies which bring us up against abrupt events that cannot be placed within timescales deriving from absolute dating methods (Gosden and Kirsanow, 2006: 36). Attempts are being made to synchronise short-term with long-term events using volcanic tephra in places such as Europe (Blaauw et al., 2007). Yet, in southern Africa as elsewhere, the tendency by archaeologists so far has been to place too much faith on absolute methods of dating to comprehend issues such as settlement hierarchy, political organisation, intra-settlement sequencing and settlement succession (Huffman, 2007; Pikirayi and Chirikure, 2011).

Because radiocarbon dates are measures of, if not exercises in, statistical probability, this over-reliance on them presents a number of challenges when reading political and social meanings from sites. For instance, an uncalibrated radiocarbon date with a plus or minus 50 years error limit quoted at 1 sigma (65%) when calibrated, and depending on the shape of the calibration curve, can translate into 200 years for the total error range. Given the short duration of historical events, the probability is equally high that we will aggregate events that took place at different times in this long period. Expressed differently, we inevitably conflate or ‘suck’ events belonging to small time segments and ‘smear’ them within the broad radiocarbon timescale, thereby promoting unrealistic reconstructions of the past (Baillie, 1991). Advances in the use of Bayesian statistics have resulted in the development of chronological models that incorporate as much prior information as possible to establish fine sequences but the overall resolution is still very coarse on a ‘human generational time scale’ (Buck, 2001). Equally, alternative dating methods such as optically stimulated luminescence provide dates with error terms that will typically cover the duration of many rulers (Ohinata, 2001; Rosenstein, 2009). This article seeks to use data from history and anthropology to develop possible ways of understanding elite settlements in southern Africa. The ultimate objective is to show that paying due regard to short duration events and their material expression to calibrate the time scales from absolute methods results in ways of studying the past that draw from the African experience.

Definitions and theory: Elite settlements, time and radiocarbon dating

The development of socio-political complexity in southern Africa during the turn of the first millennium AD resulted in the creation of archaeological sites associated with power, influence and dominance (Chirikure, 2007; Sinclair et al., 1993). In stratified societies, elite settlements were occupied by high-ranking and influential members. In general terms, archaeologists have used a number of Childean traits to define and characterise elite settlements (Childe, 1950, 1957). Following on that, elite settlements are often assumed to be the largest in their region and invariably contain prestige goods in the form of imports and other locally rare commodities (Garlake, 1973). Furthermore, elite centres are believed to participate in trade with the hinterland and have evidence of craft specialisation (McGlade, 1997, 1999). In some cases, elite settlements contain monumental architecture, both impressive and unimpressive.

Southern African elite settlements where one can tick boxes for some of these traits include first millennium AD sites such as Kadzi (Pwiti, 1996), Schroda (Huffman, 2009) and Chitope, as well as second millennium AD sites like Mapungubwe (cal AD1220–1290), Great Zimbabwe (cal AD1300–1450) (Huffman and Vogel, 1991) and Khami (cal AD1450–1650) (Huffman, 2009). According to Sinclair et al. (1993: 709), ‘elite settlements should be considered essentially as courts of paramounts and local chiefs which were occupied for single reigns within the context of a larger social and political system’. Historically, the capitals of the Venda state such as Dzata; the centres of the Ndebele state such as Old Bulawayo (Ko Bulawayo); KwaDukuza, the capital of the Zulu state under Shaka; and Old Palapye, the seat of Khama’s power, are all examples of elite settlements (Lane, 2004; Shillington, 1985). A reading of political history and anthropology intimates that these historic capitals were associated with specific and short events such as the rule of individual leaders. In short spaces of time, succession politics dictated that each of these leaders could, depending on circumstances, become the leader who had to establish a capital and supporting officials at national, provincial, district and ward levels (Beach, 1980, 1994: 40; Mudenge, 1988). Thus changes in kings or chiefs often resulted in changes in the supporting officials, although previous centres of power may have continued to be occupied. It is important to make a distinction between overall occupation of the elite sites and periods when they were capitals or centres of power. In some cases, elite sites were already being occupied before power shifted to them and may have continued to be occupied after power rotated to other centres. Because elite settlements occupy an important place as an arena for social innovations, it is important to understand them chronologically and spatially. Clearly, it is on the chronology that radiometric dating techniques have been useful in building interpretations (McGlade, 1999; Murray, 1999).

Because time is the universe in which all archaeological processes are embedded, it impacts on archaeological theory and practice. However, there is no unitary view of what time means to archaeologists. Several archaeologists have discussed the issue of time from a number of perspectives such as the palimpsest view, time perspectivism and the longue durée (Bailey, 2007; Gosden, 1994; Gosden and Kirsanow, 2006; Lucas, 2005). A single thread, however, runs through these different viewpoints – the archaeological landscape is layered with both short- and long-term events. As such, a consideration of time by archaeologists promotes a fruitful engagement with linear and multi-temporality which lies at the nexus of the relationships between archaeology, history and anthropology (Morley and Renfrew, 2009).

Time is a system of measurement used to sequence events and to compare their durations and intervals between them (Morley and Renfrew, 2009). The sequences are used to model changes in the archaeological record such that the position of past events with respect to an ever-changing present continually changes, creating what Gosden and Kirsanow referred to as the longue durée. The concept of time underlies how we practise our discipline, but an uncritical use of chronology results in inherently singular and linear interpretations of archaeological change (Lucas, 2005). By not allowing for simultaneity we may simplify the past but ultimately limit the possibility of understanding historical processes operating over different scales and durations. Archaeological landscapes represent complex and massive palimpsests of events from many periods (Bailey, 2007). This conception of time therefore challenges archaeologists to rethink social and political landscapes and how they are materialised on the physical landscape. Although archaeologists cannot operate outside chronology, a more complex approach that takes account of possible temporal disruptions and dislocations is the best approach (Lucas, 2005).

Chronometric dating techniques such as radiocarbon dating have continually been improved to take care of both long-term and short-term cultural processes. The combined use of Bayesian statics and calibration is one of the innovations made in making radiocarbon dating relevant to all ends of the time spectrum. The application of Bayesian techniques results in higher precision and, depending on the needs of the user, can accommodate long and short duration events. These improvements, however, mostly apply to new dates from stratified contexts and less to old dates from multiple contexts. Furthermore, despite their increased precision, Bayesian dates still produce a range which eclipses individual life spans and the duration of the reign of many leaders. In any case, where the dated organic material was charcoal, the resultant chronology is that of when the tree died, which may be at variance with the event itself. As a result, the use of the radiocarbon timescale often results in the nesting and conflation of events with different temporal scales and durations. This becomes a point of discomfort if one considers that radiocarbon dating is the pillar of chronology in the Iron Age in southern Africa and that no chronometric hygiene has yet been conducted on the old 14C dates used in the region. Effectively, this means that the radiocarbon scale in its current state is much broader than real life events, making it insensitive to short duration events such as variability in individual lifespan within and between societies (Gosden and Kirsanow, 2006). So far southern African archaeologists have made little attempt to grapple with this challenge.

It seems, therefore, that southern African archaeologists have not taken note of Harris’s (2006) cautionary statement that ‘in taking time scale for granted, archaeologists rarely expose its complexities and therefore overlook its crucial role in the process of representing the past’. In losing sight of timescale as a fundamental concept, archaeologists in southern Africa have exposed themselves to risks associated with falsely assigning the beginning of important innovations to sites that may be only a manifestation of innovations that started elsewhere (Shennan, 2001). This makes more sense if one considers that while innovations are short-term events in the historical and anthropological record, their effects invariably outlive the time involved in creating them. In an attempt to re-orient social and political archaeology in southern Africa, this article uses data from history and anthropology and their emphasis on short duration events to develop alternative meanings out of the broad temporal scale deriving from absolute dating methods.

History, ethnography and temporal scales: Case studies of recent elite settlements in southern Africa

By the dawn of the nineteenth century, the southern African socio-cultural landscape already contained numerous elite settlements associated with the rulers of separate chiefdom and state systems (Beach, 1980; Lane, 2004; Shillington, 1985). More importantly, radiocarbon dates are virtually useless post AD1650 because of the flatness of the calibration curve. Because the nineteenth century falls within the ambit of history and ethnography, it affords us an opportunity to trace the development of elite settlements within their chronological and spatial framework (Abraham, 1961; Sinclair et al., 1993). Perhaps, the best case study for this is the widely studied Mutapa state and the associated glut of capitals. The historical Mutapa state, which was based on the northern parts of the Zimbabwe plateau and adjacent lowlands, provides a strong example of the short-term duration of both leaders and capitals in the southern African past (Beach, 1980, 1994). Oral traditions claim that Nyatsimba Mutota, the last ruler at Great Zimbabwe, was the first Mutapa king (Mudenge, 1988). The same oral traditions have identified the capital of this ruler on the edge of the Zambezi escarpment at a Zimbabwe type enclosure known as Mutota’s Zimbabwe (Pwiti, 1996). According to Beach (1976), the Mutapa state is of special interest because it was close to Portuguese centres where information was recorded. However, before the 1760s the Portuguese on the Zambezi were unintellectual and demonstrated very little interest in Mutapa history (Beach, 1994: 213). It was only in 1763 that Mello a Castro compiled a line of Mutapa rulers.

It is important to first outline the Mutapa system of succession as enshrined in historical sources such as the widely respected accounts by Mudenge (1988) and Beach (1980, 1994), as well as anthropological observations by Holleman (1952) and Weinrich (1971). According to Beach (1994: 224), from the state’s establishment, the Mutapa succession system followed both the system of primogeniture and collateral succession. Beach (1994: 224) states that:

Basically, we have two conflicting tendencies: firstly, the tendency of fathers to pass on power to their sons and secondly the tendency of brothers to seek the title when the ruler dies. This can lead to a third tendency, of uneasy rotation of power between houses, which does not stop the first process from going on.

Huffman (2010) has wrongly accused Chirikure and Pikirayi (2008) of using the inappropriate principle of collateral succession to much older periods such as the beginning of the Mutapa state. He is convinced that Portuguese documents such as those of Dos Santos show that succession was supposed to pass from father to son, as long as the mother was a royal wife, and that collateral succession only developed in the late seventeenth century. Huffman cites Beach’s largely dissonant re-assessment of Mutapa historical sources to conclude that the first seven kings succeeded their fathers.

A re-reading of Chapter 6 in Beach (1994) and other sources such as Mudenge (1988) and documents held at the National Archives of Zimbabwe shows several discordant points in Huffman’s interpretation of historical records (see Beach, 1998, for criticism of Huffman’s unhistorical use of historical data). Firstly, Huffman does not acknowledge that there is a great deal of dissonance in the oral and written information relating to the Mutapa state. Beach’s account privileges written records over oral traditions which show power struggles between royal houses in the fifteenth and sixteenth centuries (see Mudenge, 1988). Secondly, a re-reading of Beach (1994) and Mazikana and Johnstone (1984) suggests that between 1490 and 1494, Changamire ruled as the Mutapa after assassinating the previous incumbent. The son of the slain Mutapa managed to wrestle power back in 1494 and ruled until about 1530, when the succession moved to the Karembera house (Figure 1). Thus the third Mutapa ruler was not a son of the founder of the state but a usurper. From then on succession alternated between the Chikuyo and Karembera houses and was collateral in nature. Thirdly, the reign of the first Mutapa is given variously as c.1440–50 and if one counts seven leaders from him, excluding the usurper Changamire, only the reign of Gatsi Rusere (c.1589–1623) overlaps into the seventeenth century. So, collateral succession could not have started in the seventeenth century; it was in existence much earlier together with succession by primogeniture. OPEN IN VIEWER

Admittedly, historians themselves are not agreed on the Mutapa king lists and even Beach (1994) says that his version is open to revision. Since neither Huffman nor ourselves are historians, it is important to elicit some essential principles from historic sources. The first principle is that the Mutapa succession was based on both collateral and primogeniture systems (Beach, 1994: 224; Mazikana and Johnstone, 1984). The second is that all succession systems (collateral or primogeniture) are based on rotation. The third is that a ruler did not rule from his predecessor’s residence. Instead, he ruled from his own capital (see also Sinclair et al., 1993: 709). Stretching from the mid-fifteenth to late nineteenth century, the Mutapa state is associated with multiple capitals, some of which are known though others remain unknown (Figure 2). The founder Nyatsimba Mutota established his capital at Mutota’s Zimbabwe (Mudenge, 1988; Pwiti, 1996) and was succeeded by his son Matope. All the succeeding Mutapas were descendants of Nyatsimba Mutota, with his many sons forming houses where future kings were derived following the system of rotational succession (Beach, 1994; Pikirayi, 1993; Sinclair et al., 1993). When the time for political succession came, new leaders ruled from their existing homesteads which they simply expanded to accommodate their new status. Not surprisingly, most of the elite dry stone walled sites are identified with former rulers such as Mutota, Chiwawa, Nowedza and Kasekete, among others (Pikirayi, 1993; Pwiti, 1996). It is the nature of this succession politics that therefore led to shifting capitals and the presence of many elite centres in northern Zimbabwe. If we are to consider the king lists from Abraham (1961), Mazikana and Johnstone (1984), Mudenge (1988) and Beach (1994), we realise that from 1450 to 1900, the Mutapa state had approximately 32 rulers, giving an average rule of 14 years. In practice, however, not every ruler ruled for 14 years as some rulers ruled for longer periods while succession disputes ended the rules of others prematurely. Clearly, these are all short duration events which can easily be aggregated into one time bracket using radiocarbon timescales. OPEN IN VIEWER

While skeptics may point to the possible Portuguese influence in the Mutapa area, the basic principles of shifting capitals and rotational succession are even clearer in the Buhera dynasty of south-central Zimbabwe. Being in the interior, the Buhera dynasty did not have much contact with the Portuguese. Because Buhera experienced no land alienation during colonial times, it provides a rare example of a dynasty ruling in the same area from the last period of dry stone wall construction to the present.

According to Beach (1994: 40), the history of the Mbiru-Nyashanu dynasty of Buhera is interesting in that it shows how politics may have worked during the period of Great Zimbabwe. The Hera rulers of Buhera, like their Zimbabwe predecessors before them, seem to have used different parts of the territory as their capitals. For example, Mbiru used the dry stone walled settlement of Gombe, while his successor Rukoto used the Bedza area; Mutekedza used Dukuta, while Mazazha returned to the Gombe area. This shows that the issue of shifting capitals had deeper roots in the Shona culture.

The Venda people also demonstrate this concept of shifting capitals. According to Hanisch (2008: 126–127), a number of dry stone walled capitals on the Soutpansberg range of mountains are associated with different leaders. For example, the eastern and western stone walled settlements at Tsindiendeulu are associated with different leaders. According to oral traditions, settlement was then moved to Dzata Mikondeni, where Thohoyandou built his capital before leaving it in the wake of succession disputes.

In a context different from political succession, Lane (2004) has demonstrated how the capitals of the Tswana under Khama III shifted from Shoshong to Old Palapye and to Serowe, all within a very short space of time. According to Lane (2004: 289), ‘just as the capital shifted, so did that of the outlying lands and cattle posts. The cumulative effect over time of this kind of re-arrangement was the creation of a palimpsest of overlapping categories of spaces’. If a critical perspective is not applied in the field, all these places may be treated as contemporary, producing a misleading chain of capitals and provinces. During his reign, Shaka shifted his capitals twice in a space of just under 12 years (Mountain, 1999). His successors built their own capitals and avoided those of their predecessors. Each king had his own provincial and district centres staffed with his loyalists. The archaeological implication of this is that it is difficult to identify provincial capitals associated with individual kings since they changed from time to time.

The archaeology considered: Great Zimbabwe and Mapungubwe

The short-term temporal scale from history and anthropology has clearly demonstrated that succession events often take place at a very fast pace, one which cannot be easily accessed using absolute methods of dating. Based on this information, it is important to consider the archaeology of the elite Zimbabwe-culture settlements, whose signature material remains are dry stone walls (Garlake, 1973; Pikirayi, 2001). This tradition of building elite residences is widely believed to have originated at Mapungubwe near the junction of the Shashi and Limpopo rivers before it was succeeded by Great Zimbabwe. According to Huffman and Vogel (1991), Great Zimbabwe flourished for 150 years until it was eclipsed by the Mutapa state in the north and by Khami in the south west. The tendency in the study of Iron Age settlements in southern Africa has been to identify the largest sites as capitals (Pwiti, 1996). These big centres are assumed to have presided over provincial, district and ward centres. While this is not contested, it has been historically attested that succession politics generated new provincial, district and ward centres such that these were not fixed at any given point in time. In other words, the biography or life history of settlements, like their sizes, changed over time. As archaeologists we tend to assume, however, that both capitals and subordinate centres were unchanging for the duration of the different states. Not surprisingly, Great Zimbabwe was the only capital of the Great Zimbabwe state, just as Mapungubwe was the only capital of the Mapungubwe state. Furthermore, the provincial and district centres of these states were unchanging too! What is difficult to understand is the existence of so many elite stone walled sites contemporary with Great Zimbabwe itself. There are more than 150 such sites located both near Great Zimbabwe and in regions afar (Figure 3). Based on our historically derived model, it is possible that the status of these places changed with time, both horizontally and vertically. OPEN IN VIEWER

In most cases, these elite centres appear as clusters. Great Zimbabwe itself is strewn on a 720-hectare estate which covers prominent areas such as the Hill Complex, the Valley Enclosures and the Plateau Enclosures, along with surrounding smaller walled and unwalled areas. Settlement at Great Zimbabwe lasted a considerable period of time, with the power shifting from different areas as shown by the differential durations of occupation in various enclosures (Collett et al., 1992). The main issue, however, is what was the status of smaller elite stone walled sites scattered all over Zimbabwe and adjacent areas in relation to Great Zimbabwe? Are they provincial, district and ward centres, or are they something else?

This invites us to consider the chronology of such enclosures in comparison to that of Great Zimbabwe. We gathered a series of uncalibrated radiocarbon dates for Great Zimbabwe and related settlements. These were calibrated using the Bayesian model in the program OxCal version 4 (Bronk Ramsey, 1995) and using the curve for the southern hemisphere (McCormac et al., 2004). Initially, the model was constructed by declaring each site to be a phase (i.e. making no assumptions about the chronological sequence between sites). Thereafter, another model was designed to test the validity of the dominant hypothesis, assuming that Mapungubwe gave rise to Great Zimbabwe, which in turn gave rise to Khami. The model also factored in potential overlaps and occupational longevity. The results of the modelling demonstrate that while some of the settlements are contemporary with the formative stages and florescence of Mapungubwe or Great Zimbabwe, others were established at later periods (see Appendix 1). There are some such as Tsindi which, while fitting within the chronology of Great Zimbabwe’s rise, also outlived it. The sites such as Nhunguza, the two Zvongombes and Ruanga in northern Zimbabwe, and Thulamela in South Africa, clearly emerged after Great Zimbabwe. The distributions of the dates demonstrate that the Hilltop occupation at Mapungubwe associated with the dates AD1220–1290 overlaps with places such as Mapela as well as the formative stages of Great Zimbabwe and Tsindi (see Appendix 1). The Bayesian model further indicates that the periods of occupation of Leopard’s Kopje, K2, Mapungubwe, Mapela and Taba zika Mambo overlap. This shows that the establishment of these places was coeval. More importantly, Harleigh Farm, Great Zimbabwe, Mapungubwe and Tsindi also overlap. It would seem that Gombe, Tsindi and Harleigh Farm are centres of independent polities together with Mapela and Mapungubwe (Beach, 1984). Khami settlements seem to have emerged in the mid-thirteenth century, although Taba zika Mambo had been occupied since the Leopard’s Kopje times. This picture from the Bayesian study seems to indicate that the development of the Zimbabwe culture was multi-linear, with the various centres following different trajectories.

Discussion: Reconciling the short- and long-term timescales

The historical and anthropological record has provided examples that explain the presence of many elite centres, some of which succeeded each other in quick succession. These examples are potent ways of animating the archaeological record. The most common strand running through historical examples of elite centres is that they had a short duration and often leaders shifted their capitals for political and strategic considerations. The Mutapa, Tswana and Zulu elite centre examples provide a test case for archaeology which relies on traits such as size of settlements to read political meaning and create hierarchies. The fact that Shaka’s second Bulawayo is by far the largest Zulu capital would imply that if the Zulu state was an archaeological entity, then all the other capitals would be provincial centres by virtue of their sizes. This has resonance in the Tswana towns of the Bankenveld which were massive agglomerations of people and which appeared in different sizes (see Hall et al., 2008). Marothodi, Molokwane, Kaditshwene and others are all nineteenth-century Tswana towns of different sizes. If archaeologists were to use size as an indicator then Kaditshwene would be the capital because it is the largest. Similarly, Lane (2004) has observed that the capitals of Khama III shifted during his reign. Clearly, radiocarbon chronology would place these events within one standard deviation. Changes in material culture, for example differences in glass beads, can be useful in differentiating sites chronologically, but this is not always the case since some sites are long occupied (before they became capitals) and contain a series of imported materials.

The historical Mutapa state is believed to be a direct off-shoot of the state based at Great Zimbabwe. Political succession followed the system of houses conforming to the lines of descent of sons of the founding leader. According to oral traditions, Nyatsimba Mutota is the last ruler of Great Zimbabwe and the first Mutapa king. He established his capital on the edge of the Zambezi escarpment at a place known as Mutota’s Zimbabwe. The ceramics from Mutota’s Zimbabwe are lavishly burnished with graphite and closely resemble Period IV ceramics at Great Zimbabwe (Pwiti, 1996). After his passing, political succession rotated amongst his sons, most of whom were given districts to preside over. As Beach (1994) and Mudenge (1988) have pointed out, the succession disputes associated with the passing on of Mutapa kings often resulted in collateral or primogeniture succession. Whatever the case, the new leader ruled from his own pre-existing settlement. Thus this succession system resulted in the existence of several stone walled enclosures named after former district and national leaders such as Mutota, Kasekete, Chiwawa, Nowedza and others (Pikirayi, 1993). There is a great deal of discordance among Mutapa oral traditions and Portuguese documents such that there are kings who do not feature on some published king lists and yet there are modern chieftaincies associated with them. For example, in modern day northern Zimbabwe, there are places such as Chiwawa, Matope, Katuruza and Chitsungo, among others, which represent houses from where Mutapa kings were drawn at different points in history (Mudenge, 1988). However, as is apparent from Portuguese records and oral traditions, bitter and bloody succession disputes were often fought after the passing of an incumbent Mutapa. Those who were defeated moved to new areas and established their capitals. Often they would wait for an opportune time to come back to fight and retake political control, but in other cases they established their own centres which became competitors to their original homeland. Because these events took place during a very short time, it is difficult to place them in a radiocarbon timescale which aggregates all the short duration events into one master chronology.

Because of historical continuity, we can use this Mutapa model to understand the significance of the many elite centres that are more or less contemporary with Great Zimbabwe (Figure 4). The radiocarbon dates have indicated that a number of settlements such as Tsindi and Harleigh Farm were established more or less at the same time as Great Zimbabwe. They flourished in their own areas and even outlived Great Zimbabwe itself. However, there are some sites such as Nhunguza, Zvongombe and Thulamela which emerged towards the end of Great Zimbabwe’s florescence. It is possible that these settlements represent Zimbabwe diasporas – that is, people who for one reason or another could not rule at Great Zimbabwe or who were sons given districts to rule. This resulted in the creation of many Zimbabwes. Unlike in the historical record, we cannot, however, pinpoint which ones were provincial or district centres. They may have been capitals as well at one point or another. The diaspora model is well supported by the succession dispute associated with the death of Changamire Dombo in southwestern Zimbabwe. According to oral traditions, Changamire Dombo established the Rozvi state with its centre at Danangombe (Mudenge, 1988). After his death, a succession battle developed resulting in some of his sons migrating across the Limpopo and settling at places such as Dzata. The new Singo state also had shifting capitals and dominated the Soutpansberg area (Beach, 1980). Based on this model, it would seem that some of the elite centres which are on the margins were established in this way. Others claimants to the Rozvi throne settled in the Hwange area in western Zimbabwe and are associated with the site of Bumbuzi. These new centres were independent from the core. Thus, both succession politics and environmental reasons may have precipitated the littering of the plateau and adjacent areas with elite stone walled settlements. OPEN IN VIEWER

If we are to extend this model to the Shashi-Limpopo area, we realise that there are a number of elite settlements contemporary (on the basis of 14C chronology) with Mapungubwe. As the Bayesian model has shown, these include Mapela, K2, Taba zika Mambo and Great Zimbabwe. The dates for K2 and Mapungubwe overlap for the first half of the thirteenth century, while Mapungubwe itself declines after the early fourteenth century. The redating of the sequence on the Hill Complex at Great Zimbabwe has revealed that the site was already a place of importance in the early thirteenth century AD. Thus as the Bayesian model has shown, and as supported by similarities in local and imported material culture, the development of all these polities was coeval, demonstrating that they were competing polities. If we adopt a cultural landscape perspective, one that is not bound by modern political boundaries, we see the development of a number of competing polities on the Zimbabwe plateau and adjacent areas of northeastern Botswana and northern South Africa (Figure 4). Different settlements achieved different levels of success, with some outliving others. Thus Mapungubwe outlived K2, while Mapungubwe was outlived by Great Zimbabwe and Mapela. Tsindi also outlasted Great Zimbabwe while new settlements such as Nhunguza and Thulamela emerged towards the end of Great Zimbabwe’s florescence. Also, Khami thrived alongside Great Zimbabwe for almost a century before outliving it. Comparisons of material culture from these sites show the presence of identical imports, an indicator that no single entity may have monopolised the external trade with the Indian Ocean. For the historical periods, Bhila (1982) and Mudenge (1988) have indicated that traders known as vashambadzi were the middlemen in the trade with the Indian Ocean seaboard. Such a competitive atmosphere is known to make some polities wealthier and more successful than others, as the Mayan example has shown (Schele and Friedel, 1990).

Rather than being a first capital of the Zimbabwe culture, Mapungubwe is one of the many places associated with important innovations that were taking place in the Shona world in the early second millennium AD. As Robinson (1965) has shown, Zhizo and Leopard’s Kopje communities had started building stone walls on hilltops and adjacent areas as early as AD900. There are far too many sites with stone walls and prestige goods, some of which overlap with Mapungubwe (see Appendix 1 and Manyanga, 2007). There are massive disparities in the way in which radiocarbon dates have been collected at individual sites, and yet they are the fulcrum of most interpretations. Some sites have one or two dates while others have considerably more than that, which skews our understanding of leadership processes in the past. A massive dating programme that factors indigenous conceptions of power is an urgent necessity. This will promote multi-linear understandings of the development of culture. Furthermore, it will make it easier to identify sites that overlap and to distinguish them from those that are contemporary. What we are emphasising here is that any model that is seeking to give a narrative of southern Africa’s elite settlements needs to take into account the succession politics that is mirrored in the southern African historical and ethnographic realities. Skeptics may argue that the nature of succession alone may not be the only explanation for chronological and spatial relationships between elite settlements, but it forces archaeologists to consider wider landscapes in which developments are taking place.

What therefore are the implications of our study for global archaeology? It is widely accepted that in real life events take place at a small scale but archaeologically we deal with time aggregates. Attempts have been made to correlate historical chronologies in China (Guo et al., 2000) and in the Middle East (Bruins and Plicht, 2001) with radiocarbon dates. Short-lived materials were dated using AMS dating. Two important results have emerged. First, historical chronologies are more concerned with the duration of rulers and rarely give insight into the time periods when the sites in question were not associated with leadership. Indeed, oral traditions and documentary sources suffer from the same weaknesses. Second, rather than seeing the methods as mutually exclusive, or looking at them in terms of superiority (Bruins and Plicht, 2001), it is essential to make radiocarbon dates and historical chronologies complement each other (Bronk Ramsey et al., 2010). That way, archaeologists can get to grips with the short-term and long-term events associated with archaeological sites.

Conclusion

This article has shown the difference between European and African concepts of elite centres and capitals. The southern African archaeological and anthropological record is replete with elite centres, some of which are contemporary, while others succeeded each other. For the Zimbabwe culture, the glut of elite settlements is possibly explained by succession politics which often created numerous diasporas, thereby expanding the landscape of the Zimbabwe culture. The local and imported material culture is also identical. This, when coupled to the Bayesian chronology, indicates that most sites are partially overlapping, with Mapungubwe, Great Zimbabwe and Khami demonstrating that the establishment and florescence of the Zimbabwe culture was multi-linear. However, the Bayesian chronologies, while useful, still generate age ranges which can incorporate events of different duration. Apart from using oral histories, southern Africa must develop its own proxy data from materials such as sediments. The resulting short-term chronologies can be used to calibrate the radiocarbon timescale, independent of oral traditions. The fact that we find aggregates in the archaeological record and not individual events severely weakens structuralist interpretations that assume that symbolic meaning of space was constant for the Zimbabwe culture. Such work, rather than pushing the archaeology of Great Zimbabwe and related sites forward, has suffocated and stunted it with its rigidity and failure to consider the issue of timescales and the details that can and cannot be recovered archaeologically. Key in pushing the research agenda forward is the need to overcome radiocarbon dating legacies and to move towards an ethnographically derived understanding of short-term shifting capitals with fluxing centres and dynamic shifts in local power through time and space.