Introduction: Beyond Hegemonic Discourse

In this paper, we argue for an interpretation of southern African state formation that is supported decisively by data from important but poorly known sites. This new position contradicts the longstanding linear argument which posits that over nearly 800 years, southern Africa only had three complex states: Mapungubwe (AD 1220–1290), Great Zimbabwe (AD 1300–1450) and Khami (AD 1450–1830), and that these succeeded each other in sequence, with shifts in regional power from one capital to the next. We are not the first to dispute this view about early state formation and we doubt we will be the last (Garlake 1982; Robinson 1985; Lane 1994/5, 2005; Beach 1998; Bourdillon 1998; Kim and Kusimba 2008; Prinsloo and Colomban 2008; van Waarden 2011, 2012). Bourdillon (1998) commented that anybody who understands African sociology, languages, culture and worldview in sufficient detail can easily recognise the incongruous variance between this linear framework and the African societies it was meant to explain.

Researchers familiar with the Iron Age of southern Africa would agree that there has been a very long silence concerning published field research at places such as Khami, where the major work is by Robinson (1959), but see also Thorp (1995) and Hughes (1997); Mapela (Garlake 1968) and Great Zimbabwe (Summers et al. 1961 and Collett et al. 1992). The small-scale and unrepresentative nature of excavations carried out at some of these key sites opens up doors to unlimited degrees of interpretive freedom. This is compounded by an absence of robust dating frameworks (Chirikure et al. 2012) and the dearth of artefact studies, which has created a knowledge standstill (Chirikure and Pikirayi 2008). The unfortunate truth is that if we do not know much about the sites that we argue about, then we do not know much about the evolution of sociopolitical complexity in our region, either (Chirikure et al. 2014) (Fig. 1).

Fig. 1
figure 1

Location of southern Africa showing madzimbahwe/mizinda or capitals associated with sociopolitical complexity

Full size image

In this climate of under-researched and poorly dated sites, there is no need to promote hegemonic discourse by re-affirming an unwavering commitment to the linear model for the evolution and flourishing of sociopolitical complexity in southern Africa, particularly in light of dozens of unexcavated sites (see National Museums and Monuments of Zimbabwe sites database), new evidence and new sites. The linear position is contradicted by archaeological evidence (see van Waarden 2011), our own worldview and lived experiences as Africans, Shona anthropology and oral and written history (Beach 1998). Van Waarden (2011) has identified the Tati region in northeastern Botswana and adjacent regions of southwestern Zimbabwe, the Gumanye area around Great Zimbabwe and Mapungubwe as chronologically overlapping cradles of the Zimbabwe culture.

The main problem caused by a lack of on-going research at essential sites, and archaeologists failing to revisit primary excavation texts, is that any debate inevitably becomes one’s logic and ‘experience’ against that of another. In this paper, we present new evidence from Mapela and combine it with insights from Robinson (1966), Garlake (1968) and van Waarden (1998, 2011, 2012), to argue that the idea of a power relay from Mapungubwe to Khami via Great Zimbabwe is neither supported by archaeological evidence nor Shona worldviews and histories. Sociologists of the Shona such as Bourdillon (1998, see also Beach 1998) also affirmed this position. Jahunda, cluster Tati sites, Great Zimbabwe, Khami, Mapungubwe, Mapela and many other sites have in their basal levels material that dates to the early second millennium AD, the most critical period in early state formation. It is very difficult for the same worldview to produce different material culture given the recursive relationship between the mind, the body and matter (Lakoff and Johnson 1999). Furthermore, new radiocarbon dates indicate that Great Zimbabwe and Khami overlapped chronologically between the 1400s and 1600s, also the time of the Mutapa state, showing that our region urgently requires more fieldwork and locally inspired interpretive frameworks to move beyond hegemonic discourse.

Brief Background

It is important to provide an abridged history of Iron Age research in southern Africa (see Fig. 1). Here, archaeology was established by men and women from the global north, with no meaningful involvement of local communities (Gosden 2012). Although local people provided labour, and their histories enabled interested parties to build ‘archaeological knowledge’, the resulting knowledge was never aimed for their consumption (Shepherd 2002; Chirikure et al. 2010). And yet it was a partial understanding of their worldview by outsiders that was used to interpret the archaeology. For example, Caton-Thompson (1931), the meticulous excavator of Great Zimbabwe who continues to inspire us today, clearly explained that without a solid understanding of local culture from local informants, it would have been difficult to interpret the Iron Age past. This is why even today colleagues interview local people to gain an insight into local worldviews (see Huffman 2007a, p. 175). However, the final products were neither substantiated nor verified by the local communities and providers of the information, who are still considered non-experts in their own worldview and material culture (Huffman 2015b). This created pasts which, although making sense to their archaeologist proponents, often contradict not just archaeological data but also the local worldview, customs, linguistics, histories and traditions (Beach 1998).

This generalisation aptly applies to the evolution of sociopolitical complexity in southern Africa. While researchers have long advocated a multicausal approach to the rise of complex societies in which local and external factors interacted (Garlake 1973; Sinclair 1984; Pwiti 1996; Pikirayi 2001), the development of complexity has been pigeon-holed into a linear framework where ideological transformations from K2 to Mapungubwe crystallised southern Africa’s first state at Mapungubwe (Huffman 1982, 2007b, 2015a, b) (see Gardner 1963 for an earlier version of this argument). After its demise, Mapungubwe was allegedly succeeded by Great Zimbabwe, which in turn was followed by Khami. When this speculation was made, there was very little research to support the linear model, such that a near half-century of theoretical posturing was mostly directed at validating this framework, but with no data from new sites. The result was that so many important Leopard’s Kopje, Great Zimbabwe and Khami-type sites, particularly those in northeast Botswana (van Waarden 2011) and southwest Zimbabwe (National Museums and Monuments of Zimbabwe Database, Garlake 1968), remained largely unexplored, and their contributions to understanding sociopolitical complexity absent. And yet, some of these sites have hilltop locations, poorly coursed Zimbabwe-type stone walling, glass beads and dhaka (earthern) structures. These elements are conventionally used to argue that the chronologically overlapping site of Mapungubwe is the sole cradle of state formation. The unanswered question is why do the same features so widely expressed in the broader region, have significance only when they are present at Mapungubwe?

The only answer is that the conventional framework continue to be supported despite a poor understanding of African worldviews (see Beach 1998), which is why Garlake (1982) argued that the views of archaeologists who grew up in African societies are of immense significance in calibrating archaeological knowledge. Amidst this situation, we initiated large-scale interdisciplinary research aimed at understanding the area between Great Zimbabwe, Khami and Mapela and its connection to northeastern Botswana and northern South Africa (Fig. 1). The project was aimed at training doctoral and masters students in the key areas of chronology, material culture, historical ecology and inter-site relationships. Currently, we have six students (three MAs and three PhDs) working on some of these topics. As a starting point, we revisited the original excavation reports to understand the original context, because there are massive disparities between what original excavators observed and interpreted themselves, and between their observations and what later interpreters opine. A review of this original research (e.g., Garlake 1968) shows that as far as site geography, stratigraphic sequences, radiocarbon chronologies and material culture similarities are concerned, the Mapungubwe-Great Zimbabwe-Khami transition is not sustainable by the evidence (see Chirikure et al. 2013a, b). Similarly, in the Tati cluster of northern Leopard’s Kopje sites in northeast Botswana, Van Waarden (2011, p. 60) identified 43 sites associated with hilltop occupations, 41 of which have terraced slopes. Since most of these sites exhibit features of the Zimbabwe culture in contexts that predate Mapungubwe, she therefore concluded that the origin of the Zimbabwe culture was to the north of Mapungubwe. Furthermore, she argued that Mapela was a capital of an independent state and was therefore not the middleman between Mapungubwe and gold-producing regions of the north as she previously thought (van Waarden 1998).

Note on Methodology and Approach: When Informants Become Knowledge Producers

Archaeology now has a universally acceptable methodology. However, the western epistemological ancestry of the discipline, if not properly calibrated to suit non-western contexts, often results in locally problematic interpretations (Smith and Jackson 2006). In such cases, it may be difficult for those outside a certain worldview to understand why some reconstructions, which are seemingly based on robust archaeological methodologies, gain little acceptance amongst both local people and other like-minded researchers (Smith and Jackson 2006). A number of researchers have expressed dissatisfaction with the generalising way in which ethnographies and historical data have been used to reconstruct many facets of sociopolitical complexity in southern Africa (see Lane 1994/5; Beach 1998; van Waarden 2011). After the relay from Mapungubwe to Khami was put forward (Huffman 1982), structural models were imprinted onto these and related sites. Senior scholars with experiences in African history, sociology, anthropology and archaeology challenged these on methodological bases (see Lane 1994/5, 2005; Beach et al. 1997; Beach 1998; van Waarden 2012). Indeed, a number of Africanist archaeologists have argued for the importance of understanding changing contexts, calibrated finely by local experiences in any study of the African past (Stahl 2001; Lane 2005; Schmidt 2014).

Because it fails to consider these essentials, Huffman’s interpretation does not adequately understand the dynamics of Shona worldview and associated role-playing (Beach 1998). For example, Huffman (1982, 1996) cites the twentieth-century observation that amongst the Shona, a hill is metaphorically associated with kings because they say gomo rawa—the mountain has fallen—when a king has passed on. Furthermore, the Shona also say that kukwira gomo hupoterera—which suggests metaphorically that for one to approach a king, they have to do so through a series of officials (or in a roundabout way). However, the meaning of the latter metaphor, like any other figure of speech, is contextually mediated. Those born into the Shona and African worldview and are conversant with registers and lexicography are aware that it is the king that makes a palace and not the other way round. The thinking that the shift of settlement from K2 to Mapungubwe (originally suggested by Gardner 1963) married ‘the majesty of hills with that of kingship’ (Huffman 1982) is—on the basis of worldview, metaphor and context—naive. Ethnographically and historically, there are many elite places such as the capital of the founding Mutapa, Nyatsimba Mutota known as Mutota’s Zimbabwe (Pwiti 1996), which are located on flats, and sited commoner houses that are on hilltops, too, in the Zimbabwe culture! It does not mean that Nyatsimba Mutota’s subjects would not have referred to him using metaphors of hills. Similarly, hilltop location would not have meant that commoners were now kings! As Lakoff and Johnson (1980) argue, because metaphors are a reflection of thoughts and actions, understanding role playing is important when using them. This is because a single metaphor gives people multiple ways of perceiving the world. According to Reddy (1979) and Schön (1993), privileging only one perspective as Huffman (1996, 2007a, b, 2015a, b) creates cognitive myopia, resulting in unhelpful binaries such as ‘correct’ and ‘incorrect’ positions. As a consequence, a failure to understand the dynamic nature of metaphors produces locally undesirable views (Beach 1998).

In the Shona worldview, nothing is fixed in space and time and there is a recursive and complex relationship between worldview and material culture. In fact, Lakoff and Johnson (1999) have shown that worldview and material culture influence each other in a complex way, such that a worldview produces a material culture, which in turn shapes the worldview. By arguing that we privilege material culture over worldview, Huffman (2015b) misses the point that the two are not mutually exclusive. As such, if one studies the archaeology produced by their own worldview, their interpretation flows naturally from that worldview. It is our knowledge of the Shona worldview which makes it unclear why hilltop sites such as the Tati cluster sites (van Waarden 2011) and amongst others, Mapela with Zimbabwe style dry-stone walling and dhaka, cannot also be associated with the majesty of kingship as well, given that they were authored by ancestral Shona speakers in the early second millennium AD. Most of the existing frameworks were never reviewed, validated or substantiated by local people from their worldview’s point of view. As such, there are aspects of archaeological interpretations which contradict local meaning (Beach 1998) as well as, as we have argued above, published archaeological data from the region, and it is these that we engage with in this paper.

Mapela and the Multilinear Evolution of Sociopolitical Complexity in Southern Africa

Mapela is located on a prominent hill (90 m high) in the Shashi region, on the southernmost fringes of Gwanda, southwest Zimbabwe. It is approximately 2 km east of the Shashi-Shashane confluence and approximately the same distance north of the border with Botswana (Fig. 1). The hill flattens out westwards and thins out into a small terraced hill. Immediately to the west is Little Mapela. One cannot understand Mapela without considering the history of the area, particularly the forced removals of the 1960s. Before this time, Kalanga-speaking people belonging to the Sijiye lineage lived in the area around Mapela (Headman Ragamanya, pers. comm.). However, with the outbreak of the war of liberation, the Rhodesian military established a base less than 0.5 km southwest of the Shashi-Shashane confluence. People were evicted and resettled about 15 km away. At independence, the area was designated for wildlife conservation, and today, it is used as a cattle-grazing zone.

Garlake (1966, 1968) carried out preliminary work collecting surface collections and excavating five test trenches that exposed a stratigraphy dominated by solid dhaka (earthen) floors from top to bottom. Furthermore, he recovered a significant amount of K2 and Mapungubwe beads and pottery belonging to the Leopard’s Kopje tradition. He also mapped the summit of Mapela Hill (Fig. 2) and concluded that compared to Mapungubwe, it had many walled platforms where houses were built. Further work was carried out at Little Mapela, a site with both terrace and freestanding walls. The main conclusion from Garlake (1968) was that because of Zimbabwe culture features, Mapela was a capital of an independent ancestral western Shona kingdom characterised by class and social distinction, a view shared by Beach (1984) and van Waarden (2011).

Fig. 2
figure 2

Garlake’s (1968) map of Mapela summit showing multiple platforms covering the entire hilltop. The platforms are built of poorly coursed dry-stone walls and contain solid dhaka structures. A preliminary house count identified at least 24 houses on the different platforms

Full size image

Between 2013 and 2015, we carried out dedicated archaeological research in and around Mapela, followed by three intensive intrasite surveys and excavation seasons at Mapela. The site has two types of poorly coursed dry-stone-walled platforms. The first type consists of revetment walls primarily constructed to level the slopes. On the resultant terraces, the inhabitants built the second type of platforms where houses with solid dhaka floors and features were erected (Fig. 3). The second type is present on the summit and on the slopes. Our mapping exercise produced a map of Mapela (Fig. 4) which unequivocally dwarfs that of Mapungubwe (Fig. 5) and to our knowledge is the biggest mapped Leopard’s Kopje site.

Fig. 3
figure 3

One of the multiple poorly coursed platforms on Mapela Hill summit (photo: S. Chirikure)

Full size image
Fig. 4
figure 4

Map of Mapela showing the extent of the site and a high concentration of houses on the summit. Note the sharp cliff around the summit (bold brown line; image in full colour online) (from Chirikure et al. 2014)

Full size image
Fig. 5
figure 5

The plan of Mapungubwe Hill showing that the site has only one stone-walled platform on the summit (redrawn from Huffman 2007b by F. Bandama). The perimeter wall exists only on maps; it cannot be verified on the ground. Even if perimeter wall was destroyed, Mapungubwe walling is small-scale as compared to Mapela (both summit and slopes) (Figs. 2 and 4)

Full size image

Stratigraphic excavations exposed ceramics which were analysed using Huffman’s typology that considers vessel shape, decoration placement and motif (Huffman 2007b, pp. 282–291). Iron Age archaeologists defined two clusters of the Leopard’s Kopje: a northern cluster (Mambo—phase 1 and Woolandale—phase 2) (distributed in southwestern Zimbabwe and northeastern Botswana) and a southern cluster (K2—phase 1 and Mapungubwe—phase 2) (distributed in parts of southern Zimbabwe and the Shashi-Limpopo confluence area) (Robinson 1959; Huffman 1974; van Waarden 1998). Huffman (2007b) adds Transitional K2 for the southern cluster but which has no equivalent in the northern cluster. Readers must, however, be aware that while presenting the Mambo and Woolandale motifs, Huffman (2007b, pp. 292–295) does not provide descriptions of their key features as he does for K2 and Mapungubwe. This makes the identification of the Leopard’s Kopje arbitrary and ever-changing, depending on the argument at hand (e.g., Huffman 2015b). Using Huffman’s (2007b) guidelines, the bottom layers of the excavated trenches at Mapela only contained K2 pottery which was succeeded by Transitional K2 types with Mapungubwe pots occurring in the top layers (Fig. 6). Figure 6 also shows that Mapela has recurved bowls which while present at Mapungubwe are absent in Woolandale (Huffman 2015b).

Fig. 6
figure 6

Selected K2, Transitional K2 and Mapungubwe pottery from Mapela (drawing: F. Bandama)

Full size image

Important features associated with the Zimbabwe culture such as solid dhaka floors and stone walls were found in association with the basal levels (Fig. 7) and continue throughout the sequence (see also Garlake 1968). The sequence of pottery in the stratigraphy was matched by the succession of glass beads with K2 series at the basal levels, followed by the Mapungubwe series at the top. These glass beads often devitrify and accumulate a variously coloured patina. Given that there are very few dates from our region’s most important sites, we made a modest investment in dating Mapela. The dates from the Terrace Excavation Area 1 covered levels 7 to 19 (see Table 1), because of the need to develop a tight chronology for the relationship between phases 1 and 2 Leopard’s Kopje. The lower summit excavation was also dated (Table 2). The radiocarbon dates of Mapela (both terrace and lower summit) and those of Garlake (1968) estimate the duration of occupation to between the eleventh and fourteenth century AD.

Fig. 7
figure 7

A Zimbabwe cement floor with a bevelled edge on the northern side of Terrace Excavation Area 1, Trench 1 (depth 140 cm). This 25-cm-thick floor lies below a feature resembling a granary. The southern edge of the floor was excavated so that we could continue to bedrock (photo: S. Chirikure)

Full size image
Table 1 Dates from Terrace Excavation, Trench 1 at Mapela (calibrated with OxCal version 4.4 using Hogg et al. 2013)
Full size table
Table 2 AMS dates from Mapela lower summit (calibrated with OxCal version 4.4 using Hogg et al. 2013)
Full size table

Our fieldwork has confirmed Garlake’s (1968) observations and added new insights which demonstrate that Mapela has an early Zimbabwe culture occupation that chronologically overlaps with K2 and Mapungubwe combined and with other culturally related sites in northeastern Botswana and southwestern Zimbabwe that predate Mapungubwe (van Waarden 2011).

Previously Unstudied Material Culture and New Excavations at Great Zimbabwe

Today, the basis for understanding Great Zimbabwe is the platform established mostly by Summers et al. (1961), who combined stratigraphic evidence with architectural history, radiocarbon dating and material culture evidence to develop a robust five period sequence for the site. Using indications from Garlake (1973), that imports from the Hill Complex did not date beyond AD 1450, Huffman and Vogel (1991) calibrated available radiocarbon dates at Great Zimbabwe and argued that the site flourished between AD 1300 and 1450. Apart from treating AD 1000 to 1300 and post-AD 1450 peoples as unrelated to the builders of stone walls, Huffman and Vogel’s (1991) chronology promotes the impression that all components of Great Zimbabwe were occupied and abandoned at the same time. However, as old areas were abandoned within the settlement, new ones were occupied, resulting in the artefact patterning and chronological overlaps and differences visible inside and outside the stone walls (Summers et al. 1961; Collett et al. 1992; Chirikure and Pikirayi 2008).

Besides Robinson’s (1961) excavation which has a stratified sequence, most of the individual dates came from different areas of the site, which makes it difficult to understand the evolution of various parts of the settlement. In order to have a general understanding of the chronology of Great Zimbabwe, we therefore utilised the internationally acclaimed Bayesian modelling techniques (Bronk Ramsey 1994) to have an idea of the maximum possible time that Great Zimbabwe would have been occupied as supported by material culture distribution (Chirikure et al. 2013a). The major advantage of Bayesian models is that they allow researchers to combine stratigraphic and material culture evidence with radiocarbon dates to develop well-resolved chronologies. In our model, we combined stratigraphy, not just layers but also pottery sequences, relative dates from imports, together with the architectural history. The main constraint was that there are simply too few dates on which to produce robust models: material culture indicates a longer duration than available radiocarbon dates. To add onto this work, we decided to analyse material produced during conservation-inspired excavations at different parts of Great Zimbabwe between 1990 and 2000. Before any wall restoration activities, excavations are carried out down to bedrock to expose the foundation and other important details. Such previously unpublished data provided an important cross-section of the walls and the cultural succession in the concerned area. One of the most important excavations carried out in this way were conducted on the Terrace Platform and Northern Rockshelter all located on disparate parts of the Hill Complex. In addition, we excavated a midden in the modern car park area, believed to have been occupied by commoners (a fuller report of this work is in preparation).

The conservation-inspired excavation produced evidence of Periods II, III and IV pottery in direct succession on the hill. Other finds found throughout the periods include iron slag, bronze artefacts, figurines, a significant amount of bone and various metal objects. The glass beads from the excavations followed a similar succession from Periods II to III and eventually IV. This evidence is consistent with domestic middens and not that of left overs from rituals. It also became clear that with the exception of Caton-Thompson (1931) researchers have virtually ignored the chronology of massive terraces that cover the lower sections of the hill. Caton-Thompson (1931) exposed (Fig. 8) several terrace walls covered by debris, showing that there was massive rebuilding of walls on the hill. These terraces whose dating is not known were completely covered by midden material. Therefore, it is premature to argue that the South Wall on the Western Enclosure on the Hill, believed to be the first wall in this section, was the earliest wall at Great Zimbabwe. In fact, Summers (1971) showed that massive building and rebuilding took place even here and that there are earlier walls that preceded the South Wall (Fig. 9).

Fig. 8
figure 8

Caton-Thompson’s (1931) cross-sections in test pit A3 in her terrace excavations which exposed retaining walls in association with earlier material culture (redrawn by F. Bandama)

Full size image
Fig. 9
figure 9

Plan and sections of excavations in the Western Enclosure the hill. After Summers (1971, pp. 4, 10-11) (image in full colour online)

Full size image

Most of the earlier inner walls, in Fig. 9, predate the South Wall and do not feature in Robinson’s section (Fig. 10, see Fig. 9, part of Test 1 section 1958). This inner wall does not rest 1 m above the Period III deposits but is separated by a thin layer (Fig. 9). Equally, Robinson’s (1961, p. 175) comment that there was limited floor space only applied to test 1, a 1 × 1 m trench and not the entire enclosure (see Fig. 9). In fact, the comprehensive cross-section of several test excavations produced by Summers (1971) reveals that there was more than enough floor space and gradient to build several houses in the Western Enclosure from Periods I and II through to abandonment (Fig. 9). As we demonstrated above, other areas of the hill had a widespread presence of material culture, where houses were built in Periods I, II, III and IV. Our experience of participating in Shona rainmaking ceremonies informs us that this accumulation of large middens and rebuilding of stone walls between AD 1000 and 1300 is inconsistent with the cultural practice.

Fig. 10
figure 10

This image shows structural features of the south wall in relation to Robinson’s (1961) test IV, V and VI. It excludes earlier inner walls (see Fig. 8)

Full size image

One of the common sightings at Great Zimbabwe these days is eroding archaeological deposits, in particular in the area around the modern car park. We excavated a midden associated with a number of house clusters whose pattern conforms to the typical Great Zimbabwe plan of densely clustered houses with connecting barrier walls in commoner residences (Caton-Thompson 1931). The midden had been exposed to elements for a long time such that most of its top section was gone. What started as an experiment in seeming futility sprung a very big surprise. We excavated a 3 × 1 m trench and reached bedrock at a depth of just over 60 cm in parts. The midden had extremely high levels of integrity (Fig. 11). All levels of the midden, including the basal ones, contained typical Period IV pottery dominated by thick rims and lavish graphite burnishing.

Fig. 11
figure 11

A cross-section of the Car Park Midden excavation

Full size image

Yellow cylindrical glass beads and Indian red cylinders were also found in the bottom layers together with a lot of bone. No black oblates associated with intermediate levels on the Hill Complex were found. While most of the finds were the same in the top layers, bluish green, turquoise blue and navy blue beads were the new addition. A significant amount of bronze and copper wire was found together with remnants of crucibles. Because the excavation produced burnt bone, three samples were submitted for AMS dating of which only two contained extractable collagen. One was from the 55 cm level, while the other was from the 20 cm level. The dates (Beta 396345: 320 ± 30 BP and Beta 396345: 410 ± 30 BP) estimate the age of the excavated part of the midden to be between AD 1450 and 1660. These dates decisively demonstrate that this area of Great Zimbabwe was built after the Hill Complex and likely the Great Enclosure ceased to be important residential centres. The similarity of house plans and material patterning with other areas between the outer perimeter wall and the Great Zimbabwe Hotel eastern gate shows that a large section of the site was massively occupied from the mid-fifteenth century and therefore overlaps with Khami for over 150 years. The new dates even extend the Bayesian model in Chirikure et al. (2013a) by a century, to AD 1660, demonstrating that the road to new meaning and understanding lies in sustained fieldwork in previously neglected areas. Below, we consider the implications of the new evidence from Mapela and that from Great Zimbabwe in relation to the wider southern African region.

Discussion: A Time to Move On

Excavations at Mapela and surveys in the broader region of southwestern Zimbabwe, as well as new fieldwork at Great Zimbabwe when supplemented by data from van Waarden (2011)—who argues that the Zimbabwe culture is earlier to the north (Woolandale area) than it is to the south (Mapungubwe area)—require an interpretational framework of a multidirectional evolution of sociopolitical complexity.

Our study of ceramics from Mapela using Huffman’s (2007b) typological method unequivocally identified K2, Transitional K2 and Mapungubwe pottery. After viewing 11 sherds used for illustrative purposes in Chirikure et al. (2014), Huffman (2015b) arbitrarily declares that Mapela ceramics are not southern Leopard’s Kopje but are now Woolandale (northern cluster). Subsequent to Huffman (2015b), we went back to Mapela and opened up new areas and extended Garlake’s trenches on the summit. To address the question of whether Mapela pottery belongs to the southern or northern Leopard’s Kopje cluster, we identified motifs and compared them with those illustrated by van Waarden (1998) and Huffman (2007b) (Fig. 12).

Fig. 12
figure 12

Ceramic decorative motifs from Mapela in comparison to those from southern and northern Leopard’s Kopje clusters in Huffman (2007b)

Full size image

Figure 12 demonstrates clearly that Mapela motifs are closely related to Huffman’s (2007b) motifs for the southern cluster. A comparison revealed that the motifs on Mapela pottery strongly correspond to K2 (100 % similarity), Transitional K2 (81 % similarity) and Mapungubwe (78 % percent similarity) (Table 3). This close resemblance sharply contrasts with the very low percentages of motif similarities between Mapela and the northern cluster. The percentage similarity between Mapela K2 and Mambo is 25 %, while that between Mapela Mapungubwe and Woolandale is 11 %. If Mapela was Woolandale as Huffman now claims, then the similarity ratios between the two would be very high. Table 4 shows that there are low percentages in the similarities between Huffman’s (2007b) northern and southern clusters. This confirms that Mapela pottery belongs to the southern Leopard’s Kopje and challenges the suggestion that it is Woolandale. Perhaps this confusion indicates that we must do away with these labels used to describe material culture made by related peoples.

Table 3 Percentage Mapela motifs compared to the Southern and the Northern Cluster
Full size table
Table 4 Percentage Northern Cluster motifs compared to the Southern Cluster
Full size table

Moving away from ceramics, Garlake (1968) found abundant glass beads, solid dhaka floors and extensive walling, characteristics of what is known as the Zimbabwe culture. These observations were validated and substantiated through our own research. The new map of Mapela brings into sharp focus the size and density of walling on both the summit and the slopes, which, while comparable to some sites in southwestern Zimbabwe and northeastern Botswana, dwarfs Mapungubwe by far (see Figs. 2, 4 and 5). Based on size and a major labour investment in constructing walled platforms and a retaining wall covering the entire summit, it is clear that Mapela was much bigger than previously imagined. There is a misconception that the walls at Mapela are not Zimbabwe culture because they do not have coursing. The walls were poorly coursed but in cases where big blocks were used together with smaller ones, the courses are not always obvious. The platform walls on the summit are also poorly coursed (P style walling) (see Summers et al. 1961 for wall types in the Zimbabwe culture). Our excavation of one terrace wall on the northern slopes showed that the construction technique involved stacking a core of rubble inside a neat outer face of poorly coursed stones, and this is typical of Khami walls (Summers et al. 1961; Sinamai 2014). Deposits with K2 pottery accumulated on the platform show that the wall was built in the eleventh century AD.

Amongst the Shona, stone walls, whether poorly or neatly coursed, are known as masvingo or lusvingo, and there is no term for ‘prestige walling’. In terms of Shona worldview, it was the occupants and their lineage that gave prestige to a place. Garlake (1970) demonstrated that in southwestern Zimbabwe and adjacent regions, dry-stone wall platforms dominate, while free-standing enclosures are more common in south central, northern and eastern Zimbabwe. The ideology of enclosing is associated with the latter, while the former is associated with an ideology of exposing. Therefore, the Leopard’s Kopje platforms on hilltops such as Mapela, Mapungubwe and related sites in northeastern Botswana express an exposing ideology similar to that of Khami and its platforms (Pikirayi 2013). Given the exposed nature of these platforms, ‘sacred’ leaders, if ever they existed, were not shielded from public view. In southwestern Zimbabwe and northeastern Botswana, Leopard’s Kopje and Khami overlie each other stratigraphically showing continuity in ideology (van Waarden, pers. comm.). The typology of Shona dry-stone wall masonry techniques exposes an evolutionary trajectory that progressed from poorly to neatly coursed walls (Summers et al. 1961). According to van Waarden (2011), there is evidence that Leopard’s Kopje people experimented with and developed techniques of dry-stone walling in northeastern Botswana and southwestern Zimbabwe earlier than at Mapungubwe. The weight of the early appearance of elite features at Leopard’s Kopje sites and the associated worldview in the regions north of the Limpopo (van Waarden 2011) strongly indicate the southward spread of the Leopard’s Kopje (see Robinson 1985).

Huffman (2015b) has, on the basis of insights from Google Earth (no field visit), ‘identified’ errors of observation and ‘poor scholarship’ on our part, which justifies his use of undisclosed ‘archaeological knowledge from elsewhere’ to throw away over half a metre of basal stratigraphy and the tightly clustered eleventh-century AD dates. Huffman’s view is, however, contradicted by the archaeology on the ground as well as the local history. For example, the argument that there is no suitable agricultural land around Mapela because a Google Earth scan around the site shows that no modern populations reside in close proximity of the site (Huffman 2015b), misses the historical fact that a large population was forcibly evicted from this area during the 1960s. Furthermore, Huffman (2015b) questions our ability to identify dhaka floors, but it was indeed Chirikure who provided the information that Huffman uses to describe Shona floor construction (see Huffman 2007a, p. 175). In any case, many of the excavators at Mapela dug at Great Zimbabwe, and amongst other places Khami, such that they are competent and experienced in separating Zimbabwe from non-Zimbabwe culture floors. This is why we identified Zimbabwe cement floors with typical features such as bevelled edges on the summit and slopes from the eleventh century onwards.

Our absolute chronology of Mapela is also very important. The new dates (Tables 1 and 2) were modelled in the programme OxCal using stratigraphic information, pottery typologies and bead chronologies (Chirikure et al. 2014). The results demonstrated that Mapela was occupied by Leopard’s Kopje people who abandoned it in the fourteenth century. The presence of K2 and Mapungubwe beads and pottery (see also Garlake 1968) independently supports this chronology. Huffman (2015b) offered his own chronology for the site but because of material errors and incorrect suppositions described above, he throws away over half a metre of stratigraphy using undisclosed data from elsewhere. Readers cannot verify and reproduce his chronology. We find the practice of discarding stratigraphic information and dates on the basis of undisclosed information unhelpful because it produces imaginary history (Beach 1998).

The Leopard’s Kopje, whether at Mapela, Mapungubwe or Woolandale and regardless of labels used, has a wide expression and was authored by the same western Shona people with a similar ideology. There is no reason to partition these people into little pockets that could and could not do certain things when their archaeology suggests otherwise (Hall 1984). Because Mapela has a clear stratigraphic succession from K2 to Mapungubwe on the walled terraces, and that it has eleventh-century dates for the beginning of elite Zimbabwe culture features like other sites (e.g., Van Waarden 2011), it motivates for multilinear evolution of sociopolitical complexity. The only reason why Huffman (2015b) arbitrarily somersaults from his more than 40-year view that Mapela is southern Leopard’s Kopje, is that it has Zimbabwe culture features before his K2-Mapungubwe transitions. Even if Mapela did not exist, van Waarden (2011) has already argued that Mapungubwe is later than the Tati cluster sites in northeastern Botswana and southwestern Zimbabwe, thereby making a very strong case for multilinear evolution of sociopolitical complexity.

Moving onto Great Zimbabwe, the Car Park Midden yielded cultural material belonging to late Period IV (after 1400). The complexity and construction style of the nearby house floors and their plans resemble those that were found in the elite areas since Period III times (Caton-Thompson 1931). The ceramics are dominated by thick rims and are lavishly graphite-burnished, making them Period IV. Glass beads from this area are dominated by yellow and bluish green oblates and cylinders that are commonly labelled as Khami series. The radiocarbon dates estimate the age of the deposit to be between AD 1450 and 1660. This dating as well as the glass bead types contradict the perception that Great Zimbabwe collapsed at AD 1450. As such, Khami and Great Zimbabwe co-existed for a long time.

The dating from the Car Park Midden at Great Zimbabwe is important for another reason—it exposes a chronological overlap between Great Zimbabwe and the Mutapa state. This further opens up valuable insights, specifically those related to Great Zimbabwe’s religion. Portuguese documents from the sixteenth century onwards mention that Mutapa (Shona) religion was monotheist (Mudenge 1988). At the apex was a high God, whom the living could only speak to indirectly via ancestors (plural: midzimu, singular: mudzimu). The tribal or chiefdom level midzimu, the spirits of deceased kings or chiefs (mhondoro) looked after the entire state and were important in asking for health, rain and amongst other things success in war. According to Mudenge (1988), there was a clear separation of powers between chiefs who were political leaders and mhondoro who presided over spiritual issues. Generally, mediums for mhondoro lived away from residences of the reigning kings. Huffman (1996) previously dismissed the presence of spirit mediums in the religion (including rainmaking) of Great Zimbabwe, citing its temporal distance from the Mutapa state. Given this evidence of contemporaneity, there is no reason why the Mutapa religion would be different from that of Great Zimbabwe because both were authored by the same people with a similar worldview (Mudenge 1988).

Citing the work of Schoeman (2006) and Murimbika (2006), who in turn cite Huffman (1996), Huffman believes that the kings at Great Zimbabwe directly controlled the rainmaking that was carried out in the Eastern Enclosure on the Hill. This, argues Huffman (2007b), differentiated rainmaking at Great Zimbabwe from that of the Mutapa state, which was presided over by mhondoro or mediums possessed by spirits of deceased kings/chiefs. Because in the Mutapa state rainmaking ceremonies were conducted by mhondoro who lived away from palaces, we should not expect to find clearly demarcated rainmaking spaces in former palaces such as the Valley and Great Enclosures at Great Zimbabwe. These had spaces such as back platforms inside houses (zvikuva) which would have been important in ancestor supplication. Our investigation of material culture from excavations associated with wall restorations examined a comparatively very large area of the Hill Complex, previously unconsidered. The periods predating AD 1300 are associated with terraced walls, dhaka floors and middens containing slag, bone, figurines, metal objects, pottery and much more that is consistent with domestic debris. Because rainmaking ceremonies typically left ephemeral scatters of material culture meant to be symbolic, the view that Great Zimbabwe was a rainmaking site with no settlement between AD 900 and 1300 is not supported by the evidence.

Shona people prayed to God via ancestors and as such, there are dedicated spaces for religious activities both inside and outside the houses. The Hill Complex, however, deserves special mention as the seat of power of the founder of the kingdom. Amongst the Shona, founders of chiefdoms are some of the most powerful mhondoro (Mudenge 1988). For example, the medium of Mutota, the founder of the Mutapa state, is highly revered. The residences of founding kings are also powerful shrines where his mediums ask for rain. The Hill Complex would have served this purpose even after the seat of power had moved to other parts of the site. Consultations with ancestors were not just for rainmaking but were also necessary for healing and success in war. Therefore, ritual spaces had multiple functions. For example, the back of the Great Enclosure and its conical tower, which is believed by some to symbolise fertility of the land, would have played an important role in theology when the Great Enclosure was still the residence of the king. Because Shona divinity is based on seniority, the leaders in the Great Enclosure or the Valley Enclosures would have prayed in their own homesteads, but in times of trouble would go to the hill, the home of the founder of the state. In the Mutapa state, there was a hierarchy of junior to senior mhondoro such that in times of drought the senior mhondoros were petitioned to talk to the high God Mwari. In fact, Mutota, the founder of the Mutapa state, is a very senior mhondoro when compared to his successors, such that his former capital is highly revered. Therefore, as the first seat of power, the Hill Complex would have retained much significance, such that after abandonment it became an important shrine for healing, rainmaking and other aspects of Shona theology.

Historical (Beach 1980) and archaeological (Pwiti 1996) data indicate that the Mutapa state was in existence by the mid-fifteenth century. Indeed, de Barros claimed that in the sixteenth century, Great Zimbabwe was still being occupied by Mutapa wives (Mudenge 1988). This statement should not be read literally; however, for amongst the Shona, it is common to refer to the most important, the weak and the vanquished as wives, not in a physical sense, but in a metaphoric one. Archaeologically, this is strengthened by the view that a vast area around the Car Park Midden and the Valley Enclosures was still occupied in the sixteenth and seventeenth centuries by a large number of people. A mapping of material culture distribution in these areas identified crucibles, slag, bone, dhaka, bronze objects and much more from middens associated with individual homesteads. Therefore, Great Zimbabwe was a fully functioning town, although on a much reduced scale, and was thus not a sanctuary for pregnant Mutapa wives whose husbands lived 500 km away (cf. Huffman 1996, 2007a, 2015a; see Beach 1998). This betrays Huffman’s (1996, 2007a, 2015a) poor understanding of Shona worldview and lexicography (see Beach 1998). That all this occupation took place when none of the Hill Complex and Great Enclosures was still occupied indicates that we cannot use chronology from one part of the site to estimate the chronology of the entire site. This historical shift of power to the Valley Enclosures and other parts, explains why the material culture here is not different from that of the Hill or the Great Enclosure, other former residences of successive rulers at the site (see also Beach 1998; Chirikure and Pikirayi 2008).

Huffman (2015a) takes aim at the way in which Chirikure et al. (2012) apply the principle of rotational succession, which comes in varieties such as collateral succession, to understand possible relationships between different elite sites at intra- and inter-site levels. In the Mutapa State, political succession followed the system of houses: sons of the founder of the state (Beach 1994). As such, Mutota, the founding Mutapa, was succeeded by his son: this is logical, but from his successor onward, succession was in principle to rotate to another house, i.e., that of one of Mutota’s sons. Of course, this principle was often violated through coups, succession wars and so on. However, the most important principle is that very few new kings resided in palaces of their predecessor (Beach 1994). This explains why there are so many elite centres named after Mutapa kings such as Mutota and Nowedza, to name just two. Also, because they were members of the elite, some of the district and provincial leaders could build their own homesteads that could later become centres of the state if succession moved to them by whatever means.

Furthermore, Beach (1994) also demonstrated the operation of this principle amongst the Buhera confederacy with places such as Gombe being associated with founders. Of course, Huffman (2015a) is right to argue that Dzata was occupied by successive leaders, but he has no evidence that they all occupied exactly the same spaces. He also forgot to mention that the founder of Dzata had moved away from Tshiendeulu to establish their own capital. Furthermore, succession did not remain at Dzata, for it moved as well, showing within variation that the principle works and that settlements changed status (Hanisch 2008). This partly explains why there are thousands of sites in our region. If new leaders did not establish new capitals often named after them, then we would be having only a handful of sites. Given this rich historical record, and our own lived experiences, we find Huffman’s (2015a) criticism that we cannot use Mutapa history of the early eighteenth century to understand Mutapa history and material culture of the late seventeenth-century, misplaced. Even if we remove the Mutapa, the Buhera and Venda histories demonstrate application of the same principles. We believe, however, that methodologically it is much better to use leads from the eighteenth century to understand seventeenth century history of the same people than it is to use nineteenth-century ethnographies to understand the spatial organisation of societies dating to AD 300.

In conclusion, the history of archaeology shows that interpretive frameworks must change with new evidence. Huffman (2015a, b) rebuts new evidence of early state formation so that Mapungubwe remains southern Africa’s only ‘first’ state. This position is contradicted by the wide expression of the Leopard’s Kopje and embedded elite features. It is for this reason that van Waarden (2011) argued that northeastern Botswana and southwestern Zimbabwe are credible origins of the Zimbabwe culture, pointing to multidirectional evolution of sociopolitical complexity in southern Africa. Our work at Mapela has reinforced this thinking. The chronological overlaps between multiple Leopard’s Kopje sites with elite features of the Zimbabwe culture, and the chronological overlaps between Great Zimbabwe, Khami and the Mutapa state suggests that at any given point in time, there were many and not just one powerful Shona state.

Future Pathways: Beyond Hegemonic Discourse

Perspectives from global literature indicate that nowhere in the world have researchers managed to identify the ‘first complex society’, particularly in contexts where a number of chronologically overlapping polities with identical material culture existed (Yoffee 2005). Indeed contra to the conception of powerful territorially large dominions, many early states were small city-states, engaged in struggles for dominion and dependence over each other (Stein 1998; Yoffee 2005; Kim and Kusimba 2008). There is no reason, particularly given the many chronologically overlapping Leopard’s Kopje and Zimbabwe culture sites littered across the vast area comprising southwestern Zimbabwe, northeastern Botswana and northern South Africa, why southern Africa should be an exception. Archaeologists should open up to the idea of African dynamism, because the argument that the Zimbabwe culture started at Mapungubwe—initiated by Gardner (1963) and sustained ever since—is not supported by evidence from the wider region. The way in which new evidence is rebutted with so many errors of observation and attribution exposes the entrenched nature of hegemonic discourse in southern Africa. We therefore urge associations such as the Society for Africanist Archaeologists, the Society of American Archaeology and, amongst others, the World Archaeological Congress, to follow the lead of the British Society for the Advancement of Science of over a century ago, and dispatch a team of neutral researchers to verify our observations at Mapela in the field. Such a team must also revisit ceramic classification and Iron Age sequencing because of the problems that we have exposed.

Mary Douglas (1967) argued that it is one thing to read published ethnographies or to collect ethnographies in a perfunctory way, and another to understand the language and worldview of African peoples (see also Beach 1998). On their part, Smith and Jackson (2006) stated that the issue of worldview and language is so pertinent that concepts making sense in English often do not mean much locally. Our approach to fieldwork and interpretation flows from being born and operating in the African worldview, and is inspired by those who believe that archaeologists from African backgrounds are not just informants but are knowledge producers with more to offer to Iron Age studies (see Garlake 1982; Smith and Jackson 2006; Schmidt 2009; Lane 2011). It is inconceivable that Chirikure can recount to Huffman (2007b, p. 175) how dhaka floors are made, and Huffman (2015b) turn around and suggests that Chirikure cannot identify floors. Like dozens of local people who provided information, the suggestion is that one is good as an informant but not as a knowledge producer. Meskell (2011, p. 155) has deplored such a view and argued that it is racist. When informants become knowledge producers, they interpret the past from their own worldview, experiences and understanding which necessitates the revision of incompatible theoretical frameworks. Like Garlake (1982) and Lane (2011) before us, we believe that the future of African archaeology lies in having more people who grew up in societies with an ancestral connection to various categories of sites (and share the same worldview) taking up an interest in the discipline. Any claims that anybody can know a worldview better than the people who live and experience it, are as comical as the Ugandan dictator Idi Amin’s claim to be the last King of Scotland.

Our position has always been that we must debate with civility and therefore welcome Huffman’s (2015b) call for respect. This, however, comes a long time after we have endured torrential abuse from him through the review process (evidence available on request!). To some, that an aggressor claims to be a victim may be unhelpful grandstanding, but we interpret such a gesture as a sign of good things to come. Finally, if Iron Age studies are to flourish, universities must do more to train local scholars because our success in the production of models has not been matched with the production of graduates and PhDs from all backgrounds. Our word is not the final one; we welcome critique from colleagues, but not from those who do not want to engage new data, in new conversations.