Landscapes and Ethnicity: An Historical Archaeology of Khami-Phase Sites in Southwestern Zimbabwe

Although numerous dry-stone-walled Khami-phase sites are scattered throughout southwestern Zimbabwe, their finer archaeological and historical credentials remain largely elusive. Deliberations of how several dynastic and ethnolinguistic formations that are historically linked to the rise and spread of these sites can be archaeologically defined also remain multiple and fragmentary. Correspondingly, in attempts to trace Rozvi ethnicity, a sample of three ancient Khami-phase capitals––Danamombe, Naletale, and Zinjanja––was scrutinized in the light of an agency landscape framework. For the first time, radiocarbon dates from these sites are compared with each other in order to assess the validity of oral traditions and documentary sources describing the Rozvi past.

Résumé Bien que de nombreux sites aux murs de pierre sèche de la période Khami soient dispersés à travers la région sud-ouest du Zimbabwe, leurs références archéologiques et historiques plus précises demeurent largement inconnues. Les débats sur la manière dont plusieurs formations dynastiques et ethnolinguistiques historiquement liées à l'émergence et la propagation de ces sites peuvent être définies d'un point de vue archéologique, demeurent également multiples et fragmentaires. De la même facon, dans le cadre de tentatives pour retrouver les traces de l'ethnie Rozvi, des échantillons de trois capitales anciennes de la Zambezia, there are several gray areas when it comes to illuminating the archaeological identity of associated dynastic and ethnolinguistic groupings. Diverse segments of the Shona society composed of the Kalanga, Karanga, Korekore, Torwa, Rozvi, Nambya, Singo, and Venda are widely linked to these settlements, yet their actual archaeological imprints remain mysterious (Randall-McIver 1906;Caton-Thompson 1931;Robinson 1959Robinson , 1966Mudenge 1974;Beach 1980;Pikirayi 2001; Machiridza 2005Machiridza , 2012Machiridza , 2013Huffman and du Piesanie 2011;Van Waarden 2012). These gaps, which have been glossed over for quite some time, are hereby revisited through the aid of an agency landscape framework. In particular, the Rozvi, who are historically linked to a number of Khami-phase sites, such as Danamombe, Naletale, and Zinjanja, are singled out for detailed study. To date, the origins of these sites remain poorly comprehended owing to limited research coverage and concomitant Torwa/Rozvi dynastic histories. The Torwa (Kalanga people) were the first to emerge and rule from these sites around the early 15th century A.D., but they were later conquered and assimilated by the Rozvi during the late 17th century. Apart from assimilating Torwa rulers and buildings, the Rozvi also invented numerous narratives premised on monumental buildings, such as Khami, Danamombe, Naletale, Zinjanja, Manyanga, and Great Zimbabwe (Posselt 1935;Beach 1980;Fontein 2003), which are illustrated in Figure 1.
Although the Rozvi assimilated almost everything that was Torwa, the material culture implications of this political changeover remain vague. In particular, the role of the Rozvi in the rise and spread of Khami-phase sites (A.D. 1400-1830) situated east of Khami ruins remains uncertain (Beach 1980;Machiridza 2012;Van Waarden 2012). Furthermore, the foggy Torwa past continues to constrain full archaeological interpretation of the Khami phase. Historians further compounded this challenge by postulating that the Torwa were descendants of a late 15th-century Changamire governor who rebelled against the Mutapa state ( Mudenge 1974Mudenge , 1988. In this regard, Khami-phase origins became entangled with the Mutapa state, Great Zimbabwe, and Leopard's Kopje cultures (Caton-Thompson 1931;Robinson 1959;Garlake 1973;Van Waarden 1998, 2012Chirikure, Mukwende et al. 2017). These conflicting views simply illustrate the current academic confusion surrounding Khami-phase origins. As such, the sites of Danamombe, Naletale, and Zinjanja were targeted for study in order to ascertain their origins and diagnose evidence of changeover between the Torwa and Rozvi phases. In this context, the Rozvi are conceptualized as an ethnic group, hence the sites under study and the broader southwestern landscape are scrutinized for their ethnic symbols. Since Rozvi identities were constantly constructed by reflexive agents who relied on culture and landscape for ethnic markers, their symbols of identity varied extensively from visible to abstract things. Oral traditions, documentary sources, as well as several ethnographic and anthropological sources are explored in attempts to establish the way Rozvi identities were situationally constructed and objectified through space and time. Most importantly, the prominent Rozvi dynasties coalesced around southwestern Zimbabwe because these environments were endowed with numerous strategic natural and cultural resources that were key in shaping the progression of their dynamic state.

An Agency Landscape Framework
The quest to understand the relationship among people, space, time, and material culture led, during the 1920s, to the adoption by archaeology of the notion of landscape from geography. Initially, a landscape was visualized as merely signifying geographical regions existing for purely utilitarian purposes (Trigger 1990;Anschuetz et al. 2001). Such approaches tended to emphasize the physical features of the environment at the expense of indigenous philosophies. This reality has become particularly pronounced in Africa, where calls to replace Western philosophies in archaeological practice with African paradigms and epistemologies are becoming louder and clearer (Lane 2011;Manyanga and Chirikure 2017;Sinamai 2017). Lane (2011:9) actually argues that Western philosophies have compromised archaeologists' appreciation of space, time, being, and materiality through a strict emphasis on scientific approaches. Consequently, landscapes in Africa are increasingly being viewed through indigenous community lenses that perceive them as living, spiritual, and physical places (Ndoro 2001;Fontein 2003Fontein , 2008Sinamai 2017). Sinamai (2017:405) therefore succinctly defines landscapes as a biography of space containing natural features, architectural remains, artefacts, myths, legends, memory, and feelings fostering a strong sense of belonging.
Landscapes are sites of agency: they are the hubs of power, resistance, authority, and political realignment (Monroe and Ogundiran 2012). The materiality of landscapes (people, objects, rivers, monuments, and rainmaking cults, among others) is always politically charged (Fontein 2003(Fontein , 2008. People are mutually tied to their territories; hence the dialectics between persons and their environments embody numerous sociopolitical meanings. In the simplest terms, landscapes and people define each other: even associated social meanings are premised on this intricate relationship. This explains why landscapes are always characterized by multiple narratives, most of which conflict with each other (Ndoro 2001;Fontein 2003;Munro and van der Horst 2016). Tilley (2006:14) also observes that ideas and feelings about identity are always grounded in landscapes, but this does not just happen; it requires repeated work. This article subscribes to the theoretical view, although debatable, that identities by nature are abstract and dynamic: they are first played out in the mind, then later objectified through things and places by rational agents (Tilley 2006;Hand 2015). Almost always, identities are constructed through collective traditions and imagined in an historically and materially specific way. Considering that various identities have pros and cons under different circumstances of interaction, individuals carefully choose identities to employ depending on socially imposed limits (Schortman 1989).
Ethnicity was one of the most widely used cultural categories among population segments because it evoked strong feelings of common purpose and support among people fighting for economic and political gain. The term ethnicity is quite complex and difficult to actualize archaeologically (Hegmon 1992;Emberling 1997;Jones 1997;Calabrese 2005;Machiridza 2012). In southern Africa, both historians and archaeologists have been engaging with the notion of ethnicity from quite different perspectives. For instance, Ranger (1984) perceives ethnicity as an ideology invented by 19thcentury Europeans in colonial Africa. However, this constructivist approach to ethnicity deprives precolonial Africa of collective identity, which is problematic. Archaeologically, ethnicity in southern Africa has been tackled from both a primordial and instrumentalist/ interaction approach (Huffman 1974(Huffman , 1978(Huffman , 1980(Huffman , 2007Calabrese 2000Calabrese , 2005Pikirayi 2007;Machiridza 2012Machiridza , 2013. The primordial school of thought stipulates that ethnicity results naturally from bounded, static, conservative, homogenous, closed, or rigid social units, while the interaction approach posits that ethnicity relates to a fluid, dynamic, and ongoing process of self-definition. Initial studies of ethnicity in southern Africa were strongly conditioned by the primordial and culture-historical approaches, which emphasize issues of homogeneity, chronology, and definition of cultures. Ethnic groups were widely defined on the basis of ceramic typologies and their geographical distribution (Machiridza 2012:43). In this respect, ethnic groups are situationally constructed sociopolitical and economic units that rely on ethnic boundaries or cultural diacritics to define/mark social boundaries (Barth 1969;Stone 2003;Calabrese 2005;Pikirayi 2007;Machiridza 2012). In the past, ethnicity was widely defined through culture provinces that were equated to tribes, a trend that strongly tallied with the methodological principle that cultural continuity was synonymous with ethnic continuity. This also meant modern ethnic groups could be traced back to their primordial origins (Huffman 1974(Huffman , 1978(Huffman , 1980(Huffman , 2007Veit 1989;Jones 1997;Calabrese 2005;Pikirayi 2007;Machiridza 2012). Pikirayi (2007:290) therefore laments that southern African archaeologists largely failed to appreciate the complexity between group identity and material culture variation as demonstrated by Barth (1969), Eriksen (1993), andHodder (1982). Since ethnic groups are redefined by interacting members who relentlessly reflect upon their own traditions and culture in order to maintain social boundaries, their symbols of identity were ever fluctuating. This is how ceramics, monumental architecture, and landscapes are constantly manipulated as ethnic symbols (Hegmon 1992;Calabrese 2005;Tilley 2006;Pikirayi 2007;Lane 2011;Machiridza 2012).
According to Lane (2011), people always focus on the past, especially monuments and other physical reminders, in their construction of "self." Landscapes are therefore culturally fashioned creations that act as metaphors that connect people with their past. Landscapes are constantly manipulated by human agents in attempts to preserve memories; as such they anchor memories and transform them into something tangible (Tilley 2006;Hand 2015). Since all remarkable features of the landscape prompt commentaries as to why they exist, human agents often create spurious explanations for such features, and this process is called "iconatrophy" (Vansina 1985:10). Hence, stories, myths, and legends relating to landscapes are part of community memorybuilding processes. Vansina (1985:43) further elaborates that studies of memory emphasize that remembering is action premised on cueing and scanning mechanisms. Thus, every item that is memorialized by traditional societies is assigned a cue/label that is destined to recall a particular memory. This explains why landscapes, objects, and music feature prominently as "mnemotechnic devices" in oral traditions (Vansina 1985). In recent times, oral memory has become an integral part of archaeological research because it provides opportunities to understand the ways past and present communities experienced landscapes, as well as how they created and negotiated meaning (Jones and Russell 2012:274).
Different human agents' ongoing struggle to construct meaning and maintain identities, and own and control access within and between places keeps landscapes in a constant state of becoming. Nowadays, individuals who intentionally act in pursuit of prestige and power are widely conceptualized under agency theory (Dobres and Robb 2000;Dornan 2002;Joyce and Lopiparo 2005;Machiridza 2012Machiridza , 2013. Just like ethnicity, agency theory features quite prominently in archaeological literature, yet explicit definitions of what it is and how it manifests materially remain fragmentary (Dobres and Robb 2000;Dornan 2002). Agency theory relates to the interests and actions of past social actors in dealing with social constraints/structures, such as ideology, class, tradition, age, sex, and gender (Dornan 2002). In Africa, agency theory clearly resonates with Kopytoff's (1987) model of moving frontiers and the reproduction of traditional societies. Kopytoff (1987) noted that the political history of sub-Saharan Africa was characterized by expansion, segmentation, fission, and reconstitution of polities owing to internal, contradictory tendencies within societies. Disgruntled or ambitious members were always breaking away from the metropolis into the periphery/fringes or frontiers in order to establish their own autonomy. However, the frontiers were not always unoccupied, a scenario that prompted compromises by both "newcomers" and "first comers," i.e., the groups already inhabiting the area. It therefore follows that identity renewal became a common survival strategy for most ethnic groups that converged in the frontier zones.
Perhaps it was this incessant convergence of sundry agents within the same environments that eventually prompted the evolution of Khami-phase sites in southwestern Zimbabwe. In particular, Khami-phase origins are better understood as a culmination of efforts by numerous competing agents that sought to exploit favorable natural resources found around that landscape. The mineral rich "Great Dyke," its high rainfall and cooler temperatures; very fertile shallow to moderately deep, dark or reddish luvisols; and the easily workable shallow regosols and arenosols enhanced several lucrative economic activities. All these environmental factors stimulated precolonial gold and iron mining, cereal agriculture, cattle ranching, as well as local and international trade that eventually shaped the rise of Khamiphase sites. In addition, the young foliage of Colophospermum mopane, nutritious pods of Acacia spp., and the edible parts of shrubs of the Grewia spp., together with diverse grass varieties provided excellent feed for cattle and wildlife throughout the year. The southwestern watershed temperatures, ranging below 16°C, prevented the breeding of the tsetse fly (Glossina morsitans), the carrier of sleeping sickness (Trypanosomiasis), which kills cattle. Owing to these favorable environmental factors, southwestern Zimbabwe was always free from these insects, thereby permitting extensive cattle breeding and wealth accumulation for the local inhabitants (Burrett 1998; Van Waarden 2012).
Great Zimbabwe-period settlements were, however, closely restricted in these regions because of the dry climatic conditions that once prevailed between 1290 and 1475 (Van Waarden 2012:24). Conversely, the sudden climatic shift to wetter conditions from 1500 to 1675 stimulated the rise of several Khami-phase sites around these regions (Garlake 1978; Van Waarden 2012). Interestingly, Ellert (1993:22) also comments that Butua was historically known as the "Mother of Gold." From as early as the 13th century, several gold reefs in southwestern Zimbabwe were already being exploited by Zhizo, Mambo, and Woolandale communities (Summers 1971;Swan 1994Swan , 1996Swan , 2008. Therefore the prevalence of Leopard's Kopje cultures along greenstone belts indicates their dependence on ancient gold mining, trade, and livestock production (Van Waarden 1998, 2012Chirikure, Manyanga, Pikirayi et al. 2013;Chirikure, Manyanga, Pollard et al. 2014). It was the desire to exploit these gold-rich deposits that constantly attracted numerous communities to the same landscape (Swan 2008:37). It has also been suggested that the rise of Khami during the early 15th century was triggered by the rerouting of a trade route from the east-coast trading center of Kilwa via Great Zimbabwe to the new Ingombe IIede-Quelimane-Angoche trade route via the Cuama (Zambezi) River (Newitt 1972;Beach 1980;Van Waarden 1998, 2012. The increasing control of wealth acquired through gold and the Indian Ocean trade, coupled with cattle ranching and cereal-crop production, culminated in the evolution of the Khami-phase capitals such as Khami, Danamombe, Naletale, Zinjanja, and Manyanga in southwestern Zimbabwe. In the same manner, the Rozvi, who started off as a few ordinary Karanga people in northeastern Zimbabwe ended up drifting towards these lucrative regions to settle and conquer the preexisting Torwa communities. Thus, landscapes are widely perceived as contested cultural arenas that exhibit variable histories, even though they may appear unchanged over time (Anschuetz et al. 2001;Ndoro 2001;Fontein 2003;Tilley 2006). Most often, the annexation of contested space by the victor changes the way "first comers" interact with places because new rules of access and engagement are often imposed by those assuming power (Kopytoff 1987;Ndoro 2001;Fontein 2003). These trends often constitute the materialization of power, political ideology, or ethnicity because the most strategic or prominent landscape features are appropriated as political/ethnic symbols. Archaeological literature has also focused on ceramic style as encoding cultural messages peculiar to group identities, including ethnicity. These works argue that there is a positive co-relationship between group identity and ceramic design (Hegmon 1992;Bowser 2000;Calabrese 2005;Pikirayi 2007). Pikirayi (2007:288) further posits that design styles are repetitive and integrated codes of cultural symbols that are learned and transmitted within groups. This is why these designs cut across ceramics and other categories of material culture, like wooden artifacts, headrests, drums, iron furnaces, traditional grain silos, stone walls, monoliths, and even female bodies (Bent 1892; Robinson 1965;Huffman 1989Huffman , 1996Ndoro 1991Ndoro , 1996Pikirayi 2007).
Ethnic symbols therefore crosscut other symbols of class, land fertility, gender, initiation, and sacred leadership. Although ethnic meanings are variably imbued on numerous landscape objects and features, it is impossible to objectively define all these symbols archaeologically. For this reason, the archaeological record, in tracing ethnicity, needs to be complemented by relevant anthropological and historical data. Archaeologically, style is often associated with group identities, and it is defined as the general appearance/design of an object or structure that results from the manufacturing process (Hegmon 1992; Jones 1997). Style encompasses those attributes with little or no bearing on the function of objects or structures (Wiessner 1983(Wiessner , 1985Sackett 1990;Hegmon 1992;Jones 1997;Bowser 2000). According to Calabrese (2005:61), emblemic style is widely equated with ethnicity, while heterochrestic style relates to class distinctions. In this regard, Calabrese (2005:67) confines ethnic symbolism to ceramic style and settlement layout, and heterochrestic style to elite goods and settlement layout. Although this proposition is acceptable, one can also argue that elite goods can also double as ethnic symbols because ethnicity was always situationally defined. The subsequent section concisely lays out the archaeological background of Khami-phase origins.

Khami-Phase Origins Debate
The Khami phase marks the third phase of the Zimbabwe culture, an archaeological term widely used in southern Africa to describe shifting centers of power that exhibit class distinction through the prestige construction of dry-stone walling, display of a spatial organization separating rulers from commoners, and depiction of evidence of craft specialization, external trade, and intensive cattle herding (Huffman 2000(Huffman , 2005(Huffman , 2007Chirikure, Manyanga, Pollard et al. 2014). The Zimbabwe culture has been arbitrarily divided into three phases based on the dominant capital, technique, and style of walling, as well as associated ceramic styles. Associated archaeological sites have been broadly classified under Mapungubwe (1220-1290) and Great Zimbabwe (1300-1550), which encompasses the Mutapa (1450-1900) and the Khami phase (1400-1830). Under this framework, Mapungubwe was the first state capital, followed by Great Zimbabwe, the Mutapa, and then the Torwa/Changamire states. Recent data are, however, challenging this linear model in favor of a multilinear one, where sites like Mapela and Khami are now considered as independent developments (Van Waarden 1998, 2012Chirikure, Manyanga, Pikirayi et al. 2013;Chirikure, Manyanga, Pollard et al. 2014;Chirikure, Mukwende et al. 2017). It was once thought that, upon the demise of Great Zimbabwe around 1450, some of its dynasties migrated north to found the Mutapa state, while others moved southwestward to found the Torwa state (Garlake 1973;Burrett 1998;Pikirayi 2001). As is the case with most migrationist models, the archaeological record does not support such assumptions.
In fact, both ceramic sequences and architectural typologies recorded from Great Zimbabwe and Khami contrast with each other (Van Waarden 2011, 2012Chirikure, Mukwende et al. 2017). Great Zimbabwe dates from as early as 1300 right up to the late 17th or 18th centuries, while Khami flourished between 1400 and 1650 (Chirikure, Mukwende et al. 2017). While Great Zimbabwe-tradition sites composed of freestanding dry-stone walls that formed circular to oval enclosures, concealing the ruling elites within, the Khami phase is mainly composed of dry-stone retaining walls supporting terraced platforms that exposed elite houses to public view. Khami-phase sites also contrast with Great Zimbabwe-tradition sites through check designs on dry-stone walls and profusely decorated polychrome wares. These contrasting features signify that these were two separate, but overlapping, cultural entities (Chirikure, Mukwende et al. 2017 Since Khami was the first major elite site to emerge from an advanced Leopard's Kopje chiefdom, the archaeological term "Khami phase" has gained prominence over the years in reference to sites of a similar kind (Huffman 2007;Van Waarden 2012). While extensive work has been done around Khami-phase sites in the Shashe-Limpopo Basin (Huffman 1978;Huffman and Hanisch 1987;Loubser 1989), northeastern Botswana (Van Waarden 2012), and Khami (Robinson 1959;Mukwende 2016;Chirikure, Mukwende et al. 2017), a lot more remains unknown about other sites situated east of the Khami capital or in the Rozvi core area. This has been partly perpetuated by the general assumption that, after the 1644 civil war at Khami, power shifted eastward from Khami to Danamombe. Hence, the succeeding section delves into the historical context for these Khami-phase sites.
The Torwa and Khami-Phase Sites: An Historical Perspective The first known people to be associated with Khamiphase sites were the Torwa/Togwa. Their historical origins largely remain vague owing to the scanty references stemming from Portuguese sources and oral traditions. Nonetheless, the Torwa are generally traced back to the 1490s, when a certain Changamire rebelled against the Mutapa state (Mudenge 1974;Randles 1979). By 1494 he had been subdued by Mutapa Mukombwe and forced to drink poison as punishment for his actions. Following the death of Changamire, his lands, located somewhere south of the Mutapa and variably called Guruuswa, Gunubugwa, or Kunauhwa, remained independent, but one of his sons continued fighting the Mutapa from there until 1548 (Beach 1980). The Portuguese described Guruuswa as "Butua," possibly for ease of pronunciation. By the early 16th century, the Toloa/Torwa were widely referenced as rulers of Butua, territories that coincide with Khami-phase sites (Randles 1979). Basal dates for some of these sites suggest that by 1400 or the early 15th century, they were already on the rise (Huffman and du Piesanie 2011:198). Although Torwa dynasties only appeared in documentary sources 90 years after some of the Khami-phase sites were built, they continue to be considered as the true site founders. However, if Khamiphase sites predate the Changamire rebellion against the Mutapa state, then crediting Torwa dynasties as the original Khami-phase founders is problematic.
Although it is known that Changamire's son got support from his Torwa relatives to fight against the Mutapa, who exactly these Torwa people were is still a mystery. Several propositions about their actual identity have so far been made, but none of these is really convincing. For instance, Garlake (1973) defined the Torwa as relatives of the second Mutapa who ruled the Mbire/Guruuswa province together with Changamire. Randles (1979:20) also suggests that the name Torwa implied vassalization of Khoisan people in Butua by the Mutapa Karanga people. Van Waarden (2012:51) further postulates that the Torwa were synonymous with Kalanga people who trace their origins back to Madabhale, whose "praise name" was Chibundule. On another note, Von Sicard (1948:13) argues that the Twamamba/Torwa were the original inhabitants of Butua, who should never be traced from the Mutapa state. Whatever the case might have been, the Torwa eventually became the main dynasties governing southwestern Zimbabwe. Unfortunately, there are no reliable oral traditions for the Kalanga people, who are also widely perceived as the original inhabitants of southwestern Zimbabwe since the 1400s (Beach 1980:218). Most Kalanga traditions do not even recall the Torwa; rather, they cite people like Ngelengele; a man and his sister, Nenumbata and Nemushodzi; the Nyubi under Sibuntuli; and a Kalanga Ndumba (Beach 1980). To date, only two oral account sources have mentioned the Torwa: a spirit medium of Mutota in 1958 and a Rozvi chief Mbava in 1925.
In 1925 Elaine Lloyd interviewed Chief Mbava, a Rozvi descendent from the Wedza district, who by then was over 100 years old. Following the first interview with Chief Mbava, Lloyd (1925:63) reported: Chief Mbava thinks the great chief of the Warozwi, called Togwa, was the one who built Zimbagbwe. He said the word means Palace of the Kings (Dzimbahwe). ... Togwa was a very great chief indeed, and all the chiefs went to him in Zimbagbwe to pay homage. ... King Togwa did not build Zimbabgwe only; there are many other places like Zimbabgwe which have never yet been seen by the white men. ... Besides all these places built by the Warozwi, King Togwa sent a great army of people to bring him the great hill Mutikwiri. This hill is in the country of Mawungwe (near Rusape), and he sent his people to bring it and place it near the Palace of the King (Zimbabgwe). Now, therefore, this great army of King Togwa surrounded the hill and dug around it, but failed to move it.
It is apparent that there are a lot of mix-ups and inconsistencies regarding the Torwa past. Nonetheless, what emerges is that the Torwa were of the soko (monkey) totem and they shared historical relations with both the Mutapa and Rozvi people (Beach 1983:16). Although their actual origins remain unclear, they probably predate the Mutapa state or were contemporaneous. Oral histories also emphasize that the Torwa were the original builders of most stone buildings in Butua, and that the Rozvi later invaded Torwa core territories. On a related note, although the Khami-phase sites under study formed four major clusters that coalesced around the capitals Danamombe, Naletale, Zinjanja, and Manyanga, this spatial configuration was never directly shaped by Rozvi ethnic influences. Actually, this clustered spatial patterning is better understood through the framework of Shona rotational succession systems.
The model of Shona rotational succession systems stipulates that several historical state systems, such as Mutapa, Torwa, and Rozvi, were founded by individuals who came from specific dynastic lineages (Beach 1980;Mudenge 2011;Chirikure, Manyanga, and Pollard 2012;Chirikure, Mukwende et al. 2017). These dynastic founders often appointed their close kin as provincial, district, and ward heads. When the dynastic founder died, one of his cunning sub-rulers often took over as king and ruled from his own (provincial, district, and ward) center, which he successively developed into the new metropolis. As such, the recurrence of this Shona rotational succession system gradually led to a series of overlapping centers of power. Consequently, a series of broadly similar settlements that were sometimes clustered or dispersed gradually scarred the Zimbabwean landscape (Mudenge 1988(Mudenge , 2011Chirikure, Manyanga, and Pollard 2012;Chirikure, Mukwende et al. 2017). Therefore, the spatial distribution of Danamombe, Naletale, Zinjanja, and Manyanga is better understood as an outcome of this Shona rotational succession system. It now suffices to reconnoiter the Rozvi and their historical associations with Khamiphase sites.

Trajectories of Rozvi Origins and Development
Prior to the mid-18th century, the Rozvi were simply nonexistent. This means they were historically constructed from the ordinary Karanga, Korekore, Kalanga, Zezuru, Ndau, Nambya, and Venda Shona subethnic groups. The death of a certain Portuguese missionary in the Mutapa state between 1600 and 1633 triggered a great deal of Portuguese interference in African affairs. By 1663, King Kamharapasu Mukombwe partially reversed this when he revived the Nyai client system, which drove the Portuguese away (Beach 1995;Mazarire 2009). As from the 1670s, Changamire, one of Mutapa's provincial barons, was building his own army through the Nyai client system using Mukombwe's cattle. Around the 1680s, Changamire Dombo was well advanced, so in 1684 he rebelled against the Mutapa (Beach 1980;Mudenge 1988). When Mutapa Mukombwe attacked Changamire, he retaliated by killing almost 5,000 Mutapa soldiers (Mudenge 1988). In June of the same year, he once again clashed with the Portuguese at the Battle of Maungwe (Mudenge 1988:286). He then lay low between 1685 and 1692. When Mukombwe died, the Portuguese preferred his son Nyamaende Mhande (Prince Dom Pedro) to assume the throne ahead of his uncle Nyakunembire (Mudenge 1988:287).
This offended Nyakunembire, who appealed to Changamire for help. Changamire Dombo quickly responded by attacking the Portuguese fort of Dambarare in 1693 before expanding the conflict to Mukaranga, Ruhanje, and Manyika in just two years. It was these successful military campaigns that eventually earned him and his followers the nickname varozvi (meaning the "destroyers"). Changamire Dombo later left northeastern Zimbabwe for Butua because the former territories were too close to his enemies and ravaged by wars, famine, and disease. His actual conquest of the Torwa in Butua dates between 1683 and 1696, and it is said to have gone smoothly (Beach 1980;Mudenge 1988;Pikirayi 2001). As such, what initially began as an ordinary descriptive nickname (varozvi) continuously shifted in meaning through time to accommodate and exclude other members. The name Rozvi eventually became an ethnic name. Hence, Rozvi elites began to actively exploit the name to claim territorial space. This explains why, in 1758, Ignacio Caetano Xavier reported that Changamire's lands were populated by the "Borobzes," meaning the Rozvi. Another anonymous Portuguese manuscript dated to the end of the 18th century also highlighted the following: "Urobze lies a long way from Manica, it would take a month to reach it, it is said to be nearer the Cape of Correntes. The Africans call it Goromucuro; it lies to the west of Manica. The Kingdom abounds in the rolling veld plains [campinas vistozas planices]" (Randles 1979:16-17).
The above description of the so-called Rozvi lands befits Guruuswa/Butua, which they took over from the Torwa. Ultimately the name Rozvi replaced Guruuswa/ Butua as toponym. In line with this, Guyot and Seethal (2007:3) note that the politics of name changes cannot be separated from conflicts and tensions over political and cultural hegemony. Mashiri (1999) reinforces this aspect of identity construction through place naming by stating that, among the Shona, names are important vehicles of sociocultural communication because they are used to reflect power relations. Unfortunately, the Rozvi name has always been taken for granted by most archaeologists because of the general belief that names cannot be objectively traced archaeologically. However, names are more than just labels, and they are never neutral; in fact, they are powerful vehicles of promoting identification with place and the past. In line with this, Krogseth (2012:164) posits that the named ancestral home gives space for shared memory, both in mental/ symbolic and material terms. In this case, the Rozvi actively used their name as a tool to create a lasting connection with the landscape.
The Rozvi also used mythology to forge strong ties with the landscape. Sinamai (2015) notes that myths and legends signify metaphors that symbolize historical connections between people and places. Tilley (2006:25) also adds that myths incorporate historical events and make sense of them through local understandings; they rework time so that past and present exist concurrently in relation to a desired future. Myths therefore represent the cultures that they serve; they communicate social values and memories in a special language that is only understood by community members and not outsiders. According to Sinamai (2017:407), myths and legends appropriate the landscape into community behavior, beliefs, and being; they create a body of knowledge about the landscape and environment in the minds of community members. Therefore, in order to fully understand the Rozvi, close attention must be paid to myths. Only a few examples are highlighted below in order to show how they connected symbolically with landscapes. Chief Mbava narrated one Rozvi myth as follows: He built his royal town near 'The Nameless Tree', near Marandellas. When he died, his son Munyepere was made king. ... King Munyepere had a very large army. It was he who told his men of war that he desired them to prepare ladders which should reach up to heaven, so that they could catch the moon in order that the king might make it his 'Ndoro' (a round white shell ornament) with which to adorn himself. So the whole army began to cut trees and to make ladders, and they built and they built, going towards heaven, for two to three years. By and by the ladders began to rot. There were many people building and climbing up the ladders with trees and bark rope, and the ladders broke, and a great multitude fell down and were killed. ... By and by King Munyepere died and his son Togwa was chosen to be king. It is King Togwa who, people think built Zimbabgwe and tried to dig up Mt. Mutikwiri and Mahwesimike. On the death of Togwa, Chiduku was made king. (Lloyd 1925:91-92) Through mythology, the Rozvi managed to inscribe their identities on a much broader landscape. Posselt (1935:140-143) highlights a related Rozvi tale: The Barozwi kings displayed great vanity during this period of migration and on various occasions they were filled with the desire to possess exalted "stools" or thrones. To gratify that desire they ordered their subjects to remove bodily massive hills to serve this purpose. Among others Mt. Hurungwe, a granite pinnacle of nearly 800 feet in height in the Mrewa District, is still pointed out as one so selected; the weathering of the rock is believed to be the marks of the implements of the men who tried to remove it. ... Thus Lembewu became king and they carried him to Manyanga, where Chihunduro and Tumbare were chiefs, both laying claim to a hill each, as a symbol of their royalty. The Mambo ordered Nerwande to collect earth and pile it up until it grew higher than the hills of Tumbare and Chihunduro. They piled it up as ordered. Then they took elephants' tusks and built steps to the top--hence the name "Manyanga." The myths above highlight Rozvi power and capacity to build monumental architecture scattered across the Zimbabwean plateau. Through the linking of such myths, the Rozvi managed to claim outright ownership of several dzimbahwes (monuments) that they never really built or of which they had only played a minor developmental role. Monuments were traditionally perceived as the ancestral homes of most Shona people; hence the Rozvi easily gained spiritual authority over many people who revered such places as sacred. This resonates with Sinamai's (2017) observation that, when monuments feature profusely in myths, they signify centers of life connected to a community's sources of water and food. A few other myths emphasize close Rozvi relations with Mwari (God): One of the remarkable features of the Rozi story lies in the personal nature of their relationship with Mwari. How insistent he was on their acknowledging that they lived, ruled and conquered only by his favour and help! Even the magic horns lost their power when this was withdrawn. Any hubris or trust in their own unaided strength led to the immediate withdrawal of his favour and consequent disaster. (Fortune 1956:69) Hodza and Fortune (1979:217) also narrated another Rozvi myth as follows: Their traditions tell us that they were accompanied on their travels by a voice which they called Tovela which led them on their way, keeping them safe from dangerous places. The voice could speak from any object it chose and we are told that it spoke through grass or trees or through a little child, unable as yet to speak for itself. The voice of a presence whom some say was that of the first Rozvi ever created, the founding father of the clan. Others say he was the first person ever to be created and that after his death his spirit accompanied the different groups from Guruuswa to the countries they chose. Tovela not only protected his people on their travels but gave them food as well. When they clapped in homage and supplication to him at a tree, portions of porridge, pots of milk, and combs of honey would come out of the ground. He also gave them medicines to become invincible but not inaudible to their foes and to be long-lived. Tovela came to be called by other names by different groups such as Mwari.
The Rozvi therefore initiated numerous customs of sacrificing to Mwari, especially from stone-walled sites, and this strengthened their claims of having built several such structures in Mbire/Guruuswa territories. In view of the brief historical discussion above, the relationship between the Rozvi and landscape was intimate and spiritual. Their identities were intertwined with the natural world, where trees, grass, rocks, mountains, and dry-stone-walled monumental architecture were strategically objectified as their symbols of identity. The next section describes each selected Khami-phase site and the excavations conducted thereof.

Danamombe/Dhlodhlo (19°56′48″ S, 29°19′54″ E)
One of the prominent Khami-phase sites historically entangled in Rozvi narratives is Danamombe. Its actual origins and the role of the Rozvi in its development remain uncertain, although there is a general belief that Danamombe succeeded Khami as the next Torwa state capital (Beach 1980;Huffman 1996;Pikirayi 2001;Van Waarden 2012). Danamombe is about 86 km southwest of Gweru and 92 km northeast of Bulawayo. While Danamombe covers 1.2 km 2 , constituting about 700 ha of settlement debris, the main platforms occupy just 1.21 ha. Danamombe consists of two large subrectangular artificial platforms separated by a parallel passage. About six enclosures of inferior quality define the southern and southeastern sections of this main complex, while the northern site limits have a few other platforms and enclosures. The extreme western side of the main complex is marked by semicircular, freestanding low walling with isolated bands of ironstone decorations in some sections. Previous excavations at this site uncovered a few datable imports, such as blue-and-white porcelain, blue-andwhite Nankin porcelain, and a Dutch gin bottle, that date to the late 17th and early 18th centuries (Randall-McIver 1906;Caton-Thompson 1931). These relative dates, along with the general site size, frequency of freestanding walls, and the nature of retaining walls exhibiting a greater variety of decorative designs, have been used to argue that this site is more recent than Khami. It remains difficult, however, to tell whether Danamombe started off as an extension of Khami or developed independently. Since both Torwa and Rozvi dynasties ruled from this site, several scholars have also questioned whether this political shift was materially expressed (Beach 1980;Machiridza 2012;Van Waarden 2012). The site will remain shrouded in mystery if radiocarbon dates and associated material culture are not systematically analyzed.
A 2 × 1 m Trench I, trending north-south, was laid across the middle lower platform overlooking the western courtyard (Fig. 2). This platform was excavated to retrieve securely stratified cultural and datable material. The finds were expected to address issues relating to settlement chronology and human economic activities, as well as group identities. Test Pit I was sunk 7.5 m north of the western courtyard wall on a midden trending east-west and measuring approximately 13 m wide by 22 m long. The aim was to investigate the outer elite zones and complement Trench I cultural material. Test Pit I yielded very little, so it is not discussed as part of this article. Test Pit II was located on another midden about 300 m northwest of the main platform, and it was very productive. It was excavated to recover datable material that could be used to ascertain cultural sequences at Danamombe.
For reasons of stratigraphic control, excavations proceeded by arbitrary 10 cm spits. The excavated soil was dry sieved, using both a 2.5 and 2 mm wire mesh in order to trap differently sized cultural materials. The full stratigraphic profiles of Trench I and Test Pit II are shown on Tables 1 and 2, while the distribution of selected charcoal samples is illustrated in Figures 3  and 4, respectively.

Naletale (19°53′18.38″ S, 29°31′55.49″ E)
The other conspicuous Khami-phase site widely mentioned as Rozvi in documentary and oral traditions is Naletale (Randall-McIver 1906;Huffman 1996;Burrett 1998). Just as with Danamombe, the archaeological development of this site in relation to both Torwa and Rozvi narratives remains speculative. Naletale is on a high granite kopje (hill) that rises 1465 m above sea level. It is 25 km southeast of Danamombe and 67 km southwest of Gweru in the Insiza district. The main complex at Naletale is an elliptical enclosure with two entrances, the main one facing northwest, while the other one at the back faces east. Two retaining walls form two terraces on the outer western frontage, and, of these two, the upper one is profusely decorated with chevron, herringbone, check, and isolated bands of ironstone. This west-facing upper girdle wall is also crowned by four turrets with monoliths. Internally this enclosure has a maximum diameter of 53 m, which is divided into seven sections by freestanding radiating walls emanating from a northern central platform, as illustrated in Figure 5.
Its northern and northwestern sections are characterized by a small enclosure and a few artificial platforms that were probably foundations of single huts. Based on reports of previously recovered finds, Naletale was dated to between the 17th and early 18th centuries (Burrett 1998; Van Waarden 2012). As is the case with Danamombe, the cultural sequence of this site remains speculative. Owing to previous site disturbances, only a single test pit (Test Pit II) was sunk on the northwestern platform (Fig. 4). The aim was to investigate the internal dynamics of this platform and to recover securely stratified datable material. By extension, these cultural materials were expected to broaden an understanding of past cultural and economic activities at the site. The same excavation strategies used at Danamombe were applied to this site. The stratigraphic profile and charcoal samples from Naletale Test Pit II are illustrated in Figure 6 and Table 3. 2 (10-20 cm) Soil maintained the same color although it was less compacted. Cultural finds included another brown porcelain fragment, faunal remains, potsherds, and a solid hut floor. Hut Floor 1 was well preserved and it extended into Level 3.

(20-30 cm)
About 24 cm below the surface appeared Hut Floor 2. A few potsherds, bone fragments, glass beads, and charcoal samples were recovered. Immediately below this floor at 27 cm was another Hut Floor 3.

(30-40 cm)
Compacted reddish-brown clayey soil continued into this level. New Hut Floor 4 appeared 35 cm below the surface, and it rested on packed granite blocks that were mixed with a few potsherds, a single glass bead, a shell bead, and a small bone fragment.

(40-50 cm)
Patches of fine-textured, charred thatch and ashy gray soil appearing approximately 43 cm below the surface. Several charcoal samples were collected for dating from this level. This grayish ashy layer lies above a reddish-brown sandy loam layer that also extended into large well-packed granite blocks. Finds included shell and copper beads, wound copper wire, faunal remains, and potsherds.
6 (50-60 cm) Reddish-brown sandy loam soil extended into well-packed granite blocks. A glass bead, charred shell beads, wound copper wire, and a potsherd were recovered.

(60-70 cm)
Grayish-brown soil mixed with irregular granite blocks. Finds were glass and shell beads, potsherds, a lower grinding stone, an iron fragment, and a charcoal sample.
8 and 9 (70-90 cm) Packed granite blocks of irregular size made it difficult to maintain the 10 cm spits. Finds included glass, shell and copper beads, faunal remains, an iron fragment, and charcoal.
10 (90-100 cm) Granite blocks continue into this level mixed with soil and finds including charcoal, potsherds, faunal remains, and glass, copper, and shell beads.
11 and 12 (100-120 cm) Gritty, ashy gray and reddish-brown soil mixed with large charcoal samples and ash suggest heavy firing once occurred above dhaka Hut Floor 5. Several faunal remains, wound copper wire, potsherds, a prefabricated spindle whorl, and glass, shell, and copper beads were recovered.
14 and 15 (130-150 cm) Ashy dark layer continues downwards and gives way to a compacted reddish-brown Hut Floor 6. Finds included shell and copper beads, faunal remains, and potsherds. Well-packed irregular granite blocks mixed with gritty light brown soil. A single shell and two copper beads were recovered.

(200-210 cm)
The same profile noted above continued to this level. A few potsherds and charcoal samples were retrieved here.

to 33 (210-340 cm)
Well-packed irregular stones mixed with little soil. Two shell beads were recovered at 277 cm. At 325 cm, possibly a few meters above bedrock, Trench I became extremely dangerous to work, so it was abandoned.

Zinjanja/Regina (20°08′00.02″ S, 29°21′34.98″ E)
Zinjanja is the other major Khami-phase site situated in core Rozvi territories. This site is also variably assigned to both Torwa and Rozvi dynastic histories because primary archaeological data relating to its detailed chronology remain scanty (Huffman 1996;Van Waarden 2012). This site is situated about 35 km southwest of Danamombe and almost 1315 m above sea level. The Zinjanja main complex is semioval in shape, 2 (10-20 cm) Compacted grayish soil mixed with faunal remains, potsherds, and iron fragments.

(40-50 cm)
Grayish-brown soil deposited on a soft gray ashy layer. Finds were potsherds, faunal remains, shell and copper beads, as well as charcoal samples.

(100-110 cm)
Soft grayish-brown soil. The only find was a single charcoal sample located just above the sterile layer. constituting limited Q-style 1 walling with chevron, separated herringbone, check, and cord decorations on the second tier. The basal walls defining the southeastern and southwestern sections, which are now collapsed, were probably built in the same Q style. The first two basal platforms are much wider, but the succeeding platforms gradually narrow toward the top like a wedding cake. The two upper tiers built in R style rise to over 6 m in height. The greater part of Zinjanja has Rstyle retaining walls, which are quite distinct in the eastern, northeastern, and northwestern sections. The main entrance faces the northeast and is adjacent to a small vandalized drainage hole. As the ground slopes down toward the west and northwest, lower retaining walls of the main platform become R-style freestanding walls that form an enclosure of some sort. Just as with Danamombe, the western end of the Zinjanja main platform is abutted by a wide courtyard wall of restricted height and inferior quality. The immediate northeastern sections of Zinjanja are characterized by numerous platforms. Numerous platforms and enclosures, however, extend northward into the adjacent mountain and along the western slopes for almost a kilometer. Intensive foot surveys combined with a series of shovel test pits were conducted around Zinjanja to locate potential areas for excavation. A total of five excavation units were then sampled, as illustrated in Figure 7. Although Trench I was set up on the main platform, it never yielded anything of chronological interest. The same held true for Test Pit I, sunk on a platform immediately outside the main platform. A midden situated approximately 50 m north of the main platform was then excavated to trace site origins and chronological sequences through Test Pits II and III. Since Test Pits II and III were the most productive, their stratigraphic profiles are presented in Tables 4 and 5, while the analyzed charcoal samples are illustrated in Figures 8 and 9.

Chronology and Cultural Affinities: A Comparative Approach
Owing to budgetary constraints, it was not possible to analyze all retrieved charcoal samples. Therefore, efforts were made to balance the 10 charcoal samples in relation to all excavation units. Danamombe, which had the deepest Trench I and a very productive Test Pit II, had a total of five samples, while both Naletale and Zinjanja contributed the other five samples. These samples were sent to DirectAMS Radiocarbon Dating Service in the United States for analysis, and the results are illustrated in Table 6.
The radiocarbon dates neatly show the chronological development of each site. In addition, the dates from Danamombe Trench I are quite interesting in the sense that the site has always been dated to the late 17th century. In order to ascertain the age of this site, an additional four samples were sent to the same lab, and results are shown in Table 7. 1 Whitty (1961) once developed a dry-stone-wall architectural sequence, where P (poor coursing) was the earliest, followed by PQ (mixture of poor and quality coursing), then Q (neat coursing), and finally R (rough/irregular coursing). This sequence was later refuted after the realization that stone walls never followed any chronological order. However, the P, PQ, Q, and R terms were retained for descriptive purposes.

Ceramic Analysis and Results
The fact that style is widely created and learned through close social networks and verbal communication makes ethnicity central to material culture production patterns (Huffman 1978(Huffman , 2007Plog 1983;Gosselain 2000;Pikirayi 2007). In this respect, the ceramics from the three research sites were subjected to multidimensional system analysis. Owing to the fragmentary nature of the potsherds, it was difficult to establish design layout, so vessel profile and motif combinations were prioritized in establishing ceramic typologies. The analysis also targeted other variables, such as lip form, fabric, and exterior color, because at times ethnic signatures are embedded therein. It turned out that the assemblages from the three research sites are represented by the vessel forms illustrated in Figure 10. These ceramic assemblages were further categorized into types and subtypes, as shown in Table 8. In total, 120 vessels were analyzed, and the majority were from Danamombe, which constituted 55.2% of the total, followed by Zinjanja with 31.1%, and then Naletale with just 11.4%. The most common vessel forms were globular constricted pots with rolled rims belonging to Classes 4 and 4a, of which Danamombe had the highest percentage, recording 21.7% for both classes. Zinjanja followed with only 5.8% and 4.2% of Class 4, while Naletale cluster sites just had a meager 1.2% of Class 4a. Class 2, plain, spherical, constricted pots with inward-sloping rims, was the second most common form, constituting 15% of the entire assemblage; however, this form was slightly higher in percentage at Zinjanja at 9.2%, while Danamombe only had 4.2%. The rest of the classes are not worthy of comment because of their very low frequencies. The three ceramic assemblages were essentially plain, as most pots were simply polished; only a few were graphite burnished and/or painted with red ocher (Fig. 11). At Danamombe, 49.4% of the pots were simply burnished or polished, while 43.1% had graphite, and 5.8% had red ocher burnishing. Only 1.7% of the analyzed fragments had engravings and polychrome designs. The 2 (10-20 cm) Light gray compacted soil changing to a dark gray tone. Faunal remains, potsherds, and a charcoal sample were recovered.
6 (50-60 cm) Fine-textured, light gray ash layer. Finds include faunal remains, a shell bead, and potsherds. 7 (60-70 cm) Light gray, fine-textured ash deposited on a gray compact layer. Faunal remains and a few potsherds were the only finds.
same design trend held true for Zinjanja, where 60% were simply burnished, 32.8% were graphite burnished, and 3.8% had red ocher slip. Sherds with applied designs constitute a mere 2.3% of the entire assemblage at Zinjanja. At Naletale cluster sites, a greater part of the assemblage was simply/graphite burnished, 42.9%; incised designs, 23.8%; and punctates 14.3%.

Bead Analysis and Results
So far, beads have been explored in terms of their original source areas, regional chronological sequences, implications for trade networks, technology, and symbolic meanings (Tapela 1995;Kinahan 2000;Bvocho 2005;Thondhlana 2005;Wood 2005Wood , 2011Robertshaw et al. 2010). Glass, shell, and metal beads were examined to apprehend issues of chronology and ethnic symbolism. The total number of glass beads recovered from the three sites amounted to 22. These were chemically examined using an x-ray fluorescence machine and results are illustrated in Table 9.
Most glass beads could not be clearly identified, although a majority were of European origin; these were followed by Khami series beads. However, the analyzed sample did not reflect any clear-cut bead color and shape patterns. It is notable that all the glass beads belong to the period between the 15th and 20th centuries. Although most of these glass beads came from Danamombe, the isolated beads from the Zinjanja and Naletale sites suggest that the sites were fairly recent and contemporary. Numerous shell and brass beads were also recovered and analyzed, but these did not show any salient stylistic patterns attributable to chronological or ethnic influences. Since symbolic or ethnic meanings were situationally constructed, Rozvi influences on ornaments shall be inferred from relevant historical and anthropological sources in the succeeding section.
The Rozvi: An "Imagined Community" The Rozvi past conflicts with Terence Ranger's (1984) argument that ethnicity was nonexistent on the Zimbabwean plateau prior to colonialism. In fact, extensive Rozvi citations in oral accounts collected from the Shona subethnicities that Ranger (1984) explored signifies their ethnic-invention processes. Rozvi identities   (Machiridza 2012). Since the Torwa were politically assimilated by the Rozvi, their entire landscape subsequently became Rozvi through the manipulation of naming patterns (Randles 1979:16-17). It therefore follows that the Rozvi exploited their ethnic name and multiple sources of power to forge lasting ties with monumental structures and the wider southwestern landscape. The Rozvi also relied on myths, which are subject to fiction, including the mixing and confusion of times and places and events to sustain their identities. Their survival in oral traditions suggests historical consciousness and contemporary mentalities. Thus, mountains, hills, wooden towers, trees, and grass plains that feature prominently in Rozvi narratives signify their role as mnemonic devices. In this context, Ndoro (2001:72) stipulates that Africans hardly separate humanity and nature in their perception of the world. As such, granite boulders, rockshelters, and forests always coincide with Shona sacred places.
Therefore Rozvi sociopolitical ideologies relied heavily on the landscape for ethnic objectification. However, radiocarbon dates acquired from Danamombe, Naletale, and Zinjanja revealed that Rozvi oral traditions only spoke about the very recent archaeological layers. The dates acquired from Danamombe Trench I (Level 16) indicate that by cal. 1420 or 1435 the site was well advanced. As such, its rise had nothing to do with the 1490 Changamire breakaway from the Mutapa state or the presumed dynastic migrations from Great Zimbabwe in 1450. Danamombe actually developed as a competing center of power well before Khami collapsed. While Khami was thriving in the west, Danamombe was equally vibrant in the east. In this respect, Rozvi dynasties only inherited a site that had been long established during the early 15th century, probably during the Torwa period of governance. Interestingly, the heavily burnt context of Danamombe Trench 1, Levels 11-12 and associated radiocarbon dates suggest a brief conflict ensued when the Rozvi took over. The other more recent date from Danamombe Trench I (Level 21) most probably trickled downwards during the course of excavations. The dates from Naletale indicate that this site was contemporary with Danamombe and built over a very short period. Clearly, Naletale and Danamombe evolved well before the Rozvi emergence and arrivals in the region. Zinjanja is one of the major Khami-phase settlements closely connected to the two sites mentioned above. Its main platform also promised great potential to yield cultural material for comparative study. Unfortunately, it was the nearby midden that yielded datable samples used in this study. While Van Waarden (2012:105) postulates that Zinjanja could be older than Danamombe, this view is highly unlikely. In fact, dates from Test Pits II and III indicate that Zinjanja flourished between ca. 1680 and 1926. Of course, these dates need to be handled with care because a similarly placed midden at Danamombe yielded dates much more recent than those of the main platform. It is known, however, that a date in the range of 1649-1673 was once acquired from a structural hardwood post at the Zinjanja main platform (Van Waarden 2012:100). Going by the available data, it is clear that Zinjanja overlapped temporally with both Danamombe and Naletale. Even ceramics from all three sites demonstrated an unbroken Khami-phase sequence. The most common vessels were constricted globular pots with predominantly plain or polished surfaces; just a few were graphite burnished and painted with polychrome designs. While it may appear that the transition from Torwa to Rozvi was not buttressed by ceramic stylistic changes, there is need to consider closely the cultural status of polychrome band-and-panel ware. The ubiquitous occurrence of polychrome band-and-panel ware at Khami and its rarity at Danamombe, Naletale, and Zinjanja may suggest it acquired ethnic symbolic significance later. It is hereby proposed that its reduced frequency was in tandem with declining Rozvi ethnic influences. Oral traditions actually emphasize that polychrome band-and-panel ware was Rozvi (Randall-McIver 1906;Caton-Thompson 1931;Robinson 1959Robinson , 1966. However, Van Waarden (2012:227) notes that, although this ware may have been symbolic, its multiple colors may also have been meant simply to enhance ceramic aesthetic appearence. On a related note, the long list of items that can be appropriated to symbolize ethnicity includes ornaments that were also recovered from all the research sites through excavations. With regard to ornamentation and the Rozvi, Posselt (1935:151) comments: It is said that Mambo's wives ornamented their goat skin dresses with gold beads found in the ruins, such beads being threaded to the hair on the skins. A throne built of wood and ornamented with ivory was erected at the residence, and on this the Mambo was seated while giving audience or presiding over a trial.
The analysis of beads also revealed that there were no notable stylistic changes attributable to the Torwa/Rozvi changeover. Calabrese (2005:67) also observes that glass, copper, and shell beads at elite settlements are usually expressive of heterocrestic rather than emblemic style. Ethnographically, however, it is known that among the Nambya, a post-Rozvi group, a new ruler was given a string of beads as a badge of office (Ncube 2004:12). Also, Bullock (1927:40) notes that Rozvi emissaries wore an insignia of black beads as a means of controlling other local chiefs through religion. In addition, Gann (1965:7) highlights that cloth and beads acquired through trade were often used by Rozvi chiefs to reward soldiers. Although the beads in question were not fully described archaeologically, it is known that Indian red, royal blue, light blue, yellow, and white glass beads were quite popular at Khami-phase sites (Randall-McIver 1906;Caton-Thompson 1931;Machiridza 2012;Van Waarden 2012). While the distribution of these beads was largely conditioned by external trade patterns, archaeologists should not be oblivious to the fact that through time beads also acquired numerous symbolic meanings. In addition, ivory and gold ornaments were most likely objectified as Rozvi symbols because of their presence at the three research sites (Randall-McIver 1906;Summers 1971). According to Loubser (1988:295), ivory and its ornaments were used as symbols of power by Singo chiefs, who usually claimed a single tusk from each elephant killed. Since the Singo are an offshoot of the Rozvi, the meanings of such objects among the Rozvi may have been similar. In this respect, Calabrese (2005) rightly stated that there is a very thin line between class-and ethnicity-related symbols archaeologically. Even ethnographic data gathered from post-Rozvi houses in Bikita demonstrated that objects, despite their general outward appearance, were variably imbued with social meanings (Machiridza 2012).

Conclusions
Despite the well-documented Torwa/Rozvi political transition, material culture from Danamombe, Naletale, and Zinjanja remained largely unchanged. This implies that, when the Rozvi conquered and assimilated the Torwa, they generally maintained preexisting Torwa cultural practices. This stylistic continuity could also mean that material culture by itself does not easily denote fluid ethnic identities. In a way, these results confirm the general belief held by a few scholars that, although a majority of dry-stone-walled Zimbabweculture sites were occupied by different ethnolinguistic groupings, their material culture remained basically unchanged (Beach 1980;Mudenge 1988;Muringaniza and Ruwitah 1996). On the contrary, archaeologists have always believed that objects and landscapes express ethnic identities through stylistic clusters and spatial configurations (Wiessner 1983(Wiessner , 1985Sackett 1990;Hegmon 1992;Tilley 1994;Jones 1997;Kvamme 1997;Bowser 2000;Anschuetz et al. 2001;Calabrese 2005;Johnson 2006;Huffman 2007;Hand 2015). Primarily, they postulate that when ethnic groups converge in particular geographical localities, they tend to inspire the production of distinctive ethnicity-related styles. Perhaps this entrenched theoretical perspective now needs serious reevaluation in the light of new evidence from these Khami-phase sites. Since Danamombe yielded the earliest dates, 1420 and 1435, that came from relatively later occupational layers, it was probably occupied much earlier than the other two research sites. In this regard, Torwa political power would have initially shifted from Danamombe to Zinjanja before finally shifting again to Naletale. When the Rozvi later usurped power from Torwa dynasties around the early 1680s, all these Khami-phase sites were already fully established. The Rozvi therefore simply continued with the Shona rotational succession system, but using Danamombe as the capital of their state. Until more fieldwork is done around the Manyanga site cluster, it remains very difficult to define the actual status of this cluster in relation to those in the Insiza district. All that is known is that oral traditions claim the last Changamire Chirisamhuru died on his way to Manyanga following a barrage of Nguni attacks on Danamombe (Pikirayi 2001:218). Whether these traditions are factual or not remains to be ascertained through fieldwork around the Manyanga site cluster. Therefore, the density of Khamiphase sites in the Insiza district, their proximity to one another, and their similarity in terms of architectural styles, ceramic evidence, and other ornaments largely reflects the operation of the Shona rotational succession system rather than Rozvi ethnic forces.
More interestingly, the chronometric datasets outlined in this article now support a challenge to the way that some historians and archaeologists have been reading the history of the entire landscape. These data challenge, in particular, the dominant view that Khami was the epicenter of all Khami-phase sites (Robinson 1959;Beach 1980;Van Waarden 1998, 2012Pikirayi 2001). Considering the recent radiocarbon dates from Danamombe, its general size, and the large quantities of cultural finds so far excavated there, one can tentatively argue that this site was not necessarily a successor, but a competitor of Khami. The same archaeological data challenge the traditional historical assumption that, upon the demise of Great Zimbabwe around 1450, some of the dynasties migrated north to found the Mutapa state, while others moved southwestward to found the Torwa state. In fact, Khami-phase sites in southwestern Zimbabwe were well advanced when the presumed mid-to late 15th-century dynastic migrations from Great Zimbabwe and Mutapa state took place. Hence, while Great Zimbabwe was flourishing between 1300 and the late 17th or 18th centuries, Khami-phase sites were already thriving between 1400 and the 1830s (Van Waarden 2011, 2012Mukwende 2016;Chirikure, Mukwende et al. 2017).
In view of these observations, the picture that is now emerging is that the role of state capitals, such as Great Zimbabwe, in the rise of other later state systems, such as Mutapa in the north and Torwa in the southwestern region, has been exaggerated (Pikirayi 1993;Pwiti 1996;Van Waarden 2011;Chirikure, Manyanga, Pollard et al. 2014;Chirikure, Mukwende et al. 2017). Already, some scholars are beginning to point out that both ceramic sequences and architectural typologies recorded from Great Zimbabwe and Khami project two parallel cultural developments (Van Waarden 2011, 2012Chirikure, Mukwende et al. 2017). Archaeological data gathered from the research area clearly indicate that Khami-phase sites evolved without much influence from Great Zimbabwe and the Mutapa state. This explains why some scholars are already pushing the argument that the Khami-phase evolved directly from the earlier Leopard's Kopje cultures (Van Waarden 2011;Huffman 2014;Mukwende 2016;Chirikure, Mukwende et al. 2017). This thinking is generally premised on the proximity of Khami-phase sites to Woolandale sites, their ceramic co-occurrences, as well as other broader stylistic similarities in ceramic typologies and architecture (Manyanga 2007;Van Waarden 2011;Chirikure, Mukwende et al. 2017). Considering that the proposition just mentioned above is not supported by empirical field data from Danamombe, Naletale, and Zinjanja, it is hereby argued that the Khami phase possibly emerged from a combination of numerous sociocultural factors that had emerged within the broad southwestern landscape.
It is obviously impossible to pinpoint that exact locality of "Khami-phase genesis" because the Zimbabwean plateau has always been characterized by a series of sociopolitical movements (Beach 1980(Beach , 1994a(Beach , 1994bMazarire 2009;Chirikure, Manyanga, Pollard et al. 2014). This also explains why landscapes are always in a constant state of becoming, full of ideas, memories, and feelings that bind people together. No wonder ethnic groups are ever sensitive about identity construction through space and place references. As the Rozvi gradually expanded through space and time, they actively constructed and maintained their identities by manipulating culture and landscapes. Rozvi ethnicity and power largely depended on their ability to situationally create and sustain an "imagined history/past" through the manipulation of indigenous philosophies, as highlighted above. As a result, the broad southwestern cultural landscape was systematically imbued with Rozvi identities through the mere objectification of naming patterns, totems, praise names, mythology, and invented customs. Previous researchers struggled to define the Rozvi archaeologically because they narrowly focused on their material culture while ignoring the intangible/spiritual aspects or vice versa. Although the Rozvi dynamically and reflexively connected with people, space, place, and objects through numerous strategies, their most enduring archaeological identity symbols became Khami-phase sites and polychrome bandand-panel ware.
Funding Information This research was partly funded by the University of Pretoria SanParks Bursary and the Great Zimbabwe University Research Board.

Compliance with Ethical Standards
Conflict of Interest The author declares that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.