- 9

In 1891, Captain Davis of the HMS Royalist attacked and burned a number of villages in Roviana Lagoon on the southwestern side of the island of New Georgia in the Solomon Islands' Western Province. The attack was directed at the base settlements of the Roviana head-hunters who were responsible for carrying out large scale raids throughout the Western Solomons. The Royalist attack was precipitated by the murder of a white trader at Nusa Zonga in Roviana Lagoon (Sydney Morning Herald 1891, Woodford 1890:150), but was part of a wider British agenda to bring about a Pax Britannica in the Western Province. By the late 19th century the activities of the Roviana chiefdoms were impinging on European commerce. This was done largely by creating an unstable political and economic environment, and by preventing the development of a consolidated European voice among the scattered traders (Bennett 1987:63).

The Roviana chiefdom, and its attendant ritual acts and institutionalised violence have been described by a number of contemporary observers (Guppy 1887, Woodford 1890). With few exceptions, however, the early accounts are historically sterile. They describe a repetitive cycle of raiding aimed at satisfying political, economic and ritual exigency, and at accumulating sufficient capital, as mana, to instigate and finance the next round of violence. The indigenous origin of head-hunting, and its cultural and historic setting are elided or even denied. What is left is an essentialist view of an aggressive political system that dominated the social landscape of Roviana at the time when Europeans were establishing a permanent power base in the Western Province. This system is interpreted as an anachronistic retention of darker times, spiralling out of control under the influence of Western trade and material culture, awaiting perhaps, the fatal impact of colonial pacification.

These accounts raise several problems for us. First, we are concerned to situate head-hunting, and the warfare and ritual acts observed by early visitors - 10 to New Georgia, within an indigenous historical framework. Relating to this, we question the role of Western agency in the acceleration of head-hunting activities in the 19th century. Or rather, we question the viewpoint that large-scale predatory head-hunting was solely a product of Western influence. Thus we do not wish to excise European agency from the head-hunting narratives because there is no doubt that engagement with the West had a fundamental impact on the scale and organisation of warfare, on the economic foundations of the chiefdom, and on the material culture of head-hunting. However, it is our view that the Roviana chiefdom described by firsthand observers was neither a static remnant of prehistory, nor a product of Western agency. Neither do the 50 years of firsthand observations portray a cultural trajectory usefully described in terms of impact and acculturation. Instead, the head-hunting system of the New Georgians was part of a ritually charged politico-economic system that evolved in the Western Province during prehistory. Its articulation with the West involved a process of what Carrier (1992:26) terms “generative interaction”, one involving a creative interaction between indigenous and Western systems.

As Aswani (this volume) demonstrates, one of the central elements in the emergence of a historical representation that disavows indigenous agency is the privileging of historic text. To counter this influence, Aswani examines the narratives of New Georgians and shows that the oral traditions record predatory head-hunting emerging out of indigenous political and ideological entanglements in a setting of shifting demographics and tribal interactions. The archaeological record provides a further historic “text” that informs us on social and political change in the Roviana region in the centuries preceding Western contact. In fact archaeology, although weak in comparison with indigenous narrative in providing the details of historical events and individual action, is an historical discipline which, by the nature of the data alone, emphasises long-term change above all other processes.

In the following discussion we draw on the archaeological record of Roviana and its former inhabitants to show that the head-hunting tradition developed over a period of many centuries in the lagoon region. Its development drew on basic religious practices and symbols, centring on relationships between the living and the dead, which had been present in New Georgia for centuries. Transformations in the scale and cultural setting of these ideals and symbols underpinned the emergence of the powerful, hegemonic chiefdom encountered by the first Europeans to New Georgia. The archaeological record of Roviana makes redundant any essentialist discussion of the chiefdom system and its associated head-hunting cults, and this diachronic perspective is a strength that archaeology brings to the study of head-hunting, and to Melanesian anthropology in general. Further, - 11 the archaeology shows that the notion that engagement with the West resulted in the demise of an indigenous political and ritual expression, seriously distorts a complex and creative resortment of Roviana ritual, ideational and politico-economic practices.

In this essay, head-hunting is considered to be a means by which certain political actors both acquire and manifest power. In Roviana, authority in all matters, be it head-hunting, installation of new chiefs or determination of land use rights, is bestowed by the living. But to exercise such authority requires the positive sanction of the dead, of the ancestors from whom all power ultimately derives. Sanction is demonstrated through the successful acquisition of skulls, which become a material token of the efficacy or mana bestowed by those ancestors on chiefs who organise, fund and lead head-hunting expeditions. Head-hunting connects the living and the dead and provides material evidence of the will of the ancestors in the decision-making of the living. Skulls derived from head-hunting in enemy lands, adorning the houses of the tomoko ‘war canoes’, serve as powerful symbols of chiefly efficacy. The skulls of chiefs (bangara) when conserved in shrines (hope) provide material connections to powerful ancestors.

Figure 1: Chiefly skull shrine at Kindu, Vonavona, New Georgia

- 12

In the chiefly skull shrines, the skulls sit (Fig. 1) upon a form of shell “money” (poata) known as bakiha. Bakiha are large rings (c.14cm in diameter) made from fossil tridacna shell derived from the raised coralline limestones of the Roviana Lagoon. A true bakiha, which is a marker of chiefly authority, has an orange stain, possibly derived from iron staining in the hinge portion of the fossil shell. This distinguishes it from other shell valuables made from fossil shell as well as from a variety of other valuables made from living shell of various types (Aswani and Sheppard n.d.). The relationship between poata and ancestors is not clear in Roviana, but in the closely related island of Simbo, shell valuables are recorded as living. When they die (are broken?), their spirit goes to Sondo, the residence of ancestral spirits (Hocart 1922). Shrines in Roviana contain large numbers of poata, usually broken (probably as part of a ritual of transfer of land use-rights), but in well preserved shrines complete items are common (Fig. 2). Although many shell valuables are inalienable items associated with specific social histories, others were routinely used to mark social transactions (e.g., transfer of land-access rights). A commodity-like exchange could occur whereby poata were amassed by chiefs through various mechanisms and used to finance large-scale activity such as head-hunting or as rewards to successful head-hunting participants (Aswani and Sheppard n.d., Hocart MSS n.d., Miller 1978).

Figure 2: Shell valuables from Site 94 on Honiavasa. Top row bakiha/poata, bottom row left to right two conus rings and one bareke.

The institutionalised violence associated with head-hunting has no direct archaeological signature. However, in the Roviana case it sits within, supports, and is supported by a political, economic and symbolic structure - 13 that is accessible archaeologically. Our archaeological examination of head-hunting draws on a model in which power or efficacy derived from ancestors is materialised, channelled and circulated through an interconnected set of cultural media (ancestors-skulls-shrines-priests/chiefs-exchange valuables-skulls-ancestors). A linear presentation of this process, which under-represents a complex web of relationships, is portrayed in a simplified form in Figure 3. Our point is to illustrate how the archaeologically visible component of the Roviana chiefdom system, the shrines and shell valuables, form part of a set of power relationships in which head-hunting plays a fundamental role. We contend that head-hunting developed in concert with the other elements, which can serve as a proxy for this cultural practice.

Figure 3: Simplified representation of materialised power relationships articulating with head-hunting.

In the following discussion we first document the archaeological sites historically associated with the Roviana groups who practised head-hunting in the late prehistoric and early historic periods. We look at the size and spatial layout of the settlements, and at the form and density of their internal components. Following this, we examine evidence for the development of this settlement system drawing on site data collected on the barrier islands of Roviana Lagoon, and on the coastal and interior mainland of New Georgia. We focus in particular on the shrines whose changing form, content, distribution and spatial relations we see as key to understanding the evolution of Roviana politico-religious configurations.

- 14

Figure 4: General Map of Roviana Lagoon showing sites and locations mentioned in the text.

Roviana Lagoon (Fig. 4) consists of a string of raised coral islands stretching for approximately 40km down the southwest coast of New Georgia Island. Between the islands and the mainland of New Georgia lies a shallow coastal lagoon of approximately two to three kilometres in width. New Georgia is a steep, mountainous island of volcanic origin but with a complex history of faulting, sedimentation and uplift. The vegetation cover of the mainland is dominated by lowland and montane rain forest in the interior. Along the coastal strip and lower river valleys and on the offshore islands, centuries of small-scale bush clearance, fallow regimes and bush regeneration have resulted in a distinctly anthropogenic landscape: a mosaic of garden clearances, fallow plots, scrub and fern land border stands of regenerating and mature forest. Introduced species, such as pandanus, are now dominant understory plants in some areas and groves of coconut and other nut trees (Canarium spp. and Barringtonia spp.) in forest clearings and around settlement areas attest to the role of arboriculture in traditional Roviana subsistence.

Permanent rivers transect the rugged mainland interior and flow into the lagoon every few kilometres along the coast. These once flowed across the now submerged coastal plain, meeting the ocean in what are now the major passages through the barrier islands. These reliable water sources are supplemented by springs close to the coast and there are also areas of marshy - 15 land in the lower reaches of most river systems, and located in patches behind the mangrove zone. Water supplies on the barrier islands are less reliable. There are no permanent streams, but many of the islands contain semi-permanent springs located above the high tide line. All of the barrier island water sources dry out during particularly long dry spells (July-August), and people routinely transport water from the mainland to the islands during these periods.

The coastal soils of Roviana are mainly of limestone derivation, reflecting their origins in uplifted coralline reefs, although soils inland of the mainland coastal flats are derived from basic volcanics. The limestone soils are low in nutrients but support a shifting cultivation system which is today centred around sweet potato production, supplemented by the traditional crop plants: taro, yam, banana and sago. Sweet potato is probably an early historic period introduction 1 that does well on coastal flats adjacent to the sea. Other introduced food plants include cassava, which does especially well in the low nutrient conditions, as well as a range of fruits and green leaf vegetables. Irrigated and wet-field taro production is today only practised in a few favoured locations in Roviana, although it is potentially viable along the edges of the coastal river systems. Oral traditions, however, suggest that taro production was practised more intensively in the past, particularly around the interior of north New Georgia (Tedder and Barus 1976), and we have recorded the remains of small scale irrigated systems in the upper reaches of some streams (Sheppard et al. 1998). In addition to a shifting horticultural system, the subsistence economy of the Roviana peoples focused strongly on maritime exploitation systems, reflecting the richness and diversity of the New Georgia lagoons.

To summarise the conditions of Roviana Lagoon as they apply to Melanesian economic patterns, the region has a low potential for intensive horticultural production with limitations imposed by low nutrient soils and the lack of either extensive coastal plains or swamp systems suitable for wet-field taro production. This is offset by a rich and diverse marine environment, and despite the presence of an extensive forested interior, historic Roviana was very much a maritime based society. In Melanesia, coastal chiefdom systems, such as that of Roviana, tended to develop in areas where they could dominate external trade and exchange systems (Keesing 1992a: 187). The Roviana chiefdom and head-hunting system was intrinsically linked to a political economy centred in external exchange and, while avoiding overt notions of causation, we note that the local environmental conditions no doubt facilitated these developments. Roviana Lagoon is well sheltered, with a number of deep, safe passages to the open sea, and articulates with other extensive lagoon systems to the southeast - 16 and northwest. These inshore waters offered rapid access to the major population centres of New Georgia and relatively short, straight runs from coastal New Georgia could be made to the other islands of the Western Solomons, notably Rendova, Kolombangara, Choiseul and Simbo.

The environment described above represents the physical setting upon which the Roviana chiefdom and head-hunting system developed and flourished in the centuries leading up to 1900. We have discussed elsewhere, in reference to Roviana, the manner in which the landscape plays an active role in social construction and reproduction (Sheppard et al. n.d.). Here, we discuss the evidence from landscape archaeology, specifically settlement pattern analysis, for the scale and physical structure of the Roviana polity as it existed in the proto-historic period.

Archaeological landscapes consist of a juxtaposition of enduring features from different time periods; they constitute a palimpsest of history. There are means, however, of teasing apart the various components and transforming what is essentially a synchronic artefact into a diachronic one— of breathing history into the landscape. The New Georgia Archaeological Survey (N.G.A.S.) research has recorded and mapped hundreds of archaeological sites and features relating to the Roviana chiefdom, but in this review we discuss only a small subset of these. Drawing on oral tradition, radiocarbon dating and typology we describe a sequence of change in sites, features, artefacts and their configuration. We begin by describing the archaeological landscape of Nusa Roviana Island which was the ritual and political centre of head-hunting activities in the late prehistoric and early historic era. We then discuss sites and site complexes along the barrier islands and the New Georgia mainland, and argue that a chronological sequence can be presented for these which can be used to model the development of the Roviana chiefdom. As indicated above, the key element of this chronological treatment of head-hunting lies in tracing the changing structure, content and spatial relationships of shrines, a primary ritual and symbolic focus of all power relations and acts in Roviana society.

Nusa Roviana Island is about 4km long and reaches a maximum elevation of just under 80m making it the highest of the barrier reef islands in Roviana Lagoon. It has a narrow coastal strip on three sides, rising to a central ridge about one kilometre in length. Swidden gardens are maintained on the coastal flat land and rise high up the central slopes, especially on the (northern) Kalikoqu side, where today they are encroaching into areas which contain - 17 archaeological remains of previous habitation, fortification and religious activity. The wide southeastern end of the island is low, swampy and uninhabited.

Nusa Roviana is divided into three territories, which are the settlement zones of three closely related social divisions within Roviana society. These divisions, Vuragare (ocean side), Kalikoqu (lagoon side) and Kokorapa (middle), were created through the fusion of at least three named tribal elements (Koloi, Tagosaghe, Kazukuru) as well as other coastal groups who inhabited the lagoon before the major waves of population movement. Details of this ethnohistory are found elsewhere (see Aswani this volume, Sheppard et al. n.d., Aswani and Sheppard n.d.). The sites on Nusa Roviana are clustered into four major aggregations: a hill fort, and three open settlements on the Vuragare, Kokorapa and Kalikoqu areas of the lowland.

The fortification on Nusa Roviana runs for more than 600m along the eastern end of the island's central ridge. The fort is constructed of coral cobbles which are formed into transverse walls, sometimes supplemented by ditches (Fig. 5). There are four main sections or compounds within the fort, each of which is named and associated in oral traditions with specific events and individuals. Each compound contains sets of architectural and other archaeological features, the most prominent of which are the shrines.

The shrines consist of low platforms of coral rubble and are usually rectangular to sub-rectangular in form. The occasional basalt boulder or slab is incorporated into the structure. These were brought from the mainland and are thought to be of symbolic significance. On some of the shrines sheets of lace or sheet coral mark the location of former skull houses which contained the skulls of the ancestors (in contrast, the trophy skulls taken in head-hunting raids were used to adorn the canoe houses on the coastal foreshore [see Aswani this volume]). Most of the shrines contain surface scatters of shell valuables and other artefacts, and some also contain intact or fragmentary remains of human skulls. An example of a ridge fort shrine is shown in Figure 6.

The artefacts found on the shrines were those used in the ritual and ceremonial practices associated with the ancestor cults of Roviana religion. Multiple examples of most of the shell valuable types recorded ethnographically were present on the shrines in addition to a range of ornate pearlshell ornaments (see Fig.2). Some of the shrines also contained large tridacna shells which were used as ritual objects. An important feature of many of the shrines is the presence of European trade items. On both the hill-fort and the open settlements of the Nusa Roviana coast, we recorded

- 18

Figure 5: Plan of hill-fort and Kalikoqu coastal settlement on Nusa Roviana with numbered shrine features (adapated from Nagaoka 1999, Fig. 3.17).

- 19

items such as steel axes and muskets in positions indicating that they were deposited as ritual offerings. These artefacts not only provide us with a relative date for the use of the structures, but also serve to reinforce the argument that the Roviana politico-religious system was involved in a creative interaction with the west. Artefacts which were initially introduced into Roviana as commodities appear to have been transformed, in the context of their use and perhaps their individual life histories, into inalienable or decommoditised objects (Aswani and Sheppard n.d., Thomas 1995a).

Other common archaeological features within the fort sections are residential platforms and cut stone-faced terraces that run in several tiers below the crest of the ridge and often contain deep shell midden deposits. Both of these features are interpreted as supporting the residential components of the site: the houses and ancillary structures of the domestic realm. Figure 5 presents the essential features of the hill-fort as mapped in 1996.

It has been difficult to achieve consensus among various informants about the relationship between shrines named in oral histories and specific physical structures recorded in the dense undergrowth of the hill-fort. There is, however, an essential agreement about the associations of named places, even if their locations are uncertain. This is especially true of the middle section of the hill-fort (Sites 13-19) which features prominently in Roviana tradition. Table 1 presents informants statements concerning each of the major shrines.

Approaching the complex from the modern village, the first feature associated with the fortification is a transverse ditch and bank defensive feature running across the ridge. The ditch lies on the outside of the fortification, along the northern half of the bank. The latter is constructed of mounded earth with stone facing. Excavation across the ditch and bank failed to recover any dateable material, but a shrine located just to the west of the wall, outside of the fortified area (Site 118 Ex-J), provided a calibrated age in the 13th to 14th century (Table 2). As discussed below, this shrine is typologically unlike those in the fortification and we believe it pre-dates fortification construction. Immediately to the west of the wall is a large shrine complex (Site 12, Fig.6) which is probably Olobuki, the main skull shrine of the Nusa Roviana chiefs and the only chiefly skull shrine located within the fortified area. This is a rather large complex of stepped platforms with an associated oven and probable assembly area delineated by a stone alignment to the west. A series of depressions in platforms Fl and F2 within Site 12 possibly mark the location of destroyed skull houses. 2

- 20

To the west of Site 12 are a series of shrines distributed along the narrow ridge top. Within this area there is a poorly defined rectilinear area that is a possible square assembly or dance floor, but few platforms or features other than shrines were located. Most of these shrines are associated in oral accounts with war magic and ritual; they are places where warriors acquired and then displayed their power. The two exceptions are the sites known as Zare and Zareibibu (Sites 13 and 14). These are associated with ancestral beings, a clan of mateana or ‘angels’, from whom the chiefly lines of Roviana descend (see Aswani this volume).

The next section of the fortification, Barairiranga (lit. ‘Wall for Shouting’), is isolated behind high (2-3m) stone walls. An entry-way in the western wall (Feature 1058) was excavated in 1998 and shell recovered from within the wall has been dated (WK-6757) to the 16th-17th century, suggesting possible construction of this elaborate section of the fortification in the late 16th century. The base of a stone oven (oputu, Feature 1082) closely associated with a large shrine (Site 12) located in the middle of this area has been dated (WK-6156) to the same period.

The most distinctive feature of this part of the hill-fort is the regular series of about 30 stone-faced terraces that extend down the slope on either

Figure 6: A hill-fort shrine, Site 12.

- 21

- 22

side of the ridge. These well-designed terraces, that in places have been excavated back into the rock, often have carefully constructed walkways running between them and seem to have served as residential platforms. Excavations in 1998 (Ex-I4 and I5, Feature 1063 and 1062) and 1999 (Ex-R, Feature 1091) on terraces on either side of the ridge did not reveal any posthole features indicating definite house structures. However, abundant food waste was recovered (shell, fishbone, mammal bone). Oral tradition suggests that this section served as a residential refuge when the island was under attack, but the care and design evident in this area argues for more than episodic use. The next section of the fortification contains a similar array of features to the last. It is separated from the latter by a 2m high coral rubble wall and essentially consists of a causeway-like approach along the narrow ridge. There is a cliff on the northern side and a very steep terraced slope to the south.

The highest walls on the hill-fort cut off the eastern end of the ridge, and this area contains the most isolated and probably the most sacred site on the island-the shrine of Tiola. Tiola is a central mythological figure in the Roviana religious and political system who, in at least one account (Oka in Sheppard and Walter 1998), is considered responsible for teaching the Roviana people how to construct the first war canoe, the tomoko. The tomoko is the primary symbol of Roviana power as well as being the direct physical medium for the assumption of power via its role in raiding, and in the acquisition of trophy skulls and captives. The small shrine of Tiola contains a number of shell artefacts and a small carved stone figure of a dog's head (Fig. 7), which is reportedly all that remains of a larger statue. When the - 23 Roviana polity was in danger, Tiola gave warning and the statue turned to indicate the direction from which the threat was approaching. This very small area is defined by a large (<3m high and <2m wide) massive rubble wall to the west, cliffs to the north and south, and a steep slope to a 5m deep, 7m wide, rock-cut ditch which traverses the narrow ridge to the east. To the east of this ditch is a small high platform below which, down a steep slope, are located a series of stone-faced terraces ending in a final wall. Below this point the slope drops sharply to the flat where additional platforms and terraces ring the ridge as described below.

Figure 7: Broken coral carving of Tiola.

In summary, the hill-fort is as yet poorly dated. However, two dates directly associated with the Barairiranga section, which has a design cohesiveness suggesting it was constructed during one period, indicate that construction activities involving considerable effort occurred during the late 16th century. The western end of the fortification, which consists essentially of a well-defended set of shrines, is as yet undated, although the shrines are not obviously different in form and content to those in the Barairiranga section. As will be discussed below, earlier dated shrine forms (faced shrines) do exist on Nusa Roviana and we therefore infer that most of the hill-fort post-dates the late 16th century. Drawing too on oral tradition and the presence of historic artefacts, we contend that its use continued until the late 19th century.

The open settlements of the Nusa Roviana coast are located in three distinct zones associated with the regions and tribal divisions of Vuragare, - 24 Kalikoqu and Kokorapa (Figs 4 and 5). There are additional scattered architectural features which mark out smaller settlement units that we are unable to tie as closely to oral accounts of social structure, but which were probably linked to the major settlements. Each of the three major settlement areas contains a similar rich array of architectural features as found on the ridge fort, although the components display a greater variability in form. In addition to the shrines and residential platforms, the open settlements contain wharves and a range of low mounds and platforms of unknown function. Here too, on the low coastal flats, is where the trophy skulls were displayed on the elaborate tomoko canoe houses. No archaeological remnants of the canoe houses and their trophy skulls could be unambiguously defined, but they are described in oral accounts and in the early historic records (Somerville 1897:369).

Each settlement area of the Nusa Roviana coast contains rich surface scatters of midden shell and small-scale excavations on platforms reveal much of the fill to be of cultural origin (i.e., midden taken from the surrounding living surface). The open settlements also contain evidence of manufacturing activities, including areas which were used for the manufacture of fossil tridacna shell valuables.

The sites and features described above are clearly not all contemporaneous. Excavations on the hill-fort show that major phases of fort building occurred up to 400 years ago, and this is consistent with the oral traditions and genealogy which suggest that large-scale aggregation and the centralisation of political power occurred by at least 300 years or 13 to 15 generations ago (Sheppard et al. n.d., see Aswani this volume). Yet we do believe that the Nusa Roviana archaeological landscape described above defines a critical phase in the development of the Roviana polity. This is not to suggest that a phase of stasis existed in the polity for any meaningful length of time, but rather that the cumulative archaeological record of Nusa Roviana describes a ritual and symbolic landscape that contrasts significantly with what went before. The most important element of this is the density of religious symbolism. As indicated in Figure 5, in many parts of the Nusa Roviana settlements the areas occupied by religious as opposed to secular structures are equal or greater. The shrines, including highly elaborate shrines with skull houses and large quantities of shell valuables, are juxtaposed with midden dumps, specialised activity areas, and house platforms and domestic activity zones. We do not see in the Nusa Roviana landscape the sacred/secular spatial dichotomy that is a familiar part of many Pacific archaeological landscapes, and it is tempting to extend this observation as - 25 metaphor and argue that indeed these worlds were not meaningfully differentiated during the period in which these sites were constructed and used by the Roviana people.

The archaeological landscape described above is representative of the Roviana politico-religious system at its most complex—late in prehistory and into the historic era. The spatial configuration of sites, the sheer number of shrines in relation to structures of domestic function, and the range and quantity of ritual offerings and religious paraphernalia is locally unique in time and space. Yet the individual elements of this configuration have been present in the lagoon as part of the cultural landscape for a much longer period.

Oral tradition and firsthand records indicate that Nusa Roviana was one of the main centres of head-hunting politics and ritual activity by the early 19th century. But the same traditions link the Roviana groups with communities occupying sites scattered along the barrier islands and mainland New Georgia. The following account of survey data from the barrier islands first discusses a site that was occupied subsequent to the construction of the Nusa Roviana fortification, and then two complexes occupied somewhat earlier.

Saikile (Site 47), which is located on the eastern end of Ndora Island at the eastern end of the lagoon, is a highly fortified complex over 150m long with features very similar to those on Nusa Roviana. The site is built up against a large block of makatea c. 15m high and extends out to a low coastal cliff that limits access from the sea except from a small beach on the western end of the site. To the west and east stone walls, with an interior height over 1.6m and a low probable firing step, run from the makatea cliff to the coast. Both walls are pierced by entry-ways and the eastern wall has a narrow window. The interior of the fortification is paved with shell and the debris from the manufacture of shell valuables. A series of skull shrines are located along the coastal margin and contain skulls inside sheet-coral skull houses as well as numerous shell valuables and historic artefacts. One surprising feature of the fortification is the presence, on top of the makatea block and approached via a near vertical ascent up the 15m cliff, of a series of coral cobble walls and platforms which clearly served as a final refuge from attack.

The style of the skull shrine and defences strongly suggests this fortification is at least contemporary with that on Nusa Roviana and the design of the walls suggest a possible construction date soon after the - 26 introduction of guns. Oral history supports this inference as the history of the Saikile polity (see Aswani this volume) relates that the fortification was built by Odikana, a bangara from the chiefly line on Nusa Roviana who left the island six or seven generations ago at the time when the very powerful chief Tae bangara was dominant on Nusa Roviana. The traditional history of Saikile suggests that Odikana was invited to the eastern end of the lagoon to unify the warring tribes in that region. However oral history from the western end of the lagoon suggests he fled Nusa Roviana after conflict with Tae Bangara. Whatever the cause, it is tempting to suggest that the abundant debris from bakiha manufacture and the concentration of shrines represent attempts to legitimate the mana of a new chiefly polity developing in the late 18th to early 19th century. Today the division between the Saikile chiefdom in the eastern lagoon and that of Kalikoqu at the western end of the lagoon is the fundamental power split in Roviana. The construction of the Saikile fortification is tentatively associated with the early historic period or slightly earlier at the time of Tae Bangara, who oral tradition describes as dominating the entire New Georgia Group and beyond through marriage ties and raiding (see Aswani this volume).

Just east of Honiavasa Passage a settlement complex (Site 94) is defined by coral platforms, shrines and scatters of midden. The settlement is undefended; there are no walls, ditches or other features which could be interpreted as defensive. The whole complex is spread out along the fringes of a belt of gently slopping garden soils, which are today planted by the community of Sasavele, who reside on the opposite side of the passage. The structures are dispersed over an area of approximately 10ha and occur at a very low density. There are several clusters of possible house platforms, and some of these are associated with shrines. Elsewhere clusters of shrines stand in isolation from other feature classes in contrast to the situation described above for the fortification and main settlements of Nusa Roviana. The shrines at Honiavasa contain only a small quantity of ritual goods. Shell rings and tridacna offering vessels occur on some of the features, but few shrines contain more than a single valuable, and none display the density of artefacts seen on those of Nusa Roviana.

The settlement of Kekehe (Site 46) on Ndora Island in the Saikele region of Roviana is another site complex predating the major aggregation of people and power on Nusa Roviana. Like Honiavasa, Kekehe is undefended. It consists of a series of coral and basalt platforms rising up a slope above the - 27 low coastal flat. Most site components occur either alone or in small, isolated clusters up to 70m from other structures. A number of the platforms are very large and extremely well built, with coral and basalt facing stones and basalt cobble paving. The abundance of basalt cobble constructions is unique on the barrier islands as all basalt must come from the mainland. However, the mainland opposite Kekehe is one of the few areas in Roviana where basalt outcrops occur on the coast. The largest of the basalt cobble platforms are found in a tight cluster near the base of the settlement and probably include platforms that supported houses, as well as some associated with ritual activities. A number of small shrines are reminiscent in form to those found at Nusa Roviana, but shell valuables are absent in all but one. The most significant ritual site at Kekehe is a shrine sited on a natural outcrop of raised coral on the edge of the coastal marshes. It includes two carved figures as the centrepiece to a number of small offering areas, each containing bakiha and other classes of shell valuable (Fig.8). This shrine is isolated from all other components of the site, lying at least 200m from the nearest other structure, and cannot be directly linked to the main complex. The contrast with the placement of Tiola, located adjacent to one of the most densely occupied parts of the Nusa Roviana ridge settlement, is striking. Historic artefacts were recorded on only one of the Kekehe platforms, and limited scatters of midden do not argue for prolonged or intensive domestic activity. It is possible that the main residential zones of the community were scattered over the plantation lands, rather than sited within the mapped areas.

Figure 8: Shrine (Site 40) at Kekehe on Ndora Island, Roviana Lagoon. Statues of two ancestors.

The contrast between the settlements of Honiavasa and Kekehe on the one hand, and the complex of sites on Nusa Roviana on the other is - 28 significant. On Nusa Roviana the settlement areas of the coast can be interpreted as nucleated villages. The inter-household space comprises part of the everyday living area of the community and within this zone most of the everyday domestic activities of the community are represented. This includes both the religious and secular activities, such that the shrines, domestic structures, working floors and midden occur side by side and define a contiguous living surface. In contrast, the Honiavasa and Kekehe settlements are dispersed; the household units appear to have been separated from one another by gardens and wasteland, and this space was probably crossed by pathways or tracks which linked the structures, shrines, plantations and other activity zones. Most shrines are distanced from the house platforms; they contain a comparatively meagre array of shell valuables and no historic artefacts.

Elsewhere on the barrier islands, and particularly in proximity to contemporary villages, the survey recorded a number of shrines, some lying in isolation from other archaeological features, others found in small clusters. Several contained shell valuables. The other major class of site located on the barrier islands comprises the ceramic sites in the inter-tidal zone of the lagoon foreshores. So far, we have recorded over 20 of these sites on the barrier island and mainland coasts of Roviana (Sheppard et al. n.d.). A number of the ceramic sites are located adjacent to quality springs and most are located in areas near where large population centres are found today. In other words, settlement was probably based on the same prerogatives of access to water, arable land and the reef passages. While it is clear that this ceramic phase of Roviana settlement predates anything described above, we are unable as yet to provide a strong cultural linkage between these sites and those of the Roviana polity indicative of ethnic continuity. 3 What we can say, however, is that there are no ritual sites on the coast that we can associate with the inter-tidal ceramic sites, and thus we believe the early ceramic phase of Roviana settlement predates the development of the key material and structural elements of the Roviana politico-religious system.

Oral tradition and ethnographic observations indicate that the coastal mainland contained scattered settlements throughout late prehistory and into the early historic era. Site surveys at various levels of intensity, including systematic shovel testing, were carried out along the mainland coast and along the ridges up to a distance of about 4km inland. The coastal survey resulted in the identification of only a few isolated coral cobble platforms, similar in form to some of those located on the barrier islands, but no site clusters with domestic structural units equivalent to those at Honiavasa, - 29 Kekehe or Nusa Roviana were recorded. On the ridges up to 3km from the coast, however, a series of isolated shrines and small shrine clusters were located. These differed in a number of ways from the vast majority of coastal and barrier island examples.

Most of the inland shrines were earthen mounds, up to 15m long and 8m wide, faced with basalt cobbles and columnar blocks (Nagaoka 1999:77). None of these “faced” platforms (e.g., Figs 9 and 10) contained skull houses or other elaborate ancillary components like those found associatedwith the “unfaced” coral cobble mounds which dominate the barrier islands. In particular no ovens or hearths were found in or around or the structures, which is significant since these features are recorded in oral traditions as being associated with shrine ritual.

Figure 9: Mainland type faced shrine, Site 33.

- 30

Figure 10: Photograph of Site 33 looking north.

Furthermore, only one of the shrines we recorded contained any items of material culture. The single exception was Site 57 containing one rough shell ring of the type known as bareke, which is said to be the oldest form of shell valuable in Roviana (Nagaoka 1999; see Aswani this volume, Miller 1978:293). The most elaborate of the inland shrines were stepped platforms with a small, low, square cist-like arrangement of stone slabs located on the upper platform. Often additional features including stone uprights and alignments of up to three large flat stone slabs set up on cobbles (“table” stones) were generally found at a short distance outside the platform.

Dating the inland shrines has proved problematic as they lack ovens and excavation has generally produced no cultural material. One date (NZ-6235) on canarium nutshell was obtained from a cultural layer at Site 25. This eroded midden was located c. 50m from a faced shrine (Site 24) and consisted of a band of midden-rich soils containing plainware ceramics, marine shell, nut shell and chert. The date (Table 2) calibrates at 1 sigma to the intercept ranges of A.D. 1403-1490 and A.D. 1608-1612. As discussed below, this date is comparable to those on shrines of a similar form from Nusa Roviana and we believe it is a reasonable estimate for the age of Site 24.

The isolated nature of the inland shrines suggests a deliberate separation of domestic and religious components in the settlement pattern. So far, we have failed (despite intensive survey around selected shrines) to record any platforms, terraces or other features clearly interpretable as supporting - 31 domestic activity in the vicinity of a religious structure, with the exception of the deposit noted above and one other shell midden located 80m from a small shrine complex.

If Nusa Roviana marks the symbolic endpoint of the Roviana head-hunting trajectories, there is no doubt that the settlements at Bao mark the beginning. Oral tradition states that the Roviana people have their origins in large part from a population located in the interior of New Georgia behind Munda in an area known as Bao. The movement from Bao to the coast and Nusa Roviana that genealogy suggests took place in the late 16th to 17th century A.D. (see Aswani this volume) would appear to mark the formation of the Roviana polity on Nusa Roviana. Certainly for Roviana people, the origin of Roviana correlates with the movement of Ididu Bangara to the coast some 12 generations before A.D. 1900 (Nagaoka 1999:11). In July 1999, we conducted a five day survey in the interior 9km inland from the Munda coast in the Bao region. Here on a high ridge that looks out towards the north New Georgia coast we found a series of 18 platforms located along a 600m east-west stretch of the ridge. Although analysis of the data is incomplete and radiocarbon dates are not yet available, most of these platforms are clearly shrines like those discussed above from the coastal ridges. The largest shrine at the eastern end of the ridge (Site 145) is a large basalt-faced stepped platform with a paved approach leading to a large “table” stone. No oven or shell valuables were observed in any of the sites or recovered during platform excavations. Although archaeology can not definitely link these sites to the Roviana oral tradition, it does confirm the presence of a shrine complex very reminiscent in its layout to that on the ridge at Nusa Roviana, and documents the existence of a population concentration in the area known as Bao.

The culture history and chronology of Roviana is as yet only known in broad outline. However, our research has provided the basis for a chronological framework which we feel can provide a reasonable model for the timing of the development of the Roviana system and, as we argue below, by inference the development of head-hunting. Oral traditions pertaining to surveyed sites, together with radiocarbon determinations (Table 2), allow us to define a basic typological transition from stone-faced earthen shrines, which are mainly associated with the adjacent mainland, to the unfaced coral rubble shrines of the barrier islands. The Nusa Roviana examples of the latter display a range of ancillary features, including skull

- 32

- 33

houses, and frequently contain ritual offerings in the form of abundant shell “money” and other items. With the exception of one rough shell ring (bareke) and two small tridacna shells (hio), the mainland shrines contain none of these features. We place the Bao complex near the beginning of a continuous Roviana sequence from which the intertidal ceramic sites must currently be excluded. While this positioning is not presently supported by radiocarbon determinations, it is consistent with oral tradition. Furthermore, the structures at Bao, while sharing basic attributes linking them to those at Nusa Roviana, are of the faced variety known to be the earlier form in Roviana. Like the other faced structures, they contain no ancillary features and no shell valuables or other offerings.

Although we have presented in our previous discussion a basic distinction between barrier island (including Nusa Roviana) “unfaced” coral-cobble shrine platforms versus mainland “faced” platforms, a very small number of island type shrines are found on the mainland, and most significantly a small number of mainland type shrines are found on the islands, including Nusa Roviana. Radiocarbon dates from two of the latter shrines produced the earliest dates from Nusa Roviana sites and confirm the chronological claims of our typology. Site 118 (Ex-J1) is a large platform (8.1 x 3.6m) just outside and to the west of the first wall of the Nusa Roviana hill-fort. It differs from the shrine platforms inside the fort in that it is faced with vertical slabs of coral in the fashion of the basalt facing of the mainland shrine form. Only one shell arm-ring (hokata) was recovered from this site and there were no associated ovens, skulls or lace-coral slabs commonly used in the construction of skull houses. A shell recovered from under the wall suggests that it was constructed in the 14th century (Nagaoka 1999:110). Although some might argue that this platform is not necessarily a shrine, in our view its distinctive form and construction argues for a strong link to the mainland shrine tradition.

Site 79 located on the coastal flat north of the Nusa Roviana ridge is associated in oral tradition with Ididu Bangara whose movement to Nusa Roviana, as discussed above, traditionally established the Roviana people on the island 12 generations before A.D. 1900. This shrine complex is distinctive within the Nusa Roviana context for the large amount of basalt which has been transported from the mainland and used in its construction. The complex consists of a main stepped basalt-faced platform (10.2 x 5.5m) in association with two unfaced platforms within an enclosure defined by a low coral edging. Canarium nutshell recovered from the footing trench under the main platform wall suggests a construction date (NZA-9457) after the mid-14th century. Using the Ididu Bangara association and allowing three generations per century would produce a genealogical date of c. A.D. 1500 - 34 which is in reasonable agreement with the radiocarbon estimate for this shrine.

The remainder of the dates in Table 2 are associated with unfaced shrines and the hill-fort construction and indicate a date for these developments beginning after A.D. 1500. This would mark the start of the development of skull shrines and their associated features and artefacts, including the use of shell valuable offerings.

At present this picture is fairly rudimentary, but with further study of morphological variation in shrines and associated radiocarbon dating we believe a more refined seriation of shrines and associated features from the 15th to early 20th centuries can be developed (Nagaoka 1999:108-113). However, these data provide sufficient evidence to shed light on the question of the timing of the development of head-hunting. In particular we believe that this data strongly supports an argument for the development of the religious and shell valuable component of the Roviana cultural system at some point in the 16th century A.D. and at least one period of maximum influence coeval with the development of the Saikile fortification in the late 18th or early 19th century.

The archaeological landscape of Nusa Roviana as described above represents the physical remains of a chiefdom system whose leaders derived their power through the intercession of the ancestors. The ancestral cult shrines with their skull deposits, and the canoe houses adorned with trophy skulls dominated the prehistoric settlements and bound the worlds of the living and the dead into a single whole. In the surveys of the barrier islands and the New Georgia mainland, we have demonstrated changes in the form, spatial arrangement and material culture of the structural elements of the Roviana political and religious system through, primarily, a discussion of shrines. These data support an argument for a steady growth in political and ritual complexity over a period of at least 500 years. This argument is contingent on two factors: first, that continuity in these elements can be demonstrated across time and space, and second, that a chronological site sequence can be established. In reference to the first of these issues, we have relied on oral accounts and basic site typology to situate the sites of the mainland, barrier islands and Nusa Roviana within a single cultural tradition. Oral histories in Roviana consistently refer to the movement of populations from Bao and other inland areas to the coastal mainland and the barrier islands, with the greatest aggregations of population and power occurring on Nusa Roviana in the centuries immediately preceding European arrival. In each of these places we have located and mapped sites - 35 typologically similar to those of the late Roviana phase. In reference to the second contingency, these traditions also provide a relative chronology that we have supported through limited but consistent radiocarbon dating.

The generation, acquisition and disbursement of power within the Roviana polity required the actors to negotiate ritualised pathways linking the worlds of the living and dead. A model of these networks is shown in Figure 3. Central to this model is the structural connection between shrines housing the venerated skulls of ancestors, and canoe houses with the trophy skulls of the denigrated enemy. The two ritual acts of ancestor veneration and head-hunting mediate between the worlds of life and death, and provide the paths through which mana and efficacy flow from the dead to the living. Drawing on this model we argue that the archaeological changes we document in Roviana settlement patterns, especially in reference to shrines and their associations, occurred as part of a transformation of Roviana political and religious structures of which head-hunting was an intrinsic part. This transformation involved a process of religious intensification, a process whereby the worlds of the living and the dead gradually merged, and ritual and symbolic structures, once located in separate domains, encroached on the domestic realm and came to dominate every aspect of daily life. This process of intensification is illustrated archaeologically in terms of a number of parallel trends in the internal structure of shrines and in their spatial relations. A parallel trend for canoe houses unfortunately is not open to archaeological documentation.

The shrines of the mainland are generally simple in form, with few internal features. Some of the larger structures have “table” stones or altars, but none contain skull houses or subsidiary ritual features, and none contained bone deposits or shell valuables. We were unable to locate adjacent ovens (oputu) and there is little archaeological evidence to show how these structures were used, as virtually no faunal or other cultural remains have been recovered from a series of excavations.

The shrines on Kekehe and Honiavasa showed greater variation in form and many continued a variety of internal features including cists, mounded corals, uprights and paving, although they did not contain skull houses. On Nusa Roviana the shrines reached their greatest level of diversity and internal complexity. Here the shrines commonly exhibited a range of internal features including low skull houses, uprights, paved areas and ancillary enclosures. Ritual offerings were found on most shrine in the form of bakiha, bareke and other shell valuables, as well as a range of more quotidian items. These included grinding and nutting stones, worked and unworked tridacna shell, - 36 and historic artefacts such as muskets and trade axes. Human bone was found on many shrines, others contained pig and other bone as well. Ovens were present alongside many of the shrines, providing additional evidence of ritual activity.

The other significant trend in shrine development relates to changes in their relative abundance within the settlements, and their placement within the wider settlement pattern of the Roviana communities. The settlement pattern of the mainland, as far as we can determine on the basis of oral tradition and limited survey data, is best defined as highly dispersed. The actual habitation areas have yet to be located, but houses were most likely scattered over the planting soils along the ridges but also on the lower slopes, in the valleys and along the coastal strip. The shrines, located on the ridges, were separated from both the domestic habitation zones and the horticultural land. These were areas associated with the dead and were kept isolated from the domains of the living.

At Honiavasa and Kekehe the settlement pattern is more clustered. The domestic and religious structures are spread out along the peripheries of the horticultural land and occupy the same zone within the settlement space of the community. Yet the individual structures within this settlement zone are quite scattered, and many of the house platforms and shrines are located up to 50m or more from the nearest other unit. Nevertheless, there are also areas of clustering, including areas where shrines are located within a few metres from house platforms. On Nusa Roviana the settlement pattern is highly nucleated. The ratio between shrines and house platforms is high and the shrines are scattered among the apparent house platforms. The habitation zones of the Nusa Roviana lowlands are located on the edges of the planting soils but are nucleated enough to be defined as villages. Isolated shrines are found on the planting lands and at various places around the foreshore. These are described in tradition as agricultural or fishing shrines, and many contain a specialist array of artefacts—some fishing shrines are adorned with net sinkers, for example. Thus all of the space within the Roviana settlement subsistence system was permeated with religious symbolism and acts. There we see a complete coming together of the domestic and sacred components of village life, and via the ancestral shrine and canoe houses, a merging of the worlds of the living and the dead.

Our current chronology for this development suggests that it began at some point after A.D. 1500 and well before western contact. The development of the new shrine form and its associated features, such as ovens on which offerings were prepared, clearly dates in the 16th century. - 37 So too does the clustering of shrines and residential areas, the gradual merging of the sacred and secular realms. The association of shell valuables which are found on the surface of these shrines is less certain, although the rough bareke forms associated with some early shrines argue for a long tradition of shell valuable production in Roviana. In fact, it is possible that the Roviana shell valuable tradition has its genesis in a much earlier shell ornament tradition, perhaps related to that of Lapita (Kirch 1988). Whatever the origin of the individual elements of Roviana religion, however, we see the shell valuables system, the religious intensification discussed above and head-hunting as part of a closely integrated complex. Their merger began in Roviana, and possibly spread throughout the New Georgia group and especially into the neighbouring islands of the New Georgia sound where an interaction system and a similar shrine tradition exists (Nagaoka 1999:31), starting by at least the 16th century. Together they provided a means for establishing and legitimating power over an aggregated population in the developing chiefly system of Roviana. This system continued to operate well into the historic era, until that economy was fundamentally transformed by articulation with that of the West and its new religious base opposed to skull cults and head-hunting.

The authors would like to thank the kind people of Roviana Lagoon who have assisted us in so many ways over the last four years. You have enriched our lives and for that we are truly grateful. In particular we would like to acknowledge the support of the Roviana Area Council, the Ministry of Culture Western Province as represented by our good colleague Kenneth Roga, Chief John Roni of Kalikoqu, and the Solomon Islands National Museum. We hope our interpretation of your history and custom fairly reflects all that you have tried to teach us. Leana Hola. We must also thank The Royal Society of New Zealand Marsden Fund and the National Geographic Society for supporting our research over the past two years and acknowledge the on-going support of the University of Auckland and University of Otago. Thanks to Joan Lawrence and Hamish Macdonald for figures and to referees who provided many helpful comments.

- 38 Page is blank