Volume 107 1998 > Volume 107, No. 3 > Exotic and indigenous: Ceramic sherds from the island of Rotuma, by Thegn N. Ladefoged, Jonathan Wall, Philippa Black and William R. Dickinson, p 301-312
EXOTIC AND INDIGENOUS: CERAMIC SHERDS FROM THE ISLAND OF ROTUMA
Rotuma is a relatively small isolated island located approximately 400 km northwest of Vanua Levu, Fiji. The evidence for prehistoric interaction between the inhabitants of Rotuma and other islands has to date been rather minimal. Recorded Rotuman oral traditions refer to interactions with Tonga, Samoa and Fiji (Gardiner 1898; Churchward 1937a, b, 1938; Trouillet n.d; see also Howard 1985, 1986, and Ladefoged 1993, 1995). The Rotuman language is also suggestive of multiple inter-island voyages as 40 percent of the Rotuman wordstore are Polynesian loan words (Pawley 1996:86). Archaeological evidence of this interaction has been limited to connections with Tonga, based on the morphology of Rotuman burial mounds (Ladefoged 1993, 1995) and an analysis of limited skeletal material (Shutler and Evard 1991). Recent excavations on Rotuma have recovered a range of artefactual material including, for the first time, a small pottery assemblage that sheds light on prehistoric interactions with other islands. Through a petrographic analysis of the ceramic temper, it is possible to determine if the excavated pottery was manufactured on the island or imported from afar. The geology of Rotuma dictates the availability of temper material for indigenously made Rotuman pottery, and therefore provides the context for determining if the sherds were imported.
Rotuma is composed of volcanic cones and shield volcanoes surrounded by calcareous sand beach deposits (Fig. 1). Although a small portion of Rotuma is c. 1.5 million years old, the majority of the island was formed between 200,000 and 15,000 years ago, with the beach deposits forming in the last 5000 years (Woodhall 1987:20). The east of Rotuma is a largely undissected Pleistocene shield volcano locally mantled in places by a veneer of Holocene lava flows, and dotted along its crest by approximately 25 Pleistocene and Holocene scoria cones rising above the surface of the shields. The west of Rotuma is a satellite Holocene shield topped by two cones and flanked on the east by three tuff cones.
The volcanic rocks of Rotuma are collectively known as the “Rotuma Volcanic Group” and represent an alkalic basalt assemblage of olivine basalt and hawaiite (Woodhall 1987). The volcanics include Oinafa Basalt, Noatau Scoria and Afo'a Tuff (Woodhall 1987:14). The lava flows of the main shield volcanoes are Oinafa Basalt and include both pahoehoe and ‘a’ a flows which are the products of Hawaiian/ - 302 Strombolian-type subaerial volcanism erupting magmas of low viscosity (Woodhall 1987:20). The dominant phenocrysts in Oinafa Basalt are olivine, although both plagioclase and clinopyroxene phenocrysts are also present in some lavas. Noatau Scoria consists of pyroclastic rocks that were formed during Hawaiian/Strombolian-type subaerial eruptions. Afo'a Tuff are pyroclastic rocks produced by highly explosive Surtseyan-type phreatomagmatic volcanic eruptions resulting when water invaded volcanic vents. These tuffs are composed of palagonitic tuff and lapilli tuff, with some enclosed lava blocks, of the brownish to yellowish hues characteristic of hydrated basaltic glass. Microphenocrysts in brown to yellow palagonitic glass fragments of the Afo'a Tuff are commonly olivine and only rarely clinopyroxene. The most abundant monominerallic grains of sand in detritus derived from the Rotuma Volcanic Group are thus expected to be olivine.
Archaeological Excavations and the Ceramic Assemblage
The Maka Bay site is a prehistoric settlement located beside an extensive lagoon at Sovaea in Itu' muta (see Fig. 1). The site is approximately 30 metres inland of the high tide mark, and during a six week period in 1996 12 m2 were excavated. A detailed discussion of the excavation and the analysis of the recovered faunal and artefactual material is presented in Wall (1997). The surface of the site is relatively flat, and it was excavated in 10 cm arbitrary levels which on the basis of profiles and stratigraphy were divided into three intact prehistoric cultural layers. A total of 23 ceramic sherds were recovered from the site (Table 1).
Figure 1: The geomorphology of Rotuma with the location of the Maka Bay site (after Woodhall 1987:7)
Table 1: Sherds from the Maka Bay site.- 304
The top of the site was an approximately 15 cm thick disturbed layer that contained metal fragments and other historic artefacts. Below this sediment was an intact cultural layer which extended from c. 15 cm to a maximum depth of c. 50 cm below the ground surface. The layer contained faunal remains, Tridacna shell adzes and shell ornaments, and seven ceramic sherds, including two exotic sherds, one of which is paddle-impressed. A charcoal sample (Beta-106883) from Unit-1,0 Layer 1.3 at a depth of 30 to 40 cm below ground surface produced a conventional radiocarbon age of 1050±50 B.P., which, when calibrated with version 2.18 of the OxCal program (Bronk 1994), produces a 2 sigma calibrated age of A.D. 880 to A.D. 1110.
The middle cultural layer extended from c. 50 to 70 cm below ground surface. It contained a number of hearths, post holes, pits, artefacts and considerable faunal remains. Two sherds were recovered from the layer, including one rim sherd with faint paddle-impressed banding.
The lower cultural layer extended from c. 70 cm to a maximum depth of c. 100 cm below the ground surface, with the bottom of the layer being defined by a sterile clay sediment. A Rattus exulans femur and a charcoal sample from Unit 0,0 Layer 1.8 at a depth of 88 to 98 cm below ground surface were submitted for radiocarbon dating (Ladefoged et al. 1997). The rat bone (Beta-098659) produced a conventional radiocarbon age of 1400±40 B.P., which, when calibrated with the OxCal program, gives a 2 sigma calibrated age of A.D. 560 to A.D. 680, and the conventional radiocarbon age of the charcoal sample (Beta-098660) was 1290±60 B.P., with a calibrated 2 sigma date of A.D. 640 to A.D. 880. Excavation exposed a number of features in the cultural layer including hearths, post holes and midden concentrations. Artefacts associated with the layer include Tridacna shell adzes, shell ornaments and tools, hammer and grinding stones, and 13 ceramic sherds. A large faunal collection was also recovered from the layer. An additional ceramic sherd with no provenance was recovered when profiles were being drawn.
Ceramic Temper Analysis
After reviewing the geotectonic history of West Polynesia, Dickinson et al. (1996:87) concluded “that different island groups are bound to yield temper sands that are distinctive petrographically, although to varying degrees.” It is these distinctions in temper that enable us to determine whether or not the ceramics recovered from Rotuma were made on the island or imported from afar. Five of the 23 sherds recovered from the excavations were thin-sectioned, and the tempers in the sherds were analysed under a low-power microscope to determine their petrographic composition (Table 2).
Indigenous Sherd Specimens
Four of the Rotuman sherds (FS 280-1, FS 369, FS 438, FS 625) contain mineralogically and petrologically related volcanic sand tempers interpreted as being indigenous to the island. Sand grains of three of these sherds (FS 280-1, FS 438, FS 625) are predominantly volcanic glass fragments and subordinate olivine grains, although clinopyroxene grains are present in two of the sherds in minor amounts - 305 (see Table 2). The unaltered isotropic tan-to-brown volcanic glass fragments in these sherds are characteristically microvesicular and commonly contain microphenocrysts of olivine. Derivation of the glass-rich temper sands of the three sherds from the subaerial Noatau Scoria seems likely, although ‘a’a clinker from Oinafa Basalt has a similar lithology (Woodhall 1987). Sherds FS 438 and FS 625 contain moderately sorted and subrounded medium-grained temper sands of probable beach, but possibly stream, origin, whereas sherd FS 280-1 contains a poorly sorted, fine-to-coarse-grained, subangular sand of colluvial or alluvial origin. The presence in sherd FS 280-1 of non-glassy microcrystalline volcanic rock fragments may reflect definite contributions of detritus from lava rather than, or in addition to, scoria.
Table 2: Percentages of grain types in Rotuman sherds.- 306
The fourth sherd (FS 369) contains volcanic rock fragments that were apparently derived from exposures of Oinafa Basalt lavas. The lava-derived temper of the sherd is moderately well sorted, subrounded-to-subangular sand that might have been collected on a local shoreline; but no calcareous grains of reef origin are present, and the aggregate might alternatively have been recovered from a local stream course or ravine. The most abundant grains are volcanic rock fragments with the crystalline intergranular textures typical of Rotuman alkalic basalt lava. Accompanying mineral grains of sand size which would have originated as phenocrysts in the basalt include olivine, clinopyrozene, plagioclase and opaque iron oxide. Each of these minerals also occurs sparingly as microphenocrysts in the volcanic rock fragments.
When viewed in thin section, the olivine grains and microphenocrysts have a unique internal appearance which is related to the presence of minute festoon-like or grid-like chains of skeletal but conjoined magnetite inclusions. This is a unique pattern which has not been noted in other Pacific island tempers. Woodhall (1987:32) notes that magnetite in the Oinafa lavas of Rotuma occurs partly as “inclusions within and/or intergrowths with olivine phenocrysts,” and the presence of the unusual magnetite-charged olivine grains can be provisionally taken as strong evidence for an indigenous origin of the sand temper.
The high ratio of olivine to pyroxene in the indigenous Rotuman tempers indicates a temper of Oceanic basalt type. Although Rotuma is located south of the Vitiaz paleotrench within the Fiji platform of the northern Melanesian borderland, volcanism on Rotuma was evidently influenced by sea floor spreading within the young marginal sea of the North Fiji Basin. Rotuman temper is thus more similar to the Oceanic basalt temper of Uvea than to the andesitic arc temper of Futuna and Alofi, which represent a segment of the ancestral Tongan island arc offset westward along the Fiji Fracture Zone.
Exotic Sherd Specimen
A fifth sherd (FS 152-1) from Rotuma contains a well sorted and subrounded medium grained pyroxene-rich placer volcanic sand temper. A frequency count of 415 temper grains yielded the percentages shown in Table 2. These percentages are in contrast to the high ratio of olivine to pyroxene found in all potential Rotuman rock sources, and suggests that the sand temper of the sherd, and thus the sherd itself, was derived from outside of Rotuma. The texture and placer character, together with the presence of sparse calcareous grains of reef detritus, are diagnostic of beach origin.
The sherd could have been an import to Rotuma from one of several alternative places. Futuna and Alofi just to the east of Rotuma can be ruled out as the source of the anomalous sherd because tempers from both contain abundant plagioclase feldspar mineral grains (Dickinson 1976) that are wholly lacking in the exotic Rotuman sherd. As an alternative, the sherd could have been imported either from Fiji or from Tonga.
Past research (Dickinson 1971, Dickinson et al. 1990, Dickinson et al. 1996) has suggested that the ratio of orthopyroxene (opx) grains to clinopyroxene (cpx) grains within a sherd can be used to discriminate the placer character of Fijian and - 307 Tongan tempers. The ratio defined as cpx/(cpx+opx) has been referred to as the “pyroxene ratio” (Dickinson 1971, Dickinson et al. 1990, Dickinson et al. 1996), but here we use the preferable term the “pyroxene index.”. The pyroxene index for Fijian tempers is usually quite high (Dickinson 1971, Dickinson et al. 1990), on the order of 0.91 to 1.0. In contrast, the pyroxene index for the exotic Rotuman sherd is 0.85, a value similar to that observed for beach placer sand tempers in the Ha'apai Group of central Tonga (Dye and Dickinson 1996, Dickinson et al. 1996).
However, the presence of two percent hornblende in the exotic Rotuman sherd does not support the association of the Rotuman sherd with Tonga. Ewart et al. (1977:215-16) categorically state that hornblendes are lacking in Tongan rocks, and the few Tongan sherds from Ha'apai that do contain hornblende are thought to be exotic (Dye and Dickinson 1996:160). Hornblende is, however, present in the Lau Group (Cole et al. 1990:543) and in the main Fijian islands (Wharton et al. 1995, Gill 1987:601), and has been documented in Fijian sherds (Dickinson 1997).
The petrographic analysis of five Rotuman sherds suggests that four were manufactured on Rotuma while the fifth is an exotic. The remaining 18 sherds in the assemblage were examined with a low-power microscope, and on the basis of visual similarities with the thin-sectioned sherds were classified as indigenous or exotic (see Table 1). Seventeen of the 18 sherds are thought to be indigenous, and one additional parallel-lined paddle-impressed exotic sherd was identified.
Pottery sherds made on Rotuma were recovered from all three cultural layers. This would suggest that people were manufacturing pottery from perhaps as early as the 7th century until as late as the 11th century. The earliest phase of Rotuman prehistory has yet to be documented, but based on linguistic evidence (Pawley 1996) it is probable that the island was settled by people associated with Lapita ceramics several thousand years ago. The ceramics that have been recovered from Rotuma are probably a continuation of that pottery tradition.
The upper cultural layer of the site dates to sometime between A.D. 880 to 1110 and contains evidence for the continued manufacturing of pottery and also its importation. Our results suggest that two of the sherds, one of which is paddle-impressed, are exotic. While the pyroxene index of the exotic Rotuman sherd is similar to indices for Tongan sherds, the date of the Rotuman sherd is slightly later than would be expected, given current estimates of when pottery production ceased in Tonga. Kirch (1988:246) suggests ceramic production ceased on Niuatoputapu around A.D. 800, and Burley (1994:380) notes that the conventional date for the transition from Polynesian plainware to aceramic occupation in Tonga is around A.D. 200. However, Burley (1994:393) considers the possibility of later ceramic production in Tonga, as has recently been noted for Samoa (Clark 1996). The presence of hornblende in the exotic Rotuman sherd is much stronger evidence that the sherd came from Fiji and not Tonga. Paddle-impressed ceramics were manufactured in Fiji alongside other ceramic traditions from c. 200 B.C. until the late 18th century (Best 1984:295, Crosby 1988:213-19; also see Frost 1979:66). Exotic Fijian sherds dating to c. A.D. 1000 have been recovered from Tuvalu - 308 (Dickinson 1986, Takayama et al. 1987) and Tokelau (Best 1988:113), and Best (1984:644) notes an increase in inter-archipelago interaction as evidenced by exotic ceramics and lithics in Lau after c. A.D. 1025. The parallel-lined paddle-impressed decoration on one of the exotic Rotuman sherds, the overlap in dating of the sherd with Fijian paddle-impressed ceramics, the presence of other Fijian sherds in Tuvalu and Tokelau at about the same time, and the presence of hornblende in the Rotuman sherd all lead to the conclusion that the Rotuman sherd was imported from Fiji.
Several Rotuman myths refer to interactions with other Polynesian islands. While these myths do not necessarily chronicle precise historical events, they are probably reflective of historical processes and shed light on Rotuma's past (Howard 1985, 1986; Hereniko 1991; Ladefoged 1993,1995). In a similar but more literal manner, Geraghty (1993) considers the complexity of using Polynesian myths to document interisland contact and migration, and concludes that there is validity in the practice. For Rotuma, the “origin” myth of the island credits a Samoan or Tongan chief, Raho, with creating the island by pouring sand from a basket (Gardiner 1898:503-6, Churchward 1937b: 109-16; see Howard 1985 for an insightful analysis of the myth). Parke (1969:99) places the coming of Raho and the subsequent “formation” of Rotuma in the 13th century A.D. Extending further back in time, there is reference in the legend to the hanit e ma 'us (women of the bush), who are thought to represent the original inhabitants of Rotuma (Russell 1942:232). In some legends the hanit e ma'us are associated with Fiji (Parke 1969:100). The archaeological collaboration of the myth is minimal, but the evidence does support the notion that the situation is quite complex and that there was at least interaction between Rotuma and Fiji before the later 13th century Tongan influence on Rotuma.
The variable level of prehistoric contact between Rotuma and other Pacific islands is not clearly understood. The linguistic data suggests that Rotuma was first settled at the same time as other West Polynesian islands (Pawley 1996). The uniqueness of the Rotuman language indicates that there might have been a period of low interaction just after colonisation. The archaeological evidence from the Maka Bay site suggests that interaction with other islands existed during the latter part of the middle period on Rotuma, with the most likely contact being with Fiji. The linguistic (Pawley 1996), mythological (Howard 1985, 1986; Ladefoged 1993), and archaeological (Ladefoged 1993,1995; Shutler and Evard 1991) evidence all points to relatively extensive interaction with Tonga and Samoa during the late prehistoric period. While the archaeological evidence of these shifting spheres of interaction is limited, the presence of exotic sherds on Rotuma provides a glimpse of the island's complex prehistory.
We thank Sarina Pearson for her help in the field and her valuable comments and suggestions. While we were on Rotuma, Jiare Vafo'ou graciously shared his vast knowledge, and his family was extremely hospitable. Sepeti Matararaba of the Fiji Museum assisted with the fieldwork and provided many useful insights. The chief and the people of Itu'muta welcomed us into their community and helped us - 309 in countless ways. We also thank the Rotuman Council and the Fiji Museum for their support. Peter Sheppard, Roger Green, Geoff Irwin, Melinda Allen and an anonymous reviewer have all helped us clarify our thoughts. The project was partially funded by a grant from the University of Auckland Research Committee.
- 312 Page is blank
1 Temper Analysis