Volume 104 1995 > Volume 104, No. 3 > Volcanic glass in Samoa: A technological and geochemical study, by Jeffrey T. Clark and Elizabeth Wright, p 239-266
VOLCANIC GLASS IN SAMOA: A TECHNOLOGICAL AND GEOCHEMICAL STUDY
Volcanic glass 1 was a resource long used by the prehistoric inhabitants of western Oceania. It became an item of exchange very early and appears to have been a prestige good over much of the region (Kirch 1988a). The exchange of volcanic glass has been documented for the Bismarck Archipelago as far back as 19-20,000 B.P., with the resource transported over some 350 linear kilometres (Allen, Gosden, and White 1989). Volcanic-glass use and exchange continued and millennia later were elements of the Lapita cultural complex. It is now clear, however, that those activities began long before Lapita, and continued after its disappearance. During the Lapita period, the distance over which volcanic glass was transported expanded substantially.
The involvement of volcanic glass in inter-island exchanges was most prominent in the western portion of Melanesian Near Oceania. This is not surprising given the long period of occupation in that region and the fact that three major volcanic-glass (obsidian) source areas have been identified there (Smith, Ward, and Ambrose 1977; Ambrose, Bird and Duerden 1981).In the Admiralty Islands, Lou and nearby Pam Lin and Pam Mandian, which are small islands south of Manus, were long-exploited sources of volcanic glass. The Lou Island volcanic glass was the most abundant and widely distributed. Similarly valued material in the Bismarck Archipelago came from the north coast of New Britain in the Willaumez Peninsula-Cape Hoskins area near Talasea. In the D'Entrecasteaux group, Fergusson, Dobu, and Sanaroa islands have volcanic-glass flows, but the western end of Fergusson Island has the best and most abundant glass (no Sanaroa volcanic glass has been recovered from archaeological sites) (Ambrose, Bird, and Duerden 1981:4).
Volcanic glass was used by island societies in Remote Oceania, as well. In central Remote Oceania (i.e., excluding marginal East Polynesia and Micronesia), only two source areas have been documented previously. One of these areas is in the Banks Islands of Vanuatu, where there are sources on Gaua and Vanua Lava Islands, the former being of best flaking quality. Ambrose and colleagues (1981:5) observed that these materials might better be termed “pitchstones”, and the flaking quality was inferior to that of the - 240 Lou and Talasea areas. The other source is on Tafahi Island in northern Tonga, though this source is poorly documented. Inter-island movements of volcanic glass have been demonstrated for the Reef/Santa Cruz Islands (Green 1987; Green and Bird 1989) and Fiji (Best 1984, 1987) but not for West Polynesia.
Recent archaeological investigations on the island of Tutuila in American Samoa revealed an early site at 'Aoa where volcanic-glass artefacts are unusually abundant. In fact, 'Aoa has produced more volcanic-glass artefacts than all other sites in the Samoan archipelago combined; in the Fiji-West Polynesia region, only four other sites, all on Niuatoputapu, have produced larger quantities of volcanic glass. In this paper we describe the 'Aoa volcanic-glass assemblage, including geochemical characterisations of a sample of the collection, and we compare the 'Aoa material with that from other sites in the region. We conclude that the 'Aoa artefacts reflect a lithic technology broadly similar to that found on other islands. Furthermore, Tutuila supplied volcanic glass to other islands in the archipelago and thus participated in an inter-island exchange network.
The Samoan islands sit on the Pacific plate about 120 km north of the westward-trending portion of the Tonga Trench. From east to west, the islands forming this chain are Ta'ū, Olosega, Ofu, 'Aunu'u, Tutuila, 'Upolu, Manono, Apolima, and Savai'i (Fig. 1). The first three of these islands are collectively referred to as Manu'a, which together with Tutuila and 'Aunu'u compose American Samoa; the latter four islands constitute Western Samoa. The chain comprises a series of basaltic, shield volcanoes and appears to reflect a hotspot origin (Duncan 1985). Post-erosional volcanism, however, has confused the picture somewhat. Savai'i, for example, has been blanketed recently by massive post-erosional eruptions that have obliterated most traces of any pre-existing shield volcanoes (Stearns 1944; Hawkins and Natland 1975; Wright 1986).
The subaerial shield-building lavas are mostly alkalic basalts, with Ti and large-ion lithophile element (LILE) contents that are high, even by Pacific hotspot island standards (Wright 1986). On 'Upolu, Tutuila, and Manu'a, large central calderas formed in which highly alkalic post-caldera lavas were erupted (Stearns 1944; Macdonald 1944; Stice and McCoy 1968; Natland 1980; McDougall 1985; Natland and Turner 1985). On Tutuila and 'Upolu, trachyte plugs were intruded along caldera ring faults and through the flanks of the shield volcanoes (Stearns 1944). No trachytes are found in Manu'a.
The island of Tutuila is the result of coalescence of four major shield-- 241 - 242
Figure 2. Map of Tutuila and 'Aunu'u islands, showing the major volcanic areas, trachyte plugs, 'Aoa Valley, and site AS-21-5.
volcanic centres: from east to west, the Olomoana, Alofau, Pago, and Taputapu volcanoes (Stearns 1944) (Fig. 2). While most of subaerial Tutuila formed 1.52 million years ago, the Tāfuna Plain and the nearby island of 'Aunu'u are more recent, perhaps dating to the Holocene (Stearns 1944; McDougall 1985). Only the Pago volcanic centre exhibits a central caldera structure (Pago Pago Harbor) and extensive trachyte plugs (Matafao and Pioa), though two small trachyte plugs are found in the Alofau and Olomoana volcanic complexes.
'AOA SITE AS-21-5
Despite numerous archaeological studies on Tutuila, only two sites have produced volcanic-glass artefacts, and only one of these is a residential settlement that has yielded multiple specimens. This latter site is AS-21-5, Locality 2, in 'Aoa Valley on the north coast of eastern Tutuila. 'Aoa Valley is formed by steep ridges radiating out from Olomoana Mountain to the south, which rises to about 326 m in elevation, and Le'aeno Mountain to the west, at 290 m elevation. The main part of modern 'Aoa Village occupies the eastern lobe of the valley, with a second and much smaller village, Fa'alefu, along the western margin of the valley. A few other houses are scattered through the valley.
Archaeological remains (surface features, artefacts, and midden) were found throughout much of the valley. Most of those remains compose a single site, AS-21-5, although for convenience, 16 separate localities were designated based on some geomorphic feature or geographic separation. Survey of the Puna Stream bed revealed basaltic-rock artefacts, pottery sherds, and darkly stained soil layers in the stream bank. This area was designated Locality 2 and tested through excavations.
A detailed discussion of the excavations at 'Aoa is presented elsewhere (Clark and Michlovic in press); here we will provide just a bare outline. Limited test excavations took place in 1986 and 1988, with expanded work in 1991 (Clark and Herdrich 1988; Clark 1989; Clark and Michlovic in press). In total, about 17 cu. m were excavated in eight units covering an area of 13.5 sq. m. Excavated soil was water screened through six millimetre-mesh hardware cloth. The gravelly, clay soil made the identification of small chips of volcanic glass difficult, but the water screening substantially improved recovery. Eighteen subsurface features — including postmolds, earth ovens, fireplaces, and rock alignments — were identified in the stream bank and/or encountered in excavation. The most common artefacts recovered from excavations are flakes of basaltic rock, properly termed hawaiite, which number 3566. Other hawaiite artefacts include 12 adzes, 12 preforms, - 244 15 flake tools, a hammerstone, and a grinding stone (for the manufacture of adzes). Also recovered were one siliceous (probably chert) flake and 878 pottery sherds. The sherds are generally small, lack surface decoration, and are predominantly from simple bowls. The remaining artefacts are of volcanic glass.
Two primary cultural components were identified at 'Aoa on the basis of stratigraphy, radiocarbon assays, and variation in the artefact collection. The lower component, comprising Layers VII and VIII, produced dates of cal 3455-2759 B.P. near the base of the deposit and cal 2764-2195 B.P. from higher up (at two sigma) (Clark 1993). 2 The upper component, consisting of Layers II-V, yielded dates between about 500-300 B.P. Separating these two components is a buried A-horizon, Layer VI, which appears to represent a pedologic discon-formity. The upper component contains pottery and volcanic glass in small amounts while hawaiite artefacts (notably flakes) are abundant; the lower component contains abundant pottery and volcanic glass but comparatively fewer hawaiite artefacts. This late ceramic-obsidian layer conflicts with the conventional model of Samoan prehistory in which both of these materials drop out of the material culture inventory in the first few centuries A.D. (Green 1974b). The 'Aoa evidence, however, suggests that pottery and volcanic glass use, in at least parts of the Samoan archipelago, continued for perhaps a thousand years longer than previously thought (Clark and Michlovic in press).
VOLCANIC-GLASS ARTEFACTS FROM 'AOA
Volcanic-glass artefacts recovered from the excavation totalled 276 flakes, chips, and cores. In addition, a couple of pieces of glassy-looking clinopyroxene megacrysts, black and green, were collected, as were a few weathered, natural pebbles. The distribution of volcanic-glass artefacts by unit and by layer is presented in Table 1. As indicated, volcanic glass was most common in the lower cultural component (Layers VII-VIII), and comparatively sparse in the buried paleosol (VI) and the upper component (II-V).
The flakes predominantly show typical attributes, such as identifiable striking platforms, faint bulbs of percussion, and dorsal flake scars. Also included in this category are small chips without such characteristics. Unmodified flakes were the most common type of artefact at 154 specimens (55.8%), with an additional 24 flakes (8.7%) bearing some evidence of tool use.
Cores are represented by 98 pieces (35.5%), from wasted cores to small- 245
Table 1. Stratigraphic distribution of volcanic-glass artefacts at AS-21-5.
pebbles with only a few flake scars. While the percentage of cores seems to be high, it is not unusual for Samoan collections. This seems likely to be due to the small sizes of the raw materials and the limited number of flakes that can be detached from any core. Some cores appear to be split pebbles with one or more flakes detached from the ventral side. In his discussion of the volcanic-glass artefacts from Sasoa'a (Sa-3), 'Upolu, Green (1974a:149) observed that one of the cores was comparable, though on a micro scale, to what White (1968) termed “scalar cores” in his volcanic-glass collection from Lesu, New Ireland. Several 'Aoa cores also seem to be very small examples of scalar cores.
On many of the cores the flake scars are so small that the detached flakes are unlikely to have been functional as tools (Fig. 3). Add to this the fact that the scars are usually multi-directional and it suggests that the intent of the flaking in some instances may have been to produce small multi-faceted cores rather than the detachment of usable flakes. If this is so, those cores may have been items of ceremonial and/or prestige value.
The volcanic-glass artefacts are all small, as shown by the calculated mean, minimum, and maximum measurements given in Table 2A. Mean core length and width are only slightly larger than the comparable means for flakes, but the thickness and weight measurements are noticeably larger. These sizes are comparable to the collections from other sites in Samoa, but smaller than the averages from Niuatoputapu (Kirch 1988b) (Table 2B). Size averages from Tikopia (McCoy 1982) and Reef/Santa Cruz Islands (Sheppard 1992) are comparable to those from Niuatoputapu, all three areas producing flakes and cores roughly twice the size of the Samoan material. These differences and similarities relate to the size of the raw materials and, consequently, the type of sources exploited.
During the lab analysis, each piece was examined macroscopically and microscopically (up to 80x) for minute characteristics, evidence of edge damage, and colour characterisation. This examination revealed frequent evidence of bipolar flaking (anvil use) on both flakes and cores. This evidence consists of crushing and/or chipping at proximal and distal ends, compression rings on opposite ends, and diffuse bulbs of force. The bipolar method is hardly surprising given the small size of the artefacts, which would make hand-held flaking quite difficult. The use of bipolar flaking has been claimed for 'Upolu (Green 1974a), Niuatoputapu (Kirch 1988b), Tikopia (McCoy 1982), and Hawai'i (Schousboe, Riford, and Kirch 1983). From study of the collections from Lapita sites in the Reef/Santa Cruz islands, however, Sheppard found little or no convincing evidence of the bipolar method (1992, 1993).- 247
Figure 3. Illustrations of selected volcanic-glass artefacts from AS-21-5. Probable tools are represented by a and b, cores by c, e, and h (pebble cortex shown as stippled area), a possible tool by d, unmodified flakes by f, and a scraper tool by g.
Table 2. Size characteristics of volcanic-glass artefacts; measurements in millimetres.
A. Artefact assemblages from sites in Sāmoa.
B. Artefact assemblages from three sites on Niuatoputapu (Kirch 1988b:214-217).
The cores and the dorsal surfaces of flakes commonly revealed multi-directional flake scars indicating frequent turning of the cores while working, although no pattern of turning could be distinguished. This method of reduction has been noted for Hawai'i (Schousboe, Riford, and Kirch 1983) and Tikopia (McCoy 1982).
Only one example of a definite tool could be identified in the assemblage. It is a small scraper from Layer VII with chipping along the distal end and along the sides (Fig. 3g). It is quite similar in form, though miniature in size, to hawaiite scraper/graters that are common on Tutuila. Edge damage suggestive of tool use was found on an additional 23 flakes. Based on the extent and patterning of damage, these flakes have been divided into two categories: probable tools (7) and possible tools (16). This division is somewhat subjective, but is based on the extent and patterning of the edge damage. Many other flakes also have some limited edge damage, but that damage seems more likely to be fortuitous than from use. The low occurrence of identifiable tools is consistent with findings from Western Samoa (Green 1974b), Niuatoputapu (Kirch 1988b), and Tikopia (McCoy 1982), although Best (1984:433) claims that about half of the 20 volcanic-glass artefacts from Lakeba have usewear.
From macroscopic inspection, the artefacts were divided into three groups based on colour: black, brown, and olive. Microscopic examination resulted in subgroups of the black and brown types. The vast majority of the collection (260) consists of black pieces that generally have a dull black gloss from weathering, but where fresh breaks occur the surface often has a shiny appearance. One black (Munsell colour 7.5YR2/0) subgroup is marked by small, circular, white phenocrysts. Two other black subgroups were differentiated by a more translucent quality around thin edges and slight variations in colour (7.5YR3/0). A fourth subgroup was identified by a rather distinctive molasses-like colour (5YR2.5/2) observed microscopically at the thin edges. The brown pieces (15 total) have varying amounts of brown (2.5YR3/2) to dark reddish-brown (2.5YR3/4) areas mixed with black areas. Three subgroups were differentiated by the quantity and colour of brown; a few pieces have very little black while others have very little brown. The single olive flake is actually predominantly very dark gray (5Y3/1) but there is a band of dark olive gray (5Y3/2). The brown pieces were only in Layers VI and, especially, VII, and the olive piece was from Layer VIII.
Cortex was noted on 87 (31.5%) artefacts. One type of cortex, with 48 cases, is distinguished by pocked, though smooth and flat, highly-weathered surfaces suggesting long exposure. Another ten examples have rougher surfaces that are more deeply pocked to vesicular, with the pocks often - 250 occurring in linear bands. Pebble cortex is present on 29 pieces, which have brown, crusty, rounded surfaces. No correlation between colour and cortex type could be discerned. An additional 72 pieces have at least some area with a differentially weathered surface, though it is not cortex. The high frequency of pieces with cortex or weathered surfaces suggests locally available material and comparatively little reduction of small pieces of raw material. The extraction of the raw material was probably from secondary sources, such as breccias near appropriate plugs, sediments in the area of plugs where nodules can be recovered, and possibly from stream beds. As illustrated by the work of Torrence et al. (1992) around Talasea, New Britain, such sources were widely used in the Pacific.
Fourteen samples of glass from excavations at 'Aoa were analysed for major elements using the scanning electron microscope energy-dispersive (EDAX) attachment. This work was conducted at the Department of Geological Sciences at the University of Illinois at Chicago. Three to six analyses were made of each glass sample, and the values averaged, as all samples were relatively homogeneous (within 1% for SiO2 and Al2O3, within 14% variation for CaO, Na2O and K2O). Basaltic and rhyolitic standards were used for calibration. The provenances and general characteristics of these archaeological samples are given in Table 3. These samples were selected to cover the range in excavation units, strata, and artefact characteristics. Two other glass samples were analysed, also. One is a large piece found on the ground in 'Aoa Village and said by some local residents to have been brought in from California for building purposes. One informant, however, claimed the “village obsidian” came from an outcrop, so this piece was analysed to ascertain its potential as a source for the archaeological samples. The second piece is true obsidian from the Mono Crater area of California that was included for comparison.
The geochemical measurements for the 14 archaeological samples as well as crystalline trachyte reference samples from Tutuila and 'Upolu are summarised in Table 4. The village and Californian glass samples (not included in Table 4) are more siliceous but at the same time less aluminous, sodic, and potassic than the archaeological glasses. Ratios such as Na2O/K2O, (Na2O+K2O)/SiO2, and CaO/Al2O3 clearly distinguish them from the other glasses, as do their appearance. On plots of Alkalies versus SiO2 (Fig.4), TiO2 versus FeO* (Fig. 5), CaO/Al2O3 versus Al2O3, the artefacts cluster fairly tightly together while the village and- 251
Table 3. Volcanic-glass artefacts from AS-21-5 that were geochemically analysed.
Table 4. Compositions of Volcanic Glass and Trachyte, Samoa
Californian samples are quite separate. It thus seems likely that the village volcanic glass was indeed brought in from off-island. Furthermore, on many plots, the village and Californian samples are closer to each other than either is to the Tutuila samples, lending credence to the idea of a Californian origin for the village volcanic glass.
There is no significant difference in the composition of the microscopically and macroscopically differentiated archaeological glasses. In fact, the variation between flakes is approximately the same as the variation within individual flakes, suggesting that all flakes found at 'Aoa came from the same, slightly heterogeneous, glass source. Even the distinctive brown samples fall within the cluster on all element plots. The single olive sample also fell within the cluster, though on the edge: it has the lowest SiO2, Na2O+K2O, and Al2O3, the highest FeO*, but falls right in the middle in TiO2, CaO, Na2O, and K2O. It is possible that this is a result of weathering, which reduced SiO2, Na2O, and K2O. In any case, it clearly clusters with the 'Aoa archaeological samples and not with any other known source.
When the glass samples are compared with differentiated (but not glassy) lavas from Tutuila (SAMTR-1 and TPI-8B) and 'Upolu (UPO-12F) (Natland 1980), several trends may be observed. No known 'Upolu lavas are as siliceous as the Tutuila glasses. This has been explained by Natland (1980) as the result of different primary magma compositions between the two islands, which produced independent crystallisation sequences, leading to different endpoints. As a result, differentiated Tutuila magmas are more siliceous than 'Upolu differentiates and show lower CaO/Al2O3 ratios. They are also depleted in FeO* because of the removal of titanian magnetite.
In Figures 4 and 5, along with other plots, the archaeological glasses are compared with the most-differentiated Tutuila and 'Upolu lavas sampled (Natland 1980). These differentiated lavas are crystalline trachytes, not glass; no siliceous volcanic glass has been identified in any mapped outcrop on a Samoan island. On all plots, the glass artefacts plot near the Tutuila lavas and are distinct from the 'Upolu lavas. Furthermore, the glasses plot close to Tutuila differentiation trends defined by Natland (1980) and differently from 'Upolu differentiation trends (Fig. 4). This is consistent with the work of Sheppard and others (1989) on glass samples collected from archaeological sites in 'Upolu. They note that their samples resemble Tutuila lavas in major elements. When their glass samples are compared with the glasses from Tutuila, it appears most probable that the two groups come from the same source. Artefacts from 'Upolu fall within the field defined by the Tutuila artefacts on all plots (e.g., Fig. 4).
Sheppard et al. (1989:72), however, also note that the trace element- 254
Figure 4. Plot of alkalies versus SiO2, showing fields for Tutuila and 'Upolu basalts and their differentiation trends (from Natland 1980). The Tutuila magma lineage is one of higher SiO2 at a given alkali value, and it extends to higher absolute SiO2 values.
compositions of the 'Upolu archaeological glasses may more closely resemble 'Upolu lava compositions than Tutuila compositions. But trace elements are troublesome petrogenetic indicators in highly fractionated rocks because trace elements may be strongly affected by relatively small variations in the proportions of crystallising phases and by the formation of accessory phases in late-stage magmas. Major elements are much less sensitive to local variation in accessory phases, reflecting instead the bulk crystallisation history. No lavas differentiated enough to form, upon quenching, the glass from which both Tutuila and 'Upolu artefacts were made have ever been sampled on 'Upolu, although the island is geologically well known (Stearns 1944; Kear and Wood 1959; Natland 1980; Natland and Turner 1985). In fact, Natland (1980) has suggested that magmas with the initial composition and crystallisation history of the 'Upolu lineage cannot produce late-stage lavas as siliceous as the Tutuila differentiates.
Volcanic-glass samples from 'Upolu and Tutuila may also be compared with sources of differentiated glass on other Pacific islands (Best 1987; Duerden et al. 1987). Tutuila and 'Upolu archaeological glasses are higher in Na2O+K2O and Al2O3 and lower in CaO and TiO2 than glass with comparable SiO2. They all show high Al2O3/CaO ratios (>23), whereas other Pacific glasses have Al2O3/CaO ranging from 10 to 19. As might be expected, the Samoan group is generally distinct from the other glasses in plots of alkalies vs. SiO2 (Fig. 4) and CaO vs. Al2O3. When TiO2 and FeO* are considered (Fig. 5), Samoan glasses show lower TiO2 content at all levels of FeO* perhaps reflecting fractionation of titanian magnetite.
Thus, it seems likely that the glass artefacts found in archaeological sites on Tutuila formed on Tutuila by quenching of differentiated liquids resembling those that cooled to form the quartz trachytes SAMTR-1 (from the Pioa plug) and TPI-8B (from the Matafao plug). This quenched magma would form glass, perhaps exposed near the base of one or more of these plugs as chunks in a breccia. The 'Upolu artefacts of Sheppard et al. (1989) probably came from the same source, and almost certainly came from Tutuila and not 'Upolu or any other known Pacific source. One of the key elements in sourcing the artefacts to Tutuila is TiO2, which is extremely low in both the Tutuila and the 'Upolu artefacts (it is simply given as <0.13 wt% in Sheppard et al. 1989). TiO2 is diagnostic precisely because it is so low: TiO2 was removed from Tutuila magmas by early crystallisation of titanian magnetite (Natland 1980).
Six non-archaeological samples of glassy material were also analysed, all from dykes around Tutuila: one from the east side of Ā;fono Valley, two from Masefau Valley, two from Fagasā; Valley, and one from Goat Island Point (on- 257
Table 5. Volcanic-glass artefacts from sites throughout Sāmoa.
the south coast at the Rainmaker Hotel). 5 These samples are much less siliceous than the archaeological glasses. They are basaltic glasses that reflect their sources at glassy selvedges at the edge of basaltic dykes, and such locations were not the sources of any known artefacts. Sheppard et al. (1989) also rejected Goat Island Point as the source of artefacts found on 'Upolu.
VOLCANIC-GLASS ARTEFACTS FROM OTHER ISLANDS IN THE CENTRAL PACIFIC
Volcanic-glass artefacts are rare at sites in the central Pacific (from Fiji to the Cook Islands). Many sites have not yielded any volcanic glass, some have produced small amounts, but a few sites have large quantities. A summary of sites in the central Pacific demonstrates that the volcanic-glass production from 'Aoa is unusually high.
Volcanic-glass artefacts are absent from most of the sites tested in Samoa, having been reported for only ten sites, including 'Aoa, in the entire archipelago. Furthermore, as illustrated in Table 5, at those sites that have produced these artefacts, the quantities are small, with five or fewer pieces at 70% of the sites. 'Aoa stands out prominently in the quantity of volcanic glass recovered.
The only island in Tonga that has produced volcanic-glass artefacts is Niuatoputapu, where sites have yielded the largest quantities of such artefacts in the central Pacific. Kirch (1988b:213) reports a total of 11,475 pieces from ten sites that produced from 45 to 8,778 pieces each. Kirch did not examine the Niuatoputapu collection for edge damage, but reports only occasional edge retouch and no evidence of formal tool production. When compared with the 'Aoa material, the Niuatoputapu artefacts show less cortex and are considerably larger.
Tafahi Island, an ash cone near Niuatoputapu, was observed to have abundant volcanic glass in exploitable condition two decades ago (Rogers 1974), but actual Tafahi artefacts and sites have not been reported. Poulsen (1987:214) noted two sites on Tongatapu with one unworked piece of glass each, but that material could not be local. Recent excavations in Ha'apai have not produced any volcanic glass (Dye 1987; Burley 1992). Thus, with the exception of Niuatoputapu, and presumably Tafahi, most Tongan sites lack archaeological specimens of volcanic glass.- 259
Geochemical characterisations of Niuatoputapu and Tafahi samples led Ward (1974a:169; 1974b:345) and Smith, Ward and Ambrose (1977:196) to suggest that Tafahi Island was the likely source for the Niuatoputapu artefacts. Subsequently, however, Kirch (1988b:215) located an “outcropping of volcanic glass…on the central ridge of Niuatoputapu”, and noted that while the outcrop glass has not been geochemically characterised, “it may well fall within the elemental ranges already determined.” Consequently, the actual source(s) of the Niuatoputapu glass artefacts remains to be demonstrated.
Other West Polynesia Islands
Volcanic glass has been reported in small amounts at early sites on Futuna (Kirch 1981), and Aubert de la Rue (1963:53, cited by Kirch 1976) reported black opaque “obsidian” on the island. Thus far, no volcanic-glass artefacts have been recovered from Uvea (Daniel Frimigacci, personal communication). The very limited work on Niue did not produce any volcanic-glass artefacts (Trotter 1979).
Only two islands in the Fiji group have produced volcanic glass: Lakeba, with 20 flakes from four sites, and Naigani with two flakes from one site (Best 1984, 1987). Some of the Lakeba flakes were sourced to Tafahi and others were suggested to have come from Vanuatu. Geochemical characterisations (PIXIE-PIGME) of the two Naigani flakes were compared with data from six possible source sites and matched most closely with the Talasea source in New Britain, some 3700 km to the west. The Naigani artefacts are approximately 3000 years old, indicating a very ancient pattern of exchange over very long distances, even if by down-the-line transfers.
The recent surge of archaeological investigations in the Cook Islands has substantially improved our knowledge of the prehistory of those islands. While a few pottery sherds have been found, no volcanic-glass artefacts have yet been reported. However, Allen and Steadman (1990:34) report several “unmodified nodules of poor quality volcanic glass” in deposits on Aitutaki.
DISCUSSION AND CONCLUSIONS
As can be seen from the discussion above, the use of volcanic glass in the central Pacific was highly variable in occurrence. The presence and - 260 quantities of these artefacts vary considerably from island group to island group, from island to island, and even from site to site.
Volcanic glass use also varies over time. On the basis of collections from western Samoa, Green (in Terrell 1969:168; Green 1974b:268) reported that volcanic glass in appreciable quantities occurred in association with pottery, both Lapita and Polynesian Plain Ware assemblages. Furthermore, volcanic glass use declined over time, eventually disappearing at about the same time as ceramics were abandoned, or perhaps slightly later (Green 1974a: 148). Minimal quantities have been reported from aceramic contexts (Davidson 1969:250; McKinlay 1974:33; Davidson and Fagan 1974:89), but in each of those cases the materials may have been in secondary context.
Recent work in American Samoa provides no contradictions to Green's conclusion regarding the association of pottery and volcanic glass. Excavations at' Aoa, however, indicate that pottery and volcanic glass continued in use, at some locations at least, until just a few centuries ago (Clark and Michlovic in press). At the same time, the quantity of volcanic glass dropped from the lower to the upper component, and within the upper component glass diminished from the bottom (V) to the top (II).
The paucity of volcanic-glass artefacts from most Samoan sites limits comparison. Nevertheless, the materials recovered from the various sites reflect substantial similarity. From a macroscopic examination of several 'Upolu flakes, we found them to be indistinguishable from the majority of the 'Aoa pieces. The presence of brown and olive flakes at 'Aoa marks the most pronounced difference. Similarities in the assemblages include the ratio of flakes to cores, the low occurrence of tools, the use of bipolar flaking, and the average artefact sizes. The Samoan volcanic-glass technology is not significantly different from that seen on other islands in the central Pacific. The artefacts are, however, significantly smaller than those from other sites in the region where volcanic glass is abundant. The small size of the artefacts is a product of the type of raw material available, not differences in the technologies.
'Aoa stands out from other Samoan sites—and from most sites in the region—in the quantity of volcanic glass recovered. It must be noted, however, that at most of the sites excavated in Samoa prior to 1980, the thick, clay-rich soils were not screened. Consequently, many pieces of volcanic glass undoubtedly went unnoticed, a factor recognised by Green (1974a:148). Even so, there seems little doubt that volcanic-glass artefacts are, in fact, more abundant at 'Aoa than at other excavated sites in the archipelago.
With the first discovery of volcanic-glass artefacts on 'Upolu, attempts were made to determine the source of the raw material. Kear (1967:1446) - 261 suggested that volcanic glass possibly could have come from Fagaloa Volcanics of Savai'i and 'Upolu. Green (in Terrell 1969:169) noted circumstantial evidence of a possible source in Fagaloa Volcanics surrounding Fālefā Valley of 'Upolu, including the presence of natural pebbles of volcanic glass in the alluvial sediments of the valley. But he was careful to point out that an actual source “remains to be demonstrated.” Ward (1974a) carried out trace element analysis on a sample of the archaeological collection from 'Upolu and reported that all of the volcanic glass probably came from a single source of unknown location. He also showed that the glass from the 'Upolu sites was not from Tafahi, Tonga, a conclusion that was later confirmed by Smith et al. (1977:196). Ward further concluded that while “siliceous magma required for the formation of a true rhyolitic obsidian appears to be absent from Western Samoa,” a source for volcanic glass similar to rhyolitic obsidian may still be present, especially in the Fagaloa Volcanics (Ward 1974a:167). At the same time, Tutuila was also cited as a possible source of the 'Upolu artefacts because of the presence of siliceous volcanics.
More recently, Kirch and Hunt (1988:10) listed Fagaloa on a map of “known obsidian” sources associated with Lapita sites. Duerden et al. (1987) include a volcanic-glass sample from Fagaloa in their obsidian composition catalogue, which may have lead Kirch and Hunt astray. That sample, however, actually was a natural pebble in the collection of Samoan materials at the Auckland Institute and Museum, and was not from an in situ deposit (Roger Green, personal communication). In actual fact, the evidence that Green properly stated was required before we can consider Fagaloa a source is still lacking. Moreover, on the basis of the data presented here, Fagaloa is not the source of any volcanic-glass artefacts geochemically characterised to date. Furthermore, only one Samoan site, Mulifanua on 'Upolu, has yielded characteristic dentate-stamped Lapita pottery and that site has not produced any volcanic glass. Consequently, Fagaloa is neither a known source of volcanic glass nor associated with Lapita sites.
Despite the appropriateness of the Tutuila volcanics and extensive surveys—archaeological (Clark and Herdrich 1993) and geological (Wright 1986)—no identifiable quarries, or even exploitable natural sources of siliceous rather than basaltic glass, have been found yet. Nevertheless, the geochemical evidence presented here strongly suggests that Samoans had a source of workable, quartz-trachyte, volcanic glass on Tutuila, and that this glass was exchanged between populations on Tutuila and 'Upolu, and perhaps other islands. This conclusion is bolstered by the unusual abundance of volcanic-glass artefacts from the 'Aoa site. These glasses probably formed - 262 on Tutuila by quenching of liquids resembling the differentiated lavas SAMTR-1 (from the Pioa plug) and TPI-8B (from the Matafao plug). Hence, the source is likely to have been in the Pago volcanic group, possibly in the form of silicic glass fragments found in the breccias surrounding the base of the plugs and formed during their emplacement (Stearns 1944). We hope future research will test this prediction.
This study was made possible by grants and support from the following organisations: The National Science Foundation, Washington D.C., Grant No. 911566; The Historic Preservation Office, Department of Parks and Recreation, American Samoa Government, Pago Pago; North Dakota State University; The University of Illinois at Chicago; and the School of the Art Institute of Chicago. We would also like to thank all those people who have assisted us in this work. Specific mention must go to David Herdrich, Michael Michlovic, Gene Harris, Martin Flower, Pat Bertnolli, Seth Schnieder, and Jared Erickson. We are grateful to the people of American Samoa, and in particular to the matai and other residents of 'Aoa and Fa'alefu villages for their hospitality. We specifically thank High Chief Soli Auemoeualogo and the late High Talking Chief Olomua Taua of 'Aoa.
1 This material is often referred to as‘obsidian’ While this term is technically correct in many areas of Oceania, in other areas it is not. Rather than shift terms in accord with variation in actual material, we will use the encompassing general term ‘volcanic glass’.
2 Radiocarbon dates are calibrated according to CALIB 3.03 (Stuiver and Reimer 1993) and are presented at two sigma. The slight variation of the dates presented here with those given in Clark (1993) are the result of different versions of CALIB for the two sets of calibrations.
3 refers to all iron calculated as ferrous iron. The sample U represents the average analysis of glass artifacts from sites on upolu (sheppard et al..1989). SAMTR-1 and TPI-8B (from Tutuila) and UPO-12F (from Upolu) are analyses of crystalline trachytes (NOT glasses) collected
4 Presumably these are flakes, but this was not specified in the report.
5 The five samples from the north coast (Āfono, Masefau, and Fagasā) were graciously supplied by David Herdrich.