Volume 106 1997 > Volume 106, No. 4 > The maritime distribution of Bismarck Archipelago obsidian and island Melanesian prehistory, by Clayton Fredericksen, p 375-394
THE MARITIME DISTRIBUTION OF BISMARCK ARCHIPELAGO OBSIDIAN AND ISLAND MELANESIAN PREHISTORY
Obsidian forms a rather unimpressive component of the artefact inventory of many archaeological sites in Island Melanesia. With the notable exception of certain sites in the Admiralty Islands and western New Britain, obsidian is present mainly as unmodified flakes. Yet this material has been instrumental in writing the prehistory of the region. This is for two related reasons: (1) obsidian has been found in sites covering a very broad spatial and chronological span; yet (2) the natural occurrence of obsidian in Island Melanesia is very localised, restricted to three widely separated areas (see below). This has excited prehistorians, who have viewed the Pleistocene occurrence of obsidian in occupation sites as evidence of some early form of trade (Allen and Gosden 1996:189), and its occurrence in widely separated late Holocene sites as representing a developed maritime trading system (Ambrose 1978). This article reviews current knowledge of the spatial-temporal distribution of Bismarck Archipelago obsidian in Island Melanesia and places it in the context of the ongoing debate over the origins of interisland long distance distribution around 3500 years ago.
This article is in four parts. The first introduces the location of known obsidian source areas in the region. In the second, a review of the archaeological evidence for obsidian dispersal since c. 20,000 B.P. is presented. Discussion is largely undertaken in respect of two broad phases, pre-3500 B.P. and post-3500 B.P., using the approximate time of the appearance of the “Lapita Cultural Complex” as the temporal demarcation line (while acknowledging that the radiocarbon period encompassed by this event is in the order of c. 3800-3600 B.P). The third part of the article examines chronological variation in the quantity of obsidian which was moved by sea. In the fourth part, the information is examined in the context of the current debate over whether the emergence of long distance distribution was a product of indigenous development, or a facet of the economy of a migrant population arriving out of Southeast Asia. These competing theories are critically examined in the light of the archaeological distribution of obsidian.- 376
OBSIDIAN SOURCES IN ISLAND MELANESIA
Four obsidian-bearing areas are present in Island Melanesia: the Admiralty Islands and New Britain in the Bismarck Archipelago, Fergusson Island in the D'Entrecasteaux group, and the Banks Islands of northern Vanuatu (approximately 800 km southeast of Rennell Island) (Fig. 1). After 3500 B.P, obsidian from all four source areas was distributed to various parts of Island Melanesia. Fergusson obsidian was dispersed to the islands of the Papuan Gulf and the southern Papuan coast but not beyond (with the notable exception of a single piece identified in the Reefs/Santa Cruz group, 400 km east of the southernmost Solomon islands [Green and Bird 1989]). Banks Islands obsidian was more limited in its dispersal and apparently was not transported further west than the Reefs/Santa Cruz. A spatio-temporal pattern of distribution quite different to either Fergusson or Banks sources is provided by Bismarck Archipelago obsidian. This material was first moved by sea well before 3500 B.P., possibly as early as 20,000 years ago (Summerhayes and Allen 1993), and the range over which it was distributed after 3500 B.P was far greater than that ever achieved by either Fergusson or Banks sources.
New Britain obsidian originates from two regions (Fig. 1). One is the Talasea area of the Willaumez Peninsula on west New Britain. Here there are a number of obsidian-bearing volcanoes: Baki, Hamilton, Kutau, Bao, Humgari and Gulu. The second obsidian-bearing region is at Mopir, to the east of the Willaumez Peninsula.
In the Admiralty Islands, natural occurrences of obsidian are on Lou, Pam Lin and Pam Mandian, small islands south of the main island of Manus. Obsidian has also been found on the south-western end of Manus in the vicinity of Mt Hahei. The use of Manus obsidian appears to have been very localised, as it has yet to be identified in archaeological assemblages outside the island.
PATTERNS OF SOURCE DISTRIBUTION
New Britain Obsidian
20,000-3500 B.P.: The earliest evidence of the offshore movement of Melanesian obsidians is found in the transport of New Britain obsidian to New Ireland at 20-18,000 B.P. (Summerhayes and Allen 1993, also Rosenfeld 1997:221). On current data, the geographic limit of the dispersal of New Britain obsidian prior to 3500 B.P. was New Ireland, the Arawe Islands immediately to the south and Nissan Island to the east. The Nissan occurrence represents the earliest “long distance” cross-water distribution of obsidian, with an associated date of c. 4850 B.P. (Spriggs 1991a:228). Nissan lies 110 km east of New Ireland and 57 km from the nearest land to the north-west. A claim has recently been made for the presence of Talasea obsidian at- 377
Figure 1: Island Melanesia showing main obsidian sources (indicated) and islands mentioned in the text.
6000 B.P. in the site of Bukit Tengkorak on the island of Borneo (Service 1996), approximately 3500 km west of New Britain. However, a summary of the excavation of this site places obsidian in levels dating to after 3000 B.P. (Bellwood and Koon 1989:621). The argument for an early Holocene association is therefore contentious and must be treated with caution until more contextual information becomes available.
Obsidian is present in four New Ireland sites in definite pre-3500 B.P. levels: Balof in levels as early as 7680±510 B.P. (Downie and White 1978, White et al. 1991:55), at Matenkupkum in units dating c. 12,000-10,000 B.P. (Summerhayes and Allen 1993:147), Matenbek in two phases dating c. 20,000 and 9-6000 B.P. (Summerhayes and Allen 1993:145), and possibly also at Buang Merabak in deposits dating from perhaps 20,000 B.P. to c. 4000 B.P. (Rosenfeld 1997:217). Sourcing has demonstrated that the Balof obsidian originated from Talasea, while around two thirds of the Matenkupkum obsidian was obtained from Mopir and one third from the Talasea area (plus a small percentage from two “unknown” sources) (Summerhayes and Allen 1993:147). Mopir obsidian also figures predominantly in both occupation phases at Matenbek, comprising between 70 percent and 80 percent of all analysed pieces (Summerhayes and Allen - 378 1993:146). The remainder originated from the Talasea area. At Buang Merabak only one piece was sourced to Mopir, while the remainder (13, or 93 percent) came from Talasea (Rosenfeld 1997:221). Obsidian possibly occurs in a pre-3500 B.P. context in one other New Ireland site, Panakiwuk. Here one piece was recovered from a level dating to c. 8000 B.P. (Marshall and Allen 1991:71).
On or near New Britain, obsidian is present at five localities in pre-3500 B.P. contexts. These are Buvussi, Walindi, Misisil and Yambon on New Britain itself, and Lolmo cave in the nearby Arawe Islands. The oldest occurrence is at Misisil in levels pre-dating 11,000 B.P. (Specht et al. 1981:14, Summerhayes et al. 1993:64). Two of the three pieces analysed originated from Mopir, while the remaining piece was obtained from the Talasea region (Summerhayes et al. 1993:64). Mopir obsidian, along with Talasea, is also present in early to mid-Holocene contexts at Buvussi, Yambon and Lolmo (Summerhayes and Allen 1993:144). It is, however, not represented among the eight pieces which have been analysed from Walindi (Summerhayes et al. 1993:64). Additionally, the proportion of Mopir v. Talasea obsidian appears to fluctuate between sites: in the Yambon and Lolmo assemblages, Mopir is not as well represented as Talasea (Summerhayes et al. 1993:64, 66).
Post-3500 B.P.: In the period after 3500 B.P., obsidian from the Talasea region was dispersed as far astward as Fiji (Best 1987, Kirch and Hunt 1988:21), a distance of almost 3000 km, and as far west as Borneo Bellwood 1989:153, Bellwood and Koon 1989:620). Obsidian also found its way into the eastern Highlands of Papua New Guinea (Watson 1986:4). Within the compass of this distributional region, New Britain obsidian occurs in sites on New Caledonia (Ambrose 1976:366), Vanuatu (Ambrose and Duerden 1982:84), and Tikopia (Kirch and Yen 1982:255-56) (all around 2000 km from New Britain), as well as the Reefs/Santa Cruz group (Green 1987:244) (1700 km from New Britain), Buka (Wickler 1990:147), islands between Buka and New Ireland (Ambrose and Duerden 1982:4, Spriggs 1991a:229), New Ireland and adjacent islands (Golson 1991:255, Green and Anson 1991:177, White et al. 1991:54), the Mussau Islands (Kirch et al. 1991:157), the Arawe Islands (Summerhayes and Hotchkis 1992), areas of the northern coast of Papua New Guinea (White 1996:201), and, unsurprisingly, on New Britain itself. No New Britain obsidian has been identified from the Admiralty Islands. While it can be argued that this would not be expected given that obsidian sources are present in this group, a coals to Newcastle scenario does seem to fit other obsidian distribution routes: Admiralties obsidian has been found in Lapita sites on south New Britain (Summerhayes et al. 1993:65), while obsidian from the Bismarck Archipelago is present in - 379 sites in the Reefs/Santa Cruz group (Green 1987:245) and on Tikopia (Kirch and Yen 1982:255-56), despite the much closer location of the Banks Island source (300 km south of Santa Cruz and 200 km south-west of Tikopia). The discovery of Bismarck obsidian in the Vanuatu chain itself (Ambrose and Duerden 1982:84) is further testimony to this Bismarck focus in obsidian exploitation at this time.
The Boduna and Walindi sites in the Talasea region have yielded obsidian exclusively from Talasea sources. Lapita-age levels at the Yimilo site on the southwest coast of New Britain exhibit an equal proportion of Talasea and Mopir obsidian (two pieces of each), while post-Lapita levels contained more Talasea (four pieces v. one Mopir). At the sites of Misisil and Yambon in interior southern New Britain (the Kandrian region), obsidian from Lapita-age levels was apparently derived solely from Talasea sources. However, in both sites Mopir obsidian makes a reappearance in post-2000 B.P. levels. Summerhayes and colleagues (Summerhayes and Hotchkis 1992:132, Summerhayes et al. 1993:66) make the point that Mopir obsidian appears to be relatively uncommon in Lapita contexts and propose that this may be due to the source becoming temporarily inaccessible following the eruption of Witori volcano. One interesting suggestion they make (Summerhayes and Hotchkis 1992:132) is that the appearance of Admiralties obsidian at around 3500 B.P. in number of New Britain localities (see below) may have been the result of the Admiralty Islands filling a “vacuum” in obsidian production created by the Witori eruption (see also Torrence et al. 1992:91).
Admiralty Islands Obsidian
20,000-3500 B.P.: Two sites in the Admiralty Islands have occupation unequivocally dating back to before 3500 B.P. These are the rockshelters of Peli Louson and Pamwak, both on Manus. Two small test holes excavated at Peli Louson revealed obsidian in a mid-Holocene deposit (Kennedy 1983). Most of the obsidian from this site was sourced to the Pam Islands, with the remainder to Lou (Fredericksen 1994:98). More comprehensive excavations undertaken at the Pamwak site uncovered an obsidian sequence extending back to the terminal Pleistocene (Fredericksen et al. 1993). Obsidian first occurs in levels dated to approximately 12-11,000 B.P. and was obtained from an unknown source, probably located in the region of Lou and the Pam Islands (Fredericksen 1997:71). This represents the earliest evidence of the dispersal of Admiralties obsidian. Significant changes are evident in source use at Pamwak between Pleistocene introduction and late Holocene times. At some time between c. 8000 B.P. and 6000 B.P, the Pam Islands replaced the unknown source as the major provider of obsidian (Fredericksen 1997:72). At Pamwak, Lou Island obsidian only appears in significant - 380 quantities in levels dated to between 8000 B.P. and 3000 B.P., possibly toward the upper end of this age range (Fredericksen 1997:72).
Post-3500 B.P.: The period after 3500 B.P. witnessed the first transport of Admiralties obsidian beyond the island group. Obsidian came to be distributed over an enormous range, almost equalling that of New Britain obsidian. Admiralties obsidian reached Vanuatu in the east (Ambrose 1976:362) and as far westward as the island of Borneo (Service 1996). Obsidian has been recovered from sites along the northern coast of Papua New Guinea (Ambrose 1976:359), and may have been dispersed in the opposite direction as far as Micronesia (Ambrose 1978:330), a distance of some 1000 km. Within this wide region, Admiralties obsidian has been identified in Lapita and post-Lapita contexts on the Reefs/Santa Cruz Islands (Green 1987:243), Buka (Ambrose and Duerden 1982:84), islands between Buka and New Ireland (Ambrose 1976:361, Spriggs 1991a:229), New Ireland (Golson 1991:255, Marshall and Allen 1991:71, Rosenfeld 1997:221, White et al. 1991:54), the Mussau Islands (Kirch et al. 1991:157), New Britain (Summerhayes et al. 1993:64-66), the Arawe Islands (Summerhayes and Hotchkis 1992:130), as well as throughout the Admiralties group.
A significant feature of this external dispersal is that Admiralties obsidian replaced New Britain obsidian in many sites, particularly those on northern New Ireland and smaller islands down to Buka. On the islet of Sohano in the Buka Passage, the use of Admiralties obsidian continued throughout the post-3500 B.P. sequence, while importation of obsidian from Talasea appears to have ceased in the post-Lapita period (Wickler 1990:150-51; see also Ambrose and Duerden 1982:84). On Nissan, Talasea provided between approximately one third and one half of the imported obsidian during Lapita times, but Admiralties obsidian came to predominate in post-Lapita times (Spriggs 199la:229). A similar picture of a post-Lapita emphasis on Admiralties obsidian is reflected at sites in the Feni, Lihir and Tabar island groups, northwest of Nissan (Bird et al. 1981). Similarly, Lapita occupation levels in Mussau Islands sites contain mainly Talasea obsidian, while closer-at-hand Admiralty Islands obsidian predominates in post-Lapita occupation (Kirch et al. 1991:157).
Interestingly, the “sphere of influence” of the Admiralties connection does not appear to have continued down the island chain into Remote Oceania during post-Lapita times. In the Reefs/Santa Cruz sites, the proportion of both Admiralties and Talasea obsidian declines after 2000 B.P. in favour of closer Banks Islands material (Green 1987:247). On Tikopia, a similar picture of a switch to more proximate Banks Islands obsidian is evident for the period after the early settlement phase (c. 2900-2100 B.P.) (Kirch and Yen 1982:256).- 381
Admiralties obsidian is also present in significant proportions in New Ireland sites. Four sites have obsidian in layers relating to the post-3500 B.P. period: Balof (Downie and White 1978, White et al. 1991), Lossu (formerly Lesu) (White and Downie 1980), Dori in the village of Lasigi (Golson 1991), and Panakiwuk (Marshall and Allen 1981). Although stratigraphic change in obsidian source use cannot be determined from the data, it is clear that Admiralties obsidian attains major importance in this period. Twenty pieces analysed from Balof originated from the Admiralties, while 13 are provenanced to New Britain sources (Downie and White 1978: Table 12). For Lossu, 17 of 20 sourced pieces derive from the Admiralties (Lou) (Ambrose and Duerden 1982:84). Obsidian occurs at the Dori site in all levels from approximately 2000 B.P., and of the 65 pieces analysed, 60 were transported from the Admiralties, while five originated in the Talasea area (Golson 1991:255). A small assemblage of only nine obsidian pieces was found in Panakiwuk shelter, seven of which originated in the Admiralties (Marshall and Allen 1981:71).
Two sites in the New Ireland/north-east New Britain region, Balof and Reber-Rakival on Watom Island, present exceptions to this general pattern of favouring Admiralties obsidian over geographically closer west New Britain material. At Balof, all obsidian came from New Britain with the exception of pieces in the uppermost (undated) horizon, for which two of the six pieces recovered originated on Lou in the Admiralties (White et al. 1991:54). The earliest horizon at Reber-Rakival (dated approximately 2200-2000 B.P.) revealed equal amounts of Talasea and Admiralties obsidian, plus a small proportion of obsidian from Mopir (Green and Anson 1991). The upper horizon (2100-1900 B.P.) possessed a significantly greater proportion of Talasea obsidian, while the quantity of Admiralties obsidian dropped to only a little more than that from Mopir, the percentage of which remained unchanged (Green and Anson 1991:177). Clearly, a major factor in the Watom case must be the close proximity of this island to New Britain.
For New Britain itself, an extensive sourcing project by Summerhayes and colleagues (Summerhayes et al. 1993) has revealed that only small amounts of obsidian originated from the Admiralties. The occurrence of Admiralties obsidian is noted for the Apugi site just off the Kandrian coast (southwest New Britain), where five of the 39 pieces analysed from Lapita levels were sourced to Lou. The remaining identifiable pieces originated from Talasea (21) and Mopir (two). Admiralties obsidian also makes an appearance in all four of the investigated Arawe sites (Paligmete, Adwe, Apalo and Lolmo) in an approximate 3500-2000 B.P. chronological context. However, in each site the proportion of obsidian derived from the Admiralties is small (less than 7 percent). The remaining obsidian, where it can be - 382 identified to source, comes from Talasea.
Turning to the relative importance of Lou and Pam sources, chemical characterisation analysis has demonstrated that while obsidian from both sources was used within the Admiralty Islands, the relative importance of these sources varied between sites. Analysis of post-3500 B.P. assemblages from the Mouk, Kohin, Pamwak and Father's Water sites has identified the following proportions of Pam obsidian: 50 percent, 39 percent, 38.5 percent and 7 percent (from Fredericksen 1994: Appendix B, Fredericksen 1997:72). Investigation by Ambrose and colleagues (Ambrose et al. 1981) has shown that a recent (<220 B.P.) obsidian assemblage from the islet of Ahus, off the north coast of Manus, contained only a small proportion of Pam material (8 percent). What this admittedly sketchy picture may reflect is the operation of intra-regional distribution routes based equally on Pam and Lou or focused predominantly on the Lou source (perhaps those operating in the eastern part of the island group). A change in source emphasis within the post-3500 B.P. period is always a possibility and would be obscured in a broad-brush picture such as that presented here. The difficulty is that very little information is available for this period, which is represented stratigraphically at only three sites, Pamwak, Kohin and the Mouk site. For Kohin there is an indication of increasing emphasis on Lou obsidian from the upper half of the sequence on, but no such directionality is apparent in either the Mouk or Pamwak assemblages.
Calculation of the relative amounts of Lou and Pam obsidian dispersed beyond the Admiralties is limited by the fact that many commentators do not differentiate between these two sources. A notable exception is the research carried out by Bird for the Lapita Homeland Project (Allen and Gosden 1991), but the final results of this have yet to be published. Nevertheless, some information is available in other accounts. Ambrose and colleagues identified only Lou in the Admiralty Islands obsidian at Masahet (in the Lihir group), 98 percent Lou at Lossu (Lesu) on New Ireland, and perhaps 90 percent Lou on Eloaue in the Mussau group (Ambrose et al. 1981:16). Admiralties obsidian found in sites on Buka and Nissan may have all been Lou (Wickler 1990:147, Spriggs 1991a:239). Both pieces of Admiralties obsidian found in Balof on New Ireland were sourced to Lou (White et al. 1991:54). At Panakiwuk, also on New Ireland, seven obsidian pieces were sourced to the Admiralties, of which six originated on Lou (Marshall and Allen 1991:71). All Admiralties obsidian in the Reber-Rakival site on Watom was sourced to Lou (Green and Anson 1991:177-78). Finally, the small amount of Admiralties obsidian recovered from sites in the Arawe Islands originated on Lou (Summerhayes et al. 1993:63). This evidence clearly shows that only relatively small amounts of Pam Lin and Pam - 383 Mandian obsidian were distributed beyond the Admiralties. This stands in stark contrast to dispersal within the island group (see above).
VOLUME OF DISTRIBUTION
A second aspect of obsidian distribution pertains to the actual quantity which was moved between islands. This is a more difficult issue to address, however, as it relies on calculating the relative amounts of obsidian entering sites at particular points in time. Ideally, it should be possible to undertake comparison by tabulating the amount of recovered obsidian by cubic metre of excavated deposit for each site. This method provides a relatively straightforward way to standardise for inter-site comparison. Unfortunately, this relies on a detail of stratigraphic information not available for a significant proportion of sites in western Melanesia. Nevertheless, broad comparisons can still be obtained by comparing the absolute amount of obsidian recovered in terms of the size of the excavation. This information is set out in Table 1 for 15 sites where obsidian could only be imported by undertaking a sea crossing and for which published on both amount (number) of recovered obsidian pieces and size of excavation area are available. When looking at the absolute number of pieces in a site, rather than the relatively small number of pieces on which sourcing analysis has been undertaken, it becomes possible to refine the pre-3500 B.P./post-3500 B.P. division into five periods. This enables a better appreciation of change in the quantity of obsidian moved. Of course, the drawback in adopting any method of ordering time by discrete radiometric stages is that radiocarbon age ranges for some excavation units will invariably straddle stage boundaries. The absolute dates defining the period boundaries in Table 1 should therefore be interpreted with a reasonable degree of flexibility. Additionally, it must also be borne in mind that these excavations, as with most archaeological excavation, were test investigations, and there is little way of ascertaining whether or not they provide a representative sample of past obsidian use at any particular locality for any specific time.
Bearing these provisos in mind, the presented in Table 1 show that the amount of obsidian moved by sea during Pleistocene times was in most cases not large, even allowing for the generally smaller size of excavation carried out on deeper (older) deposits. The early movement of obsidian to Matenbek and Buang Merabak, beginning perhaps 20,000 B.P., is represented by only 43 pieces in total. At Matenkupkum, where obsidian first appears close to the Pleistocene/Holocene boundary, the number of pieces deposited in the site is greater, with 166 flakes recovered. Such an amount could easily be produced during one episode of knapping. These apparently small amounts could mean a number of things, with possibilities ranging from infrequent - 384 access of obsidian sources to the relative unimportance of obsidian for activities undertaken in caves and rockshelters, the only site types represented for this period. Nevertheless, it does seem that during this early period a small but consistent seaborne movement of obsidian was undertaken, a remarkable feat in the context of world prehistory.
Pamwak on the island of Manus is the odd site out in terms of the volume of obsidian transported in the Pleistocene. Here the presence of almost 4700 pieces is far in excess of the number recovered from other sites, even when taking into account any variation in the volume of excavated deposit. The first use of obsidian at Pamwak occurred approximately 8000 years after its first transport to Matenbek, so the argument could be made that this interim saw the development of more sophisticated marine technology, facilitating easier movement by sea. However, the appearance of obsidian in Pamwak is contemporaneous (at least in archaeological terms), with its initial use at Matenkupkum, from where a much smaller quantity of obsidian was recovered. An explanation for the Pamwak evidence must lie at least partially in the close proximity of this shelter to offshore obsidian sources, perhaps no more than 20 km away. However, a simplistic distance fall-off model seems unlikely to provide a complete explanation for variation in the amount of obsidian transported to sites. For example, the site of Kohin, also on the southern coast of Manus, contained far less obsidian in late Holocene contexts than contemporaneous levels at Pamwak (Table 1). The discrepancy is far too great to be explained simply by differences in the volume of Holocene age deposits at these two sites.
From the information presented in Table 1 it is difficult to see any clear regional trend in the amount of obsidian transported during the Holocene period. The number of obsidian pieces deposited in Matenbek appears to rise dramatically in the early Holocene, but this picture is not reflected at Pamwak. But the mid-Holocene at Pamwak sees a marked increase in obsidian deposition, which is associated with the formation of an estuarine shell midden (Fredericksen et al. 1993:148). Clearly, local situational contexts will need to be examined as regards any explanations of change in the amount of obsidian being moved into sites. Again, any statements must be tempered by the fact that no standardised quantification of the amount of obsidian entering these sites has been undertaken.
The beginning of the Lapita period can be correlated with 3500 B.P. in Table 1, but for some sites the end of Lapita came after 2500 B.P. This period saw the first occupation of new localities. However, comparison on a case by case basis is hindered by the fact that our evidence of pre-Lapita occupation is almost exclusively limited to rockshelters and caves. Comparing obsidian importation and use at these kinds of sites with its - 385 reduction in open sites such as Lossu, Reber-Rakival and the Reefs/Santa Cruz sites is more than a little problematical. Almost certainly a different range of activities took place within caves and open villages. If our subset of sites is limited to open settlements for which stratigraphic variability is held constant, then the influence of location becomes clear. The distant Reefs/Santa Cruz and Tikopia sites exhibit an obsidian flake density of fewer than five per square metre. The Reber-Rakival site on strategically placed Watom Island possesses 82.5 pieces per square metre, while between 13 and 30.5 pieces per square metre are represented in the three remaining open sites (Dori and Lossu on New Ireland, and Talepakemalai in the Mussau Group).
EXPLANATIONS OF DISTRIBUTION
The above overview presents a picture of change in obsidian distribution on a pan-Melanesian scale. This can be summarised in terms of a number of major points of change or articulation.
It is clear that 1500 years before the birth of Christ a major transformation took place in maritime commodity distribution in Island Melanesia. Previously restricted distribution, which involved coastal voyages or short water crossings, was supplemented by the addition of long-distance - 386 interisland communication. Ambrose (1997:533) has recently argued that the development of seaworthy sailing craft at around 3500 B.P. provided the mechanism for the first long-range maritime movement of obsidian and other commodities. However, if people were regularly transporting obsidian across 20-30 km of sea in Pleistocene times (Summerhayes and Allen 1993, Fredericksen 1997), had colonised the remote islands of Manus (Fredericksen et al. 1993) and Buka (Wickler and Spriggs 1988) by the terminal Pleistocene, and had occupied Nissan (Spriggs 1991a) in the early Holocene, then there seems to be little technical reason (Irwin 1992:28) why they could not have successfully undertaken long voyages of trade or colonisation by at least the early Holocene. The reason for the absence of long-distance interisland obsidian distribution before 3500 B.P. appears to lie instead with a lack of social or economic incentive to undertake such journeys (at least on a sufficiently regular basis to be archaeologically visible).
The appearance of long-distance obsidian transport therefore marks the appearance of a different form of social articulation than had previously existed. There came into being a need or desire to form or maintain social/trade ties with communities inhabiting distant islands. Concomitant with this areal expansion of communication was the appearance of a number of previously unknown techno-cultural components which have been collectively labelled the “Lapita Cultural Complex” (Green 1992). Notable innovations associated with this cultural complex included decorated pottery, a range of portable artefacts, new domesticated plant and animal species, and possibly a focus on coastal habitation (see Green 1991 and Spriggs 1995 for summaries).
Two competing views have been proposed to account for the appearance of this cultural complex; these can be termed “intrusion” and “indigenist” scenarios. Proponents of the intrusion hypothesis see the prime mover for change as a migration of pottery-using horticulturalists out of Southeast Asia (Bellwood and Koon 1989, Green 1992, Kirch 1996, Spriggs 1991b, 1995, 1996). These colonists introduced new traits and a new socio-economic system which acted as a catalyst for subsequent change. By contrast, the indigenist school views the picture painted by archaeology as an outcome of largely autochthonous development, albeit with some influences derived from contact with groups to the west (Allen and Gosden 1996, Allen and White 1989, Ambrose 1997, Gosden 1991, White and Harris 1997). The final push out beyond home regions was by no means an insignificant event, and underlying the chronological continuation of this long-distance communication within Near Oceania there must have been a strong social impetus. In fact, this impetus was so strong that once it obtained its own momentum commodity distribution continued for over 2000 years until the - 387 ethnographic present (although this is not to say that the way in which distribution was carried out failed to change over this time). The ascription of prestige or novelty status to certain locally-restricted commodities may have played a role in this process, a hypothesis first mooted by Kirch (1988:113-14). It is interesting to note that more recently other commentators have pondered whether obsidian (Sheppard 1993:135, Torrence et al. 1996:221) and Lapita pottery (Ambrose 1997:531, Sand 1997:546) served a sociological rather than, or instead of, simply a utilitarian purpose.
The key to the puzzle of the 3500 B.P. geographic expansion of obsidian distribution must lie in the long period preceding Lapita (Gosden and Specht 1991:278). Allen and Gosden (1996) have advanced an argument for the development of territorial boundaries in the Pleistocene and early Holocene, and the concomitant establishment of core and peripheral occupation sites. They use the appearance of Mopir obsidian in Pleistocene deposits at Matenbek (New Ireland) and its absence at a similar time period in the more northern caves of Balof and Panakiwuk as evidence of growing territoriality. The implication here is that this trend toward defined but increasing territoriality saw its ultimate manifestation in the areally extensive distribution systems of the upper Holocene. This developmental argument for the growth of the scope of obsidian distribution has been raised by other commentators (Gosden and Specht 1991, White 1996). It must be said that arguments for indigenous economic change pose little problem for an intrusion scenario. One strong advocate of the cultural intrusion hypothesis has proposed that the terminal Pleistocene of Island Melanesia may have witnessed the first steps toward agriculture (Spriggs 1993), while another has advanced a model which incorporates bom intrusion and local innovation for later Holocene developments (Green 1991).
The overriding question in this debate is what prompted this upper Holocene expansion in the areal extent of “interaction spheres” (Allen and Gosden 1996) at around 3500 B.P. The intrusion scenario is based on the arrival of a foreign maritime-adapted people who possessed a desire or need to maintain contact between their scattered coloniser communities. This initial impetus for communication between islands ultimately became incorporated into the social fabric of island societies. This provides an explanation for the continuation of long-distance maritime trade long after the Lapita cultural complex apparently disappeared as a distinctive entity in western Melanesia. Exponents of an indigenist scenario see the impetus for long-distance maritime distribution as having developed independently within Island Melanesia. This seems to be based largely on an assumption that long-distance interisland contact would be the natural culmination of a lengthy process of technological and economic evolution within western - 388 Island Melanesia. However, the causal factors behind this development have yet to be fully argued, and it is difficult to discern from the indigenist argument what forces might have been responsible for this change.
The long sequence of obsidian dispersal in Island Melanesia can be accommodated by a hypothesis for independent evolution out of coastal and localised interisland trade/exchange. However, this explanation has yet to address the question of why a cultural impetus for long-distance maritime contact and colonisation (for Remote Oceania) appeared at around 3500 B.P. The Pleistocene colonisation of isolated Manus and Buka and the occurrence of obsidian on Nissan in mid-Holocene contexts demonstrate that long-distance voyaging was being undertaken well before Lapita times. However, the advent of Lapita did not only mark an expansion of distribution. There were also new modes of resource articulation, as exemplified by a focus on the transport of almost exclusively Lou obsidian beyond the Admiralties, despite the use of both Lou and Pam sources within the island group. It remains to be seen how factors such as these can be accommodated by a model of gradualistic development.
The evidence reviewed in this paper highlights how the Lapita phenomenon represented a point of momentous change in prehistoric obsidian distribution. A great deal more archaeological research is required before the mechanisms which brought about this transformation can be isolated and understood. However, on current evidence the most robust explanation resides in the appearance of outside elements, elements which triggered a rapid and major expansion of interisland contact. External influence therefore seems to present a strong candidate for the engine that provided the initial impetus for the long-distance maritime distribution of Bismarck obsidian throughout Island Melanesia after 3500 B.P.
This is a much revised version of a paper presented in a symposium on Pacific colonisation at the 1996 New Zealand Archaeological Association Conference. I thank Richard Walter for inviting me to take part. Thanks also to Wal Ambrose, Ian Walters and an anonymous referee for useful suggestions on various drafts. Responsibility for any errors or omissions lies solely with the author.
Table 1: Number of pieces of Bismarck Archipelago obsidian found in sites where importation was by sea. Shaded cells show periods when sites were unoccupied. (Data from sources cited in text.) The Arawes are included as watercraft were needed to convey obsidian to this group. However, only a short water crossing was involved; during mid-Holocene occupation, Lolmo cave was never more than approximately 3 km from the coast of New Britain.
- 394 Page is blank