Arboriculture and agriculture in coastal Papua New Guinea.

Gosden, Chris

A central issue in the regional prehistory over the Transition - and therefore of this whole set of papers - is the different life-ways that came to be followed in Papua New Guinea and in Australia itself; the one became agricultural, the other hunter-gatherer. There is more to the story than that divide; this is a story of a human and created world, rather than a simple response to directing environment.

In discussing the effects of the marine transgression that occurred between 15,000 and 7000 b.p. I will commit a transgression of my own. I will break the rule which divides Australia from Papua New Guinea on the lines of hunter - gatherers versus farmers. That people farm in Papua New Guinea is clear, as is the fact that they have done so for a long time. What I would question, however, is the notion that farming is the key to life in Papua New Guinea past and present and that this form of life can be divided off neatly from Australian hunting and gathering. In a classic article Peter White (1971), asking `Why was there never a "Neolithic Revolution" in Australia?', gave as his main answer that these were foragers far too affluent to countenance such excessive toil.

This characterization of the two continents has been very fruitful in providing the framework, explicit or implicit, for much research. Now, however, the time has come to rethink this framework. The reason is a growing realization amongst those working on the archaeology of northern Papua New Guinea and adjacent islands in the Bismarck Archipelago that the evidence we are unearthing will not fit well into the conventional categories of hunter - gatherer and farmer. This has led to some creative playing with these categories (Spriggs 1993), but perhaps we need to find some new bottles for the new wine. I start by discussing a different approach to problems of subsistence and its social effects, before examining how far this allows us to appreciate novel aspects of new data.

Creating a world: the deployment of materials

Arguments starting from the nature of subsistence have a particular structure to them. The requirements of the subsistence round are said to determine the nature and pattern of movement or sedentism, while the products derived from subsistence influence the levels of population possible and the amounts of trade or exchange, craft specialization etc., which are possible. Subsistence is the base on which all other elements of social life must be built.

To overcome the determinisms, environmental and economic, built into this view, let us take a different starting-point which does not necessarily privilege subsistence. People in all times and places attempt to create a world for themselves to live in, using the materials to hand and the social and technical skills at their disposal. Food and drink are a vital consideration when creating a world, but they are a necessary, rather than a sufficient condition in this shaping. The world in which people live is structured to some extent on the nature of the physical world, but has extra social dimensions. As European colonization of many parts of the world, including Australia, has shown us, different groups of people can inhabit the same areas but make totally different worlds out of them.

A humanly created world has the same dimensions as the physical world which it occupies. It has extension in space and depth in time. Extension in space depends upon the regular movements that people make in carrying out tasks and maintaining social connections. Movement and connection always have a material dimension, depending on mode of transport and the use of materials, including food to bind and divide groups. Depth in time derives from a common tradition, a basic fund of taken-for-granted and conscious knowledge, transmitted from one generation to the next which people can use and change to meet present circumstances.

A stress on creation should not blind us to the fact that people are constrained by the world they live in. Human groups may deploy materials to meet their own ends, but each set of technologies, plants and animals has its own optimum framework of space and time, so that in order best to provision a social system, certain rhythms and constraints must be observed in planting and harvesting, fishing, gathering and the production of material culture. These exigencies become etched into people's being, so that as they shape the world, the world shapes them (Gosden 1994).

Looked at from this point of view, Australia and Papua New Guinea in the present and past represent a whole series of worlds created through the combination of forms of subsistence and the uses of material culture. Because it is the combination of material things and food which is vital in creating a world, a division of social types into hunter - gatherer and farmer, may not be helpful, since it starts from the premise that subsistence is the vital foundation. In Papua New Guinea we are used to thinking of particular combinations of material culture and food as typical for the country as a whole and as the key to its prehistory. The growth of the wealth economies of the Highlands are probably the best-known historical trajectory (Golson & Gardner 1990). In the western Highlands in the present there are large densities of people engaged in extremely intensive agriculture and forms of exchange, where the production of food and the social structuring of exchange are linked through the medium of the pig. The way in which this world has been created is relatively well known concerning the Holocene period and rather less so for the Pleistocene (Haberle 1993). The pre-history of the Highlands has hardened the Australia/Papua New Guinea distinction, and the rest of Papua New Guinea has come to be seen as a paler reflection of the Highlands' mix of agricultural intensification, high population levels, and social competition through trade and ritual. What I want to do in the following account is to consider some possibilities that do not follow the Highlands' trajectories and bring out the peculiarities of the north coast and islands data more fully.

The world of the western Highlands is spatially constrained, although socially dense and extremely hectic in that people are busily involved in cycles of production and exchange. The other system allowed latitude in both senses of the word: it was an ocean world with considerable spatial extent, which was also not too constrained in terms of time (see also Allen 1993: 146 on a similar point). Before getting on to the archaeological data, some environmental background is needed to structure the discussion.

The world to be created: environmental

change

The area I consider in this paper is the north coast of Papua New Guinea from the Irian Jaya border eastwards, together with the islands of the Bismarck Archipelago and the Solomons (Figure 1), as these are the only areas of coastal Melanesia known to have human occupation prior to 7000 b.p. The main changes in this region between 15,000 and 7000 b.p. were brought about by the rise in sea-level, changes varying enormously from one area to the next. In some areas, such as the east coast of New Ireland, very little new land was added even at the lowest sea-levels (Allen et al. 1989), and little was flooded again when the sea rose. This may well have been true for much of the Bismarck Archipelago. At the other end of the scale is the case of the Sepik Basin. A drop of c. 130 m by 18,000 b.p., the best present estimate for minimum sea-level (Chappell 1993: 45), would have caused severe down-cutting of the ancestral Sepik River upstream as far as the May River (Swadling & Hope 1992). Once sea-levels began to rise again, a large area of what is today the Sepik and Ramu Basin was flooded, resulting in a huge area of enclosed sea and this has been progressively filled in over the Holocene period (Swadling & Hope 1992: 29-37). Elsewhere there have been movements of land as well as sea which complicate the picture, but only in the case of the Huon Peninsula are these relative movements understood (Chappell 1974).

Equally complex are the adjustments of biological communities attendant upon sea-level change. Growth rates of coral reef appear able to keep up with even the most rapid rise in sea-level (Chappell 1993: 46). Streams develop estuaries when the sea rises, often with mangroves and freshwater back-swamps, creating a rich and varied range of habitats (e.g. Tacon & Brockwell and Morwood & Hobbs, this volume). A falling sea causes the coastline to dry out and corals to contract to a narrow fringe on the margins of former reefs. In general, habitats are richer and more varied when the sea is rising than when it is falling, so that the periods before and after the Last Glacial Maximum may present something of a contrast (Chappell 1993: 46). Swadling (1994) has recently highlighted the potential that the analysis of shell species in archaeological sites may have for identifying different shoreline conditions. In particular, she picks out long-term sequences spanning the whole period of human occupation, from Matenkupkum and Buang Merabak on New Ireland, to look at the shift from hard to soft shorelines (Swadling 1994: figure 2). This form of analysis can obviously be extended.

Changes in temperature and rainfall through the Last Glacial Maximum are hard to summarize with any confidence on the basis of present evidence. Temperature may have been depressed at the height of the Last Glacial Maximum, although probably only by a small amount through the tropics. Estimates for sea-surface temperatures on the basis of microfauna in deep-sea cores predict temperatures only 0 - 2 [degrees] C lower than present (CLIMAP 1976; Moore et al. 1980). However, inconsistencies seem to exist between the marine and terrestrial records, although the latter are thin to say the least. Where they do exist, terrestrial records seem to indicate a greater temperature drop than is evidenced in marine microfaunas (Southern 1986). Rainfall may have declined to a greater extent than temperature, although again little direct data exists. Through general modelling (Rind & Peteet 1985) and extrapolation with the situation in northern Australia (Hiscock & Kershaw 1992: 47-9), there may have been a reduction of 25% to 50% in rainfall as the Glacial Maximum approached, although these seem to be maximum estimates.

Whether the reduction in rainfall affected the structure of lowland tropical rainforest is uncertain, but there is no present evidence that it did. The situation in the Highlands indicates that rainforest was found below 2000 m, which would encompass the whole of the lowlands and islands (Hope & Golson, this volume). In summary, we are dealing with a complex and variable situation throughout the region, for which little direct environmental evidence presently exists. Odd though it may seem, I shall only make passing mention of the nature of environmental change when discussing the archaeology. This is not because environmental variations would have been insignificant; it is merely that their effects are unknown.

One further problem is that of dating. Calibrations of radiocarbon dates, the most important source of chronology for this region, only go back to 18,000 radiocarbon years, which now calibrates to c. 22,000 calendar years (Stuiver & Reimer 1993), leading one to wonder when was the world at 18,000 b.p.? Furthermore, the Pleistocene calibrations still seem to be shaky at best, so that calibrated dates are not used here; this imperfect solution is at present the least confusing.

The world as created

One certain statement concerning the coastal environments before and after the Glacial Maximum is that they varied. Variability is also increasingly the theme running through the archaeological evidence even from the earliest periods.

In setting out the general features of the how people make a world for themselves, I said that traditions were important, as traditions represent a build-up of skills and knowledge which can be built upon and altered by each succeeding generation. To understand the world as it existed on the northern coastal reaches of Papua New Guinea at 15,000 b.p., we must look briefly at the earliest history of life in the region to get some sense of the traditions in operation.

Evidence from before the Last Glacial

Maximum (1)

Table 1 shows all the sites with evidence from the period between the earliest occupation of the area and 7000 b.p. Five sites are known to have been occupied before 30,000 b.p. - Fortification Point on the Huon Peninsula (Groube et al. 1986), Lachitu in the Vanimo region (Gorecki et al. 1991), Yombon in the interior of New Britain (Pavlides & Gosden 1994), and Matenkupkum (Gosden & Robertson 1991) and Buang Merabak (Balean 1989), both on New Ireland. Strikingly these sites are spread over a large area and represent a number of site types in different habitats. Both Fortification Point and Yombon are open sites; the former is coastal, the latter at least a day's walk inland. The other three sites are coastal caves, Lachitu on the mainland of Papua New Guinea, and Matenkupkum and Buang Merabak on the island of New Ireland. At this period people were able to enter a number of habitats, coastal and inland, to make use of specific resources, such as the fine-grained cherts of Yombon, and to cross water gaps. One site, the Huon Peninsula, has specialized tools in the form of waisted blades. Groube (1989) has proposed that these might have been used for forest manipulation or clearance, even at this early date, an idea that gains some corroboration from pollen analyses from the Highlands (Haberle 1993). Unfortunately, there are no pollen diagrams from the area being discussed here for the relevant time-periods. in terms of lithics, none of the other sites has anything other than amorphous tool forms at this early period.

Shortly after 30,000 b.p., further evidence of people's vovaging capacities comes with the first appearance of the site of Kilu on Buka Island at the northern end of the Solomons Island chain. This necessitated a journey across the horizon, which might have meant crossing a psychological barrier, if not a technical one (Wickler & Spriggs 1988). The earliest colonists of Papua New Guinea's north coast and islands do not seem to have been constrained by barriers of land or sea and moved rapidly through the area. There is also a hint in the waisted blades from the Huon Peninsula that they may have been manipulating their environment, but more evidence needs to be sought. However, there were limits on the creation of a world; there is no real evidence of people using other than local stone tools or moving resources from one place to another.

Evidence from the Last Glacial Maximum

Two New Ireland sites, Matenkupkum and Matenbek, despite being only 70 m apart, have quite different histories. The sparse faunal evidence from the lowest levels at Matenkupkum shows that the earliest inhabitants foraged for food through a number of local microenvironments, but were probably confronted with a small range of animals. Compared to Matenkupkum, two differences in the 18-20,000-year-old layers of Matenbek are quite startling. The first is that from the very bottom of the site, and continuing throughout the Pleistocene faunal assemblage, the dominant terrestrial animal present using both NISP and MNI counts is the phalanger P. orientalis (Allen in press). This animal, non-endemic to New Ireland, is considered to have been transported there by humans. Subsequent animal movements are dealt with below, but the Matenbek phalanger bones are currently the earliest evidence for the occurrence of this species in the Bismarcks.

The second difference, again continuing from the bottom of Matenbek through the Pleistocene layers, is the small but persistent presence of obsidian. In all, 37 pieces of obsidian were recovered from the Pleistocene layers of the test pit, of which 17 have been sourced using the PIXE-PIGME technique (Summerhayes & Allen 1993). Slightly more than 70% derive from the New Britain source of Mopir, the remainder from sources in the near-by Talasea area of the same island. The assemblage consists of very small flakes and tiny bipolar cores, both of which rarely exceed 2 cm in maximum dimension and 1 g in weight. It is reasonable to see this small size as a result of intensive curation. In turn, this probably reflects the distance the site is from the obsidian sources - 350 km in a straight line, including a 30-km water crossing.

The contrasts between the situations at 35,000 and at 20,000 b.p. are instructive. Whereas in the earlier period we are dealing with groups able to explore rapidly and take advantage of local resources, there is no evidence of the movement of materials from one place to another. From 20,000 b.p. people are moving animals and raw materials in order to overcome the constraints of local resources. Society is integrated over a larger area to local benefit and, in the terms of this paper, a broader world has been created through which flowed materials, people and ideas. Their ability to sail was the technical underpinning of this world, but it was equally obviously bound together by social mechanisms which allowed links between different areas to be set up and maintained. There is also evidence of a tradition of manipulating the world. The extent of this can be brought out by considering a single piece of admittedly tentative evidence. Forty-seven stone tools from Kilu cave on Buka Island were examined for use-wear and residues; of these 14 from Pleistocene layers were found to have raphides, tissues and starch grains of both Colocasia and Alocasia taro. This helps confirm both that the natural range of taros included Papua New Guinea and the Solomons (see Yen, this volume) and that these were at least a component of the diet from 28,000 b.p. (Loy et al. 1992). The evidence also implies that people knew how to process taro in order to make it edible, which would have involved various combinations of cutting, scraping, soaking and cooking so as to remove acrid or toxic compounds. The implications concerning a tradition of knowledge here are considerable.

Few sites in the area have dates which fall within the Last Glacial Maximum. Matenbek is the site that stands out, as it was first used as sea-levels were falling and greater aridity setting in. All other sites which fall within the period of interest here were either first used about 35,000 years ago and have a partial or complete break between 20,000 and c. 15,000 b.p., or were only used for the first time around this latter date (Table 1). There are various answers to the question why there is so little evidence from the Glacial Maximum period. The most obvious is that the majority of the sites are near the sea at the level it would have been before and after the Glacial Maximum. The sites we have are sampling activities taking place on the immediate coastal strip. With the sea c. 130 m lower these activities would have shifted downwards; the caves with early occupation, such as Lachitu, Matenkupkum, Buang Merabak and Kilu, now in different relationships to the coast, were either abandoned or used differently. The excavations at Matenkupkum demonstrate also problems of sampling through excavation: different parts of the cave were used at different periods, as its uses changed. This may be true of other cave sites, with material dating to the Glacial Maximum generally being further back from the entrances. Changes in the coastal relationships of caves during the Glacial Maximum may also have altered the uses they were put to by humans (see discussion in Gosden & Robertson 1991) - a possibility to be tested through further excavation. Yombon is the only site in the region from this period not within easy striking distance of the coast. Here again no dates fall within the Glacial Maximum, despite the site having been fairly extensively sampled by the standards of this region (Pavlides & Gosden 1994). While the Pleistocene layers at Yombon are generally ephemeral, the extent of this large open site allows no guarantees that all the occupation periods represented there have yet been discovered through excavation.

site name date range

Huon Peninsula 40,000
Matenkupkum 35,000-10,000
Yombon 35,000;14,000;
 Late Holocene
Lachitu 35,000; 13,900-12,300;
 9000-5600; Late Holocene

Buang Merabak 31,000-Holocene
Kilu 28,000-Holocene
Matenbek 20,000-Holocene
Panakiwuk 15,000-8000; late Holocene
Balof 14,000-late Holocene
Pamwak < 12,000-late Holocene
Misisil 11,000; Late Holocene

Table 1. Palaeolithic and early Holocene sites in
coastal Papua New Guinea and the Solomon
Islands.

In discussing forms of life within the Glacial Maximum period, I will return again to the notion of a created world. Creation of a world takes place at a number of different spatial scales, which interlock and influence each other. The smallest scale to which we have access is constituted by the internal arrangement of individual sites. Discovering sites and dating their sequences have been the main priorities in the region so far, which have resulted in relatively small excavations in most sites. Only at Matenkupkum is there a large enough area of excavation to make comments on the internal arrangement of the site. Here a trench 10 m long and 1 m wide was excavated from the mouth to the middle of the cave (Gosden & Robertson 1991: figure 2). The distribution of material, such as shell and stone, along the length of the trench and its relationship to features such as hearths change dramatically before and after the Glacial Maximum. In the earlier period there is little hint of a structure to the distribution of material along the trench, although more shell may have been concentrated in the middle of the cave and stone at the front. Build-up was also slow, with 50-60 cm of deposit laid down over 10,000 years. Between 14,000 and 10,000 b.p. deposition is more rapid, with almost 1 m of sediment and artefacts being laid down in this relatively short period. There is also evidence for structure in the deposit, with hearths in the middle of the trench and dumping of larger-sized materials behind the hearths, whilst the front of the site was kept relatively clear. This indicates a structure to activities within the cave, and some site maintenance and cleaning (Gosden & Robertson 1991: 43-4).

As well as changes in the internal use of Matenkupkum, there are indications that different sets of external connections existed than before the Glacial Maximum. Large numbers of Phalanger orientalis bones are present in the cave together with obsidian from New Britain. The 103 samples analysed showed that the Mopir sources supplied 64%, with Talasea sources contributing 30%, the remaining 6% coming from unknown sources (Summerhayes & Allen 1993: 147). We have evidence of a complex set of connections here involving multiple sources and unknown mechanisms of transport. As we have seen from Matenbek, the first evidence for the movement of obsidian starts around 20,000 years ago. Taken together, the two sites represent a considerable continuity in the movement of obsidian into New Ireland over 10,000 years.

Some 200 km to the northwest two more sites, Balof 2 and Panakiwuk, are both occupied by humans for the first time at c. 15,000 b.p. Between this time and the nominal beginning of the Holocene at c. 10,000 b.p., the human use of both sites is sporadic in the extreme, with small amounts of cultural refuse involved (Allen et al. 1989; Marshall & Allen 1991). Two aspects of these two sites are significant. Firstly, both are inland. Balof 2 is currently 2.7 km from the coast and Panakiwuk is 4 km inland. With Pleistocene sea-levels lower, the sea would have been more distant - although given the steep submarine contours of this coast, not much more distant. While marine fish and shellfish occur throughout the Pleistocene levels at Balof 2, they are absent from the corresponding

(1) The data reviewed here hare also been synthesized in a longer paper elsewhere (Allen & Gosden in press), but they have been put to different uses in each paper. layers at the more `inland' site. Secondly, both obsidian and phalangers are absent from these layers in these sites, with the exception of a single phalanger bone at Panakiwuk associated with a date of c. 13,000 b.p. It is worth noting that Kilu, on Buka may have occupied a different interaction sphere: no Bismarck's obsidian reached the Solomons until the late Holocene and Kilu has produced one Phalangerbone at 9000 b.p., with no more until much later (Wickler 1990). A complicating factor with this argument is that Yombon, itself on the obsidian-producing island of New Britain, has no evidence of obsidian until the middle Holocene (Pavlides 1993). At the same time, however, near-by Misisil cave has five obsidian flakes dated to just over 11,000 b.p., (Specht et al. 1981). Talk of differing interaction spheres may be premature until larger samples have been obtained from key areas and sites.

A more convincing case for an area forming a world apart is Manus, although it certainly also participated in larger trends. Discussion of the early Manus data are complicated because it is uncertain when Manus was first settled. As pointed out above, the currently earliest radiocarbon determination from Pamwak is c. 12,000 b.p., but this is only from part-way down the sequence, which may extend to 20,000 b.p. or earlier (Fredericksen et al, 1993: 149). The stone and shell tools are the key indicators of change here. The lowest deposit, layer 7, has mainly unmodified chert flakes, using local raw material. A similar pattern is found in layers 6 and 5 (which produced the 12,000 b.p. date), but these also produce the earliest trickle of obsidian into the site. Whether this introduction reflects new connections within the region or whether this is the period when the obsidian sources on Lou Island were first created is uncertain. Important additional first appearances at this time are the bandicoot (Echymipera kulubu) and the Pacific almond, Canarium. Both may have been human introductions from mainland New Guinea, parallelling the New Ireland introductions, although differential organic preservation in the lower levels of the site cannot be discounted. Further important changes occur in layer 4, dated by six determinations to c. 11,000 b.p. Here obsidian occurs in large amounts for the first time (Fredericksen et al. 1993: 149). It is accompanied by a range of formal tools made of stone and shell. Retouched pitchstone was made into a number of discoidal implements, some knapped from prepared cores. These were accompanied by two edge-ground shell adzes and three ground-stone axes or adzes, the earliest shell implements or ground stone reported from island New Guinea. All these types continue to be found in the upper layers of the site, layer 2 falling at the end of the period of interest here as it dates between 7000 and 5500 b.p. This layer with large amounts of mangrove shells, dates from the end of the marine transgression which brought the shoreline and mangrove communities within 1 km of the site, compared to the 10 km distance during the Glacial Maximum.

As the excavators note (Fredericksen et al. 1993: 150), not only is Pamwak unusual for its formal tools, it is also distinguished from many other Pleistocene sites in the region by the amount of stone in its deposits. Matenkupkum has by far the highest densities of the sites in New Ireland with 5600 g/[m.sup.3] stone in the deposit, but this is outweighed by the 8500 g/[m.sup.3], at Pamwak. The exact significance of this is hard to judge; one implication is that these two sites were occupied more intensively, or for longer, than sites such as Balof and Panakiwuk. The precise significance of the ground shell and stone tools is also hard to gauge, although the excavators have followed the same line of thought as Groube (1989) in seeing that they might be used for forest clearance, which might, in turn, be due to agriculture (Fredericksen et al. 1993: 150). The Pamwak data have occasioned surprise and add materially to evidence of variability within Pleistocene styles of life in the region. They have also demonstrated that shell implements are not a purely Holocene development within Oceania, but have Palaeolithic roots.

Although variability is certainly part of the story for the Late Pleistocene of this part of the world, there are also threads of similarity linking the data. Contemporary with the first reliable dates from Pamwak, changes are evident in sites elsewhere. Matenkupkum has no Holocene evidence from the front of the cave, although the back area was used in the early Holocene. In near-by Matenbek, however, deposits dating to between 9000 and 6000 b.p. reflect significantly higher rates of deposition than in the underlying Pleistocene layers. Among interesting changes in this site at this time, the obsidian assemblage maintains the same heavily reduced characteristics and derives from the same New Britain sources in similar proportions to the Pleistocene materials; however, it occurs in significantly greater numbers, and by this time had almost totally supplanted the use of local lithic materials in this site. In addition, the small test pit also uncovered an earth oven, more shell tools including a shell bead, and two fragments of edge-ground tools, probably axes.

Panakiwuk was most intensively occupied between c. 10,000 and 8000 b.p. (Marshall & Allen 1991: 89). More than 70% of faunal remains and stone tools in the site were deposited at this time. A similar pattern is reflected in Balof 2 (White et al. 1991: 56, tables 5-8). In Panakiwuk, in the middle of this period, marine shells appeared in significant numbers in the site for the first time, indicating an important change in its use which aligned it much more directly with coastal activities. In both sites phalanger bones were discarded regularly for the first time, although in far greater numbers in Balof 2 than in Panakiwuk. A single flake of Talasea obsidian was recovered from Panakiwuk in the 8000 b.p. levels, and a single flake from Balof 2 at a similar period (White et al. 1991: tables 2 & 11). Further animal movements were also recorded at this time, with the large rat, Rattus praetor, occurring at Panakiwuk and the thylogale, Thylogale brunii, appearing at Balof 2. A piece of Canarium sp. endocarp was also recovered from Panakiwuk (Marshall & Allen 1991: 88), while in Balof the early and late Holocene levels from c. 8000 b.p. onwards contained 20 elasmobranch vertebrae and 13 shark teeth (White et al. 1991: 54). At Panakiwuk at c. 8000 b.p., human occupation ceased for some 6000 years. It is less clear whether Balof 2 was also abandoned at this time there is a gap of c. 4500 years between the dates for Horizons Il and III. The excavators suggest `minor breaks before and after c. 3000 b.p.' (White et al. 1991: 48).

On New Britain, Yombon has produced evidence from the late Pleistocene period, after 14,000 b.p., but only a small number of tools were found, which may indicate the sporadic use of the landscape. More intensive use of the Yombon area is first indicated at the very end of the period discussed here, c. 7000 b.p., and this continues through into the late Holocene (Pavlides 1993). Similarly, fleeting use of the neighbouring site of Misisil occurs c. 11,000 b.p. with more definite signs of occupation from the late Holocene.

Discussion

The Pleistocene data derived from this area over the last 10 years have caused surprises. The caves of New Ireland excavated as part of the Lapita Homeland Project (Allen & Gosden 1991) and the Kilu site, dug slightly later, showed the antiquity of island occupation and the fact that people could cross the open sea to the Solomons. This evidence seemed to have a pattern to it; some parts of the pattern have remained intact and some have altered with more recent discoveries. In discussing the evidence we should also be aware of its archaeological provenance. All sites except two, Fortification Point and Yombon, are caves or shelters which are almost certainly providing only partial evidence of what people were doing in the landscape as a whole. By contrast, much of the evidence from the mid Holocene onwards is from open sites formed after the sea reached its present level. Some of the discontinuities and changes in the nature of the evidence, brought out below, would be altered in scale and significance if we had a full range of site types from each period.

The first Pleistocene New Ireland evidence appeared to show coastal locations and a reliance on marine resources; amorphous stone tool industries throughout the Pleistocene and Holocene periods; good mastery of maritime technology, and the movement of animals and raw materials after 20,000 b.p. Pamwak and Yombon, both discovered in the 1990s, throw out the notion of a purely coastal location as advanced by Gosden & Robertson (1991: 43). We can now say with some certainty that people were able to move all over both New Britain and New Ireland by 35,000 b.p. and presumably the Solomons shortly after that, so that there are no apparent breaks to movement by land or sea. The date of the settlement of Manus is vital to resolve exactly what maritime capabilities people had at this early date.

Pamwak and Yombon also cast serious doubt on the idea that amorphous tools existed throughout the Pleistocene in the islands: Pamwak has shell and stone tools of definite form by at least the end of the Pleistocene, while Yombon has well-known bifacially flaked tools from the Holocene. It remains to be seen how far back in time these types will extend when further samples are available. These finds complement the waisted blades from the Huon terraces: while (Groube (1989) and Fredericksen et al. (1993) have linked formal tools with forest clearance, one can think in terms of wood-working and the production of watercraft, such as canoes. The existence of formal tools elsewhere raises questions about the New Ireland assemblages: is New Ireland as a whole different in its stone tool technologies, or does it just happen that the sites excavated so far have no formal tools? In either case variability seems to reign and, as Torrence (1992) has pointed out, the nature of stone tool technologies may be linked to patterns of land use and subsistence varying throughout the region.

The part of that early-perceived pattern which has stayed the same is the notion that these groups were able to use and manipulate the environment for their own ends. The sea has been a source of food throughout, and the shell technologies from Matenkupkum and Matenbek may indicate more sophisticated forms of fishing than we have hitherto guessed at. The demonstration by Loy et al. (1992) that different sorts of taro may have been utilized 28,000 years ago at Kilu shows that the vegetable crops that constitute a major part of the present diet may have been utilized in some form back in the Pleistocene, together with types of taro preparation. As we have seen, Groube (1989) has put forward the possibility that forest management may date back to the earliest occupation of New Guinea. Definite evidence for the use of tree products comes from the end of the Pleistocene in the form of preserved nuts and seeds, the earliest of which is Gorecki's find of Canarium at c. 14,000 b.p. in a site in the Sepik-Ramu area (Yen 1990: 262). Shortly after this date Canarium is found at Pamwak, following a possible introduction from the mainland. By 9000 b.p. there is a small amount of evidence from a wide range of sites including Panakiwuk and Kilu. We may conclude that this important tree was being harvested for food throughout the northern coasts and islands.

This supposition is reinforced by evidence from after the period of interest here. The Dongan site, near Bosmun village in the Sepik region, has produced a large range of plant remains preserved by waterlogged conditions. These include Cocos, Canarium, Pandanus, Pometia, Aleurites, Areca Catechu and Pommetia pinnata (Swadling et al. 1991: table 2) dating to 5830 [+ or -] 90 b.p. (Beta-19077). As Swadling et al. (1991:103) note, this evidence extends the time range for tree-cropping in Melanesia considerably beyond the Lapita period from which the earliest direct evidence had previously stemmed (Kirch 1989). Although the Dongan finds are the earliest evidence of a wide range of tree crops together, there is no sign that this was the incipient stage of tree use. As Spriggs (1993: 138-9) has stressed, tree crops are important because nuts occur in relatively large numbers and can be stored for several years even in the humid tropics. The trees themselves can produce nuts and fruits for decades with the minimum of tending, which makes them a long-lasting and dependable source of food. Taken together, all these sources of food, many of which were introduced or altered by people, provide the basis for mobile and flexible ways of life.

It may now be obvious why the evidence surveyed here will not fit into the categories of hunter-gatherer or farmer. On the one hand there is a growing tradition from the Glacial Maximum onwards of the movement of animals and probably also tree crops, such as Canarium. On the other hand, there is no evidence for sedentary communities or large-scale farming until the later Holocene. Nor, as emphasized at the start of this article, should subsistence alone be considered the basis to life. Long-lasting and dependable crops, such as those provided by trees, could have been combined with casual root-cropping, hunting and fishing as part of an overall strategy of life. These were resources that had been developed to be deployed in particular ways. These were combined with forms of social connectivity, glimpsed through the movement of obsidian, which facilitated the meeting of groups and perhaps also growing forms of boundary maintenance. Some axes of connectivity can be discerned. New Britain and southern New Ireland show links through the small, but regular, movements of obsidian. The lack of obsidian at Yombon, in the interior of New Britain, raises the possibility that sea links were more important than land, and that the effective distance between interior New Britain and the north coast of the island was greater than that to New Ireland. Manus, the north coast of New Guinea and the north Solomons all seem to have created their own areas of connection, but until more sites have been excavated and more markers of connectivity found, it is difficult to say what these were.

Whatever one thinks of the above scenario, which is admittedly as provisional as it could be, a more definite statement can be made. There is no evidence that the environmental changes that occurred during the Glacial Maximum were the major causative factor in shaping people's lives. The major events or changes happen before or within the Glacial Maximum, as by this time people were using root crops and transporting animals and raw materials as they continued to do into the Holocene. More sites are known at 15,000 b.p. than before, but this is probably more to do with shifting activities associated with changing sea-levels than with a re-ordering of activities. Certainly, a Pleistocene/Holocene boundary at 10,000 b.p. means nothing in this region: it represents neither an observable environmental shift nor change in patterns of life. The really major human changes come about at the end of the period looked at here, between c. 8000 and c. 6000 b.p., when most of the New Ireland caves and Lachitu are abandoned, but much greater deposition starts at Yombon and Pamwak. Also new sites start up on small islands (Gosden et al. 1994), probably as a partial consequence of new stabilizing sea-levels and coastal ecosystems.

The papers in this volume bring out well the various trajectories of different regions and the variations in the responses human groups made to the Last Glacial Maximum. Some areas, notably northwest Australia and Tasmania, indicate that climate and attendant environmental changes were the major factors in influencing the distribution of human settlement through the period. Cutting across the general variability exhibited by ways of life in Sahul is still the dichotomy between the area that became Papua New Guinea in the early Holocene and the separate region of Australia. What appears to stand out in the Papua New Guinean evidence, from both the Highlands and islands, is the human manipulation of the environment over possibly 30,000 years. Little evidence for tree clearance, the movement of plants and animals and so on, has been adduced for Australia.

There are several possible reasons for this apparent dichotomy.

Perhaps the difference was due to the greater latitude allowed to human action by tropical environments - groups were less likely to be pulled and pushed by changes in rainfall and temperature and could do some manoeuvring of their own. However this seems truer of comparisons with the extremely arid coasts of northwest Australia, and less true in the humid tropics of Australia.

A second possible reason might be that people developed different sets of traditions in what became Papua New Guinea, even though for most of human history Sahul was a single land-mass. Yen (this volume) suggests that these traditions were to do with a greater ritual attachment to land in Australia, rather than to plants and cycles of planting and harvesting.

There is a third possibility, that different research traditions at work in the `agricultural' north and `hunter-gatherer' south, lead us to create and see the evidence in a different light.

It may be true that the two areas were always different, reflecting long-term traditions of combining land, plants, animals and material culture with ideological and ritual structures. The evidence from Sahul has so far been used to magnify the divide between hunter - gatherer and farmer, using ethnography as the measure of the past. While it may be impossible to make judgements about Sahul as a whole, the evidence from the late Pleistocene and early Holocene in northern coastal Papua New Guinea has no counterpart in the present. The Pleistocene groups of this region should not be seen solely as people on their way to agriculture in tandem with their cousins in the Highlands. Rather they can be viewed as developing a form of life sufficient unto itself which defies easy categorization. Only by rejecting pre-existing categories can we really appreciate the richness, variety and surprise contained in the late Pleistocene data from northern coastal Papua New Guinea.

References

Allen, J. 1993. Notions of the Pleistocene in Greater Australia, Spriggs et al. (ed.): 139-51. In press. The pre-Austronesian settlement of island Melanesia: implications for Lapita archaeology, Transactions of the American Philosophical Society. Allen ,J. &, C. Gosden (ed.) 1991. Report of the Lapita Homeland Project. Canberra: Department of Prehistory, RSPacS, Australian National University. Occasional Papers in Prehistory 20. In press. Spheres of interaction and integration: modelling the culture history of the Bismark Archipelago, in J. Davidson, G. Irwin & A. Pawley (ed.), Pacific culture history: essays in honour of Roger Green. Allen, J., C. & J.P. White. 1989. Human Pleistocene adaption in the tropical island Pacific: recent evidence from New Ireland, a greater Australian outlier, Antiquity 63: 548-61. Balean, C.E. 1989. Caves as refuge sites: an analysis of shell material from Buang Merabak, New Ireland. Unpublished BA-(Hons.) thesis. Department of Prehistory and Anthropology, Australian National University, Canberra. Chappell, J. 1974. Geology of coral terraces, Huon Peninsula, New Guinea: a study of Quaternary tectonic movements and sea-level changes, Bulletin of the Geological Society of America 85: 553-70. 1993. Late Pleistocene coasts and human migrations in the Austral region, in Spriggs et al. (ed.): 43-8. Climap. 1976. The surface of the ice-age earth, Science 191: 1131-7. Dodson, J. (ed.). 1992. The naive lands: prehistory and environmental change in Australia and the southwest Pacific. Melbourne: Longman Cheshire. Fredericksen, C., M. Spriggs & W. Ambrose. 1993. Pamwak rockshelter: a Pleistocene site on Manus Island, Papua New Guinea. in Smith et al. (ed.): 144-52. Golson, J. & D.S. Gardener. 1900. Agriculture and sociopolitical organization in the New Guinea Highlands, Annual Review of anthropology 19:395-417. Gorecki, P., M. Mabin & J. Campbell. 1991. Archaeology and geomorphology of the Vanimo coast, Papua New Guinea: preliminary results, Archaeology in Oceania 26:119-22. Gosden C. 1994. Social being and time. Oxford: Blackwell. Gosden, C & N. Robertson. 1991. Models for Matenkupkum: interpreting a late Pleistocene site from southern New Ireland, Papua New Guinea, in Allen a Gosden (ed.): 20-45/ Gosden, C., J. Webb, B. Marshall & G.R. Summerhayes. 1994. Lolmo cave: a mid to late Holocene site, the Arawe Islands, West New Britain Province, Papua New Guinea, Asian Perspectites 33: 97-119. Groube, L. 1989. The taming of the rainforests: a model for late Pleistocene forest exploitation in New Guinea, in D.R. Harris & G.C. Hillman (ed.), Foraging and farming: the evolution of plant exploitation: 292-304. London: Unwin & Hyman. Groube. L., J. Chappell. J. Muke & D. Price. 1986. A 40,000 year old human occupation site at the Huon Peninsula, Papua New Guinea, Nature 324: 453-5. Haberle, S. 1993. Pleistocene vegetation change and early occupation of a tropical mountainous environment, in Smith et al. (ed.): 109-22. Hiscock, P. & A.P. Kershaw. 1992. Palaeoenvironments and prehistory of Australia's tropical Top End, in Dodson (ed.): 43-75. Kirch, P.V. 1989. Second millennium BC arboriculture in Melanesia: archaeological evidence from the Mussau Islands, Economic Botany 43:225-40. Loy, T.H., M. Spriggs & S. Wickler. 1992. Direct evidence for human use of plants 28,000 years ago: starch residues on stone artefacts from the northern Solomon Islands, Antiquity 66: 898-912. Marshall, B. & J. Allen. 1991. Excavations at Panakiwuk Cave, New Ireland, in Allen & Gosden (ed.): 59-91. Moore, T.C., et al. 1980. The reconstruction of sea Surface temperatures in the Pacific Ocean of 18,000 b.p., Marine Micropaleontology 5: 215-47. Pavlides, C. 1993. New archaeological research at Yombon, West New Britain, Papua New Guinea, Archaeology in Oceania 28: 55-9. Pavlides, C. & C. Gosden. 1994. 35,000 year old sites in the rainforests of West New Britain, Papua New Guinea, Antiquity 68: 604-10. Rind, D. & D. Petet. 1985. Terrestrial conditions at the Last Glacial Maximum and CLIMAP project members sea-surface temperature estimates: are they consistenty, Quaternary Research 24: 1-22. Smith, M.A., M. Spriggs & B. Fankhauser (ed.). 1993. Sahul in review. Canberra: Department of Prehistory, RSPacS, Australian National University. Occasional Papers in Prehistory 24. Southern, W. 1986. The late Quaternary environmental history of Fiji. Unpublished Ph.D thesis, Department of Biogeography and Geomorphology, Australian National University, Canberra. Specht, J., I. Lilley & J. Normu. 1981. Radiocarbon dates from West New Britain, Papua New Guinea, Australian Archaeology 12:13-15. Spriggs, M. 1993. Pleistocene agriculture in the Pacific: why not?, in Smith et al. (ed.): 137-43. Spriggs, et al. (ed.). 1993. A community of culture. Canberra: Department of Prehistory, RSPacS, The Australian National University. Occasional Papers in Prehistory 21. Stuiver, M. & P.J. Reimer. 1993. Extended [.sup.14]C data base and revised Calib. 3.0. [.sup.14]C age calibration, Radiocarbon 35:215-30. Summerhayes, G.R. & J. Allen. 1993. The transport of Mopir obsidian to Pleistocene New Ireland, Archaeology in Oceania 28: 144-8. Swadling, P. 1994. Changing shellfish resources and their exploitation for food and artefact production in Papua New Guinea, Man and Culture in Oceania 10:127-50. Swadling, P., N. Araho A B. Ivuyo. 1991. Settlements associated with the inland Sepik-ramu sea, in P. Bellwood (ed.), Indo-Pacific prehistorv 1990:92-112. Jakarta- Indo-Pacific prehistory Association. Bulletin 11. Swadling, P. & G. Hope. 1992. Environmental change in New Guinea since human settlement, in Dodson (ed.) 13-42. Torrence, R. 1992. What is Lapita about obsidian? A view from the Talasea sources, in J.C. Galipaud (ed.), Poterie Lapita et peuplement- 111-26. Noumea: ORSTOM. White, J. P. 1971. New Guinea and Australian prehistory: the `Neolithic problem', in D.J. Mulvaney & J. Golson (ed.), Aboriginal man and environment in Australia: 182-95. Canberra- Australian National University Press. White, J.P., T.F. Flannery, R. O'Brien, R.V. Hancock &, L. Pavlish. 1991. The Balof shelters, New Ireland, in Allen & Gosden (ed.):46-58. Wickler, S. 1990. Prehistoric Melanesian exchange and interaction: recent evidence from the northern Solomon Islands, Asian Perspectives 29:135-54. Wickler, S. & M. Spriggs. 1988. Pleistocene occupation of the Solomon Islands. Melanesia, Antiquity 62:703-6. Yen, D.E. 1990. Environment, agriculture and the colonization of the Pacific, in D.E. Yen & J.M.J. Mummery (ed.), Pacific production systems- 258-77. Canberra: Department of prehistory, RSPacS, Australian National University. Occasional Papers in Prehistory 18.