Traditionally flakes possessing a length dimension at least twice that o f the width are referred to as blade flakes (Crabtree 1972:42). To avoid confusion with prismatic blades and microblades, introduced in following chapters, I will use the alternative term “elongate flake” when referring to long narrow flakes.
Elongate flakes are extremely rare in the four sampled quadrants (see Figure 4.15). In Pleistocene Phase IV, elongate flakes comprise only 2.8% (5) o f the total. For Phase III the proportion is around the same, at 2.6% (3). O f the 288 flakes associated with Phase II occupation, 9% (26) are elongate. Elongate flakes constitute 8.6% (6) o f the Phase I assemblage. A clear discrepancy exists between the proportions o f elongate flakes in Pleistocene/early Holocene occupation, and occupation during the mid to late Holocene. Between these two periods the proportion increases three-fold. Interestingly, there is no real change between Phase II (mid Holocene) and Phase I (post-Lapita).
An examination o f the entire obsidian assemblage revealed 10 microblades or microblade segments, all in mid to late Holocene deposits. Blades and microblades are here distinguished from elongate flakes by regular, parallel sides and the presence of a distinctive crest, indicative o f removal from a blade core. These are illustrated in Figures 4.17, 4.18 and 4.19. One (Figure 4.17b) has been bifacially retouched into a form resembling some blades recovered from sites on Lou (Chapter 6), although it is much smaller. The low number o f microblades demonstrates that reduction was
Excluding the retouched microblade, only five obsidian pieces in the shelter have edge scarring which can be unambiguously classed as deliberate modification. These were not shaped to any particular form and were all recovered from spits associated with the midden. Four are illustrated in Figures 4.20 and 4.21. Six other pieces exhibit less regular edge microflaking which could be use-wear. Only one, from Spit 9B in the SE quadrant o f Square 2, was in a Pleistocene context.
PATTERNS OF ARTEFACT USE
Major chronological changes took place in artefact use at Pamwak. The initial period o f site occupation, beginning possibly before 20,000 BP on the basis o f age/depth extrapolation, saw the use o f water-rolled nodules for the manufacture o f unretouched flakes. As I have mentioned previously in this chapter, stream cobbles are present in the nearby Losa River. On hand specimen comparison the nodules in Pamwak are lithologically similar to some o f the river cobbles. I make the logical assumption that the Losa or its tributaries provided the source for the water-rolled archaeological stone recovered from the shelter. In comparison with the later obsidian flakes, the chert and other stone pieces from these levels are large. No pieces possessed retouch. The picture at this time is therefore o f Pamwak being used as a stopover camp during expeditions inland.
Obsidian appears in the shelter around the time o f the Pleistocene/Holocene transition. Its appearance is associated with a rapid decline in the use o f locally available stone. Two age range estimates are available for this change, one o f 13,129-12,625 BP (ANU 8253) from Spit 14 in the SE quadrant o f Square 1 and the other o f 13,802- 13,247 BP (ANU 8242) for Spit 13 in the NW quadrant o f Square 2. These estimates are statistically significantly different at a 95% confidence level; T = 11.35, chi square
= 3.84. The summed probabilities of these two estimates provides a range o f 13,786- 12,627 BP for the period when obsidian begins to be transported to the shelter. The virtual total replacement o f local stone by obsidian after this time suggests the utilisation o f distant resources was carried out sufficiently regularly to maintain a constant flow o f material into the site. Obsidian used during Pleistocene occupation is compositionally unlike the southwest Manus source and is very similar to material from the Pam Islands. If these islands were the source, and on available evidence they are the best candidate, regular water crossings must have been made at this time. At this time crossings o f approximately 20 km would have had to have been made to collect obsidian from the enlarged landmass that now constitutes the Pam Islands and Lou.
The appearance o f retouched pitchstone tools and stone axes in the terminal Pleistocene signifies a major change in stone tool technology. Retouched implements o f the kind found in Pamwak occur nowhere else in Melanesia, although tools very similar to the discoids have been recovered from early Holocene contexts in localities further to the west in Southeast Asia (Anderson 1987:223; Glover 1977:55). Edge ground axes are documented for the terminal Pleistocene on mainland New Guinea (White 1972:95), as well as Australia (Morwood and Trezise 1989; C. White 1967) and, further afield, Japan (Blundell and Bleed 1974). The appearance o f edge ground axes in Pamwak may signify the beginning o f land clearance and, by implication, the first stages in agriculture (Groube 1989; Spriggs 1993).
The early Holocene witnessed the disappearance o f pitchstone implements but a continuation in the use o f stone axes. Tridacna shell adzes probably appeared between the early and mid Holocene. In previous publications both Spriggs
a possibility, I now consider that evidence for stratigraphic disturbance, discussed above, makes it likely that some Holocene shell adzes have been displaced downward into underlying Pleistocene deposits.
A marine focus to occupation in the early Holocene is evident in the presence o f sparse midden and trochus fishhook blanks. This economic focus is probably a reflection of rising sea levels transforming the shelter into a near-coastal locality. In the period 9400-8500 BP obsidian from known Pam Islands source deposits came to replace Source X obsidian.
Between approximately 7200 BP and 5000 BP occupation o f the shelter was heavily associated with marine exploitation. This is clear from the presence o f a dense marine and estuarine midden in all excavation squares. At this time Pamwak may have been at its closest to the shoreline, although any estimate o f distance is dependant on more information on the rate and timing o f Holocene uplift on southern Manus (Chapter 2) Shell adzes and fishhook blanks are associated with the midden. The amount of obsidian reaching the shelter increased dramatically. The Pam Islands were the major obsidian source during this occupation but obsidian from Lou begins to appear in significant quantities. At some time between approximately 5000 BP and 2000 BP obsidian from Lou became as important as that from the Pam Islands.
After the mid Holocene there appears to have been an hiatus in site occupation until the last 2000 years or so. Pottery associated with occupation at 2100 BP and 1650 BP elsewhere in the Admiralties is associated with the terminal occupation phase at Pamwak (Phase I). Generally, this occupation is ephemeral in comparison with the earlier Holocene. Edge ground tools are absent and an overall decline in the amount o f obsidian reaching the shelter is apparent. Lou obsidian has by this time become as important as that from the Pam Islands. Compositional analysis o f the single
retouched microblade segment using EDS revealed that the obsidian originated from a Pam Islands source.