In the preliminary report o f the Pamwak investigation (Fredericksen et al. 1993) seven layers were described for the site. The identification of these was made on the basis
o f visual differences in colour, texture and composition. The layers are (Figures 4.4 and 4.5):
Layer 1: a disturbed dark friable soil with scattered shell (mainly freshwater species), pottery, bottle glass, small amounts of flaked obsidian and stone, and some trade beads.
Layer 2: a dense shell midden containing marine, mangrove and freshwater species; stone included numerous obsidian flakes as well as lesser amounts o f oven stone fragments and flaked chert and other stone.
Layer 3: a mixed layer consisting o f fragmentary shell midden and dark grey sediment. Layer 4: a brown fine textured sediment with scattered shell fragments in the upper few spits and flaked obsidian and smaller amounts o f stone throughout.
Layer 5: a dark grey to brown sediment containing flaked stone (predominantly locally available chert) and small amounts o f obsidian.
Layer 6: a light brown sediment characterised by numerous small indistinct lenses and virtually no obsidian among the flake debitage.
Layer 7: a dense, compacted orange-brown clay containing flaked stone but very little organic matter.
These divisions were identified to facilitate broad comparison between squares. However, with the obvious exception of the Layer 2 midden, the distinctiveness o f these layers is probably due to pedogenic processes rather than cultural activity. Although useful for providing an overview of the stratigraphic composition o f the shelter, analysis by layer has the serious limitation that any fine-grained variation in human activity will be subsumed within a broad stratigraphic unit which may encompass a time period spanning more than a millennium (see discussion on dating below). For a more detailed explication of temporal change an analysis by excavation spit is required. In the following discussion I therefore employ individual excavation spits as the stratigraphic unit o f study. The relationship between excavation spits and
sediment layers can be ascertained by reference to the vertical scale in Figures 4.4 and 4.5 (Note that Spits 1 and 2 are 20 cm thick, the remainder are 10 cm).
DATING
A general chronology has been previously presented for Pamwak (Fredericksen et al. 1993:146-147); Layer 1 probably dates to within the last 2000 years, Layer 2 to mid Holocene times, Layers 3 and 4 to the terminal Pleistocene (c 11,000 BP for 4), while Layers 5, 6 and 7 have yet to be adequately dated but possibly represent occupation extending back to before 20,000 BP (Fredericksen et al. 1993:149). In all 32 radiocarbon estimates are now available, these are set out in Appendix A o f this thesis. Calibrated age estimates at two sigmas are shown in Table 4.1. No reliable estimates are available below Spit 21, although flaked stone is present down to Spit 37. Dates o f 6190 ± 550 b.p. (ANU 7508) and 5230 ± 450 b.p. (ANU 7509) have been obtained for Spits 34A and 28A, but these are far too recent to make stratigraphic sense. There is no obvious explanation for these anomalous dates, although sample contamination is one possibility. They have been omitted from Table 4.1. Ascertaining radiometric determinations for lower excavation spits is made difficult by the paucity of organic material from deeper deposits. Small flecks of charcoal obtained from toward the base o f the excavation have been submitted for dating by accelerator mass spectrometry, but results are not yet available. Thermoluminescence has been shown to possess great potential for determining the age o f Pleistocene deposits in this region o f the world (see Roberts et al. 1990), and a case could be made for a further expedition to Pamwak to collect samples for dating using this technique.
ranges representing the highest percentage o f the area under the calibration curve are shown (see Stuiver and Reimer 1993:8, 10). All dates are on charcoal and uncarbonised Celtis (native elm) seed. By weight Celtis seed is the most abundant non-marine dating material in Pamwak. While it is not known how groundwater uptake by a Celtis tree growing in a limestone environment will influence carbon accumulation in the tree’s seeds, dates on Celtis do generally correlate with those on charcoal. Two dates on marine shell were obtained for Spit 8 o f Square 1, on Anadara (ANU 6977) and an unidentified species (ANU 6978). These encompass an age range o f 8138-5551 BP, which is significantly earlier than the most representative probability range o f 9439-8503 BP on charcoal and Celtis seed.
Ambrose (1994) has recently undertaken hydration dating on obsidian from a 1.3 m column in the SE quadrant o f Square 2. Using measurements taken from inside the internal fissures o f flakes Ambrose found a fair correlation between age and depth. His results (1994:Table 2) are presented as a graph in Figure 4.6.
Comparison o f all dates reveals the likelihood o f disturbance within the site. An age range o f 13,110-8204 BP obtained for Spit 21 in Square 1 is much too recent and should probably be discounted Chronometrie inversion is apparent for dates in Spits 4 and 5 in Square 1, and Spits 7 and 9 in the SW quadrant o f Square 4. This is clear evidence o f deposit mixing, also evident in fluctuations in the age/depth curve o f obsidian dates for the SE quadrant o f Square 2 (Figure 4.6). In the NW quadrant o f Square 4 age ranges o f 12,273-11,048 and 9644-8415 BP were obtained on samples collected from near the base o f the Layer 2 midden. These are considerably earlier than other estimates for this deposit. Clearly, at least some o f the midden has been mixed with earlier deposits.
Despite these Chronometrie inconsistencies the general pattern is one o f increasing age with depth. Spits above the Layer 2 midden have not been dated by radiocarbon. However, the presence of distinctive pottery styles, including Sasi and Puian types (Ambrose 1991:106-109), strongly indicates occupation took place within the last 2000 years. Pottery was recovered from the uppermost spit o f the midden in most excavation quadrants. Stylistically this is similar to sherds from overlying spits and some conjoins were observed between pieces vertically separated by up to 10 cm. Rather than representing a mid Holocene pottery tradition, it seems likely that this indicates the incorporation o f later material within an earlier midden.
Radiocarbon estimates show the midden was deposited between around 7200 BP and 5000 BP. The combined probabilities for all age estimates obtained for the midden, discounting probabilities for the two anomalous old estimates for the NW quadrant o f Square 4, indicate the deposit accumulated during the period 7180-5260 BP. The midden was clearly laid down over a number of occupation episodes, during which the shelter was probably a campsite for people focussing on the exploitation o f marine and estuarine resources. Discrete shell lenses and bands o f different compactness were evident within the midden, testimony to a series o f formation events. More information on the composition o f the midden will be forthcoming after analysis o f its faunal component is completed by Corrie Williams o f Monash University.