All intact flakes from the following excavation quadrants were examined, Square 1/SE quadrant, Square 2/SE quadrant, Square 3/NE quadrant, Square 4/SW quadrant. Intact flakes are here defined as conchoidal pieces possessing both an intact platform and distal termination. In total, 649 flakes were selected for study. Three measurements were made on each. To gauge overall size each flake was weighed, while the shape o f pieces was determined by measurement of medial length and medial width. A scatterplot o f flake size and shape is illustrated in Figure 4.15.
In addition to size/shape, the presence o f dorsal detachments was also examined. These are indicative o f flake removal from a previously worked core and can be an indicator o f the intensity o f stone use. No retouching or macroscopic edge-wear was observed on any flakes from the four quadrants, although this does not necessarily mean none were used as implements (Fredericksen and Sewell 1991).
Most flakes had impact fractures on their striking platforms, large percussion bulbs, and distinctive percussion “ripples” radiating from the proximal end o f the flake. All these features suggest flakes were detached by hard hammer percussion (Odell 1989). Many broken flakes in the four sampled quadrants possess platform remnants which had been completely crushed by hammer impact. A bipolar technique has been observed in obsidian assemblages from elsewhere in the Bismarck Archipelago (Freslov 1989; Goulding 1987) but this was almost entirely absent in the Pamwak assemblage. Bipolar reduction is one way o f economising on raw material use as it enables flakes to be struck from very small cores. The virtual absence o f bipolar reduction could demonstrate obsidian was easily available at all stages o f occupation at Pamwak, although the use o f this technique can also be a response to factors other than raw material shortage (Goulding 1987:86-87).
Stratigraphic variation in the size and shape o f obsidian flakes in the four quadrants is illustrated in plots o f mean flake width and length (Figure 4.16). Although there is deviation between quadrants, the overall trend is toward shorter and narrower flakes with increasing depth. The trajectory for each quadrant demonstrates a plateau in particular spits - Square 1, Spits 5-7; Square 2, Spits 3-5; Square 3, Spits 3-4; Square 4, Spits 1-3. These spits are generally those which possess the highest concentrations o f obsidian, and are also those which either occur within or bracket the vertical distribution o f the dense midden layer (refer Table 4.5). On this evidence it therefore appears that during mid Holocene occupation and associated shell midden deposition not only was a large amount o f obsidian being imported into Pamwak, but it was being used for the production o f flakes generally larger than those made in earlier occupation.
To facilitate comparison between the four quadrants I have divided the columns o f each into four broad phases (Table 4.8). Phase I correlates with the ceramic period o f
site occupation over the last 2000 years or so. Phase II incorporates the time when the shell midden was deposited in the shelter. The top o f the midden is defined by a sharp decline in the quantity o f sherds and an increase in shell density. For quadrants in Squares 1 and 3 this approximately equates with the spit boundaries shown in Table 4.5. However, the boundary in the SE quadrant o f Square 2 occurs midway through Spit 3 while for the SW quadrant o f Square 4 this occurs midway through Spit 2 (2B). The base o f the midden is defined by a sudden decrease in the density o f shell. Phase III incorporates the early Holocene prior to the period o f midden deposition. This is distinguished on the basis o f the stratigraphic distribution o f radiocarbon dates (Table 4.1). Phase IV includes all underlying Pleistocene age spits down to where obsidian first appears in any quantity.
Table 4.9 sets out comparative data on the flake assemblage by the four occupation phases. This confirms the impression o f a stratigraphic decrease in flake size, and also indicates a trend in Phases I and II toward greater length relative to width. The large coefficients of variation in all phases reflect a wide distributional size range for flakes. To quantify the differences in flake weight, length and width between the four phases I employed a Mann-Whitney test (Siegel 1956:116ff). This is the non-parametric analog o f a t-test and compares the means of two sets o f observations to test the hypothesis that they derive from the same population. The probability values at a 95% confidence interval are presented in Table 4.10. A value o f 1.00 indicates that the two populations are identical while a 0.00 value reflects entirely dissimilar populations. For difference between assemblages to be statistically significant requires a value o f less than 0.95, which is a robust measure. It is clear from data in Table 4.10 that flake assemblages from all phases at Pamwak are statistically dissimilar. Nevertheless less variation is exhibited in weight, length and width
entirely discrete assemblages, bearing no similarity with either one another or the flake assemblages associated with Phases I and II. This reflects a trend o f decreasing flake size with increasing time depth for pieces recovered from early Holocene and Pleistocene deposits.
Table 4.11 contains information on the cores and core fragments recovered from the four quadrants. Cores are identified here as non-flake pieces possessing detachment scars on one or more o f their surfaces. The scars are placed irregularly and bear little resemblance to the patterned scars produced by implement retouching. All cores in Pamwak are multidirectional (Torrence 1992:119). As with flakes, a decrease in size is apparent with increasing stratigraphic depth. Variability is also evident in the amount o f reduction on each core and core fragment; the proportions o f cores covered entirely with flake detachment scars are 30% (6) in Phase IV, 37.5% (6) in Phase III, 44.4% (4) in Phase I, and only 7.7% (2) in Phase II occupation. This suggests that a relatively low intensity o f reduction per unit o f imported obsidian was carried out in Phase II. This could be expected if a plentiful and easily accessible supply of obsidian was available at this time.
The picture which emerges is one o f the production o f small approximately square shaped flakes in Phase IV Pleistocene occupation. Dorsal detachment scars are present on 19% (34) o f flakes in the spits o f Phase IV. This is a surprisingly low proportion and could indicate that small unworked cores were transported to the site. Early Holocene Phase III saw a general increase in flake size, as well as an increase in the proportion o f flakes with dorsal scars (29.2%). In mid Holocene Phase II occupation the pattern is toward larger and longer flakes. Dorsal scars are present on 25.7% (74) o f Phase II flakes. Flakes in Phase I ceramic period occupation are very similar in length and width to those in Phase II. However, only 5.7% (4) o f flakes
possess dorsal scars, which again could indicate the importation o f a relatively greater number o f unworked cores.