The potential value of obsidian as an indicator o f prehistoric interaction in the Melanesian region has long been recognised (Ambrose 1976; Smith et al. 1977.) Compositional analyses on obsidian have been invaluable in establishing how far (Ambrose and Green 1972; Best 1987) and how long ago (Summerhayes and Allen 1993) commodities were moved between the islands o f Melanesia. Much research has focussed on determining the extent o f chronological variation in the use o f source areas and specific source deposits, or subsources (Ambrose and Duerden 1982; Ambrose et al. 1981a, 1981b; Bird et al. 1981; Summerhayes and Hotchkis 1992; Summerhayes et al. 1993). Particularly relevant to the Pamwak situation is research
which has been undertaken on New Britain over the past few years (Summerhayes et al. 1993.; Torrence et al. 1992). This has looked at the development o f distribution over a period extending back well before the Lapita phase of prehistory. Information on the differential use of subsources during the Pleistocene has been obtained. One particularly significant piece o f information to come out o f this research is that specific Talasea subsources may have been targeted for exploitation for reasons not connected with access or quality, i.e. for extra-economic reasons (Summerhayes et al. 1993:63). This is the kind o f information which is needed for any reconstruction of developmental change in systems o f trade and exchange.
The impetus for undertaking a characterisation study o f obsidian from Pamwak was therefore the knowledge that this could reveal cultural patterning in the use of obsidian sources over time. A sample consisting o f 172 obsidian pieces was selected for analysis. The majority (159) came from the SE quadrant o f Square 2, representing 10% of all pieces in each excavation spit o f this quadrant. Although only 0.7% o f the entire obsidian assemblage by count was analysed the focus on a single column sample provided a good basis for stratigraphic comparison. I employed energy dispersive spectrometry (EDS) to measure the proportion o f major elements in each piece. A number o f statistical procedures were used to compare the compositional signature of these pieces with signatures obtained for obsidian from known deposits. In this way the Pamwak obsidian was “sourced” to geological provenance. A detailed discussion o f the EDS technique and the statistical procedures employed in the analysis can be found in Appendix B.
Results o f the characterisation analysis are summarised in Table 4.7. Three o f the 172 analysed pieces are excluded from the table as these revealed signatures well outside
including those in southwest Manus. Pieces originating from the Wekwok or Baun deposits on Lou were able to be distinguished from other Lou samples (Appendix B), but for ease of comparison I have in Table 4.7 included these with the Lou material. Three broad groups derived from the archaeological data are therefore shown in the table - pieces which correspond with obsidian from Lou, pieces corresponding with obsidian from Pam (i.e. Pam Lin and Pam Mandian) and pieces which form an entirely distinct group, which I have termed Source X. The latter is compositionally discrete from any o f the analysed reference obsidian, but does fall very close to obsidian from the Pam Islands. This can be visually determined by examining the correspondence plot in Figure 4.13. A Pam origin would therefore seem likely for Source X obsidian.
Table 4.7 shows that Source X material predominates in the Pleistocene and early Holocene levels o f the quadrant (Spits 7-16). Obsidian from Lou and Pam sources is present in only small quantities. This could indicate either only occasional use o f these sources or the intrusion of obsidian from overlying deposits. The latter is perhaps more likely as some vertical mixing was evident in this quadrant (above). A sudden change occurs in Spit 6, the unit immediately underlying the Layer 2 midden (Spits 3-5). Seven o f nine analysed samples from Spit 6 originated from a Pam Islands source. A predominance o f Pam obsidian is maintained in Spits 4 and 5. However, in Spit 3, the uppermost unit o f the midden, there is an increase in the proportion o f Lou obsidian to almost 50% of the analysed material. This pattern continues into Spit 2.
The Pleistocene occupation of Pamwak (Spits 9-16 in the SE quadrant o f Square 2, on radiocarbon dating) therefore saw an overwhelming reliance on Source X obsidian. Small amounts o f Source X obsidian are present in the upper levels o f Pamwak, but these are probably intrusive from underlying deposits. The basis for this interpretation is that no obsidian from this source has been identified by EDS or the more
discriminatory PIXE/PIGME technique in any upper Holocene sites elsewhere in the Admiralty Islands (see Chapters 5 and 7). This presents further evidence, albeit indirect evidence, for stratigraphic disturbance in the Pamwak site. I would argue that utilisation o f Source X had ceased by the mid Holocene.
The decline o f Source X need not be due to purely social factors. If this source was situated on one o f the Pam Islands its decline could easily be a result o f the inundation o f the deposit by rising Holocene sea levels. Alternatively, the volcanic activity and tectonic instability o f the St Andrews Strait islands could account for the changes in obsidian source accessibility over time. Volcanism certainly had a major impact on the exploitation o f various New Britain sources during prehistory (Summerhayes and Hotchkis 1992:132; Torrence et al. 1990; Torrence et al. 1992:91).
The Spit 6 shift to Pam Islands obsidian took place somewhere between 9440 and 8500 BP, on the basis o f summed probabilities for age ranges on three radiocarbon samples from this spit. Deriving an absolute date for the subsequent emergence of Lou as a source o f equal importance to the Pam Islands is more problematical. Spit 3 in the SE quadrant o f Square 2, the unit in which there is a sudden increase o f Lou obsidian (Table 4.7), straddles the transition between mid Holocene occupation and late Holocene ceramic occupation. The two alternatives are that Lou became an important source at either approximately 5000 BP, or at some time between this date and post-Lapita site reoccupation. The possibility therefore exists o f a correlation between the advent o f the Lapita phase o f the prehistoric sequence and an increase in the circulation o f Lou obsidian. Pamwak was not occupied in Lapita times so unfortunately a more refined chronology o f obsidian use cannot be determined on data from this shelter.
VARIABILITY IN FLAKE PRODUCTION
Analysis o f flake production was undertaken for two reasons. The first was to determine if change in lithic resource “maximisation” (Sheppard 1993) or “rationing” (Hiscock 1986) could be discerned from obsidian imported into Pamwak. Commentators have sometimes remarked that the size o f obsidian blocks traded in Melanesia decreased the further they were transported from the source (Green 1987:246; Harding 1967:42, see also Watson 1986:8), while the value o f the obsidian increased proportionally. Consumers at the tail end o f a distribution chain could therefore be expected to have access to cores smaller than those acquired by communities further up the chain, and to treat these pieces in a more economical way. One o f the few researchers to address this issue in Melanesian archaeology is Sheppard (1992; 1993). Sheppard’s work on obsidian reduction in the Reefs/Santa Cruz Islands was specifically directed toward testing the “lithic maximisation” hypothesis. Surprisingly, he found that obsidian reduction in these remote islands in Lapita times was not directed to conserving raw material, something which he (Sheppard 1993:135) attributes to the social value o f obsidian flake tools and cores being less than that o f unworked blocks.
Despite the somewhat equivocal findings of Sheppard’s research, the possibilities o f employing a maximisation model as a way of measuring the degree o f accessibility to obsidian are clear. In regard to Pamwak, analysing artefacts for variation in the economy o f reduction was carried out to ascertain if the availability o f obsidian had changed through time. I particularly wished to see if any evidence o f more economical use of obsidian was correlated with changes in the utilisation o f particular source deposits, as defined from the characterisation study. Such a correlation could provide evidence of significant change in the obsidian distribution network.
A second reason for studying variability in flake morphology at Pamwak was to determine the chronological origins o f the blade technology observed at European contact, and uncovered by archaeological research on Lou and southwest Manus (Chapter 3). Five large pitchstone blade-like flakes were present in Pleistocene deposits at Pamwak (see Plate 6 and Fredericksen et al. 1993:Figure 7), while the largest obsidian flake recovered, from an early to mid Holocene deposit, was similarly blade-like in form (Figure 4.14). The presence o f these demonstrates that the pre ceramic occupants o f Manus had the technological capability to prepare cores and remove long, narrow flakes. I wished to determine whether this technique continued through into the later Holocene by examining the shape profile o f smaller, less spectacular flakes.