Volume 100 1991 > Volume 100, No. 2 > A reappraisal of the dating for some Lapita sites in the reefs ..., by R. C. Green, p 197-208
A REAPPRAISAL OF THE DATING FOR SOME LAPITA SITES IN THE REEF/SANTA CRUZ GROUP OF THE SOUTHEAST SOLOMONS
In a review of radiocarbon dates and their calibration for sites currently assigned to the Lapita cultural complex, Kirch and Hunt (1988:28) comment on a formerly “common practice of submitting one date to represent an entire site or a local sequence”. They question the usual inference associated with this practice that one or a few dated specimens provide a valid estimate of the age of an archaeological assemblage from a given context because it is unlikely that all materials were deposited simultaneously. Moreover, they note that post-depositional alterations of Lapita sites during the last several thousand years (Green 1979:31) must be viewed as commonplace events, not exceptions, which further complicate the dating process. Consequently, they take the view that all “local sequences built on few radiometric data must be open to doubt” (Kirch and Hunt 1988:28). Of equal importance (although they say less about it), is the fact that, for sequences established more than a decade ago, a number of factors with respect to their age assessments may have significantly changed. These include the way in which former radiocarbon dates would now be interpreted, the prospect of additional dates or techniques for dating and the impact of further analyses of various items effecting how the local sequence might now be integrated.
The intent of this brief report is to update the understanding of the Lapita part of the Reef/Santa Cruz local sequence. It arises as a consequence of two new radiocarbon dates for one site, SE-RF-2, the revised calibrations now possible for all dates on sites associated with this local sequence, and additional analyses indicative of the relative age of the several excavated Lapita assemblages derived from these sites over the years.
Seventeen years ago three excavated Lapita sites from the region, SE-SZ-8, SE-RF2 and SE-RF-6, were placed in that order based on one analysis of their decorated body-sherds, pairs of radiocarbon dated charcoal samples for two of the sites (RF-2 and RF6) and some preliminary measurements of hydration rims on obsidian flakes from each of the sites. The relative order from oldest, SZ-8, to youngest, RF-6, has been supported in most subsequent analyses, but nowhere has the case made then been compiled in detail or fully evaluated. Nor have recent calibrations of the various radiocarbon dates obtained at that time, or since (Kirch and Hunt 1988: Table 2.3; Spriggs 1990: Table 1), been - 198 subject to really close scrutiny with respect to their bearing on the age structure of the entire local sequence. The following summary attempts to provide such a reassessment.
Site order based on ceramic assemblages
The original order for the three Lapita sites, SZ-8, RF-2 and RF-6, derived from the design study done by Donovan (1973) using all the decorative material for body-sherds in a rather general way. Subsequently, Green (1978) assembled a more formal set of 54 Lapita design motifs useful in establishing linkages between a number of regions with Lapita sites of which all but the Watom site were located in Remote Oceania (Green 1990: Appendix 5). These were subjected to various forms of analyses, one of which addressed the topic of decorative decay through time in the Reef/Santa Cruz region (Green 1978:12-13). In it the following points were noted for a set of 37 motifs: a) five popular motifs in the region (also found in sites elsewhere) occur only in SZ-8 and RF-2 but not in RF-6; b) a matrix of 36 motifs common to these three sites shows 32 motifs shared between SZ-8 and RF-2, 26 between RF-2 and RF-6, but only 18 between SZ-8 and RF-6; and c) that, for 13 of the 36 motifs which are restricted to the Reef/Santa Cruz sites, although five are common to all three sites, four others are shared between SZ-8 and RF-2, three others between RF-2 and RF-6, and only one other between SZ-8 and RF-6. Thus, these points were interpreted as consistent with the previous views in suggesting that RF-6 was the late site, that RF-2 and SZ-8 shared more decorative designs in common, and that, in ceramic composition, SZ-8 was the early site. This regionally focussed perspective was fully supported by two more general interregional comparisons. One was a close proximity analysis of the Jaccard coefficient data for the 54 linked motifs, which included data from these three sites, six others, and a set of Tongatapu sites as a tenth entry; it also employed a sorted matrix analysis of that same data (Green 1978: Figures 7 and 8).
Parker (1981) also studied the pottery collections from RF-2, both those excavated in 1971 and in 1976-7, as well as those of SZ-8 and RF-6, especially with respect to rim type, base form, vessel shape, and the placement of decoration on them. Among her observations were some of chronological significance. One is of three rim forms in SZ-8 not encountered in RF-2 and RF-6 (Parker 1981:70-1). Another is that, although the vessel shapes in all three sites are in general very alike, those of SZ-8 exhibit the greater range with four types found only there (Parker 1981:78, Figure 19). One of these is a ring-footed vessel (Parker 1981: Figure 30) thought to be an early attribute among Lapita vessel forms (Anson 1983:265). Thus, once again an early-to-late sequence in decorated vessel shapes from SZ-8 to RF-2 to RF-6 appeared to make good sense of these data, both in the Reef/Santa Cruz region and also more generally (Green 1990: Figures 1A and 1B).
Finally, Anson (1983:175-81 and Table XII) employed a type of decorative analysis which used quantitative data on each alloform (variation in a motif) from these three sites in the context of a comparative exercise that included six other Lapita sites, four of them from the Bismarck Archipelago in Near Oceania. The degree of similarity between various site assemblages was then calculated by a variety of statistical methods. One preliminary comparison, using only the 46 alloform variables present in sites from the - 199 Bismarck Archipelago, showed the SZ-8 and RF-2 assemblages to be extremely similar, with RF-6 a little more distant, and all three forming a discrete cluster (Anson 1983: Figure LXIV). When the entire suite of alloform frequency figures was used, however, a dendrogram employing Robinson Index values revealed RF-2 and RF-6 exhibited more in common than either had with SZ-8, all three sites again forming a separate cluster (Anson 1983: Figure LXV).
In a non-quantitative presence/absence comparison which included SE-SZ-45 as well, along with the Malo site from Vanuatu, and Natunuku and Yanuca from Fiji, RF-6 and RF-2 again fall together, following which SZ-8 and then SZ-45 joined them in that order at slightly lower values (Anson 1983: Figure LXVI). Typically, all the Reef/Santa Cruz sites clustered together. Like Anson (1983:187), I interpret the above results as showing that “geographical location and proximity exercise the predominant influence in the formation of clusters”. However, within the Reef/Santa Cruz grouping, as Anson found among sites from the Bismarck Archipelago, time appears to play a secondary role in their ordering. RF-6 is the later site most similar to RF-2, and SZ-8 and SZ-45 are more distant earlier sites. Moreover, two principal coordinate analyses indicate that it is through RF-2, not RF-6, that the linkage is established (Anson 1983: Figures LXVII and LXVIII), as was also indicated by Anson's first comparison above.
It should also be noted that, when McCoy and Cleghorn (1988:113) reported on their own work at the Lapita sites of SE-SZ-23 and SZ-45 on Santa Cruz and correlated it with the ceramic analyses of Parker (1981) and Manser (n.d.), they too concluded that these two sites were penecontemporaneons with SZ-8 and RF-2 but not RF-6.
In summary, from the first non-quantitative pottery analysis based on decorated body-sherds through various presence/absence comparisons of selected motifs to studies of rims, bases and vessel shapes and thence to quantitative analyses involving comparisons with earlier sites from the Bismarck Archipelago, support has been found for SZ-8 being the earliest site and RF-6 being the latest. The link has been through RF-2, which, although sharing strong ties with both, has frequently seemed to be somewhat more closely related to SZ-8 (and to SZ-23 and SZ-45). This ordering, of course, depends on the assumption that the earliest site in this case exhibits the most complex ceramic assemblage in terms of vessel shape and associated elements of decorative design, and that these have simplified through time in this region. These assumptions are at present generally supported by analyses of the whole Lapita ceramic system in showing distance decay in early assemblages as one moves from west to east, with each local region then exhibiting some loss of the more elaborately dentate decorated vessel shapes through time. Thus, the Reef/Santa Cruz Lapita case fits the general pattern for change in that ceramic series as it is currently understood.
Site order based on obsidian and chert assemblages
A pattern of consistent change also emerges in the excavated obsidian and chert collections when the order established by the ceramics is applied to them. One example is set out in Table 1. This shows that, in general, the same total number of pieces of pottery, obsidian and chert, circa 5900, may be expected to be recovered from the excavation of 100 m2 towards the interior of each of these sites. It should be noted that, - 200 with respect to these and the following calculations, the sample at 51 m2 for SZ-8 is representative of a surface area of 459 m2 and that of 20 m2 for RF-6 of 180 m2 (Green 1976:252 and Table 18), while the areal excavation at RF-2 in the end encompassed 153.5 m2 (Green 1986: Figure 8.2). In Table 1 results have been made comparable by reducing figures for each site to units of 100 m2 from which the percentages have then been calculated. What changes from the earliest SZ-8 site to the latest RF-6 site is a steady reduction in the number of obsidian pieces from more distant sources (all but 1 per cent from over 2000 km [Green 1987]) in relation to chert items from sources much nearer (400 and 100 km [Sheppard and Green: in press]). The major change is in RF-6, where the overall number of obsidian flakes falls off significantly so that pieces of chert and obsidian, once at a ratio of 6.5 to 1, now occur in about equal proportions.
Examining only obsidian, it has also been shown that the size of the obsidian pieces (based on weight) get smaller as one moves from SZ-8 through RF-2 to those of RF-6 (Lawlor n.d.; see Table 2). Not surprisingly then, the estimated amount of obsidian in each site, when calculated to a standard site size of 10,000 m2, shows a marked reduction in total quantity from SZ-8 to RF-6, as well as a fall in the number of pieces recovered per cubic metre excavated (Table 2).
The interpretation of these changes as being in large part of temporal origin, and not simply fluctuations arising from functional differences between the areas covered by the excavation samples, is based on the extent of the zones represented. At SE-RF-2 it is known that the assemblage derives from an 153.5 m2 zone with a large central building, an area between, and a cooking area in which the concentrations of pottery, obsidian and chert in each portion vary (Sheppard and Green: in press). The 459 m2 of area sampled at SZ-8 and the 180 m2 at RF-6 are. therefore. probably also large enough to have encompassed several different activity areas rather than reflecting one small locality of particularly high or low density.
Radiocarbon age assessments
After the two pairs of radiocarbon samples for RF-2 and RF-6 had been radiocarbon dated, and the initial ceramic analysis indicated SZ-8 as the earlier site, samples suitable for radiocarbon age determinations were sought for it. Because no charcoal samples from really secure contexts were recovered at SZ-8, marine shell samples had to be used. These produced two uncalibrated results, both of which appeared to support the view that SZ-8 was the earlier site (Green 1979: Table 2.1). At that time, suitable methods permitting reasonable corrections for ocean reservoir effect were not yet available and a need for an appropriate but unknown secular correction was, therefore, indicated by a “?”. The pairs of charcoal dates for RF-2 and RF-6, however, could be legitimately pooled and then calibrated from a 1972 MASCA table as 1130±69 B.C. and 680±67 B.C. respectively (Green 1979: Table 2.1).
Procedures for secular corrections of radiocarbon dates have greatly improved since 1972, and Table 3 sets out the best calibrations of calendrical age now available for these three sites. They are an advance on anything earlier, both in terms of the precision of the calibrations and because two more charcoal samples from RF-2 have been dated. The - 201 age for RF-6 continues to fall in the range 800-500 B.C. with a probable date in the 7th century B.C. The previous pair of charcoal dates from RF-2, one from an oven in the cooking area of the site and another from a pit between it and the large rectangular structure in the centre of the site, are now joined by two compatible results from separate areas of two pits within the central rectangular structure. Statistically the four dates could all be drawn from the same population, and stratigraphically all are associated with the undisturbed grey sand occupation layer of a site best interpreted as a coherent single phase settlement of relatively short duration (Sheppard and Green: in press). As such, it is viewed as legitimate to pool the four dates to obtain a mean age centred on 1010 B.C. (Table 3) with occupation of the site spanning a century or less on either side of it. This is about a century later than those general estimates usually given in earlier accounts. With four mutually supporting separate dates from one reasonably well controlled short-term stratigraphic context, it also makes the RF-2 Nenumbo Lapita site one of the more securely dated assemblages from this cultural complex. Certainly it is now the most accurately C14 dated Lapita site in the Reef/Santa Cruz region.
The expectation that calibrated age estimates for the SZ-8 marine shell samples, when corrected for ocean reservoir effect, would continue to be earlier than those for RF-2, has not been realised. Rather, the calendrical age estimates completely overlap with those of RF-2 at one standard deviation (Table 3). In terms of estimates based on radiocarbon determinations, the ages of the deposits at the two sites are not significantly different. Whether the values are based on a Delta R of O, something in between (which I prefer), or 100, the indications are that, if SZ-8 is (in its ceramic and lithic content) earlier than RF-2 as discussed above, the interval is something less than a century or two. Thus, some published estimates of 1300 or 1400 B.C. for the age of SZ-8 are very probably in error.
Some support for a 12th century B.C. initial Lapita settlement of the Santa Cruz Group can be garnered from another charcoal date for the base of an early plainware Lapita site (SE-SZ-47) in a cave on Tö Motu Neo islet at the northwest end of Santa Cruz itself. This yielded a result of 1439 (1261) 942 B.C. (McCoy and Cleghorn 1988; Spriggs 1990:Table 1). Although this might be interpreted as allowing one to push the beginnings of occupation on Santa Cruz back into the 13th century B.C., it is just as likely to be of 12th century B.C. age consistent with the SUA-111 date for SZ-8. Either way, the regional Lapita sequence for the Reef/Santa Cruz Islands, on the basis of calibrated C14 dates, probably does not begin before 1300 B.C. although it was certainly well under way by 1100 B.C. Moreover, SZ-8 and RF-2 are much closer in time than previously believed and are penecontemporary with the basal layers of SZ-47, SZ-45, and the undated Layer VII of SZ-33. Layer VI of SZ-33, which also contains sherds of Lapita decorated pottery, is dated to 909 (785) 399 B.C. (Spriggs 1990:Table 1) and, thus, appears to be more or less equivalent in age to RF-6. In SZ-33 the Lapita ceramic tradition seems to terminate in Layer V (McCoy and Cleghorn 1988:110). Thus, on the basis of SZ-33 and RF-6, the Lapita part of the sequence in the Reef/Santa Cruz region currently ends towards the end of the 7th century B.C. In my view, this cuts the length of the Reef/Santa Cruz sequence associated with decorated as well as plainware pottery in the Lapita style down from a previously estimated 700 to 800 years to something of the order of 500 to 600 years.- 202
Obsidian hydration dating
Attempts have also been made to obtain age estimates for RF-6, RF-2 and SZ-8 by means of intrinsic hydration rate dating of obsidian flakes (Ambrose 1976). The outcome to date has not been very successful, except perhaps for RF-2, and only the data from that site (Table 4) are, therefore, reported here.
By 1973 preliminary measurements of hydration rims on obsidian samples from three sites had been made by Ambrose on four flakes from RF-6, eight from RF-2 and six from SZ-8. The results, as Ambrose (personal communication: 19 October 1973) reported them, were “mixed”. Those for RF-2 were “right on” but those for SZ-8 and RF-6 appeared to “wander around a little”. The calculations were based on an assumed value of 25°C for the long-term storage temperature of these flakes in the site. By 1979 the soil temperatures at RF-6 and RF-2 had been measured using several sets of thermal cells in pairs. Although problems occurred in the proper collection, as well as transportation of the cells to Ambrose for weighing, the best interpretation of the results is a black layer soil temperature at 25 cm depth of 25.1°C-25.5°C at RF-2, and a 25.4°C and a 25.7°C at RF-6 for 25 cm and 50 cm depths in layers 1 and 2 of that site. All the measured flakes from RF-2 reported here were recovered from black layer 1. Thus, in Table 4 the use of k as equal to a rate of.0036 at 25°C originally applied by Ambrose is judged to be acceptable.
In Table 4 one flake (ANU-166) has not been listed as its hydration rim had weathered away, while readings on two very obviously much older flake surfaces on one specimen (ANU-168a and c) were discarded as having been created at a stage much earlier than the last use of the piece in this site. Of the other two values for more recent flake scars on that specimen, the one (ANU-168b) which overlapped with the values for the pooled radiocarbon dates has been preferred to the other one (ANU-168d) which did not. The value for one other flake (ANU-172) has also been rejected because its age range did not overlap at two sigmas with the pooled range for the radiocarbon date and because, in relation to the other six readings used, it appears to be an outlier. The one sigma intrinsic obsidian hydration age ranges for six of the seven specimens that do overlap with the pooled and calibrated age range of four C14 dates for RF-2 is then calculated and found to be entirely compatible with, and in support of, the radiocarbon results (Table 4).
This, however, is unfortunately not the outcome of the results for either RF-6 or SZ-8, both using a split image ocular rather than a single filar ocular as at RF-2. Here the obsidian hydration rim measurements yield intrinsic ages that are younger to much younger than the radiocarbon ages obtained for these sites. In micron thickness all readings are not only less or much less than the six measurements found acceptable at RF-2, but only one age estimate among the four from RF-6 (that for ANU-216) even overlaps in part with the pooled and calibrated radiocarbon date range for that site: B.C. 779 (B.C. 317) 145 A.D. for the obsidian and 791 (762, 682, 659, 631, 610, 595) 431 B.C. for the charcoal. One can only conclude along the lines suggested by Ambrose (personal communication: March 1, 1991) that the whole set of fairly unsatisfactory readings for these two sites is simply “a result of surface weathering reducing the hydration thickness. This effect seems common in most coastal sites”.- 203
Thermoluminescence dating of pottery sherds
Another potential technique, thermoluninescence, for direct dating of items from two of these site assemblages was also tried. Initially three sherds, two from SZ-8 and one from RF-2 were counted and from this it appeared that some of the sherds were datable in principle if appropriate soil background rates could be determined (Prescott et al. 1982:145). Just to be sure, two more sherds, one from SZ-8 and the other from RF-2, were later measured and yielded unpublished results that appeared to confirm the previously reported count values for the sherds from each site (Robertson, personal communication November 18, 1982). Soil samples were also supplied for layers 1 and 2 of RF-2, but using neither the count values they provided nor those which assumed the background dose for the SZ-8 samples came largely from the sherd itself produced dates in the anticipated range for either site. Thus, results proved to be closer to 500 than 3000 years ago (Robertson, personal communication, July 14, 1982). In short, thermoluminescence dating of sherds from two of these sites has not worked at all, probably for the reasons set out in Prescott et al. (1982:146): a lack of tempers with abundant quartz and alkali feldspars plus their deposition in island soils with insufficient radiation levels to fully saturate crystal lattices in the sherds.
A reasonably well dated local sequence associated with assemblages assigned to the Lapita cultural complex in the Reef/Santa Cruz region currently depends on two outcomes: fairly secure direct dating of site SE-RF-2 at Nenumbo in the Main Reef Islands, and relative dating based on ceramic and obsidian analysis for the Main Reef Island site of SE-RF-6 and the Santa Cruz Island site of SE-SZ-8 with respect to RF-2. The later positioning of RF-6 with respect to RF-2 is reasonably well supported by a pair of calibrated radiocarbon dates. A much earlier dating of SZ-8 is not supported by the recalibration of a pair of marine shell dates from that site. Rather, SZ-8 seemingly belongs in the same age range as RF-2, i.e., between 1200 and 900 B.C., with the 11th to 12th century B.C. date provided by sample SUA-111 as indicative of a lower limit for its initial occupation. The somewhat shorter and more certainly one-phase occupation of RF-2 had occurred before the end of the 10th century B.C. and that of RF-6 before the end of the 6th century B.C. The whole classic Lapita sequence involving excavated sites SZ-8, SZ-45, SZ-47, SZ-33, RF-2 and RF-6 would appear to be contained within the years 1200–600 B.C. and more probably within the range 1150–650 B.C. This is a shorter period than that entertained in earlier publications.
I would like to thank my long-term collaborator in the Southeast Solomon Islands project, D.E. Yen, for assistance with obtaining the two additional C14 dates and Wal Ambrose for his continuing assistance with all aspects of trying out intrinsic hydration rate dating on these sites.- 204
Excavated Specimens of Decorated Body-Sherds, Obsidian Pieces and Chert Items in Three Lapita Sites in the Reef/Santa Cruz Island Group
A. Number of items in relation to actual excavation area.
B. Estimated number of items recovered per 100 m2.
C. Estimates by percentage among three items recovered per 100 m2.
Estimates for the Amount of Obsidian in Three Reef/Santa Cruz Lapita Sites
Calibrated Radiocarbon Ages for Three Excavated Reef/Santa Cruz Lapita Sites
NOTE: Data based on Spriggs 1990:Table 1; Kirch and Hunt 1988: Table 2.3; CALIB 2.1 and information supplied by the ANU C14 laboratory.
Intrinsic Hydration Rate Results for Obsidian Samples from SE-RF-2
One sigma range for pooled sample of four calibrated C14 dates 1197 (1009) 926
One sigma range for pooled sample of six hydration rim dates 1033 (951) 869
NOTE: Formula x2 = kt, x = micron thickness, k =.0036 at 25°C.
- 208 Page is blank