CHAPTER 1 0 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS PATRICK V. KIRCH In 1988 I reviewed the 79 extent ra- of Lapita occupation in the Bismarck Archipelago was diocarbon dates clearly associated with to be resolved.2 For other sites from which we exca- l dentate-stamped Lapita ceramics vated key assemblages, such as ECB, EHB, EKQ, and ' (Kirch and Hunt 1988a),I an exercise EKE, it was equally important to establish age and oc- that demonstrated how inadequately cupation sequences through the use of multiple sample dated most Lapita sites were. No less suites. /jt than 66% of dated sites had only one or As shown in Table 10.1, a total of 51 samples was two 14C age determinations, and only six dated from ten excavated sites. Thirty samples (59%) sites had more than four determinations are from Talepakemalai (ECA), while the remainder are (Kirch and Hunt 1988a, fig. 2.6). Especially troubling distributed among the other nine sites. Forty-one samples was the observation that "where multiple dates have are associated with Lapita cultural materials, while 10 been obtained from a single [Lapita] site, their distribu- samples come from post-Lapita contexts. The samples tion reveals how potentially misleading a single date were processed by two laboratories, 17 dates by the drawn from that suite might be. Consequently, local se- Radiocarbon Dating Research Laboratory of the former quences built on few radiometric data must be open to Department of Prehistory, Australian National Univer- doubt" (Kirch and Hunt 1988a:28). sity (H. Polach andJ. Head), and 34 dates by Beta Ana- In the Mussau Project I resolved to establish as firm lytic, Inc. (M. Tamers). The samples dated by the ANU an "absolute" chronological framework as feasible laboratory were all obtained during the 1985 field sea- within the cost constraints of radiocarbon dating. Es- son, while those dated by the Beta Analytic laboratory pecially for the critical ECA Site, it was essential to pro- were obtained from the 1986 and 1988 field seasons. cure an extensive run of "'C dates-on different kinds Of the 51 dated samples from Mussau, 20 were previ- of sample materials from the full range of stratigraphic ously reported in Kirch and Hunt (1988b), and an addi- and depositional contexts-if the disputed chronology tional six were listed in Kirch et al. (1991, table 2); the A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 19; TABLE 10.1 Radiocarbon age determinations ens, or similar combustion features are absent. As a re- from Mussau sites. sult, only 17 of our 51 dates from Mussau are on char- coal (8 samples), wood (8 samples), or coconut en- Site Charcoal Wood Shell Other Total docarp (1 sample). The other 34 samples represent ma- rine shell and, in one case (from the EKU Site), mam- ECA 6 8 15 1 30 mal bone (presumably Sus scrofa). ECB 1 2 3 The use of marine shell samples has both advan- EHB 2 2 tages and disadvantages. The shells selected for radio- carbon dating were all of shallow-water bivalve and EKQ 5 5 gastropod species collected for food and/or for in- EKU b 1 dustrial purposes by the prehistoric Mussau inhabitants. EKO 1 1 Species used in dating include the large bivalves Tridacna spp. (8 samples) and the closely related Hippopus hippopus EKS 1 1 (3 samples), Hyotissa hyotis (7 samples), Spondylus sp. (3 EHK 2 2 samples), A nadara antiquata (2 samples), Chama sp. (2 EKE 5 5 samples), and Laevicardium sp. (1 sample), and the gas- tropods Turbo spp. (5 samples) and Strombus luhuanus (1 EKL 1 1 sample).3 All of these taxa provided important food Total 8 8 33 2 51 sources, and some (such as Tridacna, Spondylus, and Turbo) a were also used as raw materials for manufacturing arti- Coconut endocarp facts. In selecting shell samples for dating, we were careful b Mammal bone to avoid specimens that showed signs of water-rolling or weathering, and which might have been a compo- nent of the natural beach or lagoon-floor environment. remaining 25 samples are reported here for the first We are reasonably confident that all dated shell samples were originally gathered while live on the Mussau reefs, thee. either for food or for raw material, and were culturally This chapter presents an overview ofthe Mussau deposited in their respective sites. If we are correct in this radiocarbon chronology, as well as a comprehensive assertion-and we have no reason to suspect other- listing of all dated samples, including information on wise-then the clear advantage of these shell dates is provenience, sample description, laboratory treatment, that they will not have any inherent "in-built age" factor. conventional age and 613C, and calibrated age ranges That is to say, the dated age of the shell should repre- and intercepts (Appendix 1 0. 1). sent the time period at which the living mollusk was gathered by a human and taken from reef flat to occu- METHODS pation site.4 Shell samples, unfortunately, do pose a thorny prob- In selecting samples for dating, priority was given to lem with regard to calibration, and to correlation with wood charcoal or to non-carbonized wood (such as charcoal and wood samples. The calibration issues are the stilt-house post bases in Areas B and C of the ECA complicated and significant, and I discuss these exten- Site) or other organic material (such as coconut endocarp) sively below. With regard to correlation, we made a when these were available. This decision reflects the more particular effort to obtain paired sets of shell and straight-forward calibration of wood or wood char- wood/charcoal samples, and were successful in dating coal using the terrestrial calibration curve. Unfortunately, such paired sets at ECA (Areas B and C) and at ECB. charcoal or wood were not as common in the Mussau These paired sets, discussed further below, provide a archaeological deposits as they are in some other Pa- critical basis for assessing the best-fit calibration method cific islands contexts. In ECA and other Mussau Lapita to be applied to the Mussau shell samples. sites, this partly reflects the nature of the Lapita occu- Laboratory methods, as reported by the ANU and pation on stilt-houses, so that typical hearths, earth ov- Beta Analytic, Inc. labs respectively, were standard. The 198 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS ANU lab reported that the 1985 charcoal samples were nologically derived calibration curves. However, the examined for rootlets, treated with hot 10% HC1, rinsed, calibrated age ranges thus derived are essentially sets of and dried; shell samples were cleaned of their exterior probability distributions; thus as Specht and Gosden surfaces with a dental drill and crushed before combus- rightly observe, in interpreting suites of calibrated ra- tion. For the wooden posts from ECA Area B, the slightly diocarbon ages "we are matching different distributions degraded exterior surface of the wood was first scraped to look for patterns in them, rather than dealing with clean, followed by the removal of a piece of the outer point determinations which will easily give us delimited 1-2 cm of solid wood; this was chopped into small periods with clear beginnings and ends" (1997:187). fragments, washed with distilled water, and the cellu- The calibration of samples grown in a marine envi- lose extracted for dating. ronment (such as mollusks) presents additional prob- The Beta Analytic, Inc. laboratory pretreated char- lems than for wood or charcoal samples which were coal samples by first examining for rootlets, followed grown in a terrestrial environment, because marine con- by a hot acid wash to eliminate carbonates, rinsing to texts are typically not in isotopic equilibrium with at- neutrality, with a subsequent hot alkali soaking to take mospheric 14C (Stuiver and Braziunas 1993). The world's out humic acids. After rinsing to neutrality, another acid oceans are a sink for old carbon, creating a "reservoir wash followed another rinsing to neutrality. Shell samples effect" (Taylor 1987:126-32), in which marine samples were pretreated by etching away the outer layers with such as shell yield ages that are somewhat older than dilute acid; they were then attacked with further acid to their apparent true age. Unfortunately, this reservoir ef- produce carbon dioxide which was used as the carbon fect is not a constant, and is known to vary historically source for dating. All benzene syntheses and counting over time, and geographically over space. Oceanic res- proceeded normally. Any variations from these stan- ervoir effects are known to be especially salient along dard procedures are discussed for specific samples in continental coastlines where there is significant upwelling Appendix 10.1. of deep water. Whereas the "model surface ocean" (0- 75 m depth) typically yields reservoir ages between - 200- CALZBRA TION 400 years older than the atmospheric age, the "model deep ocean" can yield reservoir ages of up to 1800 As most archaeologists are painfully aware (but col- years older (Stuiver and Braziunas 1993, fig. 5A). Taylor leagues in related fields, and the reading public at large (1987:129-31, table 5.6) provides examples of paired is typically ignorant of the conversion of a radiocar- shell and charcoal dates from California demonstrating bon "date" into a calendar age is anything but a straight- that even in a single locality, the apparent ocean reser- forward process. Even leaving aside the question of voir effect can range from -800 to + 170 years! whether the dated sample properly represents the "tar- Determining the specific reservoir effect for any get" event for which an age assessment is sought (Dean particular marine environment is a complex problem, 1978), problems of calibration arise from secular effects especially when one takes into account the probability (the major secular trend in 14C, along with the so-called of temporal variation. As Taylor (1987:127) suggests, Suess and de Vries effects), from reservoir effects (particu- "one approach to investigating the reservoir effect in larly for marine samples), and from isotopicfractionation. marine shells [is] to examine the 14C activity of contem- Fractionation effects are largely controlled through mea- porary samples to determine if the initial 14C concen- surement of 683C values and conversation of the ra- tration in such materials could be significantly different diocarbon age to a "conventional" age prior to calibra- from that of standard terrestrial organics." However, tion for secular effects (Stuiver and Polach 1977). Fur- this requires the radiocarbon analysis of pre-bomb shell ther, thanks to several decades of careful radiocarbon samples, collected before 1950, after which time the dating of dendrochonological samples, secular effects widespread testing of thermonuclear weapons injected can be accounted for back to 12,000 BP, at least for large quantities of artificial 14C in the atmosphere. Un- terrestrial samples (Stuiver and Kra 1986; Stuiver, Long, fortunately, we do not know of any such pre-bomb and Kra 1993). Computer programs such as CALIB collections of marine shells from Mussau which could and OXCAL now make it possible to rapidly calibrate be analyzed so as to provide an assessment of the local conventional radiocarbon ages based on dendrochro- reservoir effect prevailing during the early 20th century. A RADIOCARBON CHRONOLOGY FOR, THE MUSSAU ISLANDS 199 Lacking such a direct assessment of the local reser- Clearly, the calibration of marine samples from voir effect specific to Mussau, one may apply the gen- Lapita sites remains a problematic issue, for several rea- eralized "model surface ocean" calibration for marine sons: (1) local reservoir effects have not been empiri- samples developed by Stuiver and colleagues (Stuiver, cally established for specific localities where Lapita sites Pearson, and Braziunas 1986; Stuiver and Braziunas are situated; (2) the possible range of temporal varia- 1993). In this model, the "surface ocean" is a zone from tion in such reservoir effects is unknown; and (3) it is by 0-75 m deep, and the resulting calibration curve is a no means evident that a "model surface ocean" based smoothed version of the atmospheric curve, offset by on a zone 75 m deep will apply to mollusks grown on an average age of -373 years from the latter (Stuiver et reef flats only a few meters deep, or in lagoons with al. 1986:982). In theory, for samples such as mollusks relatively little circulation or exchange with pelagic wa- grown in the "surface ocean," the application of this ters. It is conceivable that there may be much variability marine calibration curve to conventional 14C ages (those in reservoir effects between samples grown on exposed which have already been corrected for isotopic frac- outer reef margins (where there is considerable surge tionation effects), should bring these ages into line with action and mixing of water) and those grown on shal- those from contemporary true-age samples grown in a low reef flats, in protected seagrass beds, or in lagoonal terrestrial environment. It must be stressed, however, basins (see Chapter 2 for descriptions of these environ- that this "surface ocean" model is only a first-order ap- ments in Mussau). Indeed, as I will argue for specific proximation, with well documented local variations (as cases below, there is some empirical evidence for dif- in New Zealand, where such variation has been tested). ferential reservoir effects in different coastal situations Where the specific local reservoir effect has been em- within the Mussau Islands. pirically determined through analysis of pre-bomb ma- rine samples, the generalized model may then be ad- CORRELATING MUSSAUMA RINE AND justed through the application of a AR correction fac- TERRESTRIAL SAMPLE PAIRS tor, which either adds to or subtracts from the average -373 year offset of the model calibration curve. Spe- One potential approach to resolving the problem of cific A\R values for Pacific Ocean localities vary consid- determining a specific reservoir effect for any Lapita erably, ranging from 115 + 50 years for Hawai'i, 140 + site is to obtain paired sets of marine and terrestrial 45 years for Eniwetok Atoll 57 + 23 for Samoa, and samples from identical stratigraphic contexts, as Taylor 45 + 30 for the Society Islands, to -5 + 35 years off the (1987:130-31) has done for the University Village Site coast of Australia, and -25 + 15 or -31 + 13 for New in California, and as Hunt and Kirch (1987) did for Zealand waters (Stuiver and Braziunas 1993, fig. 16; samples from the Manu'a Islands in American Samoa.5 Higham and Hogg 1995; McFadgen and Manning 1990; The key assumption here, of course, is that such pairs Phelan 1999). of samples do indeed represent the same event, or events In calibrating marine samples from Lapita and other that were penecontemporaneous. In all cases this as- Oceanic archaeological sites-virtually all of which are sumption needs to be supported on independent ar- from localities where the specific local reservoir effect chaeological and stratigraphic criteria. has not been empirically determined-different investi- Within our suite of dates from Mussau, we have gators have applied varying AR values. Kirch and Hunt two sets of such paired marine and terrestrial samples. (1988b: 162) used a weighted average of the mid-ocean The most secure of these, in that we can be reasonably values from Hawai'i, Eniwetok, and the Society Islands confident that all samples derive from a tight cluster of (100 ? 24), while Spriggs (1990a) used a AR value of 0 behavioral events, consists of two wood and two shell + 0. Most recently, Specht and Gosden (1997:177) cali- samples from Zone of Area B, at Site ECA (samples brated a series of 14C ages from the Bismarck Archi- ANU-5790, -5791, -5081, -5082). The two wood pelago not using the marine calibration curve at all, but samples are from posts that were clearly part of the rather the bidecadal atmospheric calibration curve off- same stilt-house structure, and since they still retained set by a standard -400 years. (Their method would have their bark and had visible adzing marks cutting through roughly the same effect as using the marine calibration the bark, were clearly cut fresh for construction pur- curve with AR = 0.) poses (i.e., there is no potential "old wood" factor). 200 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS The two shell samples (of Tridacna gigas and Hyotissa is not appropriate to this situation, since it yields age hyotis valves) represent culturally-deposited midden which ranges for the shell samples that are appreciably younger accumulated directly around the posts during the initial than those for the stilt-house posts. phase of stilt-house occupation. On archaeological and Stuiver and Braziunas suggest that "with a pair of stratigraphic criteria these four samples all represent contemporaneous wood and shell samples from a single events that should have occurred within a short time location, the reservoir deficiency may be calculated with- span, and hence be of equivalent age. Figure 10.1 is a out a direct calibration to the calendar time scale" plot of the calibrated pairs of wood/shell dates from (1993:152-53). Using their figure 15 (a plot of Holocene Area B in which the shell samples are calibrated on the surface-ocean 14C ages versus atmospheric 14C ages) one "model surface ocean" curve with a AR value of 0. may convert a measured wood 14C age "to a model Clearly, the fit is not good, with a resulting summed marine 14C age, which, when deducted from the mea- probability plot (the lowest distribution curve in the fig- sured shell 14C age, yields AR" (1993:153).6 Following ure) that is strongly bimodal. Since our independent ar- this method, the mean 14C age of the two Area B wood chaeological/stratigraphic evidence tells us that these posts is 2940 years, which converts to a model marine samples must be of equivalent age, the evident conclu- 14C age of approximately 3300 years (as calculated on sion is that the "model surface ocean" calibration curve Stuiver and Braziunas' figure 15B). If we then deduct FIGURE 10. 1 OXCAL calibration plot for matched pairs of wooden posts and associated shell samples from Area B of Site ECA. The shell samples have been calibrated using the standard "mixed ocean model" marine calibration curve. Although all four samples should date the same "target events", note the discrepancy between wood and shell samples. SUM AREA B PAIRS SUM AREA B PAIRS ANU-5790 2950?80BP l ANU-5791 2930?80BP __ _ CURVE C:\OXCAL\MARINE93.14C ANU-5081 3010?80BP ___ - ANU-5082 2950?8OBP SUM AREA B PAIRS_ 2500BC 2000BC 1 500BC I1OOOBC 500BC AD Calendar date A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 201 this model age from the actual measured 14C ages for sists of wood charcoal (Beta-20453), and two of which the two shell samples, we derive AR values of -290 (for are valves of the large mollusk Hyotissa hyotis (ANU- ANU-508 1) and -350 (for ANU-5082) respectively, or 5086, -5087). In this case the three samples are not from a mean AR of -320 years. the same excavation unit, but they are from the same If we re-run the calibration of the four paired stratigraphic context since the ECB Site has only a simgle, wood/shell samples from Area B using the atmospheric shallow occupation component. Archaeological evidence decadal curve for the wooden posts, and the marine (stratigraphic and artifactual) strongly suggests that this curve with a AR value of -320 for the midden shells, small site was occupied for a relatively short time span we obtain the summary plot shown in Figure 10.2, with (see Chapter 4), and we therefore would not expect a high degree of concordance between all four samples. dating samples from ECB to display a lengthy tempo- The combined probability distribution is strongly ral distribution. Figure 10.3 is a plot of the calibrated unimodal, as should be the case for a group of samples age ranges, again applying the atmospheric decadal curve that on independent archaeological/stratigraphic crite- for Beta-20453, and the model surface ocean curve with ria represent temporally equivalent or penecontemp- AR = 0 in the case of the two marine shell samples. As oraneous events. with the previous ECA Area B test pairs, the results are A second test of this type can be run with a set of again widely divergent, yielding a completely bimodal three samples from the ECB Site, one of which con- probability distribution. There are three possible expla- FIGURE 10.2 OXCAL probability plot for the same matched pairs of wooden posts and associated shell samples from Area B, Site ECA, shown in Figure 10.1, but with the shell samples calibrated using a AR value of -320. Note that all calibrated ages are now consistent, with a unimodal summed probability. SUM AREA B PAIRS -320 TrFTT 7 1 t T TiTr- SUM AREA B PAIRS -320 I~~~~~~~~~ _ ----+e--+ W ANU-5790 2950?80BP _ WANU-5791 293080BP _ CURVE c:\oxcal\marine93.14c i DELTA R-320, SANU-5081 3010?80BP _ S ANU-5082 2950?80BP - It SUM AREA B PAIRS -320 1 @ l . I ,.. ....................._______I__i_,I_.,__._I______________________.................. . I. .. 2500BC 2000BC I1500BC I1000BC 500BC AD Calendar date 202 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS SUM ECB PAIRS r--T--- r --r T--'T-- --X- - r -- ----r--_ SUM ECB PAIRS Beta-20453 3200?709D l CURVE c:\oxcal\marine93.14c - - . 4.4 4. -4 - ---- - -t- ANU-5086 3120?80BP ANU-5087 3150?80BP - ? -t t * + t0 SUM ECB PAIRS X , , I . . , i , . , , I J * i 2500BC 2000BC 1500BC 1 OOOBC 500BC AD Calendar date FIGURE 10.3 OXCAL calibration plot of matched pairs of charcoal and shell samples from the ECB Site. Marine samples are calibrated using the "mixed ocean model" standard marine calibration curve. Note discrepancy between charcoal and shell dates. nations: (1) the site was occupied in two discrete epi- Mussau region, we can apply a -320 AR value (since sodes with a hiatus of 500-600 years, an interpretation using a higher value of -350 or -370 would essentially for which there is no archaeological support; (2) the mean that there was no reservoir effect whatsoever), to charcoal sample represents wood which was already obtain a new plot as shown in Figure 10.4. We can see 500-600 years old when it was burned, a conceivable that the three samples are now brought into reasonably but less probable scenario in the humid tropics, where close correspondence, with the charcoal sample still wood rots or is consumed by termites rapidly; or (3) as slightly older (as might be expected from burning wood with the Area B paired samples, the "model surface that was somewhat older than the target event), but with ocean" marine calibration curve has produced ages for a combined probability distribution that is unimodal, as the shell samples that are too young by several centu- we would predict on the basis of our independent ar- ries, because the local marine reservoir in which the chaeological criteria. Mussau samples were originally grown was different. These two tests with matched sample pairs indicate The third option seems to be the most probable. that the ocean reservoir effect for Eloaua Island was If we again apply the method of Stuiver and relatively slight, with an offset from the atmospheric Braziunas (1993) for the ECB paired samples, the re- curve of perhaps only 50 years. Applying a standard sulting AR values are -350 and -370 years, slightly more "model surface ocean" calibration curve (with its aver- than those derived in the case of the ECA Area B paired age offset of 373 years) is therefore inappropriate, as it samples (-320). Taking a conservative view that there yields ages that are considerably too young in compari- must be at least some reservoir effect at work in the son with wood and charcoal samples that-on inde- A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 203 pendent archaeological/stratigraphic grounds-repre- the "typical" oceanic reseroir effect of 373 years or sent the same events. more? There are several possibilities. First, recall that It is worth noting that this is not the only case in Stuiver et al.'s "model surface ocean" represents a wa- which matched sample pairs from a Pacific archaeo- ter body some 75 m thick, incorporating upwelling deep logical context have suggested a strongly negative AR ocean waters which are continually mixed with the sur- value. From the To'aga Site on Ofu Island, American face layer. Yet as we know from our own marine bio- Samoa, Kirch (1993:89) reported two samples, one of logical surveys of the Eloaua reefs carried out by C. charcoal (Beta-35603) and one of Tridacna maxima shell Catterall (see Chapter 2), the reef flats, seagrass beds, (Beta-35604), from the same stratigraphic context in and other microhabitats for Tridana, Hyotissa, and other Layer IIIB of Unit 23. Applying the method of Stuiver mollusks utilized for food and raw materials by the and Brazinus (1993) to this matched pair, one derives a Lapita people are typically only - 1-3 m deep, and are AR value of -230. Although not as strongly negative a frequently exposed or nearly exposed at low tide. More- value as those calculated for the Eloaua and Emananus over, the tidal range itself is minimal (- 1 m), and the islands samples, the To'aga samples provide additional wave energy regime unusually low, resulting in a much empirical support for the idea that the "model surface slower rate of hydraulic exchange between open ocean ocean" does not always provide a best approximation and reef/lagoonal water bodies than would be the case for calibrating prehistoric marine shell samples from in a higher-energy environment. It seems likely that these Pacific islands sites. geomorphological and bio-geochemical conditions of How can we explain this apparent deviation from the Eloaua inshore environment could minimize the FIGURE 10.4 OXCAL calibration plot of the same matched pairs of charcoal and associated shell samples from Site ECB shown in Figure 10.3, but with the shell samples calibrated using a AR value of -320. Note that all calibrated ages are now consistent, with a unimodal summed probability. SUM ECB PAIRS -320 ~TT T--T--- r - I IT T- X-r SUM ECB PAIRS -320 Beta-20453 3200?70BP CURVE c:\oxcal\marine93.14c DELTA R -320 ANU-5086 3120?80BP ANU-5087 3150?80BP= SUM ECB PAIRS -320 . I I I I I I j I I | I I I I I -1 . 0 | , 2500BC 2000BC I1500BC I1000BC 500BC AD Calendar date 204 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS overall marine reservoir effect. The shallow water depth crohabitat variation). For example, while the larger and frequent tidal exposure of reefs off Eloaua may bivalves (such as Tridacna and Hyotissa) calibrate best well create a situation in which there is substantial atmo- when a AR value of -320 is used, certain other taxa, spheric isotopic exchange, so that the shallow-water reef particularly Turbo sp., produce best fit results using AR organisms are more nearly in equilibrium with atmo- = 0. Tridacna and Hyotissa mollusks are sessile species spheric '4C levels. Of course, this is simply an hypoth- that inhabit shallow water, low energy environments, esis, but in the face of the paired sample tests discussed whereas Turbo occupies the outer reef margins where above, it offers the best available explanation.7 there is substantial wave action and open-ocean ex- Based on the linked sample-pair tests from ECA change. These different habitat preferences may well Area B and from ECB, I am reasonably satisfied that result in different reservoir effects. Certainly this is an the marine shell samples from ECA and ECB should hypothesis that bears empirical testing be calibrated with a AR value of - 320,8 and the fol- Heavy reliance on marine sample dates for the con- lowing discussion of results is based on such calibra- struction of a cultural chronology for the Mussau Is- tions.9 Doubtless, however, some colleagues will remain lands is admittedly less than ideal, given the lack of real unconvinced that the local reservoir effect in Eloaua is empirical knowledge concerning local reservoir effects significantly less than for the "model surface ocean," and their variability, both in relation to reef and will prefer to take the more conservative approach microecology and geomorphology, and over time. of applying the worldwide model average value. In the Some will argue that attempting any kind of calibration appendix, therefore, I have reported the CALIB cali- of the marine samples is pushing the data too far. My brated ages using the standard marine curve (i.e., with own view is that we must try to make the best hand of AR = 0), while the OXCAL calibration ages have been the cards that the vagaries of the archaeological record run with AR = -320 for Eloaua Island samples, or with have dealt us. Others are welcome to attempt their own AR = 0 for other localities, where the latter value pro- reworking of my calibrations on whatever bases they duces the best fit calibration. prefer;allprimarysampledataareprczided intheappendixfor While the paired sample tests from ECA Area B thispurpose. The only thing that all Pacific archaeologists and ECB-and the derived value AR = -320-bring would probably agree on is that the problem of ma- the marine samples from those sites into line with asso- rine reservoir effects and the calibration of marine shell ciated wood, charcoal, and coconut shell samples, this samples from Oceanic sites are issues urgently requiring is not the case with marine samples from sites elsewhere intensive and fine-grained research. in the Mussau group. In particular, the suite of marine sample dates from the EKQ rockshelter on Mussau ROGUE DATES AND "CHRONOMETRIC HYGIENEJ Island's NW coast display their archaeological "best- fit" calibration (when judged by the criterion of similar A final methodological matter is that of handling what ceramic assemblages) when these are calibrated using a Holdaway and Porch (1996) have called "rogue" dates, value of AR = 0. It may be that what is at play here are which in turn raises the issue of "chronometric hygiene." differences between the kinds of reef environment sur- Most archaeologists have experienced samples which rounding Eloaua-Emananus, and those found along the they initially expected would yield an accurate age esti- coastal fringe of NW Mussau. Unlike the broad, shal- mate for a "target event" (in the sense of Dean 1978), low, low-energy reef flats of Eloaua, the coastline in and then turn out to produce wholly unacceptable re- the vicinity of the EKQ Site has a narrow fringing reef sults, sometimes even wildly off the mark. This is gen- with a high-energy surge regime (in addition, the dated erally because the actual "dated event"-for a variety samples from EKQ were primarily surge-zone gastro- of possible reasons-is in fact completely separate or pods such as Turbo spp.). This situation is likely to be unrelated to the target event. Such rogue dates can re- more representative of a typical mixed "surface ocean" sult from intrusive contamination or post-depositionally reservoir condition. mixed deposits, from the burning of old wood in which Another possible factor in the complex equation the dated event is the growth period of the tree rather affecting local reservoirs may be variation between than the cultural act of burning, or for other reasons molluskan taxa (which itself could correlate with mi- such as laboratory error, improper pretreatment, and A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 205 so on. Spriggs (1989, 1990a, 1996) and Spriggs and or, for that matter, for any Lapita sites in the Bismarcks, Anderson (1993) have attempted to develop an exhaus- and it is not surprising that Specht and Gosden (1997:185 tive protocol for eliminating such potential rogue dates passim) questioned whether it accurately dates the Lapita by applying "chronometric hygiene." With Specht and occupation at ECA. It is not entirely possible to rule Gosden (1997:177), however, I prefer to apply a mini- out an "old wood" factor for ANU-5080, and it is mal degree of chronometric "flossing," focusing on the conceivable that the date might be older than the cul- more obvious problems that arise in cases of contex- tural deposit with which it is associated. On the other tual or stratigraphic security, vertical consistency of hand, we cannot reject it out of hand, for the age is not samples within a stratigraphic sequence, and the ever- wildly inconsistent with additional samples from ECA present possibility of "old wood" dates. Fortunately, Area A, or with samples from Site EHB (which on the Mussau 14C series as a whole has only a small num- ceramic criteria appears to be one of the earliest Lapita ber of such potentially problematic dates, and these are assemblages in the Bismarcks). discussed as appropriate. Two samples of marine shell (Tridacna and Hyotissa valves) were dated from Area A (ANU-5084, -5085), RESU LTS also on the paleobeach terrace, and yielded strongly over- lapping results suggesting rapid deposition. Figure 10.5 CHRONOLOGY OF THE LAP/TA SITES shows a summed probability distribution of these dates TALEPAKEMALAI (ECA) using the marine calibration curve with AR = -320. The summed calibrated range at 16 is 1500-1290 BC. This is Talepakemalai is the most intensively dated of our slightly younger than the date for TP-9, but overlaps Mussau sites, with 30 samples. It is also a large site (in- significantly with it. deed, the largest Lapita site on record), with a complex There are two additional samples from deposits and unusual horizontal stratigraphy owing to its reef- on the paleobeach terrace, both from the S end of the flat stilt-house occupations and geomorphological se- W250 transect (Beta-30676 and -30677), both on shell. quence of coastal progradation. There have been vary- The first, from a large Turbo marmoratus operculum, yields ing attempts to interpret its chronology, both prelimi- an unacceptably old age range when calibrated with AR nary efforts on my part (e.g., Kirch 1987a, 1988b; Kirch = -320 (2060-1750 cal BC), but an age which overlaps and Hunt 1988b; Kirch et al. 1991), using only the samples reasonably with the TP-9 and Area A dates when cali- available from the 1985 and 1986 field seasons, and by brated with a AR = 0 (1660-1410 cal BC). As discussed others who did not have access either to the entire suite earlier, Turbo spp. gastropods are reef-edge, surge-zone of dates, or to the full details of stratigraphic prove- dwellers which might be expected to have a different nience (Specht and Gosden 1997; Spriggs 1990, 1996). reservoir effect from shallow-water reef flat bivalves With the stratigraphic and other contextual details on such as Hyotissa, and for which the mixed open ocean record here (see Chapter 4), and the full suite of dates model calibration curve (with AR = 0) may be quite ap- available in Appendix 10.1, I now attempt a careful and propriate. The second sample, a Spondylus value, yielded reasoned definition of the chronology of this key site. I an age determination quite close to the Area A dates when begin with the three samples from that part of the site calibrated with AR = -320 (1500-1320 cal BC). which was always supratidal (i.e., the + 2 m paleobeach Taking all five samples from the paleobeach terrace terrace on which the modern airfield lies), turning then into consideration, we obtain the summed probability to Area B and the W200 transect, then to the W250 distribution shown in Figure 10.6, in which the summed transect series, and finally to Area C. age range is 1580-1310 cal BC at 1a. Admittedly, this result depends upon the differential marine calibration The Paleobeach Terrace and Area A of the Turbo and bivalve samples as noted above; some A single wood charcoal sample (ANU-5080) was ob- may find this an unacceptable manipulation of the cali- tamned from the base of the Lapita cultural deposit in bration curves. Nonetheless, and without overly push- TP-9 (Unit W400N72), which yielded a high probabil- ing the point, a case might be made for the paleobeach ity (0.97) calibrated age range at 1Gs of 1660-1440 BC. terrace deposits at ECA to be slightly earlier (by 100- Admittedly, this is one of the oldest 14C ages for ECA 150 years) than those on the adjacent reef flat (i.e., the 206 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS SUM ECA AREA A ''' 'T~- - 1n 1' r-r ' - F T m 7-"T- ^ 1-'" ''''-- T SUM ECA AREA A CURVE c:\oxcal\marine93.14c DELTA R-320 ANU-5084 3190I80BP -e----- -4--- - - + 4 4-4-+~- I - --X- ________---_ ANU-5085 3130Ot8OBP SUM ECA AREA A 2500BC 2000BC 1500BC 100OBC 500BC AD Calendar date FIGURE 10.5 OXCAL calibration plot for two shell samples from Area A of Site ECA, with summed probability distribution. Area B and W250 transect stilt-house deposits). If so, calibrated age range is 1300-1050 BC at 1Gy. this has important culture-historical implications, because For Area B we have a substantial suite of dates, the ceramics from the paleobeach terrace deposits con- including the paired marine-terrestrial set already dis- sist almost exclusively of red-slipped plainwares, with cussed above under "Calibration," as well as other shell only small numbers of dentate-stamped pottery. More- and charcoal dates, totaling 10 samples from Area B over, the single dentate-stamped sherd from TP-9 is a proper and another two samples from the 1988 Area B particularly delicate and finely-executed example with extension. For reasons made clear in Chapter 4, the in- close parallels from EHB, whose assemblage I regard ternal stratigraphic context in Area B must be taken into as likely to be the earliest in the Mussau Lapita sequence. account when interpreting this suite of dates, which rep- These culture-historical implications will be more fully resent both the stilt house itself (samples ANU-5790, - explored in Volume III, when the details of the ceramic 5791, Beta-20452), shell samples stratigraphically-asso- assemblages are presented. ciated with the posts in Zone C3 (samples ANU-508 1, - 5082), and samples from stratigraphically higher Zones Area B C2 (ANU-5078), Cl (ANU-5075, -5076, 5077, and - Moving from the paleobeach terrace toward Area B, a 5079; Beta-30673), and B1 (ANU-5083). This complex sample of anaerobically-preserved coconut endocarp stratigraphy has not been fully appreciated by others (Beta-20451) was dated from the fine "muck zone" at who have attempted to interpret the published dates - 130 cm bs in Unit W200N120. This specimen is in a from ECA (e.g., Specht and Gosden 1997), being un- context equivalent tO Zone C in Area B (although which aware that Zone C is not a single component. subzone of Zone C cannot be easily specified), but it On strictly archaeological and stratigraphic criteria, was situated 20 cm above the base of the cultural the earliest cultural activity at Area B was the construc- materials in this unit, and so should provide a less-than- tion of the stilt house represented by the tWO align- maximal date for cultural deposition in this area. The ments of posts described in Chapter 4 (see Fig. 4.18). A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 207 Figure 10.7 shows a summary plot of the calibrated field season, to obtaindatablecharcoal) whichmusthave ages for the three dated posts. The overall agreement beendqeosited intheprogradingforeshore,almostcertainlyafter between the samples is good, although the probability thestilt-houseitselfhad been abandoned. Unfortunately, all of distribution has a fairly wide temporal spread, from cal these samples have quite large associated error factors, 1390-1090 BC at 10.10 A combined plot for ANU-5790 so that when they are calibrated as in Figure 10.8, they and -5791 along with the two marine shell samples from yield long, attenuated probability distributions, spanning stratigraphically-associated Zone C3 was already pre- the entire first half of the first millennium BC. The sented in Figure 10.2, and has roughly the same tempo- summed probability distribution for these five charcoal ral distribution. samples is cal 1000-350 BC at 1o, and 1300-100 BC at Five charcoal samples were collected from Zones 2cy. While these samples are admittedly younger than C) and C1. It cannot be overly stressed that these con- those for the stilt-house, and in correct stratigraphic re- sisted of very finely dispersed charcoal flecks (collected lationship, they are less than ideal for tightly fixing an painstakingly out of near desperation, during the 1985 age span for the upper part of Zone C. One additional FIGURE 10.6 OXCAL calibration plot of five samples from the paleobeach terrace of Site ECA, with summed probability distribution. SUM ECA BEACH TERRACE .....- T r ----T" - -T--- - - T ----T-- --T-- - T SUM ECA BEACH TERRACE ANU-5080 3260?90BP l CURVE c:\oxcal\marine93.14c -~~~~~~~~ ~ - I - -t------t 4--+ --i-t---t-4- - --+-- - -- -- e------ --- - ---- Beta-30676 3590?110BP CURVE c:\oxcal\rmarine93.14c DELTA R-320 __ t.4 Ft- +s+t-.--t .--. -~ - - t- .-+- - -.-..-. - .---- - -- ANU-5084 3190?80BP ANU-5085 3130?80BP _ Beta-30677 3170?70BP v--- -+--- + -+-----4 t -+-+----- - - --- -t--- --+---!-_+ t SUM ECA BEACH TERRACE * , a . . , . I | . l I I ,____ ,______ 3000BC 2500BC 2000BC I1500BC I1000BC 500BC Calendar date 208 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS SUM AREA B POSTS - ..T -r . -T r-r----r- -r rv - SUM AREA B POSTS , , , I , I , ,~~~~I ,, .1-- 4 -, Beta-20452 3050?70BP j, l ANU-5790 2950?80BP _ ANU-5791 2930?80BP SUM AREA B POSTS I l I 'III I I I i I I I I : I 1 , I 1 2500BC 2000BC 1500BC 1 OOOBC 500BC AD Calendar date FIGURE 10.7 OXCAL calibration plot of three samples from wooden posts of Area B at Site ECA, with summed probability distribution. sample from Zone C1 is Beta-30673, a specimen of samples from Area C with a similar ceramic assem- Spondylus shell. Applying a AR = -320 correction to the blage, dominated by incised decoration. marine calibration curve for this sample yields an age The entire Area B suite of samples can be summa- range of cal 1420-1250 BC at la, or 1500-1150 BC at 16. rized as a combined calibration plot depicted in Figure Note that this range overlaps with the extended prob- 10.9, and as a set of boxed age spans plotted against ability distribution of the dispersed charcoal dates, and the atmospheric calibration curve in Figure 10.10. While while one would want to be cautious about placing too the broad temporal spans of the Zone C21 charcoal much weight on a single marine sample, it might sug- samples make these samples inadequate for precise dat- gest that the deposition of Zone C1 actually followed ing, the sequence as a whole holds together, and can be fairly rapidly after the construction, use, and abandon- reasonably interpreted. The first phase was marked by ment of the Area B stilt-house. the construction of the stilt-house, probably not much The final sample to be considered from Area B is earlier than - 1300 cal BC but also not likely to be later ANU-5083, a Hyotissa shell from Zone B1, the strati- than - 1200 cal BC. This was followed by the abandon- graphic zone in which incised motifs and notched-rim ment and, presumably, disintegration of the structure, globular vessels dominate the ceramic assemblage, hav- and by continued deposition of cultural materials in the ing largely replaced the dentate-stamped ceramics of Area B locus, marked by a shift from a dominance of Zone C. Calibrating this sample with the marine curve dentate-stamped ceramics to incised ceramics in Zone and the preferred AR = -320 value, we obtain a range B1. This upper part of the Area B sequence probably of cal 1040-830 BC at k6, or 1160-790 BC at 16. Given dates to 1000-900 cal BC, although on the basis of the the stratigraphic position of this sample, as well as the charcoal dates it could extend as late as 700 cal BC. In nature of the associated ceramic assemblage, this age short, the entire depositional sequence at Area B most seems reasonable. Indeed, it is consistent with several likely spans a period of 200-400 years duration, a not A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 209 SUM AREA B ZONE C2-1 SUM AREA B ZONE C2-1 I ANU-5075 2370012OBP ANU-5076 2430*230BP _ ANU-5077 2450I160BP ANU-5078 2600?I160BP I ANU-5079 2840?115BP SUM AREA B ZONE C2-1 I I I .. . I 11 . .1 . 2500BC 2000BC 1500BC 1000BC 500BC AD Calendar date FIGURE 10.8 OXCAL calibration plot of five charcoal samples from Zone C of Area B, Site ECA, with summed probability distribution. unreasonable estimate given the degree of ceramic an overall horizontal (and temporal) progression begin- change indicated in the sequence. ning with fine-dentate stamped pottery, changing to coarse-dentate stamped, and ending with incised ceram- The W250 Transect ics (especially in Area C at the N end of the transect). From the W250 transect we dated three wood samples This parallels the vertically expressed ceramic sequence (Beta-30681, -30682, and -30684), two from posts and within Area B, and we should thus expect that the W250 one from a smaller stake, and six samples of marine suite of dates should exhibit a similar age range. shell. Because these samples come from a number of 1 A summary plot of all nine samples from the W250 2 m excavation units spaced out over a distance of 130 transect is shown in Figure 10.11. The three wood m from N to S, precise stratigraphic correlations are samples suggest that construction of stilt houses along difficult. The specific contexts from which these samples the transect spanned a period of - 300-400 years, from were obtained have been discussed in Chapter 4, and possibly as early as 1400-1300 cal BC until - 1000 cal BC. their locations are plotted in Figure 4.31. For our pur- This is completely consistent with the age span for Area poses, it will be most useful to examine this suite of B. When we turn to the shell dates, however, it is obvi- nine samples as a group, to assess how the overall tem- ous that Beta-30676 (a Turbo shell) is either a "rogue," poral range corresponds to that derived from Area B. or that for some reason a A\R value of -320 is not ap- The ceramic sequence from the W250 transect shows propriate. Indeed, when this sample was examined as 210 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS SITE ECA AREA B, DUAL CURVES, DELTA-R -320 -T-'- F ' - -T -T - r ---nr- -----l- n>--r- Post Beta-20452 3050?70BP L J Post ANU-5790 2950?80BP LI i II Post ANU-5791 2930?80BP CURVE c:\oxcal\marine93.14c DELTA R-320 - Zone C3 ANU-5081 3010?80BPLI,Ii1 -4 + --------t - ,-.-- I---,---,-t-+---+---t--4- ------4--- -t- - -*- - ------------4- 4 - Zone C3 ANU-5082 2950?80BP r;J -$ _ 4 t + t t r - e - --4 ~ - .4 ' - - CURVE c:\oxcal\cal2O.dta Zone C3-C2 ANU-5078 2600?I160BP L , IIVIIii] Zone Cl ANU-5079 2840?115BP [Z7f1] Zone Cl ANU-5075 2370?120BP I ~ ~~~~~~~ I Zone Cl ANU-5076 2430?230BP 11111111111] Zone Cl ANU-5077 2450?I160BP CURVE c:\oxcal\manne93.14c DELTA R-320 - t - ZoneCl Beta-30673 3110?70BP711] Zone BI ANU-5083 2810?80BP LI=II ,:ngr . i ,n. . I . g I In g A 2500BC 2000BC I 500BC I OOOBC 500BC AD Calibrated date FIGURE 10.9 OXCAL summary plot of calibrated ages for 1 2 radiocarbon samples from Area B of Site ECA. A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 211 4000BP ITE ECA AREA B, DUAL CURVES, DELTA-R -320 c3500BP --'-- - - - ...iAI,... .. . l 0 X .2 oneB 2 I 5 C I Bet-30673 500BC A FIGURE 10.10OXCALsmmlt o f12 r abon s e fm Aa BANU of St E plte agis th deaa raioaro cairto cuve Not tha due to th "fatnn"o < _ 1 . . _ _ ^ = f f Cr U-SI ANU-M7 3000BP --- --- _ |e lengthened for samples in this agetrange -~ ~ ~ ~~~~ ~C l AL UI 0 7 4i + +t f 8 \~~IR 2000BP , , , , I , , , , I , , , , I , , , I , l , 2500BC 2000BC 1500BC IOOOBC 500BC AD Calibrated date FIGURE 1 0.1 0 OXCAL summary box plot of 1 2 radiocarbon samples from Area B of Site ECA, plotted against the decadal radiocarbon calibration curve. Note that due to the "4flattening" of the calibration curve between -800-400 BC the error ranges in calibrated ages become lengthened for samples in this age range. part of the suite of dates from the paleobeach terrace, plications of this possibility will be more fully treated in it was found to correlate acceptably with a AR = 0 Volume III. value. Beta-30680, when calibrated with AR = -320, is also slightly older than anticipated, although at 26 it Area C overlaps with the earliest wood sample. It is notewor- Finally, there is the suite of five samples from Area C, thy that both of these older shell dates come from the S at the N end of the W250 transect, with deposits domi- end of the transect, on the outer fringe of the + 2 m nated by incised ceramics and marked by a concentra- paleobeach terrace (also the location of the early date tion of wooden posts indicative of one or more stilt from TP-9, ANU-5080). The remaining four shell dates houses (see Chapter 4). The stratigraphic evidence at all fall within the range from - 1500-1200 cal BC, over- Area C suggested that there had been two discrete tem- lapping reasonably well with the wood sample dates. poral phases of occupation and deposition of artifac- Taken as a suite, the W250 samples reinforce the tual materials, although there need not have been a long interpretation of Lapita occupation at ECA through- hiatus between these. The dated samples from Area C out the second half of the first millennium BC. Although include two wooden posts (Beta-30686 and -30687) and the two earliest shell dates could be rogues, they might three marine shells (Beta-30674, -30675, and -30685). also lend support tO the notion that initial occupation at The calibrated age ranges for the Area C samples ECA commenced on the + 2 m paleobeach terrace, are plotted in Figure 10.12. The two wooden post where they are associated with almost exclusively red- samples yield age ranges entirely consistent with expec- slipped plainware ceramics. The culture-historical im- tations from the upper part of the Area B sequence 212 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS SITE ECA W250 TRANSECT, DUAL CURVES, DELTA R -320 T TT T T F T J T --r---1-r--a - -- T - r T FT WOOD SAMPLES Beta-30684 31 00* 1 OBP l B eta-30682 2970?50BP B eta-30681 2860?60BP p MIDDEN SHELL SAMPLES. CURVE c:\oxcatl\marine93.14c DELTA R-320 1 - t Beta-30676 3590?i110P| Beta-30680 3320?801BPL Beta-30678 3190?80BP -- Beta-30677 3170?70BP L1 I -t ,4 0 t j t i 1-. Beta-30683 3140?80BP Ii-I] Beta-30679 3080?70BP [ 2500BC 2000BC 1500BC 100OBC 500BC AD Calibrated date FIGURE 10.11 OXCAL summary plot of calibrated ages for nine radiocarbon samples from the W250 transect of Site ECA. Shell samples were calibrated with a AR value of -320. (i.e., Zone Bl). Sample Beta-30686 has an age range of calibrated with a A\R value of -320. The other two shell 1130-9 10 cal BC while Beta-30687 has a highest prob- samples, however, have calibrated age ranges that are ability (0.76) distribution centered between 850-760 cal slightly older than anticipated, although they are still BC. Shell sample Beta-30685 also accords well with the within the overall range for the Lapita occupation at wooden samples, with a range of 970-810 cal BC when ECA. It is possible that they represent midden shell A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 213 which had been discarded in the Area C vicinity prior by as much as 100-150 years; (3) the earlier phase of stilt- to the construction of the later stilt-houses associated house occupation associated with dentate stamped ceram- with the incised ceramics. Discounting these two slightly ics can be assigned to a period from - 1300-1100 cal BC; older shell samples, a reasonable overall estimate for and (4) the later phase marked by a shift to coarse- the temporal span at Area C (and for its incised ceramic stamped dentate and then nearly exclusively incised ce- assemblage) would be - 1100-800 cal BC. ramics took place from - 1100-900/800 cal BC. In ag- gregate, Talepakemalai seems to have been settled con- ECA Summary tinuously for at least 400-500 years, and possibly as long To briefly encapsulate a rather lengthy and complex dis- as 600 years. There is no evidence in the areas of the site cussion, we may reasonably conclude the following: (1) we sampled for occupation after - 800 cal BC. there is strong evidence for Lapita occupation at ECA commencing no later than - 1350 cal BC, with the dis- ETAKOSARAI (ECB) tinct possibility of occupation having commenced as early as - 1500 cal BC; (2) the earliest use of the site may Three samples (two shell, ANU-5086 and -5087; one have been on the + 2 m elevated paleobeach terrace, and charcoal, Beta-20453) were dated from this small, single could have pre-dated the Area B stilt-house construction component Lapita site. The samples were discussed FIGURE 10.12 OXCAL calibration plot of five radiocarbon samples from Area C of Site ECA. Shell samples were calibrated with a AR value of -320. SITE ECA AREA C, DUAL CURVES, DELTA R -320 WOODEN POSTS Beta-30686 2850?70BP Beta-30687 2600?60BP l MIDDEN SHELL CURVE c:\oxcal\marine93.14c DELTA R-320 Beta-30674 3110?70BP ! Beta-30675 3110?80BP p I I I I 1 I I I I I a a t l1 l l l 1 2500BC 2000BC I1500BC I1OOOBC 500BC AD Calibrated date 214 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS above (under "Calibration") because they provide one Figure 10. 13B, with a summed age range of 1460-1200 of the key sets of matched shell/charcoal pairs that cal BC at 1 6. allows us to derive a potential local (Eloaua Island) AR I am not entirely comfortable with either set of correction factor of -320 for the marine calibration calibrated ages. The values resulting from the applica- curve. A plot of the calibrated probability distributions tion of a AR = -320 seem to produce results that are for the three samples was given in Figure 10.4. As ar- slightly too old when compared with the suite of ages gued earlier, all three samples can be expected to pro- from the earliest parts of the ECA Site. At the same vide good estimates for the age of this site, and should time, given the nature of the EHB ceramic assemblage, form a single, unimodal distribution, as there is no ar- the calibrated ranges derived from using AR = 0 are in chaeological or stratigraphic evidence to suggest other my view definitely too young. My hunch is that the true than a single, relatively short-duration occupation at this age of EHB is - 1500 cal BC, a value that lies approxi- site. The charcoal sample could have a slight in-built age mately mid-way between the two sets of calibrations factor if it represented old wood, but this is unlikely to shown in Figure 10.13. On the available evidence, I do have been on the order of more than - 100 years.11 The not think it possible to resolve this issue further; again, summed, calibrated age range for ECB based on all the necessity of further research on marine reservoir three samples is 1520-1300 cal BC at 16, and 1630-1170 effects becomes patently obvious.13 cal BC at 2a. Based on the ceramic assemblage at ECB, which includes significant quantities of fine-dentate EPAKAPAKA (EKQ) AND OTHER ROCKSHELTERS stamped pottery, the occupation of this site should be expected to correlate closely with the Zone C, phase at As discussed in Chapter 5, the EKQ rockshelter con- Area B of ECA. The summed age range of 1520-1300 tains 2.6 m of stratified deposits, of which the lower cal BC is therefore a reasonably good fit. 2.2 m includes a late Lapita component dominated by incised ceramics in many respects similar to those from ETAPAKENGAROASA (EHB) Area C at the ECA Site, while the upper 0.4 m is aceramic. Four samples were dated from excavation Unit 2 at Like ECB, the EHB Site is small and was presumably EKQ, and one sample from Unit 1; all are of marine occupied for a relatively short period of time. The ce- shell. Four of these samples (Beta-20454, -21789, -25670, ramic assemblage is dominated by sherds with particu- and -25671) are associated with the incised ceramics, larly fine dentate-stamping, and by a large number of while one (Beta-25036) is from the upper aceramic de- pedestaled bowls, both traits strongly suggestive of an posit. early phase in the Mussau Lapita ceramic sequence. Un- The four samples from the lower cultural deposits fortunately, there was no anaerobically preserved wood yielded reasonably consistent conventional 14C ages of at EHB nor were we able to obtain charcoal samples. (in order from oldest to youngest) 3280 + 70 3270 + Two samples of marine shell were therefore dated, 80,3190 + 90, and 3030 + 80 BP, although they are not yielding reasonably consistent uncalibrated (conventional) in totally correct stratigraphic order. While there might 14C ages of 3470 + 90 and 3380 + 90 BP."2 have been some mixing or movement of deposit dur- Assigning a calibrated age range to Site EHB, of ing the course of occupation in this intensively utilized course, depends entirely on the AR value applied to the rockshelter, or some post-depositional disturbance (pos- marine calibration curve. If one accepts the calculation sibly by burrowing land crabs), the dates are close enough of a local AR value of -320 based on the two sets of that the most probable explanation is simply a case of matched marine/terrestrial samples from Eloaua Island very rapid deposition, as they all overlap at 26. In any sites ECA and ECB as argued earlier, the EHB shell case, there is close enough agreement in the four sample dates are calibrated as shown in Figure 10. 13A, and have ages that we are justified in combining them. a summed age range of 1860-1590 cal BC at l6. If one As with Site EHB, however, the question arises of prefers to take the more conservative (but probably what AvRvalue to apply to the marine calibration curve. inaccurate) approach of using a value of AiR = 0, then Here the local littoral-reef conditions adjacent to the the calibrated probability distributions are as shown in EKQ Site are relevant. Unlike the situation at Eloaua A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 215 SUM EHB, DELTA R= -320 SUM EHB, DELTA R= -3201 CURVE C:\oxcal\marine93. 14c DELTA R-320 i t ANU-5088 3470?90BP ANU-5089 3380?90BP e , SUM EHB, DELTA R- 3 I I I I i I I I I 3000BC 2500BC 2000BC 1500BC 100OBC 500BC Calendar date SUM EHB, DELTA R=0 T _ w ..T_ I T T- SUM EHB, DELTA R=O CURVE C:\oxcal\marine93.14c ANU-5088 3470?90BP ANU-5089 3380?9OBP SUM EHB, DELTA R=0 3000BC 2500BC 2000BC 1500BC 100OBC 500BC Calendar date FIGURE 10.13 OXCAL calibration plots for two shell samples from Site EHB. (A) Calibrations and summary probabilities using a AR value of -320. (B) Calibrations and summary probabilities using a AR value of 0. 216 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS Island, where the reef flats and seagrass beds are broad LAPITA COMPONENT AT BOLIU ISLAND (EKE) and shallow ( - 1 m depth), and the wave energy regime is very low, the coastline fronting the EKQ Site has only Because our main objectives in excavating at EKE were a narrow fringing reef (- 100 m wide) with high-en- to sample the post-Lapita phase middens, only a single ergy wave and surge action. The submarine topogra- sample (Beta-30693) associated with the older, Lapita- phy plunges steeply down immediately outside of the phase deposits at EKE was dated. This came from reef edge, conditions in which upwelling and mixing Layer II of Unit E200N175, in association with red- of deep ocean water over the reef edge can be antici- slipped, calcareous-sand tempered plainware, and yielded pated. Thus it might be expected that mollusks grown a conventional 14C age of 3420 + 70 BP. Depending on the fringing reef flat adjacent to EKQ would reflect upon the AR value applied to the marine calibration a more typical "surface ocean" reservoir effect, in which curve, one obtains age ranges of 1410-1240 cal BC (AR case we would be best advised to apply a AR value of = 0), or 1790-1600 cal BC (AR = -320) at 16. The sec- 0, rather than the -320 value derived from our matched ond of these calibrated age ranges appears to be too marine/terrestrial pairs from ECA and ECB on Eloaua old, although it is not inconsistent with the shell dates Island. Moreover, three of the EKQ samples include from the paleobeach terrace at ECA, where the depos- Turbo spp. shells which, as was seen in the case of ECA, its are similarly characterized by a dominance of red- seem to yield their best fit calibrations with a value of slipped plainware ceramics. However, a younger age AR = 0. range of 1400-1200 cal BC is also entirely reasonable. This supposition is borne out by calculating the summed probability distributions for the four EKQ CHRONOLOGY OF THE Po0 T-LAPI TA SITES shell samples from the lower cultural deposit with the two different AR values. Using AR = -320 produces a ELUNGUAI (EHK) summed probability distribution that is patently too old (1630-1320 cal BC at P3) in comparison with dates for Two sets of A nadara antiquata bivalves, both with arti- the highly similar incised ceramic assemblage at Area C ficially-chipped surfaces, were submitted for "4C dating of the ECA Site. If, however, we use AR = 0, on the from the EHK midden on Eloaua Island. Since these assumption that the reservoir effect on the Pomanai area mollusks inhabit seagrass beds or shallow-water sandy fringing reef was more typical of the "model surface flats, they are likely to have minimal reservoir effects as ocean," we get the probability distributions shown in indicated by our matched terrestrial/marine sample pairs Figure 10.14, which has summed age ranges of 1230- from ECA and ECB on Eloaua Island, and I have there- 930 (prob. 0.91) and 880-830 (prob. 0.09) cal BC at 1c6. fore calibrated them using AR = -320. The sample from These overlap with the age range for Area C at ECA, the earlier, Layer II occupation (Beta-30689) yielded an 1100-800 cal BC. The occupation of the lower com- age range of cal AD 460-620 at 16. This was associated ponent at the Epakapaka rockshelter most likely oc- with a Terebra-shell adz, the earliest dated instance of curred during the period - 1200-800 cal BC. this distinctive artifact type in the Mussau sequence. The The only other dated rockshelter is Site EKO on Layer I occupation deposit dates to cal AD 940-1070 Eloaua Island, from which a single sample of Turbo (Beta-30688), although there is no particular reason tO marmoratus shell yielded a conventional age of 3200 + believe that human use of the EHK Site was not con- 70 BP. Since Turbo shell specimens from other contexts tinuous from the mid-first millennium AD until the be- yield their best fit calibrations with AR = 0, applying ginning of the second millennium AD. The Layer I de- this value to the marine calibration curve gives a result posit contains both Terebra- and Tridacna-shell adzes, and of 1130-930 cal BC at to. This would place the use of Anadara net sinkers. this small rockshelter within the same time span as the late phase occupation at Site ECA (e.g., the Area C de- BOLIU ISLAND (EKE) posits), a temporal assignment perfectly consistent with the small assemblage of largely plain, shell-tempered The post-Lapita component at the extensive EKE pottery from EKO. midden site on Boliu Island was dated with three shell A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 217 SUM EKQ CERAMIC PHASE T m r T r-- s ~X_ TF`-v r- t -T - T_ SUM EKQ CERAMIC PHASE CURVE c:\oxcal\marine93.14c BETA-20454 3280i70BP 1 -. - --t----+ -,e- -t ..-..-. +- --X ?"4--_ BETA-25670 3270?80BP BETA-25671 3190?90BP BETA-21789 3030?80BP I ~ ~ ~ ~~ ~~ .- _-4 - - . - I ---4 SUM EKQ CERAMIC PHASE I~~~~~~~~~~~~~~-- L . 2500BC 2000BC 1500BC 100OBC 500BC AD Calendar date FIGURE 10.14 OXCAL calibration plot for four radiocarbon samples from the ceramic phase component of Site EKQ, and summary probability distribution. samples (Beta-30690, -30691, and -30692) from Layers Gibb's sampler of the OXCAL program, with 29,496 I and IC of Unit E200N175, and with one sample (Beta- iterations used). Based on these calibrated ages, we can 30694) from Unit E200N200. In the case of the EKE reasonably suggest that the post-Lapita midden-dump- Site samples, there are no clear grounds for deciding ing phase at EKE spanned a period from cal AD - 900- whether a AR value of 0 or of -320 should be applied, 1500. although it might be argued that the latter is preferable given that three of the samples are of Tridacna spp. EMUSSAU ISLAND (EKS) shell, and these taxa are represented in the Eloaua Island matched terrestrial/marine pairs that yielded the -320 The aceramic EKS midden site on Emussau Island was AR value (see "Calibration"). Calibrated age ranges us- dated with a single sample of wood charcoal obtained ing both sets of AR values are given in the appendix. from the midden-dump mound trenched by Unit 2. Figure 10.15 displays a sequence plot of the four The sample came from Layer IIC, just above the sterile post-Lapita phase 14C age determinations from EKE, beach sand (Layer IIIA) at the base of the cultural de- with the three samples from Unit E200N 175 in their posit (see Chapter 7). The calibrated probability plot correct stratigraphic order, and that from E200N200 associated with this sample has a bimodal distribution, as the most recent sample. The overall agreement of and it is probably best in this case to use the 26 range as the sequence is very high (A = 91.0%, based on the a more accurate estimate of the probable true age: cal 218 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS SEQ EKE E200N176 SEQ EKE E200N175 {A= 91 .0%(A'c= 60.0%)) -. - -~~ v- -.-.----I----------- ---1 Beta-30692 100.2 I_I Beta-30691 82.5% ___ Beta-30690 99.9 %l Beta-30694 100.2%? 600AD 800AD 1000AD 1200AD 1400AD 1600AD Calibrated date FIGURE 10.15 OXCAL sequence plot of four radiocarbon samples from Site EKE, in stratigraphic order. AD 1440-1650. This confirms the late age of the site as adzes and non-Lapita dark red sherds, such as EKE judged by the site's artifact content and by oral tradi- and EHK. tions of the Eloaua Island people that this was a known habitation locus. Essentially, then, the site dates to the POST-LAPITA COMPONENT AT EPAKAPAKA "proto-historic" period. ROCKSHELTER (EKQ) SINAKASAE (EKU) The Layer II, post-Lapita aceramic cultural deposit in the EKQ rockshelter was dated with a single sample of As related in Chapter 9, the EKU Site on Mussau Island Turbo shell (Beta-25036). As with other shell samples was test excavated in 1986 because surface collections from this site, I have applied a AR value of 0. This there revealed the associated presence of Terebra-shell yields an age range of cal AD 1510-1650 at 16, suggest- adzes with a few sherds of non-Lapita, dark red py- ing that the upper part of the site is quite late, roxene-tempered pottery. To obtain a 14C age estimate penecontemporaneous with the EKS midden site on for the thin, single-component cultural deposit at EKU Emussau Island. This also implies a long hiatus in the we submitted a sample of 9lg of mammal bone frag- use of the EKQ Site between the Lapita phase and the ments, almost certainly representing pig bones, as iden- aceramic phase. tifiable S. scrofa teeth and bones were plentiful in the deposit. The extracted collagen yielded a conventional ENUSAGILA ISLAND (EKL) age of 740 + 70 BP, with an associated age range of cal AD 1200-1300 (0.91 probability) at 16, orof cal AD As related in Chapter 9,an eroding midden on Enusagila 1120-1400 at 26. This calibrated age range is consistent Islet was tested in 1988, and a single sample of Spondylus with those from other sites also containing Terebra-shell shell was submitted for 14C dating (Beta-30695). The A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 219 conventional 14C age for this sample is 330 + 60 BP. In armrings, non-Lapita pyroxenic-tempered sherds, and this case the OXCAL calibration yields same results abundant pig bone. whether we apply AR = 0 or -320, of cal AD 1880- What is most striking in Figure 10.16 is the blank 1945, a result consistent with the discovery of a frag- middle section of the graph, the unfilled space between ment of bottle glass in the midden, indicative of a post- - 800 cal BC and cal AD - 500, representing a period of contact age. perhaps 13 centuries for which we simply have no evi- dence at present. In Volume Ill, the implications of this CONCLUSIONS gap will be further explored and discussed. For the moment, I will simply remark that this gap is not the An overall chronology for the Mussau archaeological obvious result of a sampling bias on our part, for we sequence as currently defined is summarized in Figure made real efforts to define the post-Lapita components 10.16. The Lapita period in the Mussau Islands spans a of the Mussau sequence. Rather, it may reflect a real period of at least 500 years, and possibly as long as pattern in the archaeological record of the offshore is- 700-800 years; the uncertainties are at both ends of the lands, which may have witnessed a long period of rela- Lapita period. Thus, while the dated wooden posts from tive disuse. This should not be taken to indicate the ab- Area B and from the W250 transect leave no doubt that sence of a continuous archaeological sequence in the Lapita settlements had been emplaced on Eloaua by Mussau Islands. Rather, this record tells that future re- - 1400-1300 cal BC, several shell and two charcoal dates search in Mussau will need to target the large main is- from the ECA paleobeach terrace hint at settlement one land more intensively, for it certainly must be there that or even two centuries earlier (i.e., - 1600 cal BC at the the missing middle segment will be revealed. earliest). The early dates from the ECA paleobeach ter- The 51 14C age determinations from Mussau con- race are potentially reinforced by the suite of dates from stitute one of the largest suites of dates from any single EHB and ECB, depending of course on the preferred locality in the Bismarck Archipelago, with multiple de- AR value used in calibration. What makes this possibility terminations on six different Lapita sites, and for six of an early set of dates more compelling is their asso- post-Lapita assemblages. Along with other suites of dates ciation with ceramic assemblages that are distinctive from such localities as the Arawe Islands, Kandrian and from those associated with the Area B stilt-house at Talasea regions of New Britain, Watom Island, the Duke ECA. On the ECA paleobeach terrace, these are largely of Yorks, and Nissan Island, the empirical basis for a red-slipped plainwares, while at EHB they constitute a radiocarbon chronology of the Lapita period in the distinctivey fine-dentate stamped assemblage with a high Bismarcks has improved significantly since I reviewed percentage of pedestalled bowls. the situation a decade ago (Kirch and Hunt 1988a). At the late end of the Lapita sequence, we are also Nonetheless, as a recent review of the evidence sug- confronted by some uncertainties regarding the precise gests (Specht and Gosden 1997), this expanded chro- timing of the transition from coarse-dentate stamping nometric database has raised as many new questions as (associated with flat-bottomed, flaring-side dishes) to it has answered. largely incised decoration. The best estimate for the Area While the time frame for Lapita in the Bismarcks is C and the EKQ assemblages, which are dominated by now more closely resolved than it was a decade ago incised ceramics, is that they date to - 800 cal BC, but the (we can, for example, clearly eliminate as a "rogue" the error ranges do not completely rule out a slightly later 3900 BP date [GX-5499] from ECA obtained by Egloff, date. we have now reached the point where our models re- For the post-Lapita sites, our six assemblages pro- quire chronometric resolution on the order of a single vide a largely continuous sequence from cal AD - 500 century or less. Yet as this chapter has made painfully up through the proto-historic and even post-contact evident, we still face intractable problems in the calibra- periods. The sample sizes are perhaps smaller than we tion and correlation of 14C dates derived from different would have liked, but there is considerable consistency kinds of sample materials, especially shell. No less than in the material culture assemblages throughout this 1500 two-thirds of 61 dates from Bismarck localities other year long time span, including such distinctive traits as than Mussau listed by Specht and Gosden (1997, ap- Terebra-shell and Tridacna-shell adzes, Trochus-shell pendix 3) are on marine shell, as are 65% of the samples 220 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS A ?I j Paleobeach Terrace/Area A ? . z Area B Stilt House ECA Area B W250 1 __e__ Area C EHB:? ECB LI1 EKO EKQ EKQ ?7 EKE Lapita EKE Upper Middens EHKI |[ L |- ZZ| EKU gL EKS E Hiatus/gap in sequence EKL LI] 2000 B.C. 1500 1000 500 0 500 1000 1500 A.D. 2000 FIGURE 10.16 Inferred ages of Mussau excavated sites and site components. we dated from Mussau. We have some evidence that whereas the current unresolvable variation in the marine marine reservoir effects were not the same throughout samples that make up two-thirds of our Lapita chro- the Bismarcks, and indeed the Mussau evidence strongly nometric database is on the order of + 200 years. Until suggests that there were significant local variations even these thorny problems of precise calibration and corre- within a single geographic group of sites, as well as over lation between marine and terrestrial samples are re- time. In practical terms, the situation we are confronted solved, we will continue to confront the conundrum with is this: we would like to be able to calibrate our that our models have outpaced our technical capacity to chronologies to a resolution of perhaps + 50 years, test them. NOTES TO CHAPTER I 0 1 Spriggs (1990, 1996) provided updated versions of the Kirch 2 By "disputed chronology," IJam referring to the various posi- and Hunt Lapita date list, in which he incorporates various tions on the timing of initial Lapita assemblages, as stated by plainware ceramnic sites lacking classic dentate-stamped pottery. Anson (1983, 1986), Allen (1984), Kirch et al. (1987) and oth- A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 221 ers. Much of the dispute at the time of the 1985 Lapita Home- information sufficient to indicate whether any of these samples land Project revolved around the validity of the early GX-5499 are matched pairs, in which case they might provide the addi- date obtained by Egloff from ECA (see Chapter 1). Fortu- tional independent tests of AR values for the Bismarcks re- nately-and as a direct result of our extensive dating program gion. Unfortunately, for virtually all of the other Bismarck sites at ECA-we can now definitively dismiss the Egloff date as listed by Specht and Gosden, the available samples seem to be erroneous. exclusively of charcoal or of shell, precluding the possibility of such linked sample-pair tests. 'While it might have been advantageous to select only a single taxon of marine shell for dating at all sites, this was not pos- 8 I fully appreciate that the matched pairs did not produce sible as good specimens of midden shell or industrial detritus completely consistent results, and that the value of -320 is only from a particular context which we wished to date did not an approximation. As will be clear below, however, this value always include a preferred species. Whenever possible, we chose yields calibrations that are a far better "fit" with the archaeologi- the large bivalves Tridacnaispp, Hippopus, or Hyotissa. The first cal evidence than using either the standard "model surface ocean" two genera are both in the family Tridacnidae, while Hyotissa is model, or any other AR value that has been applied to Lapita- in the family Ostreidae. age sites. Being at heart an unrepentant inductivist, I prefer to 4 . . . . . . follow the path down which the empirical archaeological record This statement requires a minor qualifier, in that the outer- itself leads me. most shell layers were etched away with acid in the radiocarbon laboratory pre-treatment process. Thus the dated shell itself In applying a AR value of -320, I have not used an additional presumably represents material that is a few years older than error factor because I have no empirical basis on which to base the actual date the mollusk was gathered; however, this differ- such a + term. Doubtless some error factor is inherent, but I ence in age is presumably only a matter of a few years or a prefer not to impose a strictly arbitrary value. decade at most, and hence is insignificant in terms of the inher- ent error range in the radiocarbon dates themselves. 10 Specht and Gosden (1997:187-88) have suggested that the 5Phelan (1999:100-101) observes that a AR value of -230 de- Area B posts "should be regarded as younger than the maxi- * . 1 . . , , \ ~~~~~~~~mum of their age rariges " on the grounds that "the nature of rived from a set of paired samples reported by Kirch (1993b) ies n bthe gu itha "the nureiof for the To'aga Site in Samoa is at variance with a AR value of the dated wood is not given, but even if it was the surviving + 57 that he has calculated from three moder samples, and on outside part of the posts, this cannot be assumed to represent this basis suggests that there may be "some problem of asso- the youngest growth of the trees prior to their being cut down." ciatio between these 2 determinations." To thecontrary, set The "nature of dated wood" is described in Chapter 4, and the clation between these 2 dleterminations." lTo the contrary, sets 'etfe yehooa'tD .Yna en of matched pairs presented here strongly suggest that the -230 species was i d b h value is valid, and that there is likely to be a greater range in AR Intsia bijuga, a hardwood described by Peekel as being "com- values than has hitherto been recognized. It must also be re- mon on the foreshore, but also not rare in inland forest" membered that AR estimations based on "modern" pre- (1984:214). Peekel also relates that the "wood is extremely hard bomb) samples do not necessarily apply to prehistoric samples, and is called'ironwood"', and is excellent for "building-tim- as it is well know that reservoir effectsdisplaytemoralber" (1984:214-16). The ECA posts had decomposed rem- as it is wellnownthareservoreffectdisplaytemporalaswell nants of bark on their outer surfaces, and the portions dated as geographic variation. did not extend more than 2 cm into the outer wood layers, as 6Stuiver and Braziunas (1993) indicate that for southern hemi- can be seen in Figure 4.25 where the sampled cuts are clearly sphere samples a further 40-year correction must be applied. visible. In short, the dated wood does indeed date theyoungest However, since Mussau is only barely S of the equator, I have growth ofthe trees prior to their being cut, and as bark was still not applied this additional correction. present there is no possibility whatsoever that old logs were used. Indeed, it is not clear why the builders of the Area B stilt 'The argument advanced here for a minimal oceanic reservoir house would possibility have wanted to use old (and hence effect on the Mussau reefs will not necessarily apply elsewhere probably rotted or termite-ridden) logs when fresh wood is so in the Bismarcks, since local reef geomorphology, tidal range, abundant along the foreshores of the Bismarck Archipelago. and energy regimes would have to be considered in each case. In sum, there is no basis whatever for Specht and Gosden's However, it is intriguing that in the date list for the FOH Site at speculations that the Area B posts should be regarded as younger Makekur in the Arawe Islands provided by Specht and Gosden than their actual age ranges. (1997, appendix 1), the marine samples when calibrated with a 400-year reservoir offset seem to yield ages that are inconsistent l' This statement is based on the absence of large, long-lived with the charcoal samples by again being somewhat too young. rainforest trees on the offshore coral islands, and on the rapid However, the authors do not present detailed stratigraphic rate of decomposition of dead wood in the humid tropics. 222 A RADIOCARBON CHRONOLOGY FOR THE MUSSAU ISLANDS 12 Specht and Gosden (1997:181-84) have questioned differing reservoir effects. One is that reservoir effects are not whether ANU-5088 was in "direct relationship" with the constant and could have varied locally over time. Another is ceramics at EHB, and "regard this sample as potentially beach that the mollusks gathered by the EHB Site inhabitants could material rather than as shell associated with the Lapita occupa- have come from a different marine microenvironment, per- tion." I can assure them that the sample was in direct asso- haps one more exposedto open ocean exchange, such as from ciation with the EHB ceramics, and that it gave every ap- the more exposed reef slopes on the S side of the Emananus- pearance of being culturally-deposited midden shell. Enusagila area, and thus be more representative of the "model surface ocean." All this is merely to point out that even in a 13 There are a number of reasons why mollusk samples from local region such as Mussau we should not necessarily expect different sites or depositional contexts in Mussau could reflect that reservoir effects will be constant over time or space.