Dating Lapita Pottery in the Bismarck Archipelago, Papua New Guinea JIM SPECHT AND CHRIS GOSDEN PREHISTORIC SITES WITH POTTERY known as Lapita have been the focus of archaeological attention in the western Pacific for more than thirty years. For much of this time the main concern has been with the relationship between Lapita pottery and the origin and spread of people who were the ancestors of the Polynesians. Whereas in the Solomon Islands and Papua New Guinea Lapita pottery appears in the archaeological record many millennia after the first human entry into the region, in Western Polynesia-and possibly also in New Caledonia and Vanuatu-the carriers and users of Lapita pottery appear to have been the first humans to colonize these island groups. For Kirch and Hunt (1988a:161), this expansion of Lapita pottery-using people from Papua New Guinea to Samoa "may be among the most rapid dispersal events in human prehistory." As Spriggs (1990: 17) has noted, this claim warrants further assessment. The reliable dating of Lapita pottery is thus important because changes in its chronology may affect interpretations of the nature and speed of its dispersal throughout the southwest ern Pacific. Kirch and Hunt (1988a) accept a date of cal. 3550 B.P. for the appear ance of Lapita pottery and propose that it spread extremely rapidly from the Bismarck Archipelago in Papua New Guinea to Western Polynesia, with no statis tically significant time difference between northern and southern sites. Spriggs (1990) prefers a slightly later starting date of cal. 3450 B.P. and sees the northern sites as slightly earlier than those to the south, thus allowing time for Anson's (1986) "Far Western Lapita" developmental stage of the decorative system in the Bismarck Archipelago. The difference between these two interpretations derives primarily from Spriggs's critical assessment of the dated samples. He rejects sev eral because of undemonstrated or doubtful association between the samples and the pottery occupations, in particular five of the earliest samples on which Kirch and Hunt base their chronology. Both papers, however, are concerned with the chronology of the pottery throughout its distribution and do not examine closely anyone area. This paper addresses issues of dating for sites in the Bismarck Archi pelago (Fig. 1), widely regarded as the "homeland" of Lapita pottery, whence the Jim Specht is Head of the Division of Anthropology, Australian Museum, 6 College Street, Sydney, NSW 2000, Australia. Chris Gosden is Lecturer and Curator at the Pitt Rivers Museum, University of Oxford, 64 Banbury Road, Oxford, OX2 6PN, England. Asian Perspectives, Vol. 36, No.2, © 1997 by University of Hawai'i Press. ASIAN PERSPECTIVES • 36(2) . FALL 1997 ADMIRALTY t'\. Mussau N ISLANDS ':g ... Islands Manus r---x--!) ECA;ECB;EKO;EHB;EKO r Island VJ--....bV GON" c!) «:)Q~ ... Balol ~Q L:n;;u~ , . ~LaSi~ELS;£.LT EFY "' D NEW ~()SSU • BISMARCK SEA IRELAND~ Watom Island. (J" SAC;SOI " \ Duke of Yorks 4. SOP;SEE;SEP;SET Nlssan ISland OFF; DG/J/2,OGW Buka ISlanfjf) DA/,OAA ~ Lapita Sites Kandrian 0'-. _~....;.;;lqOkm FFS.FLF Fig. 1. The Bismarck Archipelago, Papua New Guinea, showing the location of sites mentioned in the study. knowledge and skills for its production were carried south and east, eventually into Western Polynesia (e.g., Allen 1984,1991; Green 1979; Spriggs 1984). The main aim of the paper is to assess the dating evidence for the introduction of Lapita pottery into the Bismarck Archipelago. However, the pottery cannot be separated either physically or conceptually from the sites in which it is found or from the other artifactual elements of those sites. Taken together, the nature of the sites and their included artifacts also raise questions as to the meaning of the Lapita phenomenon in human terms. We feel that the new dates presented here, when set into the context of existing dates from the region, throw new light on whether Lapita sites or the artifacts in them were intrusive and possibly indicative of population movements into the region. We return to these issues at the end of the article. We present 24 dates (Appendix 1) for Lapita dentate-stamped pottery, or for stratigraphic contexts bearing on the date of its appearance, for sites in West New Britain Province, Papua New Guinea. We compare them with dates from other sites in the archipelago and suggest that two interpretations are possible. First, if the earliest dates accepted by Kirch are valid, Lapita may have begun in the Mussau Islands slightly earlier than in New Britain. Alternatively, if the earliest Mussau dates are not supported, Lapita pottery may have started throughout the archipelago later than the accepted calibrated date of 3450-3550 B.P. (Allen and White 1989; Green 1979: 32-34; Kirch and Hunt 1988a, 1988b; Kirch et al. 1991; Spriggs 1990). Both views raise questions about the presence of Lapita in this region. SPECHT AND GOSDEN . DATING LAPITA IN THE WEST 177 MATTERS OF CALIBRATION Spriggs (1989) introduced the idea of "chronometric hygiene" to discussions of radiocarbon dates to screen out samples of dubious material, origin, or result. This has been developed further by Anderson (1991) and Spriggs and Anderson (1993) into a protocol for the evaluation of samples and dates. Some of these criteria are not relevant for many of the samples considered here, and we employ only those relating to sample context, stratigraphic security, vertical consistency, and the possibility of" old wood." All dates discussed are calculated on the half-life value of 5568 years, adjusted for isotope fractionation with either a measured or estimated value for 013C (Stuiver and Polach 1977; Stuiver and Reimer 1993). Appendix 1 gives the mea sured values for 013C where available; estimated values are indicated by (E). Mea sured 013C values for all materials may vary slightly from the estimates, but have only a small effect on the corrected ages (Clark 1993; Taylor 1987: 122-123). The 13C corrected dates are calibrated with the CALlB 3.0.3 program (Stuiver and Reimer 1993). We do not apply the Southern Hemisphere correction factor for atmospheric samples because of the closeness of the sites to the equator, and we use a laboratory error multiplier value of K = 1.0. There is no measured Delta R value for marine samples in the Bismarck area, but Chappell and Polach (1976, 1991) have determined an Ocean Reservoir Effect value of -400 years for local ities at Huon Peninsula, near the western end of New Britain. We use this value in conjunction with the bidecadal atmospheric calibration curve (cf Bard et al. 1993; Pearson and Stuiver 1993; Stuiver and Pearson 1993). Ages are calculated by Method A and are cited as 2-sigma ranges rounded off to the nearest ten-year interval; range values ending in -5 are rounded upwards. Calibrated ages are cited as years B.P.; uncalibrated ages as years b.p. WEST NEW BRITAIN DATES Of the 24 results listed in Appendix 1, 22 have not been published previously and two (SUA 2814 and Beta 34208) have been published in summary form only (Specht et al. 1991). The results come from three sites in the Arawe Islands (FOH, FO], FOL) and two at Kandrian (FFS, FLF) on the south coast of West New Britain; one at Yombon (FGT) in the center of the island; and four on or near Willaumez Peninsula (FEA, FEL, FRI, FRL) on the north coast. Three dates are anomalous and are rejected. Site FEL is a beachside location on Garua Island, near Talasea, where dentate-stamped Lapita sherds are eroding on to the beach. Test excavations in 1989 revealed silts and slopewash containing very few sherds and obsidian flakes. This deposit extended down to a brackish aquifer with organic muds over a coral limestone substrate. Wood from the basal muds, at 125-153 em depth, gave a modern result (NZA 1979:76±56 b.p.), while charcoal from a higher level with pottery is dated to 1050 ± 80 b.p. (Beta 42526: 1130-780 B.P.). FEL in fact consists of redeposited materials from a ridge backing the area. The slopes of this ridge have slumped extensively, and the pot tery and obsidian in the FEL deposit probably derive from Lapita site FAO on the top of the ridge. The basal date suggests that this slumping was very recent. ASIAN PERSPECTIVES . 36(2) . FALL 1997 Charcoal sample SUA 2976 (480-0 B.P.) from site FEA on Boduna Island is discarded as it yielded an essentially modern result for a level expected to be ca. 2000 years old. Ambrose and Gosden (1991) obtained dates between 3150-2000 b.p. at the same level or just below this sample. The charcoal may be intrusive, reflecting disturbance of the deposit. The fourth sample rejected, NZA 450 from site FRL at Bitokara Mission near Talasea (Specht et al. 1988; cf. Torrence et al. 1990), was selected to date the palaeosol formed on a tephra from Witori volcano on Hoskins Peninsula. This tephra, known as the W-K2 tephra, covers much of central New Britain from Talasea on Willaumez Peninsula in the north to Kandrian and the Arawe Islands in the south (Gosden et al. 1994; Machida et al. 1996; Specht et al. 1991). The tephra is important because Lapita pottery is never found beneath it, and it thus sets an upper limit for the time of the appearance of the pottery in this part of New Britain. The small charcoal sample was expected to provide a minimum age for the reoccupation of the site following the tephra fall, but the modern result is clearly anomalous. All but three of the remaining 20 samples came from sites with dentate stamped Lapita pottery. The exceptions, from FGT and FRL, date palaeosols sealed by the W-K2 tephra and set a maximum age for it. Beta 57773 (4140- 3640 B.P.) and SUA 2814 (3850-3370 B.P.: Specht et al. 1991) from unit 4 at FRL (Specht et al. 1988; Torrence et al. 1990) overlap with Beta 45380 (4410- 3860 B.P.) for the equivalent level at FIF/2 near Yombon (Pavlides 1993) and with other dates for this tephra (Machida et al. 1996). SUA 2975 (3470-2750 B.P.) from FGT/V at Yombon is the youngest age yet obtained from a palaeosol immediately below the W-K2 tephra. We are uncertain how to interpret this date, since it falls between the next youngest pre-W-K2 sample (SUA 2814: 3850- 3370 B.P. at FRL) and the post-W-K2 samples from FGT and FIF (Pavlides 1993; Pavlides and Gosden 1994). The remaining 17 dates fall into two groups. The first, with four samples, ranges from 4410 B.P. down to 3470 B.P.; the second group has an upper limit of 3390 B.P. and extends to less than 2000 B.P. Three samples in the older group are from the Apalo site (FO]) on Kumbun Island in the Arawe group (Gosden and Webb 1994): Beta 55457 (3980-3640 B.P.) on wood, and Beta 54170 (4410- 3980 B.P.) and Beta 37560 (4070-3700 B.P.) on charcoal. These samples were recovered from basal contexts with obsidian flakes but no pottery. They overlap substantially with dates for pre-W-K2Ievels at FRL, FGT, and FIF/2 (see above), and are older than any accepted dates for Lapita pottery. The wood sample was taken from a timber that may have been part of a structure. This, and the two charcoal samples, is unlikely to refer to the Lapita period unless very old "old wood" was used. A simpler explanation is that the samples originated from a pre Lapita context. Pre-Lapita occupation in the Arawe Islands is demonstrated at Lolmo Cave (Gosden et al. 1994), and by a burial at the Paligmete site (FNY) on Pililo Island (Beta 27941: 4410±70 b.p. on Tridacna shell) (Gosden 1989:55, where the sample is incorrectly cited as Beta 28223 and 3960 ± 70 b.p.). The fourth date in the "older" series comes from the Lapita open site FFS at Auraruo on Apugi Island near Kandrian. This shell sample (Beta 63613: 3830- 3470 B.P.) came from just below the top of calcareous beach sand buried under clay slopewash. The Lapita occupation at FFS appears to have been on the sur- SPECHT AND GOSDEN . DATING LAP ITA IN THE WEST I79 face of this beach, although sherds and obsidian flakes occur down to 30 cm below the beach surface, presumably incorporated into the loose sand-shell-coral rubble matrix as a result of human and other activities. Beta 63613 is so much older than other accepted Lapita dates that it is likely to refer to the formation of the beach rather than to the age of the pottery. The oldest date in the "younger" series, Beta 63616 (3390-2960 B.p.-shell), refers to dentate-stamped Lapita sherds in Alanglongromo rockshelter (FLF) near Kandrian, which has a pre-Lapita occupation beginning around 4500 B.P. The sample range just overlaps with another shell date for the same layer at this site, Beta 57767 (3060-2750 B.P.), which is well within an acceptable range for Lapita. The two samples came from within 5 cm depth of each other, but in different parts of the trench. The associated pottery includes, in addition to dentate stamped designs, incision and fingernail impressions. These are generally not regarded as part of the earliest Lapita pottery. This is supported by a shell sample from the immediately post-Lapita layer (Beta 79348: CAMS 18944), which places the end of Lapita at this site prior to 2140-1890 B.P. Sample Beta 63616, there fore, could refer to pre-Lapita use of the shelter. The next oldest shell age, Beta 41578 (3260-2880 B.P.), came from the base of the FEA site on Boduna Island near Talasea, 20-30 cm below two previous shell samples in the 3060-2720 B.P. range (Ambrose and Gosden 1991). Seven other results from the Arawe Islands fall within the main range of Lapita dates: from FOH (Makekur), on charcoal, Beta 54164 (2930-2480 B.P.), Beta 54165 (3210-2770 B.P.), Beta 54166 (2960-2740 B.P.); and on shell, Beta 37561 (2750-2340 B.P.), Beta 55456 (2740-2340 B.P.), Beta 55323 (2720-2220 B.P.); from FOL (Amalut), Beta 54168 (2770-2360 B.P.). Two charcoal and Canarium sp. nutshell samples from FRI on Willaumez Peninsula, Beta 34208 (2110-1820 B.P.) and Beta 41590 (2320-1930 B.P.) (cf. Specht et al. 1991), fall at the end of the accepted Lapita range, consistent with other late dates reported for Watom Island (Green and Anson 1987, 1991). If our interpretation of the "older" group is accepted, none of the above results allows a claim for placing Lapita at these West New Britain sites earlier than ca. 3260 B.P., with the possible exception of Beta 63616 at FLF (3390-2960 B.P.). This is later than the starting date of 3450-3550 B.P. accepted by Kirch and Hunt (1988a) and Spriggs (1990), which is based primarily on Kirch's Mussau excava tions. We now look at dates for other Lapita sites in the Bismarck Archipelago and consider whether there is a gap between the Mussau and West New Britain dates. LAPITA IN THE BISMARCKS For this part of the paper we use dates already reported for the sites discussed above and other relevant dates from the Arawe sites, together with those for 14 other sites with Lapita pottery in the Bismarcks: KLK on Tuam Island in the Siassi Islands at the western end of New Britain (Lilley 1986, 1986/87); FPA on Kautaga Island and FCL on Poi Island in the Kove Islands, west of Willaumez Peninsula (Lilley 1991); the SAC and SDI sites on Watom Island (Green and Anson 1987, 1991); five unpublished dates for Lapita sites SDP, SEE, SEP, and SET in the Duke of York Islands at the eastern end of New Britain excavated by 180 ASIAN PERSPECTIVES . 36(2) . FALL 1997 J. P. White (Harris 1994; White 1995); and sites ECA, ECB, EHB, EKO, and EKQ in the Mussau Islands to the north of New Ireland (Kirch 1987, 1988; Kirch and Hunt 1988a, 1988b; Kirch et al. 1991). Two Lapita sites in North Solomons Province, DFF and DGD/2 on Nissan Island, along with four sites with the post-Lapita Buka style pottery (Y omining style of Spriggs), DGD/2 and DGW on Nissan and DAA and DAI on Buka Sohano Islands, are also included (Specht 1972; Spriggs 1991, 1994). We also consider samples from five sites on New Ireland and one on Manus associated with decorated sherds similar to Lapita or with plain sherds of uncer tain affinity: on New Ireland, Balof: EAB/2 (White 1992; White et al. 1991); Lasigi: ELS, ELT (Golson 1991, 1992, and pers. comm. 1994); Lamau: EFY (Gorecki et al. 1991); and Lossu Mound VI (White and Downie 1980); and on Manus, Kohin Cave: GDN, where Lapita sherds are bracketed by two samples (Kennedy 1981). Samples from Yombon sites FGT and FIF are included to further define the age range of the W -K2 tephra, with SUA 2975 and Beta 45380 providing maxi mum ages and the other samples dating site reuse after the tephra event (Specht et al. 1981, 1983; Pavlides 1993; Pavlides and Gosden 1994). Beta 1545 was associ ated with several undiagnostic plain sherds. All of the above samples are listed in Appendix 3. The four KLK dates differ slightly from those published by Lilley (1986/1987), o13e. which were based on assumed values for Lilley's thesis (1986:506) also provides results based on measured values; these are used here. The sets of dates differ by 10-40 years per sample. We omit ANU 4610 on the grounds that it has no cultural association (Lilley 1986: 126; Spriggs 1990) and cannot be related to the Lapita or any other use of the site. On the other hand, we include Beta 26261 from FPA and Beta 26259 from FCL, even though Lilley (1991) regarded them as not associated with the overlying Lapita deposits. At FPA Beta 26261 is stratigraphically consistent with SUA 2822 and SUA 2823 which date Lapita pottery. Beta 26261 thus provides a terminus post quem for the appearance of Lapita pottery on Kautaga Island. Beta 26259 from FCL is younger than all three FPA dates, but like Beta 26261 it provides a maximum age for Lapita at this site. Several samples are problematic. Sample ANU 2248 (4070-3480 B.P.), on shell from layer 10 of Kohin Cave (GDN) on Manus, was found with plain sherds stratigraphically below Lapita dentate-stamped sherds (Kennedy 1981). Spriggs (1990) initially accepted the date but later rejected it (Spriggs 1994; cf. Ambrose 1991 : 105-109, who regards the site as at least partially disturbed). Inspection of the sherds from layer 10 suggests that a Lapita origin for them is possible, but this needs confirmation, as does their association with the dated sample. At this stage, ± ANU 2248 and charcoal sample ANU 2212 (2310 120 b.p.: 2730-2000 B.P.) from layer 5 above the Lapita sherds provide a maximum range for the occur rence of dentate-stamped Lapita pottery at this site. On Nissan, charcoal sample ANU 6802 (4870-3740 B.P.) from DFF, with a range of more than 1000 years, is much too old for Lapita. ANU 8302 (3960- 3090 B.P.) from DGD/2, also on charcoal, starts well before any accepted Lapita date. Both samples probably represent the use of "old wood" or pre-Lap ita use of the sites and are excluded from the discussion. Charcoal sample GX 5499 (4990- 3630 B.P.) from a hearth or oven feature with Lapita sherds at the ECA site on SPECHT AND GOSDEN . DATING LAP ITA IN THE WEST 181 Eloaua Island (Bafmatuk et al. 1980) is also rejected because it is widely regarded as not referring to the pottery (Kirch and Hunt 1988a; Spriggs 1990). This sample may date old wood or refer to pre-Lapita use of the site. GX 5498 (3630-2760 B.P., on charcoal) from the same site is included because although its range of nearly 900 years makes it of limited value, its lower 5 range limit overlaps sub stantially with definite Lapita dates. Beta 26644 (3470-3160 B.p.-shell) from reworked W-K2 tephra in Lolmo Cave (FOF) (Gosden et al. 1994) is included, although the dated samples from this site are stratigraphically inverted and inconsistent. The relationship between Beta 26644, the sherds, and the tephra is unclear. The tephra may pre-date the included artifacts and shells, but it is also possible that items older and younger than the tephra became included in it during the reworking process. Beta 26644 was not associated with dentate-stamped Lapita, but with incised and fingernail impressed sherds typical ofa late stage of Lap ita (Gosden et al. 1994). The total number of dates is 112. Figure 2 shows the age ranges at 2 sigma for 62 shell samples, and 50 results on charcoal, wood, or nutshells are shown in Fig ure 3. Figure 4 shows the number of date ranges occurring in each hundred-year interval. Thus an age range of 2850-3150 B.P. is recorded in each of the intervals for 2800-2899,2900-2999,3000-3099, and 3100-3199. We have selected this approach rather than plotting the central tendency of each age (cf Kirch and Hunt 1988a) in order to provide a fuller picture of the distributions of the deter minations. By using the Method A calibration rather than the Method B proba bility approach, we avoid statistical weighting of the range determinations. A plot of the highest probability ranges using Method B would yield a result similar to that of Figure 4, but with rather fewer occurrences at the extremes of most ranges. Figure 4 does not include dates for the W-K2 tephra at FGT, FIF, and FRL; or those considered questionable or not to date Lapita, such as ANU 5339 at SAC, Beta 63613 at FFS, the earliest dates at FO], Beta 26644 at FOF, Beta 26261 from FPA, and Beta 26259 from FCL, ANU 2212 and ANU 2248 from GDN, and samples dating Buka (Yomining) style pottery in North Solomons Province. The oldest shell result, Beta 63613 (3830-3470 B.P.) from FFS, was discussed above as probably dating the beach on which the Lapita occupation took place and thus provides a maximum age for the pottery at this site. The next oldest shell date is ANU 5339 (3470-3070 B.P.) from the base of a feature in the basal zone of SAC on Watom Island. This result is suspiciously old in the light of other dates from Watom, of which Beta 16836 (2870-2380 B.P.) from SDI is the oldest. With the exception of SUA 5339, the Watom dates fall within, or are younger than, the range of 2960-2360 B.P. for the four shell dates from the Duke of York Islands (SDP: SUA 3061 [2760-2360 B.P.]; SEE: SUA 3082 [2880-2740 B.P.]; SET: SUA 3063 [2850-2550 B.P.], and SUA 3064 [2960-2750 B.P.]). Green and Anson (1987: 124) suggest that ANU 5339 may have been a shell on the beach prior to Lapita occupation which became incorporated in the infill of the feature (cf Golson 1991 for ANU 5850 from ELS at Lasigi). If that is so, Lapita on Watom Island should be younger than ANU 5339. Shell samples ANU 5088 and ANU 5089, from the Mussau area, are variously assigned to EHB (Kirch and Hunt 1988a; Kirch et al. 1991) and ECB (Kirch 1987: 168). According to the EHB attribution, ANU 5088 (3470-2980 B.P.) is FEA SOl 1500BP ELS SAC ELT I I 2000 BP FOH ECA/B KJ':'Y FOJ FCL SOP FPA SET 2500BP EKO FLF FEIA SEIE IT ECB 1"" EHB 3000BP SAC FOF FFS GON 3500BP 4000BP 4500BP Fig. 2. Radiocarbon dates on shell for Lapita sites in the Bismarck Archipelago. Siassi Islands: KLK (Lilley 1986); Arawe Islands: FNY, FOF, FOH, FO] (Gosden 1989, 1991; Gosden et aI. 1989; Gosden and Webb 1994; Gosden et aI. 1994); Kove: FCL, FPA (Lilley 1991); Willaumez Peninsula: FEA (Gosden et aI. 1989; Ambrose and Gos den 1991; Specht et aI. 1991); Watom Island: SAC, SDI (Green and Anson 1987, 1991); Duke of York Islands: SDP, SEE, SET (Harris 1994; White 1995); Mussau Islands: ECA, ECB, EHB, EKO, EKQ (Gosden et al. 1989; Kirch 1987, 1988; Kirch and Hunt 1988a, 1988b; Kirch et al. 1991); New Ireland: ELS, ELT (Gosden et al. 1989; Golson 1991, pers. comm. 1994); Manus: GDN (Kennedy 1981). EFY 1500 BP FRI ECNB II I 2000 BP LOSSU FOL FGT FIF/3 GDN FOH 2500 BP ECA EAB/II 3000 BP ECB FRL 3500 BP FGT/5 FOJ FIF/2 4000 BP 4500BP Fig. 3. Radiocarbon dates on charcoal, wood, and charred nutshells for Lapita sites in the Bismarck Archipelago. Arawe Islands: FOH, FO], FOL (Gosden 1989; Gosden et al. 1989; Gosden and Webb 1994); Yombon: FGT, FIF (Specht et al. 1981, 1983; Pavlides 1993; Pavlides and Gosden 1994); Willaumez Peninsula: FRI, FRL (Specht et al. 1991); Duke of York Islands: SEP (White 1995); Mussau Islands: ECA, ECB (Bafmatuk et al. 1980; Kirch 1987, 1988; Kirch and Hunt 1988a, 1988b; Gosden et al. 1989; Kirch et al. 1991 ); New Ireland: EFY (Gorecki et al. 1991), EAB (White et al. 1991); Lossu (White and Downie 1980); Manus: GDN (Kennedy 1981); Nissan (Gosden et al. 1989; Spriggs 1991, 1994). ASIAN PERSPECTIVES . 36(2) . FALL 1997 50 40 en w ...J 0.. ::iE 30 <C en 20 10 4000 3000 2000 YEARS B.P. Fig. 4. Frequency histogram of date ranges per hundred-year intervals. said to have come "from base of ceramic-bearing cultural deposit in calcareous sand matrix" (Kirch and Hunt 1988a: 168). In the absence of evidence for a direct relationship between the sample and the pottery, we regard this sample as potentially beach material rather than as shell associated with the Lapita occupa tion. A similar possibility can be raised for ANU 5089 (3370-2870 B.P.). Both dates overlap with ANU 5339 (3470-3070 B.P.) at SAC and Beta 63616 (3390- 2960 B.P.) at FLF, which were questioned above as possibly being from pre Lapita contexts. In all four cases, however, the lower ends of their ranges overlap with those of samples with undoubted Lapita associations. The main suite of shell samples has upper range limits of 3350 and 3260 B.P. (DGD/2, KLK, FEA) or less. The two samples from FPA (Beta 26261: 3210-2800 B.P.) and FCL (Beta 26259: 2850-2370 B.P.) in the Kove Islands described as dating "culturally sterile basal sediment" (Lilley 1991 :316), fall within this range. At EKQ, Beta 25670 (3220- 2780 B.P.), Beta 25671 (3150-2750 B.P.), and Beta 21789 (2870-2480 B.P.) come from levels 9, 13, and 17 of the same trench, but the results are inverted. The oldest and highest of the three, Beta 25670, has a range that starts about 350 years earlier than that of the lowest sample, Beta 21789. All three overlap at 2 sigma. At DGD/2 on Nissan Island, only one of the four shell dates relating to Lapita is earlier than 3300: ANU 5228 (3350-2870 B.P.). The other three shell samples from Nissan are all younger than 3000 B.P. (DFF: ANU 5221 [2770- 2350 B.P.], ANU 6804 [2120-1870 B.P.]; DGD/2: ANU 5229 [2960-2730 B.P.]). Thus seven shell results are questioned as possibly dating non-Lapita deposits. None has an upper range limit around 3550 B.P. A starting date for Lapita of ca. 3470-3390 B.P. is acceptable only if the relevant samples at FOF, FLF, SAC, and EHB do indeed refer to Lapita presence. The main series of dates, from 22 sites across the archipelago and North Solomons, including all other Mussau dates, falls later than this at ca. 3350 B.P. and younger.