CHAPTER 4 RADIOCARBON DATING AND SITE CHRONOLOGY P. V. Kirch, J. Coil, Al. I. Weisler, E. Conte, and A.J. Anderson K Wj Prior to the commencement of As stated in Chapter 1, one of our primary our project, a prehistoric chronology objectives in the Mangareva Archaeological K ;; \ for the Mangareva Islands rested Project has been to obtain new empirical evi- upon eight radiocarbon dates, all dence to refine the prehistoric chronology of from samples obtained during the Mangareva lslands, including establishing u Green's 1959 excavations at four an age for initial Polynesian discovery and colo- rockshelter sites on Kamaka and nization. Within the somewhat limited finan- Aukena islands (see Chapter 1). Five of these cial resources available to us, we have attempted dates came from Green's site GK-1 (redesig- to date as many samples from good stratigraphic nated KAM-1). The oldest date from this site contexts as possible. Here we report in detail (850 + 50 B.P., Beta-1 0901 8) along with a single on the AMS (accelerator mass spectrometry) date from the base of the nearby GK-2 (KAM- dating of 24 samples from five islands. 2) rockshelter (880 ? 70 B.P., Beta-109019) sug- gest initial occupation of Kamaka Island by the SELECTION OF RADIOCARBON SAMPLES early 13th century A.D. (Green and Weisler During fieldwork, samples for radiocarbon 2000).' As Kamaka is one of the smallest high dating were collected directly into aluminum islands in Mangareva, with limited terrestrial re- sample tins from features or contexts where the sources, it would not be expected to have been samples could be closely related to particular among the first localities colonized by early stratigraphy and associated cultural materials. Polynesians. For this and other reasons, Green A much larger number of such samples was col- and Weisler (2000, 2002) hypothesized that the lected than could be dated within the project initial discovery and settlement of the budget (these have been retained for future MIangareva Islands probablyr occurred two or analyses), and a selection of 24 samples was more centuries prior to the oldest dates from chosen for dating. In this pioneering phase of Kamaka. research, samples were chosen to represent as 95 RADIOCARBON DATING AND SITE CHRONOLOGY many different sites as possible to giv,e a broad Laboratorn prior to submission of samples to indication of the temporal framework for a Beta Analytic for sample pretreatment and ANIS Nlangarevan cultural sequence. As research in radiocarbon measurement. Using methods based Mlangareva continues it will be necessar) to run upon those described by Leney and Casteel multiple samples from the same contexts in or- (1975), charcoal fragments xvere individually ex- der to test for chronological variability within amined using two reflected light microscopes: specific sites. a Wild M15a stereoscopic for low-magnification Several recent articles have drawn attention and an Olympus BHS metallurgical scope for to the need for improving "chronometric hv- higher magnifications. Archaeological materi- giene" in Polynesian radiocarbon dating pro- als were compared with modern reference ma- grams (e.ig., Anderson 1991; Spriggs and Ander- terials consisting of Pacific Island wood thin son 1993; Dye 2000). While considerations of sections, experimentally carbonized charcoal sample stratigraphic contexts are a critical com- samples, and economic plant materials curated ponent of such an approach, the composition at the Oceanic Archaeology Laboratory. When- of radiocarbon samples can also affect the de- ever possible, charcoal fragments derived from gree to which radiocarbon dates accuratelyN re- short-lived plant taxa or from short-lived plant flect the true calendar dates of cultural events parts such as seeds or twigs, were isolated from of interest (the "target date" of Dean [1978]). the selected dating samples to minimize the po- Althou?gh charcoal is generally considered to be tential for inbuilt age to affect the resulting a good material for radiocarbon dating because dates. of its inert chemistry and the relatively simple Table 4.1 lists the provenience, sample weight, pretreatment necessary to remove modern con- laboratory identifications, ancd an assessment of taminants (Bowman 1990:29, Taylor 1987:43), the likelihood of an inbuilt age factor for all archaeological wood charcoal can also contain ANIS dated fragments. Unfortunately, it was not inbuilt age if it has been derived from the burn- possible to eliminate the potential for some ing of heartwood from long-lived tree taxa inbuilt age to exist in many of the dated samples. (Bowman 1990:15, Taylor 1987:45), or from In some cases, where samples were small, frag- wood that has been burned after a significant ile, degraded, or otherwise unidentifiable period of preservation, as with driftwood (Dye against available reference materials, dated frag- 2000:204). Dating of samples with inbuilt age ments could be identified only as "wood" or "di- can result in radiocarbon ages that are some- cotyledonous wood," while one sample (GANI- what older than the time period when the wood 7) contained questionable semi-carbonized was actually burned and deposited in its archaeo- material and one (GAM-20) contained uniden- logical context (e.g., Anderson 1991:fig.7). In tified lumps of carbonized plant matter mixed his critical analyNsis of the New Zealand ar- with sand. Those samples which could be se- chaeological radiocarbon corpus, Anderson curelr identified were primarily derived from (1991:792) concludes that "dates on charcoal economic plant genera including Aleurites, of minimal inbuilt age should be closest to the Altocarpus, Cocos, Cordyline, Hibiscus, IPanidanus, actual calendrical period; dates on marine shell and Thespesia. All of these taxa occur commonl) and moa bone collagen are less predictable but throughout the Mlangareva Islands today (see broadly in agreement; unidentified charcoal is (Chapter 2). the most problematic sample tyTpe." Because ideal materials could not be iso- To address these issues, radiocarbon lated from all of the dating contexts, al . samples from MVangareva sites were examined includes estimations of the potential for inbuilt at the U.C. BerkeleyT Oceanic Archaeology age for each of the radiocarbon results obtained. ARCHAEOLOGICAL INVESTIGATIONS IN THE MANGAREVA ISLANDS, FRENCH POLYNESIA Table 4.1. Mangareva radiocarbon dating samples: provenience and identification. Laboratory & Site No. Provenience Sample Botanical In-built Age Field Codes Weight (g) Identification Potential Beta-1 74777, 190-06-ATA-1 Unit Fl 1, Layer II, Level 8.8 Dicotyledonous wood, cf. High GAM-1 3, 50-60 cmbs (oven) Bauhine Beta-174778, 190-06-ATA-1 Layer ll, Level 1 29.6 Unident. seed tissue Low GAM-2 Beta-174779, 190-06-ATU-2 Layer I, Level 5, 52 0.2 Dicotyledonous wood High GAM-3 cmbs Beta-174780, 190-06-GAT-3 Layer II 3.4 Dicotyledonous wood, cf. High GAM-4 Bauhine Beta-174781, 190-06-GAT-3 Layer I, horizon A 0.4 Dicotyledonous wood, Low GAM-5 twig morphology Beta-174782, Akamaru, TPI Layer IB, Level 2, 30-39 2.9 Unident. seed tissue Low GAM-6 cmbs, oven Beta-174783, Akamaru, TP2 120-128 cmbs, soil + 0.1 Semi-carbonized material ? GAM-7 charcoal Beta-1 74784, 190-04-KAM-2 Layer ll, Level 3 8.1 Artocarpus wood Medium GAM-8 (no. 59) Beta-1 74785, 1 90-04-KAM-2 Layer IV, Level 2 5.7 Cocos wood Medium GAM-9 (no. 70) Beta- 174786, 1 90-04-KAM-2 Layer V, Level 2 3.5 Pandanus wood Medium GAM-10 (no. 80) Beta- 174787, 1 90-04-KAM-2 Layer VI, Level 6 16.3 Pandanus fruit (key) Low GAM-1I (no. 99) Beta-174788, Rikitea Chez Louis, Core 2, 0.1 Unident. wood High GAM-12 Transect 55-60 cmbs Beta-174789, 190-06-ATA-4 Core hole 5, 74 cmbs 0.3 Dicotyledonous wood High GAM-13 Beta-174790, 190-06-ATA-4 Core hole 6, 60 cmbs 0.2 Aleurites endocarp Low GAM-14 Beta-174791, 090-06-GAE-1 Erosional deposit with 0.2 Dicotyledonous wood High GAM-1 5 terrestrial gastropods ANU-1 1927, Rikitea trench Gley layer, 90 cmbs "Creeper twig" charcoal Low GAM-16a (split) NZA-1 5383, Rikitea trench Gley layer, 90 cmbs "Creeper twig" charcoal Low GAM-1 6b (split) Beta-168443 Rikitea trench Gley layer, 90 cmbs "Creeper twig" charcoal Low GAM-16c (split) Beta- 1901 15 190-06-ATU-I A TP-1, under pavement 0.1 Cordyline fruticosa stem Medium GAM-17 paepae Beta- 190 116 190-02-AGA-3 TP-1, interface of 0.4 Pandanus sp. wood Low GAM-18 Layers I and 11 Beta-190117 190-02-AGA-3 TP-1, base of Layer 1II, 0.5 Hibiscus tiliaceus wood Medium GAM-19 59 cmbs Beta-190I18 190-12-TAR-6 TP-2, base of Layer II, 4.4 Unknown carbon in sand ? GAM-20 58 cmbs clumps Beta- 190 1 19 190- 1 2-TAR-6 TP-2, interface of 0.5 Cf . Artocarpus wood Medium GAM-21 La|I yers I/ll Beta-1901 14 190-12-TAR-6 TP-3, Layer III, 103 cm 0.3 Seabird bone, cf. Low GAM-22 Procellariidae 97 RADIOCARBON DATING AND SITE CHRONOLOGY Of the 24 dated samples, five were considered Waikato for independent dating. All three dates to have low potential for inbuilt age (<50 years), are ANIS dates, and pretreatment methods were seven were judged to have potential for a me- comparable with acid/alkali washes. dium degree of inbuilt age (>50-10()() years), and The results of AMS dating on the 24 samples seven were considered to have significant po- are provided in Table 4.2. Somewhat surpris- tential for inbuilt age (>100 years). Two inglv-, five samples yielded ages which are re- samples (GANI-7 and GANI-20) were difficult ported in Table 4.2 as 'pMC' or 'percent mod- to evaluate based on their unusual nature (given ern carbon'. These samples are <50 vTears old, as "'unknown" in Table 4.1). These should con- meaning that there xvas a 'greater concentration servatively be considered as potentially contain- of "C in the sample than in the A.D. 1950 refer- ing a high degree of inbuilt age. These estima- ence standard (95'} (-of the "4C content of the tions need be taken into account when inter- National Bureau of Standards Oxalic Acid). preting the calibrated date ranges. For samples Remaining fragments of these samples were re- with "medium," "high," or "unknown" inbuilt turned by Beta Analytic to U.C. Berkeley for age potential, the dates obtained should be seen reexamination after AMNS dating. One of these as providing a terminus ante quenm, or "date be- (GANI-7), a sample that originallv appeared to be fore which," cultural events of site formation semi-carbonized, mayr represent stratigraphicallT occurred. It is also possible, however, that intrusive modern root material. The initial iden- samples with potential medium, high, or un- tification of the other two samples, originally known inbuilt age are actually free of such bias identified as carbonized seed tissue (GANI-2) and that their dates do in fact accurately- reflect and unknown dicotyledonous wood charcoal the calendar period when the wood was burned. (GAN-13), were reconfirmed. The reasons that these two samples returned modern ages remains DATING MIETHODt)S AND RESULTS unclear. Samples GAN1-1 to -15 and -17 to -22 were For the 19 other samples listed in Table 4.2, submitted to Beta Analytic Inc. for pretreatment we have provided the results in terms of the and ANIS radiocarbon dating. The same pretreat- measured 14C age (calculated using the Libbi, ment procedure was applied to all charcoal 14C half-life of 5568 yrs), the ratio (613C) be- samples in order to eliminate contaminants such tween 13C and 1kC (calculated relative to the as carbonates and secondary organic acids, PDB-1 international standard), the "conven- along with modern rootlets. The samples were tional radiocarbon age" (as defined by Stuiver gently crushed and dispersed in de-ionized wa- and Polach 1977), and the calibrated age range ter, folloxved byT hot HCI acid washes and alkali at I standard deviation (68' oprobability). Cali- (NaOH) washes; this was followed by, a final bration follows the calibration database and acid rinse to neutralize the solution prior to dry- methods of Stuiver et al. (1998) and of Talma ing (Darden Hood, pers. comm., Feb. 4, 2003). and Vogel (1993). All charcoal samples were For GANI-22, a sample of bird bone, bone col- calibrated using the atmospheric calibration lagen was extracted with alkali pretreatment. database INTCAL98, while a sample of sea- Three additional samples (GAMI-16a, b, c) bird bone from the Onemea site was calibrated consist of subsamples ("splits") from a single using the marine calibration curve MtARINE98 bulk sediment sample taken from a buried gleved with a AR value of 0 ? 0. We now turn to a clayr horizon in Rikitea Village. Three separate brief discussion of the various dates reported subsamples were sent to Beta Analytic, to the ra- in Table 4.2 in terms of their stratigraphic and diocarbon dating laboratory at the Australian archaeological contexts. The first five localities National University7, and to the UnivTersity of discussed below are situated on Mlan?gareva Is- 98 ARCHAEOLOGICAL INVESTIGATIONS IN THE MANGAREVA ISLANDS, FRENCH POLYNESIA TABLE 4.2 Mangareva radiocarbon dating samples: results. LABORATORY MEASURED 13C/12C RATIO CONVENTIONAL CALIBRATED AGE RANGE & FIELD CODES 14C AGE (B.P.) | (/00) 14C AGE (B.P.) () A.D. Beta-174777, GAM-1 670 ?40 -23.9 690 ?40 1280-1300 Beta-1 74778, GAM-2 113.8 ? 0.6 pMC -25.5 113.9?0.6 pMC Beta-174779, GAM-3 210 ?40 -24.9 210 ?40 1650-1680, 1770-1800, 1940- 1950 Beta-174780, GAM-4 190 ?40 -25.2 190 ?40 1660-1680, 1740-1810, 1930- 1950 Beta- 174781, GAM-5 100.1 ? 0.5 pMC -12.6 190 ?40 1660-1680, 1740-1810, 1930- 1950 Beta-1 74782, GAM-6 430 ?40 -28.2 380 ?40 1450-1520, 1590-1620 Beta-174783, GAM-7 120.7 ? 0.8 pMC -27.9 121.4 ? 0.8 pMC Beta- 174784, GAM-8 230 ?40 -26.1 210 ?40 1650-1680, 1770-1800, 1940- 1 1 1 ~~~~~~~~~~~~1950 Beta-1 74785, GAM-9 240 ?40 -25.3 240 ?40 1640-1670 Beta-174786, GAM- IO 460 ?40 -25.1 460 ? 40 1420-1450 Beta-174787, GAM- 11| 330?40 -21.2 390?40 1450-1510, 1600-1620 Beta-174788,GAM-12 860?40 -24.6 870?40 1160-1220 Beta-1 74789, GAM-13 109.3 ? 0.5 pMC -28.2 110.0?0.5 pMC ---- Beta-174790, GAM-14 650?40 -23.2 680 ?40 1280-1300 Beta-174791, GAM-15 220 ?40 -23.2 220 ?40 1650-1670, 1770-1800, 1940- 1950 ANU-H 1927, GAM-16a -24.0 320 ?180 1400-1850, (split) (estimated) 1900-1950 NZA-15383, GAM-16b 98.7 ?0.7pMC -25.8 180 +57 | (split) Beta-168443, GAM- 410 ? 40 -22.3 450 ?40 1430-1460 16c (split) Beta-190115, GAM-17 450 ? 40 -26.3 430 ?40 1430-1470 Beta-190116,GAM-18 480?40 -26.7 450?40 1430-1460 Beta-190117, GAM-19 760 ? 40 -26.3 740 ?40 1260-1290 Beta-190118, GAM-20 1010 ?40 -24.7 1010 ?40 |100-1030 Beta- 1901 19, GAM-21 | 740 ? 40 -24.0 |760 ?40 | 12 i- 1280 Beta-190114,GAM-22 |1170?40 -12.2 |1380?40 |1000-1050 99 RADIOCARBON DATING AND SITE CHRONOLOGY land, followed by those on Akamaru, Kamaka, failed to locate a substantial cultural deposit. Taravai, and Agakauitai islands (see Chapter 3 The low-WNing basin-shaped depression be- for stratigraphic details of all sites). tween the coastal sand beach ridge and the base of the colluvial slopes at Rikitea was formerly DIscus&IC)N OFT R,ADI( )( ARB( )N a major zone of wet taro cultivation. This area DATING RESULTS was originally, identified from a buried gley soil R!KTF I JTIIACE A RisL layer that formed through continuous fresh Weisler (1996:70) and Green and Weisler water saturation QTercinier 1974). As described (2000:32; 2002:232) argued, on environmental in Chapter 3, we took advantage of a 75 m long grounds, that the area of Rikitea Village on drainage trench to record the stratigraphv at a MIangareva Island was likely to have been "an locality about 10 m seaward of the base of the ideal locale for initial occupation of the cliffs inland from the MIairie. The key strati- MIangarevan group." Rikitea offers a sheltered graphic unit here was the 25 cm thick glev layer bay and canoe landing, along with one of the containing small amounts of finely dispersed largest valleyTs with manx' freshwater springs at charcoal (see Fig. 3.8). the base of Auorotini watering a swampy allu- Charcoal dispersed in sedimentarv deposits vial basin which in historic times contained the that accumulate gradually over time-such as mo)st important zone of intensive taro (&/.ocasia that represented by the Rikitea ,gleyT layer-can esculeta) cultivation. The ritual and political sig- be difficult to date reliably. This is because the nificance of this localitv also suggests a long origin of the sediments and charcoal par- historyr of settlement (see Chapter 2). As de- tidles- can change ovier time and the deposi- scribed in Chapter 3, one component of our field tional environment can be disturbed and re- strategy involved stratigraphic coring along worked centuries after first deposition. Roots multiple transects running from the shoreline of shrubs and trees can penetrate soil layers inland, cross-cutting the low accretionary beach adding more recent carbon to ancient sediments. ridge which separates the taro swamp from the This is especiall problematic when vegetation sea, and which has lilkelyT been a major locus of burns and follows the roots well belowr the sur- habitation throughout prehistory. At one face, thus adding younger charcoal to old. Nlind- transect in particular ("Chez Louis") a fairlyT ful of these potential problems, we removed a deep cultural layer, containing charcoal and bulki sediment sample from the glev layrer 90 cm overlyTing a coarse carbonate sand and grit, was below surface and processed it in the field. The encountered in rcore 2 about 15 m inland of sediment was placed on a 3.2 mm sieve and the road and on the margins of the taro swvamp. washed with fresh water. Charcoal was collected A charcoal sample of unidentified wood from with forceps and placed in a plastic bag. The 55-60 cm depth was radiocarbon dated (Beta- single sample (GANI-16) was then split into 174788, GANI-12), with a result of cal A.D. three subsamples, each being sent to a different 1160-1220. This date corresponds closely with laboratory as described above. One subsample two dates obtained by Green and Weisler (2000, (NZA-15383, GAMI-16b) yielded a "modern" table 2) from the GK-1 (190-04-lKAM-1) and age, while the other two subsamples yielded ages GK-2 (190-04-K,NAM-2) rockshelters on of 450 ? 40 and 320 ? 180 B.P. (Beta-168443, Kamaka Island (with 2a ranges of cal A.D. 1065- GANM-16c; ANtl-11927, GAMI-16a). The lat- 1294 and 10)25-1292). The Rikitea date offers ter date has a rather large standard deviation strong support fo)r the hy pothesis that early habi- but overlaps with the Beta- 168443 date which tation deposits are located here, although an ex- we take to be a best estimate for the deposition panded test excavation at Chez Louis in 20)03 of the gleyr layer. CEalibrated to A.D. 1430-1460, 100 ARCHAEOLOGICAL INVESTIGATIONS IN THE MANGAREVA ISLANDS, FRENCH POLYNESIA this date suggests that the Rikbitea taro swamp ran,ges: cal A.l). 1650)-1680), 1770-1800(, and wsin use as an agrclua se !a es 1940)-19)50. The last age can be ruled out based the mnid-15tiz centurv .) on the absence of anv modern materials; it T771.1n7SReA ~~~~~~seems fikelyT that the deposit accumulated in the Another area with si,gnificant ethnohistoricallN, late pre-contact era. documented settlement is Atituitil, to the south AnZAo-A 1/A1___y SIH-ms of Rikitea in the lee of Auorotini. A number of In the Atiaoa Valley on the northwvestern important sites are located here, including the side of Nlangareva Island, we tested a small large 1 90-06-ATU- I A paepae and associated struc- rockshelter (site 1 90-06-ATA- 1) and carnied out tures on the natural terrace called Atituiti Ruga trnetcigopainscosthcatlfa, ampe of mopnumenstai archte Icturaunqe withi locating a buried cultural deposit (site 190-06- ampl ofmonuentl arhitctur wihin ATA-4). Both sites wTere radiocarbon dated. WIngrea,bu a sggstd n hate 3 ml Two samples from the rockshelter were submit- also be the site of solstitial observations byT the ted for dating, the first (Beta-174777, GAlSI-1) N'Iangarevan priests, as described byT the earlv consisting of wood (cf. Baul)ihe) charcoal from Catholic missionaries (Laval 1938). In our 2003 an oven feature expotsed in the north profile (see test excavations at ATU-1A1, we wvere able to Fig. 3.27), the second (Beta-174778) consist- obtain several charcoal specimens from a sealed ing of an unidentified seed from the top of LarTer stratigraphic context under the basalt pavTing IIIA. The latter of these samples vielded a mod- stones in front of the large tabular boulder "seat" ern age, but the sample from the earth ovren on the paepae platform (see Fig. 3.19). A frag- yielded an age range of cal A\.D. 1280-1300). ment of Cordyline fiuticosa stem was submitted Transect coring revealed buried cultural sedi- for dating, (Beta-190115, GANI-17), with a re- metcoainghrolinazeexndg graphic conex t.D. 1430-47() probabl poest-dates about 50 m seawvard of the rockshelter (see Fig. grahicconext ths smpe pobal5Tpos-daes 3.29). Two s.amples wvere submitted for radio- the actual construction of the paepae (although carbon dating. One sample (Beta-174789, possib, ntba e ln nexa)adsod GANI- 1 3) of unidentified wood returned a mod- nrield a good estimate of the period when the ern ag;e, but the second sample of candlenut present paved surface was in use. The 15th cen- (A4leui*tes moluccana) endocarp (Beta-17479(), tury date falls into the later part of the GA I-4iedangrgofcl.. 20 l\langarevan sequence, corresponding to the 13(00, identical to that from the nearbyT period of intense inter-tribal competition and rockshelter. These tXvo acceptable dates from rivalrSy for political power so amplyT documented Atiaoa sites ATA-1 and -4 (GAINI-1, -14) are bi, Hiroa (I 938a) in his summar, of NIangarevan onlyT sfightlwr vTounger than our oldest dates fro-m oral traditions. Rikitea and the Kamaka Island rockshelters, and C)n the coastal flat called Atituiti Raro we suggest that the Atiaoa Valle), has been occu- also test excavated a buried midden deposit (site pidsnealatthlte1h nur .. 190-06-ATU-2) exposed b), coastal erosion. Our pldsnealatthlte1hcnur Ai. single 1 M2 test unit into this cultural deposit GATAvAm-. 11A1jiJE;y extended to 60 cm below surface; no artifacts GJatavake Vallei, directly across the low were found, but faunal materials and charcoal mountain pass from Rikitea and east of Atiaoa, CHAPTER 3 ARCHAEOLOGICAL FIELD INVESTIGATIONS E. Conte, P. 1V Kirch, AlI.I. Wleisler, and A.J. Anderson For reasons made clear in Chapter 1, referenced to the WGS84 datum. Sites on J.7 ,) our approach to fieldwork in the Mangareva, Aukena, and Akamaru were plot- N Mangareva Islands during our first two ted on a set of advance sheets of the new topo- field seasons has been extensive rather graphic survey of French Polynesia (1:50,000 "b than intensive. Our strategy has been scale) kindlyT made available to us by the Ser- j to sample-through both surface re- vice de Urbanisme, Pape'ete. (Unfortunately, connaissance and test excavation a such topographic maps are not available for diversitv of locales on most of the major vol- Taravai or Agakauitai.) In Atiaoa Valley and at canic islands. lntensive studies of particular lo- Atituiti Ruga, on Mlangareva, we used plane calities and extensive excavations at specific table and telescopic alidade to map architec- sites are anticipated for future phases of the tural features in detail. Other maps were made project. In this chapter, we present the results using compass, tape, and hand level. Structures of survey,s and test excavations in 2001 and were cleared, described, and photographed us- 2003, organized geographically so as to integrate ing both black-and-white (120 roll film, 35 mm), observations on surface sites, relevant environ- color slide (35 mm), and color digital cameras. mental features, and the results of tests in se- Coring operations were desiogned to investi- lected sites. We begin with the largest and cen- gate whether there were cultural deposits tral island, MIangareva, and proceed to the present in coastal beach ridges on Mlangareva smaller islands within the lagoon. and Akamaru islands, especially at depth. The equipment consisted of a Dormers Hand drill- FIELD METHODS ing rig with 6 m of aluminum rods, a 75 mm Field methods followed procedures widely sand auger, and a 75 mm Jarret loam auger. applied in Polynesian archaeologyT. Sites were Test excavations (typically 1 in2) were car- located whenever possible using a Garmin XLI 2 ried out following cultural and natural stratig- GPS receiver, with Universal Transverse raphv, and all sediment wvas screened through 5 MIercator Projection (UTMI Zone 8) coordinates mm and 3 mm mesh for recoveryT of small fau- 102 ARCHAEOLOGICAL INVESTIGATIONS IN THE MANGAREVA ISLANDs, FRENCH POLYNESIA sam pie has Ii aga-e ranges of cal A.D. 1650-1680, paeet and edging adsubsequent re-occu- 1770-1 8001, and 1940-1950; the latter can be pation in the 17-18th centuries. rejected on the total absence of recent historic materials from the deposits. The underlving Layer IV sample has an age range of cal A.D. Located along the shore of the smallest of 1640-1670. These two dates indicate that the three large bays on the northwestern coast of beach rock pavTement and the midden deposit Tarav,,ai Island', the Onemea site (190-1.2-TAR- which developed on top of it wvere deposited in 6) did not yTield many, artifacts from the two test a time frame encompassing- the 17-18th centu- pits excavated but nonetheless is remarkable for ries, i.e., the proto-historic period prior to Eu- the hi~gh concentration of bird bones found at ropean contact. The Layer V midden underly- the base of TP-2, both in the lower part of the ing the pavement, however, yTielded a signifi- cultural deposit (Layer II) and in the immedi- cantly, older age range of cal A.D. 1420-1450. atelyT underlying sand (Layer III; see Fig. 3.46). This raises the possibility of a h-iatus in use of Elsewhere in Polynesia, similar high concentra- the rockshelter between the Laver V midden and tions of bird bones have tyNpically, proved to be the construction of the Layer IV pavement. The associated with the earliest phases of human sample from one of the deep ovens returned colonization on islands (Steadman 1989, 1995; age ranges of cal \.D. 1450-151 0 and 1600-1620. Steadman and K,:-irch 1990). Hence, it seemed The earlier of these ranges overlaps with the possible that the Onemea site incorporates cul- range for Layer V7 from which the oven pits were tural deposits dating to the in'tial period of hu- cut. Mlost likel); both the Layer V midden and man occupation on Taravai Island. the ovens date to a time period of approximately, Three samples were submitted for dating, the 13-14th centuries, all from the deeper TP-2 unit at Onemea. The The presence of the large, inter-cutting ov- uppermost sample (Beta-190119, GANI-21), ens prohibited us from obtainiing a good in situ consisting of wvood tentatively idnife t1h charcoal sample from the true basal cultural genus Ai-ocdia;ps (breadfruit), came from the i'n- deposits (Green's LayTer G, see Green and terface of Layers I and II, at 20-22 cm below Weisler 2000, fig. 14). We have no reason to surface. This sample yielded an age of cal A.L). doubt the val-idity of the date obtained by Green 1250-1280, roughly the same age as the base from this deposit, calibrated to A\.D. 1025-1292, of the Kamaka Island rockshelters and of the as this is reinforced by, a date of almost identi- sample from Chez Louis at Rikitea Village. A cal age from the nearby, GIK-1 (= K,:-AM-1) second sample (Beta-1901 18, GAMI-20) was rockshelter (Green and Wesiser 2000, table 2). collected directly, from the cleaned south face An Oxcal plot of the probabil-it distributions of the stratigraphic profile after the completion of all five available dates from the IK_AM-2 of excavations, from a thin lens of carbonilzed rockshelter is provided in Figure 4.1. Our ex- material at 58 cm below surface (see Fig. 3.46). panded range of radiocarbon dates from site This sample appeared to represent burninVg of KANI-2 suggests the following temporal se- vegetative matter directly on top of the Layer quence: (1) initial occupation in the 11-13th 11I sand deposit containing the high density of centuries, followed by a possible hiatus; (2) con- bird bones, and immediately prior to the accu- tinued occupation in the 13-14th centuries, in- mulation of the Layer II cultural deposits. While 103 RADIOCARBON DATING AND SITE CHRONOLOGY Blet-a-174784 210?40BP B3eta.1,74785 240 i.-40BP Beta-174786 460?40BP Betai174787 390?40BP__ Bieta- 109019 890?70BP. I I I ~~ ~~A _j I i 1. I 500CaIAD iQOOCaIAD 150OCalIA.D 2000CaIAD Calibrated date FIGURE 4.1 Oxcal plot of five radiocarbon dates from Site 1 90-04-KAM-2. ter. This sample yielded an age of cal A.D. 1000- In Pacific archaeology, much debate has ensued 1030. The third sample (Beta- 1 90114, GAM- over what AR values should be applied to ma- 22) consisted of a complete long-bone shaft of rine samples for particular areas (see Kirch 2001 a seabird (probably, a petrel species of the fain- for discussion of the reservoir problem with re- ily, Procellariidae) from Layer III at 103 cm; the spect to Lapita archaeology). Much may depend 613C value of -12.2 obtained for this sample is on the local marine topography and environ- consistent with what would be expected for a ment', such as the presence of extensive reefs seabird subsisting on a marine diet. Using the and lagoons where there is apt to be consider- marine calibration curve with a AR value of 0 able exchange of CO 2between the upper ocean ? 0, this sample returns an age of cal A.D. 1 000- layer and the atMosphere, or the presence of 10501, essentially identical to the charcoal date steeply, shelving islands where upwelling may,, from the lens immediately, overly:ing Layer III. be signficant.2 Unfortunatel); no empirical mea- Calibration of a sample such as GAM-221, surements of the local reservoir effect are avail- using the marine (rather than atmospheric) cali- able for Mlangareva. For the Society Islands, a bration curve is affected by, the value chosen AR value of 45 ?30 has been reported (Stuiver for the ocean reservroir effect', the so-called AR et al. 1986). Recent dating of a series of 23 value. Since the world's oceans are a "sink" or coral samples from Rapa Nui (Easter Is.), per- reservoir for older carbon', marine-grown haps a better fit for Mlangareva, indicated an samples typically yield ages somewhat older than average surface ocean reservoir value of 355+ their true age (Stuiver and Braziunas 1993; 71 years (Beck et al. 2003: 102-104, table 2). Stuiver et al. 1986). The MARINE98 calibra- This would imply, a AR value of approximately tion curve uses a "model ocean surface" which -45. Manigareva, with its extensive lagoon, mayT is essentiallv' a smoothed version of the atmo- be analogous to the Mlussau situation, although spheric curve (INTCAL98) offset by, an aver- we have no idea whether the diet of the seabird age age of 400 years. Howvever, since the local dated in sample GAM-22 was primarily fish 104 ARCHAEOLOGICAL INVESTIGATIONS IN THE MANGAREVA ISLANDS, FRENCH POLYNESIA perhaps a decade or tvo older than the latter. 1460. The second sample, of Hibiscus tiliaceus This is indeed the result one would predict from wood, came from the base of cultural Layer 113, the stratigrraphic and archaeolocgical evidence. justIabove the contactwith the underlying red- As noted in (Chapter 3, the Layer lII deposit dish sediment of Layer IV, at 59 cm below sur- at Onemea is not an in fitu occupation but none- face. This sample (Beta- 190117, GANI- 19) theless shows signals of human presence. These yielded an age of cal A.D). 1260-1290. Taken to- include twvo fire-altered volcanic stones (prob- gether, these two dates suggest that the cultural ably oven stones) and a number of shells of a deposits in the Nenega-Iti rockshelter accumu- Polynesian-introduced garden snail (Allopeasgrac- lated over a period of about 200 years, from the ile). The closely, consistent results obtained for late 13th to the mid-15th centuries A.D. Expanded samples GAM-20 and GANI-22 indicate that excavations at this site should therefore provide a PolyNnesians were present on Taravai Island bv good sample of materials dating to the middle the close of the 1()th century, or first few de- phase of the Mlangarevan cultural sequence. cades of the I 1 th centuryT A.D., some 150-200 CONCLUSIONS years earlier than the initial occupation of the Kamaka Island rockshelters. This accords well The radiocarbon dates obtained as a result with the prediction of Green and Weisler (2000, of our 2001 and 2003 excavations now expand 2002) that the first part of the Mangarevan cul- the Mangarevan radiocarbon corpus by a factor tural sequence was not evidenced in the of three. Figure 4.2 provides an Oxcal plot of Kamaka sites. Based on the GANI-21 date, the all 20 calibrated dates (excluding those with occupation on the O)nemea beach ridge did not modern ages) and shows their probability dis- extend after the mid-to-late 13th century; al- tributions. As can be seen, the Mangarevan se- though it is expectable that there are later sites quence now spans a full eight centuries. The elsewhere within the Onemea Valley. In short, full implications of this corpus for outlining a Onemea is a good candidate for a site dating to prehistoric cultural sequence for the archipelago the pioneering phase of Polynesian settlement in wiii be developed in Chapter 8. Here we con- the Mangareva Islands. clude with a few ke) observations. (1) Initial Polvnesian discovery and settlement of the \TI11\ Rocro-II 7 -()C, HFS X7I JA\GANEKA L471 RocKI. HEL--R, Mang-areva Islands occurred no later than the end A GA K/i LJFE~1! isL4\lAT of the 10th century A.D., or opening decades of The Nenega-iti rockshelter (1 90-02-AGA- the I 1 th century, based on the new dates from the 3) on Agakauitai Island contains a well-strati- Onemea site. (2) By the 13th century, xve have fied, undisturbed cultural deposit extending to evidence for widely dispersed occupation-in a depth of about 60-70 cm (see Fig. 3.54). Al- both rockshelters and open sites-on though our test excavations were limited to a Mlangareva (at both Rikitea and Atiaoa), Taravai, single square meter, the site vielded the richest Agakauitai, and Kamaka islands. (3) MIonumen- material culture assemblage from any, of the sites tal architecture, as evidenced by the paepae at we sampled, with nine fishhooks or hook frag- Atituiti, was being constructed by the 15th cen- ments, 11 coral files, much worked pearl shell, tury. (4) A major episode of erosion and depo- and several other artifacts. Two samples were sition of terrestrial sediments, as evidenced at submitted for AMS dating. The first consisted Gatavake and G3aeata, was in effect by the 17- of a sample of Pand-anus wood from the inter- 18th centuries (if not earlier), indicating con- face betxveen Layers I and II (Beta-I1901 16, siderable environmental degradation and instabil- GAMI-18) and returned an age of cal A.D. 1430- it) by7 the late phase of the MIangarevan sequence. 105 RADIOCARBON DATING AND SITE CHRONOLOGY MannedfromSU%iu et al. (1998); Deta R -45;0Cal O8BrordRasey(2002) ctbr 4 sd12prcbuspicwTon 8uhaEe98 B-190114 1380?40BP _ _ Curve intcal98 B-190118 1010?40BP _ _ B-174788 870?40BP _ __ _ B-190119 760?40BP _ B-190117 740?40BP _ B-174777 690?40BP L. B-174790 680?40BP _ B-174786 460?40BP _ - B-190116 450?40BP ____ _ B-168443 450?40BP _ B- 190115 430?40BP _ B-174787 390?40BP _ _ _ B-174782 380?40BP31 ANU-11927 320?180BP Beta-174785 240?40BP - -A Beta-174791 220?40BP _ h Beta-174784 210?40BP A Beta-174779 210?40BP _ I Beta-174780 190?4OBPP Beta-174781 190?40BP _A___- 50OCalAD 100OCaLAD 150OCalAD 2000CalAD Calibrated date FIGURE 4.2 Oxcal plot of radiocarbon dates from the Mangarevan Islands, showing probability distributions. CHAPiT1ER 4 E,NDu)vNOES's 'In 2001, just prior to our own expedition, MI. Orliac (2003:159) carried out investigations at Gatavake, obtaining an additional radiocarbon date of 830 ? 70 Bx.I. (Beta-160931). 2 For example, in the M\ussau Islands of the Bismarck Archipelago (wvhere there are extensive shallow lagoons), direct comparison of paired marine shell and wvood samples from Lapita contexts strongly indicates that the reservoir effect is very slight, requiring a AR correction of -350 vears to bring the sample pairs into agreement.