Journal of Archaeological Science: Reports 3 (2015) 513–524

Contents lists available at ScienceDirect

Journal of Archaeological Science: Reports

j ourna l homepage: ht tp : / /ees.e lsev ie r .com/ jas rep

Early settlement and subsistence on Tongatapu, Kingdom of Tonga:Insights from a 2700–2650 cal BP midden deposit

Geoffrey Clark a,⁎, Elle Grono a, Ella Ussher a, Christian Reepmeyer b

a Archaeology and Natural History, College of Asia and the Pacific, The Australian National University, Canberra, ACT0200, Australiab College of Arts, Society and Education, James Cook University, Australia

⁎ Corresponding author.E-mail address: geoffrey.clark@anu.edu.au (G. Clark).

http://dx.doi.org/10.1016/j.jasrep.2015.08.0052352-409X/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 30 December 2014Received in revised form 17 July 2015Accepted 8 August 2015Available online xxxx

Keywords:PacificColonizationTongaLapitaMarine subsistenceSea-level fall

The first significant Neolithic migration of people into the Pacific was the dispersal of Lapita culture at3200–2850 cal BP that involved the colonization of previously uninhabited and large island groups. Populationexpansion was accompanied by the introduction of domesticated plants and animals, but the location andcontent of Lapita deposits frequently suggests that early subsistence focused on the collection of wild resources.The tension between models that advocate Neolithic migration as sustained by agricultural yields and archaeo-logical data that disclose rapid dispersal and a reliance on indigenous resources is particularly acute in theKingdom of Tonga. Lapita settlements located on the palaeoshoreline of Tongatapu are associatedwith extensiveshell midden deposits suggesting the establishment of permanent settlements that were located in proximity tomarine resources before human predation or a declining sea-level fall led to resource collapse and site abandon-ment. Analysis of a shell midden sample associated with ancient burials from Talasiu on Tongatapu suggests asmall sedentary occupation that lasted a few generations ~2700–2650 cal BP. Site abandonment does not appearto have been caused by a decline in marine yields and identified starch from eight food plants is the first directevidence for a broad-spectrum mixed economy. While human predation of marine resources was substantial,sea-level fall is likely to have led to the closure of the Fanga 'Uta Lagoon at ~2500 cal BP resulting in the loss ofbenthic habitats and the reduction of economically important marine taxa that sustained and structured earlysedentism.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction and associated

The degree to which large-scale prehistoric population dispersalswere sustained by the collection of wild foods relative to agriculturalyields during the Neolithic is a long-standing issue in the Indo-Pacific(Bellwood, 2008; Donohue and Denham, 2010). Around 3500–3000 years ago human populations left Island Southeast Asia andbegan to colonize distant islands in the Pacific Ocean (Clark et al.2010). The oldest of these dispersals is associated with Lapita culturefound over 4500 km from Manus Island in the west to Samoa in theeast (Fig. 1). Lapita expansion is identified primarily by sites withdecorated ceramics marked with distinctive tooth-stamped and inciseddesigns, and while themigration chronology is under constant revision,current estimates suggest that most Pacific islands in the Lapita rangewere occupied 3200–2850 years ago (Denham et al. 2012; Burleyet al. 2015).

Dickinson (2003) noted that Lapita settlement of Oceania occurredafter a mid-Holocene hydro-isostatic highstand suggesting that coloni-zation of Pacific Islandswas aided by a sea-level fall of 1–2m.A declining

sea-level caused the emergence of coastalflats on high islands, protectedlagoons and stable islets that were new and attractive environments forhuman settlement (Dickinson et al. 1994; Dickinson, 2003:498; Nunn,2007a; Nunn, 2007b:121). The transformation of near-shore environ-ments facilitated human occupation of Pacific islands, yet sea-levelfall over several centuries also had the potential to reduce the pro-ductivity of littoral marine zones that were heavily utilized by earlyprehistoric people for subsistence such as embayments, reef flats, es-tuaries and lagoons (Nunn and Carson, 2015). Lapita sites typicallycontain substantial quantities of shellfish and finfish, especially theremains of inshore reef and lagoon taxa (Kirch, 1997; Kirch, 2010;Szabó and Amesbury, 2011), and many sites are located on formershorelines that are now stranded by sea-level fall and coastalprogradation (Best, 1984).

The effects of a sea-level fall on marine dietary regimes and settle-ment location is especially evident on the island of Tongatapu – thelargest island in the Kingdom of Tonga – where dense marine shelland ceramic midden deposits are concentrated along palaeoshorelinesformed during the mid-Holocene highstand (Spennemann, 1987:82;Spennemann, 1989a; Spennemann, 1989b:178; Dickinson, 2007).Introduced plants and animals were clearly important to the economyelsewhere in the Lapita range (Horrocks and Bedford, 2004; Horrocks

Fig. 1. Top. Distribution of the Lapita expansion and location of Tonga. Bottom. Location of the Talasiu site on Tongatapu and position of early sites marked by blue circles on thepalaeoshoreline (red dashed line). Black arrows indicate the position of ground water solution channels. (For interpretation of the references to color in this figure legend, the reader isreferred to the online version of this chapter.)

514 G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

and Nunn, 2007), but on Tongatapu researchers have repeatedlycommented on the significance of marine resources. Burley et al.(2001:102) noted that massive populations of Anadara antiquata andGafrarium tumidum/pectinatum were a critical variable in the early

settlement distribution and Dickinson (2007:184) has suggested thatthe rich shellfish resources and sheltered palaeoshorelines of northernTongatapu may have attracted: “one of the densest Lapita populationsin the ancient Pacific world”.

515G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

In this paper, we present an analysis of a shell midden sampleexcavated from the Talasiu site (~2700–2650 cal BP) on the Fanga'Uta palaeoshoreline to better understand the nature of late Lapita-Polynesian Plainware occupation and subsistence. Unlike previousexcavations of Tongan sites that dry-sieved deposits, we used flotationand took sediment samples to recover the full range of midden compo-nents.Microbotanical starch remains of introduced andnative foodflorawere identified as the presence and contribution of plant foods to earlysubsistence has not been directly established on Tongatapu. Resultsindicate that small short-term settlements like Talasiu were establishedin proximity to resource-rich localities and a freshwater source, withgardens established nearby. High rates of early population mobilityand site abandonment appear to have been influenced by social factorsrather than by resource depletion. Sea-level fall led to closure of theFanga 'Uta lagoon ~2500 cal BP which likely had a dramatic impact onthemarine foodswhich sustained and structured early sedentary settle-ments on Tongatapu.

1.1 Talasiu TP.2 excavation and sampling

The Talasiu site is located on the palaeoreef-limestone shorelinenorth of Lapaha where the central place of the Tu'i Tonga chiefdom/state was built (Clark et al. 2008). Fronted by land reclaimed duringthe expansion of Lapaha as the seat of the Tongan state, the core ofthe Talasiu site consists of dense shellfish and ceramic deposit ofapproximately 450 m2 bordered in the east by the main road and tothe north by a side road running toward the lagoon that has cut throughthe midden (Fig. 2). Talasiu was first excavated in 1957 by Jack Golsonwhonoted that the site likely had an area of ~2000m2, but due to devel-opment the site is now much reduced. The 1957 excavation was3 m × 1.5 m unit in containing abundant shellfish, pottery and branchcoral, but no fishbone (Golson, 1957, field notes, Fig. 2). (Spennemann,1986; Spennemann, 1989a; Spennemann, 1989b:415) recorded thesite as TO-Mu-2 and analysed a sample of shellfish excavated by Golsonobtaining a radiocarbon date of 2800± 70 BP on oyster shell, andmadesurface collections of plain and decorated pottery, lithics and humanbone. In 2008, eroding human remains from a cremation burial wereexcavated (TP.1) and dated to 2700–2500 cal BP (Valentin and Clark,2013). A 50 by 50 cm unit called TP.2 was excavated in 2011 to obtaina detailed record of midden composition at Talasiu (Fig. 2). The TP.2unit was excavated by trowel in 5 cm spits and by natural level withall material processed by flotation except for charcoal concentrationsthat were sampled in situ. Excavated material from each spit wasweighed with a spring balance and the sediment volume measured.The sediment was divided into a number of buckets with each filled toone-third of its total volume. Water was then added until a bucketwas 90% full, and the material was gently stirred by hand to separatesediment. Themixturewas left to allowparticle separation (deflocculate)for 5–10min. Light residues were collected by decanting thewater into aflotation sieve, and additionalwaterwas added to the heavy residue untilall light residues has been collected. The remaining heavy residue waswet sieved in 3 mm mesh and dried and separated by material type;marine shell, ceramics, bone, lithics and charcoal/organics. Shell frag-ments between 3 mm and 5.6 mm were collectively weighed for eachspit but not identified. Shells and shell fragments greater than 5.6 mmwere identified from shell morphology to the lowest possible taxonomiclevel. All material was returned to the Department of Archaeology andNatural History (ANH, The Australian National University) for analysiswith fishbone and marine shell identified using the fish and molluscreference collections of Australasian and Pacific taxa at ANH.

Samples for microbotanical analysis were taken from a 10 cm widecolumn cut into the TP.2 profile with a 5 cm block of sediment collectedfrom the column every 5 cm. Starch was extracted from 19 three sub-samples using standard laboratory techniques (Horrocks et al. 2004;Torrence, 2006). Extracted starch was compared to a comparativestarch collection of 22 Pacific edible plant taxa, including a range of

primary cultigens, species of secondary economic importance and fam-ine foods along with the White Potato (Solanum tuberosum) to controlfor modern contamination. Morphological analysis of comparative col-lection starch was part of Ussher’s PhD research (Ussher, 2015). Com-parative and archaeological starch was recorded by observation underbrightfield and polarized light using a Transmitted Light Zeiss micro-scope. Only grains with birefringent properties and an extinction crosswere included in the archaeological starch assemblage. Twenty eightstarch grain variables were recorded including 2D shape, 3D shape, ori-entation, length, width, length of the extinction cross, distance betweenthe arms of the cross, distance between the hila and the end, maximumangle of the cross, presence of surface features such as an equatorialgroove, lamellae, faceting (number and type), presence and type ofhilum features such as vacuole and fissuring, and surface texturebased on previous morphometric studies (Torrence et al. 2004; Wilsonet al. 2010; Allen and Ussher, 2013).

Light microscopy and image analysis of modern starch grains wasused to create two datasets based on grain orientation/view. Each starchgrainwas rolled to view the 3D shapewith sixmetric and 22 non-metricvariables recorded from eccentric/side-on and centric/end-on views.The eccentric dataset contained starch from 16 reference species andthe centric dataset had starch from 15 reference species with sevenspecies in both. The datasets were examined with a DiscriminantFunction Analysis (DFA) analysis using the PAST software. The purposeof the DFA was first, to categorize starchmorphotypes to a comparativecollection species and second, to classify archaeological starch (afterTorrence et al. 2004; Allen and Ussher, 2013). The first DFA of thecomparative starch collection using the eccentric dataset correctlyclassified 68% of starch grains to their species of origin and the percent-age of correct classifications in the centric dataset was 57%. The refer-ence starch classifications are similar to those found in previousmorphometric studies (Torrence et al. 2004; Wilson et al. 2010).

The DFA of comparative collection starch was then used to classifyarchaeological starch grains by entering the data from extracted starchgrains as ungrouped cases. Identification of archaeological starch toa reference species was made, however, only when more than fivearchaeological starch grains from a sample were assigned to a referencespecies, and the comparative species had a successful starch classifica-tion rate greater than 60%. The starch identifications made with DFAwere then checked by comparing images of the archaeological starchgrains with images of reference starch taken with light microscopyand Scanning Electron Microscopy (SEM) to confirm grain similarity.

1.2 Stratigraphy, radiocarbon dates and material culture

The TP.2 midden stratigraphy consisted of five cultural layerscontaining varying quantities of shellfish, ceramics, charcoal and boneresting on a sterile basal clay. The upper two midden layers (Layers1–2: 0–30/35 cm) were distinguished by the presence of fragmentedshell in a grey brown silty clay. In underlying layers there was lessshell breakage with a higher proportion of whole shells and large shellfragments in a red yellow clay matrix (Layers 3–4: 30/35–70/75 cm).Layer 5 (70/75–95 cm), the lowest cultural deposit, had crushed andfragmented shell along with charcoal fragments on top of Layer 6which was a compact red-brown clay devoid of cultural material. TheTP.2 stratigraphy is similar to that reported in previous investigationswith an upper layer of fragmented midden shell and middle deposit ofdense less fragmented shell under which were sparse cultural remainsover a sterile clay (Valentin and Clark, 2013).

Three samples of charred coconut endocarp from the basal, middleand upper sections of the stratigraphy were selected for AMS dating atthe Radiocarbon Dating Laboratory (Waikato University). Age resultswere calibrated with Calib Rev 6.1.0 using the SH terrestrial curveand are similar to previous age determinations obtained from TP.1(Table 1). Calibration curve flattening for the interval 2530–2420 BPresults inmultiple calibration curve intercepts and a substantialwidening

Fig. 2. Plan an elevation of the core area of the Talasiu midden deposit showing the location of the TP.2 excavation, the TP.1 deposit where a cremation burial was found (Valentin andClark, 2013) and the approximate location of Golson's, 1957 investigation.

516 G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

in the calibrated age for determinations falling within the range whichspans 2800–2350 cal BP. There were few dentate decorated sherds inthe lower levels of the site and ceramics from the upper levels werepredominantly undecorated. The transition from decorated to undeco-rated pottery in Tonga defines the transition from the Lapita phase tothe Polynesian Plainware phase, which is associated with the develop-ment of ancestral Polynesian society and a Proto-Polynesian linguisticsubstage (Kirch and Green, 2001). A recent high-resolution Bayesianoverlap model based on U/Th ages and radiocarbon determinationsshows the decorated-to-undecorated ceramic transition on Tongataputook place at 2700–2680 cal BP suggesting that Talasiu was occupied~2700–2650 cal BP (Burley et al. 2015).

The TP.2 artefact assemblage had 467 ceramic sherds, one Conus sp.bead, a grooved piece of hematite and a fragment of a small polishedstone adze/chisel. The ceramics were mostly plain body sherds with11 rims and six sherds marked with dentate-stamping. Vessel formswere simple subglobular jars and bowls (restricted and unrestricted)

similar to other Tongan assemblages (Spennemann, 1986:12–13;Poulsen, 1987:47–87; Connaughton, 2007:208). Mean body sherdthickness (n = 211) was 7.5 ± 1.9 mm with a few thicker sherdsthroughout the sequence. The ceramic fabric under light magnificationshows ferromagnesian grains (clinopyroxene, orthopyroxene), sparsevolcanic rock fragments and rare quartz consistent with derivationfrom tephra deposited by volcanic islands to northwest of Tongatapu(Dickinson, 2006). Chemical analysis of ten sherds using pXRF (calibrationin Clark et al. 2014) supports a local island arc source for the pottery(Fe = 60,340 ± 8060, Ti = 4910 ± 490, Co = 34 ± 6, Rb = 19 ± 5,Zr = 50 ± 7, Nb = 6 ± 0.4). Simple dentate stamped motifs and toolimpressing along the carination/collar rim were mostly confined tothe basal part of the deposit (70–90 cm) possibly indicating a shortlived Lapita presence that was followed by substantial midden deposi-tion in the early Polynesian Plainware era. The pXRF composition ofthe TP.2 chisel/adze flake indicates a Tongan source (Fe = 61,242 ±1146, Ti = 4957 ± 47, Rb = 10 ± 2, Zr = 48 ± 3, Nb = 2 ± 3) while

Table 1Talasiu TP.1 and new TP.2 radiocarbon dates.

Lab number Depth below surface Material CRA 13C Calibrated 95.4%

Wk-33572 TP.2, spit 415–20 cm

Charred coconut endocarp: Shell midden 2553 ± 25 −22.6 ± 0.2 2340–2740 cal BP

Wk-33573 TP.2, spit 1150–55 cm

Charred coconut endocarp: Shell midden 2448 ± 25 −22.5 ± 0.2 2340–2680 cal BP

Wk-33574 TP.2, spit 1885–90 cm

Charred coconut endocarp: Shell midden 2504 ± 25 −23.0 ± 0.2 2360–2700 cal BP

Wk-22999 TP.1: 40 cm Tridacna sp.: Cremation grave good 2767 ± 37 2.2 ± 0.2 2450–2870 cal BPWk-23000 TP.1: 65 cm Tridacna sp.: Cremation grave good 2726 ± 30 2.8 ± 0.2 2400–2800 cal BPWk-23001 TP.1: 80 cm Conus sp.: Cremation grave good 2682 ± 30 3.5 ± 0.2 2360–2740 cal BPWk-25461 TP.1: 40–65 cm Human bone: Cremation 2499 ± 31 −12.2 ± 0.2 2360–2700 cal BPWk-22876 TP.1: 91 cm Charcoal: Burial structure 2452 ± 30 −23.5 ± 0.2 2340–2690 cal BPWk-23002 TP.1: 80 cm Charcoal: Burial structure 2562 ± 30 −25.0 ± 0.2 2370–2740 cal BPWk-28234 TP.1: 55–60 cm Charred coconut endocarp: Shell midden 2473 ± 31 −21.3 ± 0.2 2350–2700 cal BPWk-28235 TP.1: 70–75 cm Charred coconut endocarp: Shell midden 2510 ± 30 −21.3 ± 0.2 2360–2700 cal BP

1. Marine shell determinations were calibrated with deltaR set at−158 ± 68 after Petchey and Clark (2011) with the Marine 09 calibration curve.2. Human bone was calibrated with the SHCal04 Southern Hemisphere Calibration as were determinations on carbonized wood and nut endocarp following advice from F. Petchey(Waikato Radiocarbon Dating Laboratory, Waikato University).

517G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

sourcing of other stone artefacts from the site demonstrates obsidianfrom northern Tonga and lithics from 'Eua and east Fiji consistent withsignificant populationmobility (Reepmeyer et al. 2012; Clark et al. 2014).

2. Results

2.1. Results: Vertebrates

The vertebrate assemblage from TP.2 (Table 2) consisted of 2476bones dominated by the remains of fish representing six taxa with anMNI of 11 (Diodontidae 1, Serranidae 1, Scaridae 4, Nemipteridae 1,Lethrinidae 3, Labridae 1). The taxa are associated with inshore reefand soft- bottomed substrates and are commonly caught using a varietyof fishing techniques, including angling, poisoning, nets, spears andtraps (Kirch, 1988:221–225; Clark and Szabó, 2009). Vertebra fromsmall sharks/rays (Elasmobranchii), dentary and vertebra from eels(Muraenidae) suggest capture by spearing. Fish hooks are a rare artefactin early Tongan sites and the few jabbing and rotating hooks made inPinctada sp. and Turbo sp. represent location specific responses to rawmaterial and fishing opportunity (Burley and Shutler, 2007; Szabó,

Table 2Talasiu, TP. 2 vertebrate remains. Fish identifiedwith cranial elements (maxilla, premaxilla,dentary, articulate, quadrate, pharyngeal clusters). Special elements were the spines ofDiodontidae (porcupinefish), vertebrae for Elasmobranchii (sharks/rays) and Muraenidae(moray eels), and teeth forMonotaxis granoculis (sea bream). Rat bones were attributed tothe Pacific rat (Rattus exulans) and identifiable bird bone to chicken (Gallus gallus).

Spit/depth Fish Rat Sharks/rays

Bird Eel Turtle/reptile

1: 0–5 cm 7 12: 5–10 cm 32 23: 10–15 cm 36 4 14: 15–20 cm 14 15: 20–25 cm 229 12 56: 25–30 cm 35 4 67: 30–35 cm 39 18: 35–40 cm 222 4 19: 40–45 cm 114 7 1 110: 45–50 cm 419 16 1 2 411: 50–55 cm 143 12 2 3 112: 55–60 cm 207 16 1 113: 60–65 cm 245 10 114: 65–70 cm 141 15 2 315: 70–75 cm 239 10 1 1 116: 75–80 cm 79 417: 80–85 cm 43 5 318: 85–90 cm 50 2 4 219: 90–95 cm 7 1NISP = 2476 2301 126 17 16 11 5MNI = 27 11 9 1 2 2 2

2007). No fishing gear has been found at Talasiu except for tiger cowriecaps (Cypraea tigris) – probably used in the composite octopus lure rig –and perforated Anadara sp. shells that may be net weights (Poulsen,1987; Burley and Shutler, 2007). Most of the diagnostic fish headbones (see Table 2) appeared to be from small individuals and to inves-tigate fish sizewemeasured the centrum length (upper longitudinal) of574 archaeological vertebra and compared the results to Lethrinidaeand Scaridae vertebra of knownbody length. The archaeological vertebrahad a mean length of 5.7 ± 1.7 mm suggesting that most of the fishcaught had a total body length of ~10–40 cm (Fig. 3). The identifiablebird bone (n = 2) was chicken (Gallus gallus) that along with the Poly-nesian rat (Rattus exulans) was introduced to Tongatapu in the Lapitaera. In short, the TP.2 bone assemblage does not include extinct birds,the giant iguana nor large amounts of turtle bone that are a component

Fig. 3.Centrum length inmm(upper longitudinal) of 574fish vertebra (mean5.7±1.7mm)from TP.2 compared to Lethrinidae and Scaridae vertebra from individuals of known size inthe ANH Fish Reference Collection (ANU). Lethrinidae: Monotaxis grandoculis, F0198,Length = 13 cm; Lethrinus reticulatus, F0147, length = 22 cm; Lethrinus erythracanthus,F0184, length = 32 cm; Lethrinus semicinctus, F0115, length = 59 cm. Scaridae: Scarusghobban, F0331, length = 24 cm; Leptoscarus vaigiensis, F0330, length = 28 cm, Scarusghobban, F0333, length = 42 cm.

518 G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

of early Lapita sites (Dye and Steadman, 1990; Burley et al. 2001;Steadman et al. 2002), supporting a late Lapita to Polynesian Plainwaresite age (Burley et al. 2015).

2.2 Results: Marine shellfish

Almost 69 kg of marine shell was recovered from TP.2 with threeaspects of the shell assemblage reported here. First, the taxonomic com-position of the marine shells was quantified using NISP, MNI, andspeciesweight (Claassen, 1998:106–107; Szabó, 2009:186–187). Second,the impact of human predation on local mollusc populations at Talasiuwas examined by an intra-midden analysis of shell size (Jerardinoet al. 1992:58; Mannino and Thomas, 2002:458). Two bivalve specieswere measured to examine shell size through time; A. antiquata andG. tumidum. These specieswere likely to have been intentionally selectedby humans for subsistence as they are a significant component of shellmiddens on Tongatapu (Poulsen, 1987:226–233; Spennemann, 1986:89–92; Spennemann, 1987; Spennemann, 1989a:126–134, 184–195). Third,palaeoenvironmental conditions at Talasiuwere examined through speciesecology to gain insight into the types of local aquatic habitats that wereexploited during shellfish gathering (e.g., Shackleton, 1988; Szabó, 2001,2009; Morrison and Addison, 2008). In addition, the attachment ofDendostrea cf. folium to a substrate was examined to identify the presenceof now extinct coral communities.

2.2.1 MNI, NISP and shell weightOver 91,000 shell fragments (NISP) weighing 67.8 kg were recov-

ered from the Talasiu shell sample representing 9242 individuals(MNI) from only 44 taxa. MNI was calculated for bivalves from thehighest number of left or right hinges, and for species where a left/right hinge identification was difficult the number of hinges/completeshells was divided by two (e.g. D. cf. folium,Modiolus sp., Chama pacifica,Venerupis galactites). GastropodMNIwas based on the number of apicesor body whorls/bases (e.g. Trochus sp.). The ten most abundant taxa asexpressed by MNI, NISP and weight are listed in rank order in Table 3.The shell sample is dominated by bivalve species, with gastropodscontributing minimally to overall abundance. The correlation betweenrank orders derived fromMNI, NISP andweightmeasures is statisticallysignificant (p b 0.01, see Table 3 for r2 values), indicating that the threequantification units may provide comparable results (after Grayson,1984). The shell sample is dominated by Dendostrea cf. folium, with5676 individuals (MNI) and over 62,000 shell fragments (NISP) contrib-uting ~60% of the assemblage. V. galactites, a small, light-shelled clamspecies that inhabits sandy localities, is the secondmost numerous spe-cies. The large-bodied muddy-sand bivalves A. antiquata, G. tumidumand G. pectinatum are also strongly represented. Two gastropod speciesranking equal eighth in MNI abundance are Gibberulus gibbosus and

Table 3The tenmost abundant mollusc taxa in the Talasiu TP. 2 midden assemblage ranked byMNI, NI1957 midden analyses. Note that Dendostrea cf. folium, A. antiquata and G. tumidum have the saOstrea sandvichiense by Spennemann (1989a: 195), and the identification has been revised by

Taxon Shell type MNI MNI rank NISP NISP rank

Dendostrea cf. folium B 5676 1 62,488 1Venerupis galactites B 1194 2 13,554 2Gafrarium tumidum B 560 3 2383 5Anadara antiquata B 417 4 2502 4Modiolus sp. B 293 5 872 6Pinctada sp. B 204 6 4891 3Gafrarium pectinatum B 141 7 593 8Gibberulus gibbosus G 121 8Clypeomorus brevis G 121 8Trisidos semtiorta B 104 10 720 7Tellina sp. B 536 9Pitar sp. B 370 10Chama sp. B

Shell type: B = bivalvia, G = gastropoda. MNI-NISP r2 = 0.839, p = 0.002, MNI-Weight r2 =

Clypeomorus brevis. However, the latter is a small species of theCeriithidae family that rarely grows beyond 25 mm in length and mayhave been of little economic importance. As the TP.2 sample is about0.1% of the site area is necessary to consider whether the sample isrepresentative of the midden as shell dumping is not homogeneous(Meehan, 1982; Claassen, 1998). A midden sample from Golson's,1957 excavation was analysed by Spennemann (1989a:181) and alarger shell collection associated with burnt and unburnt humanremains was recorded from TP.1 (Fig. 2). As shown in Table 3, theshell samples all have D. cf. folium, A. antiquata and G. tumidum as thethree main species by weight with V. galactites ranked four or five.Although the recover methods and level of collection were differentfor each excavation the consistent results indicate the TP.2 shell assem-blage is broadly characteristic of the midden deposit.

The Evenness (E) index was used to assess the relative proportionsof taxa in a sample and was calculated using a variant of Shannon'sDiversity Index (after Morrison and Cochrane, 2008:2393, see Grono,2012). In a highly even assemblage, individuals are distributed evenlyacross taxa, and in a highly uneven assemblage, a few taxa will domi-nate the assemblage. A relatively low E index of 0.42 indicates therewas high selectivity toward particular taxa at Talasiu. Most species arepresent in small quantities and the sample is dominated by three taxawith D. cf. folium, A. antiquata and Gafrarium sp. contributing over 73%of total MNI and 88% of the assemblage by weight indicating selectiveharvesting, which has also been reported from Lapita shell middensin Tonga, Fiji and the Reef/Santa Cruz Islands (Swadling, 1986;Spennemann, 1987; Szabó, 2009). High selectivity of shell gathers atTalasiu is illustrated by a comparison with shell collections fromLapita and post-Lapita sites in Fiji, which shows that assemblageswith a comparable MNI are significantly more diverse (Fig. 4).

2.2.2 Human predation and shell sizeThe size of shell valves sampled from the lower, middle and upper

spits of the midden sequence was examined with a one-way analysisof variance (ANOVA, p b .01). The mean shell size of A. antiquataand G. tumidum did not significantly change (Table 4, Fig. 5) and theabundance of both species is relatively even (Fig. 6). Previous studiesof shell midden sites on Tongatapu have shown that G. tumidum de-creases in size over time and the abundance of A. antiquata declines(Spennemann, 1987). The absence of an intra-midden decrease inshell size and the relatively consistent occurrence of A. antiquatathrough the midden sequence was unexpected given the evidence forsubstantial shellfish harvesting. Extrapolation of the TP.2 faunal assem-blage using the Talasiu site core (450 m2) and observation of the mid-den deposit from excavation of multiple human burials suggests thatthe remaining midden deposit contains around 61 metric tons of shellrepresenting consumption of ~8.3 million shellfish. There are several

SP andWeight. The TP.2 rank shell weight for the top five species is compared to TP.1 andme rank with V. galactites ranked 4 or 5. Dendostrea cf. foliumwas previously identified asK. Szabó (pers.comm 25/6/15).

Weight (g) Weight rank TP.2 Weight rank TP.1 Weight rank 1957

39,028 1 1 12529 4 5 49456 3 3 310,057 2 2 2

1316 51125 6 4286 10

774 7

331 8304 9 5

0.772, p = 0.009, NISP-Weight r2 = 0.818, p = 0.004.

Fig. 4. Graph showing the relationship between sample size (total MNI) and shell species diversity in Lapita and post-Lapita sites in Fiji (after Szabó, 2009: Fig. 111) in comparison withTalasiu which indicates high selectivity toward particular taxa. Note the Y-axis is a log scale.

519G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

possible reasons for a lack of bivalve size/abundance variation in theTalasiu sequence including the possibility that the site was initiallyestablished in the area where the TP.2 sample was collected. If thatwas the case then midden from heavily predated shell beds might belocated in younger parts of the site. An alternative is that the durationof occupation at Talasiu was so short that locally available marineresources were not depleted.

2.2.3 Palaeoecology of the Fanga 'Uta LagoonForty-four taxa were identified in the shell sample and the propor-

tions of aquatic habitats were calculated with Dendostrea cf. foliumomitted (Table 5) as it lives on coral patches in a range of marine envi-ronments (Grono, 2012). Sandy intertidal zones are strongly represent-ed (43.7%) and include the species G. gibbosus, Lambis lambis, Pitar sp.,Tawera sp., Trisidos semitorta and V. galactites. Muddy sands and man-groves are the secondmost numerous environment (30.4%), represent-ed by the species G. pectinatum, G. tumidum and Modiolus sp., amongothers. Species such as A. antiquata derive from mixed sandy environ-ments, including sandy and muddy sand substrates and contribute19% of the sample. Overall, intertidal species that live on hard substratesoccur in relatively small quantities, comprising less than 7% of theassemblage (e.g., Planaxis sulcatus, Chama sp., Turbo sp., Morula sp.).The lack of reef species suggests that shellfish gatherers did not venturefar out to lagoon reefs in search for valued specimens, but ratherexploited in large number a limited array of intertidal shellfish taxathat were locally available in the vicinity of Talasiu.

Table 4Shell measurements of A. antiquata and G. tumidum after Claassen (1998:109) measured using dselected from the upper (spits 1–4/6), middle (spits 7–10) and lower spits (spits 12/13–18).

Species Measurement Upper spits (1–6) n = 57

(mm) Mean stdev

A. antiquata Valve length 44.8 11.2A. antiquata Valve height 34.5 9.0A. antiquata Hinge length 26.5 7.3A. antiquata Hinge width 4.8 1.9

Measurement Upper spits (1–4) n = 100(mm) Mean stdev

G. tumidum Valve length 36.6 7.3G. tumidum Valve height 30.6 5.7G. tumidum Hinge length 15.5 3.5G. tumidum Hinge width 5.2 1.1

A small oyster, D. cf. folium, was the most abundant species in theTalasiu midden and many valves had been attached to gorgonians(sea whips) or staghorn (Acropora sp.) coral. As Mannino and Thomas(2002:456) note, shellfish species, like D. cf. folium that are sessile andgregarious are highly vulnerable to human predation. The collection ofD. cf. folium from nearshore coral communities close to Talasiu likelyinvolved breakage or removal of shell-encrusted coral and processingthe shells on site (Fig. 7). There are no soft/hard corals nor D. cf. foliumin the vicinity of Talasiu today (Zann, 1984:42). The presence ofA. antiquata in shell middens along the inner lagoon also provides evi-dence for different environmental conditions in the past as the localextinction of this species has been documented by Brock (1984:58),Poulsen (1987:228–229) and Spennemann (1987:83). A. antiquatafavours sandy and sometimes muddy substrates in intertidal reefflats of the near-shore environment and prefers saline conditions(Cernohorsky, 1971:215; Spennemann, 1989b),whileGafrarium speciesaremore suited to sheltered aquatic environments such as swamps andmudflats (Spennemann, 1987).

2.3 Results: Microbotanical

The abundance of shellfish in Lapita middens on Tongatapu sug-gested that horticultural cropsmay have been aminor part of early sub-sistence and that abundant marine foods on previously uninhabitedislands facilitated rapid human dispersal (Groube, 1971; Anderson,2001). However, no direct evidence for horticulture has yet been

igital callipers accurate to two decimal places (see Fig. 5). Measured shells were randomly

Middle spits (7–10) n = 77 Lower spits (13–18) n = 80

Mean stdev Mean stdev

44.6 10.9 44.3 11.234.2 8.5 34.9 8.226.1 7.9 26.1 7.44.8 1.9 5.2 2.1

Middle spits (7–10) n = 100 Lower spits (12–18) n = 100Mean stdev Mean stdev36.3 7.4 37.1 7.330.2 5.7 30.9 5.714.6 3.0 14.9 3.15.2 1.1 5.4 1.2

Fig. 5. Shellmeasurements ofA. antiquata andG. tumidum after Claassen (1998:109)measured using digital callipers accurate to twodecimal places. VL=Valve length, VH=Valve height,HL = Hinge length, HW = Hinge width (see Table 4).

Table 5Marine shell taxa identified in the Talasiu, TP.2 assemblage, their inferred habitat andshell weight.

Taxon Habitat Weight (kg)

Anadara antiquata Mixed sand/muddy sand 10.05Cerithium columna Sandy intertidal 0.03Cerithium nodulosum Coral reef intertidal 0.06Chama pacifica Coral reef intertidal 0.31Chlamys sp. Sandy intertidal b0.01Chicoreus brunneus Coral reef intertidal 0.04

520 G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

recovered from early archaeological sites located on Tongatapu'spalaeoshoreline.

To examine this issue, starchwas extracted fromTalasiu shellmiddensediments and identified by a morphometric analysis and a micro-scopic comparison of archaeological starch grains with reference starch(Ussher, 2015). Eight economic plants were identified in sedimentswith Araceae (Colocasia esculenta, Amorphophallus paeoniifolius,Cyrtosperma merkusii) and Dioscoreaceae (Dioscorea alata, Dioscoreabulbifera) families represented in the basal and upper parts levelsof the midden deposit as well as Inocarpus fagifer and Spondias dulcis(Table 6). Along with Musa sp., the plant species include core cultigensas well as ‘inferior’ or alternative food crops such as D. bulbifera and A.

Fig. 6. Abundance (MNI) of three major food shells in the Talasiu TP.2. midden. Note therelatively consistent numbers of A. antiquata and Gafrarium spp. while fluctuation inDendostrea cf. folium probably represents variation in the number of shell encrusted coralsreturned to the site for processing.

Chicoreus ramosus Coral reef intertidal 0.07Chitonidae Rocky shore/reef intertidal 0.01Clypeomorus brevis Coral reef intertidal 0.06Conus sp. Coral reef intertidal 0.05Cymatium (monoplex) pileare Coral reef intertidal 0.05Cymatium sp. A Coral reef intertidal 0.01Cyprea sp. Coral reef intertidal 0.04Euplica varians Coral reef intertidal b0.01Gafrarium pectinatum Muddy sands and mangroves 1.13Gafrarium tumidum Muddy sands and mangroves 9.45Gibberulus gibbosus Sandy intertidal 0.28Glycydonta (timoclea) marica Sandy intertidal b0.01Lambis lambis Sandy intertidal 0.14Lucinoma sp. Mixed sand/muddy sand 0.01Modiolus sp. Muddy sands and mangroves 0.22Monilea sp. Mixed sand/muddy sand b0.01Morula biconica Coral reef intertidal b0.01Morula sp. A coral reef intertidal b0.01Nassarius clarus Muddy sands and mangroves b0.01Nassarius sp. A Muddy sands and mangroves b0.01Nerita balteata Muddy sands and mangroves 0.03Nerita signata Rocky shore/reef intertidal 0.01Oliva sp. Sandy intertidal 0.02Ostrea sp. Mixed 39.01Periglypta puerpera Mixed sand/muddy sand 0.03Pinctada sp. Mixed sand/muddy sand 1.32Pitar sp. Sandy intertidal 0.33Planaxis sulcatus Rocky shore/reef intertidal 0.02Polinices sp. Mixed sand/muddy sand 0.03Quidnipagus palatam Mixed sand/muddy sand 0.18Tawera sp. Sandy intertidal 0.07Tellina sp. sandy intertidal 0.21Trisidos semitorta Sandy intertidal 0.77Trochus sp. Rocky shore/reef intertidal b0.01Turbo chrysostomus Coral reef intertidal 0.12Turbo setosus Rocky shore/reef intertidal 0.01Vasum turbinellum coral reef intertidal 0.07Venerupis galactites Sandy intertidal 2.53

Fig. 7. Dendostrea cf. folium shells showing they have been attached to soft (Gorgonians)and hard corals (Acropora sp.).

521G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

paeoniifolius (Ferdon, 1987:206). These results demonstrate for the firsttime that a mixture of arboriculture and root cropping practices wereused on Tongatapu ~2650 cal BP (Table 6, Fig. 8).

Table 6Classification of arch aeological starch grains extracted from the TP.2 sediments with species p

SpeciesAmorphophallus

paeoniifoliusColocasia esculenta

Cyrtosperma merkusii

SpitStarch count

Elephant foot yam

Common taro

Giant swamp taro

1 11

2 19

3 33

4 43

5 11

6 11

7 24

8 13

9 5

10 19

11 5

12 46

13 5

14 16

15 8

16 37

17 15

18 23

19 85

Control 7

Starch results from the Talasiu midden support microfossil evidencefor Lapita introduction of cultigens to the Tongan archipelago. Fall(2005, 2010) and Fall and Drezner (2011, 2013) document the firstevidence for Colocasia in pollen cores taken from the Avai'o'vunaSwamp and the Ngofe Marsh in the Vava'u Group, and on 'Eua in theTongatapu Group by 2600 BP. Other introductions at this time includedCasuarina equisetifolia, Cordyline fruticosa, and Pometia pinnata, whilepollen increases in indigenous species such as Canarium harveyi, Cocosnucifera and Pandanus tectorius suggest cultivation for food. When themicrobotanical record is combined with faunal data, an overall pictureof a broad-spectrum late-Lapita mixed economy emerges. The mixed-cropping horticultural component supplemented, and may haveequaled, the contribution of marine food.

3. Discussion

There are at least 18 sites with Lapita decorated ceramics located onpalaeoshorelines in northern Tongatapu (Fig. 1) demonstrating thatproximity to marine resources was a key determinant of settlementlocation (Burley et al. 2001:100). What these sites represent is unclearwith the possibility that they were permanently occupied villagesoccupied by people with a restricted marine economy as suggested byGroube (1971) or sites were camps occupied by opportunistic foragers(Spennemann, 1989a:131). The Talasiu site is intriguing as it containsa range of material culture items (pottery, ceramics, shell ornaments,lithics) in addition to human burials in a dense shell midden consistentwith a permanent settlement. Remains of domesticated and indigenousplants indicate that gardens were likely to have been located south of

resence indicated by a black box (see Fig. 8).

Dioscorea alata

Dioscorea bulbifera

Inocarpus fagifer

Musa spp.Spondias

dulcis

Common yam

Bitter yam

Tahitian chestnut

Banana Ambarella

Fig. 8. Archaeological and reference starch: (A) archaeological starch identified as Dioscorea bulbifera (B) modern starch of D. bulbifera — cone-shaped and triangular starch grains(5–30 μm) with fissuring at the hila, (C) archaeological starch identified as Cyrtosperma merkusii, (D) modern starch of C. merkusii— small spherical grains (5–26 μm) with 2–3 pressurefacets, andpolygonal shaped grainswithflat faceting and possiblefissuring, (E) archaeological starch identified asMusa sp., (F)modern starch ofMusa sp.— cylinder, reniform, pyriformandsub-ovate shaped grains (8–61 μm)with small vacuoles and lamellae, (G) archaeological starch identifiedas Inocarpus fagifer, (H)modern starch of I. fagifer— sub-spherical or dome-shapedgrains (6–21 μm) with 1–3 pressure facets and large open vacuoles, (I) archaeological starch identified as Amorphophallus paeoniifolius, (J) modern starch of A. paeoniifolius — domed,spherical or poygonal grains (12–36 μm) with flat and curved single or multiple faceting and common longitudinal fissuring, (K) archaeological starch identified as Spondias dulcis,(L) modern starch of S. dulcis — spherical or ovate grains (13–32 μm) with a small or large vacuole, with possible lamellae and stellate or transverse fissuring, (M) archaeological starchidentified as Dioscorea alata, (N) modern starch of D. alata — ovate or conical grains (10–25 μm) with compression common and possible lamellae, (O) archaeological starch identifiedas Colocasia esculenta, (P) modern starch of C. esculenta — very small (3–8 μm) polygonal or spherical grains with flat single and multiple faceting.

522 G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

the shoreline to reduce the harmful effect of salt spray on cultigens: theTalasiu site clearly represents more than just the by-product of marineresource procurement and consumption.

Tongatapu is a limestone island without any standing sourcesof freshwater and Talasiu is located close to a ground-water solutionchannel. Nine solution channels/springs have been identified exitingthe palaeoshoreline (Spennemann, 1989a:201) and six have Lapitasites identified in their near vicinity demonstrating the importance offreshwater sources during the colonization era (Fig. 1). Marine shellfishwere a clumped and predictable resource at Talasiu that was probablyserially exploited by a few families until group mobility was stimulatedeither by ‘push’ factors such as declining subsistence returns or ‘pull’factors such as knowledge of new resource zones or a decision to joinrelated groups living elsewhere. Shellfish size did not, apparently,decrease during the occupation of Talasiu and geochemical results indi-cate significant mobility within the Tongan archipelago including toNiuatoputapu 600 km to the north and to east Fiji 400 km to the west.

It can be proposed that in Lapita times behavioural sedentism consistedof short-lived settlements adjacent to rich marine patches, often inproximity to a freshwater source with gardens established nearby.The duration of occupation at Talasiu is difficult to determine with thepresence of multiple burials at the site and a complicated mortuarysequence that included removal and reburial of skeletal elements fromprimary burial features perhaps indicating a time depth of one to twodecades at least.

Site abandonment does not appear to have been driven completelyby increasing subsistence costs of gathering wild foods although thesewere increasing. At the same time the early population of Tongatapuwas harvesting large quantities of marine shellfish it was also rapidlydepleting stocks of local vertebrates resulting in the rapid extinction of26 species of land and sea birds, several species of bat and a large bodiediguana, with evidence for significant predation of marine turtles (Dyeand Steadman, 1990; Koopman and Steadman, 1995; Steadman et al.2002; Steadman 2006). The decline of local fauna could not be fully

523G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

replaced by introduced animal protein as there is no evidence for the pig(Sus scrofa) and dog (Canis familiaris) during the colonization phase inTonga, and only the introduced Pacific rat (Rattus exulans) and chicken(G. gallus) are securely associated with Lapita occupation (Clark et al.2013). Tongatapu was first settled at 2840 cal BP (Burley et al. 2015),and Talasiu shows the persistence some two hundred years later of asubsistence pattern based on small-to-medium settlements that wereestablished in order to exploit large shell fish beds.

Marine subsistence returns were not only declining from the inten-sive exploitation of shellfish by a growing population, but also from sea-level fall which had a dramatic effect on coastal ecology. Both Dickinson(2007) and Spennemann (1987) have shown that the two arms of theFanga 'Uta lagoon were formerly open embayments harbouring deeperwater protected by offshore islets. As sea level fell, the lagoon becameincreasingly enclosed from expansion of the Nuku'alofa peninsula andthe emergence of new coastal flats. In Lapita times the Fanga 'Utaembayment had an area of ~53 km2, but the lagoon area is now only~28 km2 (Zann et al. 1984). Themodern Fanga 'Uta Lagoon is a shallow,sheltered and mangrove fringed body of water with an average depthof 1.2–1.4m that is largely closed from tidal influence. Thewater qualityin the lagoon is brackish with a mean residence time of 23 days(Kimmerer, 1984; Zann, 1984:45). Only two shellfish species occurin considerable abundance in the inner lagoon; G. tumidum andQuidnapigus palatam, both of which are muddy sand species thatlive in brackish sheltered environments (Spennemann, 1987:83).Toward the lagoon entrance, tidal access and the flow of seawaterincreases. As a result mollusc taxonomic diversity also increases, withthe occurrence of bivalve species such as A. antiquata, Fragum unedo,Tellina sp., Periglypta sp., Lucina sp., and Pecten sp. (Brock, 1984:58).Additional evidence for substantial environmental change is thatDendostrea. cf. folium and A. antiquata which prefer ocean saline con-ditions were present in large number in the vicinity of Talasiu~2650 years ago, but these species are now restricted to the lagoonentrance and open marine environments outside the lagoon (Zann,1984:41). Similarly, D. cf. folium was attached to gorgonians/Acroporasp. showing that extensive coral beds once lived in the vicinity of thesite. Such corals play a paramount role in shaping benthic communitiesby increasing local biodiversity and promoting nursery effects (Gili andComa, 1998).

The Talasiu midden results support earlier work that highlights therole of sea-level fall in the establishment of a brackish lagoon and theassociated loss of a rich marine ecosystem (Spennemann, 1987). Atthe time of human settlement ~2850 cal BP the residual mid-Holocenehighstand on Tongatapu is estimated at 1.8 ± 0.4 m declining to0.9 ± 0.3 m around 2000 years ago (Dickinson et al. 1994:Fig. 5;Dickinson et al. 1999:697). Sea-level fall in the Central Pacific is under-stood to be monotonic and linear (Dickinson, 2003), but little detailedwork has been made to examine this assumption, particularly inTonga which experienced isostatic rebound and tectonic movements.Data summarized by Woodroffe and Horton (2005) indicate a compli-cated post-highstand pattern in some parts of the Indo-Pacific withsea-level falling rapidly rather than gradually (see also Bird et al.2010). Nunn and Carson (2015) in a recent review suggest that recon-figuration of marine-focused subsistence economies in the westernPacific as a result of sea-level fall dates to 2570 cal BP. On Tongatapu,archaeological results and palynological analysis of mangrove sedi-ments suggest that a critical threshold was reached with the closure ofthe Fanga ‘Uta lagoon passage at the high-tide level by 2500 cal BP(Ellison, 1988; Spennemann, 1997). From this time on near-shore hab-itats shoaled, salinity levels and suspended oxygen decreased and silta-tion increased – all of which would have had significant negativeimpacts on the quality and quantity of benthic habitats required bymacroinvertebrates, shellfish andfinfish.While bivalves and gastropodswere collected and remained part of the prehistoric Tongan diet aftersea level fall: “… shellfish dumping, at least on any large scale, ceasedby the end of the Lapita period” (Spennemann, 1989a:245).

4. Conclusion

Shallow estuaries, embayments, reefflats and lagoons are commonlyassociated with Lapita sites in Fiji and Tonga in large part because theyheld rich and easily accessible aquatic resources (Best, 1984; Burleyet al. 2001; Clark et al. 2001). Such environments were particularlysusceptible to the effects of sea-level fall that – in tandem with humanpredation and infilling from sediments (land clearance, volcanic ash,rafted pumice) – progressively diminished the contribution of marinefoods to prehistoric subsistence. It is notable that a mid-Holocenehighstand was experienced across all of the major island groups colo-nized by Lapita groups (Dickinson et al. 1994:Fig. 4) and the magnitudeof sea level fall was significantly greater than the 60–80 cm dropthought to accompany the AD 1300 event in the Pacific (Nunn, 2007a,b). The decline of concentratedmarine resources on Tongatapu suggeststhat sea-level fall may have been awidespread factor in shifting a Lapitacolonization mode in which settlements were preferentially locatedbeside highly ranked marine resources to one in which the controlof land and management of terrestrial food plants was central tosubsistence.

Acknowledgements

We thank the Nobles of Lapaha, the Honourable Kalaniuvalu-Fotofiliand PrincessMele Siu'ilikutapu Kalaniuvalu-Fotofili and Lapaha historianMr. Nivaleti Melekiola for their permission and assistance and gratefullyacknowledge the support of Lord Vaea (Chair of the Tongan TraditionsCommittee and Minister of Internal affairs). Field work was supportedby an Australian Research Council Future Fellowship (FT0990591)grant to Clark, and was carried out under an agreement between theTongan Government, Noble Honourable Kalaniuvalu-Fotofili and Clarkto document archaeological sites for nomination to the World HeritageList.We thank David Burley (Simon Fraser University) and the reviewersfor valuable information and comments on an earlier version of themanuscript.

References

Allen, M., Ussher, E., 2013. Starch analysis reveals prehistoric plant translocations andshell tool use, Marquesas Islands, Polynesia. J. Archaeol. Sci. 40, 2799–2812.

Anderson, A., 2001. Mobility models of migration. In: Clark, G.R., Anderson, A.J., Vunidilo, T.(Eds.), The Archaeology of Lapita Dispersal in Oceania: Papers From the Fourth LapitaConference. The Australian National University, Canberra. Terra Australis, pp. 15–23 17.

Bellwood, P., 2008. Austronesian cultural origins: out of Taiwan, via the Batanes Islands,and onwards to western Polynesia. In: Sanchez-Mazas, A., Blench, R., Ross, M.D.,Peiros, I., Lin, M. (Eds.), Past Human Migrations in East Asia: Matching Archaeology,Linguistics and Genetics. Routledge, London, pp. 23–39.

Best, S. 1984. Lakeba: theprehistory of a Fijian island.UnpublishedPhD thesis, AnthropologyDepartment, University of Auckland.

Bird, M.I., Austin, W.E.N., Wurster, C.M., Fifield, L.K., Mojtahid, M., Saregeant, C., 2010.Punctuated eustatic sea-level rise in the early mid-Holocene. Geology 38 (9),803–806.

Brock, R.E., 1984. Fish and benthic communities. In: Zann, L.P., Kimmerer, W.J., Brock, R.E.(Eds.), The Ecology of the Fanga'uta Lagoon, Tongatapu, TongaSea Grant CooperativeReport UNIHI-SEAGRANT-CR-84–04. The University of Hawaii Sea Grant CollegeProgram, University of Hawaii, Honolulu, pp. 47–78.

Burley, D.V., Shutler Jr., R., 2007. Ancestral Polynesian fishing gear: archaeological insightsfromTonga. In: Anderson,A., Green, K., Leach, F. (Eds.),Vastly Ingenious: The Archaeologyof Pacific Material Culture in Honour of Janet M. Davidson. University of Otago Press,Dunedin, pp. 155–172.

Burley, D.V., Dickinson, W.R., Barton, A., Shutler Jr., R., 2001. Lapita on the periphery. Newdata on old problems in the Kingdom of Tonga. Archaeol. Ocean. 36, 89–104.

Burley, D.V., Edinborough, K., Weisler, M., Zhao, J.-x., 2015. Bayesian modeling and chro-nological precision for Polynesian settlement of Tonga. PLoS One 10 (3), e0120795.http://dx.doi.org/10.1371/journal. pone.0120795.

Cernohorsky, W.O., 1971. Marine Shells of the Pacific vol. 2. Pacific Publications, Sydney.Claassen, C., 1998. Shells: Cambridge Manuals in Archaeology. Cambridge University Press,

Cambridge.Clark, G., Szabó, K., 2009. The fish bone remains. In: Clark, G.R., Anderson, A.J. (Eds.), The

Early Prehistory of Fiji. Terra Australis 31. ANU E Press, The Australian National Univer-sity, Canberra, pp. 213–230.

Clark, G., Anderson, A., Matararaba, S., 2001. The Lapita site at votua, northern Lau islands,Fiji. Archaeol. Ocean. 36, 134–145.

Clark, G., Burley, D., Murray, T., 2008. Monumentality in the development of the Tonganmaritime chiefdom. Antiquity 82 (318), 994–1004.

524 G. Clark et al. / Journal of Archaeological Science: Reports 3 (2015) 513–524

Clark, G., Petchey, P., Hawkins, S., Reepmeyer, C., Smith, I., Masse, W.B., 2013. Distributionand extirpation of pigs in pacific islands: a case study from Palau. Archaeol. Ocean. 48(3), 141–153.

Clark, G., Petchey, F., Winter, O., Carson, C., O'Day, P., 2010. New radiocarbon datesfrom the Bapot-1 site in Saipan and Neolithic dispersal by stratified diffusion. J. Pac.Archaeol. 1 (1), 21–35.

Clark, G., Reepmeyer, C., Melekiola, N., Woodhead, J., Dickinson, W.R., Martinsson-Wallin,H., 2014. Stone tools from the ancient Tongan state reveal prehistoric interactioncentres in the Central Pacific. Proc. Natl. Acad. Sci. U. S. A. 111 (29), 10491–10496.

Connaughton, S.P., 2007. Can we dig it? Archaeology of Ancestral Polynesian society inTonga: a first look from Falevai. In: Bedford, S., Sand, C., Connaughton, S.P. (Eds.),Oceanic Explorations: Lapita and Western Pacific Settlement. Terra Australis 26. ANUE Press, The Australian National University, Canberra, pp. 199–212.

Denham, T., Bronk Ramsey, C., Specht, J., 2012. Dating the appearance of Lapita pottery inthe Bismarck Archipelago and its dispersal to Remote Oceania. Archaeol. Ocean. 47,39–46.

Dickinson, W.R., 2003. Impact of mid-Holocene hydro-isostatic highstand in regional sealevel on habitability of islands in Pacific Ocean. J. Coast. Res. 19 (3), 489–502.

Dickinson, W.R., 2006. Temper sands in prehistoric Oceanian pottery: geotectonics, sedimen-tology, petrography, provenance. The Geological Society of America Special Paper 206Boulder, Colorado.

Dickinson, W.R., 2007. Paleoenvironment of the Lapita sites on Fanga ‘Uta Lagoon,Tongatapu, Kingdom of Tonga. In: Bedford, S., Sand, C., Connaughton, S.P. (Eds.), Oceanicexplorations: Lapita and western Pacific Settlement. Terra Australis 26. ANU E Press, TheAustralian National University, Canberra, pp. 177–186.

Dickinson, W.R., Burley, D.V., Shutler Jr., R., 1994. Impact of hydro-isostatic Holocene sea-level change on the geologic context of island archaeological sites, Northern Ha'apaiGroup, Kingdom of Tonga. Geoarchaeology 9 (2), 85–111.

Dickinson, W.R., Burley, D.V., Shutler Jr., R., 1999. Holocene palaeoshoreline recordin Tonga: geomorphic features and archaeological implications. J. Coast. Res. 15 (3),682–700.

Donohue, M., Denham, T., 2010. Farming and language in island southeast Asia: reframingAustronesian history. Curr. Anthropol. 51 (2), 223–256.

Dye, T.S., Steadman, D.W., 1990. Polynesian ancestors and their animal world. Am. Sci. 78,207–215.

Ellison, J.C. 1988. Holocene sea Level Record of Tongatapu, Kingdom of Tonga, from PollenAnalysis of Mangrove Sediments. Unpublished MSc Thesis, Department of Geogra-phy, Simon Fraser University.

Fall, P., 2005. Vegetation change in the coastal-lowland rainforest at Avai'o'vuna Swamp,Vava'u, Kingdom of Tonga. Quat. Res. 64, 451–459.

Fall, P., 2010. Pollen evidence for plant introductions in a Polynesian tropical islandecosystem, Kingdom of Tonga. In: Haberle, S., Stevenson, J., Prebble, M. (Eds.), AlteredEcologies. Fire, Climate and Human Influence on Terrestrial Landscapes. Terra Australis32, ANU E Press, The Australian National University, Canberra, pp. 257–275.

Fall, P., Drezner, T., 2011. Plant dispersal, introduced species, and vegetation change in theSouth Pacific Kingdom of Tonga. Pac. Sci. 65, 143–156.

Fall, P., Drezner, T., 2013. Species origins, dispersal, and island vegetation dynamics in theSouth Pacific. Ann. Assoc. Am. Geogr. 103, 1041–1057.

Ferdon, E.N., 1987. Early Tonga: as the Explorers Saw it 1616–1810. The University ofArizona Press, Tuscon.

Gili, J.M., Coma, R., 1998. Benthic suspension feeders: their paramount role in littoralmarine food webs. Trends Ecol. Evol. 13, 316–321.

Golson, J. 1957. FieldNotes fromSurvey and Excavations inTonga. Unpublishedmanuscriptin possession of the first author.

Grayson, D.K., 1984. Quantitative Zooarchaeology. Academic Press, Orlando, Florida.Grono, E. 2012. And then therewere none. A case-study ofmaritime adaptation and resource

use in prehistoric Tongatapu, Tonga. Unpublished BPhil (Hons) thesis, School ofArchaeology and Anthropology, The Australian National University.

Groube, L., 1971. Tonga, Lapita pottery, and Polynesian origins. J. Polynesian Soc. 80 (3),278–316.

Horrocks, M., Bedford, S., 2004. Microfossil analysis of Lapita deposits in Vanuatu revealsintroduced Aracea (aroids). Archaeol. Ocean. 39, 67–74.

Horrocks, M., Nunn, P.D., 2007. Evidence for introduced taro (Colocasia esculenta) andlesser yam (Dioscorea esculenta) in Lapita-era (c. 3050–2500 cal. yr. BP) depositsfrom Bourewa, south-west Viti Levu Island, Fiji. J. Archaeol. Sci. 34, 739–748.

Horrocks, M., Shane, P.A., Barber, I.G., D'Costa, D.M., Nichol, S.L., 2004. Microbotanicalremains reveal Polynesian agriculture and mixed cropping in early New Zealand.Rev. Palaeobot. Palynol. 131, 147–157.

Jerardino, A., Castilla, J.C., Ramírez, J.M., Hermosilla, N., 1992. Early coastal subsistencepatterns in Central Chile: a systematic study of the marine-invertebrate fauna fromthe site of Curaumilla-1. Lat. Am. Antiq. 3 (1), 43–62.

Kimmerer,W.J., 1984. Circulation and Hydrology. In: Zann, L.P., Kimmerer,W.J., Brock, R.E.(Eds.), The Ecology of the Fanga'uta Lagoon, Tongatapu, TongaSea Grant CooperativeReport UNIHI-SEAGRANT-CR-84–04. The University of Hawaii Sea Grant CollegeProgram, University of Hawaii, Honolulu, pp. 11–26.

Kirch, P.V., 1988. Niuatoputapu: The Prehistory of a Polynesian Chiefdom. Thomas BurkeMemorial Washington State Museum, Seattle.

Kirch, P.V., 1997. The Lapita Peoples: Ancestors of the Oceanic world. Blackwell Press,Oxford, p. 14.

Kirch, P.V., 2010. Between garden and reef: Lapita ecology and subsistence. In: Sand, C.,Bedford, S., Chambonniere, S. (Eds.), Lapita: Ancestres Oceaniens. Chapter XV. Muséedu quai Branly, Paris, pp. 252–267.

Kirch, P.V., Green, R.C., 2001. Hawaiki, Ancestral Polynesia. An Essay in Historical Anthro-pology. Cambridge University Press, Cambridge.

Koopman, K.F., Steadman, D.W., 1995. Extinction and biogeography of bats on ‘Eua,Kingdom of Tonga. Am. Mus. Novit. 3125, 1–13.

Mannino, M.A., Thomas, K.D., 2002. Depletion of a resource? The impact of prehistorichuman foraging on intertidal mollusc communities and its significance for humansettlement, mobility and dispersal. World Archaeol. 33 (3), 452–474.

Meehan, B., 1982. Shell bed to Shell Midden. Globe Press, Melbourne.Morrison, A.E., Addison, D.J., 2008. Assessing the role of climate change and human

predation on marine resources at the Fatu-ma-Futi site, Tutuila Island, AmericanSamoa: an agent based model. Archaeol. Ocean. 43, 22–34.

Morrison, A.E., Cochrane, E.E., 2008. Investigating shellfish deposition and landscapehistory at Natia Beach site, Fiji. J. Archaeol. Sci. 35, 2387–2399.

Nunn, P.D., 2007a. Holocene sea-level change and human response in Pacific Islands.Earth Environ. Sci. Trans. R. Soc. Edinb. 98, 117–125.

Nunn, P.D. 2007b. The AD1300 event in the Pacific Basin. Geogr. Rev. 97(1): 1–23.Nunn, P.D., Carson, M.T., 2015. Sea-level fall implicated in profound societal change about

2570 cal yr BP (620 BC) inwestern Pacific island groups. Geogr. Environ. http://dx.doi.org/10.1002/geo2.3.

Petchey, F., Clark, G., 2011. Tongatapu hardwater: Investigation into the 14C marine reser-voir offset in lagoon, reef and open ocean environments of a limestone island. Quat.Int. 6 (6), 539–549.

Poulsen, J., 1987. Early Tongan prehistory, 2 vols. Terra Australia 12, Department of Prehis-tory, Research School of Pacific Studies, The Australian National University, Canberra.

Reepmeyer, C., Clark, G., Sheppard, P., 2012. Obsidian source use in Tongan prehistory:new results and implications. J. Island Coastal Archaeol. 7 (2), 255–271.

Shackleton, J., 1988. Marine Molluscan Remains from Franchthi Cave. Indiana UniversityPress, Bloomington and Indianapolis.

Spennemann, D.H.R., 1986. Archaeological fieldwork in Tonga 1985-1986. PreliminaryReport on the Archaeological Activities During the 1985/86 field Season of theTongan Dark Ages Research Programme. Department of Prehistory, The AustralianNational University, Canberra.

Spennemann, D.H.R., 1987. Availability of shellfish resources on prehistoric Tongatapu,Tonga: effects of human predation and changing environment. Archaeol. Ocean. 22(3), 81–96.

Spennemann, D.H.R. 1989a. Ata'o Tonga mo’ata’a Tonga: early and later prehistory of theTongan Islands. Volume I. A study in settlement and subsistence patterns, withspecial emphasis on Tongatapu. Unpublished PhD thesis, Department of Prehistory,Research School of Pacific Studies, Australian National University, Canberra.

Spennemann, D.H.R. 1989b. Ata'o Tonga mo’ata’a Tonga: early and later prehistory of theTongan Islands. Volume II.2. Supportive studies. Unpublished PhD thesis, Departmentof Prehistory, Research School of Pacific Studies, Australian National University,Canberra.

Spennemann, D.H.R., 1997. A Holocene Sea-level history for Tongatapu, Kingdom ofTonga. In: Sherwood, A.M. (Ed.), Coastal and Environmental Geoscience Studies ofthe Southwest Pacific Islands. SOPAC Technical Bulletin. 9, pp. 115–151.

Steadman, D.W., 2006. Extinction and Biogeography of Tropical Pacific Birds. The Univer-sity of Chicago Press, Chicago and London.

Steadman, D.W., Pregill, G.K., Burley, D.V., 2002. Rapid extinction of iguanas and birds inPolynesia. Proc. Natl. Acad. Sci. U. S. A. 99 (6), 3673–3677.

Swadling, P., 1986. Lapita shellfishing: evidence from sites in the Reef/Santa Cruz Group,Southeast Solomons. In: Anderson, A. (Ed.), Traditional Fishing in the Pacific: Ethno-graphical and Archaeological Papers From the 15th Science CongressPacific Anthropo-logical Records No. 37. Bernice P. Bishop Museum, Honolulu, pp. 137–150.

Szabó, K., 2001. The reef, the beach, and the rocks: an environmental analysis of molluscremains from Natunuku, Viti Levu, Fiji. In: Clark, G., Anderson, A.J., Sorovi-Vunidilo, T.(Eds.), The Archaeology of Lapita Dispersal in Oceania: Papers From the Fourth LapitaConference, June 2000. Terra Australis 17, Pandanus Books, Research School of Pacificand Asian Studies, The Australian National University, Canberra, pp. 159–166.

Szabó, K., 2007. An assessment of shell fishhooks of the Lapita cultural complex. In:Anderson, A., Green, K., Leach, F. (Eds.), Vastly Ingenious: The Archaeology of PacificMaterial Culture in Honour of Janet M. Davidson. University of Otago Press, Dunedin,pp. 227–241.

Szabó, K., 2009. Molluscan remains from Fiji. In: Clark, G., Anderson, A. (Eds.), The EarlyPrehistory of Fiji. Terra Australis 31, ANU E press, The Australian National University,Canberra, pp. 183–211.

Szabó, K., Amesbury, J.R., 2011.Molluscs in a world of islands: the use of shellfish as a foodresource in the tropical island Asia-Pacific region. Quat. Int. 239 (1–2), 8–18.

Torrence, R., 2006. Starch in sediments. In: Torrence, R., Barton, H. (Eds.), Ancient StarchResearch. Left Coast Press, Inc, Walnut Creek, California, pp. 145–176.

Torrence, R., Wright, R., Conway, R., 2004. Identification of starch granules using imageanalysis and multivariate techniques. J. Archaeol. Sci. 31, 519–532.

Ussher, E. 2015. Agricultural development in Tongan prehistory: an archaeobotanicalperspective. Unpublished PhD thesis, The Australian National University.

Valentin, F. and Clark, G. 2013. Early Polynesian mortuary behaviour at the Talasiu site,Kingdom of Tonga. J. Pac. Archaeol. 4(1): 1–14.

Wilson, J., Hardy, K., Allen, R., Copeland, L., Wrangham, R., Collins, M., 2010. Automatedclassification of starch granules using supervisedpattern recognition ofmorphologicalproperties. J. Archaeol. Sci. 37, 594–604.

Woodroffe, S.A., Horton, B.P., 2005. Holocene sea-level changes in the Indo-Pacific. J. AsianEarth Sci. 25 (1), 29–43.

Zann, L.P., 1984. Corals: a clue to Fanga'uta's past. In: Zann, L.P., Kimmerer, W.J., Brock,R.E. (Eds.), The ecology of the Fanga'uta Lagoon, Tongatapu, TongaSea Grant Coopera-tive Report UNIHI-SEAGRANT-CR-84–04. The University of Hawaii Sea Grant CollegeProgram, University of Hawaii, Honolulu, pp. 39–46.

Zann, L.P., Kimmerer, W.J., Brock, R.E. (Eds.), 1984. The Ecology of the Fanga'uta Lagoon,Tongatapu, Tonga.Sea Grant Cooperative Report UNIHI-SEAGRANT-CR-84–04. TheUniversity of Hawaii Sea Grant College Program, University of Hawaii, Honolulu.