- 273

Kiritimati (Christmas) Island is one of the Line Islands which stretch 2500 km northwest from French Polynesia towards Hawai'i. They are interesting strategically in relation to the prehistory of central Pacific colonisation (Irwin 1992). All of them, and most other islands in the equatorial Pacific, were uninhabited at European contact, although many contained remains of earlier settlement (Emory 1934) and constitute the largest group of the “mystery” islands of Polynesia (Bellwood 1978, Kirch 1988).

The Line Islands could have been settled from any of a number of directions at different times, including during the Lapita expansion about 3000 years ago, or in the period when colonisation expanded into Niue and the northern Cooks about 2000 years ago. Three other scenarios seem most plausible. First, the Line Islands might have been settled during the colonisation period of East Polynesia, in the first millenium A.D., either by island-hopping northward from the Societies or Tuamotus or by a relatively easy passage on the Southeast trades from the Marquesas—quite possibly during the period when Hawai'i was first settled. Secondly, the islands might have been colonised later, during two-way voyaging between Hawai'i and central East Polynesia. Thirdly, the Line Group as a whole might have functioned as a net, catching voyagers from many sources, with the larger islands being settled from different places at different times. This third scenario is the multiple-castaway hypothesis, favoured for Kiritimati Island by Emory (1934:4). Sorting between these or other propositions of - 274 colonisation prehistory for the Line Islands will require more extensive archaeological investigation beyond Kiritimati Island, but that is a suitable place to begin because it is the largest island in the Line Group, by far, and it has been in existence since the late Pleistocene.

An expedition to Kiritimati Island for three weeks in 1999 by the authors (except Fankhauser) had the following objectives: (i) to locate the archaeological sites described in Emory (1934, n.d.), and to seek previously unrecorded archaeological sites and assess them for further research, (ii) to excavate selected areas in sites which held out the prospect of useful information about the period, origins or activities of prehistoric settlers, and (iii) to find and core sedimentary deposits, including sand dunes and wet sites, in order to recover samples relevant to constructing an environmental history of the island during the Holocene Period. 1 It was anticipated that the results of this research would indicate whether the prehistory of Kiritimati Island suited Emory's castaway model or more systematic settlement, and would provide thereby a sharper hypothetical basis for understanding the settlement history of the Line Islands than currently existed. In short, what their significance is in the colonisation history of the central Pacific as examples of the “mystery” islands.

In this preliminary paper, we outline our field research and its background, summarise our initial results, including radiocarbon determinations, and outline a possible settlement sequence for Kiritimati Island. Full descriptions of field and analytical research will be published later.

Kiritimati Island is situated at 01° 40N, 158° 30'W, about 2500km south of Hawai'i. At 53km long from Northwest Point to Southeast Point, and up to 31km wide, it is the world's largest atoll, but also one of the oddest in form (Figure 1). The main lagoon, 190km2, is in the northwest, and the land area of 321km2 includes about 60km2 in numerous small saline lakes and ponds. The southeastern part of the island and the lagoon edge consist of calcite hardpans dotted with small lakes. A narrow reef flat runs around most of the island but it is rarely more than 80m across. On the northwest and southeast arms, sand plains occur and small sand dunes have formed on the east and south coasts. There are no freshwater streams or lakes, but freshwater lenses are found below the sand plains, and some springs occur in nearby limestone after rains.

Magnetic evidence suggests that the atoll has been tectonically stable and volcanic rocks may be as little as 30m below the surface in the northwest of the island. Observations by Woodroffe and McLean (1998) indicate that the main hardpans are weathering profiles dating to the last interglacial, - 275 about 125,000 years ago, or more. A series of radiocarbon ages from Porites corals and Tridacna found in situ at several sites around existing lagoons and lakes are generally 3500-3000 years old and indicate emergence of about 1 m since that time, most of it in the last 1500 years (Woodroffe and McLean 1998).

Kiritimati Island is situated about 400km south of the mean position of the Intertropical Convergence Zone and is subject to southeast trade winds and salt spray for much of the year. These provide a relatively arid environment with highly variable precipitation (<200 to >4000 mm per annum), generally reflecting the 5-7 year timescale of ENSO fluctuations. There are no cyclones. The drying atmosphere and discontinuous freshwater lenses limit plant growth, and the flora is very depauperate. Coconuts grow fairly well on the sand lenses but are scattered or absent elsewhere. Tournefortia argentea (Boraginaceae) forms an open scrub with some individual trees reaching 8m and trunk diameters around 60cm, but mostly the canopy height is 2-3m, above a discontinuous scrubland of Scaevola taccada (Goodeniaceae), with the shrub Suriana maritima (Simroubiaceae) being most common around saline lakes and on saline flats. While there is much bare ground, the sedge Lepturus repens forms small clumps and there are several other minor herbs and grasses. The ground is heavily disturbed by land crabs and in places large numbers of sea birds (booby, frigate-bird and petrels) nest in the scrub or on the ground (Hope n.d.).

When Captain James Cook discovered and named the atoll, on 24th December 1777, he saw no signs of human life, recent or ancient. Fish and seabirds were abundant and Cook's men turned about 300 turtles for fresh meat, but they counted hardly 30 coconut palms standing above the immense scrub-covered plain and could find no fresh water, even by digging for it. Yet, David Samwell reported the existence of “some ratts”, almost certainly Rattus exulans introduced prehistorically (Beaglehole 1967:1079).

The crew of the whaler Briton spent more than seven months on the island after running ashore in October 1836 (Benson 1838, Tresilian 1838). Captain Benson recorded some 2000 coconuts, but caught no more than 20 turtles. He also could find no fresh water and was fortunate to have saved the supplies from his ship, as there was almost no rain for four months. Near his wrecked vessel, on the east coast, Benson (1838: 67) saw in several places “a number of stones piled up in the form of squares; and raised about three feet from the ground”. Taking them as cultural, he dug beneath one for human bones but found nothing. While these might have been pre-European structures, they could also have been the three cairns which John Williamson, - 276 Third Lieutenant on the Resolution, had caused to be built 60 years earlier in the southern part of the island to guide several missing sailors back to the ship (Beaglehole 1967:1347).

The first informal inventory of archaeological sites on Kiritimati Island was compiled by Father Emmanuel Rougier, who leased the island to grow coconuts for the copra trade in 1913. He had explored the island in 1912, and published his diary and other information as a small book (Rougier 1914). Exploring in the northeast part of the island, he discovered a group of six stone-lined graves, presumed European, and another tomb with a stone wall and path near the Beacon. Along the east coast were two more stone tombs with human bones and ashes nearby. Further explorations (Rougier 1917) brought to light a marae on Northwest Point, elsewhere two groups of stone-lined structures that he took to be remains of villages, and some graves—some he identified as probably Polynesian but most as European. This information was passed on to Kenneth Emory (1934, n.d., Gregory 1925) who spent seven days on Kiritimati Island with the Kaimiloa

Figure 1: Kiritimati Island, showing the location of archaeological sites relocated or newly-found in the 1999 expedition.

- 277

expedition in December 1924. His annotated inventory (Emory 1934) is the standard work on the archaeology of the island and provides the background to our project.

In the absence of a site record scheme we have coded the Emory (1934, n.d.) records and new sites as KKI for Kiritimati, Kiribati. The older sites (numbered below as in Emory 1934) were mapped very approximately by Emory and Rougier, and located on the ground by proximity to kilometre cairns along the perimeter road. Sites found or relocated by us were plotted by GPS and are shown on Figure 1.

KKI-03: This is probably the second village site reported by Rougier (1917:28), although he has it at the 40km mark, whereas Emory (1934:21) reported it at the 22km mark. Emory (1934:20-21, n.d.:56) mapped the main features but did not excavate, because he regarded them as probably representing the occupation of that area by William Grieg about 1880, remains of which existed in the vicinity. Despite signs of recent test pits and shovel trenches distributed across the site, we cleared off the disturbed sand - 278 from much of the site and excavated 3.5m2 in the main structure, which appears to be a marae.

- 279

KKI-16: At Paris, on the N point of the southern arm of the lagoon, is a stone pavement extending inland from 12m to 32m from the shore and at a right angle to it. Clearance showed that this consisted of two parallel lines - 280 of flat stones, evidently part of an old road running towards the rusted remains of a former jetty, both probably built by Rougier.

Figure 2: Preform of a Tridacna shell adze found at KKI-19.

- 281

The archaeological survey of Kiritimati Island was based upon an interpretation of the landscape history in general, and more specifically at the main sites (Falkland and Woodroffe 1997, Woodroffe and McLean 1998, Hope n.d.). Our strategy was to examine each of the main geomorphological units (sand plains, sand dunes, rubble plains and beaches, relict lagoon shores,

Figure 3: A retouched Tridacna shell knife from KKI-19.

- 282

etc.) concentrating upon the most probable areas of human settlement within each. In all cases subsurface investigations, whether by coring (mainly in the dunes) or by digging, sought to establish that no earlier cultural stratigraphy had been missed: the age of first colonisation was our primary aim. Consequently, while we could not cover more than a small proportion of the island's total surface area, we surveyed a representative sample of all the potential settlement options. More sites may be found in future, but there is no particular reason to anticipate that they would be of a different age or type than those already recorded.

KKI-03, 14, 15 and 18 are on the north shore of Manulu Lagoon. They lie on and in a 50cm deep deposit of calcareous sand, formed as the sea receded after 3000 B.P., following the Woodroffe and McLean (1998) model. This overlies a massive calcrete, marking the old Pleistocene surface. Site KKI-02 extends from similar sand deposits to the floor of a lagoon embayment comprising damp calcareous sand and grit. Around the northern shores of Manulu Lagoon and others in the vicinity, the sand is often slightly mounded in low beach-lines which can be followed for a kilometre or more in places. Most of the archaeological material seems to be deposited upon or near these. The archaeological sites were clearly deposited after the island gained its present size, an event postulated by Woodroffe and McLean (1998) as occurring about 2000 years ago.

The lagoons are tidal but some sediment is building up in them. Approximately 80m southeast from the archaeological sites at Manulu Lagoon, a core 185cm in length was raised from the lagoon floor in c. 2m of water. Except for some large salt crystals in the upper 5cm, the core consisted of calcareous mud of silt and clay size containing occasional shell fragments. The muds have probably formed by algal precipitation. This core has been examined for pollen and very sparse grass pollen is the only type recovered. The levels 5-55cm seemed darker and contained scattered black flecks, but carbonised particles are extremely rare in the core so it is doubtful if a fire history is being recorded. The basal sample from the Manulu Lagoon core will be submitted for radiocarbon dating together with a modern sample of lagoon mud.

Archaeological sites on the seaward edge of the island (KKI-01, 05-11) are approximately 2m above mean sea level and consist of materials heaped upon a coral boulder rampart that was formed during the last 2-3 millennia as sea level fell. Many large shells and coral blocks still litter sand plains and coral rubble fields, especially along the south coast, where in some cases they form low ridges parallel to the shore and extending up to 600m inland.

Dunes were examined at Joe's Corner, Artemia and in the Auvergne sandhills region, but no archaeological material was noted except at KKI- - 283 17, 19. The dunes, which appear to be relict, are formed from very fine light brown sand and have a weakly developed soil with weathering and slight calcareous deposition about 15cm below the surface. The sand is extremely clean, containing only scattered crab carapace fragments and occasional pieces of mollusc and bird bone. The dunes may represent mid-Holocene deposits formed from beaches that were subsequently covered by coralline rubble.

Six sites were excavated and an initial set of radiocarbon determinations obtained from the Waikato (Wk), Uppsala (Ua) and Australian National University (ANU) radiocarbon laboratories (Table 1). The samples are all on unidentified charcoal but on pieces of small diameter in order to preclude, so far as possible, the dating of driftwood (although in terms of the alternatives for colonisation, it is worth pointing out that the true ages could only be younger than the radiocarbon determinations). A second set of samples will be dated once they have been identified as to species, a process which takes some considerable time. While this two-staged strategy is practical rather than an ideal, there is less reason to suspect significant inbuilt age in tropical Pacific charcoals than in those derived from temperate forests. On Kiritimati Island, short-lived woody shrubs are highly predominant and while there is one specimen of long-lived Pisonia grandis in existence, it is improbable that there was ever Pisonia forest, and certainly not on the coast where the sites are predominantly situated. The main sources of potential firewood are the abundant, short-lived shrubs, Tournefortia argentea, Scaevola taccada and Suriana maritima (Woodroffe and McLean 1998).

The samples for Wk-7746 and ANU-11194 came from 25cm under stones from the dismantled KKI-01 marae. The former appeared to be within the undisturbed foundations of the original structure, the latter from boulder-covered sand towards the centre of the present site. The results suggest either that disturbance was more extensive than we thought and that marae boulders had been distributed by Rougier over recent fireplaces, or that the original structure was itself of recent construction, possibly by Polynesian labourers who worked in the coconut plantations for William Grieg, beginning in 1880 (Rougier 1914).

In site KKI-02, the charcoal sample for Wk-7751 was very small (1.9g) and processed by AMS. It came from about 20cm depth in excavation of a marae which appeared quite similar in form to that in KKI-03A dated by Ua-15388 (below), and a similar result was expected. Further samples will need to be processed to check on this result.

- 284

The sites at Manulu Lagoon are well-defined chronologically by seven radiocarbon determinations. Ua-15388 is upon a sample associated with the KKI-03A marae structure and provides an age for the complex of associated structural features formed from hardpan slabs in this area (see Emory 1934, Figure 9). The structure at KKI-14A nearby has two radiocarbon determinations (Ua-15389, Ua-15390) that are on samples from either side of the feature, while from about 15m away there is a very similar age on the Wk-7747 sample, obtained at 15cm depth in one of the cluster of ovens at KKI-14B. These data, in turn, are consistent with the results from Wk-7748 and Wk-7749, each on samples from 10-25cm depth in oven clusters at KKI-15A and 15B. The broken Tridacna fragments of ANU-11176 from KKI-18 gave a determination suggesting that the material was sub-fossil.

This set of results as a whole suggests that the group of sites (KKI-03, KKI-14, KKI-15, KKI-18) extending approximately 300m along the NW shore of Manulu Lagoon represents a single complex of religious and domestic features dating to the 13th to 15th centuries, and most probably to about A.D. 1400.

The patches of Tridacna shell-working distributed about 150m along the dunes at KKI-19 are probably also of similar age. The samples for Wk-7750 and ANU-11173 are from the same oven feature. ANU-11175 was a sample of shattered and reworked Tridacna shell from the small workshop area producing the retouched knife (Fig. 3). If it was fresh shell, then this is the oldest determination from Kiritimati Island, but as ANU-11176 indicates, it is possible that it reflects the use of shell that died some years before it was taken up as raw material.

Leaving aside the modern results, this initial assemblage of radiocarbon determinations allows us to suggest two conclusions about the prehistory of Kiritimati Island, both of which, it hardly needs emphasising, are based on a small sample of data. Firstly, since similar ages were obtained on sites far apart there may have been a single period of occupation on the island rather than multiple occupations. Secondly, the age of occupation at approximately A.D. 1200-1600 is broadly consistent with evidence of settlement expanding after about A.D. 1000 into other of the very isolated or marginal East Polynesian islands, notably those in the New Zealand region (Anderson 1991, 2000), and also with the settlement of the Polynesian outliers (Spriggs 1997).

The scarcity of material culture from Kiritimati Island precludes any certainty in defining the origins of prehistoric settlement. Nevertheless, there are some clues available in the typological affinities of the structural sites.

- 285

Several types of structures on Kiritimati Island can be interpreted as ceremonial sites. KKI-01, originally a high rectangular platform of unshaped coral boulders, appears to be the only one of its type on the island, and given the radiocarbon determinations (above), its prehistoric status is uncertain. Also recorded are various low boulder-mounds and platforms of relatively undiagnostic type. The most interesting features are in sites KKI-02 and KKI-03, situated on the northern edge of the lagoon. Structures that appear to be marae have been constructed from thin coral slabs or calcrete hardpan set on edge (ahu). In the courtyard of two of these structures lay a cylindrical, water-rolled limestone boulder, presumed to have been a former upright stone. Pavements of hardpan slabs are also associated with these features.

Such structures clearly resemble others in central East Polynesia. The rectangular platform with an upright on top (KKI-01) resembles rectangular ahu platforms in the Society Islands. The building material and the location on the NE point, facing the opening in the reef, also fits the central East Polynesian pattern. The slab-constructed features, each with an upright stone on the courtyard at KKI-02 and KKI-03, are similar to ones in the Tuamotu Islands and to marae sites found in the northern Cook Islands, notably Penryhn (Tongareva) where similarities to Tuamotuan marae were noted by Bellwood (1987:89). However, the simplicity of form is such that structures of the type on Nihoa and Necker, Hawai'i (Emory 1928), for example, cannot be ruled out of comparison. Consequently, while the most likely origin of the Kiritimati Island marae types is to be found within the area of the Northern Cooks, Tuamotus and the Society Islands, a source in Hawai'i is also possible.

One potential way of establishing the origin of settlers is to determine the source of exotic lithics, for while these can be moved without human agency, stuck in the roots of drifting logs for example, it is much more probable that they were carried intentionally. On Kiritimati, the choice seems to be limited entirely to basalt, and instances of that being found are rare. Several small flakes and one core were recovered by us.

Sample CI-22 is a flake of phenocrystic basalt from the surface of site KKI-14. It appears to have undergone substantial weathering. Sample CI-32 is a fine-grained basalt core from the surface of site KKI-15. Source characterisation of these pieces was attempted by analysis of ten major elemental oxides using energy dispersive x-ray analysis (EDXA), and 30 trace elements using inductively coupled plasma-mass spectrometry (ICP-MS). Full details are available in Fankhauser (2000).

- 286

Analytical results were compared primarily with geochemical source data compiled at the University of Hawai'i (Sinton and Sinoto 1997). This contains the results of analyses for 284 single rock samples from sources, sites and 36 quarries throughout Polynesia. All of the average values from Sinton and Sinoto (1997) were incorporated into two more databases (major element oxides and trace elements) and supplemented with averages from the University of Hawai'i Database. Additional sample results from Best et al. (1992) and Clark et al. (1997) were also added. Values from single quarries or sources were averaged, based on close linking in cluster analysis. Only quarry or source samples were included in the database to avoid confusion with data from archaeological sites.

A large number of samples with trace element analyses was available for incorporation into a database. There was a problem of incompatibility between some analyses performed at different times in the same laboratory, or between different laboratories. Even though a large suite of elements was analysed in some studies (e.g., the present project, Best et al. 1992, Allen and Johnson 1997), only those elements common to all samples could be used. In addition, some of the analyses have missing values. The result was that while the trace element suite for the Kiritimati Island samples contains 30 elements, only 15, and usually 11 or 12, of these could be used for cluster analysis (Tables 2 and 3).

A plot of SiO2 versus total alkalis indicated that both Kiritimati Island samples are alkalic. The results from cluster analysis (Fankhauser MS.) reveal that neither can be matched with any quarries or sources in the current databases. Major element oxides show both samples are most closely linked with sources on Moloka'i, and islands in the Cook, Society and Samoa groups. Trace element data show CI-22 links most closely with a source on Pitcairn, a quarry in the Marquesas and quarries in the Hawaiian islands. Quarries from Samoa and the Society Islands could also be included. CI-32 is most closely linked to quarries on Tutuila and Moloka'i and, less closely, with a quarry on Pitcairn.

Some of the most closely linked quarries and sources to the Kiritimati Island samples were similar for both major element oxides and trace elements. For CI-22 the sources in common are located on Moloka'i, Tahiti, and Tutuila. CI-32 has sources on Moloka'i, Pitcairn, Easter and Tutuila in common. The samples are too dissimilar to be actually from these sources, but the data could suggest likely island groups.

Some island groups can then be eliminated by looking at the trends in concentrations of major element oxides and trace elements for island groups compared to the Kiritimati Island samples. For example, Cook Island basalts have concentrations of SiO2 that are less than in both samples from Kiritimati - 287 Island. Quarries from Pitcairn can be eliminated for similar reasons. Analysis along these lines reduces the origin of the Kiritimati Island samples to the Hawaiian Islands, with an origin in Samoa as a second choice. It must be emphasised, however, that without being able to obtain a clear match to a known source, this conclusion is tentative.

The initial results of our research indicate that the prevailing view of Kiritimati Island archaeology as the remains of castaway encampments, as implied by Emory (1934), is probably not correct. The archaeological landscape of the island is denser and more complex than his brief visit suggested. There are remains of an extensive religious and domestic complex on the NW shore of Manulu Lagoon (KKI-03, 14, 15, 18), and probably of another several kilometres to the west (KKI-02) on an embayment of the main lagoon. A shell tool manufacturing area existed on the east coast (KKI-19) and different types of marae, tombs, paths, shelters and shrines are indicated by the site records at various points elsewhere. Scarce midden recovered during the excavations has yet to be analysed, but certainly includes turtle, fish, bird and mollusc remains. There were no bones of domestic animals and, among the few artefactual remains, no sign of vegetable scrapers or graters.

The archaeological sites are distributed through all the environmental zones including the boulder beaches, dune systems and hardpan flats, and they occur throughout the island. In short, there is evidence of a complete settlement pattern, indicating that at one time in prehistory Kiritimati Island was substantially inhabited.

Neither in the nature of archaeological remains, nor in the radiocarbon determinations, is there anything to indicate that any of the occupation represented by these sites began earlier than during the period of East Polynesian expansion. Considering its size, and that unlike many other atolls it was above water throughout the period of human colonisation in Remote Oceania, Kiritimati had more chance of catching exploratory voyagers or castaways in the period 3000-1000 B.P, than many other islands. The geomorphology of the lagoon shores, in particular, indicates that the present coastal sediments were exposed by 3000 B.P. and that with continuation of sea-level recession since then, any remains of early settlement should be well preserved—none were found in 1999.

Instead, radiocarbon determinations from seven sites indicate that the period of settlement was approximately A.D. 1200-1600, with most radiocarbon determinations centred on the early 15th century. The only other radiocarbon determination in the Line Group is from Fanning Island, where - 288 Sinoto (1973) obtained ages of 1560 ± 85 B.P. (GaK-4558) and 880 ± 85 B.P. (GaK-4557) for sites where artefacts of early East Polynesian form had been recovered. Since Emory (1934:13-16) thought the ceremonial stonework on Fanning, and some of the artefacts, were of late Tongan origin, the evidence is difficult to interpret. However, the earlier radiocarbon determination looks too old for its artefactual associations and, since both are Gakushuin Laboratory determinations, they must be regarded as unreliable (Spriggs and Anderson 1993).

It is still a difficult matter to determine the origin of the initial settlers of Kiritimati Island. In the current state of knowledge, Hawai'i is the first choice, but it is too early to exclude central East Polynesia, assuming either that the Kiritimati Island basalt samples are from sources there, or possibly from Samoan sources traded into central East Polynesia. Hawai'i is quite plausible as a colonisation source, since downwind voyages on the Northeast trades from Hawai'i would take canoes to within 500km of Kiritimati Island in normal conditions and closer again with fluctuations in the position of the Intertropical Convergence Zone. There is nothing in the archaeological evidence to suggest that discovery occurred more than once, and certainly no support for the multiple-castaway hypothesis. The situation of Kiritimati Island in relation to Hawai'i appears very similar to that of southern Polynesia (Anderson 2000b), where expansion from the secondary colonisation centre of New Zealand reached all of the outlying islands, but evidently only once in each case, i.e., it was an outward expansion which did not develop into a system of inter-island voyaging. A hypothetical Hawaiian expansion needs to be tested in the remaining northern Line Islands, and other downwind archipelagos in Kiribati and the Marshall Islands.

Lastly, there is the intriguing matter of settlement extinction. Although it is a large island and lies outside the cyclone zone, Kiritimati Island was subject to one particular environmental hazard, the difficulty of obtaining fresh water. There are large bodies of fresh water beneath the sand plains in various places, especially in the southwest and northern arms of the island (almost none in the southeast), but the upper surface of these is commonly at 1.5 m depth in good conditions and it can sink to several metres or more (Deegan pers. comm., Kirby pers. comm.). At the latter depth it is out of range of coconut palms, let alone other Polynesian cultivars. It is noticeable that the main archaeological evidence, at Manulu Lagoon, is beside a spring, and has several wells sunk in the vicinity. A relatively small change in the climate, or some years of low rainfall, might have had devastating consequences for prehistoric settlers.

If horticulture was absent or difficult, then settlers must have relied mainly upon the exploitation of natural faunal resources, picking up - 289 whatever carbohydrate was available from native herbs—although it is probable that in an island as large as Kiritimati, a water crisis would not affect all coconut groves equally. The faunal resources included colonial-nesting seabirds, nesting turtles, and various reef fish and shellfish. However, an energetically-efficient foraging strategy, which took different taxa according to a preference for larger individual size (Anderson 1981), might soon over-exploit the more prominent faunal resources, reducing both biomass and diversity. In addition, the introduction of Rattus exulans would have been disastrous for ground-nesting birds. It is quite conceivable that on Kiritimati Island, as on many other of the “mystery” islands, a combination of environmental hazard, tenuous horticultural productivity and unsustainable harvesting of the natural resources, produced, eventually, a subsistence environment that was no longer attractive to settlement, even one unable to sustain settlement.

More research is needed on Kiritimati Island (indeed, in the Line Islands generally). It is so large, and the shores so scrub-covered, that it may take months of exploration to survey all of the possible areas of prehistoric settlement, although the situation of the main sites along a series of low beach ridges which extend around much of the northern lagoon shores suggests that systematic survey there would be repaid with further significant discoveries. Additional data are needed to resolve some of the outstanding questions about the nature of the settlement—were cultivars and domesticates available? What other forms did material culture and domestic structures take? Can we pin down the source of colonisation and confirm its age? And so on. However, this paper provides at least a preliminary sketch to revitalise research which has stalled since the pioneering efforts of Rougier and Emory.

We thank the Government of the Republic of Kiribati for permission to carry out research on Kiritimati Island. Assistance was kindly furnished by the staff of the Ministry of Line and Phoenix, notably Mr D. Kanono (Assistant Secretary), Mr U. Bukaireiti (Wildlife Officer) and Mr B. Smith (Wildlife Education Officer). For other assistance we thank Mr T. Awira (Conservation Support Officer), Mr J. Bryden, Mr Carl Deegan, Hydrogeologist (Douglas Partners), Mr and Mrs E. Reiti, Ms B. Weatherstone, Ms A. Gellerstig, Mr K. Anderson, Dr N. Kirby, Dr P. Colebatch and other staff at the AUSAID office, Dr L. Hartzell at the B.P. Bishop Museum, and Professor G. Ward, ANU. The artefacts were drawn by Ms Judith Sellers and the map by Ms Kay Dancey, ANU. We are grateful for the financial assistance of the Kon-Tiki Museum, Oslo, without which the field research would not have been possible. Additional funding was provided by the Department of Archaeology and Natural History, ANU, and the Royal Society of New Zealand.

- 290

• Allen, M.S. and K.T.M. Johnson, 1997. Tracking ancient patterns of interaction: Recent geochemical studies in the southern Cook Islands. In M.I. Weisler (ed.), Prehistoric Long-distance Interaction in Oceania: An Interdisciplinary Approach. New Zealand Archaeological Association Monograph 21:111-33. Auckland.
• Anderson, A.J., 1981. A model of prehistoric collecting on the rocky shore. Journal of Archaeological Science, 8:109-20.
• ——1991. The chronology of colonization in New Zealand. Antiquity, 65: 767-95.
• ——2000a. Implications of prehistoric obsidian transfer in South Polynesia. Indo-Pacific Prehistory Association Bulletin. In press.
• ——2000b. No meat on that beautiful shore: the prehistoric abandonment of subtropical Polynesian islands. International Journal of Osteoarchaeology. In press.

• Beaglehole, J.C., (ed.), 1967. The Journals of Captain James Cook on his Voyages of Discovery, Volume III, Part Two, The Voyage of the Resolution and Discovery, 1776-1780. Cambridge: Hakluyt Society.
• Bellwood, P., 1978. Man's Conquest of the Pacific. Auckland: Collins.
• ——1987. The Polynesians. Prehistory of an Island People. Revised edition. London: Thames and Hudson.
• Benson, G., 1838. Sketch of Christmas Island: With a chart of the island. The Hawaiian Spectator, 1:64-68.
• Best, S., P. Sheppard, R. Green and R. Parker, 1992. Necromancing the stone: Archaeologists and adzes in Samoa. Journal of the Polynesian Society, 101: 45-85.

• Clark, J.T., E. Wright and D.J. Herdrich, 1997. Interactions within and beyond the Samoan archipelago: Evidence from basaltic rock geochemistry. In M.I. Weisler (ed.), Prehistoric Long-distance Interaction in Oceania: An Interdisciplinary Approach. New Zealand Archaeological Association Monograph 21: 68-84. Auckland.

• Emory, K.P, 1928. Archaeology of Nihoa and Necker Islands. Bernice P. Bishop Museum Bulletin 53. Honolulu.
• ——1934. Archaeology of the Pacific Equatorial Islands. Bernice P. Bishop Museum Bulletin 123. Honolulu.
• ——n.d. Unpublished diary, Christmas Island, December 10-16, 1924. Bernice P. Bishop Museum Archives, Honolulu.

- 291

• Falkland, A.C. and CD. Woodroffe, 1997. Geology and hydrogeology of Tarawa and Christmas Islands. Kiribati. In H.L. Vacher and T. M. Quinn (eds),Geology and Hydrogeology of Carbonate Islands, Developments in Sedimentology 54. Elsevier, Amsterdam, pp. 577-610.
• Fankhauser, B., n.d. Characterisation and Sourcing of Archaeological Basalt from Christmas Island. A Report to the Indo-Pacific Colonisation Project. Indo-Pacific Colonisation Project unpublished reports, ANH, RSPAS, Australian National University, 58 pp.

• Gregory, H.E., 1925. Report of the Director for 1924. Bernice P. Bishop Museum Bulletin 21. Honolulu.

• Hope, G.S., n.d. Landscape History of Kiritimati (Christmas) Island. Indo-Pacific Colonisation Project unpublished reports, ANH, RSPAS, Australian National University, 6 pp.

• Irwin, G.J., 1992. The Prehistoric Exploration and Colonisation of the Pacific. Cambridge: Cambridge University Press.

• Kirch, P.V., 1988. Polynesia's mystery islands. Archaeology, [May-June]:26-31.

• Rougier, E., 1914. Isle Christmas, South Seas (Océanie) Coconuts, Fishes etc. Brioude: L. Watel.
• ——1917. Isle de Christmas. Bulletin de la Société d'Etudes Océaniennes(Polynesie Orientale), 1:25-30.

• Sinoto, A., 1973. Fanning Island: Preliminary archaeological investigations of sites near the Cable Station. In K.E. Chave and E.A. Kay (eds), Fanning Island Expedition: July and August 1972. Report 73-13. Honolulu: Hawaii Institute of Geophysics.
• Sinton, J.M. and Y.H. Sinoto, 1997. A geochemical database for Polynesian adze studies. In M. I. Weisler (ed.), Prehistoric Long-distance Interaction in Oceania: An Interdisciplinary Approach. New Zealand Archaeological Association Monograph 21:194-204. Auckland.
• Spriggs, M.T.J., 1997. The Island Melanesians. Oxford: Blackwell.
• Spriggs, M.T.J, and A.J. Anderson, 1993. Late colonization of East Polynesia. Antiquity, 67:200-17.
• Stuiver, M. and G.W. Pearson, 1993. High-precision bidecadal calibration of the radiocarbon timescale, AD 1950 - 500 BC and 2500 BC - 6000 BC. Radiocarbon, 35:1-23.

• Tresilian, F.H., 1838. Remarks on Christmas Island; Read before the Sandwich Island Institute, April 1838. The Hawaiian Spectator, 1:241-47.

• Woodroffe, CD. and R.F. McLean, 1998. Pleistocene morphology and Holocene emergence of Christmas (Kiritimati) Island, Pacific Ocean. Coral Reefs, 17:235-48.

- 292

ANU δ13C is estimated, with margin of error of ± 2.0 ‰. Calibrations according to Stuiver and Pearson (1993).

- 293

- 294 Page is blank