- 149

Upon landing on Kiritimati (Christmas Island), one is surprised by its vast area and large human population. Upon disembarking on the island of Tabuaeran (Fanning Island), 140 nautical miles (nm) downwind, a distance we sailed in 40 hours as an overnight voyage, one is immediately struck by its size and the luxuriance of the vegetation. This surprise is all the greater when one realises that these are two of the “mystery islands”—islands with prehistoric sites but uninhabited when re-discovered by Europeans—which are renowned for their smallness, inhospitality and remoteness (Terrell 1986, Bellwood 1987, Kirch 1988, Irwin 1992).

To question these attributes of what we might call the paradigm of the “mystery islands” led us: (i) to inventory the resources of both islands and establish that Kiritimati and Tabuaeran offered different and complementary opportunities to their prehistoric inhabitants; (ii) to define the distribution of archaeological sites in relation to these resources, and demonstrate that on Tabuaeran, sites are clustered near a freshwater marsh and good soils, where agricultural potential is highest, while on Kiritimati sites are scattered on the lagoon and ocean beaches, where birds and turtles are abundant; and (iii) to detail the nature of occupation through systematic excavations and argue for intermittent settlement on Kiritimati and “permanent” settlement on Tabuaeran. These points, developed throughout this paper, form the core of our argument that the “inhabitants” of Kiritimati were members of a larger community centred on Tabuaeran—probably also including the other islands of the northern Lines (Teraina and Palmyra)—and that the archaeological remains on Kiritimati result from its use as a bird and turtle refuge, and a way-point for long distance voyaging elsewhere in Polynesia.

Questions about the “mystery islands” have been in the air in recent years (Terrell 1986, Bellwood 1987, Kirch 1988, Irwin 1992, Weisler 1996, 1997), so perhaps it is not so remarkable that as soon as we completed our research - 150 on Kiritimati, another team of archaeologists, led by Atholl Anderson, began work on the same island, at some of the same sites. 1 What is remarkable, however, is that the interpretations developed by each team stand in sharp contrast to each other on two issues central to the “mystery island” paradigm: isolation and resource over-exploitation. In a recent paper published in this journal, Anderson et al. (2000:288) suggest that Kiritimati was discovered only once and did not become part of a system of “inter-island voyaging”. They further note that settlement of Kiritimati was “substantial” and speculate that “settlement extinction” may have resulted from over-exploitation of “faunal resources reducing both biomass and diversity” (Anderson et al. 2000:287-89). They frame their interpretation in terms of isolation and resource scarcity, confirming the existing paradigm.

We identify two points where the facts appear to be irreconcilable with this paradigm, at least when applied to the northern Line Islands. Resource over-exploitation on Kiritimati is questionable because the island still supports the most diverse, and the largest, populations of tropical seabirds in the Pacific, indicating few or no human caused extirpations in the past (Garnett 1983, D. Watling pers. comm.). Isolation, too, seems unlikely. Recent work on the geochemistry of basalt artefacts has demonstrated that both Tabuaeran and Kiritimati were part of the widest interaction sphere known to date in Polynesia (Di Piazza and Pearthree, n.d.a). In this paper, we stress interconnectedness within and beyond the northern Line Islands, rather than isolation, and resource complementarity between Kiritimati and Tabuaeran, rather than scarcity of any particular resource on either island. The hard evidence of basalt exchange, noted above, is only the tip of the iceberg. By far the more numerous voyages would have been those between neighbouring islands, giving inhabitants of certain islands, such as Tabuaeran, access to different resources, like those available on Kiritimati. By using the northern Lines as a case study, we were able to re-examine the “mystery islands” phenomenon and conclude that most of them appear to have functioned as faunal reserves or “satellites” of agricultural settlements or “mother communities” no more than about 140nm away.

Tabuaeran and Kiritimati lie at 3°50' and 1°50' North, respectively, a latitudinal zone with a steep rainfall gradient (Mueller-Dombois and Fosberg 1998:323-24,340). Tabuaeran averages 2100mm a year, while Kiritimati receives only 766mm. This disparity in rainfall accounts for the richer vegetation, soil development and agricultural potential on Tabuaeran. Plants such as breadfruit (Artocarpus altilis), bananas (Musa sapientum) and swamp taro (Cyrtosperma chamissonis) thrive in the six modern villages on - 151 Tabuaeran but at only one location on Kiritimati—near Banana village. Both islands have rich and extensive lagoons, although Kiritimati offers especially good fishing for bonefish (Albula sp.) and milkfish (Chanos sp.).

Kiritimati was once a favoured nesting site for green turtles (Chelonia mydas). Captain Cook re-provisioned his ship with hundreds of turtles on the occasion of the European discovery of the island, but during the period of whaling and the copra plantation, the population declined dramatically, probably because of over-exploitation (Garnett 1983:19). Turtles have never been recorded nesting on Tabuaeran (Garnett 1983:19,100).

Kiritimati, with its area of 341km2, still has some of the largest populations of tropical seabirds anywhere in the world (Garnett 1983:15-18), although rats (Rattus exulans) that presumably arrived with the early Polynesian explorers, copra planting, the Second World War and nuclear testing have probably all taken their toll. Rattus rattus never became established on the island (Garnett 1983:14, D. Watling, pers. comm.). Despite all of these disturbances, there were over 16 million birds of 18 species in the 1960s (Garnett 1983). Sixteen of these species nest on the ground, principally on barren rat-free motu or islets in the lagoon, and would have been easy and abundant prey to prehistoric hunters. That this island still supports such huge and diverse populations indicates that the configuration of Kiritimati, with its many motu, protected faunal biodiversity and biomass from over-predation by rats. Prehistoric hunters certainly harvested these seabirds, although apparently not intensively enough to cause any extirpations. Most of the other Line Islands with evidence of prehistoric use have only six to nine seabird species left (Emory 1934,1975; Garnett 1983). They all lack Procellaridae (petrels and shearwaters), whose large fat chicks, left unattended in shallow burrows, are easy to catch and were a favourite prey of Polynesians. That Kiritimati still has five species of Procellaridae indicates minimal human predation on these slow-breeding birds, so vulnerable to extirpation. Tabuaeran is about one tenth (34km2) the size of Kiritimati and lacks rat-free refuges. It has much smaller bird resources, only 6,000 individuals of 6 species, all tree nesting (Garnett 1983:98-99).

This inventory highlights the differences between the subsistence resources available on each island. Wet Tabuaeran offered rich agricultural potential and dry Kiritimati sheltered abundant populations of birds and turtles. Tabuaeran would have had many fewer faunal resources than Kiritimati at the time of initial settlement simply because of the disparity in size. These smaller populations would also have been more susceptible to over-exploitation. Tabuaeran consists of four forested islets, all habitable by rats and humans, and lacks the barren motu that are so numerous on Kiritimati. When we add to these observations that Tabuaeran, with its richer - 152 soil and abundant fresh water, is more attractive for human settlement than Kiritimati, it is unsurprising that turtles have been extirpated there or that there is less avian diversity.

The results of our surveys 2 and excavations, along with previously reported sites (Rougier 1914,1917; Emory 1934; Anderson et al. 2000) allow us to differentiate two narrow environmental zones on Kiritimati where archaeological sites have been found (Fig.l). The first of these zones has platforms or mounds, trails and cairns, and is confined to beach berms. All of these sites, except Site 1, are on the eastern side of the island, exposed to the northeast and southeast trade winds. The second zone has domestic features and marae, and is located along the eastern edge of the “reticulated lagoon” (as described in Woodroffe and McLean 1998). Here, freshwater springs can be found along the base of an uplifted limestone ledge marking the shore of the ancient lagoon. No prehistoric sites have been reported any distance away from either the ocean beach berm or this ancient lagoon shore, nor have any been found on the western half of the island (except Site 1). Note that Sites 17, 21 and, probably, 13 are historic. These findings do not support the assertion of Anderson et al. that “archaeological sites are distributed through all the environmental zones … and they occur throughout the island” (2000:287).

Figure 1. Archaeological sites on Kiritimati Island.

- 153

In the interests of continuity, we will refer to Emory's sites by his original designations (Sites 1 to 12), and to the newly described sites of Anderson et al. (2000) by their proposed numbering system (KKI-1 to 21). Sites or features discovered by the authors are numbered or lettered consecutive to Emory (Sites 13 to 21, Site 3 features f to j)).

The sites along the dry windward coasts seem to consist of temporarily occupied loci where people went to catch birds, turtles and fish, to build features that facilitated such activities, and to practise rituals that may well have been related to the animals hunted. Evidence of temporary use are the fireplaces and the shell workshop without associated habitation features (Site 9, KKI-19), the windbreak shelter (Site 5), and the small amounts of turtle, bird and fish bones excavated from platform fill (Sites 10 and 15). These bones were probably food remains, inadvertently incorporated during the construction of platforms at pre-existing campsites. Paved stone trails facilitating access to the beach across areas of sharp coral rubble, presumably for fishing or turtle hunting, have been reported (Sites 5,18, probably KKI-20). We might speculate that clusters of cairns along the beach ridge (Site 14, as well as the two cairns reported at KKI-21) may have served as landmarks for canoes fishing along the shore. Religious activities are indicated by burial platforms (Sites 5, 6 and 9), by platforms with upright stones (Sites 10 and 15) and by two additional mounds (Sites 19 and 20) that were not excavated. The evidence that these religious activities may be linked to the turtles and birds hunted along this coast come from the bird petroglyph surmounting a platform containing a human bone, excavated by Rougier (Emory 1934:22) and the articulated turtle carapace buried in a similar adjacent platform that we excavated. The petroglyph stone, collected by Rougier, is now on display at the “Musée de Tahiti et des lies” in Punaauia (Lavondes 1976, catalogue number 769). In total, 10 prehistoric sites have been reliably reported 3 along the dry windward coasts.

The only site (Site 1) situated on the leeward coast of the island was a rectangular platform surmounted by an upright slab (Rougier 1914:127, Emory 1934, Anderson et al. 2000). Both its architecture and its location on a beach berm are similar to the windward sites, but it differs in facing a reef pass that allows easy canoe access to the sea. Like the windward sites, it may be related to marine exploitation.

Other than the sites along the windward coasts, the only prehistoric remains are found on the northeast and southeast edges of the reticulated lagoon. They consist of “villages” or clusters of domestic features located near freshwater sources, either on the modern lagoon shore (Sites 2 and 3) or in the interior (Site 12). Emory described Site 12 as a single large cairn of piled rubble, while Rougier reported signs of “a village… a marae… and a - 154 spring” at what was probably this location (Emory 1934:23, Rougier 1917:28,). 4 Although little is known about Site 12, its location near a spring and at the easternmost coconut grove on the island would have made it an ideal base camp for hunting along the windward coasts, particularly among the large colonies of sooty terns at the southeast tip of the island (Garnett 1983:18, 85). Similarly, Sites 2 and 3 are located close to main lagoon motu with important seabird breeding colonies. They also give access to marine resources in the lagoon and to the protected ocean reef off the leeward side of the island. Emory reports “house sites” at Site 2, 5 but found little evidence of occupation. Anderson et al. (2000:277,292) noted additional features (a marae, an alignment and five ovens) “at what is probably Emory's site [2]”, although no artefacts, cultural deposits or prehistoric dates were obtained. The only village with a full range of domestic and religious features is Site 3. There, we discovered a marae, a canoe house, a cooking area, habitation features, as well as basalt and Tridacna shell workshops (see below).

Figure 2. Domestic features at Site 3 (Kiritimati Island).

It could be argued that this pattern of “villages” limited to lagoon or lake shores versus hunting camps on the windward beaches is the result of sampling error. However, the thoroughness of Rougier's efforts during his ten years of clearing and planting the island, and his interest in archaeological sites, suggest that the island has in fact been well surveyed. Moreover, during - 155 our 1999 re-survey, followed by the study of Anderson et al. (2000), only a few additional sites 6 were discovered, mostly mounds, cairns or trails. The only new domestic features found were at previously reported site locations (Sites 2 and 3).

The only complete village or hamlet is Site 3. The results of our extensive excavations there indicate that, although this site has the widest range and number of features of any site on the island, occupation seems to have been unintensive and intermittent based on the shallow cultural deposit and evidence of reuse. Anderson et al., who re-excavated some of these features several weeks later, came to different conclusions. They suggest a “single period of occupation on the island” and “substantial” settlement (2000:284,287).

Rougier described this site as “…an ancient village marvelously situated near an abundant spring, on the shore of a peninsula in the lagoon…” (Rougier 1917:28). This spring (Emory's feature a) consists of a shallow oval basin outlined by naturally cemented coral cobbles, suggesting prehistoric age. It appears to divide the village into domestic and religious zones, with habitations and activity areas just to the west and a marae isolated some hundred metres away to the east (Fig.3).

Figure 3. Marae at Site 3 (Kiritimati Island). A: Plan; B: Reconstruction in elevation.

- 156

This marae with its unenclosed court, multiple court uprights and ahu composed of a double row of uprights is similar to marae of the Leeward Society islands and the central and western Tuamotus (Emory 1970, Gérard 1974). For a different reconstruction and discussion of architectural affinities of the marae, see Anderson et al. (2000:278).

One hundred to 160 metres west of the marae, we discovered a cluster of domestic features including: a long rectangular structure outlined by postholes, a probable canoe house, a cooking area with multiple earth ovens, several small pavements partially outlined by slabs on edge (probable habitations), an isolated earth oven, a basalt debitage concentration and small clusters of Tridacna and pearlshell debitage.

- 157

Figure 4. Canoe house at Site 3 (Kiritimati Island). 1: Basalt adze, lure preform; 2,3,4,5: cut pearlshell; 6: pearlshell valve; 7: Tridacna debitage, granite pebble; 8: burnt granite abrader fragment; 9: lure preform, Tridacna debitage; 10: Tridacna debitage; 11: pumice abrader; 12: pearlshell harpoon.

We hypothesise that these features are the remains of a hamlet used temporarily and intermittently as a supply base by voyagers hunting turtles and birds. Short-term occupation is indicated by a single, shallow, cultural deposit, no more than 5cm thick (except in pit features). The total amount of stone and shell debitage is small, artefacts and midden are scarce, and - 158 habitation features are rare. The basal layer of the trash pit at the cooking area was dated to 620 ± 60 B.P. (Beta 141327). The sample consisted of one taxon, cf. Suriana maritima. The result at 2 sigma is cal. A.D. 1275 to 1425 using an estimated C13/C12 ratio of-25 0/00 (Talma and Vogel 1993, Stuiver et al. 1998). Anderson et al. (2000:Table 1) give five more dates for these domestic features, confirming the occupation of this site sometime between 1250 and 1450 A.D. Re-use or re-occupation is suggested by laminated pit features in the cooking area and in the single earth oven, by superimposed pavements in the habitation area, and by evidence of doubling or replacement of posts in the canoe house.

Occupants of this hamlet were certainly navigators. The basalt for their tools came from distant islands in central Polynesia. Their canoes were about 10m long, as inferred from the length of the canoe house, and were probably outriggers. The width of the canoe house (less than 3m) is too narrow to shelter a double canoe. This is the size of canoe favoured for voyaging (Gladwin 1970, Gillett 1987, Di Piazza and Pearthree n.d.b). Analysis of faunal remains from the cooking area and a habitation feature shows the overwhelming importance of turtles and birds in their protein subsistence. Fish bones are rare and molluscs (except for Tridacna and pearl oyster used for artefacts) appear to be absent. Coconut was an additional food source. Endocarp and husk were recovered from pit features at the cooking area as well as from postholes at the canoe house.

A good ethnographic analogue for Site 3 are the five atolls in the central Carolines (Pikelot, Gaferut, Olimarao, West Faiu and East Faiu) that do not support permanent settlements because of insufficient freshwater, but nevertheless serve as resource reserves for other islands up to 112nm away. Of these, the two best known, Pikelot and West Faiu, are visited occasionally by voyagers from surrounding islands, who sail there to feast on turtles, birds and fish, as well as to collect these resources for their home villages. Sleeping shelters and storage sheds have been built on both islands for the comfort of visitors. Each island also has a small chapel where prayers are offered for safe passages. Coconut, breadfruit and even tobacco have been planted (Thomas 1987:146-52,193-95). Canoes are carried well up the beach and put under makeshift shelters to protect them from the wind and sun. Even today, voyages to these “turtle islands” are common. Of the 73 inter-island voyages recorded over a 15-month period, 22 were to Pikelot, averaging about two weeks each (Gladwin 1970:38-39). This same pattern is found in the northern Marshalls where inhabitants travelled to uninhabited Bikar and Taongi (about 100 and 370nm respectively) two or three times a year to gather eggs, birds and turtles (Amerson 1969:332). Anutans and Tikopians also used the uninhabited island of Fataka for bird hunting (Yen and Gordon 1973:4,103).

- 159

An archaeological analogue to Site 3 may be Necker and particularly Nihoa in the Hawaiian archipelago. Both islands have numerous small marae, imported basalt tools and evidence of temporary (Necker) or semi-permanent (Nihoa) occupation, not unlike the small marae and shallow habitation sites on Kiritimati. They presumably served as bird-hunting grounds for people from the main Hawaiian islands (Cleghorn 1988). Nihoa lies about the same distance from Niihau as Kiritimati does from Tabuaeran.

On Tabuaeran, we identified a single settlement complex extending for about 1km along the lagoon shore from the old Trans-Pacific Cable Station to the abandoned airstrip north of Betania village (Fig.5). The western or inland boundary roughly coincides with an outcrop of yellow phosphate rock 130 to 200m inland (Roy 1973:193). This location is exposed to the trade winds but protected from westerly storms. Near the middle of this 15ha complex are three features of monumental scope, a marae, a large pavement and a high status burial platform. Architectural features and activity areas such as pavements, stone alignments, hearths, postholes and artefacts, as well as clusters of charcoal, fire-cracked rocks, midden and concentrations

Figure 5. Archaeological sites on Tabuaeran Island.

- 160

of pearlshell debitage are scattered throughout. Just north of the complex, graves, mounds or platforms have been recorded along the gravel ridges which border an in-filled lagoon pass (Roy 1973:193-94). This pass is occupied by a freshwater marsh (circa 6.5ha) that today supports extensive wet agriculture. A break in the reef on its ocean side forms a protected canoe landing.

An additional but much smaller habitation complex may have existed along the south edge of the main lagoon pass. Its location under the modern settlements of Baerau and English Harbour precluded excavation.

Other sites, including two probable agricultural pits (FAN1-19) and some disturbed pavements 2 to 3km north of the marsh, a complex of trails and alignments west of Baerau village (FAN3-1), a double wall of stacked coral slabs (FAN2-1), and a large rectangular platform (FAN2-2) on the northeastern islet, as well as a large rectangular enclosure of stacked coral slabs (FAN1-18) in Betania village, attest to various agricultural, fishing and, perhaps, ceremonial activities on all three islets, with no evidence of domestic features.

A more thorough survey 8 would certainly identify other remains, but it seems likely that no site comparable to the habitation complex would be found, because it occupies the area offering the best combination of canoe access, freshwater resources and soil development on the island. This pattern of an entire population concentrated in a main complex, with scattered evidence of use on other islets, is known archaeologically from Polynesian atolls such as Nukuoro, Kapingamarangi and Tokelau (Davidson 1967:366, Leach and Ward 1981:15-16, Best 1988).

Tabuaeran offers evidence not only of the largest prehistoric village on one of the “mystery islands”, but for a successful and relatively long-term settlement, a surprising finding in light of the prevailing paradigm.

For previously recorded sites (FAN1-1 to FAN1-8), see Emory (1934, 1939a) and Sinoto (1973). Additional observations on FAN1-1 and FAN1-2 are given below. New sites are numbered consecutive to Sinoto beginning with FAN1-9(1973: Fig.2).

- 161

- 162

We hypothesise that these features represent a nucleated village, continuously occupied over a relatively long period of time by a sizable population of agriculturalists, who would have occasionally visited Kiritimati to harvest turtles and birds. Continuous occupation is indicated by a deep cultural deposit (up to 35cm), without stratigraphic breaks. Long-term settlement by a substantial population is evidenced by the number and diversity of features, the large area of the complex, the abundance of artefacts and the monumental religious and mortuary structures. The sea wall too suggests high labour investment and planning for the future. That this was labour well spent is indicated by the habitation deposits that have buried the sea wall. Prolonged occupation is further supported by three radiocarbon dates, spanning the 13th to mid-14th century. Coconut endocarp from an earth oven (FAN1-13) dated to 810 ± 50 B.P. (Beta 142178). The result at 2 sigma is cal. A.D. 1155 to 1285 with an estimated C13/C12 ratio of -25 0/00 (Talma and Vogel 1993, Stuiver et al. 1998). A bulk charcoal and ash sample from a lens at a habitation site (FAN 1-9) dated to 620 ± 60 B.P. (Beta 141927). The result at 2 sigma is cal. A.D. 1275 to 1425 with an estimated C13/C12 ratio of -25 0/00 (Talma and Vogel 1993; Stuiver et al. 1998). Sinoto reports an overlapping date of 880 ± 85 B.P. (GaK-4557) on unidentified charcoal from a habitation deposit (FAN 1-6). He also gives a much earlier date of 1560 ± 85 (GaK-4558) that is an outlier from all the other archaeological dates from the archipelago (Sinoto 1973:290-91). It - 163 should be noted that the occupation of Kiritimati falls within the occupation period of Tabuaeran.

Occupants of this village were agriculturalists. Carbonised coconut endocarp and Pandanus keys as well as pearlshell coconut scrapers were recovered from habitation features (FAN 1-9, 1-10). Wet agriculture is suggested by the proximity to the freshwater marsh, used for taro (Colocasia esculenta and Cyrtosperma chamissonis) production today. Fifteen hectares of anthropogenic soil deposits at the village complex indicate dry agriculture. Protein subsistence came primarily from the sea (fish and molluscs), although small amounts of turtle and bird bones were recovered. These latter may well have been brought from Kiritimati where they would have been much more abundant.

The surveys and excavations disclosed the substantial differences in the archaeological records of these two neighbouring islands. Evidence supports minor and intermittent occupation on Kiritimati, with most sites representing specialised activities related to resource exploitation. The few habitation sites have minimal deposits. In contrast, Tabuaeran has a nucleated agricultural settlement and relatively rich and deep deposits. Subsistence on Kiritimati is based primarily on highly valued foods such as turtles and birds, whereas protein on Tabuaeran came mostly from lower status foods, molluscs and fish. Thus, we hypothesise that the people of Tabuaeran incorporated Kiritimati as an outlying islet to their own atoll, to which they occasionally sailed in order to exploit the abundant faunal resources.

Kiritimati and Tabuaeran are only 140nm apart, an overnight sail by canoe. They present wide angles, 20° and 16° respectively, using a 30nm bird-screen (the flight radius of boobies), making them easy targets. These angles are well above the lower limit of navigational feasibility (7.5°) and equivalent to the target angles of most traditional voyages (between 10° and 18°) according to Lewis (1975:222-31). The target angle of Kiritimati is even wider than the above figure if factors, such as the stationary bank of cumulus clouds 9 and the screen of frigate birds with its 75nm radius (according to Gladwin 1970:197), are taken into account. Both of these factors allowed the authors, even with their minimal navigation skills, to recognise Kiritimati from 60 miles away.

Although Tabuaeran lacks these target-expanding factors, it does have a safety net of islands lying downwind (Teraina, Palmyra and Kingman Reef). Voyages between these islands would have offered few navigational challenges, although the trade winds would have made sailing in one - 164 direction easier during much of the year (Finney 1985, Irwin 1992). The voyage from Kiritimati to Tabuaeran is usually downwind and down current. The reverse voyage “up” or against the direction of the trade wind to Kiritimati would require tacking or paddling, but more likely waiting for favourable conditions on Tabuaeran. These observations about voyaging suggest that Tabuaeran Islanders would have made the short passage to Kiritimati to hunt birds and turtles during the months of March and April, when the prevailing trades are replaced by brief spells of northerlies or southerlies (according to the Pilot Chart of the Pacific Ocean 1971), and sailed home when the trades resumed.

Kiritimati would have been advantageous, not only as a resource base for Tabuaeran, but as a way-point for voyages to more distant archipelagos. Its position, at the windward end of the northern Lines, gives the most favourable angle to begin voyages to the closest neighbouring islands, either across the wind to the Cooks or slightly into the wind to Malden, the Societies or even Hawai'i. Kiritimati is also the most prudent landfall for return voyages, again because it offers the biggest target angle and its windward position minimises the chances of being set too far downwind.

Archaeological support for inter-archipelago voyaging throughout central Polynesia, from the Marquesas to perhaps the southern Cooks, comes from the six different basalt types represented in our collections from the two islands (Di Piazza and Pearthree n.d.a). Two additional sources may be present at Site 3 (Anderson et al. 2000:285-86).

These points, taken together, suggest that Kiritimati would have served as an intermediate stop or way-point on long voyages, allowing navigators to adjust their courses, to await conditions favourable to continuing onward, to restock provisions or to offer prayers for protection. Three of the five turtle islands (Pikelot, West Faiu and Gafarut) noted above have been used as intermediate way-points for long voyages between the Carolines and the Marianas, as was Rose Atoll for trips between Pukapuka and Samoa (Lewis 1975:32,52,147,225-26; Brower 1983:110,119).

Throughout this paper we have taken a regional approach, pioneered by Weisler (1996), that allowed us to understand the contrasting settlement patterns on two adjacent “mystery islands”. In our view, wet Tabuaeran was the primary locus of settlement, with dry Kiritimati acting as a peripheral base for specialised activities. This analysis challenges the prevailing “mystery island” paradigm which focuses on resource scarcity and isolation. Not only have we noted the rich agricultural potential of Tabuaeran and the - 165 enormous faunal resources on Kiritimati, but also the pervasive evidence of long-distance interaction found on both islands.

Building on what we learned in the northern Lines helped us to recognise that the “mystery islands” should not be treated as a single phenomenon but rather as three classes, defined by the role each island played in regional interaction. These classes might be characterised as mother communities, satellites and isolates. The first two terms are borrowed from Weisler (1996), but used in a somewhat different context.

“Mother communities” would include islands that once supported sizeable resident populations and had nearby satellites. At present, there appear to be four islands in this class: Tabuaeran, Pitcairn, Orona in the Phoenix Islands and, perhaps, Caroline in the Lines. We might also add Norfolk Island to this list, although it has no satellites (Specht 1984). Regardless of size, each of these islands offers the best combination of subsistence resources, especially agricultural potential, within the range of an “overnight voyage” in their respective archipelagos (Weisler 1996, Garnett 1983). An “overnight voyage” as used here (140nm) is somewhat longer than the 100nm noted by Marck(1986).

Why these rich islands were abandoned remains a mystery, the more so when we note that some of them (Tabuaeran, Pitcairn) support large populations today and have done so for more than a century (Bryan 1942, Garnett 1983). Rather than evoking environmental crunches to explain human extinction, the numerous basalt adzes and the abundance of timber for canoe building should remind us that these people were experienced navigators, capable at any time of migrating to other archipelagos.

“Satellites” are dry islands located within an overnight voyage of a richer land, either a high island or one of the “mother communities”. Satellites of mother communities would include Enderbury, Kanton and Manra of Orona in the Phoenix group, Kiritimati and Teraina of Tabuaeran, Flint of Caroline in the southern Lines, Henderson and Ducie of Pitcairn (Bryan 1942, Emory 1975, Garnett 1983, Weisler 1997). Satellites of currently occupied islands include Temoe of Mangareva, Nihoa of Niihau, Suvarov of Nassau or Manihiki, Swains of Fakaofo, Rose of Tau and Walpole of the Isle of Pines (New Caledonia) (Emory 1939b, Cleghorn 1988, Kirch 1988). For these 14 satellite islands, “abandonment” is not really pertinent since—if the hypothesis posed here is correct—they never supported resident populations. In various parts of Micronesia today, one can observe “mystery islands” in the making. Sailors from Sonsorol or Tobi, south of Palau, rarely visit the uninhabited islands of Merir and Helen Reef to catch turtles, nor do canoes set out from Lamotrek or Etal for the five turtle islands of the central Carolines (Johannes 1981:56,87,88; Ridgell et al. 1994:202).

- 166

“Isolates” are inaccessible, tiny and/or dry islands. They comprise Necker, Maiden, Howland, Palmerston and, perhaps, Raoul (Emory 1934, Anderson 1980, Cleghorn 1988, Kirch 1988). These seem to be the only ones that fit the paradigmatic picture of a “mystery island”, and whose abandonment may be explained by scarcity of resources and/or isolation, hypotheses that have previously been applied to all the “mystery islands”.

We thank the Sous-Direction des Sciences Sociales, Humaines et de l'Archéologie du Ministère des Affaires Etrangères, Paris, and the Centre de Recherche et de Documentation sur l'Océanie (CREDO), laboratoire du CNRS, Marseille for financial support. We would also like to thank Tamaetera Teaotai, Director of Um'anibong, Tarawa, and the Ministry of Line and Phoenix Islands Development, Kiritimati, for permission to work in the Line Islands. Additional assistance was given by the Wildlife Unit and the Office of Tourism and the people of Kiritimati and Tabuaeran, as well as Roy Carvalho. We are most grateful to them all.

- 167

• Amerson, A.B., 1969. Ornithology of the Marshall and Gilbert Islands. Washington: Atoll Research Bulletin 127.
• Anderson, A., 1980. The archaeology of Raoul Island (Kermadecs) and its place in the settlement history of Polynesia. Archaeology and Physical Anthropology in Oceania, 15:131-41.
• Anderson, A., Wallin, P., Martinsson-Wallin, H. and Hope, G., 2000. Towards a first prehistory of Kiritimati (Christmas) Island, Republic of Kiribati. Journal of the Polynesian Society, 109(3):273-91.

• Bellwood, P., 1987 [1978]. The Polynesians: Prehistory of an Island People. London: Thames and Hudson.
• Best, S., 1988. Tokelau archaeology: a preliminary report of an initial survey and excavations. Bulletin of the Indo-Pacific Prehistory Association, 8:104-118.

- 168

• Brower, K., 1983. A Song for Satawal. Harmondsworth: Penguin Books.
• Bryan, E.H., 1942. American Polynesia and the Hawaiian Chain. Honolulu: Tongg Publishing Company.

• Cleghorn, P.L., 1988. The settlement and abandonment of two Hawaiian outposts: Nihoa and Necker Islands. Bishop Museum Occasional Papers, 28:35-49.

• Davidson, J., 1967. Archaeology on coral atolls. In G. Highland, et al. (eds), Polynesian Culture History: Essays in Honor of Kenneth P. Emory. Honolulu: Bernice P. Bishop Museum Special Publication 56, pp.363-75.
• Di Piazza, A. and E. Pearthree (n.d.a). Voyaging and basalt exchange in the Phoenix and Line archipelagoes: The view point from three Mystery Islands. Archaeology in Oceania. In press.
• ——(n.d.b). L'Art d'être pirogues de voyage en Océanie insulaire. Journal de la Société des Ocēeanistes. In press.
• Duff, R., 1974. Stone structure types: The Koutu Ariki or Chiefly Court. In M.M. Trotter (ed.), Prehistory of the Southern Cook Islands. Christchurch: Canterbury Museum Bulletin 6, pp.28-33.

• Emory, K.P., 1934. Archaeology of the Pacific Equatorial Islands. Honolulu: Bernice P. Bishop Museum Bulletin 123.
• ——1939a. Additional notes on the archaeology of Fanning Island. Honolulu: Bishop Museum Occasional Papers 15.
• ——1939b. Archaeology of Mangareva and Neighbouring Atolls. Honolulu: Bernice P. Bishop Museum Bulletin 163.
• ——1970. A re-examination of East Polynesian marae: Many marae later. In R.C. Green and M. Kelly (eds), Studies in Oceanic Culture History. Pacific Anthropological Records 11. Honolulu: Bernice P. Bishop Museum, pp.73-92.
• ——1975. Material Culture of the Tuamotu Archipelago. Pacific Anthropological Records 22. Honolulu: Bernice P. Bishop Museum

• Finney, B.R., 1985. Anomalous westerlies, El Niño, and the colonization of Polynesia. American Anthropologist, 87:9-26.

• Garnett, M.C., 1983. A management plan for nature conservation in the Line and Phoenix Islands. Tarawa: unpublished report.
• Gérard, B., 1974. Contribution à l'étude des structures lithiques à caractére religieux aux îles de la Société. Papeete: Unpublished Report, Orstom.
• Gillett, R., 1987. Traditional Tuna Fishing: A Study at Satawal, Central Caroline Islands. Bishop Museum Bulletin in Anthropology, I. Honolulu: Bishop Museum Press.
• Gladwin, T., 1970. East in a Big Bird. Navigation and Logic on Puluwat Atoll. Cambridge: Harvard University Press.

• Haddon, A.C., and Hornell J., 1975. Canoes of Oceania. Bernice P. Bishop Museum Special Publications 27, 28 and 29. Honolulu: Bishop Museum Press.
• Hiroa, T.R. (P.H. Buck), 1932. Ethnology of Tongareva. Honolulu: Bernice P. Bishop Museum Bulletin 92.

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

• Johannes, R.E., 1981. Words of the Lagoon. Fishing and Marine Lore in the Palau District of Micronesia. Berkeley: University of California Press.

- 169

• Kellum-Ottino, M., 1971. Archéologie d'une vallée des îles Marquises. Evolution des structures de l'habitat à Hane, Ua Huka. Paris: Publication de la Société des Océanistes 26.
• Kirch, P.V., 1988. Polynesia's Mystery Islands. Archaeology, May/June:26-31.

• Lavondes, A., 1976. La Culture matérielle en Polynésie et les collections du Musée de Tahiti et des Îles. Tahiti: Orstom. Typescript catalogue.
• Leach, F. and Ward, G., 1981. Archaeology on Kapingamarangi Atoll. Studies in Prehistoric Anthropology 16. Dunedin: University of Otago.
• Lewis, D., 1975. We, the Navigators. Honolulu: University of Hawai'i Press.

• Marck, J.C., 1986. Micronesian dialects and the overnight voyage. Journal of the Polynesian Society, 95(2):253-8.
• Mueller-Dombois, D. and Fosberg, F.R., 1998. Vegetation of the Tropical Pacific Islands. New York: Springer Verlag.

• Pilot Charts of the Pacific Ocean, 1971 NOAA, H.O. 1401.

• Ridgell, R., M. Ikea and I. Uruo, 1994. The persistence of central Carolinian navigation. ISLA, A Journal of Micronesian Studies, 2(2):181-206.
• Rougier, E., 1914. Ile Christmas, South Seas (Océanie), Coconuts, birds, fishes, etc. Brioude: L. Watel.
• ——1917. Ile de Christmas. Bulletin de la Société d'Etudes Océaniennes, 1:25-30.
• Roy, K.J., 1973. Note on phosphate rock at Fanning Island. In K.E. Chave and E.A. Kay (eds), Fanning Island Expedition, July and August 1972. Final report for National Science Foundation, pp. 193-9.

• 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. Final report for National Science Foundation, pp.283-99.
• Specht, J., 1984. The Prehistoric Archaeology of Norfolk Island. Pacific Anthropological Records 34. Honolulu: Bernice P. Bishop Museum.
• Stuiver, M., P.J. Reimer, E. Bard, J.W. Beck, G.S. Burr, K.A. Hughen, B. Kromer, G. McCormac, J. van der Plicht and M. Spurk, 1998. INTCAL98 Radiocarbon Age Calibration, 24,000-0 cal BP Radiocarbon, 40(3): 1041-83.

• Talma, A.S. and Vogel, J.C., 1993. A simplified approach to calibrating C14 dates. Radiocarbon, 35(2):317-22.
• Terrell, J., 1986. Prehistory in the Pacific Islands. Cambridge: Cambridge University Press.
• Thomas, S.T., 1987. The Last Navigator. New York: Ballantine Books.

• Weisler, M.I., 1996. Taking the mystery out of the Polynesian “mystery” islands: A case study from Mangareva and the Pitcairn Group. In J. Davidson, G. Irwin, B.F. Leach, A. Pawley and D. Brown (eds), Oceanic Culture History, Essays in Honour of Roger Green. Dunedin: New Zealand Journal of Archaeology Special Publication, pp.615-29.
• ——(ed.), 1997. Prehistoric Long-Distance Interaction in Oceania: An Interdisciplinary Approach. Auckland: New Zealand Archaeological Association Monograph 21.

- 170

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

• Yen, D.E. and Gordon, J., (eds), 1973. Anuta: A Polynesian Outlier in the Solomon Islands. Pacific Anthropological Records 21. Honolulu: Bernice P. Bishop Museum.