Ana Manuku: a prehistoric ritualistic site on Mangaia, Cook Islands.

STEADMAN, DAVID W. ; ANTON, SUSAN C. ; KIRCH, PATRICK V. 等

Introduction

Archaeological and palaeontological research has demonstrated the often profound impact of colonizing humans on previously isolated island ecosystems, including forest clearance, erosion and alluvial deposition, biotic scarcity and extinction, and introduction of non-indigenous species (Kirch 1983; 1997; 2000; Steadman 1995; Kirch & Hunt 1997). Such human-induced environmental changes are well documented on Pacifc islands such as Aneityum (Hope & Spriggs 1982), Rapanui (Flenley et al. 1991; Steadman et al. 1994), the Marquesas (Steadman & Rolett 1996; Rolett 1998) and the Hawaiian Islands (James et al. 1987; Athens 1997).

On many high islands in Polynesia, late prehistoric populations reached densities of [is greater than] 100 persons/sq. km (Kirch 2000), sustained by major agricultural developments such as dryland field systems or converting valley bottoms to irrigated pondfields for taro. While improving crop yields, these capital-intensive production systems also contributed to a sociopolitical context with considerable competition for land, and often open conflict between social groups. At their prehistoric endpoints of indigenous development (i.e. their fateful encounters with the West), many Polynesian societies had:

1 high-density populations;

2 highly intensified production systems;

3 reduced natural resources;

4 hierarchical political control and social stratification; and

5 intense competition for land and other resources.

In some societies, such as Rapanui, Marquesas, or Mangareva, this situation was also marked, according to ethographic and ethnohistoric sources, by aggression that included warfare or raiding between competing groups, taking of enemy victims for sacrificial offerings and cannibalism (Kirch 2000).

Our research on Mangaia, Southern Cook Islands, has provided substantial evidence in support of the general model of environmental and sociopolitical change outlined above (Steadman & Kirch 1990; Kirch 1996; Kirch et al. 1992; 1995). In this paper we present new data from a Mangaian rock-shelter site that further elucidate the social (and ritual?) consequences of inter-group competition in the late prehistoric period.

Mangaia (52 sq. km) is divided into six traditional districts that begin on the limestone coast and meet atop the volcanic uplands (FIGURE 1). At least 2000-3000 people lived on Mangaia at European contact (Hiroa 1934: 6), yielding a population density of 39-58 persons/sq. km. Because much of the limestone region lacks soil, the density per area of arable land exceeded 100 persons/sq. km. Prehistoric landscape and vegetational changes have been interpreted from sediment cores from all districts (Ellison 1994; Kirch & Ellison 1994). Among Mangaia's numerous archaeological sites, Tangatatau Rockshelter (site MAN-44, Veitatei District) has yielded the richest and longest (c. AD 1000-1750) prehistoric sequence of artefacts, midden bone, shell and plant material (Kirch 1996; Kirch et al. 1992, 1995). Consumption of vertebrates at MAN-44 changed considerably during this time (see Discussion).

[Figure 1 ILLUSTRATION OMITTED]

To test whether the cultural/faunal sequence at MAN-44 was characteristic of the entire island, we undertook excavations at a bone-rich rock-shelter called Ana Manuku (site MAN-84) in Keia District, 2 km north of MAN-44. Here we report the stratigraphy, chronology, faunal sequence and cultural context of MAN-84, followed by a discussion of how this new site influences our thinking on the cultural sequence of Mangaia.

Methods

Kirch excavated two 1-sq. m test units (D31, E18) at MAN-84 in 1991 (FIGURE 2). Steadman and Anton excavated nine 1-sq. m units in 1997. Because of differences in bone retrieval methods between the 1991 and 1997 excavations, the tabulated faunal data in this paper are based only on the units excavated in 1997 (C28-C32, D28-D30, D32). Major sedimentary changes were designated as layers, which we subdivided arbitrarily into levels of c. 10 cm thickness. All bones, artefacts, charcoal etc. were recovered in situ or with sieves (dry-screening through 3.3 mm mesh, subsampled by water-screening through 2.0 mm mesh). Steadman and Kirch analysed the artefacts. Steadman identified all non-human bones. For human bones, Anton determined the number of identified specimens for each layer (following cross-conjoining), the extent of heating (Buikstra & Ubelaker 1994), evidence of perimortem breakage and spiral fracture (White 1992:135; Turner & Turner 1999), and presence of stone tool marks, impact marks, flake scars, carnivore tooth marks and rodent gnawing (Binford 1981; Blumenschine 1988; Fisher 1995; Hagland 1997). Here we summarize the main points of these osteological analyses. Full details of the human bones will be reported elsewhere by Anton.

[Figure 2 ILLUSTRATION OMITTED]

Ana Manuku Rockshelter (Site MAN-84)

Site setting

Ana Manuku Rockshelter is developed at the base of the inner limestone cliff that encircles Mangaia (FIGURE 1). The floor area inside the dripline is c. 75 sq. m (FIGURE 2). Ana Manuku lies only 60 m from Keia Swamp, a large and relatively drought-resistant site of irrigated pondfield taro (Colocasia esculenta). MAN-84 also lies 75 m from a marae or temple named Tukituki Mata (described in Hiroa 1934: 139, 174-6; and Bellwood 1978: 142-, 159-60) and 600 m from marae Aka Oro. Neither Tukituki Mata or Aka Oro has been excavated.

Stratigraphy

Modern goat faeces (Layer I; thickness 10-30 cm) covered the entire cultural deposit (FIGURE 3). The two prehistoric occupation Layers (II, III) are cobbly, pebbly, slightly sandy, ashy silt deposits dominated by combustion features with concentrations of charcoal and burned oven stones (angular to subangular, mostly 8-16 cm diameter, with breakage caused by heating). We interpret these features as Polynesian earth ovens (umu) and the rake-out from these ovens. The 10-sq. m excavation block (FIGURE 4) had seven ovens in Layer II (areal extent 30x45-72x105 cm, thickness 8-20 cm) and five ovens in Layer III (areal extent 36x48-51x58 cm, thickness 12-24 cm).

[Figures 3-4 ILLUSTRATION OMITTED]

Layer II (Munsell very dark grayish brown 10YR 3/2) is more compacted with better developed bedding planes than Layer III (yellowish brown 10YR 5/4). This reflects mixing of Layer IV (strong brown 7.5 YR 5/6) into Layer III, undoubtedly while digging the ovens. Layer IV differs from both cultural strata in lacking charcoal, ash, cultural features, artefacts, marine molluscs and urchins, and in having few bones other than of birds. We interpret Layer IV as non-cultural in origin.

Chronology

The age of MAN-84 is based on accelerator-mass spectrometer (AMS) radiocarbon ([sup.14]C) dates on human bones of nine different individuals from unit C32, as well as six conventional (beta-decay) [sup.14]C dates on wood charcoal (TABLE 1). The AMS [sup.14]C dates on bone are highly consistent with each other (FIGURE 5). Three charcoal dates (Beta-48352, Beta-111939, Beta-111943) from oven features broadly overlap five human bone AMS dates (Beta-113847, Beta-113850-113853) from the same units and levels. The other three charcoal dates (Beta-111940-111942) are unlikely to be accurate age estimates of the cultural strata. A young age determination is explicable for Beta-111940, a piece of charcoal from the contact of Layers I and II, and Beta-111941, charcoal from a loose context near the wall of the rockshelter. Being from an oven, the young age estimate for Beta-111942 is less explicable, but might reflect downward displacement of the dated charcoal sample, or possibly contamination by goat urine from Layer I. That AMS [sup.14]C dates on individual bones are more reliable than those on small pieces of charcoal has been demonstrated in other cave and rock-shelter sites because charcoal tends to disperse more and to absorb contaminants more than bone (James et al. 1987; Steadman et al. 1991).

[Figure 5 ILLUSTRATION OMITTED]

TABLE 1. Radiocarbon dates, Ana Manuku, Mangaia. All dates on human bone are AMS. All dates on wood charcoal are conventional (beta-decay) determinations. All analyses by Beta Analytic Inc., Miami, Florida. CALIB = calendar date following Stuiver et al. (1998). OxCal = calendar date following OxCal version 3.3 (1999).

 layer/
lab no. material dated unit level

Beta-114104 metacarpal C32 II/3

Beta-113846 juv. proximal hand C32 II/4
Beta-113847 thoracic vertebra C32 III/5

Beta-113848 cervical vertebra C32 III/6
Beta-113849 sacral arch C32 III/7

Beta-113850 cervical vertebra C32 III/8

Beta-113851 thoracic vertebra C32 III/8

Beta-113852 thoracic vertebra C32 III/8

Beta-113853 juv. rib C32 III/8

Beta-111940 charcoal D30 II/3

Beta-111941 charcoal C30 II/4

Beta-111943 charcoal C32 III/5
Beta-111942 charcoal C32 III/7
Beta-111939 charcoal C30 III/8
Beta-48352 charcoal D31 III

 [sup.13]C/
 measured [sup.12]C conventional
lab no. [sup.14] C age ratio [sup.14]C age

Beta-114104 440 [+ or -] 30 -19.9 520 [+ or -] 30

Beta-113846 290 [+ or -] 70 -18.4 390 [+ or -] 70
Beta-113847 380 [+ or -] 50 -18.0 500 [+ or -] 50

Beta-113848 500 [+ or -] 50 -19.4 590 [+ or -] 50
Beta-113849 410 [+ or -] 50 -20.1 490 [+ or -] 50

Beta-113850 440 [+ or -] 40 -20.1 520 [+ or -] 40

Beta-113851 380 [+ or -] 40 -18.6 480 [+ or -] 40

Beta-113852 350 [+ or -] 40 -20.0 430 [+ or -] 40

Beta-113853 420 [+ or -] 40 -20.9 480 [+ or -] 40

Beta-111940 20 [+ or -] 50 -26.0 0 [+ or -] 50

Beta-111941 240 [+ or -] 60 -26.3 220 [+ or -] 60

Beta-111943 730 [+ or -] 60 -32.1 620 [+ or -] 60
Beta-111942 190 [+ or -] 50 -27.2 160 [+ or -] 50
Beta-111939 720 [+ or -] 70 -34.3 580 [+ or -] 70
Beta-48352 810 [+ or -] 90 -27.8 770 [+ or -] 90

 CALIB OxCal
lab no. cal AD (2[Sigma]) cal AD (2[Sigma])

Beta-114104 1400-1440 1320-1350 (0.10)
 1390-1440 (0.85)
Beta-113846 1420-1655 1420-1650 (0.95)
Beta-113847 1395-1470 1300-1370 (0.15)
 1380-1490 (0.80)
Beta-113848 1295-1430 1280-1430 (0.95)
Beta-113849 1400-1475 1300-1360 (0.11)
 1380-1500 (0.85)
Beta-113850 1395-1450 1300-1360 (0.19)
 1380-1450 (0.76)
Beta-113851 1410-1470 1330-1350 (0.02)
 1390-1490 (0.93)
Beta-113852 1425-1515 1410-1530 (0.86)
 1580-1630 (0.10)
Beta-113853 1410-1470 1330-1350 (0.02)
 1390-1490 (0.98)
Beta-111940 1890-1905 1680-1730 (0.18)
 1810-1930 (0.57)
 1940- ... (0.21)
Beta-111941 1650-1680 1510-1890 (0.84)
 1745-1805 1910-1960 (0.11)
 1935-1950
Beta-111943 1280-1425 1280-1420 (0.95)
Beta-111942 1655-1950 1650-1960 (0.95)
Beta-111939 1285-1450 1280-1450 (0.95)
Beta-48352 1040-1400 1030-1330 (0.86)
 1340-1400 (0.09)

The nine AMS [sup.14]C dates on human bone and three of the charcoal dates suggest that the bones were deposited in Layers II and III primarily if not exclusively from cal AD 1390-1470. Because the AMS [sup.14]C dates all overlap at 2[Sigma] (and most overlap at only 1[Sigma], we cannot rule out the possibility that all human bone was deposited at MAN-84 within a few years rather than decades. Layers II and III each may represent only a single event, perhaps in successive years or even during the same year.

Artefacts

Unlike other rock-shelter sites on Mangaia, MAN-84 lacked most of the typical Eastern Polynesian artefacts, such as adzes, fishhooks, scrapers, files, tattoo needles, etc. The only basalt artefact was a broken pounder (TABLE 2). The bifacial blade of fine-grained limestone was retouched on both edges, as was the toothed limestone coconut grater. The other four artefacts were probably ornaments. The first is a complete, two-holed barb of a trolling lure (bonito hook). This lure must be a trade item because pearlshell (Pinctada margaritifera) does not occur on Mangaia. The nearest possible sources are Rarotonga, Manuae or Aitutaki, 200-400 km to the northwest (Kirch et al. 1992,1995). Trolling lures are not among the hundreds of fishhooks recovered at MAN-44; the specimen from MAN-84 probably was a prestige item worn as an ornament. All three bone artefacts are carved too extensively to be identified reliably to species, although remnant surface textures on the two beads suggest dog bone more than pig or human bone. A similar style of bone bead is illustrated in Gruning (1937) and Hiroa (1944: figure 62). The bone ball is grooved on one side, as depicted in Hiroa (1944: figure 58e), although it lacks the lug for attachment to a necklace.

TABLE 2. Artefact summary, Ann Manuku, Mangaia, by stratigraphic Layer (II & III are cultural; IV is pre-cultural). Based on recovery in situ and dry-screening through 3.3-mm mesh.

artefact type II III IV total

pearlshell trolling lure 1 0 0 1
bone bead 1 1 0 2
bone ball 1 0 0 1
basalt pounder 1 0 0 1
limestone blade 0 1 0 1
limestone coconut grater 1 0 0 1

total 5 2 0 7

Vertebrate fauna

The 1997 excavations at MAN-84 yielded a bone assemblage (TABLE 3) very different from that at any other rock-shelter site on Mangaia. Pig is represented only by one bone (Unit D31, Layer II, excavated in 1991). Dog bone also is extremely rare, represented by two fragments (Unit D31, Layer II; Unit D29, Layer II). Human bones are abundant in Layer II and especially in Layer III, both inside and outside the ovens. All clearly in a midden rather than burial context, 96% of the 1864 human bones were recovered in situ or through dry-screening with 3.3 mm mesh. Like the human bones, fish and rat bones are rare in Layer IV, where the few such bones recovered almost certainly represent downward mixing from Layer III, because:

TABLE 3. Major categories of vertebrates, Ann Manuku, Mangaia, by stratigraphic Layer (II & III are cultural; IV is pre-cultural). Based on recovery in situ, dry-screening through 3.3-mm mesh, and wet-screening through 2.0-mm mesh. Numbers are NISP (n umber of identiffed specimens).

taxon II III IV total

human 341 1515 8 1864
fish 1134 716 45 1945
rat 565 1030 27 1622
bird 48 287 1310 1635
lizard 2 13 0 15
fruit bat 1 1 0 2
dog 1 0 0 1

total 2092 3602 1390 7084

1 they occur only in the uppermost Layer IV;

2 their preservational qualities resemble those of bones from Layer III; and

3 the human bones often conjoin with specimens from Layer III.

Bird bones are scarce in Layer II, increase in Layer III and then increase dramatically to dominate the non-cultural Layer IV. Most bird bones represent extirpated species of seabirds and landbirds, to be reported elsewhere by Steadman. The greater abundance of bird bones in Layer III than in Layer II is because bones originally deposited in Layer IV were reworked upward during digging the ovens in Layer III. Because burned bones are more likely to represent cultural than non-cultural deposition, evidence for this is seen in the percentage of burned bird bones, which decreases from Layer II (69%) to Layer III (28%) to Layer IV (0.3%).

Human bones

The 1864 human bones (TABLES 3, 4) consist of burned, highly fragmentary and isolated elements representing at least 41 individuals ranging in age from foetal/newborn to adults probably of both sexes. Minimum numbers of individuals are four adults and seven subadults from Layer II, and 10 adults and 20 subadults from Layer III. Relative to their occurrence in the complete skeleton, hand and foot phalanges are slightly overrepresented (17% vs 13.6%) and cranial bones are underrepresented (1.3% vs 10.7%), but no element is entirely absent. All human bones are in a midden rather than funerary context, and we found no evidence of burials.

TABLE 4. Human bone, Ann Manuku, Mangaia, by stratigraphic Layer (II & III are cultural; IV is pre-cultural). Fragments that cross-conjoin between Units or Layers are reported in the larger numbered Unit or Layer. Based on recovery in situ, dry-screening through 3.3-mm mesh, and wet-screening through 2.0-mm mesh. Numbers are NISP (number of identified specimens).

Unit II III IV total

C28 32 87 2 121
C29 10 285 1 296
C30 22 407 1 433
C31 10 183 0 193
C32 36 290 0 326
D28 52 29 0 81
D29 28 29 0 57
D30 50 40 0 90
D32 101 162 4 267

total 341 1515 8 1864

The human bones exhibit considerable human-induced modification. Although some pieces are unheated (Munsell 10YR 8/3-7/3), 84% are at least lightly browned (Munsell 7.5YR 7/4 or darker), consistent with heating at low temperatures. 20% are heavily heated or burnt (Munsell 7.5YR 4/4 or darker), but [is less than] 1% is calcined (e.g. Munsell gley 8/N). All adult and older juvenile limb bones are highly fragmented including spiral and step (hinge) fractures, occasionally accompanied by percussion pits, flake scars or tool striae. Other elements, particularly ribs, exhibit `peeling' indicative of fracture when fresh (White 1992: 143). Differential colouring on either side of some ancient fractures indicate that the fracture preceded heating. We found no definitive evidence of damage by pigs, although rodent gnawing occurs as do tooth marks and breakage suggesting canid activity.

The quantity and appearance of the human bones leave little doubt that people brought the human bone to the site, since the dense concentration of burned, broken and battered human remains precludes casual introduction by pigs or dogs. The non-associated, fragmentary nature of most remains is inconsistent with other mortuary practices on the island, such as burial caves. Some of the burials in Te Rua Rere cave (Tava'enga District) are coeval with the human remains at MAN-84 (Anton & Steadman in press). Thus if the site's human bones represent some form of mortuary ritual, it was not the only mortuary practice on Mangaia at that time. Since very little of the human bone is calcined a crematorium is unlikely (Correia 1997; Anton & Steadman 1998), even though other mortuary rituals were practised at that time. What seems clear is that human body parts, as well as some small fish, rats and a few birds, were cooked in earth ovens at MAN-84.

Discussion

Comparisons with Tangatatau Rockshelter (site MAN-44)

Mangaia's other extensively excavated rockshelter is Tangatatau, which has produced one of Eastern Polynesia's best chrono-stratigraphic sequences of artefacts (570 adzes, fishhooks, stone pounders, scrapers, files, bone needles, etc.), bone, shell and plant materials (Kirch et al. 1992; 1995; Kirch 1996). The prehistoric cultural strata range in age from c. AD 1000 (analytic zone 2) to AD 1750 (zone 15). Faunal evidence ([is greater than] 30,000 identified bones) suggests that the site does not reflect the earliest occupation of Mangaia because bones of non-native species (rat, dog, pig, chicken) occur regularly in zone 2. One would expect bones of these species to be scarce or absent in zone 2 if it represented initial occupation. Also, zone 2 has little evidence for heavy exploitation of seabirds, as elsewhere in Eastern Polynesia (Steadman 1995). Zone 2 of MAN-44 may represent the first long-term occupation of Mangaia's interior, since earlier habitation may have been intermittent and restricted to the coast.

While MAN-44 was occupied, consumption of vertebrates changed considerably, and bones of pig, dog, rat, and chicken increased in the middle levels (zones 8-10). All of these except rat declined in upper prehistoric levels, perhaps because resource stress forced people to eat lower on the food chain. Birds are rich in species only in early levels (c. AD 1000-1300; zones 2-4). The dramatic decline in native species of birds between zone 4 (22 species) and zone 5 (5 species) reflects extinctions as interior forests were felled to create an agricultural landscape to feed the growing population (Kirch 1996; Steadman 1997). By c. AD 1300, most species of birds were either extinct or too rare to be sampled. Zones 5 to 8 (c. AD 1300-1500) are roughly coeval with Layers II and III at MAN-84 (C. AD 1390-1470). MAN-44 differs dramatically in its scarcity of human bones and abundance of fish, pig and chicken bones (TABLE 5). Its midden deposits differ further from those at MAN-84 in their abundance of shellfish and in being more intensely burned (more ashy), with the more poorly defined ovens reflecting repeated cycles of cooking. This in turn suggests long-term occupation, as substantiated by radiocarbon dates spanning nearly a millennium.

TABLE 5. Comparison of vertebrate assemblages from coeval strata at Mangaian rockshelters: Layers II and III at site MAN-84 versus analytic zones 5-8 of site MAN-44. Minor categories (sea turtle, lizard and fruit bat) are excluded. Data for MAN-84 are from TABLES 3 and 4 herein. Data for MAN-44 are from Kirch et al. (1995).

 NISP % NISP

taxon MAN-84 MAN-44 MAN-84 MAN-44

human 1856 5 32.69 0.04
fish 1900 11,789 33.47 95.10
rat 1595 408 28.10 3.29
dog 1 5 0.02 0.04
pig 0 91 0.00 0.73
chicken 0 43 0.00 0.35
native bird 325 55 5.72 0.44

total 5677 12,396 100 100

The artefact concentration at MAN-44 (ignoring innumerable basalt flakes) is c. 35 per excavation unit, compared to less than one at MAN-84 (TABLE 2). The absence or extreme scarcity at MAN-84 of adzes, adze preforms, adze flakes, fishhooks, fishhook blanks, scrapers, coral or urchin flies, tattoo needles and stone pounders points to a lack of domestic activities and implies that it was not a habitation site.

Human remains from Mangaian rock-shelters (including MAN-44, MAN-82, MAN-83, MAN-84, and MAN-87) are not in a funerary context but are isolated, fragmentary elements most of which exhibit human modification, especially burning (Anton & Anderson 1993). Despite preservational similarities, the MAN-84 human bone assemblage is much larger (NISP = 1864) than at any other Mangaian rock-shelter (NISPs from 1-17), and indeed is even an order of magnitude greater than the largest non-human, large mammal bone assemblage from Mangaia, namely that of pig and dog from MAN-44 (combined NISP from all zones = 218). The MAN-84 human bone sample has many more and much younger juveniles than at the other rock-shelters. Also, only at MAN-84 are there stone tool marks on some of the human bones.

MAN-84 in an ethnographic context

The large number ([is greater than] 40) of human individuals associated with earth ovens suggests that MAN-84 was a special-use site, perhaps ritualistic in nature, rather than a habitation site. The Mangaian ethnohistoric record, compiled by missionaries and colonial officials, is replete with stories of: human sacrifice to ensure victory, to install a paramount chief or to ensure plentiful harvests; murder and warfare to gain arable land or to redress past wrongs; and cannibalism of either a ritualistic or nutritive nature. We review these reports and the accounts of Mangaian burial customs in order to help to understand the possible context of MAN-84.

The site's name, Ana Manuku, is interesting. Manuku means `to slip off or out, to come loose' (Savage 1980: 140). Before excavating the site, Mangaians told us that manuku means `dislocated' in the sense of a dislocated joint, such as an elbow or knee. Ana means `cave', although some Mangaians called the site Are Manuku with Are meaning `house' or `place', because the rock-shelter does not attain total darkness and is not truly a cave.

The human remains at MAN-84 are disarticulated (`dislocated'?), fragmentary, and burned but not cremated. This is not consistent with burial customs described ethnohistorically or seen in prehistoric burials caves. Usually within a day of death a corpse was wrapped in cloth, then disposed of either in a shallow pit, within a marae, or in a cave or chasm in the makatea (Hiroa 1934: 190). Those buried in the ground were placed face down in a flexed position with limbs secured by a sennit cord and the head toward the rising sun (Gill 1894). Most bodies, however, were hidden in secret caves or chasms to avoid interference with the body by enemies (Hiroa 1934: 191). Hiroa specifically suggested that a body, particularly that of a warrior, obtained by his enemies would be burned. If the human bones at MAN-84 reflect a large number of individual corpses that were ill-treated by their enemies, these might be expected to be more fully burnt, to have more complete elements and to represent solely or primarily adult males. For cave burials some secondary treatment of the corpse is acknowledged, entailing exposing it to the sun, rubbing it with oil and rewrapping it in cloth, but no dismemberment or intentional bone breakage is known (Hiroa 1934: 191; Katayama 1986; Ant6n & Steadman in press).

Sacrifice of a single human victim to the deity Rongo occurred on Mangaia to assure peace following a battle or to install a new paramount chief without warfare (Hiroa 1934: 37). Typically, a member of a defeated or weak tribe was killed by warriors of a victorious or powerful tribe, then laid out at the marae Aka Oro in Keia, some 600 m inland from MAN-84 (Hiroa 1934: 81-3, 98). The body later was transferred to the shore marae Orongo, also in Keia, where only his nose and ears were distributed among the leaders of each district and the rest of the body dumped behind the marae. The last of these sacrificial victims was said to have been taken in 1823. Although the importance of Keia District in ritual sacrifice is clear, the single (usually male) victim and the lack of dismemberment beyond the nose and ears is inconsistent with the large number of individuals, representing all age classes and both sexes, at MAN-84. Furthermore, the sacrificial victim for installation of a chief, having been dedicated to Rongo, would have been highly tapu and thus not acceptable to prepare as food.

Another marae in Keia is Tukituki Mata (MAN-28), only 75 m inland from MAN-84, on the path to Aka Oro. Tukituki Mata is described as a `secular marae' that was a favourite spot for public meetings to discuss politics or history in times of peace (Hiroa 1934: 139, 176). Such-meetings were said to have been followed by feasting on the `flesh of fish'. Tukituki means `the act of repeated pounding, bruising, braying, bumping, beating upon, slamming, banging, etc.' whereas mata means `the eyes or eyes of human beings, animals, birds and insects; the face of a human being or animal' (Savage 1980: 147,409). Tukituki Mata literally means to smash someone's face, especially the part containing the eyes. It is intriguing that the `flesh of fish' (rather than just `fish') was said to have been eaten at Tukituki Mata in late prehistoric times. On Mangaia (Hiroa 1934: 81) and in the Marquesas Islands (Handy 1923), human sacrificial victims were referred to euphemistically as `fish'.

The small area of arable land on Mangaia led to intense fighting over its control. Gill's informants recounted, in chronological order, 42 battles spanning a period of 300-400 years (Gill 1894: 308-11). Each battle led to a change in power and redistribution of arable land. A calendric chronology is not available for the first seven battles. Battle no. 8 was estimated to have been fought in c. AD 1570. The last battle was in c. AD 1828. Battles nos. 1 and 5 (before c. AD 1570) were fought in Keia District, as were nine subsequent battles. Not only were many warriors killed in most of these battles, but the surviving males and children of the defeated tribe often took refuge in the rugged limestone forest, sometimes resorting to cannibalism to survive (see below). The demographic distribution of the human bones at MAN-84 is inconsistent, however, with their being either the food remains or skeletal remains of a defeated war party. Also, the site is open on three sides, provides little concealment, and would be difficult to defend.

Among other ethnographically recorded large-scale murders to redress past wrongs, the `first oven of men' (battle no. 7) and `second oven of men' (battle no. 10) stand out because of the reportedly large numbers of victims (Gill 1894: 53-5). Both events involved revenge of the Ngariki tribe on the Ngati-Tane (Atiu). In each case, the Ngariki constructed a large communal oven in which to steam the roots of ti (Cordyline terminalis). The Atiu were invited to take part but in the final moments of construction, when the embers were hottest, the Atiu were ambushed and thrown into the oven. In the first oven of men, nearly all subadult and adult male Atiu and two male Ngariki were killed at Putoa, on the border between Ivirua and Tamarua. In the second oven, some 40 years later according to Gill, various Atiu men, women and children perished in the Angaitu area on the border of Tava'enga and Karanga. The depressions said to mark the ovens of men were examined by Gill (1894: 54-5), although these sites have not been excavated. No dismemberment or eating of the victims was reported for either oven of men, which were not located in Keia District, and thus cannot be identified with MAN-84. Regardless, the `ovens of men' suggest that large numbers of people, of both sexes and any age, could be killed and burned on Mangaia in a single, well-organized event.

The MAN-84 human bone assemblage is taphonomically consistent with having been dismembered and cooked, and perhaps by inference eaten, although other ritual activity cannot be precluded and may be suggested by the site's presence in Keia, the main ritual district on Mangaia. Cannibalism is recorded frequently in Mangaian ethnohistory in reference to fugitive tribes, isolated fugitives, people from other islands, and the proclivities of certain priests (Hiroa 1934: 66, 73). These groups or individuals practised opportunistic cannibalism of nutritive necessity (Gill 1894: chapters 14-15). Usually lone victims of any age or sex were killed, then taken to a refuge cave for cooking and eating.

The very young age of some individuals at MAN-84 is inconsistent with opportunistic cannibalism, as older juveniles or adults generally were described as targets rather than infants and young children or adults with children. Furthermore, the site is much too conspicuous to be a `cannibal lair'. All cases of `hermit' cannibalism were referred to as disgusting activities that led to, or reinforced, the social stigma and marginality of its perpetrators (Gill 1894: 86, 219). Those who cooked the body parts at MAN-84 likely did so with the assent of whoever held power in Keia at that time.

Other examples of cannibalism may have been more acceptable in Mangaian society. The priest Mautara was said to have gained his nickname from the notorious cannibal Mautara because of their shared appetite for human flesh (Hiroa 1934: 51). This priest is reported to have avenged an insult by ordering the Ngariki tribe to kill and eat their firstborn children, which reportedly was done (Gill 1894: 68-9; Hiroa 1934:73). Such a practice might account for the infant remains at MAN-84, but not the many non-infants.

Conclusions

Our excavation of Ana Manuku (site MAN-84) revealed a site unlike any other known on Mangaia, or indeed, elsewhere in Polynesia. The two occupation layers are both dominated by traditional earth ovens with abundant bones of humans, rats and fish. Nine AMS [sup.14]C dates on human bones from different individuals suggest that these bones were deposited at some time between AD 1390 and 1470, within the period of severe resource depletion on Mangaia as indicated by pollen and charcoal influx in lake sediments (Ellison 1994; Kirch & Ellison 1994; Kirch 1996) and faunal sequences (Steadman & Kirch 1990; Kirch et al. 1995; Steadman 1995).

Unusual faunal and artefactual attributes indicate that MAN-84 was a special-use site rather than a habitation site. Unlike other Mangaian rock-shelters such as Tangatatau (MAN-44), the cultural strata did not yield the artefacts of everyday life. The bone assemblage also is highly distinctive in that pig and dog are extremely rare, whereas human bones, all clearly in a midden rather than burial context, number nearly 2000.

None of Mangaia's ethnographically recorded violence is a good match for the human bones recovered at MAN-84. The site's island-wide significance is difficult to assess because of its uniqueness and because it is the only major site excavated in Keia District. To determine whether MAN-84 is representative of prehistoric rock-shelter sites in Keia, or whether MAN-44 characterizes sites in Veitatei, will require the excavation of additional sites in both districts. For now, we cannot be certain whether sites with abundant human bone in a midden context are to be found only in Keia, a district noted for its island-wide ritualistic importance, including the temples of human sacrifice. We do know that substantial body parts of at least 41 persons of all age classes were cooked at MAN-84 in what seems to be one event or two closely timed events at c. AD 1400-1450. That these body parts were eaten is suggested by the remains of other animal foods (fish, rats, birds) in the identical midden context.

Acknowledgements. We thank the many Mangaians who helped us, especially T. Arokapiti, T. George, M. Harry, T. Mautairi, P. Ngatokorua, S. Ngatokorua, D. Ngu, M. Ngu, A. Tangatakino, A. Tuara, G. Tuara and T. Tuara. Research permits were issued by T. Okotai (Office of the Prime Minister). Funding for field work and lab analyses was provided by the University of Florida Division of Sponsored Research (grant U001 to DWS, SCA & L. Norr), the National Science Foundation (grants BNS-9020750 to PVK & DWS, and EAR-9714819 to DWS), and National Geographic Society (grant 4001-89 to PVK & DWS). P. Anderson, J. Endicott, D. Sirois, A. Swartz, K.L. Weinstein and M.I. Williams assisted in the field or laboratory. We dedicate this paper to the memory of our dear friend, George Tuara.

References

ANTON, S.C. & P.K. ANDERSON. 1993. Human bone modification? an island example (Abstract). St Louis (MO): Society for American Archaeology.

ANTON, S.C. & D.W. STEADMAN. 1998. Cannibals in the Cooks? Island biogeography and hominid behavior, American Journal of Physical Anthropology Supplement 26: 64-5.

In press. Mortuary patterns in burial caves on Mangaia, Cook Islands, International Journal of Osteoarchaeology.

ATHENS, J.S. 1997. Hawaiian native lowland vegetation in prehistory, in Kirch & Hunt (ed.): 248-70.

BELLWOOD, P. 1978. Archaeological research in the Cook Islands, Pacific Anthropological Records 27.

BINFORD, L. 1981. Bones: Ancient men and modern myths. New York (NY): Academic Press.

BLUMENSCHINE, R.J. 1988. An experimental model of the timing of hominid and carnivore influence on archaeological bone assemblages, Journal of Archaeological Science 15: 483-502.

BUIKSTRA, J.E. & D.H. UBELAKER. 1994. Standards for data collection from human skeletal remains. Fayetteville (AK): Arkansas Archeological Survey. Research series 44.

CORREIA, P.M.M. 1997. Fire modification of bone: a review of the literature, in Hagland & Sorg (ed.): 275-93.

ELLISON, J.C. 1994. Palaeo-lake and swamp stratigraphic records of Holocene vegetation and sea-level changes, Mangaia, Cook Islands, Pacific Science 48: 1-15.

FISHER, J.W., JR. 1995. Bone surface modifications in zooarchaeology, Journal of Archaeological Method and Theory 2: 7-68.

FLENLEY, J.R., A.S. KING, J. JACKSON, C. CHEW, J.T. TELLER & M.E. PRENTICE. 1991. The late Quaternary vegetational and climatic history of Easter Island, Journal of Quaternary Science 6: 85-115.

GILL, W.W. 1894. From darkness to light in Polynesia. London: Religious Tract Society.

GRUNING, E.L. 1937. Notes on burial caves in the Cook Group, South Pacific, Ethnologia Cranmorensis 1: 21-5.

HAGLAND, W.D. 1997. Rodents and human remains, in Hapland & Sorg (ed.): 405-14.

HAGLAND, W.D. & M.H. SORG (ed.). 1997. Forensic taphonomy: the postmortem fate of human remains: 275-93. New York (NY): CRC Press.

HANDY, E.S.C. 1923. The native culture in the Marquesas. Honolulu (HI): B.P. Bishop Museum. Bulletin 1.

HIROA, T.R. (P.H. BUCK). 1934. Mangaian society. Honolulu (HI): B.P. Bishop Museum. Bulletin 122.

1944. Arts and crafts of the Cook Islands. Honolulu (HI): B.P. Bishop Museum. Bulletin 179.

HOPE, G. & M. SPRIGGS. 1982. A preliminary pollen sequence from Aneityum Island, southern Vanuatu, Bulletin of the Indo-Pacific Prehistory Association 3: 88-94.

JAMES, H.F., T.W. STAFFORD, JR., D.W. STEADMAN, S.L. OLSON, P.S. MARTIN, A.J.T. JULL & P.C. MCCOY. 1987. Radiocarbon dates on bones of extinct birds from Hawaii, Proceedings of the National Academy of Sciences USA 84: 2350-54.

KATAYAMA, K. 1986. Human skeletal remains of late pre-European period from Mangaia, Cook Islands, Man and Culture in Oceania 2: 57-80.

KIRCH, P.V. 1983. Man's role in modifying tropical and subtropical Polynesian ecosystems, Archaeology in Oceania 18: 26-31.

1996. Late Holocene human-induced environmental modifications to a central Polynesian island ecosystem, Proceedings of the National Academy of Sciences USA 93: 5296-300.

1997. The Lapita peoples. Oxford: Blackwell.

2000. On the road of the winds: an archaeological history of the Pacific Islands before European contact. Berkeley (CA): University of California Press.

KIRCH, P.V & J. ELLISON. 1994. Palaeoenvironmental evidence for human colonization of remote oceanic islands, Antiquity 68: 310-21.

KIRCH, P.V., J.R. FLENLEY, D.W. STEADMAN, F. LAMONT & S. DAWSON. 1992. Ancient environmental degradation: prehistoric human impacts on an island ecosystem, Mangaia, Central Polynesia, National Geographic Research and Exploration 8: 166-79.

KIRCH, P.V. & T.L. HUNT (ed.). 1997. Historical ecology in the Pacific islands: prehistoric environmental and landscape change. New Haven (CT): Yale University Press.

KIRCH, P.V., D.W. STEADMAN, V. BUTLER, J. HATHER & M. WEISLER. 1995. Prehistory and human ecology in Eastern Polynesia: excavations at Tangatatau Rockshelter, Mangaia, Cook Islands, Archaeology in Oceania 30: 47-65.

ROLETT, B.V. 1998. Hanamiai: Prehistoric colonization and cultural change in the Marquesas Islands (East Polynesia). New Haven (CT): Yale University. Publications in Anthropology 81.

SAVAGE, S. 1980. A dictionary of the Maori language of Rarotonga. Suva: University of the South Pacific.

STEADMAN, D.W. 1995. Prehistoric extinction of South Pacific birds: biodiversity meets zooarchaeology, Science 267: 1123-31.

1997. Prehistoric extinctions of Polynesian birds: reciprocal impacts of birds and people, in Kirch & Hunt (ed.): 51-79.

STEADMAN, D.W. & P.V. KIRCH. 1990. Prehistoric extinction of birds on Mangaia, Cook Islands, Polynesia, Proceedings of the National Academy of Sciences USA 87: 9605-9.

STEADMAN, D.W. & B.V. ROLETT. 1996. A chronostratigraphic analysis of landbird extinction on Tahuata, Marquesas Islands, Journal of Archaeological Science 23: 81-94.

STEADMAN, D. W., T. W. STAFFORD, JR., D. DONAHUE & A.J.T. JULL. 1991. Chronology of Holocene vertebrate extinction in the Galapagos Islands, Quaternary Research 36: 126-33.

STEADMAN, D.W., P. VARGAS & C. CRISTINO. 1994. Stratigraphy, chronology, and cultural context of an early faunal assemblage from Easter Island, Asian Perspectives 33: 59-77.

STUIVER, M., P.J. REIMER, E. BARD, J.W. BECK, G.S. BURR, K.A. HUGHEN, B. KROMER, G. MCCORMAC, J. VAN DER PLICHT & M. SPURK. 1998. INTCAL98 radiocarbon age calibration, 24,000-0 cal BP, Radiocarbon 40: 1041-83.

TURNER, C.G. & J.A. TURNER. 1999. Man corn: cannibalism and violence in the prehistoric American Southwest. Salt Lake City (UT): University of Utah Press.

WHITE, T.D. 1992. Prehistoric cannibalism at Mancos 5myumr-2346. Princeton (NJ): Princeton University Press.

DAVID W. STEADMAN, SUSAN C. ANTON & PATRICK V. KIRCH, Steadman, Florida Museum of Natural History, University of Florida, Gainesville FL 32611, USA. Anton, Department of Anthropology, University of Florida, Gainesville FL 32611, USA. Kirch, P,A. Hearst Museum of Anthropology, University of California, Berkeley CA 94720, USA.

Received 6 January 2000, accepted 7 March 2000, revised 8 May 2000 & 12 September 2000.