Volume 78 1969 > Volume 78, No. 4 > Archaeology on Rarotonga and Aitutaki, Cook Islands: A preliminary report, by Peter Bellwood, p 517 - 530
                                                                                             Previous | Next   

- 517

Over the past fifteen years, archaeological research in Island Polynesia has been extended at an ever-increasing rate. Excavations have been undertaken in most of the island groups, but a perusal of the available literature shows that a great deal of the research has been concerned with the establishment of typological sequences from artefacts and carbon dates. While a number of scholars have made detailed studies of architectural remains surviving above ground, 1 not until the publication of Green's work in the Opunohu Valley of Moorea were the results of a complete settlement pattern survey readily available. 2 Green was able to document the existence of a specific type of settlement pattern in the Opunohu Valley at the time of European contact, a settlement pattern which bears some close resemblance to that which will be described for Rarotonga.

The work here described was carried out between December, 1968, and February, 1969, by a field party from the University of Auckland. 3 - 518 The two main projects were as follows: an intensive survey of all remains in the Maungaroa Valley of Rarotonga, and a survey of all marae, as far as could be determined, on the island of Aitutaki. Several coastal structures on Rarotonga were also investigated. Little excavation was carried out, but this is planned for the 1969-70 field season.


Rarotonga is a typical high volcanic island, with central mountain ranges rising to 2,000 feet, dissected by deep narrow valleys opening into a narrow coastal strip some 200-300 yards wide. The island is small, being about five miles by four, and the fringing lagoon is very narrow except on the southern and south-western sides. Settlements and agricultural land were situated on the coastal strip and in the valleys.

In 1962-4 expeditions from the Canterbury Museum, New Zealand, led by Dr Roger Duff, excavated several ceremonial sites along the Ara Metua—the ancient road which originally ran right round the island. 4 In addition, Mr H. Parker, a member of the first expedition, located three clusters of well-preserved stone structures in the Maungaroa Valley, and these structures were the object of the survey by the Auckland party. While the coastal structures excavated by Duff in some cases yielded stratified sequences and Carbon-14 dates 5 the coastal settlement pattern has been almost totally destroyed by post-contact activities. Only in the valleys has evidence survived in quantity. Besides the Maungaroa survey, several sites were also surveyed in the Rutaki Valley, and some are reported to exist in other valleys.

The settlement pattern and economy of the coastal strip was fortunately described briefly by Williams. Crops, including taro and sugar cane, were grown right across the coastal strip to the edge of the sea.

In the first place, there are rows of superb chestnut trees, inocarpus, planted at equal distances, and stretching from the mountain's base to the sea, with a space between each row of about half-a-mile wide. This space is divided into small taro beds, which are dug four feet deep, and can be irrigated at pleasure. These average about half an acre each. The embankments round each bed are thrown up with a slope, leaving a flat surface upon the top of six or eight feet in width. The lowest parts are planted with taro, and the sides of the embankment with kape or gigantic taro, while on the top are placed, at regular intervals, small beautifully shaped breadfruit trees. . . . There is a good road round the island, which the natives call ara medua, or the parent path, both sides of which are lined with bananas and mountain plantains. . . . The houses of the inhabitants were situated from ten to thirty yards or more from this pathway. . . . Six or eight stone seats were ranged in front of the premises, by the side of the “parent pathway.”. 6

This description by Williams indicates the presence of a scattered coastal settlement pattern along the Ara Metua. He does not mention - 519 villages, and it is noteworthy that with the establishment of the Mission stations villages were actually founded. If they had existed before, then surely the Missions would have been located in them. The degree of nucleation or diffusion of the settlement pattern cannot, however, be determined with certainty, since hardly any of the house paepae remain.

Concerning the Maungaroa Valley, Williams notes the following about Tinomana Ariki and his people.

This chief, with the whole of the people of his district, were living in the mountains where Tinomana himself was born. As this was the weakest district of the three, its inhabitants were subject to peculiar oppression from their more powerful neighbours, who plundered them of their food and property with impunity. When a sacrifice was required, they would invariably seek it from this oppressed people; who, when they wanted fish, were obliged to steal down to the sea in the dead of night and return before daybreak, to avoid being plundered. . . . 7

Williams' remarks about access to the sea are important, since William Gill records that, on the founding of the Arorangi Mission village on the coast near the mouth of the Maungaroa Valley in 1828, the land was “densely covered with trees and brushwood, the growth of many generations”. 8 Williams' above remarks concerning the cultivation of the coastal strip cannot have applied to this area, and the conclusion is that Tinomana and his people had been cut off in the Maungaroa Valley for a long period—many generations, if we believe Gill. The Maungaroa settlement pattern, then, is not typical of Rarotonga as a whole, but rather represents an adaptation to a steep and enclosed valley, where flat land would be at a premium. Most of the flat land in fact seems to have been used for dwellings and marae. Owing to dense bush, no agricultural terraces were observed in the valley, although old residents reported seeing terraces many years ago, brought into relief by long shadows cast by the setting sun. Many irrigated taro terraces can be found in the Avana Valley on the other side of Rarotonga from Maungaroa, but a group of taro terraces in the lower part of the Maungaroa Valley can most plausibly be associated with Post-Contact settlement. However, the presence of irrigated taro terraces and the well-documented dependence on tree crops in most areas of central Polynesia, including the Cook Islands, is important for the study of settlement patterns. Since these resources yield a regular and normally reliable food-supply, and do not demand seasonal migration from the groups that utilise them, we may infer that ancient settlements in these areas were inhabited permanently, movement only occurring because of irregular factors such as death or war. In theory, this means that settlement pattern studies will involve fewer variables than similar studies in an area such as New Zealand, where seasonal movements were frequent.

The Maungaroa structures fall into three distinct groups—a large group of thirty-three structures in the Lower Maungaroa Valley on a flat between

- 520
- 521

the south face of Raemaru Mountain and the Muriavai Stream; a group of thirteen structures in the Upper Maungaroa Valley on the lower slopes of Maungaroa Mountain; and a small group of structures at 1,100 feet above sea level situated on a col between Raemaru and Maungaroa. Figure 1 shows the distribution of the sites in the valley, which is enclosed on all sides by relatively steep slopes, becoming very steep in the upper part of the valley. No structures were found between the Upper and Lower Maungaroa clusters. Paepae and marae have not been differentiated on the plan because of lack of space.

Structural types in the Maungaroa Valley, Rarotonga
  Structural types Lower Maungaroa Upper Maungaroa
Paepae Type A1—T-shaped pavements 4 1
  Type A2—Square, rectangular or L-shaped pavements 7
  Type A3—Simple stone-faced earthen terraces 7 6
  Type A4—Stone-faced earthen terraces with paved verandahs 1? 4
Marae Type B1—Terraced with four steps 4
  Type B2—Terraced with two steps 4 1
  Type B3—Rectangular pavement with single upright 1
  Type B4—Rectangular pavement with single atarau 1
  Type B5—Coral pavement with several uprights and atarau 1
  Type B6—Terrace with multiple uprights 1
  Unclassified isolated stone alignments 3 1
    33 14
The Lower Maungaroa cluster

This cluster consists of thirty-three structures built on a number of small flattish areas amongst the rocky lava flows extending from the south side of Raemaru. Thirty-one of these sites are situated in an area of 150 × 200 metres, and they can be described as forming a village, although topography has not allowed a large degree of nucleation. The stone structures, as in the Upper Maungaroa Valley, comprise marae and the house paepae of high-ranking members of the society (Table 1). One of the large T-shaped paepae was traditionally the home of Tinomana Ariki. The paepae themselves are pavements of basalt, edged with selected facing stones, also of basalt. Limited excavation showed that the house was built on a level area of earth behind the paepae, the latter forming a paved verandah. The houses appear to be rectangular, although evidence as yet is limited The majority of the lower Maungaroa paepae fall into types A1 and A2, although there are some terraced paepae of types A3 and A4. The largest of the T-shaped paepae of type A1 has an approach path 25 metres long, and the arm of the T, which fronted the house, is 10 metres - 522 long. Another paepae of this type consists of a basic T, with two parallel approach paths. The type A2 paepae form a varied component, and have been grouped together in the absence of any recognisable functional differences within the group. These are pavements of square, rectangular, and L-shapes, with a further example shaped like an E, but minus the middle bar. Paepae of this group are small, reaching maximum dimensions of 12 metres. Clearly, the T-shaped paepae belonged to the highest-ranking members of the society, and the shape of the T, with the approach path, must have been important for ceremonial purposes. These type A1 paepae also have a small upright just before the house area.

Eleven structures are classed as marae in the Lower Maungaroa Valley, but it is not possible to differentiate between marae and shrines as Green did for Moorea. 9 The Lower Maungaroa marae cover types B1-B5.

Type B1. These marae have four terraces, each paved with basalt slabs. The height of each terrace corresponds to one course only of facing, i.e. not more than one metre. Two of these marae cover areas of 12 × 12 metres; two are much smaller, covering 4 × 6 and 3 × 10 metres. Each has one upright, either at the foot of one of the terraces, or behind the upper terrace.

Type B2. These are like type B1, but with only two terraces. The upright is placed in front of the facing of the upper terrace.

Type B3. There is one example only, 15 × 5 metres, with an upright in one corner. This marae is basically a rectangular pavement, but a strip down the middle is slightly lowered below the general level.

Type B4. The single marae of this group is a large pavement, 23 × 13 metres, with an atarau 10 of basalt blocks situated at rear centre. This is 3.5 metres long by 1.5 metres wide, and 1 metre high,

Type B5. The one marae of this type is so unusual that it deserves description in some detail. The structures are situated on an extensive sloping terrace of coral gravel (kirikiri), and consist of five atarau of undressed basalt blocks (none of the Maungaroa structures contain any dressed stone), and two basalt uprights. Two giant natural basalt boulders seem to form the focal point of the marae. The stone alignments shown on the plan (Figure 2) are the facings of low terraces, but the placement of these bears no resemblance to the terraces of types B1 and B2. The atarau themselves have been partially collapsed, presumably with the arrival of the Missionaries, when considerable destruction took place, but they do not appear to have been more than 1 metre high. The centre platform of the line of three had a stringer course of white coral in the basalt face and this would originally have given a colourful effect.

One striking characteristic of the Maungaroa structures, and especially those of the Lower Maungaroa, is the lack of regularity. The T-paepae and the two large terraced marae were important in respect of their size, and are of distinctive morphology. The other structures, however, are more difficult to categorise.

The distribution of the sites—19 paepae and 11 marae, is not random. The settlement pattern relates to that found by Green for Moorea, and

- 523

consists of small clusters, usually of one marae and two or three paepae, each cluster separated by a short distance from the next. One might expect the settlement to be differentiated into familial units, each with the paepae of a high-ranking member of the society, of mataiapo or rangatira rank, with a small family marae. The only coastal marae which is preserved to any degree is the Arai te Tonga at the mouth of the Tupapa Valley, and this consists of a rectangular coral spread, with a small basalt atarau and an adjacent basalt upright at one end. However, this has been tampered - 524 with in modern times, and, while the structure as it now stands bears some resemblance to the type B5 marae in the Lower Maungaroa Valley, it is clear that the inhabitants of the valley evolved their own form of architecture to suit particular topographic and social conditions. With two notable exceptions, to be described in a later report, 11 the fragmentary remains of other coastal marae seem to be of Arai te Tonga type.

The Upper Maungaroa Cluster

There are fourteen structures in this cluster, mostly ranged along artificially extended terraces along the upper course of the Muriavai Stream, from 500 to 800 feet above sea-level. The house paepae of this area are of the terraced forms—types A3 and A4. Three of these are large—15 metres long—and have two paved terraces below the level of the house floor. The house on one of the structures was partly excavated by the Canterbury Museum expedition in 1964 12 and a further one was partly excavated by the present team. The houses appear to have been rectangular, and about 10 feet long by 30 feet wide. There is one T-paepae of type A1 in the cluster, but the stones used in its construction are larger and rougher than those of the Lower Maungaroa paepae of this type.

Two of the structures are marae—one of type B2 with two terraces (Plate 1A), and closely paralleled in the Lower Maungaroa cluster; while the other, of type B6, is unique, and consists of fourteen uprights arranged on a low terrace (Plate 1B). Ten of these compose one row, and there is one large single upright a little distance in front of this row. One is tempted to attribute this to Tinomana himself. Three separate uprights stand behind the main row. A parallel for this marae is reported for the Turangi Stream, with nine uprights, but the party was unable to locate this. Clearly, marae of this type, where the uprights themselves become the dominant element in the design, differ considerably from the terraced and paved marae of the Lower Maungaroa cluster.

Some of the differences between the architecture of the Lower and Upper Maungaroa clusters can be explained topographically, since terraces would be required more in the steeper upper valley. The older inhabitants of Arorangi village claim that the two clusters are contemporary and that Tinomana used the lower cluster as his main settlement, but retired to the upper cluster, defended on three sides by almost unscaleable valley walls, whenever his people were threatened. However, the absence of recorded land divisions for the upper valley, and their presence in the lower valley, may indicate that the upper cluster is the earlier. About 1600, an ancestor named Te Mutu is reported to have crossed the Upper Maungaroa Valley and to have found it uninhabited. 13 This may mean that this part of the valley was deserted by this time, but the carbon-14 dates obtained by the Canterbury expedition, - 525 from the sixteenth to eighteenth centuries A.D., 14 do not agree with this conclusion. At present, we cannot differentiate chronologically between the two clusters.

The Raemaru cluster

This consists of a small group of structures, mostly earthen terraces with paved verandahs of type A3 and A4, situated in a small col between Raemaru and Maungaroa. They are located on an excellent vantage point for observing movements on the coastal strip to the north of the Maungaroa Valley, and they clearly belong to a lookout post. Architecturally, they would appear to relate to the Upper Maungaroa cluster.


Owing to invaluable assistance offered by the Mataiapo of Aitutaki and the Resident Agent, Mr J. Macaulay, 27 classifiable sites were located and surveyed, all in the coastal regions. Aitutaki is a volcanic island with a very extensive lagoon, and the highest point is only 384 feet above sea level. In theory, the total land surface is habitable and fit for cultivation, and the island is very different from Rarotonga in this respect. For this reason, it is possible that more sites may be located in inland areas in the future.

The sites fall into three groups—marae; paepae; and unclassified sites, many of them natural features, about which traditions have survived. Sites in the latter group will not be discussed here. The types of site for Aitutaki are set out in Table 2.

Structural Types on Aitutaki
    Extan Destroyed
Paepae Type C1—Entertainment houses—Are Karioi 1  
  Type C2—Dwelling paepae 3 3
Marae Type D1—Single or paired uprights 6  
  Type D2—Alignments of standing stones 9  
  Type D3—Atarau of orthostatic construction 4  
  Type D4—Alignments and atarau 4  
  Type D5—Rectangular pavements   1
  Unclassified sites 13  
    40 4

24 of the sites surveyed were marae. Buck 15 describes two types of marae—temporary and permanent. Temporary marae, according to Buck, were founded by passing voyagers who moved on to settle elsewhere on the island. The word “temporary” is perhaps misleading, since these marae are still standing and traditions have been preserved about them to this day. The type D1 marae, of which there are six, may be of this type, having single or a pair of standing stones. The marae of Parikavakevake on Papa'u motu consisted of a standing block of coral 1.50 metres high, - 526 with no associated features. The traditional builder did in fact move on to settle on the mainland, and the other marae of this type have similar traditions.

Buck also describes another marae—Te Hautapu-o-nga-Ariki, which consisted of a pavement of basalt and coral, 63 × 22 feet. No other marae of this type were seen on Aitutaki, and this example has been totally destroyed. Therefore, I have classified this separately as type D5 (see below).

Type D2. Well-preserved marae of this group have standing stones arranged in parallel rows. Many are badly disturbed and only a few stones survive, but the best-preserved example, the marae of Ai Te Vananga near Tautu village, gives a good idea of the basic form of the group. This has 11 main stones in three parallel rows, of 4, 3, and 4 stones respectively. The rows are 2 and 3 metres apart and 11 metres long. Three separate standing stones are in the close vicinity of the main structure. The stones stand up to 1.20 metres high. The Upper Maungaroa marae of type B6 with its emphasis on uprights is the only observable parallel to the Aitutaki marae on Rarotonga, and this similarity may be due to coincidental convergence of the forms.

Type D3. The main features of marae of this group are structures corresponding to the atarau as described by Buck 16 and the ahu of Society Island marae. The best example is Toka pu moana, situated near the north tip of the island. This is basically a cist of upright basalt slabs up to 85 centimetres high, the outside dimensions being 4 × 1.20 metres—the rectangle enclosed by the slabs is left empty. Plate 1C shows a similar structure in Taravao district. It should be noted that cists of this type are structurally closely related to the ahu of Leeward Society Island marae 17 and also to the New Zealand tuahu. 18

Type D4. The four examples of this type must be recognised from their tremendous size as major ceremonial centres. The best-preserved example—Paengariki, in Taravao district at the south end of Aitutaki—consists of over 60 stones in four main rows 80 metres long, the whole complex covering an area of 80 × 40 metres (Figure 3). Some of the uprights are over 2.00 metres in height. All the sites of this group have subsidiary atarau of type D3. The presence of ceremonial complexes of this type is of extreme importance, and there seem to be no direct parallels elsswhere in Polynesia. The stones seem to line processional avenues, and they may also mark the positions of important individuals during ceremonies—several of the uprights have stone seats at their bases.


Type C1. The Are Karioi. This structure, an assembly house of entertainment, has been described by Buck. 19 The surviving remains consist of a basalt and coral pavement 25 × 9 metres. Buck's description of Te Hautapu-o-nga-Ariki resembles the Are Karioi (Type D5 above), and it may be that it had the same function and was never a marae at all. However, in view of its destruction, speculation cannot proceed further.

- 527
D4 marae, Taravao District, Aitutaki
- 528
Dwelling-house paepae—type C2

Buck 20 recorded that houses on Aitutaki were approached by a path, with two rectangular areas enclosed by slabs on either side. These areas were paved with coral gravel. The site described by Buck, at Nukunoni, has been partly destroyed by bulldozing, but some of the enclosed rectangles do survive. Unfortunately, little was found that could add to Buck's description. One site near Amuri had a pavement type of paepae, similar to those of type A2 in the Lower Maungaroa Valley, but this too was badly destroyed. Nearly all the sites shown to us on Aitutaki were marae, since these still preserve important traditions, which we recorded in all cases. However, the sites of dwelling houses may have been forgotten, and others could come to light with future exploration.


Quite clearly the stone structures of Rarotonga bear little resemblance to those of Aitutaki, despite parallels in other items of material culture. In the same way, resemblances between these two islands and the rest of Eastern Polynesia, are not present to any great degree. The type D3 marae of Aitutaki resemble those of the Leeward Societies in ahu construction, but this is only a constructional resemblance in the use of upright slabs. After all, this type of construction is world-wide. Furthermore, the alignments of type D4 are not paralleled in the Society Islands at all.

For the Maungaroa Valley, the terraced marae of types B1 and B2 are paralleled in the Marquesas, but the me'ae of this type in the Southern Marquesas are usually much larger, and may be up to 30 by 85 metres, in the case of an example surveyed in the Hanaiapa Valley, Hivaoa. 21 The terraced paepae of types A3 and A4 resemble those of the Marquesas, 22 and also the Hawaiian Islands. 23 However, the marae of type B6 is unique to Rarotonga. One could continue seeing parallels in this vein for a long time, but there are always important exceptions to the rule.

Historically, the Maungaroa sites are late. While the Aitutaki sites are undated, except by tradition, they may be assumed to be late, especially in the case of the type D4 marae. Traditions relating the structures to early ancestors may be significant, but it cannot be assumed without excavation that the present structures are not the result of rebuilding. From the point of view of recorded stone structures, which in general seem to be of late date, every island is to some extent individualistic. In theory, we would expect to find more meaningful inter-island similarities between stone structures dating from Duff's Early East Polynesian period, i.e. A.D. 500-1100, 24 except in the case of later stray voyagers building marae in the style of their homelands. This period, on present evidence, - 529 saw the spread of an early form of relatively undifferentiated Polynesian culture throughout the archipelago, and was followed by a period of individual specialisation within each island group to the period of European contact. This does not mean that voyaging ceased, for both Cook and Williams recorded drift voyages from the Society Islands to the Cooks. 25 However, with an increasing population, one may expect that society became more resistant to outside influences, and more inclined to regional specialisation. This regional specialisation was not just inter-island, but it was also present in adjacent valleys on Rarotonga. The Rutaki Valley, also belonging to Tinomana Ariki of the Maungaroa Valley, contained a number of paepae and marae surveyed by Miss Allo of the Auckland party. These are not reported in this paper, but, architecturally, the structures are different from those of the Maungaroa Valley. This situation can only be attributed to individual preference for a particular form.

With excavation and dating, it may be possible to tie in the Cook Island stone structures more closely with those of the Society Islands, which on very strong traditional and linguistic grounds were the home of the later strata of Cook Islands culture. 26 Certainly, the atarau and house paepae of the Maungaroa Valley (type A2 especially) do resemble the ahu and Group IIa paepae of the Opunohu Valley sites described by Green. But there are many absences —round-ended house floors, 27 marae enclosure walls, and stepped ahu, to name a few.

In conclusion, the main results of the past season's work on Rarotonga and Aitutaki have been to demonstrate the presence of numerous and varied stone structures, and, in the case of the Maungaroa Valley, to record in detail a virtually complete prehistoric settlement pattern.

  • BEAGLEHOLE, J. C., 1955. The Journals of Captain James Cook. Cambridge, Hakluyt Society.
  • BUCK, Sir Peter, 1927. Material Culture of the Cook Islands (Aitutaki). Memoirs of the Board of Maori Ethnological Research, Volume 1.
  • —— 1944. Arts and Crafts of the Cook Islands. Honolulu, Bernice P. Bishop Museum Bulletin 179.
  • —— 1958. The Coming of the Maori. Wellington, Maori Purposes Fund Board.
  • DUFF, R., 1965. “The Canterbury Museum Expedition to Rarotonga”. New Zealand Archaeological Association Newsletter, 8:53-58.
  • —— 1968. “Archaeology of the Cook Islands” in I. Yawata and Y. H. Sinoto (eds.) Prehistoric Culture in Oceania. Honolulu, Bishop Museum, pp. 119-132.
- 530
  • EMORY, K. P., 1924. The Island of Lanai. Honolulu, Bernice P. Bishop Museum Bulletin 12.
  • —— 1933. Stone Remains in the Society Islands. Honolulu, Bernice P. Bishop Museum Bulletin 116.
  • GARANGER, J., 1964. “Recherches Archaéologiques dans le District Tautira (Tahiti, Polynesie Française), rapport préliminaire”, Journal de la Société des Océanistes, 20:5-21.
  • —— 1965. “Fouilles Archaéologiques et restauration de trois marae, à Rangiroa”, Journal de la Société des Océanistes, 21:45-66.
  • GILL, W., 1856. Gems from the Coral Islands. London, Ward.
  • GREEN, R., 1961. “Moorean Archaeology: A Preliminary Report”. Man, article 200.
  • —— 1966. “Linguistic Subgrouping within Polynesia: the implications for Prehistoric Settlement”. Journal of the Polynesian Society, 75:3-35.
  • —— 1967. “Settlement Patterns; Four Case Studies from Polynesia”. Asian and Pacific Archaeology Series, 1:101-132.
  • GREEN, R. and K., R. A. and A. RAPPAPORT, and J. M. DAVIDSON, 1967. Archaeology on the Island of Moorea, French Polynesia. Anthropological Papers of the American Museum of Natural History, 51, 2.
  • LINTON, R., 1925. Archaeology of the Marquesas Islands. Honolulu, Bernice P. Bishop Museum Bulletin 23.
  • SINOTO, Y. H., 1965. “A Tentative Prehistoric Cultural Sequence for the Northern Marquesas Islands, French Polynesia”. Journal of the Polynesian Society, 75:287-303.
  • SUGGS, R. C., 1961. The Archaeology of Nuku Hiva, Marquesas Islands, French Polynesia. Anthropological Papers of the American Museum of Natural History, 49, 1.
  • WILLIAMS, J., 1838. Missionary Enterprises in the South Sea Islands. London, Snow.
1   e.g. Emory 1933, Suggs 1961.
2   Green 1961, 1967; Green et al., 1967; and also Garanger 1964 for the Tautira Valley, Tahiti.
3   The fieldwork was financed by grants from the New Zealand University Grants Committee, and the H. D. Skinner Fund for Physical Anthropology, Archaeology, and Ethnology. Acknowledgements are also due to the Government of the Cook Islands, the Cook Islands Library and Museum Society, and numerous private individuals resident in the Cook Islands, for invaluable help in the field. The field party consisted of the author, Misses J. Allo and T. Sparks, and Mr D. Sutton. The photographs illustrating this article are courtesy of the Anthropology Department, University of Auckland, and the line drawings are by K. M. Peters.
4   Buck 1927:211; Duff 1965, 1968.
5   Duff 1965.
6   Williams 1838:206-207.
7   Williams 1838:172.
8   Gill 1856:39.
9   Green 1961:171.
10   Buck 1944:308.
11   One notable exception is a small marae at Ngatangi'ia which appears to be unique on Rarotonga, and bears close resemblance to a marae described by Garanger for Rangiroa, Tuamotu Islands (Garanger 1965:plate 1,1—marae Pomariorio).
12   Personal communication from Dr R. Duff.
13   I wish to record my thanks to Mr R. H. Parker for bringing to my attention the existence of this tradition in the Land Court Minute Books for 1913 in Avarua.
14   Duff 1965:57.
15   Buck 1927:208-9.
16   Buck 1944:308.
17   e.g. Emory 1933:plates 16, 17.
18   Buck 1958:480.
19   Buck 1927:36-37.
20   Buck 1927:3-5.
21   Surveyed by a Bishop Museum expedition led by Dr Y. Sinoto. See also Linton 1925.
22   Suggs 1961:fig. 38.
23   Emory 1924:38-40.
24   Duff 1968: 122-3.
25   Beaglehole 1965, III:86-87; Williams 1838:199.
26   Williams 1838: 194. Green (1966) places Rarotongan tentatively in his Tahitic linguistic subgroup.
27   Duff (1965:56) excavated what appeared to be a round-ended post structure belonging to an early phase of a paepae of the Ara Metua near Arorangi, and Anderson (Beaglehole 1965, III:830) observed round-ended houses on Mangaia. However, round-ended house floors outlined by slabs of the Society Island type (Green et al., 1967:fig. 7) have not been found on Rarotonga.