TONGA, LAPITA POTTERY, AND POLYNESIAN ORIGINS

Volume 80 1971 > Volume 80, No. 3

TONGA, LAPITA POTTERY, AND POLYNESIAN ORIGINS

L. M. Groube Australian National University

The enduring problem of Polynesian origins has been entertaining readers of The Journal of the Polynesian Society for more than 75 years. It seems appropriate, therefore, that new evidence, suggesting a crucial role for Tonga in Polynesia settlement history should be first reviewed in these pages.

The East Polynesian Homeland

Linguistic and archaeological research during the last decade has clearly established the importance of Samoa in East Polynesian origins. The important review of the linguistic evidence by Pawley and Green 1 which placed the Eastern Polynesian languages in a subgroup of Polynesian which also includes Samoan, but excludes Tongan, provided evidence for Samoa as the immediate homeland from which the Proto-Eastern Polynesian speech community derived. Although linguists have not asserted that Samoa was the place where Proto-Nuclear Polynesian was spoken, it is an obvious candidate because of its size and central geographic position in relation to other members of the Nuclear Polynesian subgroup.

This identification is strengthened by the archaeological evidence. Golson, in the first archaeological work in Samoa in 1957, established at the now famous site of Vailele, that pottery was in use in Samoa in the first century A.D. 2 Similar pottery has been recovered from early levels in the Marquesas. 3 Although Golson did not recover adze evidence sufficient to establish a link with East Polynesian forms, 4 subsequent work by Green 5 disclosed a wide range of early adzes from - 279 which the diverse kit of the earliest East Polynesians (at the site of Hane in the Marquesas islands 6) could have been derived.

Although one C14 date from Samoa suggests that initial settlement was B.C. in date, 7 there are, as yet, no dates directly associated with human occupation earlier than 1950 ± 90 B.P. (0 B.C.). The probability is strong, however, that such dates will be recovered.

As the development of the distinctive linguistic innovations of the Samoic group required an interval of separation and isolation prior to the exodus to East Polynesia, the settlement date of c A.D. 400 established by Sinoto at Hane, the earliest site yet discovered in East Polynesia, is of relevance to Samoan chronology. The earlier the settlement of East Polynesia, the more probable is a B.C. initial settlement date in Samoa.

The thorough archaeological examination of Samoa, witnessed by the detailed report recently published, 8 makes unlikely an earlier belief expressed by Green that the first third of Samoan prehistory was as yet unexplored. There seems to be no reason, linguistic or archaeological, to suggest that Samoan settlement was earlier than 300-200 B.C., although dates from East Polynesia earlier than A.D. 400 would greatly change this picture.

Many problems of East Polynesia/Samoan relationships are yet to be solved, particularly the presence, in the earliest levels in the Marquesas, of sophisticated fish hooks, 9 evidence for which, apart from one doubtful example, 10 is entirely lacking from Samoa. Allied to this is the notorious absence of shell-fish middens in Samoa, 11 which, along with the evidence of early inland settlement 12 would suggest that, from a relatively early date, the prehistoric Samoan economy was oriented to an inland agricultural base, in strange contrast to the coastal, sea-bound East Polynesians.

The West Polynesian Homeland

The western boundary of Polynesia is most clearly defined on linguistic grounds: culturally the division between West Polynesia and the neighbouring islands of Fiji is slight. Late contact between Fiji, Tonga and Samoa and the possession of important features such as the Kava ceremony in common make it possible to consider the reality of a Western Pacific cultural enclave which cuts across the linguistic boundaries. 13

For this reason Fiji has always been considered a likely homeland for the Polynesians. This likelihood has become more firmly established by the linguistic and archaeological evidence assembled in recent years.

The linguistic evidence places the Fijian languages close to the Polynesian language group as descendants of a common “proto-Eastern Oceanic” language, 14 although few linguists would claim to know - 280 where this reconstructed language was spoken. Archaeological evidence is more precise as to location but it can rarely be claimed that the evidence recovered from the ground has any reference to the reconstructed languages of the linguist. It is likely, however, that unless there was an extraordinarily complex settlement history, the two forms of evidence will be inter-related.

The Archaeological Evidence

Archaeological work in the Western Pacific was initiated by the Giffords in Fiji in 1947. 15 From this work, and later surveys in New Caledonia, pottery was established as the most important diagnostic artefact in this region. In addition to this work, McKern had reported the presence of potsherds in Tonga (where pottery had been reported ethnographically). 16 This observation was confirmed in 1957 by Golson who further extended the province of pottery by discovering it at the base of the Vailele mound in Samoa.

In reviewing these early archaeological investigations, Golson emphasised the significance of certain decorations in pottery from a unique site (site 13) discovered by Gifford and Shutler in their survey of New Caledonia. 17 This site had produced the earliest dates from New Caledonia 485 ± 400 B.C. and 850 ± 350 B.C. The method of decoration (by dentate-stamp) and the motifs employed were identical to those on sherds recovered from the Sigatoka dunes in Fiji and on some sherds from Tonga reported by McKern. In addition, Gifford and Shutler had noted that pottery with similar decorations had been found on Watom Island off New Britain as well as, from McKern's data, in Tonga. The distinctive decorated ware has been subsequently labelled “Lapita ware” from the original New Caledonian site of Lapita (site 13).

Initially, as Golson hinted in 1959, this pottery style offered exciting prospects in the search for Polynesian origins. Being known on the Polynesian island of Tongatapu, it was an important early link between Polynesia and Melanesia. The distribution pattern of the known sites with Lapita ware, extending into the central Melanesian region, was close to the ideal pattern anticipated to document the movement of Polynesians through Melanesia; small impermanent settlements in an already populated region.

The belief that Melanesia was settled well before the appearance of “Polynesians” is based on sound linguistic evidence. The highly diversified Melanesian languages, according to most authorities, represent a much earlier separation and dispersal than the Polynesian group. 18 It is now argued that Polynesia is, in essence, a lower-order subgroup of a subgroup of Melanesian languages. The related belief that the Polynesians somehow moved through this populated region en route to Tonga/Samoa and the central Pacific is a hang-over from the more romantic days of - 281 speculation on Polynesian origins, when Percy Smith, for instance, traced their homeland to the Gangetic Basin. 19 The re-occurrence of Indonesian/South-east Asian adze styles in the central Pacific tended to maintain belief in the rapid migration trail of the Polynesians. Duff, for instance, claims “a very fast passage into Polynesia” to explain the diffusion of certain Indonesian adze traits into Polynesia. 20 The Lapita pottery sites, much like temporary railway encampments, were close to the expectations of diffusionist theories of Polynesian movement.

This illusion created by the distinctive Lapita distribution was weakened by the 1960 excavations of Golson on the Lapita site of St Maurice on the Isle of Pines, New Caledonia. 21 Not only was the all important rectangular-sectioned adze absent, but none of the material culture was recognisably Polynesian in character.

Subsequent excavations of Lapita sites in Fiji, 22 on Watom Island, 23 and further work by Shutler at the type-site in New Caledonia 24 failed to recover evidence that these sites were documenting the trail of Polynesian migration. Only at the Sigatoka dune site in Fiji could anything like relatives to the diversified Polynesian adze kit be claimed. 25 They show little similarity to the much-vaunted rectangular sectioned adzes of Indonesia, which, if we are to accept diffusionist claims, should have been the source of this distinctive tool.

The systematic investigation of sites of Lapita ware, launched by Golson's 1960 work on the Isle of Pines, was followed by Poulsen's excavations in Tonga during 1963-4. 26 His conclusions relevant to the problem of Polynesian origins, are summarised below:

1. That only a single pottery tradition, Lapita (with close similarities to sites in Fiji, New Caledonia and Watom Island), reached Tonga.

2. In contrast to the Melanesian islands where it was found, this tradition persisted, although weakening, to European contact.

3. During this long period of declining popularity a number of changes in pottery were detected.

• (a) The loss of the elaborate decorative motifs of the original Lapita style, with a subsequent increase in the importance of plain ware.
• (b) The progressive loss of the diversified “fancy” rims and specialised features such as flat-bottomed vessels, handles and so on. 27

These pottery changes enabled Poulsen to erect a sequence of sites, which, along with his C14 dates from various horizons, became the basis of his conclusions that the Lapita pottery tradition persisted for over 2,000 years of Tongan prehistory.

- 282

The earliest date was from the first site he excavated, To 1, where a sample of shells yielded a ninth century B.C. date of similar antiquity to that from site 13 in New Caledonia. 28 Two charcoal dates from near the base of a large refuse mound, To6, were in the fifth century B.C., while a further two dates in the fourth century A.D. from To2 and To5 apparently recorded the persistence of decorated ware into the Christian era. 29 From the top levels of his sites further charcoal samples extended the range of pottery use into the sixteenth and seventeenth centuries (Table 1).

TABLE 1

Radiocarbon Dates from Tonga (Poulsen 1963-4)

Site	Dates B.P.	Range: 2 Std Deviations, AD/BC	Median Date AD/BC
5	ANU 231/1 330 ± 100	AD 1420 - 1820	AD 1620
5	ANU 231/2 340 ± 63	AD 1484 - 1736	AD 1610
1	K 961 420 ± 100	AD 1330 -1730	AD 1530
1	NZ 597 464 ± 82	AD 1322 - 1650	AD 1486
2	NZ 635 1620 ± 60	AD 210 - 450	AD 330
5	NZ 637 1600 ± 87	AD 176 - 524	AD 350
6	ANU 24 2350 ± 200	800 - 0 BC	400 BC
6	NZ 636 2380 ± 51	532 - 328 BC	430 BC
1	K 904 2770 ± 100 1	620 - 1020 BC	820 BC
1   Shell date.

The recent dates were supported by reports of pottery in use in Tonga during the early years of European contact. In addition, subsequent evidence has established that the earliest (shell) date was close to dates from sites on neighbouring Fiji with similar concentrations of decorated Lapita sherds (see below), as well as with site 13 in New Caledonia.

Poulsen's work, therefore, was apparently a convincing demonstration of the longevity and continuity of the pottery tradition in Tonga. It also resolved the status of the Lapita-potters: they were undoubtedly the ancestors of the modern Tongans. The equation of Lapita ware with Polynesian origins, first suggested by Golson, was firmly established.

It was on the Lapita sites in Fiji that attention was next concentrated. Three sites with Lapita ware—the Sigatoka sand dunes, Yanuca Island rock shelter and Natunuku, proved especially rewarding.

From the dunes at Sigatoka, the Birkses, during two field seasons in 1965 and 1966 recovered the most complete Lapita pots yet discovered. 30 Although the bulk of the site was deflated, careful excavation recovered the remnants of three chronologically distinct pottery bearing layers. The earliest, with Lapita decorated ware, has been C14 dated to 2460 ± 90 B.P. (510 B.C.). 31 The later horizons were characterised by quite different ceramic traditions.

The second site, Yanuca Island rock shelter, duplicated the sequence established at the dune site, although the deposits, largely unstratified, - 283 did not segregate the different wares with the clarity of the Sigatoka site. The rich Lapita-decorated sherds in the bottom levels of Yanuca, although similar to the Sigatoka material, had “much more profuse and elaborate decoration.” 32 This observation was reinforced by the earlier C14 date from these deposits of 2980 ± 90 B.P. (1030 B.C.). 33 The third site, a beach midden at Natunuku has only been partially tested, but with a similar richly decorated Lapita ware it has yielded the earliest date from Fiji and the western Pacific of 3240 ± 100 B.P. (1290 B.C.). 34 These early dates are in conformity with the original dates from site 13 in New Caledonia and with the shell date from Tonga. A date from Golson's excavations at the Lapita site of St Maurice on the Isle of Pines of 2855 ± 165 B.P. (905 B.C.) 35 is also early, attesting to a rapid dissemination of this pottery across the western Pacific in the early years of the first millenium B.C. On current evidence, it survived into the Christian era only in Tonga and from a recently discovered site, 36 in the New Hebrides.

The early Lapita-dominated phase of Fijian prehistory (called the Sigatoka phase by Green) 37 was replaced by an entirely new pottery tradition, with paddle-impressed decorations and new pot-shapes. The date of replacement is uncertain: an eighth century B.C. date from Yanuca rock shelter may belong to the beginning of the impressed ware phase. 38 It may equally date the terminal Lapita occupation in the site. 39 This early date, overlapping with the Lapita dates from Tonga and Sigatoka, is in marked contrast to a date (in the second century B.C.) 40 recovered from another sample higher in the Yanuca deposits, in a secure impressed ware context. This date, in turn, is close to one recovered by Gifford at his important impressed ware site, 17A, at Navatu. 41 The middle layer with impressed ware at the Sigatoka site is even more recent, in the second century A.D., and, from Gifford's site 17B, seventh and eighth century A.D. dates 42 further extend the duration of the impressed ware phase (Green's Navatu phase). 43 This phase, on current evidence, ends as Green has suggested, about A.D. 1000-1100. 44 when a change in the proportion of paddle-impressed ware sherds is - 284 noticeable and new rim types are introduced. These C14 dates from Fiji are listed in Table 2.

The origin of the impressed ware is uncertain. There are few obvious continuities in decoration which would suggest that it had developed out of a Lapita-like base within Fiji. Similarly there are few parallels to impressed ware outside Fiji, 45 although Golson has recently noted the presence of a similar style in south-western New Caledonia. 46 For reasons to be discussed below, a New Caledonian origin for impressed ware seems unlikely. If there is some validity in the early eighth century B.C. date for impressed ware from Yanuca which overlaps with the date from the Sigatoka-Lapita, separate origins for the two contemporary wares must be postulated. If the emergence of the impressed ware, however, is as late as the first century B.C., there are 400 years of Fijian prehistory between the increasingly plain ware of Sigatoka-Lapita and the fully developed impressed ware at Yanuca which could account for the change.

TABLE 2

Radiocarbon Dates from Fiji

Site		Dates B.P.	Range: 2 Std Deviations, AD/BC	Median Date AD/BC
(I/IN) 17A Navatu	M-36	950 ± 300	AD 400-AD 1600	AD 1000
(I) 17B Navatu	M-367	1200 ± 500	250 BC-AD 1750	AD 750
(I) 17B Navatu	M-350	1300 ± 500	350 BC-AD 1650	AD 650
(I) Sigatoka	GaK 1206	1720 ± 80	AD 390-AD 70	AD 230
(I) 17A Navatu	M-351	2000 ± 500	1050 BC-AD 950	50 BC
(I) Yanuca	GaK 1228	2060 ± 100	310 BC-AD 110	110 BC
(L) Sigatoka		2460 ± 90	690 BC-330 BC	510 BC 2
(I/L) Yanuca	GaK 1227	2660 ± 90	890 BC-530 BC	710 BC
(L) Yanuca	GaK 1226	2980 ± 90	850 BC-1210 BC	1030 BC
(L) Natunuku	GaK 1218	3240 ± 100	1090 BC-1490 BC	1290 BC 3
2   Referred to by Birks and Birks, 1968:105. 3   Personal communication, B. Palmer, Fiji Museum.

(L) Associated with Lapita pottery.

(I) Associated with impressed ware.

(IN) Associated with incised ware.

The origin of the third pottery phase in Fiji, characterised by an increasing dominance of incised ware and the introduction of specialised rim and pot-forms is also unclear. The evidence from Gifford's site 26 of sherds with both paddle-impressions and “fine-line incising” 47 would suggest that this phase was a development out of the impressed ware tradition, whereas the widespread Melanesian parallels to the Fijian incised ware, particularly in the central New Hebrides, suggest external influence.

The total eclipse of the sophisticated Lapita tradition in Fiji, whether through the influence of invasion, diffusion or internal development, is - 285 one of the most remarkable events documented by archaeology in the western Pacific. Although the number of Fijian Lapita sites is small, and their distribution restricted to the coast, they nevertheless record a rich and successful culture which survived for perhaps as long as 1,000 years, and for a considerable period of which they were the only known settlements in this rich island environment. The later impressed ware sites are more frequent and widespread, being recorded not only on Viti Levu, but also on Wakaya Island 48 and Kabara (in the Lau Group). 49

The relevant facts from the Fijian sequence are:

• 1. There is no evidence for occupation contemporaneous with the earliest Lapita settlement of Fiji. Only one date from Yanuca, the pottery associations of which are not firmly established, would alter Green's claim for the commencement of the impressed ware phase in the first century B.C.
• 2. The impressed ware, apart from some similar pot-forms, shows little obvious relationship in rim types or decoration to Lapita ware.

The work in Tonga, Fiji and New Caledonia has enabled the quest for Polynesian origins to be taken one stage further back—the location of the “homeland” outside Polynesia, for which the most obvious candidate is clearly Fiji. A crucial stumbling block in further investigations is the disconformity between the early levels of Samoa and Tonga.

Unfortunately, the pottery recovered from the base of the Vailele mound in Samoa shows no distinctive Lapita decorative motifs or rim types. 50 A more sophisticated ware, also plain, recovered in the final season of Green's Samoan project at the site of Sasoa'a and another mound at Vailele (Va4), 51 although differing from the thick variety recovered by Golson in 1957, was sufficiently unlike the pottery in Tonga for Green to claim that they “are far removed from the Lapita ceramic tradition of the same date in Tonga.” 52 In contrast, the richly decorated sherds from Tonga are very close, and often identical to, material from the early Lapita sites in Fiji.

The linguistic relationship between the three island groups, a relationship which must reflect the order of isolation of the three linguistic communities, does not uphold the pottery evidence. The present-day Tongan and Samoan languages share innovations inherited from a common “proto-Polynesian” language which do not occur in the present-day Fijian group of languages. This lack of resolution between the two types of evidence is illustrated in Figure 1.

The most obvious explanation of the lack of resolution between the linguistic and pottery relationships is that the pottery movements have no historical connection with the movements which led to the isolation of the present-day speech communities.

- 286

FIGURE 1

Illustration

- 287

This situation could have arisen in several ways. The more obvious possibilities are:

1. Lapita was a trade ware exclusively (Figure 2).

Although this explanation would leave the linguistic relationship intact, other problems do occur.

(a) If the trading centre was Fiji, then the almost total absence of impressed or incised ware in Tonga could be explained only by trade

FIGURE 2

Illustration

links ceasing by about the first century B.C. In this case, Poulsen's claim for the longevity of the Lapita tradition in Tonga would be incorrect. Alternatively, the Tongan Lapita could persist as a trade ware only if the trade centre was either outside Fiji, or in some region in Fiji as yet undiscovered, where the Lapita tradition persisted unaffected by its widespread replacement by impressed ware elsewhere in Fiji. Appeal to yet another unknown island group or region to support the trade-ware hypothesis in the light of the Tongan sequence is unsatisfactory.

(b) In addition, the Samoan ware, apparently unrelated to the Lapita trade but, according to Dickinson's evidence made in Samoa, 53 would either have to be a short-lived re-invention of pottery in Samoa or Samoan-like ware will eventually be recovered from Tonga (especially in the lowest levels), where it was quantitatively submerged by the fancier imported Lapita pottery.

The acceptance of a trade-ware hypothesis is further complicated by the physical character of Tongan pottery. Key 54 has shown this contains - 288 a specific mineral sand temper which has not been located in the Tongan archipelago but is known in Fiji. 55 This fact increases the likelihood that Tongan pottery is imported from Fiji, but does not explain the presence of a different ware in Samoa or the survival of the pottery in Tonga.

A simple trade-ware hypothesis, therefore, would explain the Fiji-Tonga pottery relationships only if the use of pottery and trade ceased

FIGURE 3

Illustration

between Fiji and Tonga by the first century B.C.; it does not account for the presence of distinct pottery in Samoa.

2. The movement of the present linguistic communities was unrelated to the movements of Lapita potters (Figures 3 and 4).

- 289

(a) The simpler explanation (hinted at in Green's Samoan report) 56 (see Figure 3) would locate the origin of the Samoan ware in a tradition ancestral to the Fijian Impressed Ware. It necessarily assumes that:

• (i) Lapita pottery had nothing to do with the movements of the Polynesians. The present-day populations in Tonga and Fiji displaced a non-“Eastern Oceanic”-speaking Lapita population.
• (ii) To maintain the Tonga-Samoan linguistic relationship, the original Lapita-population in Tonga would have to have been completely replaced by “proto-Polynesian” speakers from Samoa. There is neither linguistic nor archaeological support for this total replacement hypothesis.
• (iii) Tongan pottery, if it did survive into the Christian era, would be based on the Samoan ware and the Lapita style should abruptly disappear. There are links between late Tongan pottery and Samoan ware (see below) so this possibility cannot be entirely ruled out, but once again Poulsen's Tongan sequence would require revision, and the linguists would have to accept that this replacement was so total that no linguistic clues remain.

(b) An alternative explanation (Figure 4) would connect all the known information together by assuming that Lapita ware developed out of a generalised impressed ware tradition, in a region as yet unlocated.

This explanation, however, would require a considerable extension of Samoan prehistory well before the dated 800-1200 B.C. highly developed Lapita styles of Fiji and Tonga. It would also put the ultimate origin of the Samoan ware and the Fijian impressed ware into the remote past, earlier than any known pottery from Island Melanesia. In addition, it would claim that the earliest (Lapita) settlers in Fiji were, in fact, “Tongans” (i.e. proto-Tongic speakers) who were displaced by a people derived from the same ancestral group with a less-modified pottery style. All other Lapita settlements would also have to be founded by populations with a similar language unless a trade explanation is adopted to account for the widespread distribution of the ware. Within the present facts the explanation suggested in Figure 4 seems unlikely.

Although many other more complex explanations could account for the apparent discordance between the archaeological and linguistic evidence, the above models cover most important possibilities. It is significant that in all but two of the previous explanations, the longevity of pottery in Tonga appears anomalous. The evidence from Tonga, therefore, requires critical examination.

The Field Evidence

Tongatapu, the main island of the Tongan archipelago, has been the central focus of archaeological interest. McKern ventured as far as the neighbouring island of 'Eua, but later field workers including Golson and Poulsen were restricted, because of transport difficulties, to Tongatapu.

- 290

FIGURE 4

Illustration

Tongatapu is a flat, coralline island blanketed with thick and fertile red-brown soil derived from weathered coral and volcanic ash. Low coral cliffs to the south and east are testimony to a tilting process which apparently had some significance in the pattern of settlement. The low, northern coastline is enclosed by an extensive outer coral lagoon with an inner, shallower lagoon. The mangrove-saturated foreshore around most of the inner lagoonal fringe suggests a slow infilling by coastal progradation or from tectonic movement.

For many miles along this indented lagoon shore, particularly near - 291 present-day settlements, potsherds can be recovered from the surface. Pottery, in fact, is so common along the lagoonal fringe as to be considered by Tongans as part of the soil itself. Rubbish dumps, wells, latrines, agricultural activities, house-building, earth-oven construction and the myriad destructive acts of everyday living are constantly stirring the deposits and shattering the fragments of pottery into ever smaller pieces. It is impossible in these areas today to dig a ditch or earth oven, fill in a hole or build a house platform without uncovering potsherds. This abundance of pottery, restricted to a narrow belt of highly desirable land fringing the lagoon, makes field interpretation difficult. The location of undisturbed deposits is difficult despite the profusion of material. Large sherds, from which pot-shapes can be readily inferred are rare, the vast proportion of sherdage being less than 1 square inch.

The discrete distribution of pottery-bearing sites on Tongatapu is difficult to reconcile with claims for a continuous pottery tradition. Apart from stray finds, pottery is restricted to the lagoon fringe: the southern and western perimeters and the interior are virtually bereft of pottery. It is possible to argue that this coastal profusion conforms to a prehistoric settlement pattern of villages situated along the lagoon shore. This hypothetical settlement pattern is in conflict with the clear statements of the earliest explorers in Tonga who failed to record villages (apart from the political centre of Mu'a) and described the settlement pattern as “dispersed.” 57 The present-day “ribbon development” of villages along the encircling road is a post-European development, identical to the post-European changes in settlement pattern in New Zealand and Samoa. 58 It is entirely possible, however, that the “dispersed” settlement pattern observed by Cook and the French explorers had not been a continuous feature of Tongan life, but, as Sahlins observes, the widespread distribution of resource zones in Tonga was more suitable for a dispersed than a nucleated settlement pattern. 59 Even if settlements during prehistory were predominantly concentrated along the lagoon fringe, it seems improbable pots were so seldom taken for cooking or storage to the inland plantations that their remains would be so sparse.

The most striking feature of this distribution of pottery-rich sites is the abrupt boundary between pottery rich and lean areas. During 1965-6 the author, with a party of students from New Zealand universities, conducted a detailed survey of the low-lying western tip of Tongatapu where the inner lagoon and the south-west coast are only a few hundred metres apart. 60 Pottery was detected only in the very narrow ribbon along the landward side of the Nukualofa-Kanukupolu road. On the fertile flat land between the road and the lagoon edge, over 100 metres distant, surface sherds were almost entirely absent. In contrast, on the inland side of the road potsherds could be found in vast numbers although they rapidly disappeared within a few metres of the road-edge. The - 292 puzzling but convenient arrangement, greatly facilitating pottery collection from a car, was investigated by three test excavations at different points along the road. In each of these excavations the pottery bearing soil (only half a metre thick on the deepest site) lay upon old beach sand containing water-worn shells and coral fingers washed in by wave and tidal action; the road apparently was built upon an ancient beach. The pottery was found only along this old beach-line and despite the intensive agricultural activity, only a few sherds had migrated any distance into the interior. The almost total absence of sherds on the lagoon side of the old beach-line (and road) suggested that the progradation which had produced this fertile land was more recent than the old beach with the pottery deposits. In addition, for however long this land had been exposed, no pottery had been broken upon its surface.

This situation, with pottery profusion in a narrow band some distance from the present lagoon edge, was typical of this western region. Systematic survey towards the east along the lagoon fringe confirmed the presence of the pottery on or close to the same slightly raised beach line found at varying distance from the present lagoon edge. Around the villages of Pea and Ha'ateiho, where the land is somewhat higher, a series of what appear to be old marine or lagoon terraces 10 metres above present sea-level approach close to the lagoon fringe. Here the pottery sites are sometimes found on these terraces when they are close to the present lagoon, or, as in the important Pea/Ha'ateiho region, on the flat lower land, below the terraces. Towards the eastern edges of the lagoon where detailed surveying of the ancient beach line was discontinued, the higher land beyond the marine terrace is very close to the lagoon edge. Part of the frontage of the present village of Mu'a, for instance, falls abruptly into the lagoon. Significantly, however, the pottery remains do not penetrate inland, except in the Pea/Ha'ateiho region and behind the present town of Nukua'lofa where the land is very little above present lagoon level, and the lagoon edge was probably very different when the sites were founded.

It is clear that many parts of the land along the lagoon edge post-date the beach line, presumably won from the lagoon by slow progradation (a process which is still going on today) or by tectonic activity. A large part of the present town of Nuku'alofa for instance is almost certainly founded on such land. It is significant, therefore, that in nearly three months of observation no potsherds were found in the main town area or along the waterfront. Only towards the back of the town where the new police barracks is situated, or where the Mangaia mound excavated by the Birkses is located can pottery be found. In fact although the site was missed during the initial survey, the largest pottery bearing site seen by the author in Tonga was uncovered by bulldozers clearing the land for the new police barracks.

At the time of Poulsen's work, pottery had not been reported from the northern islands of the archipelago. In 1968, a small field party from Auckland University and the Australian National University managed - 293 to hire a small boat to visit the other islands of the archipelago. 61 The reconnaissance trip, which included all the main islands plus many of the smaller ones, established the presence of pottery throughout the Tongan archipelago. Like the non-lagoonal areas of Tongatapu, however, the pottery was only in very small quantities. It took several days, for instance, in the most northerly island, Vava'u, before a single highly eroded sherd was recovered. Despite the ease with which many of these small islands could be surveyed, pottery was extraordinarily sparse, although present on almost all the islands investigated.

A rich site, comparable to many on the lagoon fringe on Tongatapu, was discovered on the outskirts of Hihifo, the main town in the Ha'apai group, and a relatively rich site was located on the island of Felevai off Vava'u. Apart from these two sites, only a few dozen sherds, always small, highly eroded and without decoration, were recovered from the entire trip.

Two significant points emerge from this reconnaissance:

• 1. Pottery, undoubtedly of the Lapita tradition, is throughout the Tongan archipelago, although in small quantities. This represents the largest concentration of Lapita pottery sites in the western Pacific. Elsewhere sites with Lapita pottery are isolated discoveries.
• 2. The argument, proposed by Golson and supported by Poulsen, that with the Lapita potters must be found the origin of the Polynesians is strengthened by the fact that on almost every island occupied by the Polynesian Tongans remnants of Lapita pottery can be found. If the pottery had been restricted to the main island, Tongatapu, it would have strengthened the possibility that the Lapita potters were alien to the present-day Polynesian inhabitants.

These field observations do little to support the argument for longevity of pottery in Tonga. If, as Poulsen claims, pottery persisted in use into the eighteenth century, it is astonishing that pottery remains are so scarce away from the lagoon fringe on Tongatapu or in the northern islands. The evidence, however, is by no means conclusive: many explanations could be offered to account for the discrete distribution of pottery sites on Tongatapu or the pottery of remains throughout the rest of the archipelago.

Apart from the late C14 dates (discussed below) the main support for Poulsen's claim, in fact, comes from the enthnographic records of the early contact period when pots were observed in use in Tonga. This enthnographic evidence, however, is at best ambiguous as most commentators were convinced that the rare pots observed were of Fijian origin. The following quotation from a forthcoming paper on the Tongan enthnographic evidence summarises the situation:

It would appear, however, from the evidence of the earliest explorers, that pottery was no longer made in Tonga, and that what pottery was seen there came from Fiji. The best description of this pottery is in Labillardière (Vol. II, p. 126, and Plate XXXI, Fig. 8), and should be quoted in full:

- 294

The art of the potter has made no great progress among these people. We saw in their possession some very porous earthen vessels, which they had baked indeed, but very slightly. In these they kept fresh water, which would have quickly filtered through them, if they had not taken the precaution to give them a coating of resin. Vessels thus made could be of no use to them in dressing victuals. The natives showed us some of a tolerably elegant form, which they said had been brought from Feejee. We saw them drinking in companies out of cups of this sort, round which they put a net of a pretty large mesh, to be able to carry them about easily. When they had emptied a few of them they went to fill them again out of little holes, which they had dug in the ground, that the water might flow into them.

This description is clearly of Fijian pottery: the Fijians used resin on their earthenware, and the water bottle shaped rather like Aladdin's lamp and illustrated by Labillardière, is obviously of a Fijian pot. (Labillardière 1800 II, plate xxxi, fig. 8; and Gifford 1951 plate 18a.)

Cook saw only two pots at Tongatapu on his first trip “the one was in the shape of a Bom-shell with two holes in it the one opposite to the other, the other was a little Pipkin which could contain about (five or six) pintes and had been in use on the fire” (Vol. II, p. 265). He surmised that these pots could not have been made locally because there were so few of them, and that they could not have come from Tasman's ship because “the time is too great for brittle Vessels like these to be preserved” (Ibid.). By the time he arrived at Nomuka he had changed his mind and decided that the few similar earthen pots seen at that island could have been made there or at some neighbouring island (Vol. II, p. 451). On Cook's final trip there are only two passing references to pots, one in Anderson's journal, where he says “though they have a few earthen globular pots they get them from Feejee” and never use them for cooking (Vol. III, pt. II, p. 942) and the other in Samwell (Vol. III, pt. II, p. 1036-1037) where he says, there are “different kinds of vessels made of clay.” Forster mentions buying some small earthen pots on 'Eua, which were soot-blackened on the outside and suspected at first that they came from Tasman's voyage “but afterwards we rather believed that they were manufactured by the natives themselves” (Forster, Vol. I, p. 471). These pots could be the same as those referred to by Cook as coming from Tongatapu, because Cook is quite explicit in saying that they saw and bought only two pots in the southern part of the Tongan group. Mariner (Vol. II, p; 284) refers only once to pottery in Tonga: “They perform the process of boiling in earthen pots of the manufacture of the Fiji islands, or in iron vessels procured from ships, or in banana leaves.” It is interesting that Dillon has an identical sentence to that in Mariner (Dillon, Vol. II, p. 81) and that D'Urville's description is almost the same (Vol. IV, p. 281): “ . . . - 295 quelquefois ils les font bouillir dans les vases en terre qu'ils tirent des isles Viti.”

Pottery, which was of such great importance, particularly in cooking, in early prehistoric times in Tonga, was obviously of little importance to the people at the time of European contact. What pots they had clearly came from Fiji at this time, and if the numbers of the pots had increased by the time Labillardière and Mariner were in Tonga, then this was presumably the result of increased trade with Fiji. 62

The Evidence of C14 Dates

It is obvious from the previous discussion that the archaeologist in Tonga is faced with one overriding difficulty: the location of relatively undisturbed sites. The narrow distribution of the sherdage, coinciding closely with present-day settlements, does not offer a favourable environment for preservation of evidence. In addition, the intensive agricultural activities of the Tongans ensure that few sites outside the villages survive intact. It is highly probable that every square inch of this extensively settled island has been turned over many times during the 3,000 years of prehistory. This situation poses an almost impossible problem for the archaeologist: how to establish the persistence or otherwise of pottery when, even today, along the rich lagoon fringe, all earthen-working activities shuffle sherds into new deposits. A house mound built in the village of Pea in 1970 will quite likely incorporate decorated potsherds identical to those recovered by Gifford at site 13 in New Caledonia or by the Birkses at Yanuca rock shelter in Fiji and dated to the middle of the first millennium B.C. Clearly this situation demands archaeological skill and finesse of the highest order to overcome the inherent difficulties of the environment in which these sherds are found.

Poulsen was well aware of these difficulties:

Stratification within these (shell middens) was so complex that in the absence of trained personnel as excavators, sites were dug in spits, the standard unit being 1 metre by 1 metre in area and 10 cms in depth. As compensation for this more destructive way of digging, profiles were drawn for each metre showing the distribution of the original layers as far as they could be recognised at all. The aim was to make it possible to allocate spits to original layers, that is to refer the artefactual evidence to its original position in the middens in a reasonable way. 63

The reinterpretation of the Tongan sequence proposed in this paper suggests that Poulsen failed in his latter objective “to refer the artefactual evidence to its original position,” although the reasons for this failure are by no means surprising with the highly disturbed sites in Tonga, where the presence of pottery in the soil, in no matter what amount, is no guarantee that the material was not mined from another archaeological context. For this reason, Poulsen sensibly concentrated on excavating - 296 shell-middens which offered the best conditions for primary deposition of potsherds. 64 This tactic proved somewhat deceptive (as described later in this paper), because it was based on the quite reasonable assumption that shell middens persisted as a characteristic feature of Tongan prehistory, and they would contain, therefore, in the form of refuse pottery, a valid ceramic sequence. There is strong evidence, however, that the persistence of shell-fish middens throughout Tongan prehistory is, like the persistence of pottery, a major interpretative problem and not a fact to be assumed.

This crucial problem of context and disturbance is important in interpreting the C14 dates from Poulsen's excavations. As any soil-moving activities in the lagoonal fringe of Tongatapu will inevitably result in sherds being incorporated into deposits, the crucial question is whether the dates refer to in situ pottery or to other activities innocently involving sherds. The full details of Poulsen's C14 submissions are available only from the Gakushuin and the Australian National University laboratories, and four important New Zealand dates are unpublished. From the published Australian National University list, however, evidence of irresolution between the pottery evidence and the C14 dates is apparent. Referring to a date of 400 B.C. from Poulsen's site To6 the commentary proceeds:

Pottery from the site suggested that it belonged to late phase of Tongan prehistory, its lower levels equivalent to upper levels at To1 and To5 (see ANU-23, A.D. 1620). However NZ636 from fire hollow K, dug into clay topsoil 1 m from present sample, gave much earlier date (i.e. 2380 ± 51 or 420 B.C.). (Poulsen commented) ANU-24 confirms NZ636. Both dated fire hollows are in corner of excavated area at Tonga 6 and appear to belong to very early phase of occupation only just touched by excavations. Some of material excavated at site can tentatively be attributed to this early phase, which however stands prior to any stage of pottery sequence established by excavation for Tongatapu. 65

The difficult situation of a pottery sequence starting with an early phase “prior to any stage of the pottery sequence established by excavation” is perhaps not unusual in archaeology. The only earlier date from Poulsen's excavations of 820 B.C. was from shell, associated in a pit with two other charcoal dates of A.D. 1530 and A.D. 1486, a similarly difficult situation for interpretation. 66 It is clear that the span of Poulsen's dates (Table 1) confirms prehistoric occupation in Tonga from c. 400 B.C. and presumably the persistence of pottery throughout this period. It is important to note, however, that, like most C14 date series, there are some inconsistencies. The inconsistencies, however, are unlikely to be with the laboratories as in all dates conflicting with anticipations, second, substantially identical, dates were established. These confirmatory dates can be usefully listed:

- 297

(1)	K961	420 ± 100	(A.D. 1530)
	NZ597	464 ± 82	(A.D. 1486)
(2)	NZ636	2380 ± 51	(430 B.C.)
	ANU-24	2350 ± 200	(400 B.C.)
(3)	ANU-23	330 ± 100	(A.D. 1620)
	ANU-23/2	240 ± 63	(A.D. 1610)

The significance of the inconsistencies between expectations and results in Poulsen's C14 submissions cannot be adequately discussed without reference to results from the author's excavations on a site substantially identical to To6.

Further Investigations in Tonga

At the commencement of my own field work in Tonga in the summer of 1966-7 none of the current issues was clear. Although the original intention had been to investigate the earthwork forts of Tongatapu, I was, like all archaeologists before me, seduced by the rich pottery remains along the lagoons. The preliminary field survey on the western half of the island (described in part above) which revealed the discrete distribution of the pottery along the old beach line was puzzling, as I had expected, with over 2,000 years of continuous pottery-making on a small island, a more even distribution of remains. Having a large party of trained students with me (mainly from Otago University), I determined to exploit their combined skill in an exhaustive area excavation. In selecting a site, preference was given to evidence of surviving stratigraphy and to the presence of plain pottery, as I assumed that Poulsen had fully explored the fancier wares. Working along the old beach line from the western tip of the island, I finished up, as had Goulson and Poulsen before me, selecting a site in the rich Pea/Ha'ateiho region.

The site, called Vuki's Mound (after the land owner), was visible from the surface only as a large flat bump on the otherwise relatively flat open compound of Ha'ateiho village. It was situated on the brink of a relatively steep slope down to mangrove mud and the modern lagoon level. The top of the barely discernible mound was 2.8 metres above present lagoon level, and barely half a metre above the surrounding flat land of the village.

Fortunately a bulldozer had cut through the mound for an access road to the lower ground. The cutting exposed over 1.5 metres of deposits, mainly shell-fish midden, but with some fascinating horizontal layers of yellow coral sand and ash plus clearly visible post holes and pits cut from various levels. The layers from top to bottom were thick with plain pottery.

Because of the complex character of this site (despite the presence of intact horizontal layers) a slow elaborate excavation procedure was adopted. Every potsherd in an intact layer was 3-dimensionally recorded, as well as being recorded and excavated by natural layers. Disturbances were carefully identified and removed as single units with pottery from these disturbed areas bulked together. This complex, multiple recording - 298 system, combining excavation by natural layers with 3-dimensional recording and careful identification of disturbances, required an extensive grid layout, based on 2 m units (10 such units were excavated), special recording apparatus and rigid discipline and field procedures by the excavators. The success of the work is largely based on their acceptance of a strict work routine (6 days a week) from early morning to late at night which would have gladdened the military mind of Sir Mortimer Wheeler.

The intention of this elaborate procedure, which required hundreds of drawings and individual paper bags for every sherd, was twofold: 1. To assess the mechanisms which were disturbing and scattering the pottery, by attempting to match sherds and assemble larger pieces or parts of pots. This was only possible with the complex 3-dimensional recording where it was possible to relocate each sherd and examine all sherds in the vicinity. This intention was only partly realised as the uniformity of the pottery made it unnecessary to reconstruct many pots all of which conformed to a few very simple types. A very large number of sherds found in close proximity, however, could be fitted together, demonstrating that the original pieces which had been deposited in the site were quite large—some clearly complete sides or bases. This was the soundest evidence that many of the more intact layers were not redeposited but contained pottery dumped at the time of the layer formation. Subsequent pressure and disturbance had dislodged many of the pieces. Particularly in the thick midden encircling the mound it was clear that very large pieces of pot had been dumped into the refuse. As this midden partially sealed over the entire site it was apparent, from early on in the excavations, that we were dealing with the remains of pottery-using people.

2. The second and equally important intention of the elaborate method was to demonstrate the utter futility of arbitrary depth excavations in the Tongan situation: indeed the futility of any method of rapid excavation in such complex deposits. With 3-dimensional data on over 22,000 sherds it is clearly possible to isolate sherds at any arbitrary depth or spit dimensions at will: an exercise which, unfortunately, was frustrated by the utter uniformity of the sherdage.

The dis-assembling of the complex activities which had gone into the formation of Vuki's Mound and its subsequent disturbance, took over two and a half months of excavation. Only 2 of the 10 squares were taken to the natural.

The most difficult task, as could have been expected, was finding, isolating and removing the multitudinous disturbances which had destroyed over 60 percent of the exposed area. It cannot be claimed indeed, despite the care in excavation, that all such disturbances were discovered. At least two areas within the gridded excavation were so completely overturned as to defy interpretation. Over 50 percent of the sherds recovered had survived the two-and-a-half thousand years of continuous earth-moving in and around the site. These sherds, surprisingly, were very similar, belonging to a simple range of undecorated cooking pots and - 299 open bowls, with no difference from top to bottom of the site apart from a thin lens of soil at the base of the site which contained a number of decorated sherds and was clearly nothing to do with the main site formation.

The low mound proved to be a series of house floors built one upon the other with an encircling collar of rich shell-fish midden deposited during and after the living activities in the house. Each house floor, where intact, was very flat and at regular intervals the prehistoric housewife had refurbished the surface with clean coral sand. It is unlikely that the house was simply for sleeping and shelter, as the abundant evidence of ash—sometimes scattered completely over the horizontal floors—and the remains of scoop fireplaces and charcoal indicated complex cooking activities. In one area the floors were stained red and black from the cindery remains of fires. It is likely, therefore, that the house was a separate cooking house (as is the present Tongan and Polynesian pattern) or that a different system prevailed during prehistory and cooking was carried out in the living/sleeping quarters. Although many clear post holes were recovered on each of the house floors, no pattern emerged.

The stratigraphy was complex and disrupted in many places by large holes and pits. The large pits, many so deep as to cut into the natural beneath the site, were all post-deposition although a few smaller pits were sealed in by the deposits. Enthnographic evidence would suggest that some of these larger pits may have been for the storage of fermented breadfruit or for ripening bananas.

The pottery throughout the series of house floors and in the encircling midden was identical, although, as could be expected, the sherdage within the house floor-area was small and, because of heating from fires and ash, in a poor state of preservation. Very few sherds from the floors could be fitted together, suggesting extensive horizontal disturbance (scattering by feet, etc.), whereas the sherds in the midden were often found in large clusters and nests, assembling into sherds of considerable size. The sherdage generally was in much better condition in the midden, where, apart from some late pits and post holes, the main cause of breakage appears to have been shattering in situ by pressure of deposits.

Three basic pot forms emerged from the surviving sherds:

• (1) A plain cooking pot with the opening very little smaller than the maximum width of the globular-shaped body. All these pots showed a degree of narrowing immediately below the rim before the body expansion. On the smaller pots this reduction was so insignificant as to be barely noticeable making the sides of the pot virtually straight. The actual rim was always flat and in the case of the larger cooking pot sometimes slightly angled outwards. The open-mouthed cooking pots varied considerably in size although the majority of sherds appear to have come from very large pots capable of holding two or more gallons of liquid.
• (2) The second common type of pot was simple bowls, once again varying in size. Some of these bowls, although undecorated, were extremely finely finished.

- 300

• (3) A small cup, similar to that recovered by Poulsen at his first site (Tol), 67 although considerably thicker-walled was also recovered.

At the base of the mound, as mentioned above, was a soil layer predating the house-building activity. A few tiny sherds of typical Lapita decoration came from this layer. From their eroded state, however, and the fact that none matched, these were hardly likely to be in primary position. Their presence, however, emphasises the communion of this site within the Lapita tradition.

In addition to pottery remains, Vuki's Mound was rich in shell and stone artefacts, largely matching Poulsen's material. 68 The most interesting forms were the shell “long units” drilled at either end, known elsewhere in the Pacific only in Fiji (at Natunuku) and from the Marianas Islands, 69 and the adzes of polished fine-grained rock. Six adzes show closest parallels to the widespread lenticular Melanesian adzes with a curved cutting edge and a base-flattened oval cross-section. The form was found in the earliest levels in the Marquesas by Suggs (his Hatiheu type), in Niue (an undoubted Tongan colony), and shares close morphological similarities with adzes from many parts of Melanesia, including Fiji. A single, small, rectangular-sectioned adze was also recovered from a reliable early context on the site.

As the argument developed in this general paper does not require the finer details of artefacts and is not intended to anticipate the full site report of Vuki's Mound, the reader will be spared the prosaic tabulation of dimensions and typologies. The site, extremely uniform in material culture, is closest to the most recent of Poulsen's sites, To6, from which he recovered at the base the anomolous C14 dates in the fifth century B.C.

Chronological Issues

The anomolous dates from the base of Poulsen's To6 site were accepted as dating an occupation at least 1,000 years earlier than the refuse dumped on top of the fireplaces from which the samples were taken. This uncomfortable conclusion was necessitated by the internal evidence for a consistent change in Tongan ceramics from a highly decorated phase to one of completely plain ware, with C14 dates spanning the entire prehistoric period. 70 Thus the two 400 B.C. dates from the base of the latest almost entirely plain-ware site To6 were out of context. The dates for Vuki's Mound, therefore, a site closely comparable in material remains, was a crucial test of this interpretation of an undocumented hiatus between the basal fireplaces and the overburden of midden at To6. These dates would also be crucial in that their context within the structure of the site could not be challenged. Over a dozen small samples from sealed fireplaces belonging to individual fires in separate house floors were recovered during excavation of Vuki's Mound. A further - 301 dozen samples taken from various sections and from the main road-cutting exposed by the bulldozers were available as a check. Some of the samples have been dated and are listed in Table 3. The latest of these dates, ANU-442, was taken from a fireplace cut into the house mound after its abandonment and therefore means little more than that the site was abandoned before this date (A.D. 800). Samples ANU 424, 429, 435, 436 and 441 are from clearly stratified sealed layers. Two of these dates (ANU 435, 436) appear anomalous but as they were extremely small samples (4 percent and 9 percent of requirement) with very high error margins, they fit the series within less than one standard deviation. This date series demonstrates, as does the uniformity of the pottery, that Vuki's Mount was occupied continuously for only a short period somewhere between approximately 550 and 250 B.C., exactly comparable, as are the material remains, with the dates from To6.

TABLE 3

Additional C14 Dates from Tonga 71 Vuki's Mound

Layer	Date B.P.	Range: 2 Std Deviations BC/AD	Median Date BC/AD
1b	ANU 442. 1150 ± 90	AD 620-AD 980 BC	AD 800
4	ANU 429. 2210 ± 145	AD 30-550 BC	260 BC
10	ANU 435. 1830 ± 800	1720-1480 BC	AD 120
14	ANU 441. 2440 ± 110	270 BC-710 BC	490 BC
14	ANU 424. 2540 ± 160	270 BC-910 BC	590 BC
15	ANU 436. 2260 ± 415	AD 520-1140 BC	310 BC

These dates from the middle of the first millenium B.C. are for plain pottery only. This strongly suggests that Poulsen's 400 B.C. dates for To6 are correct and that the more recent (A.D.) dates from Poulsen's excavations must be rejected as validly dating the pottery found with the samples. Poulsen's sequence, therefore, must be telescoped into an entirely B.C. context.

The dilemma of the remaining dates, however, cannot be easily resolved. The most recent of his dates ANU-23/1 and 23/2 (A.D. 1620, 1610) are charcoal samples from a pit cut into the site and are therefore not necessarily dating the pottery. The same is true of K-961 and NZ-597 (A.D. 1530 and 1486), and, in addition, a shell date from the same pit gave a date of 820 B.C. (K904) suggesting that the bulk of the infill of this pit (which includes shell and pottery) may have little to do with the date of the pit itself. Two further dates remain, both in fourth century A.D. from To5 and To2. They are included in the as yet unpublished New Zealand date-list and judgements as to their validity must be suspended until further stratigraphic details are available. These two dates, however, according to Golson, date pottery which can be compared to the Yanuca assemblage in Fiji “which, however, is a thousand years earlier.” Golson suggests that this anomoly may be because “the circum- - 302 stances of oceanic settlement, of isolated islands by small founder populations, fostered extreme conservatism in certain branches of culture.” 72

Fortunately, a further check on the validity of Poulsen's dates can be obtained from the recently released shell dates from the Mangaia mound (see Table 4). Three of these dates, however, are from pits cut from near the surface and cannot be accepted as reliably dating the pottery found on the site. Of the remaining two dates, N.Z. 727 is consistent with the To6 and Vuki's Mound dates. This shell sample was from the most reliable layer on the site, layer 3, a mixed soil and midden deposit underlying the mound. The pottery from this layer, with a slightly higher percentage of decorated pottery than either To6 or Vuki's Mound was assumed to be a little earlier so that the date of 680 B.C. is a useful confirmation of the dates from the other two mounds. The much later A.D. 185 date, however, came from the mound material and may not refer directly to the pottery. The exact status of this layer must await publication of the excavation report: if the pottery is in situ then it is the most reliable evidence for the persistence of plain pottery into the Christian era. 73

TABLE 4

Additional C14 Dates (Shell) from Tonga Mangaia Mound 74

Layer		Dates B.P.	Range: 2 Std Deviations AD/BC	Median Date AD/BC
2	NZ-728	1765 ± 45	AD 95-AD 275	AD 185
3	NZ-727 2630 ± 50	780 BC-580 BC	680 BC
Pit J	NZ-725 2100 ± 50	250 BC-50 BC	150 BC
Pit C	NZ-726 3130 ± 70	1320 BC-1040 BC	1180 BC
2b	ANU315 485 ± 60	AD 1345-AD 1585	AD 1465

The pottery, adzes and ornaments from Vuki's Mound and To6 can be so closely allied that little doubt can be entertained about the comparability of the sites: this is reinforced by the C14 dates. The only evidence, in fact, which militates against acceptance of the plain-ware phase of Tongan prehistory dating to the middle of the first millenium B.C. are the two dates in the fourth century A.D. from To5 and To2, purporting to date pottery closely comparable to the much earlier Fijian sites of Yanuca and Natunuku. The contextual reliability of these two dates, therefore, must be re-examined in the light of the To6 and Vuki's Mound dates. The weight of the evidence from these latter sites, from the distribution of pottery on Tongatapu and in the archipelago as well - 303 as the dates for comparable wares in Fiji and Samoa argues against the survival of decorated pottery into the fourth century A.D.

If, as the foregoing argument suggests, the pottery sequence established by Poulsen must be considerably shortened, with plain pottery dominant by 500-300 B.C., it is obviously desirable to establish more precisely the beginning and end of this sequence.

The internal witness of systematic changes in rim form and diminishing decoration established by Poulsen and the external witness of the Fijian sites of Natunuku and Yanuca argue that sites with a high percentage of decorated sherds should be among the earliest settlement sites in Tonga. The 850 B.C. date for shells contained in a pit at Poulsen's Tol site comes into immediate chronological perspective. If, as appears likely from the conflicting seventeenth century dates for charcoal from the same pit, the shells are, in fact, redeposited from the pottery-rich midden horizon into which the pit was cut, their date of 850 B.C. is entirely consistent with the array of decorated ware at the site.

To clarify this issue a sample of six Anadara shell net sinkers from the bottom spits of Poulsen's To2 site (which contained the highest percentage of decorated sherds) was submitted to the ANU laboratory and returned the date 3090 ± 95 B.P. (1140 B.C.) 75 in close conformity to the Natunuku date from Fiji (1290 B.C.) with which the pottery can be readily identified. If these shell dates can be accepted—and they are entirely consistent with the charcoal dates for the later plain-ware, the settlement of Tonga may be as early as the twelfth century B.C.

The highly distinctive dentate-stamp decoration and the complex rim forms of the twelfth century B.C. gradually declined in importance during the first centuries of settlement in Tonga. By the middle of the first millenium B.C. (if the To6 and Vuki's Mound dates are accepted), plain-ware predominated (see Figure 5). From the rich remains at the latter sites, however, it is clear that the use of pottery was still a vigorous tradition; there seems to be little evidence for any decline in use. Despite this fact, it seems probable that the use of pottery did decline by the end of the first millenium B.C., although there is no satisfactory evidence to date its final disappearance.

Evidence for this decline will never be forthcoming in the pottery-rich lagoonal fringe on Tongatapu. Assessment of the amount of pottery in any site or layer, a clue to decline in use, is unreliable. Only in areas outside this pottery profusion will the difficult question of rate of decline and date of eventual abandonment of pottery be resolved. In this regard the poverty of pottery remains on the northern Tongan islands is significant. Not only is pottery rare but it was only on Ha'apai, the closest major group to Tongatapu, that dentate-stamp decoration was found, although a single sherd with an early “fancy” rim was recovered from Vava'u. This initial reconnaissance suggests that the exploration and colonisation of the archipelago took several hundred years and that the larger northern islands were not settled until the decline in the importance of decoration was already well advanced. The poverty of pottery remains from the - 304 majority of the small coral islands would suggest that their colonisation was not effected until pottery was already declining in importance. More crucial as a test of the decline of the pottery tradition in Tonga is the evidence from the almost unexplored island of Niue, which, on linguistic evidence, is an undoubted Tongan colony. 76 A stray C14 date of shell from a midden on Niue of 1830 ± 40 B.P. (NZ-729), 77 approximately A.D. 120, confirms the linguist's claims that Tongan and Niuean have developed separately for a considerable period. So far there are no reports of pottery from Niue, although an adze with base-flattened lenticular cross-section identical to those recovered from Vuki's Mound is reported as a surface find from the island. A careful field survey of Niue offers the most immediate means of determining the chronology of pottery decline in Tonga.

The internal evidence for decline from the excavated sites is, from the nature of these sites, unsatisfactory. The A.D. 185 date from the top of the Mangaia mound is not conclusive evidence of pottery persisting into the Christian era as there is a strong possibility that this layer, part of the mound build-up, is redeposited and the sherds are not directly associated with the date. It seems clear, however, that, at least at Vuki's Mound, there was no sign of pottery by A.D. 800 when the fireplace dated by ANU 442 was dug into the house mound. The pottery in the soil sealing this fireplace is entirely derived from the deposits beneath. If, as the argument pursued in this paper requires, the two dates for decorated pottery in the fourth century A.D. must be discounted, they nevertheless do suggest that fire pits, with charcoal, were cut into deposits at these two sites without, apparently, any plain pottery in direct association. This is, perhaps, indirect evidence of the total eclipse of the pottery-using tradition by this time. In fact, there are no dates reliably associated with plain-ware other than those already quoted for To6, Vuki's Mound and Mangaia. The exact chronology of pottery decline and abandonment must await further field work in the northern islands of the archipelago and in Niue, but in the author's opinion it seems unlikely that pottery persisted for any length of time into the Christian era.

This reinterpretation of the Tongan sequence is dependent on the validity of Poulsen's claims for the gradual loss of decoration and fancy rims during Tongan prehistory with an increasing dominance of plain-ware. Comparable plain-ware, however, is an important component of all Lapita sites in the Pacific. Golson in a recent review of Lapita pottery says:

A substantial proportion of the (Tongan) excavated pottery was undecorated: Poulsen's site 2, richest in decoration, produced 88½% undecorated pottery by weight. For the Lapita site in New Caledonia Gifford and Shutler (1965:71) give a figure of 63% by - 305 weight for plain pottery. Statements to the same effect are on record for Watom (Casey 1936:14; Specht 1968:128) and for the Sigatoka sand dune site in Fiji (Palmer 1968:20). 78

It is the loss of the diagnostic decorated element which characterises the Tongan sequence. The plain-ware pot forms, however, may have changed very little: a nest of plain pots recovered at the Natunuku site in Fiji (c. 1200 B.C.) are virtually identical to the reconstructed plain pots from Vuki's Mound (c. 500-200 B.C.). This overall tendency for the persistence of the plain forms and the loss of the elaborate decoration appears to be duplicated in Fiji where at the Sigatoka site the distinctive dentate-stamp Lapita motifs have largely disappeared, replaced in some instances by new decorative devices. The possibility must remain, however, that the plain and decorated components of the Lapita pottery tradition represent a functional rather than a chronological difference, the decorated, for instance, being a trade or ceremonial ware, with the plain pottery having a more limited domestic application. If this were so, the Tongan plain-ware sites of To6 and Vuki's Mound would represent specialised domestic structures (i.e. cooking houses), an interpretation not inconsistent with the character of the sites. They could, therefore, be contemporary with other less specialised sites where the decorated ware was also in use. If these sites are specialised and the plain-ware does not represent the full range of pottery in use at that period, then the argument in this paper would be invalidated. For a number of reasons which cannot be detailed here, particularly the character of the non-pottery artefacts, this interpretation appears unlikely. 79

Although the reinterpretation of the Tongan pottery chronology outlined above has been expressed mainly in terms of the conflict in C14 dates, the real issue is the relationship between Tonga and the neighbouring islands of Fiji and Samoa. In fact, the evidence from C14 dates is, by itself, poor. Not only do the standard deviations of all the dates from Tonga overlap with wild abandon, allowing the archaeologist statistically valid choices ranging from the middle of the fourteenth century B.C. to the present day, but the argument presented here demands the rejection, as validly dating the pottery remains found with them, of no fewer than six of Poulsen's dates, one of my own dates and three of Golson's dates in favour of only three of Poulsen's dates, five of my own and only one (or perhaps two) of Golson's dates. The basis of rejection of these dates, however, is the stratigraphic evidence from each of the sites as well as the total logic of the western Pacific archaeological situation, reinforced by a common-sense appreciation of the technical difficulties of dating on the pottery-rich Tongatapu lagoon. Were it not for the late C14 dates from Poulsen's sites, it is probable that claims for the longevity of the Tongan pottery tradition would have received little prominence. Misplaced confidence in the stratigraphic integrity of Poulsen's C14 dates denied the - 306 simplest solution to the problem of Polynesian origins—that Tonga was a link between the earliest evidence from Fiji and Samoa.

A detailed examination of the material recovered from Poulsen's excavations (housed at the Department of Prehistory, ANU) 80 did not increase confidence that the dates were in primary association with the pottery recovered from the sites. Weaknesses in the original analysis of the material, particularly a failure to account for differences in the quality of the sherdage from the various horizons (apparent from its size and degree of abrasions of the sherds) frustrated one of Poulsen's original intentions “to refer the artefactual evidence to its original position in the middens.” In addition, the other artefactual evidence recovered from the excavated pottery sites in Tonga, including adzes and ornaments, did not suggest that any of the sites were recent. All the adzes recovered from Vuki's Mound, for instance, were of a form rare in surface collections from Tonga. The numerous shell “necklace units” from the pottery sites (including Vuki's Mound) were unknown in ethnographic collections from Tonga. 81

It is abundantly clear to the author that, unless new evidence is forth-coming, claims of persistence of decorated pottery in Tonga after about 500 B.C. cannot be sustained. The survival of plain pottery into the Christian era is also problematical but this issue must await evidence from areas outside the richness of the Tongatapu lagoon.

Polynesian Origins Reconsidered

The shortening of the Tongan pottery sequence resolves many of the outstanding problems of Polynesian origins. The disjunction between the linguistic data and the archaeological evidence from Fiji and Samoa, discussed in the first sections of this paper, no longer applies. Indeed, the “fit” between the Fijian, Tongan and Samoan sequences is almost too neat.

Fiji-Tonga Relationships

The new shell date of 1140 B.C. from the base of Poulsen's To2 site is in close conformity with the date for the Natunuku site in Fiji. The pottery from both sites is virtually identical in technique of manufacture, pot form, decoration and rim elaboration. There seems little reason to doubt that, by the end of the twelfth century B.C., people with Lapita pottery had penetrated into the region we now call Polynesia. In all probability, at this early date, the Fijian and Tongan Lapita populations were a closely related cultural community, the perfect candidate for (in linguistic terms), the pre-Polynesian (East-Oceanic) speech community. The subsequent isolation following separation led to the linguistic innovations which separate the Polynesian and Fijian languages. During 1,000 years of settlement in Tonga, during which the decorative

- 307

FIGURE 5

Illustration

- 308

basis of the Lapita pottery style disappeared and pottery itself became less important, the unique innovations, both linguistic and cultural, which we now identify as Polynesian, developed in the isolated archipelago. By at least 500 B.C. (the approximate date of the base of Vuki's Mound) the language in Tonga should have approached the status of “proto-Polynesian”, the ancestral language of Polynesia. In contrast, in Fiji, the characteristics of the Fijian members of the Eastern Oceanic languages should have been emerging. This linguistic divergence is exactly paralleled in the pottery evidence: the Lapita potters at Sigatoka in Fiji and Vuki's Mound in Tonga, roughly contemporary in C14 terms, had each translated the original Lapita style into divergent forms, although, as mentioned above, they shared a tendency for loss of the diagnostic dentate-stamp decoration. There seems little doubt that the independent development of Lapita pottery in the Fijian and Tongan islands is documenting the long-sought-after separation between Fiji and Polynesia, a conclusion anticipated by other archaeologists. 82

This simple solution, however, is complicated by the lack of identity of the later Fijian impressed and incised wares with the original Lapita pottery. As Green has argued, it seems likely that the latest Fijian pottery style, incised ware, which became widespread by about A.D. 1000, is intrusive and is witness of a major cultural impact upon Fiji, possibly from the New Hebrides or New Caledonia where similar pottery is known. 83 The origin of the earlier impressed ware, however, is unknown. As this pottery style, unlike the even earlier Lapita ware, became established in almost all parts of Fiji and persisted, with only minor changes, for close to 1,000 years, it must be assumed that the present-day Fijian population is largely descended from the makers of impressed pottery. The various Fijian languages, similarly, must be derived from the languages spoken by the users of impressed ware, unless we are to accept the unlikely possibility that there was a total population (or speech) displacement with the intensive influence documented by the sudden dominance of incised ware. If, as previously argued, the Lapita potters must be identified with a pre-Polynesian-Pre-Fijian speech community, and, largely on commonsense grounds, the makers of impressed ware identified with the Fijian language, the two pottery-making populations must be related. Although it is feasible that the impressed ware pottery style was also a foreign introduction imposed on an already existing community in Fiji, the new evidence from Tonga lends weight to the possibility that impressed ware is a Fijian development out of Lapita pottery, and that somewhere in Fiji sites documenting the emergence of impressed ware from Lapita plain-ware will be discovered. In fact, unless the Yanuca rock shelter date from the eighth century B.C. is associated with impressed ware rather than Lapita ware, there is a considerable gap of about 400 years between the still clearly Lapita-derived ware from the lowest level of the Sigatoka sand dune site - 309 (c. 500 B.C.) and the earliest impressed ware from Gifford's Site 17A. The derivation of impressed out of a plain ware similar to Lapita plain is not implausible in terms of pot forms, and in the absence of evidence to the contrary. The possibility that impressed ware is a local Fijian development from Lapita ware seems a reasonable hypothesis. Strangely enough, one large sherd from Poulsen's site To6 is indistinguishable from impressed ware sherds from Fiji, 84 suggesting that paddle-impressing (which produces the characteristic “wavy relief” of the style) is not foreign to the Lapita tradition.

Although the relationship between Lapita pottery and impressed ware must remain hypothetical, there is little doubt that recognisable Lapita traits had disappeared in Fiji by the first century B.C.—or, at about the same time that the Tongan pottery tradition appears to have become defunct. The apparent coincidence that Lapita ware ceases in both island groups at about the same date requires further discussion. This situation raises, inevitably, the possibility that the Tongan pottery was a trade-ware imported from Fiji and that it ceased when the source of supply was no longer available. Although this possibility seems remote with such a quantity of sherdage in Tongatapu, the results of physical analysis of Tongan pottery by Key 85 show that the sand temper in pottery from both early and late Lapita contexts in Tonga is identical, suggesting that a single homogeneous beach deposit has been the source of the temper. Unfortunately, however, a source for the unusual and characteristic pyroxene-rich mineral sand used by the Lapita potters has not been located in the Tongan archipelago where almost all the beaches are composed of yellow coral sand. One of the main purposes of the survey of northern islands in 1968 was to discover the beach from which the temper was mined: of the few mineral sand deposits located, none had the correct mineral content. The nearest beach with minerals corresponding to the temper in Tongan Lapita ware is on Vanua Levu on Fiji 86 and others may exist. Although the issues raised by the failure to locate the temper source in Tonga are irrelevant to the argument of this paper, the possibility that Tongan pottery was largely, if not totally, imported from Fiji cannot be rejected.

If the pottery was imported then trade ceased sometime about the birth of Christ, as sherds from the later Fijian ceramics have not been discovered in Tonga. A stray Lapita pottery community could perhaps have survived in Fiji after the emergence of impressed ware. Similarly, if the source of supply in Fiji did disappear at about this time, it coincides with the appearance of impressed ware, strengthening the “invasion” hypothesis to account for the origin of impressed ware. It would be convenient if the political and social upheaval in Fiji following “invasion” by the impressed ware potters had its repercussions in Tonga, as further work in Fiji would thereby resolve the difficult problem of dating the cessation of pottery in Tonga. The inevitability, from the linguistic - 310 evidence, that the people making impressed ware were descended from the Lapita potters makes the “invasion” hypothesis unlikely, and despite the puzzle of the source of the temper used by the potters, anyone familiar with the sheer quantity of pottery in Tonga shuns the thought that it was imported from Fiji.

There is another equally puzzling coincidence between the evidence from Fiji and Tonga which bears heavily upon the previous discussion.

One of the most unexpected implications of the revision of the Tongan pottery sequence is that, on present evidence, the founding of rich shell-fish middens appears to have stopped at virtually the same time that pottery ceased to be used. Pottery in Tonga is invariably associated with shell-fish midden remains: inversely there are no reported middens without pottery, although with such pottery profusion this observation may be deceptive. Although the cessation of pottery and midden building in Tonga may not be related, the evidence is sufficiently convincing to suggest that, toward the end of the first millenium B.C., there was a significant shift in the orientation of prehistoric Tongan economy, or alternatively a change in the settlement-dumping habits of the population. The cautious reader will no doubt raise the obvious alternative explanation, that local environment conditions led to a reduction in shell-fish supplies. Although this possibility cannot be ignored, the fact that shell-fish from Vuki's Mound are as large as those from the earlier sites does not suggest that the reef was being over-exploited. 87 There is no evidence for tectonic activity at about this time which could have altered the reef and disrupted the environment for the shell-fish. However, there is clear evidence for uplift preceding or coinciding with initial settlement, as all the pottery-bearing sites are founded upon an apparent old lagoon bottom of clean coral sand, with water-washed coral fingers and shells. The slow infilling of the lagoon following uplift may have altered the reef environment, resulting, by about the end of the first millenium B.C., in diminution of shell-fish supplies. There is no evidence for this in the middens. The heavy Anadara and Gafrarium shell-fish which comprise the bulk of the midden remains are still readily available from reef and lagoon.

This unexpected phenomenon, however, would have to be explained as a purely local development were it not for the coincidence that apparently there was a similar cessation of shell-fishing or change in shell-dumping pattern in Fiji. The earliest site known in Fiji, Natunuku, was rich in shell-fish remains. 88 In contrast, Gifford notes that evidence for heavy shell middens was restricted in his sites to the upper incised ware levels. 89 Green has criticised his correlation of incised sherds with shell middens, saying that “the presence or absence of shell-fish seems to depend on the position of the site with respect to the coast and the use to which the portion of it excavated was put at various points in time.” 90 The linking sites between Natunuku and Gifford's later sites - 311 are little help in resolving the dilemma of a fluctuating pattern of midden dumping: no shell or bone survived at the Sigatoka site, 91 and at Yanuca, although shell-fish was throughout the deposits, it was mixed with earth and there was no concentrated midden. 92 The Yanuca evidence, however, would suggest that shell-fishing continued uninterrupted, but Gifford's evidence, despite Green's reservations, strongly suggests a difference in shell-fish dumping patterns between levels with impressed ware and those with the later incised ware. Although, apparently, the pattern of midden dumping was eventually restored in Fiji, the absence of concentrated middens in the post-Lapita impressed ware sites, noted so consistently by Gifford, is remarkably comparable to the cessation of midden-dumping in post-Lapita Tonga.

This coincidence, although only weakly evident in Fiji, does suggest an alternative possibility for the change in shell-fishing habits in Tonga: that this phenomenon documents a shift from reef exploitation to agriculture in the latter half of the first millenium B.C. An obvious explanation for this shift could be the introduction of all or some of the major west Pacific cultigens—taro, yam, banana and breadfruit (or perhaps even the coconut), as well as the pig. The only direct support for this hypothesis is the absence of evidence of pig from the pottery sites in Tonga 93 as well as from the Lapita levels in Fiji. 94 The pig is usually associated with horticulture and its presence in the impressed and incised ware levels in Fiji is sound evidence for the establishment of a viable horticulture by this period. 95 The absence of the pig from Lapita sites, however, is not so helpful, but the unusual distribution of known Lapita sites does suggest that the makers of this distinctive ware were, either by choice or accident, restricted to littoral exploitation—usually in a coral lagoon environment. This is apparent from the invariably coastal or off-shore island location of Lapita sites—there is no evidence that Lapita potters penetrated inland. In addition, the suspicion that the base of the Lapita economy was somewhat restricted is strengthened by the astonishing failure of the Lapita potters to expand and exploit the attractive and uninhabited vastness of the Fijian islands, despite the fact they were apparently the sole occupants for over 600 years. It is not until much later, when either the Lapita style had completely changed or a new people had arrived that the rich Fijian islands began to be systematically settled—by people who were undoubtedly agriculturalists. Much the same appears to be true in New Caledonia: on current evidence, people with Lapita-style pottery were either the first, or among the first, settlers of this rich island. 96 The poverty of Lapita sites in New Caledonia suggests that these early settlers failed to - 312 expand into the agricultural interior, and were rapidly displaced or submerged by people with different pottery styles. Nowhere, except in Tonga (and perhaps in Fiji), did these earliest settlers in the western Pacific leave a noticeable impact upon the present population structure. Their success in Tonga (as in Fiji, if impressed ware is a development from Lapita) appears to correlate with a shift from lagoonal to horticultural orientation witnessed by the switch in shell-fish dumping and, probably, dependence upon reef food.

The hypothesis presented here is that the Lapita potters, initially at least, had a restricted maritime/lagoonal economy and that either the development or introduction of a more viable horticultural economy enabled them to expand and survive in Fiji and Tonga to eventually colonise the remainder of the Pacific. Careful analysis of economic evidence from other Lapita sites throughout the western Pacific should test the validity of this hypothesis: on current evidence it appears, to the author at least, to offer a convenient explanation of not only the phenomenon of midden cessation in Tonga but also of the genesis of the astonishingly rapid colonisation of the western Pacific before 1000 B.C. by people with Lapita ware. In this conception the Lapita potters would be Oceanic “strandloopers” 97 who, like the sealers and whalers in the European period, expanded ahead of colonisation by agriculturalists.

Samoa-Tongan Relationships

The rephrased Tongan sequence completely removes Green's objection to tracing the origins of the earliest Samoans to Tonga. 98 Indeed, the Samoan wares, despite some minor differences, are, as Golson has emphasised, 99 very similar to the plain pottery in Tonga with which it is now contemporary. With the important Polynesian artefact, the adze, the match between the lowest Samoan levels and those of Tonga c. 400-0 B.C. is remarkably close. Green, in a recent paper, 100 has explored the characteristics of the adze assemblages from excavations in Fiji, Tonga and Samoa. The early Samoan adze assemblages can be seen as a progressive development out of the Tongan adze forms, accelerated, perhaps, by the sharp change in available rock between Tonga and Samoa across the so-called “Andesite Line.” 101

The strangest match between Tonga of c. 400-0 B.C. and Samoa of the same date is the absence in Samoa of any evidence of shell-fish dumping. 102 If, as the previous argument suggested, shell-fishing or shell-fish dumping was declining during this period in Tonga, the absence of midden evidence in early levels in Samoa becomes, for the first time, explicable. This tendency, or rather, the increasing dominance of horticulture, was transferred from Tonga to Samoa. The evidence of early inland (and therefore horticultural) sites in Samoa 103 would suggest - 313 that the initial settlement of Samoa was by people who had already adjusted to a horticultural economy, a pattern which persisted throughout Samoan prehistory.

There are many contributory forms of evidence, and a number of important reservations about the argument presented here, which must, in a paper already over-long, be left aside. The inevitable problems which will emerge from tracing Samoan prehistory from a homeland in Tonga are for future investigations. The plausibility, linguistic and archaeological, for claiming a direct Tonga to Samoa link during the latter half of the first millenium B.C. is not outweighed by minor conflicts in evidence. It is totally dependent, however, on the validity of the compressed chronology for Tongan ceramic history.

The interpretation of Polynesian origins proposed in this paper seems to me to be required by the accumulating excavation and linguistic evidence from western Polynesia. The solution to Polynesian origins is simple and, in all respects, close to that which has been anticipated for many years by archaeologists and linguists. Only the precise location and chronology have really been in doubt. The revised picture of the relationships between Fiji, Tonga and Samoa suggested by this paper is shown in Figure 5. The correspondence of this pattern to the ideal linguistic model (Figure 1) is extremely close.

In a journal which has initiated so much research and concern for the “Polynesian problem” the evidence reviewed in this paper suggests a simple, and, in many ways unexciting, conclusion to the quest: the Polynesians became Polynesians sometime near the middle of the first millenium B.C., after over 600 years of isolation in the remote archipelago of Tonga. The Polynesians, therefore, did not strictly come from anywhere: they became Polynesians and the location of their becoming was Tonga.

REFERENCES

• BARNARD, N. (ed.), in press. Early Chinese Art and its Possible Influence in the Pacific Basin. New York, Interarts Cultural Press.
• BELLWOOD, P., 1970. “Dispersal Centres in East Polynesia, with Special Reference to the Society and Marquesas Islands”, in Green and Kelly (eds.), 1970:93-104.
• BIRKS, L. and H., 1967a. “A Brief Report of Excavations at Sigatoka, Fiji.” New Zealand Archaeological Association Newsletter, 10 (1):16-25.
• —— 1967b. “Radiocarbon Dates for Sigatoka, Fiji.” New Zealand Archaeological Association Newsletter, 10 (2):69.
• —— 1968. “Adzes from Excavations at Sigatoka, Fiji.” Archaeology and Physical Anthropology in Oceania, 3 (2):105-115.

• CLARK, J. D., 1959. The Prehistory of Southern Africa. Harmondsworth, Penguin Books.

• DAVIDSON, Janet M., 1969a. “Settlement Patterns in Samoa before 1840.” Journal of the Polynesian Society, 78 (1):44-82.
• —— 1969b. “Excavation of a Coastal Midden Deposit, SU-LO-1”, in Green and Davidson (eds.), 1969:224-52.

- 314

• DICKINSON, W. R., 1969. “Temper Sands in Prehistoric Potsherds for Vailele and Falefa”, in Green and Davidson (eds.), 1969:271-3.
• DUFF, R., 1959. “Neolithic Adzes of Eastern Polynesia”, in Freeman and Geddes (eds.), 1959:121-47.
• DYEN, I., 1962. “The Lexicostatistical Classification of the Austronesian Languages.” Language, 38:38-46.

• FREEMAN, J. D. and W. R. GEDDES (eds.), 1959. Anthropology in the South Seas. Avery, New Plymouth.

• GIFFORD, E. W., 1951. “Archaeological Investigations in Fiji.” Anthropological Records. University of California Press, 13:3.
• —— 1955. “Six Fijian Radio Carbon Dates.” Journal of the Polynesian Society, 64:240.
• —— and Dick SHUTLER Jr, 1956. “Archaeological Excavations in New Caledonia.” Anthropological Records (University of California Press), 18:1.
• GOLSON, Jack, 1959. “L'Archéologie du Pacifique Sud: Resultats et Perspectives.” Journal de la Société des Océanistes, 15 (15):5-54.
• —— 1961. “Report on New Zealand, Western Polynesia, New Caledonia and Fiji.” Asian Perspectives, 5 (2):166-80.
• —— 1970. Lapita Pottery in the South Pacific. Paper presented at Wenner-Gren conference on Oceanic Prehistory, Sigatoka, Fiji, August 1970. Draft mimeograph.
• —— in press, a. “Both Sides of the Wallace Line: New Guinea, Australia, Island Melanesia and Asian Prehistory.” Chapter 17 in Barnard (ed.), in press.
• —— in press, b. “Lapita Ware and its Transformations”, in Green and Kelly (eds.), in press.
• GREEN, R. C., 1963. “A Suggested Revision of the Fijian Sequence.” Journal of the Polynesian Society, 72 (3):235-53.
• —— 1966. “Linguistic Sub-grouping within Polynesia: The Implications for Prehistoric Settlement.” Journal of the Polynesian Society, 75 (1):6-38.
• —— 1967a. “The Immediate Origins of the Polynesians,” in Highland et al. (eds.), 1967:215-40.
• —— 1967b. “Settlement Patterns: Four Case Studies from Polynesia”, in Solheim, W. G. (ed.), 1967:101-32.
• —— 1968. “West Polynesian Prehistory”, in Yawata and Sinoto (eds.) 1968:99-109.
• —— 1969a. “Archaeological Investigation of Western Samoa Prehistory”, in Green and Davidson (eds.), 1969:3-11.
• —— 1969b. “Excavations at Va-1, 1963-64”, in Green and Davidson (eds.), 1969:114-37.
• —— 1970. “Settlement Pattern Archaeology in Polynesia”, in Green and Kelly (eds.), 1970:13-32.
• —— 1971. “Evidence for the Development of the Early Polynesian Adze Kit.” New Zealand Archaeological Association Newsletter, 14 (1):12-43.
• —— in press. “Aspects of the Neolithic in Oceanic: Polynesia”, in Barnard (ed.), in press.
• —— and Janet M. DAVIDSON (eds.), 1969. Archaeology in Western Samoa, Vol. 1. Bulletin of the Auckland Institute and Museum, No. 6.
• —— and Marion KELLY (eds.), 1970. Studies in Oceanic Culture History, Vol. 1. Pacific Anthropological Records No. 11. Department of Anthropology, Bernice P. Bishop Museum, Honolulu.

- 315

• —— and Marion KELLY (in press). Studies in Oceanic Culture History, Vol. 2. Pacific Anthropological Records. Department of Anthropology Bernice P. Bishop Museum, Honolulu.
• GROUBE, L. M., 1964. Settlement Pattern in New Zealand Prehistory. Unpublished MA thesis, University of Auckland.
• —— MS. Additional C14 dates for Plain Lapita Pottery from Tonga.
• GROUBE, Rosemary, MS. A Review of Early Ethnographic Accounts of Tonga.

• HEDRICK, J. and Mary Elizabeth SHUTLER, 1969. “Preliminary Report on ‘Lapita Style’ Pottery from Malo Island, Northern New Hebrides.” Journal of the Polynesian Society, 78 (2):262-5.
• HIGHLAND, Genevieve A., Roland W. FORCE, Alan HOWARD, Marion KELLY, Yosihiko H. SINOTO (eds.), 1967. Polynesian Culture History. Essays in Honor of Kenneth P. Emory. Honolulu, Bishop Museum Press.

• IBBOTSON, P., 1960. “Green Sands in Fiji.” Geol. Survey Report, 84:1-5.

• KEY, C., MS. Modern and Prehistoric Pottery in the S.W. Pacific.

• MCKERN, W. C., 1929. Archaeology of Tonga. Honolulu, Bernice P. Bishop Museum Bulletin, 60.

• PALMER, J. B., 1966. “Lapita Style Potsherds from Fiji.” Journal of the Polynesian Society, 75 (3):373-7.
• —— 1967. Archaeological Sites of Wakaya Island. Records of the Fiji Museum, Vol. 1, No. 2.
• —— in press. “Pottery in the South Pacific.” Chapter 16 in Barnard (ed.), in press.
• PAWLEY, A., 1966. “Polynesian Languages: A Subgrouping Based on Shared Innovations in Morphology.” Journal of the Polynesian Society, 75 (1):39-64.
• PELLETT, M. and Alexander SPOEHR, 1961. “Marianas Archaeology.” Journal of the Polynesian Society, 70 (3):321-5.
• PETERS, K. M., 1967. In Davidson, Green, Buist, Peters “Additional Radiocarbon Dates for Western Polynesia.” Journal of the Polynesian Society, 76 (2):223-30.
• POLACH, H. A., J. J. STIPP, Jack GOLSON and J. F. LOVERING, 1967. “ANU Radiocarbon Date List I.” Radiocarbon, 9:15-27.
• —— 1968. “ANU Radiocarbon Date List II.” Radiocarbon, 10 (2):179-99.
• POULSEN, J., 1964. “Preliminary Report on Pottery Finds in Tonga.” Asian Perspectives, 8:184-95.
• —— 1967. A Contribution to the Prehistory of the Tongan Islands. Unpublished Ph.D. thesis. Canberra, Australian National University.
• —— 1968. “Archaeological Excavations on Tongatapu”, in Yawata and Sinoto (eds.), 1968:85-92.

• SAHLINS, Marshall D., 1958. Social Stratification in Polynesia. Seattle, University of Washington Press.
• SHAW, Elizabeth, 1967. A Reanalysis of Pottery from Navatu and Vuda, Fiji. Unpublished MA Thesis. Auckland, University of Auckland.
• SINOTO, Y. H., 1966. “A Tentative Prehistoric Cultural Sequence in the Northern Marquesan Islands.” Journal of the Polynesian Society, 75 (3):287-303.
• —— 1970. “An Archaeologically Based Assessment of the Marquesas Islands as a Dispersal Centre in East Polynesia”, in Green and Kelly (eds.), 1970:105-131.
• SMART, C. D., 1965. Preliminary Report on Archaeological Field Work in Kabara, Southern Lau Islands, Fiji. Suva, Fiji Museum.

- 316

• SMITH, S. Percy, 1898-9. “Hawaiki the Whence of the Maori, being an Introduction to Rarotongan History.” Journal of the Polynesian Society, Vols 7 and 8.
• SOLHEIM, Wilhelm G. II (ed.), 1967. Archaeology at the Eleventh Pacific Science Congress. Honolulu, Social Science Research Institute, University of Hawaii.
• SPECHT, J., 1968. “Preliminary Report of Excavations on Watom Island.” Journal of the Polynesian Society, 77 (2):117-34.
• SPOEHR, Alexander, 1957. Marianas Prehistory: Archaeological Survey and Excavations on Saipan, Tinian and Rota. Fieldiana: Anthropology, Vol. 48. Chicago, Natural History Museum.
• SUGGS, R. C., 1961. The Archaeology of Nuku Hiva, Marquesas Islands, French Polynesia. New York, Anthropological Papers of the American Museum of Natural History, Vol. 49, Part 1.

• YAWATA, I. and Y. H. SINOTO (eds.), 1968. Prehistoric Culture in Oceania. Honolulu, Bishop Museum Press.

1   Pawley 1966; Green 1966.

2   Golson 1959:28.

3   Suggs 1961:95-6.

4   Golson 1959:29.

5   Green 1969b:130-33.

6   Sinoto 1966:1970.

7   Peters 1967:227.

8   Green and Davidson 1969.

9   Sinoto 1966, 1970; Bellwood 1970:97-8.

10   Davidson 1969b:224-45, fig. 103a.

11   Davidson 1969b:224-5; Green 1969a:5-6.

12   Davidson 1969a:60-1.

13   Green 1968:105-7.

14   Green 1966:Table 9.

15   Gifford 1951.

16   McKern 1929:115.

17   Gifford and Shutler 1956:7.

18   See, for example, Dyen 1962, Green 1967 a:232-3.

19   Smith 1898-9.

20   Duff 1959:126.

21   Golson 1961:169-70.

22   Green 1963:245-6; Birks and Birks 1967a, Palmer 1966.

23   Specht 1968.

24   R. Shutler, personal communication.

25   Birks and Birks 1968.

26   This work, carried out while a Ph.D candidate at Australian National University, is briefly summarised in two published papers, Poulsen 1964, 1968. A detailed description of the work is available in his unpublished Ph.D thesis (Poulsen 1967).

27   Poulsen 1968:86-7, 89.

28   Poulsen 1964:195.

29   Polach, Stipp, Golson and Lovering 1967:24; 1968:195.

30   Birks and Birks 1967a:18-21.

31   Birks and Birks 1967a:20.

32   Birks and Birks 1967a:24.

33   Birks and Birks 1967b.

34   Golson in press, a, b.

35   ANU 262.

36   Hedrick and Shutler 1969.

37   Green 1963:250.

38   GaK 1227:2660 ± 90 B.P.; Birks and Birks 1967b.

39   The association of this date with pottery is unclear: Palmer (in press) claims it as the earliest date for paddle impressed ware in Fiji, whereas Golson (in press a, b) suggests that it may be associated with the Lapita level. Birks and Birks (1968) do not mention this date although it was available at the time. I prefer to accept (like Golson) the probability of Lapita associations for this sample particularly with the much later (2060 ± 100 B.P.) date from higher in the impressed ware level at the same site.

40   GaK 1228:2060 ± 100 B.P.; Birks and Birks 1967b.

41   Gifford 1955.

42   Ibid.

43   Green 1963.

44   Green 1963:242.

45   Green 1963:243; Golson in press, a.

46   Golson in press, b.

47   Shaw 1967.

48   Palmer 1967.

49   Smart 1965.

50   Golson 1959:29; Green and Davidson 1969:112.

51   Green and Davidson 1969:170-75.

52   Green and Davidson 1969:174-5.

53   Dickinson 1969.

54   Key MS.

55   Key ibid, writes: “. . . the temper (of Tongan pottery) is a wave-winnowed sand, rich in heavy minerals, which occasionally contains some shell and coral debris. It is probably of andesitic origin. This temper usually consists of about 40% euhedral pyroxene fragments, both orthopyroxene and hypersthene. The rest is made up of varying quantities of clear calcic plagioclase, fragments of volcanic glass, pumice and basaltic andesite and occasional bipyramidal quartz clearly of volcanic origin.
The only heavy mineral concentrate found during this reconnaissance trip was a very localised occurrence on the shore of the lagoon on Nomuka. There the heavy minerals clearly originated from the consolidated tuffs which at the highest part of the island are approximately 100-150 ft thick. However, this heavy mineral concentration contained at least 20% olivine, a mineral not yet detected in Tongan pottery, and a good deal of shell.”
He comments also: “Several occurrences of pyroxene sand are reported in the Fijian group. One of these on Vanua Levu, at Natewa Bay is remarkably pure and contains over 90% pyroxene with some magnetite and lighter mineral fragments but no olivine (Ibbotson, 1960).”

56   Green writes: “. . . if these early 1st to 2nd century A.D. Samoan and Fijian pottery horizons are related, it would appear that it is through some common ancestral tradition, rather than by the derivation of one tradition from another.” (Green and Davidson 1969:175).

57   Green 1970:15-17; Groube, R. MS.

58   Groube 1964, Chapters II and III; Green 1967b, 1970:24-5; Davidson 1969.

59   cf. Sahlins 1958.

60   The field work was financed by the National Science Foundation of the United States of America through the Bernice P. Bishop Museum, Honolulu, Hawaii.

61   Also financed by an NSF grant through the Bishop Museum.

62   Groube, R., MS.

63   Poulsen 1968:85.

64   Ibid, 89.

65   Polach, Stipp, Golson, Lovering 1967:24.

66   Golson 1970:5.

67   Poulsen 1964, fig. 24.

68   Poulsen 1968, fig. 2, 3.

69   The units from Natunuku are on display in the Fiji Museum, Suva. Spoehr (1957:147 and Fig. 77), and Pellett and Spoehr 1961:Fig. 2c illustrate examples from the Marianas Islands.

70   Poulsen 1968:89.

71   I should like to thank Mr J. Head, of the Radiocarbon Laboratory, ANU, for his cooperation in analysing these samples.

72   Golson in press, b.

73   The excavation report on the Mangaia excavation is nearing completion. Golson (personal communication) considers that the upper layer (from which the A.D. 185 date comes) is redeposited and that the sample submitted does not date the sherds in the layer.

74   Further details on these dates and their stratigraphic relationships will be published elsewhere: Groube n.d.

75   ANU 541.

76   See Pawley 1966:39; Green 1966, Table 9.

77   This sample was collected from a shell midden on Niue by L. Birks in 1959. I should like to thank L. Birks and J. Golson for permission to publish this date as well as the dates for the Mangaia mound quoted above.

78   Golson in press, b.

79   A detailed report on the excavations on Vuki's Mound (including a re-analysis of Poulsen's material) is being prepared. It will be published in the new series Terra Australis (Department of Prehistory, Australian National University).

80   I should like to thank Dr Poulsen for his cooperation in re-examining his excavated material. His excellent field notes enabled the complexities of his sites to be re-examined in considerable detail and the sherds re-allocated to the original contexts.

81   Groube, R., MS.; Poulsen 1968:89.

82   e.g. Golson 1959; Green 1963.

83   Green 1963:243. In the sense used, Green's “Middle Period” is included in the incised ware tradition. Shaw (1967) established that the “Middle Period” hypothesised by Green had little substantiation from Gifford's excavated materials.

84   Examined at the Department of Prehistory, ANU.

85   Key, MS.

86   See f.n. 55.

87   This observation is supported by Poulsen's analysis of the shell middens from his excavated sites (personal communication).

88   The author visited the site, under the guidance of J. B. Palmer and E. Shaw in 1969.

89   Gifford 1951:203, Tables 1, 2.

90   Green 1963:237.

91   Birks and Birks 1967a:18.

92   L. Birks (personal communication).

93   Detailed support for this statement will be published in the detailed report of Vuki's Mound excavations.

94   Two pig bones have been tentatively identified at Yanuca rock shelter (Higham, personal communication), but these come from the top level associated with paddle-impressed ware: L. Birks, personal communication.

95   cf. Gifford 1951, Table 10, 11 and Tables 20, 25.

96   Golson in press, b.

97   Clark 1959:16-17.

98   See, for example, Green 1967a:234-35; Green in press.

99   Golson in press, b.

100   Green 1971.

101   Golson in press, a. Map 1.

102   Golson 1962:175; Davidson 1969:224-5.

103   Davidson 1969:60-61.

http://www.jps.auckland.ac.nz/document/Volume_80_1971/Volume_80%2C_No._3/Tonga%2C_Lapita_pottery%2C_and_Polynesian_origins%2C_by_L._M._Groube%2C_p_278_-_316/p1?page=0&action=searchresult&target=