Volume 75 1966 > Volume 75, No. 1 > Linguistic subgrouping within Polynesia: the implications for prehistoric settlement, by Roger Green, p 6 - 38
LINGUISTIC SUBGROUPING WITHIN POLYNESIA: THE IMPLICATIONS FOR PREHISTORIC SETTLEMENT
Relationships among various languages within the Polynesian subgroup have not attracted the same attention as the relation of the subgroup itself to other subgroups within the Malayo-Polynesian or Austronesian family. The concern linguistically (as well as culturally) has more often been with the origin of the Proto-Polynesian language and its speakers than with identifying the precise sequential relationship of each Polynesian language to the others by techniques of subgrouping. Even where linguistic relationships were thought to have been demonstrated their implications, when viewed in the light of other cultural or historical considerations, have often been open to misinterpretation. Thus there is reason to examine closely the more recent attempts at linguistic subgrouping in Polynesia and to outline their implications for the prehistory of that area.
This paper, stemming originally from a seminar on Polynesian cultural history at the University of Hawaii, represents an assessment of the linguistic information in the light of present archaeological knowledge. Set in similar context to two previous efforts in this direction by Elbert 1 and Emory, 2 it has benefited greatly from their work and the advice and unpublished materials which Emory, Pawley, and others have made available. 3- 7
Following a short survey of previous endeavours, I deal first with the implications for prehistory of recent studies of subgrouping in the West Polynesia, 4 the linguistic aspects of which are dealt with in detail in an accompanying paper by Pawley. A more detailed linguistic analysis of subgrouping in East Polynesia follows, from which various implications for the prehistory of that area are also drawn. In West Polynesia the main point is that while linguistic differentiation has proceeded further and the basic subgroups exhibit greater time depth, cultural differences of a similar order have been obscured by the continuous nature of more recent cultural contacts, and are only coming to light archaeologically. In East Polynesia the main points are two: (1) although there is only one main linguistic subgroup, many of its members in each island group had developed a cultural distinctiveness through a greater degree of isolation and the successive effects of the “founder” principle, 5 and (2) the sequence in which the present dialects and languages that form this main subgroup split off to form separate languages and dialects may be specified and supported to a greater degree than previously and with important implications for the cultural history of the area.
In 1911 and 1912 Churchill published historical essays on the subject of Polynesian languages, essays which were based on comparisons of words. 6 His analyses were in terms of separate bodies of lexical items within each language thought to represent separate migrations rather than in the construction of genetic relationships between the languages themselves. In his 1911 work he was concerned to document the infusion of Polynesian lexical items in Melanesian languages, particularly those of the New Hebrides. These items, he thought, resulted from a primary Polynesian migration from Indonesia identified as Proto-Samoan, which passed along this route on its way to Nuclear Polynesia (an area in which Samoa is the nucleus and Niue, Tonga, Fiji, Rotuma, Uvea and Futuna describe the perimeter). 7
In his 1912 work Churchill was more concerned with linguistic connections reflected by the languages of East Polynesia (Easter, Tuamotu, Mangareva, Marquesas, Hawaii, and Tahiti) and in particular the position of Easter Island among them. Again explanations were in terms of a series of migrations: Proto-Samoan, Tongafiti, and combinations thereof, although his work on shared lexical items and total vocabulary laid a basis for Emory's 1946 study. 8 Churchill's identifications of Nuclear Polynesia (as defined above) and Tahiti as main centres from which the other languages or linguistic influences derive, are reflected in many later interpretations of the linguistic data. His work may be criticised not only - 8 on the grounds noted by Emory 9 and Elbert 10 of failing to properly reconstruct and identify cognates, but also on the grounds of failing to conceive of relationships between languages within a genetic framework.
A long interval of seeming disinterest in the problems of historical linguistics within Polynesia followed, although Polynesia's relationship to other languages in the Pacific continued to be hotly disputed as part and parcel of the problem of Polynesian origins.
However, in the two decades which have followed the return in 1946 by Emory to the subject of subgrouping within Polynesian, some common notions and not a few divergent points of view have been expressed. Emory's unpublished dissertation in 1946 followed the lead of Churchill and concerned itself mainly with the historical relationships among Eastern Polynesian languages, as did his Science Congress paper in 1961. 11 Elbert in 1953, 12 Marshall in 1956, 13 and most recently Dyen 14 as part of a larger study have also dealt with a wide range of the languages of East and West Polynesia and at least some of those among the Outliers, proposing various subgroupings for them.
Principal among the common notions in these works is that which conceives of Eastern and Western Polynesian as major subgroups within Polynesian whose member languages coincide with those societies earlier identified by Burrows as belonging to the geographically and culturally defined areas of West and East Polynesia. In contrast to this position is the one implicit in Elbert (despite his support of the above view), 15 and considerably refined by Pawley in an accompanying paper, that seeks to demonstrate the existence of a Tongic (TO) subgroup co-ordinate with a Nuclear Polynesian (NP) subgroup. Samoic (SM) and Eastern Polynesian (EP) are seen to be subgroups within Nuclear Polynesian.
A common difficulty in these works is dealing effectively with a geographic group of islands known as Outliers, due largely to the poor quality of published material available for them until very recently. Some like Churchill, Capell, and Marshall 16 have considered them to represent remnants of the original Polynesian movement eastward leading to the settlement of Polynesia itself. Although most seem to agree that they are “throwbacks” derived from Polynesia, there has been less agreement on their precise relationship to various subgroups or languages within Polynesia. Thus some, like Marshall, have treated them as a co-ordinate subgroup within Western Polynesian, 17 while others, such as Dyen and Elbert, have differentiated among them, placing some as co-ordinate with the Eastern and Western subgroups and others within one or more of the subgroups proposed for Western Polynesian. 18 The most recent and fullest treatment of these languages and certain aspects of their cultural - 9 inventory from a lexico-statistical point of view is that of Bayard. 19 Following the earlier work of Thilenius, he concludes that all Outliers do not derive from a single source, but are examples of east to west voyaging from numerous sources. He then attempts to pinpoint various primary and secondary settlements and their sources for each of the Outliers. His principal conclusion therefore supports previous workers in the fields of linguistics, physical and cultural anthropology, such as Ray, Elbert, Shapiro, Kubary, and Hogbin, that these societies derive from Polynesia after it was settled. 20 More importantly, those sources of settlement from Polynesia which he can identify all lie within West Polynesia and support his view that more than one source in that area is involved. In general, his evidence tends to support the position of Pawley and, to a degree, of Elbert, that the relationships of those Outlier languages with sufficient data are with NP and/or one or more subgroups or languages within it, except for the EP subgroup. On the other hand, except for Anuta, thought by Bayard to have been settled within the past 300 or 400 years from Tonga, none of the languages or societies in TO subgroup seem to be involved, while Futuna, now placed in the SM subgroup is thought to have played a central role in the settlement of many of the Outliers. 21
The third major notion is that of an EP subgroup which is divisible into at least two main subgroupings, Marquesic (MQ) and Tahitic (TA). In Churchill and Emory these are referred to as the Tahitian and Marquesan provinces, and the Tahitian province is considered to be that first settled, 22 while for Elbert the principal divisions are Easter and a proto-Marquesan-Tahitian subgroup from the latter of which proto-Marquesan and proto-Tahitian derives. 23 These positions contrast markedly with that of Marshall, who has Central and Peripheral sub-grouping in which Peripheral includes Hawaiian, Maori, the Chathams and Easter, and Central is divided in turn into East and West subgroups. 24
Of the various attempts to relate these subgroupings, particularly those in East Polynesia, to the history of settlement of the various Polynesian islands, the most recent by Emory 25 is the only one which attempts to incorporate archaeological results with the linguistic data to arrive at this sequence. As it is this subject with which I am also concerned, some additional comment is called for on Emory's methodology.
Like Churchill, Emory in 1946 found Tahitian retained the greatest number of West Polynesian lexical items in its total vocabulary and for this reason considered it the nuclear language and language area from which differentiation took place in East Polynesia. 26 In his 1963 paper he retains this concept and also noted that Tahitian retains more of the PEP vocabulary on the 100 word list, and also (except for one instance involving Maori) shows higher percentages of words shared between it - 10 and Samoan (SAM) and Tongan (TON) than any other languages of East Polynesia examined on the 100, 200 and total vocabulary lists. 27 This conservatism in Tahitian, he believes, marks the Society Island as the “cradle of the Proto-East-Polynesian”, but few linguists would agree that such a linguistic phenomenon is evidence either of an ancestral language from which others derive or of a geographical location in which it was spoken. In a postscript to this article Emory modified this position and considers the possibility of the Marquesas as the place of primary settlement in Eastern Polynesia, but this change stemmed from Suggs' and his own archaeological findings and not from the linguistic evidence. 28
Distrusting the applicability of usual methods of glottochronology for Polynesian languages, Emory abandoned these procedures for a related technique. He not only used a multiple cognate method of determining cognates, as did Elbert, but because of a shorter word list, further varied this method by counting halves for words which have closely-related but not really cognate meanings or form. 29 Understandably, the resulting retention rates are far higher than those generally recorded in such studies and make any of the standard formulations for estimating time depth inapplicable. In this sense it seems incorrect that Emory, when using data from Elbert's Table 2 should have applied to it an 82% retention rate, when it too was compiled using a multiple cognate method. For, as Elbert demonstrated, when these results were compared with the single cognate method in normal use, but carried out by him for only five of the twenty languages, it indicated his results in Table 2 to be approximately 6.5 per cent too high when estimating time depths by the standard retention rate used by Emory.
In fact, the only time depths resulting from a standard application of the method of glottochronology are those of Elbert derived from comparisons on but four languages: Samoan, Marquesan, Hawaiian, and Kapingamarangi, as Tahitian had to be excluded. 30 In this context it might be well to also note that all lexicostatistical studies in Polynesia to date have had to depart to various degrees from the standard basic word lists used in other areas of the world in order to cope with particular problems of comparison in Polynesian languages, and for this reason may not give entirely comparable results to studies in other areas where the lists and their presumed rates of change have been fixed.
To provide a new basis for the differences inherent in his modification of the method, Emory turned to archaeology as one means for estimating time depth between two languages and thus for independently deriving a rate of change suited to the East Polynesian situation. On this score his innovation seems admirable. Unfortunately, his choice is not.
Having used linguistic evidence to decide that the first settlement of East Polynesia was in Tahiti, he assumed that the date of first settlement in the Marquesas of 100 B.C. (as estimated by Suggs from radiocarbon dates) actually represents the date of initial separation between the - 11 Marquesas and Tahiti. 31 Yet Suggs 32 specifically had interpreted this date and the archaeological evidence associated with it as reflecting the first settlement of the Marquesas from West Polynesia and not Tahiti. Thus there was and still is no archaeological evidence for Emory's basic initial assumption. In the postscript to this article, having noted that archaeological evidence has forced him to give consideration to the Marquesas as an earlier and more important center, he still continues to assume that the 100 B.C. date represents the split between Tahitian (TAH) and Marquesan (MQA) 33 and, further, that this implies that 700 B.C. was the very minimum for the first settlement of the Marquesas. 34 But it must be remembered that the archaeological interpretation of 100 B.C. was as an estimated date for first settlement of the Marquesas from West Polynesia, and similar materials of the same age have not as yet been found in the Society Islands, so that again archaeology furnishes support for neither of Emory's assumptions.
One result of this choice is useful, however. It leads to calculations on the above basis of glottochronological dates for splits between most of the main languages or dialects of East Polynesia and between them and TON and SAM in West Polynesia. 35 The results in many cases are far too old, as Emory noted, suggesting dates from between 2000 and 4000 B.C. for splits between languages of East and West Polynesia. They serve strongly to suggest that a glottochronological date for the split between Tahitian and Marquesan must be much later in time, and that the wrong retention rate is being applied.
In summary, this brief survey sets the scene for the discussion that follows and emphasizes how uneven in methodology, aims, and results has been the course of investigations in this field. As a result, the culturally established division between East and West Polynesia, when seemingly supported by linguistic evidence, seldom led prehistorians to worry about possible implications of greater linguistic differentiation in West Polynesia. Also overlooked was the obvious possibility that EP as a sub-subgroup was more closely related to one of the West Polynesian subgroups than to the other.
Until recently, much of the subgrouping that has been done has had lexicostatistics or glottochronology as its principal technique. Even where this has not been entirely true, other culture historical considerations have tended to override results of the comparative approach. Also studies employing lexicostatistics have varied widely in their methodology, with the result that the standard methods have never really been fully tested. Nor have the lexicostatistical results generally been successful in completely solving the problem of order of differentiation among languages or dialects of a subgroup, even in the relatively well-studied EP subgroup. Finally, historical interpretation of the linguistic evidence has not always been founded on sound linguistic or archaeological assumptions.- 12
The review above indicated that subgrouping in West Polynesia had attracted less attention than in East Polynesia. But the subgroupings of Pawley's recent comparative study, when viewed against the previous lexicostatistical results of Elbert, yields a reasonably clear set of relationships and one with important implications for Polynesian prehistory.
Linguistic differentiation among the languages of West Polynesia (especially when the Outliers are included) is, as would be expected of the area assumed to have beeen settled first, of more importance to the overall Polynesian picture than has previously been assumed, as the initial subgroupings are fundamental to an understanding of the splitting up of the entire Polynesian subgroup. This is even more true if one accepts the current position that Outliers belong with the subgroups of West Polynesian and are primarily related to the subgroups within PSM or to PNP, a point discussed above. In short, we must divest ourselves of the notion that Western Polynesian constitutes a valid subgroup, and the parallel conception that EP, which is a subgroup, holds the same rank as the principal subgroups of the West Polynesian area.
An inspection of Table 1 demonstrates that the lexicostatistical comparisons carried out by Elbert and Emory lend support to Elbert's and Pawley's position in West Polynesia that the separation of PTO from PNP is of a greater time depth than the separation of PSM and PEP from PNP.
Comparison of Samoan (SAM) and Tongan (TON) with various Eastern Polynesian (EP) languages, showing cognate per centages obtained by multiple cognate method.
The table illustrates that TON is more distant from the languages of EP than is SAM, and that MQA preceded TAH in separating from EP.
On present evidence PTO belongs in the period of earliest differentiation in Polynesian. It is characterized by those morphological and phonological features noted by Pawley. 36- 13
Elbert's lexicostatistical evidence for Niuean (NIU) suggests that this language separated between 400 A.D. and 800 A.D. (64% with TON, 72% with EUV) from Tongan-East Uvean, while on the same evidence, an approximate date for East Uvean (EUV) separating from TON is about 1200 A.D. (86% with TON). 37 While this supports Dyen's assessment of EUV as a dialect of Tongan, Pawley makes it clear that sufficient borrowing has occurred in EUV from either TON or SAM to cast doubt on the lexicostatistical results for EUV and leaves open the question of its genetic position with respect to TO and SM. 38 Obviously, its prehistoric sequence will therefore be of extreme interest to those attempting reconstructions of the Tongan and Samoan sequences and attempting to assess their relationship.
Linguistically, an initial differentiation of PPN may be explained in several ways: (1) as a result of a movement of proto-Polynesian speaking people in the area of the TO subgroup to one or more of the islands of the SM subgroup, where through innovation they came to speak proto-Nuclear Polynesian while those in the Tongan area developed PTO, (2) as a result of a group of people leaving a proto-Polynesian speaking community in the area of the SM subgroup for Tonga, where they came to speak PTO, while those who remained behind developed PNP, or (3) as two migrations from a third area (e.g. Fiji) where a related language has now disappeared. The linguistic evidence alone is not conclusive for either of the first two alternatives, and the third is least economical. Given recent archaeological and geographical evidence as well, one is presently inclined to favour the settlement of Tonga first by proto-Polynesians. 39
The PSM subgroup (with the inclusion of Futuna by Pawley) derives from NP. Within the SM subgroup the order in which East Futunan (EFU), Tikopian (TIK), and Ellice (ECE), broke away is not clear. First could have been EFU (74% with SAM) and the only member of the group retaining *?, although this feature may represent borrowing from EUV, and the lexicostatistical evidence does not yield a significant difference. 40 On Elbert's lexicostatistical evidence both TIK and ECE differentiated about the same time (78% shared with SAM), which on his estimates would be some time after 1000 A.D. 41 Bayard would place the sequence as Futunan, Ellice, and Tikopia. 42
The available linguistic evidence from West Polynesia therefore furnishes a number of reasonable and testable hypotheses for the settlement of the area which archaeological, traditional, genealogical, or other - 14 data can be used to support, modify or refute. It also furnishes us with other important evidence. This has to do with the fact that the major cultural division of Polynesia between West and East fails to conform with the basic linguistic division between the NP and TO subgroups and, instead, correlates with the lesser linguistic subdivision between SM and EP. The reason is not far to seek and is correlated with the fact that shared lexical relationships in basic vocabularies between Samoan and Tongan are generally higher than would be expected from the subgrouping evidence. 43
West Polynesia is an area where even on the most critical assessment of Polynesian navigation skills and voyaging patterns two-way contact has always been possible. 44 Traditional evidence for this contact is well known, while the distribution of cultural items both ethnographically and archaeologically attest such contact. That borrowing of lexical items between these two main groups has also occurred and has influenced the percentage of retained cognates even in the basic vocabulary resulting in too high scores is understandable.
By the single cognate method of glottochronology Elbert 45 places the linguistic split between Samoan and Marquesan at A.D. 100, so the expectations are that the split into NP and TO occurred before this, but well within the first millenium B.C., and probably in the later half of it. This is reasonably well in line with present archaeological expectations.
The overall isolation of the East Polynesian area from West Polynesia and the greater isolation of each group as it split off from the other in East Polynesia, of course, explains why the main cultural division in Polynesia correlates with the split of NP into SM and EP. The earlier cultural division between Samoa and Tonga has been obscured by later diffusion as would be expected but this has not happened between East and West Polynesia. Archaeologically, however, the evidence is encouraging. Whereas in East Polynesia as one goes back archaeologically to the earliest portions of each sequence there is a convergence of cultural forms and an increasing number of ties between these cultures, this has not been true to the same extent between Samoa and Tonga in West Polynesia. Here, as Davidson emphasizes, there are major differences for the latter periods in settlement pattern, mounds and their functions, and fortification types, and 2000 years ago this was true to a surprising degree in pottery and to a lesser extent in adze types and other items of material culture. 46 Cultural convergence attested archaeologically occurs in East Polynesia before it does in West Polynesia.
The closer lexicostatistical relationship of SAM than of TON to the languages of East Polynesia is evident in Table 1. On Elbert's glottochronological evidence the separation of Samoan and Marquesan was placed at A.D. 100, an estimate not far out of line with the archaeological - 15 estimate by Suggs for the settlement of the Marquesas from West Polynesia at 150 B.C. 47 Actually, on more recent archaeological evidence uncovered by Sinoto this archaeological estimate may prove too early, or at least in need of reassessment. 48
The great difficulty in East Polynesia, of course, has been the belief held for many years by Buck, Emory, Duff and a host of earlier scholars that Tahiti was the “Hawaiki” or primary dispersal point of all the languages and cultures of East Polynesia. As Sharp has indicated, Tahiti, with the nearer Tuamotus constitutes the one well-authenticated area of two-way voyaging in East Polynesia. 49 As such, it offered greater opportunity for the diffusion of any new cultural item throughout the area which either was received from any direction outside, or which was invented at any place within the chain. In short, Tahiti appeared, in the culture area-distribution approach then current for interpreting prehistory, to be the primary dispersal point because of its central geographical situation, the nature of voyaging pattern that obtained within the region, and the cultural diffusion that resulted from the more frequent contact possible in comparison with the greater isolation that obtained throughout much of the rest of East Polynesia. These reasons serve equally well as those of historical primacy to explain its traditional position in any distribution approach as an area of first settlement. Given the increasing archaeological evidence which Emory, Sinoto, and Suggs have assembled for the primacy of first settlement in East Polynesia in the Marquesas, Emory himself has come to favour the Marquesas. 50 Therefore, it is now the Marquesas on archaeological evidence which is seen by these scholars as a primary dispersal point in East Polynesia. If it proves correct that the Marquesas were settled before Tahiti, it will show that in finding new islands the Polynesians did not necessarily move with as much design as some assume in settling East Polynesia and Sharp's prediction concerning the primacy of the Marquesas as the initial settlement in East Polynesia will receive support. 51
Phonological justification for an EP subgroup is found in the consistent reflection of *s as h, probably influenced by the earlier loss of *h, and the merger in some or all positions in these languages of the reflexes of *s and *f as h. 52 Lexicostatistical studies by Elbert and Dyen have also served to identify EP as a subgroup distinct from those in West Polynesia. 53 It is now demonstrated by Pawley on the sounder linguistic grounds of morphological innovations in his accompanying paper.
The linguistic problem on which more attention has centred, is that of specifying the order in which the EP languages sub-divided into subgroups, languages and dialects. It is my belief that, on the basis of sound shifts and shared innovations, a portion of this sequence can be outlined and supported more precisely than before. However, closer examination - 16 of the linguistic situation in the Marquesas is called for first, as the usual treatment of Marquesan as a single language or dialect appears to be a basic difficulty in the solution of this problem.
Marquesan (MQA) includes a number of dialects, and care must be exercised in treating it as a single entity. It should be remembered that many of the “languages” of the Eastern Polynesian subgroup were, or still are mutually intelligible, and on this basis two of the main dialects in the Marquesas may be treated as “languages” that are not far apart but which occupy distinct island clusters. In this essay the main dialect on Nukuhiva 54 will be called NW Marquesan (NW-MQA) and that on Hiva Oa and Fatuhiva SE Marquesan (SE-MQA).
When 100 word lists for SE-MQA and NW-MQA are compared, using a 100-word list as in Emory, they retain 85% cognates. 55 These results are lower than many Emory obtained for other EP language/dialects normally treated as distinct, and equal to many other results he gives for various EP languages. 56 This is sufficient to indicate that linguistic differentiation in the Marquesan is equal to that between certain other “languages” of EP. This becomes important as it is intended to demonstrate below that most other subgroups or “languages” of EP had differentiated before this split into two main dialects occurred in the Marquesas, except for Hawaiian, which came after it. Using Elbert's glottochronological evidence then, differentiation in PEP of following subgroups or languages took place after the 2nd century A.D.: Easter (EAS), a Tahitic (TA) subgroup, a Marquesic (MQ) subgroup, and within the MQ subgroup, proto-Marquesan and Mangarevan (MVA).
Other evidence for considering Marquesan as two important dialects is an unusual sound shift not often encountered in Polynesian in which *ng becomes k in the dialect of the NW and n in the dialect of the SE. This shift is well attested in the dictionary of Dordillon 57 while confirmation of a former situation is attested in the Taipivai, Ho'oumi and Hatiheu valleys on Nukuhiva in which the original *ng is retained. 58 Such dialect and sub-dialect differentiation is probably explicable in the light of the geographical nature of the Marquesas and the difficulties of inter-island and inter-valley communication, plus the length of time the Marquesas have been settled. In line with the linguistic evidence above, the Marquesas is also one of the few areas in which two different but basic sets of names for the nights of the moon are associated with these two different dialect groups. 59- 17
Elbert outlined the following order in the separation of languages of the Eastern Polynesian subgroup: Easter, proto-Marquesan-Tahitian, a division of the latter into proto-Marquesan from which Marquesan and Mangareva separated, and proto-Tahitian from which Tahitian, Tuamotuan, Rarotongan, Maori, and Hawaiian separated. 60 Elbert did not use his lexicostatistical results to specify further the order for the separation of the languages within the last two subgroups. Emory, however, using both archaeological and linguistic evidence, achieved this sequence: Marquesan, Tahitian, Easter, Hawaiian, Rarotongan, Mangarevan, and Maori. 61 These results are much at variance with his interpretation of the linguistic data alone and with the interpretation presented here. On linguistic evidence alone my position is similar to that of Elbert, except for my placing Hawaiian with SE-MQA in a subgroup in the Marquesic (MQ) subgroup and considering Marquesan as two distinct “languages”. The order which I seek to demonstrate for differentiation of languages and dialects from PEP follows: First is the establishment of a Central (CP) subgroup as the result of Easter Island splitting off as a language. Next is the derivation from proto-Central (PCP) of proto-Tahitic (PTA) and proto-Marquesic (PMQ). From proto-Marquesic there appear Mangarevan (MVA) and proto-Marquesan, and from the latter NW and SE-MQA. Rather more briefly I shall then discuss evidence, which is inconclusive at present, for the separation of Rarotongan (RAR), Maori (MAO), and Tuamotuan (TUA) from the PTA subgroup beginning at a time shortly after the differentiation of MVA from the PMQ subgroup.
Examination of Table 1 reveals that most lexicostatistical results have favoured EAS and MQA over TAH and other languages of EP as among the first to differentiate. Elbert's results for EAS were sufficient to convince him that it preceded MQA, although with poor dictionary resources for EAS until recently, the very low percentages of retention obtained in comparison with other EP languages have always been somewhat suspect. 62 More recent phonological evidence for the retention in EAS of the PPN *?, however, lends weight to Elbert's position. 63 In the terminology employed here PEP first splits into proto-Easter Island (PEA) and proto-Central (PCP).
Two lexical innovations and one major phonological innovation of *hah . . . . to vah . . . . assembled in Table 2 furnish evidence for a proto-Central subgroup. In the two lexical innovations EAS appears to retain an older and probably PPN form, while it inherits the remaining PPN forms in the table according to regular sound shifts where the other EP languages exhibit a common phonological innovation.- 18
Shared innovations in PCP
In sum substantial phonological evidence, some lexical evidence, and most lexicostatistical results are agreed in placing the differentiation from EP of the language ancestral to EAS as the first to occur.
The next subgroupings within CP are usually agreed to be Marquesic (MQ) and Tahitic (TA), although the linguistic basis on which they have been established is not particularly strong, especially for the TA subgroup. As a result, I will deal at length first with the MQ subgroup and possible sub-subgroupings within it before proceeding on to the TA subgroup. The main difference here with the previous lexicostatistical results obtained by Elbert is that HAW is placed in the MQ sub-group rather than in TA.
In working with the MQ subgroup I have used a body of evidence developed by Emory in his various lists for words, some with unique forms and similar meanings and some only showing unique semantic innovations, shared between each pair of EP languages. 64 While it is not possible to reproduce his lists completely here, and in any case they are in need of reworking with the many new dictionaries now available, his results, summarized in Table 3, are a fair indication of some rather close relationships between the various languages of the MQ subgroup.- 19
Number of words with same or closely related meanings uniquely shared between various pairs of Polynesian languages (data as in Emory 1946)
Similar evidence is also found in Emory's additional lists of substantial numbers of words shared between various groupings of Mangarevan, Marquesan, Hawaiian and Easter, or the languages of his Marquesan province. 65 For instance, Emory also lists a total of 27 items shared only between SE and NW-MQA, MVA and HAW. While I have found that these listings can be reduced by a number of items, they have allowed me to identify a number of innovations to support both a MQ subgroup and to suggest various subgroupings within it. Other items presented here I have turned up in checking those given by Emory. In table 4 a group of phonological, lexical, and semantic innovations, and two shared reductions, are presented as evidence for a MQ subgroup.
Shared innovations in PMQ, sometimes noted in Moriori (MOR)
The two examples of shared reductions might be explained by convergence were it not for the fact that they are fairly wide-spread and occur only in the one of the two proposed subgroupings. The phonological innovation of r before the base for ten seems excellent evidence as the PTA subgroup shows a further reduction of the PPN form to *ngahuru. The lexical innovations for ‘full’ and ‘tears’ appear fairly sound, while the word for ‘bite’ and the related word for ‘chew’ in EP provide an interesting set of semantic innovations. Here the PPN *ngau, which appears to mean to ‘chew cane for the juice only’ takes on a new meaning in PEP of ‘bite’ and retains the sense of ‘chew’ as well (but without the restriction on juice from cane). The evidence for chew is the following: EAS, MAO, MOR ngau, HAW nau, TAH ?au, RAR ngangau. Other languages of EP retained an older form, mama, for chew. In PMQ, however, another root, probably PEP *(nga)ngahu—ngangahu, MAO to hunt with dogs, TUA to cut, was adopted for ‘bite’. Nga?u meaning bite also appears in Rapa where ? corresponds with h in other languages. The implications of MOR sharing in two of these innovations are explored below.
Subgrouping within MQ is a more difficult problem. But some phonological and lexical evidence suggests that a language ancestral to MVA was first to differentiate, and this was followed by the division of MQA into two dialects, from one of which, SE-MQA, HAW then derives. Before proceeding, however, a brief review of previous lexicostatistical results is needed. On the 100 word list used by Emory, MVA shows no significant difference in rates of basic vocabulary retention with TAH, MQA, and HAW (84-85%), while a similar result, 85%, was achieved using the same list, but by the single cognate method, when the two MQA dialects were compared. Only EAS, with 74.5%, using Emory's system of multiple cognates, plus half counts, appears to differ significantly. On Elbert's 202 word list, however, MVA exhibits a 68% retention of basic vocabulary with TAH, 69% with HAW and 73% with MQA, while MQA shares 67% with TAH and 70% with HAW. Given the postulate - 21 developed later that HAW-MQA and HAW-MVA figures are too low and HAW-TAH too high (the highest in Table 5) because of borrowing by HAW through later contact with TAH, one may see possible significance at this level of sub-grouping in the 5 to 6% difference when results for MQA and MVA on the one hand are compared with those for TAH and MQA or MVA on the other hand. Dyen's results, using a sub-adequate list from MVA, place MVA in EP where it shares its highest percentages with MQA, 64.6%, and next highest with HAW, 64.4%. These results would thus suggest the MQ subgrouping proposed here. 66 The very high figures then between HAW-TAH which led Elbert to place HAW in the TA subgroup may instead reflect later borrowing and the additional evidence cited in the paragraphs that follow also serves as further support for placement of HAW in the MQ subgroup.
Retention rate percentages among various EP languages obtained by multiple cognate methods of comparison.
(Data from Elbert 1953 and Emory 1963)
Phonological evidence for placement of HAW in the MQ subgroup as the last dialect to differentiate is based on a possible common origin for the soundshifts which it exhibits. The various soundshifts also provide some evidence for the order in which the MQ languages differentiated. The proto-forms and the shifts involved are outlined in Table 6.
Reflexes of PPN *ng, *k, *f, *r, *t, *n in the MQ subgroup
The retention of ng as a reflex for *ng by EAS, the TA subgroup, and MVA means that *ng was the proto-form for the MQ subgroup and that MVA split off from one or more of its members before the various - 22 sound shifts involving *ng took place. MVA shares uniquely some 58 lexical items with both MQA dialects, according to Emory, and in addition shares eleven important common plant and many bird names that are not widespread elsewhere in EP. 67 However, it has not been possible to demonstrate that MVA is more closely linked with one MQA dialect than the other, although Emory has suggested one name for certain nights of the moon and the meaning of marae indicate a SE-MQA connection. 68 On the other hand, if one takes the position of shared innovation rather than convergence as the explanation for some of these sound shifts, the most economical explanation for *ng>n and *k>? sound shifts in SE-MQA and HAW is common origin. In EP the sound shift of *ng>k appears in NW-MQA, the Ngai-Tahu dialect of New Zealand and a MOR dialect of the Chatham Islands. 69 It probably occurred in Tahitian as well. The *ng>n sound shift in EP occurs in SE-MQA, HAW, and the Ngati-Awa dialect of New Zealand. One or all of the New Zealand occurrences may be convergence or, less likely, of common origin. However, the two in HAW and SE-MQA, when taken in conjunction with the subgrouing evidence above for PMQ and the additional evidence which follows, are more easily explained by common origin.
Examples of the *ng to k and n sound shifts (which could be multiplied many times over) are assembled in Table 7.
Examples illustrating *ng to k or n shift in SE and NW-MQA.
HAW and SE-MQA then may be placed in a subgroup exhibiting a common sound shift of *ng to n. As Emory earlier insisted, if this were not a matter of convergence, “the only place from which the Hawaiians could have imported the ng to n shift is the southern Marquesas”. 70 Thus if “the Hawaiians were influenced from the South Marquesas in changing ng to n and k to ?, it would be after any departure of colonists to Easter or Mangarevan where such changes do not appear.” 71 At that - 23 time he saw this as an indication of influence because be believed that Hawaii had first been settled from Tahiti.
As personal experience in speaking Hawaiian, Tahitian and Tuamotuan taught Emory, the change from*ng to n could not have come from Tahiti and Tuamotuan because they never underwent it. 72 In this connection he pointed out that most of the sound shifts of Eastern Polynesian could not be reversed without detection, because once completed, all further shifts in the new merged phonemes would occur as one with language users being unable to choose consistently and correctly which former sounds occupied which positions without resorting to comparative linguistics.
The fact that HAW and SE-MQA both exhibit a shift of *k to ? may be taken as further evidence of a subgrouping relationship. Here, however, the SE-MQA evidence is not as clear as it might be. The sound shift appears to have been fairly common in SE-MQA for Dordillon says “k, Remplacé très souvent par l'accent(').” 73 Indeed, taking indisputable cases given in his dictionary, a large body of items with ? as a reflex of *k may be assembled, but there are still a fair proportion of items remaining in SE-MQA in which k is retained as the reflex of *k. Some of these doubtless result from the inconsistency with which Dordillon marked the glottal by an accent or failed to indicate the alternative form for SE-MQA where it existed. But even this does not eliminate all k noted in SE-MQA. This means it is possible that k and ? are allophones and would be in free variation, as the conditioning factors are not obvious, or at least not specifiable on existing evidence. A more likely situation, however, is one of fairly extensive borrowing of forms from neighbouring NW-MQA, which possessed k, in the same way that a few forms in Nukuhivan with ? as a reflex of *k are likely borrowings from SE-MQA.
On this view, the origin of the ? in HAW is completely different from that in TAH, where in comparatively recent times, and then only in the TAH and Austral members of that subgroup, ? has replaced both *ng and *k. Probably *ng shifted initially to k as in NW-MQA, and then ? replaced k.
Between MQA and HAW Emory lists some 80 items exhibiting uniquely shared forms and meanings or semantic innovations, which is more than the 72 he gives for HAW and TAH (Table 3). While this list may be substantially reduced by better dictionaries and more searching, it provides fair evidence for a close relationship between HAW and the two dialects of MQA. 74 Also, it is, as expected, a much larger group than the 58 items in the MQA-MVA list, or the 41 in the MVA-HAW list (Table 3), where on the present interpretation, the forms in the third language have now been lost, and would need be reconstructed to go with the smaller group noted above that he gives for all three languages. These figures suggest that HAW and both dialects of MQA form a subgroup after MVA had split off from PMQ, but before the development of two dialects in the Marquesas. From these lists I have selected a few - 24 items which seem significant as an indication of further subgrouping within PMQ.
The most important items involve SE-MQA and HAW, which Emory and others have not previously considered. Here one item is a metathesis of the word for charcoal found only in SE-MQA and HAW and set out in Table 7. As metathesis is common in PN, this could be a case of convergence, but in this context it is more economically interpreted as of common origin. More convincing, perhaps, is the word for ‘bow’ or ‘to shoot’, which in NW-MQA is pata, in MVA na?a, and in SE-MQA and HAW pana. 75 Most PN languages have a reflex of *fana for ‘bow’ or ‘to shoot’ and the MVA form, na?a, is one such metathesised from hana. As Emory notes this is a hint that the change to pana (and pata) took place after MVA split off from the MQ subgroup, but before HAW did so. 76
Innovations shared exclusively between HAW and SE-MQA are difficult to find. But a phonological innovation shared between both dialects of MQA and HAW of noni for Morinda citrifolia, where the usual form is nono or nonu, is, together with the word ‘bow’, additional evidence for Mangarevan, which possesses the common form, differentiating first.
A more difficult item to evaluate is MVA teiti and HAW keiki as a unique lexical innovation for child. Although a precisely cognate form does not now occur in MQA, NW-MQA poiti, toiki, SE-MQA, toiti, toi?i, for young child, and NW-MQA teiki and SE-MQA tei?i for name give to honour child of chief or ranking person, are later innovations based on a PMQ form, *teiti for child.
Maheama appears to be a unique base word used in various compounds for the 4th to the 6th nights of the moon in MVA and both MQA dialects. 77 In the HAW list for the nights of the moon, considered by Emory to be part MQA and part TAH in origin, 78 these forms are replaced by Tahitian ones. It stands as one of a number of such items that demonstrate a close relationship between both dialects of MQA and MVA, but do not assist much in subgrouping within MQ.
In short, subgrouping within MQ is difficult to establish, but some reasonable phonological and a little supporting lexical evidence exists that HAW together with SE-MQA form a subgroup within it. Also, good phonological evidence and lexical evidence exists that MVA either derives from one of the two MQA dialects before any of the sound shifts they now exhibit occurred, or that it derives from a language ancestral to these two dialects. On the basis of the lexical evidence presented above I tend to favour the latter interpretation.
With the detailed discussion above, a partial outline of language and dialect differentiation within the EP and MQ subgroups may now be attempted and very approximate dates assigned from Elbert's glottochronological results. 79 EP, a subgroup belonging to NP, probably split - 25 off by 100 A.D. or earlier according to glottochronological results achieved by the single cognate method of comparison between MQA and SAM. 80 In view of the fact that the earliest C.14 results to date from East Polynesia come from the Marquesas, this result does not appear unreasonable. 81
EP divides into two subgroups, CP and a language directly ancestral to EAS. Using Table 3 in Elbert for Easter, this event would be placed between 300 and 400 A.D. 82 Again, the date is reasonably well in line with the meagre existing archaeological estimates for settlement of that island. 83
CP splits into two subgroups, MQ which is reasonably well established, and TA which is not and which will be discussed in the next section. According to Elbert's scale, dates for PTA based on TAH would be after 530 A.D. and before 650 A.D. 84 Archaeological materials of this age have not yet been encountered in the Society Islands. 85
By this period the Marquesas on available archaeological evidence has definitely been settled, probably for some time, though there is no evidence that any other islands in the MQ subgroup have been. The MQ subgroup splits up, although the exact order of subgroups within it is not fully demonstrated. Likely to be first is MVA, which on Elbert's chart would separate between 800 and 850 A.D. 86 The present archaeological sequence for Mangareva extends back to approximately 1200 A.D. without encountering materials belonging to its settlement period. The 13th and 14th century items, however, clearly indicate a close affinity between Mangarevan culture of that period and the earlier Development Period materials of the Marquesan sequence from which they probably derive.
Next, from a language ancestral to the two MQA dialects, SE and NW-MQA differentiate. A similar distinction based on cultural differences between the NW and SE Marquesas has been summarized by Linton and Handy. 87 Unfortunately, the useful archaeological sequences are available only for the NW group of islands, so the point at which these differences first appear cannot yet be identified. 88
Finally, the derivation of HAW from a subgroup which also contains SE-MQA takes place, probably not long after dialect differentiation in the Marquesas. Archaeological evidence concerning Hawaii is summarized below.
At this point it would be useful to discuss the evidence for the possible location of the various proto-languages.- 26
The only conclusive case appears to be that for PMQ and the argument for it follows. Location of PMQ in Hawaii requires two long migrations to the Marquesas as well as one to Mangareva. One less long migration is involved if the location is the Marquesas or Mangareva. The earliest and longest archaeological sequences within East Polynesia are in the Marquesas, and in comparison the less complete Hawaiian and Mangarevan sequences are much shorter. Also the earliest materials presently known from Hawaii and Mangareva suggest their cultures derive, not from the materials of the Settlement Period of the Marquesas, but from the later Development Period. Finally, only the Marquesan Settlement Period materials suggest a West Polynesian source and “Melanesian” parallels. Thus it is highly likely that PMQ was spoken in the Marquesas and not in Mangareva or Hawaii.
For the present it is merely assumed that PTA was first spoken in Tahiti, although linguistically some other island groups in the sub-group are still possibilities and only New Zealand can be ruled out on archaeological evidence. Until this problem is resolved archaeologically and linguistically, it is hard to say where PEP was first spoken. On present archaeological evidence, Easter Island, Tahiti, and the Marquesas are the best candidates, with the Marquesas in the forefront and Easter Island as least likely. But linguistic evidence does not permit us to be more precise.
Differentiation in the PTA subgroup is not as well documented or studied as that of PMQ. That the four languages normally used (TAH, RAR, TUA, and MAO) may form a subgroup seems possible from previous lexicostatistical studies. Although a few shared innovations in lexical items with new meanings or new lexical forms can be identified (Table 8), the evidence is not very satisfactory and a number of problems exist which are in need of closer study.
Shared innovations in PTA
The few shared innovations identified for PTA are a mixed group. Two fairly straightforward items are (i) a further reduction in the root for ‘ten’ to ngahuru and (ii) the use of the form ki to indicate ‘full’. The use of koura, rather than PPN ?ura for the crayfish, is apparently a semantic innovation based on an existing use of the term in PCP for freshwater shrimp. The last two items for ‘spit’ and ‘mouth’ are presented because they provide conflicting evidence to the hypothesis developed above for the inclusion of HAW in the PMQ subgroup, so that some thesis is needed to account for them.
Given the evidence, both archaeological and linguistic, developed below for decisive later contact between TAH and HAW, the simplest hypothesis is borrowing. On the available linguistic evidence these two innovations are best explained as occurring in PTA because they do not occur in PEP, PCP or the PMQ languages except HAW. The alternatives are to consider their occurrence in HAW as evidence for subgrouping HAW with PTA and then to explain the far greater number of items for HAW discussed above under PMQ as either borrowing or convergence, or to consider these two items in HAW as borrowing, as is done here. As either position leaves us with evidence for borrowing in HAW, in the light of all the evidence the most economical choice appears to be borrowing from TAH.
Some relationships have been suggested among the various Austral, Cook, and New Zealand dialects, but they also raise various problems. Neither the Austral dialects nor those of the northern and southerns Cooks, except for RAR, have been used in studies of Polynesian outside of those made by Marshall and Dyen because of a lack of published material. In the Australs, except for Rapa, ? reflects *ng and *k as in TAH and this would seem to place them in the PTA subgroup. This is the position taken by Marshall in his grouping of the languages of the Australs. 89 It would also indicate, however, that these Austral dialects separated later than other languages in PTA if this sound shift had a common origin.
Support for Marshall's placement of Rapa with the languages I include in the PMQ subgroup is gained from examination of a very short vocabulary by Stokes. 90 In it taeti is listed for child, nga ?u for bite and rongo?uru for ten, along with the unique MVA word kami?a for canoe. As with MVA, Rapa retains *ng as ng and *f is reflected by ?, while as in SE-MQA some *k are reflected by ?.
Using lists supplied by Marshall, Dyen says that in the Cooks Mitiaro, Mauke and Atiu are practically the same dialect, and one closely related to RAR, with the result that, in his view, there are but two main Cook Island dialects, RAR with its sub-dialects and Penrhyn. 91 Similarly, - 28 until the recent publication of Stimson's Tuamotuan dictionary in collaboration with Marshall a study of the relationship of these dialects to PTA or the PMQ subgroups has been badly hampered. 92 Most of these dialects are in various ways related to TAH, 93 but others are not and suggest origins at some point from the MQ subgroup.
MAO in New Zealand also consists of a number of dialects and subdialects, while MOR of the Chatham Islands is even more distinctive. Some like the tribes of the west coast of the North Island, who claim descent from the Tokomaru, Aotea, and Kurahaupo canoes have a ? as the reflex of *f, as in RAR. Others reflect *f by f/h or wh/h. Other East Coast Maori dialects share an innovation with RAR, as is noted below. Still others, like the dialects of the Bay of Plenty tribes descended from the Mataatua canoe (Ngati-Awa, Ngai-te-Rangi, Tuhoe), the Ngai-Tahu tribe of the South Island, and one dialect of MOR of the Chathams, exhibit sound shifts as noted above only found in the MQ subgroup. Little attention has been given these dialect differences in New Zealand, however, the assumption usually being that all are subdialects of the one language, MAO.
In short, besides TAH, most of the comparative work in the TA subgroup had utilized restricted information on one dialect of TUA, the RAR subdialect itself, and dictionary MAO. As a consequence it is not yet possible to proceed very far in the study of differentiation among languages and dialects of this subgroup but only to note previously formulated relationships among them and to explore and speculate on some interesting data that have come to light in the course of this study with respect to possible contacts after initial settlement.
The lexicostatistical evidence presented by Emory and Elbert in contradictory as to whether MAO and RAR form a separate subgroup that derives from PTA and then differentiated into the subdialects of either language, or whether both languages split off separately from a single origin point at closely spaced intervals. 94 The same evidence is also in conflict as to which, if either, is first to differentiate. A time depth of between 900 and 1100 years ago seems to be indicated by the lexicostatistical results. 95 Archaeologically, by A.D. 900 - 1000 a form of early Leeward Society Island culture appears to have been established in New Zealand. 96 It is likely that an equally early time range and a similar culture will be established for Rarotonga. But neither the sequence of linguistic differentiation nor its verification archaeologically is now possible until the evidence for both is under better control.
One piece of comparative evidence which is suggestive has been noted by Pawley (personal communication). This is a shared innovation in East Coast MAO and RAR for the second person singular possessive marker—e.g. EC-MAO taahau, RAR ta?au, yours; and EC-MAO toohou, RAR to?ou, yours. These forms are taau, and toou or too in other MAO dialects, and in EP and Samoic generally. They imply either a separate - 29 contact and borrowing or perhaps even a separate origin for East Coast Maori from that of other groups in New Zealand.
In this regard some other rather interesting linguistic evidence appears in the language area assigned to Eastern Polynesian which has to do with the possibility of decisive contacts by people from one of the two main subgroups, Marquesic or Tahitic, in an area that appears to have been settled first from the other group. By decisive contact, I mean discovery and effective secondary settlement at an early enough time, usually before the population of the primary group was sufficiently great in relation to available land area, which itself is usually large, so that the second group has room to become established and their influence and their presence documented from the linguistic and archaeological records which now remain. In archaeological terms this is what some would call site unit intrusion or primary diffusion with continued distinctiveness of the intruding subculture before fusion into a single culture of the subcultures of the two founding groups. In Polynesia, where many contacts are likely to have been between closely-related groups, rather than with peoples of non-related cultures and languages, fusion would be an expected outcome of most such contact. The steps leading to it, however, are likely to be hard to detect, except in cases of decisive contact such as I have just outlined. Similar points have been made by Sharp in discussing the same general question. 97 However, Hawaii and New Zealand, in their relatively larger size of land area, their lateness of settlement, and increased possibilities for their multiple discovery and settlement from separate areas in an already almost completely settled East Polynesia, do seem to conform to the criteria listed above and more importantly do provide some evidence for such secondary contact during an early period of their settlement.
In recent years the Hawaiian case for two decisive settlements has been developed by Emory on the basis of linguistic and archaeological evidence. For instance, lexicostatistical scores from both Elbert's and Emory's studies provide results that are too high for TAH-HAW and consequently too low for HAW-MQA given an MQ subgroup in which it is assumed that HAW derives from SE-MQA (See Table 5). One explanation of this anomaly is that decisive later contact from Tahiti has affected the normal conditions of isolation that are assumed to have obtained in East Polynesia and provided the source not only for changes in Hawaiian material culture and many lexical items but also changes in some lexical items of the basic vocabulary. A similar result is noticeable in Emory's results for uniquely shared words (Table 3) where his TUA, which is closely related to TAH, shares only 7 words with HAW, while TAH shares some 72 words. A similar result is also revealed in Hawaiian names for the nights of the moon and the lunar calendar, which are part Marquesan and part Tahitian in origin. 98 Finally the two HAW cases noted in Table 8 are best interpreted as borrowing.
Supporting archaeological evidence has been obtained for this theory. The H-1 site, exhibiting few, or in its lowest levels no fishhooks related to the later Tahitian forms, but many hooks and octopus lures - 30 related to earlier Marqueson forms, reflects the earliest stage of materials recovered in Hawaii. 99 These materials do not represent the earliest or settlement period, but a slightly later stage when the types have been modified to conform to differences in the Hawaiian environment of available materials and the new types of fish being caught. In my view the best interpretation of the many C.14 dates for this site, some of them conflicting, is as follows: 100 Four samples provide five dates, two samples of which were the result of solid carbon analysis in its early stages and are too recent, one sample of which gives completely contradictory results on two different runs and is probably contaminated, and one sample of which yields a very early date. This last sample, however, is not stratigraphically associated with the principal culture bearing layer but under it. This leaves ten dates which consistently place the main culture bearing layers at the site between A.D. 1200 and 1550. Again, this would be a stage after settlement as the materials in the site themselves indicate.
The rapid change to the later type of fishhook on the main island of Hawaii appears to date after A.D. 1550 on the evidence of the upper layer at H-8 and the H-2 site, although low frequencies of such hooks in the lower levels of H-8 and upper levels of H-1 indicate that knowledge of this form begins earlier. 101 On the other hand, on the island of Kauai at the other end of the main chain, the change to the more recent Tahitian styles of hook-line attachment appears earlier, A.D. 1300 - 1400, the forms are more numerous from the beginning, and the changeover is completed more rapidly with few signs as yet of an earlier period. 102 Also the best evidence for the introduction of a Tahitian form of inland marae is documented from this end of the island chain on the remote island of Necker. 103 If, as Emory 104 suspects from a study of the genealogies, Tahitian chiefs arrived and established themselves between the 12th and 14th centuries A.D. and introduced these changes, it may be they were able to establish themselves first at the Kauai and outlying islands end of the Hawaiian chain and their influence spread from there, resulting eventually in fusion with dominance of the later and intruding Tahitian culture. This would compare to type A3 among possible kinds of archaeological contact situations. 105
The New Zealand case is less clear-cut but this is because most theories of settlement by secondary groups in New Zealand have concentrated on proving contact at a much later point in the sequence and usually utilized some form of a “fleet” concept. Adkin 106 is one of the major exceptions in also having two early migrations from Eastern Polynesia, but I find his use of evidence and his total scheme, which has four migrations, unconvincing. 107 Thus, most other recent theorists have argued that early settlement was from a single source in East Polynesia in - 31 their legitimate efforts to do away with earlier theories of initial settlement by Melanesians or non-Polynesians. However, in my opinion the assumption of a single source in East Polynesia for the effective settlement of New Zealand now needs closer examination and possible modification in the next few years, because some archaeological and linguistic evidence suggests it may be in error.
For instance, one cannot help but be impressed with the fact that there are two main groupings of tribes in New Zealand, those with several different Toi and those with Kupe, as origin ancestors, and these groups are in complementary distribution within New Zealand; neither origin ancestor ever appearing on a valid genealogy of the other group. 108 It should also be realized that existing archaeological evidence is not in conflict with a theory of two separate settlements, rather it contains some information supporting this view. In his writings and in personal conversations since, Parker 109 has been among the foremost exponents of two archaic, or early New Zealand-Eastern Polynesian subcultural traditions, which he called Archaic A and B. From another point of view, if the Maupiti burial site in the leeward Society Islands and the entire group's surface collections provide a reasonable inventory of the adzes, personal ornaments, and fishhooks to be expected at this time in the Society group and a culture from which New Zealand-Eastern Polynesian was derived, a number of items common in sites containing early archaic assemblages in New Zealand, particularly in the South Island, are lacking in Society Island collections and excavated assemblages but present in those from the Marquesas. Here I refer to some varieties of adzes, some items and features in fishing gear like the dorsal-ventral perforated lure shank and the harpoon, the New Zealand-Marquesan form of the tattooing chisel, the reel-shaped ornament, and some presumed early decorative motifs and art forms. It is also possible that the basic concepts of ridge pa fortification, although just possibly present in the Society Islands but certainly never developed there or in Rarotonga, come instead from the Marquesas where they are developed, called pa, and exhibit features like some fortifications in New Zealand with the same name. 110 This listing is only suggestive, but it will illustrate what I have in mind. To elucidate such a situation archaeologically in New Zealand, however, requires better control of early materials, particularly from the North Island, than we possess at present. The point is that it is still conceivable within the framework proposed by Golson 111 that two subcultures, New Zealand East Polynesian I and II were early established from different but historically and culturally closely related points in East Polynesia. One may even have introduced the sweet potato and fortifications, as some interpretations of traditions suggest.
My concern here is not to develop the archaeological side of this theme, but to call attention to possible linguistic evidence from some dialects in New Zealand and the Chathams that may point to contact with the languages of the MQ rather than the TA subgroup. First are the - 32 sound-shifts of *ng to n among the Ngati-Awa and related tribes in the Bay of Plenty, and the *ng to k shift among the Ngai-Tahu of the South Island and one dialect of MOR. 112 The first is more likely independent and a case of convergence with SE-MQA. The second may reflect a common origin between the MOR and Ngai-Tahu as there are other linguistic features common to both and it may just possibly derive ultimately from an initial innovation in NW-MQA.
In this context it is interesting that the reflex of *r and *l, which have fallen together in all NPN languages, may have been a laterally released l in the Ngai-Tahu dialect, and not a centrally released tongue-tip trill or flap (r) as in all Tahitic languages. 113 Some evidence also exists that until fairly recently l rather than r prevailed (at least in some environments) in Rapa, HAW and MQA. 114 The Ngai-Tahu also exhibit a vowel change of a to o for the causative prefix, using hoko instead of haka or faka and the hoko form appears normally in MOR. 115 It is a major form of the causative prefix in HAW, and seems to occur in a limited number of items in MQA 116 But it does not appear in this form elsewhere in Polynesia except for limited contexts in TON as foko. 117
While I am not able to provide lexical items from Ngai-Tahu, exhibiting MQ affinities, two for MOR were presented in Table 5 without comment. Those are the use of ngahu for bite and the reduction and loss of a in the word for go or walk. In his study of MOR Williams noted numerous other MQA affinities, but in checking I found few of them very convincing, though some are certainly suggestive. He ends his discussion, significantly by declaring that MOR is not correctly described as a subdialect of MAO, and has a right to as much consideration as an independent dialect as any other dialect or language in Polynesia. 118
Skinner 119 in reviewing Williams's linguistic evidence for MOR postulated its close relation to Ngai-Tahu on several grounds, including those noted above, and pointed out that Moriori material culture has its closest affinities with that of the area occupied by the Ngai-Tahu in the South Island. On the last point he may have been in error because this material culture may possibly be identified with tribes traditionally earlier than the Ngai-Tahu in the South Island who occupied the area previously, but this has not yet been demonstrated. On the other hand, his numerous identifications of affinities in the material culture of the Moriori with New Zealand Southern culture in one direction and the culture of the Marquesas, Easter and Mangareva in the other direction, which are in the MQ subgroup, furnishes considerable supporting cultural evidence for a linguistic source of this material from that subgroup area rather than the TA one.- 33
All of this discussion of the New Zealand case for decisive secondary settlement is clearly speculative. The evidence appears sufficient, however, to call for re-examination and consideration of a possible alternative to a single origin theory for settling groups, but without returning to the two stage or later Maori “fleet” concept so often employed in New Zealand culture history. Nor does it conflict with the position of those who maintain that early New Zealand culture is East Polynesian in origin. What I suggest is that decisive settlements early in New Zealand's prehistory, from both the Tahitian and Marquesan areas, could have given rise in New Zealand to two (or more) New Zealand East Polynesian subcultural traditions, and that together after long contact in favourable areas of the North Island of New Zealand, these laid the basis for the later development of Maori culture without any recourse to the concept of the fleet. In fact, two origins could be the basis for the various canoe traditions which have since been elaborated and are now wrongly interpreted by the last few generations of traditional scholars.
A last item of interest is a name for the 18th night of the moon, aniwa, a name which appears on only one of the lists from New Zealand and was supplied “by a very old ariki at Kaitoke in 1882”. 120 This list also exhibits affinities in its overall pattern with the Marquesan, Mangarevan and Hawaiian systems. In MQA aniva is the normal name for the 18th night but is otherwise unknown in PN lists, and in this respect Emory calls attention to the historical significance of its occurrence in New Zealand. 121 It suggests contact and borrowing, although without further identification of its New Zealand source this is difficult to assess.
In this essay concentration has been on the linguistic evidence and its possible historical interpretation. I have tried to use this evidence to postulate, in linguistic terms, theories capable of being tested by the now rapidly accumulating archaeological data. The results are summarized in Table 9. The archaeological data itself has not been explored in any detail. Yet nowhere have I found it to conflict significantly with the implications derived from a study of the linguistic evidence, and in several cases it furnishes clear supporting evidence. On the other hand, many of the hypotheses arrived at do not support previous and more traditional views of the broad outlines of Polynesian pre-history.
In particular, explanations of the major cultural differences between East and West Polynesia are seen not to be fundamental to the understanding of the sequence of settlement so much as they are a reflection of the extensive nature of contact that obtained both with Melanesia and within the West Polynesian area versus the more restricted nature of such contact in East Polynesia. The prehistory of West Polynesia, including the Outliers, constitute a domain that until recently has been unexplored. In fact when the Outliers are included in it, predominate movements are seen to be east to west with the winds. However, the first soundings, both linguistically and archaeologically, give results that lend- 34 - 35
new significance to questions of Polynesian origins, and to questions of historical and cultural divisions within Polynesia.
In East Polynesia the sequence of language and dialect subgrouping outlined here differs in many respects from that proposed by Emory, and although closer to that of Elbert, contains several important differences in treatment of MQA as two dialects and the placement of HAW in the MQ subgroup. On the other hand, where the archaeological evidence has led Emory to shift his ground from his earlier interpretations of linguistic evidence alone, we are now in substantial agreement, and the linguistic evidence, much of it assembled by Emory, can with these new interpretations be seen to fit present views. This is important, for the evidence for each domain serves as an independent check on the other and the interplay between them can furnish leads for the development of new theories in both domains.
At present, Elbert's lexicostatistical results by a multiple cognate method, when transformed on the basis of a more restricted sample of glottochronological dates, appear to give as reasonable results as any. On the other hand, Emory's application of modified version of the multiple cognate method (because it was rooted in an inadequately supported archaeological assumption) led to results that appear to be too early. The point is that further studies in this realm may yet provide means for independent testing of theories about the use of this method for dating in PN. But comparative linguistics and subgrouping seems to offer far greater scope for reliable propositions of a historical nature which we can then examine archaeologically. 122
Polynesia has many times been cited as a unique laboratory for testing questions of culture history by intra-disciplinary approaches within anthropology. In the fields of historical linguistics and the development of archaeologically sound historical reconstructions there lies, in my opinion, one such promising line of development. We may hope one day to reconstruct linguistically the plants and animals, the technological complexes, or the kinship systems for various subgroups of PN and trace their development from their PPN beginnings. And we may hope to test, correct, and expand on these interpretations using archaeological evidence for changes in the economy, technology, or settlement patterns, of the Polynesians with new and perhaps theoretically significant results. When we do it will be because much more work covering the entire area of Polynesia will have been done in both fields with this goal in mind, and not through isolation of fields, or concentration on particular areas, or analysis only of restricted aspects of the available evidence.
Too often, new developments in one area or in a related field affect fundamentally the interpretation of results in other realms, as recent changes in Polynesian culture history have shown.- 36
The eighty “Marquesan words shared with Hawaii only” listed by Emory (1946:104) have been rigorously checked against available Hawaiian and Marquesan dictionary sources. They have also been checked against materials filed on punched cards in the Polynesian linguistic project at the Department of Anthropology at the University of Auckland, where all available dictionary sources for a majority of Polynesian languages are now being searched for shared cognates. As a result, many items listed by Emory have been rejected, usually because either the Hawaiian and Marquesan forms were not cognate, or they were shared with some third language. The following, however, remain:
SE-MQA and HAW (3 items)
SE-MQA ?anahu, HAW lanahu ‘charcoal’; SE-MQA feei (Emory provides glottal but it is not indicated in Dordillon 1931) ‘to open eye with finger, to deny or mock’, HAW helei ‘to pull down eyelid with finger as sign of contempt’; SE-MQA, HAW pana ‘bow, to shoot with bow’.
SE and NW-MQA and HAW (13 items)
SE and NW-MQA apana/apena, HAW apana ‘a piece, fragment’; SE and NW-MQA, HAW hinihini ‘indistinct, faint (as a voice)’; SE and NW-MQA hoata, HAW hoaka ‘brightness, shining’; SE and NW-MQA, HAW humu ‘to bind (by string)’; NW-MQA kaka, SE-MQA 123, HAW ?a?a ‘sack, pocket, bag’; NW-MQA matuke, SE-MQA matu?e, HAW maku?e ‘dark brown (of people)’; SE and NW-MQA neveneve ‘very large, huge (as stomach)’, HAW newenewe ‘plump, filled out (as pregnant woman)’; SE and NW-MQA, HAW noni ‘shrub (Morinda citrifolia)’; NW-MQA pakiuma, SE-MQA pa?iuma ‘amusement consisting of striking chest in rhythm’, HAW pa?iumauma ‘chest slapping hula’; SE and NW-MQA pa?oa, HAW paloa ‘a seine’; NW-MQA pekahi, SE-MQA, HAW pe?ahi ‘fan, to fan’; NW-MQA pikao, SE-MQA(1) pi?ao ‘to wrap up, envelop’, HAW pi?ao ‘to fold ti leaves into cup-like package’; NW-MQA tekoteko SE-MQA te?ote?o, HAW ke?oke?o ‘white’.
SE and NW-MQA and HAW (12 items over which some question exists)
1 Elbert 1953.
2 Emory 1964, 1963.
3 Acknowledgment is made here of many people to whom I am indebted for information on particular points, or who have reviewed the entire manuscript and materially improved it by useful commentary and correction of errors of fact, interpretation, and presentation. They include Bruce Biggs, Kenneth P. Emory, George Grace, and Andrew Pawley, as well as Donn T. Bayard, Marian Kelly, Yoshiko Sinoto, Dave Walsh and Ralph C. White. For the use of unpublished materials and manuscripts I have to thank Dr. Emory, Mr. Pawley, Mr. Bayard, and Mr. White. To my wife I owe much in developing the theme that Tahiti was not the sole Hawaiki of East Polynesia and for support in endeavouring to provide new orientations in Polynesian culture history. For encouragement while writing this paper I wish also to thank Jesse D. Jennings. As usual, views and data as presented here remain my responsibility.
4 I use the terms West and East Polynesia to refer to the geographic areas normally thought to be culturally distinctive within Polynesia (Burrows 1938). I have abandoned Western Polynesian, however, as a term for a subgroup thought to contain the languages of West Polynesia and Outliers but retained the term Eastern Polynesian (EP) for the subgroup that contains those languages which belong to it on linguistic grounds and are concentrated in East Polynesia. Symbols and abbreviations are the same as used in the accompanying paper by Pawley.
5 Vayda and Rappaport 1963: 133-134.
6 Churchill 1911 ; 1912.
7 Churchill 1911: 179.
8 Churchill 1912. Emory 1946.
9 Emory 1963: 81.
10 Elbert 1953: 148.
11 Emory 1946; 1963.
12 Elbert 1953.
13 Marshall 1956a ; 1956b, see also Polynesian in 19th edition of Encyclopaedia Britannicaa.
14 Dyen 1965.
15 Elbert 1953: 158 suggests that Western and Eastern Polynesian had become distinguished before the differences developed among languages in either area, but in Table 4 his arrangement implies the position developed by Pawley.
16 Churchill 1911: 138-142 ; Capell 1962: 391-92 ; Marshall 1956b: 66, 70.
17 Marshall 1956b: 72.
18 Elbert 1953: 164-165. Dyen 1965: 34.
19 Bayard 1966.
20 Bayard 1966: 9-11 and fn. 7, where a detailed assessment and references to various workers may be found.
21 Bayard 1966: 85-89.
22 Churchill 1912. Emory 1946: 70, 253.
23 Elbert 1953: Table 4.
24 Marshall 1956b: 72.
25 Emory 1963.
26 Emory 1946: 70.
27 Emory 1963:86-88.
28 Emory 1963:96-97. Emory and Sinoto 1965.
29 Emory 1963:79.
30 Elbert 1953:161, 167.
31 Emory 1963:92.
32 Suggs 1961a; 1961b: 179.
33 Emory 1963: Table 5c.
34 Emory 1963:99.
35 Emory 1963: Table 5a and 5c.
36 Pawley, this issue, sec. 3.3 and fn. 27.
37 Elbert 1953: Tables 1 and 3.
38 Dyen 1965:35 ; Pawley, this issue, sec. 2.1.6 and fn. 8.
39 Earliest radiocarbon dates from Polynesia associated with pontillé (Lapita) style pottery style are from Tonga (Golson, personal communication). The connections are to west with earliest materials from Fiji and New Caledonia.
40 Elbert 1953: Tables 1, 2 and 3. Bayard 1966: 60 and B. G. Biggs, personal communication, on possibility of borrowing to explain ? in Futunan. It is important to note that if Bayard 1966:59-63 is correct in his reconstruction of the sequence of sound shifts, PEP split off from PNP before all of the Outliers except Rennell and his southern Outlier group, while Futunan would separate at a much earlier date than we have indicated using Elbert's lexicostatistical evidence. Pawley, personal communication, is also of the opinion that differentiation in the PSM subgroup is much earlier than indicated by Elbert's figures and in my Table 8. If so, the likely explanation for Elbert's figures is later borrowing.
41 Elbert 1953: Tables 2and 3.
42 Bayard 1966:93-94.
43 Elbert 1953: Table 2 gives 67% between SAM and TON compared to 61-64% for other languages of West Polynesia. Emory 1963: Table 2 gives 85.5% as compared to 62-70% for either language to other EP languages. See also Bayard 1966: 61.
44 Sharp 1963: 28-29, 32.
45 Elbert 1953: 167.
46 Davidson 1965: 69.
47 Suggs 1961b: 174-181.
48 Sinoto and Kellum 1965.
49 Sharp 1963:32.
50 Suggs 1961b: 174-179; Emory 163:96-99; Emory and Sinoto 1965: 103.
51 Sharp 1963: 133-134.
52 Elbert 1953: Table 1. Thus proto-Polynesian *s is reflected by h in the TO subgroup where it falls together with the h reflex of *h, whereas this has not occurred in PEP, suggesting that the two similar sound shifts are independent in origin.
53 Elbert 1953; Dyen 1965:34.
54 This refers to the dialect normally recorded for Nukuhiva, not those spoken in only a few of its valleys. It is like that spoken on Uahuka but today this latter speech also exhibits some SE Marquesan sound shifts and lexical items, due probably to the fact that few of the present population of this island are originally from Uahuka but from other islands.
55 I wish to thank Marimari Kellum and Yoshiko H. Sinoto who also collected the currrent 100 list vocabularies and permitted their use in comparison with my lists computed on the basis of the dictionary. The two differed only in a few items and without significantly affecting overall results.
56 Emory 1963: Table 2.
57 Dordillon 1904: 86-90.
58 Dordillon 1904: 90. M. Kellum and J. H. Sinoto, personal communication, and the 100 word vocabularies as set out in Emory (1963) collected by them in Ho'oumi, Hatiheu, and Taipivai valleys.
59 Emory 1946: Table 7. Handy 1923: 348-349.
60 Elbert 1953: Table 4.
61 Emory 1963: 83.
62 Elbert 1953: Table 2; until publication of Fuentes dictionary (1960) no really adequate source existed for EAS and main reliance was on Churchill (1912).
63 Fuentes 1960; Ward 1961; Bergmann 1963: 6-9.
64 Emory 1946:108-115.
65 Emory 1946:110-111.
66 Dyen 1965:47.
67 Emory 1946:81-82.
68 Emory 1946:86-87, 186.
69 I am indebted to Mr. D. R. Simmons for checking the occurrence of *ng to k sound shift in MOR, as Williams 1919 and Skinner 1923: 43 give only a few examples.
70 Emory 1946:20. In a series of lexical comparisons between HAW and MQA handwritten by Alexander in 1853 the closer phonological relationship of HAW to SE-MQA than to NW-MQA was commented on and demonstrated.
71 Emory 1946:28.
72 Emory 1946: 25-26 and personal communication.
73 Dordillon 1931 : 74.
74 See appendix where following re-checking and further searching, the items still remaining as valid are listed.
75 Careful checking in Dordillon (1931:76, 306, 312) reveals these forms had the distribution indicated even where he fails to mark it clearly.
76 Emory 1946:116.
77 Emory 1946:186-190.
78 Emory 1946: Table 7: 182-190, 257-60.
79 It is important to remember that Elbert established glottochronological dates for only four languages by the single cognate method, and that his Table 3 is a scale based on these results and applied to the other results obtained from the multiple cognate method.
80 Elbert 1963:167.
81 Suggs 1961b: 174-176 states his case for accepting the two early dates from the Haa-1 site, as well as two later dates. Their associations with the material they are said to date is discussed and questioned by Sinoto and Kellum 1965:37-41.
82 Elbert 1953: Table 3.
83 Heyerdahl and Ferdon 1961:527-28, Emory 1963:94-96; see also Emory 1962:884 and Golson 1965: 76-77 for critical comment. It follows, of course, that PCP would be spoken in the Marquesas or one of the members of PTA subgroup on the arguments developed here and that Easter could not be the homeland of PEP unless it can be shown archaeologically that it was settled before any other island in East Polynesia and possessed a culture at that time from which early Marquesan and Tahitian assemblages may be derived. At present neither of these conditions obtains.
84 Elbert 1953: Table 3.
85 Emory and Sinoto 1965: 95-96, 99.
86 Elbert 1953: Table 3.
87 Linton 1923:445-446: Handy 1923:21-22.
88 Sinoto and Kellum 1965:47-54.
89 Marshall 1956b: 72.
90 Stokes 1955.
91 Dyen 1965:35.
92 Stimson and Marshall 1964.
93 Stimson in Stimson and Marshall 1964: 22.
94 Emory 1963: 98, favours subgrouping RAR and MAO.
95 Elbert 1953: Table 2. Emory 1963: 98.
96 Emory and Sinoto 1964. Emory and Sinoto 1965.
97 Sharp 1961: 352.
98 Emory 1946: 257-260.
99 Emory, Bonk and Sinoto 1959. Sinoto 1962. Sinoto and Kellum 1965.
100 Emory and Sinoto 1961. Emory 1963:99 and personal discussion at B. P. Bishop Museum 1965.
101 Sinoto 1962: 164-166 and Fig. 2.
102 Sinoto 1962:166.
103 Emory 1928.
104 Emory 1963:93.
105 Willey et al. 1956:15.
106 Adkin 1960.
107 Golson 1960.
108 Simmons 1963 : 64, 91, 95, 155.
109 Parker 1962.
110 Handy 1923: 142. Suggs 1961b. Sinoto, personal communication.
111 Golson 1959. 37.
112 Williams 99, Skinner 1923:43, D. R. Simmons, personal communication.
113 B. G. Biggs, personal communication.
114 Stokes 1955: 317, who says l occurs initially and medially in Rapan and still tends to prevail over r despite TAH influence. Hale 1848 and Alexander 1853 among others for the MQA. The Rev. Ellis still used both when writing Hawaiian and after having lived in Tahiti, see Emory 1946: 16-17 for Hawaiian.
115 Skinner 1923: 43; D. R. Simmons, personal communication.
116 Dordillon 1904: 140-141.
117 Churchward 1959: 193.
118 Williams 1919: 422.
119 Skinner 1923: 43.
120 Williams 1928:343.
121 Emory 1946: 186.
122 Hoijer 1962 in a study of Athapaskan, a case not unlike that in Polynesia, reached a similar conclusion that where borrowing occurred between related languages glottochronology often failed to discriminate relationships obtainable by use of the comparative method.
123 The SE-MQA form is listed with instruction to see NW-MQA form but appropriate meaning is found only under latter. The assumption is that Dordillon dictionary is not exhaustive in listing either SE-MQA forms or their meanings.