Volume 72 1963 > Volume 72, No. 2 > East Polynesian relationships, by Kenneth P. Emory, p 78 - 100
EAST POLYNESIAN RELATIONSHIPS
Settlement Pattern and Time Involved as Indicated by Vocabulary Agreements
In this article Dr. Emory, of the Bernice P. Bishop Museum, Hawaii, applies modified glottochronological techniques to the problem of Polynesian settlement of the Eastern, Pacific. An earlier draft of this article was presented at the Tenth Pacific Science Congress.
To ANYONE intensively engaged in comparing the ancient culture of one part of Polynesia with another, it should become quite apparent that in language they are not only in remarkably close agreement, but that this also has been the most stable part of Polynesian culture. The vocabularies describing or expressing features of the culture are, more often than not, in closer agreement than the features themselves. Also, it is not only easier to compare these words than these features, but such comparisons can be handled statistically with far greater facility. For the languages of the island groups within Polynesia we possess excellent data, especially when it comes to basic vocabularies. The language data, we might say, is already well in hand. The data on the rest of the historic cultures are partially complete and that on prehistory, from the discoveries of archaeological research are still in a fragmentary stage. Certain sure conclusions and many others nearly sure can be drawn from a comparative study of the language material as to relationships, main directions of cultural spread, and relative and even absolute time involved. Let us make these studies and utilize them in the reconstruction of Polynesian prehistory.
This is a report on a study of the results of comparing a hundred basic-word vocabulary between the six most widely separated island areas of East Polynesia, and Samoa and Tonga in West Polynesia, in the hope of detecting with some finality the order in which they were settled, their prior island homes, and the arrival times. It includes a review of a two-hundred (actually 202) basic-word comparison by Samuel H. Elbert in 1953 1 and an attempt at a total word comparison by Emory in 1946, 2 having the same objectives.
The 100-basic words compared are drawn from the 100-item test list compiled by Swadesh 3 and reproduced by Hymes, 4 except for twelve of the words found unsatisfactory in one way or another for the Polynesian comparison, namely: many, dog, seed, bark, egg, horn, see, sleep, kill, walk, earth, green. Eight substitutes were drawn from - 79 Swadesh's supplementary list, namely: straight, fruit, alive, flow, grass, pull, child, wing. Four other words were added, namely: spit, rat, dirt, laugh. The entire list is to be found in Table 1, showing comparisons between reconstructed Proto-Polynesian, PPN, and reconstructed Proto-East-Polynesian, PEPN.
In reckoning equivalence, Elbert and Emory have been followed in regarding words as cognate and equivalent which are identical except for the regular, known sound shifts. Also, when two words are known for the same item, one of them cognate, the other not, I have counted a full agreement, even though the cognate is nearly obsolete, because the cognate represents a prehistoric connection. But in one respect I have departed not only from our own counting in the past, but from all others, and that is in counting as one-half agreement the following: metatheses, such as head, puoko, Easter Island, for upoko; irregularities in sound shifts such as mouth, haha, Marquesas, Easter Island, for vaha; the words having identical form but a shift to a related meaning such as kohu, “cloud”, Easter Island, while for all others kohu means a “kind of cloud”, or “fog”, and maitake, meaning “clean, beautiful”, Easter Island, but with others it means “good”. The justification for this is that in a hundred-word list, a count of zero in the cases just mentioned would remove a degree of real agreement present, and a count of one, overstresses the agreement. When it comes to glottochronological reckonings, an accumulation of 10 such cases with a choice of a 0 or a 1 count, could throw the count off 10 per cent one way or the other, and should 1 per cent be considered the equivalent of change in one hundred years, a real possibility, we would have a difference of a thousand years in counting one way as between the other. The counting of one-half, though admittedly somewhat risky and subjective, is however no more so than the other method and does have the virtue of registering a more normal agreement. In the final counting it was found that in the total of 28 comparisons made, there were 10 such related words in only one of the comparisons, 9 in three, 8 in four, 7 in nine, 6 in five, 5 in two, 4 in one, and 3 in three.
The one-hundred word list comparison made here should be more reliable than the two-hundred word list comparison in that better data has come to hand since that comparison, especially for Hawaii, 5 Tonga, 6 and Easter Island, 7 and the list of words is better adapted to Polynesia. The substitution of the ubiquitous rat for dog, for example, enables us to bring in the Marquesas and Easter Island, where the dog was absent at the time of European contact, on a better footing with the other islands. Also, to a large extent Elbert and I have worked on the list of words together.- 80
PPN AND PEPN RECONSTRUCTIONS OF THE 100-WORD LIST
check revealed that his consideration of what were cognates was not too rigorously based and in my own count I was all too conscious that I was missing many equivalents not only through inadequate data but for lack of time in searching. That study, therefore, was a rough, pilot project, to gain some idea of the situation. The fact that it foreshadowed the results from the better data and more precise methods in the 200- and 100-word comparisons, is somewhat of a validation. As a total vocabulary comparison reflects total culture more adequately than basic word comparisons, it has its own weight in determining primary relationships.
To have the necessary perspective, I have brought into the comparisons and considerations, Samoa and Tonga, and the areas chosen for East Polynesia have been Tahiti, the Marquesas, Hawaii, New Zealand, Mangareva, and Easter Island. The Marquesas and Tahiti, Samoa and Tonga, were within some effective diffusion range, and the agreement we may find between them may be closer than it would have been if this were not the case. But we can be sure that an agreement between Easter Island and New Zealand, unless accidental or due to a natural change in which both languages were headed (i.e. due to “drift”), that the identity is due to common ancestry.
Table 2 gives the results of the counts of cognate agreement on the 100-word list, on Elbert's 200-word list, and on my total vocabulary comparison.
BASIC UNITY, EAST AND WEST DIVISION
What is impressively brought out by vocabulary comparisons is the very close agreement between widely spaced island groups in East Polynesia and again in West Polynesia, and the definiteness and evenness of the gap between East and West Polynesian groups. Also revealed is a weaker agreement between Easter Island and the other groups of East Polynesia than they show among themselves, pointing to an early break from them.
The figures which support the above statements are these:
In the 100-word list (see Table 2), the East Polynesian groups, Tahiti, Marquesas, Mangareva, Hawaii, and New Zealand, show a sharing of words ranging from 81.5% to 92%, and a mean of 85.3%. The West Polynesian groups, Samoa and Tonga, show the same degree of agreement between them, 85.5%. Easter Island's vocabulary stands in weaker agreement with the other islands of East Polynesia, ranging from 69% with Tahiti, 74.5% with Mangareva, 75.5% with Hawaii, to 76.5% with the Marquesas and New Zealand, a mean of 74.4%.
The gap between the vocabularies of East Polynesia, excepting Easter Island, and those of West Polynesia is represented by a range of 60% to 71.5%, and a mean of 65.65%. Easter Island is in 62% agreement with Samoa and 64.5% agreement with Tonga, a mean of 63.25%, only 2.4% less than the 65.65% of the other groups of East Polynesia. In this it appears to share in common with them the break with West Polynesia, being a little weaker because of it being a colony of a colony.- 82
Upper numbers: 100-word list percentages of shared retention; parentheses give number of words counted as ½.
Middle numbers: 200-word list percentages of shared retention from Elbert. 10
Lower numbers: Total vocabulary comparisons: number of shared words: parentheses give number of uniquely shared words. 11- 83
POLYNESIAN SETTLEMENT PATTERN IMPLIED BY LINGUISTIC DATA ALONE.
Primary settlement pattern and estimated times as indicated by 100-word and 200-word basic vocabulary agreements and rate of change based on the earliest radiocarbon date of circa 100 B.C. for the Marquesas and circa A.D. 500 for Easter Island. Here, the Tahitian group instead of the Marquesan group is assumed to be the formative centre of what East Polynesia holds in common differentiating it from West Polynesia.
POLYNESIAN SETTLEMENT PATTERN BASED ON LINGUISTIC AND ARCHAEOLOGICAL EVIDENCE.- 84
Primary settlement pattern and estimated time involved as indicated by the most recent findings in archaeology, combined with those from glottochronology.
In Easter Island's stand of 74.4% agreement with the other areas of East Polynesia, lower by 10.9% than the 85.3% agreement they have with each other, and in its 63.25% agreement with West Polynesia, another 11.15% lower, we have indicated that about half way between the time of the separation of East Polynesia from West Polynesia and the time of the separation of the other islands of East Polynesia from each other, the language of Easter Island broke away from East Polynesia.
The 200-word comparisons yield much the same overall results. Within the same islands of East Polynesia the agreement is 67% to 76%, with a mean of 70.4%, and within the West Polynesian groups of Samoa and Tonga, it is 66%. Easter Island has an aside position with a range of 62% to 64% of words shared with the other areas of East Polynesia, and a mean of 63.2%.
The gap between East and West Polynesia is represented by a range of 45% to 60% in shared vocabulary, and a mean of 53.2%, on the part of the East Polynesian groups excepting Easter. Easter Island is in 53% agreement with Samoa, 48% with Tonga, a mean of 50.5%. This puts Easter in its closeness to the other islands of East Polynesia at one-third (34%) closer to East Polynesia than it is to West Polynesia, whereas the 100-word comparison puts Easter Island half-way between the two.
I discovered the identical pattern of agreement 12 in attempting a total vocabulary comparison, uncovering a sharing of between 1,051 and 1,439 words, and a mean of 1,191 among the same islands of East Polynesia, excepting Easter Island, the latter sharing with the others 606 to 643 words (the extreme lower number partly because of its smaller vocabulary), or a mean of 622 words. The gap between East and West Polynesia was revealed by a sharing of 776 to 900 words, a mean of 831 words, with West Polynesia, on the part of the East Polynesian Islands excepting Easter Island, and 519 words between Easter Island and West Polynesia.
CENTRE OF PROTO-POLYNESIAN AND OF PROTO-EAST-POLYNESIAN
The fundamental unity within the Polynesian family of languages, marked by an even greater unity within the islands of West Polynesia and within those of Eastern Polynesia, means a common origin for the whole family and secondary common origins for the West and the East divisions. These unities could not have been developed in widely separated island groups. In the beginning there must have been a common island area containing islands within some easy degree of intercommunication, such as that which existed between the islands of the Hawaiian group. Further, this area of primary common origin must have been within Polynesia in order to develop the unity we have just observed. Do our figures give a clue as to the location of this area for the Polynesian family as a whole? Not by themselves.
By resorting to a stratagem we can see where the home area lies for the Polynesian languages, as between East and West Polynesia, - 85 and narrow the home area for the East Polynesian division to at least between the Marquesan and Tahiti groups, with the probability centering on the Tahitian area as that in which the major moulding of the uniquenesses which the East Polynesian languages hold in common took place.
The stratagem is, out of the data at hand, to reconstruct for our 100-word test list, a Proto-Polynesian (PPN) vocabulary, or we can call it a predispersal Polynesian vocabulary, and also a Proto-East-Polynesian (PEPN) vocabulary, or predispersal East Polynesian vocabulary, then compare the present vocabularies against them on the reasonable assumption that the homeland is very likely to retain a larger percentage of original vocabulary than its “colonies”. It should follow then that the area retaining the largest percentage of ProtoPolynesian words would be the home area of the Polynesian family of of languages, and the area retaining the largest percentage of Proto-East-Polynesian words would be the home area of East Polynesia.
I am using the word “colonies” not in the usual sense of a settlement in communication with the homeland, but in the sense of the first establishment of a settlement in a distant island group which then remained in comparative isolation. In the homeland, family, district, or island, variations in a vocabulary tend to be checked from becoming the norm for the whole area because of the larger number of people familiar with and practising the original form. On the other hand, an island area will receive initially a group of people from the homeland who are likely to be from one family or district or island of the home area. Their local changes in the form of words and in meanings could now, in their isolation, be readily established and with the loss of original forms and meanings.
Not only should it be expected that the homeland would retain a greater amount of the original vocabulary than its colonies, but in the colonies themselves the vocabulary of the first settlers to establish themselves should predominate over the language of subsequent arrivals and for the same reason. The later arrivals would come in small groups, probably one canoe load at a time and less than a hundred persons. They might effect some changes in the current basic vocabulary, but would not likely replace it with their own.
If the above suppositions are correct, we should be able to detect not only the homeland of the Polynesian family of languages, and of the East Polynesian division, but also the primary settlement pattern for East Polynesia.
We will now, then, proceed to see what our figures on vocabulary agreement indicate, starting with a Proto-Polynesian and Proto-East-Polynesian vocabulary for our 100-word list. This, fortunately, can be constructed with some degree of certainty for Proto-Polynesian by counting the words in the West Polynesian list which can be traced definitely to Proto-Indonesian, as predispersal, and words shared between East and West Polynesia also as predispersal. For Proto-East-Polynesian, words shared with West Polynesia can be counted as pre-dispersal for East Polynesia and words shared widely just in East - 86 Polynesia as predispersal also for East Polynesia. Samuel Elbert and I have made this reconstruction of the Proto-Polynesian and Proto-East-Polynesian vocabulary for the 100-word list (see Table 1). I have compared the present vocabularies against these. The results are shown in Table 3.
Percentage of Proto-Polynesian (PPN) Words Retained by Individual Languages
From these comparisons it is abundantly clear that the home ground for the Polynesian family of languages is West Polynesia, where the percentage of words shared with the Proto-Polynesian vocabulary is for Tonga 91.5%, for Samoa 94%, as compared with the East Polynesian groups, which range from 68.5% to 77.5% in words retained. With the affiliations of Polynesian languages all lying to the west, it would be most unlikely in any case that East Polynesia would have been settled before West Polynesia and for the language unity to have been developed in East Polynesia and then carried to West Polynesia.
CENTRE AND SPREAD OF PROTO-EAST-POLYNESIAN
Taking as established that the early language, Proto-Polynesian, spread from West Polynesia to an island area in East Polynesia where a part of its vocabulary took on new shapes and then this differentiated vocabulary, Proto-East-Polynesian, spread to the other island areas of Polynesia as they became settled, what do our statistics indicate concerning the location of the new formation area, and the order and direction of this spread?
We must gain our sense of direction from the position of the islands themselves. From the extreme location in the north, of Hawaii, in the south of New Zealand, and in the east of Mangareva and Easter Island, it is most improbable that any of these could be the area in which the Proto-East-Polynesian language was formed and from which it was dispersed to all the other islands of East Polynesia. This would - 87 narrow the choice to the Marquesan or the Tahitian area, those nearest to West Polynesia.
Proceeding on the same assumption made in determining that West Polynesia was the area in which the over-all unity of the Polynesian language was achieved, namely, that the home area will retain more of the original vocabulary than its off-shoots, we find that, as between Tahiti and the Marquesas, the statistics for the Tahitian area fulfill this condition for being the home area for Proto-East-Polynesian, as it does retain more of the Proto-East-Polynesian vocabulary, the original vocabulary, than the other groups in Eastern Polynesia, excepting New Zealand, which retains the same amount as Tahiti. See Table 2.
In retention of the 100-word Proto-East-Polynesian vocabulary, Tahitian, with 95% retention, holds closer to the original language than the Marquesan with 87% retention. In the retention of the 2,775 Proto-East-Polynesian vocabulary, this is also the case, 72% on the part of Tahitian as compared with 60% on the part of Marquesan. New Zealand, to be sure, in the 100-word list holds as close to the Proto-East-Polynesian as Tahiti, but not on the 2,775 word list. This can be explained on the basis of New Zealand having been settled from Tahiti late in time, preserving the same characteristics as Tahiti in basic vocabulary, which is slowest to change, but revealing its weaker relationship to Proto-Polynesian in total vocabulary.
We can make and test another assumption in confirmation of the Tahitian area being the home area for Proto-East-Polynesian. On the basis that home areas retain more of the original vocabulary than offshoots, if the settlers of Tahiti came from West Polynesia, and Tahiti is the cradle of Proto-East-Polynesian, Tahitian should show a closer relationship in current vocabularies with West Polynesian than the other languages of East Polynesia show with West Polynesia. As we may see from the statistics laid out below by all three lists, Tahitian is not only closer than Marquesan to West Polynesia, seemingly eliminating that area from being the home area of Proto-East-Polynesian, but Tahitian is also closer than all the other four East Polynesian languages we have taken into consideration, though some of the others are very close.
PERCENTAGE AND NUMBER OF WORDS SHARED BETWEEN EAST POLYNESIA AND SAMOA AND TONGA
One might argue that Tahiti stands closer to West Polynesia than the Marquesas because the people migrated from West Polynesia to the Marquesas first, and have remained longer in isolation causing their speech to depart further. This thinking, which necessarily assumes a later, independent, direct migration to Tahiti, would bring about two island areas in East Polynesia where two close stages of Proto-Polynesian would continue on two independent paths, creating two divisions in East Polynesia. Under these conditions the distinctiveness which the East Polyesian languages possess in common when compared with West Polynesian languages could hardly have come about. Also, if the Marquesas and Tahiti were in the early beginning settled from West Polynesia independently, because of their geographical isolation they could not have been in close enough communication to form the Proto-East-Polynesian language. The basic East Polynesian vocabulary of the one must necessarily have been derived from the other. If it was formed in the Tahitian group, then it was carried to the Marquesan group.
It might be argued also that Tahiti stands closer to West Polynesia than the Marquesas, because of later influence. But Tahiti's distance from Samoa should preclude serious influence on its ancestral vocabulary from that direction. The lower numerical agreement in shared vocabulary between the Marquesas and West Polynesia as compared with Tahiti can, then, with some assurance be laid to a less intense connection with West Polynesia because it came through Tahiti.
POSITION OF TAHITI, MARQUESAS, NEW ZEALAND AND HAWAII IN EAST POLYNESIA
If Tahiti was indeed the island area in which the distinct characteristics common to the East Polynesian languages first appeared and were then carried to the other islands by the groups of Polynesians who first settled them, the figures we have for the amount of shared vocabulary should reflect this.
Again we must first take into account the practicalities. We can be sure the far-flung islands would not have been settled simultaneously. Some would be settled before others. Nor would the settlement necessarily be directly from Tahiti. The movement from Tahiti to New Zealand or Easter Island would not be via Hawaii, or vice versa, but the movement to Hawaii, Mangareva, and Easter could have come about via the Marquesas.
Proceeding on the basis of these practicalities, and on the assumption that the last group to break away from a common motherland would retain a larger share of the vocabulary of that motherland than groups which broke away earlier to settle and become the progenitors of the population of a new land, what do we find? Taking up the Marquesas, the nearest and most accessible group to Tahiti, we would expect that it would have been reached before New Zealand. Our figures show for the 100-word list, a 92% agreement in basic vocabulary of Tahitian with the Maori of New Zealand as compared with a 82.5% agreement with the Marquesan. The 200-word list puts these - 89 agreements at 73% and 67% respectively. The total vocabulary comparison has Tahiti share 1,348 words with New Zealand compared with 1,051 words shared with the Marquesas. Therefore, we assume that the first settlers of New Zealand broke away from Tahiti later than the Marquesans and have not had as long a time for the language to diverge. The route to New Zealand may have been via the Cook Islands, and Elbert 14 puts New Zealand closer to Rarotonga in the Cook Group than to Tahiti (85% as compared with 75%), thus opening up this possibility, which does not, however, alter our other conclusion.
What about Hawaii? It shows a larger sharing of vocabulary with Tahiti than it does with the Marquesas, an 88% with Tahiti, 85.5% with the Marquesas, by the 100-word list; a 76% with Tahiti, 69% with the Marquesas by the 200-word list; a sharing of 1,439 words with Tahiti compared with a sharing of 1,190 words with the Marquesas in the total vocabulary comparison. Therefore, we conclude that not only do the Hawaiian vocabularies owe their greater agreement with Tahitian to their first settlers having come from Tahiti, but that Hawaii was settled from Tahiti after the Marquesas had been settled from Tahiti.
Now, what we would like to know is this: was Hawaii settled from Tahiti before a group set out from Tahiti to settle New Zealand? The 100-word list comparison has New Zealand share a larger percentage of words with Tahiti, by 4%, than Hawaii, i.e. 92% compared with 88%. This would indicate that New Zealand was settled after Hawaii. However, the 200-word list reverses the situation. Hawaii shares a larger percentage of words with Tahiti by 3%, than New Zealand, i.e. 76% compared with 73%. In the total vocabulary comparison 1,439 words were found shared between Hawaii and Tahiti compared with 1,348 words between New Zealand and Tahiti, again favouring Hawaii as settled after New Zealand. However, I suspect that had my familiarity with the Maori language been equal to my familiarity with Hawaiian, I would perhaps have found as many, if not more Tahitian equivalents in New Zealand. The agreements are so close as to which was settled first from Tahiti that all we can say from our figures is that they were settled close to the same time. We know that culturally, however, Hawaii was exposed to developments in Tahiti in connection with changes in the names of months and names of nights of the moon, 15 which did not affect New Zealand, and also in aspects of worship on temples. 16 Such cultural features, however, could have been introduced by chiefs from Tahiti reaching Hawaii long after its settlement from Tahiti.
POSITION OF EASTER ISLAND AND MANGAREVA
It is perfectly clear from our statistics on vocabulary agreement that the Easter Island language departed early from the other groups in East Polynesia. These other groups agree among themselves at a - 90 mean average of 85.3%, and Easter Island agrees with them at a mean average of 74.4%, or 10.9% lower. Mangareva's agreement with the Marquesas is 85.5%, and with Tahiti, 84%. Easter Island's agreement with the Marquesas is 76.5%, or 9% lower than the Mangarevan agreement. Therefore, Mangareva must have been settled well after Easter Island and so the Easter Island vocabulary could not have been derived from this nearer archipelago.
Reckoning still by our 100-word list, Easter Island is closer to the Marquesas, with 76.5% agreement, than it is to Tahiti, with 69% agreement. The 200-word list reveals only a slight difference: 63% with the Marquesas and 62% with Tahiti. In total vocabulary agreement, 635 words were found agreeing with the Marquesas, and 610 words with Tahiti, but, very significantly, it was discovered that Easter Island shared out of these 610 words, 50 words uniquely with the Marquesas, as compared with 13 words uniquely shared with Tahiti. Thus it would seem clear that as between Tahiti and the Marquesas the first settlers of Easter Island came from the Marquesas.
The most amazing fact about Easter Island's vocabulary is its retention of the Proto-Polynesian glottal alone among the islands of East Polynesia. 17 In West Polynesia not even Samoa retained this glottal. In this regard Easter Island parallels Tonga. Eight of the words in the 100-word list which should have this glottal if it had been retained, do have it:
The above words are found in all the islands compared, but with the glottal only in Tonga and at Easter. In addition, we have the word for “spit”, a'anu at Easter, 'a'anu in Tonga, and anuanu in the Marquesas, and anu in Samoa. No other East Polynesian groups share this word with this meaning. Their word for spit is tuha. Also, Easter Island has ta'e for “no”, shared only with Tonga, which has ta'e. It does not survive in any other of the groups we are considering.
In the total vocabulary comparison it was found that Easter Island has hahie for firewood, which in Tonga is fefie, and in Samoa is fafie, but in all the other groups of East Polynesia, including the Marquesas, this word is vahie, wahie, vehie, or ve'ie.
We are hardly justified, however, on the basis of correspondence with Tonga to postulate a direct migration from Tonga to Easter Island. Easter Island shares 76.5% of its words with the Marquesas, 63.5% with Tonga, on the 100-word list, 63% with the Marquesas and 48% with Tonga on the 200-word list. Besides, the Easter Island - 91 vocabulary has picked up the East Polynesian differentiations: kiore for rat, nui for big, kiko for flesh, upoko (in the form puoko) for head, haha for mouth. Also the Easter Islanders have the East Polynesian names for the days of the month, and these correspond more closely to the Marquesan names than to the Tahitian. 18
Mangareva, in the 100-word list, shares 85.5% of these words with the Marquesas and 84% with Tahiti. In the 200-word list Mangareva shares 73% of these words with the Marquesas and 68% with Tahiti. In total vocabulary more words were found shared with the Marquesas, 1,106, than with Tahiti, 1,078, and more particularly more uniquely shared words, with the Marquesas, 126, as compared with Tahiti, 28. All lists place Mangareva closer to the Marquesas than to Tahiti, and therefore favour the Marquesas as the homeland. We know also that the plant names are distinctly Marquesan when they depart from the Proto-Polynesian or Proto-East-Polynesian names, and so can feel assured that their basic vocabulary came originally from the Marquesas. At the same time we do know, as in the case of Hawaii, that late cultural developments in Tahiti, in connection with the temple worship, reached Mangareva, as they did to some extent the Marquesas.
INTERVALS OF TIME INVOLVED AND GLOTTOCHRONOLOGY
Our examination of vocabulary agreement has brought us to the conclusion that the primary settlement pattern for Polynesia was from West Polynesia, probably Samoa, to the Tahitian area in East Polynesia, with an early break to the Marquesas followed soon thereafter by a settlement of Easter Island from the Marquesas and sometime thereafter by a settlement of Mangareva from the Marquesas. Lastly, New Zealand and Hawaii were colonized from Tahiti.
What do the figures on agreement indicate as to the relative amount of time involved between the various separations, and is there any chance of translating the intervals into absolute time? Let us begin with the 100-word list. Assuming that the primary movement was from Samoa to Tahiti, then the 71.5% agreement in vocabulary between the two can be interpreted as a shared retention of this much of the vocabulary since then. Assuming that the first break away from the Tahitian area to another far group in East Polynesia was to the Marquesas or vice versa, the 83% of words shared between Tahiti and the Marquesas would represent the amount of retention since that time. The difference between 71.5% and 83% retention on the part of Tahiti, 11.5%, would represent then the amount of change which came about during the formation of Proto-East-Polynesian, and 17% the change which took place in Tahiti and the Marquesas since then. More change, therefore, would have taken place since the occupation of the Marquesas from Tahiti, or vice versa, than since the occupation from Samoa. If the change had proceeded at the same rate, then in fact, we could even say that this prior time was two-thirds as long as the time since.- 92
Is there any means of translating the amount of change into years? If the rate were uniform, the answer would be yes, because through radio-carbon dating we have evidence that the Marquesas were inhabited by at least 2,000 years before 1900, or about 100 B.C. 19 We can take that date as a minimum time since settlement and obtain a minimum rate of change. The shared retention of basic vocabulary since then, since the theoretical settlement from Tahiti, on the 100-word list is 83%, the present amount of agreement between Tahiti and the Marquesas. This means a shared retention rate of 91.25% in 1,000 years. On the 200-word list, the retention obtained is 67%, a rate of 82% in 1,000 years. Elbert 20 used the rate of change worked out for a European language, i.e., at the rate of 82% in 1,000 years, within the same language. This means a shared retention rate of 64.5% for 1,000 years. Obviously, if the rates used here for East Polynesia are anywhere near correct, the Polynesian vocabularies have been changing at a much slower rate than European vocabularies. This is in all probability true.
I am aware that C. Douglas Chrétien 21 claims to have demonstrated mathematically that the fundamental formula of glottochronology has no mathematical basis whatsoever. Certainly, as we shall see, we seem to be getting much exaggerated lengths of time as the percentages of shared retention descend. The formula chosen above seems to give the least distortion. However, vocabularies do change in time, and this change is a function of time, and there seems to be in Polynesia assignable rates of change giving useful approximations. If mathematicians have not devised a means of determining these and applying them, perhaps they can discover better ways of examining the data presented here.
Applying the rate of 91.25% and 82% to its appropriate list and using the formula given by Swadesh, 22 t = log C/log r2 multiplied by 1000, in which t is the time, C the percentage of agreement, and r2 the rate of shared retention in 1000 years, we come out with the years shown in Table 5 for the various separations we wish to consider.
DISPERSAL DATES FOR POLYNESIA
(a) Based on the settlement of the Marquesas from Tahiti 2,000 years before A.D. 1900, giving a shared retention rate of 91.25% per 1,000 years for the 100-word list, and 82% per 1,000 years for the 200-word list.
(b) Based on the settlement of Easter Island from the Marquesas 1,400 years before A.D. 1900, giving a shared retention rate of 82.5% on the 100-word list, and 72% on the 200-word list.
(c) Based on same retention rate as (a) but with movements as indicated on page 99.
It is clear from the above figures that the rate obtained for the 200-word list was on the basis of those particular words, a particular method of determining agreement, and a particular method of scoring. Therefore, the rate for the 200-word list cannot be applied to the 100-word or vice versa. Furthermore, a study of the years obtained from applying the proper rate for each list to the percentage of change found in the comparisons brings results which, at best, give only an approximation, in terms of absolute time or dates, as wide as the plus and minuses in radio-carbon dating with a chance of their falling outside this range. Again, we can well suspect that the rate of change has not been even, and that some of the colonies, though completely independent since separation, have changed at a faster rate than others. We suspect this, particularly of Easter Island, where, if our dates were taken literally, we would have Easter Island colonized from the Marquesas before the Marquesas were inhabited by the East Polynesian speaking people. Nevertheless, we can gain from these percentages of agreement and their conversion into rates of change and into years based on radio-carbon dates of early settlement in East Polynesia, very useful time perspectives.
Radio-carbon dating in Hawaii already puts the settlement of this archipelago beyond a thousand years ago and of New Zealand almost that far. In the Marquesas we are back to the beginning of the Christian era and Easter Island, perhaps back to A.D. 400, or earlier, and Mangareva to nearly a thousand years ago. We know that some time must have been involved in the development of Proto-East-Polynesian from Proto-Polynesian. All of our calculations based on the three lists indicate that this time was considerable, more so if it were from Tonga than Samoa, and that this time was from one-fourth to two-thirds as long as the period since dispersal. By considering the Tahitian area the dispersal area to the Marquesas, Hawaii, and New Zealand, we come out with a time perspective more in accordance with what we know of the cultural relationships in general than we would if the Marquesas were taken as the dispersal area. In the latter event, the time for the formation of the East Polynesian language is longer than it would be if Tahiti were the area, and it seems unreasonably long. Also, both Hawaii and New Zealand would have been settled from the - 94 Marquesas before Tahiti, which we know could not be the case. So, our figures steer us to the alternative course: West Polynesia to Tahiti, Tahiti directly to the Marquesas, to Hawaii and to New Zealand. Here we come out with years and dates that seem more logical. This puts the settlement of the Marquesas after Tahiti followed by the settlement of Hawaii and New Zealand, with considerable agreement as to the time of Hawaii's settlement (A.D. 500 by the 100-word list; A.D. 510 by the 200-word list; A.D. 250 by the total vocabulary comparison).
It is the New Zealand position that differs: the total vocabulary comparison would have it at A.D. 150, a little before Hawaii at A.D. 250, and the 200-word comparison at A.D. 310, before Hawaii's date of A.D. 512. However, I regard the 100-word comparison as the most reliable and it happens to fit in with the results so far obtaining in radio-carbon dating, where we are getting earlier dates for Hawaii than for New Zealand. But if this reflects the true situation, then we must assume later contact also between Tahiti and Hawaii to explain some cultural developments which took place in Tahiti after the Maoris left and which affected Hawaii but not New Zealand.
Easter Island of necessity branched away from East Polynesia, but to reconcile its glottochronological date with the motherland, the Marquesas, we must assume that instead of being settled by 2474 B.C. or 1403 B.C., that it was settled after the Marquesas, or after 100 B.C. A radio-carbon date of A.D. 380 ± 100 has been obtained for Easter Island. 23 Were Easter Island settled A.D. 500, then in the 1400 years since, it has retained 76.5% of its mother vocabulary, and its rate of retention or change would then be 82.5% per thousand years on the 100-word list, and 72% per thousand on the 200-word list as against the seemingly normal rates of 91.25% and 82% respectively. Its rate of retention, then, on both lists is approximately 10% lower.
We seem to have another clear case of accelerated change taking place when the population is small and therefore the control is not so restricting at the Polynesian outlier, Kapingamarangi. There the agreement with West Polynesia on our 100-word list I found to be 67.5% with Samoa, and 65.5% with Tonga, which at the rate of 91.25% shared retention in a thousand years, would place its occupation at an unreasonable distance back in time. The 200-word list, giving a 53% agreement with Samoa and a 49% agreement with Tonga, does not alter the situation. We have very good reason to believe that the ancestors of Kapingamarangi came from the Ellice Islands, but the agreement with them on the 200-word list is 58%. Applying the normal rate of change we had worked out for this list, 82% per 1,000 years, would give Kapingamarangi a settlement date of 852 B.C., whereas A.D. 852 would be a far more reasonable date to expect. Kapingamarangi seems to have had a physical limitation on the population, holding it below 500 and reducing it at times of famines to well below 300. 24 Perhaps wars or famines at Easter Island would have kept the population at a low level for long periods, thus accelerating change.- 95
If Easter Island had an accelerated rate of change because of a small population, Mangareva also may have experienced the same acceleration, above and beyond the change taking place in the islands capable of holding a larger population. If this were so, our dates of A.D. 560 and A.D. 310 would be exaggerations of the length of time they have been occupied, and it would seem from Roger Green's radiocarbon date 25 (A.D. 1120 to A.D. 1280 for charcoal from the bottom of the cultural layer in a fisherman's shelter), is in accord with applying the accelerated rate worked out for Easter Island. That is, the rate of 82.5% on the 100-word list and 72% on the 200-word list would give us the dates for Mangareva of A.D. 1263 and A.D. 943, which are in harmony with the radio-carbon date.
It is certain that we have to accept a lower rate of retention and therefore an accelerated rate of change for the Easter Island vocabulary, placing it at the rate we have on the basis of radio-carbon dates. This gives for the branching from the Marquesas to Easter Island about the same time as the branching from Tahiti to Hawaii. At the time the Easter Islanders left, the Proto-Polynesian glottal was surviving in the Marquesas. It follows from this that at the time the East Polynesians from Tahiti who may have settled in the Marquesas found already there a contingent from Tonga in the process of obtaining a foot-hold, but not yet in full occupation. The group that migrated to Easter Island came certainly from a group in the Marquesas which not only retained this glottal, but also adopted most of the East Polynesian variations and the East Polynesian names for the days of the month. In turn, the newcomers must have adopted their word for breadfruit, mei; the Tongan name (mai). The Tahitians had the Samoan name 'ulu, which was carried to Hawaii.
It would seem strange that in all the long period of the formation of the eastern branch of the Polynesian language, if it were in the Tahitian area, that the not-so-remote Marquesas Islands would have remained unreached and unsettled, either from Tahiti or West Polynesia. In either case, if that had happened at all early in the history of East Polynesia, we would have had a gap between the Marquesan and the Tahitian languages, comparable to, though smaller than, the gap we have found existing between East Polynesia and West Polynesia. There is no such gap.
Viewed through the most conservative part of Polynesian culture, the basic vocabulary, the time perspective we gain through assuming some even rate of change over the years and tying the rate to time indicated by radio-carbon dates, appears in bold relief to be this:
The order of dispersal in East Polynesia appears to be first from the Tahitian area to the Marquesas, with the real possibility that Tongans of West Polynesia had already reached the Marquesas and had begun to establish themselves in one or more of its islands. The next dispersals were to New Zealand and Hawaii from the Tahitian area and to Mangareva and Easter from the Marquesas, not later than circa A.D. 500 to A.D. 1000. In the dispersal to Easter and Mangareva it is clear that the branch to Easter was established first, and it would seem possibly from that part of the Marquesas which had been occupied by Tongans. In the dispersal to New Zealand and Hawaii, the vocabulary agreements with Tahiti are so close as to be inconclusive as to which was first, but Hawaii is fixed as settled by A.D. 500, and New Zealand from a little before A.D. 1000. Radio-carbon dating to the present has indicated A.D. 500 for Hawaii and A.D. 1000 for New Zealand as likely dates of settlement, and at any rate, that Hawaii was reached before New Zealand. What seems probable is that the ancestors of the Maoris of New Zealand left at about the same time, but arrived in New Zealand later through having paused in the Cook Islands which lie in between, and that the Hawaiians received later migrants from Tahiti who introduced developments in Tahitian culture which had taken place since the Maoris left, such as innovations in the names of months and days of the month, and in the forms of the temples. Hawaii, also, through the presence of uniquely shared words and traits, seems not to have been beyond influence from the Marquesas.
This is the picture which emerges from the vocabulary comparisons. It should be followed up with closer scrutiny of the words themselves, in tracing irreversible changes which give proof of direction, and independent comparisons to check the range which must be allowed for varying judgements as to what constitutes cognation or agreement and how it should be counted.
Since the above was presented in brief at the Tenth Pacific Science Congress held in Honolulu in August, 1961, an archaeological discovery made by Bishop Museum's expedition in the Society Islands has supplied positive evidence that New Zealand was, in the beginning, settled from the Society Islands. It has also opened up the possibility that Hawaii was first settled from the Marquesas rather than the Society Islands, and revealed, in any case, that there was cultural contact between the Hawaiian Islands and Tahiti after Hawaii's first settlement and subsequent to the spread from Tahiti to New Zealand.
The discovery was made on the island of Maupiti. It consisted of a prehistoric burial containing ornaments, adzes, and fishhooks. The ornaments and hooks coincided in their specialized forms with those of the - 97 Moa Hunters of Wairau in the South Island of New Zealand. One of the adzes had its equivalent at Wairau, but the other two, by their prototype forms, suggest that the berial even preceded the cultural spread to New Zealand. Both the adzes and the fishhooks were not identical in their forms to the early Hawaiian adzes and hooks, but later Tahitian forms of hooks appear in later Hawaiian archaeological deposits, radiocarbon dated between the 12th and 14th centuries. The earlier Hawaiian fishhooks are quite similar to early, contemporaneous hooks which have been found in the Marquesas Islands. Thus, the Marquesas Islands may be the source of Hawaii's first culture rather than the Society Islands.
An article on the Maupiti discovery presenting illustrations of the artifacts and comparing them in detail with New Zealand, Maori, and Hawaiian artifacts as mentioned above, has been prepared by Dr. Yosihiko Sinoto and myself. We hope this will appear presently in the Journal of the Polynesian Society.
What I wish to point out here is that the implications of these findings are not in serious conflict with the figures gathered in the language comparisons just made, but do throw a quite different light upon their interpretation. Although on the whole, in choosing between the Society Islands and the Marquesas, the statistics have favoured the Society Islands as the formative centre of the East Polynesian family of languages, small shifts could turn the balance in favour of the Marquesas.
We have seen that Easter Island has a vocabulary, which in all probability came from the Marquesas, and which, through the amount of its divergence and its retention of a Proto-Polynesian glottal, appears to have broken away from East Polynesia before the Hawaiian or the Maori. This indicates the Marquesas as a very early dispersal area. If then, Hawaii was first settled from the Marquesas, rather than from Tahiti, we can understand the Easter Island vocabulary being closer to the Hawaiian than to the Tahitian, through common origin in the Marquesas. In the 100-word list Easter Island shares 75.5% of its vocabulary with Hawaii and 69% with Tahiti. In the 200-word list Easter Island shares 64% of its vocabulary with Hawaii and 62% with Tahiti.
If, as indicated by archaeological findings, Tahitian chiefs arrived in Hawaii between the 12th and 14th centuries and brought about changes in the language to correspond with their speech at that time, regardless of whether the previous people spoke a language derived from Marquesan or an earlier speech derived from the Society Islands, then our percentage of shared vocabulary would represent a divergence from the one or the other, modified by later contact between Tahiti and Hawaii. The amount of agreement between Hawaiian and Tahitian, then, would be lower than it would have been if these late-comers from Tahiti had been the first settlers. This would explain why, with New Zealand settled before these late Tahitian migrants went to Hawaii, the Maori language could be closer to the Tahitian than the Hawaiian. Although this is not so in the 200-word list, in which Maori shares 73% of its vocabulary with Tahitian as compared with Hawaiian sharing - 98 76%, in the 100-word list, probably a more reliable indicator, the Maori shares a larger percentage of the vocabulary with Tahiti than does Hawaiian, i.e., 92% compared with 88%.
It might be argued that the coming of the “fleet”, or those traditional canoes from which Maori tribes trace descent, brought fresh influences from the Society Islands. But we do not seem to find in New Zealand cultural history, as represented archaeologically, changes in the East Polynesian culture, brought by the first settlers, to later phases of Society Island culture, as we do in Hawaii. Perhaps migrants from the Society Islands paused for some time in the Cook Islands before moving to New Zealand. This could result in their having separated from the Society Islands before the departure of the migrants from Tahiti to Hawaii and yet having arrived in New Zealand after the Tahiti-Hawaii contact. The culture then, that would have been taken to New Zealand would not have represented the late Tahitian culture that was carried to Hawaii. With the Cook Islands as a stopping place in mind, it is interesting to examine vocabulary agreement with the Cook Islands. Elbert included Rarotonga of the Cook Islands in his 200-word list and I have worked out the percentages of agreement in the 100-word list and made the following comparisons:
In the above we see that New Zealand is closer to Rarotonga than to Tahiti and Tahiti is closer to Rarotonga than to Hawaii. Hawaii is closer to Tahiti than to Rarotonga, as we should expect on the 100-word list, but on the 200-word list this is reversed, again indicating that the 100-word list is more reliable.
With the vocabulary evidence favouring a stop-over in Cook Islands for the settlers of New Zealand, it will not be surprising if evidence for this will be found by the archaeological team working presently in Rarotonga under Dr. Roger Duff of the Canterbury Museum, in co-operation with a programme directed by the Bishop Museum.
In our consideration of glottochronology, the time scale was anchored on the Marquesan earliest radio-carbon date of approximately A.D. 100, for settlement from Tahiti. But, if the Marquesas were the formative centre for the East Polynesian family of languages, and the spread was then to Tahiti, this date, even for the separation of the two, becomes wholly arbitrary unless we find another basis to support it. By a happy circumstance this can be done.
As we are assured that the culture first brought to New Zealand was derived from the Society Islands, we can base our time scale on the - 99 radio-carbon dates for the early settlement of New Zealand. For South Island they are close to 1000 A.D. Roger Green, 27 therefore, has quite reasonably fixed upon a date of A.D. 900 for the probable beginning of Polynesian history in New Zealand. Thus, the 92% of shared vocabulary between Tahiti and New Zealand, on the 100-word list can become the shared retention rate of 92% for 1900 minus 900, or a thousand years. This is so slightly different from the 91.25% rate which was used in our tables that we need not refigure our calculations in fixing upon A.D. 100 as the time of separation of the Marquesas from Tahiti. A.D. 100 then becomes the theoretical date for the spread from the Marquesas to Tahiti, and our 700 B.C. as the very minimum time for the first settlement of the Marquesas. This would allow for the settlement of Easter Island from the Marquesas even before the time of its earliest radio-carbon date, A.D. 380, and for the settlement of Hawaii from the Marquesas after Easter Island was settled.
On the 200-word list, when its percentages of shared retention are compared with those of the 100-word list, as we have just done with Rarotonga, the agreement between Tahiti and New Zealand of 73% is seen to be abnormally low and results in a glottochronological date of A.D. 314, for the separation between the two, which is much lower than radio-carbon dates indicate. So, if we were to accept for the 200-word list 73% as the shared retention rate for 1,000 years, instead of the 82% which we used and which was based on the Marquesan separation from Tahiti, it would lengthen the time of the separation beyond reasonableness. Therefore, it is best to let our calculations, based on the original estimate of 82% stand.
Because the glottochronological rate that we have worked out cannot apply to Hawaii, how are we going to translate the shared retention percentages? Presumably, if the Hawaiian language had an original beginning through migrants from the Marquesas and was later modified by Tahitian influence, the agreement between Hawaii and Tahiti would be less than it would have been had these Tahitians been the first settlers of Hawaii, hence the change translated into years at our rates, based on change since separation from a single mother area without further contact, will give dates which neither represent the coming of the Tahitians or the time of first settlement. The dates for both will be earlier than they should be. We have seen that using the 100-word list results in a date of A.D. 504 and the 200-word list of A.D. 517 for the spread from Tahiti to Hawaii, and A.D. 267 and A.D. 103, respectively, for the spread from the Marquesas to Hawaii. While our Hawaiian radio-carbon dates for our earliest sites, on South Point, Hawaii, have been giving such erratic results that we cannot trust them and will not know how to evaluate them until we have discovered the causes, we can be fairly sure that at least the island of Hawaii was well settled by A.D. 750, if not earlier, and that fresh Tahitian influence is apparent between the 12th and 14th centuries. Thus our glotto-chronology dates give us what we expect: too early a date for settlement from the Marquesas and too early a date for the coming of - 100 migrants from Tahiti. Does this not amount to evidence that the Hawaiian vocabulary is not purely derived from either Tahiti or the Marquesas, but that its speakers have the multiple derivation as indicated by its archaeology?
Thus we have seen the necessity and profit of integrating findings of comparative language and comparative archaeological research. Each can act as a corrective in interpreting the results of the other.
From the latest findings of archaeology, we see that the Marquesas instead of the Society Islands may have been the centre for the early formation of East Polynesian culture, and this is not incompatible with our vocabulary findings. Both linguistic and archaeological evidence together, support migration to Easter Island before migration to Hawaii and New Zealand, and the branching of Tahitian culture to New Zealand before later Tahitian culture reached Hawaii. The combined findings give a picture of order of dispersal and time illustrated in the figure on page 83. These tentative pictures which are taking shape are certain to be subject to modification as research proceeds.
1 Elbert 1953.
2 Emory 1946.
3 Swadesh 1955.
4 Hymes 1960.
5 Elbert and Pukui 1957.
6 Churchward 1959.
7 Ward 1961.
8 Emory 1946.
9 Churchill 1912.
10 Elbert 1953, p. 159.
11 Emory 1946.
12 Emory 1946.
13 Figures in parentheses are the number of words scored as 1/2.
14 Elbert 1953, p. 159.
15 Emory 1946, pp. 181-208, 256-266.
16 Emory 1933.
17 Ward 1961.
18 Emory 1946, Table 7, p. 182.
19 Shapiro, H. L. and R. C. Suggs 1959.
20 Elbert 1953
21 Capell 1962, p. 398.
22 Swadesh 1955, p. 123.
23 Heyerdahl and Ferdon 1061:394.
24 Emory 1957.
25 Suggs 1961:92
26 Figures in parentheses indicate the number of words scored as a half.
27 Green and Shawcross 1962:219.