- 33

The island of Erromango in the southern part of Vanuatu possibly has the dubious honour of having suffered the greatest amount of linguistic devastation in the region of Oceania outside of Australia. Depending on how many languages were originally spoken on the island, somewhere between five-sixths and two-thirds of the original number of languages are now extinct, or very nearly so.

These drastic changes in the linguistic ecology of the island have taken place within the last century and a half. In the crucial early stages of this process, a number of European observers recorded some aspects of relevant historical circumstances, while present-day oral tradition provides some previously unrecorded information about the linguistic past as well. Those early European observers have also provided us with, in some cases, the only linguistic information that we are ever likely to have about some of these now extinct speech varieties.

This article addresses a number of more general issues in its historical account of the evolving linguistic demography of the island. First, it aims to document as fully as possible, on the basis of surviving linguistic evidence, oral tradition and documented historical sources, what the pre-contact linguistic situation on Erromango was like. This discussion deals with a number of methodological issues in relation to the linguistic interpretation of sources that were not intended specifically for linguistic exegesis. Second, it attempts to fit the reconstructed linguistic developments of the last century and a half into a more general framework of language contact and linguistic speciation. Third, it also touches tangentially on the issue of ethnic identification and the expression of identity through the use of language names. The article examines the issue of why the modern Erromangan language no longer has a generally known name, despite the earlier proliferation of named speech-forms.

Erromango is the third largest island in Vanuatu. Its current population is only about that of the more northerly island of Paama, though Erromango is about thirty times the geographical size of Paama. According to the 1989 census, the total population of Erromango was about 1250. There is essentially a single language spoken on the island today, which I will refer to for the time being as Erromangan. At the same time, there is a residual group of about half a dozen people in their seventies living in Unpogkor - 34 (Dillon's Bay) and Potnarvin who speak, in addition to Erromangan, a language known as Ura, which was once more widely spoken in the north. Map 1 locates the places just mentioned, as well as all other places referred to in the text of this paper.

About the only thing that we can say with absolute certainty about the linguistic and demographic history of Erromango is that there was once a significantly larger population, and that there was once significantly more linguistic diversity on the island than at present. Lynch (1983a) and Taki and Tryon (1994) refer to the massive depopulation that affected the island in the second half of the 19th century. From an estimated mid-19th century population of up to 6000, the nadir was reached in 1931, when the population was only 381. Since then, the population curve has shown a normal demographic increase, and there are now about as many Erromangans as there were at the turn of the century. The original population estimate of up to 6000 for the island is certainly plausible as the island is large and lush. Nearby Tanna has just over half the land area of Erromango, yet today it supports a population of about 20,000. 1

The initial reported population loss was extremely rapid. Robertson (1902:72-73) indicates that in January 1861 measles was introduced from a recently arrived ship and the disease immediately spread throughout the island with a high mortality rate. The weakened population was then struck by a cyclone later in the same month, and several other cyclones struck again in March. Within just a few months, about a third of the entire population had died of disease and hunger. Other epidemics of dysentery, diphtheria, influenza and venereal diseases saw large numbers of people die, and birth rates dropped. 2

Such massive demographic changes must have meant considerable disruption to traditional lifestyles. There is little information in historical sources about how this affected traditional residence patterns, 3 but oral tradition on Erromango paints a picture of villages repeatedly becoming too small to be viable and the remaining groups of people moving to other locations to reconstitute larger villages. This period of depopulation coincided with the beginning of the operation of the Presbyterian mission on the island. One of the policies of Christian missions in Vanuatu was to amalgamate the populations of small hamlets into larger villages which could be serviced by a single church. The island is now covered with named sites which represent abandoned 19th-century village locations. The interior of the island—much of which was presumably once occupied, as we now find on Tanna—is now almost completely devoid of villages.

- 35

Map 1: Erromango and its location within Vanuatu

- 36

We have a number of documented observations from recorders of the linguistic situation on Erromango from a time when the island was linguistically more diverse than at present.

The published source of earliest provenance is the 1861 account of a two-day visit by the Rev. George Turner to the Rev. George Gordon, who was living on Erromango in 1859. Turner (1861:488) observed:

Another visitor to Gordon shortly afterwards, A.W. Murray, published an account of his visit in which he had this to say about the linguistic situation on the island:

These early quotes refer to visits which took place before the first reported measles epidemic mentioned above. The interesting point is that there is already seemingly contradictory evidence being presented in these sources about the linguistic situation on the island at the time, as one writer suggests the existence of a second language that was not hinted at by the other.

George Gordon and his wife Ellen were killed on Erromango in 1861. In 1864, George's brother, James, arrived on the island to continue his brother's work. He wrote a brief account of the linguistic situation on the island shortly before he too was killed on the island in 1872. This was published posthumously in 1889 in a volume edited by his missionary colleague from Efate, the Rev. D. MacDonald.

Gordon (1889:78) refers to a number of “Eromangan dialects”, though from the examples that he provides of some of these speech-forms, and from later published comments by other observers, it is clear that these “dialect” names involved a number of mutually unintelligible speech-forms. The names of the speech communities that he recorded, along with his estimate of the number of people belonging to each at the time, were:

- 37

Some of these language names appear to have had meanings as independent words in these languages. In Yocu/Enyau and Sye/Sorug, the first person singular possessive forms were indicated by Gordon as having been enyau and sorug respectively. There is evidence—presented below— that first person singular possessive forms were used as alternative names for other speech varieties as well. The name Sye as an alternative to Sorug appears to be identical to an interrogative word of the form sye ‘what’.

Gordon's published account does not give any indication as to the location of these speech communities. Given that the speech-form that had been adopted from the outset for mission use was Yocu/Enyau, we can probably safely assume that this was the language of the area where his brother had previously begun the task of Bible translation that he continued, i.e., Unpogkor (Dillon's Bay). Although James Gordon eventually ended up living (and being killed) near Potnarvin on the other side of the island, his first years on Erromango were spent at Unpogkor, where he learned the language into which the New Testament was eventually translated (Miller 1981:57-59).

Since Gordon wrote his sketch shortly before he was killed in 1872, it is clear that the situation that he was describing was that which pertained after the measles epidemic and cyclone-induced famine of the early 1860s. The precise extent of depopulation by 1872 is difficult to ascertain, though the figures assembled by McArthur and Yaxley (1968:4) suggest that the population had dropped from as high as 6000 to between 1000 and 2000.

Since the populations for Gordon's various speech communities at that time totalled 3550, it is necessary to account for the apparent contradiction between this figure and the lower total population indicated by McArthur and Yaxley. One possible explanation is that Gordon may have been enumerating people who spoke more than one language twice. Given that Gordon's largest speech community claimed 2000 speakers, it would be possible to reconcile the two sets of figures if we were to assume the following demographic profile according to speaking ability:

According to this interpretation, one of the Erromangan languages would have been universally known as a second language among speakers of all of the other languages, with a minority of the population being monolingual in the lingua franca.

- 38

whatever the precise population of the various speech communities of Erromango, it is clear that there had already been major depopulation and linguistic realignment on the island by the 1870s. The Novulamleg speech community had already disappeared completely, though the existence of this as a name for a speech-form is confirmed by modern oral tradition of some elderly people who originate from northern Erromango.

Utaha was clearly moribund by the early 1870s. With only 50 speakers, it was probably only spoken in a single village, or more likely by small numbers of people scattered throughout a number of villages. Matthew Spriggs (pers. comm.) informs me that on the basis of information supplied by Willy Tovovur, an elderly Ura speaker, the last speaker of Utaha was James Nalig, who died in 1954. Given that he was then reported to be in his eighties, he was probably born during the period 1870-80. He could therefore well have been one of the remaining Utaha-speakers enumerated by Gordon. 5 Some of the remaining speakers of Ura retain a knowledge of a few words and phrases of Utaha, and Jerry Taki of the Vanuatu Cultural Centre is in the process of recording as much of this vocabulary as can be remembered.

There is no way of knowing if signs of the ultimate fate of Ura were already manifest in the situation that Gordon was describing, apart from its relatively small number of speakers. There are many viable languages in Vanuatu today with similarly small populations, so the small size of the speech community cannot, by itself, be taken as indicative of shift in progress. The final shift away from Ura must have taken place from the 1920s, given that the present group of elderly speakers are in their seventies. We could therefore guess that by the 1870s at least some Ura families were probably already producing non-Ura-speaking children.

Inglis was based on Aneityum but visited James Gordon at Dillon's Bay. Robertson (1902:127) printed an account written by Inglis of the arrival of James Gordon at Dillon's Bay in 1864. This includes mention of the following incident in which the man responsible for the murder of Gordon's brother and sister-in-law was present to welcome Gordon to the island:

- 39

This extract makes no explicit reference to the linguistic make-up of the island, though it does offer some material for consideration. The Mr. Ella referred to here was a European who operated the missionary printing press on Aneityum, who clearly had a knowledge of Samoan. Mana was a local Erromangan convert who spoke the language of Dillon's Bay, i.e., Yocu/Enyau. The impudent fellow who had murdered Gordon's family members was a Sye/Sorug speaker from the south, and he presumably received his telling-off in Yocu/Enyau.

Several different interpretations of this statement are possible. (1) The impudent Sye/Sorug-speaking fellow may simply have not understood his telling off at all when it was given to him in Yocu/Enyau. This seems unlikely, or the writer would presumably have commented on his bemusement. (2) Possibly Yocu/Enyau and Sye/Sorug were linguistically similar enough to be considered as mutually intelligible dialects of a single language. (3) If Yocu/Enyau and Sye/Sorug were not mutually comprehensible, then at least some Sye/Sorug speakers (and perhaps many, or even all) were bilingual in Yocu/Enyau. (4) There may have been a tradition of dual-lingualism on Erromango—such as we find on Tanna today—in which people could speak their own language and be passively understood by speakers of other languages. (5) The impudent fellow spoke only Sye/Sorug, but Mana was bilingual in Yocu/Enyau and Sye/Sorug.

Unfortunately, from the context in which this statement was drawn, it is not possible to decide which of these interpretations is the most likely.

The Rev. Hugh Robertson, James Gordon's successor after his murder, spent 30 years on the island continuing the missionary work of his predecessors. In 1902, he published a history of the Presbyterian mission on the island, in which he likened Erromango to the linguistically diverse island of Tanna—which linguists now regard as having five distinct languages—rather than to monolingual Aneityum: 6

The extent of the linguistic diversity on Erromango is hinted at by the following discussion of his murdered predecessor, James Gordon.

- 40

Since Yocu/Enyau was the language that was adopted by the Presbyterian mission on Erromango for evangelical purposes, it is likely that by the late 1800s, this language was spoken—or at least understood—by the entire population. Although there were smaller populations of people reported to speak Sye/Sorug and Ura, they were probably largely bilingual in Yocu/Enyau, though we cannot know if this involved a passive or an active knowledge.

The earlier quote by Turner referred to above is consistent with an interpretation that a single language was used as a lingua franca from the very beginning of the missionary period (in addition to a number of other local languages, to which he did not refer). If this interpretation is correct, then the replacement of smaller languages by Yocu/Enyau is something that may have commenced even before the arrival of the missionaries. Thus, the missionaries may simply have promoted the development of a process that was already underway before their arrival.

There is, however, one significant contradiction between what Gordon had said about the distribution of languages on the island and Robertson's later comments. I interpreted Gordon's words earlier as possibly meaning that Sye/Sorug was used as a lingua franca, while Yocu/Enyau was used only in a particular area. Robertson, on the other hand, says that Sye/Sorug (and Ura) were understood by only “a few”, while Yocu/Enyau was “known all over the island”. There are several possible explanations for this discrepancy. (1) Gordon had simply switched the names and the population figures for Sye/Sorug and Yocu/Enyau. In endnote 4, I indicated that the published version of Gordon's notes contained a number of obvious errors, and this might have been just one more. (2) Gordon may have simply overestimated the population of Sye/Sorug speakers in the first place, as I have already mentioned. Sye/Sorug was spoken in the south, which is the part of the island that he was least familiar with, so his figures from that area were perhaps more likely to be subject to error. (3) In the intervening period, epidemics may have affected the south more severely than the north. This is a plausible suggestion as the mission station provided health care facilities in the north, which may have reduced mortality there. The unchristianised south had no direct access to such facilities. (4) In the 30 or so years since Gordon's sketch was written and Robertson's comments were published, a - 41 major language shift may have taken place, with Sye/Sorug speakers shifting massively to Yocu/Enyau, in the same way that Utaha and Ura speakers had already shifted, or begun to shift, to Yocu/Enyau. (5) Yocu/Enyau and Sye/Sorug may have “merged” as speech communities, with people simply redesignating themselves as speakers of Yocu/Enyau rather than Sye/Sorug over the intervening period.

Humphreys (1926) was writing around the time that the population had reached its nadir, and there were possibly genuine grounds at that time for fearing that the Erromangan people might disappear completely. The original population of up to 6000 had fallen to less than 400, and Humphreys repeatedly described the Erromangans in the 1920s as a dispirited people who had suffered massive disruption to their traditional lives.

Although he was very much a latecomer to the linguistic make-up of the island, any elderly people he spoke to almost certainly had been born before the death of James Gordon in 1872, and possibly even before the death of James' brother and sister-in-law in 1861. Thus, while they would not necessarily have had firsthand knowledge of the pre-contact distribution of languages, they could well have had direct information about the situation at the time that Gordon was writing. On the basis of information provided by such people, Humphreys made the following observations:

Humphreys' Eniau and Sorung clearly represent Enyau and Sorug respectively. The Utaha and Ura first person singular possessive pronouns are etiyo and aryau respectively, which presumably represent Humphreys' Etio and Adiau, based on the analogy of Enyau and Sorug having been named after their first person singular pronominal forms. Thus, Etiyo would appear to be an alternative name for Utaha, while Aryau represents an alternative name for Ura. I have not been able to find any independent confirmation for the existence of Humphreys' speech-forms Seimo and Tanempenum.

- 42

Capell conducted some linguistic fieldwork on Erromango in 1958. He published no account of the language, though he did produce a manuscript grammar (Capell n.d.), as well as a short published article on phonology (Capell 1972). The language that he generally referred to as Eromangan is the same language that earlier sources refer to alternately as Enyau or Yocu. His discussion of the linguistic geography of the island appears to contain no observations based on his own fieldwork, and all of his statements are simply citations from the published work of von der Gabelentz (1861, 1873), who also had no direct contact in the field, with his conclusions being based simply on his own interpretation of written sources available to him at the time.

Lynch (1983a) indicates that speakers who he interviewed in the 1970s used Sie—more correctly Sye—as the name for the modern Erromangan language in contrast to the moribund Ura. In the late 1970s, William Mete from Unpogkor recorded a tape which provided a couple of hundred Ura words with their equivalents in the modern Erromangan language, which he referred to as Sye. 9 William Mete referred on that tape to a third—and extinct—Erromangan language, Utaha, which had already been referred to by Gordon.

The situation today has evolved somewhat beyond that which was described by Lynch (1983a). As the moribund status of Ura has become even more apparent, the need for a distinct name for modern Erromangan appears to have diminished. The name Sye is now little more than an archaic name for the modern Erromangan language, and has largely disappeared from everyday usage. The great majority of Erromangans today are in fact surprised to hear that their language had once been called Sye (or, indeed, anything at all). People on Erromango today generally refer to their language when speaking to each other simply as nam Eromago ‘Erromangan language’, or as nam horet ‘our (inclusive) language’, and to outsiders, in Bislama, as lanwis Eromango ‘Erromangan language’.

The name Ura survives, and it is widely known that this is the name of an originally larger language once spoken in the north of the island. Apart from this, during my own fieldwork on the island, I did hear some people speak occasionally of Yocu, Horug 10 and Enyau as older names for speech-forms. Nobody had any firm idea about the linguistic grouping or the geographical provenance of Yocu, while Horug and Enyau are generally associated with the eastern and western areas of the southern part of the island respectively.

The most detailed information of all about the linguistic geography of the northern part of Erromango derives from interviews conducted by Matthew Spriggs with Willy Tovovur in 1983 in conjunction with his - 43 archaeological fieldwork. According to this information, Ura was one of three mutually intelligible speech-forms spoken in the north, along with Uravat and Novulamleg. These names appeared in print in Spriggs and Wickler (1989:86). 11 The name Uravat does not appear in any of the earlier sources. It may be that this is the additional unnamed speech community referred to by Humphreys, though this would be nothing more than a supposition.

Taking into account all of the available sources referred to above, there are 12 names recorded for earlier speech-forms on Erromango: Sorug/Horug, Sye, Enyau, Utaha, Ura, Novulamleg, Uravat, Aryau, Etiyo, Yocu, Seimo and Tanempenum. As has already been demonstrated, the various primary sources offer plenty of evidence for the existence of earlier linguistic diversity. However, there are real problems in deciding the actual extent of original diversity on Erromango. The main question that needs to be answered is: how many mutually unintelligible speech-forms—i.e., “languages”—are represented by this diversity of names?

Defining the boundary between “language” and “dialect” is sometimes difficult enough when we have complete information, but the situation in the case of Erromango is made rather more complicated because there is considerable variability in the sources about how these two terms are used.

The frequent use of the term “dialect” rather than “language” in older sources should not be taken as meaning that two different speech-forms necessarily differed only dialectally as modern linguists would understand the term. It has long been common practice to refer to the small languages of pre-literate and technologically unsophisticated people as dialects rather than languages, even when there was clearly no mutual intelligibility. In any case, Robertson (1902) alternates in his account between the terms “dialect” and “language”. The practice of overgeneralising the use of the term “dialect” still continues to some extent among laypersons today.

In contrast, the local understanding of the term “language” tends to overemphasise dialect differences as language differences. People in Vanuatu—indeed in Melanesia in general—often extend the use of the term “language”, using it to refer even to mutually intelligible speech-forms. The distinction between “dialect” and “language” is one that is not usually lexically encoded in either Bislama or Melanesian vernaculars. In Bislama, for example, it is reasonably common for people to refer to mutually intelligible dialects of the same language simply as being different toktok or lanwis.

- 44

These facts mean that the terms “dialect” and “language” in all sources— whether published or derived from oral tradition from present-day Erromangans—should simply be equated with a term such as “speech-form”, which is indeterminate with regard to dialect or language status. Any decision as to the status of particular pairs of speech-forms as constituting dialects of a single language or separate languages should be made on the basis of observations of relevant sociolinguistic data.

The various primary sources referred to earlier provide plenty of evidence of apparent contradiction, both in terms of the extent of earlier linguistic diversity and in the populations of earlier speech communities. This contradictory evidence can only be resolved by making some kind of informed interpretation of these sources.

For example, it will be remembered from the earlier quotes that Turner pointed to the existence of one speech-form on the island in the early 1860s, while Murray referred to two. While these two quotes are seemingly contradictory, Turner does not actually deny the possibility of there being additional speech-forms. He could just as easily have been saying that if there were several speech-forms on the island, there was one that was known by everybody. Presumably, the speakers of any other speech-forms could have been bilingual in the speech-form that had become the basis for the work of the missionaries, as I suggested above.

Murray's comment that there were two speech-forms on the island appears to represent a contradiction with Gordon's and Robertson's indications of a greater amount of linguistic diversity. However, Murray derived his information directly from Gordon himself, and he only visited Erromango for a very short period. He may therefore simply have been summarising (or even misrepresenting) what Gordon had told him about other speech-forms that were either extinct at the time, or at least moribund. He had no direct information of his own about the local situation, so we should give greater weight to the words of observers such as Gordon and Robertson, who spent years, rather than days, on the island.

We should probably also place greater reliance on observations of greater antiquity than observations based on more recent information. The information provided in Humphreys (1926), as well as any information that comes from any modern Erromangans, derives from a time when the original linguistic situation had already been massively disrupted. While it would be unwise to completely ignore such information, it should certainly be weighed carefully against earlier sources.

- 45

I will now now turn my attention to interpeting what Gordon and Robertson had to say about the original extent of linguistic diversity on Erromango. Gordon (1889:80) comments on the difference between Enyau (which he indicates was alternatively known as Yocu) and Sye (or Sorug, which is now reflected as Horug) in the following words: “The changes from Yoku to Sie are slight.”

This, in itself, of course, does not allow us to conclude whether we are simply dealing with differently named dialects of a single language, or two different —but presumably closely related—languages.

On a priori grounds, we could perhaps argue that because these two speech-forms had different names implies that they were different languages. However, there are areas in Vanuatu where speech-forms that are similar enough to be treated as dialects of a single language are named separately, such as some parts of southern Malakula (Charpentier 1982).

The question that now arises is how distinct the various named speech-forms actually were on Erromango. For some of these speech-forms we have more documented information about their linguistic nature than we have for others. Only the most fragmentary of materials were recorded from Sye/Sorug and Utaha. In the case of Sye/Sorug, hardly more than a dozen lexical items were recorded, while for Utaha, about three dozen words were recorded (though, as I mentioned above, some additional lexical material in Utaha has been recorded by Jerry Taki, but this has not yet been published). We also have a published possessive pronominal paradigm for Sye/Sorug. For Utaha, we have a parallel possessive paradigm, as well as the paradigms for the independent pronouns, the dative forms of pronouns, and the numbers 1-10. For Ura, we are in a somewhat more fortunate position. Gordon (1889) provided parallel paradigms to those that he had given for Utaha, as well as just over three dozen lexical items. Lynch (1983b) provides a considerably larger vocabulary. Combined with unpublished material that has been gathered recently by Jerry Taki and myself, we have a total lexicon of about a 1000 words in this language, as well as a fair amount of structural information (Taki and Crowley 1996).

For Yocu/Enyau in these early sources, we are also quite well off. Gordon's sketch provides a fuller range of grammatical information than he gives for Ura, Utaha and Sye/Sorug, and he also gives just under a 100 lexical items from Yocu/Enyau. This number can be supplemented by about 150 lexical items supplied by Robertson scattered throughout the text of his book, as well as a fair number of words provided by Humphreys (1926).

Lynch (1983a:4-5) presents the following lexicostatistical comparisons on the basis of existing materials (including all sources for Ura, and regarding - 46 sources for modern Erromangan as directly reflecting what is referred to in the earlier sources as Yocu/Enyau):

On the basis of these figures, it is obvious that Ura, Utaha and Enyau constituted three separate languages, in that they are all well below the usual cut-off point of 80 percent cognate-sharing for dialects of the same language. Willy Tovovur, in his discussions with Matthew Spriggs in 1983, also indicated that these names represented three mutually unintelligible speech-forms. William Mete, in his tape of 1979, also referred to three separate languages having formerly been spoken on Erromango: Ura, Utaha and Sye (which is the modern derivative of Enyau).

These conclusions about the separate linguistic status of Ura, Utaha and Enyau are consistent with what grammatical information we have on these speech-forms. From the information in Gordon 1889, Lynch 1983b, and Taki and Crowley 1996, it is possible to deduce the following paradigms for the independent pronouns and the corresponding dative forms for the three named speech-forms, and these exhibit the range of variation that one might expect for three different languages: 12

Lynch (1983b) includes a 14-page sketch of Ura, which was based on a short period of fieldwork with one speaker, as well as Capell's notes from - 47 his fieldwork in 1958, and the work of Gordon. Taki and Crowley (1996) also include a number of sentences in Ura, with their equivalents in modern Erromangan (i.e., what was known in the 19th century as Enyau). These data are all consistent with the conclusion that Ura is grammatically quite a different language from Enyau. In 1996 I recorded a short text in Ura from Ukai Ndaleg of Potnarvin, and when this was played to speakers of modern Erromangan, it was not understandable, which further emphasises the separate linguistic status of Ura

Because of the highly restricted amount of information on Sye/Sorug, it is more difficult to be definitive about its status vis-à-vis Yocu/Enyau. The following represents an exhaustive lexical comparison between 19th century sources for these two speech-forms (with forms presented in this case in the spelling of the original source rather than in phonemic form):

Allowing for the possibility of hū being a misprint for hī, representing therefore hai according to Gordon's orthographic principles, these lists are similar enough that the data could be interpreted as coming from two dialects of a single language. At the same time, however, it would be not too difficult to find 13 words from any pair of closely related languages which would allow such a conclusion to be drawn.

Even in cases where there are differences between the two, it appears that Gordon may have exaggerated the extent of the difference by ignoring variants within Yocu/Enyau when they happened to look more like Sye/Sorug. When he was explicitly contrasting the two speech-forms, he compared the Yocu/Enyau form that he wrote as tiē ‘what’ with the Sye/Sorug form that he - 48 wrote as siē. Elsewhere in his discussion, however, the form for ‘what’ in Yocu/Enyau was given as siē, which is exactly the same as the Sye/Sorug form.

A possible explanation for this is that both siē and tiē may have been used in Yocu/Enyau, with the former being the dominant variant. 14 In order to maximise contrast with Sye/Sorug, Gordon may have chosen to present tiē, even though it was possibly less frequent. Similarly, he explicitly contrasts the Yocu/Enyau form sukrim ‘five’ with the Sye/Sorug form of the same meaning, siklim. However, in the 1864 translation of Luke into Yocu/Enyau, he used the form siklim. Thus, both siklim and sukrim may have been used in Yocu/Enyau as variants, with the more divergent form being chosen again to emphasise linguistic differences that may have been tendencies rather than absolute differences.

There was clearly one significant phonological difference between Yocu/Enyau and Sye/Sorug. On the basis of modern distributions, it is possible to argue that a change from s to h had probably begun to spread in Sye/Sorug by the 1870s, while in 19th century Yocu/Enyau, original s was retained intact. However, in Sye/Sorug, s and h had apparently not fully acquired their modern distribution, as the possessive markers from the paradigm of which sorug ‘my’ was an exponent had not yet been affected. 15

The only grammatical information that is comparable between the four named speech-forms are the possessive paradigms. On the basis of information provided by Gordon (1889), Lynch (1983b) and Taki and Crowley (1996), the following forms can be deduced:

The fact that Sye/Sorug and Yocu/Enyau are recorded as having had distinct paradigms here is consistent with a conclusion that they may have been separate languages.

Gordon states that Yocu/Enyau possessive forms were based on a paradigm of which enyau was the first person singular exponent, while Sye/Sorug possessive forms were based on a paradigm of which sorug was the form - 49 meaning ‘my’, hence the language names. However, Gordon's examples indicate that, at least in Yocu/Enyau, forms from both paradigms were used in alternation.

On the basis of the linguistic data alone, as presented above, it is impossible to decide whether Yocu/Enyau and Sye/Sorug were dialects of a single language or separate languages. It is quite conceivable for a single language to have dialects that differ lexically to the extent illustrated above, for there to be a correspondence between s and h in lexical items, and for there to be a difference in possessive forms. The only way that we can conclude if we are dealing here with two languages or one is to allow the linguistic commentaries of firsthand observers to decide. If it emerges that we are dealing with two languages, they were certainly more closely related to each other than either is to both Ura and Utaha.

Gordon (1889:80) stated that the differences between the Yocu/Enyau and Sye/Sorug were “slight”, though there was of course no indication of how slight the differences were. Robertson (1902:136) said that “the sorūng and the ura … are only understood by a few”, which should probably be taken as indicating that both of these speech-forms were mutually unintelligible with Yocu/Enyau. Although he had become proficient in Yocu/Enyau, Robertson (1902:135-36) explicitly stated that he had never learned either Ura or Sorug, which again suggests that both were sufficiently different to have constituted separate languages. All of the observations just mentioned are consistent with the conclusion that Yocu/Enyau and Sye/Sorug were separate languages, though they fall short of constituting definitive proof.

The remaining lect-names need to be reconciled with the linguistic entities that I have described above, i.e., Ura, Utaha, Enyau and Sorug. As I have just indicated, some speech communities were referred to by stereotypical features with which they are associated. In addition to the use of first person singular possessive pronouns as lect-names, the name Sye appears to derive from the exclusive use of sye ‘what’ in Sye/Sorug, in contrast to the variable use of tye/sye in Yocu/Enyau for the same meaning. I noted above that Etiyo was apparently an alternative name for Utaha, while Aryau was an alternative for Ura.

Finally, there are a number of references to named speech-forms for which we have no linguistic information whatsoever, so we are not in any position to draw any conclusions about the status of these entities. These mysterylects are Humphreys' Seimo and Tanempenum. We have no way of knowing if these are simply dialectal variants of established languages, four separate additional languages, or alternative names for a smaller number of separate languages. It is also possible that these were not language names at all, but names of places or descent groups.

- 50

Taki and Tryon (1994) report Uravat as being “almost certainly” a sixth language. However, no evidence was presented for upgrading Spriggs and Wickler's more circumspect status for Uravat as a “speech-form” (and Spriggs' own fieldnotes clearly indicate that Willy Tovovur regards Ura and Uravat as having been mutually intelligible speech-forms, along with Novulamleg). Lynch (pers. comm.) suggests a possibility that -vat may be cognate with a Tannese morpheme meaning ‘shorewards/eastwards’. If this is correct, the name Uravat could simply have been a descriptive term for an eastern variety of Ura.

Novul is recorded as a word for ‘language’ in Ura (Lynch 1983b:173). Emlag means ‘lost’ in modern Erromangan, and this may be cognate with amleg. Novulamleg may be a name for an additional dialect of Ura which is etymologically unrelated to the word emlag ‘lost’, or it may simply have been a descriptive reference to some lost speech-form that people had once heard about, but which was already extinct by Gordon's time.

In summary then, the linguistic situation on Erromango as far as it can be established at the time of initial European contact appears to have been as represented schematically in Figure 1 (with references to the unverifiable entities referred to by Humphreys being ignored). Solid lines indicate definite language boundaries, while dotted lines indicate boundaries that are indeterminate with respect to the difference between language and dialect on the basis of the information that we have.

Figure 1: Early contact-era linguistic diversity on Erromango

- 51

The discussion in the preceding sections has made some reference to the geographical provenance of a number of speech-forms, though I have not yet discussed this issue in detail. The only published early source which says anything at all about the geographical location of any speech community is Murray's (1863) brief reference to a separate (and relatively small) language spoken in the northeast of the island. However, this source does not provide names for either of these speech-forms.

We can draw the inference that because the early missionaries established themselves at Unpogkor, the Yocu/Enyau language that they adopted as their medium was the language of this area. Modern Erromangans associate the now moribund Ura language with the northernmost part of the island (which is now sparsely populated), so it is conceivable that Ura was already showing signs of becoming moribund by the time of Murray's visit, and that this was the second language to which he was referring.

Lynch (1983a:5) indicates that the most detailed early observations about the distribution of languages on the island comes from an unpublished manuscript which appears to have been written either by the Rev. James Gordon or by his wife before he was killed in 1872. 16 He extracted the following information from this source (Lynch 1983a:5):

• (a) Eniau (Yoku): spoken in Dillon's Bay and extending about eight miles north and ten miles south of Dillon's Bay; it intermingles with Utaha on the north and Sorung on the south and south-east.
• (b) Sorung: spoken between Unepang, twelve miles south of Dillon's Bay, and Traitor's Head, a distance of about fifty miles; it intermixes with Eniau on the one hand and Ura on the other.
• (c) Ura: spoken from Portinia Bay to Uarringrie [Warigi Bay], about twentyfive miles more or to where it mixes with Eniau.
• (d) Utaha: spoken by a small tribe about five miles north of Dillon's Bay, but even then all but extinct.

According to information contained in Spriggs (1983), Willy Tovovur provided the following specific information about the location of the northern speech-forms on the basis of modern oral tradition:

• (a) Ura: Elizabeth Bay to the western side of Ponusia.
• (b) Utaha: from Umrup Point to Tugomle (i.e. the river at the southern end of Elizabeth Bay), and inland to Unavwolu and the area known as “Whitegrass”.
• (c) Uravat: from Sumprin to Potnuma and inland up the Potnuma River into the large volcanic caldera in the interior.
• (d) Novulamleg: starting on the western side of Ponusia and ending just north of the river at Sumprin.

- 52

Map 2: Reconstructed linguistic boundaries on Erromango at initial European contact.

Map 2 presents the locations of these original speech-forms according to the combined information just presented. Boundaries along the coast, which can be considered definitive, are marked by means of solid lines. Boundaries in the interior are inferred from either current landholding boundaries or geographical probability, and these are indicated instead by dotted lines.

Interestingly, the notes cited by Lynch (1983a) generally refer not to definitive boundaries as such. Rather, the use of the terms such as “mixes” and “intermixes” suggests that there were gradual rather than abrupt transitions between speech-forms. Such a situation would not be too surprising, of course, as this is not too dissimilar from what we find on Tanna today, as well as in many other parts of Melanesia. Obviously, the details of such linguistic transitions can never be recovered.

Modern Erromangan is clearly quite distinct from both Ura and Utaha. It is therefore the relationship between 19th century Enyau and Sorug and modern Erromangan that requires particular attention in determining the source of the modern language.

- 53

Modern Erromangan emerged somehow out of 19th century Yocu/Enyau and Sye/Sorug at a time when the population was falling rapidly. Under these sorts of circumstances, there was considerable amalgamation of villages as depopulated villages became unviable, with some movement of people between the original Yocu/Enyau- and Sye/Sorug-speaking areas as people relocated.

We have no explicit demographic statement about the distribution of the population of Erromango at the time that its population was close to its nadir. However, we do have a number of brief observations by Humphreys (1926:123) from which we can deduce some demographic information. The figures presented by McArthur and Yaxley (1968:4) indicate that the total population at the time that Humphreys was making his observations was in the high 400s or low 500s. Humphreys carried out cranial measurements on the following numbers of adult males in a variety of locations, as follows:

His total sample of adult males on the island comes to 59, which was apparently close to exhaustive. He reported a “preponderance” of females in all villages that he visited. Assuming that there was a total population at the time of about 500, these 59 adult males would have been outnumbered more than 7 to 1 by women and children. Erromango was clearly seriously “undermanned” at the time. On the island today—which reflects a more normal sort of demographic distribution—men are outnumbered by women and children by less than 3.5 to 1 (Statistics Office 1991).

Humphreys indicated that the only large settlement on the island at the time was Unpogkor (Dillon's Bay). Extrapolating from his figures on the basis of the ratio just mentioned, this village may have had a population of about 185 out of the total Erromangan population of about 500. He reported that settlement in the area of Unpag and the south was dispersed into smaller villages. There were also two smaller villages located in the north of the island.

Apart from the disproportionately small population of adult males in Humphreys' time, the distribution of the population showed a number of similarities with the present-day pattern. Unpogkor is still the largest settlement on the island, though instead of accounting for just over a third - 54 of the total population as it did then, it now contains just under a quarter of the population.

Thus, as the population of the island has increased, it has become geographically more dispersed. As in Humphreys' time, there is a scattered and very small population to the north of Unpogkor. In the area known as Unpag to the south, there is now a large number of small and geographically scattered hamlets, located mostly along the top of the coastal escarpment, as well as a few larger villages. Finally, the lack, or near lack, of an inland population in Humphreys' time directly mirrors the situation that we find today.

The one major difference between Humphreys' account and the modern situation is that there was no mention of any significant population at all in the east. In this area today, we find the two quite large villages of Potnarvin and Potac (Ipota). In the 1920s, however, it appears from Humphreys' account that the east and the inland areas were only sparsely populated.

The main lexical differences between modern spoken Erromangan and the earliest materials produced by Gordon in the early 1870s are the following:

In addition, Modern Erromangan has h in many words corresponding exclusively to s in 19th century Yocu/Enyau.

I do not propose to discuss the source of the lexical items in this list. It is easy enough for such forms to be transferred from one language to another, so I do not see any great need to try to picture scenarios which would have enabled these kinds of lexical developments to take place. In any case, it is quite possible that the original sources have oversimplified the situation as it was in the 1870s with respect to the distribution of these lexical items, as I have already indicated was the case with the items sye/tye ‘what’ and sukrim/siklim ‘five’.

What is of greater interest is the source of the modern distribution of h, and the distribution of the modern possessive paradigms. With regard to the spread of h in modern Erromangan, it is possible to make a case that this change was already under way in Sye/Sorug in Gordon's time, but that - 55 speakers of Yocu/Enyau in Dillon's Bay were probably largely quarantined from any influence from this change because Sye/Sorug speakers had not yet converted to Christianity. This change apparently only began to take off at Dillon's Bay after the 1920s, about half a century later than in the south.

Speakers of Sye/Sorug were converted to Christianity during Robertson's time on the island, i.e., between the 1870s and the turn of the century. This would have opened up the northerners to contact with the southerners. This was also a time of continued depopulation. Unpogkor (Dillon's Bay) did not represent the site of a traditional village. The illustration in Turner (1861:facing 486) clearly indicates that there was no village at the site of modern Unpogkor at the time of his visit. The demographic picture painted above suggests that there was probably a later agglomeration of the population at the mission station at Dillon's Bay, which drew people from inland areas, the east, and probably to a lesser extent also the south, a view which is consistent with local oral tradition today. Thus, there would have been a considerable influx of Sye/Sorug-speakers into the relatively large Yocu/Enyau-speaking village at Unpogkor.

If Yocu/Enyau and Sye/Sorug were separate languages, there must have been substantial sharing in their lexicons for the change from s to h to spread from one to the other as quickly as it did. If one speech community has nahiven as the word for ‘woman’ and the other has nasiven, it is easy enough to imagine one pronunciation influencing the other, and for the h to spread into the same structural position in the speech community where the word was originally nasiven. When there are larger numbers of non-cognate forms, it would presumably be more difficult for such a change to diffuse generally through the lexicon from one speech-form to another.

Whatever the precise details, h continued to spread in Sye/Sorug, to the point where the possessive paradigm was also affected, as well as a significant proportion of the general lexicon. After the 1920s, the same change began to infiltrate through the vocabulary of Yocu/Enyau, such that the two speech-forms have to a significant extent converged on this phonological point. There are now few noticeable differences between the distribution of s and h between Unpogkor and other parts of the island. Thus, Unpogkor appears to have caught up with the rest of the island in terms of the distribution of s and h in the intervening period.

The other major difference between Yocu/Enyau and Sye/Sorug in the 19th century involved the distribution of the two possessive paradigms. It is also possible to argue that the opening up of contact between the south and the north, along with substantial population realignments that accompanied continued depopulation, could easily have triggered an influx of southern possessive forms into the north.

- 56

I would be the first to concede that this article contains more than an average amount of speculation. However, unless there are remarkable discoveries from hitherto unseen papers of the 19th century missionaries on Erromango, we will probably never know any more than I have presented here. I have tried to clearly indicate what represents speculation and what represents fact.

That there were three distinct languages on Erromango is beyond dispute: Ura (or Aryau), Utaha (or Etiyo), and Enyau (or Yocu). There are other names recorded in the literature which may represent separate languages, or they may simply be local dialect names, place or area names, names of descent groups, or even simply descriptive terms of some kind. In the complete absence of evidence upon which to base any kind of further speculation in this regard, it would perhaps be wiser to relegate questions relating to the status of these terms to the unanswerable.

There is evidence from 19th century observers which suggests—though it falls short of proving—that Sorug (or Sye) may have been an additional language, distinct from Yocu/Enyau. If it was not a separate language, then these two speech-forms must have been dialects of a single language.

Without doubt modern Erromangan emerged after contact and population mixing between speakers of the two in the late 19th century, and this linguistic contact was clearly associated with a movement of linguistic features from the south to the north. Without too much difficulty, then, modern Erromangan can be seen as being a direct descendant of 19th century Yocu/Enyau, with a bit of phonological influence from the south, as well as the introduction of a competing possessive paradigm. This is consistent with the scenario presented by Lynch (1983a).

However, if Sye/Sorug really were a distinct language, then presumably it should have had other grammatical features that would have been just as susceptible to transfer as the possessive paradigm. Lynch and Tepahae (forthcoming) have argued for the possibility of an additional earlier language on Aneityum, partly on the basis of the existence of frequent alternations in competing forms in the modern language. Their argument is basically that where there are competing forms in modern Aneityumese, these may have arisen as a result of features from two different languages coming together in some kind of fused language consisting of a random mixture of certain elements from both of the original languages.

- 57

Since we know for certain that Sye/Sorug and Yocu/Enyau came into contact in Erromango in the late 19th and early 20th centuries, one might expect to find exactly the kind of situation in modern Erromangan that we find in Aneityumese today. However, the only point of comparison between the two situations is the alternation between the two possessive paradigms that we find in modern Erromangan. This re-opens the possibility that Yocu/Enyau and Sye/Sorug may not have been originally structurally so distinct after all.

If the two speech communities had originally been structurally nearly identical, except for differences in the possessive paradigm (along with perhaps a few other relatively minor structural and lexical differences), I can fairly easily visualise the current situation having evolved out of a mixing of the two populations. Under this kind of scenario, however, Yocu/Enyau and Sye/Sorug should probably be described as dialects of a single language, possibly verging on mutual unintelligibility. The major difference between the two would presumably have been lexical rather than structural if there was any mutual incomprehensibility.

I concede that I am now entering again into the realm of speculation, but perhaps the fact that 19th century European observers regarded Yocu/Enyau and Sye/Sorug as different languages was because they themselves could not understand Sye/Sorug speakers, even if Yocu/Enyau and Sye/Sorug speakers could understand each other. It is possible to imagine native speakers of Yocu/Enyau and Sye/Sorug under these sorts of circumstances experiencing less difficulty communicating across this lectal boundary as they could have used their greater knowledge of structure, as well as greater sharing of cultural context, in interpreting what was being said.

Of course, the widespread tendency of Melanesians in multilingual locations to have dual-lingual competences in which different people speak their own languages while having a passive command of other languages would make the boundary between language and dialect that much more difficult to draw. Under such circumstances, of course, the distinction between “language” and “dialect” becomes purely academic. For the missionaries (and perhaps also for linguists who can transport themselves back in time), Yocu/Enyau and Sye/Sorug may well have been separate languages. For Erromangans, however, the two may have been similar enough to have been dialects of a single language, albeit dialects that were teetering on the brink of diverging into separate speech communities.

The effect of the population mixing of the late 19th century then would have been to reverse a pre-contact trend towards linguistic diversification, - 58 turning the direction of change from linguistic speciation to unification. It is tempting to ask: “What happened to Sorug?” Sorug did not become extinct; nor did Enyau. What did become extinct is the name Sorug, as did the name Enyau. But the lects for which these terms were once used as labels have continued to be spoken uninterrupted. Some features have crossed from original Sorug into Enyau. This has resulted in a reversal of a pre-contact trend for the two to become less alike, with the two now becoming more alike.

What was once Enyau is now probably simply the Dillon's Bay dialect of modern Erromangan, while what was once Sorug is now probably the southern and eastern dialect of the same language. The reason for the loss of the original lect-names also makes sense according to this kind of scenario. It is no longer necessary to distinguish the two dialects on the basis of their possessive markers because both sets are used throughout the modern speech community.

According to 19th century sources, the name Sye was originally used as an alternative to Sorug. Lynch (1983a) records Sie (more accurately, Sye) as the name used to refer to modern Erromangan. In coming to be used to refer to both northern and southern varieties of the language, this name has also “jumped” across the original lect boundary. Instead of expressing a contrast between Sye (or Sorug) and Enyau (or Yocu), once the boundary between these two lects had become sufficiently blurred, this term was freed to perform the function of contrasting Ura from that which was not Ura.

Given the imminent loss of Ura, it is arguably redundant for there to be any distinct language name to refer to the language on the island which is not Ura. This would account for the fact that the name Sye, along with any of the other original candidates as a lect-name, all appear to be heading for oblivion. It is almost completely unambiguous now for Erromangans to refer to their language simply by the descriptive label nam Eromaga ‘Erromangan language’.

The question that I want to address now is a rather different one. Historical linguists sometimes reconstruct an earlier phonological or grammatical stage of a language on the basis of some kind of synchronic structural untidiness, on the assumption that certain kinds of complex distributions of forms are likely to derive from an earlier less complex situation. The question that I would like to pose now is: on the basis of synchronic evidence (along with - 59 any recent historical information that I have been able to deduce in this discussion), can we reconstruct any aspects of the sociolinguistic situation on Erromango before the first European missionaries arrived?

Given that new villages, such as Unpogkor (Dillon's Bay), have developed in the last century or more in places where there were originally no villages, populated initially by people who spoke different—though still probably fairly closely related—languages, the situation would presumably have been be ripe for language contact phenomena to come into play.

When dialects of a single language or very closely related languages come into contact in these sorts of circumstances, we might expect some kind of koine to emerge. Such a variety is typically characterised by a mixture of structural features and vocabulary from the various speech-forms which come into contact. It will also typically contain a number of original features that reflect forms that are structurally intermediate between the various original structures, or which represent simplifications of the original structures (Trudgill 1986:83-126).

It is difficult to know if these sorts of processes have taken place in the history of modern Erromangan, given that we know so little of the nature of the various speech-forms that were used in the middle of the 19th century. That there are competing possessive paradigms in modern Erromangan is at least consistent with the notion of koineisation having taken place.

However, if structural simplification has taken place, there is little evidence for this in modern Erromangan. Morphologically, this is one of the more complex languages of the Oceanic subgroup. There is a considerable amount of morphophonemic variation, much of which is not phonologically predictable. There are a large number of inflectional categories in the verbal morphology of the language, with a considerable number of discontinuously marked categories.

If any structural simplification has taken place, it certainly did not go as far as it might have. Also, if there has been any structural simplification, it must have taken a more complex language than modern Erromangan as its starting point. On comparative grounds at least, there is no evidence that modern Erromangan was ever any more complex than it is at the moment. Comparison with other Oceanic languages would suggest that the complexities of Erromangan are in fact relatively recent developments.

Thus, it would seem that levelling has not played a major part in the emergence of modern Erromangan. This also suggests that, despite the loss of a number of languages on Erromango, as well as the clear evidence that we have for massive demographic change, languages such as Ura and Utaha (and possibly other languages) never came into significant general contact with the larger language. If there had been such multilingual contact on a - 60 major scale, we would expect to find significant evidence of levelling.

The earliest quote about the linguistic situation on Erromango by Turner (1861), which dates from before the first of the reported epidemics, already refers to a single dominant language on the island. I have suggested several times that speakers of Ura, Utaha and Sye/Sorug even at that early stage were possibly universally bilingual in Yocu/Enyau. We could account for the fact that levelling and structural simplification did not take place as speakers of these minority languages would simply have stopped using their languages when they moved to new villages and adopted their second language as their primary language. This implies that Yocu/Enyau had possibly come to be dominant around Erromango even before the missionaries arrived. The establishment of their main centre of operations at Dillon's Bay would therefore have simply sped up a process that was already under way before evangelisation.

It would be interesting to speculate how such a situation might have arisen. One possibility is that there may have already been major unreported demographic disruption before the arrival of the first missionaries. Sandalwood traders, for example, had been operating on Erromango between 1854 and 1864 (Shineberg 1967:251-52), which might have provided the locus for the earlier introduction of diseases. If the original population had been substantially higher than the estimate of up to 6000 that I have presented—and it will be remembered that the island could have supported substantially more than this number—rapid (and unreported) depopulation might account for this kind of situation.

Many thanks to John Lynch for helpful comments on earlier drafts of this paper, and to Matthew Spriggs for supplying copies of relevant sections of his 1983 fieldnotes. Final responsibility for all interpretation is my own.

- 61 Page of endnotes

- 62

• Capell, A., 1972. Some phonological phenomena in Eromangan. In J. M.C. Thomas and L. Bernot (eds), Langues et techniques, nature et société, vol. 1: Approche linguistique. Paris: Klincksieck, pp. 49-56.
• —— n.d. Eromanga. Typescript.
• Capell, A. and John Lynch, 1983. Sie vocabulary. In John Lynch (ed.), Studies in the Languages of Erromango. Pacific Linguistics C-79. Canberra: Australian National University, pp. 75-147.
• Charpentier, Jean-Michel, 1982. Atlas linguistique du Sud-Malakula/Linguistic atlas of South Malakula (Volume 1). Langues et cultures du Pacifique 1. Paris: Société d'Études Linguistiques et Anthropologiques de France.
• Crowley, Terry. 1995a. Erromangan Dictionary. Typescript.

• Gabelentz, H.C. von der, 1861, 1873. Die Melanesischen Sprachen nach ihrem grammatischen Bau und ihrer Verwandtschaft unter sich und mit den malaiisch-polynesischen Sprachen. Leipzig: S. Hirzel.
• Gordon, Rev. J.D., 1889. Sketch of the Eromangan grammar. In Rev. D. MacDonald (ed.), Three New Hebrides Languages (Efatese, Eromangan, Santo). Melbourne: Melbourne Public Library, pp. 61-84.

• Humphreys, C.B., 1926. The Southern New Hebrides: An Ethnological Record. Cambridge: Cambridge University Press.

• Lynch, John, 1983a. The languages of Erromango. In John Lynch (ed.), Studies in the Languages of Erromango. Pacific Linguistics C-79. Canberra: Australian National University, pp. 1-10.
• —— 1983b. Ura grammar sketch and vocabulary. In John Lynch (ed.), Studies in the Languages of Erromango. Pacific Linguistics C-79. Canberra: Australian National University, pp. 148-83.
• Lynch, John and Phillip Tepahae, forthcoming. Digging up the linguistic past: The lost language(s) of Aneityum, Vanuatu. In R.M. Blench and M.J.T. Spriggs (eds), Archaeology and Language, Vol. 2. One World Archaeology Series. London: Routledge.

• McArthur, Norma and J.F. Yaxley, 1968. Condominium of the New Hebrides: A Report on the First Census of the Population, 1967. Sydney: New South Wales Government Printer.
• Miller, J. Graham, 1981. Live: A History of Church Planting in the New Hebrides (Now the Republic of Vanuatu) to 1880, Vol. 2. Sydney: Bridge Printery.
• Murray, A.W., 1863. Missions in Western Polynesia, Being Historical Sketches of these Missions, from their Commencement in 1839 to the Present Time. London: John Snow.

• Robertson, H.A., 1902. Erromanga: The Martyr Isle. New York: A.C. Armstrong and Son.

• Shineberg, Dorothy, 1967. They Came for Sandalwood: A Study of the Sandalwood Trade in the South-west Pacific 1830-1865. Melbourne: Melbourne University Press.
• Spriggs, Matthew, 1983. Vanuatu Fieldnotes. Copy held in the Vanuatu National Museum.

- 63

• Spriggs, Matthew and Stephen Wickler, 1989. Archaeological research on Erromango: Recent data on southern Melanesian prehistory. Indo-Pacific Prehistory Association Bulletin, 9:68-91.
• Statistics Office, 1991. National Population Census, May 1989. Vol. VI: Tafea. Port Vila.

• Taki, Jerry and Terry Crowley, 1996. Netai ra nam Ura/Diksenari blong lanwis blong Ura. Typescript.
• Taki, Jerry and Darrell Tryon, 1994. The lost languages of Erromango. Paper presented at the World Archaeological Congress, December 4-11. New Delhi.
• Trudgill, Peter, 1986. Dialects in Contact. Oxford: Basil Blackwell.
• Tryon, D.T., 1976. New Hebrides Languages: An Internal Classification. Pacific Linguistics C-50. Canberra: Australian National University.
• Turner, Rev. George, 1861. Nineteen Years in Polynesia: Missionary Life, Travels and Researches in the Islands of the Pacific. London: John Snow.

- 64

Illustration

Available now from all good booksellers, $69.95

The result of over two decades of intensive and wide-ranging research, Tokelau: A Historical Ethnography is both a comparative ethnographic study of the three tiny Polynesian atolls of Tokelau, and a narrative record of their past.

The ethnographic study, set in the years around 1970, gives a detailed description of life in the villages of each atoll and compares the environment, the different systems of government that have evolved, gender roles, kinship ties and the relationship of young to old. The second part of the book traces the histories of the three atolls, with local narratives and records complementing foreign documents to tell the separate and combined stories of the atolls' traditional and colonial pasts. Throughout, the differences and interrelationships between the three islands are highlighted.

This is a comprehensive and substantial work, in which history informs ethnography and ethnography history. Illustrated with photographs, many by Marti Friedlander, it is written to appeal to a wide readership, as well as to the anthropologist and historian.

AUCKLAND UNIVERSITY PRESS

Illustration