Volume 85 1976 > volume 85, No. 1 > Review article: Green, R. C., and M. Kelly (eds.)Studies in Oceanic Culture History, by George W. Grace, p 103-112
                                                                                             Previous | Next   

- 103
REVIEW ARTICLE

GREEN, R. C., and M. KELLY (eds.): Studies in Oceanic Culture History, Volume 3, Pacific Anthropological Records 13. Honolulu, Bernice P. Bishop Museum 1972. vii, 160 pp., 10 tables, 8 figures. Price (U.S.) $4.00.

This volume consists of revised versions of two papers presented at the Wenner-Gren Symposium on Oceanic Culture History held at Sigatoka, Fiji, in August 1969. The first, “On the Internal Relationships of Eastern Oceanic Languages”, by Andrew Pawley, has been greatly expanded over the original conference paper and fills 142 of the 160 pages of the volume. The second paper, by Bruce G. Biggs, is entitled, “Implications of Linguistic Subgrouping with Special Reference to Polynesia”.

Biggs' paper does not require lengthy discussion. It is addressed to an audience of non-linguists who are concerned with making inferences from linguistic data regarding the prehistory of peoples and cultures. He provides a clear discussion of the family tree model for genetic subgrouping in linguistics and shows why the idea of language mixture does not accord with that model.

As regards prehistoric population movements, he cites evidence that some Polynesian island groups have been reached by more than one significant migration. He finally makes a number of interesting comments on the application of the Wörter und Sachen technique in Polynesia. Of particular interest is his observation that, although the sound correspondences permit the reconstruction of a Proto-Polynesian word for the sweet potato (which other evidence argues to have been a recent introduction), the reconstructed form is of a phonological shape uncharacteristic of Proto-Polynesian words—a fact that renders it suspect.

The paper is short and reads easily. For those who are, or should be, interested in these matters it is well worth the reading.

Pawley's paper is an important work. Its scope is that of a full-fledged monograph even though the mode of its publication makes it technically an article. It makes claims and raises issues that require full discussion, and is, in fact, a major contribution to Oceanic linguistics.

It is divided into 10 sections. The first provides general background on previous attempts at classification. Pawley accepts the hypothesis of an Oceanic subgroup of Austronesian languages and surveys the proposals regarding groupings within Oceanic. There are brief discussions of phonological (pp. 7-8), grammatical (pp. 8-9), and lexicostatistical (pp. 9-11) evidence for the existence of an Eastern Oceanic (EO) subgroup.

The second section outlines his approach. Briefly, his subgrouping hypotheses are based on shared innovations, and his search for shared innovations is made primarily in the grammar. As a basis for the search for shared innovations, he undertakes first a reconstruction of the proto-language (i.e., Proto-Eastern - 104 Oceanic [PEO]). Pages 20-3 identify the languages selected to represent EO in the study.

The following three sections deal with EO phonology, grammar, and lexicon, respectively. Section 3 (pp. 24-31) deals with the PEO sound system and sound correspondences within and without. Section 4 on PEO grammar is by far the longest section of the entire study. It includes a sketch of PEO grammar (pp. 32-44) followed by (pp. 45-90) a tabular presentation of the reconstructed grammatical elements with the sets of cognate morphemes in the languages of the sample.

Section 5 (pp. 91-7) gives some comparative vocabulary. Sections 6 to 9 (pp. 98-135) present Pawley's proposed subgrouping of EO with a detailed discussion of the evidence, and the final section presents his conclusions.

I will devote a good deal of space to the Eastern Oceanic hypothesis itself, reviewing what has been said about its membership, considering the nature of the supporting evidence and its possible interpretations, and criticising specifically one of Pawley's arguments. I will then discuss his internal groupings. Finally, I will comment on his reconstruction of PEO.

The Eastern Oceanic hypothesis has never been given a very careful formulation. The term “Eastern Oceanic” and the hypothetical grouping which it designates derive from Biggs 1965. 1 The working hypothesis that some such grouping exists proved very useful in that study and continues to be useful for some linguistic purposes (including, I will maintain, Pawley's grammatical reconstructions in the present work). For linguists it has been a fruitful strategy to proceed as if such a grouping existed and to postpone for the time being any serious attempt to give greater precision to the hypothesis or to subject it to critical scrutiny.

However, the intended audience for the work under review consists principally of non-linguists who are interested in the culture historical implications of linguistic subgroupings. For their purposes such a postponement is not satisfactory; the status of EO is of immediate concern. If there is in fact a group of languages comprising Polynesian, Fijian and some but not all other Oceanic languages which have uniquely shared a portion of their history, it is of considerable importance to know just what other languages are involved.

Pawley nowhere provides a straightforward discussion of his ideas concerning the boundaries of EO. In fact, he attempts to escape the question entirely by defining the term “Eastern Oceanic” as being (p. 7) used “specifically to denote the 31 languages” dealt with in the study. Of course, if EO consisted simply of an arbitrary set of 31 languages, it would be of no conceivable interest to anyone. However, Pawley makes it clear that he thinks there actually is a subgroup of which his 31 languages are members, and at times it is apparent that he is using the term “Eastern Oceanic” to refer to that supposed subgroup itself.

In introducing his 31 language sample, Pawley distinguishes nine subregions “within the EO-speaking area” (p. 21). The 31 languages are grouped according to the subregions which they represent. The regions are: (1) Polynesia (represented by Proto-Polynesian as reconstructed), (2) Fiji (represented by Bauan and Wayan), (3) Micronesia (Gilbertese), (4) Rotuma (Rotuman), (5) Banks Islands [a northern extension of the New Hebrides] (represented by four languages), (6) Northern New Hebrides (seven languages), (7) Central New Hebrides (five languages), (8) San Cristobal-Malaita [south-eastern Solomons] (six languages), (9) Guadalcanal, Florida, South-east Santa Ysabel [south-eastern Solomons] (four languages).

In summary, of the 31 languages, 26 are concentrated in a segment of the arc - 105 formed by the main Melanesian chain. Generally each of these languages is accompanied in the sample by a group of others which are its close linguistic relatives and geographical neighbours. Of these 26 languages, 10 are distributed over the south-eastern half of the Solomon Islands, and from all indications one may reasonably suppose that all of the languages within that area would be classified as EO. The remaining 16 languages are distributed over the northern two-thirds of the New Hebrides.

It is not entirely clear whether Pawley thinks all of the languages of the latter area are EO. In his discussion of the evidence for EO (pp. 7-11) he shows a particular concern for criteria distinguishing the EO languages from their immediate neighbours in the southern New Hebrides and the western Solomons. The discussion there might lead one to infer that perhaps all of the languages between are thought to belong to EO. However, he comments elsewhere (p. 113) that the New Hebrides languages in his sample are not representative of the region as a whole (it is not clear whether the region referred to is the entire New Hebrides or just the central and northern New Hebrides), and that they represent languages which appeared most similar to Polynesian. I believe also that his principal phonological criterion for membership in EO would exclude certain northern New Hebrides languages (see below). For the culture historian in any case it would seem to be of significance to know whether EO languages occupy the entire central and northern New Hebrides area or whether, on the other hand, they are scattered among non-EO languages throughout the area.

On the basis of previous work and additional evidence presented here Pawley considers Fijian and Polynesian to form an exclusive grouping called Central Pacific. The remaining two languages in the sample, Rotuman and Gilbertese, appear rather anomalous. Neither makes any significant contribution to the reconstruction of PEO, and Pawley does not succeed in determining their position within the group. They are simply left as unclassified members of EO.

Pawley discusses phonological, grammatical, and lexicostatistical evidence for EO. The grammatical evidence consists of 18 grammatical morphemes which are reconstructible by his criteria for PEO, most of which are reflected in a considerable number of the languages of his sample, and none of which he has observed in languages outside the sample. Although a search for such morphemes in the present state of the literature is a very difficult task, and he makes no claim of having systematically conducted a search, the accumulation of examples is still striking.

The lexicostatistical evidence comes from Dyen. 2 The grammatical and lexicostatistical evidence may be said to constitute more systematic presentations of the kinds of observations that led to the EO hypothesis in the first place. I believe it is fair to say that with these two kinds of evidence Pawley has provided a reconstruction of the reasoning behind the original proposal. Although this evidence is relevant to the question of the existence of such a grouping—in particular to a closer relationship between Polynesian along with Fijian and languages of the northern New Hebrides-south-eastern Solomons region of the Melanesian chain—it does not provide a good sorting criterion for distinguishing EO from non-EO languages. (Pawley does set up a phonological criterion and appears to suggest that it can be so used. I will discuss this below.)

The almost complete absence of phonological evidence for EO is striking. The identification of subgroups has always been a difficult undertaking and the most convincing arguments usually seem to have been based on phonological evidence. - 106 In fact, phonology has always had a very prominent role in comparative linguistics.

Why has this been so? The answer, I believe, lies in the nature of the phenomenon itself, or at least in our way of perceiving it. We have been able to conceptualise phonological change as permitting a system to persist through a gradual evolution. Thus we distinguish an invariant (the system) and transformations (the changes).

In this conception the sound system of a language contains a determinate number of elements, each item of the vocabulary of the language is signalled by a fixed sequence of one or more of these elements, and the sole type of function which the elements have is that of providing such contrastive signals. Specifically, each element has such a signalling function for each vocabulary item in which it is involved. Therefore, we have a closed set of a small number of elements, but a given element typically is involved in a large number of such element-function relations.

Phonological change is then, mainly, a matter of transformation by general rules of the phonetic substance leaving invariant the system of element-function relations. The element-function relations are, moreover, almost entirely arbitrary and presumably non-adaptive, and not subject to areal influence.

This way of looking at the matter permits us to consider phonological change as the evolution of a system as opposed to the sporadic changes in grammar and vocabulary. The large number of functions involving phonological elements facilitates documentation, and the arbitrariness of the element-function relations lessens the danger that similarities are due to diffusion.

The use of lexical or grammatical evidence involves two additional complications: neither can be regarded as immune to areal influence and neither involves a closed set of elements or an indefinitely persisting system of form-function relations.

The susceptibility of lexicon to borrowing is notorious. It is sometimes said that grammar is relatively impervious to borrowing. However, abstract (i.e., not involving cognate morphemes) grammatical features such as word order, grammatical gender, etc., are very subject to diffusion (or better, to calquing) and such features have generally been rejected as valid evidence for genetic relations. The unreliability of such evidence has been made even more conspicuous recently since Gumperz 3 has cited a case of what appears to be calquing to the point of identity of the grammatical structures of a dialect of an Indo-Aryan language (Marathi) and a dialect of a Dravidian language (Kannada).

Even when the comparisons are based on cognate morphemes, we cannot be sure that calquing is not involved, especially if the languages share much vocabulary and are geographically close. This question arises with regard to the EO pronouns (see below).

Moreover, grammar as represented by grammatical morphemes is not a closed system. What is a grammatical function at one stage in the history of a language may be performed by lexical means at a later stage, or indeed may not even be identifiable as a single function at all (e.g., at one stage future tense may be a compulsory category expressed by an affix, but subsequent changes might eliminate the affix as well as the compulsory expression of tense and leave the language with only ad hoc means for expressing futurity). Moreover, entire grammatical systems can undergo fundamental changes (as in the wholesale loss of inflections).

Areal influences can presumably serve either to inhibit or to hasten change according to the existing fit between the language and the areal environment.

- 107

The foregoing remarks were directed primarily at the problem of determining the membership of EO. However, I should point out here that Pawley relies mainly on grammatical evidence for his internal subgrouping of EO, and that he argues that the systematic nature of grammar “consisting of relatively small numbers of elements whose structural relationships are sharply defined” (p. 20) makes its evidence rather like that of phonology. Moreover, I must concede that he has a point; much of the weight which I assign to the innovations that he identifies is related to the fact that he places them within the framework of a general system. But there is a sort of circularity here; these languages (except for Rotuman and Gilbertese, which he cannot classify) were chosen in the first place mainly because they did have similar grammatical systems. None the less, I think his point should be acknowledged; where we have sufficiently comparable grammatical systems, grammatical evidence does have a kind of systematic character which differentiates it from lexical evidence and makes it somehow more like phonological evidence.

One further point should be made regarding the evidence for EO; it is especially difficult to make a convincing case that a particular language is not a member of a particular subgroup in a study that does not consider other sub-groups. In a study encompassing an entire language family one may argue cogently that language A is not descended from Proto-X by showing that its present characteristics are more satisfactorily explained by the assumption that it is descended from a competing proto-language, Proto-Y. If, however, the scope of the study is restricted so that the question of a Proto-Y cannot be raised, the only possible evidence that A is not descended from Proto-X must consist in a lack of evidence that it is descended from Proto-X. Except where Proto-X is characterised by an irreversible innovation, such a lack of evidence is difficult to document satisfactorily.

I think it is highly likely that all of the languages in Pawley's sample do belong to a subgroup of Oceanic from which some Oceanic languages are excluded. However, it still seems very possible that the particular languages considered in his study are a relatively small sample of the most conservative languages of a subgroup which is more extensive than is suggested by anything Pawley says.

The hypothesis of an Eastern Oceanic group was originally, as I understood it, a working hypothesis providing a framework for comparing Polynesian and Fijian (and in the original instance, Rotuman) with those languages which seemed most apt for the purpose—i.e., those which offered the most numerous and the most transparent points of comparison with them. The results of these comparisons were systematically represented in reconstructions labelled Proto-Eastern Oceanic.

The languages most often used in the comparisons might have been predicted since various discussions of the position of Polynesian over the years have singled out roughly the same languages as being most like Polynesian. These persistently include Fijian, Rotuman, and languages of the central and northern New Hebrides and of the south-eastern Solomons. The Heonesian group in Dyen's (1965) lexicostatistical classification 4 consists largely of languages from the same set.

However, the same languages which suggested the Eastern Oceanic hypothesis have also been regarded as being most like Indonesian languages. In fact, the Heonesian of Dyen, which has frequently been identified with EO, is in that study classified together with substantially all of the Indonesian languages over against the remaining Oceanic languages. Furthermore, scholars who believe that the Melanesian languages originated through the pidginisation of Indonesian - 108 languages generally single out the same Melanesian languages as having best preserved the original Indonesian Sprachgut.

If these judgements that these languages are more like the Indonesian languages than are the other languages of Oceanic are correct, then either (1) they must have separated from the Indonesian languages more recently than did the others, or (2) if they separated from them as early as did the others, they must have changed less in the ensuing period of separate development.

If they actually did share a longer common history with the Indonesian languages, then the Oceanic hypothesis is false; there is no Oceanic subgroup but rather a group along the general lines of Dyen's Malayopolynesian Linkage.

If, on the other hand, the Oceanic hypothesis is correct, and the “Eastern Oceanic” languages have been separate from Indonesian as long as the other languages of Oceania but have conserved an atypically large portion of the original vocabulary and structure, then this conservatism is just as apt to be the explanation of their resemblance to one another as of their resemblance to Indonesian languages. In sum, they might represent not a linguistic subgroup at all but just the set of the most conservative languages within Oceanic.

It is also possible, of course, that both factors are involved—that these languages do belong to a subgroup of Oceanic but that they are a selection of the most conservative members of that subgroup. If so, the number and identity of the other, less conservative, members remain unknown.

It is appropriate to point out again that for the purpose of reconstruction it is often a good approach to work with a sample of more conservative languages for the first approximations. For the exploratory work in reconstruction which Pawley undertakes, his sample of languages seems to be well chosen. However, they were chosen to represent the internal rather than the external history of the group.

Pawley's phonological evidence for EO was mentioned, but not discussed, above. He was able to propose only a single kind of phonological evidence for the grouping. The phoneme inventory reconstructed for PEO shows no changes from that reconstructed for POC. However, Pawley argues that there is phonological evidence for EO in the loss of (originally) final consonants in some vocabulary items where the consonants are retained in some other Oceanic languages. Unfortunately, he does not make his hypothesis very specific. He cites 10 items for which he gives Proto-Austronesian (PAN), Proto-Oceanic (POC), and PEO reconstructions. Generally, the PAN reconstructions have a final consonant, while he reconstructs the POC form with the consonant plus (in parentheses) a vowel (identical to that preceding the consonant in question), and the PEO form with neither the consonant nor the vowel.

His hypothesis seems to be that in the development from PAN to POC final consonants in many words had been lost, but in others they had been retained (perhaps with a supporting vowel added). The conditions governing loss or retention are not apparent, and loss and retention may be considered to have been sporadic. After the break-up of POC such sporadic loss of final consonants continued to occur in various Oceanic languages. PEO is distinguished by having lost all of them.

Pawley has done well to call our attention to the complexity of the development of the AN final consonants in OC languages. However, what he proposes as an explanation is essentially the null hypothesis—that there are no regularities to be discovered, only numerous sporadic events.

On the other hand, the unreliability of this phenomenon as evidence for subgrouping is indicated in Pawley's own remarks (p. 7, emphases mine), “The final —C(V) is retained in most or all of the cognates I have found in many - 109 languages of Papua, the Western Solomons, New Caledonia and in at least some languages of the north coast of New Guinea, New Britain, New Ireland and southern New Hebrides.” My own hasty check of vocabularies readily accessible to me suggests that we may go farther and simply dismiss this criterion entirely. For example, I find no evidence for the preservation of the final consonants in the languages of either end of the Oceanic area in Melanesia—the south of New Caledonia on the one hand or the Oceanic languages of Irian Jaya on the other—as well as in many languages in between, especially in the New Guinea area. On the other hand I have found three cases of apparent preservation in the northern New Hebrides. They are: (in the orthography of the sources) from Malekula, Big Nambas manoch and Atchin nananko “bird” from PAN manuk, and from the Torres Islands in the extreme north, Lo gilit “skin” from PAN kulit (despite the fact that from the skimpy evidence available Lo otherwise appears to be a quite close relative of the Banks Islands languages in Pawley's sample). Examples such as the above may explain the previously noted uncertainty as to whether Pawley considers all northern New Hebrides languages to be EO.

This phonological criterion would not merit so much attention except that in some places in Pawley's work (e.g. pp. 20, 130, 134) it seems to be regarded as a reliable touchstone for distinguishing between EO and non-EO languages. Pawley's statement (p. 130) that: “Rotuman meets the most important criterion for inclusion in EO in the loss of [these final consonants]”, when taken in context suggests that this is in fact his principal argument for retaining Rotuman as EO. A very similar statement, suggesting the same conclusion, appears (p. 134) with respect to Gilbertese.

Pawley's subgrouping of EO technically claims, of course, only to be a sub-grouping of the languages in his sample. Nevertheless, some unquestionably significant results emerge. As was noted above, he was not able to classify Gilbertese or Rotuman, and the subgrouping presented therefore involves only 29 languages representing seven of his nine regions.

In all but one case the regional groupings originally set up turn out to be subgroups at the lowest level. The one exception involves regions 5 (Banks) and 6 (northern New Hebrides). Pawley finds that the languages representing the two regions fall into a single subgroup called Northern New Hebrides-Banks, of which no further subdivision can be made. He points out (p. 115) that this differs from the comparable grouping in my classification 5 in including the languages of Espiritu Santo whereas I had thought the latter to be most closely linked with the languages of northwest Malekula (none of which are included in Pawley's sample). I should comment here that I had little data on Santo languages then, and so felt some uncertainty about their position, but also that the data available to Pawley are still limited, and that I do not find his case overwhelming. Fortunately, the prospects of our having much better information on Santo languages very soon, thanks to W. G. Camden, Jacques Guy, and Darrell Tryon, are very bright. In fact, there is good reason to hope for rapid progress with all of the various problems of New Hebrides language classification as a result of Tryon's work.

There are, thus, six (always excluding Gilbertese and Rotuman) lowest level subgroups, two in the Solomons, two in the New Hebrides, and two to the east of the Melanesian chain. Subsequently, the two Solomon Islands groups are grouped together as Southeast Solomonic, the two New Hebrides groups as North Hebridean, and of course the remaining two, Fijian and Polynesian, as Central Pacific. Finally, Central Pacific is grouped together with North Hebridean - 110 into the North Hebridean-Central Pacific group. Pawley acknowledges that the arguments for these groupings are not conclusive in all cases. The case for Central Pacific is certainly quite strong. From the standpoint of culture history the most interesting question doubtless is that of the immediate external relations of Central Pacific. Pawley has compiled a rather impressive case that they are with the New Hebrides rather than with the Solomons as some have proposed. If he had accomplished no more than that his efforts would have been worth while.

Undoubtedly Pawley's grammatical reconstructions, even though they appear here only as a step in an argument directed at other questions, will in the long run prove to be the most important contribution of the work. This is a trailblazing effort which will serve as a model for future efforts. It is impossible to give a thorough review here. In fact, review and revision will surely continue for a long time as new data become available. I will restrict myself here to a few critical remarks which I hope may contribute to a more rigorous approach in subsequent studies.

First, a few words should be said about the phonological reconstructions. The reconstructed phoneme inventory is essentially the same as had usually been proposed for PEO (and for POC). Pawley's positions on certain matters where there has not been complete agreement are as follows: he distinguishes *r from *d and *ñ from *n. He also recognises two labiovelars, which he writes *mw and *pw. *mw is the labiovelar counterpart of *m and *pw apparently is conceived as the labiovelar counterpart of *p rather than of *mp. Even though he is quite explicit about this (cf. p. 2), I assume that it is simply an oversight (and that the hypothetical *mpw which appears on p. 116 is an isolated lapse growing out of the original oversight).

As is to be expected, the tables giving the reflexes of PEO phonemes in the 31 daughter languages have a more or less tentative character. Frequently more than one reflex is accepted for a given proto-phoneme in a given language where the conditions are not yet known. The imprecision that results is complicated further by the fact that Pawley apparently has sometimes obtained data for a language from different sources using different orthographies (compare these tables with the tables of cognates, pp. 45-97).

Pawley differs from Dempwolff 6 in finding no distinction in Sa'a between *s and *ns, and in fact cites the loss of this distinction as one of the innovations characterising the San Cristobal-Malaita group. However, he does not comment on the disagreement or criticise Dempwolff's evidence.

The reconstruction of PEO grammar is impressively detailed. Detailed proposals as to morpheme classes, phrase structures and sentence structure are put forward and specific grammatical morphemes are reconstructed. I count a total of 137 such morphemes for which data are given (65 if numerals and personal pronouns are excluded).

The supporting data are presented in tabular form with the reconstructed morphemes across the top and the reflexes in the languages of the sample appearing in the columns below the reconstructions. Occasionally, non-cognate forms having a like function in a particular language are given in brackets. The forms cited without brackets under a particular reconstruction are, then, claimed to be reflexes of that reconstructed form, and, in fact, presumably constitute the data for that reconstruction. Pawley nowhere discusses the methodology of his reconstructions, and the relations between the precise shape of a reconstruction and the supporting data are sometimes not entirely clear. It is clear that one cannot always derive a particular reflex from the reconstruction by the simple - 111 application of the sound correspondences as they appear in the tables of reflexes of the PEO phonemes.

It seems obvious that if a study of this scope is to be successful it will be obliged to tolerate a certain amount of unexplained irregularity in the sound correspondences in particular items. However, there is a price that must be paid for relaxing the requirement that the reconstructions must be determined by the data, and that the data must be predictable from the reconstructions. The price is the slackening of empirical restraints on what can be claimed and a loose relation between the reconstructions and the data on which they are based. I will discuss instances of several different kinds of problems that result.

1. The languages have a rich pronominal system. A total of 60 personal pronominal forms (or more precisely, pronominal slots, since in a few instances alternative forms are reconstructed for the same slot) are reconstructed. The reconstructed pronouns are given in a table on p. 37 and again (pp. 61-75) in the tables giving the reflexes in the daughter languages. For 18 of the 60 slots there is some difference between the reconstructions as they appear in the two places. Most of the differences are minor. Some are undoubtedly due to faulty proof-reading. But since the reconstructions are so loosely associated with the data, it generally is not clear that one is correct and the other in error.

2. Pawley comments (p. 119) that, “. . . the type tamoli undoubtedly reflects PEO *tamwane ‘man, person’ with irregular changes *mwa > mo, *n > l, *e > i.” But what is actually the basis for the assurance that it reflects *tamwane? Do the central New Hebrides forms in question fit *tamwane significantly better than they fit another reconstruction, *tamwata “person” (cited on p. 32)? Or how can we be sure that a third reconstruction is not required, especially in view of Rotuman famori “man” (miscited [131] as famoli) which he also attributes to *tamwane, but which must, in view of his subgrouping, represent a convergent development if the proposed etymologies are correct?

3. There is a “stative derivative” prefix that is reconstructed in two (freely alternating?) shapes, *ta and *tapa. The two shapes are required to account for the data. However, for the causative prefix (reconstructed as *paka) there is equally good evidence for reconstructing an alternant *pa, but no such reconstruction is made, and the forms which would reflect *pa are given as reflexes of *paka without comment. (If the forms in question which appear in nine widely scattered languages are really interpreted as irregular reflexes of *paka they must be regarded as either a striking piece of evidence for sub-grouping or as a remarkable set of convergences.)

4. One of the principal problems involved in the reconstruction of grammar is that of distinguishing between inherited and areal features. A case in point is the system of personal pronouns. Most Oceanic languages have a set of dual pronouns, and these generally appear to have been formed from the corresponding plural pronouns followed by the root for “two”. Many, including most of the languages of Pawley's sample, have a further set of trial or paucal pronouns formed on the plural pronouns and the root for “three”.

Although these general patterns may be recognised, their historical explanations are not equally clear. The sound correspondences connecting the comparable forms in different languages are extremely irregular. Did the protolanguage in fact have the four number system or are the dual and trial (or the latter alone) areal features resulting from a diffusion by calquing? A glance at the tables makes it clear that the modern forms have not all evolved by regular sound changes from the reconstructed forms. For example, to cite some of the more extreme cases, we see Oroha aru, a appearing as the reflexes of *mudua or Wayan oba as the reflex of *(ka)mutolu. In several languages the roots for - 112 “two” and “three” have come to be prefixes, rather than suffixes. In Wayan the numeral elements are prefixed, but also the root for “three” has been replaced by what is said to be an obsolete form of the word for “four”. In sum, although Pawley's treatment suggests that all of the forms cited are continuations of the respective parts of a similar system in the proto-language, I do not find it easy to discern what elements actually have been retained from the proto-language and what elements are innovated, much less to judge which innovations were independent and which result from stimulus diffusion. I must, therefore, regard the question of the kind of pronominal system that should be attributed to the language being reconstructed as still not finally resolved.

In conclusion, I am obliged to point out that this work is marred by insufficient care in editing and proof-reading. I have found numerous errors. Most of those that I have found, of course, are obvious and can be corrected by anyone who knows English. What is more disturbing is that errors occur also in the cited linguistic data. I cannot be sure how frequent they are, but I know of no reason to suppose that they are less common there than in the English text. Readers should be warned against citing data from this work without confirming them in the original sources. If proper caution is not exercised, erroneous data tend to be perpetuated in the literature.

I will make no attempt to enumerate the typographical and other errors I have found. There are, of course, many other minor points that could bear comment. I will permit myself one such comment. I should like to point out that it has not been demonstrated that Rotuman ko is a borrowing from Polynesian (p. 121). The Polynesian and Rotuman forms can be quite naturally explained as straight-forward developments of a hypothetical PEO ŋko.

In spite of its various shortcomings, this is an important work which contributes not only to the study of linguistic relationships in eastern Oceania but also to the antecedents of Central Pacific grammar.

REFERENCES
  • BIGGS, Bruce, 1965. “Direct and Indirect Inheritance in Rotuman”. Lingua, 14:383-415.
  • DEMPWOLFF, Otto, 1937. “Vergleichende Lautlehre des austronesischen Wortschatzes”; 2er Band, Deduktive Anwendung des Urindonesischen auf austronesische Einzelsprachen. Berlin. Beihefte zur Zeitschrift für Eingeborenen-Sprachen, Heft 7.
  • DYEN, Isidore, 1965. A Lexicostatistical Classification of the Austronesian Languages. Bloomington, Memoir of the International Journal of American Linguistics, 19.
  • GRACE, George W., 1955. “Subgrouping of Malayo-Polynesian: A report of Tentative Findings.” American Anthropologist, 57:337-9.
  • GUMPERZ, John J., 1967. “On the linguistic markers of bilingual communication.” Journal of Social Issues, 23 (2):48-57.
1   Biggs 1956.
2   Dyen 1965.
3   Gumperz 1967.
4   Dyen 1965.
5   Grace 1955.
6   Dempwolff 1937:155-8.