Chapter 3. The Prehistory of Oceanic Languages: A Current View

Andrew Pawley, Malcolm Ross

Table of Contents

Introduction
Some questions
The Comparative Method
The Oceanic Subgroup and Proto-Oceanic Culture
The Dispersal and Diversification of Oceanic Languages
Locating the POc language community and directions of dispersal
Modes of Oceanic dispersal
Change in the Lexicons and Cultures of Oceanic Speakers
Conclusions
References

The paper sketches the linguistic comparative method and reports on current results of its application to the Oceanic group of Austronesian languages. We give a brief account of the culture of Proto-Oceanic speakers as it is revealed by language, then outline our present view of Oceanic subgrouping, explaining what this implies about the location of Proto-Oceanic and the dispersion of Oceanic speakers into the Pacific. We examine patterns discernible in this dispersion, and address the question of why the languages of Melanesia have changed more than others.

Introduction

Some questions

Proto-Oceanic (POc) is the ancestor of some 450 Austronesian languages of Melanesia, Micronesia and Polynesia. Darrell Tryon has outlined in his paper for this volume the origins and position of the Oceanic subgroup within the wider Austronesian language family. Our concern here is with what historical linguistics has to say about the development of the Oceanic languages and the cultures of their speakers from Proto-Oceanic times onward.[1]

We will address the following questions:

  1. How solid is the integrity of the Oceanic subgroup?
  2. What was the culture of Proto-Oceanic speakers like?
  3. What was the order of genetic splits among the languages of the Oceanic subgroup?
  4. Where were Proto-Oceanic and later interstage languages spoken?
  5. How did Proto-Oceanic and later interstage languages break up? Is a recurrent pattern discernible?
  6. Why have some Oceanic languages of Melanesia changed much more than others?
  7. How and why did Proto-Oceanic culture change in the daughter communities?

First, however, a few remarks on the methods of historical linguistics are in order.

The Comparative Method

The historical linguist’s main tool is the Comparative Method.[2] It needs to be stressed that, in spite of its name, this method is completely different from the comparative method used in ethnological reconstruction, which is based on typological similarities.

The linguistic Comparative Method gains its strength from four facts associated with the sound systems of language or with the relation between sounds and meanings. Firstly, each word or morpheme in a language consists of a sequence of sounds which themselves have no meaning. Secondly, the association of meanings with sound sequences in a particular language is in almost all cases arbitrary. Thirdly, sound changes occur in all languages over time and, fourthly, these changes are, typically, regular. Regularity of sound change means that within a language community pronunciations change systematically, such that sound x becomes sound y under statable phonological conditions not just in a few words but in all words that meet those conditions. Together, these facts allow us to identify genetically related languages and to recognize genetically related morphemes or cognates shared by sister languages, as opposed to borrowed morphemes and accidental resemblances.

The comparative linguist takes sets of words and looks for regular sound correspondences among them. For example, among the cognate sets from widely distributed Oceanic languages in Table 1, we see a sound correspondence recurring in medial position in the words for ‘paddle’, ‘rain’ and ‘name’ between Manam -r-, Takia -Ø-, Mangap -z-, Motu -d-, Mekeo -k-, Bali-Vitu -z-, Tolai -Ø-, Nyindrou -r-, Gela -h-, Lau -t-, Bauan -∂-, and Tongan -h-. In these and later comparisons, Ø indicates zero, i.e. loss of a sound present at an earlier stage.

Another correspondence set is reflected by initial v- in the Gela items for ‘full’, ‘turtle’, and ‘paddle’. We find Manam Ø-, Takia f-, Mangap p-, Motu h-, Mekeo p-, Bali-Vitu v-, Nyindrou b-, Gela v-, Lau f-, Bauan v- and Tongan f-. Notice, however, that initial v- also occurs in the Gela items for ‘woman’ and ‘stone’, but here the Takia correspondence is p-, not f-. This is the result of phonological conditioning: Takia has f- before what was once -o-, and p- before -a-. Similarly, medial -v- in Gela vaivine ‘woman’ corresponds with Takia -Ø- in pein, since POc *p is lost between vowels in Takia.

Where sound correspondences like those exemplified in Table 1 recur throughout the vocabularies of the languages concerned, we know that the languages are genetically related to one another. That is, we can demonstrate that the related words are formally derived in a regular manner from a common ancestral language.

Table 1. Some Oceanic cognate sets.(a)

‘full’

‘turtle’

‘paddle’

‘woman’

‘rain’

‘stone’

‘name’

POc

*ponuq

*poñu

*pose

*papine (b)

*qusan

*patu

*qasan (c)

Papua New Guinea north coast

Manam

uni

ore

aine

ura

ara-

Takia

fon

fei

pein

ui

pat

Mangap

pin

pen

peze

(waine)

pat

za-

Papua New Guinea south coast

Motu

honu

hode

hahine

lada-

Mekeo

poŋu

poke

papie

aka-

Bismarck Archipelago

Bali-Vitu

Vonuku

Bonu

vozere

(tavine)

γuzaŋa

vatu

iza-

Tolai

pun

vo

vavina

vat

iaŋ

Nyindrou

boi

bos

ur

bek

ñara-

Southeast Solomon Islands

Gela

vonu

vonu

vohe

vaivine

uha

vatu

aha-

Lau

fuŋu

fonu

fote

uta

fou

hata-

Central Pacific

Standard

vonu

i-vo∂e

u∂a

vatu

ya∂a-

Fijian

 

Tongan

fonu

fonu

fohe

fefine

?uha

Notes

(a) A blank indicates that the appropriate cognate does not occur in the linguist’s data, perhaps because the cognate has been replaced by some other word, perhaps because the linguist failed to collect the cognate due to a shift in its meaning or for some other reason.

(b) From data in other languages, we know that Mangap waine is derived from an alternant *wapine, Bali-Vitu tavine from *tapine.

(c) Most of the words for ‘name’ have a hyphen on the end: this indicates that they normally occur with a suffix indicating the person and number of the possessor: ‘my’, ‘thy’, ‘our’, ‘their’, etc.

Having charted the sound correspondences among contemporary languages, one can infer the sound system of earlier stages by various established procedures, e.g. noting which languages retain distinct correspondences in sets of words that cannot be accounted for except by positing an original distinction. From data of which Table 1 gives only a tiny sample, Oceanic linguists are able to reconstruct the words[3] *ponuq ‘full’, *poñu ‘turtle’, *pose ‘paddle’, *papine ‘woman’, *qusan ‘rain’, *patu ‘stone’ and *qasan ‘name’ for POc.

At the same time, where a word shows an apparently irregular sound correspondence, the linguist must investigate whether this is the result of an unrecognized regularity, or whether it is the result of borrowing from a language which has undergone different sound changes, or of analogical change, or perhaps simply a chance resemblance. The careful reader will notice that Nyindrou bos ‘paddle’ does not conform to the pattern displayed in other cognate sets, an indication that this item could be a borrowing from another Oceanic language rather than a direct continuation of the ancestral form. There are a number of cases in the Pacific of patterned irregularity, where a language shows two or more sets of correspondences, each reflecting the reconstructed sound system of POc. A classic case of this is Rotuman, studied by Biggs (1965), where there are three sets of correspondences: one set occurs in directly inherited words, the other two in words borrowed from Polynesian languages. Another case is Wagawaga in southeast Papua, whose vocabulary comes from two clearly different sources.

Subgrouping is usually a much harder task than establishing genetic relationship. In the Comparative Method, subgrouping is done by working out which languages share innovations relative to an earlier ancestor.[4] For example, we saw in Table 1 that Lau fote ‘paddle’ and uta ‘rain’, descended from POc *pose and *qusan respectively, have undergone an innovation whereby POc *-s- has become Lau -t-. This innovation is one of several sound changes shared by a number of the languages of Malaita and San Cristobal (Makira) Islands in the southeast Solomons, indicating that they form a genetic subgroup, i.e. that they are probably descended from a more recent ancestor, ‘Proto-Cristobal-Malaitan’, in which this innovation took place.

Reconstruction and subgrouping are delicately connected: if we reconstructed POc *-t- instead of *-s- for this correspondence, then the Malaitan languages would not be innovators in this case, but the other languages in Table 1 would be instead. Comparative linguists use various techniques to work out the probable directions of phonetic change: some sound changes are generally known to be unidirectional (e.g. Manam -r- from -s- is an instance of a fairly common change in human language, also reflected in Latin operis, the genitive form of opus ‘work’), and the reconstruction of POc *s is also supported by non-Oceanic Austronesian languages which indicate its presence in proto-languages of higher order than POc.

Subgrouping under the Comparative Method can be applied recursively to identify subgroups within subgroups, that is, to construct what is conventionally called the “family tree” of a set of genetically related languages.

The Comparative Method produces several kinds of result in addition to subgrouping. Firstly (and we return to this in the section on kinds of subgroups) different kinds of subgroup, with different kinds of history, can be identified. Secondly, reconstructed vocabulary, and especially reconstructed terminologies, can illuminate the culture and environment of the people who used them. And thirdly, like other scholars, linguists assume that dispersal centres are most likely to be in the areas of greatest genetic diversity. Thus, if there are two primary subgroups in a family which occupy contiguous geographical areas, the most likely dispersal centre is around the geographic seam.

One thing that comparative linguists cannot do is to provide absolute dating of the linguistic splits posited in a family tree hypothesis. Reconstructing sequences of putative splits provides only relative chronologies.[5] To give absolute dates to prehistoric linguistic events, we need to be able to relate them to archaeological events.