-
Armu. Rev. Anthropol. 1993. 22:425-59Copyright 1993 by Annual
Reviews Inc. All rights reserved
AUSTRONESIAN HISTORICALLINGUISTICS AND CULTUREHISTORY
Andrew Pawley and Malcolm Ross
Department of Linguistics, Research School of Pacific Studies,
Australian NationalUniversity, Canberra, ACT 0200, Australia
KEY WORDS: Austronesian, historical linguistics, culture
history, methods
INTRODUCTIONAbout one sixth of the world's languages are
Austronesian (AN), but it is theircultural and biological diversity
and their predominantly insular distribution,and not their numbers,
that have made the Austronesian-speaking peoples ofgreat interest
to anthropologists. In westem Melanesia, for instance, there
aremany small, culturally and biologically heterogeneous
communities living insustained intensive contact with speakers of
non-AN languages and with eachother. In Polynesia, on the other
hand, sister populations have diverged inisolation. "Islands as
laboratories" has long been a popular catchcry amongscholars
working in Polynesia (69, 108, 113, 161, 179-181, 195, 227,
228).For those searching for principles of change or seeking to
reconstruct events ofculture history, each isolated Polynesian
isleuid or island group has the value ofbeing a relatively
independent witness to the effects of variables such asgeography,
technology, population size, and time on a common ancestral
base.
Although it lacks the elegant simplicity of its Polynesian part,
the widerAN-speaking region provides extremely rich material for
culture historiansand typologists. Perhaps 5000-6000 years ago
Proto Austronesian (PAN) wasspoken, almost certainly somewhere in
East or Southeast Asia, by a neolithicpopulation (7, 8, 236). Some
AN speakers became the world's first efficientlong distance
navigators (7, 153, 206) and over several millennia the family
0084-6570/93/1015-0425$05.00 425
-
426 PAWLEY & ROSS
scattered around two thirds of the earth's circumference, from
Madagascar toEaster Island, and over 70 degrees of latitude, from
Taiwan and Hawaii toNew Zealand. In Island Southeast Asia,' AN
languages largely replaced in-cumbent speech traditions. In the
Central Pacific and in Madagascar, ANspeakers settled islands that
previously had lain beyond the reach of Homosapiens. In westem
Melanesia, known to have been settled for more than40,0(X) years
(3), AN speakers had a lesser impact but established themselvesin
many coastal regions of New Guinea and on almost all the smaller
islands.During this diaspora, as the colonists encountered other
peoples and culturesand adapted to new ecological contexts, there
was a considerable diversifica-tion in technology, social
organization, cosmology, and biological makeup, aswell as language
(7, 9,10, 34,72,161, 162,171,192,200, 244).
The culture historian seeks to make sense ofthe similarities and
differencesexhibited by AN speakers, aligning evidence provided by
different disciplines.This task is made difficult by two types of
methodological problems. First,there are some large gaps in the
data provided by each of the contributingfields of study. Second,
there are problems of synthesis. Whereas each disci-pline and
subdiscipline has it own kinds of data and an array of techniques
forinterpreting these data, culture history has no adequate
procedures for marry-ing the evidence of different disciplines.
The AN region does provide some especially favorable contexts
for com-paring the findings of two ofthe pillars of culture
historical research: archaeol-ogy and historical linguistics.
However, relations between the disciplines areuneasy. For example,
Diebold (74:19-20) refers to a "chronic alienation"between
archaeologists and linguists concerned with reconstructing
Indo-European prehistory. Green (131) observes that many Pacific
archaeologistsbelieve that each discipline should stick to its last
t^hat synthesis is rarelyfeasible because archaeology and
linguistics operate with such different dataand methods. There are
linguists who think likewise.
In most syntheses of Holocene culture history in Island
Southeast Asia andthe Pacific (7-11, 29, 150, 204, 205, 220, 232,
234, 239) historical linguisticshas provided much of the main
storyline. This is particularly true for nonmate-rial culture: no
other discipline has quite such coherent tales to tell. Thecomments
of an archaeologist writing about Indo-European prehistory fit
theAustronesian scene equally well: "there is a sort of horrible
irony in the factthat, while modem archaeologists are greatly
interested in reconstructing tbesocial systems of prehistoric
peoples, historical linguists offer the archaeolo-gists such
detailed reconstructions that they are still beyond
archaeologicalretrieval even when we know what to look for"
(175:122-123). Archaeology
1Island Southeast Asia comprises Indonesia (except Irian Jaya),
Malaysia, the Phillippines and
Taiwan.
-
AUSTRONESIAN HISTORICAL LINGUISTICS 427
in the Austronesian region started much later than linguisticsin
most areaslittle or no work was done before the 1950s or 1960sand
it is understandablethat many archaeologists at this stage prefer
to write their own stories. But it isdifficult to see how culture
historians can avoid the responsibility of evaluat-ing competing
interpretations of their materials against the full range of
evi-dence provided by all the relevant disciplines.
Blust (29) points out instances in which archaeological and
linguistic evi-dence may yield directly comparable evidence and so
corroborate or contradicteach other's conclusions. For example, the
suite of artifacts recovered from asite may be compared with the
names for artifacts reconstructible from cognatesets and
attributable to an earlier linguistic stage. There are other
domainswhere their testimony may be usefully complementary. Unlike
reconstructedlanguages, archaeological assemblages may be
associated with secure loca-tions and reasonably secure dates and
the artifacts themselves will revealdetails of culture and culture
change not recoverable from linguistic compari-sons. On the other
hand, in any prehistoric culture there are many elementssuch as
social categories, belief systems, and easily perishable artifacts
that arenot directly accessible to archaeological methods but may
be partly recover-able from reconstructed vocabulary. Then again,
archaeologists sometimesfind it hard to decide whether a sequence
of assemblages in a region showscontinuity of tradition (with
internally generated change) or discontinuities(the intrusion of
foreign traditions). A good example from the Pacific is therecent
debate on the origins of the cultural complex associated with
Lapitapottery.
This highly distinctive ceramic tradition appears in the second
half of thesecond millennium BC in South Pacific sites spread over
4500 km from theBismarck Archipelago to West Polynesia. One school
of thought believes thatthe core elements of the Lapita cultural
complex derive, with some localadaptation, from an intrusive
cultural tradition brought into the BismarckArchipelago by AN
speakers from Southeast Asia. Another school argues thatmost
elements in the Lapita assemblages continue traditions that were
estab-lished in the Bismarck Archipelago or New Guinea area well
before theappearance of Lapita pottery. Proponents of the AN
intrusion interpretationargue that historical linguistics strongly
supports their view. Supporters oflocal continuity prefer to treat
the Lapita issue as a purely archaeologicalmatter, to which
linguistic evidence is irrelevant, or to aver that the
linguisticevidence is itself unsound or inconclusive. Certainly,
the appeal to linguisticevidence can only be decisive in this
debate if particular archaeological assem-blages can be securely
associated with particular linguistic traditions.
Linguistics and comparative ethnography each contribute in
different waysto reconstructions of social organization: one
reconstructs terminologies forparticular linguistic stages and the
other elucidates the range of behaviors and
-
428 PAWLEY & ROSS
ideologies cissociated with types of terminologies. Many
anthropologists andlinguists have risen to the challenge presented
by the AN speakers' diversity insocial structure (e.g. 34, 51, 97,
98, 100, 108, 112, 113, 177, 183, 185, 186,227-229). Some studies
rely on typology as the basis for historical inferences.Murdock
(185, 186) believed that his sample of AN-speaking societies
pro-vides forceful evidence supporting his method of reconstructing
the develop-ment of kinship system types. In a study of "adaptive
radiation" in Polynesiansocieties, Sahlins (228) sought to
correlate differences in social stratificationwith differences in
economic and social conditions imposed by environmentaldifferences,
e.g. between large, fertile, high islands and small islands
andatolls. In another paper (229), he argued that Polynesians and
Melanesianshave different types of political structures with
different potentials for change.
It is noticeable, however, that much writing on culture history
is marred bya weak understanding of linguistic methods, which
sometimes results in anuncritical use of linguistic data or an
uncritical acceptance of speculativeproposals made by certain
linguists. There is uncertainty about what each ofthe methods of
comparative linguisticsprincipally, the genetic comparativemethod,
internal reconstruction, dialect geography, typological
comparison,lexicostatistics, and the age-area methodis good for.
The most recurrentmisuses of linguistic evidence in reconstructing
social organizations are prob-ably those that stem from a confusion
between the genetic and typologicalcomparative methods of
reconstructing elements of prehistoric cultures. It isperhaps not
widely understood that the genetic comparative method operateson
radically different principles from the typological method and
that, whenthe data it requires are plentiful, the former is a far
more reliable instrument ofhistorical reconstruction. (Equally, it
must be said that without ample data, thegenetic method cannot be
applied rigorously.)
The limitations of comparative typology for sociological
reconstruction arewell illustrated by attempts (94,177, 186) to
reconstruct early Malayo-Polyne-sian or early Oceanic patterns of
sibling classification on the basis of thegeographical distribution
and statistical frequency of certain structural types.The
conclusion was drawn that early Oceanic speakers did not
distinguishterms for older and younger same-sex siblings or
separate male and femaleterms for cross-siblings. As several
commentators have observed (41, 60, 62,66), these conclusions are
powerfully contradicted by the distribution acrosssubgroups of
cognate terms for these same distinctions.
In this review we examine research in AN historical linguistics
that carriesimplications for culture history, focusing equally on
the facts and theories thathave emerged from empirical research and
on questions of method and evalu-ation. We treat four linguistic
domains especially germane to culture historicalreconstruction: I.
the genetic comparative method, 2. subgrouping, 3.
lexicalreconstruction, and 4. continuity and change. These domains
bear on several
-
AUSTRONESIAN HISTORICAL LINGUISTICS 429
broad questions: Where and when were PAN and later interstages
spoken?What can be inferred about the society, technology, and
physical environmentof speakers of PAN and AN interstage languages
from reconstructions ofvocabulary? Which features of a particular
culture (contemporary or attestedby archaeological or linguistic
reconstructions) represent continuity of an an-cestral tradition
and which represent innovations? Why have some AN lan-guage-culture
systems been much more conservative than others?
A NOTE ON RECENT GROWTH IN AUSTRONESIANLINGUISTICS
The existence of the AN family was recognized as early as 1708,
from wordlists brought to Europe from Madagascar, Indonesia, and
Polynesia (144,216).During the nineteenth century and the first
half of the twentieth, research onthe family went on at a leisurely
pace. The high point of comparative work inthis period was the
publication in the 1930s of a three-volume study of ANhistorical
phonology and the reconstruction of some 2000 PAN roots
withsupporting cognate sets by the German scholar, Dempwolff
(73,73a,b).
In the 1950s and 1960s the pace picked up. Indonesian studies
have longhad academic centers at the Universities of Leiden and
Hamburg, but until themiddle of the twentieth century, most of the
descriptive and comparative workin AN was done by amateurs, chiefly
missionary scholars. After World War na handful of professional
linguists entered the field in the United States, theUnited
Kingdom, and the Antipodes, and highly productive research
centersemerged at the Universities of Auckland and Hawaii, the
Australian NationalUniversity, and later, at the Language Center in
Jakarta. A number of joumalsand publication series specializing in
Austronesian or Pacific languages werefounded, including Oceanic
Linguistics, Pacific Linguistics, Philippine Jour-nal of
Linguistics, Nusa, and Language and Linguistics in Melanesia.
Regularinternational conferences have been held since 1974. Of the
1600 or so entriesin a recent select Austronesian bibliography
representing research done overthe last 150 years (56), over 70%
are dated after 1970.
Looking at works written around 1970 gives a sharp reminder of
how muchhas changed in the Austronesian linguistic scene over the
last few decades.Even the perceived size of the family has doubled.
The number of languageswas frequendy estimated at close to 500
(86,119,198). The most authoritativerecent surveys (132, 250) give
figures of 1000-1200, making AN the world'slargest well-established
language family, rivaled only by the less secure Niger-Congo
grouping. Both are far ahead of the next largest established
families,which are in the 150-200 range.
Why such dramatic growth in the perceived size of a family with
a historyof study going back more than 200 years? Until recendy,
several regions.
-
430 PAWLEY & ROSS
particularly in Melanesia and eastem Indonesia, remained poorly
known tocomparative Austronesianists, who were few in number and
who concentratedtheir attentions on a small selection of well-known
languages. During thescores of regional surveys carried out over
the last few decades (e.g. 17, 54,132, 177, 211, 219, 230, 231,
248, 251, 255, 263, 265, 266) hundreds oflanguages spoken by small
communities have come to light. Some 460 lan-guages are spoken in
Oceania (Melanesia, Micronesia, and Polynesia). TheIndo-Malaysian
region contains some 500 AN languages, the Philippinesabout 150,
and Taiwan 17, while a handful of others are spoken in
Vietnam,Thailand, and Madagascar. The ambitious two-part Language
Atlas of thePacific Area (265, 266), though already out of date in
some details, gives afairly accurate picture ofthe number and
locations of languages.
THE GENETIC COMPARATIVE METHOD OELINGUISTICS
The genetic comparative method is the fundamental method of
historical hn-guistics. Often simply called "the comparative
method," it differs sharply fromthe typological comparative method.
Typologists compare structural systems,seeking to determine
universal principles of association between types ofcategories or
subsystems. The statistical frequency of particular associationsmay
form the basis of historical inferences about possible earlier
systems anddirections of change. A crucial weakness of the
typological method, as aninstrument of culture history, is that it
treats types and frequencies but nothistorical particulars. The
high frequency of a particular structural type in aparticular
region or language family may be due not to its great antiquity but
torecent diffusion or to the expansion of one subgroup of the
population.
The genetic comparative method deals with historical
particulars. Its coresubject matter is the body of morphemes
(smallest meaningful elements) thatrelated languages have inherited
from a common ancestor. The strengths of themethod stem from
several peculiar characteristics of human language thatmake certain
components of languages strikingly like species in their mannerof
continuity and diversification. A meshing of these features defmes
thenotion "genetic relationship" for languages.
A key peculiarity of language is that its morphemes are composed
of unitsof sounds (phonemes) that by themselves have no meaning.
Because a lan-guage has from a dozen to several dozen phonemes and
most morphemescomprise several phonemes, there is a vast range of
possible morphemicforms. Consequently, its morphemes (other than
onomatopoeic elements) willshow a high degree of arbitrariness in
their sound-meaning pairings. Thus,different languages are most
unlikely to have more than a tiny percentage of(non-onomatopoeic)
morphemes that resemble each other by chance. The fmal
-
AUSTRONESIAN HISTORICAL LINGUISTICS 431
peculiarity, crucial for the genetic method, is that sound
change in the lexiconof any well-defined speech community is
largely regular. Sometimes otherchanges, such as borrowings or the
loss of old words, obscure the details ofsound changes, but
ordinarily, the linguist can identify recurrent sound
corre-spondences between related languages. Proof that morphemes in
differentlanguages are cognate (related by direct inheritance from
a common ancestor)rests not at all on superficial similarity in
form but on whether the morphemesdisplay regular sound
correspondences. These factors make it possible todistinguish
genuine cognates from convergences and borrowings. The chancesare
virtually nil for languages to independently develop such regularly
corre-sponding roots as the following set: Tagalog hipag
'brother-in-law,' Malayipar 'related by marriage,' Sa'a (Solomon
Is.) ihe 'brother-in-law,' Wayan (W.Fijian) iva 'son-in-law;' or
the set Tagalog ba:go, Malay baharu, Sa'a haalu,Wayan vou, Tongan
fo'ou, all meaning 'new.'
Commentators have occasionally suggested that although the
principle ofsound change regularity is the key to unraveling the
history of Indo-European,it will not hold for language families
such as AN (111). In AN, as in Indo-European, intensive borrowing
between dialects and neighboring languageshas sometimes created a
tangle that is almost impossible to unravel (123,125),but the
principle has proved highly effective in making sense of the
history ofAN languages. In fact, in no group of languages is the
regularity principlebetter exemplified than in the isolated
languages of the Polynesian Triangle(21).
The linguist uses recurrent sound correspondences, together with
a logicthat specifies the conditions under which phonological
distinctions must beattributed to the parent language in order to
construct a theory of the soundsystem of the proto-language. This
yields a formal reconstruction of the proto-form for each cognate
set. Reconstruction of the meaning attributable to aproto-form is
straightforward if all witnesses agree. Otherwise the meaningmay be
determined by a theory of semantic change together with
subgroupingconsiderations (44,46,51,131).
Unlike the typological method of historical reconstruction, the
genetic com-parative method classifies languages not on the basis
of their shared similari-ties in structure but according to the
distribution and weighting of sharedchanges to a reconstructed
ancestral language. Armed with a theory of theproto-phonology and
proto-lexicon, the linguist may be able to identify inno-vations
peculiar to certain members of a family and thus arrive at a
reasonablefamily tree. A mass of uniquely shared innovations or a
smaller number ofunusual ones indicates a period of common
development apart from otherlanguages. The most significant
innovations are usually certain kinds of regu-lar sound changes,
idiosyncratic sound changes in particular words, andchanges in the
structure of morphological paradigms.
-
432 PAWLEY & ROSS
The genetic method has strict limitations. It is only
secondarily concemedwith the comparison of noncognate bits of
language. It is not equipped toinvestigate structural convergence
due to bilingualism, genetic drift, or chance.The genetic method
can, however, be allied usefully with other methods. Forexample,
evidence about possible or favored kinds of structural change,
takenfrom typology, may help linguists to choose between competing
historicalinterpretations of comparative evidence.
AUSTRONESIAN SUBGROUPING AND DISPERSAL
Subgrouping (cladistics or family tree construction) plays a
central role in thecomparative method, especially in lexical
reconstruction and the chronologyof changes. Although the sound
system of the ancestral language of a hetero-geneous family can, in
principle, be reconstructed without reference to sub-grouping
assumptions, this is not the case for the lexicon. Each phoneme
oftheproto-language has a determinable outcome or reflex in every
daughter lan-guage in those morphemes that are retained from the
ancestral stock. (Even if aphoneme is lost, that loss is detectable
as a zero reflex, i.e. nothing, in theposition where other
languages have a positive reflex. For example. Way anand Tongan
zero correspond to Tagalog g, Malay r, and Sa'a / in the
compari-sons given earlier.) Unlike the phonemes, not every
morpheme or word of theproto-language is continued in each daughter
language. Inferences about theantiquity of a word shared by two
members of a fannily (a cognate set) dependon how closely the two
languages are thought to be related, i.e. about thestructure of the
family tree.
Subgrouping also provides a relative chronology for changes in a
speechtraditionwhether these are generated internally or borrowed
from other tra-ditions. Finally, by indicating the geographic
centers of genetic diversity, afamily tree may give strong clues
about directions of language dispersal. Dyen(81) has developed a
set of procedures for inferring the most probable disper-sal
centers or "homelands" of language families or subgroups whose
intemalrelationships are known, based on the principle of fewest
moves.
At least five radically divergent hypotheses about the high
order subgroupsof AN have been proposed. Until the 1930s views on
AN high order subgroupswere ill-defined but most commentators
(without providing any sound justifi-cation) spoke of four main
branchesIndonesian, Melanesian, Micronesian,and Polynesian. The
last three groups corresponded to the familiar geographicregions,
and "Indonesian" (better labeled "Western AN") encompassed all
thelanguages not in the other groups, including the languages of
Formosa, thePhilippines, and Madagascar. However, as we shall see,
the status of theMelanesian languages was particularly
controversial.
-
AUSTRONESIAN HISTORICAL LINGUISTICS 433
The Oceanic HypothesisA radically different view was
foreshadowed by Kern (157) and developed byDempwolff (73).
Dempwolff argued that the Melanesian, Polynesian, and allbut a few
Micronesian languages fall into a single subgroup (today
called"Oceanic") apart from all Western AN languages. The
Micronesian outsidersare Chamorro, Belauan, and possibly Yapese,
all spoken on the western mar-gin of Micronesia. The Oceanic
hypothesis, now accepted by virtually allAustronesianists, has
powerful implications for Pacific culture history. If allAN
languages of the southwest and Central Pacific derive from a
singlelinguistic interstage exclusive of the rest of the family,
the implication is thatthere was a single effective AN colonization
of this area.
Dempwolff based his Oceanic group on the impressive number of
sharedphonological innovations in roots retained from PAN. He first
reconstructed aPAN sound system from comparisons of what he
believed to be a repre-sentative sample of languages. The many
changes uniting the Oceanic lan-guages in his sample included
several mergers (where two or more PANphonemes became a single
phoneme) such as the mergers of PAN *b and *p;of *c, *j, *s, and
*z; of *e and *aw, and of *uy and */. Dempwolff did not saywhether
he regarded Oceanic as a primary branch of AN or as a subgroup
withcertain Western AN languages. But in a postscript (73a: 193-94)
he proposedan explanation for the apparent biological differences
between Polynesiansand Melanesians as well as for the greater
diversity of the Melanesian lan-guages. Rephrased in modem terms:
the AN speakers who entered the Pacific,and who came eventually to
speak Proto Oceanic, were light-skinned, straight-haired people who
had little resistance to malaria. Some Pacific islands werealready
inhabited by dark, frizzy-haired people with stronger resistance
tomalaria and other local diseases. Proto Oceanic speakers were in
contact withsuch people, intermarried with them and were influenced
by them linguisti-cally. Oceanic speakers then spread out over
Melanesia and such contacts wererepeated. In places where malaria
was rife, the malaria-resistant genes of thedarker, frizzy-haired
people dominated. In places free of malaria, such asPolynesia, the
original gene pool of the Oceanic-speaking colonists was
betterpreserved.
The Oceanic hypothesis received further support from early
postwar re-searchers (115-117, 182, 184) who drew the boundaries of
the group moreprecisely and noted certain irregular lexical changes
diagnostic of Oceanic. InMelanesia the boundary between Oceanic and
non-Oceanic was placed justeast of Cenderawasih Bay at the western
end of New Guinea (32, 120).Linguists have tinkered a good deal
with Dempwolffs PAN and Proto Oce-anic sound systems and a number
of refinements have been generally accepted(20, 32, 79, 80, 85,
166, 219, 220). These modifications have reduced the
-
434 PAWLEY & ROSS
number of shared sound changes defining Oceanic, but they still
leave asizeable residue. Several grammatical innovations have been
claimed to bediagnostic of Oceanic (197, 207) together with certain
irregular changes in theform of words (117).
Challenges to the Oceanic HypothesisIn the 1960s the Oceanic
hypothesis was challenged on lexicostatisticalgrounds. A study by
Dyen (82, 84) showed that the AN languages of Melane-sia are
exceptionally diverse in terms of percentages of cognates in a
standardlist of some 200 basic vocabulary meanings. Over 30 of the
40 lexicostatisti-cally defmed first-order branches in Dyen's
classification were confined toMelanesia, a finding that led one
linguist and a few anthropologists (82, 186,244) to propose
Melanesia as the most likely homeland of the Austronesians.This
classification was dramatically at odds with innovation-based
sub-groupings and has gained few supporters. Subsequent
lexicostatistical studiesin AN have shown that different meanings
in the standard 200-word list havewidely varying rates of
replacement (92). They have shown that languagesvary a lot in their
overall lexical replacement rates (37), thereby raising thequestion
why some languages have been much more lexically conservativethan
others.
The lexical diversity of the AN languages of Melanesia was
earlier given aquite different explanation. Some scholars, chiefly
Ray (210) and Capell (53),argued that the Melanesian languages are
not Austronesian in the same senseas other members ofthe family.
Instead, the languages of Melanesia were seenas hybrids, the
products of contact between Indonesian (Westem AN) andPapuan
(non-AN) languages that followed migrations of Westem AN speakersto
various parts of Melanesia. The typical result of such contact in
each regionwas a pidgin or mixed language with an AN superstrate
and a Papuan sub-strate. The Papuan language was thought to differ
radically for each localityand this was the main reason why the
non-Austronesian content of Melanesianlanguages differed so greatly
from region to region. A second part of theargument, spelled out in
detail by Capell, denies that the Melanesian andPolynesian
languages form a subgroup apart from Indonesian. The AN ele-ments
in the Melanesian languages, he said, stem from multiple sources:
anearly movement from Indonesia established AN languages in parts
of Melane-sia, then people from the Philippines, central Sulawesi,
and so on settled indifferent places. Capell specified groups of
words that he believed stemmedfrom particular regions.
The Ray-Capell theory has been attractive to some
anthropologists who seeit as a ready explanation for the perceived
biological and cultural differencesbetween Melanesian and other AN
speakers. For many years the theory wascited in anthropological
surveys as if it were the standard or most probable
-
AUSTRONESIAN HISTORICAL LINGUISTICS 435
interpretation ofthe facts (190). However, there are some
fundamental fiaws inthe linguistic argument. An objection to the
first partthe hypothesis thatPapuan speakers in various parts of
Melanesia adopted AN languages withsubstrate residueswas that in no
case was the Papuan substrate languageidentified (119, 120); it
remained a deus ex machina. Although Papuan-ANbilingualism in the
New Guinea area is common and several recent studies(e.g. 76, 246)
have shown that some AN languages have been greatly affectedby
contact with neighboring Papuan languages, we are asked to believe
thatlanguage shift happened again and again, that all the
Melanesian languages arethe product of pre-AN communities adopting
AN languages. Yet there areparts of Melanesia^Vanuatu, New
Caledonia, and Fijithat probably had nopre-AN inhabitants. The
other and quite crucial objection is that the secondpart of the
Ray-Capell theorymultiple colonization by AN speakers fromdifferent
parts of Island Southeast Asiadirectly counters the evidence for
theOceanic hypothesis. In any case, statistical analysis (184)
shows that there isno solid basis to Capell's correlations between
particular words and particularregions.
The present consensus is that although AN-Papuan contacts have
not con-tributed to the genetic diversity of AN languages, which
the comparativemethod defines in terms of subgrouping relations,
they have contributed muchto the structural and lexical diversity
of languages in the Melanesian area (171,219, 225, 246) and to the
sheer number of discrete languages spoken by smallcommunities.
Recent Proposals about High Order Subgroups
During the 1960s and 1970s a more complex theory of AN high
order sub-groups emerged from work on historical phonology and
morphology. Thepoorly documented Formosan languages, completely
left out of Dempwolff scomparisons, became key witnesses in the
reconstmction of PAN. Severalchanges to Dempwolffs proposed PAN
sound system have been made in thelight of Formosan testimony (71,
84, 87, 222). Dyen noted the possibility of aprimary split in AN
between {a) some or all Formosan languages and {b) agroup
containing all other AN languages (84,87) on the grounds of
phonologi-cal mergers common to all the extra-Formosan languages.
Dahl (71) arguedforcefully for such a primary split. Blust (30, 49)
named the extra-Formosanbranch "Malayo-Polynesian" (MP) and gave a
morphological argument sup-porting it. Although several scholars
have expressed strong reservations (89,91,262), the hypothesis has
gained increasing acceptance (140,212, 223, 243,270).
There are several variants of the Formosan/Malayo-Polynesian
hypothesis(223). According to Blust (30, 49), Harvey (140), and
Reid (212) the Formo-
-
436 PAWLEY & ROSS
PAN
Formosan
Central/Eastem MP
Lesser Sundas,Maluku, etc.
S. HalmaheraAV.New Guinea
Oceanic
Figure J Subgrouping of Austronesian languages after Blust.
san languages may comprise more than one first-order branch of
AN, perhapsdividing into Atayalic (northern), Tsouic (central), and
Paiwanic (southern)groups. Another point of debate (48, 212, 269)
is whether the languages of thePhilippines form a subgroup within
MP. Although most Philippine languagesseem at least superficially
similar to each other, Reid (212) suggests that theyhave no
significant innovations in common. Zorc (269) challenges Reid,
argu-ing that Philippine languages share numerous lexical
replacements and thatthese constitute innovations defining a
Philippine subgroup. The problem inthe Philippines, as in many
other compact regions, is to distinguish innova-tions from
borrowings among related languages that have been in contact
formillennia. A recent study of Tiruray (Mindanao) vocabulary (50)
shows thatthis Philippine language has replaced neariy 30% of its
basic vocabulary withloans from its neighbors.
Blust (30, 31, 37-39, 42, 49) has proposed a more detailed
family tree(Figure 1). In this tree the Westem MP comprises chiefiy
the languages of thePhilippines, Malaysia, westem Indonesia
(including Sulawesi) as far east asmid-Sumbawa, and Madagascar and
Central MP comprises approximately thelanguages of eastem Indonesia
east of Sumbawa and Sulawesi excluding
-
AUSTRONESIAN HISTORICAL LINGUISTICS 437
Halmahera. Oceanic remains, but it has been demoted to something
like afourth-order subgroup.
A caveat is in order conceming Figure 1. It is probably fair to
say that ofBlust's proposed subgroups, MP and three of its
daughtersEastem MP,South HalmaheraAVest New Guinea, and Oceanicare
rather widely acceptedbecause each is based on a significant body
of diagnostic innovations. WestemMP, Central MP, and Central/Eastem
MP, on the other hand, are much moreproblematic. The difficulties
in finding innovations encompassing the entireputative Central MP
group very likely refiect the existence of an earlier exten-sive
and longstanding dialect network in the eastem Indonesian region
(133).What one finds is overlapping innovations, each covering part
of the region.As a whole, Westem MP languages seem to inherit only
the innovationsshared by all MP languages, i.e. those attributable
to Proto MP (PMP). Thissuggests that there was no Proto Westem MP,
but rather that PMP divergedinto a number of dialects, one of whose
descendants became Proto Cen-tral/Eastem MP. The Westem MP
languages are simply those MP languagesthat do not belong to the
Central/Eastem group. In the same vein. Central MPlanguages may be
just those Central/Eastem languages that are not membersof Eastem
MP.
It has long been recognized that linguistic splits are often
imperfecti.e.not sudden and complete, but entailing the gradual
divergence of a chain oflocal dialects. A corollary is that
subgroups are often imperfecti.e. theproduct of a split in a
homogeneous proto-language. Many imperfect sub-groups are formed by
the spread of innovations over parts of a chain ofdialects, over
many centuries, before this chain finally breaks up into
discretelanguages. AN comparativists have begun to give more
attention to the meth-odological problems that ancient dialect
chains present in subgrouping andreconstmction. The most incisive
application ofthe methods of dialect geogra-phy to a single AN
region is Geraghty's (101) study ofthe history ofthe
Fijianlanguages but there have been other studies that examine
dialect chains, con-temporary or past (2,67,70,91,168, 177,
189,205,209, 219,230,248). Ross(219) attempts to build into his
classification of Westem Oceanic languages adistinction between
perfect subgroups, which result from complete splits, andimperfect
subgroups.
Some Rival HypothesesSome scholars have argued for a
Formosan-Philippine subgroup, as op-
posed to the notion of a primary division between Formosan and
the rest. TheFormosan-Philippine hypothesis (83, 90, 91, 93, 253)
is based on what itsproponents see as an impressive number of
exclusively shared cognate sets.They argue that some of these
cognate sets must refiect lexical innovationscommon to the
languages of Taiwan and the Philippines. Dyen (91) has
-
438 PAWLEY & ROSS
recently proposed a variant of this hypothesis whereby the
Formosan-Philip-pine group is part of a chain that also includes
western Indo-Malaysian lan-guages. He labels this chain
"Indo-Formosan." In other words, he implies thatProto Indo-Formosan
remained a dialect chain for many centuries, stretchingover the
long string of islands from Formosa to westem Indo-Malaysia.
Thequestion is whether a dialect chain could have been maintained
over a chain ofislands as far flung as those of Taiwan, the
Philippines, and westem Indo-Ma-laysia and if so, for how long.
A seductive feature of the Formosan-Philippine and Indo-Formosan
hy-potheses is that the languages in the proposed groups show quite
strong simi-larities to each other (93,261). But the case should
not be overstated. Similari-ties in vocabulary are more obvious
across Philippine languages and certainlanguages of southeastem
Taiwan than across the rest of Taiwan. Stmcturalsimilarities, and
in particular the elaborate system of verbal "focus" (wherebya
range of semantic rolesactor, undergoer, location, instmment, or
benefici-ary^may occur as the "topic" or subject of a clause, with
the role of the topicmarked by a distinctive affix on the verb),
occur across a range of Philippinelanguages and some Formosan and
westem Indo-Malaysian languages (48,201, 223, 243,259,260).
As we have mentioned, the genetic comparative method subgroups
lan-guages by identifying shared innovations attributable to a
common ancestor.The Formosan-Philippine and Indo-Formosan groups
are supported only byputative lexical replacements inferred on the
basis of exclusively shared cog-nate sets, which tend to be shaky
evidence, especially when not supported byinnovations of other
kinds. For one thing, words are the most easily borrowedelements of
a language and it is sometimes difficult to distinguish
betweenshared inherited replacements, which are diagnostic of a
subgroup, and bor-rowings, which are not. More serious is the
difficulty of determining what hasreplaced what in a family with
only two first-order subgroups. Taking cognatesets meaning 'leaf,'
for example, Formosan languages have a cognate setrefiecting
*wciSaw 'leaf (49) whereas MP languages show a distinct cognateset
refiecting *daSun 'leaf.' Did PAN have both forms or only one? If
onlyone, which? When we encounter a sound correspondence like /s/
in LanguageA versus /h/ in Language B, we can be reasonably certain
that the sound intheir common proto-language was *s and that the
direction of change wasfrom *s to /h/ (95). But given the present
data, there is no analogous way ofdetermining whether *daSun
replaced *waSaw, or vice versa, or whether bothwere present in
PAN.
Of course, arguments of this sort are not resolved by a single
piece ofevidence. The proponents of a Formosan-Philippines group
appeal to the rela-tively large number of exclusively shared
cognate sets supporting it. But we
-
AUSTRONESIAN HISTORICAL LINGUISTICS 439
have no means of knowing that these are not shared retentions
from PANand/or borrowings, rather than innovations (51).
The Eastern Malayo-Polynesian Hypothesis and the
IntemalClassification of OceanicA significant body of shared
innovations indicates that the immediate relativesof Oceanic
consist of a South Halmahera-West New Guinea (SHWNG) group,whose
members are spoken in the southem half of Halmahera and
aroundCenderawasih Bay, at the northwestem end of Irian Jaya, close
to the Bird'sHead of New Guinea (31). The name "Eastem
Malayo-Polynesian" is nowgenerally applied to the putative
SHWNG/Oceanic group.
Except for the well-defined Polynesian group and the fairly
well-markedNuclear Micronesian group it has been hard to identify
subgroups of Oceanicthat cover a wide area. However, substantial
evidence has recently beenbrought forward indicating several large
high order subgroups in westemMelanesia (i.e. New Guinea and the
westem Solomon Islands). Ross (219)argues for two such high order
subgroups: a Westem Oceanic group, derivedfrom a dialect network,
and an Admiralties group. He divides Westem Oce-anic into three
large groups: the North New Guinea Cluster, consisting of allthe AN
languages of the north coast and offshore islands of Papua
NewGuinea from Aitape to the Huon Gulf together with the languages
of thenorthem coast of New Britain west of the Willaumez Peninsula
and much ofsouthem New Britain; the Papuan Tip Cluster, containing
the Oceanic lan-guages of southeastem and central Papua; and a
third very large group, calledMeso-Melanesian, encompassing the
languages of northem New Britain eastof the Willaumez Pensinula,
the Bali-Vitu group. New Ireland, and the westemSolomons. Each of
the three larger groups appears to derive from an olddialect
network. Evidence is insufficient to classify the languages of the
St.Matthias group, to the north of New Ireland, but these may well
constitute aseparate high order subgroup within Oceanic.
No very large subgroups have been clearly identified elsewhere
in Melane-sia. The more important currently accepted groups include
Southeast Solo-monic, centered in Guadalcanal, Malaita and San
Cristobal (168, 196, 251);Central and Northern Vanuatu (67, 196);
Southem Vanuatu (170, 172); andNew Caledonia-Loyalties (103, 142).
A Nuclear Micronesian subgroup isgenerally recognized (17, 18,
154), comprising the languages of geographicMicronesia excluding
Chamorro, Belau, and Yapese. Its center of geneticdiversity is in
the east, in the region of Kiribati, Kosrae, Pohnpei, and
theMarshalls (154).
Much effort has gone into fmding the immediate relatives of
Polynesian.The consensus is that the Fijian languages and Rotuman
are, by a smallmargin, its closest kin and the name "Central
Pacific" is given to this putative
-
440 PAWLEY & ROSS
Proto Oceanic
Admiralties Southeast Nuclear Central/NorthSolomonic Micronesian
Vanuatu
NewCaledonia-Loyalties
Westem Oceanic Central Pacific
North NewGuinea
Meso-Melanesian Fijian Polynesian
RotumanPapuan Tip
Figure 2 A partial subgrouping of Oceanic languages.
subgroup (101,102,115,118, 196,199). Geraghty (101) suggests
that Polyne-sian and the eastem Fijian dialects were once a unity
distinct from westemFijian. The first split in Polynesian appears
to have been between a Tongicbranch comprising Tongan and Niuean,
and a Nuclear Polynesian branchcomprising the 28 or so remaining
Polynesian languages (21, 63, 64, 193,194).
A wider "Eastem Oceanic" subgroup, comprising at least Central
Pacificplus Central and Northem Vanuatu and perhaps Southeast
Solomonic has beenproposed (101, 174, 196) but the evidence for
such a group is so far uncon-vincing. Figure 2 gives a summary of
important Oceanic subgroups that havesome degree of general
acceptance.
Dispersal Centers: TheAge-Area MethodOccam's Razor tells us that
the most likely primary dispersal center for agenetic group of
species or languages is the area of its current greatest
geneticdiversity. The subgrouping outlined in Figure 1 places the
most likely primarydispersal center for Austronesian in the region
of Taiwan and the northemPhilippines, where the Formosan and
Malayo-Polynesian groups are contigu-ous. The same subgrouping also
implies (a) a Philippines dispersal center forthe Malayo-Polynesian
branch with a subsequent fanning out across the Indo-Malaysian
archipelago, {b) an eastem Indonesia dispersal center for Central
&Eastem Austronesian, and (c) a dispersal center either in
northem Halmaheraor Cenderawasih Bay in New Guinea for Eastem MP.
Blust (32) favors Cen-
-
AUSTRONESIAN HISTORICAL LINGUISTICS 441
derawasih Bay because of the apparently greater genetic
diversity of its lan-guages.
Taken together. Figures 1 and 2 imply a primary dispersal center
for Oce-anic in westem Melanesia (specifically, the Bismarck
Archipelago or thefacing northem cocist of New Guinea). The
structure ofthe Oceanic family treehas been interpreted by some
(200, 205) as indicating a rapid dispersal ofOceanic-speaking
peoples from northwestern Melanesia across southemMelanesia and
into the Central Pacific following the breakup of Proto
Oceanic.This interpretation is based on the observation that there
is no well-definedcenter of genetic diversity within Oceanic
itself. If there had been an initialbreak of Oceanic into several
languages in northwestem Melanesia, followedmuch later by movements
into southem Melanesia (the southeast Solomons,Vanuatu, New
Caledonia, and Fiji), Polynesia, and Micronesia, one wouldexpect to
find the languages of the later-settled regions subgrouping with
oneor another of the major northwest Melanesian groups, but this is
not the case.
LEXICAL RECONSTRUCTION AND CULTURE HISTORY
Lexical reconstmctions were first used over a century ago to
elucidate earlyAN culture (156) but a much richer body of evidence
has now accumulated.As Dempwolffs (73) phonological and lexical
reconstmctions included noFormosan languages they are now generally
attributed to PMP rather thanPAN. In a series of papers Blust
(25-28, 33, 39, 43, 45) has increased thenumber of cognate sets
requiring PMP or PAN lexical reconstructions to over4000. He is
consolidating this material into a massive comparative
dictionarythat will include reconstmctions for various stages from
PAN down. At thePAN level, constraints of method and data continue
to hamper reconstmctions.Only those cognate sets represented in
both Formosan and Malayo-Polynesiancan be attributed to PAN etyma.
The small number of Formosan languages andthe poor quality of most
descriptions means that the number of Formosancognates with
Malayo-Polynesian is modestperhaps little more than 1000.
A Proto Polynesian lexical file, begun in 1965 at the University
of Auck-land as part of a broad Polynesian culture history project,
has grown to over3000 reconstmctions (23). A Proto Micronesian file
of some 1300 roots hasbeen assembled at the University of Hawaii. A
Proto Oceanic lexicon andthesaums with perhaps 2000 lexemes is in
preparation at the Australian Na-tional University.
A good deal of lexical work has been concemed chiefly with
historicalphonology and has paid little attention to the fine grain
of semantics. There isnow a move to examine closely the history of
particular semantic fields,especially those of interest for culture
history, and to specify reconstmctedmeanings more precisely. Recent
studies have examined terms for horticulture
-
442 PAWLEY & ROSS
(99), wet rice cultivation (215), plant names (24, 249), canoe
parts and sailing(206), buildings (44), cooking techniques (169),
fish and/or fishing (105, 149,257), birds (65, 68), dog and pig
(173), and aspects of early AN social organi-zation (34, 35, 38,
40, 51, 131, 201, 167). Blust (39) reconstructs about 180terms for
marine and land-based flora and fauna, and for climate and
topogra-phy, showing the levels of his family tree (PAN, PMP, etc)
at which they arereconstructible. Zorc (270) has taken about 1300
PAN and PMP reconstruc-tions as the basis of a sketch of early AN
culture history. Chowning (61)surveys a wide range of cultural
fields, noting cognate sets within Oceanic thatare candidates for
Proto Oceanic status.
Dispersal Centers: The Evidence of Lexical
ReconstructionsLexical reconstructions for environmental features
sometimes provide evi-dence for or against centers and directions
of dispersal. Cognate words forcertain plants and animals
characteristic of the Indo-Pacific tropics and sub-tropics are
found in regions as widely separated as Formosa and Polynesia.Such
widespread terms do no more than indicate an AN homeland
somewherein this vast region (29,40, 86,156, 270). However, Blust
(38) does better withcognate sets for certain mammals restricted to
one side of the Wallace Line,which separates the Asian from the
Australian (including eastem Indonesian)faunal zones. Cognate names
for the scaly anteater (reflecting a proto-form*qaRem), a monkey
taxon (HUCUT)), and ruminants (probably deer)[*(qa)Nuarf 'ruminant
taxon,' *salajer) 'male, of ruminants,' and *(q)uReT)'horn'] are
found in languages of Formosa, in certain parts of westem
Indo-Malaysia, and (except for the scaly anteater) the Philippines.
If one accepts thehypothesis of a primary division between Formosan
and Malayo-Polynesian,one must attribute all of these etyma to PAN.
The inference can then be drawnthat PAN was spoken west of the
Wallace Line.
Blust (40) also points out that PAN had many etyma associated
with the seaand a system of orientation distinguishing between
*daya 'landward,' and*lahud 'seaward,' characteristic of people who
live on the coast or on islands.Other words were *baRiuS 'typhoon'
and *qamiS(-an) 'north' or 'cold sea-son.' The last two etyma taken
together indicate a homeland north of theequator and perhaps on the
margin of the tropical zone, for which southemTaiwan or the northem
Philippines are reasonable candidates.
Technology and the Austronesian DispersalSeveral commentators
(10, 12, 126, 127, 205, 232) have argued that horticul-ture was
another key element in the Austronesian diaspora. Possession of
avariety of crops may have enabled ANs to replace or marginalize
non-farmingpopulations in Island Southeast Asia and to survive on
small islands withimpoverished biota in the Central Pacific (126,
129). Blust (29) attributes to
-
AUSTRONESIAN HISTORICAL LINGUISTICS 443
PAN a cluster of terms for rice and millet: *pajay 'rice plant,
paddy,' *beRas'husked rice,' *Semay 'cooked rice,' *ZaRami 'rice
straw,' and *zawa 'millet'(all reflected both in Formosan and
Indonesian witnesses), as well as *qumah'garden, cultivated field'
(reflected also in Oceanic). Rice is present in earlyNeolithic
sites in Taiwan, the Philippines, and Borneo (15). However,
riceagriculture goes back at least 6000 years in mainland Southeast
Asia (15), withwet field (paddy) rice probably preceding dry field
(swidden) rice, and itwould be surprising if PAN speakers lacked
terms for wet rice technology. Itmay be significant that northem
Luzon languages continue PAN *pajay and*Semay by terms that mean
'paddy rice' and 'swidden rice,' respectively(215). A group of
terms for root crops are attributable to PMP but not to PAN:*tales
'taro: Colocasia sp.', *qubi 'yam: Dioscorea sp.,' and *biRaq
'giantarum: Ahcasia sp.' All the root crop terms persist in
Oceanic, but none ofthose for grain crops do, indicating that rice
and millet were not part of theProto Oceanic economy. Names for
sago (PMP *Rambia), bananas (PMP*pu(n)ti), sugar cane (PAN *tebuS),
and a range of tree and ground fruits canalso be attributed to PMP
or PAN.
Over 20 boat and sailing terms can be reconstructed for PMP
(206). Theseinclude terms for several outrigger parts, hull
planking, and sail, indicating thatthese were part of PMP
technology as one would expect from the comparativeethnographic
evidence (134). None of the terms for outrigger parts have
cog-nates in mainland Formosan languages. Linguistics cannot tell
us whetherFormosan peoples lost the outrigger technology or whether
they never had it.
Austronesian Social OrganizationLinguistic evidence has been at
the center of an ongoing debate about thenature of kinship groups
and marriage systems in early AN society. Using acomparative
typological method, relating a variety of social factsrules
ofresidence, marriage and descent, kinship terms, etcMurdock (185,
187)concludes that early Austronesians had bilateral kindreds and a
Hawaiian-typekinship terminology rather than descent groups and
distinct terms for mother'sand father's siblings. Goodenough (112),
an isolated partial dissenter from thisview, argues for
land-holding descent groups, which by definition could nothave been
ego-centered kindreds. He notes a cognate set indicating
*kainaya'land-holding descent group,' attributable to the common
ancestor of Polyne-sian and Nuclear Micronesian languages.
Blust (34) argues that careful attention to linguistic evidence
suggests aradically different set of conclusions from those reached
by Murdock. Draw-ing on lexical agreements between Oceanic and
Philippine-lndo-Malaysianlanguages he reconstructs a system of
terms that are attributable to PMP andsometimes to PAN. Unless
otherwise indicated, the following are all attribut-able to PMP:
*Rumaq 'house, lineage,' *datu 'lineage or clan priest or offi-
-
444 PAWLEY & ROSS
cial' (with compound terms meaning 'male *datu' and 'female
*datu' or'*datu ofthe upper vs lower half), *suku 'quarter, limb;
section in a quadri-partite society,' *na Se(m)pat na baiay 'four
houses; four parts of a society,'*Sua(n)ji 'blood kinsman,'
*ma(n)tuqa 'MBAVF,' *dawa 'ZS/DH,' *[aya'FZS/ZH (ms),' and PAN *aya
'FZ.' Drawing on Murdock's work Blust notes(34:220) that the PMP
system of kin terms "shows a negative statististicalcorrelation
with exclusive organization around the bilateral kindred, and
henceimplies the coetaneous presence of descent groups." He
concludes that, takentogether, the linguistic material and the
theory of social types require theinferences that PMP society 1.
had descent groups that were subsequently lostin a broad culture
area extending over the Philippines and westem Indo-Ma-laysia, 2.
practiced preferential marriage to a classificatory mother's
brother'sdaughter, 3. proscribed marriage to father's sister's
daughters, 4. ordered line-ages into a dual division, with its
associated universal cosmological scheme,and 5. recognized a
four-part division of "houses," representing descent orresidential
groups. He also suggests that there is a strong case for
inferringasymmetrical exchange of wives between marriage
classes.
Blust regarded his study as an illustration of the greater
reliability of thegenetic comparative method vs comparative
typology for reconstructing his-torical particularities. His paper
provoked a lively response from a number ofcommentators (e.g. 1,
59, 96-98), who suggested that some of his inferenceswere too
strong for the evidence. Sticking to his guns, Blust has developed
theargument in subsequent papers (34,36,51). The debate has
centered on (a) theprecise meaning and antiquity of certain terms
and (b) whether specific kindsof terminologies imply specific norms
of social behavior.
Sahlins' influential paper (229) on political types in Oceania
is more con-cemed with structural principles governing the
evolution of polities than withthe history of particular Oceanic
societies. However, his broad equation ofPolynesia with hereditary
chieftainships and of Melanesia with big-man lead-ership threatened
for a time to be taken as implying distinct historical originsfor
Polynesian and Melanesian societies. But critics have pointed out
thatmany Austronesian-speaking societies of Melanesia have
hereditary chiefs(58, 75). Pawley (201) attributes to Proto Oceanic
a pair of terms *qa-lapas'chief, person of chiefly rank' and
*qa-riki 'first bom son of chief,' indicatingthat a system of
hereditary rank was found in the society whose language
wasancestral to all AN languages of Melanesia, Polynesia, and
Nuclear Microne-sia. Drawing on Malaita cognates, Lichtenberk (167)
modifies the forms to*ta-la(m)pat and *qa ariki for which he offers
the more cautious glosses'leader' and 'oldest child.' It has been
argued (143) that a measure of socialstratification was necessary
to organize colonizing voyages of the scale associ-ated with the
settiement of the Central Pacific. Kirch (158) and Green (131)have
used archaeological evidence to assert that what they term
Ancestral
-
AUSTRONESIAN HISTORICAL LINGUISTICS 445
Polynesian Society was built in part on an inherited framework
of status andrank.
Rapid Dispersal, Mobility, and Linguistic DivergenceThe initial
AN expansion in Oceania was the work of sailing people whoappear to
have moved quickly. For at least the first few centuries after
theLapita expansion, some degree of contact was often maintained
between set-tlements along island chains extending up to 600 km
(126).
If the initial AN dispersal across Island Southeast Asia
followed a similarpattern, how might this have shaped the pattem of
linguistic divergence? Thebest contemporary analogs to the pattem
of early AN differentiation withinlarge islands and compact island
groups in Southeast Asia may well be foundin certain regions
settled quite late by AN speakers. Madagascar (about thesize of
Sumatra and larger in land area than the Philippines) and the
HawaiianIslands (much smaller but scattered) were both probably
settled about AD 600,and New Zealand (about the size of the
Philippines) about AD 1000-1200(242). Madagascar contains a network
of dialects, but only at the margins ismutual intelligibility low,
with less than 60% of cognates on the SwadeshlOO-word list (88,
256). Hawaiian speech shows little regional variation. TheMaori
dialects of the North Island of New Zealand are lexically quite
diversebut all dialects have a high degree of mutual
intelligibility (22).
These cases indicate that AN colonizers were generally mobile
enough tomaintain fairly cohesive dialect networks over large
islands and island groupsfor up to 1000 years or so. Work on the
Fijian and westem Polynesian archi-pelagos (72, 101, 104, 151, 209,
230) and the disti-ibution of languages inPolynesia (200) reinforce
this view. However, to draw any more precise impli-cations from the
Madagascar and New Zealand cases for the early history ofAN we need
careful studies of the dialect geography of these two regions.
CORRELATING ARCHAEOLOGICAL AND LINGUISTICEVENTS
A set of principles has been proposed for connecting
archaeological and lin-guistic traditions in the Central Pacific or
Remote Oceanic region (161, 204).Geographic isolation of the major
islands groups and the fact that colonizationdid not begin until
about 3000 years ago has favored continuity with gradualchange of
the founding archaeological and speech traditions. There is
virtuallyno doubt that the first humans to enter the Fiji-westem
Polynesia region, andlater eastem Polynesia, spoke languages
ancestral to the present day languagesof the Central Pacific
subgroup.
Between 3600 and 3000 BP variants of the Lapita cultural complex
appearacross a wide belt ofthe South Pacific (4, 127, 130, 153,
158, 159, 162, 235,
-
446 PAWLEY & ROSS
237, 239, 240). The principal markers of Lapita sites are
elaborate and highlydistinctive dentate-stamped pottery motifs and
a variety of vessel shapes, butthese markers are usually associated
with other features (127, 130, 159, 162,240) including large
coastal settiements, very often situated on small islandsand always
handy to beaches that would provide good launching sites forboats;
a tool kit containing characteristically shaped stone and shell
adzes andscrapers; obsidian and chert flake tools, often imported
from remote sources;one-piece shell trolling hooks; pearlshell
knives and scrapers; various kinds ofconus shell disks and
pendants; earth ovens; and middens full of bones fromlagoon fish
and turtle and containing chicken and pig bones. The earliestLapita
sites, dating to about 3600 BP, are in the Bismarck Archipelago.
Therethe complex geometric dentate stamped decorative style appears
full blown, ina variant called Early Westem Lapita. By 3200 BP a
slightly modified form ofLapita was present in Santa Cmz, Vanuatu,
and New Caledonia, where it wasevidently the founding culture. By
about 3000 BP another variant, EastemLapita, appears in Fiji,
Tonga, and Samoa, and this is clearly ancestral to laterFijian and
Polynesian material cultures (127-129,161).
This swift spread of Lapita culture across Island Melanesia and
into west-em Polynesia, following a perhaps 400 year period of
earlier development inthe Bismarck Archipelago, is consistent with
the pattem of Oceanic sub-grouping we outlined previously. The
subgrouping indicates a period of Oce-anic unity, most likely in
westem Melanesia, where Oceanic has its immediateextemal relatives,
followed by the breakup of Proto Oceanic into a number ofwidely
dispersed subgroups that are either coordinate or close to
coordinate.
A long period of common development apart from the rest of
Oceanic,perhaps on the order of a thousand years, is indicated by
the phonological,grammatical, and lexical innovations of the
Polynesian group. The archae-ological record shows a
correspondingly long pause between the Lapita hori-zon in westem
Polynesia and the settlement of eastem Polynesia, which ap-pears
not to have begun until early in the first millennium AD (10, 128,
161,204, 242). Irwin (152,153), however, questions whether the
pause is real or anartifact of archaeological sampling and
visibility.
The equation of Lapita with intmsive AN languages is less
straightforwardin the Bismarck Archipelago, where human settlement
goes back more than30,000 years (3), than it is in the Central
Pacific. Countering this equation,proponents of a predominantly
westem Melanesian origin of Lapita (4, 114,258) point out that no
completely satisfactory ancestral tradition for Lapita hasbeen
found either in Southeast Asia or in Melanesia. They argue that
some ofthe elements of the Lapita complex, including horticulture,
lagoon fishing,inter-island trading of obsidian, and earth ovens
predate Lapita in westemMelanesia. A radical suggestion has been
made (245) that in westem Melane-
-
AUSTRONESIAN HISTORICAL LINGUISTICS 447
sia, Lapita may be no more than a kind of pottery, widely traded
and superim-posed on diverse local cultures.
This last suggestion, though perhaps made tongue-in-cheek,
raises severalquestions: How well are archaeologists able to
distinguish between inheritedand intrusive elements in an
archaeological tradition? Under what conditions,if any, does it
make sense to speak of genetic relatedness or continuity
betweenarchaeological assemblages, i.e. implying the transmission
of a coherent(though not unchanging) tradition from generation to
generation within asociety? The term "culture" is notoriously
vague. What assurance do we havethat the archaeologist's "culture"
has the same kind of coherence and transmis-sibility that a
language has? Kirch & Green (161) attempt to answer
thesequestions in the Polynesian context.
Those who think the major elements of Lapita culture originated
in South-east Asia (7, 8, 16, 238, 240) argue that Lapita is not
just pots but a coherenttradition with clear antecedents in Island
Southeast Asia. In westem Melanesiaand further east, Lapita sites
form a well-defined cultural horizon. Theseresearchers assert that
the continuities with pre-Lapita technology in the Bis-marck
Archipelago have been exaggerated.
Even if Southeast Asian antecedents are found for Lapita
technology, canwe be sure that the technology was spread by the
migration of a society ofpeople rather than by the diffusion of
useful elements through trade, etc? Herethe archaeological evidence
is suggestive but at present not decisive. Propo-nents ofthe
migration view appeal to linguistic evidence as the clincher:
wholelanguages do not spread by diffusion. The subgrouping evidence
stronglyindicates that Oceanic speakers entered Melanesia from the
west, and the massof cognate sets for many facets of social and
economic life among contempo-rary AN languages from all regions
supports the argument for migration. Theancestors of the Proto
Oceanic speech community may not have brought alltheir culture with
them but they brought a large part of it. There is an impres-sive
persistence of vocabulary for various cultural domains in some
contempo-rary Oceanic languages of Melanesia, in most of the
Nuclear Micronesianlanguages, and in all of the Polynesian
languages (see references in the pre-vious section on lexical
reconstmction). But cognate sets generally do notindicate the size
and layout of settlements or the fine details of technologymotifs,
adze forms, etcby which archaeologists often unite or
distinguishtraditions.
For much of Island Southeast Asia the archaeological record for
the neo-lithic is sparse. The earliest pottery-bearing sites in the
Philippines and Indone-sia date to around 5000 BP (236). The
earliest sites in Island Southeast Asia,dating just prior to 5000
BP, are found in Taiwan where they are associatedwith the Dapenkeng
culture (57, 160, 236). A number of archaeologists favorTaiwan as
the AN dispersal center on the grounds that the pottery and
material
-
448 PAWLEY & ROSS
culture tradition is probably derived from cultures that existed
in South Chinaat a slightly earlier date (8, 9, 57, 160, 252). Not
all archaeologists, however,support a Taiwanese dispersal center
(178,234).
QUESTIONS OF CONTINUITY AND CHANGEWe tum fmally to questions of
continuity and change: Why have some ANlanguages been extremely
conservative and others extremely innovative? Whyhave AN languages
virtually taken over Island Southeast Asia, with the
partialexception of the eastem islands of Indonesia, while in
westem Melanesia theyshow a much more restricted distribution? And
does continuity of languagenecessarily imply continuity of
community?
Language change is most easily quantified in the domain of basic
vocabu-lary. Blust (37) shows that Malayo-Polynesian languages vary
enormously intheir retention of forms for meanings on a 200-item
basic vocabulary list. (Hisstudy does not include Formosan
languages.) Stmctural change is harder toquantify, but there is no
doubt that some languages have been more conserva-tive than others
in phonology (47, 71, 73) and morphology (195, 196, 219,221).
Island Southeast Asia, New Guinea, and the larger islands in the
BismarckArchipelago and the Solomons have been occupied by Homo
sapiens for atleast 40,000 years (3, 4, 9). As AN speakers
colonized Island Southeast Asiaand westem Melanesia they
undoubtedly encountered established populations.These populations
are perhaps best represented today by the Negrito groupsscattered
through the Philippines, the Malaysia/Thailand border area, and
theAndaman Islands, and by the Papuan-speaking peoples of eastem
Indonesiaand westem Melanesia (8). No indigenous non-AN languages
survive today inTaiwan, the Philippines, or westem Indo-Malaysia.
By contrast, in eastemIndonesia and westem Melanesia numerous
non-AN languages (convention-ally referred to as "Papuan") continue
to be spoken along probable AN migra-tion routes.
There seems to be a rough correlation between the degree to
which ANlanguages have replaced previous languages in a region and
their degree oflexical and structural conservatism. Lexically, the
most conservative Malayo-Polynesian languages are concentrated in
the Philippines and the westem partof the Indo-Malaysian
archipelago (centered in Malaysia, Sumatra, Kaliman-tan, and
Sulawesi). The most innovative languages are concentrated in
eastemIndonesia (Halmahera and Irian Jaya) and Melanesia. In terms
of grammaticalstmcture, the most conservative languages are
concentrated in Formosa, thePhilippines, Madagascar, and to a
lesser extent, in the westem part of theIndo-Malaysian archipelago.
Within Oceanic one can also distinguish betweenconservative and
innovative languages (121, 125). The most conservative
-
AUSTRONESIAN HISTORICAL LINGUISTICS 449
languages, lexically and grammatically, are concentrated in
Fiji, westem Poly-nesia, the westem Carolines, the southeast
Solomons, northem Vanuatu, thenorthem coast of New Britain, and the
nearby Bali-Vitu islands. Within Mela-nesia, the more innovative
languages are found mainly on the northem coast ofthe New Guinea
mainland and in the Markham Valley and Huon Gulf hinter-lands, in
southem New Britain, southem Vanuatu, and New Caledonia.
If there is a single dynamic behind these two sets of facts,
what might it be?It probably has to do with the nature of the
populations encountered byimmigrant AN speakers. From linguistic
and ethnographic evidence, Reid(213,214) reconstmcts a scenario for
early Negrito/AN contact in Luzon in thenorthem Philippines. He
concludes that the Luzon Negritos were hunter-gath-erers,
relatively few in number, and readily dominated by the AN
agricultural-ists. He infers that they shifted at an early date to
AN languages, since the ANlanguages spoken by the Negritos today,
although ultimately subgrouped withother Luzon languages, are not
closely related to any of them. Some Negritosretained much of their
distinctive way of life after the language shift. In othercases
they were absorbed into non-Negrito AN-speaking communities. Onecan
assume that what happened in the northem Philippines was repeated
inwestem Indo-Malaysia, although in this case all non-AN speakers
ultimatelyeither shifted to AN languages or died out. Either way,
the pre-AN populationlacked the demographic and economic muscle to
have much impact on ANlanguages in this region.
The situation in New Guinea was quite different. Evidence
indicates thatpre-AN populations in New Guinea were in many cases
horticulturalists, culti-vating a number of root and tree crops
(109, 110, 268). They were probablymore numerous and sedentary and
therefore much less readily dominated byimmigrants than
hunter-gatherers would have been. There were probably sub-stantial
populations in the hinterlands of the northem coast of New Guinea
andon all the large islands of westem Melanesia when AN speakers
arrived (4).There is little evidence for pre-AN horticulture in the
Bismarck Archipelagobut the proximity to New Guinea suggests that
horticulture may have beenthere. AN immigrants, who traditionally
exploited coastal resources, wouldnot have competed strongly for
territory far from the coast; indeed, the Lapitapeople seem to have
largely skirted the New Guinea mainland, keeping tooffshore
islands.
The distribution of early Lapita sites in westem Melanesia and
the dates fortheir eastward spread suggests that some of the Lapita
people of westemMelanesia were for a few centuries able to keep
their distance from the non-AN peoples inhabiting the main islands.
Before long one or more groups ofthese Lapita island dwellers moved
east. Their descendants who reached Fiji,Polynesia, and Micronesia
had the whole field to themselves and retained ituntil European
contact. It appears that they also found Vanuatu and New
-
450 PAWLEY & ROSS
Caledonia uninhabited, although a degree of doubt remains in
these cases(131).
This separation between immigrants and indigenes in westem
Melanesiawas not maintained for long. AN languages remain dominant
in Island Mela-nesia, but on the New Guinea mainland there is
evidence that communitieshave occasionally shifted from an AN to a
neighboring Papuan language.Dutton (76,77) documents an area of
southeastern Papua where such shifts arein progress. In AN
languages spoken on and close to the New Guinea main-land, contacts
between AN and Papuan speakers have at times led to
profoundstructural changes and to heavy lexical borrowing
(53,55,171,218,246,247).There are numerous AN languages and groups
on or near mainland PapuaNew Guinea that show features attributable
to Papuan contact: verb-final wordorder (replacing Proto Oceanic
verb-medial or verb-initial word order), pre-posed possessors,
postpositions, and loss of Proto Oceanic derivational mor-phology.
Only in the Bismarck Archipelago and the Solomons, where there isno
firm evidence for agriculture in pre-Lapita times and where
communitieswere therefore probably smaller than on the New Guinea
mainland, might weregularly expect to find shifts from a Papuan to
an AN language or AN-in-duced structural change. Madak on New
Ireland is an AN language whosespeakers appear to have once been
Papuan speakers (224). Papuan languageshave partly held their
ground in New Britain and Bougainville. In other partsof westem
Melanesia, Papuan languages have been almost completely re-placed
by AN languages. In general, we do not know how the
replacementswere effected.
The precondition for contact-induced structural change is
widespread biUn-gualism. When a substantial number of speakers are
bilingual in the languagethat is emblematic of their own ethnicity
and in the language of neighbors withwhom they have frequent
contact, the linguistic effects of this contact are oftenfar more
radical than those of a culturally dominant language (that is
oftenspoken by only a small minority of the dominated). The less
extreme outcomeof such bilingualism is contact-induced change in
the emblematic languagetoward the norms of the neighbor language.
The more extreme outcome, asnoted above, is language shift: the
emblematic language is relinquished infavor of the neighbor
language.
After bilingualism is established as a social norm, bilingual
speakers seemto construct meanings in increasingly similar ways in
both their languages.This convergence probably has both a cognitive
and a social foundation. Onone hand, bilinguals tend to integrate
the different conceptual resources of-fered by their two languages
into a single intellectual system. On the otherhand, when
foreigners talk to each other, they adapt to each other's
usages,choosing common ground and chopping off the noncongruent
edges. Bilin-gualism in premodem societies has received little
scholarly attention, despite
-
AUSTRONESIAN HISTORICAL LINGUISTICS 451
an upsurge of interest in it in metropolitan settings.
Exceptions are Grace'sbooks (121, 124), which are based on the
author's experience as a field linguistin the Pacific.
Contact-induced changes form a progression that begins with the
adoptionof discourse-level markers (for example, aria 'O.K.,'
'Let's go,' in both ANand Papuan languages of the Madang area) and
a tendency for "the way thingsare worded" (124) in the two
languages to become increasingly similar. Thur-ston (246, 247)
provides several complex examples of wording convergencesin both AN
and Papuan languages in northwestern New Britain. Similarity
ofwording then leads to increasing similarity in morphosyntactic
structure. TheBel languages, a group of AN languages spoken around
Madang on the north-em coast of Papua New Guinea, have adopted the
clause structure of neighborPapuan languages (218, 226).
AN-Papuan contacts over several millennia in parts of Melanesia
must haveproduced changes in many domains of culture, from
economics and kinship tocosmology. Such changes often show
linguisdc traces in the form of borrowedwords and formulae; yet
there have been few systematic studies attempting toestablish
directions of cultural borrowing or convergence from linguistic
evi-dence.
There are two other varieties of contact that have changed AN
languages.The first may be encompassed under the rubric "foreign
domination," and thesecond entails contact and bilingualism in two
(or more) AN languages. For-eign domination, which takes the form
of political or cultural domination byspeakers of another language,
u
.sually manifests itself in the shape of lexical borrowings, but
if theseborrowings are sufficiently intense, changes in morphology
and phonologyalso take place. A glance through the etymologies in
the Comparative Aus-tronesian Dictionary (250) reveals the
pervasive effects of Sanskrit and Arabicborrowings in westem
Indo-Malaysian languages through first Indie, thenIslamic, cultural
dominance, as well as considerable borrowing from the lan-guages of
European colonizers. Intense Sanskrit borrowing has introduced
acontrast between dental/alveolar /t, d/ and retroflex /2t, Idl
intoMadurese, Javanese, and Balinese (212, 223). Cases of pervasive
borrowingfrom a politically or culturally dominant language have
also been documentedin a number of pre-contact societies in
Melanesia, Micronesia, and Polynesia(20, 105, 139,148,191).
The most detailed study of AN contact-induced change is
Pallesen's (192)examination of contact between Tausug, a
Meso-Philippine language, andvarious Sama dialects (whose genetic
affiliations lie somewhere in westemIndonesia) in the Sulu
Archipelago. Most other studies of contact-inducedchange in AN
languages concern Oceanic cases. Rotuman has been reshapedby
extensive borrowing, at two different periods, from two Polynesian
Ian-
-
452 PAWLEY & ROSS
guages (20). Labu, an AN language spoken at the mouth of the
MarkhamRiver in Papua New Guinea, has been radically restructured
on the model ofBukawa, a member of a different AN subgroup (148).
West Uvean, a Polyne-sian language in the Loyalty Islands, has
added several new consonant pho-nemes as a result of borrowing
lexical items from its radically different ANneighbor, Iaai (191).
The phonological complexity of a number of languagesin New
Caledonia and the difficulty of applying the comparative method
tothem has been attributed to massive contact-induced change among
divergentAN languages (122, 123,125,141, 217).
CONCLUSION
The problem of culture history is that it is an
interdisciplinary enterprise, butthe methods and data used by each
of its major constituent disciplines are notreadily comparable.
Nonetheless such comparisons are necessary in order toevaluate
competing hypotheses within disciplines and to gain a more
completepicture of the past than any single method can provide. The
AN-speakingregion offers exceptionally favorable conditions for
such interdisciplinary re-search. Until recently, most prominent
hypotheses about the culture history ofthe AN-speaking regions
originated in the data of comparative linguistics orcomparative
ethnography, with scholars from these two disciplines
generallyworking independently. Archaeology has been a vigorous
latecomer. Earlyattempts at integrating linguistic and
archaeological evidence concentrated oncenters and directions of AN
dispersal, with archaeology providing a chrono-logical framework
for linguistically-based scenarios. Currently, the focus ofculture
historical syntheses is shifting toward comparisons of the lexicons
ofreconstructed languages with the content and environmental
contexts of vari-ous archaeological assemblages. There has been no
serious attempt to squarethe recent findings of historical human
biology with those of other disciplines,but there are signs that
this too is under way (11,131,146).
ACKNOWLEDGMENTS
We are grateful to Lois Carrington for carefully checking and
editing the textand references; to Roger Green and Matthew Spriggs
for providing detailedcomments on a draft; and to Robert Blust,
Ross Clark, James Eox, PaulGeraghty, Jeff Marck, Matthew Pawley,
and Lawrence Reid for many usefulcriticisms and
suggestions.Literature Cited
1. Aberle DF. 1980. Comment. Curr. Anthro- its Morphology.
Canberra: Pacific Linguis-pol. 21(2):226-30 tics, C-l 19. 253
pp.
2. AdcUaiKA. 1992. Proto-Malayic The Re- 3. Allen J, Gosden C,
White JP. I989.Humanconstruction of its Phonology and Parts of
Pleistocene adaptations in the tropical is-
-
AUSTRONESIAN HISTORICAL LINGUISTICS 453
land Pacific: recent evidence from NewIreland, a greater
Australian outlier. Antiq-uity 63(240):548-60
4. Allen J, White P 1989. The Lapita home-land: some new data
and an interpretation.y.Po/y/i. Soc. 98:129-46
5. Anson D. 1986. Lapita pottery of the Bis-marck Archipelago
and its affinities. Ar-chaeoi. Oceania 21:157-65
6. Baldi P, ed. 1990. Linguistic Change andReconstruction
Methodology. Berlin/NewYork: Mouton de Gruyter. 752 pp.
7. BellwoodPS. \97S. Man's Conquest of thePacific: The
Prehistory of Southeast Asiaand Oceania. Auckland: Collins. 462
pp.
8. Bellwood PS. 1984. A hypothesis for Aus-tronesian origins.
Asian Perspect. 26:107-17
9. BellwoodPS. ]9S5. Prehistory ofthe Indo-Malaysian
Archipelago. Sydney: Aca-demic. 370 pp.
10. Bellwood PS. 1987. The Polynesians: Pre-history of an Island
People. London:Thames & Hudson. 175 pp.
11. BellwoodPS. 1989. The colonization ofthePacific: some
current hypotheses. See Ref.146, pp. 11-59
12. BellwoodPS. 1991. The Austronesian dis-persal and the origin
of languages. Sci. Am.265:88-93
13. Bellwood PS, ed. 1991. Indo-Pacific Pre-history 1990. Proc.
14th Congr. Indo-Pa-cific Prehist. Assoc.
Canberra/Jakarta:Indo-Pacific Prehist. Assoc.
14. Deleted in proof15. Bellwood PS, Gillespie R, Thompson
GB,
Vogel JS, Ardika IW, Datan I. 1992. Newdates for prehistoric
Asian rice. Asian Per-spect. 3\(2):\6\-10
16. Bellwood PS, Koon P 1989. 'Lapita colo-nists leave boats
unburned!' The questionof Lapita links with Island Southeast Asia.A
6 3 2 O 6^ O )
17. Bender BW. 1971. Micronesian languages.SeeRef. 231, pp.
426-65
18. Bender BW, Wang J. 1985. The status ofProto-Micronesian. See
Ref. 203, pp. 53-92
19. Deleted in proof20. Biggs BG. 1965. Direct and indirect
inheri-
tance in Rotuman. Lingua 14:383-41521. Biggs BG. 1978. The
history of Polynesian
phonology. See Ref. 264, pp. 691-71622. Biggs BG. 1989. Towards
the study of
Maori dialects. See Ref. 138, pp. 61-7623. Biggs BG. 1990.
Proto-Polynesian lexical
file. Printout. Auckland: Univ. Auckland,Dept. Anthropol.
24. Biggs BG. 1993. New words for a newworld. See Ref. 208. In
press
25. Blust RA. 1969. Some new Proto-Aus-tronesian trisyllables.
Ocean. Linguist.8:105-84
26. Blust RA. 1970. Proto-Austronesian ad-denda. Ocean.
Linguist. 9(2): 104-62
27. Blust RA. 1972. Proto-Oceanic addendawith cognates in
non-Ocean Austronesianlanguages: a preliminary list. Work
Pap.Linguist. 4(l):l-43
28. Blust RA. 1976. Additions to Proto-Aus-tronesian addenda.
Work. Pap. Linguist.5(3):33-61
29. Blust RA. 1976. Austronesian culture his-tory: some
linguistic inferences and theirrelations to the archaeological
record.World Archaeoi. 8(l):19-43
30. Blust RA. 1977. The Proto-Austronesianpronouns and
Austronesian subgrouping: apreliminary report. Work. Pap.
Linguist.9(2):1-15
31. Blust RA. 1978. Eastem Malayo-Polyne-sian: a subgrouping
argument. See Ref.264, pp. 181-234
32. Blust RA. 1978. The Proto OceanicPalatals. Auckland: Polyn.
Soc, Monogr.43
33. Blust RA. 1980. Austronesian etymologies.Ocean. Linguist.
19:1-181
34. Blust RA. 1980. Early Austronesian socialorganization: the
evidence of language.Curr. Anthropol. 21(2):205-47
35. Blust RA. 1980. Notes on Proto-Malayo-Polynesian phratry
dualism. Bijdr. Taat-Land- Volkenkd. 136:215-47
36. BlustRA. 1981. Dual divisions in Oceania:innovation or
retention? Oceania 52(1):66-80
37. Blust RA. 1981. Variation in retention rateamong
Austronesian languages. Presentedat 3rd Int. Conf. Austronesian
Linguistics,Bali, Indonesia
38. BlustRA. 1982. The linguistic value oftheWallace Line.
Bijdr. Taal- Land- Volkenkd.138(2-3):231-50
39. Blust RA. 1984. Austronesian etymolo-gies-II. Ocean.
Linguist. 22-23:29-149
40. Blust RA. 1984. The Austronesian home-land: a linguistic
perspective. Asian Per-spect. 26:45-67
41. Blust RA. 1984. Comment. See Ref. 177,pp. 626-28
42. Blust RA. 1984. More on the position of thelanguages of
eastem Indonesia. Ocean.Linguist. 22-23:\-2&
43. Blust RA. 1986. Austronesian etymolo-gies-III. Ocean.
Linguist. 25:1-123
44. BlustRA. 1987.Lexicalreconstmctionandsemantic
reconstruction: the case of'house' words. Diachronica
4(l-2):79-106
45. Blust RA. 1989. Austronesian etymolo-giesIV. Ocean.
Ljnguist. 28:1-180
46. Blust RA. 1990. Linguistic change and re-construction
methodology in the Austrone-sian language family. See Ref. 6, pp.
133-
-
454 PAWLEY & ROSS
47. Blust RA. 1991. Sound change and migra-tion distance. See
Ref. 52, pp. 27-42
48. Blust RA. 1992. The Greater Central Phil-ippines hypothesis.
Ocean. Linguist.30:73-129
49. BlustRA. 1992. The position ofthe Formo-san languages:
method and theory in Aus-tronesian comparative linguistics. See
Ref.165, pp. 21-55
50. Blust RA. 1992. On speech strata inTiruray. In Papers in A
ustronesian Linguis-tics, No. 2, ed. MD Ross, pp. 1-52. Can-berra:
Pacific Linguistics, A-82
51. BlustRA. 1993. Austronesian sibling termsand culttire
history. See Ref. 208. In press
52. Blust RA, ed. 1991. Currents in PacificLinguistics: Papers
on Austronesian Lan-guages and Ethnolinguistics in Honour ofGeorge
W. Grace. Canberra: Pacific Lin-guistics, C-l 17. 560 pp.
53. Capell A. 1943. The Linguistic Position ofSouth-Eastem
Papua. Sydney: Aus-tralasian Medical
54. Capell A. \962. A Linguistic Survey of theSouth-Western
Pacific. Noumea: S. PacificCommission
55. Capell A. 1976. General picture of Aus-tronesian languages.
New Guinea area. SeeRef. 263, pp. 5-52
56. Carrington L. 1990. Select Austronesianbibliography. See
Ref. 250. In press
57. Chang KC. 1992. Taiwan Strait archaeol-ogy and the
Protoaustronesians. See Ref.165, pp. 56-81
58. ChowningA. 1979. Leadership in Oceania./ Pac. Hist.
14:66-84
59. Chowning A. 1980. Comment. Curr. An-thropol.
21(2):231-33
60. Chowning A. 1984. Comment. Curr. An-thropol. 21(2):628
61. Chowning A. 1991. Proto Oceanic culture:the evidence from
Melanesia. See Ref. 52,pp. 43-75
62. Clark R. 1975. Comment on Polynesiansibling terms. Am.
Anthropol. 77:85-88
63. ClarkR. \976. Aspects of Proto-PolynesianSyntax. Auckland:
Linguistic Soc. NZ 178 pp.
64. Clark R. 1979. Language. See Ref 155, pp.249-70
65. Clark R. 1982. Proto-Polynesian birds. SeeRef. 233, pp.
121^3
66. Clark R. 1984. Conmient. Curr. Anthropol.21(2):628-29
67. Clark R. 1985. Languages of north andcentral Vanuatu:
groups, chains, clustersand waves. See Ref. 203, pp. 199-236
68. Clark R. 1993. Evolution, migration andextinction of Oceanic
bird names. See Ref.208. In press
69. Clark JT, Terrell J. 1978. Archaeology inOceania. Annu. Rev.
Anthropol. 7:293-319
70. Collins JT. 1983. The Historical Relation-ships ofthe
Languages of Central Maluku,
Indonesia. Canberra: Pacific Linguistics,D-47. 168 pp.
71. Dahl OC. 1976. Proto-Austronesian. Oslo:Studentlitteratur.
Revised ed. 1976
72. DavidsonJM. 1977.WestemPolynesiaandFiji: prehistoric
contact, diffusion and dif-ferentiation in adjacent
archipelagos.World Archaeoi. 9(l):82-94
73. Dempwolff O. 1934. Vergleichende Lautle-hre des
austronesischen Wortschatzes.Band I: Induktiver Aufbau einer
indonesis-chen Ursprache. Beiheften zur Zeitschriftfiir
Eingeborenen-Sprachen 15. Berlin: Re-imer
73a. Dempwolff O. 1937. Vergleichende Laut-lehre des
austronesischen Wortschatzes.Band 2: Deduktive Anwendung des
Urin-donesischen auf austronesische Einzel-sprachen. Beiheften zur
Zeitschrift fiirEingeborenen-Sprachen 17. Berlin: Re-imer
73b. Dempwolff O. 1938. Vergleichende Laut-lehre des
austronesischen Wortschatzes.Band 3: Austronesisches
Worterverzeich-nis. Beiheften zur Zeitschrift fiir
Einge-borenen-Sprachen 19. Berlin: Reimer
74. Diebold AR Jr. 1987. Linguistic ways toprehistory. In
Proto-Indo-European: TheArchaeology of a Linguistic Problem:
Stud-ies in Honor of Marija Gimbutas, ed. SNSkomal, EC Polome, pp.
19-71. Washing-ton, DC: Inst. Study Man
75. Douglas B. 1979. Rank, power, authority:a reassessment of
traditional leadership inSouth Pacific societies. /. Pac. Hist.
14:2-27
76. Dutton TE. 1976. Magori and similar lan-guages of south-east
Papua. See Ref 263,pp.581-636
77. Dutton TE. 1982. Borrowing in Austrone-sian and
non-Austronesian languages ofcoastal south-east mainland Papua
NewGuinea. See Ref 135, pp. 109-77
78. Dutton TE, Tryon DT, eds. 1993./.anguageContact and Change
in the AustronesianWorld. Berlin: Mouton de Gruyter. In press
79. Dyen I. 1951. Proto-Malayo-Polynesian"Z." Language
27:534-40
80. Dyen 1. 1953. The Proto-Malayo-Polyne-sian Laryngeals.
Baltimore: Linguist. Soc.Am. 65 pp.
81. Dyen I. 1956. Language distribution andmigration theory.
Language 32:611-26
82. Dyen I. 1962. Lexicostatistically deter-mined borrowing and
taboo. Language37:38-46
83. Dyen I. 1963. The position of the Ma-layopolynesian
languages of Formosa.Asian Perspect. 7(1-2):261-71
84. Dyen 1.1965. Formosan evidence for somenew
Proto-Austronesian phonemes. Lin-gua 14:285-305
85. Dyen I. 1965. Alexicostatisticalclassifica-
-
AUSTRONESIAN HISTORICAL LINGUISTICS 455
tion of the Austronesian languages. Int. J.Am. Linguist. Memoir
19
86. Dyen I. 1971. The Austronesian languagesand
Proto-Austronesian. See Ref. 231, pp.5-54
87. Dyen I. 1971. The Austronesian languagesof Formosa. See Ref.
231, pp. 168-99
88. Dyen I. 1971. Malagasy. SeeRef. 231, pp.211-39
89. Dyen I. 1982. The present status of someAustronesian
subgrouping hypotheses. SeeRef. 136, pp. 31-35
90. Dyen I. 1990. Homomeric lexical classifi-cation. See Ref. 6,
pp. 211-30
91. Dyen I. 1992. The internal and externalclassification of the
Formosan languages.SeeRef. 165, pp. 162-96
92. Dyen I, James AT, Cole JWL. 1967. Lan-guage divergence and
estimated word re-tention rate. Language4'i(\).\5Q-l\
93. Dyen I, Tsuchida S. 1991.Proto-Philippineas the closest
relative of Proto-Formosan.SeeRef. 137, pp. 85-101
94. Epling PJ, Kirk R, Boy