Hawaiki, Ancestral Polynesia: An Essay in Historical Anthropology [1 ed.] 9780521788793, 052178879X, 0521783097, 9780521783095, 9780511067006

Citation preview

This page intentionally left blank

Hawaiki, Ancestral Polynesia An Essay in Historical Anthropology

The power of an anthropological approach to long-term history lies in its unique ability to combine diverse evidence, from archaeological artifacts to ethnographic texts and comparative word lists. In this innovative book, Kirch and Green explicitly develop the theoretical underpinnings, as well as the particular methods, for such a historical anthropology. Drawing upon and integrating the approaches of archaeology, comparative ethnography, and historical linguistics, they advance a phylogenetic model for cultural diversi®cation, and apply a triangulation method for historical reconstruction. They illustrate their approach through meticulous application to the history of the Polynesian cultures, and for the ®rst time reconstruct in extensive detail the Ancestral Polynesian culture that ¯ourished in the Polynesian homeland ± Hawaiki ± some 2,500 years ago. Of great signi®cance for Oceanic studies, Kirch and Green's book will be essential reading for any anthropologist, prehistorian, linguist, or cultural historian concerned with the theory and method of longterm history. patrick vinton kirch is Professor of Anthropology, and Director of the Phoebe A. Hearst Museum, at the University of California at Berkeley. A member of the National Academy of Sciences, he has authored some ten previous books on Paci®c archaeology and prehistory, including Anahulu: The Anthropology of History in the Kingdom of Hawaii (1992) (co-authored with Marshall Sahlins), which won the J. I. Staley Prize in Anthropology. roger c. green is Emeritus Professor of Prehistory at the University of Auckland, New Zealand. A member of the National Academy of Sciences and a Fellow of the Royal Society of New Zealand, he is the author of several important monographs on Paci®c Islands archaeology and prehistory.

Frontispiece: Mata o Tangaloa (``Face of Tangaloa''), by Fatu Feu'u

Hawaiki, Ancestral Polynesia An Essay in Historical Anthropology

PAT R I C K V I N TO N K I RC H University of California, Berkeley

and RO G ER C . G R E EN University of Auckland, New Zealand

   Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo Cambridge University Press The Edinburgh Building, Cambridge  , United Kingdom Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521783095 © Cambridge University Press 2001 This book is in copyright. Subject to statutory exception and to the provision of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published in print format 2001 - isbn-13 978-0-511-06700-6 eBook (NetLibrary) - isbn-10 0-511-06700-3 eBook (NetLibrary) - isbn-13 978-0-521-78309-5 hardback - isbn-10 0-521-78309-7 hardback - isbn-13 978-0-521-78879-3 paperback -  paperback isbn-10 0-521-78879-X Cambridge University Press has no responsibility for the persistence or accuracy of s for external or third-party internet websites referred to in this book, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.

To TheÂreÁse and Valerie, for their love and support; and to the late Bruce Biggs, preeminent Polynesian linguist

Contents

List of ®gures page viii List of tables x Preface xiii List of language abbreviations

xvi

Prologue: on historical anthropology

1

Part I The phylogenetic model: theory and method 1 2 3

The phylogenetic model in historical anthropology 13 Methodologies: implementing the phylogenetic model 32 Polynesia as a phylogenetic unit 53 Part II Rediscovering Hawaiki

4 5 6 7 8 9

Introductory remarks 95 The Ancestral Polynesian world 99 Subsistence 120 Food preparation and cuisine 143 Material culture 163 Social and political organization 201 Gods, ancestors, seasons and rituals 237 Epilogue: on history, phylogeny, and evolution Notes 285 Glossary of terms 313 References 317 Subject Index 357 Index of Proto Polynesian Reconstructions

vii

369

277

Figures

Frontispiece Mata o Tangaloa (``Face of Tangaloa''), page ii by Fatu Feu`u 1.1 Map of the Polynesian triangle and the Polynesian Outliers 17 1.2 Kirch's 1984 model of phylogenetic differentiation in Polynesia 20 2.1 The higher-level subgrouping of the Austronesian languages, down to the Oceanic level 39 2.2 The geographic distribution of higher-level subgroups in the Austronesian phylum 40 3.1 The major subgroups of Oceanic form a ``rake-like'' tree structure 56 3.2 The geographic distribution of major subgroups within the Oceanic branch of Austronesian languages 57 3.3 The Proto Central Paci®c dialect chain 58 3.4 North±south dialect differentiation within Proto Polynesian 59 3.5 A ``family-tree'' type classi®cation of the Polynesian languages 61 3.6 Islands in the Fiji±Western Polynesian region linked by voyaging circles of 24 hours or less 62 3.7 The Paci®c region with Near Oceania, Remote Oceania, and the Andesite Line and ``continental'' type islands indicated 64 3.8 Canoe regions of the Paci®c 67 3.9 The geographic distribution of sibling classi®cation types in Oceania 68 3.10 Relationships among Polynesian biological populations as indicated by distance analysis of thirty-eight non-metric cranial traits 75 3.11 A graphic representation of the ``density'' of available archaeological information for major Polynesian cultural sequences 76 3.12 Locations of key archaeological sites dating to the Ancestral Polynesian phase 84 4.1 The central Paci®c region, showing the location of the Andesite Line 108 4.2 The hierarchical structure typical of folk biological classi®cations 110 viii

List of ®gures

5.1 5.2 6.1 6.2 6.3 7.1 7.2 7.3 7.4 7.5 7.6 8.1 9.1 9.2 9.3 9.4 9.5

Cobbles with ®nger-grips from site FU-11, Futuna, interpreted as hammers for opening hard-shelled nuts, such as Canarium Turbo shell ®shhooks from the To`aga site The earth oven, a central feature of Ancestral Polynesian cooking, attested by an example at the Lolokoka site (NT-90) on Niuatoputapu Ethnographic examples of coconut graters, made up of a stool or other wooden base to which a shell grater is lashed Straight-sided pits, lacking evidence of burning, may have been used as silos for the fermentation and storage of breadfruit paste Pottery vessel shapes in Ancestral Polynesia Conceptual terms for Proto Polynesian containers, and their realization in plainware pottery vessels of the Ancestral Polynesian culture Industrial tools: adzes in Ancestral Polynesian culture Industrial tools: Saw, ®les, whetstones, grinding stones, stone and coral abraders, drill points and bow drill Ornaments from archaeological sites of the Ancestral Polynesian phase Excavation plan of the Sasoa`a site in the Falefa Valley, Samoa Social groups and leadership roles in Ancestral Polynesian societies Plan of a Tikopia fare house with attached marae Annual tributary presentation of the ®rst yams on the ceremonial plaza (malae) at Mu`a, Tongatapu Perspective renderings of three variants of Tuamotuan marae The southern sky as it would have appeared an hour before sunrise on May 16, 500 BC, from an island in Western Polynesia, showing the heliacal rising of the Pleiades (*Mataliki) Diagrammatic summary of the reconstructed Ancestral Polynesian ritual cycle and calendar

ix

124 133 148 153 161 169 174 179 181 188 195 236 250 252 253 266 274

Tables

2.1 3.1 3.2 4.1 4.2 4.3 5.1 5.2 5.3 5.4 6.1 6.2 6.3 6.4 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 8.1

POLLEX database entry for PPN *waka, `canoe' page 47 Cultural traits distinguishing Western and Eastern Polynesian regions 72 Selected archaeological sites and assemblages associated with the Ancestral Polynesian period 82 Selected Proto Polynesian terms for the physical world 103 Proto Polynesian life-form terms 111 Proto Polynesian terms for reef and shoreline invertebrates 112 Proto Polynesian crops 123 Proto Polynesian terms associated with horticulture 127 Distribution of ®shing methods in tropical Polynesia 136 Proto Polynesian terms associated with marine exploitation 138 Proto Polynesian terms for raw, cooked, and taste 145 Proto Polynesian terms associated with the cookhouse, earth oven, and cooking equipment 150 Proto Polynesian terms for food preparation and cooking methods 155 Proto Polynesian terms associated with the pudding complex 158 Perishable and durable components of Polynesian material culture inventories 165 Proto Polynesian terms for things 166 Proto Polynesian terms for containers 167 Proto Polynesian terms for industrial tools 176 The Proto Polynesian bark cloth complex, clothing, ornaments, and tattooing 186 Proto Polynesian terms for warfare, sports and games, and musical instruments 191 Proto Polynesian terms relating to household units and their architectural features 194 The Proto Polynesian canoe complex and cordage 198 Proto Austronesian (PAN), Proto Malayo-Polynesian (PMP), and Proto Oceanic (POC) words for settlements and architecture 206 x

List of tables

8.2

Linguistically indicated changes in architectural forms from Proto Oceanic to Proto Polynesian interstages 8.3 Re¯exes of Proto Polynesian *kainanga, including extra-Polynesian witnesses 8.4 Re¯exes of Proto Polynesian *kaainga 8.5 Re¯exes of Proto Polynesian *saqa, `social group' 8.6 Proto Polynesian terms relating to exchange or trade 8.7 Proto Polynesian terms for persons 8.8 Proto Polynesian kinship terms 8.9 Re¯exes of Proto Polynesian *qariki 8.10 Re¯exes of Proto Polynesian *fatu, leader of the *kaainga 9.1 Proto Polynesian terms relating to gods, spirits, and ancestors 9.2 Proto Polynesian terms relating to ritual practitioners and spaces 9.3 Proto Polynesian terms associated with ritual 9.4 Key aspects of Polynesian calendrical systems 9.5 Selected Fijian and Polynesian lunar calendrical lists 9.6 Reconstructed lunar month names for various Polynesian proto-languages 9.7 Probable reconstruction of the Proto Polynesian lunar calendar, and its transformations in subsequent Polynesian proto-languages 9.8 Some post-Proto Polynesian lexical and semantic innovations in ritual terminology

xi

207 212 216 219 221 222 223 229 233 240 246 258 262 268 270 271 275

Preface

Enchanted by the seductively salubrious atmosphere of California's Napa Valley, we gazed over sun-drenched vineyards with the 1993 harvest ripening on the vine, sipping the last of a lush Cabernet while intently arguing the intricacies of some Proto Polynesian term. Perhaps ± given the blissful feeling this setting inspired ± we might have been excused our conceit that we would conspire to write ``a little essay between covers.'' The notion, naive in retrospect, was to expand slightly on our 1987 article on ``History, phylogeny, and evolution in Polynesia'' (Kirch and Green 1987), so as to address certain critiques of the phylogenetic approach to historical anthropology, and to elaborate what we call a ``triangulation method'' for historical reconstruction. The proposition seemed straightforward enough. Yes, a ``little essay,'' perhaps a hundred pages or so. Over plates of roast Petaluma duck and grilled sword®sh, our wives had seconded the idea, insisting that we should keep the essay lean and trim. Nearly a decade later, our ``essay'' has taken shape as a book, a more ponderous volume than we at ®rst envisioned. Its writing has occupied far longer than anticipated, requiring several international trips and much longdistance collaboration. Yet we do not regret the transformation that our project has undergone, because out of it we have gained a deeper respect for the possibilities of a truly integrative historical anthropology. We were trained (at Penn and Yale, New Mexico and Harvard, respectively) in the classic holistic perspective of Americanist anthropology, and although we are both primarily archaeologists of the Paci®c, each of us in our respective careers has endeavored to bring a full spectrum of anthropological evidence and approaches to bear in our research programs. Green early on incorporated historical linguistics into his models of Polynesian settlement (e.g., Green 1966), while Kirch integrated ®eld ethnography into his work on prehistoric ecology and economy (e.g., Kirch 1994a). This book re¯ects the maturing of those long-standing interests, a statement of our conviction that anthropology at its best is always holistic and integrating. At a time when at least one prominent biologist is crying out for ``consilience'' between the social and biological sciences (Wilson 1998), we would point out that anthropology has always heeded that call. xiii

xiv

Preface

While engaged in drafting several chapters during June of 1997, in Berkeley, we became overtly conscious of how our respective ethnographic and linguistic experiences in a diversity of Polynesian venues critically aided the construction of the arguments we were striving to advance. Comparative ethnography can, in theory, be carried out by the proverbial ``armchair'' scholar, but there can be no doubting the value of personal ethnographic experience over a range of Polynesian cultures and societies. The most astute comparativists in the Oceanic ®eld themselves had the advantage of original ®eldwork in at least two or more locales: Handy, Hiroa, Burrows, Emory, Oliver, and Sahlins, among them. As with our predecessors, we likewise have spent much time residing and working in many Polynesian societies, including: Anuta, Tikopia, Taumako, Tonga, Futuna, Samoa, `Uvea, Mangaia, Mo`orea, Mangareva, Aotearoa, Rapa Nui, and Hawai`i. Between us we speak or have made signi®cant efforts assembling vocabularies of the following Polynesian languages: Anutan, Tikopian, Taumako, Futunan, Tongan, Samoan, Tahitian, Mangarevan, and Hawaiian. This ethnographic and linguistic background, acquired through a combined total of seven decades of continuous effort in the Polynesian ®eld, has proved invaluable for the task we set ourselves. All this, need we say, has been in addition to our primary efforts as archaeologists in the same islands, where we have endeavored to generate materially documented historical sequences of cultural change. We underscore this point here not to assert our authority, but rather to stress the necessity in historical anthropology of erudition based on broad comparative knowledge. Quite possibly, the kind of work we would wish to see undertaken and extended is, in fact, only possible through collaboration, for it is doubtful that any one individual can command either the necessary depth of methodological and theoretical expertise, or the range of speci®c knowledge acquired through ®eld or library research. Writing this book has been a true collaboration. But one of us writes books, having honed the necessary skills, while the other does not; the order of authors recognizes that reality. Of course, each of us read, emended, edited, and critiqued the drafts of the other, so the ®nal book truly re¯ects a joint effort. Acknowledgments Green thanks the Miller Institute for Basic Research in Science, University of California at Berkeley, for a Visiting Miller Professorship which brought him to Berkeley in the fall of 1994, and allowed us to begin our collaboration. Kirch gratefully acknowledges the support of the Center for Advanced Study in the Behavioral Sciences, Stanford, California, which provided him

Preface

xv

with ideal working conditions during the ®nal stages of writing and editing. Kirch also thanks the National Science Foundation, which partially funded his 1997±98 CASBS Fellowship (Grant No. SBR-9601236). We owe a great debt to our colleagues in Paci®c historical linguistics, without whose decades of careful work in lexical reconstruction we would not have been able to undertake this book. In particular, the late Emeritus Professor Bruce Biggs of the University of Auckland provided a major underpinning for our research through his POLLEX database of Proto Polynesian reconstructions which he has tirelessly compiled since 1965. Professor Biggs gave us free access to his computerized database, for which we are immensely grateful. It was with great sadness, as this book was in ®nal proof, that we learned of his passing. Other linguists, especially Andrew Pawley, Malcolm Ross, Ross Clark, and Bob Blust, have provided us with information, insights, and helpful critiques over the years. We are especially grateful to the following colleagues who took the time to read and critique draft versions of various chapters: Peter Bellwood, Bob Blust, Janet Davidson, Ward Goodenough, Steve Hooper, John Moore, Frank Lichtenberk, Andrew Pawley, and Marshall Sahlins. David Tuggle kindly provided simulated southern hemisphere sky charts for the mid-®rst millennium BC, including that reproduced as Figure 9.4. Hans Schmidt kindly provided us with his transcriptions, in English and Rotuman, of selected excerpts from the manuscript notes of A. M. Hocart, housed in the Alexander Turnbull Library. Serge TcherkeÂzoff shared with us a copy of his manuscript paper on Samoan matai. In the ®nal stages of manuscript preparation, Sara Diamond (Berkeley) and Dorothy Brown (Auckland) provided invaluable assistance with word processing and bibliography. Joan Lawrence prepared the illustrations from our rough copy. It gives us great pleasure to dedicate this book to our wives, TheÂreÁse Babineau and Valerie Green. They shared our early enthusiasm, encouraged us through the rough spots, and reminded us of the larger signi®cance of our project. Patrick Vinton Kirch Roger C. Green

Abbreviations

Language abbreviations Proto-language abbreviations PAN Proto Austronesian PCE Proto Central Eastern Polynesian PCEMP Proto Central-Eastern Malayo-Polynesian PCP Proto Central Paci®c PEC Proto Ellicean PEP Proto Eastern Polynesian PMP Proto Malayo-Polynesian PMQ Proto Marquesic PNP Proto Nuclear Polynesian POC Proto Oceanic PPN Proto Polynesian PTA Proto Tahitic PTO Proto Tongic Modern language abbreviations, and geographic af®nity AIT ANU AUS EAS ECE EFU EUV FIJ HAW KAP MAE MAO MFA MIA

Aitutaki (Cook Is.), Central Eastern Polynesia Anuta (Cherry Is.), Outlier Austral Is. (French Polynesia), Central Eastern Polynesia Easter Is., Marginal Eastern Polynesia Tuvalu (Ellice Is.), Western Polynesia East Futuna (Horne Is.), Western Polynesia East Uvea (Wallis Is.), Western Polynesia Fiji Hawai`i, Marginal Eastern Polynesia Kapingamarangi, Outlier Emae (Vanuatu), Outlier New Zealand Maori, Marginal Eastern Polynesia Mele-Fila (Vanuatu), Outlier Mangaia (Cook Is.), Central Eastern Polynesia xvi

List of language abbreviations

MKI MOR MQA MQN MQS MRA MVA NIU NKO NKR OJA PEN PIL PUK RAR REN ROT RUR SAM SIK TAH TAK TIK TOK TON TUA WFU WUV or WEV WYA

xvii

Manihiki (Cook Is.), Central Eastern Polynesia Mooriori (Chatham Is.), Marginal Eastern Polynesia Marquesas (French Polynesia), Marginal Eastern Polynesia Northern Marquesan dialect (French Polynesia), Marginal Eastern Polynesia Southern Marquesan dialect (French Polynesia), Marginal Eastern Polynesia Manihiki/Rakahanga (Cook Is.), Central Eastern Polynesia Mangareva (French Polynesia), Central Eastern Polynesia Niue Is., Western Polynesia Nukuoro, Outlier Nukuria (Solomons), Outlier Luangiua (Ontong-Java, Solomons), Outlier Penrhyn (Cook Is.), Central Eastern Polynesia Pileni (Solomons), Outlier Pukapuka (Northern Cook Is.), Central Eastern Polynesia Rarotonga (Cook Is.), Central Eastern Polynesia Rennell and Bellona Is. (Solomons), Outlier Rotuma (Fiji) Rurutu (French Polynesia), Central Eastern Polynesia Samoa, Western Polynesia Sikaiana (Solomons), Outlier Tahitian (French Polynesia), Central Eastern Polynesia Takuu (Solomons), Outlier Tikopia (Solomons), Outlier Tokelau Is., Western Polynesia Tonga, Western Polynesia Tuamotu (French Polynesia), Central Eastern Polynesia West Futuna (Vanuatu), Outlier West Uvea (Ouvea, New Caledonia), Outlier Waya, Western Fiji

Prologue: on historical anthropology Our problem may be metaphorically de®ned as the translation of a two-dimensional photographic picture of reality into the threedimensional picture which lies back of it . . . The gaining of an historical perspective will mean the arrangement in as orderly temporal sequence as possible, within as de®nitely circumscribed absolute time limits as circumstances will allow, of the processes studied by our science, the carriers of these processes being generally de®ned more inclusively than in documentary history.

sapir 1916:2

Polynesians called it Hawaiki (or sometimes, Kahiki, or Pulotu), the distantly remembered homeland, source of their ancestors, mythical site of the creation of culture, and spirit realm to which their own souls would voyage after death.1 They honored this ancestral homeland in chant and song, and named newly found islands after it: Savai`i in Samoa, and the large island of Hawai`i, among them. But was there ever in reality such a ``Hawaiki,'' or does it exist only in the shadowy realms of cosmogonic myth? Archaeologists, after a half-century of intensive pursuit of the question of Polynesian origins, would answer af®rmatively. More precisely, they would ®x the coordinates of this ancestral homeland in time and space: the archipelagos of Tonga and Samoa (with their immediate smaller neighbors), in the ®rst millennium BC. Through an unbroken sequence of cultural change that begins with the arrival of small groups of Early Eastern Lapita peoples around 1100±1000 BC, a distinctive Ancestral Polynesian culture had developed four to ®ve centuries later. While archaeologists con®dently point to various settlements and sites of this period and to their characteristic material assemblages of Polynesian Plainware pottery and plano-convex adzes, securely ®xed in time by numerous radiocarbon dates ± what do we really know about this Ancestral Polynesian world, this Hawaiki? Is it possible to move beyond the strictly material evidence of potsherds, adzes, and shell ®shhooks, postmolds and earth ovens? Simply stated, this is the problem that has energized us to write this book, for we would maintain that twentieth-century anthropology has 1

2

Hawaiki, ancestral Polynesia

indeed developed powerful tools and methods for recovering and writing the deep history of ``peoples without history.'' Yet we are perturbed that as the twenty-®rst century dawns, the academic and scholarly rush toward specialization and even sub-specialization (not to mention the current postmodern conceit that ``culture'' or ``history'' are anything other than academic constructions) threatens to erode the essential strength of a holistic vision of anthropology as an integrated set of perspectives and methods trained upon a diversity of evidence. The founders of the unique Americanist tradition in anthropology ± Boas, Kroeber, Sapir, and others ± reacted in part to the theoretical excesses of a generalizing ``evolutionary'' approach, and advocated a more rigorous ``historical particularism.'' They saw the advantage to be gained from multiple lines of investigation and evidence, and thus bundled ethnography, archaeology, linguistics, and physical anthropology together in a way that the European academic world never fully embraced. Eighty years ago Edward Sapir advanced a charter for historical anthropology in his short monograph on Time Perspective in Aboriginal American Culture: A Study in Method (Sapir 1916). This paper ± once famous but now seldom cited ± laid out the potential contributions to historical reconstruction to be made by combining the direct evidence of documentary writings, native testimony, and archaeological ®nds, with the inferential evidence provided by physical anthropology, ethnology, and linguistics.2 Sapir envisioned a historical anthropology that ± as a joint intellectual enterprise ± required contributions from all of these ®elds, each with its own unique evidential sources. The historical goals that motivated Sapir have waxed and waned in anthropology over the intervening decades, and the paradigms and methods of the ``sub®elds'' (archaeology, ethnology, biological anthropology, and linguistics) have also changed dramatically.3 Despite some interesting proposals in the interim (e.g., Romney 1957; Vogt 1964, 1994a), few integrated data-rich explorations along the lines conceived by Sapir have evolved. Nonetheless, in the ®rst decade of the twenty-®rst century a renewed interest in matters historical may be discerned in the several sub®elds into which anthropology has been partitioned. These trends lend cautious optimism that our present endeavor ± fundamentally similar to Sapir's, but here applied to Polynesia ± may be of more than strictly regional interest.4 Like Sapir, we aim to advance a historical anthropology, but one that brings to bear the myriad advances in data, methods, and theory developed throughout the twentieth century. Sapir devoted most of his attention to linguistics and ethnology; he only brie¯y mentioned documentary sources, oral history, and physical anthropology, and relegated archaeology to a single page of his monograph. Sapir's ethnolinguistic bias is understandable, given the embryonic state of New

Prologue: on historical anthropology

3

World prehistory in 1916. Even for the Old World, where archaeology had an earlier start, existing knowledge was then encompassed within the boldest of schemes: Palaeolithic, Neolithic, Bronze Age, and Iron Age. But a growing subdiscipline of anthropological archaeology, especially in North America, increasingly became the main player in historical anthropology, where during the ®rst half of the twentieth century it struggled to develop methods for establishing relative or absolute chronology (Taylor 1948; Trigger 1989a). At the same time that archaeology concentrated on cultural homologies (similarities due to common ancestry) and synologies (similarities produced by diffusion or borrowing), within what became known in North America as ``culture history,''5 ethnology increasingly rejected historical reconstruction. Following Radcliffe-Brown's pejorative characterization of ethnology's earlier efforts in this direction as ``conjectural'' or ``pseudohistory'' (1941:1, 1950:1±2), developments in social and cultural anthropology moved steadily toward synchronic orientations.6 In the Paci®c, the ethnographies of Raymond Firth, Gregory Bateson, and Margaret Mead provide examples. Interest in historical sources and problems was largely relegated to the temporally restricted topic of ``ethnohistory'' (Dening 1966). Attempts to weld the shorter-term perspective of ethnohistory to the longerterm trajectories revealed by archaeology, proposed by some North American scholars, came to be known as the ``direct historical approach'' (Wedel 1938; Steward 1942; Strong 1953). Although the direct historical approach fell out of favor in the post-World War II era, it now shows signs of renewed application (Lightfoot 1995). Archaeology too, at least in North America, went through its own phase in which the particular contingencies of history were devalued in favor of a more ``scienti®c'' orientation that sought universal ``laws'' of cultural process. The New Archaeology of the 1960s and 1970s replaced the earlier emphasis on homologous change with a concern for analogous change, driven in part by a paradigm of archaeology as an experimental and even predictive social science (e.g., Watson et al. 1971). Anthropological linguistics, in contrast, has always retained to varying degrees its historical component (Hock 1986:v±vi), even while it underwent a range of transformations in its more mainstream descriptive, theoretical, and sociological varieties (Hymes 1964). These continuing historical linguistic enterprises ± largely independent of archaeology ± have culminated in a series of language-family histories based on genetic subgroupings, for many of the world's languages (Blench 1997: table 2). Finally, like linguistics, biological anthropology has long maintained its evolutionary interests in the genetic history of human populations.7 In spite of these varied efforts in anthropological history over the course of the twentieth century ± or perhaps just because they remained largely

4

Hawaiki, ancestral Polynesia

uncoordinated as the subdisciplines burgeoned and specialized ± a genuinely systematic, methodologically rigorous, and theoretically sophisticated historical anthropology of the kind that Sapir envisioned eighty years ago failed to materialize. However, that situation has begun to change, and especially in the Paci®c. The varied strands of a new historical orientation are contained within what Trigger (1989a, 1989b, 1991) calls ``holistic archaeology,'' an approach he sees as forming ``a new synthesis for archaeological explanation.'' Echoing Sapir, Trigger proposes to combine archaeological data with the ®ndings of historical linguistics, oral traditions, historical ethnography, and historical records so as to produce a more rounded view of prehistory, as well as of ethnohistory and historical archaeology. Trigger (1991:562) argues that such interdisciplinary approaches ®rst developed as early as the 1950s, citing examples from Africa (e.g., Murdock 1959; McCall 1964; Trigger 1968). Early efforts were, however, largely rejected by the emerging and rapidly dominant ``processual'' archaeologists. Renewed efforts at tackling sequences of homologous change are noted by Trigger as recurring in the late 1970s and early 1980s in North America, the Mayan region, and Polynesia, as well as in Africa.8 They are one basis for Trigger's claim that ``the direct historical approach is perhaps the most challenging and potentially important task confronting archaeology today,'' requiring archaeologists to become ``still more open to using non-archaeological forms of data to study the past'' (Trigger 1991:563). Other recent examples include the collaborative works of Kent Flannery and Joyce Marcus (1983; Marcus and Flannery 1996) on the long-term historical evolution of the Zapotec and Mixtec peoples of Mesoamerica, and Kirch and Sahlins' collaborative work on the Hawaiian Kingdom (1992).9 Calls for a renewed historical orientation within anthropology are not limited to archaeology. Throughout the 1980s some sociocultural anthropologists became increasingly historicized (Ohnuki-Tierney 1990:1±6), taking their lead in part from the well-developed Annales tradition of encompassing social history as practiced by Marc Bloch, Fernand Braudel, Georges Duby, and others. Marshall Sahlins incorporated and modi®ed aspects of Braudel's (1980) famous ``wavelength'' scheme of history in his brilliant work on Captain Cook and the con¯uence of Hawaiian and British cultures in 1778±79 (Sahlins 1981, 1985, 1995). At the same time, Greg Dening ± a historian with anthropological training ± was moving in his studies of Marquesan ethnohistory and early European contacts in the Paci®c toward what he calls ``history's anthropology'' (1980, 1988, 1992). The pioneering efforts of Sahlins and Dening have been extended by others (e.g., Linnekin 1990; Thomas 1991, 1997). Such historicization of social anthropology was, moreover, by no means con®ned to the Paci®c arena (see Cohn 1980, 1981;

Prologue: on historical anthropology

5

Ohnuki-Tierney 1990). Biersack, in her introduction to Clio in Oceania, a book with the notable subtitle ``Toward a Historical Anthropology,'' writes: In varying degrees, the issues of history and theory rehearsed herein bear on other branches of anthropology [in addition to archaeology] and serve as core issues around which the sub®elds of anthropology may coalesce and enter into collaboration . . . Positioned among historical and cultural studies and at a powerful con¯uence of subdisciplines within anthropology, historical anthropology provides a forum within which to perpetuate the debates of the last two decades but on new and less parochial terrain. To historical anthropology is thus transferred the theoretical commissions of the discipline: past, present, and future. (1991:25)

A concrete expression of these merging historical interests within social anthropology and archaeology is the collaborative work of Kirch and Sahlins, Anahulu: The Anthropology of History in the Kingdom of Hawaii (1992). This project ± combining the data and perspectives of a historical ethnologist and an archaeologist, focused on a particular geographic and historic space, the Anahulu Valley ± is a book-length example of research that purposively merges subdisciplinary approaches. That more collaboration between archaeologists and historical ethnographers has not been undertaken may re¯ect a long-standing ± and in most cases implicit rather than explicit ± bias toward those last few hundred years of global European expansion, and an implicit privileging of textual records (Wolf 1982).10 Thus Sahlins, while discovering that the ``peoples of the Paci®c I had studied indeed had a history,'' could still remark that ``these exotic histories . . . as recorded do not go very far back'' (1985:xviii). And Dening can claim that ``the history of Polynesian cultures could only be written out of sources that were European'' (1991:372, emphasis added). These comments for the Paci®c are echoed in Ohnuki-Tierney's more general remark that ``the longue dureÂe is not easily accessible for histories of nonliterate peoples'' (1990:3, fn. 2). Thus turning their backs to archaeological colleagues often housed in the very same academic departments of anthropology, historical ethnographers have often haughtily disdained anything except the documentary form of the literate world's historical texts, usually European-authored. In such agendas, the archaeological record is assumed to be either irrelevant to history, or relevant only to a short segment of it.11 But the historical ``texts'' of the longue dureÂe are encoded not just in the ciphers of Western scribes; they exist equally as material traces dispersed over landscapes and sedimented in their depths, no less as patterns of cognate words in the linguists' comparative lexicons, or as indigenous traditions transmitted orally over long generations. Only when archaeologists, as valued interpreters of their unique historical ``texts,'' are accorded seats in the same seminar room will historical

6

Hawaiki, ancestral Polynesia

anthropology truly be able to encompass the longue dureÂe of nonliterate societies. Also damaging to the effort to develop a historical ethnography has been the postmodernist critique in anthropology (e.g., Clifford and Marcus 1986), which among other things has eschewed or rejected regional and comparative perspectives.12 For a Paci®c example, in his book on South Coast New Guinea Cultures, Knauft struggles with the problem of describing and comparing ethnographic regions in the face of the postmodernist stance that such regions in and of themselves are no more than ``the result of a Western academic discourse that projects its own cultural biases and assumes incorrectly that these characterizations re¯ect other people's reality'' (1993:3; see also Knauft 1999). Signi®cantly, Knauft ®nds a key to the reinvigoration of ethnographic comparison in the analysis of ``historical context.'' While we do not dispute the potential validity of the critique that concepts such as ``cultural regions'' are anthropological constructions, we do ®nd disturbing the postmodernist tendency to dismiss such constructions out of hand, rather than on the basis of a critical examination of empirical validity. With respect to linguistics, we detect a renewed and more weighty interest in the intersection of its disciplinary contribution to the historical concerns among the various subdisciplines of anthropology. An example from the 1970s, notable for its methodological rigor, is Dyen and Aberle's (1974) reconstruction of Proto Athapaskan kinship systems. Marshall (1984) offered an exposition on the culture history of structural patterning in Oceanic sibling classi®cations, a line of inquiry more recently taken up by Hage and Harary (1996). A return to an interest in linguistics and archaeology is evident as one major theme selected for the 1994 World Archaeology Congress, stimulated in part by provocative ideas of Colin Renfrew (1987, 1989, 1992) on the spread of Indo-European (Blench and Spriggs 1997, 1998). Two of the most robust regional endeavors linking archaeological and linguistic evidence focus on Africa, and on the Paci®c. The ®rst includes the work of Ehret and his collaborators (Ehret and Posnansky 1982; Ehret 1998) on Mashariki Bantu origins and their spread in sub-Saharan Africa, and on Nubian speakers in the Sudan. In the Paci®c, collaborative linguistic, archaeological, and anthropological research has burgeoned since the 1970s. In his extensive writings leading toward the reconstruction of the Proto Austronesian lexicon, Blust (e.g., 1980, 1985, 1987, 1995a) advances many important hypotheses regarding early Austronesian social organization and culture, as well as the locations of homelands and particular protolanguages, stimulating new archaeological research. The Comparative Austronesian Project of the Australian National University (Fox, ed., 1993;

Prologue: on historical anthropology

7

Pawley and Ross 1994; Fox and Sather 1996; Ross et al., eds., 1998) has likewise adopted a research methodology explicitly incorporating a historical perspective, and drawing upon linguistic, comparative ethnographic, and archaeological approaches. Some of these trends in the study of the Austronesian language family and culture history are reviewed by Pawley and Ross (1993). Recently, McConvell and Evans (1997) attempt to bring archaeology and linguistics closer together, with a geographical emphasis on Australia. For those who, like us, would advance anew the cause of historical anthropology, Pawley and Ross (1993) make several salient claims. Although they concur that the job of the culture historian is to make sense of resemblances as well as differences by aligning the evidence compiled by various disciplines, Pawley and Ross point out a number of methodological challenges. One is the sizable gaps in the data sets provided by each contributing ®eld of study. A second issue ± the problem of synthesis ± is more serious and not so readily corrected. Whereas each discipline and subdiscipline has its own kinds of data and particular array of methods for their interpretation, historical anthropology (or ``culture history'' in their terms) as yet has no equally reliable procedures for marrying the evidence of different disciplines.13 A third problem is ``that much writing on culture history is marred by a weak understanding of linguistic methods'' (Pawley and Ross 1993:428). Nonetheless their conclusion is worth quoting in full: The problem of culture history is that it is an interdisciplinary enterprise, but the methods and data used by each of its major constituent disciplines are not readily comparable. Nonetheless such comparisons are necessary in order to evaluate competing hypotheses within disciplines and to gain a more complete picture of the past than any single method can provide. The AN[Austronesian]-speaking region offers exceptionally favorable conditions for such interdisciplinary research. Until recently, most prominent hypotheses about the culture history of the AN-speaking regions originated in the data of comparative linguistics or comparative ethnography, with scholars from these two disciplines generally working independently. Archaeology has been a vigorous latecomer. Early attempts at integrating linguistic and archaeological evidence concentrated on centers and directions of AN dispersal, with archaeology providing a chronological framework for linguistically-based scenarios. Currently, the focus of culture historical syntheses is shifting toward comparisons of the lexicons of reconstructed languages with the content and environmental contexts of various archaeological assemblages. There has been no serious attempt to square the recent ®ndings of historical human biology with those of other disciplines, but there are signs that this too is under way. (1993:452, emphasis added)

In sum, not since Sapir has there been such renewed interest in developing an interdisciplinary approach to historical anthropology. What Trigger, an

8

Hawaiki, ancestral Polynesia

archaeologist, espouses under the umbrella of ``holistic archaeology,'' the social anthropologist Biersack advocates under the rubric ``historical anthropology,'' while linguists Pawley and Ross label the same endeavor a kind of ``culture history.'' (Biological anthropologists might subsume it all under ``co-evolution'' and wonder about all the fuss.) This kind of ``culture history,'' moreover, is quite different from (although a congruent development out of ) ``traditional archaeology'' (Feinman 1997; Renfrew and Bahn 1991:407) or ``Americanist culture history'' (Willey and Sabloff 1980; Lyman et al. 1997) of the ®rst half of the twentieth century. One would be tempted to call such a project a ``New Culture History,'' were that label not already appropriated by others (e.g., Hunt, ed., 1989). Although the current emphasis on history has its ``new'' elements, its roots in anthropology run deep indeed, as a rereading of Sapir reminds us; the adjective ``new'' is hardly required. We thus ®nd the rubric ``historical anthropology'' elegantly suited to our purposes. These varied subdisciplinary efforts, not always coordinated but clearly tending toward a common direction of historical anthropology, might be seen on a larger canvas of late twentieth-century science as part of a movement toward increased sophistication of the ``historical sciences.'' Thus Stephen Jay Gould has drawn a distinction between two modes of science (1989:277±91).14 The ®rst mode (including traditional physics and chemistry, for example) is the Newtonian form concerned with universal laws of invariant expression, able to make predictions about a deterministic universe. In these largely experimental sciences, time is motion, and history is irrelevant. The second mode, of which geology is a good exemplar, is thermodynamically based, concerned with open (rather than closed) systems in which time and history ``matter'' (Gould 1986). This is the terrain of the historical sciences including cosmology, historical geology, evolutionary biology and ± notably ± archaeology and historical linguistics, in which retrodiction rather than prediction must be to the fore. As Gould (1980), Ernst Mayr (1982, 1997), and others have eloquently argued, in such historical sciences the recognition of contingency and a historical narrative mode of explanation become not only philosophically valid, but essential. As Gould cogently writes, ``If the primacy of history is evolution's lesson for other sciences, then we should explore the consequences of valuing history as a source of law and similarity, rather than dismissing it as narrative unworthy of the name science'' (1986:68). Our book integrates a study in method with a substantive, data-rich case: the reconstruction of the world of the Ancestral Polynesian homeland, of ``Hawaiki.'' Polynesia offers exceptionally favorable conditions for historical anthropology, a model region in which to investigate the congruence of

Prologue: on historical anthropology

9

history, phylogeny, and evolution (Kirch and Green 1987). We intend to explicate more fully the theoretical issues involved, as well as the methodological procedures required to forward a phylogenetic approach in historical anthropology. Biersack (1991:25), commenting on our 1987 contribution in Current Anthropology, wrote that ``judging by the responses to their . . . article, the effort [of Kirch and Green] to produce a historical archaeology . . . will prove as theoretically and methodologically challenging and as fraught with contention as the parallel effort in cultural anthropology has proved.'' The contention is anticipated. Such is inevitably the case with scholarship that aims, not to sit conformably and comfortably within its own disciplinary cocoon, but rather to reach across disciplinary boundaries, to engage in dialogue across ingrained scholarly traditions. We have written a work that dares to draw upon not just the theoretical perspectives and methodological approaches of our own ®eld of archaeology, but also those of historical linguistics and comparative ethnography. Our hope is that this effort will inspire a renewed appreciation of the power of a holistic, ``historical anthropology.'' Most importantly, if this book manages to move us closer to the kind of integrative anthropology envisioned decades ago by Edward Sapir, we shall be pleased.

part i

The phylogenetic model: theory and method As a problem, recognized since Aristotle, natural similarities come in two basic, largely contradictory styles. We cannot simply measure and tabulate; we must factor and divide. Similarities may be homologies, shared by simple reason of descent and history, or analogies, actively developed . . . as evolutionary responses to gould 1986:66 common situations.

Chapter 1

The phylogenetic model in historical anthropology Physical type and language, we would say, have no causal relationship; there is no functional reason why a given physical type should occur within a given language family. Therefore, when these two variables do show signi®cant concordance in their distribution this may well represent an important historical fact, namely that the explanation for their concordance can be traced to a common point somewhere in the past. A demonstration that these two factors are also uniquely accompanied by a systemic culture pattern . . . romney 1957:36 strengthens the belief in a common origin.

The ``phylogenetic model'' has a long pedigree within historical anthropology, traceable in its essentials to Sapir's 1916 monograph. In the 1950s, it was formally developed under the label of ``the genetic model,'' a term that might be confused with a strictly biological perspective of somatic (genetic) inheritance, and which we (Kirch and Green 1987) therefore replaced with ``phylogenetic model.'' This revised label emphasizes historical sequences of cultural differentiation or divergence within related groups, regardless of the mechanism of transmission. Indeed, in the complexities of human history, both somatic and extra-somatic modes of trans-generational inheritance are salient (Durham 1982, 1991; Boyd and Richerson 1985). Thus a phylogenetic model within historical anthropology must incorporate data and perspectives from the full range of anthropological subdisciplines, including biological anthropology, archaeology, historical linguistics, and comparative ethnology. In this introductory chapter we will sketch the intellectual history of the phylogenetic model within anthropology, including its applications in Polynesia; discuss some current issues surrounding its applicability; compare its principal methods with phylogenetic (cladistic) approaches in biology and linguistics; and, ®nally, argue the fundamental signi®cance of a phylogenetic understanding of homologous change within historical anthropology. Our aim, in short, is to lay out the theoretical background and framework necessary for developing speci®c methods for an integrated historical anthropology, especially as these apply to Remote Oceania and Polynesia. 13

14

The phylogenetic model: theory and method

A brief history of the phylogenetic model Kim Romney (1957) ®rst delineated the speci®c criteria for cultural phylogenetic units, building upon Fred Eggan's proposals (1954) regarding ``controlled comparison'' in anthropology (see also Goodenough 1957). Basing his argument upon the fundamental anthropological observation that there is no necessary relation or correspondence between language, biology, and culture, Romney outlined the essentials of what he termed the ``genetic model'': The genetic model takes as its segment of cultural history a group of tribes which are set off from all other groups by sharing a common physical type, possessing common systemic patterns, and speaking genetically related languages. It is assumed that correspondence among these three factors indicates a common historical tradition at some time in the past for these tribes. We shall designate this segment of cultural history as the ``genetic unit'' and it includes the ancestral group and all intermediate groups, as well as the tribes in the ethnographic present. The genetic unit represents a substantive segment of cultural history while the term ``genetic model'' refers to the conceptual framework which serves as a tool to order the data. (1957:36)

Romney's seminal proposals were expanded and re®ned by Evon Vogt (1964) in his introductory essay to a volume on Maya cultural development (see also Vogt 1994a). Vogt, like Romney, stressed that a ``common historical tradition'' in any area, such as the Maya, would need to be de®ned on independent criteria of (1) common physical type, (2) common systemic cultural patterns, and (3) genetically related languages. Vogt elaborated on the theoretical implications of the ``genetic model,'' explicitly comparing it to models of adaptive radiation in biology: In brief, the genetic model assumes that genetically related tribes, as determined by related languages, physical types, and systemic patterns, are derived from a small proto-group with a proto-culture at some time in the past. The model resembles that of the zoologist who views a certain species of animal as evolving and making an adaptive adjustment to a given ecological niche and then radiating from this point as the population expands into neighboring ecological niches. As the population moves into different ecological settings, further adaptive variations occur in the species. But these variations are traceable to the ancestral animal, or, in other words, back to the proto-type. In the genetic model, as applied to human populations, we assume that a small proto-group succeeds in adapting itself ef®ciently to a certain ecological niche and in developing certain basic systemic patterns which constitute the basic aspects of the proto-culture. If the adaptation proves to be ef®cient, the population expands, and the group begins to radiate from this point of dispersal. As members split off from the proto-group and move into neighboring ecological niches, they make appropriate adaptations to these new situations and begin to differentiate ± that is,

The phylogenetic model in historical anthropology

15

there are adaptive variations from the proto-type over time as the members of the genetic unit spread from the dispersal area. (1964:11±12)

Vogt moved the ``genetic'' model beyond a strictly theoretical concept, and proposed a series of methodological steps and procedures for its implementation in historical anthropology. These required ``the combined use of a number of linguistic, archaeological, physical anthropological, ethnological, and historical methods bringing to bear the full range of anthropological data as these become available from ®eld and archival research'' (Vogt 1964:12). Vogt gave primacy to the evidence of language, suggesting that an anthropologist should commence with ``the de®nition of genetic units in terms of genetically related languages.''1 As Sanders put it: ``Methodologically speaking, the basis of de®ning such genetic units should be linguistic because of the relative exactness of linguistic methods as compared to those of ethnography and archaeology'' (Sanders 1966). However, this was primacy only in using linguistic data to de®ne which groups to include within a speci®c genetic unit; linguistic data were always to be cross-checked against those provided by other sub®elds, and thus were not in any ultimate sense privileged.2 Vogt (1964:10±13) advocated eight steps for the application of the ``genetic'' model to a speci®c ``segment of cultural history'': 1 2 3 4 5 6 7 8

plot the geographical distribution of related languages; calculate time depth, using lexicostatistics and glottochronology; locate the dispersal area and spread of the proto-group; reconstruct the proto-language and proto-culture using the linguistic methods of lexical reconstruction; use archaeological data to test speci®c hypotheses generated by steps 3 and 4; check the sequences of divergence derived from linguistic and archaeological analyses with the independent evidence of physical or biological anthropology; use ethnohistorical materials to ``provide readings on the various branches of the genetic unit'' between the time of ®rst European contact and the present; and add ethnographic data on contemporary communities to ``map variations in systemic patterns that have survived from earlier time levels and to detect cultural `drifts' or trends that are still occurring in these living systems.''

These steps constitute an integrated methodology for delineating an evolutionarily meaningful unit, one whose branches have diverged from a

16

The phylogenetic model: theory and method

common ancestor, according to a historical sequence that can be temporally and geographically de®ned. Despite some necessary modi®cations and re®nements to Vogt's research procedure (to be discussed below), Vogt's methodology remains eminently sound and reasonable. In effect, Vogt proposed to bring the full holistic power of twentieth-century anthropology to bear on the problems originally outlined by Sapir (1916) nearly a halfcentury earlier. Despite its potential, Vogt's research strategy was not widely applied, in part because of the move by sociocultural anthropologists away from an interest in historical and evolutionary issues, as discussed in our Prologue. Archaeologists, too, increasingly downplayed history and homologous change. Flannery and Marcus (1983), however, explicitly used Romney and Vogt's ``genetic model'' in their insightful study of divergence among the Zapotec and Mixtec populations of Mesoamerica. It was an initially independent reading of their work that inspired the two of us to collaborate on a joint application of the phylogenetic model to Polynesia (Kirch and Green 1987). Controlled comparison in Polynesia Edwin G. Burrows (1938a, 1940) ®rst championed the Polynesian cultures as an exemplary unit for controlled comparison. His classic monograph, Western Polynesia: A Study in Cultural Differentiation, drew explicitly on Sapir's methodology (1916) and established Polynesia as a cultural area (Figure 1.1). Burrows lacked the advantages of a developed archaeological record, and of careful historical linguistic analyses of relationships between Polynesian groups; these were to come only later. Thus, his evidence was con®ned to comparative ethnography, examining the distribution of a range of cultural ``traits,'' including material culture, kinship systems, cosmogony, and religious beliefs. Nonetheless, Burrows deduced a series of ``historical processes which had apparently brought about the differentiation of western from central-marginal Polynesia'' (1938a:92), including diffusion, local development, and abandonment or rejection of speci®c cultural traits.3 In the context of a renewed emphasis on Paci®c regional studies after World War II, Ward Goodenough (1957) authored a programmatic agenda for comparative research in Oceania. Marshall Sahlins (1958) produced the ®rst new comparative study of Polynesia, now theoretically situated within a cultural evolutionary framework (Sahlins and Service 1960), and explicitly invoking a phylogenetic analogy by describing Polynesian cultures as ``members of a single cultural genus that has ®lled in and adapted to a variety of local habitats'' (1958:ix). Sahlins, however, was interested neither in phylogenetic analysis per se, nor in the reconstruction of historical trajectories of change

Fig. 1.1

Map of the Polynesian triangle and the Polynesian Outliers.

18

The phylogenetic model: theory and method

within Polynesia. Nor were such historical issues the main concern of Goldman (1955, 1970), who like Sahlins regarded Polynesia as a group of genetically related societies admirably suited for comparative analysis. Goldman (1970) nonetheless incorporated newly emerging archaeological data into his work, which, along with genealogically based oral traditions and related ethnohistorical records, provided a historical context for his comparative analysis of the Polynesian ``status system'' and of descent group organization. Motivated by broad theoretical questions of cultural evolution, both Sahlins and Goldman were more interested in ``process'' (analogic change) than in particular sequences of homologous change. Moreover, their enterprises relied on synchronic data sets ± the ethnographic record ± only minimally integrating information from historical linguistic and archaeological sources. As a consequence, particular ethnographic endpoints in their evolutionary schemes inevitably stood as exemplars of putative earlier stages in the historical process. To use Goldman's model as an example, the ``Traditional'' societies of Tikopia, Pukapuka, or Ontong-Java represented an original, ancestral form of Polynesian society. Such a strategy ± we now know ± simply will not work for, to paraphrase the great evolutionist George Gaylord Simpson, ``one cannot be one's own ancestor.'' Ethnographically attested societies are not the changeless descendants of their ancestors, even though they may be assessed as culturally conservative. Thus the pioneering strategies of Sahlins and Goldman ± although they yielded valuable insights through systematic ethnographic comparison ± are not suitable as models for a theoretically rigorous historical anthropology. The phylogenetic model applied to Polynesia In the early 1980s, Kirch (1980, 1984a) attempted a broad synthesis of historical change within Polynesia, using an explicitly comparative and evolutionary approach, and according archaeological evidence primacy over the ethnographic data emphasized by Burrows, Sahlins, or Goldman. Although unaware at the time of Romney's or Vogt's ``genetic'' models, Kirch (1984a:5±8) proposed an essentially identical type of ``study of internal differentiation of Polynesian societies,'' one designed to draw on the power of holistic anthropology: Precisely because Polynesia as a region consists of a series of discrete, but historically related societies ± all derived from a common ancestor ± and because there was direct historical continuity between the ``ethnographic present'' and the prehistoric past, we are in an excellent position to draw upon ethnohistoric, ethnographic, and linguistic data, as well as upon strictly archaeological evidence in an attempt to understand the region's prehistory. The Polynesian ethnographic baseline does not

The phylogenetic model in historical anthropology

19

provide mere analogies for the interpretation of archaeological data; it illuminates directly the endpoints of indigenous developmental sequences. (Kirch 1984a:5)4

A graphic model, reproduced here as Figure 1.2, illustrated the process of differentiation within Polynesia. It was fundamentally a phylogenetic model, in which the ethnographically attested Polynesian cultures and societies were regarded as ultimately derived from a proto-group, termed ``Ancestral Polynesian Society'' (abbreviated APS in the diagram). Although Kirch emphasized a series of successive colonization events (migrations out of the original APS homeland, and out of later daughter communities), and the effects of subsequent isolation between descendent populations, he explicitly pointed out that cultural contact and borrowing had occurred between some island groups (as depicted by the double-arrow linking W3 and X3 in the diagram). Kirch singled out the reconstruction of Ancestral Polynesian Society as a critical step in any evolutionary study of cultural differentiation within Polynesia. Knowledge of the APS ``baseline'' was necessary in order to assess or measure later historical changes in the descendent cultural traditions within Polynesia. Only by ®rst having some reasonable idea of the social and technological bases of APS would it be feasible to determine which later features were retentions, adaptations, or elaborations of older patterns, and which were entirely new innovations, borrowing, or at times convergences. Recognizing that archaeology alone was insuf®cient to reconstruct some aspects of APS (such as social structure), Kirch drew upon evidence from linguistics to outline important aspects of APS, including technology, production systems, and social relations (Kirch 1984a:53±69). His methods for such reconstruction were in retrospect insuf®ciently developed, and some of his reconstructions were later challenged (e.g., Sutton 1990, 1996). We will redress these initial methodological shortcomings, and signi®cantly extend and improve his interpretations of Ancestral Polynesia in Part II. In 1987, Kirch and Green put forward an analysis of Polynesia, for the ®rst time explicitly referring to Romney's ``genetic'' model. Others had already delineated Polynesia's advantages for historical anthropology; we sought to extend those advantages by adopting a formal set of procedures under the label of the ``phylogenetic model.'' These derived initially from the analytical steps outlined by Vogt (1964), to whose work we had been introduced through our reading of Flannery and Marcus' monograph, The Cloud People (1983). We demonstrated how Vogt's procedural steps could be more rigorously applied in Polynesia, but also argued some ``initial propositions'' regarding ``evolutionary process'' within the Polynesian region. We emphasized the critical importance of ``establishing homologies, thus clearing the path for the analysis of evolutionary process'' (1987:432). We

20

The phylogenetic model: theory and method

Fig. 1.2

Kirch's 1984 model of phylogenetic differentiation in Polynesia (from Kirch 1984a: ®g. 1).

The phylogenetic model in historical anthropology

21

found inspiration in Stephen Jay Gould's assertions regarding what he called the ``triumph of homology, or why history matters'' (Gould 1986). In Gould's succinct phrasing, ``once we map homologies properly, we can ®nally begin to ask interesting biological questions about function and development ± that is, we can use morphology for its intrinsic sources of enlightenment, and not as an inherently ¯awed measure of genealogical relationships'' (1986:68). Although Gould wrote as a biologist, his arguments apply equally to issues of cultural evolution (with the proviso that similarities in culture can result from contact and borrowing, as well as from descent). By ®rst determining the sequence of phylogenetic divergence within Polynesian societies, and by reconstructing as thoroughly as possible the baseline of Ancestral Polynesian culture, we could ``get our history under control,'' thus enabling far more rigorous analyses of cultural evolutionary process. Although we hardly intended our 1987 paper as an exhaustive study of such evolutionary processes (let alone a thorough reconstruction of APS) we did outline several ``mechanisms of divergence,'' ``parallel evolutionary processes,'' and processes of ``convergence,'' which we thought would prove to be of signal importance in Polynesia. Among potential mechanisms of divergence, we discussed: (1) isolation; (2) founder effect or drift; (3) colonization, including adaptation to new and contrastive environments; (4) long-term environmental selection resulting both from natural and from human-induced environmental changes; and (5) external contact.5 Among the parallel evolutionary processes alluded to in our 1987 paper are: (1) demographic factors, such as a cultural analogue of the well-known r/K selection continuum in evolutionary ecology; (2) intensi®cation and specialization of production; and (3) increased competition over time. We also commented on the ``analogic emergence of similar traits or structures'' in Polynesian societies that had not been in direct contact with each other, and where such structures were evidently convergences. A key aspect of the phylogenetic model insuf®ciently developed in our 1987 paper was the procedure for reconstructing the ancestral culture and proto-language, that stand at the base or root of any phylogenetic unit. This procedure, which we here call the ``triangulation method,'' is a main thrust of this book, and the subject of Chapter 2. Necessary modi®cations to Vogt's methodology By 1987, we were both aware of problems or inadequacies in Vogt's (1964, 1994a) proposed phylogenetic methodology. We especially took issue with his reliance upon lexicostatistics and glottochronology, not only to determine the subgrouping relationships among the languages under consideration, but also to calculate the time depth of cultural differentiation within a particular

22

The phylogenetic model: theory and method

``segment of cultural history.'' Vogt's recommended procedure had been used by Marcus (Topic 2 in Flannery and Marcus 1983) to outline a sequence of linguistic divergence among the ``Otomangueans'' who include the modern Zapotec and Mixtec. In reading Flannery and Marcus' otherwise brilliantly argued volume, we realized that Polynesian scholars had surpassed our Mesoamerican colleagues, because historical linguists in the Paci®c had already recognized the pitfalls associated with a strictly lexicostatistical approach. Paci®c lexicostatistical studies in the 1950s and 1960s (e.g., Dyen 1965) had failed to provide the same level of acceptable propositions about subgrouping as the comparative method. Lexicostatistics and glottochronology have little following today among Paci®c scholars, and we do not employ them, for reasons detailed below.6 In retrospect, we should have made this point more prominently in our 1987 paper; the subgrouping model for Polynesian languages that we used there (Kirch and Green 1987: ®g. 2, after Biggs 1971) was based not on lexicostatistics, but rather on a great deal of careful comparative work on phonology, lexicon, and morphology.7 It was a subgrouping model based on signi®cant numbers of shared innovations (as opposed to retentions, or generalized similarity), a methodological advance comparable to the paradigm shift from phenetic to cladistic methods in biological phylogenetics (see discussion, below). Furthermore, in Polynesia we did not rely on the questionable method of glottochronology to calculate the time depth of differentiation, as archaeological work in the Western Polynesian ``homeland'' area had independently established this temporal frame on the basis of radiocarbon dating (Green 1981). We cannot overstress the importance of applying the ``genetic comparative method'' in historical linguistics when developing a phylogenetic model in historical anthropology. For reasons detailed in Chapter 2, this is the approach we follow and strongly advocate for historical anthropology in general.8 It is the only approach that leads directly to a corpus of soundly based lexical reconstructions for a proto-language's content (Ross 1997:254). In Polynesia, the comparative method has provided increasingly precise models of subgrouping, and in turn a large and well-attested set of lexical reconstructions. A similar situation obtains even in the more subtle and entangled dialect continuum of Fiji, where a strict family tree model is even less appropriate (Geraghty 1983; Pawley 1999). The same is true at higher levels in the Austronesian language family, where the comparative method yields both the major Oceanic subgroup (Pawley and Ross 1993:433±34), and various internal subgroups within it (Ross 1988; Pawley and Ross 1993:439±40; ®g. 2; Green 1997b, ®g. 4).9 The value of subgrouping based on the genetic comparative method likewise holds for the highest-order subgroups within Austronesian (Pawley and Ross 1993:435±39). Beyond the

The phylogenetic model in historical anthropology

23

Austronesian region, experience with several language families (IndoEuropean, Austronesian, Niger-Congo, etc.) again indicates that the comparative method has the most to recommend it, especially when integration with archaeological and other lines of evidence is intended. Dendritic versus reticulate models in historical anthropology After decades of relative neglect, a phylogenetic perspective in historical anthropology has again become the focus of debate.10 In two notable papers, John Moore (1994a, 1994b) challenges the view that human societies have typically differentiated in a branching or phylogenetic manner, arguing instead that rhizotic or reticulate processes of ``ethnogenesis'' are more common in human prehistory. Moore's critique, which extends generally to the application of cladistic methods within historical anthropology, was motivated in part by adverse reactions to the overextended claims of CavalliSforza and others associated with the Human Genome Diversity Project (Cavalli-Sforza et al. 1988, 1994); these authors implied that virtually all of human history can be reduced to a kind of master ``family tree'' with unambiguous links between biological populations and major language families. Moore's ``ethnogenetic'' approach is in¯uenced by his own ethnographic and ethnohistorical research among Native American populations in central and southeastern North America, where, for example, he convincingly demonstrates historical fusion as well as ®ssion among groups such as the Lakota and Cheyenne. Nevertheless, Moore does not wholly reject a phylogenetic or cladistic model, and he points to Polynesia as one region where such models are useful and appropriate (Moore 1994b:14). In contrast, Ruth Mace and Mark Pagel (1994) just as cogently argue the necessity of applying a formal comparative method within anthropology, one that explicitly incorporates a cladistic or phylogenetic method. They write: The critical point . . . is that the validity of comparative methods for anthropology depends upon correctly counting independent instances of cultural change. Independent instances of cultural change, in turn, cannot be identi®ed without the construction of a phylogeny (or cladogram) showing the patterns of hierarchical descent of the cultures being studied. (1994:551)

As in Vogt's method (which they seem to be unaware of and do not cite), Mace and Pagel (1994:552) argue that ``all the evidence available, be it genetic, linguistic, archaeological, historical, or cultural'' should be drawn on in the complex task of constructing phylogenies. As with Vogt and ourselves, Mace and Pagel see linguistic evidence as occupying a critical position, for ``whilst being far from perfect, [language] may offer the best

24

The phylogenetic model: theory and method

general method of reconstructing cultural phylogenies that we have'' (1994:553). But Mace and Pagel are not so naive as to think that the construction of cultural phylogenies is identical to the process of phylogeny construction in biology. This is due to the problem of ``horizontal transmission'' between human cultures, what anthropologists (and linguists) classically labeled ``diffusion'' or ``borrowing'' (what we here call synologies). They thus recognize the issues raised in Moore's critique, but do not regard them as overwhelming. Dendritic versus rhizotic ± or phylogenetic versus ethnogenetic ± models have been explicitly debated by historical anthropologists working within the vast Austronesian-speaking region, including Polynesia. Bellwood et al. (1995), for example, see much merit in applying a phylogenetic model to the entire realm of Austronesian-speaking societies and cultures.11 For Madagascar, Dewar (1995) argues the relative strengths of reticulate versus dendritic processes in that island's cultural history. John Terrell and his associates (Terrell 1986, 1988; Welsch et al. 1992; Welsch and Terrell 1994; Welsch 1996), on the other hand, insist that there is no phylogenetic patterning discernible in the region of Near Oceania, and Terrell et al. (1997) extend this argument to the whole of Oceania, proclaiming the death knell of a phylogenetic and comparative approach. They would see the Paci®c as a sea of islands criss-crossed by so many reticulate pathways of interaction that any attempt to infer cultural phylogeny is inherently doomed to failure. Yet Terrell et al.'s own north coast New Guinea research, which they hold up as a model case in point (Terrell et al. 1997), fails to support their contentions. Rather, as Moore and Romney (1994, 1996; Roberts et al. 1995) have cogently demonstrated, the New Guinea data set developed by Terrell and Welsch yields strong patterns of association between language and culture, even though much contact (``horizontal transmission'') between Austronesian and Non-Austronesian cultures has occurred in the region for at least 3,500 years. Moreover, recent molecular genetic work (Merriwether et al., 1999) indicates strong associations between Austronesian languagespeaking populations and certain genetic traits, such as a particular 9-base pair deletion, reinforcing earlier indications of strong biological±linguistic correlates in dermatoglyphs and other features (Froehlich 1987). Thus, although rhizotic or reticulate processes of cultural interaction have been signi®cant in the cultural history of Near Oceania over the past several millennia (Kirch 2000), in our view these were not so pervasive as to have wholly masked or eliminated underlying phylogenetic patterns. Bellwood (1996a) casts the debate over phylogeny versus reticulation more broadly, arguing that both approaches have their place in the repertoire of historical anthropological models, a point we enthusiastically second. As

The phylogenetic model in historical anthropology

25

Bellwood observes, much depends on the temporal and spatial scales at which an investigator works. As in Moore's case examples from the American southeast, we think that Bellwood is right to point out that these concern ``societies already under the in¯uence, whether bene®cial or malignant, of colonial authorities,'' and these societies, moreover, had suffered from several centuries of severe demographic collapse. Societies under adverse colonial in¯uence differed entirely from those associated with earlier expansions of agricultural peoples in various parts of the world, beginning in the early Holocene, fueled by demographic increase. Bellwood expresses our own viewpoint quite eloquently: Large-scale and fairly integrated colonizations did happen in prehistory; human cultures and languages can, to varying degrees depending upon time and space coordinates, be organized in phylogenetic arrays. The generation of human diversity in the past has not been entirely reticulate and dependent on processes of in situ interaction between peoples of different ethno-linguistic background. Neither has it been entirely radiative and dependent upon adaptation in isolation. But to rule out phylogeny as of any signi®cance in the patterning of difference and similarity between human cultures is surely no more than a ``whimsical view.'' (Bellwood 1996a:888)

Phylogenetic analysis in biology, linguistics, and anthropology A phylogenetic model of cultural evolution obviously has close parallels with its more widely understood application in evolutionary biology, as the study of divergence or radiation among groups of historically related lineages (Gould 1980; Mayr 1982). Yet there are signi®cant differences, resulting from the nature of ``dual transmission'' in cultural evolution (Durham 1982, 1991; Boyd and Richerson 1985) and from the complications arising from ``horizontal transmission,'' as discussed above. A brief comparison of basic theoretical tenets of modern phylogenetic theory and practice in biology (what is today commonly referred to as ``cladistics'') with those underpinning the genetic comparative method in historical linguistics is instructive. Such comparisons reveal the same fundamental emphasis on procedures for discovering characters that are truly innovative and hence exclusively shared by a set of sister groups (whether these be species or languages); they also emphasize that truly evolutionary phylogenies must be based on analysis of ``derived'' characters. As Mayr (1982:209±21) points out, Darwin recognized that ``af®nity is proximity of descent,'' and that phylogenetic systematics should therefore be based on a principle of monophyletic classi®cation, but later practitioners of macrotaxonomy largely ignored the ``splendid start'' he had made. The

26

The phylogenetic model: theory and method

inconsistent procedures for macrotaxonomy that developed among systematists throughout the ®rst half of the twentieth century set the stage for a new approach, which seized hold of the opportunities for large-scale numeric quanti®cation and statistical comparison provided by the development of high-speed computers. This approach, at ®rst called ``numerical taxonomy'' and later ``numerical phenetics,'' or simply ``phenetics'' (Sokal and Sneath 1973), used large arrays of metric or discrete characters, for which statistical assessments of ``similarity'' were then calculated ``by the mechanical operations of the computer'' (Mayr 1982:222).12 The phenetic approach enjoyed a short period of dominance in the ®eld of biological taxonomy, but it was doomed to failure because it equally weighted all characters, and it wilfully ignored phyletic information. Speci®cally, phenetics made no attempt to discriminate characters that were shared retentions of a much older common ancestor (and thus found in more taxa than just those of the sister groups under comparison), from characters that were uniquely inherited from the immediate ancestor of the species being considered. The greatest advance in phylogenetic analysis in biology, which has come to be known as ``cladistics,'' was initially developed by Willi Hennig (1965, 1966).13 Hennig's advance ± which was in effect merely a return to the principles advocated by Darwin ± was to hold that phylogenetic classi®cations should seek to represent genealogy, or common descent. Hennig (1965:101±2, ®g. 1) distinguished between three quite different categories of morphological resemblance, categories that had not been suf®ciently discriminated among by prior generations of systematists, and worse, had been wholly ignored by the pheneticists. Hennig termed these categories: (1) plesiomorphy, similarity that results from using ancestral characters which may be shared among taxa well beyond the ``sister group'' under consideration; (2) convergence, similarity owing to analogous (as opposed to homologous) innovations; and (3) synapomorphy, similarity that truly re¯ects mutually exclusive derivation among a sister-group of taxa, and which therefore is the only kind of similarity which can truly be used for phylogenetics.14 Since Hennig's initial proposals and the development of a core methodology, there has been a vast outpouring of literature on the theory and methods of cladistics, much of which is worthy of consideration by historical anthropologists (and indeed has in¯uenced such scholars as Mace and Pagel [1994]).15 A thorough review of these materials would digress well beyond our intended scope. Here we simply point to the core of Hennig's cladistic methodology, in Ernst Mayr's clear prose: The crucial aspect of cladistics is the careful analysis of all characters in the comparison of related taxa and in the partitioning of these characters into ancestral (plesiomorph) and uniquely-derived (apomorph) characters. Branching points in the phylogeny are determined by the backwards tracing of uniquely derived characters

The phylogenetic model in historical anthropology

27

(synapomorphies) because such apomorph characters are believed to be found only among the descendants of the ancestor in which the character ®rst occurred. (1982:227)

The development of historical linguistics ± a discipline also fundamentally concerned with the methodology of historical reconstruction ± displays remarkable similarity to that of biological macrotaxonomy during the course of the nineteenth and twentieth centuries.16 The early philologists, beginning with Jacob Grimm, Friedrich Schlegel, and Franz Bopp, and continuing with others such as Brugmann (1884), had worked out a series of principles of historical reconstruction and of the subgrouping of related languages. These were codi®ed in several classic texts of the early to mid-twentieth century (e.g., Bloom®eld 1933; Hoenigswald 1960), and laid out such principles as the working out of regular sound correspondences, based on careful, systematic comparison of cognate morphemes from a set of languages hypothesized to be related (Hoenigswald 1963; Allan 1994). In the 1950s and 1960s, motivated by the development of radiocarbon dating in archaeology and by the desire for a comparable method that could give quick and precise results, some historical linguists began to advocate a set of strictly quantitative procedures, called ``lexicostatistics.'' In many respects, this paralleled the phenetic approach in biology. The inventor and chief proponent of this approach was Morris Swadesh (1952, 1955; see also Gudchinsky 1956), who extended the method of numerical comparison to a core list or ``basic lexicon,'' which he reasoned could then be used as a method for absolute dating of pairwise language separations. As Trask (1996:362) points out, however, lexicostatistics ``can be applied only after the languages of interest have been shown to be related and after cognate words have been securely identi®ed''; that is, after the basic hard work of the comparative method has been done. Unfortunately, lexicostatistics has sometimes been used as a substitute for the comparative method itself. One example was Dyen's (1965) vast statistical comparison of Austronesian languages (requiring access to newly developed high-speed computing), leading to a ``family tree'' or classi®cation that has proven to be completely at odds with the tree resulting from application of the comparative method (Grace 1966; Guy 1980; Blust 1981b). For reasons discussed by Wang (1994:1,446), glottochronology ± as an extension of lexicostatistics ± was also fraught with uncritical assumptions, and is little used today. Most historical linguists have now returned to the properly cladistic methodology for subgrouping which has been available to them for more than a century (i.e., since Brugmann 1884). Lexicostatistics and glottochronology were remarkably parallel, in both theory and method, to the approach of the ``numerical taxonomists'' or

28

The phylogenetic model: theory and method

``pheneticists'' in biology. Neither paid proper attention to the weighting of characters (in linguistics, lexemes), or to a method for discriminating ancestral (i.e., plesiomorphic) from derived (i.e., synapomorphic, or innovative) characters, so essential in the construction of rigorous phylogenetic models. This brief comparison of phylogenetic theory and method in biology and historical linguistics has critical implications for historical anthropologists who would seek to advance phylogenetic analysis and reconstruction using archaeological and ethnographic data. The signi®cance of phylogeny for historical anthropology We believe that a phylogenetic approach should be a central component of the tool kit of the historical anthropologist; we do not claim it should be the only component, or even necessarily the most important component, in that kit. To draw an analogy again with biology, our position closely parallels that of Eldredge and Cracraft (1980), who argue the necessity of integrating phylogenetic systematics with the study of evolutionary process, especially macro-evolutionary process.17 The cultural equivalent of microevolution is probably best observed in the short-term ethnographic (or ethnohistoric) record, and hypotheses concerning cultural change in the short term do not necessarily require close historical analysis (i.e., synchrony can be assumed). When we turn to the macroevolutionary equivalent in cultural evolution ± the longue dureÂe of change that leads to the emergence of new cultural con®gurations out of older ± control over history becomes essential. Boyd et al. (1997:376) make this same point when they state that ``reconstructing the histories of peoples without written records requires that one distinguish between homologies . . . analogies . . . and synologies (similarities produced by diffusion or borrowing).'' Just as the testing of macroevolutionary hypotheses in biology, then, ``require[s] a cladogram'' (Eldredge and Cracraft 1980:327), we would argue that the testing of hypotheses of long-term cultural change necessitates the careful working out of cultural phylogenies. And we hasten to add ± lest we be misunderstood on this point ± that phylogenetic analysis would certainly include the testing of the null hypothesis that any particular group of cultures under consideration does not, in fact, constitute a monophyletic unit, or that similarities between them have arisen from signi®cant interaction, diffusion, or borrowing (rather than from common descent). Many contemporary archaeologists ± if not all those in related disciplines who might consider themselves to be ``historical anthropologists'' ± operate within one or another evolutionary theoretical framework. But these frameworks vary considerably in various aspects of their epistemology and in their

The phylogenetic model in historical anthropology

29

premises about the fundamental processes of evolution with respect to culture (Spencer 1997). Although most archaeologists would now probably regard a unilineal or progressive evolution by stages (sometimes called, unfortunately, ``cultural evolution'') as naive and outdated, they are by no means agreed on what model should replace it. At least two principal camps can be discerned, termed the ``processualists'' and the ``selectionists'' by Spencer (1997:210).18 It seems to us, however, that the essential problem of disentangling homology from analogy or synology is equally critical to both schools, and as Neff and Larson write in a recent review of the methodology of comparison in evolutionary archaeology, ``evolutionary archaeology is as much about history as it is about ecology and economics'' (1997:83±4). They go on to note that ``history matters . . . because use of comparative data to test hypotheses about cultural adaptation requires independence of the cases being compared.''19 Lyman and O'Brien, representing the contemporary selectionist school of evolutionary archaeology, likewise insist that ``history matters'' (1998:622±23). Certainly not all historical anthropologists are interested in evolution, and not even all archaeologists. But the importance of ``getting one's history right,'' of working out particular historical trajectories of cultural change and sequences of cultural differentiation and divergence, is not con®ned to evolutionary approaches within anthropology or archaeology.20 In our view it applies equally to schools of thought that regard themselves as distinct from either evolutionary ``processualists'' or ``selectionists,'' for example explicitly Marxist theorists (McGuire 1992). It was, after all, Marx himself who said that men make their own history not ``just as they please,'' but rather ``under circumstances directly found, given and transmitted from the past.'' In his famous phrase, ``the tradition of all the dead generations weighs like a nightmare on the brain of the living'' (Marx 1978:595). We will not further belabor our point, but simply conclude by stressing that, for us, the determination of phylogenetic sequences is not an end in itself, although it has intrinsic interest. Rather, phylogenetic analysis in historical anthropology provides the means to the following ends: (1) it offers a degree of methodological rigor necessary for the reconstruction of ancestral cultural patterns; and (2) it allows for a more controlled and rigorous study of change among descendent groups. Objectives of this book Ward Goodenough, an early champion of controlled comparison, more recently proclaimed that: ``Remote Oceania, where we can presume that we are dealing with phylogenetically related cultural traditions, as well as with phylogenetically related languages, is an excellent area in which to conduct

30

The phylogenetic model: theory and method

. . . systematic examination of structurally homologous traditions'' (1997:24). In our view, this exercise requires a well-developed program in historical linguistics (using the genetic comparative method), paralleled by an ethnographic corpus that can be engaged with historical issues, and by an archaeology whose outcomes are solid enough to address wider historical interests. We agree with Pawley and Pawley that ``for doing culture history several disciplines are, ultimately, better than one'' (1998:209). Sapir wisely concluded his pioneering essay, Time Perspectives in Aboriginal American Culture, with the comment that ``anything like real completeness is, of course, entirely out of the question'' (1916:460). Heeding Sapir's words, we do not aim here at a full and systematic method for historical or evolutionary analysis. Our more modest objective is to explore some fruitful avenues of phylogenetically based reconstruction that have emerged in Polynesia, part of the Remote Oceanic region viewed by Goodenough as particularly favorable to phylogenetic study. A truly sophisticated methodology for historical anthropology is something that will develop only over time and through practice, as others apply it to varied units of time and space, in regions well outside the Paci®c. An exhaustive application of the phylogenetic model would require tracing each different local branch and divergence within Polynesia historically, and examining each instance of a parallel trend or convergence. Such a grand enterprise is not our purpose here (although we hope to see it someday accomplished). Rather we will explore more fully the ``commonly inherited structural base'' (Kirch 1984a:262), what we call Ancestral Polynesian culture. We limit ourselves to this aspect of the phylogenetic model in Polynesia for two reasons. First, reconstruction of the ancestral, structural base from which all later Polynesian societies and cultures arguably descended is an essential step that must be accomplished before more sophisticated analyses of cultural divergence and evolution can proceed. Second, believing that the methods for historical reconstruction themselves need further re®nement, we hope to contribute to the general ®eld of historical anthropology by working these methods out through their application to Polynesia. These may then be applied beyond the particular timespace constraints of our own study region. In Chapter 2, we consider in detail the theoretical principles and methodological procedures for correlating linguistic and archaeological evidence, speci®c problems of lexical and semantic reconstruction, and the application of the direct historical approach, all in the context of Polynesia. This is followed, in Chapter 3, by a discussion of the speci®c linguistic, biological, ethnographic, and archaeological evidence permitting us to circumscribe Polynesia as a discrete and well-de®ned phylogenetic unit, and to delineate its major internal branches. These two chapters develop the

The phylogenetic model in historical anthropology

31

conceptual and methodological tools allowing us to tackle the substantive problem in Part II, the historical reconstruction of Ancestral Polynesian culture and societies. It may be that in Polynesia we have chosen an unusually simple case, and some will say that it would be dif®cult or impossible to apply our approach in more complicated situations. We are not so pessimistic. By focusing on a ``peripheral'' region such as Polynesia, where time depth is relatively shallow and where linguistic diversity is restricted to a single subfamily of Austronesian, we can work out our principles and methods most clearly.21 But we would argue ± as has Goodenough (1997:22±24) ± that the phylogenetic approach is just as applicable elsewhere in Remote Oceania. We believe it can be applied productively even in Near Oceania, complicated as that region is by a time depth of more than 30,000 years, by signi®cant cultural complexity, and by great linguistic diversity at the family level. Moreover, we are convinced that a phylogenetic approach in historical anthropology holds great promise for other regions of the world.22

Chapter 2

Methodologies: implementing the phylogenetic model Anthropologists have long recognized that cultural traditions in different societies can be related phylogenetically in that they derive historically from a common ancestral tradition in the same way that languages can be related phylogenetically. Problems of method for convincingly establishing such relations for cultural traditions have remained unresolved. Remote Oceania, where we have reason to assume that nearly all existing cultural traditions are phylogenetically related, offers possibilities for comparative study to illustrate the goodenough 1997:16 methodological issues to be resolved.

Sophisticated application of a phylogenetic model in historical anthropology requires a rigorous methodology. Using our Polynesian case, we begin with the problem of correlating linguistic and archaeological evidence, followed by possible models for linguistic divergence. We will show that such models are not limited to strictly ``tree-like'' branching models; historical linguists have recently developed ``network-breaking,'' or ``dialect-chain'' models that are more appropriate to Paci®c history. This in turn leads to matters of de®ning ``homelands,'' and of establishing time depth. All of these issues are essential to establishing a proper phylogenetic framework. The second half of this chapter examines what we call the ``triangulation method'' in historical anthropology, which allows one to achieve a reconstruction of the ancestral culture which lies at the root of a phylogenetic ``tree.'' The procedural details of the triangulation method will take us into the realm of lexical and semantic reconstruction, especially the ``recovery of meaning'' by using comparative ethnographic materials to develop speci®c semantic history hypotheses. Correlating linguistic and archaeological evidence in Polynesia In his pioneering formulation of a phylogenetic method, Vogt (1964; see also 1994a) advocated commencing with linguistic evidence, in order to de®ne the scope of a particular ``segment of cultural history'' (see Chapter 1). 32

Methodologies: implementing the phylogenetic model

33

However, matters are more complicated than Vogt imagined, necessitating some modi®cations. For reasons explained in Chapter 1, we reject lexicostatistics and glottochronology as inadequate for determining the time depth of linguistic splits within a language family. Moreover, we see the need to identify linguistic ``homelands'' ± within which one can search for archaeological correlates of any given proto-language ± as a necessary part of a linguistic methodology. For Remote Oceania including Polynesia, Pawley (1966, 1967) and Green (1966) initially applied family tree models based on the comparative method, and tried to work out their implications for Polynesian archaeology and prehistory. Building upon these early endeavors, Pawley and Green (1973) outlined ®ve principles, which we also take as our starting point for applying linguistic evidence in historical reconstruction. They began by labeling ``the language of the ®rst well-established population in an island the foundation language, and the language of any later arrivals an intrusive or invading language'' (1973:38). Equivalent terms can be applied to archaeological sequences with the ®rst well-established cultural traditions in a region being the foundation culture, and later traditions designated as intrusive or invading cultures. Pawley and Green equated the foundation culture with the earliest cultural tradition appearing in the archaeological record. For the western sector of Remote Oceania (i.e., the Reef/Santa Cruz Islands, Vanuatu, Loyalty Islands, New Caledonia, and Fiji±Western Polynesia) this foundation culture is the Lapita cultural complex (Green 1979a, 1997b; Spriggs 1984, 1991, 1997; Kirch 1997a). Green (1997b) argues that the only apparent intrusive cultures in this sector occur on the island of NendoÈ (Santa Cruz) and in the main Reef Islands, during the last 1,000 years. Possible intrusive cultures occur from Vanikoro to Tikopia and through northern and central Vanuatu to New Caledonia in the form of the Mangaasi culture,1 while invading cultures are well documented for various of the Polynesian Outliers such as Tikopia and Anuta (Kirch and Yen 1982; Kirch 1984b). For this same region, the foundation languages all fall within a set of closely related subgroups deriving from a late stage of Proto Oceanic (Ross 1988; Green 1997b).2 Again, intrusive language replacements include only Reef/Santa Cruz (Wurm 1969, 1970, 1978, 1992),3 while invading languages are represented by the various Polynesian Outlier languages (Pawley 1967), and by the northern Tongan case described by Pawley and Green (1973:39). Thus, within the vast Remote Oceanic area ± permanently settled only 3,200±3,300 years ago on radiocarbon chronology ± the linguistic and archaeological histories coincide remarkably, allowing the following principles to be advanced as guidelines (the following is paraphrased from Pawley and Green 1973:39±41):

34

The phylogenetic model: theory and method

Principle 1. Under the conditions obtaining in the Paci®c in pre-contact times, the foundation language of a remote island group could seldom be replaced by an intrusive language. A remote island group is one isolated by more than 450 km of open sea. The principle does not apply to non-remote island groups, although it may be that the de®nition of ``remote island group'' can be made more sophisticated, and that under certain conditions a distance of less than 450 km was suf®cient to reduce the frequency of two-way contact. The same claim cannot be made about the kinds of materials with which the archaeologist deals. Fusion of distinct cultural traditions occurs commonly, and a small intrusive population may have a large effect on the material culture of a region. For these reasons an analogous principle cannot be applied to material culture, or at least can be applied only with important reservations. Principle 2. Once a language X has become established on two island groups, separated from each other by more than 450 km of open sea, linguistic splitting (gradual divergence into separate dialects Y and Z) is inevitable. Principle 3. After 1,000 years Y and Z will have diverged to the point of being separate languages or will be very close to that point. (There are various criteria for determining whether two speech traditions are separate languages or dialects of one language. A low degree, or lack of, mutual intelligibility is one criterion of distinctness; some would give less than 80 percent, or less than 70 percent, cognation in basic vocabulary as another, more arbitrary, index.) Principles 2 and 3 make predictions about diversi®cation only between relatively isolated islands, not within an island group or archipelago. One might reasonably ask how long the language of a Neolithic society can persist as a unity in an island group where some neighboring islands are separated by shorter distances of 50 or 100 km. Is linguistic splitting inevitable under such conditions? The Tongan and Samoan cases show that linguistic unity can be maintained over single archipelagos of widely dispersed islands for more than 2,000 years. In fact, there are no known cases of separate languages developing in the same archipelago anywhere in Polynesia.4 We now turn to the contribution archaeology can make to the dating of linguistic splits. The following principles are suggested as the most economical way of relating certain archaeological and linguistic facts for a given island group. Principle 4. For a given island group, the foundation language can be equated with the foundation culture in the archaeological sequence, provided that the latter is well established and widely distributed throughout the group. Principle 5. If the archaeological sequence in a remote island group is continuous, i.e., if one tradition has not replaced another at some stage, then the foundation language is ancestral to the present language or subgroup spoken in that region.

Methodologies: implementing the phylogenetic model

35

In sum, there are logical reasons for expecting that the language of a remote island or island group will in most cases be descended from the languages spoken by the ®rst society to establish a sizable population there. Yet, even with these principles, more realistic models than those conveyed solely by classic ``family tree'' diagrams are essential to the joint archaeolinguistic enterprise. Linguistic models of divergence Historical connections among related languages are typically conveyed by the graphical device of the ``family tree,'' or ``dendrogram.'' The linguistic splits thus depicted are usually taken to imply that a high degree of isolation (or, conversely, a low level of continuing contact and lexical borrowing) obtained between each of the branches. Such claims usually have a strong geographical and distributional underpinning (cf. Terrell 1986:248±49, ®gs. 87, 88). But, as Terrell (1986:247) ± reiterating Biggs (1972) ± observed, scholars working in the Paci®c have repeatedly warned against converting such family tree subgroupings into ``simple A to B to C models'' of prehistory. Ross (1997:213) likewise comments, ``although the model is isomorphic enough with a wide-angle view of linguistic prehistory, it can be inimical to more narrowly focused research.'' Green (1981) documented the use (and misuse) of a narrowly focused branching model, its problems, and the need for an alternative in the search for the putative location of the Polynesian ``homeland.'' As the ``wave theorists'' and dialect geographers of Indo-European and other languages have long been aware (Diebold 1987), linguistic histories are typically more complicated than implied by the family tree model. Such trees are indeed ``often unsatisfactory for making sense of and for representing historical relationships among languages'' (Pawley and Ross 1995:51). Thus, Austronesian comparativists have given increased attention to the problems that ancient dialect chains present in determining subgrouping models, and in lexical reconstruction (Pawley and Ross 1993:437). Pawley and Ross cite numerous instances where ``in early Oceanic, dialect differentiation and network-breaking were the rule rather than the exception'' (1995:51). For example, it would be quite inappropriate to transform a typical branching family tree diagram of the closely related Polynesian, Fijian, and Rotuman languages into a simple sequence of settlement steps commencing in island A as the homeland of the founding proto-language, proceeding on to island B, and thence to C, D, and so on. Pawley and Green (1984) addressed this problem directly by setting out two models within which the linguistic evidence for Oceania may be framed:

36

The phylogenetic model: theory and method

In work on Austronesian subgrouping since the 1950s there has been a tendency to conceive of subgroups as being formed by what we will call the radiation model. This model posits an initial period of uni®ed development undergone by a localized, homogenous language community, followed by a period of geographic expansion, leading to the creation of dispersed, isolated daughter communities which develop independently from the time of dispersal. (1984:138)

However, as more detailed evidence accumulated, it became apparent within many Austronesian-speaking regions that other conditions have frequently obtained, and that linguists had ± especially for Remote Oceania ± underestimated the capacity of the early Austronesians to spread quickly over a vast area, and to maintain a fairly uni®ed speech tradition across a network of local communities dispersed across an archipelago ± a unity that may last for many centuries, even millenia [sic], before there is a decisive divergence of local dialects. In some cases it is clear that the location of the ancestral language was approximately equal to the area now occupied by all of its daughter languages. That is to say, it is not necessary for there to be a geographic expansion for the proto-language to break up. A gradual weakening of ties between the network of sister dialects suf®ces: eventually, sharp language boundaries appear. Call this the network-breaking model. (Pawley and Green 1984:138±39, emphasis added)

A network model is especially relevant for ascertaining the probable linguistic correlates of: (1) the Lapita cultural complex which spread rapidly through much of Remote Oceania and was ancestral to Polynesian culture (Pawley 1981; Green 1997b); (2) the Ancestral Polynesian stage centered in the western Polynesian archipelagos (see Chapter 3); and (3) the early Eastern Polynesian stage centered in the Cook±Society±Marquesas region (Kirch 1986a; Green 1988). The formation and break-up of proto-languages in a networking model assumes a ¯ow of lexical and grammatical innovations across the dispersed local communities. What one is really studying are situations of dialect geography that require a different approach, as Geraghty (1983) amply demonstrates in the Fijian and pre-Polynesian case. While the spread of innovations will never be entirely constant or always congruent throughout the network, as long as they continue to be fairly even, the entire dialect chain will continue to change as a unity or partial unity. In these circumstances, there is no single point in time that can be equated with the breakup of the proto-language. Instead, there is a period during which unity declines; when one dialect has ceased to be intelligible to the rest, or has ceased to take part in the ¯ow of innovations, the breakup may be said to be complete. (Pawley and Green 1984:139)

For these reasons, Pawley and Green (1984:139) argued that ``if pressed, most linguists would probably have always conceded that subgroups are

Methodologies: implementing the phylogenetic model

37

sometimes formed in this general manner.'' However, in doing lexical reconstruction it is often more convenient (and hence less ``messy'') to simply assume that each descendent of the proto-language is an independent witness to the parent stage. In 1984, Pawley and Green found that with respect to the Oceanic branch of Austronesian languages, it is necessary ``to employ both the radiation and network-breaking models (for different stages). Both permit the construction of family trees and proto-languages'' (1984:139). In many of the Oceanic cases listed by Pawley and Green (1984: fn. 30), both kinds of models are required to account for different phases of the family tree sequence. Similarly, in the classi®cation of western Oceanic languages, Ross (1988) drew a distinction between the well-marked subgroups that result from complete splits, and those that form imperfect linkages. Such alternative models to the radiation or family tree types in Oceania have now reached considerable sophistication, and differ from the IndoEuropean ``wave model'' which does not seek family trees and is best used in dialect studies (Pawley and Ross 1995:64±65, fn. 9). For the Oceanic languages, we now have innovation-de®ned subgroups based on isolating mechanisms, and innovation-linked subgroups where continuing linkages are to the fore (Pawley and Ross 1995:50±51; Ross 1997:222±25). Pawley (1997:472±73) calls these ``perfect'' and ``imperfect'' subgroups (the latter category including persisting dialect linkages), and has shown that the second category has major application not only for Oceanic, but for the higher-order Malayo-Polynesian level as well. Ross (1997) has formalized and theorized the entire process in a ground-breaking article on social networks and on kinds of speech community events, a paper we recommend to all ``linguistic prehistorians'' (1997:211), those who would practice any form of archaeolinguistics. The current situation in the ®eld can be summed up in two quotes from Ross: It is only in the last twenty or so years that linguistic prehistorians have become more adept at recognizing that there are different kinds of linguistic divergence, and that there are also varieties of linguistic convergence re¯ecting community contact and integration (1997:213, emphasis in original). It is obvious that the social network diagram can not replace the family tree as a means of representation. (1997:254)

In sum, the family tree: (1) functions as the presentation of event sequences and the subgroups that are their outcomes; (2) determines what words are reconstructable in the proto-language; and (3) is crucial to locating the homeland of a proto-language (Ross 1997:254±55). We cannot overstress that both linguistic strategies ± trees as well as networks ± are required; it is not a case of one or the other. In Oceania we have both

38

The phylogenetic model: theory and method

dispersed dialect chains, and localized homogeneous languages and protolanguages, and we ®nd extended periods of continuing contact as well as points of fairly sharp diversi®cation when a population moved to a distant island group creating a sharp break.5 Dispersal centers and homelands Sapir (1916:456±57) discussed some of the principles for inferring ``historical centers of distribution'' for related languages, based on historical evidence for signi®cant linguistic differentiation in certain regions. Dyen (1956) formalized these and related propositions into a set of speci®c procedures. The objective is to infer from the linguistic evidence the most probable dispersal centers or homelands for any given language family, once the main internal relationships within the family have been established. Inferences about movement are based on identi®cation of those regions exhibiting the greatest genetic diversity according to a subgrouping of the languages concerned, rather than simply those displaying the greatest number of daughter languages. Invoking the principle of parsimony, regions that served as primary dispersal centers should also require the fewest subsequent moves for all subgroups and daughter languages to have achieved their contemporary geographic distribution. Lexical reconstructions ± for a given protostage ± of words for the environment and distinctive physical features, as well as for plants and animals of restricted range, can also be explored to test hypotheses of former homelands of a language family or its subgroups (the WoÈrter und Sachen method). However, in our opinion these should not be relied on as a primary means of homeland identi®cation. Rather, as Ross (1997:255) indicates, ``a well-argued tree is . . . crucial to locating the homeland of a proto-language.'' Pawley and Ross (1993:440±42) discuss the historical conclusions drawn from the evidence for the higher-level subgroupings within Austronesian (see Figures 2.1 and 2.2). At the level of Proto Austronesian they start in Taiwan and the northern Philippines, and proceed south and east through several interstages to northern Halmahera and Cenderawasih Bay in New Guinea, a likely homeland for Proto Eastern Malayo-Polynesian. They place the primary dispersal center for the Oceanic subgroups in western Melanesia, speci®cally the Bismarck Archipelago possibly including the northern coast of New Guinea (1993:441). In Chapter 3, using similar procedures, we will consider probable dispersal centers in Remote Oceania for the various eastern Oceanic subgroups, especially Proto Central Paci®c and Proto Polynesian.

Methodologies: implementing the phylogenetic model

Fig. 2.1

39

The higher-level subgrouping of the Austronesian languages, down to the Oceanic level (after Pawley 1996a).

Establishing time depth Linguistic methods for establishing time depth can yield only relative sequences of events, not precise chronology. Only a general sense of age is obtained from the internal structure of family trees; for example, it is clear that the closely related Polynesian languages have differentiated more recently than those of most other Oceanic or other higher-level subgroups of Austronesian. Likewise, Polynesian ties to the various western Austronesian subgroups (e.g., the Formosan languages of Taiwan) where the languages are far more differentiated, are even more remote in time. However, precise chronological estimates for dispersals from a series of centers by linguistic methods alone are not possible, as the Austronesian case indicates. Grace (1964) originally proposed a time frame of roughly 1500 BC to AD 500 for the movement of Malayo-Polynesian speakers into Oceania. Drawing on lexicostatistical analysis and referring to the great diversity among Austronesian languages of western Island Melanesia, Dyen (1971) suggested that Oceanic speakers of Austronesian had been present in that region for at least 5,000±6,000 years.6 By attempting to correlate the linguistic evidence from Remote Oceania (and especially Fiji and Polynesia) with the archaeological evidence, Pawley and Green (1973:53) reduced

Fig. 2.2

The geographic distribution of higher-level subgroups in the Austronesian phylum.

Methodologies: implementing the phylogenetic model

41

Dyen's estimate to 4,000±5,000 years, and subsequently on further consideration (Pawley and Green 1984) to c. 3,600±4,000 years. In Chapter 3, we will revisit in greater detail the evidence for dating the Proto Central Paci®c subgroup of the Oceanic languages through its correlation with radiocarbon-dated archaeological assemblages of the Eastern Lapita complex, found in the Fiji±Western Polynesian region. Our view is that key intersections between an independently dated archaeological record and speci®c internal subgroups within a language family can yield convincing estimates of time depth for a sequence of proto-languages. As stressed earlier, while lexicostatistics and glottochronology might be consulted as independent lines of evidence, these are marred by inherent methodological ¯aws, and cannot be relied on as the primary means for calculating time depth. Terminology and units of analysis It is essential to de®ne unambiguously a few key terminological units that we employ, terms suggested previously (Kirch and Green 1987:434, 451). Drawing upon the ®eld of historical linguistics, we use the well-established convention of a proto-language to refer to each separately reconstructed node or stage in a linguistic sequence, whether based on a radiation or on a networkbreaking model. In this study, we focus primarily on the Proto Central Paci®c (PCP) and Proto Polynesian (PPN) stages, and to a lesser extent the Proto Nuclear Polynesian (PNP) and Proto Tongic (PTO) stages which followed upon the breakup of PPN. The period of accumulating language change following a previous node, and leading up to the next, is usually described as the pre-stage (i.e., pre-Polynesian or pre-Central Paci®c). To avoid any confusion between reconstructions which derive from separate sets of data, we strongly inveigh against attempts to use these linguistic terms simultaneously as cultural or biological designations. Thus we employ the tripartite terminological set of: proto-languages, parental populations, and ancestral cultures (and societies) to clarify this distinction. At any given time in the past, a parental population of interbreeding individuals, organized into one or more social units, spoke a common proto-language, and shared an ancestral culture. Having been remonstrated with by several colleagues over the label ``Ancestral Polynesian Society,'' we emphasize our view that archaeological units must ± in a phylogenetic approach ± be on a level commensurate with those used by historical linguists (i.e., languages, dialects, dialect chains, and communalects), and by biological anthropologists (i.e., inter-breeding or potentially inter-breeding populations). As we wrote in our earlier paper: Polynesian culture and society will not do. Rather what we are talking about are Samoan, Tongan, and other ancestral societies, and not ancestral Polynesian society,

42

The phylogenetic model: theory and method

Lapita, or any other societal or cultural entity in the general sense. (Kirch and Green 1987:451, emphasis added)

One way of addressing this issue is to use existing archaeological designations ± Lapita cultural complex (and its regional variants such as Eastern, Western, Far Western, or Southern Lapita; see Kirch 1997:69±74), and Ancestral Polynesian culture ± as the temporally successive cover terms for these two broadly based archaeological entities, each of which is known to exhibit variation from island group to island group. It is then possible to speak of Ancestral Polynesian societies in the plural, recognizing the variation that obtained from site to site, between localities, and within and between regions, as well as to more accurately re¯ect the small size of the social and political entities and therefore settlement units involved. In short, this focuses discussion at the community and island or island group level. The triangulation method and its application to the phylogenetic model We seek to develop a triangulation method in which the subdisciplines of historical linguistics, archaeology, comparative ethnology, and biological anthropology independently contribute their data and assessments to the common objective of historical reconstruction. We derive the label from a surveying metaphor, which should be immediately understood by most ®eld archaeologists at least.7 In the classic method of survey by triangulation, sightings are taken from two or more points along a known baseline to an unlocated point which one wishes to ®x in space. As these sightings begin to converge on that point, a ``triangle or polygon of error'' is de®ned, within which the real point lies. So it is with our proposed triangulation method. Our ``sight-lines'' are those provided by the independent evidence of historical linguistics, archaeology, comparative ethnography, biological anthropology, or even oral traditions. As these converge and cross-check each other, the target of our sightings ± some aspect of the historical record ± comes increasingly into focus, and the ``polygon of error'' decreases in size. Of course, our focus on the historical ``reality'' may never be crystal clear, but, as in surveying, triangulation is always preferable to estimating the position of a point from backsights taken from a single station. While Sapir (1916) distinguished ``direct'' historical evidence (i.e., written documents, oral history, and archaeological materials) from ``inferential'' evidence (i.e., data from ethnology, linguistics, and living biological populations), the situation is arguably more complicated. Archaeologists have come to realize that their data too exist in the present, and that interpretations of past events are accomplished through a variety of inferential techniques.

Methodologies: implementing the phylogenetic model

43

The same restrictions apply to the study of ancient skeletal remains (although biological anthropologists also have information derived from living populations). In both cases, once the effects of taphonomy are removed from remains created at some past time ± that time being determinable through a variety of ``absolute'' dating techniques such as radiocarbon ± analysis of such remains does offer a kind of ``direct'' evidence on those events. In contrast, the evidence from ethnology or linguistics permits only a relative sequence of past events to be inferred, and chronological dates must be obtained through correlations with non-ethnographic or linguistic evidence of a historical kind. The triangulation method brings to bear these two distinct classes of evidence, those which are truly diachronic in that they derive directly from past events and can be ``dated'' in real chronological time, and those that permit the construction of relative time sequences through systematic comparison of synchronic, but historically related, evidence.8 The analytical power of the triangulation method ± and the robustness of the historical reconstructions derived from it ± only holds, however, if one treats each data source separately, respecting the relevant subdisciplinary methods, inferences, and conclusions as they are developed independently, based exclusively on the evidence from that ®eld.9 A brief example from Polynesia ± the problem of reconstructing ancestral ®shing strategies (Green 1986) ± illustrates this principle of independence. Based strictly on the evidence of ethnographic observations, one might conclude that simple (one-piece) ®shhooks, consistently absent in museum collections from Western Polynesia, were an Eastern Polynesian innovation (Anell 1955). Indeed, initial archaeological excavations in such Eastern Polynesian islands as Hawai`i and the Marquesas yielded an abundance of shell and bone hooks, and, combined with the absence of such hooks from sites in Tonga or Samoa, seemed to con®rm this hypothesis. However, the independent evidence of historical linguistics indicated that a term *mataqu, `®shhook,' could be reconstructed at the Proto Polynesian stage ancestral to both Western and Eastern Polynesian cultures. Of course, the reconstructed word ± while evidently meaning some kind of ®shhook ± gave no speci®c clues as to form, materials of manufacture, or frequency of use. It was only after more extensive archaeological excavations in Western Polynesia increased our sample sizes, that simple one-piece ®shhooks in Turbo shell were shown to have been present in the Western Polynesian region at an early time depth (Kirch and Dye 1979). These archaeological discoveries were consistent with the Proto Polynesian reconstruction from linguistics, permitting us to infer that one-piece ®shhooks had indeed been part of the Ancestral Polynesian ®shing kit, but had later been abandoned in Western Polynesia. Moreover, the archaeological evidence also indicates that the ubiquitous simple ®sh-

44

The phylogenetic model: theory and method

hooks of Eastern Polynesia (present in both early and late contexts) derive from the ancestral Turbo-shell hook tradition in Western Polynesia (Kirch et al. 1990:11), but for environmental reasons (Allen 1992) came to be made predominantly of pearl shell (Pinctada sp.) in the eastern archipelagos. This case, albeit a simple example, illustrates how alternative historical inferences can be developed on independent lines of evidence, and when taken together and submitted to cross-examination, permit a more rigorous reconstruction. The advantage of the triangulation method is underscored by the philosopher of science Allison Wylie (1989, 1992, 1993, 1999), discussing problems of con®rmation in archaeology. Disavowing either an objectivist or a relativist stance, Wylie notes that employing independent sources of data and analysis in any given reconstructive-evaluative argument ensures that the constituent arguments are not merely mutually reinforcing but also ± and crucially ± mutually constraining. We are in¯uenced ± but not monolithically controlled by ± our current understanding which makes us largely see or understand what our background knowledge and theoretical commitments prepare us to see. For this reason, each subdiscipline initially arrives at a slightly different conclusion regarding a particular historical phenomenon. But we can also assess our data differently when the circumstances change, and we can be forced by the evidence to consider alternative interpretative possibilities to those we have previously entertained (Wylie 1989:16). In short, ``if diverse evidential strands all converge on a given hypothesis ± if you can use different means to triangulate on the same postulated set of conditions or events ± then you may be able to provide it decisive, if never irreversible, support simply because it is so implausible that the convergence should be the result of compensatory error in all the in¯uences establishing its evidential support'' (Wylie 1992:28). In Part II we employ this triangulation method to develop a set of hypotheses about a given ``domain'' in the ®eld of Polynesian culture (such as subsistence, cooking, or social organization). Lexical reconstruction and meaning In their assessment of Austronesian historical linguistics, Pawley and Ross (1993) observe that most lexical work has concerned itself chie¯y with historical phonology, paying little attention to the ®ne grain of semantics. Fortunately, increasing attention is now paid to ``the history of particular semantic ®elds, especially those of interest for cultural history, and to specifying reconstructed meaning more precisely'' (1993:441). These endeavors have focused particularly on lexicons for the Proto Austronesian, Proto Malayo-Polynesian, Proto Oceanic, and Proto Polynesian stages.10 Our concern is with Proto Polynesian (PPN), and our approach follows the

Methodologies: implementing the phylogenetic model

45

practice of proceeding by semantic ®eld or domain, as will be evident in Part II. Different procedures for the reconstruction of meaning have been developed, beginning with Dyen and Aberle's (1974) pioneering work on Athapaskan kinship, and more recently by Dyen (1985), Blust (1987), and Ross, Pawley, and Osmond (1998:4±7). Green (1994) examines these methods from the point of view of an archaeologist concerned with ``culture history'' (what we prefer to call ``historical anthropology''). Unfortunately, there is a high degree of skepticism among Paci®c archaeologists concerning the use of linguistic evidence, partly because of the problems posed by meaning.11 The focus of such archaeological criticism of linguistic culture history is not on the morphological or phonological reconstruction of a given lexical form ± a task most archaeologists are happy to leave to the linguists ± but on the assignment of likely former meanings to these reconstructed etyma, given the often multiple meanings for cognates in the various daughter languages. The problem, naturally, is one of concern both to the archaeologist cum historical anthropologist, and to the historical linguist. Following Blust (1987:81), we make a critical distinction between lexical reconstruction and semantic reconstruction as separate approaches.12 Semantic reconstruction asks the question, ``what was the probable meaning of protomorpheme `X' within a given semantic ®eld,'' while lexical reconstruction asks the question, ``what was the proto-morpheme which probably meant `X.' '' Green, favoring the methodology of semantic reconstruction, observed that ``this method is more likely to provide the kind of former meanings and subsequent changes in them which archaeologists would be inclined to explore using their own data, rather than the limited, rather stark and still currently preserved sameness of meaning resulting from the Dyen and Aberle approach'' (1994:177).13 In Green's view there is a need for wellworked-out versions of what Dyen (1985) called ``semantic history hypotheses,'' or what Diebold (1987:56) refers to as a ``strong diachronic semasiology.'' Historical linguists making lexical reconstructions need to spend more time and effort attending to how they construct their meaning glosses; so do archaeologists who wish to employ their data. Most recently, Ross, Pawley, and Osmond (1998:4±7) have elaborated a methodology for ``terminological reconstruction,'' which is similar to Blust's semantic reconstruction. Essentially, they use semantic history hypotheses informed not only by linguistic evidence, but by comparative ethnography, as we do also in Part II of this book. Their method is as follows: First, the terminologies of present-day speakers of Oceanic languages are used the basis for constructing a hypothesis about the semantic structure of corresponding POc terminology, taking account of (i) ethnographic evidence . and (ii) the geographical and physical resources of particular regions of Oceania .

as a .. ..

46

The phylogenetic model: theory and method

Secondly, a search is made for cognate sets from which forms can be reconstructed to match each meaning in this hypothesized terminology . . . Thirdly, the hypothesized terminology is re-examined to see if it needs modi®cation in light of the reconstructions. (Ross, Pawley, and Osmond 1998:4)

In the ®rst of a projected ®ve-volume set dealing with the lexicon of Proto Oceanic, Ross et al. (eds., 1998) elaborate a series of terminological reconstructions dealing with aspects of material culture, such as architecture, horticulture, ®shing, canoes, and related domains. Careful semantic or terminological reconstruction provides historical anthropologists with strong arguments, particularly when there is what Dyen (1985) terms ``prime semantic agreement'' (PSA), a sameness of meaning (homosemy) ``between cognates belonging to members of different branches'' of a language family or subgroup (1985:358). Examples of stable prime semantic agreement from Proto Polynesian down through other interstages to the various modern daughter languages include PPN *talo, `taro' (Colocasia esculenta), and *toki, `adz/axe/chisel' (Green 1994:178±79). However, PSA applies in only a limited number of cases, and when there is disagreement in the meanings of the cognate terms a ``semantic history hypothesis'' is required to explain how multiple and often quite varied daughter language meanings (among words still judged to form a cognate set) could have developed.14 The issue of meaning must always be kept to the fore when one is doing semantic reconstruction in historical anthropology. The POLLEX project Polynesianists ®nd themselves especially well positioned to apply the triangulation method because ± in addition to having a sophisticated subgrouping model based on the genetic comparative method ± the work of lexical reconstruction for the ancestral PPN interstage is well advanced. Thanks to decades of careful research by Bruce Biggs and his students and colleagues,15 a lexical database called POLLEX (Biggs 1998) has been developed, containing more than 2,300 speci®c PPN reconstructions, a monumental achievement.16 Table 2.1 shows a typical data array for one such PPN term, *waka, `canoe.' The POLLEX ®le lists the re¯exes of *waka for thirty-two modern Polynesian languages plus a few external witnesses (keyed according to standard three-letter abbreviations for language names, see list of language abbreviations (pp. xvi±xvii)), along with brief glosses derived from dictionary or other sources. Since *waka is a case of prime semantic agreement, there is little hesitancy in accepting Biggs' reconstructed PPN gloss of `canoe.' 17 Note also that in this case PPN *waka is itself derived from yet older cognate terms that can be reconstructed for antecedent interstages of Austronesian,

Methodologies: implementing the phylogenetic model

Table 2.1. POLLEX database entry for PPN *waka, `canoe' .AN *0 *4 *41 *5 *6 = *8 *9 *PN* ANU SAAEAS ECE EFU EUV FIJ HAW KAP MAE MAO MFA MQA MQA1 MTA MVA NIU NKO OJA PEN RAP PUK RAR REN ROT SAASAM SIK TAH TAK TIK TOK TON TUA WEV WFU WYA

WAKA.A 5/11/96. POC *wa9ka ``boat'' (Ply. 1973). PCEMP *wa9ka ``canoe'' (Bst. 1993a). PMP *va9ka9 (Dpf ). PAN *wa9ka ``boat'' (Ply. 1973), *ba9kaq ``canoe'' (Rss. 1988). Note. (Bst. 1993a) argues that *wa9ka9 (Dpf ) is a Chinese loanword. PBN *vaka ``ship'' (Lvy. 1979). :Canoe. Vaka. :Canoe (Yen). . Baka. :Boat, vessel. Vaka. :Canoe. Vaka. :Canoe. Vaka. :Canoe. Waqa. :Canoe. Wa`a. :Canoe. Waga. :Canoe (Lbr). Vaka. :Canoe, ship (Clk). Waka. :Canoe. Vaka. :Old word for canoe. Vaka (MQN). :Canoe (Bgs). Va`a. :Canoe. Aka. :Canoe (Cdn). Vaka. :Embarcation, navire, radeau, pirogue (Rch). Vaka. :Canoe. Vaga. :Big ship/Canoe. Va`a. :Canoe. Vaka. :Canoe. Vaka. :Canoe (Stk). Vaka. :Canoe (Bge). Vaka. :Canoe. Baka. :Canoe, vehicle, receptacle (Ebt). Vaka. :Canoe. . VA`A. :Canoe. Vaka. :Canoe. VA`A. :Canoe. Va`a. :Canoe. Vaka. :Canoe (Fth). Vaka. :Canoe. Vaka. :Canoe. Vaka. :Canoe, boat, ship, vessel etc. (Stn). Vaka. :Embarcation, pirogue, bateau (Hmn). Vaka. :Canoe. Waqa. :Canoe.

Note that ``9'' in POLLEX stands for ``ng'' sound.

47

48

The phylogenetic model: theory and method

such as POC and PCEMP (Proto Central-Eastern Malayo-Polynesian), but not for PMP or PAN (Pawley and Pawley 1998:178). These deeper (higherlevel) lexical reconstructions are listed at the top of the data set, although the extensive list of cognates in extra-Polynesian languages from which these pre-PPN reconstructions are derived are not given in POLLEX. Without POLLEX, we would have hesitated to even attempt the sort of historical analysis undertaken in Part II. Nonetheless, the POLLEX project has concentrated primarily on lexical reconstruction, with (until recently) minimal attention to semantic or terminological reconstruction, and no effort to spell out detailed semantic history hypotheses for the many cases in which the varied glosses for the modern cognates do not closely agree in meaning. Taking the lexical items in POLLEX as our starting point, we have paid close attention to semantic reconstruction. Moreover, consistent with our triangulation method, we have done this not by relying on dictionary glosses, but by systematically mining the rich ethnographic corpus for Polynesia. Ethnographic evidence If dictionaries provide less than ideal sources for semantic reconstruction, a major resource for expanding on the meanings encapsulated in words are ethnographies. These, however, do far more than simply ®ll out the domain of meaning covered by a term; they also contextualize the lexicon within the broader systemic cognitive patterns and behaviors of a particular society and its culture. Unlike Vogt (1964), for whom the use of ethnohistorical and ethnographic data formed later steps (7 and 8) of his phylogenetic method, our strategy considers these sources in conjunction with the initial linguistic data (steps 1±4). The outstanding comparative ethnographic sources begin with E. G. Burrows' monograph (1938a), Western Polynesia: A Study in Cultural Differentiation, an explicit analytical adaptation of some of the methods outlined by Sapir (1916). Burrows examined distributions of distinct resemblances between geographically adjacent cultures to infer historical relationships (1938a:7). Differential distributions of a highly selected set of cultural traits were then interpreted within the historical processes of diffusion, local development, and abandonment or rejection (1938a:92). Everything from items of material culture through kinship terms to the ``nights of the moon'' is included in these distributional analyses. Burrows argued that all three kinds of historical processes played their part in the differentiation of Western Polynesian culture from either a central or a central-marginal grouping, together with unspeci®ed responses to a changed natural environment (1938a:153), thus laying the long-standing basis for the distinction between Western and Eastern Polynesia.18 Burrows' 1938 study contrasted markedly

Methodologies: implementing the phylogenetic model

49

with the two- and three-strata migration models previously employed by anthropologists such as E. S. C. Handy (1930). Indeed, Burrows rejected such multiple-strata migrations and concluded with respect to ``the remote period of ®rst settlement'' that a ``fundamental unity of Polynesian culture'' was indicated (1938a:156), and further suggested how that unity might be demonstrated. Subsequent explorations of systematic regularities within Polynesian societies and cultures, such as those of Sahlins (1958) and Goldman (1970), drew upon Burrows' notion of an underlying Polynesian unity. Sahlins highlighted the importance of natural resource distribution and its variations in Polynesia, and the role that a hitherto neglected environmental component had played in the elaboration of social arrangements. In a similar vein, Goldman's exercise in carefully controlled comparisons revealed orderly patterns of social variability in Polynesia (1970:545±49), and principles of status differentiation (1970:551±54) which he regarded as the outcome of tension within the status system (1970:567). The most recent contribution in this comparative genre is Howard and Borofsky's (eds., 1989) Developments in Polynesian Ethnology. The contributors to this volume examine social organization, mana and tapu, chieftainship, art and aesthetics, and other topics within several kinds of framework, all having one or more of three goals: (1) they should aim at illuminating underlying structural patterns shared among Polynesian groups as well as explaining variations on common themes; (2) they should strive to illuminate key variables that have facilitated continuity and change through time; and (3) they should look for similarities and differences between Polynesia and other areas within Oceania and beyond. (Borofsky and Howard 1989a:287)

Within these goals two types of comparisons are delineated (Borofsky and Howard 1989a:287±88): (1) controlled comparisons of particular island groups with similar institutions where convergences (as well as differences) arise from parallel conditions or constraints acting on a commonly inherited structural base; and (2) broader multi-island comparisons of the Burrows/ Sahlins/Goldman kind which explore general patterns within Polynesia as a whole. Borofsky and Howard (1989a:289) also draw attention to the possibilities for broader comparisons between Polynesia and other parts of Oceania. In short, a rich tradition of ethnographic analysis employing comparative frameworks and goals is available to inform a renewed historical anthropology of Polynesia. These works, of course, are not without recognized defects. As one of the more telling critiques suggests, problems with these works ``go deeper than has been previously acknowledged'' so that speci®c

50

The phylogenetic model: theory and method

misinterpretations are consistently generated, even though they cannot be totally written off on grounds of theoretical or ideological contamination (Thomas 1989:79). Thus it is claimed that ``the earlier forms of ethnography have been discredited to an extent quite out of proportion to the actual differences between such texts and modern anthropological writing'' (Thomas 1989:79); this applies particularly to the Polynesian museum ethnographies of the 1920s and 1930s (1989:32±33). In his phylogenetic model, Vogt (1964) called for separating ethnohistory (his step 7) from more recent ethnographic endeavors (his step 8), in order to assess changes during the period of European contact, colonialism, and emergence of traditional societies into the present world system (see Chapter 1). Some of the problems posed by this era of rapid change are covered in Borofsky and Howard's chapter (1989b:241±75) on the early contact period in Polynesia. Discussion of this complex period within Polynesia has reached a new level of sophistication, moving well beyond pre-World War II decades of ``museum ethnographies'' much criticized by Thomas (1989:32±33, 41±49). We now have the ``island-centered'' approaches of the Paci®c historians, with the attempts at careful reconstruction by anthropologists of selected Polynesian cultures or their institutions at the time of initial Western contact (e.g., Oliver 1974; Valeri 1985), as well as a multi-faceted approach to that period from both a European and local perspective of the agents and processes of change involved in the late eighteenth and nineteenth centuries (e.g., Dening 1980; Sahlins 1981, 1985; Hooper and Huntsman 1985; Salmond 1991, 1997; Kirch and Sahlins 1992). If these improved reconstructionist enterprises (Vogt's step 7) are thus combined with the greater depth in content offered by modern ethnographic perspectives (Vogt's step 8), the latter can allow us ``to gain insights into the cultural logics formerly at work in Polynesian societies'' (Borofsky and Howard 1989b:248), thus highlighting some of the enduring structures of history.19 While Thomas (1989:32±33) rightly criticizes the bookshelf of museum ethnographies of Polynesia for what they failed to do, and for their naõÈvete in viewpoint and approach to their data, this corpus remains much valued by archaeologists dealing with indigenous material culture. Just as stark linguistic meanings taken from dictionaries can often be ampli®ed and enhanced through ethnographic inquiry, so also are the methods for production, function, and use of archaeologically recovered objects greatly illuminated by museum collections and descriptions of them in these sources. Nonetheless, as with the reassessments undertaken for ethnographic accounts, a new level of evaluation has been required in the study of museum-based collections. Te Rangi Hiroa certainly distinguished contact and earlier period pieces from those of later manufacture (Hiroa 1944:412±13), but a great deal of work has been required since to

Methodologies: implementing the phylogenetic model

51

distinguish and to authenticate the full range of genuine early items (Kaeppler 1971, 1978b). With better documentation, these museum collections time and again prove their worth not only to archaeologists seeking to understand the manufacture and use of the bits and pieces which they recover through excavation, but also as to how these traditional productions once ®tted within the now-transformed societies. Granted that Te Rangi Hiroa's comparative analysis at the conclusion of Arts and Crafts of the Cook Islands (1944:410±526), with its three-period temporal separation of material culture traits into early, late, and postEuropean contact, has been completely superseded, but the basic descriptive data on the known distribution and the techniques of manufacture, use, and function of these items within Polynesian traditional societies during the early contact period have not. One has only to cite the work of Duff (1956, 1959), especially that on stone adz heads, to see how the approaches of Burrows and Hiroa in¯uenced some of the initial interpretations of archaeologically recovered materials within a prevailing ethnological paradigm that lasted well into the 1960s. Interpretations of archaeological data typically have deep roots within ethnology, as is evident in Polynesia where the transition from one data set to another is virtually seamless. Archaeology and the direct historical approach Several factors favor a direct historical approach for Oceania. The ®rst is strong support, especially in Polynesia, for continuity in dated, well-attested, and connected local chronologies ranging from 1,200 to 3,300 years in length and exhibiting changes over time, but with little or no signs of population replacement. Such continuities, indeed, seem to be a feature of most local sequences in Remote Oceania during the past 3,200±3,300 years (Green 1997b); they also occur, at least in parts of Near Oceania, over the last 2,000 to 6,000 years (Spriggs 1997; Kirch 2000). While there certainly are instances of contact between related societies and their cultures within various subregions of Polynesia (see Chapter 3), evidence for exchanges with either extra-Oceanic or more distantly related cultures (e.g., South American, Fiji, Eastern Micronesia) is fairly circumscribed prior to the eighteenth century. Greater intersocietal contact, and even occasional displacements by distantly related or even unrelated societies and their cultures, is evident for western Remote Oceania, while in Near Oceania the need to consider contacts between totally unrelated societies and cultures (and occasions of replacement of one by another) rises sharply. In sum, a main contribution of archaeological data to an integrated historical anthropology is to reveal local, materially evidenced cultural sequences providing a wide spectrum of information on technological,

52

The phylogenetic model: theory and method

economic, social, political, and even religious practices. These sequences serve as an independent basis for formulating constructions about some period or periods in the past, as well as for testing and cross-checking constructions based on linguistic or ethnographic data sources. Despite its invariably partial nature (see Chapter 7), the archaeological record provides historical anthropologists with the means to convert comparative synchronous, or only relatively sequenced diachronous, constructions of former events and cultural patterns as offered by linguists, ethnographers, and biological anthropologists, into accounts that truly re¯ect deep time.

Chapter 3

Polynesia as a phylogenetic unit

The Polynesian cultures derive from a common source; they are members of a single cultural genus that has ®lled in and adapted to sahlins 1958:ix a variety of local habitats. . . . in many instances race, language, and culture do coincide to establish an authentic ethnic coherence. Polynesia seems to be such goldman 1970:xxiii an instance.

Space, time, and content must be the prime de®nitional axes for any phylogenetic entity in the kind of historical anthropology for which this book is an extended argument. We insist that the issue itself requires multiple perspectives: linguistic, ethnological, biological, and archaeological. Having laid out speci®c methods and procedures by which a phylogenetic model can be promulgated, we now scrutinize the Polynesian case, searching for congruence among independent lines of evidence to test whether Polynesia meets the criteria for a discrete ``segment of cultural history.'' We will demonstrate, ®rst that the linguistic subgrouping of Polynesian languages is closely mirrored by systemic ethnographic patterns across Polynesian societies, and second that there is biological coherence as well among the various Polynesian populations. It is also essential for us to de®ne precisely what is meant by ``Ancestral Polynesian'' culture and societies. This is not some vague notion of a putative cultural stage loosely corresponding to a set of reconstructed Proto Polynesian words. On the contrary, we will identify a discrete group of archaeological sites and their artifactual assemblages, dating to the second half of the ®rst millennium BC, as the physical manifestations of those Ancestral Polynesian societies and communities that ¯ourished in the Tonga±Samoa homeland region prior to the diaspora that ultimately put Polynesians on virtually every habitable island in the vast eastern Paci®c. These archaeological materials allow us to ground the root of the Polynesian phylogenetic tree in time and space. 53

54

The phylogenetic model: theory and method

Polynesia as an emic category The very term Polynesian is, need we say, a European construction which nonetheless has acquired a certain cogency since its de®nition by the French explorer Dumont d'Urville (1832) in the early nineteenth century. Anthropologists have consistently regarded Polynesia as a group of related societies within a geographic area possessing a high degree of homogeneity in language, culture, and human biology. But from an insider perspective, the concept of a region inhabited by people of like culture, speech, and physical appearance is a matter of substantial antiquity. Thus in the Proto Polynesian lexicon itself we ®nd the term *tangata ma(a)qoli, which probably meant something very close to `indigenous person, one of our own kind.'1 And present-day, indigenous populations in various islands groups within Polynesia use re¯exes of *maqoli to demarcate themselves as maaori (New Zealand, Cook Islands), maohi (Tahiti), or maoli (Hawai`i) in opposition to nonindigenous occupants, bringing the concept squarely into the present. In ancient Polynesia, where the *tangata ma(a)qoli were invariably the ®rst human occupants on island after island as these were discovered and settled, it was unnecessary to have a contrastive term, hence the confusing diversity of appellations which were given to the European `invaders' when these appeared from the seventeenth century onwards: papalagi, popa`aa, papa`aa, pakeha, haole, and so on. The obvious af®nities of language, culture, and biology that link the various Polynesian groups have probably always been recognized by the indigenous peoples of this region, as they continue to be today. Moreover, this native perspective was not lost on early European explorers, traders, missionaries, and settlers, and would only be further reinforced by later nineteenth-century investigators such as Abraham Fornander or Percy Smith, and ®nally encoded within the canons of early twentieth-century anthropology. Thus the notion of a fundamental category ± call it Polynesian or call it Maoli ± has from an ethnic point of view substantial and longstanding historical relevance (Green 1987), suggestive of its potential claim to the status of a phylogenetic unit. To forward that claim now requires that we examine the category ``Polynesia'' from the standpoints of linguistics, comparative ethnology, archaeology, and some aspects of human biology. Linguistic perspectives Linguistic relatedness contributes in high degree to attributions of ethnic identity, and is one basis for promulgating historical connections among now-dispersed peoples who speak related languages. Fortunately, the place of

Polynesia as a phylogenetic unit

55

the Polynesian (PN) subgroup of languages within the larger Austronesian (AN) language family is unusually well understood.2 The Oceanic subgroups The Austronesian language family, dispersed geographically from Madagascar to Easter Island and encompassing most of the indigenous languages of the Paci®c island world, includes several high-level subgroups (Blust 1985), of which the most easterly is Oceanic (OC; Pawley and Ross 1995:43). The distribution of these major subgroups within the western part of the Paci®c (see Figure 2.2) ± with the Oceanic subgroup occupying the entire eastern portion ± points to an Austronesian homeland in Island Southeast Asia. On more detailed linguistic evidence, the favored immediate homeland region for Oceanic is the Bismarck Archipelago (speci®cally the Admiralty [Manus] and Mussau Island groups), although the region extending along the north New Guinea coast remains a possibility (Pawley and Ross 1995:57±58; Pawley 1997). Oceanic itself encompasses approximately nine higher-order subgroups (Figure 3.1).3 Five of these lie in Remote Oceania, a region not inhabited until about 3,300 years ago; one subgroup (Southeast Solomonic) is situated at the most eastern end of Near Oceania, occupied for at least 6,000 years and probably much longer (Figure 3.2). Attempts to show that this cluster of six easterly subgroups of Oceanic constituted a single higher-order unit (termed Eastern Oceanic) have not proved convincing (Pawley and Ross 1995:65, fn. 6). Rather, these six subgroups constitute the widely dispersed set of language groups in Oceanic that cannot be included in any of the internally complex high-order subgroups to the west (i.e., in Near Oceania), the region which is thought to be the homeland region for Oceanic itself. The best interpretation of this cluster of six Oceanic subgroups extending eastwards from the southeast Solomons is that their wide distribution over the previously unoccupied island groups of Remote Oceania re¯ects a former dialect chain or linkage. This linkage resulted from an early, but not initial, stage of Oceanic speakers who ®rst expanded into this region from the southeast Solomons, and thereafter differentiated into the ®ve primary subgroups found today in Remote Oceania (Pawley 1981). These became the foundation languages for each of the separated regions, replaced at a much later time only in the case of an intrusive group of non-Austronesian languages in NendoÈ (Santa Cruz) and the Main Reef Islands, and on some of the Polynesian Outliers (Green 1997b). The spread of these Oceanic languages from eastern Near Oceania throughout Remote Oceania has been linked with the initial expansion of the Lapita cultural complex through the Outer Eastern Islands of the Solomons group, to Vanuatu and

56

The phylogenetic model: theory and method

Fig. 3.1

The major subgroups of Oceanic form a ``rake-like'' tree structure.

New Caledonia, between about 3,300 and 3,200 years ago (Pawley and Green 1984; Green 1997b; Kirch 1997a; Spriggs 1997). This colonization proceeded uninterrupted into the Fijian archipelago and to the islands of Western Polynesia (Tonga, Samoa, Futuna, `Uvea). The archaeological assemblages of dentate-stamped Lapita pottery which mark initial settlements throughout Remote Oceania are dated to a fairly narrow time horizon of 200±300 years duration (Kirch 1997a:57±63). Using the principles we set out in Chapter 2, it is possible to link the foundational Oceanic language subgroups with the foundational Lapita cultural assemblages in Remote Oceania.4 These were the cultural complexes upon which all later cultural developments would build. Proto Central Paci®c and the emergence of Proto Polynesian The Central Paci®c (CP) subgroup of Oceanic (see Figure 3.2) includes the languages of Rotuma and Fiji, and all those of Polynesia. The Proto Central Paci®c (PCP) interstage, however, is only weakly attested by lexical innovations (Geraghty 1983, 1996a; Pawley 1996a, 1997), suggesting that it was temporally ephemeral. Proto Central Paci®c consisted of a geographically extensive dialect chain stretching throughout the rapidly settled Fiji±Western Polynesian region. In the terminology of Pawley and Ross (1995:52), Central Paci®c is an innovation-linked subgroup, and language differentiation within that dialect chain did not follow a strict family tree type of model (Geraghty 1983; Pawley 1996a, 1999). Rather, there was a process of network-breaking and reemergence, resulting in successive subgroups whose boundaries

Fig. 3.2

The geographic distribution of major subgroups within the Oceanic branch of Austronesian languages.

58

The phylogenetic model: theory and method

Fig. 3.3

The Proto Central Paci®c dialect chain.

shifted and re-formed through time (Figure 3.3); the details need not detain us here. The soon-to-be Polynesian portions of this PCP dialect chain, at its eastern geographic end (what Geraghty [1983] calls Tokalau±Fijian±Polynesian), accumulated in their pre-Polynesian stage a signi®cant range of innovations not shared with Rotuman or any languages in the Fijian region (Pawley 1996c). These extensive innovations include ®fteen phonological, fourteen morphological, and eight syntactic changes exhibited only by languages in the Polynesian subgroup (Pawley 1996a:392±95). More importantly, the POLLEX lexical ®les contain upwards of 1,300 reconstructed Proto Polynesian (PPN) lexemes lacking cognates outside of Polynesian, or which continued old forms with new semantic innovations. 5 Thus Polynesian (PN) constitutes a well-marked, innovation-de®ned language subgroup. In sum, out of dialectal variation in PCP, one innovation-de®ned subgroup of languages ± called Polynesian ± ®rst differentiated in the eastern part of the region covered by the original PCP dialect chain. Yet this Polynesian subgroup already exhibited some dialectal variation, and as Pawley (1996a:401) remarks, the elegant family trees that are often drawn by historical linguists do not satisfactorily re¯ect the processes of differentiation

Polynesia as a phylogenetic unit

Fig. 3.4

59

North±south dialect differentiation within Proto Polynesian (after Pawley 1996a).

involved. In its pre-PN stage, internal dialectal variation within the subgroup can be detected, with northern and southern dialect clusters (Figure 3.4). These clusters were to become the basis for the two highest-order subgroups within PN itself, the Tongic (TO) and Nuclear Polynesian (NP) subgroups (Pawley 1996a:401±2). Thus linguistic variation was present right from the beginning in the formation of the Polynesian subgroup. What is normally reconstructed as PPN, therefore, represents the point at which this subgroup ®nally broke apart; this would also have been at the approximate time that parts of central Eastern Polynesia and certain Polynesian Outliers6 to the

60

The phylogenetic model: theory and method

west were initially settled.7 It is to this late stage of PPN that the approximately 2,300 lexical reconstructions in POLLEX, assembled by Bruce Biggs, belong. Internal classi®cation of Polynesian The long-standing internal genetic classi®cation of the Polynesian languages (Elbert 1953; Green 1966; Pawley 1966), usually graphed as a family tree (e.g. Clark 1979), has recently undergone some modi®cations (Figure 3.5).8 The Eastern Polynesian branch remains more-or-less intact, with only minor alterations at the lower levels (Marck 1996c), and has been considerably strengthened by additional evidence. However, modeling Proto CentralEastern Polynesian (PCE) and its dissolution requires application of a dialect chain-network breaking model (similar to that discussed above for PCP; see Green 1988; Pawley 1996a:403; Marck 1999b). These revisions are informed by evidence of grammatical (especially pronominal) innovations, and for geographical distributions (isoglosses) of irregular changes in some lexemes (Wilson 1985; Marck 1999b). This evidence allows us to revise signi®cantly: (1) how the numerous Outlier languages group (or do not group) under the Nuclear Polynesian umbrella; (2) how the Proto Ellicean (PEC, Tuvalu) subgroup includes certain Outlier languages, as well as Tuvalu and Tokelauan under one sub-branch, and Samoa under another; and, (3) how Eastern Polynesian joins these two as a third sub-subgroup category. The implications of these revisions are multiple (Marck 1999b, in press), but most importantly they demonstrate that all the higher-order subgroups within Polynesian ± and thus true genetic diversity, as opposed to simple language diversity ± lie in the Western Polynesian region, and unequivocally mark that zone as the PPN homeland. This conclusion has long been supported by environmental lexical evidence. Words such as PPN *malau, `megapode, incubator bird,' and the archaeologically recovered bones for one extant and several extinct megapode species in the Fiji±Tonga±Samoa region but not elsewhere in Polynesia, provide one example (Clark 1982; Steadman 1997:74; see Chapter 4). Likewise the PPN word *palolo for the marine seaworm, present only in the central islands of Western Polynesia, is telling among a long list of such lexical items (Pawley and K. Green 1971:21±23). However, such lexical environmental evidence is hardly necessary to identify the Proto Polynesian homeland, de¯ecting any critique from those who regard WoÈrter und Sachen methods as questionable (e.g., Renfrew 1987; see Mallory 1997). Far more important for Polynesian are arguments that depend on language distribution, delimiting the area of greatest genetic diversity (Sapir 1916), and invoking the principle of parsimony that requires the

Polynesia as a phylogenetic unit

Fig. 3.5

61

A ``family tree'' type classi®cation of the Polynesian languages (modi®ed after Marck 1999a).

fewest and shortest moves necessary to account for the geographic distribution of the higher-order Polynesian subgroups (Dyen 1956). In short, language admirably identi®es Polynesia as a discrete phylogenetic unit. Moreover, it isolates that part of the vast region now covered by its daughter languages (the well-known Polynesian ``Triangle,'' plus the

62

The phylogenetic model: theory and method

Fig. 3.6

Islands in the Fiji±Western Polynesian region linked by voyaging circles of 24 hours or less (after Marck 1999a).

``Outliers'' in Melanesia and Micronesia) in which the initial homeland lay, and shows how early dialect variation within PPN became the basis for later, high-level subgroup emergence within the homeland zone. The process of language differentiation does not require an A to B to C sequence of moves from one island group to another, as Biggs (1972) cautioned against (see also Green 1981). Differentiation in the early stages of Polynesian was not strictly the outcome of isolation, but occurred within a zone over which interaction was continuous for some 3,000 years (Green 1975, 1996), a zone in which an overnight voyage permitted few real isolates (Marck 1999a:13, map 1.3), yet encouraged dialectal differentiation into discrete subgroups, then languages, over several millennia (Figure 3.6).9 Ethnological perspectives Cultural regions in Oceania No consensus exists regarding major cultural areas in the Paci®c. In anthropology, many scholars have de®ned Oceania as encompassing the

Polynesia as a phylogenetic unit

63

nineteenth-century categories of Melanesia, Micronesia, and Polynesia. Others have included Australia within the Oceania framework (e.g., Oliver 1989). But it has become clear that such categories are of limited value for historical analyses. Green (1991, 1994), in trying to overcome some of these problems, identi®ed a vast region in the western Paci®c ®rst settled by humans c. 50,000±40,000 years ago, which he termed Ancient Near Oceania.10 Archaeological evidence suggests that by about 6,000 years ago ± still some millennia before people began to expand into the farther reaches of the central Paci®c ± the Ancient Near Oceanic zone might have been divided into three distinct cultural entities: Island Southeast Asia (Bellwood 1997), Australia (Lourandos 1997), and modern Near Oceania (Kirch 1997a; Spriggs 1997). Each of these units, which are de®nable by 6000 BP, was to have a somewhat separate history thereafter (Figure 3.7). The large island of New Guinea, along with the immediately adjacent Bismarck Archipelago and Solomon Islands, is usually regarded as a linked but environmentally divergent part of ``Sahul'' or ``Greater Australia'' (White and O'Connell 1982; Allen 1993, 1996). ``Island Melanesia'' has recently been accorded its own separate archaeological analysis (Spriggs 1997). Delineating possible phylogenetic units within such broad zones as these will be challenging, although the Comparative Austronesian Project has shown the value of language as an index.11 Two possible units for application of a phylogenetic approach, as indexed by language, are Oceanic (see Kirch 1997a for discussion), and the Trans-New Guinea Phylum which occurs within New Guinea (Pawley 1995; Ross 1995b). Echoing the views of Bellwood, Fox, and Tryon (1995:3±4) that phylogenetic models have considerable analytic power for Austronesian studies as a whole, we are con®dent that historical analyses informed in part on such language relationships, as well as other kinds of indices, will emerge in time. 12 Whatever analytical units may develop through future research, these surely will not be the overworked and outdated categories of ``Melanesia'' and ``Micronesia.'' From our knowledge of the histories and diversity of these two regions ± whether in linguistic, biological, or cultural terms ± there can be no doubt that these units are fatally ¯awed (see Kirch 2000). We employ instead the distinction between Near Oceania and Remote Oceania (Fig. 3.7). Polynesia, the phylogenetic unit of interest to us, geographically occupies a large part of the Remote Oceanic zone, and it is the only one of Dumont D'Urville's original tripartite classi®cation of Oceanic peoples to have survived the test of anthropological scrutiny. Based on an increasingly sharp pattern of radiocarbon dates, human expansion into Remote Oceania began approximately 3,300±3,200 years ago, proceeding as a rapid series of colonization events (Irwin 1992, 1997; Kirch 1997a, 2000; Green 1997c). Initial expansion into the southwestern

Fig. 3.7 The Paci®c region with Near Oceania, Remote Oceania, and the Andesite Line and ``continental'' type islands indicated (after Green 1994).

Polynesia as a phylogenetic unit

65

part of Remote Oceania encompassed the Reef/Santa Cruz, Vanuatu, Loyalty, New Caledonia, Fiji, Tonga, and Samoa archipelagos. Occupation of Nuclear Micronesia and of Eastern Polynesia did not occur until somewhat later. The reason for this pause in the geographical advancement of people, its duration, and its explanation are all hotly debated issues among Paci®c prehistorians.13 Within Remote Oceania, we can de®ne four regions with separate and internally intertwined historical trajectories, and these exhibit striking differences ethnographically: (1) Nuclear (sometimes referred to as Central/ Eastern) Micronesia; (2) western Remote Oceania (encompassing the Outer Eastern Islands of the Solomons, Vanuatu, Loyalties, and New Caledonia); (3) the Fijian archipelago; and (4) Polynesia. In our view, each of these regions potentially constitutes a valid phylogenetic unit. However, our concern in this book is only with Polynesia. Within Remote Oceania, Polynesia is perhaps the most viable, ethnographically well-attested unit, comprised of related societies possessing numerous cultural regularities, many of which were surely inherited from the ancestral culture. Many of the speci®c systematic patterns that mark Polynesia off as a distinctive unit will be discussed in detail in Part II. In the following section we will support our claim that at their ethnographic endpoints (i.e., from the time of effective European contact to the present), the various Polynesian societies may be grouped together as a phylogenetic unit. Systemic cultural patterns that de®ne Polynesia We agree with Shore (1989:164) when he writes that ``no coherent vision of local variation in Polynesia is possible without a prior clari®cation of what common characteristics make it a real culture area.'' Certain ethnographic domains more explicitly set Polynesia off from both Island Melanesia and Fiji than others. We will canvas just a few of these, seeing no point in attempting to be exhaustive. In the domains of food and food production, Polynesia does not particularly stand out from regions to the west, except perhaps in its attention to the breadfruit, something also found in Micronesia (Yen 1971, 1973; Oliver 1989:188). When one turns to domestic architecture and domiciliary arrangements, however, Polynesia exhibits a contrasting set of patterns from Fiji, or from those of societies in Island Melanesia. Thus Oliver (1989:339±41) treats Fiji's nucleated settlement pattern and its distinctive house forms, men's clubhouses, and a temple-like god house as quite different from the Micronesian or Polynesian cases (1989:337±38). Oliver paraphrases Green (1970; see also Green 1986) in describing some regularities in physical layout of tropical Polynesian communities:

66

The phylogenetic model: theory and method

Throughout Polynesia the most common pattern consisted of a dispersed or loosely clustered neighborhood of relatively small households each located on its own kinconnected land and all focused, socially and religiously, on a common complex of edi®ces that usually included, or was close to, the household of the community's chief. In most places the household buildings housed groups of extended-family composition. They consisted of one or more sleeping houses, a cookhouse, a canoe shed (if located near the coast), occasionally a separate hut for menstruating women, and occasionally a family god house or some other type of religious shrine. In most Polynesian societies the household compounds of chie¯y persons tended to be larger, mainly because of the larger number of persons in residence, but it was only in the more highly strati®ed societies (Tonga, Society, Hawaii) that the chie¯y domestic buildings were superior in terms of workmanship or elaborateness. (Oliver 1989:348)

One could add other public buildings for secular assemblies, and sleeping houses for bachelors and/or young people, in contrast to the strongly gender-restricted men's houses of Fiji or Island Melanesia. Sea-craft and ocean travel in the Paci®c have been ethnographically surveyed many times (e.g., Hornell 1936; Haddon 1937; Haddon and Hornell 1938; Doran 1974; Oliver 1989:361±422), and the topic has been recently studied from the viewpoints of historical linguistics (Pawley and Pawley 1994, 1998) and of prehistoric Paci®c colonization strategies (Irwin 1992; Finney 1994, 1996). As Doran (1974) shows, a distinctive Polynesian sea-craft complex, overlapping in some of its elements with Fiji, may be mapped from the available ethnographic sources. We reproduce Doran's map here as Figure 3.8. External (long-distance) exchange systems throughout Island Melanesia have long been seen by ethnographers as contrasting with the lack of similar networks in Polynesia. Of Fiji, Oliver (1989:577) writes: ``it has separate exchange institutions of both Melanesian and Polynesian types, along with some that contained a blend of both.'' Polynesia exhibits several noteworthy contrasts, and Oliver lists some reasons for this (1989:563±65). For one, there are much greater traveling distances in Polynesia, that not only would have minimized the quantities of cargo able to be transported, but which also would offer fewer possibilities for the development of distinctly localized production, and hence a ``goods-focused'' exchange. The exceptions are in the Tonga±Samoa-Fiji area (Kaeppler 1978a; Kirch 1984a:217±42), and in the Tahitian±Tuamotuan sphere (Oliver 1989:565±66, 1,177±81). Classic ethnographies of Polynesian societies contain relatively little information about inter-community circulation of goods other than in chie¯y tribute or in exchanges between chie¯y persons as part of marriage (Oliver 1989:564±65). Archaeology, using advanced methods of sourcing unevenly distributed resources such as ®ne-grained basalt, is now beginning to

Polynesia as a phylogenetic unit

Fig. 3.8

67

Canoe regions of the Paci®c (after Doran 1974).

demonstrate that there may have been more extensive or frequent interisland exchanges at deeper time levels in Polynesian prehistory (Weisler and Kirch 1996; Weisler 1997; 1998). With regard to internal as opposed to external exchange,14 however, ethnographic data indicates that the former is of widespread signi®cance within Polynesia, and hence a systemic pattern: ``Formalized exchange is an essential part of social life in Polynesia and operates at every level of society, from the domestic to the apically political'' (Howard and Kirkpatrick 1989:84). In short, the ethnographic differences between the exchange systems of Polynesia and Melanesia lie in how they were structured ± in the one case largely internally, and in the other externally ± and the ways in which contrast in emphasis in the two regions based on this distinction has been developed.15 To several generations of Oceanic scholars, the chief/big man distinction was paradigmatic of the differences between Polynesia and Melanesia. Marcus (1989:178±80) revisited the concept of chieftainship in Polynesia,

Fig. 3.9

The geographic distribution of sibling classi®cation types in Oceania (after Marshall 1984: ®g. 7).

Polynesia as a phylogenetic unit

69

looking in particular at what he calls its ``kingly'' and ``populist'' sides, which effectively collapses the chief/big man distinction used by Sahlins (1963). Marcus suggests ``not only that chiefs who share much in common with Melanesian big men are to be found in Polynesia, but that big men who share much in common with chiefs, and in fact are chiefs, are to be found in Melanesia'' (1989:180). Yet the Polynesian boundary is maintained through a concept of chieftainship rooted in the institution of kingship throughout the region, and if Sahlins has its analysis correct, one that is of ``a distinctive pan-Polynesian form'' (Marcus 1989:181). In short, Polynesian chieftainship has a recurring set of characteristics that contrast with those of the chie¯y institutions of Fiji (Oliver 1989:1,174±76), and with those occurring on occasion in some societies to the west. A wide range of other social institutions exhibiting ethnographically attested patterns held in common have also been identi®ed as characteristically Polynesian. Thus the Fijian method of classifying cognatic kin, and the resulting kin types, are sometimes described as ``Dravidian'' (Oliver 1989:1,167±68), and are quite distinct from those of Polynesia, which are generally classi®ed as ``Hawaiian'' following Morgan (1871; see also Murdock 1949). The Polynesian sibling term pattern is characteristic of what Marshall (1984) called ``Type 10,'' though a few societies in Western Polynesia and some Outliers were of his Types 3 and 4. In contrast, the close kin terminologies of Fiji and most Island Melanesian societies were of Type 6, with a scattering of other types dominated by Type 3 (Figure 3.9). In this respect, Polynesian kinship is largely of generational type (Marck 1996b), organized on two principles ± seniority and gender duality ± the one dominant in Eastern Polynesia and the other in Western Polynesia (Howard and Kirkpatrick 1989:65). This results in kinship groupings and descent units that are more varied than might be expected, and which do not permit of ready generalities. Once regarded as ``almost invariably based on common descent from an ancestor in the male line'' (Howard and Kirkpatrick 1989:52±4), since the 1950s anthropologists have regarded Polynesian social groups as based on a nonunilineal principle.16 Oliver (1989:938, 1,028) indicates that in only three Polynesian societies was descent-unit af®liation normatively unilateral; in all others a person was entitled to af®liate with the descent unit of either parent, or of both. This contrasts with Melanesian societies, in which the principal kinds of descent units were typically unilineal (1989:1,028). Oliver (1989:935, 937) summarizes Polynesian descent groups in terms of three general features: (1) most widely identi®ed with descent units, corporately, was land; (2) the collective activity they engaged in most widely was religious; and, (3) their structure in relation to kin was pyramidal. Thus he sees Polynesian societies as ``all very much alike in some fundamental

70

The phylogenetic model: theory and method

features of social structure, especially those having to do with kinship'' (1989:883). Howard and Kirkpatrick provide an even more sweeping view, following their review of the recent literature on Polynesian descent groups: The cultural perspective argues that Polynesians carried with them a set of principles for interpreting the world and organizing their social lives. From this standpoint Polynesian social formations are expressions, under a variety of historical and ecological conditions, of a basic world view that includes speci®c notions about kinship, relationships between human beings and ancestral gods, and a host of related beliefs. (1989:59)

Shore (1989) explores the maintenance of such a basic world view among Polynesian societies under the widely known concepts of mana and tapu, advancing eleven propositions of systemic patterns holding throughout Polynesia in respect to these categories, and demonstrating how even in concepts more widely shared throughout the Oceanic world, Polynesians are distinguished by their speci®c interpretations in practice. Other kinds of regularities and systemic patterns could be cited, such as Goldman's (1955, 1970) attention to status rivalry and status lineages in Polynesia, or widespread practices in child rearing and adoption (Borofsky and Howard 1989a:288). All of these systemic patterns make it abundantly evident that, from the ethnological perspective, Polynesia quali®es as a ``segment of cultural history'' eminently suited for study as a phylogenetic unit. Cultural differentiation within Polynesia While systemic cultural patterns help to de®ne Polynesia as a discrete and robust phyletic unit, an ethnographic perspective also provides evidence for internal differentiation within the Polynesian ``cultural clade.'' That is to say, ethnographers have long recognized that cultural traits among the various Polynesian societies are not randomly distributed, but show regularities of clustering. Thus the ethnographically attested societies of Western Polynesia (Tonga, Niue, Samoa, Futuna, `Uvea) share a number of cultural patterns ± such as elaborate kava ceremonial, the fahu privilege of the sister's son, or the pasting method of bark cloth joining ± which are not found in Polynesian societies outside of this core region. On the other hand, many of the societies to the east, north, or south of Western Polynesia (collectively referred to as the Eastern Polynesian societies) similarly share traits or patterns not found in Tonga, Niue, Samoa, Futuna, or `Uvea. Examples include the pantheon of ®rst-order anthropomorphic gods (especially Tu, Tane, and Rongo; see Marck 1996a), the architectural elaboration of walled or platform ritual structures (marae), or the method of bark cloth retting.

Polynesia as a phylogenetic unit

71

These intra-Polynesian geographic associations or clusters of traits began to be formally recognized as one outcome of the systematic ethnographic survey of Polynesia initiated by the Bernice P. Bishop Museum in 1920, through the Bayard Dominick Expeditions and successive ®eld studies (Kirch 2000:20±24). Ralph Linton (1923), for example, used a wide array of material culture traits from six Polynesian cultures (Tonga, Samoa, Society Islands, Marquesas, Hawai`i, and New Zealand) to infer a set of culturehistorical relationships; Handy (1930a, 1930b) did much the same based on an expanded trait list that included social organization and religion. These early attempts, however, suffered from a total lack of archaeological perspective and proper time depth, from a strongly diffusionist theoretical perspective, and from undue in¯uence of the then-current racial classi®cations of the somatologists. (Linton, for example, was convinced that the clustering of material traits would have to be explained as the outcome of three distinct migrations of ``negroid, Caucasic, and Indonesian races'' [1923:465].) Interestingly, however, Linton's data were later used by Driver and Kroeber (1932:220±25) in a pioneering quantitative study to assess statistically the validity of inferred cultural relationships. Their analysis con®rmed that the Western Polynesian cultures (in this case represented by Tonga and Samoa) grouped most closely with each other, while the Eastern Polynesian cultures' similarity displayed the highest statistical af®nities with each other. Burrows (1938a) formally extended and re®ned these early efforts, comparing the distribution of cultural traits within Polynesia as a whole, and leading him to de®ne Western Polynesia as a distinctive subregion, marked by a constellation of cultural traits that set it off from other subregions (which he named Intermediate, Central, and Marginal). Table 3.1 provides a selection of these traits, adapted from Burrows (1938a:88±90). The range of traits includes aspects not only of material culture, but also of kinship, religion, and calendrics. What was particularly path-breaking about Burrows' approach, however, was his rejection of multiple migrations as an explanatory device; rather, Burrows emphasized the role of internal processes of cultural change, not only ``diffusion'' (what we would now call ``horizontal transmission''), but also local development and abandonment or rejection of traits. Burrows' (1938a) delineation of a culturally based major internal division within Polynesia ± the Western Polynesian/Eastern Polynesian distinction ± continues to be well supported (Howard and Kirkpatrick 1989:69; Marcus 1989:179; Shore 1989:165). But since Burrows' pioneering effort we have learned that although systemic patterns with great time depth in Polynesia may be expected to appear in both Western and Eastern Polynesia (and in the Outliers to the west), in the core region of Western Polynesia some

72

The phylogenetic model: theory and method

Table 3.1. Cultural traits distinguishing Western and Eastern Polynesian regions (adapted from Burrows 1938a) Cultural trait

Western Polynesia

Eastern Polynesia

One-piece ®shhooks Ruvettus hooks Bonito hooks Stone or wood food pounders Bark cloth retting Bark cloth joining Bark cloth watermarking Bark cloth decoration Right-angle plaiting Coiled basketry Twining in kilts Stone adzes Canoe hulls Canoe planks Outrigger attachment Canoe sail Carved human ®gures Dart Wooden slit-gong Drum Formal kava ceremonial Chief 's ``language'' Brother±sister avoidance Vasu (fahu) privilege of sister's son Tu, Tane, Rongo as major gods Myth of origins of humankind Nights of the moon Name of ancestral homeland or underworld

0 0 proximal projection 0 0 pasting 0 tablet rubbing + + 0 tangless low ends ¯ange lashing indirect Oceanic lanteen 0 composite + 0 + + + + 0 evolutionary 0

+ + distal projection + + felting + stamping 0 0 + tanged upturned ends right-through lashing direct Oceanic spritsail + simple 0 + 0 0 0 0 + procreative +

Pulotu

Hawaiki

ancient patterns may have been lost or obscured; it will require the techniques of historical linguistics and of archaeology to fully recover them. Thus a strictly ethnographically based attempt to determine which systemic patterns are truly homologous as opposed to convergences (analogous), or the mechanical application of the age-area hypothesis, is not suf®cient for a rigorous historical anthropology. This caution is not meant as grounds for dismissing these approaches as of little account in historical work, for they are often useful starting places. Only when cross-checked with the independent evidence of archaeology and historical linguistics, however, can the

Polynesia as a phylogenetic unit

73

hypotheses generated by comparative ethnography be advanced with con®dence to the status of historical narrative explanation. Polynesia as a biological unit The immediate as well as remote biological origins of the Polynesians have been the subject of much debate (see Howells 1979; Terrell 1986; Green 1989a; Houghton 1996). We aver that paleo-populations associated with the Lapita cultural complex in Near and Remote Oceania gave rise to the founding parental Polynesian and Fijian populations from which all later populations in that region derive (Houghton 1989, 1996; Pietrusewsky 1989; Serjeantson and Hill 1989; Kelly 1996; Clark and Kelly 1993; Martinson 1996; Kirch 1997a; Lum and Cann 1998; Lum et al. 1998). At the time of initial Lapita settlement of Fiji±Western Polynesia, the founding or parental population constituted what could be called (after Howells 1979) a ``prePolynesian'' phenotype. The phenotypic differentiation toward ``Polynesian'' was a slow evolutionary process.17 From morphological evidence of skeletal remains dated to between 2200 and 1700 BP, the Lapita-descended people of Watom (Bismarcks), Natunuku, Sigatoka, Waya (Fiji), and several later populations of Polynesia were still much alike (Visser 1994:195±219, 248±49). Thus the biological separation so evident today was accomplished by a fair degree of secular change in one direction within Fiji during the past 1,500 years, presumably with the input from the west of new genetic material into that archipelago (Visser 1994:249), and change in another direction in Tonga (Van Dijk 1993) and in other Eastern Polynesian islands. As a result of these later changes, Polynesian populations came to be marked off phenotypically from those elsewhere in the Paci®c. We cannot neglect the role of strong genetic bottlenecks, ®rst in coastal Papua New Guinea (Redd et al. 1995:611), and later in Fiji±Western Polynesia (Flint et al. 1989; Martinson et al. 1993; Harding and Clegg 1996:591, 593). Recent genetic evidence strongly indicates that the ancestors of the Polynesians passed through a constricted demographic bottleneck in which a parental group of small size (and one showing the genetic traces of initially having resided in a malarious region [Kelly 1990; Clark and Kelly 1993; Martinson 1996]) served as its founding colony.18 Together with the well-known founder effect (Dobzahnsky 1963), this combined to make Polynesians relatively homogeneous, a biologically more uni®ed set of populations. The bottleneck evidence has a second important implication, that the parental populations were of extremely small size, perhaps less than 100 individuals19 at the onset of established settlements in the Fiji±Western Polynesian region. One result was a largely homogeneous set of late Polynesian populations in a physical and genetic sense, now perhaps best

74

The phylogenetic model: theory and method

typi®ed by the phenotype exhibited by people of Eastern Polynesian descent. Thus major later changes within Polynesia from this small founding parental base follow the general pattern of an east/west split, with three isolates (Easter Island, New Zealand, Hawai`i) attached to the central Eastern Polynesian cluster (Matisoo-Smith 1990), and with Fiji separated off from the Western Polynesian cluster (or on some traits grouping with populations even further to the west). Because these biological processes and outcomes do not directly affect how one reconstructs the Ancestral Polynesian societies and their culture, we will not further consider strictly biological aspects of the phylogenetic approach (as in Vogt's step 6, see Chapter 1). But the small number of people involved as a founding population in the Fiji±Western Polynesian region ± strongly supported by the biological evidence ± does help to condition our views of population size in these societies. Beyond that, we will merely comment that a variety of multivariate studies of Polynesian populations, using anthroposcopic traits as well as metric and non-metric traits of crania, yield dendrograms in which the Western Polynesian populations of Samoa and Tonga cluster together (with Fiji joining them in some analyses), while the Eastern Polynesian populations also tend to cluster in various arrangements (Pietrusewsky 1970, 1971, 1996; see Howells 1979, ®g. 11.3). An especially clear example is the dendrogram of relationships deriving from Pietrusewsky's analysis of thirty-eight non-metric cranial traits (1996: ®g. 1), in which Tonga±Samoa link closely with Fiji, forming one major clade, while the Eastern Polynesian groups of Hawai`i, Marquesas, Easter Island, Chatham Islands, Society Islands, New Zealand, and Tuamotu link together to form a second, distinct clade (Figure 3.10). Summarizing his most recent thinking, Pietrusewsky writes that ``although samples remain small for Western Polynesia, differentiation between Western and Eastern Polynesia is suggested'' (1996:351). Thus whatever continued biological evolution there has been among Polynesian populations over the past 2,500 years, this has been conditioned by the relative isolation between Western and Eastern Polynesian subregions. Archaeological perspectives The key contribution of archaeology, in the process of de®ning a phylogenetic unit, is to offer temporal and spatial (geographic) controls that cannot be provided by either linguistic or ethnographic data. For example, while lexicostatistics and its derivative, glottochronology, offer only the crudest approximations of time depth, archaeology through its use of radiocarbon and other ``absolute'' dating methods can ®x the temporal dimension of a phylogenetic unit fairly precisely. Archaeology also furnishes evidence to

Polynesia as a phylogenetic unit

Fig. 3.10

75

Relationships among Polynesian biological populations as indicated by distance analysis of thirty-eight non-metric cranial traits (after Pietrusewsky 1996: ®g. 1).

de®ne the spatial boundaries of a phylogenetic unit which have been tentatively sketched from linguistic and ethnographic distributions, and may more accurately de®ne the probable homeland region for the unit. Once a phylogenetic unit has been de®ned, and its branching structure worked out, archaeology also contributes substantially to the history of the individual cultural trajectories represented by those branches. Fixing Ancestral Polynesia in time and space Archaeological evidence de®nes a boundary between Fiji and Polynesia, discernible in material culture, beginning around 2500±2200 BP, which continued to be maintained thereafter (Green 1981). Direct continuity in the archaeological sequences and developmental trajectories within Western and Eastern Polynesian archipelagos and islands convinces most archaeologists that there are strong ties between speci®c foundation cultures and their ethnohistoric endpoints throughout Polynesia (for further details see Kirch 1984a; Kirch and Green 1987). There is no indication of cultural replacements within island groups, and certainly none of major intrusions from outside of the Polynesian phylogenetic unit itself. This is not to deny evidence for occasional, if signi®cant, extra-regional contacts with Fiji, Eastern Micronesia, and even South America;20 nonetheless, there can be little doubt that archaeologically one is dealing with a discrete phylogenetic unit.

Fig. 3.11

A graphic representation of the ``density'' of available archaeological information for major Polynesian cultural sequences (modi®ed after Kirch and Green 1987: ®g. 4).

Polynesia as a phylogenetic unit

77

The archaeological record of three millennia-long continuous sequences in Western Polynesia, compared with sequences ranging from about 2,000±1,000 years duration in Eastern Polynesia and the Outliers (Figure 3.11), supports the view that the Ancestral Polynesian homeland was situated in Western Polynesia. In reviewing the Polynesian homeland issue, Green (1981) concluded: (1) that this homeland had to encompass a region made up of several island groups including Tonga, Samoa, `Uvea, and Futuna, rather than being centered on a single island or group (e.g., Tonga) as some had argued (Groube 1971; Pawley and K. Green 1971); (2) that a model for the differentiation of cultural (and linguistic) entities throughout the homeland region had to incorporate continuing contact and interaction among the societies involved, rather than proceeding strictly by isolation; and (3) that demarcation between the historical trajectories of the Fijian societies to the west and the Polynesian societies to the east was not well attested in the archaeological record until around the latter part of the ®rst millennium BC. Green (1987) also argued that there was no a priori reason that initial biological, linguistic, and cultural differentiation among Ancestral Polynesian societies had proceeded as strictly contemporaneous events or processes, as long as within a reasonable period of time all three began to cohere. That is, the emergence of a parental population, proto-language, and ancestral culture into what became a foundation phylogenetic unit ± prior to its expansion over the larger region it eventually occupied ethnographically ± need not have been a synchronous event, but rather a process of historical differentiation. Indeed, attempting to get too close a chronological ``®t'' between the different biological, linguistic, and cultural ®elds is probably an unwarranted expectation, and not even a realistic occurrence historically. The small founding biological populations discussed in the previous section were associated with the ®rst cultural (and archaeological) assemblages distributed throughout the Fiji±Western Polynesian region. Referred to as Early Eastern Lapita (Kirch 1997a:73), this regional variant of the larger Lapita cultural complex is marked by ceramic assemblages characterized by particular vessel forms, decorated in a distinctive subset of the dentate-stamped Lapita design system (Mead et al. 1975). Recent evidence from Ha`apai (Burley 1998), combined with data from Fiji, Niuatoputapu, Futuna, and Samoa, now indicates that the period of dentate-stamped ceramics in the Early Eastern Lapita style is con®ned to a narrow time span, beginning not earlier than about 1100±1000 BC and ending by 800±700 BC (Kirch 1997a:66±69; Burley 1998). The archaeological evidence suggests that these colonizers were socially organized into small-scale communities (see Chapter 8). The earliest archaeological assemblages, and their associated radiocarbon ages, give us a ®rm temporal ®x on the initial occupation of the Fiji±Western

78

The phylogenetic model: theory and method

Polynesian region. It is precisely this ``horizon-like'' portion of the Eastern Lapita cultural complex21 that we would associate with the founding Proto Central Paci®c (PCP) innovation-linked language subgroup (see above). Linguistically, we visualize this entity to have constituted a kind of rapidly dispersing dialect chain. As the chain lengthened through the rapid and progressive colonization of islands to the east, the main variation in dialects occurred between its western and eastern (Tokalau±Fijian/Polynesian) portions. Some dialect differences seem to have rapidly emerged between speakers of communalects in the northern as opposed to southern zones in the eastern portion of the PCP chain (i.e., along its Tonga±Samoa alignment). Given direct archaeological evidence for such north±south variation in another communication medium ± that of the Lapita decorative design system as expressed on pottery (Kirch 1988:187±88, ®g. 114) ± the possibility of corresponding lexical variation is noteworthy. As Pawley (1996a:401±2) suggests, such pre-Polynesian dialectal differences probably laid the bases for the highest-order internal splitting within the Polynesian subgroup of languages, just as early cultural variation between the ceramic design systems of northern and southern groups re¯ects a parallel process, one which is the basis for the later distinctions drawn between Samoan and Tongan ``cultural provinces.''22 If the eastern portion of the PCP dialect chain became increasingly marked as an innovation-de®ned Polynesian subgroup, how long did this process of differentiation take? Pawley's (1996a:395, 400) best estimate from linguistic information (especially his admittedly shaky use of glottochronological computations) is that 400 years is possible but improbably rapid, whereas a more reasonable estimate is at least 800 years, and conceivably longer. To allow only 400±500 years would in Pawley's view suppose a rate of language innovation accumulation and lexical change ``probably unparalleled in the subsequent history of any of the 30 individual Polynesian languages'' (1996a:400). Pawley's construction ± the most sophisticated presently available ± accommodates the dialect chain concerns adumbrated by Dyen (1981:97±99) and Rensch (1987; see also Pawley 1996a:406, fn. 17, 18). It sees Polynesian (a subgroup of languages already exhibiting dialectal variation at the pre-Polynesian stage) emerging during the mid-®rst millennium BC as it simultaneously split into two ®rst-order subgroupings: Tongic and Nuclear Polynesian (see Figure 3.5). It is likewise in the mid to later ®rst millennium BC that archaeologists are able to demonstrate the development of a distinctive Polynesian adz kit within Western Polynesia, and of a type of pottery termed Polynesian Plainware (for details, see Chapter 7). More importantly, Kirch (1981, 1988) and Sand (1990, 1992), using ceramic assemblages from Futuna and `Uvea, as well as from Tonga and Samoa, demonstrate island-speci®c variation in

Polynesia as a phylogenetic unit

79

pottery. Thus, archaeologists are able to de®ne a material form of early Polynesian culture which emerged toward the middle of the ®rst millennium BC, ®ve to seven centuries after initial Lapita colonization. The correspondence between linguistic and archaeological evidence here is essential: both occur in the same set of islands and during the same temporal interval. Archaeology and historical linguistics converge on the later part of the ®rst millennium BC as the formative period for Ancestral Polynesian societies and their culture, and on the chain of islands stretching from Tonga to Samoa but also incorporating Futuna and `Uvea as their geographic location. The breakup of Ancestral Polynesia and subsequent dispersals The end of the formative stage in Polynesia was marked by a series of events that portend ± from the linguistic perspective ± the break-up of Proto Nuclear Polynesian (PNP), and thus the endpoint of Proto Polynesian (PPN) as Biggs (1971) employs the term. These events are beginning to be recognized archaeologically through settlement sequences for various island groups lying beyond the boundaries of the core Western Polynesian area. Speci®cally, these were the discovery and colonization of the most northerly atolls in Western Polynesia (the Tuvalu and Tokelau groups), of some of the Polynesian Outliers further west, and of the central islands of Eastern Polynesia. While an ``absolute'' radiocarbon chronology for these events is still in its early stages of development, and the subject of some debate (e.g., Kirch 1986a; Spriggs and Anderson 1993; Kirch and Ellison 1994), we would bracket these events to the period dating from about 2200±1900 BP. This makes them consistent with Pawley's (1996a:399) glottochronological estimate of about 50 BC to AD 100 (with a scatter of median dates from 550 BC to AD 400). While admitting the controversial dating of the initial settlement of central Eastern Polynesia, we are convinced that this process must have started by 1600 BP, the date now minimally attested for early human activity in the Society Islands and Mangaia (Ellison 1994; Kirch and Ellison 1994; Lepofsky et al. 1996).23 Whatever the date of ®rst settlement in central Eastern Polynesia eventually proves to have been (i.e., for the area occupied by the Cook, Austral, Society, Tuamotu, and Marquesas archipelagos), there is ®rm evidence that all of these islands were well populated by AD 800. Moreover, recent advances in the sourcing of basalt adzes and other kinds of artifacts (Weisler 1997, 1998) demonstrate that central Eastern Polynesia constituted a complex network of communities linked by frequent inter-island and interarchipelago exchanges. This archaeological evidence for an early interaction network linking the central Eastern Polynesian islands meshes well with the

80

The phylogenetic model: theory and method

linguistic evidence for an innovation-based Proto Central-Eastern Polynesian (PCE) speech community, in all likelihood a dialect chain or linkage. It was presumably within and across such a linkage that a number of phonological, syntactic, lexical, and semantic innovations arose, as well as various cultural developments (``synapomorphies,'' in cladistic terminology) that are uniquely shared by the modern Eastern Polynesian cultures descended from this interstage in the Polynesian phylogenetic ``tree.'' The farthest-¯ung and most remote of the Eastern Polynesian islands and archipelagos (those termed ``Marginal Polynesia'' by Burrows [1938a]) lie beyond the central Eastern Polynesian core, including Hawai`i to the north, Rapa Nui to the southeast, and New Zealand and the Chathams to the southwest. There is substantial archaeological as well as paleoecological evidence con®rming Hawaiian settlement no later than AD 800, and quite possibly as early as AD 300±500 (Kirch 1985; Athens 1997). The immediate source of the colonizing population in Hawai`i is likely to have been the Southern Marquesas, but continued contact between Hawai`i and islands in the core region is indicated by linguistic evidence (lexical borrowings from the Tahitic subgroup), abundant oral traditions (Cachola-Abad 1993), botanical indications, uniquely shared mtDNA sequences in populations of the Paci®c Rat (Matisoo-Smith et al. 1998), and possibly some archaeological style changes as well. However, long-distance voyaging between Hawai`i and the central Eastern Polynesian core became less frequent after about AD 1200, and was little more than a memory encoded in Hawaiian oral traditions by the time of European contact. Rapa Nui seems to have been settled as early or earlier than Hawai`i, as re¯ected in its conservative language which retains phonological and lexical forms lost in other Eastern Polynesian dialects (Green 1966, 1988, 1998a).24 Archaeologically, the earliest period on Rapa Nui is imperfectly attested, although excavations at Anakena (Steadman et al. 1994) con®rm occupation by AD 900, and indications of human disturbances to the island's vegetation in the lake cores are dated even earlier, c. AD 700±800 (Flenley 1996). It is doubtful that Rapa Nui was ever connected with the central Eastern Polynesian core area by regular two-way voyages, although the possibility of a limited number of post-colonization contacts should not be ruled out (Green 1998a). Finally, current archaeological evidence holds that the vast temperate islands of New Zealand were the last landfalls within Eastern Polynesia to be discovered, explored, and eventually colonized, around AD 1000±1200 (Anderson 1991; Sutton 1994). The degree of continued, two-way voyaging contact between New Zealand and the central Eastern Polynesian core region is uncertain, but seems unlikely to have been extensive. As with Hawai`i, New Zealand Maori societies had become wholly isolated from the rest of Polynesia by the time of European arrival.

Polynesia as a phylogenetic unit

81

Without belaboring the details of the Eastern Polynesian archaeological record, the overall sequence of island colonizations just summarized provides an independent model of the most recent stages in the differentiation of the Polynesian cultures, a process that involved migration and expansion into previously uninhabited lands and proceeded in a series of stages lasting perhaps a millennium. Isolation by distance became a signi®cant factor in this process of differentiation with respect to the most marginal islands (Hawai`i, Rapa Nui, and New Zealand), but played a lesser role within the central Eastern Polynesian core, where early interaction spheres are being indicated by archaeological evidence (Weisler 1998). In this core area, interaction over complex networks allowed for the spread of numerous lexical as well as cultural innovations which, however, did not spread westwards as far as the Tonga±Samoa region, and thus contributed to the marked differentiation between Western and Eastern Polynesian subgroups. In short, in their general outlines the archaeological and the linguistic models for the ®nal stages of phylogenetic differentiation in Eastern Polynesia are remarkably concordant. Ancestral Polynesian sites and assemblages Having established a temporal and geographic framework, we can nominate speci®c sites and archaeological assemblages relevant to the construction of the cultural content of Ancestral Polynesian societies and their culture. Hence we respond, in part, to the criticism of Sutton (1996:377±78) that in our earlier writings on Ancestral Polynesian Society we had failed to de®ne ``APS'' in discrete archaeological terms. In Table 3.2, we list numerous sites and assemblages dating to the time period de®ned above for the emergence of a distinctive innovation-linked Polynesian language group, and which demonstrate material culture innovations (in ceramics and adzes, particularly) that mark these assemblages off from those of earlier time periods, as well as from later sites (see Figure 3.12 for geographic locations of sites). The selection of sites and assemblages listed in Table 3.2 is conditioned not solely on the radiocarbon evidence, but also on internal evidence of a consistent (if variable) set of artifacts. The most important of these artifacts are the ceramics known as Polynesian Plainware (see Chapter 7). It is from these thirty-odd sites and assemblages that we adduce archaeological evidence for the reconstruction of certain cultural domains within Ancestral Polynesia, in the analytical chapters of Part II. Many of the sites listed in Table 3.2 are known only by their ceramic assemblages; for others, there is additional and useful evidence. In large part, this re¯ects the excavation strategies that archaeologists have used in Western Polynesia, at least for sites of this time period. Many sites have been

82

The phylogenetic model: theory and method

Table 3.2. Selected archaeological sites and assemblages associated with the Ancestral Polynesian period Ceramic phase representedb

References

Open midden 2500±2400 Open midden 2400±2200 Numerous open ceramic scatters

LEL LEL

Groube 1971 Poulsen 1967, 1987

LEL to PPW

Spennemann 1989

Coastal midden 2750±2550

EEL to PPW

Coastal midden 2600±2500

PPW

Dye 1987a; Shutler et al. 1994; Burley 1998 Burley 1998

Coastal midden 2600±2500

EEL (?) to PPW Shutler et al. 1994; Burley 1998 EEL to PPW Shutler et al. 1994; Burley 1998 EEL to PPW Dye 1987a; Shutler et al. 1994; Burley 1998 EEL to PPW Burley 1998

Island and site

Site type

Tongatapu Vuki's Mound To.6 Tongatapu: general Ha`apai Group Tongoleleka (Lifuka Is.) Holopeka (Lifuka Is.) Faleloa (Foa Is.)

Age range (BP)a

Pukotala Coastal midden 2750±2500 (Ha`ano Is.) Vaipuna (`Uiha Is.) Coastal midden 2750±2650 Mele Havea (Ha`afeva Is.)

Coastal midden 2750±2500

Vava`u Group Falevai (Kapa Is.) Open site (TO-Va-19, -20) Pangaimotu Open site (`Utungake Is.)

±

PPW

Davidson 1971; Kirch, ®eldnotes 1976 Davidson 1971; Kirch, ®eldnotes 1976

±

PPW

Niuatoputapu Lolokoka (NT-90) Open midden Lotoa (NT-100) Open midden Pome`e (NT-93) Open midden

3200±1800 2800±2000 2500±2000

EEL to PPW LEL to PPW PPW

Kirch 1988 Kirch 1988 Kirch 1988

Tafahi Fatuloa

Open site

±

PPW

Dye, in Kirch 1988

Futuna and Alo® Tavai (FU-11) Asipani (SI-001) Mamalua (AL-09) Alo®tai (AF-34B)

Buried midden Buried midden Open site Coastal midden

2100 2200±2000 ± 2350

LEL LEL to PPW (?) PPW PPW (?)

Kirch 1981 Sand 1990 Kirch 1975 Sand 1990

`Uvea Utufua Atuvalu Utuleve

Open site Open site Open site

± ± ±

PPW PPW EEL to PPW

Kirch 1975 Kirch 1975 Frimigacci et al. 1984

Polynesia as a phylogenetic unit Samoa Sasoa`a (SU-Sa-3) Inland site

1900±1800

83 PPW

Vailele (SU-Va-1) Base level under 1900 mound site Vailele (SU-Va-4) Base level under mound site Potusa (SM17±1) Coastal midden 1800

PPW

Falemoa (SM17±2) Faleasi`u

Coastal midden

PPW

Basal layers of mound Buried midden on now in®lled bay Buried coastal midden Buried coastal midden

PPW

`Aoa Valley (Tutuila) Ofu Is., To`aga (AS-13±1) Ta`u Is., Si`ufaga (AS-11±51)

PPW PPW

2900±?

LEL to PPW

3100±2000

LEL to PPW

2300

PPW

Green and Davidson 1974 Green and Davidson 1969 Green and Davidson 1969 Jennings and Holmer 1980 Jennings and Holmer 1980 Jennings and Holmer 1980 Clark and Michlovic 1996 Kirch and Hunt, eds., 1993 Hunt and Kirch 1988

a

Ranges given are approximations in calibrated years BP, based on available radiocarbon age determinations. b Abbreviations: EEL, Early Eastern Lapita; LEL, Late Eastern Lapita; PPW, Polynesian Plainware.

de®ned only by limited test excavations. Where formal sampling strategies have been applied (e.g., Kirch 1988 on Niuatoputapu; Kirch and Hunt, eds., 1993 at To`aga), these have had geomorphological or other objectives in mind, rather than extensive areal exposure of former living surfaces. We currently lack ± for any site of this critical time period ± extensive horizontal excavation and exposure of deposits, such as have been undertaken for some later sites. Because of these limitations, the archaeological record for Ancestral Polynesia is not as extensive as we would like, or as we imagine might be obtained from future ®eldwork using different excavation strategies. We are cognizant of the problems of sampling error, and of differential survivability of items of material culture, and do not minimize these here (see Chapter 7). Rather, we hope that the current limitations on archaeological evidence will stimulate archaeologists to turn ®ner-grained attention to sites in the age range bearing on Ancestral Polynesian culture. Isolation, interaction, and phylogeny We cannot end our consideration of Polynesia as a phylogenetic unit without discussing the linked issues of isolation and interaction, especially given that

Fig. 3.12

Locations of key archaeological sites dating to the Ancestral Polynesian phase.

Polynesia as a phylogenetic unit

85

recent critiques have pointed to inter-island and inter-archipelago interaction as posing fundamental or even insurmountable problems for a phylogenetic approach (Terrell et al. 1997; Hunt et al. 1998). We have already shown that sudden and absolute isolation is unlikely to have been the case during the early stages of human colonization in the Fiji±Tonga±Samoa region. Instead of an A ? B ? C sequence of colonization with discrete breaks, the formation of linked communities and dialect chains is more plausible, and a gradual process of network-breaking accounts for the formation of dialectal and cultural differentiation within the Polynesian homeland. But what of later, and continued, contact or interaction among related Polynesian societies? Would such interaction, combined with ``horizontal trait transmission'' such as lexical borrowing, or the diffusion of technological innovations, lead to a total masking of the phylogenetic patterns of shared ancestry and inheritance? This is the position argued by Hunt et al., that phylogenetic ``trees re¯ect an unknown and unknowable mixture of ancestry and later sharing'' (1998:3; their emphasis). Terrell et al. (1997) opine that Paci®c Island societies have never been ``primitive isolates,'' and that some degree of interaction and communication has always been present between island communities. We concur.25 Indeed, a decade earlier, Kirch (1986b) offered a detailed analysis of Tikopia as a case study of ``inter-island contact in the transformation of an island society.'' Arguing that islands were rarely, if ever, ``closed systems,'' Kirch wrote that ``even in the more geographically remote islands of Eastern Polynesia, the notion that island societies developed in vacuo, as it were, deserves on recent evidence to be seriously questioned'' (1986b:33). The issue is not whether such contact or interaction occurred, but more speci®cally what effects it had on island languages, cultures, and gene pools, and whether interaction would inevitably and inexorably lead to the dissolution of patterns of homologous traits, fatally undermining a phylogenetic model. In this regard, we do not share Terrell et al.'s (1997) pessimism. Ethnohistoric and ethnographic evidence for interaction within the Polynesian region demonstrates that degrees of isolation varied signi®cantly. Despite oral narratives of an earlier period of two-way voyages between Hawai`i and an ancestral land called ``Kahiki,'' for example, the Hawaiians had not been in contact with other Polynesian groups for at least several centuries prior to Cook's arrival in 1778 (Cachola-Abad 1993). Similarly, there is no evidence that the Rapa Nui people had maintained regular communication with other islands after the period of initial settlement by their founding ancestor Hotu Matu`a (MeÂtraux 1940). New Zealand, too, had been isolated for some time prior to European voyaging. Thus the marginal sectors of Eastern Polynesia, set off from the core by open-ocean

86

The phylogenetic model: theory and method

distances ranging from 2,000 to 4,000 km, were unquestionably isolated, and had not been in regular contact with the Eastern Polynesian homeland for many centuries at least. The situation was different in the tropical core, where inter-island distances were typically in the range of 200±600 km. Two major spheres of regular inter-island voyaging can be identi®ed in the ethnohistoric record: (1) a formal exchange system linking Tonga with Samoa and other Western Polynesian islands, and with Fiji; and (2) a less formalized system linking the Society Islands, directly or indirectly, with the Tuamotus, Australs, Marquesas, and southern Cooks. The Tongan system is sometimes known as the ``Tongan maritime empire,'' and has been analyzed by Kaeppler (1978a), Kirch (1984a:217±42), and Hage and Harary (1991:16±20). What is crucial, for our purposes, is that the Tongan exchange network was operated by a small number of Tongan elites, and involved the transfer of prestige goods (such as mats and feathers) and the marriage of limited numbers of highranking spouses. The central Eastern Polynesian network, best known perhaps from the famous map dictated by the Tahitian priest-navigator Tupaia to Captain Cook (Dening 1962), is less well documented, but also involved elites and was probably restricted to prestige goods. Neither of these systems involved large numbers of people, or high-frequency movement of goods and materials. We infer that the impacts of such exchange were principally con®ned to the elite sectors of society, where they doubtless did in¯uence political affairs and, at times, religious ideology. That such inter-island contacts did not lead to wholesale horizontal transmission of traits (linguistic or cultural) is patently obvious in that the interacting groups maintained distinctive languages and cultural patterns. Thus, despite limited inter-marriage between high-ranking Tongan males and Samoan chie¯y women, Tongans and Samoans retained their own cultural distinctiveness (as in bark cloth designs, club forms, or monumental architecture, to name just a few). Tongan and Samoan languages, likewise, remained discrete and distinctively separate, even though there was some borrowing between them. Marck (1999a:134) identi®es ninety-nine words in the ``metropolitan Western Polynesian vocabulary which may constitute post Proto Tongic and/or post-Proto Nuclear Polynesian borrowings around Western Polynesia.'' While signi®cant, this hardly constitutes massive linguistic impact. The same can be said for the maintenance of cultural and linguistic differences among the various central Eastern Polynesian cultures. Indeed, various Polynesian groups consciously maintained distinctive cultural and linguistic identities, quite the opposite of willy-nilly borrowing of every new word or thing they heard or saw on voyages to other islands. While the ethnohistoric and ethnographic evidence demonstrates that some degree of isolation did separate various Polynesian groups, and that

Polynesia as a phylogenetic unit

87

interaction both was restricted to certain (elite) segments of society and did not inexorably lead to cultural homogenization, we can go farther. Contrary to the assertion of Hunt et al. (1998) that the ``mixture of ancestry and later sharing'' is ``unknowable,'' historical linguists and archaeologists alike have developed empirical methods for determining historical contacts between groups. In historical linguistics, ``borrowing'' can be detected through the presence of irregular sound correspondences, and sometimes ``doublets,'' resulting in what Biggs termed ``direct and indirect inheritance'' in his classic study of Rotuman (Biggs 1965).26 Given the ethnohistoric evidence for the Tongan ``maritime empire,'' one would predict the presence of Tongan loan words in the languages of other groups with whom the Tongans interacted; this is precisely the case (Clark 1979:264; Biggs 1980; Dye 1980:352; Marck 1999a:137±43). Indeed, with East `Uvea, which was conquered and politically dominated by the Tongans for some centuries, extensive lexical change did occur, although Pawley (1967) could still determine convincingly that `Uvean was fundamentally a Samoic language with a heavy Tongan overlay. The respective contributions of ancestry or shared inheritance could be unambiguously separated from those of culturecontact. Similarly, for Hawai`i, Green (1966; see also Marck 1999a:146±7) was able to point to a small number of Tahitic borrowings in Hawaiian, a language that otherwise groups with the Marquesic branch of Eastern Polynesian. Again, shared retentions can be distinguished from the effects of later contact. With the recent development of archaeometric techniques for sourcing prehistoric artifacts, and especially the application of x-ray ¯uorescence (XRF) techniques to the sourcing of basalt adzes in Polynesia (Weisler and Woodhead 1995; Weisler and Kirch 1996; Weisler, ed., 1997; Weisler 1998), archaeologists are now able to empirically trace and de®ne ancient spheres of interaction. Weisler (1998) has recently shown that basalt adzes from Eiao Island in the Marquesas moved as far as Mo`orea and Mangareva, in agreement with ethnohistoric predictions of a central Eastern Polynesian exchange network. Likewise, Best et al. (1992) demonstrated that adzes produced at the Tatagamatau quarry on Tutuila in Samoa radiated outwards to Tuvalu and Tokelau, as well as to Tonga, Fiji, and even the eastern Solomons. The archaeological sourcing evidence, still in its early stages of development, will in time provide an invaluable data array for de®ning prehistoric interactions within Polynesia. It is essential to keep in mind, however, that the transport of a handful of stone artifacts from one island or archipelago to another implies exactly the kinds of relatively low-frequency, elite-centered voyaging demonstrated by the ethnohistoric record. Moreover, the emerging archaeological evidence supports a model of initial integration of some

88

The phylogenetic model: theory and method

island populations into an interaction sphere, followed by later isolation. This is the case, for example, with some of the Southern Cook Islands, such as Mangaia and Ma`uke, where importing basalt and pearl shell occurred in the prehistoric sequences, followed by a long period of non-importation and exclusive use of local resources (Kirch et al. 1995; Walter 1998; Sheppard et al. 1997). Indeed, Mangaian oral traditions (Hiroa 1934) speak to several attempts by groups from Rarotonga, Aitutaki, and Atiu to invade Mangaia, and the repulsion of such groups by Mangaian warriors. Yet a third body of independent evidence bearing on the question of interaction versus isolation in prehistoric Polynesia comes from recent studies of biological variation. With regard to mtDNA variation in human populations, Lum and Cann (1998:116) ®nd ``signi®cant correlations between genetic and linguistic distances,'' which they interpret as ``evidence of isolation between populations.'' The one area in the Paci®c where they ®nd evidence of extensive gene ¯ow, central Micronesia, ®ts the ethnographic pattern there of extensive two-way voyaging (see also Lum 1998). Referring to Terrell et al.'s (1997) critique of the so-called ``myth of the primitive isolate,'' Lum and Cann tellingly urge ``caution against the adoption of panmictic alternatives'' (1998:109). Additional biological evidence comes from the recent analysis of variation in the mtDNA of the Paci®c Rat (Rattus exulans), a commensal species transported by Polynesians from island to island (Matisoo-Smith et al. 1998). Mitochondrial DNA phylogenies of R. exulans populations from Polynesian islands provide evidence of both isolation and interaction. For example, ``the close relationships between the R. exulans sequences of the Cook and Society Islands suggest a broad central east Polynesian interaction sphere encompassing the Southern Cook and Society Islands'' (Matisoo-Smith et al. 1998:15,146), agreeing with what we know from ethnohistory, linguistics, and archaeological sourcing studies. On the other hand, the rat mtDNA data also suggest signi®cant isolation for the Chatham Islands and the Marquesas. And for Hawai`i, links with both the Marquesas and the Society±Cook Islands are indicated, again independently con®rming the evidence of historical linguistics. In sum, the attack on the validity of a phylogenetic model ± on the grounds that in the absence of total isolation, massive interaction overwhelmed patterns of shared ancestry ± proves to be thoroughly unconvincing, indeed anthropologically naive. While regular patterns of interaction certainly characterized parts of Polynesia, other more geographically remote sectors were signi®cantly isolated. Where interaction was ongoing and is ethnographically described, it involved small numbers of elites. Furthermore, the claim that the effects of interaction are ``unknowable'' is uninformed, for linguists, archaeologists, and biological anthropologists have an arsenal of

Polynesia as a phylogenetic unit

89

empirical methods to track interaction. Correspondence between the results of their independent analyses is striking. In the last analysis, we need only point to the kinds of deep divisions within Polynesia, so well analyzed by Burrows (1938a; see Table 3.1), to realize that interaction did not produce a homogenized culture. The major and pervasive distinctions between Western and Eastern Polynesia are suf®cient to differentiate two major clades or branches within the Polynesian phylogeny. Polynesian societies were never ``primitive isolates,'' nor were they a panmictic meÂlange. Their history, as de®ned by shared ancestry, by innovation, and by interaction, is both real and knowable. What is essential is good theory, and rigorous method. Phylogenetic differentiation in Polynesia: a summary We now brie¯y restate the essential points emerging from our analysis of Polynesia as a phylogenetic unit distinct from other such units within Remote Oceania. (1) On linguistic, ethnographic, biological, and archaeological criteria, the Polynesian cultures and societies comprise a distinct phylogenetic unit set apart from the rest of Remote Oceania, a consequence of their shared history. (2) The formative or Ancestral Polynesian stage, at the root or base of the Polynesian phylogenetic tree, can be geographically situated within the region including Tonga, Samoa, and its near neighbors such as Futuna and `Uvea (but excluding Fiji), which constitute its immediate homeland. (3) It was in this homeland region that Ancestral Polynesian culture and societies emerged out of an immediately preceding Early Eastern Lapita precursor, the date of such emergence being the mid to later ®rst millennium BC. (4) Initial dialectal as well as cultural differentiation within the Ancestral Polynesian region occurred not as a series of discrete branches as in a classic dendritic model, but rather as an overlapping series of innovations over a linkage, network, or interaction sphere encompassing many islands and at least several dozen individual social communities. (5) A further stage of linguistic, cultural, and biological differentiation within Polynesia, beginning roughly 2000 BP, was accelerated through the expansion of Polynesian populations into previously uninhabited islands and archipelagos outside of the core Ancestral Polynesian homeland. This included the discovery and eventual colonization of islands both in Eastern Polynesia, and to the west (the Polynesian Outliers). (6) Once the central Eastern Polynesian core archipelagos had been settled, there was considerable isolation between these and the original Ancestral Polynesian homeland region. Much like the Tonga±Samoa interaction sphere, the central Eastern Polynesian islands constituted a second, complex, linked network among which various new innovations spread, leading to marked cultural as well as

90

The phylogenetic model: theory and method

linguistic differentiation between Western and Eastern subgroups of Polynesia. (7) The settlement of the most geographically remote outposts of Eastern Polynesia, which were only infrequently or minimally linked to the central Eastern Polynesian core, led to additional ``branches'' of the Polynesian ``cultural tree,'' which through time and as a consequence of increasing isolation also developed their own distinctive cultural patterns. (8) Finally, within the core of central Eastern Polynesia, regular inter-island communication began to decline after about AD 1400 and, although this did not completely cease, encouraged social isolation and some degree of cultural differentiation. By the time of regular European intrusion into Polynesia beginning in the late eighteenth century, the island groups of central Eastern Polynesia, as well as those of Western Polynesia and the Outliers, were each distinguished from the other by myriad local variations in speech, dress, material culture and architectural styles, details of sociopolitical organization, and so forth. A ®nal note on method We have repeatedly pointed to various lines of evidence ± not just linguistic, but ethnographic, archaeological, and biological ± indicating not only that Polynesia as a whole constitutes a robust phylogenetic unit, but that at least two internal subdivisions are also well marked: the Western and Eastern Polynesian regions. This brings us to one additional methodological detail: the importance of using information from both Western and Eastern Polynesian subgroups when reconstructing the society and culture ancestral to both. Given the traditional family tree for the Polynesian languages (see Figure 3.5), a particular lexeme or word may be considered to have been part of the PPN vocabulary if it is attested by cognates in the two main branches of Polynesian languages: Tongic and Nuclear Polynesian. Or, in the case where such cognates are missing in one of these branches, an external witness from either Fijian or Rotuman (the other members of the Central Paci®c group of languages) is required to demonstrate that the lexeme under question was present at the PPN stage. However, we can see that a potential problem arises in the case of terms that satisfy these requirements, but for which the re¯exes are found exclusively in Western Polynesian languages (e.g., Tongan and Futunan). In such cases, the term may potentially be a later innovation, after the breakup of PPN, which was then borrowed into one or more other Western Polynesian languages, given that the speakers of these languages have been in intermittent contact throughout prehistory. Addressing such problems in the context of a comparative study of Polynesian cosmogonic traditions, Marck proposed two principles of histor-

Polynesia as a phylogenetic unit

91

ical reconstruction, which we will also follow. These principles, as stated by Marck, are: 1. If a feature . . . is universal or occurs in Tonga and Eastern Polynesia the feature will be suggested to have occurred amongst the beliefs of the PPN speakers. (Agreements between Tongan and Samoan will be considered possible borrowings unless there is agreement from Eastern Polynesia as well.) . . . 2. Similarly, if a feature occurs in two widely separated groups not otherwise known to have borrowed from each other, the feature will be suggested to have occurred in the community of speakers of their common proto-language rather than to be borrowed. (1996a:218±19)

As an example of the ®rst kind of feature, Marck gives the cosmogonic ``belief that the sky was close to the earth at the time of creation . . . and that an early act of the early gods was to raise the sky into its present position'' (1996a:219). Because versions of this belief are attested in Tonga and Samoa, and from Eastern Polynesia, Marck regards it as an aspect of the belief system of the PPN speakers. As an example of the second principle, he gives the naming of the male of the Primordial Pair as Papa-adjective in both Samoa and the Marquesas. We will also apply these principles in our work of cultural reconstruction to the Ancestral Polynesian level. Conclusion Based on the conjunctions of independent sets of evidence ± derived from linguistics, comparative ethnology, and archaeology, with more limited input from biological anthropology ± we have de®ned a phylogenetic unit called Polynesia. Within the vast geographic region now occupied by Polynesian societies and populations, we have moreover identi®ed a homeland region situated in the core islands of Western Polynesia, an inference supported by both linguistics and archaeology. Polynesian culture emerged out of the Early Eastern Lapita cultural complex, representing the founding settlement of the Fiji±Western Polynesian area at the beginning of the ®rst millennium BC. A distinctive Polynesian culture is archaeologically recognizable by the mid-®rst millennium BC, and continued no later than the ®rst centuries of the Christian era. After this time it is no longer reasonable to speak of an Ancestral Polynesian culture or societies, for differentiation within the homeland had proceeded suf®ciently for us to distinguish distinctive Tongan and Samoan cultural provinces. By the close of the ®rst millennium BC, an expansion of Polynesian populations had commenced, out of the homeland into central Eastern Polynesia, as well as into the Outliers. Having thus delimited our units in time and space, including the main branches of the Polynesian phylogeny, we proceed to the challenging task of de®ning the cultural content of Ancestral Polynesian societies.

part ii

Rediscovering Hawaiki

Hawaiki, as one of the names of the homeland, has been carried along and applied to various islands in memory of the past, but the original Hawaiki lies buried under the accretions of time.

hiroa 1945:12

If there is a word that can persist in the memory of Polynesians in their centuries of wandering, it would have to be a word such as Hawaiki: a word of origin that tells of greatness and divine ancestry; a word of identity stamped with a glorious past; a word which, whatever the particular cognate and its language-speci®c associations, may be taken to symbolise the journey of Polynesians taumoefolau 1996:406 through time and space.

Introductory remarks

We have chosen the title for Part II of our book ± Rediscovering Hawaiki ± quite consciously. We might have called it ``reconstructing'' Hawaiki, for one aim of the triangulation method within a phylogenetic model is to determine as fully as possible the ancestral culture which stands at the root of any ``segment of cultural history.'' But every label carries its own intellectual baggage, and the baggage of ``reconstruction'' includes an ubiquitous usage during the heyday of the New Archaeology, thus tending to be associated with an uncritical form of positivism. We recognize that history (or prehistory) is ± to a certain degree ± ``constructed'' by its practitioners, although we would strongly assert that what makes anthropological history a science and not an art is that its constructions of the past are self-consciously and continually constrained by the evidence. What is more, a central goal of triangulation is to assure that this array of evidence is as broad as possible, that it is not con®ned to any single line of inquiry, but draws upon the full spectrum of anthropological insight. Thus we prefer rediscovering Hawaiki, a simple phrase that, in our view, more aptly conveys the essence of our enterprise. Hawaiki was, after all, in the indigenous Polynesian conception of their history the original homeland, a place ± an island or islands ± that their own ancestors had ®rst discovered and populated (see below). Historical anthropology can only hope to ``rediscover'' this ancestral space and time, to recover only partially the material culture, the social and political organization, the calendar and rituals, and other aspects of life in the Ancestral Polynesian homeland some two or more millennia ago. Hawaiki as the ancestral Polynesian homeland Hawaiki is a word deeply ingrained in the annals of Polynesian scholarship. Smith (1921:35 passim) regarded Hawaiki as the name of a place which had been ``occupied by the people in the remote past,'' noting that it ``is known to nearly every branch of the race, though it varies in form from island to island according to the changes that have taken place in the language since the dispersion.'' Williamson (1933, 1:312±13) pondered the matter at 95

96

Rediscovering Hawaiki

length, concluding that ``we are justi®ed by the evidence . . . in believing that the name Havaiki . . . was in some islands given to one or another ± sometimes to more than one ± place which was regarded as being an ancestral home; and that in some islands certain souls were supposed to pass to an ancestral home, which was sometimes called Havaiki, and was sometimes given another name.'' Burrows (1938a:73±76, diagram 17) pointed to the ``distinct regional distribution'' of the names Hawaiki and Pulotu, the latter being a strictly Western Polynesian concept of an ``island in the west, home of gods and of the elect after death.''1 While the two names thus have similar meanings, they are in complementary distribution in Eastern and Western Polynesia, respectively. As Burrows comments, Pulotu ``is as de®nitely western Polynesian as the concept of Hawaiki as underworld or ancestral home is central-marginal'' (1938a:76). Geraghty (1993) deals at length with Pulotu while Taumoefolau (1996) reconsiders Hawaiki, both invoking detailed linguistic arguments. The fascinating details of their respective arguments need not concern us here, except to note that Geraghty regards Pulotu as the original (pre-)Polynesian homeland, which he speci®cally situates in the Eastern Lau Islands, perhaps on Matuku. Taumoefolau offers an intriguing etymological hypothesis for the origin of Hawaiki, suggesting that it was originally a loan word from PTO into PNP, and that it can be traced back to a compound term in PPN, *sau ariki, meaning `chie¯y/ancestral/traditional ruler.' She believes that ``simultaneous with the modi®cation of the proto-form, there has been a corresponding shift in the proto-meaning from the `title of ancestral rulers' to the `name of the ancestral land' '' (1996:398). These lexical and semantic shifts would have occurred ``some 2,000 years ago, perhaps earlier, after PPN separated into PTO and PNP'' (1996:405). Our position in this continuing discussion over the naming of the Polynesian homeland is as follows. We think the most reasonable explanation for the distribution of the terms Pulotu and Hawaiki, and their varied meanings, is that Pulotu is the more ancient, reconstructable to PCP as *burotu, and which for the PPN speakers probably referred to the Fijian archipelago, their immediate homeland and place of the ancestors. *Sawaiki was a lexical innovation at the PNP interstage, corresponding to the period when the original PPN speech community (or dialect chain) was breaking up as a coherent unit. Its origin at this time is signi®cant: to the immediate descendants of the Ancestral Polynesians who began expanding out of the Samoa±Tonga homeland region, *Sawaiki/Hawaiki indexed that homeland, again the abode of the ancestors. Thus the origin of Hawaiki marks the end of the Ancestral Polynesian period. The name would be carried by Polynesian voyagers throughout virtually the whole of Eastern Polynesia, where it was variously

Introductory remarks

97

given to islands or places (e.g., Hawai`i, `Avaiki), and where it would ®gure in local religious ideology as the ancestral homeland, and abode of spirits of the dead. We thus use the word in much the same sense that we infer the PNP speakers to have done, as the name of the ancestral homeland in which a distinctively Polynesian culture arose. Plan of Part II Each of the six chapters to follow addresses a domain of Ancestral Polynesian culture. We begin with the physical world of the archaic homeland, its islands and seas, ¯ora and fauna, and how these were culturally categorized. We next turn to the economic basis of Ancestral Polynesian life, to cultivation practices, and strategies for exploiting the bounty of reef, lagoon, and open sea. This will lead in turn to a quintessentially cultural domain, the culinary complex. From the world of the ancient Polynesian cookhouse, we proceed to other aspects of technology and material culture. Social and political organization are then addressed, and ®nally, the spiritual world-view and ritual practices which structured the seasonal round, year after year. We are aware that these six major domains do not exhaust the possibilities for historical analysis; ``music'' and ``myth,'' for example, are domains we have left for others to pursue. For each of our chosen domains, we endeavor to show how the different strands of evidence ± linguistic, comparative ethnographic, and archaeological ± can be brought to bear through triangulation. In some cases, the direct evidence of archaeology plays a strong role, while in others we must depend almost exclusively on linguistic clues aided by comparative ethnography. In every case, the comparative ethnographic corpus is essential for robust semantic reconstruction. We hope to show, however, that there is no domain for which at least some reasonable hypotheses cannot be put forward. Although the ``rediscovery'' of Ancestral Polynesian culture is an intellectually suf®cient goal in its own right, as a substantive contribution to cultural history it is just as much the means to larger ends. These are the study and explanation of cultural change, ``evolution'' if you will, represented by the diversi®cation and differentiation of the myriad Polynesian cultures that descended from the ancestral ``Hawaiki.'' By de®ning the cultural content of the ancestral node or root, we set the stage for a more comprehensive analysis of subsequent cultural change, through the individual trajectories of daughter cultures and societies. But these larger ends extend far beyond our goal in this book. We will return to these future possibilities in the Epilogue, and offer there some thoughts on where the phylogenetic approach in historical anthropology might lead us in future endeavors.

98

Rediscovering Hawaiki

A note on orthography and abbreviations In the following chapters, we consider a great many Polynesian (PN) and Proto Polynesian (PPN) words, for which the following conventions apply. Reconstructed lexemes (whether attributable to PPN or other interstages of Austronesian) are designated by an initial asterisk (*). For PPN reconstructions, a glottal stop is indicated by the letter q, following Biggs' practice in POLLEX. For modern Polynesian language words, we use the more familiar symbol (`) to indicate the glottal stop. The velar nasal we have rendered consistently as ng (although in some modern Polynesian orthographies, such as for Samoan, it is indicated with a g; linguists denote it with h). The principal proto-languages or interstages with which we will be concerned are the following, with the abbreviations used for them: Proto Austronesian (PAN); Proto Malayo-Polynesian (PMP); Proto Oceanic (POC); Proto Central-Paci®c (PCP); Proto Polynesian (PPN); Proto Nuclear Polynesian (PNP); Proto Ellicean (PEC); and Proto Central-Eastern Polynesian (PCE). For modern Polynesian or other Oceanic languages, we use the standard three-letter abbreviations widely adopted by Austronesian linguists. A complete listing of these languages and their abbreviations is given in the list of language abbreviations (pp. xvi±xvii).

Chapter 4

The Ancestral Polynesian world

The lexical reconstructions indicate that the PPN speech community were ®shermen-horticulturalists, familiar with a typical tropical IndoPaci®c high island environment and also with certain objects found natively only on certain islands of this category, including the balolo worm, the pearl oyster, such land animals as snakes, pestiferous mosquitoes, bats, owls, rails, pigeons, parrots, and [a] moderately pawley and k. green 1971:23 diverse land ¯ora . . .

The Ancestral Polynesian homeland Linguistics, archaeology, and comparative ethnography converge to situate the Ancestral Polynesian homeland in space and time: the region known today as Western Polynesia. A WoÈrter und Sachen1 approach to locating a proto-homeland, well known to Indo-European specialists (Diebold 1994) is thus unnecessary. Nevertheless, it may be instructive ± from a theoretical perspective ± to ask whether the classic evidence of ``words and things'' independently agrees with the conclusions derived from archaeology and linguistic subgrouping. Some years ago, Pawley and K. Green (1971) queried the evidence of PPN lexical reconstructions, to ask where the homeland of the PPN speakers was most likely to have been located. Drawing on a preliminary version of POLLEX (Biggs et al. 1970), they discussed a range of relevant terms, contextualizing these within a set of postulates. For example, their Postulate 4 stated that ``the presence in any proto-language of a term denoting a category of objects is taken as indicating that the referents were familiar to the speakers of the language, either as part of their own immediate environment or as part of a nearby environment'' (Pawley and K. Green 1971:17). They correctly realized that such ``referents'' might include species introduced to the islands by humans themselves. We will not discuss all of their thirteen postulates, but simply note that the method applied by Pawley and Green was an exemplary instance of circumscribing the assumptions underlying a WoÈrter und Sachen approach. Among the many PPN lexical reconstructions considered by Pawley and K. Green are several words indicating that the Ancestral Polynesian home99

100

Rediscovering Hawaiki

land was situated ``on or near a high island or island group,'' and which at the same time preclude the possibility of this island(s) being located within Eastern Polynesia (1971:21). One of the most intriguing terms is PPN *palolo, re¯ected by TON, EUV, and SAM cognates, with an outside witness in FIJ balolo (indicating that the word was a PPN retention from the immediately preceding PCP interstage). As Pawley and Green relate, ``this `sea-worm' [Nereis sp.] lives in the crevices of coral reefs in certain regions of the Paci®c. The egg-swollen tail segments of the balolo rise to the surface around dawn on the 8th or 9th day after the ®rst full moon in October or November (sometimes in both months), and are taken in hand-nets as a highly prized food'' (1971:21). The Nereis sea-worm occurs only in Fiji, Tonga, and Samoa, and nowhere in Eastern Polynesia. Cognates of PPN *palolo are found in some Eastern Polynesian languages, but the reference is usually to a lunar month in the calendric system (see Chapter 9), or as in ECE to a small marine insect eaten by birds. The distribution of palolo terms, when compared with the geographical distribution of the sea-worm itself, unambiguously supports a PPN homeland in the Tonga±Samoa region. Similar evidence is provided by PPN *ngata, ``snake.'' The term is retained in TON, SAM, EFU, and EUV, along with TOK and a few Outlier languages, but snakes themselves are distributed only as far eastwards as Samoa (Loveridge 1946). As Pawley and K. Green observe, ``this does not imply that Samoa is the only possible homeland within the Triangle region, for snakes were evidently known by reputation to the inhabitants of some other island groups near to Samoa or Fiji'' where snakes also occur (1971:22). But it precludes the Eastern Polynesian archipelagos, where snakes are entirely absent, as possible candidates for the homeland of PPN speakers. Other lexical evidence delimiting the geographic region in which the PPN speakers resided includes an extensive list of plants and other organisms commonly found ± if not always exclusively distributed ± in Western Polynesia. This lexical evidence converges on the same homeland archipelagos indicated independently by both archaeology and by linguistic subgrouping. Beyond the shadow of a doubt, the Ancestral Polynesian homeland consisted of ``the central area of West Polynesia, i.e. the area bounded by Samoa, Uvea, Futuna, and Tonga'' (Pawley and K. Green 1971:23). We will now examine in greater detail the physical and biotic environments of this homeland, characterizing the opportunities as well as constraints or challenges posed by this environment, and exploring the ways in which Ancestral Polynesians categorized and conceptualized the natural world. We rely heavily on a careful review of the PPN lexicon, but do not neglect archaeological evidence, which provides a critical perspective on dynamic landscape and biotic changes, modi®cations to the Ancestral Polynesian

The Ancestral Polynesian world

101

environment which often resulted from human actions and land-use practices. The physical environment The Ancestral Polynesians inhabited a geographically diverse island world. The volcanic ``high'' islands included Futuna, `Uvea, Niuatoputapu, and the Samoan group, the makatea or upraised-coral islands were represented by Tongatapu and Vava`u, and the low-lying atoll or atoll-like islands by the Ha`apai group. Each type posed its own environmental challenges, and offered unique resources. The high islands, with their fertile volcanic soils and rainforests, were well suited to shifting cultivation (see Chapter 5). On the smaller coral islands, cultivation must have been more limited, but intensive forms of tree cropping (especially of coconuts and breadfruit) would have been favored. Yet the coral islands had their own advantages, especially their vast reefs and lagoons rich in ®sh and shell®sh. This spectrum of environmental variability, necessitating particular human adaptations, encouraged cultural variation among the Ancestral Polynesian societies. In moving eastwards beyond the large Fijian islands of Viti Levu and Vanua Levu, the immediately pre-Polynesian (Early Eastern Lapita) peoples crossed a fundamental threshold of Paci®c geology and biogeography. The ``Andesite Line'' which runs between Fiji and Western Polynesia (see Figure 4.1) marks the tectonic divide separating the vast Paci®c Plate on the east, from the Fijian Plate on the west. As the Lapita peoples expanded out of the Bismarck Archipelago, down through the Solomon Islands and out to Fiji, they encountered islands that, while different in minor respects, were all of andesitic-arc geological origin (Nunn 1994). Such islands have long and complex geologic histories, composed of a diverse suite of rocks, including both igneous and meta-volcanic types, as well as sedimentary formations. East of the andesite line, the islands are exclusively formed by ``hot spot'' or mid-plate volcanic eruptions, largely of basaltic composition. Erosion and subsidence of these volcanic islands gradually converts some of these to atolls, or when they are elevated, to makatea islands. The consequences of moving over this transition from island-arc to midplate volcanic islands were far-reaching. For one, the islands east of the andesite line are generally much smaller than those to the west. Within the Ancestral Polynesian homeland, the largest island is Savai`i with 1,600 square kilometers, a good-sized land mass but considerably smaller than Viti Levu at 10,388 square kilometers. Most other islands in the Polynesian homeland are much smaller, such as Niuatoputapu at 15 square kilometers, or many of the islets in the Ha`apai group. Second, the lithic resources

102

Rediscovering Hawaiki

available for stone tool production were limited to dense, relatively hard-towork basalt or hawaiite;2 a low-silica type of obsidian also occurred at a few localities, such as on Tafahi Island. This restricted suite of available rock types inspired changes in the way in which stone tools, especially adzes, were manufactured (see Chapter 7). Third, biotic diversity (both terrestrial and marine) declines as one moves eastwards out of Fiji into Samoa and Tonga. Although the ¯ora and fauna of the Ancestral Polynesian homeland is part of the Indo-Malasian and Indo-Paci®c provinces from which the Lapita peoples had hailed, many of the useful or edible species found in the larger westerly archipelagos were missing in Samoa and Tonga. How did the Ancestral Polynesians conceptualize this new world of properly ``oceanic'' islands? The PPN reconstructions pertaining to the physical world open a window into the Ancestral Polynesian mind, yielding the very semantic categories into which this world was partitioned and classi®ed. Table 4.1 presents the set of PPN words for aspects of the physical world.3 Because this is the ®rst of many such tabular presentations of PPN lexical items we will use, a few brief explanations are offered. The ``probable gloss'' is our best judgment of the meaning of the PPN term. The POC and PPN reconstructions are as given in POLLEX (Biggs 1998), augmented by other sources, especially for POC (e.g., from the chapters in Pawley and Ross [1994], and in Ross et al. [eds., 1998]). If cognate POC and PPN terms are given, the PPN term must obviously be a retention, whereas if only a PPN form is given, it is probably an innovation at the PPN (or in some case PCP) interstage. Table 4.1 and others to follow include columns bearing the codes ``NCOG,'' ``P1,'' ``P2,'' and ``PSA.'' NCOG is the number of cognates within Polynesian languages (excluding extra-Polynesian languages such as FIJ and YAS). If NCOG is high (~20±30), the PPN term has been retained in the majority of Polynesian languages, while if it is small (