Exploring spatial relationships between material culture and language in the Upper Sepik and Central New Guinea.

Exploring spatial relationships between material culture and language in the Upper Sepik and Central New Guinea. LAPITA AND THE EXPLORATION OF RELATIONSHIPS BETWEEN LANGUAGE ANDCULTURE IN MELANESIA One of the most important and enduring debates associated with theprehistory of Melanesia has concerned the migration of Austronesianspeakers into the region. The geographical distribution of this familyof languages ranges from Taiwan in the north to New Zealand in thesouth. Its east-west distribution is significant, spanning the areabetween Madagascar and Easter Island, and includes the IndonesianArchipelago, the Philippines and Polynesia. It is thought thatmigrations resulting in this distribution began somewhere around 5,500Before Present (BP) and culminated in a final wave of migration fromMelanesia into remote Oceania around 3,200 BP (Bellwood 1991). Evidence for retracing the origin and course of Austronesiandispersals has been sought initially through assessments oflexicostatistical relationships among the hundreds of Austronesianlanguages (Gray & Jordan 2000). This evidence has given rise to anumber of models of Austronesian dispersal (Greenhill & Gray 2005:32-3). Three of the models have the Austronesian origin as Taiwan fromwhich populations moved into South East Asia then east to Melanesia andPolynesia. (1) The only significant differences between these 'Outof Asia' models are the length of time in which the migration isassumed to have taken place and the number and intensity of migrationepisodes. (2) The 'Entangled Bank' hypothesis provides an alternativemodel which proposes that present population and language distributionswithin Melanesia and Polynesia are best explained by demographic changesoccurring within established populations in Melanesia inclusive of anintense network of interaction between Austronesian speakers and localpopulations across Melanesia and east South East Asia (Hurles et al.2003; Terrell 1988; Terrell et al. 2001). (3) It has been suggestedthese factors would have given rise to the kinds of seafaring andagricultural innovations that enabled populations to access and coloniseremote Oceania. Those who support the 'Out of Asia' models have proposedthe Lapita complex (3200-2600 BP) as a cultural marker for theAustronesian migration eastwards across the north coast of New Guineaand into the Bismarck Archipelago. Elements of this complex includedomesticates and agriculture: dogs, pigs, chickens and a distinctiveadze technology, as well as the introduction of a new settlement patterncomprising clusters of stilt houses built over protected and shallowcoastal waters (Gosden & Webb 1994; Spriggs 1993: 192). The mostsignificant material characteristic of the complex, however, is adentate-stamp decorated ceramic ware that has been found atarchaeological sites almost entirely within Melanesia but includingFiji, Tonga and Samoa. Its most westerly occurrence is on the northcoast of New Guinea. The interpretation of Lapita as a marker for the arrival ofAustronesian peoples in Melanesia derives from a traditionalarchaeological viewpoint that abrupt occurrences of cultural phenomenain the temporal record are best explained by the arrival of a newpopulation (Jones 1997: 16-7). The lack of any pre-Austronesian ceramicprecursor to the elaborate incipient Lapita ware, coupled with theco-occurrence of other parts of the complex, has been seen as evidencethat it was brought into the area fully developed (Gosden & Webb1994; Kirch 1996; Kennedy 1981). Uncertainty as to the function of theceramic ware, its distinctive design and stylistic qualities as well asthe tradition's relatively short span of less than five hundredyears, has led to the proposition that it was an importantsocio-cultural marker for the colonising populations (Sand 1997: 4-6).It is assumed that there was a significant pause and consolidation inthe Bismarck Archipelago during the colonisation process that gave riseto the emergence and florescence of a network of exchange andinteraction, including a development of a localised Lapita culture,before population movement continued south-east to Vanuatu, NewCaledonia and Fiji and eastwards into western Polynesia (Green 1991).Once Lapita had been distributed as far as Fiji and Samoa, isolation andlocalised spheres of interaction resulted in a number of stylisticprovinces and gradually the ornate design gave way to plainware (Anson1986; Green 1978; Kirch 1990) A contrasting theory supported by the proponents of the'Entangled Bank' model sees Lapita as an innovationoriginating in the Bismarck Archipelago that diffused along overlappingspheres of interaction between existing and recent arrivals westwardsalong the north coast of New Guinea and south-eastwards to Vanuatu, NewCaledonia and Fiji (Allen 1984). (4) This model is also supported bysome important archaeological facts: there has been no Lapita stylepottery found in the supposed origin of Austronesian peoples in SouthEast Asia; Lapita has been found only as far east as Samoa and Tongawithin the Polynesian sphere; and both physical and decorative qualitiesof Lapita ware exhibit characters reflecting an inclination for locallysourced materials and techniques (Ambrose 1997; Chiu 2003; Galipaud1988). Nevertheless, it is possible that elements of the Lapita traditionmay have derived from a number of sources and questions have arisen asto what kind of features belonging to Lapita ceramics are pertinent toexploring questions surrounding the arrival of Austronesian speakers(Green 1988). Most attention has been focused on the dentate designs andtheir component elements but in reality there are a bewildering numberof stylistic, technological and material features that demonstrateconsiderable diversity throughout Melanesia. As Green (2003: 102) hasrecently argued, there is a need to sort elements of the Lapita Cultural Complex into separate piles as a strategic part of the approach to improving understanding of the various processes at work with respect to the history of each of the topics, traits, clusters or particular elements under discussion. There have been some attempts to isolate particular material,morphological and technological elements of Lapita ware and track theirdistributions but these have been limited in their regional scope (e.g.Chiu 2003; Galipaud 1988; Summerhayes 2000). There remains considerablework to do, not only in assessing the range of styles and techniquesassociated with the tradition, but also in the task of constructing arange of attribute levels that can be tracked co-ordinately acrossLapita sites throughout the Pacific. THE QUANTIFICATION OF RELATIONSHIPS BETWEEN MATERIAL CULTURE ANDLANGUAGE ON NEW GUINEA'S NORTH COAST The Lapita puzzle has significant relevance to the questions thathave preoccupied anthropologists and archaeologists concerning theemergence, spread and evolution of cultural packages and whether suchpackages can be used to identify populations who may share genes,languages and indeed common histories (Bellwood 1996; Terrell 2001). In an attempt to provide answers to such questions, a team at theField Museum, Chicago, undertook an investigation of the relationshipsbetween material culture, language and distance along a 600 kilometrestretch of the north coast of New Guinea by using a dataset of over6,000 items collected in the region by the ethnographer Albert B. Lewisbetween 1909 and 1913. (5) The environment of this region is relativelyhomogeneous and without a need to factor in the influence of a varyingenvironment and geography, the relationship between material culture andlanguage distribution was presumably easier to explore. The Field Museum team used the dataset to gauge the level ofsimilarity between assemblages from villages within the study area. Eachobject within an assemblage was classified according to itsfunctional/operational mode. This classification was seen as a method ofindicating the presence of certain cultural practices that involved theuse of those objects. Presence and absence of object classes was used todetermine assemblage similarity or dissimilarity. Even though theassemblages used in their study largely consisted of objects collectedby one person, they found marked differences in the number of objectsand the number of classes of objects in the samples representing eachlanguage area. They also found that the number of classes present ateach village was closely related to sample size and accordingly factoredthis into the analysis (Welsch et al. 1992:581). (6) Welsch et al. (1992: 582) constructed a language matrix fromexisting assessments of the language relationships. (7) Geographicaldistance between each pairing of villages was measured and used as thecounter variable to language in the analysis. Multiple and partialregressions were then carried out to measure the relationships betweenthe variables. They found that sample size accounted for most of thesimilarity between assemblages; the rest demonstrated a greaterrelationship with distance than with language (Welsch et al. 1992:583-88). The Field Museum study was seen by many as significant because ithas brought to anthropology and archaeology [a] model set of data that represents an era that is of historical importance and could not be duplicated today. Given the quality, complexity, and completeness of the data together with the theoretical relevance of the variables of propinquity, language, and material culture, it is certain to become a classic data set (Moore & Romney 1994: 370-71). But the Field Museum method has since come under criticism for tworeasons: firstly, the study did not compare morphological differencesamong class members and by not doing this the Field Museum study failedto acknowledge that objects that are considered functionally equivalentmay vary in form. Appropriately, Peter Bellwood (1995: 777) asked ... what would the Welsch et al. analysis have shown had it been focused not on gross functional categories of items but on fine-tuned stylistic variations (eg. in shape and decoration) within these categories? What would be the patterns with respect to linguistic differentiation if one were to extend this kind of analysis to the whole of Melanesia, including the New Guinea Highlands? Functionally/operationally equivalent objects are often made with arange of different materials and techniques at different localities.These often provide a better means of establishing cultural ties, indeedas A. B. Lewis himself pointed out: [The design on string bags] and the character of the weave of the narrow band around the mouth of the bag, indicate the local place of origin. A native who is familiar with these styles can tell what village a bag came from by looking at it (cited in Welsch 1998:119-20). Secondly, in using matrices of assemblage similarity anddissimilarity constructed from presence/absence binary data, they didnot quantify the frequency distribution of individual classes whichwould have enabled the exploration of more discrete relationshipsbetween sites. Soon after the Field Museum analysis, anthropologists of theDepartment of Anthropology, University of California, Irvine, undertooka reanalysis of the Field Museum data (Moore & Romney 1994; Robertset al. 1995). Rather than using presence and absence to gaugesimilarity/dissimilarity, they used class frequency data in bothcorrespondence analyses and log-linear modelling and found that languageand distance equally accounted for the total variance in the data. More recently, Stephen Shennan, and Mark Collard (2005), both ofthe University College London, applied a range of multivariate analyseson the Field Museum dataset. They were able to demonstrate a significantdifference between the assemblages of Papuan and Austronesian speakingcommunities. (8) Importantly, nearest-neighbour tests established thatAustronesian villages were more likely to be similar to each otherregardless of propinquity. As Austronesians were relatively new arrivalsthey also argued that a phylogenetic signal may be evident in theircultural set. Accordingly, they undertook cladistic analysis and foundthat branching (phylogenesis) rather than blending (ethnogenesis) betterexplained the Austronesian set. Nevertheless, while the University College London and University ofCalifornia analyses may have provided a better means to quantify andanalyse the distribution of functional classes, the fact that they wereunable to quantify the formal relationships between members of the sameclass means that the majority of potential variability within the A.BLewis dataset has remained unexplored. THE UPPER SEPIK-CENTRAL NEW GUINEA PROJECT AND ITS OBJECTIVES The Upper Sepik-Central New Guinea Project (USCNGP) is amulti-disciplinary project undertaken jointly by the South AustralianMuseum and the Department of Geographical and Environmental Studies atthe University of Adelaide. Its purpose is to quantify and explorerelationships between material culture, geographic distance, language,and the environment. (9) It has sought to further contribute to thedebates that have arisen from the Field Museum study by applying similaranalyses to a dataset from the Upper Sepik and Central New Guinearegions of Papua New Guinea. This paper presents an overview of results from preliminaryanalyses undertaken by the Project. These analyses focus on the two mostubiquitous and important functional/operational classes from the studyarea. One class comprises arrows made by men and used by them forhunting and fighting, the other the diverse range of string bags made bywomen, which are essential in the daily and ritual lives of both womenand men. Importantly, each class provided a comprehensive range oftechnical and stylistic variability that was able to be codified. Among others that will be discussed below, there are two crucialreasons why the use of these two classes is seen as having greatpotential. Firstly, due to their complexity and ubiquity they offer afine-grained measure of cultural similarity/dissimilarity between thecommunities of the study area. Secondly, as these classes areexclusively products of either gender their respective attributedistributions are likely to reflect differences that exist between thegenders' social patterns. In doing so they should not only providean insight into the differences between the way men and women socialiseand move through geographical space but they would also enable betterdiscrimination of more discrete cultural patterns, most importantlythose that best reflect either the language picture or relative spatialpositions. THE GEOGRAPHY AND ENVIRONMENT OF THE STUDY REGION The Project's study area can be defined as the regionintersected by the course of the Sepik River and its tributariesupstream from its confluence with the May River to its source in thehighlands where it shares a large watershed with the Fly and Digulrivers. The Bewani Mountains form the northern limit of the study area(thus precluding the north coastal zone of West Sepik Province), and thesouthern slopes of the highlands of Central New Guinea (CNG) define thesouthern limit. The area can be subdivided into three geographical regions: adramatic mountainous zone consisting of around 13,000 square kilometresof CNG, a lowland zone immediately to the north that encompasses over6,000 square kilometres of the Upper Sepik Basin (USB), and a hilly zoneto the northwest of around 1,200 square kilometres known as the BorderMountains (Maps 1 and 2). Central New Guinea The highlands of CNG are essentially a cordillera running east-westwith an apex that follows the Western Province and West Sepik Provinceborder. The Star Mountains form the highest range in the system,reaching an altitude of around 4,000 metres in the western extreme ofCNG in Papua New Guinea. They continue into Indonesia where they coverapproximately 2,600 square kilometres of [West] Papua (Brongersma &Venema 1962). This geologically diverse and rugged region is characterized bylimestone karst formations with massive escarpments along the southernedge (Gillieson & Hope 1990: 28; Reynders 1962: 46). The slopes ofthe central range are serrated and complex, consisting of sedimentaryrock which is unstable and prone to slippage due to the heavy rainfall(Hyndman 1979: 59; Hyndman & Menzies 1990; Morren 1986: 67). Alongthe centre of the highlands is a series of broad inter-montane valleysbetween 1,450 and 1,800 metres above sea level, forming the sourcebasins for the Strickland, Sepik and Digul Rivers. Hyndman and Menzies (1990) have identified three zones of forestfor CNG. The foothills north and south of the highlands (100 to 500metres), including the West Range, provide stands of sago palms in thevalley floors and tropical forest of mixed composition on the slopes andridges (Hyndman 1979: 79; Hyndman & Menzies 1990: 245). The mid-altitude zone (500 to 1,500 metres), south and north of thecentral range, is rich in wild plant resources and various animals, butcloud and excessive rain restrict the success of tuber crops such astaro and sweet potato (Gillieson & Hope 1990: 29-30; Hyndman 1979:61,78ff; Hyndman & Menzies 1990: 244; Hyndman & Morren 1990:16-8; Morren 1986: 68-71). [ILLUSTRATION OMITTED] [ILLUSTRATION OMITTED] In the highlands above 3,000 metres cloud forest and alpine tundradominate; these are generally poorer in plant resources but the formersupports a greater diversity of marsupials desirable as prey (Morren1986: 71). The inter-montane valleys, located around 1,500 metres,provide relatively stable micro-climates; within them are situatedprimary and secondary stands of temperate rain forest especially on thesurrounding slopes. Considerable tracts of cane and kunai grass(Miscanthus and Imperata sp.) caused by centuries of sustained shiftingagriculture cover much of the valley floors (Gillieson & Hope 1990:29). In CNG, most people live between 1,250 and 1,800 metres. Themid-altitude fringe is lightly populated with settlements mainlyconcentrated in the larger valleys. The most densely populated areas arefound within the inter-montane valley system as these broad and openvalleys have the lowest precipitation of the highland region, as well asbetter exposure to sunlight, and therefore provide the best environmentfor shifting agriculture (Cranstone 1990: 35-6). The Upper Sepik Basin In contrast to the highlands of CNG and the Border Mountains, theUpper Sepik Basin (USB) is a relatively featureless alluvial plainsituated between 50 and 100 metres with a few isolated hills exceeding200 metres. It is drained by several tributaries of the Sepik River andlarge tracts are flooded in the wet season (November to April). There are three major environments in the USB. One consists oflightly timbered swamps that are recharged by flooding in the wet seasonand floodplain forest with a rather low canopy. Within these areextensive stands of sago palms and swamp grasses. Another environmentconsists of mature or secondary rainforests with a diversity of plantspecies and trees (Morren 1986: 68). Two subdivisions of such foresthave been identified and described by Reiner and Robbins (1964: 32-3)for the lowland hills and plains of the middle Sepik to the east and,from descriptions found throughout patrol reports and ethnographies, theforest structure of the USB conforms to their observations. Alluvialforests exist in the lowlands above the flood zones and a lowland hillsrainforest of more mixed composition, which is relatively analogous tothat in the hills proper, is found at the higher fringes of the plain(Clunie 1978: 4-6). By and large both these forest subdivisions havethree strata with canopies reaching up to fifty metres. The last major environment type consists of stands of anthropogenicgrassland also comprising rhizomatous grasses. Two very large stands arelocated away from the major riparian flood zones near the Green Riverand Yellow River stations (Jeffries 1950; Kelm & Kelm 1980: 3-10). The most variable environmental section of the Basin, for whichThurnwald (1914: 5-6-translated by Harry Beran) provides an excellentdescription, is on and near the banks of the Sepik and the other majorrivers: The banks are similar all along the upper course of the main river: steep and wooded on the outside of the curves, while on the inside, sandbank or wild sugar cane and behind, young forest with numerous wild breadfruit trees. One frequently comes across signs of alterations in the river's course--new breaches and devastated woods, the mouths of old streams, and lagoons. Because much of the Basin lacks the environmental diversity foundin the hills and mountains, communities traditionally took advantage ofareas where riparian zones, alluvial forest and swamp were adjacentbecause diverse resources could be accessed relatively easily.Consequently, large tracts of swamp were left relatively unpopulated. The Border Mountains The topography of the Border Mountains, like that of the slopes andfoothills of CNG, is complex (Gell 1975: 8-9; Huber 1979; Peter 1990:245). From swampy forests at around 50 metres, thirty kilometres east ofthe border with Indonesia, they rise to a maximum of just over 1,000metres at the border, some forty kilometres north of the Sepik. Fromhere they extend a similar distance into [West] Papua. In total theyspan approximately fifty kilometres north to south. Huber's (1979: 131) description of Anggor territory on theeastern flank of the mountains provides a vivid and accurate picture forthe region: ..... the terrain is steep and irregular. It consists of an intricate dendritic drainage pattern of small, even miniscule watercourses. These are separated by steep narrow ridges, forming an outline that articulates and complements the stream pattern. In any locality the overall altitude gradient is obscured or distorted by small, prominent mountains and by the myriad oblique intersections of streams and ridge crests. The ridge crests are narrow, five to ten meters, and the hillsides are commonly in excess of thirty degrees in slope. Much of the Border Mountains is composed of limestone outcrops withvalley floors and hollows composed of fine sedimentary clays. The lowestground is generally swampy and includes stands of sago palms. Denseprimary and secondary forest dominates the rest of the terrain and thereare no significant grasslands. Most settlements are located on ridgesbetween 300 and 500 metres. The Languages of the study area A recent assessment of the study area's linguistic situationwas undertaken by Martin Steer for the Project. He summarises (2005: 4): The languages of the upper Sepik and of CNG differ markedly in the character of their genetic relationships. The Sepik catchment exhibits a degree of genetic diversity unequalled anywhere in the world with 200 languages belonging to perhaps as many as ten unrelated families together with several genetic isolates, and the upper Sepik has diversity commensurate with this. The CNG region, by contrast, is linguistically relatively homogeneous. Sixteen languages are found in the USB and Border Mountains (seeMap 3). The majority of these were originally assigned by linguists tothree phyla (Laycock 1973; 1975; Wurm & Hattori 1981): * Trans-New Guinea (Waina and Amanab of the Bewani Family and Deraand Anggor of the Senagi Family); * Kwomtari (Fas and Kwomtari in the Kwomtari Family, and Baibai andBiaka in the Baibai Family); * Sepik-Ramu (Abau constituting its own Family within the UpperSepik Stock and Namie, Ak and Awun in the Yellow River Family/Stock). Four languages were unclassified and now considered to be isolates,i.e. unrelated to any other languages (Yuri in the Border Mountains,Nagatman and Busa in the Upper Sepik lowlands, and Amto in the WestRange). Recent linguistic research has called for revisions to theclassification of these languages. Membership of the Senagi Family inthe Trans-New Guinea Phylum has been questioned with suggestions thatthese languages are more likely to be distantly related to languages ofthe Sepik (Steer 2005:18-9). Also Foley (2005) has questioned theintegrity of the Sepik-Ramu Phylum and has placed the languages into twoseparate 'families' (Lower Sepik-Ramu and Sepik); this doesnot affect the determination of relationships between the languages ofthe study area. At present eight languages are identified for CNG; all of thembelong to the Trans-New Guinea Phylum (Map 3). The Ok Family oflanguages has been sub-divided into Mountain Ok and Lowland Ok (Healey1964), the former are located in CNG and the latter inhabit the lowlandsto the south. The Mountain Ok Sub-Family has three apparent divisions inlinguistic relatedness: one includes the Ngalum in the west; anotherincludes the Mianmin in the north; and the last includes all remainingMountain Ok language found in the central and eastern part of theregion: Kauwol, Tifal, Telefol, Faiwol and Bimin. Oksapmin is aTrans-New Guinea Phylum-level isolate that is spoken in the extreme eastof CNG. Steer has provided the Project with a matrix of relatedness (Tables1 and 2). It 'implies a relative chronology' and wasconstructed on the basis of shared cognates with some consideration ofmorphology and structure (Steer 2005: 7-8; see also Laycock 1973: 70-1). [ILLUSTRATION OMITTED] REGIONAL SUBSISTENCE PATTERNS Subsistence systems found in the Border Mountains, USB and CNG havebasic similarities and dissimilarities. The most significant similarityis that all populations practise shifting agriculture where suckers orcuttings are transplanted with the use of digging sticks, although thereis variation as to whether people use mulching or burning in thepreparation of gardens (Hyndman 1979:89-91; Huber 1974: 66-9). Anotherimportant similarity is that there is usually only one cropping before afallow period of between fifteen and thirty years. However, in parts ofthe highlands where the cultivation of sweet potato is more important,fallow periods are somewhat shorter and multi-cropping does occur (Allenet al. 2002; Bourke et al. 1993). The major difference between the three regions at the time thecollections were made was the degree to which people relied onhorticulture for their staples. People of the lowland plains and to alesser extent the hills environment of the West Range and BorderMountains, intensively harvested stands of sago and complemented thiswith a diversity of fruit and nuts that could be collected during thesago harvesting process. Consequently gardening was only a subsidiaryactivity that included a diversity of cultivars such as taro, sweetpotato, yams, breadfruit, bananas and coconuts (Craig 2002: 2; Huber1974: 62-81; Jeffries 1950; Kelm & Kelm 1980: 115; McCarthy 1936).It is also important to note that the climate in the lowlands ismoderately seasonal; planting occurs with the coming of the wet and thismeans that the diet varies somewhat over the course of a year. In contrast, in large parts of CNG, people practised a virtualmonoculture dominated by taro. (10) Their gardening regimes involvedshorter cycles of shifting cultivation with a series of plantings beingmade throughout the year, providing gardens at various stages ofmaturity (Ohtsuka 1994). Over the last four or five decades, however,there has been an increasing reliance on sweet potato which, unliketaro, matures within a single year, thrives above 1,500 metres, can growon lower valley slopes and floors that have poor soils unsuitable fortaro, and provides excellent pig fodder (Eggertsson 2003: 19-22; Hyndman& Morren 1990; Jones 1980:11-6; Kuchikura 1990; Morren 1986: 90-100,Poole 1976: 255-300). The use of sweet potato provides another measure of difference inthe study region, as it was important in some areas during the period inwhich the major collections were made. It must be pointed out here thatthe economic and social changes brought about by sweet potato for otherNew Guinea highland societies--what has been termed the 'Ipomeanrevolution'--had not occurred in CNG (c.f. Watson 1965). Rather,the adoption of sweet potato has been argued to have been either aresponse to critical limitations of taro horticulture in the face ofenvironmental degradation (Hyndman & Morren 1990:21); or to a needto increase domesticated pig populations to replace wild animal stockslost through loss of forest (Craig 1990). While it must be acknowledged that there have usually been higherpopulation densities in other regions, where sweet potato is thedominant staple, in CNG there was no large scale gardening that couldprovide a surplus and lead to major changes in social patterns. On thewhole, there are no major differences in social complexity between CNGgroups that favoured either sweet potato or taro, although sweet potatodominant groups do tend to have smaller settlements. In both the lowlands and highlands, hunting involved theexploitation of a broad range of marsupials, cassowaries, birds andfish, but the main focus was on the hunting of feral pigs, which arefound across the range of environments (Huber 1980; Jorgensen 1981,1983; Kelm and Kelm 1980: 59). The major difference was that lowlandgroups had access to a greater diversity of mammals and reptiles, and,most importantly, feral pigs were more abundant, as was fish, as analternative protein source. Because of the relative paucity of game, highland groups developeda form of pig husbandry involving a stock of semi-domesticated breedingsows which for much of the time were allowed to roam free, enabling themto consume wild fodder and to be impregnated by feral boars (Cranstone1990: 40; Morren 1986: 88-9). While lowland groups did not traditionallypractise pig husbandry, they sometimes supplemented their protein byrearing captured feral piglets (Gell 1975: 17; Juillerat 1982: 287). In essence, differences between the geography and environments ofthe regions is reflected in some crucial differences between thesubsistence patterns. In the relatively homogenous environment of thelowland plains, resources are more evenly distributed and accessible,and subsistence activities are less regimented and require littleintensification. In the relatively heterogeneous environments of CNG andBorder Mountains, wild resources and arable land are distributed moreirregularly and are limited. Subsistence activities, undertaken by thefamily or spousal unit, were more geared towards horticulture andrequired more mobility. The hills environments are a transitional zone but more closelyresemble that of the plains because the exploitation of sago was thedominant subsistence activity, one that also provided opportunities tohunt and gather wild foods. It is important to note here that somehighland fringe groups, such as the Mianmin in the northern limit ofCNG, in more recent times have moved into lowland zones in the foothillsof the West Range and consequently have adopted some lowland subsistencestrategies (Morren 1986: 283). SETTLEMENT TYPES Population density throughout the study area was, and still is,relatively low averaging below three persons per [km.sup.2]. When mostof the ethnographic material was collected, around 20,000 people livedin the USB and Border Mountains, and around 30,000 in CNG. Unlike somevillages further down the Sepik River and in the highlands to the eastthat have populations up to a thousand people, settlements throughouthave comprised relatively small populations of between fifty to 250persons. Traditional settlements in the study area followed four basic formsthat correspond to particular intervals in demographic size (e.g. Allen1983: 18). These four classes of settlement and their composition can bedescribed thus: 1) homesteads: two to three houses, twenty to thirty people; 2) large communal houses: one house, twenty to fifty people; 3) hamlets: up to five houses, forty to ninety people; 4) villages: more than five houses, 100 to 250 people. In villages, where there were many houses, they were usually setaround a plaza and some larger settlements included non-familyresidential buildings such as permanent ritual structures, menstrualhuts and separate men's houses (Craig 1969). Access to land for gardening or hunting and gathering determinedthe position of the settlement. All settlements had a defined territorywith land dedicated to horticulture and to hunting and the gathering ofwild resources (Huber 1980: 45; Morren 1986: 249; Ohtsuka 1994). Therewere significant ecological constraints determining settlement size,most importantly the nature of rainforest soil, which is usuallyexhausted after a year of cropping. Even where gardens did provide agreater proportion of the diet, the short period of cropping and longfallow required that gardens be established at increasing distances fromthe settlement. In some CNG and Border Mountains communities, thisproblem was overcome somewhat by garden housing (eg. Jorgensen1981:151). In other areas, the settlements, or parts of them, wererebuilt close to the new gardens every generation, or at least onceevery second generation (Barth 1971: 188; Morren 1986: 198-202). The fewlowland groups that were more reliant on gardening shifted every fewyears (Thurnwald 1914: 342-43). However, where sago was the main stapleand horticulture of less significance, settlement relocation was moreoften related to the depletion of wild game and social instability(Barry Craig pers. comm. 2006). Sometimes the pressure to relocate provided the opportunity for asplit in the community resulting in part of the population moving to newterritory or joining another community (Juillerat 1992a: 5; Morren 1986:245-47). Major settlement shifts were also caused by misfortunebefalling a community, such as sickness (sorcery), a perceived declinein fecundity, or conflict (Barth 1975: 22). In parts of CNG, the Border Mountains and in the Yellow River areaof the USB, communities lived in dispersed affiliated hamlets or hamletclusters of up to 200 people, or villages of 100 to 200 people. Hamletswere the most common settlement form throughout CNG and the northernregion of the Border Mountains, while villages were more apparent in theeastern Border Mountains, Yellow River and in the inter-montane valleysand southern slopes of CNG. In the inter-montane valleys, villages andhamlets were located on the wide valley floors (Craig 1969). In thenorthern slopes of the highlands, and in the hilly country of the BorderMountains, villages and hamlets were located on ridges above narrowvalleys for security, while still providing reasonable access to gardenson the slopes and to stands of sago in the valley bottoms (Bercovitch1989; Huber 1974: 33; Morren 1986: 72-5). Both of these settlement patterns, the village and hamlet cluster,could involve a dual residence pattern, the settlement proper oftenremaining relatively vacant with a large proportion of residents livingas conjugal or family units in garden houses; a practice that causeddifficulty for anthropologists and patrol officers alike (Huber 1980:46). This was done to not only tend gardens but also so that individualsand couples could enjoy a period of relative privacy (Barth 1975: 26;14; Huber 1980: 46; Gell 1975: 79; Jorgensen 1981: 151). Confederacies of villages, hamlets or hamlet clusters representedthe largest form of communal unit, commonly referred to as a'parish' in the literature, due to members being linkedthrough local cult obligations, fictive descent and a network ofmarriage ties (R. Craig 1969; Morren 1974; Perey 1973; Pouwer 1964).Parishes were geographically delimited within large valleys andwatersheds and were relatively stable. In CNG, all parishes included acentrally located settlement with a cult house or cult houses wherepeople from settlements within the parish congregated for ritualoccasions and male initiation (Barth 1971: 174; Craig 1969; Hyndman1979; Jones 1980: 25). Thus social and domestic organisation waspartitioned according to gender. The village living space was dividedinto male and female zones: the women, girls and pre-pubescent boyslived in individual family houses set around the central plaza and themen slept together in one or two men's houses located at the'top' (upstream) end of the settlement (Craig 1988: 26;Jorgensen 1981 : 149-55). (11) Many communities of the USB are now located along the course of theSepik River and its tributaries. This is a relatively recent developmentmade possible by the pax Australiana. Previously, most riverinesettlements were small, set a little distance from the river on naturallevees or high ground between the rivers and swamps. This provided adegree of security against flooding and raids from the river (Thurnwald1914: 6). Elsewhere in the USB, settlements were located on the fringeof grassland or lowland alluvium forest close to swamps where stands ofsago grew (Bragge 1964). In much of the USB, people lived in smallhomesteads, or large communal houses that contained a number of familiesor extended families (Conrad & Dye 1975: 11; Schultze-Jena 1914:passim). However, independent family dwellings have become more commondue to government and missionary influence in the area since the 1950s(Craig 2002: 1-2; Craig 2008; Gell 1975: 11, Peter 1990: 247). SOCIAL STRUCTURE AND DESCENT Weakly asserted descent units are features of many societiesthroughout New Guinea and kinship has often been framed as a product ofa tenuous dialectic between obligation and optation that results in a'looseness' or 'flexibility' in social relationships(Barnes 1962; Brown 1962; Harrison 1985: 415; Heider 1970: 6; Kaberry1967; Lepervanche 1967; Watson 1970). Such observations are no lesscommon, indeed are very apparent, in ethnographies for the study area.Reckoning of descent is usually patrilineal, with bilateral associationsoften given considerable weight. There appear to have been few socialconventions to obstruct individuals who wished to establish ties beyondtheir close kin (Bercovitch 1989: 76; Craig, R. 1969: 177; Huber 1974:9; 20; Hyndman 1979: 49; Jorgensen 1981: 180; Morren 1986: 175; Pouwer1964: 144). An important feature of many New Guinea societies that has beenseen as undermining the objectification of descent relationships is therelative lack of formal leadership coupled with an ideal of equivalencewhich meant that there was little to be gained from filiation(Bercovitch 1994; Read 1959: 427-29; Scheffler 1985). This can beconsidered even more so for the study area for there was no counterpartto the 'Big Men', evident in other highlands societies ofPapua New Guinea, who commonly accumulated wealth (Bercovitch 1994: 504;Brumbaugh 1980: 29; Huber 1974: 214). (12) Therefore there was noincentive to trace descent to any depth in these societies. Suchknowledge was a residual of any continuity in residential patternsthrough generations and most informants were unable to recountgenealogies beyond three generations (e.g. Brumbaugh 1980: 109;Jorgensen 1981: 211). What is most apparent in the ethnographic accounts is that thesibling and cousin relationship was pivotal to the formation of socialunits. These relationships united conjugal units for the formation ofextended corporate units and the cross-sex sibling relationship was, ofcourse, crucial in linking men from different social units (Brumbaugh1980: 86, 112; Gell 1975: 43; Jones: 1980: 68; Juillerat 1996: 337ff;Pouwer 1964: 141; Wheatcroft 1975: 94). The most important and 'conspicuous' social unit for thestudy area was the nuclear family as it was an 'independentlyoperating work, residential, consumption and property-owning unit,usually with its own house, garden plot, and routine' (Pouwer 1964:141). Cooperation between them most commonly involved gardening or sagoprocessing. Families usually formed households whose adult membership comprisedbrothers or cousins and their wives or, where endogamy, bilocalityand/or uxorilocality were in operation, a married brother, his sisterand her husband. These units were important because they comprised twosets of producers one of which could provide a regular supply of food,when sickness or garden failure afflicted members of the other(Brumbaugh 1980: 30; Gell 1975: 10; Juillerat 1996: 92-99; Pouwer 1964:141). Although social relationships were weighted according to thesebilateral associations, even for the most mobile of populations in thestudy area residency was largely determined by a male's patrilinealassociation because this usually determined rights to land and thestructure of land guardianship (Juillerat 1996:128). A wife's ormother's father's land could be accessed but this wouldusually require approval by their close male relations (Brumbaugh 1980:60). The territory through which a community moved was superimposed witha mosaic of different land-holding histories connecting male members ofthe larger communal unit. The consequence of this system of land tenure was that almost allsocial groups for which data are present comprised patrilineagesresulting from the succession of male siblings maintaining residencyover a number of generations. In the USB these lineages were exogamousand often synonymous with the settlement. Nevertheless, they were by andlarge small and shallow, links to a common ancestor commonly went backonly three or four generations and almost all numbered below fiftypeople and commonly comprised less than twenty (Fyfe 2009:106-15).Therefore it is clear that the lifespan of these lineages was relativelyshort, no doubt largely due to high mortality, sibling rivalry and theconstant fissioning of communities. Clan structures were reported for the Border Mountain and somesectors of CNG, most commonly for settlements of more than sixty people.In most cases these were the same size or barely larger than the largestlineages and had very little function beyond providing an alliancestructure and performing the role as the exogamic unit where lineageswere particularly small (Gell 1975: 37-40; Juillerat 1996: 291. What isclear, however, is that the strength of affinal ties between lineages ofdifferent clans determined the composition of future clans after acommunity fissioned. Some Mountain Ok populations, most specifically Telefol and Tifalspeaking groups, lacked segmentation, and large cognatic descent unitsof approximately one hundred to two hundred people were reported; oftenwith members in several hamlets and villages (Brumbaugh 1980: 109;Jorgensen 1981:160). (13) Sometimes, as was often the case for Tifalspeakers, cognatic descent units were synonymous with the parish(Hyndman 1979: 53; Wheatcroft 1975: 29; 96). MARRIAGE Most information about marriage systems in New Guinea has come fromthe highlands where exchange tended to take place between relativelylarge segmentary cognatic descent units. Such systems included anemphasis on establishing and extending new ties countered by aprerogative for maintaining some ties to groups with whom there was ahistory of association (Barnes 1962; Glasse & Meggitt 1969;Strathern & Strathern 1969: 154). Large clans may have had a numberof exchange relationships, the majority involving neighbouringsettlements with whom they maintained alliances underpinned by ritualand military obligations. A minority of exchanges were distant and thesetended to have greater economic significance (Barnes 1962: 8). Where there were large cognatic descent units, these tended to besegmented and relatively endogamous, with taboos specifically preventingindividuals from marrying too close (Cook 1969: 105; Glasse 1969).Marriage was regulated according to a prescribed number of generationsor by a system of delayed and sometimes indirect exchange--andaccompanied by bride-wealth--in order to better distribute ties amongthe small agnatic units within the larger community (Strathern 1984;45-6; Strathern & Strathern 1969: 141). Where information has been provided for the sparsely populatedlowland areas of inland New Guinea, connubial patterns have beendescribed as localised and involute, whereby contiguous larger communalentities or small agnatic residential units are engaged in cycles ofreciprocal exchange of women. Neighbouring groups form a sphere of ties,the strength of which may lead to some residential aggregation of smallsocial units or movement of social segments between communities (e.g.Stasch 2003). Apart from the northern and eastern Border Mountains and parts ofCNG, there are mostly qualitative observations concerning theconstitution and boundaries of exogamic units, the preferred exchangemodes, the extent of marriage taboos, and information as to howrigorously the existing marriage rules were followed (Fyfe 2009:118-26).What is apparent from the data is that differences between marriagepatterns were a corollary of demographic and settlements patterns, whichwere strongly regulated by the environment. Exchanges could take placebetween proximal but autonomous communal units or between lineages orkindred units in villages or parishes. Where settlements or more mobilesocial aggregates were small (less than sixty persons), they appear tohave been almost exclusively exogamic, but where settlements were over ahundred individuals, the majority of marriages were endogamous. On the whole the small exogamic segments, the bilateral nature ofkinship, the loose definition or lax compliance to marriage taboos, andthe tenuous nature of communal unity contributed to a situation whereaffinal ties and filiative ties were somewhat superimposed. It appearsthe majority of marriages did take place within a relatively stablenetwork of ties, regardless of whether communities were composed ofclans, lineages or kindreds. Indeed, in many circumstances restrictedconnubial spheres may have been seen as not only more expedient but alsoas vital for maintaining security and demographic viability (Jorgensen1981: 223; Juillerat 1996: 291,298). Communities at some distance fromlinguistic boundaries seemed rarely to have had an opportunity, orreason, to undertake marriage exchanges with communities that spoke adifferent language. As indicated above, where there was someconcentration of marriage ties, these were commonly followed by someresidential aggregation of segments belonging to the communitiesinvolved. Essentially larger communities were constituted according tomatrilateral and affinal ties linking men. THE RELATIVE MOBILITY OF MEN AND WOMEN One important conclusion stemming from this investigation of theethnographic data was that, at least at the time the data werecollected, it is likely that the patrimony of men and women followedrelatively similar patterns of residency regardless of the rules ofmarriage. Essentially, women were exchanged between two exogamous unitseither within a communal unit comprising a single settlement orsettlement cluster, or exogamous units belonging to two contiguoussettlements or settlement clusters. As these units had a socialstructure with a strong bilateral emphasis, membership within such unitswas somewhat tenuous and with any social instability there was atendency for these units to fission. Subsequent emigration, asindividuals, family groups or kin group segments, was determinedaccording to ties that had been established by marriage prior tofissioning. Therefore any migration of men essentially retraced thepaths of their mothers. After migration, subsequent generations of men were able to gainrights to land near the communities where they were born and sometimesnew lineages were formed which in some contexts became an exogamous unitor part of a larger descent unit. These could be involved in exchangeswith contiguous groups and because of the bilateral structure ofrelationships, after a few generations they too could wane or split andrecombine, or be absorbed into another segment or group. In terms of this history, it is unlikely there would be muchdifference in the way that cultural practices belonging to men and womenwould have been patterned. Rather than residency, it would have been therelative mobility of women and men that would have determined anydifference in the way that each gender's cultural practices weredistributed through space. It can be argued that men's mobility demonstrated a differentspatial dynamic and range to that of women. In terms of factors such asritual and trade, there would have been several incentives for men toextend their social networks to groups outside their immediate affines(Fyfe 2009: 128-54). Importantly, male mobility started early. Indeedthe greatest amount of mobility between friendly communities isattributed to young men, or bachelors (Morren 1986: 242). In societieswhere bride-wealth was expected, many bachelors started their tradingapprenticeship largely due to the need to accumulate or ensure their ownbride-wealth (Brumbaugh 1980: 42). At maturity, men were expected todemonstrate some facility for securing ties in other communities and aman's social worth was sometimes measured according to the numberand value of these allegiances. Ethnographies suggest that communitiestypically contained men of remarkable sociability who ended up havingreciprocal partners in many villages and were instrumental infacilitating and preserving ties between more distant communities; inmany cases there were no marriage relations between these communities(Bercovitch 1989: 258; Brumbaugh 1980: 41; Juillerat 1996: 225). Alongwith trade items important men conveyed ideas and knowledge acrosssocial boundaries that most other members of their communities rarely ifever crossed. Women's mobility, both before and after marriage, seems to beentirely different. As Litteral (1978: 28) pointed out for the Anggor ofthe Border Mountains, women 'were not direct participants in theintertribal communicative network'. Indeed, Bercovitch observedthat among the Atbalmin 'the majority of women have never travelledfurther than a day's walk from their local..... valley [while] themajority of men have travelled many times to areas at least four or fivedays' walk away'. A woman's network of friends would havealso been much smaller, rarely extending beyond close kin (Bercovitch1989: 282). An endogamous woman would have very little reason to travelbeyond her own village unless relations were bad with both her husbandand consanguines, thus convincing her to flee her village (Huber1974:178). It is also important to point out that exogamous women oftenmaintained a strong relationship with their natal village, mostespecially through the enduring and complementary relationship they heldwith their brothers (Juillerat 1996: 341). Indeed, many seemed to havereturned to their natal village and even remarried there when theirhusband died or the marriage failed in their affinal community(Juillerat 1996: 295). This would have restricted somewhat the socialconnections between exogamous and local women and inhibited transmissionof ideas between them. Consequently, women's cultural forms and theideas behind their conception would more likely be restricted to smallerspheres of contiguous communities, especially considering the relativelow level and scope of regional trade (Fyfe 2009:128-41). Therefore, ifmaterial culture is able to provide clues to historical relationshipsbetween populations it would seem more likely that these would be foundin artefacts generated by women. MATERIAL CULTURE Around 11,000 objects made up the Project's larger dataset.These collections, the approximate number of objects they hold, and themuseums in which they are housed, are given in Tables 3 and 4. (14)Approximately fifty five percent are from the USB and Border Mountainsand forty five percent from the highlands of CNG. (15) The collectioncame from 237 settlements/location points across the study area. The sample includes all manner of objects from the most basicutilitarian tools to highly elaborate paraphernalia used in ritual. Manyutilitarian classes were common throughout the study area because ofregional similarities in hunting, gathering and horticulture. Theseinclude readily made items such as diggings sticks, bamboo knives andfire tongs to technologically complex items such as stone adzes, stringbags and arrows. Throughout the study area there was also little in theway of clothing. The woman's skirt, of which there were variousforms, was the major form of clothing in the study area, although in CNGa type of bark cape was worn by women for protection against wind andrain. The men, on the other hand, had no major form of body coveringother than the phallocrypt, which functioned simultaneously as aprotective device and a form of ornament or insignia. While there is little in the way of clothing there is a significantassemblage of body ornaments and personal items made with such materialas feathers, seeds, bones, dogs' teeth, pigs' tusks andshells. These sometimes required minimal manipulation of materials whilesome were intricately fashioned. Where such ornaments consisted ofvaluable and exotic items such as teeth and shells, they were sometimesused for marriage payments. It is also important to note that thefoundation for many body ornaments was provided by looped string bandsmade by women. Two important personal items included the smoking-gourd and bamboosmoking tubes, crafted locally by men for the consumption of tobaccothroughout the study area. The other important piece of apparatus forintoxication, ubiquitous throughout the lowlands, was the locallycrafted lime-gourd for carrying lime used in chewing the areca('betel') nut. These were ubiquitous throughout the lowlandsand were also crafted locally. As for musical instruments, hand drums and jaw harps were foundthroughout the study area. Hand drums were hourglass-shaped and madefrom immature tree trunks hollowed out with the use of embers andgougers. Their distal ends were decorated with carved and painteddesigns and their tympana made of lizard skin tuned with wax knobs. Theonly other musical instrument of significance was the wooden trumpetfound in some parts of the USB and the Border Mountains. Hand drums andtrumpets were used by men during ceremonies, while jaw harps, made froma short section of bamboo, were usually played by men and boys duringtimes of leisure. (16) Apart from the bow and arrow, carved and painted wood shields wereused in warfare throughout most of the study region. They ranged in sizeand shape and were used as a defensive shelter from arrows rather thanto parry (Beran & Craig 2005: 75-7; Craig 1967; 1970; 1976; 1988:31-44; 2005a: 117-26. Where the shield was absent, especially throughoutthe Border Mountains, the cuirass (rattan body armour) was used forprotection (Tiesler 1984). In CNG some groups used both shields andcuirasses. As far as the visual arts are concerned, the region lacked thefigural traditions and the monumental architecture and sculpture foundin the lower and middle Sepik regions. The decorative programs found inthe Upper Sepik are predominantly geometric and were carved and paintedwith mineral and organic pigments. The common media for decorative artwere arrows, hand drums, shields and, in CNG, architectural facades(house boards) with designs similar to those of the shields (Craig 1970;1988; Cranstone 1967; 1968). The other important media for painting tradition were masks andplaques used in sickness-curing and sago-fertility rituals performedthroughout the USB and Border Mountains. Masks often consist offlattened sheets of sago petioles or coconut palm fibre, and werepainted with a range of natural pigments. They were commonly attached toconical frames constructed of bamboo or rattan. In the Border Mountains,the designs tended to be more figurative than those in the lowlands(Gell 1975: 174-5; Gell 1992: 130-31; Juillerat undated; Juillerat1992b: 27-42). Other forms of decorative art can be found on gourd phallocrypts,gourd lime containers and bamboo smoking tubes. These were eitherpyrograved or incised with vaguely figurative or geometric patterns andstained with organic or mineral pigments (Craig 1988; 1990; 2005b). TWO CLASSES OF OBJECTS SELECTED FOR ANALYSIS: STRING BAGS ANDARROWS There were several reasons why string bags and arrows were seen asoffering good potential for the identification of socially significantpatterns in material culture distributions. Firstly, they were both ubiquitous to the study area and crucial tothe most important activities undertaken by men and women: gardening,hunting and fighting. Secondly, string bags and arrows were important personal andcultural items as well as being particularly visible (Barth 1975: 68,160-61, 188,233; MacKenzie 1991: 136-38). Whenever a man left hissettlement he carried his bow and arrows; both men and women carried apersonal string bag during daily activities (Gell 1975: 142-43;Jorgensen 1981: 68; MacKenzie 1990: 94-5). Essentially, they wereexcellent vehicles for individuals to express and measure socialdistance. Thirdly, these classes include structurally and technologicallymore complex artifacts than most others found in the region. Theirmanufacture was labour intensive and a single example may have had anumber of technical features, each of which required different skills tocomplete. They were also often highly elaborated, either with intricatearrangements of applied materials, such as feathers, shells, or seeds,or with patterns, such as those carved into the heads or foreshafts ofarrows, or looped into the fabric of bags using string dyed with naturalpigments. Therefore they potentially offer a wealth of formalvariability that can he used in this kind of comparative exercise. Lastly, as they were made exclusively by either men or women, theconceptual and technical qualities apparent in these objects potentiallyreflect the different social prerogatives and transmission paths of eachgender, therefore enabling an additional means of determining socialfactors affecting cultural variability. String bags Throughout the study area, the creation of string bags is theprimary objective of the craft of looping. String bags provide aflexible and portable container for which there is no comparablesubstitute. Some large bags are specifically made to carry gardenproduce, firewood and babies. Other smaller bags are designed to holdmore personal possessions such as tools, smoking tubes and, in thelowlands, areca nut chewing paraphernalia. In CNG personal bagsbelonging to men are often provided with feathers to signify theirinitiation status. The smallest range of bags are a form of amuletcontaining magic charms worn to provide success in hunting andgardening, to cure illness or even to secure a sexual partner. All string bags within the study area are constructed withhand-spun string made from the treated inner bark (bast) of saplings. Inthe lowlands, fibre from the bark of the tu-lip (Gnetum gnemon) isreported as having been the most commonly used material (Kelm & Kelm1980: 178; Kooijman 1962: 21), although Phaleria sp. has been reportedfor some groups in the Border Mountains (Juillerat undated). In thehighlands, fibre traditionally comes from Ficus sp. (Hyndman undated;MacKenzie 1991: 69-70). Before the spinning process, the bast could either be treated bysmoking or by soaking in water for a period of time. In the highlands,it was the women who treated and shred the bark while in the lowlandsthe bast was sometimes beaten and prepared by men (Kelm & Kelm 1980:178; MacKenzie 1991: 73-4). To make the string, the treated fibre wasrolled and twisted with the flat of the hand on the upper thigh (Kelm& Kelm 1980: 178; MacKenzie 1991: 78-9). Across the study region, there are numerous techniques used to formstring bags. Maureen MacKenzie's fieldwork in CNG has provided themost comprehensive body of work from the study area concerning stringbag production and the variability of techniques in the region (1990;1991). Characteristics MacKenzie identified as diagnostic of particulargroup preferences include: * the nature of bag components (e.g. mouthbands, edgings and straptypes); * the method of construction and co-joining; * bag shape or form. Following from MacKenzie's work, lowland bags were assessed ina similar manner (Bolton & Fyfe forthcoming; Fyfe 2009: 198-203). Inline with MacKenzie's findings, nine different looping techniqueswere identified as being used in the study area to construct the fullrange of bag components. Additionally, there were some different ways inwhich straps and mouthbands were attached from those discussed byMacKenzie. It was found that some techniques were more suited to aparticular range of bag sizes but there was significant overlap. Arrows As string bags were for women's craft, the arrow was thecenterpiece of male craft. The degree of attention and effort that mendedicated to the creation of arrows was considerable (Craig 1988: 47;Jorgensen 1981: 68; Morren 1986: 272-73). High quality arrows conferredprestige on their makers and owners, and were readily sought out by menand sometimes traded and gifted between men from different communities(Bush 1985: 257; Cranstone 1990: 38; Jorgensen 1981: 69). Arrows can be divided into functional sub-classes, most of whichare ubiquitous to the study area. Major sub-classes of arrows can bedetermined according to whether arrows have a single pointed tip,multiple tips (prongs) or are blunt (percussive); the latter two werereserved for small animals. For this analysis only single tipped arrowswere used as they were by far the largest part of the sample and more orless operationally analogous. Single tipped arrows have two or three of the following structuralcomponents: * a 'head' or 'blade' that varies in size andshape according to the prey and the type of wound intended; * a weighting device; * a shaft. As the arrows are not equipped with fletching to stabilise theirflight they are long, relative to fletched arrows, and weighted towardsthe tip. The most commonly used methods of weighting involved either: * using a foreshaft made from palmwood; * binding a small stone to the base of the arrow blade; * applying a paste made from marl or lime to the binding whenjoining the shaft to the head; or * adjusting the length of the head, where the head was made of ahard heavy wood. The arrow shafts were made with a cane from reeds or sword grass.(17) They were cut to length according to the types of head that theywould carry and the mode of weighting used. Arrowheads were made of bamboo or a hard wood, typically one of themany varieties of palmwood that grow in the lowlands and lower slopes.(18) Bamboo arrowheads are lancet or blade-like and have cross-sectionaltendencies that produce either a deep piercing or a gash-like wound.(19) Throughout the study area, bamboo-blade arrows were most commonlyreported as being intended for war, or for hunting pigs and cassowaries.Those with thick, heavily beveled culm sections were invariably reportedas fight arrows in the lowlands while sharp wide and open blade arrowswere more commonly reported as being intended for pigs and cassowaries.In the highlands, on the other hand, where thick, beveled bamboo-bladearrows were less common, semi-tubular and open blades were both reportedas fight arrows. (20) A feature of the bamboo-blade arrows from thestudy area is that sometimes barbs are cut into one or both edges of theblade of those intended for fighting. Palmwood arrowheads have the greatest diversity in morphology.Palmwood heads can comprise a simple tapering length of round or ovalcross-section, or be carved with a range of barb forms andcross-sections. Barbs follow a range of tendencies. Sometimes the wholelength of the head is carved with barbs while some arrows have just oneor two barbs near the end. In other cases heads are incised so that thetips will break leaving a portion in the victim's body. (21) The third most significant class of arrowhead consists of palmwoodheaded fight arrows with a detachable bone tip; these tips are intendedto stay in the body once the arrow has been removed to cause infectionin the wound of the victim. (22) The most important technical features used to secure the variousarrow components are the bindings. These are composed of braided, wound,or knotted strips of plant material. The number and position of bindingsare determined by the structure of the arrows. The largest range ofbinding positions, and also techniques, is found on bamboo-blade arrowswith foreshafts. These usually have at least two, and sometimes up tofive, independent bindings that vary according to their function andtheir position on the arrow. There is also a range of possible bindingtechniques for each position and it is apparent that while variation inbraiding, whipping or knotting is somewhat determined by choice, thenature of the head and the structure of the arrow to some degreedetermine the method of binding. For example, wide blades were morelikely to be fastened with whippings using straight or alternatingwinding techniques while, where narrower arrowheads and foreshafts havebeen used, and inserted into the shaft, the juncture may be reinforcedwith a small braided ring on the shaft that acts as a ferrule. ANALYSING THE MATERIAL CULTURE ATTRIBUTE TRAITS The preliminary analysis involved languages for which both stringbags and arrows were well represented. Included were all six CNGlanguages on the Papua New Guinea side of the border, two adjacentgroups from the lowland plains, Abau and Namie, and the contiguouslanguage of the Border Mountains, the isolate Yuri. There were 397 bagsand 1,389 arrows in the sample. To create the dataset each arrow and string bag was initiallyconsidered in mechanical terms. This led to the identification of acomplete set of possible components belonging to the two classes'samples which provided a set of possible 'attributes' arrangedaccording to a logical order of functional contingency: a particularcombination of attributes was seen as indicative of discrete functionaland operational properties. Each attribute level was then assessedtechnically and a set of variants was established for each. Thisprovided values to assign to each artefact where the relevant attributeswere present--a more detailed account of this process is given in Fyfe(2009: 176-214). Separate analyses were then carried out for each string bag andarrow attribute level. Assessments were made by comparing the relativeposition of each language in terms of material culture similarity tothat of their respective linguistic relationship, shown in the languagematrix (Table 2). Assessment of the effect of distance was determinedfrom a comparison with the languages' respective geographicalpositions (Map 3). CORRESPONDENCE ANALYSIS The first analyses focused on counts of attribute states wherebycorrespondence in the proportions of counts would be the measure ofrelatedness. Correspondence analysis (CA) was chosen because it isconsidered a useful tool to explore the association between discrete(categorical) values, and therefore ideal for quantifying the formal andmodal attributes derived from the sample of arrows and string bags(Sherman 1988: 283-88). CA both quantifies association between discretecategories by measuring relationships between rows and columns intwo-way or multiple-way frequency tables. This enables the establishmentof scaled distances between categories according to their relativescores along the sequence of corresponding rows and columns. As expectedvalues are those in which rows and columns demonstrate completeindependence from one another, the sum deviation from expected valueswill give an [x.sup.2] value. Dividing [x.sup.2] by the number in thesample (n) provides a measure of this deviation. CA provides a means ofextracting several factors and their dimensions which individuallyaccount for a percentage of this deviation: the two dimensionsaccounting for the greatest percentage of this inertia can be used asaxes to construct a map onto which the column or row values can beplotted thus providing a visual representation of the relationships. ANALYSIS OF VARIANCE The quantification of metric data involved the use of analysis ofvariance (ANOVA) which is appropriate for exploring relationshipsbetween metric variables and discrete variables (Foster et al. 2005:16ff). These tests would determine whether there was any significantdifference between language group means in both the string bag and arrowsamples. Analysing variance within the sample is done by comparingmeans' differences with variation due to random error (within-groupsum of squared deviations (SS)). In such tests, overall variance for thesamples' dependent variable(s) mean is calculated as SS from theoverall mean, divided by the number of the sample minus one (n-1).Adding together the sums of squares within each group provides the Errorvariance, and subtracting this from the overall SS provides the MeansSquare Effect. Under the null hypothesis, within group andbetween-groups variance should be statistically equivalent; an F test isused to compare the two estimates of variance to establish whether thereis a significant difference between groups. Aside from the determination of significant difference betweengroup means, another important feature of an ANOVA is that if the testsproved to be significant, the dependent variables can then be submittedto post-hoc tests that provide pair-wise comparisons of groups'means. These result in homogeneous subsets of groups whose means are notsignificantly different under the terms of the chosen [alpha]-level.(23) RESULTS: CORRESPONDENCE ANALYSIS The most significant trend in the analyses was the persistentrelationship between the language groups' geographical positionsand material culture attribute variability; essentially, lowland andhighland languages clustered according to their relative positions ineither of the study area's major geographic regions. (24) Therewas, however, only minor evidence for a relationship between languageand material culture for either class. Nevertheless, there weredifferences in the degree to which attributes from either classexhibited a relationship to distance suggesting that techniquesresponsible for these attributes have been differentially transmittedaccording to the range of social relationships and contexts thatcommonly exist for individuals living in these societies. Mostimportantly, in terms of factoring in the possible effect of differencein mobility between men and women, attribute levels of each classexhibited different patterns in the way language groups clustered: arrowattributes demonstrated a more general relationship to the geographicdistribution of language groups and tended to form fewer and lessdefined clusters while string bag attributes, particularly in thehighlands, exhibited more discrete clustering that better reflected thegroup's geographical positions. Arrows The following three CA plots are illustrative of patterns found inthe majority of plots associated with arrow attributes. The firstconcerns cross-section tendencies found for carvedpalmwood-head/bone-tipped arrow head-cores. A 'core' in thisregard refers to the part of the head prior to the tapering that resultsin the point. Cores can also have a lot to do with the primarydifferences between the barb carving techniques where barbs exist alongthe head and point. Two basic tendencies associated with carving arewhether barbs are cut into a broad pre-shaped core, a process thatleaves barb-like incisions, or whether the barbs are a result of thecraftsman cutting back an un-shaped core at the end of a shaped stem, aprocess that leaves barbs projecting from a narrower cylindrical core.There was a greater preference for the former in the highlands while thelatter was the preferred process in the lowlands. Where barbs were notpresent there was a smaller range of cross-sections. In all, eight corecross-sections were identified for the sample and included in theanalysis (Fyfe 2009: 207). As can be seen from plot (Figure 1) associated with thesecross-section tendencies, the CNG groups are clustered to the right handside with lowland groups spread out to the left. There is no furtherpattern to be determined for the highland groups but, in line with thegeographical situation, the Abau are nearest to this highland cluster,with the Yuri positioned nearest to the Abau; the relativelygeographically and socially remote Namie (Mitchell 1975) are positionedsome distance from the rest. The second plot (Figure 2) concerns bamboo-blade arrow structuresof which three were present in the sample: arrows with foreshafts,arrows with cane shafts attached directly to the head, and arrows withpure palmwood shafts. As can be seen, this plot produces an even denserhighland cluster, one largely resulting from a shared preference forforeshafts. Again there is no suggestion of additional divisions withinthis cluster and lowland languages remain positioned relatively as theydo in geographic space. The most distinct outlier is the Mianmin which,although close to the other Mountain Ok languages along Dimension 1,demonstrate a divergence along Dimension 2 due to a number of purepalmwood-shafted arrows collected from this group. [FIGURE 1 OMITTED] The third plot (Figure 3) shows relationships determined for thepresence and types of binds placed on the head/foreshaft near thejuncture with the shaft: a binding position that is used as a bufferingdevice; or as a decorative device derived from a buffering antecedent(Fyfe 2009: Appendix 27g, BIND B). This feature is common for palmwoodheaded arrows which, unlike bamboo-blade arrows, always have the headdirectly inserted into the culmpith cavity and therefore have greaterneed for such a buffering device. Importantly, it usually involvedcomplex binding techniques involving braiding. In this regard mosthighland samples have braids almost exclusively at this position while asignificant proportion of lowland arrows with this bind have those thatinvolve whipping and knotting. The Mianmin sample falls between thesetwo extremes while Oksapmin arrows are rarely equipped with a bind atthis position. [FIGURE 2 OMITTED] While this plot again demonstrates strong correlation with therelative geographical positions, there is arguably a possible signal forlanguage when one considers the minor divergence of the Mianmin andrelative isolation of the Oksapmin group from an otherwise tighthighland cluster. There was one sequence of bamboo-blade arrow attributes for whichdiscrete clustering within the highland group was evident, although itdid not reflect linguistic affinities or any specific geographicalfactor beyond relative contiguity. This persistent clustering ofhighland groups was due to the presence/absence of a bamboo-blade arrowtechnology whereby blades were slotted into a notch carved into thepalmwood foreshaft (Fyfe 2009: Appendix 26f, BBA-F-BATT 2), a featurealmost absent from all other groups in the study area who typicallybound the blade stem onto the outside of the foreshaft/shaft. The plot shown on page 146 (Figure 4) concerns affinitiesdetermined by shared tendencies in bamboo-blade cross-sections.Bamboo-blade arrows had either a broad, fine and flat open blade; athick strongly beveled blade; or a semi-tubular blade facilitated bybamboo cane with smaller diameters. Notched foreshafts, which arguablyresulted in a more secure blade attachment, were particularly practicalwhen using the former. Therefore greater proportions of fine and flatopen blades were associated with the distribution of this technology. [FIGURE 3 OMITTED] [FIGURE 4 OMITTED] While the plot again shows lowland groups spread out and set awayfrom those from the highlands, two distinct clusters are evident for thehighland groups. One, consisting of those who share the notchedforeshaft technology--the northern Central New Guinea language groups,Telefolmin, Oksapmin and Mianmin--is positioned in the top right handcomer of the plot. The other, for which the technology is relativelyabsent, is shown at the bottom right hand comer, and includes threeneighbouring groups all of which border the southern flank of thecordillera, Tifal Faiwol and Bimin. This division of highland groups is intriguing, for the groups thatshare the technology also had a significant degree of mutual enmity withlittle in the way of exchange taking place between them (Morren 1986:271ff; Perey 1973: 112). Importantly, the linguistic distance betweenthem is as great as possible for pairings of highland languages. It ispossible that interest in each other's arsenal lead to thisdistribution although their regional geographical position is animportant factor as the notched foreshaft innovation is also found forthe adjacent Asabano (Duranmin) population to the northeast and groupseast of the Strickland River (Roscoe pers. comm. 2008). (25) String bags Like the arrow analyses, string bag plots persistently positionedlowland languages apart from highland groups, and most commonly inrelation to their positions in geographic space. Unlike those forarrows, however, string bag plots exhibited a persistent clustering ofhighland groups in line with a south-east/north-west division of CNG.This division is important as there are significantly more topographicalobstacles between the south-east and north-west regions than there arebetween groups within these regions. Not surprisingly there is alsogreater ethnographic evidence for interaction, involving trade,marriage, ritual co-participation, or conflict, between groups withinthese clusters than between groups belonging to either cluster (Fyfe2009:passim). (26) The first CA plot (Figure 5) for string bags involved the waystring bags were constructed. Four basic methods were discerned andthese concerned either: the looping of a bag up from a chain or cord, bystarting directly from a looped row and spiraling up; looping a cylinderthen sealing the bottom; or looping a panel that was the sealed on thesides (Fyfe 2009: Appendix 24b). It was thought that the patterning ofthese fundamental processes of creating the bags' basic forms, eachof which in essence encompasses an extended procedure relying on arelatively inscrutable set of motor skills and motor adjustment, wouldbe more likely to reflect deeper social relationships as they wouldconceivably require more intensive learning techniques (e.g. Pryor andCarr 1995). Indeed, this appeared to be the case as the plot, more thanothers associated with string bags, conforms somewhat to the linguisticpicture: the Namie and Abau are now proximal, with the Yuri and Oksapminpositioned further towards the fringes. The next sequence of plots is illustrative of the typical patternfound for string bag attributes. The first represent group relationshipsaccording to choice of looping methods used for string bag bodies. Therewere four methods identified in the sample and a description of these isgiven by MacKenzie (1991: 215-217, Appendix 2a) and Bolton and Fyfe(forthcoming). As can be clearly seen in the plot (Figure 6), the above mentioneddivisions between the south-east region (Faiwol, Bimin and Oksapmin) andthat of the north-west region of CNG (Telefol, Tifal and Mianmin) arevery apparent. Importantly, the lowland groups remain somewhat isolatedwith the Namie now positioned well away from the rest. The onlydeviation from an otherwise unequivocally geographical orientation isthe positioning of the Yuri among the north-west CNG cluster. This stronger relationship between material culture affinity andgeography is more asserted in the following two plots. The firstconcerns looping techniques found for mouth finishing: edgings andmouthbands (Figure 7). Two edging and four mouthband looping techniqueswere found in the sample; the most common mouthband looping techniquewas present for all groups (Fyfe 2009: Appendix 25c). Significantdifferences between the highland clusters largely concerned the choicebetween the use of either mouthband or edging. The sample from thesouth-east cluster included a greater amount of bags, both large andsmall, with edgings and included an additional edging technique.Differences between the north-west and lowland samples were also due todifference in the relative proportions of edgings and mouthbands as wellas mouthband techniques. [FIGURE 5 OMITTED] [FIGURE 6 OMITTED] [FIGURE 7 OMITTED] The second plot concerns the manner in which straps were attachedto the bags (Figure 8). Seven different techniques were identified (Fyfe2009: Appendix 25f). Five of these were used in the case of mouthbands,and all five were used in the case of the most common mouthband loopingmethod. Four techniques were used where edging or no mouth finishing waspresent. As can be seen, in both plots not only are the south-east andnorth-west divisions within CNG maintained but the Yuri, Abau and Namieassume their relative positions away from the highland clusters. Clearlythese plots suggest that the distribution of string bag attributes moreaccurately reflects geographic positions, at least within CNG. One maybe tempted to argue that some of this difference between arrow andstring bag attribute patterns may be due to arrows being more regularlytraded over greater distances but there is little to suggest that thiswas the case. Besides, this would still not detract from the implicationthat men were more likely to travel to and beyond contiguous communitiesand exchange information concerning material culture in the process.Therefore this difference must be due to women being restricted tosmaller spheres of interaction. [FIGURE 8 OMITTED] RESULTS: ANOVA Arrows In terms of evidence for a relationship between material cultureand ecology the greatest was found in the relationship betweenaltitude/subsistence patterns and arrow length/head weighting. Asdiscussed above, homogeneous subsets of related means can be created aspart of the ANOVA procedure. Table 5 provides homogeneous subsets oflanguages with related means for bamboo-blade arrow lengths with [alpha]= 0.1 ; similar patterns were found for palmwood-head arrow and headlengths (Fyfe 2009: Tables 20.7-10). In simple terms, high altitudegroups invariably had shorter and lighter arrows than low altitudegroups, with the mid-altitude fringe Mianmin having mean lengths andweights between those two; although across all arrow classes the Yurivaried in having arrows/arrowheads either slightly shorter or longerthan the Abau. The Sepik languages, Abau and Namie, were closer to eachother in this respect than to any other language, but rather thanconstituting evidence of any genetic relationship, this associationlikely reflects similarity in lowland subsistence strategies. What must be considered when interpreting such results is that,firstly, highland peoples in New Guinea are commonly shorter in staturethan those in the lowlands. For one thing, this is partly due to dietarydifferences, the most important of which is the normally smaller proteinintake for highland populations (Dennett & Connell 1988: 275).Secondly, people of the lowland tropics tend to have longer limbs tofacilitate better regulation of body heat (Eveleth & Tanner 1976:271). But what is additionally clear in these data is that sweetpotato-dominant highland groups are most affected in terms ofphysiological restriction, suggesting that the greatest proportion ofsweet potato uptake has not resulted in a channelling of surplus to pigproduction but occurred as an adjustment to the decline of taroproduction due to environmental degradation and other demographicfactors. The Tifal speaking groups are third highest in this tendency.As pointed out by Fyfe (2009: 38), the ethnographic evidence suggeststhat the Atbalmin, a large Tifal speaking population nearest the borderwith [West] Papua, was tending towards a sweet potato dominant regime.The data therefore confirm such observations. They also suggest that theleast horticulture-intense, sago exploiting societies are the bestnourished, although those living in the hilly environment of the BorderMountains, who have ready access to both sago and viable land forhorticulture, would be very close. String bags Given the great variance in size within the string bag sample, andthe obvious relationship between bag size and function, the relationshipbetween size and language was not considered for analysis in the mannerfor arrows. Rather boxplots of bags' sizes were created for bothlanguage and string bag width and height ratios to gauge proportionaltendencies and determine whether sampling problems had resulted in theunequal distribution of possible functional subclasses across languagesamples. A boxplot for size to language (Figure 9) revealed that therewas considerable difference in bag size between language groups as wellas a number of outliers. Clearly, irregular sampling had occurred across the study area anda number of functional subclasses were likely to be poorly representedfor some languages; this was most evident for Faiwol and Oksapminsamples. The boxplot for bag construction method (CM) and size (Figure10), two variables likely to be strongly related, indicated similarproblems although it was apparent that two out of the four CMs were morelikely to be used for much smaller bags. A boxplot (Figure l1) for CM, this time involving height to maximumwidth (HTMXW), indicated that bag proportions did not vary greatlyaccording to the way that the bags were constructed. Nevertheless greatvariance was found for CM 2, which was understandable because theproportions of very small bags, such as the amulet bags for which thismethod has been commonly used, are not overly constrained by functionalconcerns, indeed curious shapes are commonly used for such bags (e.g.Fyfe 2009: appendix 20e). [FIGURE 9 OMITTED] [FIGURE 10 OMITTED] [FIGURE 11 OMITTED] [FIGURE 12 OMITTED] A second boxplot for bag construction method and proportions(Figure 12), this time involving minimum to maximum width (MNTMXW),revealed the existence of a similar lack of constraint for CM 2 while atthe same minor variance in dimensions for CM 4, a method also used forsmall pocket bags. In this case the lack of variance for CM 4 is afunction of the construction method as it involves an initial panel oflooping followed by folding and sewing of the sides and therefore littlescope for the proportions to diverge from a one to one ratio. The first solution to the problem of testing the sample for anysignificant variance between groups was therefore to remove CM 2 and CM4 from the sample. The second solution was to remove significantoutliers for CM 1 and CM 3 evident in the first boxplot (Figure 9) andin the second (Figure 10). It was decided to leave in outliers for CM 1and CM 3, apparent in the third (Figure 11) and fourth (Figure 12)boxplots, for they were not as extreme as the others and less likely dueto functional difference. This left a sample size of 305, a reduction of88. Considering that the sample had been reduced to belonging to eitherof two structural forms associated with construction method it wassurprising to find significant non-random patterns apparent in the posthoc tests for between group differences concerning MNTMXW means. Therelative differences in between group means appeared to be stronglyrelated to distance and groups at the extreme ends of the distancescale--Namie, Oksapmin and Bimin--demonstrate the greatest differencefrom overall group comparisons. The table for post hoc homogeneoussubsets again demonstrates the same lowland and highlands clusters asfound for CA using nominal variables (Table 6). At an [alpha] -level of.05, subsets again reflect the south-east and north-west clusters thatwere observed for many of the string bag CA plots. Beyond these clustersit is important to note Subset 3 which includes all languages of theMountain Ok sub-family. This result was compelling because it revealed a significantpattern for a continuous variable associated with bag proportions thatwas not explicable to factors such as function and ecology (physiology).Rather it was one that was clearly related to geographical and socialdistance. Clearly bag shape was directed by socio-cultural factors ashas previously been suggested by MacKenzie (1991 : 33). CONCLUSION Results for the Project so far reflect the findings of theresearchers at the Field Museum for the New Guinea north coast data inthat distance appears to be a stronger factor in the distribution ofcultural traits than any particular relationship implied by languageaffinity. Overall, the transmission of a full range of attributesbetween groups appears to have been relatively strong for both stringbags and arrows, and suggests that cultural differences between groupsspeaking closely related languages have largely emerged as a result ofaccess to innovations from other groups rather than because of anyinternal factors that one would associate with the process of drift. However, our findings go beyond those of the Field Museum. We havedemonstrated that a material culture dataset can provide a number ofdifferent patterns some of which can provide means to determine aneffect caused by more discrete social factors. We have done this byfocusing on technology and material culture classes whose variance hasimportant and arguably dichotomous social implications. The data haveshown that for men's culture, specifically in the form of arrows,the intensity of social ties is less important, as attributes appear tobe widely and indiscriminately disseminated. As concerns women'sculture, however, attribute patterns are clearly more reflective ofclose social relationships between groups: ones that would entail somepermanent movement of personnel between adjacent communities. Inconclusion, women's crafts are likely to be a more appropriate toolthan men's for determining the strengths of social and possiblyhistorical ties between populations. These observations should therefore have relevance to questionsconcerning Lapita. Clearly, material culture has been strongly shaped byinteraction in the study area but in different ways for men's andwomen's crafts. It would be interesting therefore to considerregional distributions of Lapita attributes in light of what has beenrevealed here. Do the spatial patterns of Lapita attributes suggest apervasive transmission of culture via large seafaring networks? Or, doLapita attributes tend to form smaller clusters composed of relativelycontiguous sites. For one thing Summerhayes (2000: 235) concluded thatLapita ware affinities across sites in West New Britain appear not toreflect broader exchange networks but rather more discrete 'socialprocesses' including 'spouse exchange'. If this is so,these findings from the Upper Sepik and Central New Guinea shouldprovide a compelling set of ideas for future Lapita research. ACKNOWLEDGEMENTS This paper is based on the data gathered by the Upper Sepik-CentralNew Guinea Project (USCNG). The Project was funded by an ARC-LinkageGrant LP0455756, which provided the APAI scholarship for my doctoralresearch. The USCNGP has been extended by a second ARC-Linkage grantLP0883050. The Linkage Partners for both grants are the South AustralianMuseum and Ok Tedi Mining Ltd. I wish to acknowledge the significantinput of Barry Craig who is the founder of the USCNG Project. I wouldalso like to thank the Principal Investigator for the Project, GraemeHugo, for his advice and encouragement. Thanks should also be given toMartin Steer, Andrew Pawley and Bryant Allen from the Research School ofPacific and Asian Studies, Australian National University, forlinguistic and ecological data. Special thanks are extended to JillBolton, Stephen Sherman, Malcolm Fyfe and Paul Roscoe REFERENCES ALLEN, B. 1983. 'Human Geography of Papua New Guinea',Journal of Human Evolution 12(1): 3-23. ALLEN, B.J., R.L. HIDE, R.M. BOURKE, W. AKUS, D. FRITSCH, R. GRAU,G. LING, and E. LOWES. 2002. Agricultural Systems of Papua New Guinea.Working Paper No. 4. Western Province. Rev. edn. Canberra: Department ofHuman Geography, Australian National University. ALLEN, J. 1984. 'In search of the Lapita homeland',Journal of Pacific History 19:186-201. 1996. The pre-Austronesian settlement of Island Melanesia:Implications for Lapita Archaeology', Transactions of the AmericanPhilosophical Society 85(5): 11-27. AMBROSE, W.R. 1997. 'Contradictions in Lapita pottery, acomposite clone', Antiquity 71 (273): 525-547. ANSON, D. 1986. 'Lapita pottery of the Bismarck Archipelagoand its affinities', Anthropology in Oceania 21: 157-6. BARNES, J.A. 1962. 'African Models in the New GuineaHighlands', Man 62: 5-9. BARTH, F. 1971. 'Tribes and intertribal relations in the Flyheadwaters', Oceania 41:171-91. 1975. Ritual and Knowledge among the Baktaman of New Guinea. NewHaven: Yale University Press. BELLWOOD, P. 1991. 'The Austronesian dispersal and the originof languages', Scientific American 265: 88-93. 1995. Comment on: 'Predicting similarity in material cultureamong New Guinea villages from propinquity and language: a log-linearapproach', Current Anthropology 36(5): 769-88. 1996. 'Phylogeny versus reticulation in prehistory',Antiquity 70(270): 881-90. BERAN H. and B. CRAIG (eds). 2005. Shields of Melanesia. Adelaide:Crawford House Publishing Australia. BERCOVITCH, E. 1989. 'Disclosure and Concealment: A Study ofSecrecy among the Nalumin people of Papua New Guinea'. PhD Thesis,Anthropology, Stanford University. 1994. The Agent in the Gift: Hidden Exchange in Inner NewGuinea', Cultural Anthropology 9(4): 498-536. BOLTON, J. and A. FYFE (forthcoming). Knots, Loops & Braids:exploring threads of history, gender & sociality in Upper Sepikfibre crafts. BOURKE, R.M., B.J. ALLEN, R.L. HIDE, D. FRITSCH, R. GRAU, E. LOWES,T. NEN, E. NIRSIE, J. RISIMERI and M. WORUBA. 1993. Agricultural Systemsof Papua New Guinea. Working Paper No. 3. West Sepik Province. Canberra:Department of Human Geography, Australian National University. BRAGGE, L. W. 1964. Patrol Report, Green River No. 7, 1963/64. BRONGERSMA, L.D. and G.F. VENEMA. 1962. To the Mountains of theStars. London: Hodder & Stoughton. BROWN, P. 1962. 'Non-agnates among the patrilinealChimbu', Journal of the Polynesian Society 71: 57-69. BRUMBAUGH, R.C. 1980. 'A secret cult in the West SepikHighlands'. PhD Thesis, Anthropology, State University of New Yorkat Stony Brook. BUSH, T. 1985. 'Form and decoration of arrows from thehighlands of Papua New Guinea'. Australian Museum Records 37:255-93. CHIU, S. 2003. Social and economic meanings of Lapita pottery: aNew Caledonian case, in Pacific Archaeology: Assessments and Prospects:Proceedings of the International Conference for the 50th Anniversary ofthe First Lapita Excavation, Kone Noumea 2002. Noumea:Nouvelle-Caledonie: Ddpartment Archdologie, Service des Musees et duPatrimoine, pp. 159-82. CLUNIE, N.M.U. 1978. Introduction, in J.S. Womersley (ed.),Handbooks of the Flora of Papua New Guinea. Melbourne: MelbourneUniversity Press, pp. 1-11. CONRAD, R. and W. DYE. 1975. 'Some Language Relationships inthe Upper Sepik Region of Papua New Guinea', Papers in New GuineaLinguistics No. 18, Pacific Linguistics; Series A 40. Research School ofPacific Studies, Australian National University. COOK, E.A. 1969. Marriage among the Manga, in R. M. Glasse and M.J. Meggitt (eds), Pigs, Pearlshells and Women : Marriage in the NewGuinea Highlands. Eaglewood Cliffs: Prentice-Hall, pp. 138-58. CRAIG, B. 1967. 'Houseboards of the Telefomin Sub-District,New Guinea', Man 2: 260-73. 1969. Houseboards and Warshields of the Mountain-Ok of Central NewGuinea: Analysis of an Art Style. M.A. Thesis, Anthropology, SydneyUniversity. 1970. Art Styles of the Upper Sepik. Unpublished Manuscript inpossession of author. 1975. The Art Style of Yellow River, West Sepik District, Papua NewGuinea'. Baessler-Arehiv 23: 417-45. 1976. The Keram Style in the Abau area of the Upper Sepik,PNG'. Baessler-Archiv 24:181-95 (with a 1983 postscript in Vol.31:351-55). 1988. Art and Decoration of Central New Guinea. Shire Ethnographypublication Nr. 5. Aylesbury, Bucks.: Shire. 1990. Relic and Trophy Arrays as Art Among the Mountain-Ok ofCentral New Guinea, in F.A. Hanson and L. Hanson (eds), Art and Identityin Oceania. Bathurst: Crawford House Press, pp. 196-210. 2002. Abau, in M. Ember and I. Skoggard (eds), Encyclopedia ofWorld Cultures. Supplement. New York: MacMillan, pp. 1-5. 2005a. Upper Sepik area, in H. Beran and B. Craig (eds), Shields ofMelanesia, Adelaide: Crawford House, pp. 75-89. 2005b. What can material culture studies tell us about the past inNew Guinea?, in A. Pawley, R. Attenborough, J. Golson and R. Hide (eds),Papuan Pasts: Cultural, Linguistic and Biological Histories ofPapuan-Speaking Peoples.. Canberra: Pacific Linguistics, Research Schoolof Pacific and Asian Studies, Australian National University, pp.493-513. 2008. Sorcery divination among the Abau of the Idam Valley, UpperSepik, Papua New Guinea', Journal of Ritual Studies 22, pp. 37-51. CRAIG, R. 1969. Marriage among the Telefolmin, in R. M. Glasse andM. J. Meggit (eds), Pigs, Pearlshells and Women : Marriage in the NewGuinea Highlands'. Eaglewood Cliffs: Prentice-Hall, pp. 176-97. CRANSTONE, B.A.L. 1964b. Field Notes of the British MuseumEthnographical Expedition to New Guinea. Archives of the British Museum,London. Unpublished Manuscript. 1967. Some boards from a New Guinea haus tambaran', Man 2:274-7. 1968. War Shields of the Telefomin Sub-district, New Guinea',Man 3: 609-24. 1990. The Tifalmin and their Environment: Material Culture of aMountain-Ok People, in B. Craig and D. Hyndman (eds), Children of Afek:Tradition and Change Among the Mountain-Ok of Central New Guinea.Sydney: Oceania Monographs No.40, pp. 35-53. DENNETT, G. and J. CONNELL. 1988. 'Acculturation and Health inthe Highlands of Papua New Guinea: Dissent on Diversity, Diets andDevelopment', Current Anthropology 29(2): 273-96. DIAMOND, J. M. 1988. 'Express train to Polynesia', Nature336: 307-8. EGGERTSSON, S. 2003. Transforming Skins: Kwermin Sequences ofGrowth. Reykjavik: University of Iceland Press. EGLOFF, B. 1975. Archaeological investigations in the coastalMadang area and on Eloaue Island of the St Matthias Group. Port Moresby:Papua New Guinea Public Museum and Art Gallery. EVELETH, P.B. and J.M. TANNER. 1976. Worldwide Variation in HumanGrowth. Cambridge & New York: Cambridge University Press. FOLEY, W. A. 1986. The Papuan Languages of New Guinea. Cambridge:Cambridge University Press. 2005. Linguistic prehistory in the Sepik-Ramu basin, in A. Pawley,R. Attenborough, J. Golson and R. Hide (eds), Papuan Pasts: Cultural,linguistic and biological histories of Papuan-speaking peoples; PacificLinguistics 572. Canberra: Pacific Linguistics, Research School ofPacific and Asian Studies, Australian National University. FOSTER, J.J., E. BARKUS and C. YAVORSKY. 2005. Understanding andUsing Advanced Statistics. Thousand Oaks: Sage. FYFE, A. 2009. Gender; mobility, and population history: exploringmaterial culture distributions in the Upper Sepik and Central NewGuinea. PhD Thesis, Geographical and Environmental Studies, Universityof Adelaide. GALIPAUD, J.C. 1988. The physio-chemical analysis of ancientpottery from New Caledonia, in M. Spriggs (ed.), Lapita Design, Form andComposition; Proceedings of the Lapita Design Workshop, Canberra,Australia, December 1988. Canberra: Department of Prehistory, ResearchSchool of Pacific Studies, Australian National University, pp. 139-42. GELL, A. 1975. Metamorphis of the Cassowaries: Umeda Society,Language and Ritual. London: The Athlone Press, University of London. 1992. Under the sign of the cassowary, in B. Juillerat (ed.),Shooting the Sun: Ritual and Meaning in West Sepik. Washington: Smithsonian Institution Press, pp. 125-43. GILLIESON, D. and G. HOPE. 1990. Development of the Mountain-Okenvironment, in B. Craig and D. Hyndman (eds), Children of Afek:Tradition and Change Among the Mountain-Ok of Central New Guinea.Sydney: Oceania Monographs No.40, pp. 27-34. GLASSE, R.M. 1969. Marriage in South Fore, in R. M. Glasse and M.J. Meggit (eds), Pigs, Pearlshells and Women : Marriage in the NewGuinea Highlands. Eaglewood Cliffs: Prentice-Hall, pp. 16-37. GLASSE R.M. and M.J. MEGGIT. 1969. Pigs, Pearlshells and Women :Marriage in the New Guinea Highlands, R. M. Glasse and M. J. Meggit(eds). Eaglewood Cliffs: Prentice-Hall. GOLSON, J. 2005. Introduction to the chapters on archaeology andethnology, in A. Pawley, R. Attenborough, J. Golson and R. Hide (eds),Papuan Pasts: Cultural, linguistic and biological histories ofPapuanspeaking peoples; Pacific Linguistics 572. Canberra: PacificLinguistics, Research School of Pacific and Asian Studies, AustralianNational University, pp. 221-33. GOSDEN, C. and J. WEBB. 1994. 'The creation of a Papua NewGuinean landscape: archaeological and geomorphological evidence',Journal of Field Archaeology 21(1): 29-51. GRAY, R.D. and F.M. JORDAN. 2000. 'Language trees support theexpress-train sequence of Austronesian expansion', Nature 405:1052-5. GREEN, R.C. 1978. 'New sites with Lapita pottery and theirimplications for an understanding of the settlement of the westernPacific', Working papers in Anthropology, Archaeology, Linguisticsand Moari Studies. Auckland: Department of Anthropology, University ofAuckland. 1988. Lapita design analysis: the Mead system and its use; a pottedhistory, in M. Spriggs (ed.), Lapita Design, Form and Composition;Proceedings of the Lapita Design Workshop, Canberra, Australia, December1988. Canberra: Department of Prehistory, Research School of PacificStudies, Australian National University, pp. 33-52. 1991. The Lapita cultural complex: Current evidence and proposedmodels', Bulletin of the lndo-Pacific Prehistory Association l1 :295-305. 2003. The Lapita horizons and traditions--signature for one set ofOceanic migrations, in Pacific Archaeology: Assessments and Prospects:Proceedings of the International Conference for the 50th Anniversary ofthe First Lapita Excavation, Kone Noumea 2002. Noumea:Nouvelle-Caledonie: Department Archeologie, Service des Musees et duPatrimoine, pp. 95-117. GREENHILL, S.J. and R.D. GRAY. 2005. Testing population dispersalhypotheses: Pacific settlement, phylogenetic tree and Austronesianlanguages, in R. Mace, C. J. Holden and S. J. Shennan (eds), TheEvolution of Cultural Diversity: A Phylogenetic Approach. London:University College London Press, pp. 31-52. HARRISON, S. 1985. 'Ritual Hierarchy and Secular Equality in aSepik River Village', American Ethnologist 12(3): 413-26. HEALEY, A. 1964. The Ok Language Family in New Guinea'. PhDThesis, Linguistics, Australian National University. HEIDER, K.G. 1970. The Dugum Dani: A Papuan Culture in theHighlands of West New Guinea. New York: Wenner-Gren Foundation ForAnthropological Research. HUBER, P. 1974. "Identity and Exchange: Kinship and SocialOrder Among the Anggor of New Guinea'. PhD Thesis, Anthropology,Duke University. 1979. 'Anggor floods: reflections on ethnogeography and mentalmaps', Geographical Review 69(2): 127-39.' 1980. The Anggor bowman: ritual and society in Melanesia',American Ethnologist 7(1): 43-57. HURLES, M.E., L. MATISOO-SMITH, R.D. GRAY and D. PENNY. 2003.'Untangling Oceanic Settlement: the edge of the Knowable',Trends in Ecology and Evolution 18(10): 531-40. HYNDMAN, D. 1979. 'Wopkaimin Subsistence: Cultural Ecology inthe New Guinea Highland Fringe'. PhD Thesis, Anthropology andSociology. University of Queensland. HYNDMAN, D.C. (undated). Untitled documents: collection notes heldat the Papua New Guinea National Museum & Art Gallery. UnpublishedManuscript. HYNDMAN, D.C. and J.I. MENZIES. 1990. 'Rain forests of the OkTedi headwaters, New Guinea: An ecological analysis', Journal ofBiogeography 17(3): 241-74. HYNDMAN, D.C. and G.E.B. MORREN. 1990. The human ecology of theMountain-Ok of Central New Guinea: a regional and inter-regionalapproach, in B. Craig and D. Hyndman (eds), Children of Afek: Traditionand Change Among the Mountain-Ok of Central New Guinea. Sydney: OceaniaMonographs No.40, pp. 9-26. JEFFRIES, A. 1950. Patrol Report, Green River No. 1, 1950/51. JONES, B. 1980. 'Consuming Society: Food and Illness among theFaiwolmin'. PhD Thesis, Anthropology, University of Virginia. JONES, S. 1997. The Archaeology of Ethnicity. ConstructingIdentities in the Past and Present. London: Routledge. JORGENSEN, D. 1981. 'Taro and Arrows: Order, Entropy andReligion among the Telefolmin'. PhD Thesis, Anthropology,University of British Columbia. 1983. Principal Telefol Gardening and Hunting Areas, in P. Swadling(ed.), How Long Have People Been in the Ok Tedi Impact Region. Recordsof the Papua New Guinea National Museum No.8. Waigani, Papua New GuineaNational Museum. p. 133. JUILLERAT, B. 1982. 'Note sur les rapports de production dansl'horticulture-arboriculture Yafar (Nouvelle-Guinee)', Journald'Agriculture Traditionelle et de Botanique Appliquee 29(3-4):285-93. 1992a. Introduction, in B. Juillerat (ed.), Shooting the Sun:Ritual and Meaning in West Sepik. Washington: Smithsonian InstitutionPress, pp. 1-19. 1992b. "The mother's brother is the breast": incestand its prohibition in the Yafar Yangis, in B. Juillerat (ed.), Shootingthe Sun: Ritual and Meaning in West Sepik. Washington: SmithsonianInstitution Press, pp. 20-124. 1996. Children of the blood: society, reproduction and cosmology inNew Guinea, trans. N. Scott. Oxford: Berg. (undated). Notice de la basede gestion des collections: Collection Bernard Juillerat; Musee du QuaiBranly, Paris. Unpublished Manuscript. KABERRY, P. 1967. The plasticity of New Guinea kinship, in M.Freedman (ed.), Social Organisation: essays presented to Raymond Firth.London: Cassell, pp. 105-23. KELM, A. and H. KELM. 1980. Sago und Schwein. Ethnologie vonKwieftim und Abrau in Nordost-Neuguinea. Wiesbaden: Franz SteinerVerlag. KENNEDY, J. 1981. 'Lapita Colonization of the AdmiraltyIslands?', Science 213 (4509): 757-9. KIRCH, P.V. 1990. 'Specialisation and exchange in the Lapitacomplex of Oceania', Asian Perspectives 29:117-33. 1996. 'Lapita and its aftermath: the Austronesian settlementof Oceania', Transactions of the American Philosophical Society86(5): 57-70. KOOIJMAN, S. 1962. 'Material Aspects of the Star MountainsCulture', Nova Guinea 10(2): 15-44. KUCHIKURA, Y. 1990. 'Subsistence activities, food use, andnutrition among the Mountain Ok in Central New Guinea', Man andCulture in Oceania 6:113-37. LAYCOCK, D.C. 1973. Sepik Languages: Checklist and PreliminaryClassification. Pacific Linguistics B 25. Canberra: Research School ofPacific Studies, Australian National University. 1975. Isolates: Sepik Region, in A. Wurm (ed.), New Guinea arealanguages and language study, vol 1: Papuan languages and the New Guinealinguistic scene; Pacific Linguistics. Series C. Books ; no.38 S,.Canberra: Research School of Pacific Studies, Australian NationalUniversity, pp. 881-86. LEPERVANCHE, M. de 1967. 'Descent, residence and leadership inthe New Guinea Highlands', Oceania 38: 134-158,163-189. LITTERAL, R. 1978. Changes in the Bibriari communicative system, inJ. Murane (ed.), Language Planning and Grammatical Typologies.Workpapers on Papua New Guinea languages 24. Ukarumpa: Summer Instituteof Linguistics, pp. 25-30. LOHMANN, R.I. 2000. Cultural Reception in the Contact andConversion History of the Asabano of Papua New Guinea. PhD Thesis,Anthropology, University of Wisconsin-Madison. MACKENZIE, M. 1990. The Telefol string bag: a cultural object withandrogynous forms, in B. Craig and D. Hyndman (eds), Children of Afek:Tradition and Change Among the Mountain-Ok of Central New Guinea.Sydney: Oceania Monographs No.40, pp. 88-108. 1991. Androgynous Objects: String bags and gender in central NewGuinea. Chur, Switzerland: Harwood Academic Publishers. MCCARTHY, J.K. 1936. Report of patrol to the Yellow River BaseCamp, Sepik Division. MITCHELL, W.E. 1975. Culturally contrasting therapeutic systems ofthe West Sepik: the Lujere, in T. R. Williams (ed.), PsychologicalAnthropology. The Hague: Mouton. pp 409-39. MOORE, C.C. and A.K. ROMNEY. 1994. 'Material Culture,Geographic Propinquity, and Linguistic Affiliation on the North Coast ofNew Guinea: A Reanalysis of Welsch, Ten'ell, and Nadolski(1992)', American Anthropologist 96(2): 370-92. MORREN, G.E.B. 1974. 'Settlement Strategies and Hunting in aNew Guinea Society'. PhD Thesis, Anthropology, Columbia University. 1986. The Miyanmin: Human Ecology of a Papua New Guinea Society.Ann Arbor: UMI Research Press. OHTSUKA, R. 1994. 'Subsistence ecology and carrying capacityin two Papua New Guinea populations', Journal of Biosocial Science26: 395-407. PEREY, A. 1973. 'Oksapmin Society and World View'. PhDThesis, Anthropology, Columbia University. PETER, H. 1990. Cultural changes in Gargar society, in N.Lutkehaus, C. Kaufmann, W. E. Mitchell, D. Newton, L. Osmundsen and M.Schuster (eds), Sepik Heritage: Tradition and Change in Papua NewGuinea. Durham, N.C.: Carolina Academic Press, pp. 245-50. PETREQUIN P. and A.M. PETREQUIN. 1990. Fleches de chasse, flechesde guerre: Le cas des Danis d'Irian Jaya (Indonesie). Bulletin dela Societe Prehistorique Frangaise 87:484-511. POOLE, F.J.P. 1976. 'The Ais Am: An Introduction to MaleInitiation Ritual among the Bimin-Kuskusmin of the West'. PhDThesis, Anthropology, Cornell University. POUWER, J. 1964. 'Social system in the Star Mountains, towardsa reorientation of the study of social systems', AmericanAnthropologist 66(4): 133-61. PRYOR, J. and C. CARR. 1995. Basketry of Northern CaliforniaIndians: Interpreting Style Hierarchies, in C. Carr and J. E. Neitzel(eds), Style, Society, and Person: Archaeology and EthnographicalPerspectives. New York & London: Plenum, pp. 259-96. RAPPAPORT, R.A. 1968. Pigs for the ancestors: ritual in the ecologyof a New Guinea people. New Haven: Yale University Press. READ, K.E. 1959. 'Leadership and Consensus in a New GuineaSociety', American Anthropologist 61 (3): 425-36. REINER, E.J. and R. G. ROBBINS. 1964. 'The Middle SepikPlains, New Guinea: A Physiographic Study', Geographical Review54(1): 20-44. REYNDERS, J.J. 1962. 'Shifting cultivation in the StarMountains Area', Nova Guinea 10: 45-73. ROBERTS, J.M., C.C. MOORE and A.K. ROMNEY. 1995. 'Predictingsimilarity in material culture among New Guinea villages frompropinquity and language: a log-linear approach', CurrentAnthropology 36(5): 769-88. ROSS, M. 1988. Proto Oceanic and the Austronesian languages ofWestern Melanesia, Pacific linguistics. Series C no. 98. Canberra:Pacific Linguistics, Research School of Pacific and Asian Studies,Australian National University. SAND, C. 1997. 'The chronology of Lapita ware in NewCaledonia', Antiquity 71 (273): 359. SCHEFFLER, H.W. 1985. 'Filiation and Affiliation', Man 20(1): 1-21. SCHULTZE-JENA, L. 1914. Forschungen im Innern der Insel Neuguinea:Bericht des Fuhrers fiber die wissenschaftlichen Ergebnisse derdeutschen Grenzexpedition in das westliche Kaiser-Wilhelmsland 1910, inMitteilungen aus den deutschen Schutzgebieten. Erg 11. Berlin. SHENNAN, S.J. 1988. Quantifying Archaeology. Edinburgh: EdinburghUniversity Press. SHENNAN, S.J. and M. COLLARD. 2005. Investigating processes ofcultural evolution on the north coast of New Guinea with multivariateand cladistic analyses, in R. Mace, C.J. Holden and S.J. Shennan (eds),The Evolution of Cultural Diversity: A Phylogenetic Approach. London:University College London Press, pp. 133-64. SILLITOE, P. 1988. Made in Niugini: Technology in the Highlands ofPapua New Guinea. London: British Museum. SPRIGGS, M. 1993. Island Melanesia: the last 10,000 years, in M.Spriggs, D. E. Yen, W. Ambrose, R. Jones, A. Thorne and A. Andrews(eds), A Community of Culture: The People and Prehistory of the Pacific.Canberra. Department of Prehistory, Research School of Pacific Studies,Australian National University, pp. 187-205. STASCH, R. 2003. 'The semiotics of world-making in Korowaifeast longhouses', Language and Communication 23 (3-4): 359-83. STEER, M. 2005. Linguistic report for the Upper Sepik--Central NewGuinea Project. Canberra, Research School of Pacific & AsianStudies, Australian National University. (unpub.). STRATHERN, A. and M. STRATHERN. 1969. Marriage in Melpa, in R. M.Glasse and M. J. Meggitt (eds), Pigs, Pearlshells and Women: Marriage inthe New Guinea Highlands. Eaglewood Cliffs: Prentice-Hall, pp. 96-116. STRATHERN, M. 1984. 'Marriage Exchanges: A MelanesianComment', Annual Review of Anthropology 13:41-73. SUMMERHAYES, G. 2000. Lapita Interaction; Terra Australis 15,Canberra: Department of Archaeology and Natural History and Centre forArchaeological Research, Australian National University. TERRELL, J.E. 1988. 'History as a family tree, history as anentangled bank: constructing images and interpretations of prehistory inthe Pacific', Antiquity 62(237): 642-57. 2001. Introduction, in J.E. Terrell (ed.), Archaeology, Language,and History: Essays on Culture and Ethnicity. Westport: Bergin &Garvey, pp. 1-10. TERRELL, J.E., K.M. KELLY and P. RAINBIRD. 2001. "Foregoneconclusions? In search of "Papuans" and"Austronesians", Current Anthropology 42(1): 98-124. THURNWALD, R. 1914. Entdeckungen im Becken des oberen Sepik,Mitteilungen aus den deutschen Schutzgebieten 27(3): 338-48. TIESLER, F. 1984. Die Kurasse Neuguineas, Abhandlungen und Berichtedes Staatlichen Museums fur Volkerkunde, Dresden 41:46-85. TOWNSEND, P. K. W. 1969. 'Subsistence and Social Organisationin a New Guinea Society'. PhD Thesis, Anthropology, University ofMichigan. WATSON, J.B. 1965. 'The significance of recent ecologicalchange in the Central Highlands of New Guinea', Journal of thePolynesian Society 74: 438-50. WATSON, J.B. 1970. 'Society as an organised flow: the Tairoracase', Southwestern Journal of Anthropology 26(2): 107-24. WELSCH, R. L. 1998. An American anthropologist in Melanesia : A.B.Lewis and the Joseph N.Field South Pacific Expedition, 1909-1913.Honolulu: University of Hawai'i Press. WELSCH, R.L., J. TERRELL and J.A. NADOLSKI. 1992. 'Languageand Culture on the North Coast of New Guinea', AmericanAnthropologist 94: 568-601. WHEATCROFT, W.G. 1975. 'The Legacy of Afekan: CulturalSymbolic Interpretations of Religion among the Tifalmin of NewGuinea'. PhD Thesis, Anthropology, University of Chicago. WURM, S.A. and S. Hattori (eds). 1981. Language Atlas of thePacific Area. Canberra: Australian Academy of the Humanities/JapanAcademy, Australian National University. NOTES (1.) What gave rise to the predominant theory that Austronesianshave migrated from North Asia is that Austronesian languages outsideTaiwan, all part of the subfamily of Malayo-Polynesian, are closelyrelated while all other subfamilies and their languages are found onTaiwan (see Greenhill & Gray 2005, 37). (2.) The first of these was the 'Express Train' (Diamond1988). Then there was the Voyaging Corridor Triple I model (VC TripleI). This model assumes five migration pulses with large pauses inbetween migrations. The second pulse, around 3,200 BP, coincides withLapita complex which is seen as a result of the Triple I process(Intrusion, Integration and Innovation) (see Green 2003); The other isthe 'Slow Train' which includes the same sequence as theExpress Train, but with greater pauses between migrations. (3.) This process had its origins in the initial human migration tothe area around 40,000 BP (Golson 2005). (4.) This was made possible by innovations in seafaring technologyalready taking place around the Bismarck group prior to the Austronesianarrival (Allen 1996; Egloff 1975). (5.) From Jayapura in Irian Jaya ([West] Papua) to Madang in PapuaNew Guinea (Welsch et al. 1992). (6.) They then constructed three subsets of assemblages containing39, 47, and 54 classes respectively by discounting assemblages that hada 'very small number of objects as well as ... a very small numberof object classes' (Welsch et al. 1992, 571). The subset consideredmost meaningful contained 47 and represented 31 villages. They also usedDriver's G equation as one of four measures of similarity toaccount for missing data. Driver's G discounts similarity based onshared absences and this was seen as a way of eliminating flaws in themeasure created by poor sampling. (7.) Wurm & Hattori (1981) for both Papuan and Austronesian forone measure; Foley (1986) and Ross (1988) for Papuan and Austronesianrespectively for the other. (8.) Principal components and multiple regression analyses werecarried out using logged artefact frequency data. (9.) For more information concerning the Project see website:www.uscngp.com. (10.) Indeed some groups in CNG have over fifty varieties of taro(Hyndman & Morren 1990: 22). (11.) In the Yellow River area, the arrangement was similar: thewomen and children lived in family houses set around a plaza and the menslept together in a large house at one end of the settlement (Craig1975: 419-20, 440-41 ; Mitchell 1975:417; Thurnwald 1914: 347). (12.) There was a prohibition concerning the use of staples such astaro as an exchangeable commodity; food was to be repaid only in kind.Negative reciprocity in the exchange of valuables was also deemed asunacceptable between members of the same community (Brumbaugh 1980: 29). (13.) Jorgensen (1981: 163) has argued that the cognatic descentunits of the Telefolmin of the Ifitaman Valley (known as tenum miit)arepositioned to form a symmetrical and complementary division of thecommunity in fulfillment of ritual obligations and this led to, orrather maintained, some degree of integration across parishes.Co-participation in ritual activity, especially those involvinginitiation rites, provided the mechanism through which members ofdescent units interacted and maintained ties (Barth 1971 ; Brumbaugh1980: 303ff; Craig 1969b: 71; Hyndman 1979: 53). Extensive genealogiesdid not exist for such units; at best they comprised sequences ofnotable men going back half a dozen generations. Each member'sgenealogical position was obscure and membership was passed from eitherparent through both male and female children and sometimes an individualcould identify more than one way of tracing links (Brumbaugh 1980:108-109). (14.) These figures are approximate because some registered objectshave as yet to be located, some objects identified as belonging to thecollection yet to be registered, and a number of objects identified asbelonging to a specific collection have yet to be corroborated. (15.) Additional information concerning the material culture,including photographs and an overview of the distribution of some formscan be found on the website: http://uscngp.com (16.) Slit gongs were found in some Sepik River communities of theUSB but they were relatively uncommon and probably derivative (Craig2002: 2). (17.) The most commonly identified are Miscanthusfloridulus andSaccharum sp. (Cranstone 1964b; 1990: 37; Kelm & Kelm 1980: 67;Hyndman 1979: 216) (18.) Hyndman (1979: 216) provides the most comprehensive list forspecies of palmwood used by communities within the study region (e.g.Hydriastele, Nengella, Caryota). (19.) Sillitoe (1988: 134-44) reported similar tendencies in Wolabamboo-blade arrows and also observed that different varieties of bamboowere sometimes chosen for the different blade types. (20.) This is in keeping with Petrequin and Petrequin's (1990)observations concerning variability in what was considered appropriatefight arrowhead types across Dani communities in [West] Papua. (21.) Some tendencies in head modification are found throughout theregion, demonstrating that people readily adopt new weapons technologywhen it is assumed to be advantageous--a kind of 'arms-race'to counter the emergence of a technological advantage in warfare,coupled with the tendency to associate prestige with the ownership ofsuch weaponry. (22.) This device is used in many parts of New Guinea; in someareas the tips were made from sections of human fibulae but morecommonly cassowary spurs, or a long thin section of fibulae bone of thecassowary, wallaby or tree kangaroo (Sillitoe 1988:145-51). In theUSCNGP area, only tips made from fibulae are evident and the availabledata suggest these are marsupial fibulae (Kelm & Kelm 1980: 69). (23.) In this study homogeneous subsets of related means for arrowlengths were created using Tukey's Honestly Significant Differences(HSD) Test. (24.) In the subsequent text concerning CA results, 'lowlandgroups' will also refer to the Yuri. (25.) The Asabano had strong exchange relations with the Oksapminand antagonistic relations with Telefol speaking groups during thisperiod (Lohmann 2000: 33-41). (26.) To some degree this result is attributable to smaller amuletbags, which involve techniques that sometimes do not occur for largerbags, being more commonly collected in the south-east sector (see Figure9). However, this still would not explain the extent of the divisionbetween these clusters, besides, for most part, different collectorswere responsible for each language sample and therefore there was littlescope for bias; accordingly any consistent difference in the content ofeither sample may be considered significant in itself. Andrew Fyfe University of AdelaideTable 1. Scale of Linguistic relatedness for study area languages1. Unrelated2. Very distantly related (not enough putative cognates to establishsound correspondences or to reconstruct the phonological history or aprotolexicon).3. Belonging to the same subgroup but extremely disparate, indicatingat least 3000 years of separation.4. Belonging to the same subgroup and showing systematic similaritiesin numerous subsystems, indicating separation within the past 1500-3000years.5. Belonging to the same subgroup and showing evidence of having beenmutually intelligible as recently as 500-1500 years ago.6. Well-differentiated dialects of the same language.7. Little-differentiated dialects of the same group.Table 2. Matrix of linguistic relatedness for study area languagesAmanab4 Waina1 1 Baibai1 1 2 Biaka1 1 2 3 Kwomtaril 1 1 1 1 Anggor1 1 1 1 1 3 Dera *1 1 1 1 1 1 1 Abau1 1 1 1 1 1 1 21 1 1 1 1 1 1 21 1 1 1 1 1 1 21 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 1Ak3 Awun3 3 Namie1 1 1 Oksapmin1 1 1 2 Ngalum1 1 1 2 4 Mianmin1 1 1 2 4 4 Bimin1 1 1 2 4 4 5 Telefol1 1 1 2 4 4 5 51 1 1 2 4 4 5 51 1 1 2 4 4 5 51 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 11 1 1 1 1 1 1 1Tifal5 Faiwo 15 5 Kauwol1 1 1 Amto1 1 1 1 Busa1 1 1 1 1 Yuri1 1 1 1 1 1 Nagatman* The Dera language was not inc1uded in Steers reviewTable 3. Ethnographic collections from the USB/Border Mountains.Collector No. of Year Museum repository objectsBuller, A. 190 1959 Museum der Kulturen, BaselCraig, B. 1120 1968, PNG National Museum and 1969, 1972-3 Art Gallery (PNGNMAG), WaiganiCraig, B. 680 1968,1969 Australian Museum, SydneyCraig, B. 680 1968,1969 Ethnologisches Museum, BerlinCraig, B. 680 1968,1969 R.v.Volkenkunde-RVM, LeidenEve, H. D. 320 1938 Australian Museum, SydneyGerrits, G. 170 1972 PNGNMAG, WaiganiHuber R. P. 140 1970 PNGNMAG, WaiganiJuillerat, B. 255 1973-74 Musee du Quai Branly, ParisJuillerat, B. 255 1973-74 PNGNMAG, WaiganiKelm, A. & H. 450 1970 Ethnologisches Museum, BerlinKelm, A. & H. 75 1970 PNGNMAG, WaiganiLewis, M. J. 85 1963 South Australian Museum, AdelaideMiles, D. 550 1965 Australian Museum, SydneyPeter, H. 750 1969-74 M.f. Volkerkunde, ViennaPeter, H. 85 1969-74 PNGNMAG, WaiganiPeter, H. 150 1969-74 Gottingen University MuseumPeter, H. 70 1969-74 Museum der Kulturen, BaselSchuster, M. & G. 205 1965 Museum der Kulturen, BaselWomersley, J. S. 100 1949 South Australian Museum, AdelaideMiscellaneous 100 Various VariousTOTAL c.6855Table 4. Ethnographic collections from Central New Guinea.Collector No. of Year Museum repository objectsBarth, F. 90 1968 Bergen University MuseumCampbell, S. 125 1935-36 Australian Museum, SydneyCraig, B. 320 1964 Australian Museum, SydneyCraig, B. 120 1972 PNGNMAG, WaiganiCraig, B. 85 1983 PNGNMAG, WaiganiCranstone, B. 490 1964 British Museum, LondonCranstone, B. 300 1964 PNG National Museum, WaiganiStar Mtns Exped'n 480 1959 R.v.Volkenkunde-RVM, LeidenEggetsson, S. 86 1994 Personal Collection, ReykjavikFriend, M. 160 1981 British Museum, LondonFriend, M. 200 1981 James Cook Universty Museum, TownsvilleHyndman, D. 100 1971 Queensland University Museum, BrisbaneHyndman, D. 70 1971 PNGNMAG, WaiganiMacKenzie, M. 110 1982-84 Australian Museum, SydneyMorren, G. 100 1969 Personal Collection, New JerseyMorren, G. 105 1969 PNGNMAG, WaiganiPerey, A. 255 1967 PNGNMAG, WaiganiSchuster, M. 480 1965 Museum der Kulturen, BaselWheatcroft, W. 90 1969 Penny Collection, SAM, AdelaideWilliams, W. 450 1935-6 Los Angeles County MuseumMiscellaneous 200 Various VariousTOTAL c.4095Table 5. Homogenous subsets of languages with related means forbamboo-blade arrow whole length, a = 0.1.Tukey HSD n 1 2 3 4 5BIMIN 8 137.38OKSAP 17 140.88 140.88TIFAL 46 141.63 141.63FAIW 5 147.8 147.8TELEF 73 152.74 152.74MIAN 35 156.89NAMIE 56 164.89ABAU 211 166.56YURI 66 172.64Sig. .575 0.041 0.362 0.61 .012Table 6. Homogenous subsets of languages with related means for stringbag maximum to minimum width, a = 0.5.Tukey HSDLANGUAGE n 1 2 3 4NAMIE 51 70.10ABAU 49 78.39 78.39YURI 26 78.50 78.50TIFAL 47 85.45 85.45TELEF 42 85.83 85.83MIAN 11 85.91 85.91FAIW 12 93.33 93.33BIMIN 34 93.68 93.68OKSAP 33 99.24Sig. .533 .678 .562 .890