CHAPTER 3 EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD: AN INTRODUCTION PATRICK V. KIRCH * During the course of our three expe- ditions to Mussau, 13 archaeological sites were excavated (Table 3.1), with sample sizes ranging from a single 1 2 test as in sites EHN and EKL, up to the 84 m2 excavated at Talepakemalai (ECA), the largest Lapita site. Sites cho- sen for excavation-on Eloaua, Emananus, Boliu, Emussau, Enusagila, and Mussau is- lands (Fig. 3.1)-include open shell middens, rockshelters, and a cave site.' Although our research objectives dictated that our primary focus would be on sites containing Lapita pottery, six non-Lapita sites were investigated in order to extend the definition of the Mussau cultural sequence over the past two millennia. Chapters 4 through 9 provide detailed excavation re- ports for each of these 13 sites, describing site location and geomorphological settings, stratigraphy, features, cultural content, and chronology as determined by ra- diocarbon dating. To avoid repetition, common aspects of excavation strategy and methods are summarized here; only procedures specific to individual sites are dis- cussed in the appropriate excavation reports. SAMPLING THE ARCHAEOLOGICAL RECORD: SITE CHOICE AND EXCAVATION STRATEGIES Archaeologists make sampling decisions-both con- sciously and unconsciously-at a hierarchy of levels during the course of their field and laboratory investi- gations. The choice of areas to be surveyed (and the survey methods used), selection of sites for subsurface testing and excavation, excavation strategy, recovery tech- niques used and materials retained as well as discarded, and decisions made in the laboratory regarding which materials to study intensively and by what methods-all of these impinge on the sample of the archaeological record from which we ultimately draw our conclusions and interpretations of the past. In the Mussau Project, sampling decisions at each of these levels evolved over the course of three field seasons, as our knowledge of the local archaeological record expanded, and as our research goals developed (see Chapter 1). When we com- menced work in Mussau in 1985, the islands' archaeo- logical record was virtually unknown. The extensive ECA Lapita site had been tested by Egloff (1975; Bafmatuk EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD TABLE 3.1 Features of Mussau archaeological site classes. Site Class Lapita Non-Lapita Terebra-Shell Obsidian Pottery Pottery Adzes Open Sites 1 . Large shell middens A + - - + 2. Large shell middens B - +/- + + 3. Small shell midden scatters - - - +/- Rockshelters +/- +/- +/- + Explanation: + present; - absent; +/- may be present or absent in particular sites. et al. 1980), who also reported the presence of a sec- ond, smaller Lapita site on Eloaua Island (ECB). How- ever, the size and extent of site ECA itself were un- known, and the chronology of its occupation sequence was controversial (see Chapter 4). Beyond these two sites, a one week reconnaissance byJ. Allen andJ. Specht in 1984-preparatory to the Lapita Homeland Project-had recorded 16 new sites, although in most cases the information obtained consisted of no more than a map coordinate locality and a few brief lines of description (Allen et al. [1984:8-11]; see also site inven- tory forms in the Prehistory Section, Papua New Guinea National Museum). For the most part, then, our 1985 season was carried out in archaeological terra incognita. In 1985 and 1986, while excavating the ECA Lapita site and testing the smaller ECB and EHB Lapita sites, we also reconnoitered Eloaua, Emananus, Boliu, Ekaleu, Emussau, and Ebolo islands, with brief forays onto the main island of Mussau at Lomakunauru and Palakau. We utilized informants' knowledge of the landscape by inquiring about locations of old village sites, and about localities where they had observed pottery, obsidian, shell midden, or other cultural artifacts while gardening, or hunting in the forest. When time permitted, we walked the coastal terraces and interior portions of the offshore islands looking for surface indications of sites. By the close of the 1985 season, we had reached some tentative generalizations concerning the surface archaeo- logical record of Mussau. The archaeological landscape appeared to consist of several primary kinds of sites:2 (1) open shell middens containing classic dentate-stamped Lapita pottery, represented by sites ECA, ECB, and EHB; (2) open shell middens with occasional non-Lapita pottery sherds, or with pottery wholly absent, but with Terebra-shell and Tridacna-shell adzes on their surfaces; (3) small shell-midden and obsidian scatters, lacking pottery or shell adzes; and (4) rockshelters and caves situated in limestone escarpments. The first two classes of sites were frequently spatially extensive, with areas ranging from 1,000 up to 72,000 i2; these sites were invariably found on the immediate coastal terraces, - 1- 2 m asl. The small shell and obsidian scatters, in con- trast, usually covered only a few tens of square meters, and were generally situated on the upraised, interior por- tions of the islands (- 10-40 m asl), in terrain utilized today for shifting cultivations.3 Rockshelters are not com- mon in the areas surveyed, and their presence depends upon the necessary geological conditions, such as verti- cal escarpments in the upraised reef limestone, where former wave-cut notches or solution caves produced habitable spaces. It was also apparent by the close of the 1985 season that other classes of archaeological fea- tures-although common on some Pacific islands- were wholly absent from the Mussau archaeological land- scape. For example, there was no evidence of con- structed stone architecture: no walls, stone-faced ter- races, platforms, pavings, or other constructions that could represent either habitation or agricultural activi- ties. Rather, the cultural remains produced by past gen- erations of islanders in Mussau were largely concen- trated in a limited range of discrete, depositional envi- ronments: primarily large coastal shell middens of both Lapita and non-Lapita affiliation, and to a lesser extent in smaller shell-midden scatters and in rockshelters. Some of the key features of these major site classes revealed during initial reconnaissance are summarized in Table 3.1. Our excavation strategies obviously had to be for- mulated specifically for these kinds of depositional con- texts. This knowledge of the Mussau archaeological land- 58 EXCAVATING THE MUSSAU Ar 03 04 05 P A C I t Q.CEAN+ M U S S A U 'I S L A N D' 7~ - 9^ E---/ X / /~ / br - - 1\ 2/ \ / \ / "'N GN MN I I I l I FIGURE 3.1 Map of the Mussau Islands, showing the locations of excavated and tested sites. - - Rocl< * Sites EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD scape gained during our 1985 reconnaissance-ampli- fied and refined by more intensive surface survey in the 1986 and 1988 seasons4-informed our choice of sites to be excavated. Since our primary interest was the "Lapita period,"5 it was necessary that all sites yielding Lapita ceramics be investigated by subsurface sampling. This was a main objective of both the 1985 and 1986 field seasons. It was also essential to set the study of the Lapita period in Mussau within a broader temporal con- text, by identifying depositional contexts pre-dating Lapita (if these existed), as well as contexts bridging the Lapita period sites with the ethnohistoric period. This entailed subsurface sampling of open shell midden sites lacking Lapita pottery, as well as rockshelter deposits. Sampling of potential non-Lapita sites began during the 1986 season, with the testing of several caves and rockshelters (sites EHM, EHN, EKO, EKP, EKQ) as well as the EKS and EKU open midden sites.6 During the 1988 expedition, further sampling of non-Lapita sites became a major objective, with excavations at the EKE, EHK, and EKL shell midden sites (Table 3.2). In sum, at the first level of sampling-decision mak- ing-that of site choice for excavation-we were guided by two main considerations: (1) the nature of the ar- chaeological landscape and of depositional contexts as informed by reconnaissance and intensive survey, and (2) our specific research goals. The focus on Lapita re- sulted in a greater emphasis on sites containing Lapita pottery, but the sample of six non-Lapita sites also pro- vided valuable information on the later prehistoric pe- riod. Having selected sites for subsurface investigation, a second level of sampling decisions involved excavation strategy. Here again, both the nature of the archaeologi- cal landscape as well as specific research objectives came into play. To a certain extent, each site had to be taken on its own terms, and the particular excavation strategy and procedures employed are discussed in the site re- ports that follow (Chapters 4 to 9). However, all exca- vations were guided by several common strategic goals: (1) to determine the geomorphological context respon- sible for a site's depositional history; (2) to determine the depth and nature of cultural deposits, and where feasible, the spatial boundaries and areal extent of the site; and (3) to define micro-stratigraphic and deposi- tional history. For certain sites, a fourth goal was also defined, to assess the nature of internal patterning or differentiation in cultural content. For the large shell middens, these research objec- tives were best addressed through the use of a systematic transect excavation strategy, in which 1-m2 excavation units were excavated along transect lines carefully posi- tioned in relation to geomorphological features and to surface archaeological distributions.7 The use of system- atic transects is also effective in defining site boundaries and in providing an overview of stratigraphic varia- tion. This systematic transect strategy had been used by me in previous archaeological studies on Tikopia (Kirch and Yen 1982), on Niuatoputapu (Kirch 1988a), and in the Manu'a Islands (Xirch et al. 1990; Kirch and Hunt, eds., 1993) in coastal beach-ridge and terrace environ- ments, and its application in Mussau was based on these prior experiences. The cave and rockshelters required different strategies, usually of deep vertical tests within the level floor areas with their concentrated deposits. Likewise, once we had discovered the presence of wa- ter-logged, anaerobically preserved wooden post ar- chitecture in the ECA Site, our excavation strategy was modified to include areal exposures in order to expose a large contiguous area of deposit. EXCAVATION METHODS All of the excavations carried out during the Mussau Project followed standard methods and techniques. Horizontal metric grid control was established prior to the beginning of an excavation, and sites were mapped either with compass-and-tape, or with the aid of a Leitz telescopic level and stadia rod. Elevation profiles were taken along key transects, and correlated to sea-level through the use of elevation readings with the aid of a tide chart.9 On large shell midden sites, a datum peg was established, and excavation units were designated with a coordinate system; e.g., a unit designated W250N150 would be situated 250 m W and 150 m N of the site datum. Excavation always proceeded by "natural" stratig- raphy, that is, by sedimentary units of deposition. How- ever, if a natural stratum proved to be fairly thick (e.g., more than about 10 cm) it was usually sub-divided into two or more levels. The level is thus the minimal unit of vertical control in our excavations. A single level may or may not correspond to a stratigraphic layer, depending upon whether the latter was subdivided into more than one level. However, levels never cross stratigraphic boundaries. 60 EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD TABLE 3.2 Summary of Mussau excavations, 1985-1988. Island Locality Eloaua Talepakemalai Eloaua Etakosarai Site Site Type ECA Open midden ECB Open midden 1 areal ex. 1985, 86 Transect Emananus Etapakengaroasa Eloaua Eloaua Eloaua Mussau Mussau Emussau Mussau Boliu Eatulawana Epakapaka Sinakasae Eloaua Enusagila TOTAL EHB Open midden EHM EHN EKO EKP EKQ EKS EKU EKE Cave Rockshelter Rockshelter Rockshelter Rockshelter Open midden Open midden Open midden EHK Open midden EKL Open midden The main tool used to excavate was the Marshailtown trowel,10 aided by whisk brooms and paint brushes for fine work. Excavated sediment was transferred to buck- ets using plastic dust pans, and the bucket contents then sieved (see Recovery Methods, below). Artifacts found in situ were three-dimensionally plotted; all other mate- rials were provenienced by unit and level, with each lot or item numbered sequentially within each level. Thus the field code ECA/W250N150/6/42 would refer to the 42nd item or lot1" recovered from level 6 of unit W250N150 in Site ECA. Vertical control was main- tained either by recording the depth below surface, bs, (from the SW corner of the unit), or by measuring depth with telescopic level and stadia rod and converting this to depth below site datum. During and after excava- tion, a photographic record was made using both black- and-white and color 35-mm film. Upon the comple- tion of excavation, at least one stratigraphic profile of each unit was drawn at 1:10 scale, and the layers de- scribed according to a standard set of criteria, includ- ing grain-size, lithology, color (Munsell system), struc- ture, nature of stratigraphic boundaries, and so forth. Sediment samples were taken from selected stratigraphic profiles for laboratory analyses of grain-size and other characteristics. Sediment sampling followed the proce- dures outlined by Stein (1985, 1987).12 Certain terms widely used in archaeological par- lance nonetheless vary considerably in their definitions, and therefore we wish to be explicit regarding their meaning in these monographs: Grid Unit (or simply, unit): the horizontal unit of control, a lxl m block. Units may be further specified as test units or transect units. Level: the vertical unit of control within a unit. Lev- els never cross natural stratigraphic boundaries, although they may be of artificially defined thickness within natu- ral strata. This was necessary sometimes to subdivide a thick deposit for purposes of vertical control." Levels are numbered with Arabic numerals from top to bot- tom. Layer: a natural sedimentary and depositional stra- tum, defined in terms of lithology, structure, grain-size, color, boundary contacts, and similar criteria. One layer may correspond to one or more levels. Layers are num- bered with roman numerals from top to bottom; fa- cies distinctions are sometimes designated with letters (e.g., Layer IA). Analytic Zone: an aggregate of levels across two or Year Dug Method Area 2) 1m) 985, 86, 88 Transect, 84 Periods Represented Lapita 1 9 Lapita 9 Lapita 1985, 86 1986 1986 1986 1986 1986 1986 1986 1988 1988 1988 Transect Test pits Test pit Test pits Trench Test pits Test pits Transect Transect Transect Test pit 3 1 2.5 5 2 4 5 19 5 159.5 Lapita Lapita Lapita Post-Lapita Lapita Post-Lapita Post-Lapita Lapita to Post-Lapita Post-Lapita Histonc 61 EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD more units, combined for purposes of analysis and re- porting of cultural content. Levels combined into an analytic zone are regarded as being chronologically and stratigraphically equivalent. Analytic zones are designated with capital letters, from top to bottom; subzones are numbered with numeric subscripts. RECOVERY TECHNIQUES AND BIASES Although in a few regions of the world archaeological sites continue to be excavated without screening, most prehistorians would not dream of digging into the earth without also passing their loosened dirt through a sieve. The object of this time-consuming task, obviously, is to recover artifacts and other culturally deposited materi- als not detected during the process of digging itself. Yet, as a spate of experimental studies have demon- strated, the mere application of screening in-and-of- itself does not assure a high rate of recovery, and vari- ous factors including mesh size, use of water, and me- chanical actions, can significantly affect recovery rates. Moreover, the use of different screening techniques ei- ther within a single site, or between sites, can thwart inter-site comparability. Butler (1988) and Nagaoka (1988) have discussed such problems of non-compa- rability of faunal data sets from various Lapita sites, resulting from a variety of screening methods and mesh SZS14 sizes.4 As the Mussau Project progressed, we became in- creasingly concerned with the problems of recovery and sample bias, especially after Butler (1988) and Nagaoka (1988) undertook systematic reviews of the Lapita faunal evidence prior to the 1988 expedition. In 1985, we had little choice in recovery methodology, because the screening equipment had been provided in advance by the organizers of the LHP. We were sup- plied with several sieves of 7 mm and 5 mm mesh, and a single sieve of 3 mm mesh.15 We used the 5 mm mesh as much as possible, taking control samples with the 3 mm mesh to determine whether there was significant size bias in the assemblages retained by the 5 mm sieves. These tests indicated that the 5 mm mesh sieves were retaining most of the small faunal remains, as well as all potsherds and obsidian flakes, and in 1986 we there- fore equipped ourselves with a number of new 5 mm mesh sieves, as well as additional 3 mm sieves.16 In 1986, we also initiated the regular use of wet-sieving at the ECA Site and, where feasible, at other sites (such as EKS and EKQ). We remained concerned about possible bias in the dominant use of 5 mm mesh in our excavations, how- ever, and at the urging of Virginia Butler conducted additional comparative sieve tests in 1988 (see Volume II). Butler's analysis of faunal suites recovered through the 5 and 3 mm meshes revealed no statistically signifi- cant differences between the fish remains recovered; we are thus satisfied that our screening methodology has not introduced significant bias into our assemblages. For persons working with the Mussau Project collec- tions in the future, a database has been compiled indi- cating mesh size, and use of wet or dry sieving, for all units and levels of all sites. Thus sampling decisions can be made explicitly on the basis of recovery technique. During the 1988 season, we devoted substantial time and labor to the systematic subsampling of every unit and level excavated at the ECA Site with 1-liter bulk sediment samples, which were both floated and fine- sieved through 0.125-inch mesh to recover minute flo- ral and faunal materials. This labor-intensive work yielded only marginal improvements in recovery rates, but re- assured us that our main sieving techniques were not yielding biased results. Recovery is affected not only by excavation and screening methods, but by what is retained for analysis and further study. I determined early on that we would save for analysis 100% of all cultural materials retained in whatever sieve size was being utilized, for every ex- cavated unit.17 This procedure reflects the view that ex- cavation itself is already an act of sampling a universe (the site) of unknown-but almost certainly not ran- dom-distribution, and that a priori sub-sampling for faunal remains or other particular classes of material would only lessen our ability to understand site struc- ture. We thus concur with Leach and Leach (1979) in their stated justification for collecting "massive residues" from excavations at Palliser Bay, New Zealand: this approach was adopted in the belief that the analysis of cultural samples will only reveal information about the samples and not the population from which they derive. Cultural evidence is not distributed randomly; therefore, random sampling will not reveal cultural distributions (1979:5). The wisdom of saving all cultural materials retained by the sieving operations was questioned by team mem- bers more than once as-already exhausted by a long EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD day of digging in the tropical heat and humidity-we worked long into the night sorting, weighing, and count- ing staggering quantities of shell midden. Indeed, the processing of invertebrate midden materials consumed a large amount of time and labor; at the ECA Site alone, we processed 1.43 metnrc ton of shell midden during the three seasons. Although 100% of all excavated and sieved cul- tural materials were returned to the field laboratory, budget restrictions dictated that only some materials could be shipped to the United States, for study and curation. Thus, sampling decisions again had to be made to ana- lyze under field conditions certain classes of cultural materials which would then be discarded. The follow- ing classes of materials were retained for permanent curation: (1) non-ceramic artifacts; (2) diagnostic sherds, as well as large representative samples of plain body sherds;18 (3) vertebrate faunal remains; and (4) paleobo- tanical remains except for unmodified wood and co- conut endocarp (found in site ECA only), which were sampled. Procedures for processing invertebrate faunal materials varied from site to site, and will be discussed in Volume II. Minimally, shell midden from each level was bulk weighed prior to discarding. For a large num- ber of units, shell midden was first sorted to taxonomic class, each class being weighed individually. During the 1988 field season, this procedure was extended by also counting shell MNI for individual taxa, and by taking extensive series of individual shell measurements (see Volume II). RECORDING SCHEMES AND DATABASES The field and laboratory recording schemes and data- bases used by archaeologists in the complex process of converting an in situ depositional context of cultural materials-a site-to a curated "archaeological record" are not trivial matters, for such systems of recording, index- ing, and retrieving data impose a certain structure by their very existence.'9 The importance of standardized record- ing procedures was recognized by the organizers of the 1985 Lapita Homeland Project, who provided all field teams with pre-printed excavation recording forms, in self-carbon duplicate sets. The intent was that each field team would retain the top copy, submitting the dupli- cate to the Australian National University, where data would be coded and entered into a LHP-wide data- base using the MINARK software program." For the 1986 and 1988 expeditions, we modified the field recording forms, with improvements based on our 1985 experience. The "level form," an example of which is shown in Figure 3.2, provides the primary record for each excavated level. The sheet records pro- venience data, sieve size and wet/dry recovery data, whether depth was recorded below datum or below surface, start and end depths for the level, diagrams of all features at the start and end of excavation, descrip- tions of sediment, recovered finds, etc., and a listing of all objects provenienced by x, y, and z (depth) coordi- nates. These forms were kept on a clipboard and filled out by the excavator. In addition, the site director and supervisors maintained daily logs, and recorded strati- graphic profiles of excavated units in standardized bound field notebooks.2" In addition, three preprinted forms were used to assist in processing and recording of excavated materials in the field laboratory: (1) The first was a "small finds form" on which the washed and sorted pottery sherds from each level were enu- merated according to sherd type (rims, body sherds, etc.) and decorative technique (these sets of classes were not mutually exclusive). Also recorded on the small finds sheet were data on coral oven stones, and on worked shell debris (chipped Trochus and Conus shell detritus from the manufacture of shell artifacts) determined during the processing of shell midden. (2) The second form recorded weight and NISP/MNI counts of mollusks. (3) The third form (used in 1988 only) recorded stan- dard measurements on individual midden shells. Ex- amples of these field laboratory forms are illustrated in Figure 3.2. Upon our return from the field, finds were un- packed and checked against the level and small finds forms, catalog numbers written on all objects with in- delible ink, and the following standard data set was coded for entry into the computer database: (1) item code, or catalog number, based on the site/unit/level provenience; (2) item type, whether artifact, faunal or floral sample, or other material; (3) raw material; (4) class (e.g., rim sherd, flake, adz, or specific taxon in the case of floral and faunal materials); (5) count (1 in the case of individual artifacts); (6) weight (in grams); and (7) a free-form comments field. These data fields were coded not only for all excavated artifacts, but for all floral and faunal materials (including those discarded in the field), and other kinds of samples. 63 EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD 19S MUSSAU EXEDmIoIN sits _ A_a _ Grid so sow _ ___ 1i MUSSAU EXPEDMON Sswdve ply P wet a p -- Dw.du Swfacs p Dm p 16 MUSSAU EXPEDITION MomA Ah A Ipamumm~~~~~~~~~~~~~~~~~ GCO aOSe Pgmd 7it f am d.aT dm Chdd,bds Cood Tabwm 12AL & OPD .sm SBIV_LVES ;MA- t_ _dMadk Cgm Aic} v I~ ~ TOA=-VLE 1. TMAL KO_LL-L1"-~~~~~~~~~~~~~~~~~~~~~~~~~~~ 19J MUSSAU EIXPEDMON . .. I I I:LEl1 I I I I === A -= I --= I I I == - I7 i _ I I D _"_- . M.t ,- ==j~andmm SPa gL 7 8d Pwk _ A i p. No. I I I I I, D s bwkoffom aso escribe amWorsvm A c_d ewtOckoomnswdm. FIGURE 3.2 Examples of standardized field recording forms used during the Mussau Project: (A) to maintain vertical and horizontal excavation control, and to record finds in situ, (B) the small finds form, used in the field laboratory; (C) the form used to record shell midden analyzed in the field laboratory and discarded; and (D) the form used in the 1988 season to record detailed metric data on individual midden shells. Sift ,,At Gtd Spk S& iZ 7- ad L- Sm LAves EndLAves 0My11~~~~~y MmSIZ.__ cm momZE __ cm att ump ks. cm I e s 1 9 2 Re mo ad : s __ _ A B C 64 D-ft . I I - I i I I I - I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I# I I I I I I i I I I I I i I i I i i I I I I I I I I EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD The software and hardware utilized for this Mussau Project Database have changed several times over the life of the Project, reflecting the rapid changes in com- puter technology during the mid-1980s to late 1990s. A prototype of the database with over 14,000 entries was developed at UC Berkeley using the UNIX platform of the Quantitative Anthropology Laboratory. Data entry was conducted mainly with two networked SUN 3/50 workstations in the Oceanic Archaeology Labo- ratory; the database software used was Sybase SQL. While SQL software is powerful, it also has a steep learning curve. Furthermore, the multi-user environment was frequently over-taxed, resulting in delays of up to 45 minutes for complex relational queries of the data- base. In 1991, therefore, as data analysis was in full swing, we transferred the database (which now exceeded 16,000 records) to a stand-alone 386-processor 33-MHz PC, with a math coprocessor and a 120 MB hard disk. The new database software chosen was Borlan's PARA- DOX for DOS. This software proved much easier for the laboratory staff to learn, while continuing to allow completely relational queries. Although this dedicated machine was, in theory, slower than the SQL server, independence from the over-taxed network environ- ment meant that all queries could be accomplished in a matter of minutes. Furthermore, data were readily trans- ported to Borlan's QUATTRO-PRO spreadsheet pro- gram, which provided a variety of basic statistical rou- tines and graphical displays. For more complex statisti- cal analyses, we used the SPSS and SAS software pack- ages. More recently, given rapid improvements in the de- sign and costs of PC computers, we upgraded this con- figuration to a Pentium II-processor PC computer (233 MHz, 64MB, 6.4GB), with the software also up- graded to PARADOX 7 for Windows. The final configuration of the Mussau Project Data- base is diagrammed in Figure 3.3. The core of the data- base is the MUSSAU.DB table, which includes a total of 22,727 records representing some 231,874 individual arti- facts, faunal and floral remains, radiocarbon samples, and other materials. Each record in the MUSSAU.DB table consists of entries for: (1) item code, or catalog number; (2) item type; (3) raw material; (4) class; (5) count; (6) weight; and (7) free-form comments. Linkedtables (e.g., CLASSDB, RAW.DB, SITE.DB, ZONE.DB, SITENAME.DB, and MOLCLASS.DB) provide additional data on analytic zones and extended labels. Detailed provenience data are linked to MUSSAU.DB by the special tables OFFSET.DB and XYZ.DB while information on sieve sizes and wet/dry sieving methods for all units and levels are provided in the SIEVE.DB table. In addition to the main MUSSAU.DB database table, the project database also includes several FIGURE 3.3 Structure of the Mussau database system. I Parent/Class Codes I MOLCLASS.DB = I Molluscan I (self-join) r Phylogenetic Class I - 1-n Concordances 152 Records Class Code Unit/LevelAtem Codes n-n 1-1 . OFFSET.DB Spatial Concordances Unit/Level Codes for Various Years - = of Excavation n-1 29 Records MUSSAU.DB Main Object Table All Sltes 9') 797 Rp-r-nrrl I Unit/Level Codes Unit Code SITENAME.DB Class Code Raw Material Code I I = Site Code to Name = = n-1 n-1 Concordances n-1 n-1 24 Records CLASS.DB RAW.DB SIEVE.DB ZONE.DBSIEDSieCd Class Code to Name Raw Materal Code to Sieve Size and Type Analyic Zone SiTence SitetCode Concordances Name Concordances Data for Mussau Sites Concordances for for Vaious Sites n-1 264 Records 24 Records 623 Records Site ECA fo157 Records S 294 Records15Reod 65 EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD eciaized data analysis tables, including MUSCERAMDB and ECASHRD.DB which provide ceramic analysis data, and FISHECA.DB and OBSID.DB which provide data on fishbones and obsidian. It is our intention that the computerized database created for the Mussau Project should become a part of the permanent archival record of our excavations and analyses. Copies of the database have been depos- ited with the Archaeological Research Facility at the University of California, Berkeley. NOTES TO CHAPTER 3 All sites have been labled with the site designation system in use by the National Museum of Papua New Guinea. In this system, the initial letter "E" stands for New Ireland Province, while the second and third letters indicate the individual site, beginning with "AA" for the first site record (EAA, thus being the first site reported for New Ireland Province). 2 In his preliminary report on archaeological investigations in the Arawe Islands, Gosden (1989, 1991a) questions the basic notion of the archaeological site: "the last 10 years have seen a shift towards the analysis of regional artefact distributions as a whole, rather than a concentration solely on the archaeological 'hot spots' defined as sites" (1991:205). While I agree with Gosden that it is essential to take a 'regional'-as well as geo- morphologically informed-approach to archaeological land- scapes, the fact remains that in both Mussau and the Arawe islands (as in much of the Bismarcks), 99% of the archaeologi- cal record is concentrated in 'hot spots' with discrete bound- aries. Sites do exist, despite the fact that they need to be concep- tually and analytically seen as components of larger archaeologi- cal landscapes. The apparent reality of sites in the Arawe Islands is evident in Gosden's own report, for in spite of his theoretical discussion, Gosden proceeds to use the 'site' as his primary conceptual unit. I Many of these shell scatters were identified by informants as former locations of temporary garden houses, with ages of no more than a few decades. 4In all, we have formally recorded 29 sites, primarily on the SW, coral limestone islands of Eloaua, Emananus, Emussau, Boliu, Ebolo, Ekaleu, Ebanalu, and Enoanaulu. While we do not claim to have discovered or recorded every small shell midden scatter on these islands, we believe that most large open shell midden deposits have been identified. However, on the main island of Mussau our survey remains limited to reconnaissance of a few areas, primarily at Lomakunauru, Roitano, the coast S of Palakau, and the Tanaliu area in the NW part of the island. I I define this term minimally as the period of time during which pottery of Lapita style, in the sense of Mead et al. (1975), was produced and used. 6 Some of these did prove to have Lapita occupation compo- nents. 7For example, as most of the open shell middens were situated on coastal terraces, the transects were usually oriented so as to bisect the terrace, providing an array of excavation and strati- graphic samples that cross-cut the geomorphological features of the site. I I cannot overly stress the effectiveness of the systematic transect strategy for the first phase of subsurface sampling at ECA, for it was this strategy that led to the discovery of the waterlogged Area B deposits. The failure of Egloff and his colleagues to discover this key part of the ECA Site, in contrast, is largely a reflection of their use of a strategy confined to the testing of surface midden concentrations, and not informed by a geomor- phological analysis of the site. 9 We used the NOAA tide chart (NOAA 1987), with our tidal information based on Truk Island Publon Harbor), applying the necessary temporal and height correction factors stipulated for Emirau (station 3089; NOAA 1988:331). By recording the time at which sea level readings were made, and by checking this with multiple readings at different times and on different days, it was possible to tie our site elevation profiles to modem mean sea level with a minimal error factor. 10 In 1985, the experience of having cheaply manufactured, riv- eted trowels break one after another convinced me that the 66 EXCAVATING THE MUSSAU ARCHAEOLOGICAL RECORD slight cost of shipping Marshalltown trowels (the long-stand- ing tool of choice in North American excavations) to PNG was worth the expense. 11 The term "lot" refers to a set of items of the same class, e.g., a group of plain body sherds, or of obsidian flakes. 12 We also took control sediment samples along two modern beach slopes where the depositional environment was known, for comparison with the archaeological samples; see Chapter 4 for details. 13 Because our 1985 excavations were undertaken as a part of the Australian-organized LHP, we fell into the habit of referring to levels as "spits," using the Old World term common in Austra- han archaeological jargon. Our field forms likewise use the term "spit." In this monograph, however, we exclusively use the more common American term "level." 14 Screen mesh sizes reported in the literature as having been applied on Lapita sites include: 10-12 mm (Liley 1986), 9 mm (Best 1984), 7.1 mm (Best 1984), 6 mm (Liley 1986), 5 mm (Green 1986, Best 1984; Kirch 1987a), 2.5 mm (Best 1984), 1/ 2" (Gifford and Shutler 1956), and 1/4" (Kirch and Yen 1982: Kirch 1988a). Other excavators have not bothered to report screen mesh size at all (e.g., Poulsen 1968). 15 The single 3 mm sieve was unfortunately in poor condition, and broke before the end of the field season. 16 In North America, mesh sizes of 1/4" and 1/8" are routinely used, with most zooarchaeologists insisting that the 1/8" mesh is essential to avoid bias toward larger elements in faunal as- semblages. The 5 mm sieves we used in Mussau are slightly finer than 1/4" (being approximately 3/16"), while the 3 mm sieves are slightly finer than 1/8". 17 By "cultural materials," we mean anything not obviously deposited in the site by natural processes. Cultural materials therefore includes not only ceramic sherds, lithic flakes, and all other classes of artifacts sensu strictu, but coral oven stones, non-local stone manuports, vertebrate and invertebrate faunal materials, and preserved plant remains. 18 Diagnostic sherds include rims, carinations, bases, and any sherd bearing decoration. 100% of plain body sherds from both the 1985 and 1986 excavations were saved, but only repre- sentative samples of plain body sherds from the 1988 season at ECA. 19 Some 30-plus years excavating experience has convinced me of the importance of formal recording procedures in archaeo- logical excavation. Depending solely upon "free-form" note taking is a certain recipe for ensuring that important data will be lost. This does not by any means lessen the value of free-form observations, which are to be encouraged by all participants, but which should in my view be in addition to formally struc- tured records. 20 Regrettably, this well-planned scheme did not materialize, through no fault of the organizers, as some team directors did not use the pre-printed forms, and funds were not available for data coding and computer entry. 21 We used the excellent "Anthropology Field Notebook" de- signed and published by the staff of the Bishop Museum in 1974. These books, hardbound with 47 leaves, are printed on water-proof paper, have recto pages in metric graph paper (suit- able for 1:10 section drawings), and verso pages in two ruled columns. 67