12n CHEMICAL CHARACTERIZATION AND PROVENANCE OF MANU'A ADZ MATERIAL USING A NON-DESTRUCTIVE X-RAY FLUORESCENCE TECHNIQUE MARSHALL WEISLER INTRODUCTION THE COLONIZATION SiRATEGIEs employed during the settlement of Polynesia and the subsequent diversification of island societies are key issues in Oceanic prehistory. Yet it is only recently that lcsearchers have begun to amass empirical evidence of inter-island communication-throughout island sequences-that undoubtedly influenced the direc- don and tempo of post-settlement island histories. I refer here to recent finds of Tongan pottery in the Cook Islands (Walter and Dickinson 1989:465), of Samoan adz material in Fiji (Best 1984; Best et al. 1992) and the Cook Islands (Weisler 1993), and of ier-island transport of volcanic glass and fine- grained basalt in Hawaii during late prehistory (Weisler 1990; et al. In prep. [For a review of Polynesian basalt adz provenance studies, see Weisler In press a.]) . Isolation has been invoked as '"e most fundamental" mechanism of divergence in Polynesia (Kirch and Green 1987:440), but "how . . . different extremes of isolation have influenced the evolution of human diversity from island to island- ad perhaps even on the same island" (Terrell 1986:122-23) remains a problem that must be addmessed for each island sequence. Dissimilarities may also originate because of continuing contact (see Terrell 1986:147). Isolation may have contrib- uted to regional variations in portable artifacts, architecture, and language in New Zealand (Prickett 1982), Society Islands (Emory 1933), and Hawaii (Kirch 1985, 1990). In contrast, inter-island commu- nication is inferred by parallel histories in subsis- tence and technological transformations in the northern and southern Marquesas (Rolett 1990:363) and similar changes in ceramic styles and adz evolution in the Samoa-Tonga-Fiji area at similar times imply continuous inter-island contact (Davidson 1977, 1978, 1979). Tracking the spread and subsequent communi- cation of prehistoric Polynesian societies has been difficult without the widespread occurrence of pottery and obsidian, artifact classes that have proven especially useful for demonstrating interac- tion in the southwest Pacific. Without inferring contact from similar styles of fishhooks, adzes, and architecture, Polynesian archaeologists are left with few items to track empirically the intra- and inter- island movement of things. Connecting stylistic nodes, however, may only reflect convergent evolution (Kirch and Green 1987). I may risk being accused of taking a hard-line empiricist view of culture contact and change, but I think it crucial to confront the problem with separating stylistic from functional dimensions (Dunnell 1971:26-30) in relation to documenting the movement of exotic raw s - - 168 The To'aga Site materials and finished objects (e.g., pearishell and industrial stone) between and within island groups. The spatial and temporal dimensions of contact spheres (Irwin 1990:92; Walter 1990; Weisler In press b), can be delimited more accurately using exotic materials than by connecting artifact styles. Consequently, the distribution of exotic materials provides a better framework for evaluating the role of isolation and communication in shaping historical developments of island societies. This preliminary effort to document inter-island communication during the prehistory of the Manu'a Islands should take into account the region's geography. These islands are situated 100 km east of the most important source of adz material in central Polynesia, the Tatagamatau adz quarry, Tutuila (Leach and Witter 1987, 1990). The dis- tance and often turbulent ocean conditions between Tutuila and the Manu'a Group may have reduced the frequency of contacts (Hunt and Kirch 1988:155). THE COLLECTIONS Two questions are addressed by XRF analysis of the Ta'u and Ofu island assemblages and source material: (1) Did any artifacts originate at the Tatagamatau adz quarry complex on Tutuila Island located 100 km to the west?; and (2) What was the geochemical variation of adz material and unmodified basalt flakes from Manu'a sites? Al- though it seemed unlikely that I would be able to assign a provenance to all artifacts, I could determine if specific geochemical groups were correlated with certain tool classes. Interaction spheres could be delineated by identifying similar rock at different sites and, perhaps, different temporal periods. An additional goal of the XRF analysis was to generate geochemical data for the important Tatagamatau quarry complex on Tutuila and two local To'aga sources that may have been used. The Archaeological Sample The archaeological contexts of the Ta'u and Ofu island assemblages are described elsewhere (Hunt and Kirch 1988; Kirch et al. 1990; chapters 5 and 11, this volume) and are not reiterated here. The artifacts selected for compositional analysis con- sisted of all whole or fragmentary adzes, all flakes exhibiting polished surfaces and assumed to be adz fragments, and a sample of unmodified basalt flakes which represented the macroscopic vanability present. Table 12.1 presents characteristics of the analyzed artifacts. Specimens ranged from 1.2 to 421.9 gins and 22.08 to 136.30 mm in length. The thinnest sample was 3.53 mm. Source Material During their 1987 field season, Kirch and Hunt collected nine samples of source rock from the Tatagamatau quarry complex on Tutuila (Best et al. 1989; Leach and Witter 1987). On Ofu Island, they collected four samples from dike swarms at Mako Ridge and three samples from Fa'ala'aga. The Tatagamatau quarry complex consists of numerous cut-and-fill terraces, stone-working areas, pits, and possible fortifications along several ridges and scree slopes which together may cover more han 1 10 acres (Best et al. 1989; Leach and Witter 1987, 1990). The raw material derives from a dike com- plex at the foot of a waterfall at the head of a small gulch (P. Kirch, personal communication 1990, Leach and Witter 1987:39). Fine-grained dike rock may have been removed from dikes but most likely was prised from the soil and rock scree of slopes and from streambeds. Although I have not visited Tatagamatau, the geological setting appears to be similar to the Waiahole quarry on windward O'ahu (Dye et al. 1985), albeit on a much smaller scale. Two collections of source rock were made by Kirch and Hunt during a visit on July 3, 1989. Eight samples, consisting of large primary or secondary flakes with obvious bulbs of percussion, were collected at three locations along the main spur of the complex which trends NE-SW, and an additional flake was retrieved from Area 3 (figure 12.1). The exterior color of the source rocks varied from dark brown (lOYR 3/3), dark gray (1OYR 4/1), to dark grayish-brown (2.5Y 4/2) due to weathering and contact with lateritic soil. Fresh saw-cut surfaces are mostly dark bluish-gray (Gley card, SB 4/1) grading to dark gray (7.5R 4/0 and 2.5Y 4/0). Even prior to geochemical analysis, it was not likely that nine samples would document the total variability of quarry rock from so large a quarry, but this collection did contribute significantly towards developing a strategy for collecting additional material. Chemical Characterization of Manu'a Adz Material 169 Table 12.1 Basalt Artifacts Analyzed by EDXRF Lab Artifact Number Number Weight (g) Description Fiti'uta 1 1 1-1-S-4 11-2-S-1 11-2-S-2 11-2-S-3 11-2-S-8 11-2-9 11-52-S-9 13-S-3 13-S-5 13-S-6 13-S-9 13-1-S-36 13-2-3-1-1 13-1-9-2-4 13-1-13-87 13-1-16-1 13-1-16-11 13-1-16-11-81 13-1-16-12-83 13-1-20-4-118 13-1-20-5-121 13-1-20-5-122 13-1-20-5-123 13-1-20-6-139 13-1-20-6-140 13-1-20-6-141 13-1-20-7-143 13-1-20-9-147 13-1-21-7-151 13-1-22-2-171 13-1-22-2-173 13-1-22-2-174 13-1-23-7-194 13-1-27-5-39 13-1-27-5-40 13-1-27-5-41 13-1-27-6-42 24.8 70.9 41.4 35.3 21.4 33.1 60.9 133.1 48.4 44.7 33.8 44.4 63.9 63.4 421.9 0.7 5.5 1.5 127.9 20.5 3.4 1.2 4.4 11.3 11.4 128.4 56.7 3.2 20.6 59.3 3.7 2.1 0.4 211.4 89.4 128.1 2.1 9.5 Quadrangular, reworked adz Quadrangular adz Quadrangular adz, front section Quadrangular adz, butt section Quadrangular adz, front section Quadrangular adz, front section Quadrangular adz, butt section Quadrangular adz, front section Quadrangular adz, front section Quadranglukar (?) adz, front (?) section Quadrangular adz, front section Quadrangular adz Plano-convex adz Quadrangular adz, midsection Plano-convex adz Adz flake Adz fragment Unmodified flake Retouched flake Retouched flake Adz flake Unmodified flake Unmodified flake Uninodified flake Unmodified flake Plano-convex adz, front section Retouched flake Unmodified flake Unmodified flake Used flake Unmodified flake Adz flake Adz flake Plano-convex (?) adz Quadrangular (?) adz Quadrangular (?) adz, butt (?) section Unmodified flake Unmodified flake * = Probable source is Tatagamatau. 89-22* 89-25 89-26* 89-31 89-30* 89-28 9030* 90-33 89-18 89-34* 89-33* 89-32 89-19 89-17* 90-22 90-23 90-42 90-46 90-45 90-38 90-27* 90-37 90-41 90-34 90-48 90-29 90-36 90-40 90-43 90-49 90-44 90-25 90-24* 90-26 90-31 90-39 90-47 90-35 170 The To'aga Site Situated at about 350 m elevation and exposed by a recent road-cut, the Mako Ridge dike swarm on Ofu consists of relatively dense basalt. Weathering has smoothed the noimally angular, loose dike rock into sub-rounded cobbles which lie buried within a thick soil matrix. It is doubtful that this material was used for making stone tools, but four samples were collected for chemical analysis. The weathered exterior surface was gray in color (7.5YR 5/0) and fresh breaks, dark gray (2.5Y 4/0). A large concentration of dikes is found along a steep ridge towards the east end of Ofu. The Fa'ala'aga dike swarm (Stice and McCoy 1968) may have been a more important source of local, me- dium-grained basalt during prehistory. Angular rock from the eroding dikes form a large scree slope which descends to the To'aga coastal zone where stone is readily available (see Kirch, chapter 2). Three fresh samples were removed from dikes exposed by a modem road-cut (see fig. 12.2) to provide the general chemical composition of the dike swam although many dozens of dikes may exist each with varying chemical compositions. Surface color of these rock samples is very dark gray (2.5Y 3/0) with fresh breaks being gray (2.5Y 510). METHODS Because this specific technique of non-destruc- tive, energy-dispersive XRF had not been used previously with adz material from Oceania, each individual source sample was divided and analyzed by two techniques: destructive XRF using pelletized samples and non-destructive XRF using whole specimens. Destructive XRF analyzes a wider range of elements, and results are fully quantitative for most elements and comparable to other data sets now being developed for adz source rock in Polynesia The non-destructive technique focused on those elements that are more reliably detected with this procedure, and results presented are considered here as semi-quantitative. Destructive XRF Source samples from Tatagamatau, Mako Ridge, and Fa'ala'aga were analyzed by Dr. Peter R. Hooper at te Department of Geology, Washington State University, on 16 and 17 October 1989. The equipment and operating procedure are paraphrased from Hooper and Johnson (1987; this manuscript is available from Dr. Hooper or myself). Twenty- seven elements were analyzed on an automatic Rigaku 3370 spectrometer. For the oxides, SiO2, A1203, TiO2, FeO*, MnO, CaO, MgO, K20, Na2O, and P205, the weight percents are presented. For the other seventeen elements, the weight percent or ppm (parts-per-million) of the element are provided. All elements are analyzed on a single 2:1 lithium tetraborate:rock powder fused disk. Each element analysis is fully corrected for line interference and matrix effects of all other analyzed elements. The results are normalized and printed out with total iron expressed as FeO*. Whole rock samples were reduced with a tungsten carbide jaw crusher to small chips generally about 1 cm in average size. Chips were hand picked and reduced further in a Tema swingmill shatterbox with tungsten carbide surfaces where they were milled for two minutes. Rock powder weighing 3.5 grams was mixed with pure lithium tetraborate and emptied into a 34.9 mm-intemal-diameter, graphite crucible. The samples were fused in a muffle furnace at 1000(C. After grinding, the samples were fused a second time. The lower, flat surfaces of the fused disks were ground on coarse (240) grit and fine (600) grit. Then, they were washed in an ultrasonic cleaner, dried, and loaded into the XRF spectrometer. The x-ray intensity of each element from unknown samples was measured and compared to values of two beads from each of eight international rock standards (U.S. Geological Survey [U.S.G.S.] standards PCC-1, BCR-l, BIR-l, DNC-1, W-2, AGV-1, GSP-l, and G-2). Standards were run and recalibrated after processing between one and two hundred unknown samples. U.S.G.S. BCRP-84 and GSP-1 were used as internal standards and routinely run after every twenty-eight samples to check instrument performance. Drift between standard calibrations was due almost entirely to iron which produced slightly higher values for other elements presented as oxides, weight %. When thirty analyses of the U.S.G.S. standard BCRP were compared to accepted values, the precision or standard deviation of oxide values to known measures is 0.22% ppm. Values reported for thirteen trace elements in ppm were below 6% ppm at one standard deviation, except barium (16.2%) and vanadium (10.0%1o). MTe highest precision was achieved with Rb, Sr, Zr, Y, Chemical Characterization of Manu'a Adz Material 171 Map of Tatagamatau quarry showing locations of source samples (after Best et al. 1989, fig. 21). Oa,Cu, and Zn. Ni, Cr. Sc, V, and Ba should be xgarded as semi-quantitative below the 30 ppm kvel. Rb, Sr, Zr, Nb, and Y have satisfactory precisions and accuracies down to one to three ppm, while Y and Nb could be measured to 0.1 ppm. Evaluation of accuracy suggested that variation between different samples of standard powder or nonhomogeneity resulting from sample preparation is greater tanm inaccuracies caused by inadequate matrix and interference correction. Non-destructive Energy Dispersive XRF Portions or splits of all source samples from Tatagamatau, Mako Ridge, and Fa'ala'aga were analyzed by Weisler along with thirty-eight artifacts during four runs in late 1989 and early 1990. Sam- ple preparation consisted of submerging specimens Figure 12.1 172 The To'aga Site 70 T Tatagamtau3 subsource s Mako (x) A Tatagamatau A Fa'ala'aga (x) ;ubsource Mako (x) A A Fa'ala'aga (X) 40 w 5 . . a 50 60 70 80 90 100 110 120 130 Nb/Sr X 1000 Figure 122 A compaison of source rocks analyzed by non-destructive x-ray flourescence using whole specimans and XRF with fused disks (indicated by italic type). in a sonic bath of distilled water for up to one hour then air drying. Some artifacts, deeply stained with latentic soil, were scrubbed with brushes as well. Carbonate erutstations on a few artifacts were removed with a 10% solution of HCI and then rinsed in distilled water. The EDXRF instrument is limited to the maximum weight and size of specimen that can be analyzed. Although the opening of the sample holder where the x-ray beam is directed though to the specimen is 32 mm in diameter, careful placement of the artifact or rock in the EDXRF instrument may accommodate samples (as in this analysis) up to 421 gins and 136 mm long. Molding clay can be used to secure specimens on the tray which can hold up to twenty samples. Each artifact was carefully examined to locate the flattest surface for analysis that could be accommodated within the space parameters of the sample chamber (see illustration in Bouey 1991:fig. 5). Source samples were cut to size and only fresh saw-cut surfaces were analyzed. Laboratory facilities and equipment were provided by the Department of Geology and Geo- physics, University of California, Berkeley. The same equipment and nearly identical operating conditions were used as reported by Hughes (1986:esp. 25-30) in his comprehensive study of California and Oregon artifactual and source obsid- ian. The XRF spectrometer consisted of a Spectrace 440 energy dispersive machine and 572 power supply (50kV, imA), 534-1 pulsed tube control, 588 bias/protection module, 514 pulse processor or amplifier, Tracor Northern 1221 100 mHz ADC converter, and a Tracor Northern 2000 computer based analyzer with an LSI-1 1 microcomputer (Hughes 1986:25). The Si(Li) solid state detector with 144 eV resolution (FWHM) at 5.9 keV in a 30 mm squared area was used for detecting all x-ray intensities. For analysis of trace elements in the mid- z energy range, a rhodium (Rh) x-ray tube was used for primary x-ray excitation at 30.0 keV, .20 mA pulsed, with a .05 mm Rh primary beam filter in an air path at 300 seconds livetime. Analytical lines used for analysis were Ni (Ka), Cu (Ka), Zn (Ka), Ga (Ka), Pb (Lb), Th (La), Rb (Ka), Sr (Ka), Y (Ka), Zr (Ka), and Nb (Ka). For rare earth elements, an Am24l 100 mCi radioscope source was used in the 20-60 keV range at 300 seconds livetime (Hughes 1986:26), and analytical values were derived from the Ka lines of Cs, Ba, La, Ce, Pr, Nd, and Sm. The x-ray tube was operated at 15.0 kV, .40 mA pulsed, with an Al primary beam filter in a vacuum path at 200 seconds livetime for Ka lines of Fe, Mn, and Mi. The elemental data as reported represent one analysis per specimen. While Bouey (1991) has demonstrated some variability with multiple analy- ses of the same obsidian specimens when values axe presented in strictly quantitative (ppm) data, these 0 0 Co L.. N 60 a 50' 40' Chemical Characterization of Manu'a Adz Material 173 effects are minimized when presenting data as ratios (see, for example, Jack and Carmichael 1969). The XRF technique has its greatest accuracy in detecting elements in the mid-z range (e.g., ru- bidium, strontium, yttrium, zirconium, and niobium) which was confirmed by this study for both destruc- tive and non-destructive sample preparations analyzed on their respective instruments. It is fortunate that these elements are of particular interest to igneous petrologists (Cox et al. 1979:332). For pressed pellets or fused disks these particular mid-z elements have satisfactory precisions and accuracies down to 1 - 3 ppm, while niobium and yttrium could probably be measured to 0.1 ppm (Hooper and Johnson 1987). In basalt, mid-z elements are also present in sufficient concentrations to be easily measured, and detection in whole, unaltered speci- mens ranges from 10 ppm to 100% with an accuracy of ? 2 - 5% under favorable conditions (Parkes 1986:153). Lighter elements are not only harder to detect and measure by XRF, but they are usually not present in great abundance. Iron, titanium, and magnesium values, reported here, were detected under vacuum. Lead (Pb), although detected during the mid-z analysis, was not used to discriminate sources or characterize artifacts due to its presence in low abundance and its susceptibility to atmospheric contamination (Flanagan 1969:82). Rock standards have been used in routine XRF analyses for about twenty-five years Flanagan 1969), and currently at least 272 geostandards are in use worldwide (Govindaraju 1989). Standards are important for calibrating the XRF instrumentation for matrix effect corrections and for monitoring the precision and accuracy during analysis (Germanique and Briand 1985). Using standards thought to be close in composition to the unknowns limits the efficacy of matrix correction programs to a restricted range of elements. Conversely, many different standards provide a greater range of elements and values for calibrating the program and evaluating unknowns. Therefore, more than ten internal standards were used to calibrate the Spectrace 440 machine used in this study. U.S.G.S. standard RGM- 1, a rhyolite from Glass Mountain, California, was used to monitor precision and accuracy during analysis and the results are presented in table 12.2. Ppm values reported by Govindaraju (1984, 1989) ar preferred "working values" which are the average of at least forty results from more than four techniques of analysis (1989:7). The accuracy and precision values reported for this study in ppm and selected ratios are reasonably close to accepted standard "working values." Due to its cryptocrystalline texture and homoge- neous distribution of elemental abundances, it is not surprising that obsidian has garnered most analytical attention in XRF studies. Large grain sizes can Table 12.2 Evaluation of Analytical Accuracy and Precision for U.S.G.S. Standard RGM-1 Govindaraju Govindaraju This study (n=4) 1984 1989 Accuracy Precision Rb 155 149 159.9 156.3-163 Sr 100 108 107.3 106-108 Zr 200 219 227.2 225.3-229.9 Y 25 25 27.7 25.9-29.2 Nb 9.4 8.9 10.3 9-11.7 Pb 21 24 28.6 27.1-31.1 Th 15 15.1 20.4 18.9-22.1 Zr/Sr 2 2.03 2.12 Nb/Sr 0.094 0.082 0.096 Govindaraju (1984, 1989) pressed powder samples; this study, whole specimens analyzed. 174 The Tolaga Site distort XRF analysis of whole samples. Pressed pellet samples are pred by reducing whole rock to a fine powder estimated to be 90% less than 400 mesh (Bice 1980:19) or ca. 40 microns. However, very fine-grained basalt may have more than 130 grains per mm (ca. 160 microns per grain). While this grain-size is early four times larger than pressed pellet samples, it has not been demonstrated that grains of this size adversely effect XRF analysis. This subject should be investigated further. The distribution of elemental abundances within a single basalt specimen may not be as homogeneous as obsidian, but a comparison of the date in table 12.3 is quite instructive. Here, the distribution of selected elements within a quarry (1.8 hectares in size) are indeed quite regular, and perhaps this is more so for individual rocks or artifacts. Another factor which can affect XRF analysis is an uneven specimen surface. Analyzing adz mate- rial has a distinct advantage, however, especially over bifacially flaked obsidian artifacts, because, almost by definition, most adz surfaces are ex- tremely flat and, in many cases, are ground to a near mirror-like finish. Recalling that fused disks are finished by 240 (coarse) and 600 (fine) grit prior to XRF analysis (Hooper and Johnson 1987), 1 exam- ined under 10-40X magnification source rock from eight west Moloka'i Island basalt quarries whose material exhibited a range of textures and phenocryst sizes and densities. These samples were polished with 600 grit mesh and compared to the adz material in the present study. The prepared specimens had uniform, smooth surfaces and occasional striations formed by disintegrating phenocrysts that had been trapped between the rock and grinding plate. Al- though the artifacts were not as uniform in contour and had many more striations, portions of the artifacts were as smooth-if not more finely pol- ished-an the prepared specimens. Therefore, careful selection of artifact surfaces for EDXRF analysis may limit the amount of analytical distortion caused by uneven sample surfaces. To reduce or eliminate the effects of uneven sample surface, many researchers have advocated presenting elemental abundance values as ratios. "In spite of variations in effective sample surface of randomly broken pieces or loosely packed grains, relative intensities may be very precisely deter- mined" (Jack and Carmichael 1969:30; see also Table 12.3 Variation of Oxides and Elements from the Mo'omomi Basalt Quarry, Hawaiian Islands Mean Range n=10 Oxide (weight %) S102 46.16+0.12 AL203 15.72 t 0.04 T102 4.17 ? 0.01 FEO* 14.21 ? 0.20 MNO 0.19 ? 0.01 CAO 8.53 ? 0.03 MOO 6.32 ? 0.08 K20 0.92 ? 0.02 NA20 3.52 ? 0.07 P205 0.62 ? 0.01 Element (ppm) Ni 68.30 ? 1.90 Cr 5.40 ? 1.69 Sc 18.50 ? 2.58 V 302.20 ? 7.70 Ba 215.50? 20.69 Rb 14.40 ? 0.92 Sr 786.00 ? 6.62 Zr 272.50 ? 2.91 Y 37.80 ? 0.98 Nb 27.09 ? 1.19 Ga 23.30 ? 1.79 Cu 11.80 ? 5.62 Zn 133.30 ? 1.90 Pb 4.30 ? 1.35 La 18.80? 9.05 Ce 62.00 ? 9.89 Th 0.80 ? 0.98 45.96-46.36 15.66-15.81 4.157-4.183 13.91-14.04 0.178-0.202 8.50-8.58 6.18-6.42 0.90-0.95 3.40-3.67 0.613-0.634 64-70 3-8 15-23 283-310 187-258 12-15 775-798 268-278 37-40 25.0-28.9 21-26 4-21 131-137 2-6 2-31 49-80 0-3 *total iron Analyst: Dr. Peter R. Hooper, Dept. of Geology, Washington State University Stross et al. 1968:82). Sheets et al. (1990:149-50) concur that errors intmduced by sample size and shape are largely cancelled by 'the use of abundance ratios of elements having nearly the same energy (e.g., Rb, Sr, and Zr)." These observations were Chemical Characterization of Manula Adz Material 175 taken into consideration for the present study. Ever since the pioneering work of Parks and Tieh (1966), selection of elements for obsidian analysis has been fairly standard. Regarding data reduction, however, two "camps" have emerged somewhat recently. For reasons cited above, elemental abundances are usually presented in semi- quantitative ratio form including ternary diagrams (e.g., Best 1984; Jack and Carmichael 1969; Shackley 1988; Stross et al. 1968) or scatter plots (see fig. 12.3, this study; Jack and Carmichael 1969). Hughes (1986, 1988a, b) suggests that simple bivariate plots of zirconium and rubidium in ppm are sufficient for distinguishing raw material and determining artifact provenance in some regions. However, Bouey (1991) demonstrated that multiple analyses of the same specimens produced "widely divergent determinations in ppm concentrations," while ratio level presentation reduces this error significantly (cf. Jack and Carmichael 1969; Sheets et al. 1990). Hughes (1986) and Walter (1990) have applied statistical clustering programs to define geochemical groups. Multivariate statistics are quite appropriate for some problems, however their use eliminates any consideration of the elemental abundances (either in ppm or at a nominal level) as having any geological value. For example, on west Molokai, Hawaiian Islands, relatively high values of Y signal one particular quarry (David Clague, personal communication, 1989) and, simple bivariate plots of silica and total alkalis can be very informa- tive for determining sources and characterizing geochemical groups. Statistical clustering tech- niques should only be used after the data have been examined for geological information that can inform on the "provenance environment," that is, the geology of the suspected interaction sphere. Con- versely, groups created by statistical manipulation are an end in themselves. RESULTS The results of te fused disk XRF analysis of sixteen source rocks from Tatagamatau, Mako Ridge, and Fa'ala'aga are presented in tables 12.4-6 and summarized in table 12.7. According to a recent comprehensive and international evaluation of the systematics of igneous rocks (Le Maitre 1989), the Tatagamatau and Mako Ridge rocks are classified .125T .115 .105 .095 .085 CO, 0 D .075 .065 .055 .045 0 Tatagamatau Fa'ala'aga (x) subsource * /v /0.0 0 @0 0 * 0 A Mako (W I . -I .045 .055 .065 .075 .085 Y/Sr Figure 12.3 A wide dispersion of data points results when ratios of rubidium, strontium, and yttrium are used to assign speci- mens to possible sources. chemically as hawaiites and the Fa'ala'aga rock as basalt. Rock names were assigned by plotting silica values between total alkalis. Between sources, marked differences are found with the oxides, A1203, FeO, CaO, and P205, and most trace ele- ments. The samples from the Tatagamatau source reveal important intra-source variability between Areas I and 3. Oxide values for A1203, MnO, Na2O, and P205 demonstrate marked differences as well as the trace elements Ni, Cr, and Cu. Until additional samples collected from several areas of the 10-acre Tatagamatau quarTy complex are analyzed and the geochemical variability of source rock is understood in greater detail, Ni and Cr may well be significant trace elements for demonstrating intra-source variability; tatw is, at least between Areas I and 3. 176 The To'aga Site t- C t. .l NO tn _ C-4- e qt qC CoS Wo C4 Co o ooo en IC Nb %A xob o4 %A WI-- ,, %enao%0 oooo x ?l_ or_ t} N -O O- so 0 O- t- N 'v . 0\ w C2 q0 - O-" t N A t- m _) _- t- w e-O X m W)0i T _4 - , A ^ ml C 14 1", ,r ? ? 8 tx "" ? A"'EDn Rt en 4 % o -i i t- r- t (n ."-4 "4 en "4 - ?. in C1I4 0% O 00 '0 ' q en t . f r - en 4 i fo 0 _ - CD *b qt *) X X wi 00 00 e C C4 Co 8 7- c7 %n _ 4 _- "' o 8 _- M x oi co 8 v4 W) wr t- w )00% "4 ~ t..: V 4 qt 0 0 O* m W m t- F mO -n -O O-~o "-4 "-4 0 l m N 4 o \0 C4 t oW CNqt 0 t-X t4 Wo tn ON O "400C~~0~%00 " rT . . -j _ f C t _ ^ _ o0 0 J; e O en tl-W W-O 0% 0% %0t _4 *j %n " 4 . r 0D 0%N. to Xou en 0 4 0 0 0 - e (C: A X Xo? t-O V- o o, Ir) V"4 co 1? -4 00 -4 te Ic- ) _ " 4n ? - Q m x) w ( nx t_4 N en %n qt 00C) C en)" en C w 000% o 00 %o "400 n-t 4 - 00 0 ~ ie~ f~'% 0 X ) " 4 t-) " O- > Bo "I ' 8 W) t-- r -g C 00 O o m Ch N en 00 ) o4 x en o NW^ o o 0 0t o Rt en t " _ o en N El> n (s ;x o7A 0x o o r 0 cx " N en qt Out en 14 ;^xe %000% 0 e ooQC "40 0 0 o O- > . _ t t o -O A LL iu i izso "4 "4 _ w o eq w o t" 4 a 9 w V- ONI _I ne TO 0 U r ~o cn 4E. PC N 8 N E as 0 a- co e4 N- "4 a 00 "4 "M4 A "o V0% A 0a 0% 00 0% A 00 00 00 No 00 0% 00I 01 z U) 00 ON "- %0 * - "4 S W) 00 I :a r- *A, *Is Chemical Characterization of Manu'a Adz Material 177 Table 12.5 Geochemistry of Mako Ridge Source Rock (Fused Disks) Sample No. Range Mean 89-1 89-2 89-3 89-4 Oxide (Weight %) SiO2 46.48 A1203 18.22 TiO2 3.552 FeO* 12.10 MnO 0.21 CaO 6.65 MgO 4.44 K20 1.89 Na2O 3.45 P205 0.988 Total 97.980 Element (PPM) Ni 0 Cr 3 Sc 23 V 228 Ba 433 Rb 46 Sr 660 Zr 433 Y 97 Nb 96.3 Ga 36 Cu 0 Zn 163 Pb 7 La 86 Ce 141 Th 7 49.84 17.22 3.238 10.68 0.193 8.42 4.41 1.79 4.18 0.897 100.868 0 0 13 206 421 39 883 397 46 82.7 30 2 154 3 68 121 7 43.44 19.72 3.886 13.20 0.231 5.50 4.42 1.98 2.55 1.070 95.997 0 8 19 246 471 52 481 469 61 104.8 39 0 188 7 62 144 8 48.12 18.33 3.473 11.48 0.208 7.29 4.24 1.37 3.84 0.960 99.311 0 1 12 204 450 14 856 429 58 89.3 35 0 167 6 87 161 6 43.44-49.84 17.22-19.72 3.238-3.886 10.68-13.20 0.193-0.210 5.50-7.29 4.24-4.44 1.37-1.98 2.55-4.18 0.897-1.070 0 0-8 12-23 204-246 421-471 14-52 481-883 397-469 46-97 82.7-104.8 30-39 0-2 154-188 3-7 62-87 121-161 6-8 46.97 18.35 3.537 11.87 0.211 6.966 4.38 1.76 3.51 0.979 98.533 0 3 16.75 221 443.75 37.75 720 432 65.5 93.275 35 0.5 168 5.75 75.75 141.75 7 * total iron Analyst: Dr. Peter R. Hooper, Dept. of Geology, Washington State University, 16-17 October 1989. EDXRF analysis of Tatagamatau, Mako Ridge, ad Fa'ala'aga source rock are presented in tables 12.8-10; mean and range for sources are summarized in table 12.1 1. Elemental values are presented with one standard deviation which represents the uncer- tainty of counting statistics at 300 seconds livetime. A comparison of EDXRF and fused disk samples for source means and the Tatagamatau source envelopes is illustrated in figure 12.2. Zr/Sr values are very comparable, whereas Nb/Sr ratios are higher for fused disk values. This probably corresponds to the greater detection efficiency (machine sensitivity) for niobium by the Rigaku spectrometer (Hooper and Johnson 1989); or it could be a sample thickness problem since niobium x-rays are excited as deep as 5-7 mm into the specimen (David Clague, written 178 The To'aga Site Table 12.6 Geochemistry of Fa'ala'aga Source Rock (Fused Disks) Sample No. Range Mean 89-14 89-15 89-16 Oxide (Weight %) SiO2 46.89 A1203 13.68 TiO2 5.412 FeO* 13.04 MnO 0.172 CaO 10.54 MgO 4.72 K20 1.68 Na2O 3.00 P205 0.674 Total 99.808 Element (PPM) Ni 57 Cr 24 Sc 21 V 344 Ba 300 Rb 45 Sr 644 Zr 323 y 34 Nb 68 Ga 28 Cu 104 Zn 128 Pb 5 La 40 Ce 115 Th 4 46.69 13.79 5.426 12.93 0.178 10.56 4.8 1.73 3.13 0.684 99.918 88 25 25 357 305 46 642 330 36 71.1 30 117 138 7 45 108 6 46.79 13.80 5.335 13.06 0.174 10.75 4.76 1.60 2.97 0.627 99.866 72 38 19 359 307 42 651 313 35 66.6 29 101 129 6 31 104 3 46.69-46.89 13.68-13.80 5.335-5.426 12.93-13.06 0.172-0.178 10.54-10.75 4.72-4.80 1.60-1.73 2.97-3.13 0.627-0.684 57-88 24-38 19-25 344-359 300-307 42-46 642-651 313-330 34-36 66.6-71.1 28-30 101-117 128-138 5-7 31-45 104-115 3-6 46.79 13.76 5.391 13.01 0.175 10.62 3.16 1.67 3.03 0.662 98.268 72.3 29 21.7 353.3 304 44.3 645.7 322 35 68.57 29 107.3 131.7 6 38.7 109 4.3 * total iron Analyst: Dr. Peter R. Hooper, Dept. of Geology, Washington State University, 16-17 October 1989. communication, 1990). The larger source envelope for EDXRF may relate to the greater variability in specimen surface. After selecting elements for analysis (based on analytical precision, accuracy, and sufficient elemen- tal concentrations), assigning artifacts to source was facilitated initially by trial and error. Figure 12.3 illustrates a wide dispersion of data points with most artifacts plotting outside the source envelope when Rb/Sr and Y/Sr ratios are used. Figure 12.4, how- ever, using ratios of Zr/Sr and Nb/Sr, is much mole useful for assigning artifacts to the Tataganatau source. The source envelope is delimited by taking into account the variability in analytical precision. Chemical Characterization of Manu'a Adz Material 179 Table 12.7 Geochemistry of Fa'ala'aga, Mako Ridge, and Tatagamatau Source Rock (Fused Disks) Fa'ala'aga (n=3) Mako Ridge (n=4) Tatagamatau (n=8) Tatagamatau sub-source Range Mean Range Mean Range Mean (n=l) Oxide (Weight %) SiO2 46.69-46.89 A1203 13.68-13.80 TiO2 5.335-5.426 FeO* 12.93-13.06 MnO 0.172-0.178 CaO 10.54-10.75 MgO 4.72-4.80 K20 1.60-1.73 Na2O 2.97-3.13 P205 0.627-0.684 Element (PPM) Ni 57-88 Cr 24-38 Sc 19-25 V 344-359 Ba 300-307 Rb 42-46 Sr 642-651 Zr 313-330 Y 34-36 Nb 66.6-71.1 Ga 28-30 Cu 101-117 Zn 128-138 Pb 5-7 La 31-45 Ce 104-115 Th 3-6 72.3 645.7 322 35 68.57 29 107.3 131.7 6 38.7 109 4.3 0 0 1t-JL 481-883 397-469 46-97 82.7-104.8 30-39 0-2 154-188 3-7 62-87 121-161 6-8 ) 1 .0 720 432 65.5 93.3 35 0.5 168 5.8 75.8 141.8 7 0 0 1.1 18.8 207.8 290.5 _J 42.9 694-708 702.9 348-362 355.6 45-49 47.6 52.4-56.0 54.5 27-31 29.4 0-9 1.5 165-178 171.3 4-8 6.1 13-50 35.3 88-110 100.3 3-7 5.25 * = total iron Analyst: Dr. Peter R. Hooper, Dept. of Geology, Washington State University, 16-17 October 1989. Nine artifacts fall within the envelope and four others are very close. Taken together, fifty percent of the adzes and other artifacts with one or more polished surfaces (adz flakes) can be assigned to the Tatagamatau quarry on Tutuila. Weathered surfaces of these specimens were gray (2.5Y 5/0; 2.5YR 5/0; 7.5YR 5/0), dark gray (2.5Y 4/0), to very dark gray (7.5R 3/0; 2.5Y 3/,0 7.5Y 3/0). Freshly broken surfaces on two specimens were very dark gray (2.5Y 3/0; 7.5Y 3/0). Additional samples from this quarry will undoubtedly define a much larger source envelope since the Tatagamatau sub-source is well outside the limits of the eight samples used to define the geo- chemical dimensions of the quarry. This under- scores the need to collect sufficient samples to define the geochemical variability of adz quarry sources. The diagonal line in figure 12.4 separates 46.79 13.76 5.391 13.01 0.175 10.62 3.16 1.67 3.03 0.662 43.4449.84 17.22-19.72 3.238-3.886 10.68-13.20 0.193-0.210 5.50-7.29 4.244.44 1.37-1.98 2.554.18 0.897-1.070 46.97 18.35 3.537 11.87 0.211 6.966 4.38 1.76 3.51 0.979 49.56-50.56 15.35-15.62 3.385-3.467 12.34-12.90 0.173-0.179 7.56-7.70 4.55-4.77 1.54-1.62 3.95-4.09 0.783-0.814 50.02 15.45 3.413 12.68 0.177 7.63 4.69 1.58 4.02 0.800 49.32 13.86 3.649 12.40 0.167 7.80 7.06 1.79 3.45 0.710 173 181 27 233 331 50 683 325 37 47.9 26 31 153 5 43 99 2 180 The To'aga Site 0 r go mg 00 mt 00 eq o d) el C-4 4 o en , ent _- en x m 0 ~o 0 in ON x - 0 t- m m o t- V4 en oo1 06 t- _ _ cl en cli cr er t-: o6 > V- -4 qt n en % C1 t- en V-4 .- 00 ^o 4 e n7 en t ^ V -4xv CD t -4 Q v- 6 6 + +1 WI)x C CDC ( q ur + +1 *i 6 + +1 * i C> - Q M o m o m; *1 * W) c5s * * Q W- C>W e en ON tn 4 eq en +1 en %n 0 etiM t - 1- t.o +1 ,4 -H C-4 9Cs +1 M t e oN m o" Go ok Co ot o8 C4 %n W r- (: 6 od W) in in ene ,t? +1 wI +0 08 W en 00 .1 + WI Ch Ie +I t r- %O ON O- V-4 P - eq 0 ON z o C' q to- 00 + +1 eno +, +1 + *o v (D t- 6 +1 C C V-4 C114 >o t- IC *4 +14 +D C% F_ w :;V- c>WqN t t- 6 Cf oo +1 ql - V t Mt 0e C> r- 0 O' oo: cr ~ n ON r- V) n o4 en 0 -- 0 r- 1* o os t M 00 e4 TNO 0 :a co F* .2 C1 WI, en +1 C14 eq IT +1 Nt 0 N t1 m ?- C4 +1 _ z *- * * r.8 ? - + qq H Q W) 6 +1 T-4 0 14t 6 +1 W) 14 C q1t C; +1 10 -4. cr) -t o) o o ,t _ 0 Vt ._ 0 I 0 00 ~o C It +1 +1 +1 C- 4o tt +1 kn V) X- j" ON W) 1 qqt C4C1 (I in tn 0 It 0 +1 C x %6 C 00C 00 en o It CN r- en A C14 CN (ON M m +I I V.-I ko 00 C1 m A tn Af C114 0 p 0 -2 = IL,2 I en I en eq M .ON .en .C , M It tn M M -4 C4 WI In C4 V-4 12,  s en V-48 C14 Chemical Characterization of Manu'a Adz Material 181 O XD tn CN 00 W N m It V- b E 0 N ofi -- M u 'r(..if0 0% N ur a tno F-~t % - 06 t "- r-- 'R W ,^ 0 r- 41 N- 0; 00 00 e 00 CN V0 OQ- t- 0\NXN0 ( - 0_ 44414 i 414444cq -o ON 0 N t W- V - o r- eni C1 Ch Wo< o= 'Rt cn --q . 141 m414 Co- Sf m ) q rt 0 en m ON et 414 "i4 m * N " - X00 414n 14141 *e *i *i 0 * * *en -'0 00' t- -- q ~t r 4 004 s ON 11- qq cn W) (14"t At x r r- It v N-v00 O~N D V _ I- N '0 e.. Ce i m X 't W 'o 00 i _ +I - 't c +1 + vo +1 +1 + _- t_ O en C- _. _~ t- ..4 _ te V- N m )N V-4 o 00 O -cr tn 4-141 4 4 W~ Vt- M IR It i m t- c 00 en so " _ V N cf CN _ ) It M -r '0 o * v - wi -_ * * 4141414141414 0 00 00~ +14+1 CN4 00 - - N * e 0 r~ 414+1 q* en t o cn N _n ^ t mO - "- m 6 ~o N 0 o x0 ,t o'4 x- N O N _4 * -- 6 V-O ._ Ut I- 0 oof. 0 0 Is 0% am' 41 n +00 '0 ~o ri * . ONO%\ ON + d 4 1004 01 41~ ,; fi > ON aN ? i o O in o C41 6 41 00 o o 0 z Oa E as W) le co en 00 N 00 41 - V- 'I.) * 4 Ch4 41 o . 6 41 en _4 (Ni 41 VNI (Ni4 oN 00 c0 0. 0 0 C.) P v. 2 w z m 1: (n t? >. z 0 u 9 u u ;zQtaC < 182 The To'aga Site 70. Tatagamatau source rock from adz material deriving . / . from the dike sources of Fa'ala'aga and Mako Ridge Mako M on Ofu and from all unmodified flakes. The source 8 Tatagamatau - for the flakes was probably the scree slope below the X_ ss * **A Fa'ala'aga R) Fa'ala'aga dike swarm. There are indeed several L A :.. flakes that plot close to this source mean. The dike rUi * subso~r :e * * /;;; * Ssource rock and flakes recovered from the site are all 40 medium- to coarse-gniined, and eight specimens dike rock (61.5% of the total flakes) have one or more natu- *0 rally flat surfaces characteristic of dike rocks. 50 60 70 80 90 Several flakes, however, exhibit retouching and one Nb/Sr X 1000 distal margin is smoothed from use. These medium- to coarse-grained rocks were probably not conducive Figure 12A Aiac and source rocks plotted by raios of to adz manufacture, and it is perhaps noteworthy that zirconium, standun, and niobiumn rslng all fine-grained rocks were manufactured into, or are in many specirnes plottng widin the Tatagamatau source envelope. Table 12.10 Geochemistry of Fa'ala'aga Source Rock (Whole Specimen) Sample No. Range Mean (PPM) 89-14 89-15 89-16 Oxide TiO2 50476.8 ? 358 50757.1 ? 366.8 47928.1 ? 334 47928.1-50757.1 49720.7 FeO 168791.1 ? 1153.7 171686.3 ? 1191.4 156480.4 ? 1047.1 156480.4-171686.3 165652.6 MnO 2391.4 ? 44.6 2396 ? 45.5 1978.4 ? 40 1978.4-2396 2255.3 Element Ni 84.6 ? 7.9 103.2 ? 8.1 89.7 ? 7.1 84.6-103.2 92.5 Ba 234.4 ? 6.1 230.3 ? 7.1 233.3 ? 6.1 230.3-234.4 232.67 Rb 45.4 ? 2.4 48.4 ? 2.6 42.9 ? 2.4 42.9-48.4 45.57 Sr 691.6 ? 6.2 675.3 ? 6.6 704.8 ?6.3 675.3-704.8 690.57 Zr 362.4 ? 5.3 359.2 ? 5.7 341.7 ? 5.2 341.7-362.4 354.43 Y 40 ? 2.8 36.1 ? 3 39.1 ? 2.7 36.1-40 38.4 Nb 51.8 ? 3.7 60 ?4 52.3 ? 3.6 51.8-60 54.7 Ga 25.7 ? 2.7 30.1 ? 2.6 33.9 ? 2.5 25.7-33.9 29.9 Cu 135.9 ? 6.5 134.7 ? 6.4 134.4 ? 5.7 134.4-135.9 135 Zn 143.1 ? 5.5 137.6 ? 5.4 115.9 ? 4.6 115.9-143.1 132.2 Pb 7 ? 1.8 7.4 ? 2 7.6 ? 1.8 7-7.6 7.33 La 40.1 ? 3.3 39.7 ? 3.9 26.8 ? 3.2 26.8-40.1 35.53 Ce 70.2 ? 3.9 81.3 ? 4.5 91.2 ? 3.9 70.2-91.2 80.9 Nd 43.7 ? 3.9 40.8 ? 4.4 34.4 ? 3.9 34.4-43.7 39.63 Cs 0 0 0 0 0 Fr 8.3 ? 3.5 7.9 ? 4 9.2 ? 3.5 7.9-9.2 8.47 Th 0 0 0 0 0 Co 0 0 0 0 0 As 1.7 ? 0.5 1.8 ? 0.4 1.6 ? 0.4 1.6-1.8 1.7 Analyst: Marshall Weisler, September 1989. Chemical Characterization of Manu'a Adz Material 183 Table 12.11 Geocheniistry of Fa'ala'aga, Mako Ridge, and Tatagamatau Source Rock (Whole Specimen) Fa'ala'aga (n=3) Mako Ridge (n=4) Tatagamatau (n=8) Tatagamatau sub-source (PPM) Range Mean Range Mean Range Mean (n=1) Oxide 1i* 4.793-5.076 4.972 3.033-3.393 3.147 3.160-3.490 3.320 3A14 FeO* 15.65-17.17 16.57 13.65-14.78 14.33 14.60-16.85 16.00 15.55 MnO* 0.198-0.240 0.226 0.238-0.285 0.258 0.206-0.243 0.230 0.220 Element Ni 84.6-103.3 92.5 nd nd nd nd 188.0 Ba 230.3-234.4 232.6 327.1-354.2 340.0 209.2-262.4 226.7 259.2 Rb 42.9-48.4 45.6 14.6-52.2 40.4 36.4-46.0 40.7 56.5 Sr 675.3-704.8 690.6 551.6-959.7 791.9 718.1-769.7 750.3 738.6 Zr 341.7-362.4 354.4 466.1-497.1 477.7 376.4-409.6 395.8 357.0 Y 36.1-40.0 38.4 48.2-94.8 66.1 47.1-53.1 49.7 38.8 Nb 51.8-60.0 54.7 66.4-75.3 71.2 41.0-47.3 44.4 32.7 Ga 25.7-33.9 29.9 32.1-33.9 33.0 27.3-36.8 30.8 28.3 Cu 134.4-135.9 135.0 8.3-26.4 16.2 19.1-95.3 42.0 61.2 Zn 115.9-143.1 132.2 161.6-178.3 168.7 159.0-191.1 177.7 149.2 Pb 7.0-7.6 7.3 5.9-9.8 7.8 0.0-13.1 8.0 nd La 26.8-47.6 38.0 47.2-66.5 57.3 28.3-41.4 33.9 33.1 Ce 70.2-81.3 77.6 115.2-131.3 121.8 70.5-84.5 75.7 75.9 Nd 34.443.7 39.6 60.2-75.3 68.8 34.1-52.3 45.1 38.3 Cs nd nd 0.0-2.4 0.6 nd nd nd Fr 7.9-9.2 8.5 10.1-15.1 11.6 0.0-19.4 8.9 nd Th nd nd 0.0-13.8 3.5 0.0-13.1 1.6 nd Co nd nd nd nd nd nd nd As 1.6-1.8 1.7 1.2-2.2 1.6 0.8-2.1 1.3 1.5 * x 1i04 Analyst: Marshall Weisler, September 1989. the by-products of, adz production. Geochemical data for all artifacts are presented in table 12.12. Nine artifacts assigned to the Tatagamatau adz qwary complex on Tutuila are from hee archaeo- logical sites on Tau Island and two localities within the large To'aga coastal habitation area (see table 12.1). Unfortnately, only three artifacts are from excavated contexts, and the remaining are surface finds. All stylistically diagnostic specimens are quadrangular-sectioned adzes or fragments dating to the late prehistonic penod (Green 1974; Green and Davidson 1969). By connecting the six sites with adz material that originated from the Tatagamatau quarry, we can document several nodes of an interaction sphere between the islands of Tutuila, Ta'u, and Ofu. Theseyresults seem quite promising for additional provenance studies of the kind em- ployed here. 184 The To'aga Site Table 12.12 Geochemistry of To'aga and Ofu Island Basalt Artifacts (Whole Specimen) Trace Element Concentrations (PPM) Artifact Lab Number Number Pb Th Rb Sr Y Zr Nb Fiti'uta 1 89-22 11.3 ? 1.3 0.0 65.4 ? 2.4 832.5 ? 6.3 52.0 ? 2.7 438.8 ? 5.2 50.5 ? 3.3 11-1-S-4 89-25 11.1 ? 2 0.0 68.1 ? 6 724.0 ? 6 51.9 ? 2.7 449.1 ? 5.4 51.1 ? 3.4 11-2-S-1 89-26 11.0 ? 1.6 0.0 52.4 ? 2.2 812.8 ? 6.1 49.6 ? 2.6 397.6 ? 5 49.4 ? 3.2 11-2-S-2 89-31 7.7 ? 1.7 0.0 49.0 ? 2.2 782.3 ? 6 41.1 ? 2.6 369.4 ? 4.9 43.2 ? 3.2 11-2-S-3 89-30 11.1 ? 1.7 0.0 50.5 ? 2.3 816.8 ? 6.2 40.3 ? 2.6 422.8 ? 5.2 46.9 ? 3.3 11-2-S-8 89-28 0.0 0.0 30.2 ? 2.1 578.0 ? 5.3 39.2 ? 2.6 339.1 ? 4.8 36.7 ? 3.3 11-2-9 90-30 10.4 ? 2.1 0.0 57.8 ? 2.4 815.7 ? 6.4 51.1 ? 2.8 427.4 ? 5.4 49.9 ? 3.5 11-52-S-9 90-33 8.9 ? 1.8 0.0 59.3 ? 2.5 750.7 ? 6.2 48.0 ? 2.8 457.0 ? 5.5 53.1 ? 3.5 13-S-3 89-18 6.5 ? 1.6 0.0 48.7 ? 2.3 823.3 ? 6.5 43.7 ? 2.7 406.9 ? 5.3 44.3 ? 3.4 13-S-5 89-34 11.7 ? 1.8 0.0 49.0 ? 2.2 824.4 ? 6.1 50.4 ? 2.6 438.0 ? 5.1 50.9 ? 3.3 13-S-6 89-33 6.6 ? 2 0.0 43.0 ? 2.3 772.1 ? 6.2 44.0 ? 2.7 413.1 ? 5.3 43.9 ? 3.4 13-S-9 89-32 7.6 ? 2 0.0 46.3 ? 2.4 691.2 ? 6 34.7 ? 2.7 322.5 ? 5 37.4 ? 3.4 13-1-S-36 89-19 5.3 ? 1.9 0.0 38.9 ? 2.4 635.8 ? 6.2 39.1 ? 2.9 365.1 ? 5.6 53.0 ? 3.8 13-1-3-1-1 89-17 9.2 ? 1.6 0.0 58.9 ? 2.4 854.1 ? 6.5 50.9 ? 2.7 464.6 ? 5.5 54.3 ? 3.4 13-1-9-2-4 90-22 0.0 0.0 48.9 ? 2.5 894.5 ? 7.2 49.0 ? 3 527.1 ? 8.2 64.5 ? 3.8 13-1-13-87 90-23 12.0 ? 2.1 0.0 26.4 ? 2.2 660.0 ? 5.8 31.6 ? 2.6 292.1 ? 4.8 27.8 ? 3.3 13-1-16-1-84 90-42 8.3 ? 1.8 0.0 31.5 ? 2.3 545.5 ? 5.6 33.6 ? 2.8 329.7 ? 5.1 34.1 ? 3.6 13-1-16-11 90-46 9.6 ? 2 0.0 32.5 ? 2.2 638.5 ? 5.8 32.4 ? 2.6 293.2 ? 4.8 49.7 ? 3.5 13-1-16-11-81 90-45 9.6 ? 2.2 0.0 40.8 ? 2.6 856.3 ? 7.5 35.3 ? 3.1 333.8 ? 5.7 53.0 ? 4 13-1-16-12-83 90-38 9.7 ? 1.8 0.0 37.3 ? 2.3 967.8 ? 7.3 34.7 ? 2.6 331.0 ? 5.2 51.1 ? 3.5 13-1-20-4-118 90-27 6.9 ? 1.7 0.0 48.5 ? 2.3 772.8 ? 6.2 40.8 ? 2.7 386.0 ? 5.1 44.4 ? 3.4 13-1-20-5-121 90-37 10.9 ? 2.1 0.0 33.5 ? 2.3 691.7 ? 6 36.8 ? 2.6 317.7 ? 4.9 48.3 ? 3.5 13-1-20-5-122 90-41 0.0 0.0 38.7 ? 2.4 733.0 ? 6.7 29.3 ? 2.8 321.3 ? 5.4 58.5 ? 3.8 13-1-20-5-123 90-34 9.1 ? 1.7 0.0 37.2 ? 2.2 700.9 ? 5.9 33.9 ? 2.6 324.5 ? 4.9 56.6 ? 3.5 13-1-20-6-139 90-48 8.0 ? 1.8 0.0 40.1 ? 2.4 739.8 ? 6.5 33.2 ? 2.7 335.7 ? 5.2 52.1 ? 3.7 13-1-20-6-140 90-29 7.4 ? 1.6 0.0 46.4 ? 2.1 724.4 ? 5.5 35.7 ? 2.4 368.9 ? 4.6 36.0 ? 3 13-1-20-6-141 90-36 11.8 ? 2 0.0 31.1 ? 2.3 758.6 ? 6.6 37.0 ? 2.8 342.1 ? 5.3 61.7 ? 3.8 13-1-20-7-143 90-40 9.1 ? 1.7 0.0 35.6 ? 2.2 694.4 ? 6.2 27.9 ? 2.7 314.2 ? 5.1 52.0 ? 3.6 13-1-20-9-147 90-43 11.4 ? 1.9 0.0 69.0 ? 2.8 969.8 ? 7.9 44.7 ? 3.1 401.7 ? 6 63.4 +4 13-1-21-7-151 90-49 8.7 ? 2 0.0 49.8 ? 2.5 524.1 ? 5.5 24.4 ? 2.8 264.6 ? 4.9 41.4 ? 3.7 13-1-22-2-171 90-44 8.8 ? 1.7 0.0 52.3 ? 2.4 786.8 ? 6.3 37.5 ? 2.7 385.8 ? 5.2 49.6 ? 3.5 13-1-22-2-173 90-25 8.7 ? 1.9 0.0 46.1 ? 2.7 722.4 ? 7 45.2 ? 3.1 335.4 ? 5.7 40.8 ? 3.9 13-1-22-2-174 90-24 12.8 ? 2.4 0.0 41.5 ? 2.7 712.3 ? 6.8 49.6 ? 3.1 361.9 ? 5.8 43.6 ? 3.9 13-1-23-7-194 90-26 7.6 ? 1.9 0.0 48.2 ? 2.2 689.7 ? 5.6 47.1 ? 2.6 415.3 ? 5 45.4 ? 3.3 13-1-27-5-39 90-31 8.6 ? 1.9 0.0 49.6 ? 2.3 793.6 ? 6.2 50.2 ? 2.7 455.4 ? 5.4 53.7 ? 3.4 13-1-27-5-40 90-39 7.6 ? 1.6 0.0 50.5 ? 2.2 729.7 ? 5.7 50.0 ? 2.6 440.0 ? 5.1 47.3 ? 3.2 13-1-27-5-41 90-47 7.1 ? 1.9 0.0 38.1 ? 2.2 676.3 ? 5.7 30.8 ? 2.5 298.1 ? 4.6 48.9 + 3.3 13-1-27-6-42 90-35 9.1 ? 1.8 0.0 43.4 ? 2.4 703.6 ? 6.4 40.7 ? 2.8 392.7 ? 5.6 43.0 ? 3.6 Analyst: Marshall Weisler, September 1989 to February 1990. Chemical Characterization of Manula Adz Material 185 SUMMARY AND CONCLUSIONS Defining the spatial and temporal dimensions of intra- and inter-island communication is an impor- tant precursor for evaluating historical developments on insular landscapes. Because Polynesia generally lacks the "footprints of pottery" throughout se- quences of most island groups, and large quantities of obsidian are absent in this geologic province, fine- grained basalt manufactured into adzes and widely distributed must of necessity fonn an important basis for tracking regional interaction. Knowledge of the "provenance environment" or geologic features and their chemical signatures are essential aspects for identifying and predicting the occurrence of fine- grained, stone-tool-quality basalt. Two material characterization and provenance techniques have been described and evaluated, and geochemical analysis is argued to be the most profitable for long- term and regional-scale distributional studies as: (1) the results are reproducible; (2) instrument operating conditions can be reported in full facilitating com- parison of regional databases; (3) identification of elements is not subject to human error as with thin- section descriptions; (4) elemental abundances can be reported with precision and accuracy values for specimens and standards; and (5) geochemical sampling locales on specimens more closely repre- sent the population rather than petrographic thin- sections which are limited by two-dimensional surfaces. The efficacy of non-destructive x-ray fluorescence spectroscopy has been demonstrated and its continued use for distributional studies of Polynesian adz material is warranted. Although the technique, as applied here, is limited to elements that can be analyzed with high precision and accuracy (those in the mid-z range), future applications should seek to expand these limits. A major benefit of EDXRF is the wide range of specimen sizes that can be analyzed without destruction. In this study, artifacts ranged from as small as 2.1 grams up to 421.9 grams with lengths of 22 to 136 mm. Accom- modating a wide range of specimen sizes can be advantageous when analyzing assemblages with a high proportion of small adz flakes, a common situation with most collections of adz material. In fact, with nearly 1000 m2 of archaeological excava- tions on the island of Moloka'i, small adz flakes outnumber whole or stylistically diagnostic adzes by more than 10 to 1. Moreover, most adzes are surface finds whereas adz flakes are more often from excavated contexts. Application of the non-destructive, energy- dispersive technique to source matenal and artifact assemblages from several sites on Ta'u and Ofu islands has permitted delineation of an interaction sphere during late prehistory that links habitation complexes on these islands to the large adz quarry complex at Tatagamatau, Tutuila Island, some 100 kn to the west. Local dike rock from Ofu Island is chemically similar to most of the non-polished and unground flakes. Geochemical analysis has shown that these medium- to coarse-grained rocks were not fashioned into adzes, but this material was restricted to a few retouched, "awl-like" tools and one flake with use-wear along a distal margin. The vast majority, however, were unmodified. EDXRF is not the answer to all provenance studies, but the results reported here suggest that the technique merits further use and refinement with adz material from Polynesia and possibly throughout Oceania (e.g., Weisler 1993; Weisler et al. In prep.). As the ownership and preservation of archaeological sites and museum collections become increasingly the focus of controversy between the scientific and native communities, discovering more about the past without destroying the evidence itself, may help to bridge our common goals. ACKNOWLEDGMENTS I thank David Clague (Director, Hawaii Volca- noes Observatory), Joachim Hampel (Department of Geology, University of California, Berkeley), M. S. Shackley (Lowie Museum, University of Califomia, Berkeley), Pat Kirch, and Roger Green for com- ments on an earlier version of this paper. I am grateful to Joachim Hampel who was especially helpful during my XRF analyses, and I thank Pat Kirch and Terry Hunt for inviting me to participate in the Manu'a archaeological project. REFERENCES CITED Best, S. 1984. Lakeba: The prehistory of a Fijian island. Unpublished Ph.D. dissertation, Depart- ment of Anthropology, University of Auckland. Best, S., H. Leach, and D. Witter. 1989. Report on the second phase of fieldwork at the Tataga- matau site, American Samoa, July-August 186 The To'aga Site 1988. Report on file, Historic Preservation Office, American Samoa. Best, S., P. Sheppard, R. Green, and R. Parker 1992. Necromancing the stone: Archaeologists and adzes in Samoa. Journal of the Polynesian Society 101:45-85. Bice, D. 1980. XRF data storage/matrix correction programs. Ms. on file, Department of Geology and Geophysics, University of California, Berkeley. Bouey, P. 1991. Recognizing the limits of archaeo- logical applications of non-destructive energy- dispersive x-ray fluorescence analysis of obsidians. Materials Research Society Sympo- siwn Proceedings 185, Materials Issues in Art and Archaeology II. pp. 309-320. Pittsburgh. Cox, K., J. Bell, and R. Pankhurst 1979. The Inter- pretation oaf Igneous Rocks. London: George Allen and Unwin. Davidson, J. 1977. Westem Polynesia and Fiji: Prehistoric contact, diffusion and differentiation in adjacent archipelagos. World Archaeology 9:82-94. . 1978. Western Polynesia and Fiji: The archaeo- logical evidence. Mankind 11:383-90. . 1979. Samoa and Tonga. IN J. Jennings, ed., The Prehistory of Polynesia, pp. 82-109. Cambridge: Harvard University Press. Dunnell, R 1971. Systematics in Prehistory. The Free Press, New York. Dye, T., M. Weisler, and M. Riford 1985. Adz quarries on Moloka'i and Oahu, Hawaiian Islands. Report on file, Historic Sites Section, Department of Land and Natural Resources, Honolulu, Hawaii. Emory, K. 1933. Stone Remains in the Society Islands. Bernice P. Bishop Museum Bulletin 116. Honolulu. Flanagan, F. 1969. U. S. Geological Survey stan- dards-U. First compilation of data for the new U.S.G.S. rocks. Geochi&ica et Cosmochinica Acta 33:81-120. Gemianique, C., and B. Briand 1985. XRF determi- nation of Zr, Nb, Y, Sr, Rb, Zn, Pb, in fifteen international geochemical reference samples. Geostandards Newsletter 9:31-34. Govindaraju, K. 1984. 1984 compilation of working values and sample description for 170 interna- tional reference samples of mainly silicate rocks and minerals. Geostandards Newsletter Special Issue 8. . 1989. 1989 compilation of working values and sample description for 272 geostandards. Geostandards Newsletter Special Issue 13:1- 113. Green, R. 1974. Review of portable artifacts from Western Samoa. IN R Green and J. Davidson, eds., Archaeology in Western Samoa Vol. II, pp. 245-75. Bulletin of the Auckland Institute and Museum 7. Green, R., and J. Davidson, eds. 1969. Archaeology in Western Samoa. Bulletin of the Auckland Institute and Museum 6. Auckland, New Zealand. Hooper, P., and D. Johnson 1987. Major and trace element analyses of rocks and minerals by automatic x-ray spectrometry. Ms. on file, Department of Geology, Washington State University, Pullman. Hughes, R. 1986. Diachronic variability in obsidian procurement patterns in northeastern California and southcentral Oregon. University of Califor- nia Publications in Anthropology 17. Berkeley: University of California Press. . 1988a Archaeological significance of geo- chemical contrasts among southwestern New Mexico obsidians. The Texas Journal of Science 40:297-307. -. 1988b. The Coso volcanic field reexamined: Implications for obsidian sourcing and hydra- tion dating research. Geoarchaeology 3:253-65. Hunt, T., and P. Kirch 1988. An archaeological survey of the Manu'a Islands, American Samoa. Journal of the Polynesian Society 97:153-83. Irwin, G. 1990. Human colonisation and change in the remote Pacific. Current Anthropology 31:90-94. Jack, R., and I. Carmichael 1969. The chemical 'fingerprinting' of acid volcanic rocks. Special Report 100:17-32. California Division of Mines and Geology, Sacramento. Kirch, P. 1985. Feathered Gods and Fishhooks: An Introduction to Hawaiian Archaeology and Prehistory. Honolulu: University of Hawaii Press. . 1990. Regional variation and local style: A neglected dimension in Hawaiian prehistory. Chemical Characterization of Manu'a Adz Material 187 Pacific Studies 13:41-54. Kirch, P. V., and R. Green 1987. History, phylog- eny, and evolution in Polynesia. Current Anthropology 28:431-56. Kirch, P. V., T. Hunt, L. Nagaoka, and J. Tyler. 1990. An ancestral Polynesian occupation site at To'aga, Ofu Island, American Samoa. Archaeology in Oceania 25:1-15. Leach, H., and D. Witter 1987. Tataga-matau "rediscovered." New Zealand Journal of Archaeology 9:33-54. . 1990. Further investigations at the Tataga-matau site, American Samoa. New Zealand Journal Qf Archaeology 12:51-83. Le Maitre, R. ed. 1989. A Classification of Igneous Rocks and Glossary of Terns, Recommenda- tions of the International Union of Geological Sciences Subcommission on the Systematic of Igneous Rocks. Oxford: Blackwell Scientific Publications. Parkes, P. 1986. Current Scientific Techniques in Archaeology. London: Croom Helm. Parks, G., and T. Tieh 1966. Identifying the geo- graphical source of artifact obsidian. Nature 211:289-90. PNckett, N., ed. 1982. The First Thousand Years: Regional Perspectives in New ZealandArchae- ology. New Zealand Archaeological Associa- tion, Palmerston North, New Zealand. Rolett, B. 1989. Hanamiai changing subsistence and ecology in the prehistory of Tahuata (Marquesas Islands, French Polynesia). Unpublished Ph.D. dissertation, Yale Univer- sity, New Haven. Shackley, M. 1988. Sources of archaeological obsidian in the Southwest: An archaeological, petrological, and geochemical study. American Antiquity 53:752-72. Sheets, P., K. Hirth, F. Lange, F. Stross, F. Asaro, and H. Michel 1990. Obsidian sources and elemental analyses of artifacts in southem Mesoamerica and the Northem Intermediate Area. American Antiquity 55:144-58. Stice, G., and F. McCoy 1968. The geology of the Manu'a Islands, Samoa. Pacific Science XXII:427-57. Stross, F., J. Weaver, G. Wyld, R. Heizer, and J. Graham 1968. Analysis of American obsidians by x-ray fluorescence and neutron activation analysis. Contributions of the University of California Archaeological Research Facility 5:59-79. Berkeley. Terrell, J. 1986. Prehistory in the Pacific Islands. Cambridge: Cambridge University Press. Walter, R. 1990. The southern Cook Islands in eastern Polynesian prehistory. Unpublished Ph.D. dissertation, University of Auckland, New Zealand. Walter, R., and W. Dickinson 1989. A ceramic sherd from Ma'uke in the southern Cook Islands. Journal of the Polynesian Society 98:465-70. Weaver, J., and F. Stross 1965. Analysis by x-ray fluorescence of some American obsidians. Contributions of the University of California Archaeological Research Facility 1:89-93. Berkeley. Weisler, M. 1990. Sources and sourcing of volcanic glass in Hawai'i: Implications for exchange studies. Archaeology in Oceania 25:16-23. .1993. Long-distance interaction in prehistoric Polynesia: Three case sudies. Unpublished Ph.D. dissertationUniversity of California, Berkeley. In press a. Provenance studies of Polynesian adze material: A review and suggestions for improving regional data bases. Asian Perspec- tives 32(1). In press b. The settlement of marginal Polynesia: New evidence from Henderson Island. Journal of Field Archaeology 24(1). Weisler, M., P. Kirch, and J. Endicott In press. The Mata'are basalt source: Implications for prehistoric interaction studies in the Cook Islands. Journal of the Polynesian Society. Weisler, M., J. Sinon, and Y. Sinoto In prep. ?fhe Polynesian basalt artifact geochemical da .