Download - Source analysis of Central Plains tradition pottery using neutron activation analysis

325

Donna C. Roper et al. Source Analysis of Pottery Using Neutron Activation Analysis

Plains Anthropologist, Vol. 52, No. 203, pp. 325-335, 2007

Source Analysis of Central Plains Tradition PotteryUsing Neutron Activation Analysis:

Feasibility and First Results

Donna C. Roper, Robert J. Hoard, Robert J. Speakman, Michael D. Glascock,and Anne Cobry DiCosola

We report the results of a study applying instrumental neutron activation analysis (NAA) topottery from eight sites assigned to the western part of the Central Plains tradition (Upper Repub-lican and Smoky Hill phases) and six components identified as High Plains Upper Republican. Ourpurpose is to test the feasibility of using NAA to trace interactions among people of the CentralPlains tradition and between the Central Plains tradition people and their counterparts on theHigh Plains. Results of the statistical analysis, which was performed using the chemical data forboth newly sampled sites and previously studied sites (as reported by Cobry), suggest that NAA is ausable method for evaluating the movement of pottery for at least parts of the Central Plains tradi-tion. Samples from four sites on Medicine Creek as well as the Albert Bell and LeBeau sites formeda single homogeneous group referred to as the Central Plains Reference Group. Samples from theother sites, however, formed distinct groups for each site and also reflected some interaction withtheir contemporaries. We discuss the implications of these results, some of the questions that re-main, and the need for continued sampling.

Keywords: Central Plains tradition, ceramics, neutron activation analysis, interaction

Donna C. Roper, Department of Sociology, Anthropology and Social Work, 204 Waters Hall, Kansas State University,Manhattan, Kansas 66506, [email protected], 785.776.3772

Robert J. Hoard, Kansas State Historical Society, 6425 SW 6th Avenue, Topeka, Kansas 66615,[email protected], 785.272.8681 x269

Robert J. Speakman, Museum Conservation Institute, Smithsonian Institution, Museum Support Center,4210 Silver Hill Road, Suitland, Maryland, [email protected], 301.238.1242

Michael D. Glascock, Research Reactor Center, University of Missouri, Columbia, Missouri 65211,[email protected], 573.882.5270

Anne Cobry DiCosola, Department of Anthropology, 3525 Faner Hall, Southern Illinois University, Carbondale,Illinois 62901-4502, [email protected], 618.453.0273

PROBLEM DEFINITIONFor as long as archaeologists have recognized

the “cultures,” now usually designated phases, thatcollectively compose the Central Plains tradition,these phases often have been treated as if they rep-resented distinct, discrete, bounded, small village-dwelling societies analogous to the tribes, or atleast to self-identified bands within the tribes, ofthe historic period. Our on-going work with the

western part of the Central Plains tradition—theSmoky Hill and Upper Republican phases—ispainting a different picture. As a basic point, wefind that we must renounce the notion of smallvillage organization in the Central Plains traditionand replace it with one of economically autono-mous farmsteads spread along the major streamvalleys. This is important because even if someadjacent farmsteads were contemporaneous, we

326

PLAINS ANTHROPOLOGIST VOL. 52, NO. 203, 2007

have no evidence for a village level ofsociopolitical organization on the central Plainsat this time. As we begin to build a new concept ofCentral Plains tradition social organization, thescale, degree of localization, and nature of inter-actions among members of the communities (andit is the members of the communities, not thecommunities per se, that interact) are topics ofconsiderable interest.

The area of the western Central Plains tradi-tion, i.e., the portion of northern Kansas and south-ern Nebraska from about the Blue River valley innortheast Kansas west to the Medicine Creek lo-cality in southwest Nebraska, is an area that ex-hibits a strong ecological gradient. The notablevariation in precipitation, vegetation, and faunalresources within this region is cross-cut by thevalleys of trunk streams—the Smoky Hill, Saline,Solomon, and Republican rivers—that rise on theHigh Plains or Colorado Piedmont and flow inrather straight and parallel courses to the east. Eachriver valley provides a series of closely-spacedlinear environments composed of the stream it-self; the stream banks; timber belts that line thestreams, if discontinuously; the prairie-coveredstream terraces; and the uplands separating the riv-ers. Throughout the western Central Plains tradi-tion, the farmsteads were small, usually rectangu-lar, lodges spread along the front edges of the ter-races in the main river valleys or the valleys oftheir larger tributaries. The most productive zonesof the river valleys were the river meander belts,the stream banks, and the timber belts lining theriver edges. Each farmstead’s immediate catch-ment, therefore, contained most of the full diver-sity of resources on which the people of the Cen-tral Plains tradition depended for food and rawmaterials. Only two classes of resources wouldhave required a walk of more than a kilometer orso to procure. One of these was the larger fauna,such as bison—an animal not always extensivelyused by western Central Plains tradition people.The other was, in most places, chert and other rockand mineral raw materials, potentially includingsuitable pottery clay.

Despite the lack of actual villages, we haveno doubt that farmsteads were linked by notionsof community. In the face of the strong ecologi-cal gradient, however, the degree of localization

of these communities appears variable. Precipi-tation in the lower valleys of the main streams ofthe central Plains, of course, varied from year toyear, but it is likely that it did not often drop tosuch low levels that crops failed or resulted inseverely reduced yields, and, if it did, it is not likelythat it did so over runs of years. In this area, whereresource yields were reasonably stable, commu-nities may have been strongly localized, with farm-steads within a given valley comprising dispersedcommunities that, among other functions, main-tained and reinforced exclusive claims to and ac-cess to its resources. Yet in the absence of actualvillage organization, farmsteads within valleysmay have interacted primarily with their neighborsfor several kilometers in each direction, result-ing in “fuzzy” and probably continually shiftingcommunity “boundaries.”

Farther west, the situation may have beensomewhat different. Here, resource yields prob-ably were not so stable and, moreover, runs of badyears probably did occur, potentially making con-tinuous occupation of farmsteads precarious. Infact, as is well-known, elements of the distinctivematerial culture complex of this westernmostCentral Plains tradition phase, what is often calledthe Upper Republican phase, are found on the HighPlains in situations without lodges and withoutevidence of agriculture. Recent assessments ofthe relations among the westernmost lodge sites,such as those in the Medicine Creek valley, andthe High Plains sites (Roper 2007; Scheiber 2006)suggest these are all the same people who, whenconditions were favorable, occupied lodges andtended crops (and of course consumed other lo-cal resources) but, when conditions were less fa-vorable, became mobile hunter-gatherers movingthrough a region where bison were abundant. Com-munities on this part of the central Plains werenot localized and interaction probably was morefluid and situational.

We have several means of evaluating the util-ity of this conception of Central Plains traditionorganization. One is detailed ceramic style analy-sis. This has been initiated and will be continuingover the coming years. A second approach is theuse of chemistry-based approaches, such as neu-tron activation analysis (NAA) to study the move-ment of pots and, by extension, people through

327


the region.NAA has been proven as an effective analyti-

cal tool to characterize the chemical compositionof prehistoric pottery and to relate sherds to theirsources through their chemical composition(Glascock 1992; Neff 1992). Given the chemicalcomplexity of clay and temper materials, sensi-tive and precise measures of compositional ele-ments are required to discriminate between pro-duction-source areas. The relation is, however,empirical and its sensitivity in any given regionmust be established through analysis of carefullyselected samples. In this study, therefore, we re-port the results of a pilot study that tests the fea-sibility of using NAA to trace the interactionsamong people of the Central Plains tradition andbetween these people and their counterparts onthe High Plains. Our basic strategy is to selectmaterials from sites in most of the major rivervalleys forming the upper Kansas River basin, andto sample the valleys at several points along theireast-west course. Our basic research question iswhether pots appear to have moved west-to-eastalong drainages, which is consistent with the or-ganizational model sketched above and with theresults to date of the style analysis. Alternatively,interactions could haveoccurred across drain-ages in more of a north-south direction, as im-plied by the traditionaltaxonomies assigningthe central Kansas sitesto the Smoky Hill phase(regardless of which val-ley they lie in) and thewestern Kansas/south-western Nebraska sitesassigned to the UpperRepublican phase (re-gardless of which valleythey lie in).

MATERIALS ANDMETHODS

This is not the firststudy of material in thebroad set of sites we areconsidering. A recent

study by Anne Cobry (1999; see also Cobry andRoper 2002) analyzed sherds from two UpperRepublican sites in the Medicine Creek locality—a lithic work area at 25FT30 and House 5 at25FT39—and from three sites on the High Plainsof northeast Colorado and southeast Wyoming—Donovan (5LO204), Seven Mile Point(48LA304), and Gurney Peak Butte (48LA305).Cobry also analyzed clay samples from four sepa-rate sampling points in the Medicine Creek local-ity. Three of these were from B horizons of soilsformed on the surface of the T-1. The fourth wasobtained from an outcrop of weathered Pierreshale just below Medicine Creek Dam. Other clay-source samples included in Cobry’s analysis weretaken near the Donovan site in northeast Colorado,from alluvial context near Muddy and Lodge Polecreeks, both near Seven Mile Point in southwestWyoming; from Pine Bluffs Volcanic Ash, alsonear Seven Mile Point; and from Gurney PeakButte, also in southwest Wyoming (Figure 1).

Cobry’s (1999; see also Cobry and Roper2002) results suggested that some of the HighPlains pots apparently were coming from Medi-cine Creek, but that some were made of clays fromdifferent sources that were not identified. Con-

Figure 1. Sites from which pottery samples were obtained and of the Central Plains ReferenceGroup (see text for details of how this group is formed).

328


versely, it also appeared that some of the pots fromone of the Medicine Creek sites, House 5 at25FT39, might have been produced from claysderived from sources other than the MedicineCreek valley, implying importation of pots and notclays, but it could not be determined where thosepots might have come from. These results wereand are credible, particularly given that the exca-vated material from the 25FT39 lodge includeslarge quantities of western lithic raw materials,especially Flattop Butte chalcedony.

The study reported here sought to expand thesampling and analysis of pottery to the east, over-lapping Cobry’s study area along the western edgeof our sampling area (Figure 1). Thus, we selectedsamples of 10 to 30 sherds from each of sevenlodge sites in four river valleys. We expanded thesampling of the apparently diverse Medicine Creekvalley by selecting sherds from two additionalsites, a midden associated with Houses 1 and 4 on25FT22 and what is thought to be a plowed-downhouse on 25FT167. We also included sherds fromthe LeBeau site (14NT301) located in the valleyof Prairie Dog Creek. Both Prairie Dog Creek andMedicine Creek are tributaries of the RepublicanRiver. Moving south, we selected material fromtwo sites in the Solomon River drainage. Thewesternmost of these two is Albert Bell(14SD305), an Upper Republican site in the val-ley of Museum Creek, a tributary of the South Forkof the Solomon River. The Albert Bell site is situ-ated at the western edge of the Central Plains tra-dition and located due south of the Medicine Creekvalley site group. From farther downstream in theSolomon River valley, we selected samples fromMinneapolis (14OT5), an important Smoky Hillphase site. From a short distance south of theMinneapolis site—in the lower valley of theSmoky Hill River—we sampled the ceramics fromthe Kohr House #1 (14SA414).

Finally, we included pottery from the Coal-Oil Canyon site (14LO1/401). Coal-Oil Canyonis not a lodge/agriculture site, but it is not exactlylike others of the reported Upper Republican siteson the High Plains. The 1950s excavations at Coal-Oil Canyon (Bowman 1960) resulted in the re-covery of a large quantity of pottery that can beaccommodated by Central Plains tradition potterytypologies. However, our recent re-examination

of the numerous rim sherds indicates that the as-semblage is very diverse, with decorative simi-larities to vessels from Medicine Creek valley,Solomon River drainage, and Smoky Hill Rivervalley sites. Coal-Oil Canyon is beyond the west-ern limits of the horticultural farmstead sites (thatlimit is at about 100° W longitude) and is func-tionally distinct from them. It is near historic trailsand seemingly represents a place used repeatedlyfor short, perhaps transitory, occupation. Our pur-pose in including samples from this site was tohelp confirm whether this key western Kansas site,uniquely situated near trails, is indeed at a cross-roads, or whether it better fits with other SmokyHill River valley sites. To address this question,we sampled as best we could across the range ofvariation to see if that variation is the result ofpots moving to the site from different areas or ifthis locality was used largely by the people of onecommunity.

Roper (2006:111–112) summarizes the char-acteristics of Central Plains tradition pottery.Overall, this pottery usually is grit- or sand-tem-pered. Vessels are globular to sub-globular withdirect flaring or collared rims. Decoration, whenpresent, and except in rare instances, is confinedto the rim and/or lip. Exteriors are predominantlycord-roughened. The High Plains material frombeyond the area of the lodge sites is consistentwith this general description. Shell-tempered pot-tery occurs in sites of the eastern part of the Cen-tral Plains tradition and is represented in smallproportion as far west as the Minneapolis site.Shell-tempered pottery was not considered in thisstudy. All sherds used for this study were grit- orsand-tempered body sherds.

In addition to the pottery samples, fivesamples from clay sources in the Albert Bell sitevicinity were analyzed and compared against thesherds from that site to determine if they werelocally manufactured. Two of these samples camefrom a playa in native rangeland 14.5 km south-west of the site. One came from a pit that washand-excavated to obtain a sample from a Bkt ho-rizon on a small terrace just south of the site. An-other was taken from a small slack-water area inthe Museum Creek channel. The final sample isfrom slack-water in an unnamed, slow-movingtributary of the Solomon River 1.9 km northeast

329


of the site. It is believed that these clay sourceswould have been accessible to prehistoric pottersfrom Albert Bell.

SAMPLE PREPARATION ANDIRRADIATION

Sample preparation and analysis were con-ducted at the University of Missouri ResearchReactor Center (MURR). The pottery sampleswere prepared for INAA using standard MURRprocedures (Glascock 1992; Neff 1992, 2000).Fragments of about 1–2 cm were removed fromeach sample and the surfaces were abraded usinga silicon carbide burr to remove any materials ad-hering to the surface, thereby reducing the risk ofmeasuring contamination. The samples werewashed in deionized water and allowed to dry inthe laboratory. Once dry, the individual sherds wereground to powder in an agate mortar to homog-enize the samples. Archival samples were retainedfrom each sherd (when possible) for future re-search. Clay samples were fired in a furnace to700 degrees Celsius for one hour.

Portions of approximately 150 mg of pow-der were weighed into small polyvials used forshort irradiations at MURR. At the same time, 200mg of each sample were weighed into the high-purity quartz vials used for long irradiations. Alongwith the unknown samples, reference standards ofSRM-1633a (coal fly ash) and SRM-688 (basaltrock) were similarly prepared, as were qualitycontrol samples (e.g., standards treated as un-knowns) of SRM-278 (obsidian rock) and OhioRed Clay.

Neutron activation analysis of ceramics atMURR, which consists of two irradiations and atotal of three gamma counts, constitutes a supersetof the procedures used at most other NAA labora-tories (Glascock 1992; Neff 1992, 2000). As dis-cussed in detail by Glascock (1992), a short irra-diation entails irradiation of each sample for fiveseconds by a neutron flux of 8 x 1013 n cm-2 s-1.The 720-second count yields gamma spectra con-taining peaks for short-lived elements aluminum(Al), barium (Ba), calcium (Ca), dysprosium (Dy),potassium (K), manganese (Mn), sodium (Na), ti-tanium (Ti), and vanadium (V). The samples en-capsulated in quartz vials are subjected to a 24-hour irradiation at a neutron flux of 5 x 1013 n cm-

2 s-1. This long irradiation is analogous to the singleirradiation utilized at most other laboratories. Af-ter the long irradiation, samples decay for sevendays, and then are counted for 2,000 seconds (the“middle count”) on a high-resolution germaniumdetector coupled to an automatic sample changer.The middle count yields determinations of sevenmedium half-life elements, namely arsenic (As),lanthanum (La), lutetium (Lu), neodymium (Nd),samarium (Sm), uranium (U), and ytterbium (Yb).After an additional three- to four-week decay, afinal count of 9,000 seconds is carried out on eachsample. The latter measurement yields the follow-ing 17 long half-life elements: cerium (Ce), co-balt (Co), chromium (Cr), cesium (Cs), europium(Eu), iron (Fe), hafnium (Hf), nickel (Ni), rubidium(Rb), antimony (Sb), scandium (Sc), strontium(Sr), tantalum (Ta), terbium (Tb), thorium (Th), zinc(Zn), and zirconium (Zr).

STATISTICAL ANALYSIS OF THECHEMICAL DATA

The analyses produced elemental concentra-tion values for 32 or 33 elements in most of theanalyzed samples. Data for As, Ni, Sb, and Tb werenot considered during the statistical analysis ofdata given that these elements generally have pooranalytical precision. As is customary in ceramicprovenance studies at MURR, parts-per-milliondata were converted to base-10 logarithms. Useof log concentrations rather than raw data com-pensates for differences in magnitude between themajor elements, such as iron, and trace elements,such as the rare earth or lanthanide elements(REEs). Transformation to base-10 logarithmsalso yields a more “normal” distribution for manytrace elements.

Data were examined using principal compo-nents analysis (PCA) and through inspection ofbivariate plots (see Baxter 1992; Bieber et al.1976; Bishop and Neff 1989; Neff 1994, 2000,2002; and Neff et al. 2006 for detailed discus-sions of various statistical approaches commonlyused in NAA studies of pottery). Samples wereassigned to groups based on patterning observedin PCA and bivariate space. Mahalanobis distanceprobabilities were used to evaluate the statisticalvalidity of the proposed group structure. Use ofMahalanobis distances usually results in a subset

330


of specimens that cannot confidently be assignedto any group. As Neff et al. (2006) indicate, thesespecimens may be statistical outliers from one ofthe defined groups, may represent different pot-tery production practices or diagenic anomalies,or they may pertain to sources sampled so sparselythat they cannot be recognized as distinct groups.In the current study, specimens were left unas-signed if they were marginal to all groups or ifincluding a specimen in the group to which it ap-parently belonged obscured distinctions betweengroups that were otherwise well discriminated.Although unassigned specimens are problematic,the approach taken herein is similar to that takenby most NAA laboratories and serves to minimizeincorrect group assignments by leaving marginalspecimens unassigned (Neff et al. 2006). Ideally,additional analyses of pottery will permit classi-fication of the unassigned specimens.

RESULTSThe current study reports on the analyses of

the 87 artifact and 13 source clay samples newlysubmitted for this study and the data generated forthe 56 sherds and 11 source clay samples reportedin Cobry’s earlier study, thus producing a trulyregional feasibility study.

Analysis of potteryEight compositional groups were

identified in the current study. Cobry(1999; Cobry and Roper 2002) iden-tified two compositional groups,Medicine Creek 1 and 2, from potteryproduced at sites along MedicineCreek. The analysis of additionalMedicine Creek pottery samples aswell as analysis of samples from theAlbert Bell, LeBeau, and Coal-OilCanyon sites has necessitated revalu-ation of the distinctiveness of thesetwo groups. With the increasedsample, groups designated MedicineCreek 1 and Medicine Creek 2 are nolonger considered distinct. Instead,the two groups are combined with allpottery from the Albert Bell andLeBeau sites, and the vast majority ofpottery from the Medicine Creek

sites and renamed the “Central Plains ReferenceGroup,” which contains 70 pottery specimens (seeFigure 1 for the spatial extent of this referencegroup). Other compositional groups are formedfrom pottery samples from the Seven Mile Point,Kohr House #1, Minneapolis, Coal-Oil Canyon,and Donovan sites. Pottery from Donovan site lev-els 1 and 4 each formed distinct groups. Theseresults corroborate the intuitive notion that pot-tery was manufactured locally from nearby clays.Twenty-two pottery samples are unassigned.

The subgroup structure in the complete cen-tral and High Plains regional potsherd data set isvisually portrayed in three bivariate plots (Figures2–4; Speakman and Glascock 2004) and is sum-marized in Table 1. The table lists group member-ship based on Mahalanobis distance probabilities;these probabilities are based on the 29 elementsretained for analysis. The group membership as-signments listed in Table 1 confirm that the sub-group structure proposed herein is statisticallyviable, with the possible exception of the speci-mens from Coal-Oil Canyon that exceed one per-cent probability of membership in the CentralPlains Reference Group. These specimens sug-gest that the chemical composition of the two

Figure 2. Bivariate plot of sodium and vanadium base-10 logged concentrationsfor pottery samples. In this projection the Kohr and Donovan groups (level 1 andlevel 4) are shown to be separate from all other groups. Ellipses represent 90%confidence interval. Unassigned samples are not plotted.

331


groups overlaps to some extent.The group structure for Gurney

Peak Butte, Seven Mile Point, andDonovan has not changed significantlywith the analysis of additionalsamples. This is not unexpected giventhat all new samples were from eitherthe eastern edge of the area coveredby Cobry’s samples or from sites evenfarther to the east and probably did notaffect the grouping of those samples.A few samples were reassigned, butthese reassignments do not affect theinterpretations made by Cobry andRoper (2002). Of particular note hereis that one sample from Coal-Oil Can-yon (CPT051), which consistentlygroups with Donovan-1 pottery. How-ever, this assignment is tenuous atbest given that the small number ofsamples in the Donovan-1 group pro-hibits any rigorous statistical testingof the member in the group.

Results for the central Kansaslodge sites are of considerable inter-est. Seven of the nine samples ana-lyzed from the Minneapolis site forma distinct compositional group. Six ofthe 10 samples from Kohr house alsoformed a distinct compositionalgroup, one sample is unassigned, andthree have a high probability of mem-bership in the Central Plains Refer-ence Group. Additional analyses ofpottery from Minneapolis and adja-cent sites would help to refine thisgroup and would permit more rigor-ous statistical testing of the Minne-apolis and Kohr groups. Given that adistinct group comprised of Kohr pot-tery was identified, the assignment ofthe three Kohr House samples to theCentral Plains Reference Group po-tentially indicates a west-to-eastmovement of some pots to the lowerSmoky Hill valley. Roper in fact hasseen in collections from this area occasionalsherds whose ceramic fabric macroscopicallybears a strong resemblance to that from Medicine

Creek and other sites to the west. It is highly sig-nificant, however, that neither the Kohr site northe Minneapolis site appears to have pottery from

Figure 3. Bivariate plot of calcium and europium base-10 logged concentrationsfor pottery samples. In this projection Gurney Peak and Minneapolis Groups areshown to be separate from the CP Ref. and COC groups. Ellipses represent 90percent confidence interval. Unassigned samples are not plotted.

Figure 4. Bivariate plot of barium and cesium base-10 logged concentrations illus-trating the best possible separation of the CP Ref. and COC groups. Ellipsesrepresent 90 percent confidence interval. Unassigned samples are not plotted.

332


the other site. This is consistent with our predic-tion of more interaction along than between rivervalleys. Considerable additional work will have tobe done to strengthen this conclusion, includingshowing the contemporaneity of the two sites andincreasing the sample size of analyzed sherds, butthe results obtained here at least suggest that thesesites can be differentiated by neutron activationanalysis and that expanded sampling and analysisshould be productive.

Of the 26 pottery samples analyzed fromCoal-Oil Canyon, 15 are assigned to a new group:the Coal-Oil Canyon Group (COC). Of the other10 samples, one is assigned to the Central PlainsReference Group, one sample is assigned toDonovan-1, and nine are unassigned. Pottery fromthe Coal-Oil Canyon Group is so similar to pot-tery assigned to the Central Plains ReferenceGroup that barium appears to be the only elementcapable of discriminating between these twogroups, and even then the separation is not unam-biguous. A discriminant analysis failed to show agood separation for these groups and severalspecimens assigned to the Coal Oil-Canyon groupexceed one percent probability of membership inthe Central Plains Reference Group. It is possiblethat analysis of a larger sample of pottery fromCoal-Oil Canyon would help to refine the COCgroup so that a statistically viable group (or groups)

could be defined. Given the high number of unas-signed specimens, increased sampling might alsoaid in identification of additional COC subgroups.

Analysis of clay samplesFor the earlier study by Cobry and Roper, in-

vestigators sampled clay sources from a T-1source in the Medicine Creek locality. This claysource should have been accessible to prehistoricpotters. Roper and some of her students previouslyhad made and fired pots made from these claysand had added temper similar to that seen in pre-historic pots from the area. The vessel fabric mac-roscopically appeared very like Upper Republi-can pottery, both on the surface and on freshbreaks. Samples of weathered shale in the regionwere tested in similar fashion with poor results. Itwas difficult to produce vessels with thicknesssimilar to that of Upper Republican pottery, andthe clay body cracked extensively. Moreover, thevessel fabric did not look like that of Upper Re-publican pottery. Clay samples from near theAlbert Bell site collected for the current studywere selected based on presumed availability toprehistoric potters (terrace cut, stream bed, andplaya) and their relatively high clay content.

Based on the available data, it appears that clayresources in this region exhibit similar ranges ofvariation across broad geographic areas. The

Table 1. Samples from Sites and Compositional Group Membership as Determined by Chemical Com-position Acquired by Neutron Activation Analysis. Group Membership Is Based on Mahalanobis Dis-tance Probabilities

Sites and Samples Compositional GroupsSites N CPR1 SMP2 GPB3 D14 D45 COC6 M7 K8 Unassigned25FT22 10 1025FT30 10 9 125FT39 10 1025FT167 10 8 2LeBeau 12 12Albert Bell 10 10SMP2 13 7 3 1 1 1GPB3 10 6 4D14 5 5D45 5 1 4DL99 3 3COC6 26 1 1 15 9M7 9 1 7 1K8 10 3 6 11 Central Plains Reference Group, 2 Seven Mile Point, 3 Gurney Peak Butte, 4 Donovan 1, 5 Donovan 4, 6 Coal-Oil Canyon, 7Minneapolis, 8 Kohr House #1, 9 Donovan Level 9.

333


chemical similarity and the small number of pot-tery samples analyzed from the Albert Bell (n =10) and LeBeau (n = 12) sites preclude the iden-tification of distinct reference groups for theMedicine Creek, LeBeau, and Albert Bell sites andthe evaluation of the movement or lack of move-ment of pots among these valleys.

In nearly all cases, our clay samples do notgroup with the pottery samples, and thus are notincluded in the principal components analysis ofpottery samples. The mismatch between potteryand clay samples suggests that we have not sampledthe sources that were used by prehistoric potters.Of the analyzed clay samples, the closest matchesto the Central Plains Reference Group are thosethat were collected from a playa near the AlbertBell site. In general, most pottery from the AlbertBell and LeBeau sites is lower in aluminum thanpottery from the Medicine Creek sites. Addition-ally, cesium, lutetium, and zirconium are gener-ally higher in pottery from the LeBeau site sug-gesting a possible separation of this pottery fromthat represented at Albert Bell. Most of the pot-tery in this dataset is tempered with a high silicasand. When added to clays, the quartz in the sandseffectively serves to dilute most other elements.The inclusion of possible sand tempers in the nextstage of this study will permit us to “mathemati-cally temper” the clay data so that clays can bemore effectively compared to the pottery refer-ence groups. Because of this, it seems likely thatadditional sampling of Albert Bell site pottery willshow that the clay from the vicinity of the AlbertBell site best matches the pottery from this site.Also, we are optimistic that additional samplingof Albert Bell and LeBeau site pottery will lead tothe identification of distinct compositional groups.

CONCLUSIONSThe results are notable on two fronts, one

methodological and one substantive. Method-ologically, the study shows that NAA is a viablemethod for chemically discriminating CentralPlains tradition pottery and evaluating its move-ment through at least some parts of the CentralPlains tradition. The chemical similarities betweenthe Central Plains Reference Group and the Coal-Oil Canyon group present some difficulties. Thisdivision possibly could disappear with additional

sampling. Should this happen, it would be evenmore difficult than it currently is to evaluatemovement between sites on the western edge ofthe Central Plains tradition and the High Plainssites. The chemical similarities between these twogroups also make it difficult to determine how theceramic assemblage at Coal-Oil Canyon came tobe as it is. On the other hand, the lack of associa-tion of ten Coal-Oil Canyon sherds with the Coal-Oil Canyon Group may be an indication that theassemblage is as eclectic as it appears on stylisticgrounds. More concretely at the present time, thechemical distinctness of the sites downstream onthe Solomon and Smoky Hill rivers suggests thatwe can evaluate the directions of the interactionsof people centered at these two sites. Samplingof more sites in this downstream area should tellus much about interactions among people in thisarea.

This project has been gratifying in that sev-eral new compositional groups have been identi-fied and we now have a better understanding ofthe range of chemical variability in this area ofthe central Plains. Unfortunately, LeBeau, AlbertBell, and Medicine Creek sites form a single ho-mogeneous group indicating that ceramic sourc-ing between sites within this area is difficult atthis time, but perhaps not impossible. Petrogra-phy may provide a viable and comparable methodfor investigating provenance in this area. Analysisof additional samples from Coal-Oil Canyon andsites in western Kansas may help to resolve lin-gering questions about pottery from this area. Fur-ther clay source sampling and the analysis of tem-pering material may also help document theamount of variability expected among the locali-ties and may even indicate whether we should ex-pect to be able to differentiate geographicallyseparate sites within this large group.

Substantively, though, some movement ofpottery is suggested by the data and some indica-tions of the directionality of that movement areshown. The seven sherds of Central Plains Refer-ence Group pottery at the Seven Mile Point site,the three at 14SA414 (Kohr House #1), and theone at Coal-Oil Canyon suggest general east-to-west or west-to-east movement of pottery fromthese clays. While seven sherds from the centralPlains reference area are found on the High Plains

334


(Seven Mile Point), only one sherd from theDonovan 1 compositional groups was found fur-ther east, at Coal-Oil Canyon. This is noted byCobry and Roper (2002), though with the data theyhad at hand they stated that Medicine Creek pot-tery was found on High Plains sites. The currentanalysis shows the clays and pottery found at Medi-cine Creek are compositionally similar over anarea that extends minimally south to the SolomonRiver and, should the Coal-Oil Canyon ReferenceGroup prove to be a part of this broader group,possibly to the Smoky Hill River. Further, we havealready noted that the downstream sites (e.g., Min-neapolis and Kohr) show west-to-east, but notnorth-south, movement of some pottery.

Finally, clay samples from the general vicin-ity of the Albert Bell site, taken from streambedsand from clay horizons buried in a terrace, showedno relation to the Central Plains Reference Groupusing Mahalanobis distance calculation with theexception of two samples taken from the B hori-zon of an upland playa. This is based on a very smallsample but suggests that people on this part of thecentral Plains were using clays developed in playasoils rather than streambed or terrace deposits.Clay samples from the vicinity of Coal-Oil Can-yon showed a low affinity toward the CentralPlains Reference Group overall, but a sample fromthe B horizon of the upland Richfield silt loam,mixed with about 10 percent local sand showed a2.09 percent probability of membership in thegroup.

Selecting clay samples for this type of studyis speculative. Unlike stone sources, for which weoften have evidence of prehistoric use of a sourcebased on waste flakes, it is difficult to determinewhat sources were used to make pots and investi-gators must use intuitive notions of what claysmight be best in a given region. Because there maybe multiple sources of clay with differing parentmaterials in any given region, it may take manysamples to come up with a match to the prehis-toric pottery under study. Finally, unlike stonesources, clay sources may be severely affectedby the passage of time. Streambed clays, for ex-ample, surely have been affected by intensive,machine-assisted agriculture of the twentieth cen-tury and subsequent erosion. Because of this, claysampling is, in many cases, an attempt to discover

in a very general sense what types of sources (al-luvial, weathered bedrock, or in the case of pla-yas, aeolian) might have been used rather than try-ing to pinpoint specific sources. This is the ap-proach we used here and it is apparent that we havenot yet located the right type of source. Our re-sults are interesting, though, in that the playa claysamples proved to have the closest relationshipto the region’s prehistoric pottery. Perhaps theaeolian nature of these clay deposits explains whythe pottery from the broad Central Plains Refer-ence Group area is chemically similar. Playas arecommon in the south-central United States in themixed-grass and short-grass prairies that coverportions of western Texas, Oklahoma, Kansas, andNebraska and the eastern portions of New Mexicoand Colorado, and southeastern Wyoming (Smith2003:6), and wetlands with similar soils are widelydistributed.

We have no doubt that significant questionsremain, that alternative scenarios must be evalu-ated by additional research, and that the samplingmust be expanded. This study, however, shows thatNAA is a feasible approach to the research prob-lem. A priority will be determining if the apparenthomogeneity of the Central Plains ReferenceGroup is a result of the chemical similarity ofavailable source clays or sufficient movement ofpeople that the group actually is a composite ofmany geographically separated sources. It shouldalso be determined if the Coal-Oil Canyon groupsimilarly is a composite, for, on other grounds,we are not at all sure that the pottery found at thissite actually was made there. Roper (2007) sug-gests that the people in the Medicine Creek-LeBeau-Albert Bell site area may be far moremobile than the presence of farmsteads and lodgesmight lead one to believe. We must therefore con-sider the possibility that people were movingaround considerably and possibly exploiting manyclay sources across the landscape, thus creating alarge group that is impossible to differentiate. Thedistinctiveness of the groups from sites down-stream (Kohr and Minneapolis) is consistent withindications, again on other grounds, that thesepeople were considerably less mobile than theircontemporaries to the west. The lack of interac-tion between these two sites is provocative, butmust be supported by increased sampling from

335


both sites, sampling of additional sites in the vi-cinity of these two sites, and a demonstration thatthe two communities were contemporaneous.Sampling sites in other localities along the lowervalleys of these rivers should also be indicative.

ACKNOWLEDGMENTSThis project was supported in part by the National Science

Foundation through a grant to the MURR Archaeometry Lab(grant no. SBR-0102325) and by the Bureau of Reclamationthrough a Cooperative Agreement (5-FC-60-06030) with KansasState University for which Roper is the Principal Investigator.We thank Bill Chada of the Bureau of Reclamation for allowinguse of cooperative agreement funds for this purpose and foragreeing to destructive testing of selected samples from theMedicine Creek sites. Jim McDowell and Bobby Tricks of theNRCS located and acquired Sheridan County playa clay samples.Mary Adair allowed use of archaeological samples from the14LO1/401(Coal-Oil Canyon) collection held in theArchaeological Research Center at the University of Kansas.Samples from the Albert Bell, LeBeau, Kohr, and Minneapolissites were drawn from collections held by the Kansas StateHistorical Society. The Donovan, Seven Mile Point, and GurneyPeak Butte samples included in Cobry’s original analysis wereprovided courtesy of Charles Reher and the High PlainsArchaeology Project at the University of Wyoming. DiCosolaalso thanks the Petsch family for access to their ranch to collectclay samples from the Gurney Peak Butte area.

REFERENCES CITEDBaxter, Michael J.

1992 Archaeological Uses of the Biplot—a NeglectedTechnique? In Computer Applications and QuantitativeMethods in Archaeology, 1991, edited by Gary Lock andJonathan Moffett, pp.141–148. BAR International SeriesS577, Tempvs Reparatvm, Archaeological and HistoricalAssociates, Oxford.

Bieber, Alan M. Jr., Dorthea W. Brooks, Garman Harbottle, andEdward V. Sayre

1976 Application of Multivariate Techniques to AnalyticalData on Aegean Ceramics. Archaeometry 18:59–74.

Bishop, Ronald L., and Hector Neff1989 Compositional Data Analysis in Archaeology. In

Archaeological Chemistry IV, edited by Ralph. O. Allen,pp. 576–586. Advances in Chemistry Series 220, AmericanChemical Society, Washington, D.C.

Bowman, Peter W.1960 Coal-Oil Canyon (14LO1): Report on Preliminary

Investigations. Bulletin 1. Kansas AnthropologicalAssociation, Topeka.

Cobry, Anne M.1999 Sourcing the Middle Ceramic Period: Investigations of

Upper Republican Ceramics on the Central and High Plains.Unpublished Master’s thesis, Department of Anthropol-ogy, University of Wyoming, Laramie.

Cobry, Anne M., and Donna C. Roper2002 From Loess Plains to High Plains: The Westward

Movement of Upper Republican Pots. In GeochemicalEvidence for Long-Distance Exchange, edited by Michael

D. Glascock, pp. 153–166. Greenwood Publishing Group,Westport, Connecticut.

Glascock, Michael D.1992 Characterization of Archaeological Ceramics at MURR

by Neutron Activation Analysis and Multivariate Statistics.In Chemical Characterization of Ceramic Pastes inArchaeology, edited by Hector Neff, pp. 11–26. PrehistoryPress, Madison, Wisconsin.

Neff, Hector1992 Introduction. In Chemical Characterization of Ceramic

Pastes in Archaeology, edited by Hector Neff, pp. 1–10.Prehistory Press, Madison, Wisconsin.

1994 RQ-mode Principal Components Analysis of CeramicCompositional Data. Archaeometry 36:115–130.

2000 Neutron Activation Analysis for ProvenanceDetermination in Archaeology. In Modern AnalyticalMethods in Art and Archaeology, edited by Enrico Cilibertoand Giuseppe Spoto, pp. 81–134. John Wiley and Sons,Inc., New York.

2002 Quantitative Techniques for Analyzing CeramicCompositional Data. In Ceramic Production andCirculation in the Greater Southwest: SourceDetermination by INAA and ComplementaryMineralogical Investigations, edited by Donna M.Glowacki and Hector Neff, pp. 15–36. Monograph 44,Cotsen Institute of Archaeology, University of California,Los Angeles.

Neff, Hector, Jeffrey Blomster, Michael D. Glascock, Ronald L.Bishop, M. James Blackman, Michael D. Coe, George L.Cowgill, Richard A. Diehl, Stephen Houston, Arthur A.Joyce, Carl P. Lipo, Barbara L. Stark, and Marcus Winter

2006 Methodological Issues in the Provenance Investigationof Early Formative Mesoamerican Ceramics. LatinAmerican Antiquity 17:54–76.

Roper, Donna C.2006 The Central Plains Tradition. In Kansas Archaeology,

edited by Robert J. Hoard and William E. Banks, pp. 105–132. University Press of Kansas, Lawrence, in associationwith the Kansas State Historical Society.

2007 100o W Longitude: Exploring Cultural Dynamics at theWestern Edge of the Central Plains Tradition. CentralPlains Archeology 12 (in press).

Scheiber, Laura L.2006 The Late Prehistoric on the High Plains of Western

Kansas: High Plains Upper Republican and Dismal River.In Kansas Archaeology, edited by Robert J. Hoard andWilliam E. Banks, pp. 133–150. University Press of Kansas,Lawrence, in association with the Kansas State HistoricalSociety.

Speakman, Robert J., and Michael D. Glascock2004 Instrumental Neutron Activation Analysis of Central

Plains Pottery. Report submitted to the Kansas StateHistorical Society and the Department of Sociology,Anthropology, and Social Work, Kansas State University,Manhattan.

Smith, Loren M.2003 Playas of the Great Plains. Peter T. Flawn Series in

Natural Resource Management and Conservation, NumberThree. University of Texas Press, Austin.

336


Future Plains Conference Meetings

66th Plains Conference, Laramie, Wyoming, Marcel Kornfeld, GC Frison Institute, Box 3431,University Station, University of Wyoming, Laramie WY 82071, 307-766-2473, [email protected],date and location pending.

67th Plains Conference, Norman, Oklahoma, Robert Brooks, Oklahoma Archeological Survey, Uni-versity of Oklahoma, Norman, OK 73019, [email protected], date pending.