Madisonville and Madisonville Horizon Fort Ancient Ceramics

Davidson

Preliminary mineralogical and chemical study of Pre- Madisonville and Madisonville horizon Fort Ancient ceramics

Michelle M. Davidson

Faculty Mentors: Tammie L. Gerke, PhD (Department of Physics and Geology) A. Gwynn Henderson, PhD (Kentucky Archaeological Survey) Northern Kentucky University Highland Heights, KY 41099 Abstract

Fort Ancient ceramics can be divided into two distinct groups referred to as Pre-Madisonville (A.D. 1000-1400) and Madisonville (A.D. 1400-1750), based on stylistic features. Ceramics from these two periods were analyzed to test the hypothesis that Pre-Madisonville peoples utilized a large number of local clay sources and that Madisonville peoples utilized a limited number of clay sources. X-ray diffraction and x-ray fluorescence analyses were conducted on 15 ceramic sherds from the Thompson, Fox Farm, and Petersburg sites. Each site is a multi-component Fort Ancient site located in the Middle Ohio River Valley.

The mineralogy of the ceramics from all three sites are indistinguishable both temporally and spatially and contain quartz, micas, and trace amounts of feldspars. Compositional variation for the ceramics was documented between the sites using multivariate statistics and bivariate plots. Ceramics from the Petersburg site appeared to be distinguishable temporally, but not spatially. Ceramics from the Thompson and Fox Farm sites are distinguishable spatially, but not temporally. Thus, it appeared that the changes in the style of ceramics between the Pre-Madisonville and Madisonville correspond to a change in clay sources at the Petersburg site, but not at the Thompson and Fox Farm sites.

Introduction al 1995; Lynott et al 2000; and Steponaitis et al 1996). Archaeologists consider the introduction of However, these types of studies have been conducted ceramics a profound technological leap. Ceramic in areas with a limited number of clay sources. Miner- styles evolved following the introduction of ceramic alogical analyses of clays and ceramic pastes could technology. This evolution of ceramic technologies help establish the source of clays used in the ceramic and styles gives ceramic artifacts high diagnostic industries of certain cultures, temporal periods, or spe- value. Ceramic sherds and vessels can be easily associ- cific sites. This might shed some light on prehistoric ated with specific prehistoric cultures and/or time peri- trade, travel routes, and procurement strategies. ods, and in some cases, with specific activities. Thus, ceramics are some of the most analytically useful arti- The manufacture of ceramics began in the Middle fact classes found on prehistoric sites. Ohio River Valley (MORV) (Figure 1) approximately 3000 years ago. Several different prehistoric cultures Archaeologists have traditionally categorized made pottery in this region, such as the Adena, ceramic types based on qualitative observations of Hopewell, Newtown, and Fort Ancient cultures. The temper, surface treatment, and morphology. Although most recent MORV ceramic producers were the Ft. the value of this kind of analysis has been recognized, Ancient peoples (A.D. 1000-1750). The high number identification of the mineralogy of ceramic paste has of possible clay sources in the MORV (Birkeland been less often used (Carr and Komorowski 1995; 1999) may have contributed to the lack of sourcing Isphording 1974; Kamilli and Steinberg 1985; Lizee et studies in this region.

Norse Scientist 91 Davidson

Figure 1. Modified map of the Middle Ohio River Valley from Henderson, 1992. Site locations are indicated by filled triangles.

.The Ft. Ancient culture was not static throughout shell ornaments possibly from Mississippian chiefdom its 750-year history (Henderson 1998; Pollack and societies in eastern Tennessee; and Henderson 1992, 2001). Archaeologists have documented many important changes in Ft. Ancient culture 5) A marked change in ceramics both stylistically and after A.D. 1400. Archaeologists refer to the period of in vessel form diversity. Ft. Ancient culture from A.D. 1400 to approximately 1750 as the Madisonville Horizon. The Ft. Ancient period circa A.D. 1000-1400 will be referred to as the Several Pre-Madisonville sub-regional ceramic Pre-Madisonville Period for the purposes of this study. styles, such as Anderson, Fox Farm, Baum, and Jessa- The shift from the Pre-Madisonville period to the Mad- mine, give way to the pan-regional Madisonville style isonville horizon is marked by: (Sharp 1990; and Turnbow and Henderson 1992). Jars, which predominated during the Pre-Madisonville 1) Changes in physical layout of villages from those period, are joined by bowls, pans, and colanders during that were circular with a central plaza to villages com- the Madisonville Horizon. Archaeologists postulate posed up of scattered house clusters; that greater interaction among and between Ft. Ancient villages (i.e., potters sharing information) may account 2) An increase in village size with a corresponding for the pan-regional ceramic homogeneity in ceramic decrease in the number of villages, which may reflect style and vessel form that developed after A.D. 1400. an aggregation of small villages; The most obvious change in Madisonville Hori- 3) Changes in burial practices from placement of bod- zon ceramics is a thinning of the vessel walls and an ies in low earthen mounds, around village central pla- appearance more refined to the modern eye. This may zas, or the dead being buried in areas adjacent to house reflect an improvement in firing techniques. It is also clusters; possible that whereas the Pre-Madisonville people utilized local, low quality clay sources, the Madisonville 4) An increase in exchange of exotic goods with other people utilized a limited number of high quality clay cultures outside of the MORV, as evidenced by marine sources. Because archaeologists have not incorporated

92 Norse Scientist, Northern Kentucky University Davidson the identification of the mineralogy and chemistry of Petersburg. Petersburg is located in northwestern ceramic pastes and clay sources in their analyses of Ft. Boone County, Kentucky. It lies in the floodplain Ancient ceramics, they do not know whether available immediately adjacent to the south bank of the Ohio clays at Ft. Ancient can be distinguished from each River in the Hills of the Bluegrass region (Henderson other on a sub-regional basis or, most importantly, 1993). Boone County is underlain by Upper Ordovi- whether the changes identified in Ft. Ancient ceramics cian limestones, shaly limestones, and shales. Streams also extend to the kinds of clays Ft. Ancient potters in this area contain glacial gravels, sands, and silts used. (Webb and Funkhouser 1932). In addition, Petersburg is underlain by Pleistocene and Holocene alluviums Did all Ft. Ancient potters after A.D. 1400 use (Swadley 1972). clay from a few particular sources (which might be reflected in ceramics with uniform pastes), or did Ft. Sample Selection and Preparation Ancient potters continue to use the same sources of clay they had used previously? The purpose of this Sample Selection. Five body sherds were obtained study is to determine whether changes in the use of from each Pre-Madisonville Period and Madisonville clay sources mirror the changes documented in Ft. Horizon component at each site (Table 1). All of the Ancient ceramic technology, vessel form, and style sherds were shell tempered and ranged in size from 4 after A.D. 1400. to 6 cm2 (Figure 2). Recognizing the diversity of vessel types in the Madisonville Horizon, care was taken to Methods select sherds exhibiting definite morphological charac- teristics of a jar. Site Descriptions The northern Kentucky three, multi-component Potsherd Preparation. A portion of each potsherd sites chosen for this study, were Thompson (15Gp27), was ground to a particle size of 5 to 10 microns using a Fox Farm (15Ms1), and Petersburg (15Be6, Figure 1). tungsten-carbide ball mill. A portion of the powder Potsherds were chosen from these sites because they from each potsherd was pressed into a pellet using a are within or adjacent to the Middle Ohio River Valley Spex 3624 B X-Press for x-ray fluorescence analysis. and contain vertically and/or horizontally stratified, Each pellet was held at 20 tons for one minute. chronometrically dated Pre-Madisonville and Mad- isonville components. Analytical Conditions

Thompson. Thompson is located in northwestern X-Ray Diffraction Analysis. The powders were then Greenup County, Kentucky. It lies in the floodplain on analyzed using a Seimens D-500 x-ray diffraction sys- the south bank of the Ohio River. The site is situated tem with a copper x-ray tube in the Geology Depart- near the western edge of the Cumberland Plateau ment at the University of Cincinnati. The 2θ ranged where it borders the Knobs (Henderson, 1992). The from 2o to 50o with a step of 0.05o. A count time of 1 bedrock of the Cumberland Plateau consists of Penn- o sylvanian and Mississippian sandstones, conglomer- second was used at each step. The 2θ range of 2 to o ates, and shales alternating with high grade coal beds 50 was chosen because it covered a wide range of (Bailey and Winsor 1964). clay and other silicate mineral compositions.

Fox Farm. Fox Farm is located in southern Mason X-Ray Fluorescence. A Rigaku 3070 x-ray fluores- County, Kentucky. It lies in an upland setting approxi- cence spectrometer, located in the Geology Depart- mately 2.5 km south of the North Fork of the Licking ment at the University of Cincinnati, was used to River (Turnbow 1992). It is located in the Outer Blue- analyze all of the potsherds. The samples were ana- grass physiographic region, which is underlain by lyzed for SiO2, TiO2, Al2O3, Fe2O3 (total iron), MnO, Upper Ordovician limestones (McFarlan 1961). This MgO, CaO, Na2O, K2O, P2O5, Nb, Zr, Sr, Rb, th, Pb, site sits on a broad, gently rolling ridge top that exhib- Zn, Cu, Ni, Cr, and V. Intensity data were converted to its mild karst topography (Turnbow 1992). weight percent, or parts per million (ppm), using multi- ple regressions applied to USGS and NIST soil and

Norse Scientist 93 Davidson

sedimentary rock standards and loss of ignition (LOI) Table 2 lists theresults of the XRF analysis and LOI data. A small portion of each ground sherd was heated values for each sherd. to 1000 oC for one hour to determine the LOI values.

94 Norse Scientist, Northern Kentucky University Davidson

a) Thompson Pre-Madisonville sherds.

b) Fox Farm Pre-Madisonville sherds.

c) Petersburg Madisonville sherds

Figure 2. Examples of fresh water shell tempered sherds from each site.

Results the x-ray signal and the horizontal axis represents the 2θ range in each figure. All traces contain a strong cal- XRD Results and Interpretation cite peak, because of the fresh water shell temper The mineralogy of the Pre-Madisonville and Mad- present in all of the sherds. There may be calcite in the isonville sherds from the Thompson, Fox Farm, and sherds, which is inherent to the source clays, but I was Petersburg sites are presented in Figures 3 through 5, not able to distinguish it from the shell temper. The respectively. These figures were used to assess the predominant mineralogy for both the Pre-Madisonville temporal variability of the mineralogy of the sherds. and Madisonville sherds for all three sites is quartz, Each figure contains a representative XRD trace for mica (probably chlorite), and trace amounts of feld- the Pre-Madisonville and Madisonville sherds for a spars. Thus there appear to be no temporal changes in given site. The vertical axis represents the intensity of the mineralogy of the sherds within each site.

Norse Scientist 95 Davidson

96 Norse Scientist, Northern Kentucky University Davidson

Figures 6and 7 depict representative traces from become vitreous and form porcelain. When the sherds each site. Figure 6 is an analysis of the Pre-Madison- in this study were analyzed there were no peaks for ville sherds, and Figure 7 is an analysis of the Mad- clays or glasses. Thus the sherds in this study were isonville sherds. These figures were used to assess the probably fired above 500 oC but below 900 oC regional variability of the mineralogy of the sherds. The mineralogy for both the Pre-Madisonville and the XRF Results and Interpretation Madisonville sherds from all the sites is predominantly Major, minor, and trace element data are presented quartz, micas (probably chlorite), and trace amounts of in Table 2. PCA analysis was applied to assess if the feldspar. Thus, it is clear that there is little to no temporal and regional relationships of the sherds were regional variability in the mineralogy of the sherds. reflected in their chemistry. Based on their chemistry, divergence in the samples was assessed by applying a Also, based on the mineralogy of the sherds, I variance-covariance matrix to the data set. Fresh water determined an approximate temperature range for their shells and clays incorporate calcium (Ca) and stron- firing. Clays, which were used to make the sherds, are tium (Sr) into their structures. Concentrations of these typically referred to as hydrous silicates (DHZ 1992). elements can be as high or higher in shell material than When heated above 500 oC the hydroxyl component is in clays (Lizee et al 1995), therefore, they were driven off. This results in a collapse of the clay struc- excluded from this analysis. Phosphorous was also ture, which results in mica-like structures. If the clays excluded because of atypically high concentrations for are heated to approximately 900 oC or higher, the clays

Figure 3. Representative XRD traces for the Pre- Figure 4. Representative XRD traces for the Pre-Madisonville Madisonville and Madisonville sherds from the Thompson and Madisonville sherds from the Fox Farm site. Q = quartz, site. Q = quartz, M = micas, and F = feldspars. M = micas, and F = feldspars.

Norse Scientist 97 Davidson

20.8 ppm while the Madisonville sherds displayed a larger range from 9 to 22.9 ppm. Figure 10 is the Petersburg site (see Figure 1). The Pre-Madisonville sherds had high Cr (39.1 to 73.0 ppm) and Pb (13 to 16.6 ppm) concentrations compared to the Madison- ville sherds (Cr 12.4 to 48.8 ppm and Pb 10.1 to 11.9 ppm). The Pre-Madisonville and Madisonville clusters have the best separation at the Petersburg site and the greatest degree of overlap at the Thompson siteNext, the spatial variability was evaluated (Figure 11). I observed that the Fox Farm and Petersburg sites overlapped between 40 to 60%. The Petersburg site appeared to share a common boundary with the Thompson site. I observed that the Fox Farm and Thompson sites did not overlap. Based on the small data set, I distinguished sherds from the Fox Farm and Thompson sites based on their Cr and Pb concentrations. However, I could not distinguish sherds from Fox Farm and Petersburg and probably would not be very successful in trying to distinguish Petersburg and Thompson site sherds based on their Cr and Pb concentrations.

Conclusions We reached the following conclusions:

Figure 5. Representative XRD traces for the Pre-Madisonville 1) The mineralogy of the Pre-Madisonville and Mad- and Madisonville sherds from the Petersburg site. Q = isonville ceramic pastes are indistinguishable both quartz, M = mica, and F = feldspars. temporally and spatially. MORV clays, which suggests that it might also be a 2) The ceramic pastes in this study were probably fired temper component rather than a clay component.. to about 700 oC. This is based on the comparison of the The population of sherds were best separated by XRD traces. chromium (Cr) and lead (Pb). First, the temporal variations for each site were evaluated. Figure 8 is the 3) The chemical composition of the of the Pre-Mad- Thompson site (see Figure 1). The Pre-Madisonville isonville and Madisonville ceramic pastes are not tem- and Madisonville clusters overlapped between 33 to porally unique for Thompson and Fox Farm but are 66%. The Pre-Madisonville cluster displayed the spatially unique from each other. I could not identify greatest range in both Cr (11.2 to 65.3 ppm) and Pb them temporally on the basis of chemical composition (14.9 to 53.5 ppm) concentrations compared to the alone. Madisonville cluster (Cr 36.7 to 56.9 ppm and Pb 21.2 to 46.6 ppm). Figure 9 is the Fox Farm site (see Figure 4) The chemical composition of the Pre-Madisonville 1). The Pre-Madisonville and Madisonville clusters and Madisonville ceramic pastes are unique temporally overlap strongly. The Pre-Madisonville sherds had Cr for Petersburg but are not unique spatially from the concentrations that range from 52.4 to 100.1 ppm com- Fox Farm or Thompson site. Thus one could not iden- pared to the Madisonville sherds that range from Cr tify them spatially on the bases of chemical composi- 66.5 to 126.2 ppm. The Pb concentrations displayed tion alone. greater variations between the two clusters. The Pre- Madisonville cluster had a limited range from 15.4 to 5) At the Fox Farm and Thompson sites the Pre-Mad- isonville and Madisonville Ft. Ancient peoples

98 Norse Scientist, Northern Kentucky University Davidson

Figure 6. Representative XRD traces for the Pre-Madisonville Figure 7. Representative XRD traces for the Madisonville sherds from all three sites. Q=quartz, M=mica, and F= sherds from all three sites. Q=quartz, M=mica, and F= feldspars. feldspars.

.appeared to have used the same clay source(s). How- Acknowledgements ever, at the Petersburg site it appeared that the Mad- I would like to thank the William S. Webb Museum of isonville peoples used different clay sources than the Anthropology for the use of Ft. Ancient potsherds, and Pre-Madisonville peoples. the Department of Geology at the University of Cin- cinnati for use of the XRD and XRF. I would also like

Norse Scientist 99 Davidson

0 10 15 20 25 30 35 40 45 50 55 60

Pb ( ppm)

Figure 8. Test of temporal variability: Thompson. The filled diamonds represent Pre-Madisonville and the filled squares represent Madisonville sherds. to thank Drs. Warren D. Huff and J. Barry Maynard at [dissertation]. Lexington, KY: University of the University of Cincinnati for their time and conver- Kentucky. sations. Henderson AG. 1993. Prehistoric research at References Petersburg (15BE6), Boone County, Kentucky: archaeological report 289. Program for Cultural Bailey HH, Winsor JH. 1964. Kentucky soils. Misc. Resource Assessment. Lexington, KY: University Pub. 308. U.K. Agricultural Experiment Station in of Kentucky. Cooperation with U.S.D.A. Soil Conservation Service. Henderson AG. 1992. Introduction. In: Henderson AG, editor. Fort Ancient cultural dynamics in the Birkeland PW. 1999. Soils and geomorphology. middle Ohio valley. Madison: Prehistory Press. Oxford: Oxford University Press. 103 p. p. 2. Carr C, Komorowski JC. 1995. Identifying the clay Henderson AG, Pollack D. 1992. The Thompson Site. mineralogy of rock temper in ceramics using x- In: Henderson AG, editor. Fort Ancient cultural radiography. Am Antiquity 60(4): 723-749. dynamics in the middle Ohio valley. Madison: Deer WA, Howie RA, Zussman J. 1992. An Prehistory Press. p. 31-50. introduction to the rock-forming minerals 2nd ed. Isphording WC. 1974. Combined thermal and x-ray New York: Longman Scientific and Technical. diffraction technique for the identification of 1992. 696 p. ceramic ware temper and paste minerals. Am Henderson AG. 1993. Middle Fort Ancient villages Antiquity 39 (3): 477-483. and organizational complexity in central Kentucky

100 Norse Scientist, Northern Kentucky University Davidson

140

120

100

20 0 5 10 15 20 25 Pb (ppm)

Figure 9. Test of temporal variability: Fox Farm. The filled diamonds represent Pre-Madisonville and the filled squares represent Madisonville sherds.

0 6 8 10 12 14 16 18 20

P b (ppm)

Figure 10. Test of temporal variability: Petersburg. The filled diamonds represent Pre-Madisonville and the filled squares represent Madisonville sherds.

Kamilli DC, Steinberg A. 1985. Archaeological Lizee JM, Neff H, Glascock MD. 1995. Clay Geology. New Haven: Yale University Press.. acquisition and vessel distribution patterns: neutron activation analysis of late Windsor and

Norse Scientist 101 Davidson

140

120

100

0 8 13182328333843485358 Pb (ppm)

Figure 11. Test of spatial variability: All the sherds. The sherds are represented as filled triangles for Thompson, filled squares for Fox Farm, and filled diamonds for Petersburg.

Shantok tradition ceramics from southern New Steponaitis VP, Blackman MJ, Neff H. 1996. Large- England. Am Antiquity 60(3): 515-530. scale patterns in the chemical composition of Lynott MJ, Neff H, Price JE, Cogswell JW, Glascock Mississippian pottery. Am Antiquity 61(3): 555- MD. 2000. Inferences about prehistoric ceramics 572. and people in southeast Missouri: results of Swadley WC. 1972. Geologic map of parts of the ceramic compositional analysis. Am Antiquity Lawrenceburg, Aurora, and Hooven quadrangles, 65(1): 103-126 Boone County, Kentucky. Washington, D.C.: U.S. McFarlan AC. 1961. Geology of Kentucky. Kentucky Geological Survey. Department of Economic Development in Turnbow CA. 1992. The Fox farm site. In: Henderson cooperation with the Kentucky Geological AG, editor. Fort Ancient cultural dynamics in the SurveyPollack D, Henderson AG. 1992. Insights middle Ohio valley. Madison: Prehistory Press. p. into Fort Ancient culture change: A view from 51-68. south of the river. In: Cultures before contact: the Turnbow CA, Henderson AG. 1992. Ceramic analysis. late prehistory of Ohio and surrounding regions. In: Henderson AG, editor. Fort Ancient cultural Ohio Archaeological Council. p. 194-227. dynamics in the middle Ohio valley. Madison: Pollack D, Henderson AG. 1992. The possible role of Prehistory Press. p. 113-135. salt production in Fort Ancient cultural Webb WS, Funkhouser WD. 1932. Archaeological development from A.D. 1200 to 1550. Research in survey of Kentucky. Reports in archaeology and Economic Anthropology, Supplement 6. p. 77-99. anthropology. Lexington, Kentucky: University of Sharp WE. 1990. The Fort Ancient period. The Kentucky. archaeology of Kentucky: past accomplishments and future directions. Kentucky Heritage Council, State Historic Preservation Comprehensive Plan Report No. 1. 2: 467-471.

102 Norse Scientist, Northern Kentucky University