A Chronology of Middle Missouri Plains Village Sites Craig M. Johnson with contribution by Stanley A. Ahler, Craig M. Johnson, Herbert Haas, and Georges Bonani s m i t h s o n i a n c o n t r i b u t i o n s t o a n t h r o p o l o g y ? n u m b e r 4 7 Smithsonian Institution Scholarly Press SERIES PUBLICATIONS OF THE SMITHSONIAN INSTITUTION Emphasis upon publication as a means of ?diffusing knowledge? was expressed by the fi rst Secretary of the Smithsonian. In his formal plan for the Institution, Joseph Henry outlined a program that included the following statement: ?It is proposed to publish a series of reports, giving an account of the new discoveries in science, and of the changes made from year to year in all branches of knowledge.? This theme of basic research has been adhered to through the years by thousands of titles issued in series publications under the Smithsonian imprint, com- mencing with Smithsonian Contributions to Knowledge in 1848 and continuing with the following active series: Smithsonian Contributions to Anthropology Smithsonian Contributions in History and Technology Smithsonian Contributions to the Marine Sciences Smithsonian Contributions to Paleobiology Smithsonian Contributions from the United States National Herbarium Smithsonian Contributions in Visual and Material Culture Smithsonian Contributions to Zoology In these series, the Institution publishes small papers and full-scale monographs that report the research and collections of its various museums and bureaus. The Contributions Series are distributed by mailing lists to libraries, universities, and similar institutions through- out the world. Manuscripts submitted for series publication are received by the Smith- sonian Institution Scholarly Press from authors with direct affi liation with the various Smithsonian museums or bureaus and are subject to peer review and review for compliance with manuscript preparation guidelines. General requirements for manuscript preparation are on the inside back cover of printed volumes. For detailed submissions require- ments and to review the ?Manuscript Preparation and Style Guide for Authors,? visit the Submissions page at www.scholarlypress.si.edu. smithsonian contributions to anthropology ? number 47 Washington, D.C.Was i , .C. 2007 a Chronology of Middle Missouri Plains Village sites Craig M. Johnson with contribution by stanley a. ahler, Craig M. Johnson, herbert haas, and georges Bonani aBstRaCt Johnson, Craig M. A Chronology of Middle Missouri Plains Village Sites, with contribution by stan- ley a. ahler, Craig M. Johnson, herbert haas, and georges Bonani. smithsonian Contributions to anthropology, number 47, 344 pages, 69 figures, 34 tables, 2007. ? a comprehensive and systematic research effort focusing on refining the chronology of individual Plains Village tradition sites from the Middle Missouri subarea of the great Plains relies on a number of absolute and relative dating tech- niques. seventy-four conventional and aMs radiocarbon dates from short-lived materials (seeds, corn) and charred ceramic pot residues are used in conjunction with ceramic ordinations of 225 components assigned to nine cultural variants or phases. site stratigraphy, Euro-american trade materials, historic documentation, oral traditions, historical linguistics, and craniometric distance are also employed to help interpret temporal information derived from the ceramic ordinations and radiocarbon dates. the emphasis of this research is on the southern part of the Middle Missouri subarea, namely that portion along the Missouri River that flows through south Dakota and southern north Dakota. this area is the ancestral homeland of the Mandan and arikara. the existing chronology of the hidatsas who oc- cupied the Knife region in north-central north Dakota is integrated into the Mandan and arikara cul- tural sequence. other cultural chronologies from the northeastern (Cambria, Mill Creek, great oasis, lower James River) and Central Plains (Lower Loup/historic Pawnee) are also related to the Middle Missouri and Coalescent tradition sequences of the Dakotas. this information is used to reconstruct the settlement history of the Mandan, hidatsa, and arikara in the Middle Missouri subarea, which is divided primarily into 50- and 100-year time segments. areas of future research that would improve the Plains Village chronology or would benefit from a refined chronology are also reviewed. the re- sults of the radiocarbon dating, which reviews 301 extant dates and the 74 Plains Village Dates (PVD) obtained during this study, suggest that 50 percent of the existing dates can be accepted, whereas 76 percent can be accepted for the PVD dates. a limited number of dates from the same site contexts also indicate that charred residues from ceramic vessels date 100 to 250 years earlier than their aMs coun- terparts on seeds and corn. the results also suggest a revision in the time span of the cultural variants to the following ranges: (1) initial Middle Missouri, ad 1000?1300; (2) Extended Middle Missouri, ad 1200?1400; (3) terminal Middle Missouri, ad 1400?1500; (4) initial Coalescent, ad 1300?1500; (5) Extended Coalescent, ad 1400/1450?1650; and (6) Post-Contact Coalescent, ad 1650?1866. the distribution of sites along the Missouri River from ad 1000 to 1866 is complex and dynamic, reflecting multiple Late Woodland origins, locally available resources, cultural continuities and discontinuities, village consolidations and dispersals, widespread warfare, and the exposure to epidemic diseases and to Euro-american trade. Published by smithsonian institution scholarly Press P.o. Box 37012 MRC 957 Washington, D.C. 20013-7012 www.scholarlypress.si.edu Library of Congress Cataloging-in-Publication Data Johnson, Craig M. a chronology of middle Missouri Plains village sites / Craig M. Johnson ; with contribution by stanley a. ahler . . . [et al.]. p. cm. ? (smithsonian contributions series. smithsonian contributions to anthropology ; no. 47) includes bibliographical references and index. 1. indians of north america?great Plains?antiquities. 2. indians of north america?great Plains?history?Chronology. 3. Excavations (archaeology)?great Plains. 4. Radiocarbon dating? great Plains. 5. great Plains?antiquities. i. ahler, stanley a. ii. title. E78.g73J583 2007 930.1?dc22 2007022132 the paper used in this publication meets the minimum requirements of the american national standard for Permanence of Paper for Printed Library Materials Z39.48?1992. List of figuREs ix List of taBLEs xiii PREfaCE xv 1 intRoDuCtion 1 objectives 2 Environmental and Cultural setting 3 2 PREVious CuLtuRE-histoRiCaL REConstRuCtions 9 Description and Early Classification Period (1883?1945) 9 salvage archeology and Culture history Period (1945?1970) 10 Contract archeology and Cultural Ecology Period (1970?Present) 11 3 MEthoDs of Dating 17 Ceramic ordination 17 applications in the Middle Missouri subarea 17 initial Middle Missouri 18 Extended Middle Missouri 19 initial Coalescent 19 Extended Coalescent 19 Post-Contact Coalescent 20 Knife Region and Related areas 21 Methodological Considerations 22 theoretical Considerations 26 Correspondence analysis 28 Dendrochronological Dating 31 Radiocarbon Dating 31 site stratigraphy 34 Euro-american trade Materials 35 historic Documentation 35 oral traditions 36 historical Linguistics 36 Craniometric Distance 36 Contents i v ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y 4 Data aCquisition 39 Descriptive Rim sherd Categories 45 Radiocarbon samples 49 field sampling strategies 51 5 RaDioCaRBon Dating REsuLts Stanley A. Ahler, Craig M. Johnson, Herbert Haas, and Georges Bonani 53 Review of Criteria for sample selection 53 Mechanics of sample selection and Processing 54 Evaluation of Results of the Radiocarbon Dating Program 57 sites with Complex or Extensive Date series 67 helb (39Ca208) (Extended Middle Missouri) 67 Jake White Bull (39Co6) (Extended Middle Missouri) 68 Lower grand (39Co14) (Extended Coalescent) 70 Walth Bay (39WW203) (Extended Coalescent) 71 sommers (39st56) (initial Middle Missouri) 72 huff (32Mo11) (terminal Middle Missouri) 73 other Dated initial Middle Missouri sites 73 Jones Village (39Ca3) (initial Middle Missouri) 73 fay tolton (39st11) (initial Middle Missouri) 74 Jandreau (39LM225) (initial Middle Missouri) 74 Dodd (39st30) (initial Middle Missouri) 74 Pretty head B (39LM232) (initial Middle Missouri) 75 Cattle oiler (39st224) (initial Middle Missouri) 75 Jiggs thompson B (39LM208) (initial Middle Missouri) 75 other Dated Extended Middle Missouri sites 75 havens (32EM1) (Extended Middle Missouri) 75 Paul Brave (32si4) (Extended Middle Missouri) 76 Vanderbilt Village (39Ca1) (Extended Middle Missouri) 76 Calamity Village (39DW231) (Extended Middle Missouri) 76 sully school (39sL7) (Extended Middle Missouri) 76 Cheyenne River (39st1) (Extended Middle Missouri) 76 Black Widow Ridge (39st203) (Extended Middle Missouri) 77 terminal Middle Missouri sites 77 shermer (32EM10) (terminal Middle Missouri) 77 initial Coalescent sites 78 arzberger (39hu6) (initial Coalescent) 78 Lynch (25BD1) (initial Coalescent) 78 other Extended Coalescent sites 78 Demery (39Co1) (Extended Coalescent) 78 Potts Village (39Co19) (Extended Coalescent) 79 Meander (39LM201) (Extended Coalescent) 79 hosterman (39Po7) (Extended Coalescent) 79 sully (39sL4) (Extended Coalescent) 79 over?s La Roche (39st9) (Extended Coalescent) 80 sites not Dated by the Plains Village Dating Program 80 Broken Kettle West (13PM25) (initial Middle Missouri-great oasis) 80 Williams (13PM50) (initial Middle Missouri-great oasis) 81 n u m b e r 4 7 ? v Larsen (13PM61) (initial Middle Missouri-great oasis) 81 Cambria (21BE2) (initial Middle Missouri-Cambria) 81 Jones (21BE5) (initial Middle Missouri-Cambria) 81 Price (21BE25) (initial Middle Missouri-Cambria) 81 great oasis (21Mu2/17) (initial Middle Missouri-great oasis) 81 Packer (25sM9) (initial Middle Missouri-great oasis) 81 Crow Creek (39Bf11) (initial Middle Missouri, initial Coalescent) 81 swanson (39BR16) (initial Middle Missouri) 82 arp (39BR101) (initial Middle Missouri) 82 Mitchell (39DV2) (initial Middle Missouri) 82 st. John (39hu213) (initial Middle Missouri-great oasis) 82 Bloom (39hs1) (initial Middle Missouri) 82 twelve Mile Creek (39ht1) (initial Middle Missouri) 82 King (39LM55) (initial Middle Missouri) 83 antelope Dreamer (39LM146) (initial Middle Missouri) 83 Langdeau (39LM209) (initial Middle Missouri) 83 heath (39Ln15) (initial Middle Missouri-great oasis) 83 over?s La Roche C (39st9) (initial Middle Missouri) 83 Breeden (39st16) (initial Middle Missouri, Post-Contact Coalescent) 83 Eagle feather (39st228) (initial Middle Missouri) 83 stony Point (39st235) (initial Middle Missouri, Post-Contact Coalescent) 83 Clark?s Creek (32ME1) (Extended Middle Missouri) 84 White Buffalo Robe (32ME7) (Extended Middle Missouri) 84 Bendish (32Mo2) (Extended Middle Missouri) 84 32Mo291 (unclassified/transitional Extended Middle Missouri?terminal Middle Missouri) 84 Cross Ranch (32oL14) (Extended Middle Missouri) 84 fire heart Creek (32si2) (Extended Middle Missouri) 84 Ben standing soldier (32si7) (Extended Middle Missouri) 84 south Cannonball (32si19) (Extended Middle Missouri) 85 travis i (39Co213) (Extended Middle Missouri) 85 indian Creek (39st15) (Extended Middle Missouri) 85 Ketchen (39st223) (Extended Middle Missouri) 85 Durkin (39st238) (Extended Middle Missouri) 85 Whistling Elk (39hu242) (initial Coalescent) 86 Elbee (32ME408) (Extended Coalescent-Related) 86 Molstad (39DW234) (Extended Coalescent) 86 McClure (39hu7) (Extended Coalescent) 86 Little Pumpkin (39hu97) (Extended Coalescent) 86 Bowman (39hu204) (Extended Coalescent) 86 39sL24 (Extended Coalescent) 86 Eagle feather (39st228) (Extended Coalescent) 86 Bower?s La Roche (39st232) (Extended Coalescent) 86 Rattlesnake Keeper (39LM160) (Post-Contact Coalescent) 86 ghost Lodge (39st20) (Post-Contact Coalescent) 87 v i ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y summary of Carbonized sherd Residue Dating 87 summary of Plains Village Dating Program Results 87 Dates from Plains Village Dating sites 88 Dates from non?Plains Village Dating sites 89 6 CERaMiC oRDination REsuLts 91 great oasis 91 taxonomy 91 origins 92 internal Development 92 Demise 93 Radiocarbon Dates 93 analysis 93 Craniometric Distance 97 initial Middle Missouri 98 taxonomy 98 origins 98 internal Development 99 Demise 100 Radiocarbon Dates 101 analysis 101 Craniometric Distance 108 Extended and terminal Middle Missouri 109 taxonomy 109 origins 109 internal Development 111 Demise 111 Radiocarbon Dates 112 analysis 113 Craniometric Distance 118 initial Coalescent 119 taxonomy 119 origins 119 internal Development 120 Demise 120 Radiocarbon Dates 120 analysis 122 Craniometric Distance 122 Extended Coalescent 124 taxonomy 124 origins 124 internal Development 125 Demise 125 Radiocarbon Dates 126 analysis 126 Craniometric Distance 137 Post-Contact Coalescent 137 taxonomy 137 origins 139 internal Development 139 n u m b e r 4 7 ? v i i Demise 140 Radiocarbon and Dendrochronological Dates 140 analysis 141 Craniometric Distance 147 Post-Contact Microstylistic Variation 158 Dating the Plains Village Pattern Variants 163 7 sEttLEMEnt histoRy 167 Period 1 (ad 1000?1100) 168 Period 2 (ad 1100?1200) 172 Period 3 (ad 1200?1300) 174 Period 4 (ad 1300?1400) 178 Period 5 (ad 1400?1500) 178 Period 6 (ad 1500?1550) 185 Period 7 (ad 1550?1600) 188 Period 8 (ad 1600?1650) 191 Period 9 (ad 1650?1700) 191 Period 10 (ad 1700?1750) 194 Period 11 (ad 1750?1785) 197 Period 12 (ad 1785?1830) 199 Period 13 (ad 1830?1886) 202 8 futuRE REsEaRCh 203 Chronological issues 203 great oasis 203 initial Middle Missouri 204 Extended and terminal Middle Missouri 206 initial Coalescent 207 Extended Coalescent 208 Post-Contact Coalescent 209 Woodland to Plains Village transition 211 subsistence-settlement systems 212 Lithic technology 215 Ceramic technology 217 Bone tool technology 217 Cultural interaction 218 Euro-american trade impacts on native tool technologies 219 subtribal Ethnic identities 219 human skeletal Remains 220 Mitigation of Continuing site Destruction 220 aCKnoWLEDgMEnts 223 aPPEnDix a: CERaMiC Data MatRiCEs 227 aPPEnDix B: CRoss REfEREnCE of DEsCRiPtiVE RiM shERD CatEgoRiEs anD CERaMiC tyPEs 245 aPPEnDix C: taBLEs of RaDioCaRBon DatEs 277 aPPEnDix D: DEnDRoChRonoLogiCaL DatEs 295 REfEREnCEs 299 inDEx 321 figures 1. the Middle Missouri subarea with archeological regions, counties, and reservoirs depicted 4 2. Location of selected Plains Village archeological sites from the Middle Missouri, Central Plains, and northeastern Plains subareas 5 3. Plains Village sites located in south Dakota and southern north Dakota employed in the ceramic ordinations 6 4. Cultural chronology of the Middle Missouri subarea 14 5. Decadal radiocarbon calibration curve 33 6. Plot of calibrated radiocarbon date ranges from initial Middle Missouri variant components based on the decadal calibration curve as implemented in CaLiB 3.0.3 94 7. Plot of great oasis components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 96 8. Plot of initial Middle Missouri variant components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 104 9. Plot of calibrated radiocarbon date ranges from Extended and terminal Middle Missouri variant components based on the decadal calibration curve as implemented in CaLiB 3.0.3 112 10. Plot of Extended and terminal Middle Missouri variant components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 115 11. Plot of Extended Middle Missouri variant components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 117 12. Plot of calibrated radiocarbon date ranges from Coalescent tradition components based on the decadal calibration curve as implemented in CaLiB 3.0.3 121 13. Plot of initial Coalescent variant components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 123 x ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y 14. Plot of Extended Coalescent variant components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 129 15. Plot of southern Extended Coalescent variant components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 131 16. Plot of northern Extended Coalescent variant components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 135 17. Plot of Post-Contact Coalescent Le Beau phase components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 143 18. Plot of the relative amounts of metal modified bison scapula hoes, bone knife handles, bone awls, and trade metal artifacts, Post-Contact Coalescent Le Beau phase 145 19. Plot of Post-Contact Coalescent Bad River phase components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 150 20. Plot of Post-Contact Coalescent talking Crow phase components and descriptive rim sherd categories on axis 1 and axis 2, detrended correspondence analysis ordination 156 21. Plot of the relative amounts of metal modified bison scapula hoes, bone knife handles, bone awls, and trade metal artifacts, Post-Contact Coalescent talking Crow phase 159 22. Plot of Post-Contact Coalescent components and rim sherd cord impressed motifs on axis 1 and axis 2, detrended correspondence analysis ordination 162 23. Distribution of Plains Village components within the Middle Missouri subarea during Period 1 169 24. aerial photographs of initial Middle Missouri sites 171 25. Distribution of Plains Village components within the Middle Missouri subarea during Period 2 173 26. Distribution of Plains Village components within the Middle Missouri subarea during Period 3 175 27. aerial photographs of Extended and terminal Middle Missouri sites 176 28. aerial photographs of Extended Middle Missouri sites 177 29. Distribution of Plains Village components within the Middle Missouri subarea during Period 4 179 30. Distribution of Plains Village components within the Middle Missouri subarea during Period 5 181 31. aerial photographs of heart River, Painted Woods, and Knife River complex sites 182 32. aerial photographs of initial, Extended, and Post-Contact Coalescent sites 183 33. aerial photographs of Extended Coalescent sites 184 34. Distribution of Plains Village components within the Middle Missouri subarea during Period 6 186 35. aerial photographs of talking Crow phase, heart River, and Knife River complex sites 187 36. Distribution of Plains Village components within the Middle Missouri subarea during Period 7 189 37. aerial photographs of late Extended and Post-Contact Coalescent sites 190 38. Distribution of Plains Village components within the Middle Missouri subarea during Period 8 192 39. Distribution of Plains Village components within the Middle Missouri subarea during Period 9 193 40. aerial photographs of post-contact period sites 195 41. Distribution of Plains Village components within the Middle Missouri subarea during Period 10 196 42. Distribution of Plains Village components within the Middle Missouri subarea during Period 11 198 43. Distribution of Plains Village components within the Middle Missouri subarea during Period 12 200 44. Distribution of Plains Village components within the Middle Missouri subarea during Period 13 201 Appendix B Figures B.1. initial Middle Missouri rim sherds 251 B.2. initial Middle Missouri rim sherds 252 B.3. initial Middle Missouri rim sherds 253 B.4. initial Middle Missouri rim sherds 254 B.5. initial Middle Missouri rim sherds 255 B.6. initial Middle Missouri rim sherds 256 B.7. initial Middle Missouri rim sherds (Cambria site) 257 n u m b e r 4 7 ? x i B.8. initial Middle Missouri (Cambria site) and great oasis (l?o) rim sherds 258 B.9. initial Middle Missouri (great oasis) rim sherds 259 B.10. great oasis (a?h) and Extended/terminal (i?o) rim sherds 260 B.11. Extended/terminal Middle Missouri rim sherds 261 B.12. Extended/terminal Middle Missouri rim sherds 262 B.13. initial Coalescent rim sherds 263 B.14. initial Coalescent rim sherds 264 B.15. Extended Coalescent rim sherds 265 B.16. Extended Coalescent rim sherds 266 B.17. Post-Contact Coalescent rim sherds 267 B.18. Extended Coalescent rim sherds 268 B.19. Post-Contact Coalescent rim sherds 269 B.20. Post-Contact Coalescent rim sherds 270 B.21. Post-Contact Coalescent rim sherds 271 B.22. Post-Contact Coalescent rim sherds 272 B.23. Post-Contact Coalescent rim sherds 273 B.24. Post-Contact Coalescent rim sherds 274 Appendix D Figure D.1. Plot of the ranges of the latest dendrochronological dates from Plains Village sites within the Middle Missouri subarea 296 tables 1. Commonly used Middle Missouri subarea taxonomic systems 15 2. two-way weighted averaged (uk) table of selected Le Beau phase components, first iteration of correspondence analysis based on descriptive rim sherd frequencies 30 3. List of Plains Village tradition components used in the ceramic ordinations by taxonomic assignment, extent of fieldwork, and source of the ceramic data 40 4. Mean orifice diameters (centimeters) by descriptive rim sherd category for six components 48 5. List of Plains Village tradition components with a potential for radiocarbon dating listed by region, variant, and estimated date based on existing radiocarbon dates and ceramic seriation 50 6. intermediate working list of sites and components to be dated prior to the start of sample selection, february 1992 55 7. Radiocarbon dates from Plains Village sites, smithsonian institution Repatriation office Plains Village dating program 58 8. Data from test excavations in three villages in north Dakota 65 9. Results of dating pottery sherd residues with more routine aMs dates from the same contexts 87 10. Evaluation of radiocarbon dates produced in the Plains Village dating program and previously existing dates from the same or different sites as the Plains Village dating program 88 11. site component and ceramic type scores from the detrended correspondence analysis of great oasis components 95 12. site component and ceramic type scores from the detrended correspondence analysis of initial Middle Missouri components 103 13. site component and ceramic type scores from the detrended correspondence analysis of Extended/terminal Middle Missouri components 114 14. site component and ceramic type scores from the detrended correspondence analysis of Extended Middle Missouri components 116 x i v ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y 15. site component and ceramic type scores from the detrended correspondence analysis of initial Coalescent components 122 16. site component and ceramic type scores from the detrended correspondence analysis of Extended Coalescent components 127 17. site component and ceramic type scores from the detrended correspondence analysis of southern Extended Coalescent components 130 18. site component and ceramic type scores from the detrended correspondence analysis of northern Extended Coalescent components 134 19. site component and ceramic type scores from the detrended correspondence analysis of Le Beau phase components 142 20. frequencies of temporally sensitive post-contact period artifacts and their ratios to ceramic vessels 146 21. site component and ceramic type scores from the detrended correspondence analysis of Bad River phase components 149 22. site component and ceramic type scores from the detrended correspondence analysis of talking Crow phase components 155 23. site component and ceramic type scores from the detrended correspondence analysis of Post-Contact period cord impressed motifs 161 24. Comparison of Middle Missouri subarea Plains Village chronologies (years ad) 163 Appendix A Tables a.1. Data matrix of ceramic rim sherd descriptive categories by component, initial Middle Missouri variant (great oasis) 228 a.2. Data matrix of ceramic rim sherd descriptive categories by component, Middle Missouri tradition (initial variant) 229 a.3. Data matrix of ceramic descriptive rim sherd categories by component, Middle Missouri tradition (Extended and terminal variants) 231 a.4. Data matrix of ceramic descriptive rim sherd categories by component, Coalescent tradition (initial variant) 232 a.5. Data matrix of ceramic rim sherd types by component, Coalescent tradition (Extended and Post-Contact variants) 233 a.6. Data matrix of ceramic rim sherd cord impressed motif categories on straight or braced rim forms, Post-Contact Coalescent variant 239 Appendix B Table B.1. Correlation of descriptive rim sherd categories and defined ceramic types, Middle Missouri subarea 246 Appendix C Tables C.1. summary of sample location and sample type information for radiocarbon samples dated under the smithsonian institution Repatriation office Plains Village 278 C.2. List of radiocarbon dates from Plains Village components 282 C.3. Radiocarbon dates from Plains village sites, combines smithsonian institution Repatriation office and preexisting dates 290 Preface t he funding for this project was initiated through the efforts of Dan Rogers of the smithsonian institution. in november 1991, a purchase order (fP2016180000) to develop a chronology of archeological sites located in the Middle Missouri subarea of the great Plains was granted to the author by the repatriation office of the Department of anthro- pology, smithsonian institution. a second one was written for stanley a. ahler to assist the author in the radiocarbon chronology of the study. the terms of these purchase orders were fulfilled in December 1994 when a draft report was submitted. a final report was submitted early in 1995. in July 1995 Diane tyler, managing editor of the smithsonian Contributions and studies series, expressed an interest in publishing the report. Based on her comments, a revised report was submitted to the Department of anthropology in august 1996. By october 1996, 20 additional radiocarbon dates became available for a series of Middle Missouri tradition sites, as part of an independent effort by ahler. thirty-four additional dates later became available for site 32Mo291 and the huff site as part of separate contract completion reports. this necessitated some revisions of the 1996 report. a decision was made to submit a third revised report in June 2001, a fourth in february 2003, and a fifth and final manuscript in July 2003 for publication. the 2001 version was sent out for peer review that year. although this study was initiated in 1991, in another sense this research really began years ago when i started my graduate studies in the Department of anthropology, university of nebraska-Lincoln. Lincoln, which is the former headquarters for the smithsonian institution?s Missouri Basin Project in salvage archeology, served as a backdrop for my immersion into Middle Missouri arche- ology. My initiation began when i helped to computer code the ceramic vessels from the Walth Bay and Lower grand sites under the direction of f.a. Cala- brese of the Midwest archeological Center, national Park service, Lincoln, ne- braska. in 1975, i began my long association with Carl R. falk, formerly chief of the Midwest archeological Center and subsequently director of the Division of archeological Research, Department of anthropology, university of nebraska- Lincoln. at this time i began the research for my thesis involving the analysis of the ceramic assemblage from the Medicine Crow site (39Bf2), a Post-Contact Coalescent talking Crow phase earthlodge village in the Big Bend reservoir. x v i ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y this collection, then curated at the Midwest archeological Center, was one of many excavated under the auspices of the interagency archaeological salvage Program (isaP). falk and ahler secured funding from interagency ar- cheological services (ias), national Park service in Den- ver, Colorado, to complete the site report from Medicine Crow, which contained an important pre-ceramic compo- nent of particular interest to ahler. then with the illinois state Museum, ahler acted as principal investigator on the project, mostly because of his interest in the pre-ceramic components present at the site. additional contracts to complete the site reports from the Larson (39WW2), h. P. thomas (39st12), and sommers (39st56) Plains Village sites were taken on by falk as the principal investigator through ias with the help of Jack Rudy, J. J. hoffman, and others of ias-Denver. these projects, in addition to inventory surveys and excavations along the east bank of the oahe and Big Bend reservoirs for the u.s. army Corps of Engineers, served as a springboard into Middle Mis- souri archeology for me and for others, such as Dennis toom, terry steinacher, and Darcy Morey. My involvement with falk and ahler initiated a pe- riod of particularly intense research during my graduate and post-graduate career, focusing on statistics, electronic data processing on mainframe computers (then state-of- the-art), and detailed data collection and analysis of large numbers of lithic and ceramic artifacts. an integral part of my work on three of the four sites involved intersite stylis- tic comparisons of ceramic assemblages. this led me into various multivariate statistical techniques, such as factor and cluster analysis, to assess stylistic variability between sites. a portion of the data that appears in the present study was collected during this time (1976?1979). the re- sults of these early efforts are similar to those presented herein and set the foundation for much of my subsequent research. the research emphasis was largely inductive in nature, focusing on a number of basic research questions important to understanding the occupational histories of these sites, the kinds of activities that took place within them, their relationships to other sites, and their culture- historical context. although this occurred at a time when the ?new archeology? was in vogue, this theoretical school had little impact on my day-to-day research activities. the emphasis was mostly culture-historical in nature. in many respects, this report is a culmination of my first 20 years of Middle Missouri research and represents a logical conclusion to this chapter of my professional and personal life. During this period, my interactions with falk and ahler have been intellectually satisfying and broaden- ing experiences, providing me not only with the tools to further my career, but also with an appreciation for quality research and the effort and personal sacrifices it requires takes. i was not the only student at the university of ne- braska who might consider falk and/or ahler as mentors. among others who owe their early research interests to them include Dennis toom, Darcy Morey, Lynn snyder, and Rob Bozell. all worked on various aspects of falk and ahler?s long-term research program in the Middle Missouri, which had its beginning while they were graduate students under Ray Wood at the university of Missouri-Columbia. the enthusiasm and interest that Wood, falk, and ahler generated in their students and colleagues resulted in an espirit de corps among a small group of passionate and dedicated Middle Missouri archeologists. it is very satisfy- ing to know that others share a deep professional and per- sonal commitment to Middle Missouri archeology and are willing to make numerous sacrifices to further knowledge of its prehistory. hopefully, our dedication to understand- ing the history of the native americans who once lived along the Missouri River in the Dakotas will continue to be passed on to future generations. given the continuing problems in funding research in the Middle Missouri sub- area and other extenuating circumstances, this goal will be even more difficult to attain in the future. although the authorship of this report reflects the ac- tual division of writing responsibilities, the effort put into chapter five by ahler, haas, and Bonani far outweighs their scheduled commitments. ahler in particular put in a major effort at selection, preparation, and transmittal of datable samples, far exceeding his planned workload. haas and Bonani were very accommodating in their work on the project. this point can be underscored by the fact that haas agreed to prepare, run, and transmit samples despite a heavy work schedule and during a time when his laboratory at southern Methodist university was being terminated. Because this project renewed ahler?s interest in refining the Plains Village chronology in the Middle Missouri subarea, he secured additional funding from university of north Dakota sources to run an ad- ditional 19 radiocarbon dates from six sites. these data were submitted for analysis on 21 february 1996 and 11 March 1996 and were reported later that year. the re- sults of these dates, together with those presented in this report and elsewhere, form the basis of a larger synthetic treatment of radiocarbon dating of Plains Village manifes- tations within the Middle Missouri subarea. Preliminary results appeared in ahler et al. (1995). Despite his death earlier this year, stan?s inspiration lives on through me and others as we fulfill our shared commitment to under- standing Middle Missouri Plains village archeology. n u m b e r 4 7 ? x v i i any archeologist who has been immersed into a par- ticular research area is well aware of the dangers of losing one?s perspective on life. this problem commonly results in reducing or virtually eliminating one?s personal rela- tionships during particularly intensive periods of research and can result in strains within the family. this research effort is no exception. throughout the 10 years of work on this project, my wife and children have coped during the times i was traveling or working evenings or weekends in my office. as much as anyone, Charlene has endured these hardships and bore the brunt of child rearing during this extended period of time. she has been responsible as much as anyone for maintaining my sanity during the course of this project. although my daughters, alicia and Rachel, were too young to fully appreciate their sacrifices, perhaps some day they will realize what it means to have an archeologist for a father. My mother and father, Eileen and Reuben, also deserve credit for taking care of the chil- dren on saturdays and other particularly crucial periods of research activity. to all of those who have played such an important role in my professional and personal life, i dedicate this report. i t has been more than 35 years since the publication of Donald J. Lehmer?s (1971) Introduction to Middle Missouri Archeology, which has been used by a generation of archeologists to organize and interpret the prehistory of the northern Plains. this publication shared with his earlier synthesis (Lehmer, 1954a:118?159) the same basic taxonomic structure and sequence of cultural events, but it added detail and an absolute time scale from information collected during the intervening years. in this sense, Lehmer?s culture-historical model of the Middle Missouri villagers, the Mandan, hidatsa, and arikara and their prehistoric and protohistoric ancestors, is a formulation that had been defined very early from a limited number of often inadequately reported sites. Most of the research within the Middle Missouri subarea uses a model that is 50 years old. Logan (1977) discussed the events leading to the publication of Lehmer (1971). to the uninitiated, the availability of a comprehensive treatment of Middle Missouri archeology tends to give the impression that the basic culture history of the subarea is complete. other subarea specialists who continue to employ the Lehmer model of 1971 for lack of a suitable alternative, particularly those con- cerned with the cultural sequences in south Dakota, reinforce this perception. Breaking away from a mold that many archeologists have grown up with is a difficult task. Examinations of large series of radiocarbon dates also tend to sub- stantiate a portion of Lehmer?s (1971, table 2) chronology (thiessen, 1977:64? 65), despite problems with dating the cultural complexes of the subarea (Lehmer, 1971:57?58). the thought among some subarea researchers is that ?the basic archeological complexes are blocked out, and by now, most are reduced to a fair degree of temporal control? (Caldwell and D. R. henning, 1978:139). this view may be justified to some extent, for when salvage archeology ended with the dissolution of the smithsonian institution River Basin surveys program in 1969, the pace of field investigations dramatically declined, and with it the timely dis- semination of site reports, radiocarbon dates, and other associated information. in short, many casual readers of the literature of the Middle Missouri Plains villagers may be lulled into a sense of complacency, or overconfidence, that the culture history of these peoples has essentially been completed (see also toom, 1994:483). nothing could be further from the truth. Craig M. Johnson, Minnesota Department of Transportation, 395 John Ireland Boulevard, St. Paul, Minnesota 55155-1899. Stanley A. Ahler (d. February 2007), formerly of Paleo Cultural Research Group, P.O. Box EE, Flagstaff, Ari- zona 86022. Herbert Haas, RC Consultants Inc., 2846 Marida Court, Las Vegas, Nevada 89120. Georges Bonani, Institute of Particle Physics, HPK-H30, ETH H?nggerberg, CH- 8093, Zurich, Switzerland. Manuscript received 11 September 2003; accepted 18 March 2004. 1 introduction 2 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y soon after its publication, some portions of Lehmer?s (1971) synthesis of native american life on the upper Missouri River began to be questioned by a small group of active and dedicated researchers. these inquiries fo- cused on the taxonomic assignments of a number of vil- lages and the time they were occupied (ahler, 1977a:144; falk and Calabrese, 1973; falk and ahler, 1988). More recent research has redefined the temporal limits of some taxonomic units, challenged the validity of others, and cre- ated new culture-historical frameworks (ahler, 1993a:38; Lovick and ahler, 1982:57?65; steinacher, 1983:91?93). it is clear that an integrated program that focuses on the dating of a select number of the hundreds of excavated sites is needed to refine the Plains Village chronology. Objectives the objective of this research was to develop a refined chronology of Plains Village pattern components from the Middle Missouri subarea of the great Plains. the project had six specific goals: (1) assign as many components to specific time periods as possible; (2) establish dating pa- rameters for each of the six cultural variants; (3) evaluate the feasibility of dating short-lived macrobotanical remains and carbon residues on pottery; (4) assess the reliability of radiocarbon dates run by various laboratories; (5) con- struct a settlement history of the Plains villagers who lived along the Missouri River in the Dakotas; and (6) recom- mend future research within the Middle Missouri subarea, particularly as it applies to chronology. these goals were accomplished by employing eight interrelated absolute and relative dating techniques: (1) intersite ceramic ordi- nations; (2) radiocarbon dating; (3) site stratigraphy; (4) Euro-american trade materials; (5) historic documenta- tion; (6) oral traditions; (7) historical linguistics; and (8) craniometric or biological distance. Most previous efforts to refine the chronology of the Plains villagers focused on taxonomic units rather than on individual sites, with the exceptions of ahler and haas (1993). the research pre- sented herein attempts to greatly refine the chronology of the Plains Village pattern by assigning a large number of components to 50- and 100-year periods. the 225 site components involved in this study are from extant archeological collections recovered by vari- ous governmental agencies, universities, museums, and individuals. twelve regional ordinations or seriations, distinguished by employing traditionally defined ceramic types from various phases, variants, and traditions, were the focus of this research. in several instances, the inter- site ordinations were interpreted with the aid of intrasite ordinations, site stratigraphy, and other temporal indi- cators. absolute dates were assigned to the temporal di- mensions in each ordination by dating a select number of components through the use of radiocarbon dating of short-lived organic materials, such as botanical remains, grass, and twigs. Previously dated components also were included in this study. historic documentation of village sites occupied after about ad 1800 was used to anchor the late end of the protohistoric or Post-Contact period components in time and to link the long prehistoric and protohistoric cultural sequences to the tribal groups who occupied the Middle Missouri subarea in historic times, namely the Mandan, hidatsa, and arikara. the resulting temporal sequences also were compared with the related Mill Creek and Pawnee/Lower Loup chronologies, whose sites in iowa and nebraska fall outside of the Middle Mis- souri subarea. a secondary goal of this project was to provide the office of Repatriation of the smithsonian institution with information useful in determining which modern native american tribes are culturally affiliated with the occu- pants of specific Plains Village archeological sites or the taxonomic units created to assign these sites to an overall culture history framework. the Mandan, hidatsa, and arikara were the three ethnic groups in the Middle Mis- souri subarea participating in a Plains Village lifestyle at the time of European and american contact. of particu- lar interest to native americans are the excavated human skeletal remains and associated funerary objects of their ancestors and the ultimate repatriation of these remains. as a consequence, an emphasis in this report is placed on those villages and associated cemeteries containing human skeletal remains. the chronology developed in this report facilitates the assignment of particular sites to the various ethnic groups and tribes who occupied the Middle Mis- souri subarea during the prehistoric, protohistoric, and historic periods. this effort is seen as providing a culture- historical context for the human skeletal remains, thereby assisting the office of Repatriation of the smithsonian in- stitution in fulfilling its responsibilities under the national Museum of the american indian act (20 u.s.C., section ? 80q). the large quantity of human skeletal remains and associated artifacts from Plains Village sites in the Middle Missouri subarea can only be put in their proper context by a methodical examination of the context or location within each village or cemetery. this stems from three basic facts of Middle Missouri archeology: (1) many of the archaeological sites within the subarea and included in the current study have not been fully reported; (2) many sites contain a number of discrete and often mixed occupations n u m b e r 4 7 ? 3 or components; and (3) there is no a priori reason to as- sociate any particular cemetery population with a specific village or component. again, only after a detailed analysis of the archaeological context of the recovered skeletal re- mains and associated artifacts can any assurance be made concerning their relationship with a component. of par- ticular interest to the smithsonian institution is the chron- ological placement and cultural relationships through time of the sully (39sL4), Mobridge (39WW1), Black Widow Ridge (39st203), fay tolton (39st11), Buffalo Pasture village and cemetery (39st6/39st216), Cheyenne River (39st1), Leavitt (39st215), indian Creek (39st15), and swan Creek (39WW7) sites.1 in addition to providing information useful in de- termining the ancestral relationships between the sites in this study and today?s native american tribes, establish- ing refined and site- specific chronologies is crucial to any analyses of culture process. up to this point, most of the chronologies have focused on the phase or variant level, providing only the coarsest of temporal controls and little specific information about sites other than those that are directly dated. this is a particular problem for some stud- ies that propose cultural or biological process models of varying sophistication without chronologies of comparable refinement ( Jantz, 1973; Key and Jantz, 1981; Ramenof- sky, 1987; Rogers, 1990, 1993; Zimmerman and Bradley, 1993). it is no exaggeration to reiterate that ?archeology is a historical science, so time and temporal variability in ar- cheological phenomena are of paramount importance . . . we must keep striving for better temporal control over our data, for without improved chronologies our cultural- historical scenarios lack the precision necessary to guide processual studies into meaningful lines of inquiry? (toom, 1992a:126). envirOnmental and cultural setting the Middle Missouri subarea of the Plains (Wedel, 1961, fig.1; 1983, fig. 6.1; Lehmer, 1971, fig. 20) is de- fined as the area along the Missouri River where it flows through north and south Dakota (figure 1). the subarea consists of four distinct physiographic zones, including the uplands, the breaks, the river terraces and floodplain, and the river itself (Lehmer, 1971:49?53). the valley is greatly modified by the construction of five mainstem dams oper- ated by the u.s. army Corps of Engineers. these dams created reservoirs that have destroyed and/or inundated thousands of archeological sites. a number of sites are currently being directly affected by these reservoirs and by vandals, particularly in the tailwaters of Lake oahe near Mobridge, south Dakota, and in the Big Bend reservoir (Lake sharpe). the interagency archeological salvage Program (iasP) was established to salvage a small por- tion of these cultural resources. Between 1945 and 1969 the iasP coordinated the efforts of (1) the smithsonian institution River Basin surveys (si-RBs) Missouri Basin Project; (2) cooperating institutions, such as universities and historical societies; and (3) the national Park service Midwest archeological Center (nPs-MWaC) (ahler, 1993a:30; thiessen, 1994a:18, 1999). Most of the data used in this report was derived from iasP. Details of the administration of the iasP program, as well as life in the field, are found in Wedel (1967), Lehmer (1971:1?23), Jennings (1985), smith (1992), thiessen (1994a, 1994b, 1999), Wood and hoffman (1994), hurt (1995), Wood (1995), gradwohl (1997), and grange (1997). in his synthesis of Middle Missouri archeology, Lehmer (1971, fig. 21) divided the subarea into the Big Bend, Bad- Cheyenne, grand-Moreau, Cannonball, Knife-heart, and garrison regions. the portion of the river below the Big Bend region was not included within Lehmer?s (1971:37) regional scheme, although it is apparent that the area, herein labeled the fort Randall region (figure 1), has cul- tural relationships to the Big Bend region, as well as to other plains subareas. in an earlier report of Middle Missouri archeology that proved to be a draft of his 1971 synthesis, Lehmer (1965:K28) defined the Lower fort Randall dis- trict or region as encompassing the area between the White and niobrara Rivers. one year later, Lehmer and Caldwell (1966) dropped the ft. Randall and garrison districts in their application of the Willey and Phillips (1958) taxon- omy to the Middle Missouri subarea. ahler (1993b:58?59) proposed two additional tentative regions, not depicted in figure 1 (Lower yellowstone, Little Missouri), as well as a separation of the Knife-heart region into two distinct areas. Related sites located off the Missouri River that had been assigned to the initial variant of the Middle Missouri tradition (Knudson, 1967:278?280; Lehmer, 1971:98; alex, 1981a; tiffany, 1983; toom, 1992a; a. Johnson, 1993), as well as the Coalescent Lower Loup phase and the historic Pawnee (grange, 1968; o?shea, 1989) extended the area of interest into western and southeastern south Dakota, northwestern iowa, southwestern Minnesota, and nebraska (figure 2). Most of the sites that form the basis of this study are located along the Missouri River in south Dakota (figure 3). sites from other Plains subareas are in- corporated for comparative purposes. the Plains villagers of the Middle Missouri subarea lived in circular and rectangular lodges arranged in a va- riety of village plans and subsisted on a dual economy of 4 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y Figure 1. the Middle Missouri subarea with archeological regions, counties, and reservoirs depicted. n u m b e r 4 7 ? 5 Figure 2. Location of selected Plains Village archeological sites from the Middle Missouri, Central Plains, and northeastern Plains subareas. 6 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y Figure 3. Plains Village sites located in south Dakota and southern north Dakota employed in the ceramic ordinations. n u m b e r 4 7 ? 7 bison hunting and horticulture (Wilson, 1917; Lehmer, 1954a:139?140; Lehmer and Wood, 1977; toom, 1992b). the hunting of smaller animals and the gathering of plants added to this subsistence base, which revolved around an annual cycle (see hurt, 1969). the relative importance of their subsistence activities is a source of debate, with recent evidence suggesting a greater reliance on bison in one case (tuross and fogel, 1994:288). Villages were lo- cated on a series of Missouri River terraces and consisted of from several to 200 or more houses, many fortified by a combination ditch and inner wood-post palisade sys- tem. historically, the houses were constructed from wood frames covered with earth (Wilson, 1934). Most villages were probably occupied for one or perhaps two genera- tions (toom, 1992a:124, 1995:363), but some may have been inhabited for a period of two centuries or more, as indicated by thick midden deposits (ahler, 1993b, table 25.2). Rectangular Middle Missouri tradition houses usu- ally ranged in length between 30 and 50 feet and were normally 20 to 30 feet wide. Circular, and more rarely square, Coalescent tradition lodges were usually 30 to 40 feet in diameter but could have been as much as 70 feet. Most houses were semi-subterranean and probably earth covered with a central fire hearth and a number of storage/refuse pits dug into their floors. Most village sites contained large quantities of broken grit-tempered ceramic vessels, ground and chipped stone tools, chipped stone flaking debris, bone tools, fire-cracked rock, and unmodified vertebrate remains indicative of rather inten- sive occupations. small amounts of Euro-american trade items made from metal or glass have been found in villages dating after about ad 1600/1650. human remains were occasionally found on house floors or in pits but were most commonly found in separate cemeteries associated with late Extended Coalescent and Post-Contact Coales- cent period villages (ad 1600?1780). summaries of both the Middle Missouri and Coalescent traditions have been reported in Lehmer (1954a, 1971, 2001), Winham and Lueck (1994), toom (1996), C. Johnson (1998a, 1998c), Winham and Calabrese (1998), Wood (1998, 2001), and Krause (2001). Descriptions of the history and cul- ture of their Mandan, hidatsa, and arikara descendents have been published by Wood and irwin (2001), stewart (2001), and Parks (2001), respectively. Note 1. the site numbers appearing in this report are based upon a trinomial, or smithsonian institution, site designation system. the three components are codes that represent the state, county, and site within each county. the 48 contigu- ous states are listed in alphabetical order and are given a number between 1 and 48, with alaska and hawaii given the final two state numbers. Counties within each state are given a two letter designation. sites within each county are given a number that is generally based upon their order of discovery. the five state numbers in this report are iowa (13), Minnesota (21), nebraska (25), north Dakota (32), and south Dakota (39). thus, using this system, Cheyenne River (39st1) is the first site in stanley County, south Da- kota, to be given a site number. 2 Previous Culture-historical Reconstructions t hroughout the history of systematic archeological investigations in the Middle Missouri subarea of the Plains, archeologists have focused on developing Plains Village culture taxonomies and chronologies (Wood, 1969). these have taken the form of culture-historical syn- theses, which were based mainly upon intuitive understandings of the infor- mation at hand and data-based ceramic seriations or ordinations. the former efforts were in vogue from about 1940 to 1971, whereas ceramic ordinations, in conjunction with radiocarbon dates and other temporal indicators (e.g., site stratigraphy, historic documentation, metal trade artifacts), generally began to be used after that period. the addition of radiocarbon dating during the 1960s facilitated the construction of the most commonly accepted taxonomic and chronological framework (Lehmer, 1971). Work at the Knife River indian Villages national historic site (KnRi) resulted in a revised developmental se- quence for the Knife and heart regions (ahler, 1993b). ahler (1993a:32?41), Billeck and Byrd (1996:4?30), and toom (1996:60?65) presented summaries of archeological research and taxonomy within the Middle Missouri. these and other taxonomies provide the historic context of the present study. DescriptioN aND early classificatioN perioD (1883?1945) this period began with the explorations of the northwestern archaeological survey by theodore Lewis in 1883 and continued through the work of george Will and herbert spinden at Double Ditch and other sites in north Dakota. William D. strong (1940) formulated the first widely accepted chronology of Plains Village sites within the Dakotas. he assigned the villages to the Man- dan, hidatsa, and arikara through the direct-historic approach, using a variety of historic, protohistoric, and prehistoric components that he excavated in the 1930s. a related effort by Wedel (1940) dealt with cultural developments in the Central Plains. the next synthesis to appear (Will and hecker, 1944) was based upon a large number of sites in north Dakota and north-central south Dakota (c.f. Will, 1924). they assigned those sites to three sequential periods (archaic 1 0 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y Mandan, Middle Mandan, Later heart River) that are the approximate equivalents of the Extended Middle Missouri variant, terminal Middle Missouri variant, and the heart River phase, respectively. Bowers (1948a) developed a culture-historical model of the Mandan and hidatsa on the basis of ethnographic information and data recovered from test excavations at a large number of sites in north and south Dakota from 1929 to 1933 for the Logan Museum, Beloit College, Be- loit, Wisconsin. he organized the components into 15 foci (arzberger, Cannonball, Cheyenne, Crow Creek, ft. sully, heart River, huff, La Roche, Lower grand, Mey- ers, Painted Woods, Pierre, sommers-fort-george, upper grand, Woodlands), some of which have been recognized in modified form in later syntheses. some of these units have been identified with particular subgroups of the Mandan and hidatsa. Bowers? cultural reconstruction, based upon the direct-historic approach, never received broad acceptance because neither his dissertation nor the primary data upon which it was based were ever published (ahler, 1993a:35). his model was later eclipsed by the one proposed by Lehmer (1954a, 1971). subsequent work at the Knife River indian Villages national historic site adds credibility to Bower?s model (ahler, 1993a:35). salvage archeology aND culture history perioD (1945?1970) the construction of dams and reservoirs in the Mis- souri River basin by the u.s. army Corps of Engineers and the Bureau of Reclamation after World War ii prompted archeologists to begin to survey and excavate numerous Plains Village sites. this was done under the auspices of the interagency archeological salvage Program (iasP), a multi-agency program composed of the smithsonian insti- tution (which sponsored the smithsonian institution River Basin surveys [si-RBs]) and various cooperating entities (e.g., universities and historical societies) (thiessen, 1999). the accumulating mass of information from this program prompted Robert stephenson (1954), along with other ar- cheologists, to develop a cultural taxonomy of the Middle Missouri and Central Plains subareas that was based upon the Midwestern taxonomic system. this model has five aspects and at least eight foci that focus on the Plains Vil- lage components concentrated in the Bad-Cheyenne and Big Bend regions. Many of these divisions are recognized today, although the names and assigned components have changed. the five aspects of concern here and their current equivalents are the Pahuk (talking Crow phase, historic Pawnee-Lower Loup phase), sanish (Bad River phase), Chouteau (Extended Coalescent variant), aksarben (ini- tial Coalescent variant), and Chamberlain (anderson phase) aspects. a sixth unnamed aspect includes five over focus components (swanson, Mitchell, Brandon, twelve Mile Creek, Dinehart), which are now assigned to the ini- tial Middle Missouri variant. a good summary of many of these taxonomic units from the perspective of the arz- berger site appears in spaulding (1956:65?110). During the 1950s, Wesley hurt produced a series of site reports on his work in south Dakota in which he de- fined a number of foci. the reports in this series contained essentially the same chronology chart dealing with the full range of Middle Missouri and Coalescent taxonomic units, including those defined by hurt?s contemporaries. the over focus of the Mill Creek aspect, initially formu- lated by over and Meleen (1941:40?41) from villages along the James and Big sioux Rivers (Brandon, Mitchell, twelve Mill Creek), was expanded to include the swanson and Brandon sites (hurt, 1951:15?16). hurt?s work at the thomas Riggs site formed the basis for the thomas Riggs focus (hurt, 1953:47?48), a unit now included within the Extended Middle Missouri variant. he also postu- lated the Bennett, La Roche, and akaska foci to include Coalescent components in the Chamberlain, Pierre, and south Mobridge areas (hurt, 1957:26?30, 80). today, these taxonomic units are part of the Extended variant of the Coalescent tradition. his Le Beau focus, currently the Le Beau phase of the Post-Contact Coalescent, was based upon excavations at the swan Creek site (hurt, 1957). at about the same time, Donald Lehmer (1954a: 114?159; 1954b:138?159) developed a culture-historical taxonomy that gained widespread acceptance by incor- porating some more recent minor modifications. this model proposed a Plains Village pattern encompassing all semi-permanent villages in the Central Plains and Middle Missouri subareas. this pattern, or tradition as it later became to be known, is composed of the Middle Missouri, Coalescent, and Central Plains traditions. the Plains Village chronological sequence in the oahe Dam area was designated the fort Pierre Branch in an earlier synthesis (Lehmer, 1952a), a short-lived term that is no longer used. Based upon the superpositioning of circular Coalescent earthlodges upon rectangular Middle Missouri tradition houses at the Dodd site (39st30), Lehmer was able to develop a relative chronological relationship be- tween the two traditions, anchoring the late Coalescent sites into historically documented Mandan, hidatsa, and arikara villages. he envisioned the Coalescent tradition as a blending, or ?coalescence,? of cultural traits of the n u m b e r 4 7 ? 1 1 Middle Missouri and Central Plains traditions. a series of radiocarbon dates were instrumental in establishing ap- proximate temporal spans of the two traditions and their variants (Lehmer, 1971:33). a dendrochronological dat- ing program in the Middle Missouri subarea was under- taken by Weakly (1971), although the results did not gain widespread acceptance because of a variety of problems (Caldwell and snyder, 1983). thiessen (1977) and toom (1992a, 1992b) completed examinations of Middle Mis- souri tradition radiocarbon dates. their results generally supported Lehmer?s temporal positions of Middle Mis- souri tradition variants. Most taxonomies developed in the 1960s dealt with low-order integrative units, such as the phase. Most no- table among these are the grand Detour phase (Caldwell and Jensen, 1969), the Bad River phase (hoffman and Brown, 1967), and the felicia focus (Caldwell, 1966a:80). a broader based taxonomic system, incorporating the Willey and Phillips (1958) system, appears in Deetz (1965:7?18). Deetz discussed two regional sequences, in- corporating two phases (aksarben, Lower Loup) and the historic Pawnee from the Central Plains, and five phases (Middle Missouri, arzberger, La Roche, stanley, snake Butte) from the Middle Missouri region. soon after, Lehmer and Caldwell (1966) and Lehmer (1966:53?54) presented a revised version of the Lehmer (1954a:114? 154) taxonomic scheme, incorporating the units of the Willey and Phillips (1958) system. Lehmer (1966:56?60), in his analysis of the fire heart Creek site, defined the fort yates phase to include all Extended Middle Missouri tradition sites in the Knife-heart and Cannonball regions, then called districts. he also drew comparisons to similar sites in the Bad-Cheyenne district, which was assigned to the thomas Riggs focus. Except for a few modifications made since then, the Lehmer and Caldwell system remains essentially unchanged. Wood (1967:112?146), in a report of his huff site excavations in the Cannonball region, pro- posed the huff focus. it represents the final phase of the Middle Missouri tradition stretching back to the Extended Middle Missouri thomas Riggs focus and the initial Mid- dle Missouri Chamberlain aspect found to the south in the Bad-Cheyenne and Big Bend regions. Wood?s thomas Riggs focus included all sites currently included within the Extended Middle Missouri. Drawing on his earlier syntheses (Lehmer, 1954a:114? 159, 1966:53?67), the taxonomic outline by Lehmer and Caldwell (1966), and a summary of Middle Missouri ar- chaeology (Lehmer, 1965), Lehmer (1971) presented his most influential and lasting contribution to the archeology of the subarea. Because his 1971 synthesis was formulated on information available prior to this time, it is placed in the ?salvage archeology and Culture history Period.? in it, Lehmer defined six regions of the subarea (garrison, Knife-heart, Cannonball, grand-Moreau, Bad-Cheyenne, Big Bend) that are almost identical in their boundaries to the earlier defined districts (Lehmer and Caldwell, 1966). the Plains Village pattern or major cultural tradition is divided into two distinct traditions, the Middle Missouri and the Coalescent, that are in turn subdivided into seven variants, which supersede the earlier defined horizons (Lehmer, 1971:28?33). coNtract archeology aND cultural ecology perioD (1970?preseNt) this period begins with the dissolution of the si-RBs program and the beginning of problem-oriented research. although this transition began several years prior to 1970, this date is chosen as an approximate demarcation point. from 1968 until his death in 1975, Donald J. Lehmer of Dana College in Blair, nebraska, together with W. Ray- mond Wood and his graduate students from the univer- sity of Missouri-Columbia, conducted field investigations at more than 20 Plains Village sites in the upper Knife- heart region (Lehmer, 1967; Lehmer et al., 1978:v?ix; Wood, 1986a:x?xi). this region was chosen because its archeology was largely unknown, and it is the only large portion of the Middle Missouri subarea unmodified by the construction of the mainstem reservoirs. national Park service-Midwest archeological Center excavations from 1969 to 1973, under the overall direction of Carl R. falk and stanley a. ahler, targeted three Plains Village sites (helb, Walth Bay, Lower grand) in the upper oahe Res- ervoir (falk and ahler, 1988; thiessen, 1999:59?61). un- like the earlier work in the upper Knife-heart region, these investigations focused on intensive large-scale excavations. the two programs were not only influential in refining field procedures and the cultural sequences in the Knife, Cannonball, and grand-Moreau regions, but more impor- tantly, they introduced a number of Wood?s graduate stu- dents to Plains Village archeology. some of these students, such as stanley a. ahler and Carl R. falk, have continued their interest in the subarea, representing a ?third genera- tion? of Middle Missouri archeologists who, in turn, be- came mentors to and influenced their own students and colleagues, including this author. Ray Wood?s enthusiasm for research in the subarea has encouraged a number of his students to pursue research into the expanded fields of ethnohistory, cultural ecology, and disease ecology. 1 2 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y the results of some of the 1968 field work in the upper Knife-heart (i.e., Knife) region, as it pertains to cul- ture history and archeological taxonomy, is summarized in Lehmer (1971: 203?206). two studies, Lehmer et al. (1978) and Wood (1986a:7?24), have portions devoted to cultural chronology and taxonomy based upon Lehmer?s framework. Lehmer et al. (1978) focused on the last pe- riod of Plains Village life, the Knife River phase, whereas Wood (1986a:7?24) summarized the culture history of the region from the Middle Missouri through the Coalescent traditions, ending with the non?Plains Village historic Da- kota occupations at several sites. Despite the fact that his synthesis largely reflects earlier culture history formula- tions, Wood extended the heart River phase, originally placed fully within the protohistoric period at ad 1675 to ad 1780 by Lehmer, well back into the prehistoric period at ad 1450 to ad 1780. in summary, both the Lehmer and Wood syntheses have been eclipsed by recent devel- opments stemming from the research program within the Knife River indian Villages national historic site. the validity of the Lehmer culture-historical model, particularly as it pertains to the Coalescent tradition, also has been called into question as a result of recent research within the Knife River indian Villages national historic site (Lovick and ahler, 1982:57?65; ahler, 1993a:38?39). this problem was traced, at least in part, to the fact that Lehmer?s scheme developed from his early work in south Dakota, where the Caddoan-speaking arikara were the dominant group. on the other hand, the north Dakota pre- historic record is largely based upon the siouan-speaking Mandan and hidatsa. the results indicate that the Coales- cent tradition, as it is defined by Lehmer (1971:115?120), is most similar to the Willey and Phillips (1958:33) hori- zon and does not account for the cultural diversity present within the Knife and heart regions of the Middle Missouri subarea. there appears to be at least three cases in which the process of coalescence or diffusion occurred, but they did not occur sequentially nor did they always involved the same groups: (1) when Caddoan speakers moved from the Central Plains to the Middle Missouri; (2) when the southern Mandan adopted architectural and ceramic traits from others while in the Black hills; and (3) when the three groups of hidatsa and two groups of Mandan inter- acted in the Knife and heart regions (ahler, 1993a:38). according to John Ludwickson, it is diffusion, precipi- tated by climatic change (?little ice age?), large-scale warfare, and pre-European epidemic diseases that led to widespread borrowing among groups. Diffusion and not micro-evolution within ethnic groups led to the dramatic material culture changes encompassed under the banner of the ?Coalescent tradition? (pers. comm., 1993). Lehmer (1971:99, 104, 125?128) also identified a number of cases of borrowing between different variants. Bowers? (1948a, 1965:10?25, 476?489) model, al- though limited by a lack of taxonomic rigor, is better able to accommodate the complex cultural developments within the Knife and heart regions. as a result, five partially con- temporaneous and sequential cultural complexes (Charred Body, Middle Missouri, Painted Woods, heart River, Knife River), encompassing 11 phases, are defined for the Knife and heart regions (ahler et al., 1991:27; ahler, 1993b). the Middle Missouri complex includes the fort yates phase and is followed by the huff phase in the Cannonball region, which was assigned by previous researchers to the Extended and terminal variants of the Middle Missouri tradition. the Charred Body complex is partially contem- poraneous with the fort yates phase and represents the earliest villagers in the Knife, heart, and garrison regions. the next two phases, Clark?s Creek and nailati, have com- ponents assigned to both the Middle Missouri and Painted Woods complexes. the scattered Village and Mandan Lake phases also are assigned to the Painted Woods complex, with the latter also containing components of the heart River complex. the heart River complex includes almost all late villages in the heart region assigned to the heart River phase (Lehmer, 1971:203?205). Because there is so little systematic research within the heart region, phases are not defined for the heart River complex there. Data from only two of these sites, Double Ditch and slant, have been reported (Will and spinden, 1906; ahler, schneider et al., 1980; ahler, 1997). heart River complex compo- nents in the Knife region are included in the Mandan Lake, hensler, and Willows phases. some components of the Wil- lows phase also are assigned to the Knife River complex, as are all succeeding Minnetaree, Roadmaker, and four Bears phase villages. Knife River complex components represent the protohistoric villages of the Mandan and hidatsa, in- cluding remnant arikara groups from south Dakota. this complex most closely corresponds to Lehmer?s (1971:203? 206) heart River 2 and Knife River 1 and 2 subphases. the latter two subphases are assigned to Lehmer?s Disorganized Coalescent variant. the dating of the late village compo- nents, traditionally assigned to the Coalescent tradition in the Knife and heart regions, also has been modified. Many were occupied up to 200 years earlier than previously be- lieved (ahler, 1993b, figs. 25.1?25.2). other efforts at refining the Middle Missouri tradition taxonomy focused on initial variant sites along the lower James River in southeastern south Dakota and in the Mill Creek sites in northwestern iowa (alex, 1981b:171?186; n u m b e r 4 7 ? 1 3 tiffany, 1982:93?98, 1983). alex included the Mitch- ell, Bloom, twelve Mile Creek, and goehring sites in the Lower James phase. tiffany proposed adding the swanson and Brandon phases along with the conversion of three foci to phases (anderson, Cambria, thomas Riggs). the Chamberlain and Mill Creek variants, two spatial vari- ants of the initial Middle Missouri, also were added to the growing list of taxonomic units. the Chamberlain variant is similar to hurt?s (1953, chart iii) Chamberlain aspect. tiffany?s two new spatial variants are further solidified with the differentiation of the initial Middle Missouri variant into western and eastern divisions (toom, 1992a, 1992b). steinacher (1990, table 2) presented a summary of initial Middle Missouri taxonomy. in addition to the work at the Knife River indian Vil- lages national historic site, a series of research results from sites in south Dakota has raised questions about the validity of the widely accepted Lehmer model of Plains Village taxonomy and development within the Middle Missouri subarea. one of these challenges the concept that the earliest Coalescent tradition villages, those assigned to the initial variant, represent admixtures of Central Plains and Middle Missouri tradition cultural traits (steinacher, 1983:91?95). Despite earlier pronouncements by Lehmer (1971:111), the results of this research indicate that at least some of these initial Coalescent villages are essen- tially Central Plains tradition sites that lack evidence of a blending of traits. additional radiocarbon dates also in- dicate that the initial Coalescent began about ad 1300, some 100 years before the commonly accepted date for the variant (toom, 1992b, table 1; C. Johnson, 1998a:308), and the time span for the initial Middle Missouri variant along the Missouri River was compressed from Lehmer?s ad 900 to ad 1400, to ad 1000 to ad 1300 (tiffany, 1982, fig. 27; 1983, fig. 5; toom, 1992a: 125, 1992b, table 1). other archeologists have questioned the taxonomic as- signments of Extended Middle Missouri villages in south Dakota (falk and Calabrese, 1973; ahler, 1977a:14?145; a. Johnson, 1977). the implications of these results are addressed in chapter 6. these developmental sequences and taxonomies, en- compassing the period between ad 900 and ad 1886, is reconstructed in figure 4. it is based upon existing radio- carbon dates and the discussions by Lehmer (1971), stein- acher (1983), tiffany (1983), toom (1992a, 1996), ahler (1993b), and C. Johnson (1998a) but does not include the findings in the present study. the resulting taxonomy, composed of the Middle Missouri and Coalescent tradi- tions, is essentially unchanged from Lehmer?s (1971:33), except for the Knife, heart, and Cannonball regions, where ahler?s system was adopted. the contrast between the Lehmer and ahler systems is presented in table 1. the dating of the Middle Missouri and Coalescent traditions and the variants and phases they subsume, however, is in the process of being revised. six of Lehmer?s seven variants (initial, Extended, and terminal Middle Missouri; initial, Extended, and Post-Contact Coalescent) also are recog- nized, with the Disorganized variant absorbed into the Post-Contact Coalescent. the initial variant of the Middle Missouri tradition is divided into nine phases included within eastern (great oasis, Big sioux, Little sioux, Bran- don, Cambria, Lower James phases) and western (swan- son, grand Detour, anderson phases) divisions following the conventions of tiffany (1983) and toom (1992a, 1992b). the Extended Middle Missouri variant, or Middle Missouri complex, includes four phases (thomas Riggs, fort yates, Clark?s Creek, nailati), whereas huff is the sole phase of the terminal variant. the initial Coalescent variant is divided into two phases (Campbell Creek and arzberger), with the anoka phase formulated to include the Lynch site in nebraska (not depicted in figure 4). the Extended variant of the Coalescent tradition is separated into a large number of foci and phases. the most recent, although poorly defined, scheme (Lehmer, 1971:120) pro- posed six phases for the Extended Coalescent (shannon, Bennett, La Roche, akaska, Le Compte, Redbird). four of Lehmer?s (1971:201?206) six Post-Contact Coales- cent phases (felicia, talking Crow, Bad River, Le Beau) also were recognized, pending future modifications. his Knife and heart River phases are eliminated in favor of the Knife and heart River complexes of ahler (1993b). Detailed discussions of the cultural taxonomy of the Plains Village pattern are presented with each ceramic ordination in chapter 6. 1 4 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y Figure 4. Cultural chronology of the Middle Missouri subarea. n u m b e r 4 7 ? 1 5 TabLe 1. Commonly used Middle Missouri subarea Plains Village taxonomic systems (Lehmer, 1971; ahler, 1993b).* Lehmer Taxonomy (all Middle Missouri subarea regions) Tradition Variant Phase/Focus/aspect Type Sites/important Sites Middle Missouri initial (AD 900?1400) anderson Dodd, Breeden, fay tolton (AD 1000?1300) grand Detour Jiggs thompson, Pretty head, Langdeau swanson swanson, Crow Creek, King Lower James Mitchell, Bloom, twelve Mile Cr., ghoering, Ethan Brandon Brandon Cambria Cambria, Price Big sioux Broken Kettle, Kimble Little sioux Phipps, Brewster, Wittrock, Chan-ya-ta great oasis great oasis, Broken Kettle W., hitchell, oldham Extended (AD 1100?1500) Clark?s Creek Clark?s Creek (AD 1100?1500) nailati Cross Ranch fort yates fire heart Creek, Paul Brave, Jake White Bull thomas Riggs thomas Riggs, Pitlick, Cheyenne River terminal (AD 1500?1675) huff huff, shermer (AD 1500?1650) Coalescent initial (AD 1400?1500) arzberger arzberger (AD 1300?1500) Campbell Creek talking Crow, Crow Creek anoka Lynch Extended (AD 1550?1675) Le Compte Potts, Molstad, Moreau River (AD 1500?1650) akaska swan Creek La Roche scalp Creek, La Roche, McClure, Robinson Bennett Black Widow B, Meyers shannon spain a, Clarkstown B Redbird Redbird ii Post-Contact (AD 1675?1862) Knife Deapolis, Rock Village, nightwalker?s Butte (AD 1650?1886) heart River Double Ditch, slant, Boley, Mandan L, smith farm Le Beau four Bear, anton Rygh, spiry-Eklo, swan Creek Bad River Dodd, Buffalo Pasture, Philips Ranch talking Crow Medicine Crow, talking Crow, oacoma, oldham felicia two teeth, Cadotte, Black Partizan Ahler Taxonomy (Cannonball, heart, Knife and garrison regions only) Complex Phase Type Sites/important Sites Charred Body (AD 900/1000?1200) ? flaming arrow, Menoken Middle Missouri (AD 1000?1450/1500) fort yates (AD 1000?1300) fire heart Creek, Paul Brave, Jake White Bull huff (AD 1300?1500) huff Clark?s Creek (AD 1200?1300) Clark?s Creek, steifel, grandmother?s Lodge nailati (AD 1300?1400) Cross Ranch heart River (AD 1450?1785) Mandan Lake (AD 1450?1525) Mandan Lake hensler (AD 1525?1600) Lower hidatsa, White Buffalo Robe Willows (AD 1600?1700) Lower hidatsa Painted Woods (AD 1300?1525) Clark?s Creek (AD 1200?1300) Clark?s Creek, steifel, grandmother?s Lodge nailati (AD 1300?1400) Cross Ranch, White Buffalo Robe, amahami scattered Village (AD 1400?1450) Big hidatsa, upper sanger, Mandan Lake Mandan Lake (AD 1450?1525) Mandan Lake, Mahhaha Knife River (AD 1600?1886) Willows (AD 1600?1700) Mahhaha, Big hidatsa Minnetaree (AD 1700?1785) Lower hidatsa, Big hidatsa, Molander Roadmaker (AD 1785?1830) Big hidatsa, sakakawea, Deapolis, ft. Clark four Bears (AD 1830?1886) ft. Clark, Deapolis, Rock Village, Like-a-fishhook * the first series of dates for the Lehmer variants are those appearing in Lehmer (1971, table 2). the second series below these are those proposed by toom (1996, table 8). Lehmer?s Disorganized Coalescent variant is included in the Post-Contact Coalescent. some phases and sites included in the Lehmer scheme have been added by later research- ers. Lehmer did not include the Cambria, Big sioux, Little sioux, or great oasis phases in his taxonomy; they were added later by other archeologists. 3 Methods of Dating a rcheologists divide dating techniques into two general classes, abso- lute and relative. absolute techniques rely on a number of procedures, such as radiocarbon dating, to determine the age of a particular sam- ple. Relative techniques, one of the most popular being seriation, or ordination, place various strata, artifacts, or groups of artifacts from a common provenience unit in an order that is interpreted to be temporal in nature. Because ceramic ordination plays a major role in refining the Plains Village chronology of the Middle Missouri subarea, the subject of this study, several theoretical and methodological issues are addressed. ceramic orDiNatioN archeologists have employed seriation, or ordination, as a relative dating or ordering technique since the nineteenth century, beginning with the work of a number of old World archeologists. Willey and sabloff (1993) discussed early applications in the new World. James a. ford, together with several of his col- leagues, applied the technique to a large number of collections from the lower Mississippi River valley (Phillips et al., 1951). one of his more influential studies described the technique in detail (ford, 1962). as a consequence of the work of ford and others, and of concurrent advances in seriation techniques by Brainerd (1951) and Robinson (1951), a large number of early chronologies were built upon ceramic seriations, including a number from the Middle Missouri subarea of the Plains. applicatiOns in the middle missOuri subarea the role of pottery in assessing the similarities between components and what these relationships entail for cultural chronologies and taxonomies has been explicitly addressed in a number of seriations and ordinations within the Middle Missouri subarea. these efforts, focusing on intrasite and intersite ceramic vari- ability, provide a context for the ceramic ordinations developed for this report. Beginning with modest efforts in the early 1960s, ceramic seriation was applied 1 8 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y with increasing frequency because of the development of various computerized and noncomputerized quantitative techniques. the first of these (smith, 1960, 1963) relies on the ford (1962) technique of visually arranging ceramic types from various components into lenticular or ?battle- ship? curves. shortly thereafter, the techniques devel- oped by Brainerd (1951), Robinson (1951), and Meighan (1959, as modified by ascher, 1959) began to be used with increasing regularity as additional sites were excavated. several of the more notable analyses in this regard are by hoffman (1963a, 1972). More recently, multivariate sta- tistical techniques, such as principal components analysis, have been used to assess the chronological relationships between large numbers of provenience units within partic- ular sites or between site components (C. Johnson, 1977a, 1977b; falk, Johnson, and Richtsmeier, in prep.; falk, steinacher, and Johnson, in prep.; falk and Johnson, in prep.). the following discussion briefly summarizes these studies for each Plains Village variant, including related sites not situated along the Missouri River in southeastern south Dakota and northwestern iowa. Initial Middle Missouri the earliest attempt to use a quantitative statistical measure to assess the ceramic relationships between pro- venience units at a site was by Lehmer (1954a:73?83), for artifacts excavated at the Dodd site (39st30), a multicom- ponent village containing two and perhaps three distinct Plains Village occupations. as noted earlier, the data ob- tained from the excavations at this site also was instru- mental in developing Lehmer?s cultural taxonomy within the subarea. Lehmer used the mean standard error and the index of association to assign particular provenience within the site to three components, one Post-Contact Coalescent component and two initial Middle Missouri components. his analysis, not a seriation in the strictest sense of the word, was the first to evaluate the nature of intrasite variations with a full range of material classes. Because of its limited applicability (Robinson, 1952), the mean standard error did not become popular, and Lehm- er?s early interest in quantitative analyses waned in his later research. a study by Ehrenhard (1971) used a factor analysis of ceramics in an attempt to determine if the King site (39LM55) represents a hybridization of the initial and Extended Middle Missouri, as represented by the som- mers and thomas Riggs sites, respectively. Like Lehmer?s Dodd site study, this was not a seriation, but it represents the first study to use factor analysis to evaluate intersite relationships within the Middle Missouri subarea. several analyses by Johnson (falk and Johnson, in prep.; falk, Johnson, and Richtsmeier, in prep.) and stein- acher (falk, steinacher, and Johnson, in prep.) used multi- variate statistical techniques to order initial variant sites along several dimensions of variability. Johnson?s analyses used an r-mode principal components procedure of five ceramic types to order nine Missouri River components along two factors based upon their factor scores. one of the extracted factors was interpreted to be temporal in na- ture, although there was an insufficient amount of inde- pendent information, in the form of site stratigraphy and radiocarbon dates, to support this interpretation. stein- acher?s study used a variety of techniques (cluster analysis, principal components analysis, multidimensional scaling) to arrange a larger set of initial Middle Missouri, great oasis, Mill Creek, and Extended Middle Missouri compo- nents. only the initial variant components located along the Missouri River, including the Mitchell site, were sub- jected to an intensive analysis. the final analysis used 18 of these components and a number of ceramic types and attributes to distill what steinacher interpreted to be tem- poral dimensions in the data, supported by radiocarbon dates. the results conflict with Johnson?s earlier findings. in his intrasite analysis of the sommers site pottery, stein- acher (1990) used the Brainerd-Robinson technique, prin- cipal components and cluster analysis, and a number of additional statistical measures to assess stylistic variability and complexity. his goal was to test four models of site settlement, with the results indicating that the village was a single large settlement that subsequently contracted into a smaller fortified area. several ceramic seriations have used initial Middle Missouri sites located off of the Missouri River along the lower James River in southeastern south Dakota and with Mill Creek components from northwestern iowa. alex (1981b) used the ascher (1959) modification of the Mei- ghan (1959) seriation technique to order units, some com- posed of very small amounts of pottery, from the Mitchell, goerhing, and 12 Mill Creek sites based upon the percent- ages of two wares. a number of seriations of Mill Creek components and/or provenience units also have been ac- complished. the first of these (flanders, 1960) used the Brainerd-Robinson approach on a number of surface and excavated collections from the Little sioux locality. flan- ders? analysis formed the basis of a number of subsequent seriations. the second effort to seriate Mill Creek compo- nents was by Vis and D. R. henning (1969), who used the ascher and Brainerd-Robinson techniques on as many as 11 components. a key to this analysis was the inclusion of two stratigraphically related zones (upper and lower) from n u m b e r 4 7 ? 1 9 each of two sites, Wittrock and Phipps. their results sup- port a division of the Little sioux sites into early and late phases. in a similar seriation, anderson (1981:104?109) used percentages of the same two pottery types (Kimball and Mitchell Modified Lip) on the same basic set of site units and on an expanded number of proveniences from the Brewster site (13CK15). the Brewster site included a series of stratigraphic units used to support the order of the seriation. anderson (1981:108) concluded that there was no evidence to support a separation of the Little sioux phase into early and late subphases. tiffany (1982:56?66) performed an r-mode principal components analysis of 14 pottery types from the Chan-ya-ta site in order to deter- mine the occupational sequence at the site. his claim of two separate components, however, cannot be evaluated with- out a re-analysis of the distributions of ceramic types within the Brewster provenience units, perhaps by an r-mode principal components analysis linked with factor scores of these units. tiffany?s intersite analysis of Little sioux locality sites was similar to the previous studies, using the ascher-Meighan approach. he concluded that his se- riation reflected a developmental, diachronic sequence as opposed to anderson?s interpretation that the seriation simply reflected variations of contemporaneous village groups. finally, D. R. henning and King (1982:113?130) performed an intrasite seriation of 10 provenience units from the Larsen site (13PM61), a great oasis/Mill Creek site from the Big sioux locality in northwestern iowa. they used the Brainerd-Robinson and Meihgan techniques and concluded that the site had a single component amalgam of great oasis and Mill Creek ceramics. Extended Middle Missouri in contrast to the initial variant of the Middle Missouri tradition, there are virtually no ceramic ordinations of Ex- tended Middle Missouri villages. Calabrese (1972:49?65) performed a discriminant analysis of various metric pot- tery attributes from three sites (fire heart Creek, Paul Brave, Cross Ranch), although his intention was not to develop a chronological ordering but to determine which attributes best discriminate between these sites and how the sites related to each other in discriminant space. it also is of interest to note that he conducted similar analyses with arrow points. Lippincott (1970) ordered a number of Extended Middle Missouri and Post-Contact Coalescent Plains Village components from the upper Knife-heart re- gion by using the ford technique on attributes instead of types. C. Johnson (1999) presented a detrended correspon- dence analysis of 13 Extended Middle Missouri tradition components from north and south Dakota based upon generalized pottery types. two dimensions were extracted from the data, one relating to spatial variation between the sites. the lack of significant amounts of ceramic varia- tion within the Extended variant and radiocarbon dates prevents any conclusive statements regarding the relation- ships between chronology and ceramic relationships be- tween the components. Initial Coalescent the earliest attempt to seriate initial Coalescent pro- venience units was by smith (1960, 1977:141?149) from his work at the talking Crow site. he used pottery types and body sherd surface treatment to order up to 21 units from this partially mixed multicomponent initial and Post- Contact Coalescent site. steinacher (1983) performed an intersite seriation of five initial Coalescent components, including the Whistling Elk site (39hu242), using the Brainerd-Robinson procedure on types and attributes. the ceramic assemblages from the ordered sites were compared with five Central Plains tradition sites from nebraska and Kansas, and one initial Middle Missouri site (Langdeau) based upon attributes. steinacher concluded that the earli- est initial Coalescent site, Whistling Elk, was very similar to the Central Plains tradition sites and that this affinity decreased through time within the initial Coalescent. the results of this analysis contradicted Lehmer?s (1971:111) notion that the initial Coalescent, in its earliest manifes- tations, represented an admixture or blending of Central Plains and Middle Missouri cultural traits. Extended Coalescent smith (1960, 1963) and hoffman (1963a) conducted the first seriations of Extended Coalescent components, both relying on the ford technique to arrange their units. smith?s seriation included three Extended variant compo- nents and three early post-contact (felicia phase) sites in the Big Bend Region. the results support the assignment of the sites to their respective taxonomic units. hoffman?s tempo- ral ordering of the Chouteau aspect (Extended Coalescent) was more ambitious; 11 Extended and seven Post-Contact Coalescent components, from almost the entire length of the Missouri River in south Dakota, were the basis for the analysis. he discovered that these components, although varying along a single temporal continuum, were divided into three temporal units. hoffman (1967:59?61, 78?79; 1968:65?66, 73?75) pursued his early interests in seria- tion as a way of assessing the ceramic similarities between 2 0 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y sites in his reports on the La Roche and Molstad Village sites. Both efforts compute Brainerd-Robinson coefficients between the two reported sites and from 8 to 19 addi- tional components based upon traditionally defined pot- tery types. his report of the La Roche sites also included an intrasite seriation of four separate and pooled prove- nience units in order to determine the number of compo- nents present at the two adjacent La Roche sites. from his seriation of Molstad Village, hoffman concluded that the results support previously defined taxonomic units, lead- ing him to postulate a Le Compte focus consisting of the Molstad, no heart Creek, and possibly Potts Village sites. the seriation in the La Roche site report also was used to assign components to taxonomic units and to assist in the dating of these villages. Brown (1967:153?162) also took a similar approach in determining the relationship between component B at the Chapelle Creek site and eight Extended Coalescent and felicia phase villages. More recently, C. Johnson (1977a:22?24; 1977b) advocated arranging components in multidimensional space through the use of principal components analysis. this approach recognizes the possibility that a number of underlying dimensions in a data set might be isolated, instead of traditional seriation techniques that attempt to reduce all ceramic variability down to a single, temporal dimension. Johnson?s approach was to use groups of pottery types from sites located along the entire length of the Missouri River in south Dakota. More than 32 Extended Coalescent and felicia phase components were included in these analyses (C. Johnson, 1984a). the results isolated a single major source of varia- tion, interpreted to be temporal in nature, although the corroborating information (i.e., radiocarbon dates) did not unequivocally support this conclusion. Johnson (1988) also arranged a number of Extended Coalescent and felicia phase components based upon a plot of the percentages of three ceramic types. all of Johnson?s findings are very similar to those of past investigators, such as smith and hoffman. Post-Contact Coalescent no other variant or taxonomic unit defined for the Middle Missouri subarea has received as much attention from archeologists seeking to order and date villages as has the Post-Contact or protohistoric period. there is a long list of researchers involved in a number of attempts to arrange or group sites and provenience units within sites, using a variety of techniques. the earliest of these seriations includes those by Lehmer (1954a:80?83), smith (1960), Krause (1967:182?192), and Pollnac and Pollnac (1969). the Lehmer and smith analyses are discussed above. Krause seriated 12 post-contact villages, includ- ing Leavenworth, using spaulding?s (1960:81?82) index of ceramic likeness, which is almost identical to the one developed by Robinson (1951). Krause presented a partial matrix of these values for the sites, although he appar- ently did not attempt any formal arrangement of the full matrix. it is unclear how these indices were computed, but presumably a number of ceramic modes were used in the computation (Krause, 1967:244?252). he also assigned absolute dates to these components. the Pollnac and Poll- nac study was based upon a q-mode factor analysis of the data presented in Krause (1967:244?252). it is a study of the variation of these modes in time and space rather than a seriation of sites, because they use the same village dates as Krause. hoffman (1970b:290?311, 1972) relied on the Brainerd-Robinson and ascher modification of the Meighan techniques to order eight Bad River phase vil- lages, based upon the percentages of stanley ware types from the villages. hoffman used this seriation to question Lehmer and Jones? (1968:97?100) division of the Bad River phase into two temporal subphases. in his report of the Chapelle Creek, Brown (1967:147?153) calculated indices of likeness (spaulding, 1960:81?82) between the site and six additional Bad River phase components based upon four ceramic types. from this analysis, he developed a relative order of the villages, although it is unclear how this sequence relates to his seriation. D. M. healan per- formed a q-mode factor analysis of five lodges and a pit from the Biesterfeldt site on the basis of seven ceramic attributes (Wood, 1971). he concluded that the pottery from the site exhibits a high degree of homogeneity from one provenience unit to the next. Lippincott (1970) or- dered a number of Extended Middle Missouri and Post- Contact Coalescent Plains Village components from the upper Knife-heart region by using the ford technique on attributes instead of types. in his report on the McClure site, Johnston (1982:42?50) used a computerized seriation program (Craytor and Johnson, 1968) to arrange 16 late Extended and Post-Contact Coalescent components on the basis of ceramic type percentages. from this chronology, he then reconstructed the changing settlement patterns in the Big Bend region during late Coalescent times. C. Johnson (1977a, 1977b), C. Johnson et al. (1995), falk and Johnson (in prep.), and falk, Johnson, and Rich- tsmeier (in prep.), in a series of intrasite and intersite ce- ramic analyses, performed r-mode principal components. they used both traditionally defined pottery types (inter- site) and attributes (intrasite) to arrange components or provenience unit factor scores in multidimensional space. n u m b e r 4 7 ? 2 1 all of these studies isolate what is interpreted to be tem- poral dimensions in the data, partially supported by site stratigraphy, radiocarbon dates, and historic documenta- tion. in a recent review, toom (1994:483) criticized pre- vious ceramic seriations by Johnson because they lacked corroborative temporal evidence and were too broadly based. as the analyses presented in chapter 6 demonstrate, there is supporting evidence for interpreting some kinds of ceramic variation along temporal lines, recognizing that there are other types of variability that can complicate this general and somewhat simplistic perspective. the results of these analyses indicate that various levels of confidence can be assigned to each ceramic ordination, depending on the amount of independent chronological controls. a somewhat different approach to dating was proposed by grange (1981), who used the mean ceramic date ap- proach initially developed for historic European ceramics. his analysis was an outgrowth of his interests in develop- ing a Pawnee-Lower Loup temporal sequence through ce- ramic formula dating (grange, 1968, 1984). he calculated a series of mean dates for a large number of Coalescent components, focusing on the Post-Contact period, but also included a smaller set of initial and Extended variant vil- lages. grange then proposed a series of conclusions about the origin of particular ceramic wares and types, and the settlement of the Missouri River valley in south Dakota through time. the last major attempt to order Post-Contact period villages was by Ramenofsky (1987, appendix C). as a way of establishing temporal control in her study of the spread of contagious Euro-american derived diseases among na- tive americans, she seriated sites from a full range pre- historic and protohistoric contexts, including Middle Missouri and early Coalescent villages. the ceramic data used in these seriations was based upon the percentages of three classes of attributes (rim shape, extra clay to rim, rim decoration). the particular technique used in these se- riations is not made explicit, although it appears that the ford (1962) technique was used (Ramenofsky, 1987, fig. 22). additional supporting information used to develop her five period chronology (ad 1030?1359, 1416?1649, 1649?1780, 1780?1837, 1845?1885) included radiocar- bon dates, seriation of metal occurrences, and cumulative frequencies of trade glass. Knife Region and Related Areas Principal components analysis and other analytical techniques are used to arrange components within the Knife River indian Villages national historic site (KnRi) based upon both ceramic attributes and types. the first of these evaluates the relationships of the anomalous Elbee site, a component with close ties to Extended Coalescent develop- ments in south Dakota (ahler, 1984a). using a series of de- scriptive ceramic categories (types), ahler concluded that Elbee is most similar, in terms of Euclidean distances, to the Mondrian tree in north Dakota, Demery, and swan Creek a components from north-central south Dakota. an r-mode image factor analysis of the percentages of the pot- tery types from 17 components yielded a factor interpreted to be temporal in nature. on this dimension, Elbee is simi- lar to the Payne and Demery sites, both Extended Coales- cent villages in north-central south Dakota. in his analysis of the ceramics from taylor Bluff, ahler (1988) conducted two cluster analyses and one r-mode principal components analysis of 11 components, includ- ing greenshield and fort Clark, based upon 11 ethnic and temporally sensitive variables and attributes. Both green- shield and fort Clark were occupied at various times by the Mandan and arikara. Leavenworth, a historically documented arikara village in the Cannonball region of south Dakota, was used for comparative purposes but not formally included in the multivariate analyses. this analy- sis, for the first time within the Middle Missouri subarea, attempts to systematically evaluate the ethnic identity of the occupants of a particular village (i.e., taylor Bluff). he concluded that it was most likely occupied by the hidatsa (awatixa and/or hidatsa proper subgroups) and not by any other groups, including the arikara. ahler?s analysis of the taylor Bluff pottery represented a concern with as- signing components to the various hidatsa and Mandan ethnic groups in the Knife region, an interest he peruses in the ceramic synthesis of the KnRi project. in his analysis of the Mondrian tree site ceramic as- semblage, C. Johnson (1983) performed a principal co- ordinate analysis or multidimensional scaling of nine components both within and outside of the Knife River indian Villages national historic site on the basis of ce- ramic types. a cluster analysis also was performed on this data. Results indicated that Mondrian tree was most similar to scattered Village Complex components within KnRi, and that two dimensions are tentatively defined as temporal in nature. in their synthesis of ceramic variation within the upper Knife-heart region, ahler and swenson (1993) used 42 metric and nonmetric variables and attributes of various artifact classes in an r-mode principal components analysis and several cluster analyses of 78 ceramic batches or com- ponents. the principal components analysis isolated three temporally sensitive factors and the variables making up 2 2 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y these factors. Plotted factor scores of the batches agree with the placement of the components when site stratigraphy, historic documentation, and a large series of radiocarbon and thermoluminescent dates are considered. Cluster analy- sis was used to group the batches into eight temporal periods and subperiods, forming the basis of a culture taxonomy (ahler, 1993b; ahler and swenson, 1993). a cluster analy- sis of 43 late period (heart River and Knife River phase) batches was able to discern tribal and subtribal ceramic differences. these differences are helpful in assigning the batches to ethnic groups. ahler and swenson defined tem- poral trends in ceramic variation within KnRi, including differences among contemporaneous batches attributable to ethnic groups. their results have four broad-ranging im- plications for ceramic seriation within the Middle Missouri subarea and beyond: (1) ceramic types mask important tem- poral and ethnic group variation that are more adequately explored by a consideration of variables and attributes; (2) ceramic variation because of differences between con- temporaneous villages occupied by different ethnic groups can lead to incorrect temporal assignments without sup- porting chronometric dating techniques; (3) multivariate statistical techniques, such as principle components analy- sis, is able to isolate varying sets of temporally sensitive ce- ramic variables and attributes depending on the time period involved, assuming adequate independent temporal mea- sures are available; and (4) there is a close correspondence between native oral tradition history and culture-history derived from the archeological record. any attempts to reconstruct late prehistoric culture-history without a firm basis in oral traditions may be severely compromised, de- pending on the nature, specificity, and depth of these tradi- tions. the impact these observations have on the current project is dealt with in greater detail in chapter 6. methOdOlOgical cOnsideratiOns seriation is defined as ?the placing of items in a se- ries so that the position of each best reflects the degree of similarity between that item and all other items in the data set? (L. Johnson, 1972:310). similar definitions have ap- peared in spaulding (1971:10), Cowgill (1972:381), and Marquardt (1978:258). ordination, on the other hand, may be viewed as a more general approach in which the units of interests are positioned in two or more dimen- sions rather than along a single scale (see also sneath and sokal, 1973:245). the ?items? in the current study are site components, and the ?degree of similarity? is measured between all components on the basis of ceramic descrip- tive category or ?type? frequencies, making it a frequency (similarity) seriation rather than abundance (presence/ absence) seriation (Rowe, 1961:326; Dunnell, 1970:308). this analysis relies on detrended correspondence analysis to order the components in multivariate space represented by a reduced number of dimensions from a larger number of ceramic types. Because these analyses are not necessarily limited to a single dimension or ?series? as are more tra- ditional seriation techniques, the depiction of components in relation to each other is best described as an ordination. interpretation of one or more of these dimensions along temporal grounds is called a seriation, however. Dunnell (1970:305?306) pointed out that in the past, seriation was treated as a technique, when in reality it should be viewed as a method with its own series of underlying assumptions and corollaries that are organized for the solution of a par- ticular problem (i.e., chronology building). these assump- tions are the focus of the following discussion. in their influential analysis of ceramic surface collec- tions from the lower Mississippi River valley, Phillips, ford, and griffin (1951:219?223) presented seven as- sumptions necessary for a seriation to reflect chronology: (1) geographically stable population; (2) short period of site occupation; (3) gradual change in pottery over time and through space; (4) defined ceramic types are sensi- tive at measuring spatial and temporal variation; (5) types exhibit unimodal, lenticular or battleship curves through time if assumptions 1 and 3 are true; (6) if assumption 5 is true, the unimodal distributions of the pottery types at any particular time forms a unique pattern; and (7) the combined site collections, consisting of random samples of at least 50 sherds/site, is representative of the entire range of temporal variability within the study area so that there are no gaps in the seriation. Many of these assump- tions were subsequently discussed by ford (1962:42?43), Dunnell (1970), and Mcnutt (1973:46). to this list Dun- nell (1970:311, 318) added several additional conditions for seriation to reflect a chronological ordering: (1) all groups (e.g., components) must be of comparable dura- tion; (2) components must belong to the same cultural tradition; (3) components must come from the same local- ity; and (4) deviations from unimodal type distributions or the model of continuous artifact changes are due only to sampling errors. Deetz and Dethlefsen (1965:196?196, 206) also pointed out that seriation assumes equal rates of ceramic change, meaning that any type employed origi- nates and spreads from a single locus and that sites farther away from the locus exhibit frequency occurrences later in time compared with those near the center of origina- tion (i.e., Doppler effect). this time lag in the popularity of a decorative type is used in chapter 6 to correlate the n u m b e r 4 7 ? 2 3 northern Extended Coalescent temporal sequence with the hensler phase. several archaeologists think that chronologies are more discriminating if the artifact classes, types, or attributes have short temporal spans or change relatively rapidly (Meighan, 1959:210; ford, 1962:43). a number of researchers note problems in their own results, commonly attributed to vio- lations in the above assumptions. Most problems seem to arise when dimensions other than time are represented in the data (Brainerd, 1951:305; Cowgill, 1968:374; Renfrew and sterud, 1969:271; gelfand, 1971a:271?273, 1971b:199; Kruskal, 1971; L. Johnson, 1972:324; Johnson and John- son, 1975:283; LeBlanc, 1975:33). the ways this study adheres to or violates these as- sumptions is presented in the following discussion. the as- sumption of a geographically stable population is related to the condition of restricting a seriation to a confined spatial unit, such as the locality. this is to prevent the problem of the confounding temporal effects of diffusing styles through space (i.e., the Doppler effect). that is, as a particular temporally sensitive trait diffuses, the period when it reaches its height of popularity is generally later the farther a particular site is from the point of trait origin. this principal was demonstrated by Deetz and Dethlefsen (1965) in their study of the spread of historic grave head- stone styles in the northeastern united states. in addition, if there are population movements into and out of a local- ity or region, the pottery of each village group presumably reflects the development of their ceramic tradition from their point of origin or homeland. this study attempts to partially control the confounding effects of time and space by limiting, to the extent possible, each ordination to a region or a phase. this approach, along with the use of correspondence analysis, has the potential of beginning to unravel multiple sources of variation in the archeologi- cal record. this partially ameliorates a recent criticism by toom (1994:483) of this methodology. Because of the small number of components in some regions, the stability of potting traditions across regions, and the desire to cor- relate regional chronologies, some chronologies combine components from different regions. finally, geographic stability (i.e., movements) of the Plains villagers must be assumed to be equal through time and space because it is very difficult to control for this factor with the avail- able information. Certainly within the late protohistoric and early historic period, there are numerous cases of vil- lage movements (Wedel, 1955:77?84; ahler and swenson, 1993:135?139). it is suspected that migrations occurred prehistorically, sometimes across long distances (toom, 1992b). it also is assumed that all components have rather short occupations and comparable time depth. Even though this may not be true in some cases, a number of techniques are used to satisfy this condition. De Barros (1982) addressed the ways that variable site-duration affects frequency seria- tion. his simulation study concluded that errors in seriation ordering, using pottery types, occur under the following conditions: (1) varying rates of ceramic change; (2) tran- sition periods, especially short ones, during the rise and decline of a type?s production; (3) increasing rates of type replacement; (4) lower intervals of time between site oc- cupation dates; and (5) short pottery-type spans in relation to the duration of site occupation. More specifically, if ce- ramic types are replaced rather slowly (e.g., every 75?100 years) and have relatively long life spans (e.g., 180 years) compared with site occupation spans (e.g., 10?60 years), few seriation errors occur. on the other hand, long-lived types having slow replacement rates result in seriations with low temporal resolutions (cf. Baxter, 1994:122). in addition, considerable overlap in site occupation and in- creased numbers of contemporary types does not result in more seriation errors, although the detection of these er- rors is sometimes difficult to verify. one solution that de Barros (1982:310?313) suggested, namely to seriate short- term provenience units of comparable duration (e.g., pits, houses), using temporally sensitive ceramic attributes with multivariate techniques (e.g., factor analysis), is beyond the scope of this study. the studies by C. Johnson (1977b) and steinacher (1990) tried to minimize temporal depth by conducting intrasite analyses using discrete provenience units. the present study, by its nature and scope, must rely on traditionally defined types from a large number of sites, so the solution that de Barros recommended awaits a more exhaustive and comprehensive analysis of ceramic variabil- ity among a smaller number of Middle Missouri Plains Vil- lage sites. in addition, many sites do not have large enough samples from short-term contexts, such as pits or house floors, to perform these analyses. this problem is one of mixed materials from different time periods. one solution proposed a technique to detect mixed samples (Kohler and Blinman, 1987). this approach is thought to be inappli- cable to the present study because it relies on establishing calibrated (i.e., dated) ceramic data sets for various time periods, apparently relying on ?unmixed? units for estab- lishing characteristics for each of them. in the present study, several of the above seriation er- rors are minimized at sites with long periods of discon- tinuous occupations by dividing them into their respective components. in cases in which there is a lengthy, continu- ous occupation by peoples of the same cultural tradition 2 4 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y or variant, subdivisions into shorter temporal segments are made. this is accomplished by two procedures: (1) limiting each seriation to the same tradition, variant and or phase, thereby fulfilling Dunnell?s second condition; and (2) divid- ing sites with long and continuous occupational histories into shorter temporal units, as in the case with some Le Beau phase villages, such as Larson (39WW2), spiry-Eklo (39WW3), anton Rygh (39Ca4), sully (39sL4), and swan Creek (39WW7). on occasion, this procedure is not pos- sible because of inherent limitations in the nature of some site assemblages (e.g., Mobridge site). these examples are discussed in conjunction with each seriation. With these exceptions, all components are treated as single units with no further temporal divisions. this approach is deemed necessary because of low sample sizes within individual site provenience units, lack of uniform and meaningful provenience units, and the desire to include as many com- ponents as possible within the funding and time constraints of the study. in essence, most of the village ceramic as- semblages are treated as ?surface? collections; hence, the concern with the seriation assumptions of Phillips, ford, and griffin (1951:219?223). additional variables contrib- uting to potential seriation errors, such as the longevity of pottery types, varying rates of ceramic change and type replacement, and variation in the interval between village occupation, are largely uncontrolled in the current study. instances in which these factors are particular problems are discussed so that ordination results can be evaluated. the assumption that ceramic change is gradual over time and space was demonstrated in many archaeologi- cal contexts and in several ethnoarcheological studies (graves, 1985:32?33). graves also pointed out that rapid population loss and other factors might result in the dis- ruption of traditional inter-cohort transmission of design information, resulting in dramatic ceramic change. Plog (1980:108?111) indicated that ceramic change in the amer- ican southwest can occur rather quickly, perhaps within a period of about 75 years or less or can proceed at low rates, such as in the Valley of oaxaca. Cleland (1972:209) also suggested that objects with short life-spans, such as pottery, change rapidly over time because of stylistic drift. Changes in the intensities of village interaction and the rates of ceramic change can vary through time, causing additional problems in interpreting a seriation or ordina- tion along temporal grounds. Controlling for the effects vessel size, function, and form among contemporaneous prehistoric communities would help to determine the in- fluence of these variables on style (see graves, 1985:31). Ceramic variability within the Middle Missouri subarea is characterized by conservative potting traditions broken by a number of relatively rapid temporal changes or non- continuous clinal variations through space. these changes generally occur at divisions between various taxonomic units at the phase or variant level, the typical subject of individual ordinations. the requirement that the defined ceramic types, de- scriptive categories, or attributes be sensitive to temporal and spatial variation is self-explanatory. types or attri- butes that were diagnostic space-time indicators were des- ignated as ?historical? by Rouse (1960:317). those that are not were referred to as ?descriptive? types or attri- butes. a summary of these two views of types appeared in Dunnell (1986:167?176). Marquardt (1978, 1979) also emphasized the search for temporally sensitive attributes before the seriation process begins. another area of con- cern during the artifact classification process involves the search for types or attributes (i.e., mental templates) that conform to those of the makers of the artifacts. Dunnell (1986:176?182) summarized these efforts in terms of etic and emic approaches. further consideration of these dis- tributions is beyond the scope of this study other than to say that, except at the initial and most inclusive stages of classification into wares, ceramic typologies within the Middle Missouri have focused on descriptive types, with a few possible exceptions (e.g., smith, 1977:51?52). several researchers have commented that chronologies are more discriminating if the artifact classes, types, or attributes have short temporal spans or change relatively rapidly (ford, 1962:43; Meighan, 1959:210). LeBlanc (1975:29) also indicated that reliance on too few pottery types does not result in the discrimination of fine time increments. in the present study, it is impossible to determine which ar- tifact types or attributes are temporally sensitive before a formal analysis is completed. in addition, the time span of ceramic types or their rates of change in the Middle Mis- souri remains largely unknown because of an inadequate number of absolute dates. only a massive dating program, much larger in scope than the present study, can begin to address these questions. these issues are discussed in greater detail in chapter 8. there are no all-inclusive tests to determine if the es- tablished ceramic types or descriptive categories from the Middle Missouri subarea are sensitive to the spatial or tem- poral dimensions. Many of the types, or combined types (i.e., descriptive categories) do vary significantly through space, as measured by their relative frequencies from re- gion to region. there also are major differences in their oc- currence from variant to variant, and less so from phase to phase. What has not been accomplished is to demonstrate that these types or categories consistently vary through n u m b e r 4 7 ? 2 5 time within phases or variants so as to provide a means of making fine-scale temporal discriminations between indi- vidual components. some studies, however, have isolated temporally sensitive attributes that relate to traditionally defined types (C. Johnson, 1977b:43?44; ahler and swen- son, 1993:93?135; Johnson and toom, 1995:263). in a sense, this study not only seeks to order components along a time scale, but it also explores the usefulness of types in making fine-scale chronological discriminations. there also is some evidence to indicate that types, whether from the Middle Missouri subarea (steinacher, 1990) or other areas (hegmon, 1992) reflect social interaction, historical context, symbolic, or emblematic expression. the work within the Knife region is an exception to this generaliza- tion, providing fairly precise measures of ceramic change on the attribute level while controlling for the effects of ethnic group variation (ahler and swenson, 1993). several intrasite ordinations using attributes demonstrate that not only is there quantifiable ceramic change through time, but also that attributes play significant roles in the formu- lation of traditional types (C. Johnson et al., 1995:239? 240). the results of the present study begins to evaluate more fully the ability of these established types, or their descriptive category reformulations, to help to make finer temporal divisions than previous seriations. there is no assurance in the current study that the as- sumption of a unimodal type distribution through time is adhered to. the underlying premise that components can be ordered into a chronological sequence depends upon this assumption, although various internal checks are used to determine if this is indeed the case. techniques for test- ing this assumption and for providing independent relative and absolute dating controls on each seriation include site stratigraphy and radiocarbon dating. stratigraphy is dis- cussed by a number of archaeologists as an independent technique for verifying seriations (Brainerd, 1951:305; Vis and D.R. henning, 1969:255?257; LeBlanc, 1975:32). By including stratigraphically related components from the same site containing a long and continuous occupation, a test of the temporal ordering of components and the na- ture (e.g., lenticularity) of ceramic change is accomplished. adding radiocarbon dates provides an absolute time scale to each seriation. Deetz (1968) also presented an example in which the assumption of unimodality is violated. Most of the components used in this study have ce- ramic assemblages consisting of at least 50 rim sherds to ameliorate misrepresentative, unreliable, and highly er- ratic type frequencies inherent in small sample sizes. the use of attributes instead of types was demonstrated by LeBlanc (1975:26) to increase sample size when missing data present a problem but, as is outlined above, collecting this data is beyond the scope of the current study. Cowgill (1968:372?373) pointed out that correlation coefficients are dependent upon sample size, with smaller samples yielding lower correlations. hatch et al. (1982) had sug- gested that only sample sizes in the thousands begin to reduce error or noise in frequency seriations. several com- ponents that contain small sample sizes in the present analysis are included because they have a pre-existing suite of radiocarbon dates or are considered to be important to the overall interpretation of the study. in general, the smaller the sample size in relation to the population of rim sherds from a village, the less likely the sample reflects the actual variability in the entire site. these cases are pointed out during the analysis portion of this report. finally, Dunnell?s condition that all components rep- resent occupations of equal duration is related to the life spans of ceramic vessels. there are a large number of ethno- archeological studies indicating that different sized pots are broken and enter the archeological record at different rates (David, 1972:141; David and hennig, 1972:18?19; DeBoer, 1974; DeBoer and Lathrap, 1979; Longacre, 1985, table 13.1, 1991:7; arnold, 1988, table 3). Breakage rates depend on how the pots are used. those that are used and transported more often tend to break more frequently than those that are not. in general, the largest pots are broken less frequently and therefore have longer life expectancies compared with the smaller ones, on the order of 2.5?5 times (0.2?13 years). C. Johnson (1983, table 9.14) and Wilson (1977) discussed the ways pots were used among the Mandan, hidatsa, and arikara. if different size pots are correlated with variations in decorative style or any other variables, then the qualities of smaller pots are overrepre- sented in the archaeological record compared with their relative abundance within a household or village at a given point in time. the longer a site is occupied, the greater the disparity between the two figures, assuming of course that the greater rim circumference of larger pots when they are broken does not outweigh their lower breakage rates when rim sherds rather than vessel counts are used (see ford, 1962:38). several examples from the Middle Missouri subarea support the idea of a relationship between length of oc- cupation and material content. undecorated wares, which appear to be used most often for physical utilitarian func- tions, tend to break at higher rates than decorated wares (steinacher, 1990:226?231). another example is demon- strated at the Whistling Elk site (39hu242). this village was probably occupied for a very short time as reflected in the presence of a large number of chipped stone tool 2 6 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y preforms, unused, and fully functional stone tools made from nonlocal resources (toom, 1983a). this pattern dra- matically contrasts to almost every other village within the subarea. another site that could represent a short occupa- tion is the fay tolton site (39st11), where a number of human skeletons and complete pots were recovered from several houses (Wood, 1976). an unusually light density of occupational debris characterizes the site. there is virtually no way to control for the effects of occupational duration on the ceramic assemblages of Middle Missouri villages given the existing information. as noted above, villages with comparatively long occu- pational histories are divided into a number of shorter temporal units, although no systematic efforts to study this problem have been undertaken beyond that of stein- acher (1990); however, the relationship between the rates at which vessels are broken and discarded and style can be indirectly examined. if it is assumed that vessel size has something to do with the way pots are used, then this characteristic (as measured by orifice diameter) is easily compared with the ceramic types used in this study. this relationship is examined in chapter 4. theOretical cOnsideratiOns in order to place the results of the ceramic ordinations in chapter 6 of this study into their broader archeologi- cal context, it is useful to review some of the important concepts that form the theoretical basis of this study. nor- mative and social interaction theory provide the most use- ful frameworks through which to view the results of this study. normative theory is the oldest of the perspectives and is the basis of much of the artifact and assemblage-pattern recognition studies in archaeology so crucial in inferring social boundaries used to reconstruct culture history (stein- acher, 1990:99?100). normative theory, as summarized by Binford (1965:203?204), is concerned with the ide- ational basis of human life (i.e., the normative concepts in people?s minds that structure the way they behave). for- mal or stylistic variation in material culture is an expres- sion of these normative ideas. the cohesive set of shared ideas that make up culture are transmitted through learn- ing between the generations or by diffusion between non- breeding social units, with an inverse relationship between transmission and the social distance between the groups. spatial discontinuities in this pattern are the result of mi- gration and natural or psychological barriers. Cultural in- novation originates at ?culture centers? and spreads out, blending with surrounding cultures. social interaction theory, which relies on many of the same assumptions as normative theory, maintains that aspects of style are communicated from individual to in- dividual based upon their intensity of interaction (J. n. hill, 1985:364). individuals or groups that interact more frequently share more aspects of style than those who do not. stylistic elements diffuse by exposure; there are no im- plications that these elements are adaptive for the groups who use or adopt them. styles remain stable or change at certain rates unless there are accidental or random errors in communication, or other factors disrupt the transmis- sion process. in this sense, it is similar to normative theory in its emphasis on the transmittal of information through learning, borrowing, and diffusion. unlike the normative perspective, social interaction theory elaborates on the mechanisms by which style is transmitted, leading many early proponents to reconstruct social structure from ce- ramic variation (Deetz, 1965; Whallon, 1968; J. n. hill, 1970; Longacre, 1970). some of the later analyses involv- ing this perspective (Plog, 1976, 1980) have related simi- larities and differences in ceramic style to various levels of social integration through time. a key concept involved in both theories is ?social distance.? in summarizing the normative theory of cul- ture, Binford (1965:204) stated that culture is transmit- ted between social units through diffusion and across generations by learning that is inversely proportional to the degree of social distance between them. (a study by Roberts et al. (1995), of 47 artifact types among 31 new guinea villages indicated that distance and language each account for slightly more than 25% of the variability in the artifacts, whereas combining the two accounts for an additional 10%.) the relationships between various mea- sures of social distance (e.g., ceramic similarity) between villages and their locations in space might point to discon- tinuities because the interaction between contemporane- ous communities is assumed to be inversely related to the physical distances between them. a lack of correspondence between physical and social distance, as measured by ce- ramics, might be interpreted to be the result of migration, varying degrees of intervillage social interaction, or some other phenomenon, assuming time is not a factor. Popula- tion movements into, out of, or within the Middle Mis- souri subarea during the prehistoric, protohistoric, and historic periods were discussed by Wedel (1955:77?84), stewart (1974), Wood and Downer (1977), ahler et al. (1991), toom (1992b), and ahler (1993b). in terms of the present study, social distance between villages is measured by the degree of similarity between their rim sherd assemblages, as reflected in ceramic types. n u m b e r 4 7 ? 2 7 assemblages that are most similar are thought of as being temporally and/or socially proximate, assuming a rela- tively uniform degree of social interaction between villages through time and space (see Plog, 1976, 1983:126, for a discussion of factors relating to unequal rates of interac- tion). Close similarities may be caused by the sharing of a potting tradition by distinct and contemporaneous or nearly contemporaneous village groups, or by direct an- cestral relationships between them. in this sense, a primary goal of this study of Plains Village chronological relation- ships is the reconstruction of culture-history, one of the three goals (along with reconstruction of past lifeways and the study of culture process) of archeology (Binford, 1968:8?16). these chronologies are taken to the next level of abstrac- tion by interpreting the ceramic relationships between vil- lages in terms of ethnic origins, migrations, and interactions between peoples (see Binford, 1962), or what hegmon (1992:518) called time-space systematics. the investigation of these similarities characterizes culture-historical archeol- ogy (Conkey, 1990:5). as hegmon noted, this approach dominated traditional archeology and the early years of the new archeology. Because style is treated as a passive phenomena and an integral component of material culture, a close analytical link between the objects of study or pat- terns of formal variation and subject (style) is maintained (hegmon, 1992:518). other theoretical perspectives, such as information exchange (Wobst, 1977), take a more active ap- proach to style, relying on complex bridging arguments relat- ing subject to object that are much more difficult to establish (hegmon, 1992:518?519). this also includes the evolution- ary approach of Dunnell (1978) and ecological theory (J. n. hill, 1985; o?Brien and holland, 1990; neff, 1992). another important topic to consider is the framework in which ceramic variation is viewed. formal or morpho- logical variation in pottery or any other material class is viewed as consisting of the components of style and func- tion. the relationship between these two dimensions is a major point of debate in archeology, resulting in a num- ber of theoretical formulations. Early discussions make a relatively sharp division between the two, confining func- tion to the utilitarian use of tools, such as ceramic ves- sels to cooking, storage, or transportation, and confining style to a residual category. Recent perspectives consider the overlap between the two concepts, preferring to view style as performing a broad array of specific social func- tions within the context of production and use of ceramic vessels. this discussion is of particular interest to ceram- icists who perform ordinations because of the desire to determine whether the variation between ordered units is the result of some relatively broad-scale stylistic changes over time or a relationship to the way pots were used in the cooking, storage, or transportation of their contents. a relationship between utilitarian function and purely stylistic variation might result in spurious or inaccurate chronologies if these factors are not controlled or at least considered. these relationships have not been systemati- cally investigated in the Middle Missouri subarea and are beyond the scope of this study. Binford (1962) made one of the first attempts by an archaeologist to deal with the ways material items function within the total cultural system. in a classic discussion, he defined three kinds of artifacts that differ in the ways they were used: (1) technomic artifacts, which have their pri- mary use in dealing directly with the physical environment; (2) sociotechnic artifacts, which serve primarily within the social subsystem of culture to articulate individuals to one another into cohesive groups; and (3) ideotechnic artifacts, which function primarily within the ideological subsystem to provide ideological rationalizations for the social sys- tem and a symbolic milieau for the enculturation of its participants. formal or stylistic characteristics, defined as qualities that cannot be directly explained in terms of the physical aspects of artifacts or variation in the technologi- cal or social subsystems, are labeled as stylistic. stylistic attributes are best studied when questions of ethnic origin, migration, or interaction between groups are considered. in a later article Binford (1965) made a distinction, when discussing ceramic variation, between primary functional variation (specific use) and secondary functional variation (by-product of the social context in which the vessel was made). secondary functional variability arises from the traditional ways of doing things within a family or larger social unit, or may be an expression of between-group solidarity. sackett (1977) viewed function as operating in all areas of the cultural system: technological, societal, and ideational. artifact function has an active (technological) role when dealing with the environment, whereas style operates in a non-utilitarian or passive sense, reflecting a particular culture-historical context within the societal or ideational spheres of culture. similar to Binford?s concept of secondary functional variation, sackett (1977, 1990:33) introduced the term ?adjunct? form to account for vari- ability in artifacts that are included within the cultural subsystem opposite of which the object finds its major function. in the archeological record, adjunct form is most widely manifested in decoration. this idea was reinforced with the definition of style by Davis (1983:55) as formal similarities not related to the physical or mechanical ef- fectiveness of artifacts. Rye (1981:3) and Braun (1983:13, 2 8 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y 1991:363) considered decoration to be a non-essential characteristic of pots that was not necessary for them to be used as tools, although the physical characteristics of pottery put constraints on the ways in which it was deco- rated. Ceramic variation in this ?secondary,? ?adjunct,? ?non-essential,? or stylistic realm is one of the key con- cepts underlying the present study. a somewhat simplistic but crucial assumption is made herein that Plains Village life throughout all periods and places in the Middle Missouri subarea was relatively uni- form in basic subsistence pursuits (see Wood, 1962, 1974), including the ways ceramic vessels were used and the role that style played. from this perspective, ceramic style is measured from rim sherds that vary in terms of form or shape and decoration. With the exception of Deetz?s (1965) work, efforts at defining and interpreting stylistic ceramic variability within the Middle Missouri subarea have not been very ambitious. aside from the intersite ceramic seriations dis- cussed above, several studies examine intrasite variation at several Plains Village communities. steinacher (1990) explored ceramic variation at the initial Middle Missouri sommers site (39st56) in an attempt to evaluate various occupation scenarios at the village. C. Johnson (1977a, 1977b) and Johnson et al. (1995) reexamined the Medi- cine Crow site (39Bf2) that formed the basis of Deetz?s work. their goals were much less ambitious than Deetz?s, focusing on arranging the various provenience units at the site into a chronological sequence. although a temporal sequence was established, uncontrolled variation because of the occupation of the site by two or more distinct ari- kara village groups was acknowledged to be a problem (C. Johnson, 1977b:47?48). it was not until sometime later (ahler, 1992; toom, 1995) that additional tempo- rally sensitive data was used to question Johnson?s con- struct of village occupation. these interpretations of the data point to an occupation of the two main site areas by contemporaneous village groups rather than a sequential use of the areas by the same group. During the historic pe- riod, the arikara consolidated into one major village, the Leavenworth site. the site was divided into two spatially and ethnically separate communities, often designated as an upper and lower village (Bowers, 1935:5; Krause, 1972:16). a comparison of ceramic modes from these two parts of Leavenworth suggested very little difference be- tween them (see Krause, 1972, table 5). absolute temporal control, in the form of a large se- ries of radiocarbon and thermoluminescent dates, was brought to bear on the problem of ceramic variation be- tween contemporaneous hidatsa villages from the Knife region by ahler and swenson (1993, fig. 17.16). their re- sults indicated that there are sufficient differences between the potting traditions of the different hidatsa subgroups to render traditional ceramic seriations misleading when interpreted as reflecting only temporal variations. in an- other study, ahler (1988:108?119) demonstrated that specific ceramic attributes reflect temporal and ethnic variability among the Mandan, hidatsa, and arikara and their various ethnic subgroups. these attributes are used to assign specific villages to these ethnic entities. finally, in his analysis of the settlement pattern within the sommers site (39st56), steinacher (1990:220?238) indicated that there are a number of factors that contribute to ceramic variability at the site, including length of occupation. cOrrespOndence analysis Correspondence analysis is a member of a family of pattern-seeking data reduction multivariate statistical techniques. other more familiar techniques to archeolo- gists, such as factor analysis, principal components analy- sis, and multidimensional scaling, belong to this general group designed to extract a reduced number of factors, dimensions, or axes from a data matrix or matrix of similarity/dissimilarity coefficients (see sneath and sokal, 1973:245?250; shennan, 1988:241?297). implicit in the use of these techniques is the assumption that there is more than one underlying dimension or pattern of variability in the data. in the case of this study of Plains Village ceramic variation, such dimensions may be related to temporal and spatial patterned variability, or perhaps to some other fac- tors. Multivariate techniques should not be viewed as a panacea because they mask patterns in data and do not allow the exploration of other dimensions of variabil- ity available through other techniques (Plog, 1983:131). steinacher (1990) presented an example of the use of other techniques and alternative interpretations of ceramic vari- ation within one initial Middle Missouri village. Correspondence analysis, also referred to as recipro- cal analysis, was first widely applied in the field of ecology to study variation in vegetation along a number of under- lying gradients, such as available moisture (M. o. hill, 1973, 1974). in these applications, the incidence (presence/ absence) or abundance of a number of plant species are recorded at a number of locations. Each location is referred to as a sample. the resulting species by sample data ma- trix is then input into correspondence analysis to extract underlying gradients, such as moisture, slope, plant suc- cession, and plant communities. in this sense, it has most often been used to generate empirical generalizations of the n u m b e r 4 7 ? 2 9 data by induction. it is interesting to note, however, that a number of articles on correspondence analysis by ecolo- gists also refer to seriation applications in archeology, with Kendall?s (1971) work being cited most often (M. o. hill, 1974:344, 351?354; hill and gauch, 1980:47; Warten- berg et al., 1987:438). Correspondence analysis was first introduced to the larger archeological community through the efforts of Bolviken et al. (1982). it has gained wide- spread application in old World archeology in the 1980s, but only recently in north america (Madsen, 1988; Bax- ter, 1994:133?139). Correspondence analysis is a nonparametric ordination technique developed primarily for the analysis of contin- gency tables, although it can be applied to both incidence and abundance data matrices (M. o. hill, 1974:348?349; greenacre, 1984:55?58, 308?315; ter Braak, 1985:859, 1987:96?97; Weller and Romney, 1990:72?74; Ringrose, 1992:616?617). it also has been used with continuous data (M. o. hill, 1974:348?349; Davis, 1986:589?594). Corre- spondence analysis is considered to be a variant of principal components analysis based upon a chi-square distance met- ric (Bolviken et al., 1982:43; gauch, 1982:148; Wartenberg et al., 1987:437), with both techniques resulting in similar solutions in many applications (M. o. hill, 1974:340, 346 349; ter Braak, 1985:871; Ringrose, 1992:615?616). Like principal components, correspondence analysis derives a series of axes or eigenvalues from a data matrix with each axis accounting for a decreasing amount of variation in multidimensional space as additional axes are extracted. these axes explain as much variation in a data matrix as possible from a ?null hypothesis? of no association be- tween the matrix rows and columns (Baxter, 1994:114). unlike principal components, the procedure uses a two- way weighted averaging algorithm or formula applied to a series of iterations to reach a solution (ter Braak, 1987:97? 103). as greenacre (1984:54) described it, correspondence analysis displays the rows and columns of a data matrix as points in dual low-dimensional vector spaces. in this sense, it orders or arranges a data matrix composed of both the species (or variables, ceramic types) and sample (or cases, site components) scores in terms of their similarities to other species or samples simultaneously (hence the term two-way or reciprocal analysis). it thus provides a useful way to explore the relationships between both variables and cases in a single procedure (Bolviken et al., 1982:43; gauch, 1982:144?147; greenacre, 1984:60?61). this fea- ture is apparent once the actual ordinations that are the subject of this study are presented in chapter 6. once an initial solution is found, the reordered data matrix is re- entered into the algorithm until a more precise ordering is achieved. this iterative process is done a number of times until a final stable ordered matrix is achieved, indicating that the first axis is successfully extracted. once this first axis is extracted, a second axis or dimension orthogonal or independent of the first is extracted by applying sev- eral intermediate steps to the process before the two-way weighted averaging algorithm is once again applied to the data matrix through a number of iterations. this process of orthogonalization and recalculation of the weighted species and sample scores is applied any number of times, depending on how many axes or eigenvalues are desired. these scores are standardized to a mean of zero and a vari- ance of one (z-scores). Correspondence analysis is based upon a chi-square metric, so low-expected cell frequencies should be avoided in any analysis (Weller and Romney, 1990:71; Baxter, 1994:113?118). there appears to be some ambiguity in the literature as to whether this applies to inferential or hypothesis testing during correspondence analysis, or whether it applies to use of the technique in a purely de- scriptive sense. Weller and Romney (1990:71) and Baxter (1994:113) correctly stated that inferences based upon ta- bles with low expected cell frequencies can invalidate for- mal chi-square testing. use of correspondence analysis as an exploratory, data-reduction, descriptive tool, eliminat- ing formal testing of the null hypothesis of no association between variables and cases, could permit low-expected cell frequencies. the problem arises when low frequencies in a particular row or column unduly influence the solu- tion of a correspondence analysis (Baxter, 1994:115). al- though this is acknowledged to be a problem, an example consisting of a single infrequently occurring artifact type that Baxter (1994:113, 115, 118) used demonstrates that low frequencies sometimes have little effect, or ?inertia? on an analysis. in the present study, a number of components and pottery types are characterized by low frequencies. the approach is to include as many of them, within reason, into the analysis because (1) they are radiocarbon dated, (2) their potential interpretive value, (3) including many components can help to delineate changes in regional set- tlement patterns, and (4) the importance of some pottery types in temporal-spatial dynamics is unknown. a num- ber of additional analyses not reported in this study were made eliminating infrequently occurring pottery types, resulting in little change in the overall placement of com- ponents or types. in addition, the downweighting option of the detrended correspondence analysis was chosen in several analyses, also resulting in little change in the final results. 3 0 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y in order to illustrate the correspondence analysis pro- cedure, a data matrix consisting of 11 components and 12 ceramic types composing a portion of the Le Beau phase sample used in this study is passed through one iteration or step in the algorithm by manual calculations. this pro- cedure follows the one described by ter Braak (1987:97? 107). table 2 lists a first iteration ordering of these 11 components, initially placed in numerical order by site number: anton Rygh Ri (39Ca4), anton Rygh Rii?iV (39Ca4), Bamble Early (39Ca6), Bamble Late (39Ca4), Red horse hawk a (39Co34), four Bear (39DW2), spotted Bear (39hu26), sully Late (39sL4), sully Middle (39sL4), swan Creek C+D (39WW7), and swan Creek B (39WW7). according to step 2 (ter Braak, 1987, table 7), the weighted-average site scores (uk in table 2), are cal- culated by multiplying the frequency of each type for a given component by the numerical position of these types in table 2 (i.e., 1 = cord-impressed, s-shaped rim through 12 = undecorated straight rim). these summed values are then divided by the sum of all types for a particular com- ponent. as an example, the uk for Bamble Late is com- puted as follows: [(117 ? 1) + (27 ? 2) + (16 ? 3) + (1 ? 4 ) + (14 ? 5) + (17 ? 6) + (0 ? 7) + (7 ? 8) + (260 ? 9) + (106 ? 10) + (243 ? 11) + (384 ? 12)] / [(117 + 27 + 16 + 1 + 14 + 17 + 0 + 7 + 260 + 106 + 243 + 384)] = 11132/1192 = 9.34 an examination of the first iteration of this matrix places components in a close approximation of their final order (see chapter 6). the purpose of achieving this ar- rangement of values in a data matrix is to model the pat- tern in ecology in which successive replacements of species usually show gaussian, unimodal, or bell-shaped distribu- tion curves with respect to environmental gradients (hill and gauch, 1980:49; ter Braak, 1985:859, 1987:95?96; Wartenberg et al., 1987:434?435). the reader familiar with the literature on seriation in archeology recognizes that ceramic types or attributes are thought to rise, peak, and decline through time resembling unimodal or ?battle- ship? curves (Brainerd, 1951:304; Robinson, 1951:293; ford, 1962:39?40; Dunnell, 1970:309; Mcnutt, 1973). several popular methods of arranging similarity matrices of site components focus on achieving a similar pattern with high values along the diagonal and correspond- ing lower values at off-diagonal positions (Renfrew and sterud, 1969; Robinson, 1951; gelfand, 1971a, 1971b). hill and gauch (1980) proposed a modification of the ordination technique, detrended correspondence analy- sis, shortly after correspondence analysis was introduced in ecology. the modification is designed to alleviate two problems inherent in correspondence analysis or recipro- cal averaging: the arch effect and the distortion of relative distances of species and samples along their axes (hill and gauch, 1980:47?49; gauch, 1982:150?152; greenacre, TabLe 2. two-way weighted averaged (uk) table of selected Le Beau phase components, first iteration of correspondence analysis based on descriptive rim sherd frequencies. Components arranged according to increasing uk. (C.i. = Cord impressed; h.i. = horizontal incised; D.i. = Diagonal incised; hR.i. = herringbone incised; t/f i. = tool or finger impressed; un = undecorated; Dh.i .= Diagonal or herringbone incised; t.i. = tool impressed; f.i. = finger impressed.) straight/Curved/flared/simple rim forms s-shaped/Collared/Compound rim forms Exterior rim Exterior rim decoration decoration Lip decoration Component C.i. H.i. D.i. Hr.i. T/F i. uN DH. i. H.i. C.i. T.i. F.i. uN uk swan Creek B 103 28 7 4 32 0 0 27 60 134 101 15 7.02 Bamble Early 13 3 2 0 1 1 0 1 7 5 11 7 7.10 anton Rygh Rii-iV 79 22 20 3 7 3 0 18 18 76 56 95 7.51 swan Creek C+D 256 181 47 26 55 1 0 48 403 393 580 141 8.05 sully Middle 12 16 51 3 1 7 3 28 41 125 96 21 8.49 anton Rygh Ri 31 1 2 2 1 0 0 2 50 23 69 23 8.74 Bamble Late 117 27 16 1 14 17 0 7 260 106 243 384 9.34 spotted Bear 20 2 4 0 0 2 0 0 319 327 106 34 9.50 sully Late 12 0 17 4 1 0 2 1 20 71 102 44 9.66 four Bear 22 4 24 0 2 1 0 0 126 229 393 368 10.50 Red horse hawk 5 0 0 0 0 0 0 0 78 58 227 166 10.82 n u m b e r 4 7 ? 3 1 1984:232; ter Braak, 1987:105?107). the arch, or horse- shoe, effect occurs when axes extracted by correspondence analysis are dependent upon each other, resulting in a cur- vilinear plot of samples along two or more axes. horseshoe patterns are commonly found in archaeological applica- tions using multivariate data reduction techniques, result- ing in curved patterns of components, samples, or cases in two dimensions (Kendall, 1971; LeBlanc, 1975:35; Dren- nan, 1976; Marquardt, 1979:320; ahler and swenson, 1993, fig. 17.2). the second problem of distance distortion is related to the arch effect. species (or ceramic types) and samples (or components) at the ends of the extracted axes or dimensions exhibiting this horseshoe effect tend to be more compressed relative to one another compared with those nearer the center of the axes. these two problems arise because the underlying model of correspondence or reciprocal analysis differs from the mathematical proper- ties of unimodal or gaussian distributions encountered with real data (gauch, 1982:150). as a result, some re- searchers (hill and gauch, 1980:47; Peet et al., 1988:926) thought that these problems were more a artifact of the ordination technique used than a real property of the data. Wartenberg et al. (1987), among others, contended that the arch effect is a real property of the data that should not be ignored. the solution that hill and gauch (1980:48) devised to ameliorate these inherent problems in corre- spondence analysis was to divide an axis into a number of segments and center the species or sample scores to a mean of zero (see also gauch, 1982:152?157; ter Braak, 1987:105?107). this rescaling or detrending of species and sample scores between axes makes all axes orthogonal and independent of one another, removing any systematic re- lationships between them (hill and gauch, 1980:47?48; gauch, 1982:153). one of the most popular detrended correspondence analysis computer programs is DECoR- ana (DEtrended CoRrespondence analysis) developed by M. o. hill (1979). this program is used in all but one of the ordinations of components presented in chapter 6. DeNDrochroNological DatiNg Dendrochronology in the Middle Missouri subarea has provided some of the earliest dates from sites, par- ticularly those from north Dakota (Will, 1946, 1948). faced with an extensive salvage archeology program at numerous sites in the Dakotas in conjunction with dam building after World War ii, there was a need to date these sites. Dendrochronology was one of several dating tech- niques incorporated into the Missouri Basin Chronology Program with the intended purpose of developing inter- related chronologies, using a variety of techniques (ste- phenson, 1958). the results of this program were reported by Weakly (1971) and in the Missouri Basin Chronology statement no. 3. Caldwell and snyder (1983) reviewed the topic and concluded that the dendrochronological dates from the Middle Missouri subarea are unreliable. they concluded that all radiocarbon and dendrochrono- logical dates from the Middle Missouri subarea must be questioned. this was supported by a comparison of three radiocarbon and dendrochronological dates on the same wood samples from the sommers site, producing widely divergent dates that fall outside the temporal range of variant sommers is assigned to. Dendrochronological dates from specific sites are discussed in chapter 6. raDiocarboN DatiNg Broadly speaking, the purpose of the radiocarbon dat- ing portion of this chronology program is to integrate to the maximum extent possible an accurate, high-precision calendar-based chronology with the ceramic ordination process. to achieve this, the following two specific goals in the area of radiocarbon dating were accomplished: (1) the evaluation of the relatively large number of exist- ing radiocarbon dates for accuracy and precision, and the reduction of this array of existing dates in to a smaller set considered to be most reliable and informative about the cultural components and time period under study; and (2) the generation of new, maximally precise and accu- rate radiocarbon dates for as many previously undated or poorly dated components as possible. Broad developments in three aspects of radiocarbon dating during the past decade conditioned the specific approach to these goals. first, development and routine application of the accelerator mass spectrometry (aMs) dating method makes it possible to date many components more precisely compared with conventional means (Creel and Long, 1986; aitken, 1990:76?85). Because aMs dat- ing uses an atom counting procedure, it is used with small samples, as little as several milligrams, within a short time span (aitken, 1990:76?77). Conventional beta counting requires larger samples, on the order of about 10 grams. the aMs procedure facilitates the dating of small samples that would ordinarily not be suitable for conventional ra- diocarbon dating. this development has a direct bearing on the ability to date many components in the study area, which previously would not be accessible to radiocarbon assessment. 3 2 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y second is the general understanding of the importance of focusing dating on short-lived botanical remains un- equivocally associated with human activities (e.g., culti- gens) as opposed to dating the often more abundant wood or wood charcoal, which is subject to a much greater pos- sibility of discontinuity between the growth event and the cultural event under study. Error potentially assignable to ?old wood? sources has been demonstrated to be substan- tial and confounding in several contexts in which the wood was well preserved in the environment (e.g., schiffer, 1984, 1987:308?311; smiley, 1985; Dean, 1991). in the present study area, old wood (wood predating the cultural event that is to be dated) undoubtedly exists in the inner rings of large timbers used as house posts, and dates on posts form a large fraction of previously existing radiocarbon dates. in addition, cultural practices compound the ?old wood? problem. Cultural use of scavenged wood and river drift logs or salvaging of old house timbers would make most radiocarbon-dated samples appear to be older than the events (i.e., village occupation) with which they are asso- ciated. these practices have been documented among the hidatsa (Wilson, 1934:359, 372?376). it is therefore clear from cultural practices within the region, as well as from case studies outside the region, that compelling reasons exist for reevaluating existing dates that were based upon wood and for reorienting any new radiocarbon dating toward more ideal sample materials. fortunately, the advent of routine aMs dating allows for a focus on dating materials that often occur in relatively small samples sizes in regional study sites. in our case, these newly accessible target materials include cultigens, wild plant food remains, and carbonized cooking residues occurring in small quantities on pottery vessels. Dates on maize are now well accepted when corrected through iso- topic fractionation (Creel and Long, 1986). Dating of pot- tery residues seems practical, but it is far less established by test studies. Dating of pottery residues in the present study should therefore be considered an experimental approach. a third advancement in radiocarbon dating that has direct bearing on this study is the development of routinely accessible computer programs that rapidly perform several very important functions, such as (1) testing for contempo- raniety among a suite of dates, (2) averaging two or more dates, and (3) calibrating individual or mean dates into a calendar age by basing them upon the varying amounts of radiocarbon in the atmosphere through time. We refer in particular to components of the desktop computer program CaLiB 3.0 and its successors (stuiver and Reimer, 1993). Regarding the first point, the test of contemporaniety is used as a screening tool to eliminate from consideration suites of dates that contain too large an amount of error to accurately reflect the true age of the associated archaeo- logical context based on a careful assessment of archaeo- logical data independent of radiocarbon results (see shott, 1992, for an excellent example of a systematic compari- son of radiocarbon results and archaeological evidence to achieve the best possible age estimate for a site). Error as- sessed in this fashion includes a combination of association error (poor temporal association between the growing date of the sample and the cultural event), error introduced dur- ing laboratory processing (interlaboratory differences (in- ternational study group, 1982), as well as intra-laboratory error?all in addition to counting error), and error from other unknown sources. after suites of dates are screened for accuracy, the averaging routine is used to generate a mean date with the highest possible level of precision for a given cultural component. although it has been known for more than 30 years that differing amounts of radiocarbon in the atmosphere and oceans can affect the relationship between the radio- carbon age and the calendar age of dated materials, only recently have computer programs become available to correct for these variations. Program CaLiB (stuiver and Reimer, 1993) greatly facilitates the task of calendrically calibrating the 375 dates employed in this study. the pro- gram is based upon a recent effort to calibrate the radio- carbon time scale (stuiver and Becker, 1993). a graphical representation of a portion of one of the calibration curves upon which this program is based is presented in figure 5. it is observed that this decadal tree-ring curve fluctuates or ?wiggles? more in some time periods than others. this has implications for the precision of corrected radiocar- bon dates that are based upon the curve and can mean that the number of curve intercepts and the 1-sigma or 2-sigma ranges of dates varies depending on their position on the curve. generally, dates that fall within the straight portions of the curve calibrate to smaller ranges than those on other more ?jagged? parts. once a new set of screened, averaged, and calendri- cally calibrated radiocarbon dates are obtained for the largest possible number of components, the radiocarbon dates are integrated with ceramic ordination results in the following manner. Components with radiocarbon dates are used in conjunction with ceramic ordination to assign, within certain error parameters, many of the undated com- ponents included within the ceramic ordinations to 50 and 100 interval time periods. this is done by determining if an ordination reflects a temporal dimension or pattern by appealing to site stratigraphy, historic documentation, ra- diocarbon dates, and other temporal indicators. once this n u m b e r 4 7 ? 3 3 Figure 5. Decadal radiocarbon calibration curve (adapted from stuiver and Becker, 1993: fig. 2a?c). 3 4 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y is accomplished, the radiocarbon dates associated with a number of components included within each ordination are used to estimate early and late dates of each temporal axis that is interval scaled by detrended correspondence analysis. once this is done, assigning ?absolute? dates (or more appropriately hypothesized dates) to each component is a matter of interpolating between two or more dated components in each ordination to other undated compo- nents. this technique, which is explained in more detail in chapter 6, results in an ordering of village components along an ?absolute? time scale that is divided into 50- and 100-year increments. only with a massive absolute dat- ing program, far beyond the scope of the current study, could a more precise temporal ordering of components be achieved, assuming an appropriate amount of datable ma- terial and archeological context information is available. as a result, the temporal orderings presented in chapters 6 and 7 are considered to be hypothesized chronologies to be tested more thoroughly with additional radiocarbon dates and other independent bodies of temporally sensi- tive artifacts. Various levels of confidence also can be as- signed to each variant or phase chronology depending on the comprehensiveness of available dates and the nature of ceramic variability within each one. the current radiocarbon dating program is considered to be the third systematic effort to date sites of the Plains Village pattern within the Middle Missouri subarea. the first of these was the Missouri Basin Chronology Program, which was intended to develop inter-related chronologies, using a variety of techniques and data sets, including ra- diocarbon dating, dendrochronology, geological-climatic dating, and ceramic typology (stephenson, 1958). the ob- jectives were to develop a general chronology of the sub- area that could be used to date subsequently excavated sites and be related to cultural developments in other parts of north america. it is particularly relevant to the current study that ?the type of ceramic analyses done by Dr. a. o. shepard in conjunction with the type of ceramic seriation that has heretofore been done in the area? was a compo- nent in the program (stephenson, 1958:4). the ceramic aspect of the chronology program did not progress be- yond the prospectus dealing with technology and how clay sources could be used to identify trade vessels (huscher, 1958). other portions of the chronology program yielded substantive results however, particularly in the area of ra- diocarbon dating. a summary of many of the radiocarbon dates from the Missouri Basin Chronology Program ap- pears in neuman (1967). the second integrated dating program to be under- taken in the Middle Missouri subarea was in conjunction with the first phase of research at the Knife River indian Villages national historic site (ahler and haas, 1993). the cultural chronology that grew out of this and related anal- yses formed the basis of the reconstruction of Plains Vil- lage occupations in the Knife and heart regions employed in the present study (see ahler, 1993b). the chronology of the remaining regions within the Middle Missouri subarea (Cannonball, grand-Moreau, Bad-Cheyenne, Big Bend, fort Randall) is the result of the current study. site stratigraphy When a number of stratigraphically related prove- nience units within a particular site are grouped together into components or other cultural-temporal units, a very powerful tool is created to aid the interpretation of the results of any archeological ordination that includes these units. interpreting the various dimensions, factors, axes, or patterns of a multivariate ordination procedure that includes some variables thought to be related to time ul- timately depends upon information independent of the actual data used to construct these ordinations. Like the absolute dates provided by radiocarbon dating, stratigra- phy is an interpretive tool to help in the construction of chronologies relying on ceramic ordination. for the most part, the limited stratigraphic informa- tion available for this study seems to be an incidental by- product of the excavation of a number of villages with relatively thick midden deposits. Even in these cases, the emphasis of the interagency archeological salvage Pro- gram (iasP) program was to uncover as many houses as possible, within the existing budgetary constraints; areas between houses received far less attention it is no exag- geration to state that good stratigraphic control usually can be attained with relatively small excavations placed to maximize this information. the excavations at some of the villages in the Knife River indian Villages national historic site dramatically demonstrates this point (ahler, 1993c). it is emphasized that the stratigraphically related units sought in the current study are those that are included within the same ordination, usually confined to those com- ponents assigned to the same variant or phase. there is a wealth of stratigraphic information from many excavated sites within the Middle Missouri subarea, although it usu- ally involves a short time period or components assigned to different variants or traditions that cannot be included within the same ordination. Most of the stratigraphic controls for the present study are derived from villages containing relatively thick n u m b e r 4 7 ? 3 5 midden deposits (i.e., those occupied shortly before and during the protohistoric or Post-Contact period). the excavations at the Big hidatsa and Lower hidatsa sites within the Knife River indian Villages national historic site is a case in point. other villages providing some de- gree of stratigraphic control include a number assigned to the late Extended Coalescent variant (swan Creek, anton Rygh), Le Beau phase (anton Rygh, Bamble, Lar- son, spiry-Eklo, swan Creek, sully), talking Crow phase (Medicine Crow), and to a lessor degree the Extended Coalescent (Walth Bay, Lower grand), and initial Middle Missouri (Jiggs thompson, Pretty head, sommers) vari- ants. given the magnitude of this study and the number of components included within it, this list is relatively small and inadequate. Because there are a large number of Le Beau components that provide stratigraphic control, the ordination of this phase, along with the existing heart River and Knife River complex sequences from the Knife and heart regions, is firmly established. affixing an ab- solute time scale to the Le Beau sequence relies on dating parameters developed largely from components occupied prior to these villages, because they were occupied most intensively after the latest feasible date for radiocarbon dating (i.e., after ad 1650). euro-americaN traDe materials artifacts of Euro-american derivation, either in their original manufactured form or modified by native ameri- cans, serve as a horizon marker for the beginning of the protohistoric or Post-Contact period in the Middle Mis- souri subarea. some of the more common artifacts include copper, brass or iron knives; arrow points; tubes; beads; awls; fishhooks; conical tinklers; and scraps cut up from kettles and other trade items. glass beads also are commonly found in Post-Contact period sites. Billeck (2000) is the only one to have systemat- ically dated sites in the Middle Missouri subarea using glass beads. his study provided an independent dating source for the post-contact sequences in this study. the actual process by which Euro-american trade items are eventu- ally incorporated into the archaeological record is complex (Ray, 1978; toom, 1979). generally, sites later in time have larger numbers and more varieties of Euro-american trade items than those dating to the early part of the proto- historic period. there also are qualitative differences, as Billeck (2000) demonstrated. these trends provide ways to estimate the relative and absolute temporal placement of sites in this period. one of the major limiting factors is the field recovery techniques employed in excavations and their consistency of use, as pointed out by Billeck (2000:1) and in chapter 4. Because almost all of the sites used in this study were not screened during excavation, most of the Euro- american trade artifacts, which tend to be small, were not recovered. only those items that were fortuitously found during excavation or were excavated with more attention to detail (e.g., human burials) remain in extant collections. Comparisons between Billeck?s chronology and this study also are limited by the difference in archaeological context; this study is based upon villages, whereas Billeck?s study fo- cused on their associated cemeteries. this is not a problem when the villages were occupied for relatively short periods; however, some were occupied for long periods spanning the late prehistoric and protohistoric periods. Dates based upon beads probably tend to place these sites (i.e., the latest oc- cupations) later than the dates derived from ceramics. Even when different components are dated in the present study, it is unclear if these occupations are defined precisely enough to allow for a one-to-one comparison with their associated cemeteries. Many of the late multi-component sites used in this study were excavated under salvage conditions that did not allow enough time for fine-scale separations of long, continuous occupations. the focus of excavations at many sites were contexts, such a houses (rather than thick mid- dens), that could not be related stratigraphically. still other sites either lack reports (e.g., sully, Mobridge, Cheyenne River, Black Widow, 39st25), do not adequately justify their separation of components (e.g., swan Creek, spiry- Eklo), or are reported based upon grossly inadequate field records (Larson). all of these factors add more uncertainty in the dating process. nevertheless, comparisons are made in this report between the chronologies based upon ceram- ics and those based upon beads. historic DocumeNtatioN historic documentation of Mandan, hidatsa, and arikara villages not only provides information about the tribes that occupied particular regions and specific village sites along the Missouri River in the Dakotas, but also when these villages were inhabited. this information is most complete for the eighteenth and nineteenth century Mandan and hidatsa communities in the Knife and heart regions, but information also is included for a number of arikara villages in north and south Dakota. an increasing interest in ethnohistory during the past 30 years has generated a number of summaries of the perti- nent primary sources. these include Chomko (1986), Wood 3 6 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y (1986a:25?58; 1986b), and thiessen (1993b), as well as the earlier studies of Wedel (1955:77?84), hughes (1968:19? 24), and Lehmer (1971:164?179, 1977a, 1977b). these compilations of European and american observations, aided by native informants, form the bulk of the historic documentation relating to the time of occupation and aban- donment of specific Plains Village sites within the subarea. oral traDitioNs oral histories provide another source of information useful in reconstructing Mandan, hidatsa, and arikara origins, their cultural growth, and eventual decline. Most important for the current study is their record of where they lived during these periods and whether these loca- tions or specific village sites can be tied to the archeo- logical record. this historical record is critical not only as a source of independent chronological information, but also as a guide in assigning certain village components, archeological phases, complexes or other units to par- ticular tribal entities or ethnic subdivisions within these groups. the further these oral histories are extended back in time, the more secure are the ethnic assignments of the prehistoric and post-contact villages. Making these links also is invaluable in interpreting the archeological re- cord, a fact not lost on the adherents of the direct-historic approach to archeology on the plains (Wedel, 1938, 1940:296; strong, 1940:353?354). studies, such as the present one, which rely on multiple lines of archeological, historic, ethnographic, linguistic, and physical anthropo- logical evidence in their reconstructions, follow the gen- eral methods set forth by these early pioneer archeologists more than 50 years ago and bring it full circle to these early efforts. syms (1985) argued for a return to these various sources of information in the interpretation of the archeological record. it becomes apparent later in chapter 6 that these various sources of interpretive information vary in detail and usefulness, depending on the tribe or ethnic group under consideration. the historic record is most detailed for the Mandan and hidatsa, as are their oral traditions. historical linguistics and physical anthropology are best documented for the arikara and their ancestors. on one level, more is known about the archeology of the proto- historic and prehistoric arikara than the Mandan and hidatsa, although this balance has tipped in recent years as the Knife River indian Villages national historic site is interpreted through the archeological, historic, and ethnographic records of the hidatsa (thiessen, 1993b). Much of what is known about the Mandan and hidatsa is the result of Bowers? (1950, 1965) historical and ethno- graphic work. a comparable study of the arikara has not been conducted, although holder (1970) provided some relevant information as more continues to be made avail- able, if only in piecemeal fashion (Parks, 1979a). Bowers keen interest in setting the prehistoric context of the Man- dan and hidatsa, so useful in recent cultural-historical reconstruction (ahler, 1993a:34?38), has not been dupli- cated for the arikara. this information may exist in the unpublished notes of Preston holder, although efforts by this author to gain access to that source have not been productive. historical liNguistics Linguistic studies that outline the temporal relation- ships between various northern Plains language groups (hughes, 1968; Parks, 1979b; hollow and Parks, 1980; springer and Witkowski, 1982) provided useful informa- tion relating to the distributions and splitting of various Caddoan (e.g., arikara and Pawnee) and siouan (Man- dan, hidatsa) language groups. to the extent that such studies can date the approximate times of language group splits, the more useful this information is when combined with other studies of the affinities of certain biological populations, or ceramic relationships between villages or complexes of villages. the results of these historical lin- guistic studies are integrated with other temporal data in chapter 6. craNiometric DistaNce through the efforts of Bass (1964), Jantz (1972, 1973), and a number of their students (summarized in Bass, 1981), craniometric analyses focus on morphology to infer biological distance between skeletal populations associated with certain northern Plains Villages. among the findings in these studies is what various investigators interpret to be temporal variability between the popula- tions. Despite the fact that the absolute dates they em- ployed were only estimates that were based upon very limited information ( Jantz, 1972:24?25, 1973, table 1; owsley and Jantz, 1978, table 2; Key and Jantz, 1981, table 1; Jantz and Willey, 1983, tables 1 and 2), these studies do provide relative orderings between the popula- tions and a measure of biological distance between them that are in turn related to other factors. these results are n u m b e r 4 7 ? 3 7 particularly intriguing for the present study that tries to examine the flip side of the affinity question (i.e., the ceramic or inferred ?social distance? between villages). Clearly, the complementary nature of these two data sets, ceramic and skeletal, and the inferred social and biologi- cal affinities are important to the overall effort to date and otherwise relate the villages and their associated re- mains, even though the underlying processes of ceramic and skeletal change are probably very different. 4 Data acquisition t able 3 lists a total of 225 site components used in this chronology of Middle Missouri Plains Village sites. the accompanying data matri- ces, listing the percentages and frequencies of various ceramic types, and cord-impressed decorative attributes, appear as a series of tables in appendix a (tables a.1?a.6). site assemblages containing 50 rim sherds or more appear in this table, although there are a few exceptions in which smaller samples are included. Components not included in the analysis are discussed later. also not included in the study are a number of components from the Knife and heart regions in north Dakota; the established chronology for this region (ahler, 1993b) is integrated into the present study. Related Mill Creek (flan- ders, 1960; D. R. henning, 1971; anderson, 1981; tiffany, 1982) and Lower Loup/historic Pawnee (grange, 1968, 1984; o?shea, 1989) chronologies also are considered herein, although no formal re-analyses are attempted. the com- ponents in table 3 are organized by traditionally defined taxonomic units, the variant and phase, with additional information pertaining to the degree of the fieldwork and source of the information. Variant designations follow Lehmer (1971), and taxonomic assignment is listed only in cases in which the first re- porting of the site designated a phase, focus, aspect, or variant. the intensity of the field effort includes four categories: (1) surface or beach collections; (2) ini- tial reconnaissance tests; (3) major excavations not resulting in the exposure of houses; and (4) excavations uncovering houses and the number of partial or complete houses excavated. the source of the information listed in table 3 deserves additional clarifica- tion. there are two major sources of actual rim sherd counts: (1) published and unpublished literature; and (2) personal inspection by the author, the latter desig- nated as ?Johnson? followed by the institution in which the collections were cu- rated at the time they were examined. Explanations of all abbreviations appear at the end of this table. in the case of the cord-impressed motif data in table a.6, all collections were personally examined except for Deapolis, amahami, Rock Vil- lage, nightwalker?s Butte, and Biesterfeldt, reported in Wood (1971) and Lehmer et al. (1978). When more than one source for each component appears in table 3, the first listed provides the actual rim sherd counts, whereas the second gives back- ground information on the site. in some instances, the second source also provides 4 0 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y TabLe 3. List of Plains Village tradition components used in the ceramic ordinations by taxonomic assignment, extent of fieldwork, and source of the ceramic data. Dashes indicate not assigned or not available. Taxonomic extent of Variant/Component1 assignment2 Field Work3 Source4 InItIal MIddle MIssourI Broken Kettle West (13PM25) great oasis 1 house Johnson, 1974 Williams (13PM50) great oasis Excavations Williams, 1975 Larsen (13PM61) great oasis Excavations henning and King, 1982; henning 1996 Cambria (21BE2) Cambria Phase Excavations Johnson (Mhs;uMM); Knudson,1967 Price (21BE25) Cambria focus Excavations scullin (1979) great oasis (21Mu2) great oasis Excavations Johnson (uMM); Johnson, 1969 ferber (25CD10) great oasis Excavations Edwards, 1993 Packer (25sM9) great oasis Excavations Johnson (nshs); Bozell and Rogers, 1989 Crow Creek (39Bf11) ? test hanenberger, 1986; Kivett and Jensen,1976 Pretty Bull (39Bf12) ? (6) Excavations Johnson (sDaRC) akichita (39Bf221) ? Excavations Johnson (sDaRC); husted, n.d. swanson (39BR16) over f. 4 houses Johnson (sDaRC); hurt, 1951 Jones Village (39Ca3) iMM tests Johnson (sDaRC); Johnson, 1986 Pease Creek (39Ch5) great oasis Excavations Johnson (nMnh) oldham (39Ch7) great oasis (6) Excavations Johnson (nMnh); huscher, 1957 hitchell (39Ch45) great oasis (6) Excavations Johnson (nMnh); Johnston, 1967 39Ch205 great oasis Excavations Johnson (nMnh) Mitchell (39DV2) over f./Lower James Ph. 2 houses Johnson (sDaRC); alex, 1981b; Meleen, 1938 Chapelle Creek C (39hu60) over f. (5,6) Excavations Brown, 1967 st. John (39hu213) great oasis tests Johnson (nMnh); Jensen, n.d.a twelve Mile Creek (39ht1) over f. 1 house Johnson (sDaRC); Meleen, n.d. Dinehart (39LM33) ? 2 houses Johnson (sDaRC) King (39LM55) iMM 1 house Johnson (nMnh); Ehrenhard, 1971 Jiggs thompson a (39LM208) ? tests Caldwell & Jensen, 1969 Jiggs thompson B (39LM208) grand Detour Phase 2 houses Caldwell & Jensen, 1969 Langdeau (39LM209) grand Detour Phase 5 houses Caldwell & Jensen, 1969 Jandreau (39LM225) ? 2 houses Johnson (sDaRC); Jones, n.d. gilman (39LM226) ? tests Johnson (sDaRC) Pretty head a (39LM232) grand Detour Phase 1 house Caldwell & Jensen, 1969 Pretty head B (39LM232) over f. 2 houses Caldwell & Jensen, 1969 heath (39Ln15) great oasis 1 house Johnson (aCaL); hannus et al., 1986 Brandon (39Mh1) over f. 7 houses Johnson (sDaRC); over & Meleen, 1941 fay tolton (39st11) anderson Phase 2 houses Johnson, 1976 h.P. thomas 1 (39st12) iMM Variant (5,6) Excavations falk, Johnson, and Richtsmeier, in prep. Breeden a (39st16) anderson Phase (6) 4 houses Brown, 1974 Dodd (39st30) Monroe/anderson foci (6) 11 houses Johnson (nMnh; sDaRC); Lehmer, 1954a sommers (39st56) iMM 8 houses steinacher, 1990; falk, steinacher, and Johnson, in prep. Cattle oiler Early (39st224) iMM (2,5) 6 houses Johnson (nshs); Moerman & Jones, 1966 gavins Point (39yK203) great oasis surface Johnson (nMnh); Brown, 1968 extended MIddle MIssourI havens (32EM1) fort yates Phase 4 houses sperry, 1968a; 1995 tony glas (32EM3) ? 2 houses Johnson, 1999; howard, 1962a Clarks Creek (32ME1) fort yates Phase tests Calabrese, 1972 White Buffalo Robe (32ME7) nailati Phase 1 house Lee, 1980 amahami (32ME8) nailati Phase 1 house Lehmer et al., 1978 Bendish (32Mo2) fort yates Phase 2 houses thiessen, 1976 Cross Ranch (32oL14) nailati Phase 2 houses Calabrese, 1972 fire heart Creek (32si2) fort yates Phase (6) 3 houses Lehmer, 1966 Paul Brave (32si4) thomas Riggs focus 3 houses Wood & Woolworth, 1964 Ben standing soldier (32si7) ? 3 houses hoffman, 1970a south Cannonball (32si19) EMM Variant 7 houses griffin, 1984 McKensey (39aR201) ? 1 house Caldwell, 1966a 39aR210 ? tests Caldwell, 1966a Vanderbilt Village (39Ca1) EMM tests, surface Johnson, 1986; Jensen, n.d.b Keen Village (39Ca2) EMM tests, surface Johnson, 1986; Jensen, n.d.b helb (39Ca208) fort yates Phase Excavations falk & Calabrese, 1973 n u m b e r 4 7 ? 4 1 Jake White Bull (39Co6) EMM Variant tests/surface ahler, 1977a travis i (39Co213) EMM Variant (5) Excavations Johnson, 1982 Calamity Village (39DW231) ? 1 house Johnson (sDaRC); Mcgowen, 1973 thomas Riggs (39hu1) thomas Riggs focus 6 houses Johnson (sDaRC); Kleinsasser, 1953 sully school (39sL7) ? (5) 1 house Johnson (sDaRC); Mcnutt, n.d.a C.B. smith (39sL29) thomas Riggs focus (5) tests Johnson (sDaRC); Mcnutt, 1958 Zimmerman (39sL41) ? ? Johnson (sDaRC) Cheyenne River Early (39st1) ? (5,6) 5 houses Johnson (nMnh) indian Creek (39st15) EMM horizon (6) Excavations Lehmer & Jones, 1968 Blk. Widow Ridge E. (39st203) ? (6) 2 houses Johnson (nMnh) Cattle oiler Middle (39st224) EMM (1,6) 1 house Johnson (nshs); Moerman & Jones, 1966 extended/terMInal MIddle MIssourI 32Mo291 ? Excavations ahler et al. (2000) terMInal MIddle MIssourI shermer (32EM10) tMM 4 houses sperry, 1968b huff (32Mo11) huff focus 10 houses Wood, 1967; ahler & Kvamme, 2000 InItIal CoalesCent Lynch (25BD1) anoka focus 5 houses Witty, 1962 talking Crow ii (39Bf3) Campbell Creek Ph. (6) 4 houses smith, 1977 arzberger (39hu6) ? 4 houses spaulding, 1956 Whistling Elk (39hu242) ? 2 houses steinacher, 1983 Black Partizan B (39LM218) arzberger Phase (6) 2 houses Caldwell, 1966b extended CoalesCent Redbird ii (25ht2) Redbird focus 1 house Wood, 1965 Elbee (32ME408) scattered Village Excavations ahler, 1984a no heart Creek (39aR2) Le Compte focus/Phase 1 house hurt, 1970 39aR7 Le Compte focus surface Johnston & hoffman, 1966 anton Rygh RV-Vii (39Ca4) akaska focus (6) Excavations Johnson (nMnh); Knudson et al., 1983 39BR10 ? surface Johnson (nMnh) Locke Creek (39Ca201) ? surface Johnson (Mt) Demery (39Co3) Chouteau a. 4 houses Woolworth & Wood, 1964 fort Manuel (39Co5) ? Excavations Johnson (MWaC); smith & Ludwickson, 1981 Leavenworth Early (39Co9) ? (6) Excavations Johnson (nsM); Krause, 1972 Lower grand (39Co14) ? 5 houses Johnson, 1988 Bellsman Creek (39Co17) ? surface Johnson (Mt) 39Co18 ? surface Johnson (Mt, nMnh) north White Bull (39Co41/207) ? surface Johnson (Mt) Potts Village (39Co19) Chouteau a./Le Compte f. 2 houses stephenson, 1971 h & h (39Co78) ? surface Johnson (Mt) travis i (39Co213) EMM (2) Excavations a. Johnson, 1982 Moreau River (39DW1) ? tests Johnson (nMnh); Winham, 1984 39DW217 ? surface Johnson (nMnh) fox island (39DW230) EC horizon 1 house Kotch & starr, 1968 Molstad (39DW234) Chouteau a./Le Compte f. 4 houses hoffman, 1967 39DW253 ? surface Johnson (Mt) 39DW254 ? surface Johnson (nMnh) scalp Creek a (39gR1) La Roche focus 10 houses hurt, 1952 Pierre school-south (39hu10) EC Excavations Johnson (aCaL) Robinson (39hu15) ? surface Johnson (nMnh); george, 1949 Chapelle Creek B (39hu60) Chouteau aspect (1,6) Excavations Brown, 1967 Little Pumpkin (39hu97) ? surface Johnson, 1984b; toom, 1992d Little Cherry (39hu126) EC Variant Excavations Kapler, 1991 Bowman (39hu204) ? 1 house Johnson (MWaC); Roetzel & strachan,1986a standing Bull (39hu214) ? surface/tests Johnson, 1984b fry a (39hu223) ? 1 house Jensen, n.d.a 39hu241 ? surface Johnson (nMnh) stricker B (39LM1) La Roche Phase 1 house smith, 1975 Bice (39LM31) ? 1 house Johnson (MWaC) Clarkstown B (39LM47) ? surface smith & grange, 1958 Meander (39LM201) Chouteau aspect tests husted, 1965a 39LM222 ? 3 houses Johnson (MWaC) spain a (39LM301) Chouteau a./shannon f. 1 house smith & grange, 1958 hosterman (39Po7) ? 5 houses Miller, 1964 (continued) 4 2 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y TabLe 3. (continued) Taxonomic extent of Variant/Component1 assignment2 Field Work3 Source4 extended CoalesCent (continued) gettysburg (39Po209) EC horizon tests Coleman, 1968 fairbanks (39sL2) ? tests Johnson (nMnh) sully Early (39sL4) ? (6) Multiple houses Johnson (nMnh) sully school (39sL7) ? (2) 1 house Johnson (MWaC) 39sL8 ? tests Johnson (MWaC) 39sL12 ? tests Johnson (MWaC) 39sL23 ? tests Johnson (MWaC) 39sL24 ? 1 house Johnson (MWaC) C.B. smith (39sL29) Bennett focus (2) 1 house Mcnutt, 1958 Cheyenne River Middle (39st1) ? (2,6) 1 house Johnson (nMnh) Black Widow Early (39st3) ? (6) 3 houses Johnson (nMnh) over?s La Roche (39st9) EC horizon 7 houses hoffman, 1968 Meyer (39st10) ? surface Johnson (nMnh); hoard, 1949 h.P. thomas 2 (39st12) EC Variant (1,6) Excavations falk, Johnson, and Richtsmeier, in prep. Cooper (39st45) ? 2 houses Johnson (MWaC) Leavitt (39st215) EC horizon 1 house Lehmer & Jones, 1968 Cattle oiler Late (39st224) ? (1,2) Excavations Johnson (nshs); Moerman & Jones, 1966 Bower?s La Roche (39st232) EC horizon 2 houses hoffman, 1968 swan Creek a (39WW7) akaska focus (6) Excavations hurt, 1957 spiry (39WW10) akaska focus 1 house Baerreis & Dallman, 1961 Walth Bay (39WW203) EC 7 houses Johnson 1988; ahler, 1975, 1975b 39WW300 akaska focus tests hurt, 1957 Payne (39WW302) Chouteau a./akaska f. 2 houses Wilmeth, 1958 FelICIa Phase 39BR201 ? 2 houses Weakly, 1961 Cadotte (39hE202) felicia Phase tests/surface smith & Johnson, 1968 Black Partizan a (39LM218) felicia focus (4) 3 houses Caldwell, 1966b 39LM219 a ? 1 house husted, 1965b Crazy Bull (39LM220) Chouteau aspect 1 house frantz, 1962 Mc Clure (39hu7) felicia Phase 2 houses Johnson (MWaC); Johnston, 1982 talkIng Crow Phase Medicine Crow 1-5 (39Bf2) talking Crow Phase 17 houses Johnson & toom, 1995 talking Crow iii (39Bf3) talking Crow Phase (4) 9 houses smith, 1977 39Bf4 ft. thompson focus Excavations Kivett & Jensen, 1976 Pretty Bull (39Bf12) ? (1) 1 house Johnson (MWaC) fire Cloud (39Bf237) Pahuk a./ft. thompson f. test Karklins, 1970 sanitarium (39BR6) ? 1 house Kruschwitz, n.d. 39BR13 ? tests/surface Johnson (nMnh); Witty, 1960 oldham (39Ch7) ? (1) 13 houses Johnson (nMnh); huscher, 1957 hitchell (39Ch45) fort. thompson f. (1) 2 houses Johnston, 1967 iron shooter (39hu217) ? tests/surface Johnson, 1983b; Peterson, n.d. amos shields (39hu220) ? tests/surface Johnson (MWaC);Roetzel & strachan, 1986b hawk (39hu238) ? 1 house Jensen, n.d.a ocoma (39LM26/27) ? 10 houses Johnson (nshs); Kivett, 1958 39LM34 ? surface Johnson (MWaC) Peterson (39LM215) fort thompson f. 1 house Jensen, 1966 Breeden B (39st16) talking Crow Ph. (1) Excavations Brown, 1974 fort george Village (39st17) Bad River Phase 6 houses hoffman, 1970b Bad rIver Phase Pascal Creek (39aR207) stanley/snake Butte f. tests Johnston, 1966 Chapelle Creek a (39hu60) Bad River Phase (1,5) 5 houses Brown, 1967 Coleman (39sL3) ? 1 house Johnson (MWaC) Little Bend (39sL13) ? tests Johnson (MWaC) Madison (39sL19) ? tests Johnson (MWaC) 39sL28 ? tests Johnson (MWaC) Cheyenne River Late (39st1) ? (2,5) 2 houses Johnson (nMnh) n u m b e r 4 7 ? 4 3 Black Widow Late (39st3) ? (5) 1 house Johnson (nMnh) Dan Donavan (39st5) ? surface Johnson (nMnh) Buffalo Pasture (39st6) Bad River 2 5 houses Lehmer & Jones, 1968; howson, 1941 h.P. thomas 3 (39st12) Bad River Phase (1,5) 5 houses falk, Johnson, and Richtsmeier, in prep. Phillips Ranch (39st14) snake Butte f. 10 houses Lehmer, 1954a indian Creek a (39st15) Bad River 1 (2) 2 houses Lehmer & Jones, 1968 gillette a (39st23) stanley/snake Butte f. 2 houses Brown, 1966 39st25 ? Excavations Johnson (nMnh) Dodd a (39st30) stanley focus (1) 10 houses Lehmer, 1954a 39st50 ? 1 house Johnson (nMnh) 39st51 ? surface Johnson (nMnh) Johnston (39st244) Bad River Phase 2 surface Lehmer & Jones, 1968 Blk. Widow Ridge L. (39st203) ? (2) 2 houses Johnson (nMnh) le Beau Phase anton Rygh Ri-iV (39Ca4) ? (5) Excavations Johnson (nMnh) anton Rygh upper-Lower (39Ca4) Le Beau f., Rygh f. 8 houses Knudson et al., 1983 Bamble Early (39Ca6) Le Beau focus 3 houses Baerreis & Dallman, 1961 Bamble Late (39Ca6) Le Beau focus 3 houses Baerreis & Dallman, 1961 nordvold i (39Co31) ? ? Johnson (unL); Krause, 1962 Red horse hawk (39Co34) Bad River Phase 15 houses hoffman, 1970b four Bear (39DW2) four Bear focus 1 house hurt et al., 1962 oahe Village (39hu2) ? 2 houses Johnson (sDaRC) Mush Creek (39hu5) ? tests Johnson (MWaC); Bleacher, 1980 Pierre school 1987 (39hu10) ? (5) Excavations Johnson (sDaRC; aCaL) Pierre school 1990 (39hu10) ? (5) Excavations Johnson (aCaL) 39hu22 ? tests/surface Johnson (nMnh) spotted Bear (39hu26) spotted Bear focus 5 houses hurt, 1954a steamboat Creek (39Po1) ? tests/surface Johnson (MWaC, nMnh) Rosa a (39Po3) Le Beau/Bad River Ph. 2 houses hurt, 1974 artichoke Creek (39sL1) ? surface Johnson (nMnh) sully Middle (39sL4) ? (5) Multiple houses Johnson (nMnh) sully Late (39sL4) ? (5) Multiple houses Johnson (nMnh) Mobridge (39WW1) ? surface Johnson (nMnh); Parmalee, 1979 Larson 1-4 (39WW2) Le Beau Phase 7 houses falk and Johnson, in prep. spiry-Eklo Late-Early (39WW3) Le Beau focus 2 houses Baerreis & Dallman, 1961 swan Creek B (39WW7) Le Beau focus (5) 1 house hurt, 1957 swan Creek C+D (39WW7) Le Beau focus (5) 2 houses hurt, 1957 39WW301 akaska/Le Beau f. tests hurt, 1957 knIFe rIver Phase Deapolis (32ME5) Knife River Phase surface Lehmer et al., 1978 amahami (32ME8) Knife River Phase 4 houses Lehmer et al., 1978 Rock Village (32ME15) Knife River Phase 13 houses Lehmer et al., 1978 nightwalker?s Butte (32ML39) Knife River Phase 28 houses Lehmer et al., 1978 ProtohIstorIC Cheyenne Biesterfeldt (32RM1) PCC 7 houses Wood, 1971 late ProtohIstorIC/hIstorIC arIkara greenshield (32oL17) ? tests Johnson (uMC); nicholas & Johnson, 1986 fire heart Creek Late (32si2) ? (2) 6 houses Lehmer, 1966 Leavenworth (39Co9) ? (5) 7 houses Johnson (nsM); Krause, 1972 Blue Blanket island (39WW9) ? 1 house stephenson, 1969 1 Component abbreviations: E = Early, M = Middle, L = Late. single occupations at multiple component sites are frequently designated by letter codes (e.g. a, B, C, D). 2 taxonomic abbreviations: Variant assignments and taxonomic abbreviations follow Lehmer (1971). iMM=initial Middle Missouri; EMM=Extended Middle Missouri; tMM=terminal Middle Missouri; iC=initial Coalescent; EC=Extended Coalescent; PCC=Post-Contact Coalescent. a=aspect; f=focus; Ph=phase; h=horizon. taxonomic as- signments are from site reports. numbers in parentheses under taxonomic assignment refer to additional components present at a site and included in this analysis. 1=iMM, 2=EMM, 3=tMM, 4=iC, 5=EC, 6=PCC. 3 Extent of field Work: four levels of fieldwork are included. surface are those collections from surface cutbank or beach areas. tests are usually small, limited scale investiga- tions. Excavations represent more extensive excavations than tests but did not result in the discovery of any houses. house excavations are denoted by the number of complete or partial houses excavated. additional units were often excavated along with houses at the same site. 4 source of Data: Mt=Marion travis Collection; nMnh=national Museum of natural history (smithsonian institution); nsM=nebraska state Museum; nshs=nebraska state historical society; sDaRC=south Dakota state archaeological Research Center; MWaC=Midwest archeological Center (national Park service); unL=university of nebraska-Lincoln; Mhs=Minnesota historical society; uMC=university of Missouri-Columbia; uMM=university of Minnesota-Minneapolis; aCaL=augustana College archeology Laboratory. all references to Johnson refer to Craig M. Johnson, the author of this report. semi-colon between multiple sources indicates first source for ceramic data, second source as additional background. Dash between sources indicates use of both as sources of ceramic data. 4 4 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y a rim sherd count, but this information is rejected in favor of the author?s personal data. sources separated by a dash indicate that both are independent samples used in the tab- ulation of rim frequencies. Ceramic assemblages inspected by the author are currently curated by a number of institu- tions, with several housed at Marion travis? residence in Mobridge, south Dakota. the author examined many of the ceramic assemblages during a period of almost 20 years, beginning in 1975. since this time, most of the archeologi- cal collections curated at the Midwest archeological Cen- ter, national Park service in Lincoln, nebraska, have been transferred to the south Dakota state archaeological Re- search Center in Rapid City, south Dakota. there were two major periods when ceramic assemblages were examined by the author, twice in 1982 at the Midwest archeological Center (national Park service) and the national Museum of natural history (smithsonian institution), and again in 1991 and 1992 at the national Museum of natural his- tory, the south Dakota state archaeological Research Center, the university of Minnesota, and the Minnesota historical society. a few ceramic assemblages not included during these periods were examined at various times since 1982. the author privately funded the 1982 effort with the assistance of Carl R. falk, whereas the Repatriation of- fice of the smithsonian institution supported the 1991 and 1992 trips. Most Coalescent tradition assemblages were examined in 1982, whereas the 1991?1992 effort focused primarily on pottery from Middle Missouri tradition com- ponents and the selection of botanical remains for the radio- carbon dating portion of the project. there are a number of ceramic assemblages not listed in table 3. a more complete list of Plains Village components appears in Lehmer (1971, figs. 38, 39, 76, 77, 79, 82, 111, and app. 1). additional small, partially documented, or unlocated collections were made from a variety of survey and testing projects sponsored by the in- teragency archeological salvage Program, the u.s. army Corps of Engineers-omaha District, Logan Museum- Beloit College (Bowers, 1948a), over Museum of south Dakota (sigstad and sigstad, 1973) and other institutions and individuals (Petsche, 1968) . some of these are dis- cussed in a series of published and unpublished reports including Bowers (1930, 1948a), howson (1941), hughes (1955), Kuhn (1961), howard (1962b), hoffman (1963b, n.d.), Caldwell et al. (1964), neuman (1964), Johnston and hoffman (1966), gant (1967), Lehmer and Jones (1968:75), smith and Johnson (1968:41?43), sigstad and sigstad (1973), a. Johnson (1979), C. Johnson (1986), Roetzel and strachan (1986a, 1986b), toom (1990:138? 147), fritzen (n.d.), hillman (n.d.), Jensen (n.d. a,b), Mcnutt (n.d. b?f), Reed (n.d.), Richards (n.d.). two pre- liminary reports (Moerman and Jones, 1966; Jones, 1969) exist for the Cattle oiler site (39st224); only one is listed in table 3. in addition, data from a recently released re- port on the havens site could not be incorporated (sperry, 1995). an older version of the report was used to com- pile the ceramic information from the site (sperry, 1968a). Many of the unpublished reports are in the library at the Midwest archeological Center, national Park service, Lincoln, nebraska. in addition, the initial Middle Mis- souri and initial Coalescent components from the Crow Creek site reported by Kivett and Jensen (1976) could not be used because of insufficient descriptive information and problems separating components (see introduction to ap- pendix a). hanenberger?s (1986) description of the initial Middle Missouri pottery from Crow Creek is used in this analysis. the initial Coalescent component from the farm school site (39Bf220), reported by neuman (1961), also is eliminated because of mixing of components (steinacher, 1983:64?66). the result is to reduce the small number of initial Coalescent villages to five for this very important taxonomic unit. the latest Redbird components also are rejected because they contain large amounts of undeco- rated straight-rimmed vessels, unlike the late Extended and Post-Contact Coalescent in south Dakota (see Wood, 1965:99?103). the assemblage from the Medicine Creek site (39LM2), which contains initial Middle Missouri, initial Coalescent, and Extended Coalescent components (Lehmer, 1971:195), has been in the possession of Eliza- beth Potter henning since the early 1970s and could not be used despite a concerted attempt to examine it. Plains Village components from the Knife, heart, gar- rison, and Little Missouri regions, subjects of recent re- ports (Woolworth, 1956; Cooper, 1958; Metcalf, 1963a, 1963b; Wheeler, 1963; ahler, Weston et al., 1980, ahler, schneider et al., 1980; ahler and Weston, 1981; ahler and Mehrer, 1984; Kuehn et al., 1984; ahler and swenson, 1985), are not formally included in the present analysis, except for the Clark?s Creek, Cross Ranch, White Buffalo Robe, amahami, and Elbee sites. the Deapolis, amahami, Rock Village, nightwalker?s Butte, greenshield, and Bies- terfeldt sites from north Dakota are included in the analy- sis of cord-impressed motifs. the Plains Village sites from the Knife and heart regions, forming the basis of a re- cent synthesis of the Knife River indian Villages national historic site (thiessen, 1993b), also are not formally in- cluded in the present study. the cultural chronology from the area (ahler, 1993b) is incorporated into the settlement history in chapter 7 and is relied upon to help interpret the sequences from south Dakota. n u m b e r 4 7 ? 4 5 Descriptive rim sherD categories Ceramic classification in the Middle Missouri subarea has resulted in more than 200 crosscutting and sometimes duplicate ceramic wares, types, subtypes, varieties, classes, groups, and other categories (Wheeler, 1954; Lehmer, 1971:22?23; Calabrese, 1977; C. Johnson, 1980; Butler and hoffman, 1992). the plethora of pottery categories tends to inhibit comparisons between site assemblages, al- though some broader studies have been accomplished (see chapter 2). the inclusion of additional assemblages in the classification process beyond those originally used to define ceramic types has had a subtle, yet pronounced effect on the actual composition of these groups. as van der Leeuw (1991:22) asserted, the inductive definition of types changes when newly classified artifacts are included within them. the reasons for the explosive growth of ceramic wares and types remains unclear, but the shear number of excavated sites, and their geographic and temporal distributions con- tributed to the problem. the large number of archeologists working in the subarea, the evolving nature of ceramic classification, and the availability at any one time of only one or several ceramic assemblages during the classifica- tion process certainly aggravated the situation (C. Johnson, 1980:1?2). Despite one attempt (C. Johnson, 1980), there were no productive efforts to integrate, reconcile, or revise the ceramic classification systems used in the Middle Mis- souri subarea. Without a single consistent system, any in- tersite quantitative analyses are doomed to failure. it is the intent herein to begin to reconcile the multitude of ceramic classification or grouping systems by reducing the myriad number of types into a smaller and more man- ageable number of groups, termed descriptive rim sherd categories. Because these categories are designed for use in broad cultural comparisons, they represent the efforts of a classic ?lumper? as compared with a more fine-grained ?splitter? approach to ceramic classification. the author uses the latter approach in intrasite ordinations (C. John- son, 1977b; Johnson et al., 1995; falk and Johnson, in prep.; falk, Johnson, and Richtsmeier; in prep.), although analyses that are based upon attributes take enormous amounts of time not available for the present study. a good example of the utility of using attributes in chrono- logical ordering is by ahler and swenson (1993). a recent discussion of attribute versus type approach in microseri- ation was presented by Duff (1996) who concluded that types can provide as accurate results as attributes. one of the results of the lumper approach is to emphasize the affinities between certain ceramic assemblages at the ex- pense of variation between them. this approach also is deemed necessary to maximize the number of components used in the study. in short, the approach is to include as many of the radiocarbon dated components into the ce- ramic ordinations as possible. this is done to evaluate the relationship between ceramic variation and radiocarbon dates, and to provide data for the settlement model pre- sented in chapter 7. another result of the present study is to identify a small number of ceramic assemblages that can be used to address chronology problems, using more detailed attribute-based studies (see chapter 8). the percentages and frequencies of these descriptive categories listed for each component are presented in ap- pendix a. the resulting five data matrices, consisting of 225 components and the frequencies of 53 descriptive rim sherd categories, are the basis of the ceramic ordinations. Examples from most of these categories are illustrated in appendix B. the components in appendix a are ordered by traditionally defined taxonomic units (Lehmer, 1971). the five matrices list components assigned to the following units: (1) great oasis phase of the initial Middle Missouri variant; (2) initial Middle Missouri variant; (3) Extended and terminal Middle Missouri variants; (4) initial Coales- cent variant; and (5) Extended and Post-Contact Coales- cent variants. the latter matrix is further divided into the Extended variant, four post-contact phases (felicia, talk- ing Crow, Bad River, Le Beau), and the ?Late arikara? vil- lage sites. the late arikara sites are included in the analysis of the Bad River Phase. separate intersite seriations, based upon these variants and phases, are the focus of this report. Components assigned to some variants also are segregated by region or locality and are included within their own regional ordinations to partially control for the effects of spatial variation. several components are included in more than one regional ordination to cross tie each one and to serve as internal checks on the orderings. an additional data matrix, listing a variety of cord-impressed motifs from 31 Post-Contact Coalescent components, also is included in the analysis (table a.6). this matrix is based upon motif divisions within one descriptive rim sherd category (straight/curved/flared/outcurved/simple rim forms with cord-impressed lips or rim braces) included in the more in- clusive intersite analysis. Methodological and other consid- erations inherent in the use of ceramic types in ordinations are addressed in chapter 3. the descriptive rim sherd categories employed in this analysis are listed in a hierarchical fashion in table B.1 by variant or phase as they appear in the data matrices. this table also lists the major corresponding previously defined types included within these descriptive categories. these categories follow the traditional and most widely accepted 4 6 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y method for classifying ceramics from the Middle Missouri subarea (Lehmer, 1954a:3; C. Johnson, 1980:2?13). in short, pottery associated with a particular variant or phase is generally assigned to a number of wares based upon rim form. types within each ware are defined on the presence or absence of decoration, the location of decoration, the techniques used in the decoration process, and the result- ing motif or overall pattern. the result is a hierarchy with wares at the top, followed by types based upon decora- tion, or lack of it, and the decoration location, technique, and motif. the descriptive system used herein recognizes a number of these attributes that had been more fully dis- cussed by C. Johnson (1980:37?42), arranged in a hierar- chical system most closely corresponding to the majority of typologies used in the past by Middle Missouri arche- ologists. Examples included within the resulting categories are illustrated in appendix B. the first division in the descriptive system includes two major rim forms: straight/curved/flared/outcurved/simple (abbreviated to straight), and s-shaped/collared/compound (abbreviated to s-shaped). two additional minor forms, rolled rims and bowls, occur in small or moderate quantities in some initial Middle Missouri assemblages. the straight and s-shaped rim forms do not take into consideration the orientation of the rim to the plane of the lip (e.g., straight versus outflared) but rather the nature of the curved-rim surface from the vessel neck to the lip. straight rims are those that have surfaces that do not change direction or have inflection points, as contrasted with s-shaped rims having surfaces that change direction once and sometimes twice, resembling the letter ?s? on profile or cross-section (appendix B). Braced rim forms, used as the defining at- tribute of stanley Braced Rim ware (Lehmer, 1951; 1954a; Lehmer and Jones, 1968), is not recognized here but rather included within the straight rim category because of inher- ent inconsistencies in its definition and difficulties in sepa- rating it from closely related types (Baerreis and Dallman, 1961: 439?441). Rolled rims have no rim heights or very low ones and have round and exterior thickened bulbs of clay at their orifices. Bowl-rim or vessel forms are those that have straight or curved vessel walls up to the lip but lack a separate, definable neck or rim. Rim sherds in each rim form are next divided by the presence or absence of decoration, and whether decoration is present on the lip/lip margins or exterior rim surfaces. no further divisions are made in the rolled or bowl forms, although some typological systems have made additional distinctions. shoulder decoration is not considered in this system because of the fragmentary nature of the vessels and because it has not been used in the classification systems. undecorated rim sherds are assigned to separate descrip- tive categories, whereas decoration technique and motif further subdivide those with lip or exterior-rim decora- tion. two exceptions to this procedure deal with s-shaped rims from Extended Coalescent and Extended Middle Missouri variant contexts. these rims lacking exterior-rim decoration, with or without decorated lips, have usually been classified into one, infrequently occurring type. Rim sherds with straight rims and decorated only on the lip, or on the interior and/or exterior lip margins, are assigned to descriptive categories based upon decoration technique and occasionally by motif as well. there are three major decorative techniques (tool impressing, cord impressing, and finger impressing), all illustrated in ap- pendix B. the differentiation between tool and finger im- pressing has, with a few exceptions (Brown 1974:16?17; steinacher, 1990, table 14), not been made on rim sherds from initial Middle Missouri, Extended Middle Missouri, and initial Coalescent contexts. in these cases, tool- and finger-impressed lips on straight rims constitute a single descriptive category. Pinching, a special case of finger im- pressing, is consistently defined from Extended Middle Missouri assemblages and is maintained in this study. Rim sherds associated with initial Middle Missouri components having crosshatched and incised lips also are maintained herein as a separate descriptive category. tool impressions generally consist of short and relatively narrow diagonal or inward oriented incisions or punctates in a variety of mo- tifs made with a bone, chipped stone, shell, or wood tool. finger impressions are relatively broad decorations applied to the top of the lip or in an alternating pattern on the interior and exterior lip margins, creating a wavy or ?pie crust? effect on the lip. Cord impressions consist of single twisted cords applied a number of times in horizontal, ver- tical, or diagonal motifs on vessel lips. Cord impressions also occur on the exterior rims of many of these straight- rimmed vessels but have not been considered in typological classifications. they are combined with rims with cord im- pressions only on the lip (or rim brace) into a single cord- impressed category. a few straight rim sherds recovered from Extended Coalescent components have cord impres- sions only on the rim exteriors and either undecorated lips or lips decorated with another technique. these rim sherds also are combined with the rims having cord-impressed lips because of the overall similarity of the types. there are a relatively large number of descriptive rim sherd categories within each rim form. these categories are based solely upon exterior rim decoration technique and motif. in cases in which this portion of the vessel is deco- rated, other areas, such as the lip, also may be decorated n u m b e r 4 7 ? 4 7 but are ignored in the classification process, following the procedures used in previous typologies. the over-riding criterion in these cases is rim decoration. straight and s-shaped rims share many common exterior rim decora- tion techniques and motifs. for the purposes of the pres- ent analysis, rim sherds of the same rim form (straight or s-shaped) having different decorative techniques are main- tained as separate descriptive categories, although they may or may not include rims with similar motifs. these incon- sistencies or exceptions, like those previously discussed, are the legacy of ceramic classification in the Middle Missouri subarea. a number of the more common decorative motifs included within the descriptive categories are illustrated in appendix B. a variable used in comparisons among ceramic assem- blages involves the use of rim sherds versus vessel counts. Determining the minimum number of vessels within a par- ticular provenience unit within a site, such as a house, is crucial in the study of other problems, such as site forma- tion processes or intrasite stylistic variations. it is impos- sible to determine the degree to which the assemblages being studied were subjected to vessel reconstruction, although it appears that only the most obvious matches were considered during laboratory processing to be single vessels (steinacher, 1990:120?121; C. Johnson and toom, 1995:249). a number of ceramic assemblages (falk and ahler, 1988:66; steinacher, 1990:120?121; Johnson and toom, 1995:249; falk and Johnson, in prep.; falk, John- son, and Richtsmeier, in prep.) were reduced to smaller vessel counts by systematic matching efforts of conjoined and nonconjoined rim sherds. a comparison of type per- centages from the large Medicine Crow site (39Bf2) as- semblage before and after the matching process indicates that there is little difference between the two data sets. it might be anticipated that the smaller a particular as- semblage, the greater the differences in type or descriptive category percentages between matched and unmatched collections. if this factor is combined with the probable, but unknown, bias in field selection techniques discussed below, assemblage size must be considered in the ordina- tion process. Determining the minimum number of vessels at a particular site is a particularly crucial problem when the assemblage consists of highly fragmented rim sherds that are decorated around the vessel circumference by regular recurring motif segments. for example, great oasis as- semblages have relatively large quantities of rim sherds with horizontally incised or trailed lines encircling the en- tire vessel upon which are imposed diagonal lines, form- ing various triangular motifs at regular intervals. Many rim segments that are classified as horizontally incised in this study are probably from areas between these diago- nal motifs. the result is to artificially inflate the number of rim sherds assigned to the horizontally incised descrip- tive category in highly fragmented assemblages, such as the heath site (39Ln15). another of the many examples involves Post-Contact Coalescent assemblages containing rim sherds decorated only with finger-impressed (pinched) lips. these broad and sometimes shallow impressions often occur at fairly large intervals (4?8 centimeters). this results in the classification of small rim fragments from areas be- tween these impressions into an undecorated category. the author cannot recall any assemblages with a particularly high number of small rim fragments. Because screens were generally not used during excavations (see the discussion below), only the larger rim sherds were recovered. these two examples illustrate the problems in using ceramic assemblages recovered using a variety of field techniques. in the current report, it is assumed that the ef- fects of unequal fragmentation are random and balanced out for the vast majority of the assemblages. the defini- tion of the descriptive categories also has tried to minimize these ?fragmentation effects.? for example, Post-Contact Coalescent assemblages have cord-impressed s-rims with a wide variety of motifs. all of these were combined under a single descriptive category, despite earlier attempts to subdivide them (hurt, 1957:36?44). another factor that may have an effect on the vessel versus rim sherd counts is vessel size. Larger vessels gener- ally have larger orifice circumferences, although this rela- tionship has not been systematically studied for the entire range of variation of constricted-mouth, globular jars most commonly found in Plains Village sites in the Middle Mis- souri subarea. a study by Rogers and Brewster (1990) of Post-Contact period pots associated with arikara villages demonstrated a positive curvilinear relationship between orifice diameter and volume. Larger vessels, therefore, are probably more likely to break up into a larger number of rim fragments, with all other factors, such as rim thickness, vessel use, and post-depositional factors, being equal. Ves- sel size and breakage rates may tend to cancel one another, with larger, thicker vessels less likely to break into more pieces after they are discarded and enter the archeological context than smaller, thinner ones. the way vessels are used also is related to their break- age rates. a number of ethnoarcheological studies have concluded that large storage vessels are broken less often than those used for cooking or other purposes (see chapter 3). if there is another relationship between vessel size and the decorations applied to the rims, differential breakage 4 8 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y rates and the corresponding archeological assemblages might be very different than an ?in use? village assem- blage. the relative frequencies of vessels of particular styles or types used at any particular time might be very different from the broken vessels left by the same village group in even relatively short periods of time, perhaps on the order of 10 years. in an ideal world, all of the ?non- stylistic related? factors, such as vessel size (as measured by orifice diameter or rim thickness), might be controlled in comparisons of style from one village to the next. Despite the fact that such an analysis is beyond the scope of the present study, data relating to variation in orifice diameter by the descriptive categories used in this study was gathered for a number of sites, presented in table 4. this means that there is some minor varia- tion, not only among sites but also between descriptive rim sherd categories. in general, s-shaped rims from Lower grand, Walth Bay, and Larson are larger than their straight-rim counterparts, whereas tool-impressed straight rims are usually larger than the more highly dec- orated, cord-impressed or incised straight rims. several reasons for the small size of cord-impressed, straight- rimmed vessels may relate to their use. these vessels have a higher incidence of handles, making them more suitable for carrying and suspending. these activities may require small pots. in addition, some of these vessels may be in- TabLe 4. Mean orifice diameters (centimeters) by descriptive rim sherd category for six Plains Village components (sample size in pa- rentheses). (iMM =initial Middle Missouri component, EMM = Extended Middle Missouri component.) Lower Walth Medicine Cattle Oiler Descriptive rim sherd grand bay Larson Crow iMM eMM Coalescent Tradition s-shaped/Cord impressed Rim 18.0(37) 16.3(15) 16.9(72) ? ? ? s-shaped/horizontal-Diagonal Rim 18.0(123) 17.3(120) 15.9(56) ? ? ? s-shaped/tool-finger impressed Rim 19.0(1) 13.8(14) ? ? ? ? s-shaped/undecorated 17.4(7) 16.8(4) 16.4(7) ? ? ? s-shaped/all Categories ? ? ? 15.6(14) ? ? subtotal s-shaped rims 18.0(168) 16.9(153) 16.5(135) 15.6(14) ? ? straight/Diagonal incised Rim ? ? ? 10.1(1) ? ? straight/horizontal incised Rim 14.5(68) 14.2(78) 17.0(32) 12.2(4) ? ? straight/Cord impressed Lip 13.2(13) 10.8(8) 14.2(586) 13.3(39) ? ? straight/tool impressed Lip 16.0(48) 15.5(73) 16.7(289) 14.7(397) ? ? straight/finger impressed Lip 21.0(1) ? 17.1(737) 18.8(5) ? ? straight/undecorated 16.2(14) 11.8(19) 16.7(167) 12.5(24) ? ? subtotal straight rims 15.1(144) 14.3(178) 16.1(1811) 14.5(470) ? ? total all rims 16.7(312) 15.5(331) 16.1(1946) 14.5(484) ? ? Middle Missouri Tradition s-shaped/horizontal Cord impressed Rim ? ? ? ? 17.2(29) 10.1(1) s-shaped/triangular Cord impressed Rim ? ? ? ? 16.0(16) ? s-shaped/horizontal incised Rim ? ? ? ? 18.9(8) ? s-shaped/triangular incised Rim ? ? ? ? 18.0(8) ? s-shaped/Cross hatch incised Rim ? ? ? ? 25.5(2) ? s-shaped/undecorated Rim ? ? ? ? 11.0(1) ? subtotal s-shaped rims ? ? ? ? 17.4(64) 10.1(1) straight/tool-finger impressed Lip ? ? ? ? 19.3(13) 19.7(12) straight/Cross-hatch incised Lip ? ? ? ? 14.6(13) ? straight/Cord impressed Lip ? ? ? ? 12.0(1) ? straight/undecorated ? ? ? ? 18.5(36) 17.8(12) subtotal straight rims ? ? ? ? 17.8(63) 18.8(24) Rolled Rim ? ? ? ? 14.0(4) ? total all rims ? ? ? ? 17.5(131) 18.4(25) n u m b e r 4 7 ? 4 9 tended for burial, as it is this author?s impression that the majority of vessels that accompany burials during the protohistoric period are straight rimmed and cord im- pressed. Because of lower sample sizes, the intertype vari- ation in table 4 is less conclusive among the two Middle Missouri components at Cattle oiler, although these pots are larger than their Coalescent counterparts. Variation in size also is apparent among the four Coalescent vil- lages of Lower grand, Walth Bay, Larson, and Medicine Crow. the Post-Contact Coalescent assemblage from h.P. thomas (39st12) has a comparable overall orifice diameter of 15.8 centimeters (n = 174). a more striking difference in inferred vessel size is apparent when sev- eral Extended and terminal Middle Missouri and heart River phase villages are compared with those above. ori- fice diameters from these former villages include Bendish (32Mo11) at 25.8 centimeters (n = 28), huff at 19.1 centimeters (n = 41), nailati phase at 22.0 centimeters (n = 184), and heart River phase at 21.7 centimeters (n = 244) (Wood, 1967:64?69; thiessen, 1976, table 19; C. Johnson, 1983, table 9.13). the greater sizes of these vessels compared with their Coalescent counterparts may reflect household size or simply the degree of rim flare. it is unclear at this time what the role that overall vessel shape plays in these calculations and if the ratio of orifice and maximum vessel diameters differs between these sites (i.e., wide- versus narrow-mouthed jars). only examina- tions of complete or nearly complete vessels (i.e., those for which maximum vessel diameters are known) will re- solve this issue. raDiocarboN samples the selection of samples to be submitted for radio- carbon dating for this study was determined by a number of explicit criteria that maximized the ability to assign absolute time scales to the ordering of components de- rived from the ceramic ordinations. the selection process is detailed in chapter 5. the intent was to avoid problems of cultural association, laboratory procedures, and single dates in the selection process (thomas thiessen, pers. comm., 1992; toom, 1992a:121?122). it is desirable to obtain many samples of short-lived botanical remains, such as corn, squash, seeds, and ceramic vessel residues, so as to minimize the dating of aged samples, and also to allow for examination of other attributes of these samples (see chapter 5). first, a priority was placed on samples from components that were suspected of falling at either end of the temporal range within each variant and those located at the maximum geographic limits of each taxo- nomic unit. it was the intent to date those villages in each variant that were suspected of falling at the extreme tem- poral and geographic limits of these taxonomic units. ide- ally, this would allow for the dating of both ends of each regional variant and would tie these into the ceramic or- dinations of components so that a relatively accurate tem- poral sequence of a much larger number of villages could be established. as a rough guide, a list of components assigned to each variant by region (table 5) was compiled prior to the radiocarbon sample selection process on the basis of previous ceramic ordinations and existing radio- carbon dates. second, archeological context is carefully considered in the sample selection process because many villages have long occupational sequences, some being inhabited by native american village groups of different cultural traditions. to be able to relate the results of the radio- carbon dating to the ceramic sequences requires an abil- ity to associate a particular dated sample with a specific ceramic component assemblage recovered from each site. this problem has sometimes been overlooked in past dat- ing efforts (thomas thiessen, pers. comm., 1992; toom, 1992a:121?122). the sample-selection process focuses on those sites that have a single component. When this is not possible, only samples from contexts that are assigned to a specific component are chosen for dating. finally, most samples are selected from components that could be included in the ceramic ordinations. in addi- tion, samples from components that have relatively large ceramic assemblages are chosen instead of those from smaller assemblages. furthermore, sites that have pub- lished or unpublished reports are generally favored instead of unreported ones. this is important because understand- ing the context of each dated sample is made clearer when a formal site report exists. these three criteria for sample selection were compromised once an examination of avail- able short-lived botanical and ceramic vessel residue spec- imens proceeded. some of the ideal village components could not be dated or subjected to multiple dates because datable material was either absent or of insufficient quan- tity (e.g., arzberger, Lynch, Black Widow Ridge). at sites with multiple prehistoric and post-contact occupations, most of the botanical remains are associated with the later component. these late villages cannot be reliably dated (i.e., Cheyenne River, Dodd sites). some multicomponent sites are so badly mixed or poorly understood so as to preclude associating some datable samples with one com- ponent or another without an intensive reanalysis of each one (e.g., Crow Creek). 5 0 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y TabLe 5. List of Plains Village tradition components with a potential for radiocarbon dating listed by region, variant, and estimated date based on existing radiocarbon dates and ceramic seriation. extended Middle initial Middle Missouri Missouri initial Coalescent extended Coalescent BIg Bend regIon Early unknown time unknown time Early sommers (st56)* Cattle oiler (st224) Medicine Creek (LM2) Bowers LaRoche (st232)* akichita (Bf221) useful heart (LM6) overs LaRoche M/E (st9) Cattle oiler E/M (st224) standing Bull (hu214) Middle Middle Middle King (LM55) Lynch M/L (25BD1) Redbird ii (25ht2) Chapelle Creek C&D (hu60) Crow Creek (Bf11) hu15/hu241/hu223 Langdeau (LM209)* Late Late Late Late Dinehart (LM33) Durkin (st238) arzberger (hu6)* spain (LM301) Jandreau E/L (LM225) stricker B (LM1) Jiggs thompson (LM208)* McClure (hu7) Pretty head L/M (LM232) scalp Creek (gR1) Crow Creek M/L (Bf11) 39BR201 hu204/LM47/LM201/LM31 Bad-Cheyenne regIon Early Early Early Breeden a (st16)* Cheyenne River (st1) no heart Creek (aR2) Dodd E/L (st30) thomas Riggs L/E (hu1) Black Widow (st3) Cheyenne River (st1) sL24/sL8/sL7/st10 Middle sully (sL4) sL23 Late Late McKensey (aR201)* sully (sL4) indian Creek (st15) gettysburg (Po209) st45/sL2 unknown Black Widow Ridge (st203) grand-Moreau regIon Early Early travis i (Co213) Lower grand (Co14) Potts (Co19) Molstad M/E (DW234) DW230/DW217/DW254 unknown time Calamity Village (DW231) Middle Walth Bay (WW203)* WW10/WW302 Late hosterman (Po7) WW7/WW300/WW301 CannonBall regIon Early helb (Ca208)* Jake White Bull (Co6)* Paul Brave E/M (si4) Middle Middle Bendish (Mo2)* Demery (Co1)* havens (EM1)* Ben standing soldier L/M (si7) n u m b e r 4 7 ? 5 1 fielD sampliNg strategies Because most of the ceramic assemblages and dat- able material employed in this study were collected during the 1950s and 1960s, a brief discussion of field collecting strategies is in order. the major focus of most of the past research in the Middle Missouri subarea was on develop- ing culture-historical chronologies and taxonomies (Wood, 1969). Ceramic variation and differences in house form were identified early on as key components in these efforts during the salvage period in Middle Missouri archeol- ogy (see Lehmer, 1954a:118?159). the work of Donald Lehmer and other pioneering archeologists, such as Car- lyle smith and Wesley hurt, emphasized the primacy of ceramic and architectural variability instead of other data in their culture-historical reconstructions, and shaped the way archeologists excavated and interpreted sites for the next generation. at the time, this perspective appeared to answer all the relevant questions of the currently accepted theoretical framework. the emphasis on architecture re- sulted in the excavation of houses, with little or no work accomplished in other site areas. this point is particularly important for the present study, because a number of extra- mural contexts, such as midden deposits, can provide use- ful stratigraphic information crucial in the study of ceramic change without many of the post-depositional disturbances associated with activities, such as house construction on top of older cultural deposits. the utility of a small num- ber of test excavations strategically placed within stratified deposits to develop site chronologies is demonstrated by work at the Knife River indian Villages national historic site (ahler, 1993c:73?76). the emphasis on architecture and ceramics meant that screening of site matrix was not particularly important in the recovery of the data in question, especially considering the voluminous amounts of large pieces of pottery recov- ered at Plains Village sites. Excavation techniques involved shovel skimming of general site matrix and troweling fea- tures. a popularized version of these techniques appeared in Deetz (1971:57) and was evaluated by ahler, falk et al. (1995:30?33). Certainly, the burden of salvaging infor- mation from hundreds of sites containing large quantities of cultural materials with limited funding led to excava- tion techniques that would generally not be used today or in areas at the time when cultural materials were not so abundant (see Lehmer, 1971:14). there are only a few instances known to the author in which screens were em- ployed (Kivett and Jensen, 1976:20; owsley et al., 1977, fig. 1; ahler, falk et al., 1995:30), and these occurred in unique situations. the result of these field techniques was that large quantities of small chipped stone, ceramic, and bone artifacts were not recovered, in addition to botani- cal remains, trade artifacts, and unmodified vertebrate re- mains. shortly after the interagency archeological salvage Program was terminated in 1969, systematic screening was added to the repertoire of field recovery techniques in the Middle Missouri (Lehmer et al., 1978:149; falk and ahler, 1988:26; ahler et al., 1991:108). in those instances in which 1/4 inch hardware cloth and 1/16 inch window screen was used, enormous and sometimes overwhelming quantities of cultural material were recovered, creating their own mas- sive problems of sorting, processing, and data recordation. occasionally, even small seeds were recovered using these techniques (haberman, 1984:273). these changes in field south Cannonball (si19)* fire heart Creek (si2)* tony glas (EM3) Late (terminal Middle Missouri) huff (Mo11)* shermer (EM10) * Denotes agreement of 14C and ceramic seriation dates. E/M, M/L, E/L, M/E, L/M, L/E denote disagreement between 14C and seriation dates. first letter designates radiocarbon date, last letter seriation date. E = Early, M = Middle, L = Late. Components are listed by group according to priority. the ones at the top of each group have the largest sample sizes, are single component sites and/or have been reported. the following sites, in the Knife-heart and garrison regions, are not included in this list: closely related Lower Loup sites in nebraska; over focus sites in southeastern south Dakota; Mill Creek sites in northwestern iowa; the Cambria phase sites in southwestern Minnesota;, great oasis sites in south- western Minnesota, northwestern iowa, and northeastern nebraska; and select Central Plains tradition sites in nebraska. sites separated by slashes (/) contain small ceramic samples. 5 2 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y procedures were designed to answer questions that went be- yond the traditional concerns of tribal identity, taxonomic assignment, and cultural chronology, focusing on new ap- proaches to describing and analyzing cultural materials. the Walth Bay (39WW203), Lower grand (39Co14), and helb (39Ca208) sites served as a springboard for ex- amining lithic raw material procurement (ahler, 1977b), a systematic approach to stone tool function (ahler, 1975b), and the use of unmodified vertebrate fauna to study cli- matic change and subsistence-settlement patterns (falk, 1977; semken and falk, 1987, 1991). at this time, it is impossible to determine how these varying field methods affected the ceramic data used in this study. there is no question that the lack of systematic screening resulted in the loss of many, if not most, arti- facts from the sites. a comparison of the estimated yield of vessels greater than 0.5 inch recovered from the screened matrix of house 12 excavated during 1969 at the Lower grand site, 0.55 vessels/square foot (2,273 vessels/4,150 square feet excavated), and all vessels from a typical un- screened si-RBs excavation at the Medicine Crow site, 0.20 vessels/square foot (2,306 vessels/11,351 square feet excavated from 14 houses), indicates a considerable loss of data. from these figures, the average yield of ceramic vessels/house at Medicine Crow is 165, a figure much lower than the 2273 vessels recovered from house 12 at Lower grand. it seems that on an intuitive level there would not be a systematic field selection of rim sherds assigned to a particular type or descriptive category instead of other ones, particularly when large assemblages are considered. field selection might result in a bias in favor of one or more ceramic categories in smaller assemblages. this situ- ation also might apply to surface or beach samples, al- though this does not appear to be the case with Marion travis? collections (ahler, 1977a:58, 62). a comparison of smaller items, such as chipped-stone arrow points, could not be performed on these same sites, although data is available from the two separate excava- tion programs at the Lower grand site (falk and ahler, 1988, table f.1). these are the 1962?1964 excavations by alfred Bowers and the 1969 excavations by the uni- versity of Missouri. the earlier excavations appear to be unscreened, whereas artifacts from the latter fieldwork were recovered, using 0.25 inch and finer hardware cloth. a total of 228 arrow points were recovered from Bow- ers? excavations, consisting of 3.5 houses and another area equivalent in size to a house. this contrasts to a total of 1260 complete or partial points recovered from house 12 in 1969 (stanley ahler, pers. comm., 1993). although pre- cise figures cannot be calculated, it is apparent that smaller artifacts, such as projectile points, are under-represented to an even greater degree than ceramics, perhaps by a magnitude of about 20 times. Even smaller artifacts, such as trade materials (glass beads, small pieces of brass, cop- per, or iron) found in protohistoric contexts, are probably even more underrepresented in extant collections because of field sampling methods. a casual inspection of the num- ber of glass beads recovered from the recent excavations at the Knife River indian Villages national historic site (ahler and Dreybred, 1993) and any Post-Contact period assemblage from non-screened excavations anywhere in the subarea illustrates this point. finally, unscreened exca- vations tended to be highly selective in favor of patterned tools (C. Johnson, 1995: 267). 5 Radiocarbon Dating Results stanley a. ahler, Craig M. Johnson, herbert haas, and georges Bonani review of criteria for sample selectioN as a prelude to the presentation of the radiocarbon dating program re- sults, it is worthwhile to briefly review the criteria already discussed in chapter 4 that guided site and sample selection. Because funds for radiocarbon analy- sis conducted as part of this program were limited, a strategic sampling and site selection process was followed in which some sites and components were given priority over others. several general and specific criteria served to guide the selection process. these are ideal parameters, and in some cases they were changed. these modifications are explained in the discussion of specific samples and components subjected to dating. 1. site selection focused on dating Plains Village sites and collections housed in the national Museum of natural history (nMnh). this provided maximum information regarding sites subject to actions under the smithsonian institution Repatriation Program, which meant that the sites under consideration were pri- marily in south Dakota. 2. sites chosen for dating were only those from which an analyzable ceramic sample also exists (generally n>50 rim sherds). this allowed maximum integra- tion of the radiocarbon dating results with the ceramic ordination and seriation aspect of the Plains Village chronology program. Because far more components were included in the ceramic analysis than in the radiocarbon study, the radio- carbon sample selection was seen as a precision tool to be used with great selec- tion and as a critical adjunct to the ceramic ordination program. 3. site selection considered key components in the Cannonball region in north Dakota, but did not focus on the heart and Knife region sites in north Dakota. few heart region sites have been investigated, and the chronology for sites in the Knife region is reasonably well developed through the focused chronometric and research program centered at the Knife River indian Villages national historic site at the mouth of the Knife River (ahler and haas, 1993). the project?s limited resources were best utilized by concentrating on the large series of sites and samples in the less well-dated regions of south Dakota. 4. in the interest of thoroughness, intended to complement the scope of the ceramic analysis program, priority in radiocarbon sample selection is given to 5 4 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y components that (1) lie near the limits of the known geo- graphic range for a given cultural variant (e.g., Extended Middle Missouri, initial Coalescent), and (2) are thought to lie near the respective early and late temporal extremes previously assigned to a given variant based upon all previ- ous information and preliminary ceramic seriation results. thus, defined variants are accepted as a midlevel organi- zational unit useful for both the ceramic and radiocarbon analyses. the focus is on refining the chronological limits of each unit and the potential interactions among spatially proximate components of various units. By taking all of these things into account, a list presented in table 6 was developed as a working set of sites considered for radio- carbon sample selection. several additional site-specific and sample-specific factors were considered: 5. in general and if available, multiple samples were run from each dated site, thereby providing internal checks on date reliability. 6. additional attention was given to dating at a small number of sites that were qualitatively superlative in terms of the type of overall archaeological information avail- able for further study. these included four village sites that were subjected to excavations of various scale with fine-screen artifact recovery: (1) Lower grand (39Co14) (falk and ahler, 1988); (2) Walth Bay (39WW203) (falk and ahler, 1988); (3) helb (39Ca208) (falk and Cala- brese, 1973, falk and ahler, 1988); and (4) Jake White Bull (39Co6) (ahler, 1977a). Each of these villages was subjected to previous dating programs. additional dating allowed a check on the validity and reliability of existing date series considered acceptable (Lower grand and Walth Bay), as well as existing radiocarbon date series that were internally inconsistent and badly in need of more complete reevaluation (helb and Jake White Bull). several samples were dated from each site to provide a large enough set for comparison with previously run dates. 7. sample selection focused on short-lived organic materials, such as carbonized seeds, corn, and other food plants, grasses, and twigs and small diameter branches that occur occasionally in burned house roof-fall debris. Po- tentially aged material, such as wood in large posts, were completely eliminated, and charcoal potentially from in- terior rings in large trees was avoided if possible, in order to minimize problems with dating materials whose age is not closely tied with the time of cultural use of the dated material. 8. Relatively routine, as well as somewhat experi- mental aMs techniques were applied to date previously unstudied samples of very small size. this included rela- tively routine aMs dating of seeds and vegetal remains in good cultural association (preferably cultigens or charred wild plant food remains) and somewhat more exploratory aMs dating of carbonized organic residues adhering to pot sherds. Charred residues occurred in abundance on freshly excavated potsherds, and such residues were con- sidered to be a potentially untapped source of firmly asso- ciated cultural material residing in the available museum collections. 9. Whenever possible, a single lab was used for radio- carbon analyses in order to reduce sources of error intro- duced by interlab variability. ahler?s explicit role in the project was to provide rec- ommendations for laboratories to conduct the dating, to be directly involved in sample selection and transmittal, and to assist in evaluation of results. ahler?s unequivocal first choice of labs for conventional radiocarbon dating was the facility run by herbert haas at southern Methodist uni- versity (sMu). this choice was based upon the excellent consistency and reliability of conventional dating provided by haas in the past, as was particularly evident in the re- sults of the radiocarbon dating program conducted for the Knife River indian Villages nhs project (ahler and haas, 1993). this choice also was based upon haas? attention to detail, particularly in the domain of sample pretreatment, which previous work had demonstrated to be a critical step in achieving good radiocarbon-dating results. mechaNics of sample selectioN aND processiNg after some discussion between ahler and haas, haas agreed, despite a heavy workload, to take on the conventional dating aspects of the program at his sMu lab. after further discussion haas agreed to carry out all pretreatment for aMs samples, with all aMs work to be done under the direction of george Bonani at Eth- h?nggerberg in Zurich, switzerland, with haas serving as coordinator for all aMs work done at the Eth lab under the Plains Village dating (PVD) project. a multi-step approach was taken for the sample selec- tion and dating process: (1) Johnson and ahler made gen- eral sample selections based upon availability of materials and the sample and site criteria listed above; (2) ahler ex- amined more closely actual material remains available for dating and prepared both conventional and aMs sample materials, transmitting them to haas; (3) haas conducted all conventional radiocarbon work at the sMu lab; (4) haas removed carbonized residues identified by ahler from potsherds and conducted all pretreatment of all n u m b e r 4 7 ? 5 5 TabLe 6. intermediate working list of sites and components to be dated prior to the start of sample selection, february 1992. sta- tus of datable materials column added after inspection of collections. (Collection location abbreviations: sDaRC=south Dakota ar- chaeological Research Center, Rapid City; nMhn= national Museum of natural history, Washington, D.C.; nshs=nebraska state historical society, Lincoln; MWaC=Midwest archeological Center, nPs, Lincoln; unD=university of north Dakota, grand forks; unL=university of nebraska, Lincoln; shsnD=state historical society of north Dakota, Bismarck.) Variant and relative Site Collection place in variant Site name number location Status of datable material Initial Middle Missouri Early *Jandreau 39LM225 sDaRC none exist Early *sommers 39st56 nMnh no addtional samples known Early swanson 39BR16 sDaRC none located ? *Cattle oiler 39st224 nshs unknown Late *Pretty head B 39LM232 nMnh none exist ? *Dodd 39st30 nMnh no samples exist Middle *fay tolton 39st11 sDaRC unknown Early Jones Village 39Ca3 sDaRC abundant conventional exist Extended Middle Missouri Early? *Cattle oiler 39st224 nshs unknown Early? thomas Riggs 39hu1 sDaRC none located ? *Cheyenne River 39st1 nMnh none exist Late McKensey 39aR201 nMnh unknown ? *Calamity Village 39DW231 nMnh unknown Early Paul Brave 32si4 shsnD several conventional exist Early *helb 39Ca208 MWaC abundant aMs/conventional exist Early *Jake White Bull 39Co6 unD aMs samples exist Late? *havens 32EM1 shsnD several aMs/conventional exist Terminal Middle Missouri ? *shermer 32EM10 shsnD several aMs samples exist Initial Coalescent Early Black Partizan B 39LM218 nMnh unknown-mixed components Late *arzberger 39hu6 sDaRC no samples exist ? *Lynch 25BD1 unL none exist ? useful heart 39LM6 nMnh unknown-not examined Extended Coalescent Early *over?s La Roche 39st9 nMnh unknown Late scalp Creek 39gR7 sDaRC none located-mixed components Late McClure 39hu7 nMnh unknown-not examined Mid/Late *sully 39sL4 nMnh none exist Early *Lower grand 39Co14 MWaC abundant aMs/conventional exist ? *Potts 39Co19 nMnh none exist Middle *Walth Bay 39WW203 MWaC abundant aMs/conventional exist Late *hosterman 39Po7 nMnh abundant maize-quality doubtful Late? *Demery 39Co1 shsnD one aMs/one conventional exist Unkown Late hintz 32sn3 nMnh Potential aMs exists * = sites actually selected for dating. 5 6 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y aMs samples (sherd residue and otherwise), forwarding pretreated materials to georges Bonani; (5) Bonani con- ducted all aMs dating in the Eth-hoenggerberg lab, for- warding all aMs results to haas; and (6) haas reported all conventional and aMs results to ahler for wider proj- ect dissemination. initially, it was estimated that available funds and an- ticipated sample availability would allow for dating ap- proximately 18 conventional samples and 23 aMs samples. several additional samples were added to the program when we were informed by the omaha District of the u.s. army Corps of Engineers (usaCoE) that additional financial support was available for dating samples from archaeologi- cal collections belonging to the Corps (specifically, samples from the Lower grand, Walth Bay, helb, and Jake White Bull site collections). the usaCoE did not maintain this commitment, and the necessary financial support was pro- vided by the Department of anthropology at the university of north Dakota-grand forks (ahler?s home institution at the time). as it turned out, the actual selection of samples and the choice between conventional and aMs dating was determined not only by funding constraints but in equal measure by the availability of dateable sample materials from key sites. By the end of the program sponsored by the smithson- ian institution, 15 conventional samples were dated at the sMu lab and 42 samples were submitted for aMs dat- ing. as the samples were prepared and submitted by ahler to haas, they were assigned sequential sample numbers for tracking purposes: PVD-1 through PVD-57 (PVD = Plains Village Dating). Because of complications stemming from southern Methodist university?s decision, midway through this project, to cease operation of herbert haas? radiocarbon operation, 38 rather than 42 aMs samples were actually dated at the Eth lab. in the end, 54 dates were produced on 53 samples in the original PVD series (PVD-1 through PVD-14, PVD-16 through PVD-53, with PVD-37 dated twice, and PVD-56). the sample selection process conducted specifically for the smithsonian started early in 1992, with the last dates from this group of 57 (PVD 1-PVD 57) samples reported by the labs in february 1994. table C.1 in appendix C provides an inventory of information about all 57 prepared and submitted samples. ahler and Johnson maintained a continuing interest in developing a more complete chronology for Plains Village sites in the Dakotas, and wished to resolve certain gaps and problems in the date series available in 1994. Consequently, 19 dates were run in 1997 after the completion of the Plains Village Dating program, as reported by C. Johnson (1996) to the smithsonian. these dates also are listed in table C.1, appendix C (PVD-58 through PVD-76), and are discussed here as having been developed under the PVD program. this continued dating work was supported by funding from the Department of anthropology, university of north Dakota, grand forks. additional dates were run for the Jake White Bull (39Co6), Jones Village (39Ca3), Vanderbilt Village (39Ca1), shermer (32EM10), Paul Brave (32si4), and huff (32Mo11) sites, all assigned to the Middle Missouri tradition. the continuing work was managed in the same fashion as the early part of the program. herbert haas con- ducted all conventional dating at his lab, now relocated at the Desert Research institute in Las Vegas, nevada (these conventional dates bear DRi lab numbers in table C.1). herbert conducted pretreatment of all aMs samples, with accelerator dates again produced by georges Bonani at the Eth-h?nggerberg facility in Zurich. table C.1 is organized by recognized cultural variant according to Lehmer?s (1971) taxonomic system. for each sample listed, the following information is given: (1) recog- nized cultural variant; (2) site name and number; (3) PVD number assigned to the sample; (4) institutional source of the material selected for dating; (5) field or lab catalog number(s) previously attached to the specific specimen or material used as a sample; (6) within-site provenience of the sample material; (7) material submitted for dating (e.g., charred wood, maize, charred residue); (8) more detailed information about the size of the sample and how it was selected, derived, or prepared; and (9) a brief note about why that sample or component was selected for dating. We conclude with a few words about sample avail- ability and some of the sites targeted for study but left undated. Based upon firsthand observations of abun- dant charred residues on recently excavated village pot- tery samples, ahler and Johnson entered the project with the anticipation that carbonized residues on sherds would likely be the panacea for aMs radiocarbon dat- ing of Plains Village museum collections. it was expected that residue-bearing sherds would be abundant in the old collections under consideration. With surprise and disap- pointment we learned that the actual availability and qual- ity of residue-bearing sherds was very low, mostly because of the intensive scrubbing and cleaning processes that the pottery collections of this era had been subjected to. it was clear that almost all rim sherds had been routinely and intensively cleaned to enhance examination of decora- tive details and for illustration. in addition, some sherds were subjected to acid baths to remove carbonate depos- its (Jake hoffman, pers. comm., 1992). the few sherds that did retain residues usually had very thin, low qual- ity deposits tightly adhering to one sherd surface. in such n u m b e r 4 7 ? 5 7 circumstances it was difficult to remove enough material for dating without danger of removing some of the clay matrix, as well. Most of the suitable residue deposits oc- curred as small remnants imbedded in decorative inden- tions or incisions on rim sherds, apparently small, deeply seated bodies that had accidentally escaped the intensive sherd scrubbing process. a few site collections contained additional short-lived, aMs-datable, botanical remains in some abundance, whereas others contained nothing suitable for further ra- diocarbon analysis. among the former group are Lower grand, Walth Bay, helb, and Jake White Bull that during excavation were all subjected to fine-screen field recovery procedures that retrieved abundant charred food remains and other datable materials. among the latter are several si-RBs era collections that contain little more than cleanly scrubbed sherds and occasional remnants of house posts. table 6 also presents a listing of all site collections care- fully examined or considered in the present sample selec- tion process, giving a brief note on the suitability of the collection for further radiocarbon dating using presently available techniques. it is noted that the aMs samples submitted in this study generally consisted of a fraction of available charred seeds or maize cob and cupule frag- ments. in almost all cases, an equal or greater amount of aMs-datable material was retained for future study. such was not the case for aMs sherd residue samples; virtually all available residue was removed from the sample sherds for study. Backup charred seed/cupule/cob samples can generally be assumed to exist for all dated samples listed in table C.1; such backups are not noted in table 6. evaluatioN of results of the raDiocarboN DatiNg program table C.1 presents a reference to the radiocarbon labo- ratory number (date number) assigned to the 74 dates that were run on the 72 samples and the conventional radio- carbon age reported by the lab (corrected for 13C carbon isotope fractionation). table 7 presents more complete in- formation on the radiocarbon dating results, with site and site samples listed in the same order as in table C.1. table C.2 lists all dates, PVD or non-PVD, considered in this report (PVD dates are in bold). it also presents the carbon isotopic fractionation value or 13C value reported for each sample. the chi-squared test as provided in the CaLiB 3.0.3 program (stuiver and Reimer, 1993) was used to test for sample contemporaneity among two or more samples from a single site or component, prior to calendrical cali- bration of radiocarbon ages. the default value for this test in CaLiB is at the 0.05 level of significance and this was accepted as the standard for this study. in general, after applying the test for sample contemporaneity, a weighted mean date was developed for dates thought to be from the same population within each site or component, fol- lowed by a calendar age calibration of all averaged PVD dates or individual date (if only one was run from a site) using the decadal tree-ring data set (data set 2) curve as provided in the program CaLiB 3.0.3 (stuiver and Re- imer, 1993). single or multiple calendar age cross points or intercepts and one and two sigma calendar probability age ranges are provided in table 7 as generated from the CaLiB 3.0.3 program (Method B). to more fully utilize and interpret the results of the present dating program, it was necessary to relate the cur- rent dating results to existing dates and date series from several of the previously dated sites. in general, the end goal is to compute, when possible, a mean for all dates considered to be reliable and valid from a particular site, and to use this mean value (if more than a single date is involved) in the interpretation of ceramic ordination stud- ies. Because of the complexity of the decision making processes used in determining what are considered to be reliable means in some cases, it is best to proceed from this point forward with a fairly detailed discussion of the dating results and the critical evaluation of those results on a site-by-site basis. this discussion is best prefaced by making clear some assumptions and sources of informa- tion used in the evaluation of all existing dates and date series from individual sites. several elements are particu- larly pertinent: 1. the construction and settlement of an earthlodge village is perceived as a planned and organized commu- nity event, requiring the combined efforts of a relatively large workforce to cut and move substantial amounts of timber and to construct houses (Wilson, 1934). it is highly likely, judging from the planned nature of most villages, that the typical village was constructed during a relatively short period (perhaps a few weeks or less) and was occu- pied relatively quickly by a substantial, aggregated popu- lation. When the village was abandoned, it is likely that most of the village was evacuated in a single event, and that the majority of inhabitants moved simultaneously to construct a new village at a new location. it is therefore assumed that most or all lodges within a typical village were constructed and abandoned in a synchronous man- ner, and are essentially contemporaneous from a radiocar- bon dating point of view. there are exceptions to this, as when villages were continuously occupied during longer T a b L e 7 . R ad io ca rb on d at es f ro m P la in s V ill ag e si te s, s m it hs on ia n in st it ut io n R ep at ri at io n o ffi ce P la in s V ill ag e da ti ng p ro gr am ( co rr ec ti on s ar e ba se d on C a L iB 3 .0 .3 in s tu iv er a nd R ea m er 1 99 3, d ec ad al t re e- ri ng d at as et , c al cu la ti on m et ho d B - p ro ba bi lit ie s) . D at es ( a .D .) a nd r el at iv e ar ea s PV D L ab a ge C al ib ra ti on c ur ve un de r pr ob ab ili ty d is tr ib ut io n V ar ia nt a nd s it e no . nu m be r (r C Y b P) in te rc ep ts ( a .D .) 68 .5 % P 95 .4 % P In It Ia l M Id d l e M Is so u r I Jo ne s V ill ag e (3 9C a 3) 71 D R i- 31 15 89 1 ? 45 11 60 ,1 17 3, 11 88 10 49 ?1 08 9 .2 9 10 31 ?1 14 4 .4 9 11 20 ?1 13 9 .1 4 11 46 ?1 22 4 .4 5 11 51 ?1 21 4 .5 7 12 29 ?1 24 4 .0 3 12 46 ?1 25 7 .0 3 72 E t h -1 60 74 91 5 ? 65 10 73 ,1 07 7, 11 35 ,1 15 5 10 35 ?1 10 1 .4 6 10 03 ?1 00 8 .0 1 11 13 ?1 14 2 .2 0 10 19 ?1 22 5 .9 3 11 48 ?1 16 4 .1 1 12 25 ?1 25 8 .0 6 11 66 ?1 19 4 .1 8 11 98 ?1 20 7 .0 5 73 E t h -1 60 75 94 5 ? 65 10 42 ,1 09 1, 11 18 ,1 14 0, 11 50 10 24 ?1 15 9 .9 9 98 9? 12 22 .9 9 11 83 ?1 18 5 .0 1 12 34 ?1 23 7 .0 0 74 E t h -1 60 76 98 0 ? 65 10 27 10 00 ?1 01 3 .0 9 90 0? 91 7 .0 2 10 16 ?1 07 1 .4 1 99 7? 12 16 .9 8 10 80 ?1 12 7 .3 5 11 36 ?1 15 4 .1 4 75 D R i- 31 16 88 1 ? 37 11 63 ,1 16 9, 11 91 ,1 20 3? 12 05 10 66 ?1 08 6 .1 6 10 36 ?1 10 0 .2 4 11 22 ?1 13 8 .1 2 11 14 ?1 14 2 .1 2 11 52 ?1 21 8 .7 2 11 48 ?1 25 9 .6 4 76 D R i- 31 17 89 8 ? 30 11 59 ,1 18 2, 11 85 10 65 ?1 08 7 .2 2 10 34 ?1 14 2 .5 2 11 22 ?1 13 8 .1 6 11 47 ?1 21 7 .4 8 11 52 ?1 20 9 .6 2 Ji gg s t ho m ps on B ( 39 L M 20 8) 40 E t h -1 10 41 77 0 ? 60 12 65 ,1 26 6, 12 77 12 15 ?1 29 0 1. 00 10 73 ?1 07 7 .0 0 11 28 ?1 13 5 .0 1 11 54 ?1 32 6 .9 5 13 52 ?1 36 1 .0 1 13 66 ?1 38 9 .0 4 Ja nd re au ( 39 L M 22 5) 51 E t h -1 10 50 10 15 ? 5 5 10 20 98 3? 10 55 .6 6 89 7? 92 1 .0 5 10 89 ?1 12 1 .2 3 94 3? 11 60 .9 3 11 39 ?1 15 1 .1 0 11 73 ?1 18 8 .0 2 Pr et ty h ea d B ( 39 L M 23 2) 33 sM u -2 73 4 10 02 ? 4 2 10 22 99 4? 10 43 .6 0 90 3? 90 9 .0 1 10 90 ?1 11 9 .2 8 98 1? 11 58 .9 9 11 39 ?1 15 1 .1 2 11 81 ?1 18 5 .0 0 fa y t ol to n (3 9s t 11 ) 13 E t h -1 01 13 86 5 ? 60 12 09 10 68 ?1 08 4 .1 1 10 35 ?1 14 2 .3 6 11 24 ?1 13 7 .0 8 11 48 ?1 26 5 .6 1 11 53 ?1 25 9 .8 1 12 66 ?1 27 7 .0 3 so m m er s (3 9s t 56 ) 34 sM u -2 73 6 97 2 ? 73 10 30 ,1 14 4, 11 46 10 01 ?1 01 1 .0 5 89 9? 91 9 .0 2 10 17 ?1 15 6 .9 5 96 2? 96 7 .0 0 97 6? 12 23 .9 6 12 31 ?1 24 2 .0 1 12 47 ?1 25 7 .0 1 35 E t h -1 10 37 94 0 ? 60 10 44 ,1 08 9, 11 20 ,1 13 9, 11 51 10 27 ?1 15 8 1. 00 99 5? 12 20 1. 00 12 54 ?1 25 5 .0 0 36 E t h -1 04 18 10 85 ? 6 0 98 4 89 3? 92 7 .3 0 78 0? 79 1 .0 2 93 9? 10 20 .7 0 80 2? 10 40 .9 5 10 93 ?1 11 8 .0 2 11 49 ?1 15 0 .0 1 37 E t h -1 04 19 83 5 ? 60 12 16 11 59 ?1 18 3 .1 9 10 36 ?1 09 7 .1 4 11 85 ?1 26 5 .7 5 11 15 ?1 14 2 .0 7 12 67 ?1 27 7 .0 6 11 48 ?1 28 8 .8 0 37 E t h -1 10 38 10 65 ? 6 0 99 3 89 5? 92 5 .2 4 98 2? 78 8 .0 1 94 0? 10 27 .7 6 80 9? 82 4 .0 1 82 6? 84 4 .0 2 86 6? 10 68 .8 7 10 85 ?1 12 5 .0 7 11 37 ?1 15 3 .0 3 C at tl e o ile r (3 9s t 22 4) 39 E t h -1 10 40 69 0 ? 60 12 92 12 79 ?1 32 7 .6 0 12 23 ?1 23 1 .0 2 13 51 ?1 36 3 .1 1 12 38 ?1 25 0 .0 2 13 66 ?1 39 0 .2 9 12 56 ?1 40 7 .9 6 e x t e n d e d M Id d l e M Is so u r I h av en s (3 2E M 1) 48 E t h -1 10 47 56 0 ? 55 14 08 13 25 ?1 35 3 .3 6 13 00 ?1 37 5 .5 0 13 60 ?1 36 7 .0 8 13 76 ?1 43 9 .5 0 13 88 ?1 43 1 .5 6 49 E t h -1 10 48 61 5 ? 55 13 27 ,1 35 0, 13 90 13 03 ?1 33 3 .3 7 12 88 ?1 41 8 1. 00 13 39 ?1 37 2 .4 2 13 82 ?1 39 9 .2 0 Pa ul B ra ve ( 32 si 4) 61 D R i- 32 06 75 5 ? 63 12 80 12 14 ?1 29 7 1. 00 11 57 ?1 32 9 .9 0 13 47 ?1 39 2 .1 0 62 D R i- 32 07 78 3 ? 47 12 62 12 21 ?1 25 3 .5 2 11 59 ?1 18 0 .0 4 12 56 ?1 28 1 .4 8 11 86 ?1 29 7 .9 6 63 D R i- 32 08 67 4 ? 62 12 96 12 84 ?1 32 7 .5 4 12 25 ?1 22 7 .0 0 13 51 ?1 39 0 .4 6 12 43 ?1 24 6 .0 0 12 57 ?1 41 3 .9 9 V an de rb ilt V ill ag e (3 9C a 1) 64 D R i- 31 13 67 3 ? 44 12 96 12 86 ?1 32 5 .6 1 12 65 ?1 26 6 .0 0 13 54 ?1 35 8 .0 6 12 77 ?1 33 4 .5 2 13 68 ?1 38 8 .3 3 13 38 ?1 40 4 .4 8 65 E t h -1 60 69 74 5 ? 65 12 81 12 15 ?1 30 2 .9 6 11 59 ?1 18 3 .0 3 13 73 ?1 38 0 .0 4 11 85 ?1 33 1 .8 3 13 44 ?1 39 4 .1 4 66 E t h -1 60 70 75 0 ? 65 12 81 12 13 ?1 30 1 .9 8 11 57 ?1 33 1 .8 8 13 74 ?1 37 8 .0 2 13 45 ?1 39 4 .1 2 (c on ti nu ed ) T a b L e 7 . ( co nt in ue d) D at es ( a .D .) a nd r el at iv e ar ea s PV D L ab a ge C al ib ra ti on c ur ve un de r pr ob ab ili ty d is tr ib ut io n V ar ia nt a nd s it e no . nu m be r (r C Y b P) in te rc ep ts ( a .D .) 68 .5 % P 95 .4 % P e x t e n d e d M Id d l e M Is so u r I ( co nt in ue d) h el b (3 9C a 20 8) 42 E t h -1 10 43 51 5 ? 55 14 23 13 31 ?1 34 5 .1 6 13 03 ?1 37 2 .2 5 13 94 ?1 44 4 .8 4 13 83 ?1 48 8 .7 5 43 E t h -1 10 44 60 0 ? 55 13 30 ,1 34 6, 13 43 13 05 ?1 30 8 .0 4 12 92 ?1 42 5 1. 00 13 16 ?1 37 1 .6 8 13 86 ?1 40 8 .2 8 Ja ke W hi te B ul l ( 39 C 06 ) 1 sM u -2 66 3 74 7 ? 64 12 81 12 15 ?1 30 1 .9 8 11 59 ?1 18 3 .0 3 13 74 ?1 37 8 .0 2 11 85 ?1 33 0 .8 4 13 45 ?1 39 4 .1 3 2 E t h -1 01 09 72 0 ? 60 12 86 12 47 ?1 30 9 .7 6 12 11 ?1 33 2 .7 5 13 56 ?1 38 3 .2 4 13 41 ?1 39 9 .2 5 3 E t h -9 23 8 97 5 ? 75 10 29 ,1 14 5, 11 45 10 00 ?1 01 3 .0 7 89 8? 92 0 .0 2 10 16 ?1 07 5 .3 9 94 4? 95 2 .0 1 10 75 ?1 15 5 .5 4 95 7? 97 1 .0 1 97 6? 12 22 .9 5 12 33 ?1 23 9 .0 1 12 50 ?1 25 6 .0 1 58 D R i- 32 04 77 0 ? 41 12 65 , 1 26 6, 1 27 7 12 22 ?1 23 3 .1 6 11 64 ?1 16 7 .0 0 12 38 ?1 25 0 .1 9 11 93 ?1 20 1 .0 1 12 56 ?1 28 5 .6 4 12 06 ?1 29 8 .9 8 C al am it y V ill ag e (3 9D W 23 1) 4 E t h -1 01 10 79 5 ? 60 12 60 11 64 ?1 16 6 .0 1 10 45 ?1 04 6 .0 0 11 94 ?1 19 7 .0 1 10 47 ?1 08 9 .0 4 12 07 ?1 28 7 .9 8 11 19 ?1 13 9 .0 2 11 51 ?1 30 2 .9 3 13 73 ?1 38 0 .0 1 5 E t h -1 01 11 66 5 ? 60 12 98 12 87 ?1 32 7 .5 1 12 62 ?1 41 1 1. 00 13 50 ?1 39 0 .4 9 su lly s ch oo l ( 39 sL 7) 52 E t h -1 10 51 77 5 ? 55 12 64 ,1 27 0, 12 76 12 19 ?1 28 5 1. 00 10 75 ?1 97 5 .0 0 11 58 ?1 32 0 .9 8 13 70 ?1 38 7 .0 2 53 E t h -1 10 52 79 5 ? 55 12 60 12 11 ?1 28 3 1. 00 10 66 ?1 08 6 .0 2 11 21 ?1 13 8 .0 2 11 52 ?1 30 1 .9 6 13 74 ?1 37 6 .0 0 C he ye nn e R iv er ( 39 st 1) 17 E t h -1 04 13 67 5 ? 50 12 95 12 85 ?1 32 6 .5 9 12 64 ?1 27 1 .0 2 13 53 ?1 36 0 .0 8 12 76 ?1 33 4 .5 0 13 67 ?1 38 8 .3 2 13 36 ?1 40 5 .4 8 B la ck W id ow R id ge ( 39 st 20 3) 56 E t h -1 10 53 55 5 ? 55 14 09 13 26 ?1 35 2 .3 4 13 00 ?1 37 5 .4 8 13 62 ?1 36 6 .0 5 13 76 ?1 44 1 .5 2 13 89 ?1 43 3 .6 1 t e r M In a l M Id d l e M Is so u r I sh er m er ( 32 E M 10 ) 46 E t h -1 10 45 62 0 ? 55 13 26 ,1 35 2, 13 63 ,1 36 6, 13 89 13 01 ?1 33 3 .3 8 12 87 ?1 41 6 1. 00 13 41 ?1 37 4 .4 1 13 78 ?1 39 6 .2 0 47 E t h ?1 10 46 78 5 ? 55 12 62 12 15 ?1 28 4 1. 00 10 70 ?1 08 1 .0 1 11 24 ?1 13 6 .0 1 11 53 ?1 30 4 .9 7 13 72 ?1 38 3 .0 1 59 D R i? 32 05 51 8 ? 51 14 21 13 32 ?1 34 2 .1 2 13 06 ?1 30 9 .0 1 13 95 ?1 44 2 .8 8 13 14 ?1 37 1 .2 3 13 85 ?1 48 4 .7 6 60 E t h ?1 63 94 43 5 ? 45 14 44 14 27 ?1 49 4 .9 3 14 13 ?1 52 4 .8 1 16 03 ?1 61 3 .0 7 15 64 ?1 57 5 .0 2 15 75 ?1 62 7 .1 7 60 R E t h -1 75 11 66 0 ? 50 12 99 12 90 ?1 32 6 .5 1 12 80 ?1 40 7 1. 00 13 52 ?1 36 3 .1 4 13 66 ?1 38 9 .3 5 h uf f (3 2M 01 1) 67 D R i- 31 14 59 8 ? 39 13 30 , 1 34 6, 1 39 3 13 20 ?1 36 9 .7 0 12 97 ?1 41 5 1. 00 13 87 ?1 40 8 .3 0 68 E t h ?1 60 71 66 0 ? 60 12 99 12 89 ?1 32 8 .4 9 12 63 ?1 27 3 .0 2 13 50 ?1 39 1 .5 1 12 75 ?1 41 2 .9 8 69 E t h ?1 60 72 73 0 ? 65 12 84 12 21 ?1 25 3 .2 3 11 63 ?1 16 8 .0 1 12 56 ?1 30 5 .6 0 11 92 ?1 20 3 .0 1 13 10 ?1 31 6 .0 4 12 06 ?1 33 3 .7 7 13 71 ?1 38 5 .1 2 13 39 ?1 40 2 .2 1 70 E t h ?1 60 73 49 0 ? 60 14 33 13 33 ?1 33 9 .0 4 13 06 ?1 31 0 .0 0 13 99 ?1 45 5 .7 6 13 14 ?1 37 1 .1 3 14 55 ?1 48 1 .2 0 13 85 ?1 52 2 .8 1 15 82 ?1 62 5 .0 6 In It Ia l C o a l e sC e n t L yn ch ( 25 B D 1) 50 E t h -1 10 49 78 0 ? 55 12 63 ,1 27 3, 12 75 12 17 ?1 28 4 1. 00 10 72 ?1 07 8 .0 0 11 25 ?1 13 6 .0 1 11 54 ?1 30 5 .9 7 13 12 ?1 31 6 .0 0 13 71 ?1 38 5 .0 1 a rz be rg er ( 39 h u 6) 14 E t h -1 01 14 44 0 ? 55 14 43 14 19 ?1 49 7 .8 7 14 08 ?1 53 1 .7 6 15 14 ?1 51 5 .0 1 15 51 ?1 63 4 .2 4 16 00 ?1 61 7 .1 2 41 E t h -1 10 42 34 0 ? 60 15 23 ,1 56 5, 15 78 ,1 62 7 14 94 ?1 60 3 .8 1 14 42 ?1 66 0 1. 00 16 12 ?1 63 8 .1 9 (c on ti nu ed ) T a b L e 7 . ( co nt in ue d) D at es ( a .D .) a nd r el at iv e ar ea s PV D L ab a ge C al ib ra ti on c ur ve un de r pr ob ab ili ty d is tr ib ut io n V ar ia nt a nd s it e no . nu m be r (r C Y b P) in te rc ep ts ( a .D .) 68 .5 % P 95 .4 % P e x t e n d e d C o a l e sC e n t D em er y (3 9C o 1) 44 sM u -2 80 0 50 9 ? 47 14 26 13 34 ?1 33 9 .0 6 13 23 ?1 35 4 .1 4 14 00 ?1 44 4 .9 4 13 57 ?1 36 8 .0 2 13 88 ?1 47 7 .8 5 45 sM u -2 78 9 58 3 ? 58 13 33 ,1 33 9, 14 00 13 16 ?1 37 0 .6 4 12 95 ?1 43 2 1. 00 13 86 ?1 41 5 .3 6 L ow er g ra nd ( 39 C o 14 ) 10 sM u -2 68 9 39 5 ? 57 14 83 14 42 ?1 52 2 .6 7 14 35 ?1 63 9 1. 00 15 82 ?1 62 5 .3 3 11 sM u -2 72 5 32 6 ? 66 15 29 ,1 55 6, 16 32 14 94 ?1 60 3 .7 7 14 40 ?1 67 2 .9 7 16 13 ?1 64 7 .2 3 17 81 ?1 79 9 .0 2 19 45 ?1 95 3 .0 1 12 E t h -1 01 12 47 0 ? 55 14 37 14 10 ?1 48 6 1. 00 13 29 ?1 34 8 .0 4 13 92 ?1 52 4 .8 6 15 64 ?1 57 3 .0 1 15 75 ?1 62 7 .1 0 16 E t h -1 04 12 28 0 ? 55 16 48 15 19 ?1 59 5 .5 1 14 54 ?1 45 8 .0 0 16 22 ?1 66 9 .3 9 14 76 ?1 68 2 .8 2 17 83 ?1 79 7 .0 8 17 37 ?1 80 7 .1 3 19 48 ?1 95 2 .0 3 19 36 ?1 95 5 .0 5 Po tt s (3 9C o 19 ) 23 E t h -1 01 20 23 5 ? 55 16 62 15 35 ?1 53 6 .0 1 15 00 ?1 50 8 .0 1 16 37 ?1 68 3 .3 5 15 17 ?1 59 7 .1 2 17 34 ?1 80 8 .4 8 16 20 ?1 70 6 .3 1 19 32 ?1 95 5 .1 6 17 20 ?1 81 7 .3 8 18 33 ?1 87 9 .0 4 19 15 ?1 95 5 .1 4 24 E t h -1 01 21 41 5 ? 55 14 49 14 36 ?1 51 9 .7 9 14 21 ?1 53 3 .6 4 15 95 ?1 62 2 .2 1 15 42 ?1 63 6 .3 6 M ea nd er ( 39 L M 20 1) 19 E t h -1 01 16 35 0 ? 55 15 20 ,1 59 3, 16 22 14 90 ?1 53 2 .3 2 14 46 ?1 64 9 1. 00 15 47 ?1 63 5 .6 8 20 E t h -1 01 17 39 5 ? 55 14 83 14 42 ?1 52 2 .6 8 14 35 ?1 63 8 1. 00 15 83 ?1 62 5 .3 2 21 E t h -1 01 13 27 5 ? 55 16 49 15 20 ?1 59 2 .4 6 14 78 ?1 68 4 .7 8 16 23 ?1 67 1 .4 0 17 34 ?1 80 8 .1 6 17 81 ?1 79 9 .1 1 19 32 ?1 95 5 .0 6 19 46 ?1 95 3 .0 4 h os te rm an ( 39 Po 7) 22 E t h -1 01 19 29 5 ? 55 16 43 15 01 ?1 50 9 .0 5 14 50 ?1 67 6 .9 2 15 17 ?1 59 8 .6 1 17 76 ?1 80 3 .0 6 16 19 ?1 66 3 .3 4 19 40 ?1 95 5 .0 3 30 sM u -2 72 8 74 7 ? 62 12 81 12 16 ?1 30 0 .9 9 11 59 ?1 17 7 .0 2 13 75 ?1 37 6 .0 1 11 86 ?1 33 0 .8 5 13 46 ?1 39 3 .1 2 31 sM u -2 73 1 53 0 ? 64 14 14 13 26 ?1 35 2 .2 6 12 97 ?1 48 6 1. 00 13 62 ?1 36 6 .0 3 13 89 ?1 44 3 .7 1 32 sM u -2 73 2 23 2 ? 56 16 63 16 38 ?1 68 5 .3 4 15 03 ?1 50 5 .0 3 17 33 ?1 80 9 .4 9 15 18 ?1 59 6 .1 1 19 30 ?1 95 5 .1 7 16 21 ?1 71 0 .3 0 17 18 ?1 81 9 .3 9 18 34 ?1 88 1 .0 5 19 13 ?1 95 5 .1 5 su lly ( 39 sL 4) 27 E t h -1 04 16 36 5 ? 55 14 94 ,1 60 1, 16 16 14 54 ?1 45 7 .0 3 14 43 ?1 64 3 1. 00 14 78 ?1 52 9 .3 8 15 56 ?1 63 2 .5 9 28 E t h -1 04 17 26 5 ? 55 16 51 15 21 ?1 58 4 .3 5 14 84 ?1 68 9 .7 1 16 25 ?1 67 5 .4 1 17 30 ?1 81 0 .2 1 17 77 ?1 80 2 .1 6 19 24 ?1 95 5 .0 8 19 41 ?1 95 5 .0 8 29 sM u -2 72 6 40 0 ? 63 14 54 ,1 45 7, 14 78 14 40 ?1 52 3 .6 7 14 28 ?1 64 1 1. 00 15 81 ?1 62 6 .3 3 o ve r? s L a R oc he ( 39 st 9) 25 E t h -1 04 14 25 0 ? 50 16 55 15 29 ?1 55 6 .1 4 14 92 ?1 60 2 .2 2 16 32 ?1 67 9 .4 3 16 16 ?1 69 2 .3 5 17 63 ?1 80 5 .3 0 17 28 ?1 81 2 .3 1 19 38 ?1 95 5 .1 3 19 22 ?1 95 5 .1 2 26 E t h -1 04 15 42 0 ? 55 14 48 14 33 ?1 51 8 .8 1 14 16 ?1 53 2 .6 7 15 96 ?1 62 1 .1 9 15 45 ?1 63 5 .3 3 38 E t h -1 10 39 34 5 ? 60 15 22 ,1 58 4, 16 25 14 93 ?1 53 3 .2 9 14 42 ?1 65 6 1. 00 15 42 ?1 63 6 .7 1 W al th B ay ( 39 W W 20 3) 6 sM u -2 68 5 32 0 ? 63 15 32 ,1 54 7, 16 35 14 95 ?1 60 1 .7 7 14 42 ?1 67 1 .9 6 16 16 ?1 64 9 .2 3 17 81 ?1 79 9 .0 3 19 45 ?1 95 3 .0 1 7 sM u -2 68 7 32 1 ? 57 15 31 ,1 54 8, 16 35 14 98 ?1 51 3 .1 1 14 46 ?1 66 7 .9 9 16 15 ?1 60 0 .6 6 17 85 ?1 79 5 .0 1 16 18 ?1 64 7 .2 3 19 52 ?1 95 2 .0 0 8 sM u -2 67 7 32 2 ? 64 15 31 ,1 55 0, 16 34 14 94 ?1 60 2 .7 7 14 41 ?1 67 1 .9 7 16 15 ?1 64 8 .2 3 17 81 ?1 79 9 .0 3 19 45 ?1 95 3 .0 1 9 sM u -2 67 8 27 9 ? 46 16 48 15 22 ?1 58 2 .4 7 14 85 ?1 67 8 .8 6 16 25 ?1 68 8 .4 4 17 63 ?1 80 5 .1 0 17 84 ?1 79 6 .0 7 19 38 ?1 95 5 .0 4 19 50 ?1 95 2 .0 2 (c on ti nu ed ) T a b L e 7 . ( co nt in ue d) D at es ( a .D .) a nd r el at iv e ar ea s PV D L ab a ge C al ib ra ti on c ur ve un de r pr ob ab ili ty d is tr ib ut io n V ar ia nt a nd s it e no . nu m be r (r C Y b P) in te rc ep ts ( a .D .) 68 .5 % P 95 .4 % P Po st -C o n t a C t C o a l e sC e n t D od d (3 9s t 30 ) 18 E t h -1 01 15 16 0 ? 55 16 83 ,1 73 5, 18 07 , 1 93 4, 19 54 16 71 ?1 70 3 .1 9 16 64 ?1 89 3 .8 3 17 23 ?1 78 1 .3 5 19 08 ?1 95 5 .1 7 17 99 ?1 81 5 .0 9 18 39 ?1 87 7 .1 9 19 17 ?1 94 6 .1 7 19 53 ?1 95 5 .0 1 n u m b e r 4 7 ? 6 5 periods. separate rules for evaluating these sites are pre- sented later. 2. furthermore, ethnographic evidence (cf., Will, 1930; Wilson, 1934; Weitzner, 1979) and data from modern lodge reconstructions at sites, such as fort abraham Lin- coln state Park, north Dakota (C. Erikson, pers. comm. to s. ahler, 1988), indicate clearly that the useful life of an earthlodge is about 20 years or less, due primarily to the rotting of posts and beams that form the main structural elements in the lodge. thus, if a village was used for more than about 20 to 30 years, there should be clear evidence of such a length of occupation in the form of lodge repairs or replacements or multiple lodge floors and architectural plans built on the same or superimposed locations. it is assumed, therefore, that architectural information can be used to assess the approximate duration of occupation of a site and, on that basis, to suggest what to expect from multiple radiocarbon dates from that site. 3. in addition, data from relatively well-dated historic and protohistoric village occupations of various durations are beginning to yield information about what to expect re- garding midden accumulation rates within an earthlodge vil- lage. this in turn is used to assess the approximate duration of occupation of a prehistoric village, and thereby to predict what to expect regarding the spread of radiocarbon dates from that site. table 8 contains data about rates of mid- den accumulation within three historic and protohistoric hidatsa villages at the mouth of the Knife River in north Dakota. Data are presented for 15 individual excavation lo- cations in these three villages, all in outside-house contexts (data are taken from excavation reports of ahler, Weston et al., 1980; ahler and Weston, 1981, and ahler and swenson, 1985, in conjunction with more refined chronological as- sessments for each site in ahler and haas, 1993). the mid- den accumulation rates range from 0.250 to 1.985 m per 50-year interval. the rate for unit 4 at Big hidatsa (1.68 m per 50 years) is obviously an outlier within that site, and rates for sakakawea in general are substantially higher than for the other two sites. in fact it appears that the rates for deposits accumulating within the fully historic period (post? ad 1790) are substantially higher than for all deposits that have some component earlier than that in time. the overall mean of all 15 rate measurements in table 8 is 0.827 m per 50-year interval. if the five historic period measurements from sakakawea and the one fully historic period sample from Big hidatsa (unit 4) are eliminated, the nine remaining TabLe 8. Data from test excavations in three villages regarding the rate of outside-house midden accumulation. excavation Midden depth Midden rate, Village number (m) Dates (a.D.) Years m/50yr sakakawea 3 1.35 1800?1834 34 1.985 sakakawea 8 1.20 1800?1834 34 1.765 sakakawea 9 0.75 1800?1834 34 1.103 sakakawea 11 0.60 1800?1834 34 0.882 sakakawea 12 1.00 1800?1834 34 1.471 sakakawea all units 4.90 170 1.441 Big hidatsa 1 1.70 1600?1790 190 0.447 Big hidatsa 4 1.85 1790?1845 55 1.682 Big hidatsa 6 1.25 1600?1745 145 0.431 Big hidatsa 7 1.15 1600?1745 145 0.397 Big hidatsa 8 1.05 1745?1830 85 0.618 Big hidatsa 9 0.90 1650?1830 180 0.250 Big hidatsa 11 0.80 1745?1845 100 0.400 Big hidatsa all units 8.70 900 0.483 Lower hidatsa 1 1.10 1600?1740 140 0.393 Lower hidatsa 3 2.40 1560?1780 220 0.545 Lower hidatsa 4 1.80 1525?1780 255 0.353 Lower hidatsa all units 5.30 615 0.431 Mean of 15 units, all sites 0.827 Mean of 9 units, prehistoric and protohistoric 0.426 6 6 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y sample units have a mean accumulation rate of 0.426 m per 50-year interval. this is thought to be a realistic estimate of midden accumulation rates in prehistoric earthlodge vil- lage sites. Based upon this, it is assumed that an occupation of less than 50 years would yield a midden less than about 40 cm in maximum thickness in outside house locations. it can be noted that the vast majority of known earth- lodge village sites appear to lack midden deposits deeper than 30?40 cm or so, and also lack clear evidence of lodge rebuilding on the same or overlapping locations. thus, the norm is for a village to be occupied for a half-century or less (perhaps only for a decade or two). as a result, radio- carbon dates from village sites should be evaluated in the context of what to expect in the dates relative to other in- formation about the duration of occupation of the sites. 4. Recent work in the Knife region of north Dakota resulted in a regional chronological sequence and taxon- omy, based in large measure upon radiocarbon dating and ceramic variation, in which phase units are 50, 75, or 100 years in duration. these are discrete, definable units based upon changes in ceramic attributes, lithic raw material selection and reduction patterns, settlement patterns, and lodge construction patterns. it is reasonable to expect to identify and define taxonomic systems or units of culture- historic construction in other regions in the subarea that have equal temporal precision, being no more than a cen- tury in duration. 5. When this information is combined with the ob- servation that most earthlodge villages were occupied less than 50 years (a single component has an associated time span of 50 years or less), it is apparent that radiocarbon dates that have an associated error (sigma or standard de- viation) of 100 years or more are of extremely little use for furthering our understanding of cultural chronology and change in the subarea. if the temporal measuring instru- ment has a built-in error of plus or minus one hundred years, then this instrument is of little value for defining and discriminating among units of study that may lie less than one hundred years apart on the time scale. therefore, it is appropriate to, from the start, eliminate from consid- eration any reported radiocarbon date that has a standard deviation of 100 years or greater. By screening dates in this manner, the focus shifts to evaluation of radiocarbon dates that have the greatest potential to contribute to re- solving the chronological problems in the subarea. 6. therefore, after excluding all reported dates that have sigmas of 100 years or greater, the next step in the evaluation process involves a test of contemporaneity or an estimation of the probability that, when more than one dates exists, that all are drawn from a single population. this involves testing the variance among two or more dates relative to what their expected variance is, based upon direct archaeological evidence from a given site. it is assumed that if there is no firm archaeological evidence that a site was occupied during a long period, the expected duration of occupation of a village would be 50 years or less. as a consequence, the expectation would be that all reliable dates from the site would date what is essentially a single event (drawn from a single population), therefore passing the test of contemporaneity as discussed by Ward and Wilson (1978) and as computed in stuiver and Re- imer (1993). acceptable evidence of long-term or multiple occupations occurs only in a few forms: (1) ceramic ag- gregates that contain sizable numbers of types distinctive of different cultural variants in the region; (2) evidence of superimposed houses, house floors, and/or extensive re- building of houses on a single location; and/ or (3) deep and extensive midden or trash accumulations of 40 cm or more in thickness and many square meters in extent. this is called primary evidence of long-term occupation (PELto). it is therefore only in instances in which PELto is clearly present that multiple dates from a site should not be expected to pass the test of contemporaneity. 7. if a site does not contain PELto, and if at the same time dates from a single site having sigmas smaller than 100 years do not pass the test of contemporaneity, it is concluded, lacking clear, direct evidence for other sources of error, that the failure of the contemporaneity test indi- cates a problem in the reliability of the radiocarbon dates themselves. this perspective is the opposite of the more conventional approach, a willingness to do the reverse? to accept almost all dates except the wildest extremes as meaningful, and to look for some age/association problem in the dated material itself to explain internally inconsis- tent radiocarbon results. as thomas thiessen recently stated (t. thiessen, pers. comm. to s. ahler, 1994), there has for many years been a near-consensus among archae- ologists that ?every date that is produced must be used or explained in some manner??dates simply cannot be ignored. after all, individual radiocarbon dates are too costly for them not to be used in some fashion. the present perspective?questioning the dates them- selves rather than the dated material or its association? comes from a growing body of evidence that indicates that many radiocarbon dates for Plains Village sites in the Da- kotas run by many different labs (even many dates recently produced on carefully selected short-lived materials) are simply not as precise or reliable as reported error values would indicate. there is growing evidence, grounded in in- creasingly careful scrutiny of materials submitted for dating n u m b e r 4 7 ? 6 7 and better understanding of the archeological context (cf. shott, 1992) that individual labs vary a great deal in terms of the reliability and accuracy of the dates they produce, regardless of reported error values. there is empirical evi- dence (e.g., ahler and haas, 1993) that it is not inherently difficult to repeatedly produce sets of internally and exter- nally consistent dates for single Plains Village sites. it has been known for some time within the radiocarbon com- munity that reported error rates from some labs are greatly underestimated because they often include only counting error but no other sources of error (see haas, 1995). it also is clear from a growing body of archaeological evidence and by experience with different labs that some (and per- haps many) labs simply are not applying the routine ana- lytic procedures and quality controls necessary to ensure the production of reliable dates for such sites. Knowing this, the usefulness of the results produced by various labs, based upon primary archaeological data and knowledge of the archaeological record, can be used as a beginning point in the assessment of radiocarbon dates from the study area. this is the approach that is followed here, focusing first and foremost on the results of the PVD date series. With this background in mind, the evaluation of radio- carbon dates begins with the sites in which larger series of dates (include some PVD dates) were run, and proceeds to the less intensively dated sites about which something can be said in terms of critical evaluation. in general, the first assessments begin with large PVD date sets from individ- ual sites that had been previously extensively dated or well dated. this permits the simultaneous evaluation of PVD dating results and the evaluation of previously run dates. in some cases, this process yields information useful for farther-reaching decisions about the reliability or validity of sets of dates from specific labs, helpful in reaching some conclusion about the advisability of using any or all dates from specific labs. this in turn affects the way previously run or existing dates from several sites that were not dated in the PVD dating program but that are used in the ceramic analysis aspects of this study are chosen for consideration. the abbreviations associated with archeological con- texts are house (h), feature (f), and excavation unit (xu) numbers. feature number, depending on the excavator of the site, sometimes designates houses. sites with cOmplex Or extensive date series Helb (39CA208) (Extended Middle Missouri) the helb site was selected for dating for several reasons. foremost, it has produced a large, well-controlled excavated collection, potentially contributing a great understanding of Plains Village lifeways. an important data set, such as helb, cannot be left undated. Equally important, the site remains the focus of much discussion and controversy regarding its age, taxonomic affiliations, and role in the dynamics of cul- tural interaction in the subarea. Lehmer (1971, fig. 79) considered helb, discovered in 1966, to be a terminal Middle Missouri variant village, and one of the southernmost, if not the southernmost, ex- ample of such, a village in a prime location to elicit conflict with competing Extended Coalescent populations to the south. falk and Calabrese (1973) reevaluated helb as an Extended Middle Missouri village, apparently occupied relatively early in time, and with little evidence for later components at the site. thiessen and nickel (1975) as- sessed a large series of radiocarbon dates from the site and evaluated the radiocarbon evidence as indicating the pres- ence of two components, one very early in the Extended Middle Missouri period (ca. ad 1000), and the second in the sixteenth century ad. More recently, Kay (1994) speculated that there were three distinct periods of use of the site by Extended Middle Missouri populations who, when using the site for more than five centuries, retained a strikingly uniform ceramic tradition. thus the site has the possible distinction of simultaneously containing one of the earliest Extended variant components within the entire subarea, as well as documenting a late terminal Middle Missouri population driven from the locality by expanding Extended Coalescent peoples holding territory just to the south. Regardless of how the speculative, fanci- ful, and imaginative arguments about the chronological significance of certain archaeological details within the site are perceived (e.g., house size [thiessen and nickel, 1975:309]; house pit creep rates [Kay, 1994, cited in Rose- brough, 1994:66]; house floor depths [Kay, 1994:19?20]; and pottery vessel thickness [Rosebrough, 1994:66]), it is believed that all investigators would agree that there is no primary evidence of long-term occupation (PELto) at the helb site. all of the speculation about time depth at helb has its roots in a single data set that is taken at face value: the existing series of radiocarbon dates whose central ten- dencies alone span from ad 970 to ad 1620 (thiessen and nickel, 1975; thiessen, 1977) (table C.2). Lacking PELto, an evaluation of the reliability of the dates pro- duced for the helb site is needed. Cultigen (maize) samples from helb are the focus of the current dating effort with one aMs sample each drawn from two trash-filled pits associated with two very similar and adjacent rectangular houses excavated in 1973 (f 417 with house 15 and f 352 with house 14). five cupule fragments 6 8 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y constituted one sample and a single cupule the other. two wood charcoal samples from within each of these same pits had previously been dated (table C.2), and collectively, these four previous dates reflect the wide range of dates available from the site as a whole. the PVD dates provide a direct check on the previously run dates from these two contexts. Lacking PELto, it is expected that all dates from the individual labs would pass the test of contemporaneity. several points can be made from our evaluation: (1) the two PVD dates from f 417 and f 352, re- spectively (Eth-11043 and Eth-11044) pass the test of contemporaneity at the 0.05 level (t'=1.06, df=1, critical x2=3.84). thus, they are consistent with each other, as is expected from all archaeo- logical data, and they appear to be reliable given the lack of PELto for helb. (2) the two Radiocarbon Limited dates from helb (RL-298 and RL-299) fail to pass the test of con- temporaneity (t'=15.33, df=1, x2=3.84). Lacking PELto, there is therefore reason to question the reliability of these dates. there is no particularly good reason to choose one date as more accurate than the other. one perspective is simply to con- sider the lab error factor to be highly underesti- mated by the reported sigma. using CaLiB 3.0.3 (stuiver and Reimer, 1993) it is determined that an error multiplier K value of about 2.1 must be applied to the reported dates for them to pass the test of contemporaneity. this means that when lab error and counting error are both taken into ac- count, sigmas on the order of ?190 years are ap- propriate for the reported RL dates. (3) four of nine nebraska Wesleyan university (nWu- 39, 40, 45, and 55) dates have reported sigmas of 100 years or greater (table C.2). Based upon as- sumptions made previously, these are dropped from consideration as being too imprecise for contribu- tion to the chronological assessment of the site. (4) the remaining five nebraska Wesleyan university dates from the helb site (nWu-38, 46, 52, 53, and 54) fail the test of contemporaneity (t'=35.46, df=4, critical x2=9.49). Lacking PELto, this se- ries of nWu dates appears to be unreliable. using procedures in CaLiB 3.0.3, the reported sigmas must be increased by a lab error factor of approx- imately K=2.0, meaning sigmas on the order of 150 to 190 years rather than the reported 70 to 90 years, in order for the date series to pass the test of contemporaneity. other more specific evaluations of the remaining low- sigma nWu dates also can be made: (5) the two previously run nWu dates from a single context, f 417 in h 15 (nWu-52 and nWu-46), do not pass the test of contemporaneity (t'=8.97, df=1, critical x2=3.84). thiessen and nickel (1975: 308?309) noted this discrepancy and chose to speculate that the two nWu samples from the single pit are reliable but derived from carbon of different ages. the preference here is to assume, lacking PELto, that all of the carbon from the single pit feature is effectively the same age (small twigs were selected for the nWu samples) but that lab error values (sigmas) reported for these sam- ples are greatly underestimated. this conclusion is consistent with that reached for the nWu date series from helb as a whole, but provides a re- fined assessment of this generality based upon two nWu dates from a single context. these two dates (nWu-46 and 52) are judged to be unreliable. (6) the remaining three low-sigma nWu dates (38, 53, and 54) (table C.2) do not pass the test of contem- poraneity (t'= 21.77, df=2, critical x2=5.99). Lack- ing PELto, these three dates are considered to be unreliable, eliminating them from the analysis. (7) in summary, lacking PELto for helb, the two Eth dates for helb are viewed as the most reliable and valid radiocarbon dates for the site. the weighted average of these two dates (558 ?41 bp) is used as the basis for developing a calibrated calendar age for the site of ad 1409 (crossover point) and a 2-sigma probability range of ad 1304 to ad 1433. Jake White Bull (39CO6) (Extended Middle Missouri) the Jake White Bull site is an analog to helb in many ways, originally being evaluated by Lehmer (1971:122) as one of the southernmost terminal Middle Missouri vari- ant sites known, then having been reevaluated as represent- ing an early component of the Extended Middle Missouri (ahler, 1977a). the Jake White Bull site lacks PELto so it is expected, all else being equal, that all series of dates from the site will pass the test of contemporaneity. in the initial testing program at the site, excavations were made near and within one house (h 4) in the north- central part of the cutbank and at the intersection of the south fortification ditch by the cutbank. initially, the uni- versity of georgia lab dated six radiocarbon samples from various places in both of these contexts. these six dates n u m b e r 4 7 ? 6 9 (uga-1488 through uga-1493 in table C.2) have a wide dispersion, some are unrealistically old, and as a group they fail the test of contemporaneity (t'= 65.82.43, df=5, critical x2=11.07). in discussions between stanley ahler and Dr. Betty Brandeau of the uga lab, Brandeau iden- tified three dates as being unreliable (uga-1488 through uga-1490), with the remaining three being more reliable. as a test of this proposition, ahler submitted a seventh sample for dating at uga (uga-1558) that was comprised of a composite of charcoal from the same contexts repre- sented in the three dates identified by Brandeau as unreli- able (from f 2, f 4, and f 7). this in effect was a re-date of the three unreliable samples. Based upon what Bran- deau had stated, strikingly different results were expected, compatible with the three dates identified by Brandeau as being more reliable. this was precisely the result received. the re-date result (uga-1558) was in fact much younger than previous runs on the same material, and the date was compatible with the three other uga dates identified by Brandeau as more reliable. the test of contemporaneity on these four samples (uga-1491, 1492, 1493, and 1558) (t'=0.90, df=3, critical x2=7.82) shows them to be drawn from the same population with a high degree of probabil- ity. in assessing the site, ahler (1977a) used the mean of the four dates (ca. ad 1013?43) as the best chronometric estimate of the age of the site, placing it very early in the Extended Middle Missouri sequence in the subarea. unfortunately, the check date run by the uga lab (uga-1558) has a sigma of 105 years, a value too large to be considered useful for developing a refined chronology for village sites, as discussed previously. in keeping with the general assumptions guiding this renewed evaluation, it is appropriate to eliminate this date from consideration, treat- ing it, in essence, as a non-contributor to the dating pro- gram. When this is done, the three dates identified by the uga lab as the most reliable (uga-1491, 1492, and 1493) are accepted. these have a weighted average of 943?33 years bp. in retrospect, there has always been the nagging ques- tion in ahler?s mind of why the three unreliable dates from the uga lab were not reported as such from the outset, rather than having been identified as unreliable only after the archaeologist questioned the results. in addition, ahler learned that the university of georgia radiocarbon labora- tory never routinely conducted pretreatment designed to extract soluble humic acids (Dennis toom, pers. comm. to s. ahler, 1993). haas? experience with Plains Village samples from north Dakota indicates that humic acid ex- traction is sometimes a time-consuming but critical step in pretreatment, with samples often containing large amounts of soluble humates requiring many cycles of treatment in naoh for removal. in hindsight, substantial doubt re- mains regarding the reliability of the uga dates from Jake White Bull, and this is the context in which new samples from the site were chosen for dating in the PVD program. four samples from Jake White Bull are selected for dating in the PVD program (table C.1). these consist of a conventional sample of wood charcoal from the midden outside of h 4 (material from the same context as uga- 1493), a conventional sample from the south fortification ditch, an aMs sample that is a composite of fragments of charred maize cupules from four subfloor pit features asso- ciated with h 4 (closely comparable to sample uga-1558 (rejected for having a large sigma) but dating presumably better, shorter-lived materials), and an aMs sample of carbonized residues from the lip and exterior shoulder of a large Riggs ware pottery fragment from within f 3, a sub- floor pit within h 4. the latter sample is best considered experimental in nature, being the first carbonized sherd residue date to be run in the PVD program. if reliable, the result from this sample should be compatible with the other three more conventional PVD sample dates. these four dates (sMu-2663, Eth-10109, Eth-9238, and DRi-3204) pass the test of contemporaneity (t'=7.47, df=3, critical x2=7.82), yielding a weighted average of 786?30. Even though they pass the test, the sherd resi- due date is at variance with the other three dates. Lacking PELto at the site and given the experimental nature of the aMs sherd residue-dating program, the date on residue is the primary choice for possible error of unknown source. When the residue date (Eth-9328) is set aside, the remain- ing three PVD dates (sMu-2663, Eth-10109, DRi-3204) also pass the test for contemporaneity (t'= 0.40, df=2, criti- cal x2=5.99). the weighted average of these three dates is 751?33 bp, some 224 radiocarbon years more recent than the date from the carbonized sherd residue date from basi- cally the same context. tentatively then, the sherd residue date is omitted from consideration (Eth-9328), accepting the remaining three PVD dates as useful. a final evalua- tion of the sherd residue date awaits further comparisons of residue and non-residue dates from the same contexts at other sites. the next step is to compare the newly run and accept- able PVD dates with the previously run and accepted uga dates from Jake White Bull. the weighted means of these two sample sets (mean of three uga dates=943?33 and of three PVD dates=751?33) do not pass the test of contem- poraneity (t'= 12.18, df=1, critical x2=3.84), suggesting that one or the other mean is not accurate. furthermore, individual tests of contemporaneity applied to pairs of 7 0 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y dates from essentially the same contexts indicate that one or the other date in each pair is not accurate (Eth-10109 vs. uga-1558 yields t'=4.19, df=1, critical x2=3.84 (even with the large sigma for the uga date); sMu-2663 vs. uga-1493 yields t'=3.53, df=1, critical x2= 3.84). given the concerns about reliability and pretreatment processes conducted by the uga lab, as discussed in the preceding paragraph, the PVD data are considered to be the most accurate of the available dates. the weighted mean of these three dates (751?33) generates a calibrated cross- over point of ad 1280 and a 2-sigma probability range of ad 1217 to ad 1297. Lower Grand (39CO14) (Extended Coalescent) the Lower grand site is targeted for attention in the present program because: (1) it contains one of the larger, well-controlled, excavated samples of all sites in the sub- area; (2) it lies near the northern limit of the geographic extent of Extended Coalescent components; and (3) it has previously been radiocarbon dated and assessed as being the earliest Extended variant component in the subarea, an unusual circumstance given its far upstream geographic position. it is therefore potentially a key site for under- standing interactions among the Extended Middle Mis- souri and the Extended Coalescent variants, and it plays a potentially central role in understanding the culture history of both the arikaras and the Mandans (Bowers, 1948a:96?99, considered Lower grand to be a prehistoric, southern Mandan subgroup village, a serious proposition still potentially valid, yet untested). there is evidence from fortification systems that the village area enclosed by fortification ditches expanded through time, with the ditch reexcavated to enclose a larger area later in time. in addition, there is architectural evidence in h 6 and h 17 that at least these two structures were substantially repaired by the addition of numerous small support posts during the period of their use. house 12, a much larger structure, shows less clear evidence of substantial maintenance. Middens in various parts of the site reach about 50 cm in depth, but not deeper (falk and ahler, 1988:30?37). Present ceramic data do not support the presence of multiple taxonomic units at the site. thus, there is clear evidence that the village occupation was not brief, but primary evidence of long-term occupation re- mains equivocal. it is almost certain that the village was inhabited for more than a decade but less than 100 years; an occupation span of about 50 to 60 years is estimated for the site, with no clear evidence of this having occurred along with substantial periods of site abandonment. thus, the radiocarbon dates from the site are expected to pass the test of contemporaneity. four radiocarbon dates had previously been run for the Lower grand site (ahler, 1975a, table 4; falk and ahler, 1988:64?65) (RL-300 through RL-303, table C.2). these samples were selected to represent several contexts within the site, specifically focusing on the question of the duration of site occupation: two are from h 12, one is from h 6, and the fourth is from a midden in the north- ern part of the site, north of the interior fortification ditch in a part of the site that might have been used latest in time. one of these dates is eliminated from consideration (RL-301) based upon its large standard error of 120 years. the remaining three dates pass the test of contemporaneity (t'=0.87, df=2, critical x2=5.99). there is no clear reason not to average them, and they yield a weighted mean of 560? 53 bp. this is the earliest reported date for an Ex- tended Coalescent component in the subarea. four additional dates were run from Lower grand in the PVD program (table C.1 and table 8). one (sMu-2689) is a conventional sample of wood charcoal from a large pit (f 59) just outside h 12, in a context comparable to f 102, another pit of great research interest. two dates (sMu-2725 and Eth-10112) are on short-lived material from f 102, outside h 12. the final date (Eth-10412) is on maize from a pit feature (f 3) in h 11, the first dated sample from the site from excavations by alfred Bowers. these four dates pass the test of contemporaneity (t'=6.12, df=3, critical x2=7.82). the weighted mean of these four dates is 366?30 bp. a span of 194 years exists between the weighted mean of the three Radiocarbon Limited dates from Lower grand and the four PVD dates, with the RL dates being the older. the two respective weighted means do not pass the test of contemporaneity (t'=9.80, df=1, critical x2=3.84). a more specific comparison is made by comparing the single low-sigma RL date from h 12 (RL-300) with the mean of three PVD dates from h 12 (sMu-2689, sMu-2725, Eth-10112). the two sets of dates pass the test of con- temporaneity (t'=3.46, df=1, critical x2= 3.84). Even so, there is a 186-year difference between these two values, and there is a probability of between 5% and 6% that they are drawn from the same population. although the dates from each lab series are internally consistent (reli- able), they are not comparable enough for an averaging of all dates (one set or the other is probably inaccurate). Because the RL dates from the helb site are internally in- consistent and unreliable, greatest weight in the present context is given to the PVD dates. in addition, the PVD dates are for the most part derived from short-lived mate- rials, whereas the RL dates are not so derived. thus, it is n u m b e r 4 7 ? 7 1 concluded that the most accurate date for Lower grand as a whole is the mean of the four dates produced in the PVD program, 366?30 bp. When calibrated, this date crosses the calibration curve at ad 1494, 1602 and 1615, and has a 2-sigma probability range of ad 1449 to ad 1635. two additional dates from wood charcoal became available with the completion of a report of salvage ex- cavations at the site conducted by the u.s. army Corps of Engineers in 1984 (Winham, 1995:187). Winham?s analysis of the dates indicates a difficulty in assessing the two dates (510?90, 270?80) because of their lack of overlap. When these dates are entered in CaLiB, they just barely past the test of contemporaneity (t'= 3.83, df=1, x2=3.84). Based upon this consideration and the fact that the four PVD dates are from short-lived materials (twigs, corn), we decided not to combine the two dates reported by Winham with the four acceptable PVD dates. Walth Bay (39WW203) (Extended Coalescent) the Walth Bay site is selected for additional dating be- cause it has produced a very large, well-controlled sample of archaeological material and it has previously been dated with apparently acceptable results. the site presents an in- teresting example of superimposed architectural features in combination with an apparent short-term occupation. the largest excavated residential unit, h 21, is highly patterned in form and was apparently briefly used. h 19, slightly smaller, was built on the same location and in the same ori- entation as h 21, presumably by the same group of people. Evidence suggests that h 19 was built and used immedi- ately after the destruction of h 21. h 9 is very small and built in a shallow pit superimposed on remnants of both h 19 and h 21 (falk and ahler, 1988:61?63). Despite this clear sequence in house construction, there is no clear evi- dence that the Walth Bay site was used for a lengthy period of time. Midden deposits are generally dense but very shal- low (less than 0.3 m in thickness), and there is no clear evidence in ceramic remains of substantial time passage during use of the site. thus, there is no clear indication of PELto, with the expectation that the radiocarbon dates from the site are drawn from a single population. the Radiocarbon Limited lab had previously run four dates on samples from Walth Bay (ahler, 1975a, table 4; falk and ahler, 1988:64?65). these were intentionally selected from widely separated locations within the site to test the possibility that measurable time depth was re- flected in the samples locations. one sample (RL-304) is from h 19, the earliest of three overlapping house fea- tures, a second (RL-305) is from h 9, the latest of the three; third and fourth samples (RL-306 and RL-307) are from tests in spatially divergent parts of the village. the four dates (table C.2) pass the test of contemporaneity (t'=1.62, df=3, critical x2=7.82), and produce a weighted mean of 396?46 bp. at face value, there is no reason not to accept this series as reliable and accurate. four new dates were run on Walth Bay materials in the PVD program. the focus of this dating effort is on selected contexts associated with h 19 or undifferentiated h 19/21, primarily as a test of the reliability of dates being produced in the PVD program. Because it is believed that the period of use of these two houses was very short term, the new dates, if reliable, should cluster tightly. one sample, sMu- 2685, is essentially a split sample of the same h 19 roof- fall material dated in RL-304. a second, sMu-2687, also is small twigs and branches comprising roof fall from the same burned house that was swept into an open trash pit. a third, sMu-2677, is a conventional charcoal sample from f387, a previously undated pit associated with the use of houses 19 and/or 21, and the final sample, Eth-10112, is a maize sample from the same pit context. the PVD results are amazingly consistent internally (table 7). Central tendencies for the two h19 short-lived, roof-fall samples differ from each other by one year. the central tendency for the conventional f387 sample differs from these two dates by one and two years, respectively. the aMs date on maize from the same pit feature is 43 years younger than the conventional charcoal date, sug- gesting that the corn may have grown a few years after the branches or saplings were cut to construct the roof of h19. not surprisingly, these four dates pass the test of contemporaneity (t'=0.17, df=3, critical x2= 7.81), hav- ing a weighted mean of 312?28 bp, clearly increasing con- fidence in the reliability of the dates being produced in the PVD program, whether conventional or aMs. the remaining question then centers on a comparison of the two date series from the RL and PVD program. Comparison of the weighted means for the two sets of four dates (the mean for the RL dates is 91 years older than the mean for the PVD dates) indicates that they pass the test of contemporaneity (t'=2.37, df=1, critical x2=3.84). a more precise comparison is made between the single RL date from h19 (RL-304, 450?90 bp) and the mean of the two PVD dates on roof fall from the same house (mean of sMu-2685 and sMu-2687, 321?42 bp). Despite the 129- year difference in these dates, they pass the test of contem- poraneity (t'=1.61, df=1, critical x2=3.84). it is of note, however, that the mean for the RL dates from Walth Bay is about 30% older than the mean of PVD series from the site. this is similar to the relationship between RL dates 7 2 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y from Lower grand and PVD dates from that site, although in that case the difference is greater (the RL mean is 57% older than PVD mean). is it possible that the RL dates in general are systematically older than they should be? it is believed that this is a distinct possibility, and on that basis the mean of the PVD dates alone is used as the best esti- mate of the age for the Walth Bay site (mean=312?28 bp). this date calibrates to intercepts of ad 1535 and ad 1638 and a 2-sigma probability range of ad 1492 to ad 1653. Sommers (39ST56) (Initial Middle Missouri) the sommers site is selected for additional study be- cause it is a possible early example of the initial Middle Missouri variant. it was subjected to a thorough assessment of ceramics and intrasite variability (steinacher, 1990), and, even though extensively dated, its chronological placement has not been established through chronometric means. the sommers village consists of a scatter of more than 100 house depressions, some 17 or 18 of which appear to have been enclosed within a fortification system, with the others lying outside a fortification ditch. steinacher (1990) conducted a thorough assessment of all available stratigraphic, architectural, and ceramic evidence concern- ing the settlement dynamics and internal chronology of the site. he concluded that there was some evidence in ceramic remains to indicate that the village was large and that it contracted in size through time, with houses within the fortified area having been used the longest. although the evidence indicates that the village may have had a dy- namic settlement history, there is nonetheless little clear evidence for long-term occupation. although there is evi- dence that one of seven excavated houses may have been rebuilt, there is no evidence from midden accumulations or ceramic data that any given house unit or house area was used for a long period. there are, however, several small midden areas within the fortified portion of the site. among the dated samples from the site, those in xu17 and xu18 (table C.2) are from within the fortified part of the site, whereas those from xu70 are from a house lying outside the fortified area. from the point of view of what is expected in series of radiocarbon dates from the site, it is best to study sets of dates from respective labs and respective inside-ditch and outside-ditch areas separately. Because houses within the fortification (xu17, xu18) may have been used longer, there is the possibility that dates from that part of the site may not pass the test of contemporaneity and yet may be reliable and accurate, whereas dates from xu70 outside the fortification should definitely pass the test of contemporaneity. it is with these expectations that all of the radiocarbon information from the site is evaluated. two samples from house posts within two structures inside the fortification ditch were dated many years ago (si-314 and si-315; table C.2). these dates have sigmas of 100 years or greater, and are eliminated as not particu- larly useful given the arguments presented previously. the Beta analytic lab ran three dates on outer rings from post samples from within a single house, xu18, inside the fortification system. these three Beta dates (Beta-1901, 1902, 1903; table C.2) pass the test of contemporaneity (t'=3.37, df=2, critical x2=5.99), and are accepted as reli- able for the moment. three additional Beta dates (Beta-1904, 1905, 1906) were run on outer ring wood from posts in a single house, xu70, outside the fortification that are expected to pass the test of contemporaneity. these three do not pass the test of contemporaneity (t'=26.38, df=2, critical x2=5.99), and are therefore considered to be unreliable. they are eliminated from further consideration. in the PVD dating program, a concerted effort was made to date short-lived materials from the sommers site and to date previously dated contexts in order to untangle the chronometric confusion that exists for the site. four samples were submitted, consisting of charred grass and twigs (sMu-2734), as well as maize (our PVD sample no. 37) from xu18 (which may or may not be contemporane- ous), and two maize samples (Eth-11037, 10418) from a single house, xu70 (which should be contemporaneous). Because of a mix-up in the lab, the maize sample PVD- 37 from xu18 was actually aMs-dated twice, yielding dates Eth-10419 and Eth-11038. these two dates on the same material do not pass the test of contemporaneity (t'=8.32, df=1, critical x2=3.84), and both are rejected as being unreliable. the single date from burned roof debris (sMu-2736) is therefore taken as the only reliable PVD date for this context, xu 18. When this date (972?73 bp) is compared with the mean of the three apparently accept- able Beta dates for this sample context (Beta-1901, 1902, 1903; mean of 1129?38 bp), they pass the test of con- temporaneity (t'=3.26, df=1, critical x2=3.84). given the substantial difference in these dates (the sMu date is 157 years more recent than the mean of the Beta dates), the less than 10% chance that the dates are drawn from the same population, and the fact that the sMu date is on highly preferable short-lived grass and twig material from roof fall, whereas all of the Beta dates are on less preferable posts (possibly aged wood), it is thought that the single sMu date is probably the far more accurate assessment for the age of xu18. our preference is to use this date n u m b e r 4 7 ? 7 3 alone for xu18 and not to use the Beta dates from the same context. the two remaining PVD dates from xu70 (Eth- 11037 and Eth-10418; outside the fortification) pass the test of contemporaneity (program CaLiB 3.0.3 in stuiver and Reimer, 1993; t'=2.62, df=1, critical x2=3.84), and yield a weighted mean of 1011?45 bp as a good estimate of the age of xu70. When this mean is compared with the single acceptable date for xu 18 (sMu-2736), they pass the test of contemporaneity (t'= 0.18, df=1, critical x2=3.84), as do all three dates when compared individu- ally (sMu-2736, Eth-11037, and Eth-10418, t'=2.81, df=2, critical x2=5.99). thus, when only the most accept- able and reliable dates from the site are considered, there seems to be no evidence for long-term occupation. it is optional whether the mean of the single sMu date from xu18 averaged with the mean of the two Eth dates from xu70 (mean=999?42 bp) is accepted or the mean of all three dates outright (1001?39) is considered to be the best assessment of the age of the site. in this case the latter is chosen, yielding a calibrated calendrical age of ad 1022 and a 2-sigma probability range of ad 982 to ad 1157 for the sommers site as a whole. Huff (32MO11) (Terminal Middle Missouri) Based upon the PELto criteria discussed earlier in this chapter, this site lacks any primary evidence for long- term occupation. the presence of one square house with rounded corners and La Roche (Extended Coalescent) pot- tery suggested to Wood (1967:105?107, 134?136, 159) that there was possibly some interaction between the huff villagers and contemporaneous Caddoan (arikara) groups from south Dakota. at the present time, huff is generally thought of as representing a single, relatively short-term occupation by terminal Middle Missouri peoples (stanley ahler, pers. comm., 2000). the huff series of 20 dates represents one of the largest and most diverse in the Middle Missouri subarea. Work- ing with only five of these 20 dates (si-178, 179, 180, 182, 183) Wood (1967:114?116) rejected several of them and posited a range of ad 1400 to ad 1600 for the occupation of the site. a later effort by ahler et al. (1996:182?183) discarded all but two dates (si-179 and usgs-29), yield- ing a calibrated curve intercept of ad 1441 and a 2-sigma probability range of ad 1406 to ad 1629, very close to Wood?s (1967:116) estimate for the site of ad 1400 to ad 1600. since then, two additional groups of dates have been run for the site. the first group of four is considered to have been produced as part of this program and utilized material excavated by Wood (DRi-3114, Eth-16071, Eth-16072, Eth-16073; table C.1). three averaged dates result in a 2-sigma range of ad 1288 to ad 1400 for house 6. one date from house 9 yields a 2-sigma range of ad 1321 to ad 1487, a value that is incompatible with the mean for house 6. a second group of six dates was produced in conjunction with a limited geophysical and archaeological testing program undertaken at the site in 1999 (ahler and Kvamme, 2000). six aMs dates were obtained on maize kernels or cobs from two pit features (ahler, 2000a, tables 4 and 5). the results of ahler?s analysis suggests that all six of these most recent dates pass the test of contemporaneity (using CaLiB 4.1.2) and therefore can be averaged. using CaLiB 3.0.3 (the version used in the current study), all six dates (Eth-21581 through Eth-21586) also pass the test of contemporaneity with this version of the program (t=8.31, df=5, critical x2=11.07). the weighted mean of bp 424?22 calibrates to an intercept of ad 1446 and a 2- sigma probability range of ad 1431 to ad 1496. ahler and Kvamme (2000:66) noted that these results are inconsistent with even the dates run in 1996. following ahler (2002), we will treat the mean of the six most recently run samples as the most reliable date for the site, and will exclude all other dates from further consideration. Other dated initial middle missOuri sites Jones Village (39CA3) (Initial Middle Missouri) there are ten radiocarbon dates available from this site, three from work stemming from the 1979 university of nebraska survey of the east bank of Lake oahe and six made available in 1997 as a follow-up to the smithsonian- sponsored PVD program. an earlier assessment of the site (C. Johnson, 1994:156, appendix D2; 1996:173?174, ap- pendix E) focused on the three university of georgia dates (uga-3359, uga-3360, uga-3358) available at the time this study was initiated and how they relate to the perceived site stratigraphy. at that time, there appeared to be two Middle Missouri tradition occupations at this site (initial and Ex- tended), each characterized by their own unique ceramic assemblages and different stratigraphic positions. it was concluded that there was PELto at the site. Johnson (1994, 1996) assigned each of the four dates from the site to either the initial (uga-3357) or Extended Middle Missouri (uga- 3358, 3359, 3360) component. the test of contemporane- ity for the three Extended variant dates indicated that they were drawn from different populations (t'=43.24, df=2, critical x2=5.99). a single date (uga-3359) was considered to be an outlier and, together with the other date from h 15, 7 4 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y was eliminated from the study for failing to pass the test of contemporaneity for this structure (t'=26.41, df=1, criti- cal x2=3.84). this left a single date (uga-3358) associated with h 10 of the proposed Extended component. this date (bp 1045? 65), had a calibrated intercept of ad 998 and a 2-sigma probability range of ad 869 to ad 1188. the single date from the initial Middle Missouri occupation (710?65 bp) corrected to a calibration curve intercept of ad 1288 and a 2-sigma probability range of ad 1214 to ad 1405. the second series of dates, run after the completion of the final report ( Johnson, 1996), consists of three conven- tional and two aMs dates on wood charcoal (mixed sizes, small diameter); a sixth date was run on charred residue from a cord roughened body sherd. all samples were col- lected in 1979 as part of the university of nebraska Lake oahe survey (falk and Pepperl, 1986). these samples were selected from provenience units (based upon a com- bination of depth of features and ceramic content) thought to be associated with the initial (DRi-3115, Eth-16074, Eth-16075, Eth-16076) and Extended (DRi-3116, DRi-3117) Middle Missouri occupations. surprisingly, all six dates form a fairly tight cluster from bp 881?37 to bp 980?65, indicating no radiocarbon support for the two- occupation hypothesis. Emergency salvage excavations conducted almost 20 years later from a different portion of the site (C. Johnson, 1997, 1998b) also provided little support for two Middle Missouri tradition occupations. Rather, a single component assigned to an initial Middle Missouri occupation is present. there is some evidence for PELto at the site, with two separate fillings of at least one abandoned house and extramural pit features origi- nating at various depths. in any case, there is no direct field evidence to suggest the six most recent dates are from provenience units dating to different occupations of the sites. as a consequence, all can be assigned to the initial variant occupation. the single sherd-residue date is ex- cluded from the analysis, consistent with decisions to set aside such dates from other sites based upon systematic differences between residue and non-residue dates (see later discussion). the five dates (DRi-3115, Eth-16074, Eth-16075, DRi-3116, DRi-3117) pass the test of con- temporaneity (t'=0.73, df=4, critical x2=9.49), yielding a weighted average of 899? 22 bp. this calibrates to cross- over dates of ad 1158 and ad 1185. the 2-sigma prob- ability ranges from ad 1036 to ad 1216. Fay Tolton (39ST11) (Initial Middle Missouri) this initial Middle Missouri village was the scene of conflict and apparently catastrophic abandonment (Wood, 1976). it is dated previously with four dates run by three labs (table C.2). one reported date, nWu-50, has a reported sigma of 170 years and is eliminated from consideration here for reasons discussed previously. a single PVD sample was submitted from this site, primarily to run a check on the previously run and apparently consistent dates for this important site, this time using short-lived materials maxi- mally conducive to accurate date association. five charred chokecherry seeds were submitted, having been recovered from inside a pottery vessel near Burial 18 in h 1. the two Wisconsin lab dates reported for the site (Wis-722 and 728) pass the test of contemporaneity (t'=0.18, df=1, critical x2=3.84), as would be expected because they are from the same intrasite context. When considered with the other two dates from the site (M-1082 and Eth-10113), all four pass the test of contempora- neity (t'=0.51, df=3, critical x2=7.82), and appear to be drawn from a single population. this is consistent with a lack of PELto for the site (Wood, 1976:42?43). these four dates have a weighted mean of 885?32 bp; this cali- brates to crossover dates of ad 1162, 1171, 1190 and a 2-sigma probability range of ad 1036 to ad 1258. this is taken as the best radiometrically determined age for the site (table C.3). Jandreau (39LM225) (Initial Middle Missouri) a single date existed beforehand for this initial Mid- dle Missouri site, this being an early run si date on a house post (si-337) (table C.2). this date is not considered to be useful because of its large standard error (150 years). suitable additional datable materials are practically non- existent for this site, so two sherd residue samples were submitted for dating in the PVD program. Because of a mix-up in the Eth lab, one sample (PVD-15) was not dated. the single date returned is 1015?55 bp; this cali- brates to ad 1020 and a 2-sigma probability range of ad 897 to ad 1188. an analysis of ceramic residue and non- residue dates from four sites (Jake White Bull, Calamity Village, arzberger, Jones Village) presented later in this chapter indicates that residue dates may be systematically earlier than nonresidue dates. as a result, all residue dates are excluded from further consideration, including the single date from Jandreau. Dodd (39ST30) (Initial Middle Missouri) an attempt was made to date the anderson focus component at this type site for the initial Middle Missouri variant. Prior to the PVD program, a single date existed n u m b e r 4 7 ? 7 5 on a house post (M-843). Because of the large standard error for this date (100 years), this date was not consid- ered to be useful in the present chronometric program. a large number of maize samples exist in the nMnh collec- tions for the Dodd site. the proveniences of these were re- viewed in detail, and a single sample was isolated as being associated within the initial Middle Missouri component rather than the Post-Contact period component. Parts of several maize cupules were submitted as an aMs sample. the results (Eth-10115) are so recent as to suggest that the sample was actually associated with the post-contact component at the site. the radiocarbon date of 160?55 bp falls in a portion of the calibration curve where mul- tiple crossovers occur and completely ambiguous calendri- cal ages are computed (ad 1683, 1735, 1807, 1934, and 1954). the 2-sigma probability dates range from ad 1671 to ad 1955. Pretty Head B (39LM232) (Initial Middle Missouri) PELto exists at this site, consisting of several super- imposed features, about 60 cm of midden deposit, varia- tions in pottery content, and variation in architectural patterns. the site investigators (Caldwell and Jensen, 1969:70?71) distinguished two components, B being early and a being late, which were not thought to be separated by a great length of time. two dates exist for this site, one each for the two initial variant components distin- guished at this location (table C.2). these two dates are compatible with each other, but one from component a has an unacceptably large sigma (140 years; si-166), the remaining single date from component B (si-165; 520?80 bp) suggests a relatively late initial Middle Missouri pe- riod of site use. two PVD samples were submitted from component B at the site, these being from the same con- text, a stratigraphically deep trash filled pit (f 68) beneath a substantial midden deposit. this context was not given a specific component assignment by Caldwell and Jensen (1969:11?12), but its stratigraphic position in the depos- its clearly indicates that it should date early in the period of site occupation, presumably with component B. one charred wood sample was submitted from this context, and a second sherd residue sample also was submitted, these providing some check on internal reliability. unfor- tunately, the sherd sample (PVD-54) could not be pre- treated at the sMu lab and was not submitted for aMs dating, leaving only a single PVD date for evaluation. the conventional date (sMu-2734) is almost 500 years earlier than the si date from component B (sMu-2734=1002?42 and si-165=520?80 bp). these two dates obviously do not pass the test of contemporaneity (t'= 25.31, df=1, critical x2=3.84). if it is assumed that they should be drawn from the same population, component B at the site, then one or both are obviously in error. subject to further evaluation if the archeological context is clarified, our choice is to accept the sMu-2734 date as the most accurate for the early component at Pretty head. this date has a calibra- tion curve cross-point of ad 1022 and a 2-sigma range of values from ad 903 to ad 1185. Cattle Oiler (39ST224) (Initial Middle Missouri) Cattle oiler has two components (initial Middle Mis- souri and Extended Middle Missouri variants), and five previous dates for the initial variant component. three dates have sigmas ranging from 100 to 140 years and are not considered to be useful for dating (si-316, 317, 318; table C.2). the remaining two dates (si-474 and si-475) do not pass the test of contemporaneity (t'=9.54, df=1, critical x2=3.84). Both dates are discounted as being un- reliable. this leaves no chronometric data for assessment except a single sample run in the present program (Eth- 11040) on charred reeds that dates 690?60 bp. Because of the relatively late age for this date and the presence of an Extended Middle Missouri component at the site, its con- text and association with the initial Middle Missouri com- ponent should be reevaluated if at all possible. this date calibrates to ad 1292 and a 2-sigma probability range of ad 1223 to ad 1407. Jiggs Thompson B (39LM208) (Initial Middle Missouri) there are two previous dates from the initial variant component at this site (table C.2), but both existing dates (i-1186, 1187) have 120-year sigmas that are considered of little value for the present chronometric analysis. a single charred maize cob sample was dated in the present program, yielding a date of 770?60 bp (Eth-11041). this calibrates to calibration curve intercepts of ad 1265, 1266, and 1277. the 2-sigma probability range is from ad 1128 to ad 1389. Other dated extended middle missOuri sites Havens (32EM1) (Extended Middle Missouri) two dates were previously run for this site, but both (M-2362 and 2363) have sigmas of 100 years (table C.2), eliminating them from further consideration here. two samples were submitted in the PVD program from the ha- vens site, these consisting of charred maize cob and cupule 7 6 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y fragments from two separate contexts (table C.1). the resulting dates (Eth-11047 and 11048) pass the test of contemporaneity (t'=0.44, df=1, critical x2=3.84) and are considered to be reliable. the weighted mean of these two dates is 588?41 bp; this calibrates to calendar age of ad 1332, 1342, and 1396 with a 2-sigma range of ad 1300 to ad 1422. Paul Brave (32SI4) (Extended Middle Missouri) three samples from the Extended Middle Missouri variant type site were submitted as part of the PVD pro- gram (DRi-3206, DRi-3207, DRi-3208). all samples are on unidentified wood charcoal from three separate under- cut pit features. two previous dates from the university of Michigan are eliminated because of high 1-sigma errors of 100 years. the three dates pass the test of contemoraneity (t"=1.75, df=2, critical x2= 5.99). the weighted average of the dates is 744?35 bp. this calibrates to a crossover of ad 1282 and a 2-sigma range of ad 1214 to ad 1302. Vanderbilt Village (39CA1) (Extended Middle Missouri) this site lacks any indication of PELto. there are five dates available, one of which is disregarded because it has a 1-sigma error of 105 years. the remaining dates (uga- 3355, DRi-3113, Eth-16069, Eth-16070) pass the test of contemporaneity (t'=3.54, df=3, critical x2=7.82) with a weighted average of bp 685?30. a systematic examina- tion of university of georgia dates from three other sites included in this study (sommers, Jake White Bull, Whis- tling Elk) indicates that the georgia dates are unreliable for a number of reasons. With this in mind, the single georgia date (uga-3355) is excluded from this analysis in favor of using the three remaining dates. these three dates pass the test of contemporaneity (t'=1.19, df=2, critical x2=5.99), yielding an average date of 710?34 bp. When this date is calibrated it gives a crossover date of ad 1288 and a 2-sigma range of ad 1257 to ad 1391. Calamity Village (39DW231) (Extended Middle Missouri) a single previous date exists for the Calamity Village site (si-375; table C.2). having a standard error of 200 years, it is considered to be non-useful for dating. two new samples are dated from this site, one comprised of a carbonized maize cob fragment and the second consist- ing of charred sherd residues (tables C.1 and 7). the two dates (Eth-10110 and 10111) pass the test of contempo- raneity (t'=2.09, df=1, critical x2=3.84). Lacking PELto for the Calamity Village site, it is legitimate to average the two dates for estimation of a calendrical age for the site. the weighted mean is 729?45 bp; this calibrates to a calendar age of ad 1284 and a 2-sigma probability range of ad 1215 to 1389. the sherd residue date from this site is 130 years or about 20% older than the maize date from the site. Because the sherd residue dating is taken as ex- perimental in nature (see later discussion), the maize date might be considered to be the more accurate of the two. this date (bp 665?60) alone (Eth-10111) calibrates at ad 1298 with a 2-sigma probability range of ad 1262 to ad 1411. although it is a single date, it is currently ac- cepted as the most reliable date from the site. Sully School (39SL7) (Extended Middle Missouri) no dates had been run from the sully school site pre- vious to the PVD program. two aMs samples were sub- mitted for dating, each consisting of carbonized residues on potsherds. there is no indication of PELto at this site, and the two dates (Eth-11051 and 11052, tables C.1 and 7) pass the test for contemporaneity (t'=0.06, df=1, criti- cal x2=3.84). the weighted mean of these dates is 785?42 bp; this calibrates in calendar age to 1262. the 2-sigma probability range is from ad 1162 to ad 1293. an analysis of ceramic residue and nonresidue dates from four sites (Jake White Bull, Calamity Village, arzberger, Jones Vil- lage) presented later in this chapter indicates that residue dates may be systematically earlier than nonresidue dates. as a result, all residue dates are excluded from further consideration, including the two from sully school. Cheyenne River (39ST1) (Extended Middle Missouri) three village components are present at this site, an Extended Middle Missouri occupation represented by rectangular houses, and Extended and Post-Contact Co- alescent components with circular houses. a review of the maps, aerial photos, and notes available from this site per- mits the assignment of several excavated house units to the Extended Middle Missouri component. three different labs had previously run a large series of 10 dates on sam- ples from contexts apparently assigned to the Extended Middle Missouri variant component (table C.2). Lacking a detailed report on the site, it is assumed that the village as a whole exhibits PELto, but that subparts of the site assigned to discrete components based upon architectural association do not exhibit PELto. therefore, dates from individual components are evaluated with the expectation that they were drawn from a single population. n u m b e r 4 7 ? 7 7 three of the existing dates (M-840, i-581, and si-119) have standard error values of 100 years or greater, and these are not considered further here for reasons stated above. the remaining seven dates are from three different house contexts, and it is of value to evaluate the samples by house context. two dates from two different labs (i-582 and si-116, table C.2) were run on post material from f 5, an Ex- tended Middle Missouri variant house. these two dates do not pass the test of contemporaneity (t'=17.80, df=1, critical x2=3.84), and given that these samples are from the same house context, both are dismissed as unreliable. four of the five remaining dates are from f 34, an Ex- tended Middle Missouri variant house. a single PVD sam- ple from this same house context is dated (tables C.2 and 7), and it is of interest to evaluate the dates from this single context. Multiple PVD dates would have been submitted from this context, except that no additional short-lived ma- terials could be located for dating. the four smithsonian institution lab dates from this house (si-12, si-17, si-117, and s-118) technically pass the test of contemporaneity at the p=0.05 level, yet the results (t'=7.36, df=3, critical x2=7.82) indicate that there is only about a 7% chance that the dates as reported reflect the same population. if the single additional si lab date from this same context (si-119), previously excluded because of its high standard error value, is re-included then the five si dates from f 34 fail the test of contemporaneity (t'=14.24, df=4, critical x2=9.49), this despite the high error factor associated with the fifth date. together, these results indicate little confi- dence should be placed in the si date series from f 34 as a whole, even if the four dates with smaller sigmas pass the test of contemporaneity at the p=0.05 level. for the sake of further comparisons, the small-sigma si dates from f 34 (si-12, 17, 117, and 118) yield a weighted mean of 889?32 bp that is compared statistically with the single PVD date from this same feature (Eth-10413; 675?50 bp). this weighted mean and the single PVD date do not pass the test of contemporaneity (10.68, df=1, criti- cal x2=3.84), indicating that one or the other should be rejected, lacking PELto from this context. given the wide dispersion exhibited by si lab dates that were compared, only the Eth date is accepted as the best temporal esti- mate for f 34 within the Cheyenne River site. a single si lab date exists that has not been evaluated, this being si-15 from f24. this context cannot be con- firmed to be an Extended Middle Missouri house location. thus, its association is less that certain. although an evalu- ation of si-15 and the single PVD date, Eth-10413, indi- cates that they pass the test of contemporaneity (t'=2.24, df=1, critical x2=3.84), there is a reluctance to give this date equal weight to the Eth date. the preference, given the results of overall assessment of date reliability from various labs, is to use the Eth-10413 date alone as the best estimate of age for the Extended Middle Missouri component at Cheyenne River. this date (675?50 bp) cali- brates to a calendar age of ad 1295 and a 2-sigma prob- ability range of ad 1264 to ad 1405. Black Widow Ridge (39ST203) (Extended Middle Missouri) Both an Extended Middle Missouri variant compo- nent and a Post-Contact Coalescent variant component are present at this site. the site is not reported, and it is assumed that the ceramic sample and dated material that are associated with the Extended Middle Missouri com- ponent lack PELto. two samples were selected for dat- ing from this site, one comprised of charred maize cob fragments and the second a substantial layer of residue on the interior of a ceramic body sherd. the sample pair was selected to be a clear test comparison of short-lived food remains and what appeared to be well preserved and abundant sherd residue material. unfortunately, the sherd residue sample could not be pretreated and submitted as originally planned, because of the unanticipated closing of the sMu lab. the charred maize cob fragment was dated, yielding a value of 555?55 bp (Eth-11053; tables C.1 and 7). this date yields a calibrated age of ad 1409 and a 2-sigma range of dates from ad 1300 to ad 1441 as the best and only estimate for the age of this site. terminal middle missOuri sites Shermer (32EM10) (Terminal Middle Missouri) two samples were submitted in the PVD program from the shermer site. the site has not been previously dated. Excavation data from the site (sperry, 1968b) in- dicate the occupation may have extended for a few or even several decades, based upon the density of storage pits within some houses, about 1.8 feet of midden in one area, and the extensive rebuilding of houses 4 and 6. the dated samples consist of fragments of charred maize cu- pules from two subsurface features associated with two different excavated houses at the site (table C.1). the two dates produced (Eth-11045 and 11046) fail the test of contemporaneity (t'=3.96, df=1, critical x2=3.84). two additional dates from the site were obtained in an attempt to clarify its chronological position. these were run on charred wood (DRi-3205) and maize (Eth-16394). the 7 8 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y test of contemporaneity was performed on all four dates, resulting in the rejection of the hypothesis that they date the same event (t'=22.09, df= 3, critical x2=7.82). Be- cause Eth-16394 was much later than the other dates, a remaining portion of the maize used for this date was re-run, producing a later date (EtRh-17511). this date can be substituted for Eth-16394 and averaged with the three other dates. these dates also fail the test of contem- poraneity at the 95% level of confidence (t'=11.07, df=3, critical x2=7.82). an internally consistent set of dates has apparently not been obtained from the site. this may be due to long-term occupation, as suggested by information from excavation. initial cOalescent sites Arzberger (39HU6) (Initial Coalescent) the arzberger site lacks PELto, so it is anticipated that the dates from it will pass the test of contemporaneity. two dates previously exist for this location (M-1126 and 1126a), each originally reported at two sigma but listed with their 1-sigma values in table C.2. two additional dates were run from arzberger in the PVD program, one on charred residue on a potsherd and the second on un- charred wild sunflower seeds. one of the Michigan dates (M-1126a) is eliminated from further consideration be- cause of a high 1-sigma error. this leaves a single Michigan date and two PVD dates to be tested for contemporaneity. the test indicated all were drawn from the same popula- tion (t'=2.91, df=2, critical x2=5.99). the two Eth lab dates also pass the test of contemporaneity (t'=1.41, df=1, critical x2=3.84) and yield a weighted mean of 391?42 bp. a weighted mean of the three dates was computed, yield- ing a date of 416?37 bp. this calibrates to calendar age of ad 1449 and a 2-sigma probability range of ad 1428 to ad 1628. it is of note that the date on the sherd residue from the arzberger site (Eth-10114; 440?55 bp) is 100 years or 29% older than the date on short-lived food re- mains (Eth-11042; 340?60 bp). this general relationship pertains in several other instances in which sherd residues were aMs-dated (see the analysis later in this chapter). as a result of this analysis, all residue dates are excluded from this analysis, including the one from arzberger. the test of contemporaneity between the two remaining dates (M-1126 and Eth-11042) passes at the 95% probability level (t'=2.62, df=1, critical x2=3.84) with an average of 400?48 bp. Calibrating this date yields crossover points of ad 1454, 1457, 1478, and a 2-sigma range of ad 1433 to ad 1636. Lynch (25BD1) (Initial Coalescent) this is an initial Coalescent variant site in nebraska that lacks PELto. three dates were run previously on samples from Lynch, and two of these three are dismissed because of large standard errors (M-1127 and M-1128; table C.2). this leaves a single previous date, M-842, 250?75 bp, for further evaluation. a single PVD date was run for the Lynch site, this being an aMs assess- ment on charred residue on pottery (Eth-11049; tables C.1, C.2). this date is quite divergent from the existing Michigan lab date (the former is 780?55 bp). Considered together, these two dates fail the test of contemporaneity (t'=30.48, df=1, critical x2=3.84). one or both of these dates is undoubtedly inaccurate. the Eth date is much closer to what are considered to be valid dates for the early part of the iC variant in the region (e.g., at Whistling Elk, table C.2), particularly when sherd residue dates may in fact date systematically 20% to 30% too old (see the discussion at the end of this chapter). With adjust- ment for this systematic error in the sherd residue date, the age of Lynch might be in the range of 550 to 625 bp. this would place the site around the dates for Whistling Elk and Crow Creek (table C.3). thus, the Eth date is considered to be closer to an accurate estimate of the age of the Lynch site. nonetheless, to be consistent with other decisions made on residue dates in this study, this date is excluded from any further discussion. the single Michigan date is rejected as unrealistically late, leaving the site undated. Other extended cOalescent sites Demery (39CO1) (Extended Coalescent) the Demery site is of particular interest because it is the northernmost component of the Extended Coalescent variant, with the possible exception of the Elbee site near the mouth of the Knife River in north Dakota (Woolworth and Wood, 1964; ahler, 1984a). the ceramic collection from this location also is of interest. Various assessments by Johnson (this report) and Woolworth and Wood (1964) judge the ceramic collections to be typical Extended Co- alescent pottery. surface collections examined by ahler (admittedly, less intensively and with a less experienced eye) appear to have strong similarities to Extended Mid- dle Missouri pottery, particularly in the high frequency of what appears to be Riggs ware. so, the chronological placement of the Demery site is important to the overall interpretation of cultural change in the area. n u m b e r 4 7 ? 7 9 a single previous radiocarbon date exists for the site (usgs-168; table C.2). two conventional dates were run in the PVD program, consisting of charred wood selected for short-lived small twigs in one sample, and charred maize cobs in the other (sMu-2800 and 2789, respectively; tables C.2 and C.3). there is no indication of PELto at the Demery site, so all dates would be expected to pass the test of contemporaneity. the two sMu lab dates do in fact pass the test (t'=0.85, df=1, critical x2=3.84). all three dates considered together also pass the test (program CaLiB 3.0.3 in stuiver and Reimer, 1993; t'=1.35, df=2, critical x2=5.99). it is reasonable to compute a weighted mean for all three samples, which is 523?31 bp. this cal- ibrates to a calendrical date of ad 1418 and a 2-sigma probability range of ad 1330 to ad 1442. Potts Village (39CO19) (Extended Coalescent) no dates had previously been run for the Potts site. two aMs dates on maize were run in the PVD program (Eth-10120 and 10121; tables C.2 and C.3). there is no indication of PELto for the Potts site, so the expectation is that the two dates are drawn from the same population. these two dates do not pass the test of contemporane- ity (program CaLiB 3.0.3 in stuiver and Reimer, 1993; t'=5.07, df=1, critical x2=3.84). therefore, both dates are rejected as unreliable. Meander (39LM201) (Extended Coalescent) there are no existing dates from the Meander site. three aMs dates were run in the present program, two on charred maize fragments and third on a charred fruit pit (Eth-10116, 10117, and 10118, respectively; tables C.2 and C.3). the archeological evidence from the site is too limited to fully evaluate the presence or absence of PELto. all three dates pass the test of contemporaneity (t'=2.27, df=2, critical x2=5.99). it is therefore reason- able to average the dates; the weighted mean of 338?33 bp calibrates to a series of four curve intercepts (ad 1524, 1564, 1575, and 1627) and a 2-sigma probability range of ad 1478 to ad 1647. Hosterman (39PO7) (Extended Coalescent) no dates had previously been run for the hosterman site. Charred botanical remains, particularly maize, are abundant in the existing archaeological collections, leading to the expectation that sound dates were possible for this site. PELto is lacking for hosterman, with the expectation that all dates derive from a single population. three conven- tional samples of charred maize cobs from a single pit fea- ture were dated (sMu-2728, 2731, and 2732), and a single aMs date (Eth-10119) was produced on a charred maize cob fragment from a different pit feature context within the site (table C.2, table 7). the three conventional samples are from successive 0.5 ft thick levels in a single, deep pit feature. they are from a large concentration of corn cobs, and likely were deposited as a single depositional event. it is expected that these three dates would cluster tightly and be drawn from the same population. this expectation is not met. these three dates (sMu-2728, 2731, 2732) fail the test of contemporaneity (t'=36.19, df=2, critical x2=5.99), and furthermore, the trend in relative ages is directly the oppo- site of what would be expected based upon their relative stratigraphic position within the pit feature. therefore, all three of these dates must be rejected as unreliable. an explanation for the variance in the dates is not readily ap- parent. one possibility is the presence of some form of petroleum-based contaminant in the samples, perhaps in the form of a solvent and hardener applied to the corn cobs. such a preservative was not visible on the specimens, but use of a chemical hardener may have been deemed ap- propriate to stabilize the large numbers of exceptionally well-preserved cobs in the samples. the fourth PVD program date from the site, Eth- 10119, 295?55 bp, from a different context, is for the time being treated as the only reliable date from the hosterman site. it seems inappropriate to compare this date with the other three from the site run by the sMu lab. should other dates become available from this site, they may prove ap- propriate for use in tests of reliability. this date calibrates to a calendar age of ad 1643 and a 2-sigma probability range of dates from ad 1450 to ad 1676. for the time being, this is used as the best age estimate for the site al- though it is very late for an Extended Coalescent site if it extends into the seventeenth century. Sully (39SL4) (Extended Coalescent) Excavations at the sully site, although not reported, seem to document a relatively long and continuous his- tory of occupation by successive Extended Coalescent and Post-Contact Coalescent peoples. this is the picture gleaned through the study of site records, ceramic collec- tions, and a partially completed manuscript on the site and other available information. therefore, the site as a whole exhibits PELto. it is most useful, therefore, to attempt to isolate and date discrete components at sully, components 8 0 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y within which it is a reasonable assumption that PELto is lacking. no previously run radiocarbon dates are available for the sully site. four samples were prepared during the PVD program, and three were dated (table C.1). two samples (one conventional sample of charred maize cobs and a sec- ond aMs sample of uncharred squash seeds) (sMu-2726 and Eth-10417, respectively; table C.1) were drawn from individual cache pits in a single excavation context thought to represent a later Extended Coalescent compo- nent at the site. in terms of the ceramic ordination to be presented, these two pits fall within the ?Middle? com- ponent at the site. a single aMs date (Eth-10416) was produced on charred maize from a context also assigned to this middle of three occupational periods at sully. all three dates pass the test of contemporaneity (t'=2.85, df=2, critical x2=5.99). the average of these three dates is 337?34 bp, computing to calibration curve intercepts of ad 1525, 1563, and 1628. the 2-sigma probability range is ad 1477 to ad 1648. Over?s La Roche (39ST9) (Extended Coalescent) this is the final Extended Coalescent site to be con- sidered in the discussion of PVD program results. the site investigator (hoffman, 1968) chose to divide the sev- eral houses excavated at the site into two components, a (later) and B (earlier), both Extended Coalescent compo- nents. this distinction is based partly upon architectural evidence, partly on ceramic evidence, and partly on radio- carbon dates from the site. the site was cultivated prior to excavation, and there was no indication of PELto from any excavated part of the site. for this study, the compo- nent separation provided by hoffman (1968) is used, with the expectation that the radiocarbon dates within each component are drawn from a single population. five dates had been run previously on materials from the La Roche site. three are associated with component B (si-95, 104, and 169; table C.2). the si-169 date is removed from consideration because of the large standard deviation of 120 years, based upon reasoning discussed above. the two remaining dates from component B do not pass the test of contemporaneity (t'=3.89, df=1, critical x2=3.84 ). these two dates are therefore rejected as being unreliable. three samples were dated from component B in the PVD program, these consisting of charred maize cupules and charred fruit pits (Eth-10414, 10415, and 11039; tables C.1, C.2, C.3). these dates just pass the test of contemporaneity (t'=5.02, df=2, critical x2=5.99). a weighted mean is computed for these three dates, 330?32 bp. this calibrates to ad 1528, 1558, and 1631. the 2- sigma probability range varies from ad 1485 to ad 1649. this is the best age estimate for component B at the site. two dates were previously run on Component a (table C.2). these two dates (si-97 and 106) pass the test of contemporaneity (t'=0.06, df=1, critical x2=3.84). it is appropriate to average these two dates. their weighted mean is 301?41 bp. this calibrates to a calendrical age of ca. ad 1641, with a 2-sigma probability range of ad 1484 to ad 1668. this is the best estimate of the calendar age of component a. it is of note that the radiocarbon dates do not reflect any measurable or significant difference in the age of component a and component B, despite the ar- chitectural differences associated with this distinction by the site investigator (hoffman, 1968). a test of contempo- raneity between all of the dates from the sites, excluding si-169 with its 120 year 1-sigma value, supports this con- clusion (t'=5.38, df=6, critical x2=12.59). in this case, the weighted mean of the seven dates is 319?25 bp. sites nOt dated by the plains village dating prOgram the following sites were not dated by the Plains Vil- lage Dating (PVD) program sponsored by the smithson- ian institution. these sites and their existing date data are included in this evaluation because of their usefulness in assessing the results of the ceramic ordinations that follow this presentation. they also are considered because they help to round out the dating of many components and some taxonomic units not included in the PVD program. all individual dates are listed in table C.2 and averaged in table C.3 (appendix C). Broken Kettle West (13PM25) (Initial Middle Missouri-Great Oasis) there are nine dates available from this site, all run by the university of Wisconsin (table C.2). the site contains a number of rectangular houses; C. Johnson (1974) reported the partial contents of one (house 4). there is no evidence for PELto at the site, so all nine dates are assumed to be drawn from the same population. these dates fail the test of contemporaneity (t'=21.97, df=8, critical x2=15.51). sepa- rating the dates by structure (h 2 or h 3) and running and testing for contemporaneity fails to produce two series of dates from either one that passes this test. the final proce- dure that is carried out with this series of dates is to test for contemporaneity between three dates (Wis 452, 488, 499) from feature 25 within house 3. these three dates pass the n u m b e r 4 7 ? 8 1 test of contemporaneity (t"=1.26, df=2, critical x2=5.99) and are accepted as the average date (912?34 bp) from the site. Calibrating this date yields correction curve intercepts of ad 1075, 1076, and 1155, and a 2-sigma probability range of ad 1031 to ad 1212 (table C.3). Williams (13PM50) (Initial Middle Missouri-Great Oasis) two dates are available from this site that lacks evi- dence for PELto. When these dates are entered into the CaLiB program for testing contemporaneity, they pass with an average date of 889?40 bp (t'=0.00, df=1, critical x2=3.84). Calibrating this date produces intercepts of ad 1161, 1172, and 1189, and a 2-sigma probability range of ad 1033 to ad 1243. Larsen (13PM61) (Initial Middle Missouri-Great Oasis) this site has a ceramic assemblage consisting of a mixture of pottery typically found at great oasis and Mill Creek sites (D. R. henning and King, 1996). it is unclear whether there are two components or a single one repre- sented by a mixture of these ceramics. the evidence sug- gests that the site does not possess PELto. as a result, all four available dates were entered into the testing portion of CaLiB, passing the test of contemporaneity (t'=0.13. df=3, critical x2=7.82). this yields an average of 727?30 bp. Entering this date into the CaLiB calibrating pro- gram yields a calibrated curve intercept of ad 1285 and a 2-sigma probability range of ad 1222 to ad 1386. Cambria (21BE2) (Initial Middle Missouri-Cambria) the two dates available from this site have 1-sigma errors in excess of 100 years and therefore were eliminated from any further consideration. Jones (21BE5) (Initial Middle Missouri-Cambria) there are five dates available from this site. three have 1-sigma errors 100 years or greater and can be eliminated from the analysis. the two dates (Beta 83242 and 113877) with 1-sigma errors less than 100 years were entered into the test for contemporaneity. the dates fail this test (t'=3.87, df=1, critical x2=3.84) and therefore cannot be averaged. Price (21BE25) (Initial Middle Missouri-Cambria) there are three dates available from this site, all from a single pit feature. the site lacks evidence of PELto. these dates pass the test of contemporaneity in the CaLiB 3.0.3 program (t'=1.87, df=1, critical x2=3.84). the av- erage of 910?48 bp yields an intercept of ad 1156 and a 2-sigma probability range of ad 1022 to ad 1256. Great Oasis (21MU2/17) (Initial Middle Missouri-Great Oasis) the two dates from this site, which is designated by two different names and site numbers, appear to be associated with its major component. this great oasis component ap- pears to lack PELto. there is evidence to suggest that the site was much less intensively occupied by Late Woodland and Cambria peoples, although no comprehensive report is available to evaluate the overall occupational history of the site (anfinson, 1987:161?164). the two dates from the site pass the test of contemporaneity (t'=0.65, df=1, criti- cal x2=3.84), indicating that they can be averaged together to produce a composite date. this date (1015?47 bp) pro- duces a calibration curve intercept of ad 1020 and a 2- sigma probability range of ad 899 to ad 1158. Packer (25SM9) (Initial Middle Missouri-Great Oasis) three dates are available from a single storage pit from this site. as a result, the dates are taken as repre- senting a short-term occupation. all other evidence from the site suggests an absence of PELto, although there are Woodland and Central Plains tradition components pres- ent. the test for contemporaneity between the dates in- dicates that they can be averaged (t'=0.47, df=2, critical x2=5.99). this date, ad 937?42, when entered into the calibration portion of CaLiB yields intercepts of ad 1057, 1088, 1121, 1139, and 1151. the 2-sigma probability range is ad 1022 to ad 1208. Crow Creek (39BF11) (Initial Middle Missouri, Initial Coalescent) this site contains an initial Middle Missouri and ini- tial Coalesecent component. Both components lack evi- dence of PELto. three dates are available from the initial Middle Missouri occupation. one of these dates (M-836) has a 1-sigma value of 100 years and is eliminated from further consideration. the remaining two dates pass the test of contemporaneity (t'= 0.18, df=1, critical x2=3.84), indicating that they can be pooled together to form a com- posite date of 875?59 bp. Calibrating this date produces intercepts of ad 1164, 1167, 1194, 1198, and 1207, and a 2-sigma probability range of ad 1030 to ad 1276. 8 2 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y two dates are associated with the initial Coales- cent occupation of the site (M-1079a, Wis-1074). When tested for contemporaneity, both dates are drawn from the same population (t'= 0.26, df=1, critical x2=3.84). the weighted average (592?47 bp) corrects to calibration curve intercepts of ad 1331, 1343, and 1394, and a 2- sigma probability range of ad 1297 to ad 1423. Swanson (39BR16) (Initial Middle Missouri) By most criteria (thick middens, multiple superimposed features), this site lacks PELto. there is some evidence to suggest that several house-building episodes occurred at the site, although the period is unknown (hurt, 1951:3). this includes the re-use of posts from earlier houses in the con- struction of later ones. the 13 dates from the swanson site represent one of the largest series available from the subarea. one date, M-839, has a 1-sigma error of 125 years and is eliminated from further consideration. When entered into CaLiB?s date averaging program, the remaining 12 dates fail the test of contemporaneity (t'=85.52, df=11, critical x2=19.68). Dividing them by provenience unit (houses 1 and 2) also suggests that the dates within each of the two structures were drawn from a different population and can- not be averaged. a further subdivision is accomplished by testing for contemporaneity between dates from the same posts within h 2. three dates each are available from post C (Wis-526, 553, 660) and post D (Wis-524, 551, 650) within h 2. the post C dates pass the test of contemporene- ity (t'=2.82, df=2, critical x2=5.99), yielding an average of 891? 34 bp, curve intercepts of ad 1160, 1173, and 1188, and a 2-sigma range of ad 1033 to ad 1256. the post D dates also pass the test of contemporaneity (t'=.04, df=3, critical x2= 7.82). a weighted average of these dates yields a value of 1090?39 bp, that translates into a calibrated inter- cept of ad 984 and a 2-sigma range of ad 877 to ad 1024. there is no overlap in the 2-sigma ranges between these two series of dates and the intercept values differ by about 200 years. this suggests that there is a strong possibility that the posts used to construct h 2 are of different ages. the present tentative interpretation is to view post D as aged wood from either a previously occupied house or from another source. Post C might represent the true age of the structure. for now, both series of dates are accepted for the village, suggesting that it was occupied on at least two occasions. Arp (39BR101) (Initial Middle Missouri) the two dates from this site have 1-sigma errors of 100 years or greater and are eliminated from further con- sideration. Mitchell (39DV2) (Initial Middle Missouri) this site apparently has PELto based upon excavations in 1996 (see Winham, 1996). Parts of the village have some midden accumulation attributable to intentional banking of deposits against the sides of houses (Baerreis and alex, 1974:144?145). there are 10 dates available from the site from varying contexts, most from h 4. it is assumed that these dates are from a single, relatively short-term event as- sociated with the occupation of the house. Considering all dates, a run through the test of contemporaneity portion of the CaLiB program fails to indicate that they were drawn from the same event (t'=23.92, df=9, critical x2=16.92). there also is no apparent trend in the dates and their depth below the surface within h 4. there is no rationale for test- ing two or more dates from h 4 based upon the currently available information from the site (there is no compre- hensive report of excavations at the site). the earliest and most divergent date of the series (Wis-570) is based upon willow wands, a short-lived material expected to date later than the wood charcoal dates. it was decided to arbitrarily exclude Wis-570 from consideration, realizing that this is not a completely acceptable procedure. in reality, this re- jected date may be more closely tied to the occupation of the site, based as it is upon a short-lived material. the test of contemporaneity of the remaining nine dates results in accepting them as dating the same event (t'=12.39, df=8, critical x2=15.51). the weighted average of these dates is 908?20 bp, not very different when all 10 dates were aver- aged (887?19 bp). When the former average is calibrated, it results in an intercept of ad 1156 with a 2-sigma prob- ability range of ad 1036 to ad 1212. St. John (39HU213) (Initial Middle Missouri-Great Oasis) only one date is available from this site (1180?60 bp). it corrects to an intercept of ad 881 and a 2-sigma prob- ability range of ad 692 to ad 990. Bloom (39HS1) (Initial Middle Missouri) a single date of 1050?50 bp is available from this site. When calibrated, the date has an intercept of ad 997 and a 2-sigma probability range of ad 885 to ad 1153. Twelve Mile Creek (39HT1) (Initial Middle Missouri) this site apparently contains no indication of PELto, although no formal site report exists. there are two dates available from the site that, when entered into the CaLiB n u m b e r 4 7 ? 8 3 program?s test of contemporaneity, suggests that they are drawn from separate populations (t'=9.47, df=1, critical x2=3.84). as a result, they are eliminated from any further consideration in this study. King (39LM55) (Initial Middle Missouri) this site lacks any indications of PELto. therefore, the two dates are entered into the test of contemporaneity. as a result, both dates are accepted as being drawn from the same population (t'=0.08, df=1, critical x2=3.84). the averaged date of 842?45 bp has an intercept of ad 1214 and a 2-sigma probability range of ad 1045 to ad 1282. Antelope Dreamer (39LM146) (Initial Middle Missouri) there are six dates from this site that lacks evidence for PELto (toom, 1990). this series of dates is rather unique for it represents the first attempt in the Middle Missouri subarea to run dates on corn and wood charcoal from the same site. all corn dates are from h 11, whereas the wood charcoal dates are from a different structure, h 15. toom?s (1990, tables 92, 93) segregation of these dates is main- tained for the purposes of this report. the test of contem- poraneity of the three corn dates (uCR 2308, 2309, 2310) indicates that they are dating the same event (t'=1.08, df=2, critical x2=5.99) as does the test of the wood char- coal dates (uCR 2311, 2312, 2313) (t'=0.33, df=2, critical x2=5.99). the average corn date of 704?31 bp calibrates to an intercept of ad 1289 and a 2-sigma range of ad 1261 to ad 1390. the weighted average of the wood charcoal dates is 827?49 bp, yielding an intercept of ad 1227 and a 2-sigma range of ad 1057 to ad 1288. a comparison be- tween these two dated series indicates a difference of about 50 years. When all six dates are combined, they pass the test of contemoraneity (t'=5.82, df=5, x2=11.10), yield- ing an average of 740?27 bp. Calibrating this date yields a 2-sigma probability average of ad 1222 to ad 1298 and a curve intercept of ad 1282, the accepted date for this site. Langdeau (39LM209) (Initial Middle Missouri) Based upon the description of this site by Caldwell and Jensen (1969:14?26), there is no indication of PELto. therefore, all three dates are considered in the test to de- termine if they are drawn from the same population. When this is done, all pass the test of contemporaneity (t'=2.19, df=2, critical x2=5.99). the resulting average of 871?37 bp calibrates to curve intercepts of ad 1165, 1195, and 1208, with a 2-sigma probability range of ad 1030 to ad 1263. Heath (39LN15) (Initial Middle Missouri-Great Oasis) a single date (i-9499) available from the site is 940?195 bp (Winham, Lueck et al., 1992:80). Because it has a 1-sigma error greater than 100, it is eliminated from further consideration. Over?s La Roche C (39ST9) (Initial Middle Missouri) this site, together with nearby Bower?s La Roche (39st232), is considered to be a single site complex. a major Extended Coalescent component and a smaller initial Middle Missouri occupation are present at the site. two of the three dates associated with the initial variant component from the site (si-170, 242) are excluded from further con- sideration because they possess 1-sigma errors of 100 years or more. this leaves a single date of 570?55 bp to date the initial Middle Missouri component at the site. By entering this date into the CaLiB 3.0.3 program, a calibration curve intercept date of ad 1406, with a 2-sigma range of ad 1299 to ad 1436, was produced. Because this sample is the most recent of all initial Middle Missouri dates and comes from a site with an extensive Extended Coalescent component, mixture with more recent materials cannot be ruled out. Breeden (39ST16) (Initial Middle Missouri, Post-Contact Coalescent) this site contains PELto in the form of initial Mid- dle Missouri and Post-Contact Coalescent components. all three dates are associated with the earliest or initial variant occupation of the site. there is no indication that the initial Middle Missouri occupation was a particularly lengthy one. applying the test of contemporaneity to these three dates indicates that they are not drawn from the same population (t'=7.04, d.f=2, critical x2= 5.99). as a result, the entire series is eliminated from further consideration. Eagle Feather (39ST228) (Initial Middle Missouri) a single date from this site, 760?40, is available and corrects to a calendrical age of ad 1279 and a 2-sigma probability of ad 1208 to ad 1378. the site also contains a dated Extended Coalescent component. Stony Point (39ST235) (Initial Middle Missouri, Post-Contact Coalescent) Very little is known from this unreported site except that it contains initial Middle Missouri and Post-Contact 8 4 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y Coalescent components. two dates (uCR 2314, 2315) are associated with the initial Middle Missouri occupa- tion that contains no indication of PELto. a test of con- temporaneity between the two dates indicates that they are drawn from the same population or date the same event (t'=3.36, df=2, critical x2=5.99). the weighted av- erage of the two dates is 774?42 bp, calibrating to inter- cepts of ad 1264, 1269, and 1276 and a 2-sigma range of ad 1163 to ad 1297. a single date (gx-13406) ap- pears to date the Post-Contact period occupation of the site. this date (295?75 bp), when run through the CaLiB program, produces a single calibration curve intercept of ad 1643 and a 2-sigma probability range of ad 1441 to ad 1955. Clark?s Creek (32ME1) (Extended Middle Missouri) this site lacks any evidence for PELto. there are three dates available from the site, two of which (M-2366, 2367) are rejected because of 1-sigma errors of 100 years or more. this leaves a single date of 750?50 bp that cor- rects to a calibration curve intercept of ad 1281 and a 2- sigma probability range of ad 1193 to ad 1389. White Buffalo Robe (32ME7) (Extended Middle Missouri) White Buffalo Robe contains Extended Middle Mis- souri and Post-Contact period, Knife River phase com- ponents. the five dates from the relatively short term Extended Middle Missouri occupation are the focus of this evaluation. these dates were included in a previous effort at averaging and calibration (ahler and haas, 1993, table 8.3). the dates were entered into the testing portion of the CaLiB 3.0.3 program to determine if they were dat- ing the same event. the results suggest that they are indeed drawn from the same population (t'=4.80, df=4, critical x2=9.49) and can be averaged. the resulting weighted av- erage of 660?28 bp corrects to a calibration curve inter- cept of ad 1299 and a 2-sigma probability range of ad 1285 to ad 1394. Bendish (32MO2) (Extended Middle Missouri) all except one of the four dates from this site were eliminated from further consideration because of 1-sigma errors equal to or greater than 100 years. the single re- maining date (730?80 bp) calibrates to an intercept of ad 1284 and a 2-sigma probability range of ad 1155 to ad 1412. a somewhat different approach to dating the site was taken by thiessen (1995:170?171). 32MO291 (Unclassified/Transitional Extended Middle Missouri?Terminal Middle Missouri) this site has a ceramic assemblage similar to huff and shermer in the heart region and scattered Village phase sites located further north in the Knife region (ahler and Metcalf, 2000:243?245). twenty-one dates are available from the site (14 aMs, 7 conventional) and, by a series of analyses, 11 of these were selected by the site investigators as best representing the period of occupation (ahler and Metcalf, 2000:89?93, table 5.6). further analysis indicated that these dates doc- ument perhaps intermittent occupations extending across decades, from early to late in the ad 1400s (ahler and Metcalf, 2000:93?98). although the 11 dates were not drawn from a single population, there is some value to averaging them to arrive at a central tendency to use in the present ceramic ordination study. the weighted average is 471?17 bp. Calibrating this average yields a crossover date of ad 1437 and a 2-sigma range of ad 1413 to ad 1451, the date that is used herein. Cross Ranch (32OL14) (Extended Middle Missouri) there are six dates available from this site. gx-19395 and Beta-66015 are from Milepost 28, a subarea within the site. three (M-2368, M-2369, gx-19395) are eliminated from further consideration because of 1-sigma errors of 100 years or greater. the remaining three dates are tested for their contemporaneity at this site that lacks PELto. the re- sults of this test indicate that they are drawn from the same population and can be averaged (t'=3.46, df=2, critical x2=5.99). the weighted mean of these three dates is 613?31 bp that calibrates to intercepts of ad 1328, 1350, and 1391. the 2-sigma probability range is ad 1299 to ad 1407. Fire Heart Creek (32SI2) (Extended Middle Missouri) the single date available from this site is one of those from the smithsonian institution chronology program (si-213). When calibrated, the 720?80 bp date yields a curve intercept of ad 1286. there is a 98% chance that the 2-sigma probability places the site in the range of ad 1186 to ad 1412. Ben Standing Soldier (32SI7) (Extended Middle Missouri) all seven dates from this site have 1-sigma errors of 150 years and therefore are eliminated from consideration in this study. n u m b e r 4 7 ? 8 5 South Cannonball (32SI19) (Extended Middle Missouri) the occupational history of this site was reconstructed by griffin (1984:108?111), who interpreted it to reflect a series of house rebuilding episodes and/or reoccupations in an attempt to explain the temporal variation in the ra- diocarbon dates. an examination of the site report indi- cates that there is little indication for PELto, despite the interpretation by griffin. the series of 14 dates from this site represents one of the largest in the Middle Missouri subarea. one of these, i-4205, is eliminated from further consideration because of a 1-sigma error of 100 years. the remaining 13 dates are evaluated in terms of their con- temporaneity, resulting in accepting them as being drawn from the same population (t'=18.54, df=12, critical x2= 21.03). the weighted mean of these dates is 655?20 bp and intercepts the calibration curve at ad 1301, 1374, and 1378. the 2-sigma probability range is ad 1289 to ad 1392. these dates are accepted as an approximate chrono- logical placement for the site. a somewhat different picture emerges when dates from separate houses are examined. two dates from h 1 (i-4202, Wis-1103) were tested for contemporaneity, resulting in the conclusion that they are drawn from dif- ferent populations (t'=6.18, df=1, critical x2=3.84). next, the two dates from h 2 are tested for contemporaneity, resulting in the conclusion that they also date different events (t'=5.56, df=1, critical x2=3.84). tests of con- temporaneity also were conducted for the four remain- ing houses at the site. houses 3 (t'=0.01, df=2, critical x2=5.99), 4 (t'=0.30, df=2, critical x2=5.99), 5 (t'=0.17, df =1, critical x2=3.84) and 7 (t'=0.08, df=1, critical x2=3.84) pass the test of contemporaneity of dates within each house. the weighted averages from these four houses exhibit about a 150-year spread from house 7 (585?51 bp), house 5 (642?58 bp), house 4 (647?43 bp) and house 3 (754?48 bp). the nine dates from these four houses, when tested for contemporaneity, results in accepting all as drawn from the same population (t'=6.67, df=8, criti- cal x2=15.51). the weighted average of these nine dates is 660?24 bp, only 5 years from the weighted average of the 13 dates discussed above. the approach to average the dates from houses 3, 4, 5 and 7 as outlined above, was advocated by thiessen (1995:173). Travis I (39CO213) (Extended Middle Missouri) two dates are available from this site that lacks any evidence for PELto. Both dates are from different fea- tures but are identical in value (780?70 bp). averaging the two dates together yields a value of 780?59 bp that corrects to calibration-curve intercepts of ad 1263, 1273, and 1275, and a 2-sigma probability range of ad 1069 to ad 1387. Indian Creek (39ST15) (Extended Middle Missouri) there are four dates (i-18,039, i-18,040, i-18,041, i-18,042) available from this multiple component site. the main component consists of a Post-Contact Coalescent, Bad River phase occupation. there are several houses and a fortification ditch that are attributed to a the earlier Ex- tended Middle Missouri occupation, a component that most likely contained more houses partially obliterated by the later occupation (see Lehmer and Jones, 1968:64?66). the dates (330?80 bp from feature 43, 450?80 bp from feature 30, 300?80 bp from feature 7, 330 ?80 from fea- ture 5) from a relatively tight cluster and are determined to be statistically the same when entered in CaLiB for the test of contemporaneity (t"=1.97, df=3, x2=7.81). this test yields an average of 351?40 bp. Calibrating this gives curve intercepts of ad 1519, 1595, and 1622 with a 2-sigma error range of ad 1452 to ad 1643. these clearly fall outside the range of dates from other Extended Middle sites (ad 1200?1400) and are earlier than the Bad River phase (ad 1650?1785). it is unclear why these dates are so late, although contamination from the later Bad River phase component cannot be eliminated. Based upon these concerns, the dates from the site are excluded from further consideration. Ketchen (39ST223) (Extended Middle Missouri) this site lacks PELto and has four dates associated with it. one of these (si-378) is excluded from further consideration because of a 1-sigma error of 140 years. the remaining three dates were input into the testing proce- dure of CaLiB 3.0.3, resulting in accepting all of them as dating the same event (t'= 1.86, df=2, critical x2=5.99). the weighted average of 782? 35 bp has calibration curve intercepts of ad 1263 and ad 1275, and a 2-sigma prob- ability range of ad 1208 to ad 1292. Durkin (39ST238) (Extended Middle Missouri) this site also lacks any indication of PELto. the two dates available from the site, when tested for contempora- neity, indicate that they were drawn from the same popu- lation (t'=0.03, df=1, critical x2=3.84). the weighted average date from the site is 647?42 bp and intercepts 8 6 ? s M i t h s o n i a n C o n t R i B u t i o n s t o a n t h R o P o L o g y the calibration curve at ad 1303, 1372, and 1383. the 2- sigma probability range is ad 1286 to ad 1405. Whistling Elk (39HU242) (Initial Coalescent) this initial Coalescent site is one of a few villages, to- gether with fay tolton (39st11), that is characterized by a very short-term occupation (steinacher, 1983). Whistling Elk also has a large and somewhat diverse series of dates. the seven uga dates, when tested for contemporaneity, indicate that they cannot be accepted as dating from the same event (t'=55.52, df= 6, critical x2=12.59). these results are similar to those arrived at by toom (pers. comm. to C. Johnson, 1991), suggesting that all but the three university of Califor- nia-Riverside (uCR) dates be excluded from further consider- ation. applying the test of contemporaneity to the three uCR dates indicates that they date the same event (t'=2.44, df=2, critical x2=5.99). the weighted mean computes to 638?40 bp, calibrating to intercepts of ad 1317, 1370, and 1386. the 2-sigma probability range is ad 1290 to ad 1405. Elbee (32ME408) (Extended Coalescent-Related) all indications from this site suggest that it lacks PELto. three dates are available from the site, all but one (sMu-797) being rejected by ahler and haas (1993, table 8.2). the single acceptable date (440?40 bp), when entered into CaLiB 3.0.3, has a calibration curve intercept of ad 1443 and a 2-sigma range of ad 1412 to ad 1624. Molstad (39DW234) (Extended Coalescent) there is no evidence from this site to suggest that it was occupied for a lengthy period because it lacks PELto. four dates are available from the site, with one being re- jected from further consideration because of a 1-sigma error of 100 years. another (i-721) also is excluded be- cause of possible contamination by modern organic ma- terials (hoffman, 1967:45). the remaining two dates were entered into the CaLiB 3.0.3 program?s test of con- temporaneity. the results indicate that the two dates are drawn from the same population (t=0.05, df=1, critical x2= 3.84). the resulting weighted mean of 365?45 bp calibrates to intercepts of ad 1494, 1601, and 1616. the 2-sigma probability range is ad 1448 to ad 1638. McClure (39HU7) (Extended Coalescent) the single date from this site is rejected from further consideration because of a 1-sigma error of 140 years. Little Pumpkin (39HU97) (Extended Coalescent) three dates are available from this site that lacks PELto. one date (sMu-2629) is eliminated from further consideration because of a 1-sigma error of 100 years. the remaining two dates pass the test of contemporaneity, in- dicating that they date to the same event (t'=1.30, df=1, critical x2=3.84). the resulting weighted mean (377?44 bp) has a calibration curve intercept of ad 1490 and a 2- sigma probability range of ad 1442 to ad 1637. Bowman (39HU204) (Extended Coalescent) a single date from this site run on corn has a ?mod- ern? date. Because isotope fractionation was not considered when the date was run, the ?modern? determination could be caused by the failure to correct for this factor. in any case, the date is eliminated from further consideration. 39SL24 (Extended Coalescent) the single date from this site (i-614) has a RCyBP date of 240?80 bp. Entering this date into the calibration program yields a curve intercept date of ad 1660, placing it just outside the acceptable range of the Extended Coalescent variant. the date has a 2-sigma probability range of ad 1489 to ad 1955. if the date is confined to the 81% probability range within the 2-sigma error, an earlier date of ad 1489 to ad 1823 results. the earlier half of this range, up to ad 1650 is acceptable. Eagle Feather (39ST228) (Extended Coalescent) the Extended component at this site also containing an initial Middle Missouri occupation is dated at 340?60 bp (uCR-2448). this translates into calibration curve in- tercepts of ad 1523, 1565, 1578, and 1627. the 2-sigma range is ad 1442 to ad 1660. Bower?s La Roche (39ST232) (Extended Coalescent) there are two dates from this site. si-214 has a 1-sigma error of 210 years, eliminating it from further consideration. the remaining date (si-215) is 710?90 bp, well outside the range of acceptable dates for the Extended variant of the Co- alescent tradition. it also is discounted in any further analysis. Rattlesnake Keeper (39LM160) (Post-Contact Coalescent) three dates are available from this poorly known site that lacks PELto. two of the three dates (uCR-2382b, n u m b e r 4 7 ? 8 7 2382) are rejected because of imprecise bp dates (i.e., <150 years). the remaining date of 210?60 bp has five calibra- tion curve intercepts (ad 1669, 1783, 1797, 1948, and 1952), with a 2-sigma probability range of ad 1528 to ad 1955. Ghost Lodge (39ST20) (Post-Contact Coalescent) this site has a single date of 175?80 bp associated with it. this date calibrates to curve intercepts of ad 1678, 1765, 1775, 1805, 1939, and 1954. the 2-sigma prob- ability dates range from ad 1531 to ad 1955. summary of carboNizeD sherD resiDue DatiNg there were four instances in which an aMs date was run on charred organic residues scraped from the surface of a pottery sherd, and for which there is at least one companion aMs date on other material from the same or a closely equivalent within-site context. these situa- tions allow comparison of the results of dating sherd resi- dues with dating of more typical aMs materials, such as charred plant remains. these instances are listed in table 9, showing the radiocarbon age of the more typical mean aMs age (or mean age) and the aMs date on charred residues, and the difference between these values in each paired case. al- though the number of examples is small, in three of four cases the dates on residues are 20% to 35% older than the companion dates on less experimental aMs material. this suggests that there may be a systematic error in the ages derived from charred residue on pottery, providing a direction for further evaluative studies. summary of plaiNs village DatiNg program results it is useful to provide a summary of the results of evalu- ation of the dates produced in the PVD program, as well as the evaluation of previously existing dates. table 10 gives such a summary, broken down by dating lab and showing the number of dates by lab that are rejected and accepted. three kinds of date rejection are listed, deriving from the evaluation process discussed herein. one (R-1) is be- cause of overly large standard errors associated with a date, as reported from the lab. if a standard error of 100 years or greater is reported, the date is dismissed as non-useful. a second type of rejection (R-2) comes from a failure to pass the test for contemporaneity in instances in which more than one date exists from a context lacking PELto. in this case, the date is rejected as being unreliable, on sta- tistical grounds, in essence having substantially too small an error factor reported with the date based upon relevant archeological evidence. a third type of rejection occurs (R- 3) when internal evidence within a series of dates from a single lab indicates they are drawn from the same popula- tion, passing the test for reliability, but that external evi- dence (often in the form of an equally or more compelling series of dates from the same context run by another lab) suggests that the dates from the first lab are inaccurate or invalid (the mean age is not an accurate estimate of actual context age). this occurs in several cases in which a series of dates from more than one lab were run on samples from the same context. two kinds of date acceptance occur. the first and more meaningful (a-1) is when multiple dates are run from a sin- gle context by a single lab, the context lacks PELto, and the dates pass the test of contemporaneity. in other words, mul- tiple dates cluster reasonably closely, as would be expected TabLe 9. summary comparisons of results of dating pottery sherd residues with more routine aMs dates from the same contexts. (n = number of samples.) Charred sherd Difference: routine aMS residue aMS +=residue older; Percentage Site date (n) date (n) ?=younger difference Jones Village 899 BP (5) 898 BP (1) ?1 year