University of Tennessee, Knoxville TRACE: Tennessee Research and Creative Exchange
Doctoral Dissertations Graduate School
12-2002
Biological Affinities of chaicAr Period Populations from West- Central Kentucky And Tennessee
Nicholas Paul Herrmann University of Tennessee - Knoxville
Follow this and additional works at: https://trace.tennessee.edu/utk_graddiss
Part of the Anthropology Commons
Recommended Citation Herrmann, Nicholas Paul, "Biological Affinities of chaicAr Period Populations from West-Central Kentucky And Tennessee. " PhD diss., University of Tennessee, 2002. https://trace.tennessee.edu/utk_graddiss/2123
This Dissertation is brought to you for free and open access by the Graduate School at TRACE: Tennessee Research and Creative Exchange. It has been accepted for inclusion in Doctoral Dissertations by an authorized administrator of TRACE: Tennessee Research and Creative Exchange. For more information, please contact [email protected]. To the Graduate Council:
I am submitting herewith a dissertation written by Nicholas Paul Herrmann entitled "Biological Affinities of chaicAr Period Populations from West-Central Kentucky And Tennessee." I have examined the final electronic copy of this dissertation for form and content and recommend that it be accepted in partial fulfillment of the equirr ements for the degree of Doctor of Philosophy, with a major in Anthropology.
Lyle W. Konigsberg, Major Professor
We have read this dissertation and recommend its acceptance:
Richard L. Jantz, Walter E. Klippel, Kenneth H. Orvis
Accepted for the Council:
Carolyn R. Hodges
Vice Provost and Dean of the Graduate School
(Original signatures are on file with official studentecor r ds.) To the Graduate Council:
I am submitting herewith a dissertation written by Nicholas Paul Herrmann entitled “Biological Affinities of Archaic Period Populations from West-Central Kentucky And Tennessee.” I have examined the final electronic copy of this dissertation for form and content and recommend that it be accepted in partial fulfillment of the requirements for the degree of Doctor of Philosophy, with a major in Anthropology.
Lyle W. Konigsberg Major Professor
We have read this dissertation and recommend its acceptance:
Richard L. Jantz
Walter E. Klippel
Kenneth H. Orvis
Acceptance for the Council:
Anne Mayhew Vice Provost and Dean of Graduate Studies
(Original signatures are on file with official student records.)
BIOLOGICAL AFFINITIES OF ARCHAIC PERIOD POPULATIONS FROM WEST-CENTRAL KENTUCKY AND TENNESSEE
A Dissertation Presented for the Doctor of Philosophy Degree The University of Tennessee, Knoxville
Nicholas Paul Herrmann December 2002
Copyright © 2002 by Nicholas Paul Herrmann All rights reserved.
ii Dedications
I wish to dedicate my dissertation to two individuals who have influenced my life in different ways:
First, I dedicate this to my mother who has supported me throughout my graduate studies. The “big paper” is finally finished.
Second, I dedicate this to Leonard Blake, who recently passed away. During my years at Washington University, Leonard was always in the Archaeology Lab working and answering questions. He was an incredible person, dedicated researcher and inspiration to all who knew him
iii Acknowledgements
This dissertation is the result of years of work. Without the help of numerous friends and colleagues this document would not have been possible. First, my committee provided invaluable advice and suggestions. Dr. Kenneth Orvis of the Department of
Geography has provided insightful questions and additional thoughts on how the Green
River shell mound populations should be viewed. Dr. Lyle W. Konigsberg has supported my research efforts throughout my tenure in the Department of Anthropology. Dr.
Konigsberg has included me in several research projects and provided numerous professional opportunities through presentation or publication. He also provided invaluable statistical advice and Fortran code for performing several of the analyses described in this dissertation. Dr. Richard L. Jantz has also supported my efforts.
Through several research projects and contracts, he has provided employment and research opportunities. Dr. Walter Klippel has been an incredible influence on my academic career at the University of Tennessee. His approach to archaeological research is impeccable. Dr. Klippel has always had just one more thought provoking question that needs to be addressed. I view Drs. Konigsberg, Jantz and Klippel as valued mentors, colleagues and friends.
Numerous individuals assisted my efforts with the Green River collections, but two individuals stand out. First and foremost, Dr. Jim Fenton provided invaluable advice on how to “get it done” and a unique perspective on Green River archaeology and skeletal biology. I always had a place to stay in Lexington during my numerous visits. I thank him for his hospitality, academic advice, archaeological thoughts and friendship. I
iv acknowledge Valerie Haskins for sparking my interest in the Green River collections way back in 1989. We inventoried the Read (15BT10) collection as part of my Master’s research. She provided access to the Read skeletal material when it was temporarily curated at Western Kentucky University. Dr. May Stanford of the University of North
Carolina at Greensboro granted provided access to the Barrett (15McL4) collection. Dr.
David Hunt provided access to the Indian Knoll (15OH2) material curated at the
Smithsonian Institution. Dr. Robert Mensforth provided access to the Ward (15McL11) collection when it was temporarily housed at Cleveland State University and graciously provided living accommodations during my stay. I thank a succession of museum directors who granted access to the Green River skeletal collections housed in the
William S. Webb Museum of Anthropology at the University of Kentucky. These individuals include Dr. Mary Powell, Nancy O’Malley, Dr. Sissel Schroeder, and Dr. Jim
Fenton. Drs. Lynn Sullivan and Susan Frankenberg provided access to the Eva skeletal material curated at the McClung Museum at the University of Tennessee. Dr. Sarah
Sherwood offered helpful advice from her recent dissertation experience. The Kentucky
Heritage Council provided funding for radiocarbon dates from Indian Knoll, Ward and
Barrett.
I thank my family and Sherri, my wife, for the support they have shown. My parents as well as Sherri’s family here in Knoxville have been incredibly supportive throughout my graduate career. Finally, Sherri has provided unending support throughout this process – It has taken a little longer than we had initially planned, but we have had good times with good friends.
v Abstract
The Green River Archaic period skeletal collections represent one of the largest regionally specific aggregate hunter-gatherer sample available for study. These collections have been the focus of numerous studies on paleopathology and paleodemography. Indian Knoll (15OH2) is the largest collection with over 1000 individuals. These burials were recovered from two primary excavations directed by
Clarence B. Moore and the Work Progress Administration (WPA) in the first half of the nineteenth century. The WPA excavated numerous sites along the Green River and it’s tributaries resulting in additional skeletal collections from sites such as Barrett (15McL4),
Carlston Annis (15BT5), Chiggerville (15OH1), Read (15BT10) and Ward (15McL11).
Besides the skeletal collections, the archaeological data from Green River Archaic sites has played a pivotal roll in the interpretation of Archaic period subsistence and social interaction throughout the southeastern United States and Eastern Woodlands.
This study details the results of a biological distance study of these skeletal collections based on cranial non-metric traits. Recent quantitative genetic methods and theory is employed in the analysis of these quasi-continuous traits in an effort to derive meaningful biological relationship. This study is important within the southeastern
United States given that most biological distance studies of Archaic period populations focus on numerous sites spread across a large area (i.e. the entire Southeast region) or are site specific. This study examines the biological relationships of relatively contemporaneous Archaic period skeletal collections from the middle Green River drainage.
vi A series of 24 traits were coded for all adult individuals (>~15 years old) from the
Green River skeletal collections as well as from skeletal material recovered from the Eva site (40BN12). The Eva site represents a middle to late Archaic stratified shell midden located on the lower Tennessee River. Mahalanobis D² values were calculated according to methods described by Blangero and Williams-Blangero (1991; see Williams-Blangero and Blangero, 1989) and Konigsberg et al. (1993). Effects of age, sex and temporal trend on the expression of specific traits were accounted for within the model. Biological distance measures were compared to temporal and geographic matrices in an effort to elucidate the effects of isolation by temporal and spatial distance within the samples. The results indicate a strong geographic influence on the biological distance structure, but the temporal influence is more confounding. The extent of trait variation hints at greater female mobility within the mating network which is indicative of a patrilineal/patrilocal society, but these tests are inconclusive.
vii Table of Contents
CHAPTER 1. INTRODUCTION...... 1 CHAPTER 2. SHELL MOUND ARCHAIC REGION AND RESEARCH...... 12 REGIONAL SETTING ...... 13 MID-HOLOCENE REGIONAL ENVIRONMENT...... 17 EARLY EXCAVATIONS AND PUBLIC FUNDED WPA AND TVA ARCHAEOLOGY ...... 19 PERSPECTIVES ON PREHISTORIC HUNTER-GATHERERS AND SMA RESEARCH...... 24 INDIVIDUAL SITE DESCRIPTIONS...... 35 CHAPTER 3. RADIOCARBON DETERMINATIONS AND TEMPORAL MATRIX...... 61 RADIOCARBON DETERMINATION FOR SMA SITES...... 65 TEMPORAL MATRIX CONSTRUCTION ...... 76 CHAPTER 4. QUANTITATIVE AND POPULATION GENETICS FOR NON- METRIC TRAITS ...... 81 QUANTITATIVE GENETIC THEORY ...... 82 METHODOLOGICAL APPROACHES ...... 96 CHAPTER 5. PREVIOUS BIODISTANCE RESEARCH ON SMA POPULATIONS...... 98 EARLY POPULATION STRUCTURE RESEARCH OF SMA GROUPS ...... 99 RECENT BIOLOGICAL DISTANCE STUDIES OF SMA GROUPS ...... 104 CHAPTER 6. SAMPLES AND ANALYTICAL METHODS...... 107 SAMPLES...... 107 NON-METRIC TRAIT SELECTION...... 110 TRAIT CODING AND SIDE SELECTION...... 111 AGE AND SEX EFFECTS ...... 112 DISTANCE CALCULATION ...... 114 TEMPORAL AND GEOGRAPHIC DISTANCES...... 116 MATRIX COMPARISON ...... 117 TEMPORAL TREND ...... 120 COVARIANCE MATRIX DETERMINANT COMPARISONS ...... 122
R MATRIX AND FST STATISTIC CALCULATION ...... 124 CHAPTER 7. RESULTS...... 126 SAMPLE AND TRAIT STATISTICS ...... 126 AGE AND SEX EFFECTS ...... 134 DISTANCE STRUCTURE ANALYSIS ...... 139 TEMPORAL TREND ...... 154 DETERMINANT RATIO BOOTSTRAP TESTS...... 159 viii DISCUSSION ...... 169 CHAPTER 8. CONCLUSIONS...... 173 LOOKING TO THE FUTURE ...... 176 REFERENCES CITED...... 180 APPENDIX. NON-METRIC TRAIT CODING ...... 203 VITA ...... 208
ix List of Figures
FIGURE 1-1. MAP OF SITE LOCATIONS FROM KENTUCKY AND TENNESSEE...... 7 FIGURE 1-2. TOPOGRAPHIC VIEW OF THE MIDDLE GREEN RIVER REGION WITH THE DISTRIBUTION OF ARCHAIC PERIOD SITES. SITES WITH SHELL MOUND ARCHAIC COMPONENTS ARE HIGHLIGHTED IN RED AND THE SITES EXAMINED FOR THIS STUDY ARE IN BLUE...... 9 FIGURE 2-1. PHYSIOGRAPHIC REGIONS OF KENTUCKY BASED ON FENNEMAN (1938)...... 14 FIGURE 2-2. GEOLOGY OF THE MIDDLE GREEN RIVER. (A) FORMATIONS OF THE MIDDLE GREEN RIVER. (B) GEOLOGIC SURFACE FAULTS IN THE REGION...... 15 FIGURE 2-3. WPA EXCAVATIONS AT INDIAN KNOLL. WILLIAM S. WEBB MUSEUM OF ANTHROPOLOGY, NEGATIVE NO. 4201...... 36 FIGURE 2-4. PLANVIEW OF INDIAN KNOLL (15OH2) WITH BURIALS, CONTOURS AND EXCAVATION BLOCK...... 38 FIGURE 2-5. PLANVIEW OF READ (15BT10) WITH CONTOURS, BURIAL LOCATIONS AND SHELL AREAS DEPICTED...... 42 FIGURE 2-6. PLANVIEW OF CARLSTON ANNIS WITH TOPOGRAPHIC LINES, STRUCTURE LOCATIONS AND BURIALS...... 47 FIGURE 2-7. PLANVIEW MAP OF CHIGGERVILLE (15OH1) WITH ONE-FOOT CONTOURS AND THE BURIAL LOCATIONS...... 49 FIGURE 2-8. DISTRIBUTION OF PROJECTILE POINTS FROM GREEN RIVER SITES GROUPED BY ROLINGSON'S CLUSTERS. CHIGGERVILLE (OH1) IS VERY SIMILAR TO INDIAN KNOLL (OH2) AND CARLSTON ANNIS (BT5)...... 52 FIGURE 2-9. PLANVIEW OF THE WARD SITE WITH BURIALS, CONTOURS AND EXCAVATION BLOCK SHOWN...... 53 FIGURE 2-10. PLANVIEW OF THE BARRETT SITE, 15MCL4. BURIALS 100 AND 87 REPRESENT THE TWO BURIALS THAT HAVE BEEN RADIOCARBOIN DATED...... 56 FIGURE 2-11. PLANVIEW OF THE EVA SITE WITH CONTOURS, EXCAVATION BLOCK AND BURIALS...... 59 FIGURE 3-1. CALIBRATED RADIOCARBON DATES FOR EVA, BARRETT, WARD AND CARLSTON ANNIS. HIGHLIGHTED BOX INDICATES THE SUMMED CALIBRATED RANGE...... 66 FIGURE 3-2. CALIBRATED RADIOCARBON DATES FOR READ, INDIAN KNOLL AND KIRKLAND. HIGHLIGHTED BOX INDICATES THE SUMMED CALIBRATED RANGE...... 67 FIGURE 3-3. PROJECTILE POINT RECOVERED WITH BURIAL 612 AT INDIAN KNOLL...... 71 FIGURE 3-4. PROJECTILE POINTS RECOVERED WITH BURIAL 827 AT INDIAN KNOLL...... 72 FIGURE 6-1. PARADOX DATA ENTRY FORM FOR SMA NON-METRIC TRAITS...... 113 FIGURE 6-2. GEOGRAPHIC DISTANCES MEASURED BY RIVER MILES AND DIRECT POINT-TO- POINT DISTANCES...... 118 FIGURE 7-2. PLOT OF THE FIRST TWO EIGENVECTORS DERIVED FROM THE R-MET ANALYSIS OF NINE MEASUREMENTS AVAILABLE FROM INDIAN KNOLL (N=344, SNOW 1948), CHIGGERVILLE (OH1, N=11, SKARLAND 1939), AND EVA (BN12, COMPONENTS: EVA=4, THREE MILE [TM]=20, BIG SANDY [BS]=7, LEWIS AND LEWIS 1961).
x THESE TWO EIGENVALUES ARE DERIVED FROM THE BIASED R MATRIX AND ACCOUNT FOR 88% OF THE VARIATION. THE INDIAN KNOLL SAMPLE IS DIVIDED BY DEPTH. OH2-1 REPRESENT THE UPPER 2.5 FEET OF THE MIDDEN. OH2-2 REPRESENTS INDIVIDUALS 2.5 FEET OR MORE BELOW SURFACE. OH2-UK REPRESENT INDIVIDUALS OF UNKNOWN PROVENIENCE...... 141 FIGURE 7-3. PRINCIPAL COORDINATE PLOT OF THE FULL SAMPLE RELATIONSHIPS. THE ‘+’MARKS THE CENTRIOD OF THE TWO-DIMENSIONAL COORDINATE SPACE...... 146 FIGURE 7-4. PLOT OF THE FIRST TWO PRINCIPAL COORDINATES DERIVED FROM THE DIVIDED DISTANCE MATRIX. THE ‘+’ MARKS THE CENTROID...... 153 FIGURE 7-5. PRINCIPAL COORDINATE PLOT OF THE MODIFIED DISTANCE MATRIX DERIVED FROM THE TEMPORAL TREND ANALYSIS. THE “+” MARKS THE CENTROID...... 158 FIGURE 7-6. A PRINCIPAL COORDINATE PLOT OF THE GREEN RIVER SITES BY SEX. THE “+” MARKS THE CENTROID...... 161 FIGURE 7-7. DETERMINANT RATIO BOOTSTRAP DENSITY PLOTS FOR CARLSTON ANNIS. THE UPPER PLOT REPRESENTS THE FULL SAMPLE. THE DASHED LINE REPRESENT THE OBSERVED RATIO. DOTTED LINES REPRESENT THE STANDARD 95% CI...... 164 FIGURE 7-8. DETERMINANT RATIO BOOTSTRAP DENSITY PLOTS FOR READ AND CHIGGERVILLE. THE DASHED LINE REPRESENT THE OBSERVED RATIO. DOTTED LINES REPRESENT THE STANDARD 95% CI...... 165 FIGURE 7-9. DETERMINANT RATIO BOOTSTRAP DENSITY PLOTS FOR INDIAN KNOLL. THE DASHED LINE REPRESENT THE OBSERVED RATIO. DOTTED LINES REPRESENT THE STANDARD 95% CI...... 166 FIGURE 7-10. DETERMINANT RATIO BOOTSTRAP DENSITY PLOTS FOR WARD AND BARRETT. THE DASHED LINE REPRESENT THE OBSERVED RATIO. DOTTED LINES REPRESENT THE STANDARD 95% CI...... 167 FIGURE 7-11. DETERMINANT RATIO BOOTSTRAP DENSITY PLOTS FOR EVA. THE UPPER PLOT REPRESENTS THE FULL SAMPLE. THE DASHED LINE REPRESENT THE OBSERVED RATIO. DOTTED LINES REPRESENT THE STANDARD 95% CI...... 168
xi List of Tables
TABLE 1-1. SKELETAL SAMPLES EXAMINED FROM KENTUCKY AND TENNESSEE...... 6 TABLE 1-2. ADDITIONAL ARCHAIC PERIOD SKELETAL SAMPLES FROM THE GREEN RIVER DRAINAGE EXCAVATED BY THE WPA...... 10 TABLE 3-1. RADIOCARBON DATES FOR SHELL MOUND ARCHAIC SITE IN KENTUCKY AND THE EVA SITE...... 62 TABLE 3-2. TEMPORAL MATRIX FOR THE GREEN RIVER ARCHAIC MIDDENS AND EVA.... 78 TABLE 3-3. TEMPORAL MATRIX FOR DIVIDED GREEN RIVER ARCHAIC MIDDENS AND EVA...... 79 TABLE 3-4. CHI-SQUARE TEST OF THE RADIOCARBON DATES BY SITE...... 80 TABLE 6-1. SKELETAL SAMPLES EXAMINED FROM KENTUCKY AND TENNESSEE...... 109 TABLE 6-2. MID-LINE AND BILATERAL TRAITS CODED IN THE SMA SAMPLES...... 111 TABLE 6-3. GEOGRAPHIC DISTANCE MATRICES IN KILOMETERS. UPPER TRIANGLE IS THE STRAIGHT-LINE DISTANCE AND THE LOWER TRIANGLE IS THE RIVER DISTANCE...... 117 TABLE 7-1. TRAIT FREQUENCIES FOR READ...... 127 TABLE 7-2. TRAIT FREQUENCIES FOR CARLSTON ANNIS...... 128 TABLE 7-3. TRAIT FREQUENCIES FOR WARD...... 129 TABLE 7-4. TRAIT FREQUENCIES FOR BARRETT...... 130 TABLE 7-5. TRAIT FREQUENCIES FOR CHIGGERVILLE...... 131 TABLE 7-6. TRAIT FREQUENCIES FOR INDIAN KNOLL...... 132 TABLE 7-7. TRAIT FREQUENCIES FOR EVA...... 133 TABLE 7-8. COMPARISON OF THE AVERAGES OBSERVED TRAIT FREQUENCY BY SITE...... 135 TABLE 7-9. OBSERVED TRAIT FREQUENCIES FOR THE ENTIRE SAMPLE (N = 1358)...... 135 TABLE 7-10. AGE AND SEX EFFECTS AS DETERMINED THROUGH THE UNIVARIATE PROBIT TESTS...... 137 TABLE 7-11. THRESHOLD VALUES FOR EACH SITE DERIVED FROM THE ANALYSIS OF THE ENTIRE SAMPLE. TRAIT NUMBERS FOLLOW TABLE 7-10...... 142 TABLE 7-12. POOLED TETRACHORIC CORRELATION MATRIX DERIVED FOR THE FULL SAMPLE ANALYSIS. TRAIT NUMBERS FOLLOW TABLE 7.10...... 143 TABLE 7-13. DISTANCE MATRIX CALCULATED FOR THE FULL SAMPLE ANALYSIS. SAMPLE SIZES ARE IN PARENTHESES...... 144 TABLE 7-14. MANTEL MATRIX COMPARISONS OF THE FULL SAMPLE BIOLOGICAL DISTANCE MATRIX TO SPATIAL AND TEMPORAL DISTANCES...... 147 TABLE 7-15. THRESHOLD VALUES FOR THE DIVIDED SITE ANALYSIS USING 13 TRAITS. DIVISIONS ARE DESIGNATED BY U AND L IDENTIFYING UPPER AND LOWER STRATIGRAPHIC UNIT, RESPECTIVELY...... 149 TABLE 7-16. POOLED TETRACHORIC CORRELATION MATRIX FOR THE DIVIDED SITE ANALYSIS...... 150 TABLE 7-17. DISTANCE MATRIX FOR DIVIDED SITE ANALYSIS. THE LOWER TRIANGLE REPRESENTS THE DISTANCES AND THE UPPER ARE THE SIGNIFICANCE VALUES BASED ON DROSSLER’S F-TEST. SHADING REPRESENTS NON-SIGNIFICANT DIFFERENCES. SAMPLE SIZES ARE IN PARENTHESES...... 152
xii TABLE 7-18. MANTEL MATRIX COMPARISONS FOR THE FULL SAMPLE DIVIDED BY STRATIGRAPHIC UNITS AT INDIAN KNOLL, EVA, AND CARLSTON ANNIS...... 154 TABLE 7-19. PEARSON CORRELATION COEFFICIENTS FOR THE TEMPORAL TREND ANALYSIS OF THE DIVIDED SITE SAMPLE. SHADED CELLS REPRESENT SIGNIFICANT VALUES. .. 155 TABLE 7-20. DISTANCE MATRIX FROM THE REDUCED PRINCIPAL COMPONENTS ON EIGENVALUES AFTER THE TEMPORAL TREND ANALYSIS...... 157 TABLE 7-21. BIOLOGICAL DISTANCE MATRIX OF THE GREEN RIVER SITES DIVIDED BY SEX. SAMPLE SIZES ARE IN THE SECOND ROW OF THE TABLE...... 160 TABLE 7-22. COVARIANCE MATRIX DETERMINANT RATIOS AND STANDARD 95% CONFIDENCE INTERVALS FOR EACH SITE AND STRATIGRAPHIC UNIT. THE COVARIANCE MATRIX COULD NOT BE CALCULATED FOR THE UPPER STRATIGRAPHIC UNIT AT CARLSTON ANNIS. THE MALE AND FEMALE MATRICES ARE NOT POSITIVE SEMIDEFINITE (I.E. EACH HAS ONE INVARIANT TRAIT). THE BOOTSTRAP COULD NOT BE PERFORMED FOR THE EVA UPPER STRATUM...... 163
xiii
Chapter 1. Introduction
The Shell Mound Archaic (SMA) sites along the Green River in west-central
Kentucky have captivated avocational and professional archaeologists for nearly a century. In 1916, Clarence B. Moore excavated several “mounds” along the Green River.
Building on Moore's work, William S. Webb directed the investigations of numerous
Green River sites with the support of the Works Progress Administration's (WPA).
Beginning in the early 1970s, William Marquardt and Patty Jo Watson went to the Green
River shell mound in hopes of possibly identifying early horticultural activities associated with these ancient populations. Marquardt and Watson viewed the populations of the
Green River Archaic as the logical antecedents to the intrepid early cavers of the
Mammoth and Salts cave systems. In her revolutionary cave research, Watson found that the Early Woodland cavers possessed a diverse suite of indigenous cultigens. In an effort to identify the origins of these horticultural activities, Watson and Marquardt shifted their research focus slightly west, down stream from Mammoth Cave to the Big Bend region of the Green River.
To Marquardt and Watson (1983) the shell middens of the Green River epitomized locations for the development of early horticultural activities. These sites represented seasonal aggregation locations where groups probably gathered for extended periods of time to exploit specific resources, feast, trade, conduct rituals and bury the dead. Smith (1986; 1992) views such intensively utilized locations as “domestilocalites,”
1 and he considers them as critical in the coevolutionary relationship of plants and people.
The only problem with Smith’s model is that the archaeological investigations at the
Green River middens have failed to produce substantial quantities of early domestics.
These plants have proven elusive. However, these sites do document an increase in social
interaction and an intensification of resource exploitation during the Late Archaic (ca.
3000 – 1000 B.C., Jefferies, 1990) in Kentucky. As such, they testify to a shift in social
interaction from the Early/Middle Archaic (ca. 8000 – 3000 B.C., Jefferies, 1990) to more sedentary communities in the Late Archaic.
Given that the Green River sites have yet to yield significant evidence of plant husbandry, the shift in settlement and subsistence strategies must have entailed a combination of intensification, diversification and communication. Groups certainly intensified their use of specific species and diversified the suite of wild plant and animal species exploited throughout the year. Population increases would have required increased social interaction and communication among groups (Goad, 1980; Jefferies,
1996, 1997; Sassaman, 1994).
Jefferies (1996, 1997) believes the shift in settlement from “a residential mobility strategy to a logistically organized one” occurred in the midsouth during the late middle
Archaic (6000-5000 BP). This type of settlement pattern shift requires greater interaction and social networking by the more sedentary hunting and gathering groups in an effort to reduce subsistence risks within the network. Jefferies (1997) documents the increase in social interaction through the study of bone pin styles. Jefferies sees bone pin styles in a very similar light to biological data in that a style, like a gene, can be passed from one
2 group to another. Bone pins can move in the form of trade goods or a craftsperson can
move through migration or mate exchange. Based on his examination of bone pins
recovered from a series of archaeological sites located in the Middle Mississippi and
Lower Ohio River valleys, Jefferies concludes that social networks extended over much
of the midsouth by the late Middle Archaic, but he believes regional/territorial boundaries
did exist. Therefore, the level of social interaction occurring at this time throughout the
Southeast would have had a dramatic effect on the population biology and the biological
relationships of these hunter-gatherer groups.
But how would such interactions be manifested in the skeletal biological distance data? Was it merely a down-the-line exchange system where one could expect simple isolation by distance, both temporal and spatial? Powell (1995) documented this exact pattern for several mid-Holocene sites using a diverse regional sample spanning from
Florida to southern Illinois. A similar pattern has also been demonstrated from a large temporally diverse but regionally circumscribed series from the lower Illinois River valley (Konigsberg, 1987; Steadman, 1997, 2001).
The late Middle to Late Archaic middens along the Green River drainage provide an exceptional research area to examine these questions. A majority of the larger sites within the drainage have been completely or partially excavated, primarily during the
WPA. Often cultural deposits remain at these sites, and current researchers have tried to tie recent stratigraphic and radiometric data to the older excavations. The skeletal
samples recovered from these sites do provide invaluable information concerning the
activities and interactions of these early populations. Osteological studies have focused
3 on diet, health and demography of these Archaic hunter-gatherers (Haskins and
Herrmann, 1996). Often these studies compare the Green River Archaic material to later,
temporally distinct samples, such as the Fort Ancient sample from Hardin Village
(Cassidy, 1972; Knutson, 1982; Wilczak, 1998).
Interestingly, no large-scale systematic study of the genetic structure of these
skeletal lineages has been undertaken. The lack of interest in biological distance studies
revolves around the widely held assumption that all eastern Woodland Archaic groups
share close genetic relationships (Long, 1966; Wyckoff, 1977). Studies of the Green
River populations are hindered by the lack of complete craniometric data sets. The latter
is a product of poor preservation and fragmentation seen at most Green River sites.
In this study, an assessment of the biological structure of six mid to late Holocene
Archaic skeletal samples from Kentucky and one from Tennessee is performed based on a series of non-metrical cranial traits. Studies of morphological traits are based on the assumption that the observed features are genetically driven. The exact amount of the genetic component expressed in cranial non-metric traits is difficult to ascertain and is often viewed as inferior to craniometric data. Given the limitation of the Green River skeletal collection for metric analyses, the analysis of cranial non-metric traits affords a good opportunity to examine a large percentage of the individuals and capture most of the variation within each sample. Employing recently developed quantitative trait methods, biological distance matrices are calculated from the non-metric data for various sample configurations examining differentiation by sex and stratigraphic association. Variations in temporal placement and geographic location of the sites will be considered in reference
4 to a genetic model of isolation by distance (Wright, 1943; Falconer, 1989). Konigsberg
(1990b) and Powell (1995) have demonstrated the utility of this type of analysis with
discrete cranial and dental morphological traits. Through the examination of this
variability, I hope to elucidate patterns of population history and regional interaction.
Non-metric cranial trait data and basic demographic parameters were collected from six SMA skeletal series curated at the William S. Webb Museum of Anthropology at the University of Kentucky in Lexington and the Eva sample from Tennessee housed at the Frank H. McClung Museum at the University of Tennessee in Knoxville. These sites include six shell middens and one rock midden (see Table 1-1). The site locations span the middle Green River drainage and the lower Tennessee River basin (Figure 1-1).
In his introduction to Webb’s classic reprinted Indian Knoll volume, Winters
(1974) classified the major Green River sites as either base camps, settlements, hunting camps or transient camps. Hensley (1994) redefined Winters’ “settlement” category as
“aggregation” sites vis-à-vis Hofman (1986) and reclassified Barrett, Ward and Carlston
Annis as aggregation sites along with Indian Knoll, Chiggerville, Read and Eva.
Therefore, this research examines all six major Green River sites identified as
“aggregation” locales by Hensley. The Eva site, a classic middle to late Archaic SMA site from Tennessee, is also examined. The Ward site is different in that it represents a dirt-rock midden rather than a shell midden. The artifact assemblage is similar to the other SMA sites, but the midden lacks a significant shell component. One important question to be addressed by this research is the relationship of Ward to the shell middens.
In addition, Ward and Read are positioned on bluff tops adjacent to the river bottoms, but
5
Table 1-1. Skeletal samples examined from Kentucky and Tennessee. Burials Site Excavated Burials Site Number by WPA Examined Reference Carlston Annis (KY) 15BT5 3901 209 Webb (1950a) Read (KY) 15BT10 247 82 Webb (1950b) Barrett (KY) 15McL4 412 189 Webb and Haag (1947) Ward (KY) 15McL11 433 203 Webb and Haag (1940) Chiggerville (KY) 15OH1 114 61 Webb and Haag (1939) Indian Knoll (KY) 15OH2 8802 506 Moore (1916), Webb (1946), Snow (1948) Eva (TN) 40BN12 180 110 Lewis and Lewis (1961) Total 2656 1360 1 Additional burials recovered during SMAP project (Marquardt and Watson 1983) 2 Moore (1916) also excavated 298 burials. Portions of seventy individuals are curated at the Smithsonian National Museum of Natural History. These materials were examined for this study.
6 Green Ohio River River $Z15McL11 15McL4$Z $Z15Bt5 15Oh2$Z$Z $Z 15Oh1 15Bt10 Mammoth Cave NP Kentucky
Mississippi River Tennessee
40Bn12 $Z
Tennessee River
$Z Site Examined N 20 0 20 Miles 40 0 40 Kilometers Green River Shell Mound Area
Figure 1-1. Map of site locations from Kentucky and Tennessee. they exhibit similar artifact suites and complementary mortuary programs to the flood
plain shell middens. Comparison of the Kentucky Archaic samples to the Tennessee
sample provides a regional link for all the Kentucky Green River samples with Eva
serving as an outgroup for this analysis.
Numerous Archaic period sites are located within the middle Green River drainage (Figure 1-2), and several smaller SMA sites have been identified throughout the valley (e.g. Hockensmith et al. 1985). The WPA excavated several of these smaller SMA sites along the Green River (see Table 1-2). Most contained human remains and produced similar artifact collections as the larger sites examined for this study. The skeletal samples from these sites are, however, too small or extremely fragmentary to be included in this analysis. The Butterfield site does appear to have a large number of individuals, but this material had not been inventoried or sorted since the WPA excavations.
In an effort to visualize and manage both archaeological and mortuary data from these sites, burial location information from the original WPA field notes was converted into coordinate databases for mapping and data management. Each site map is based on available burial records and topographic information. Similar maps have been produced by Rolingson (1967) for all the Green River sites, Hensley (1994) and Milner and
Jefferies (1998) for Read (15BT10), and Crothers for Chiggerville (15OH1). Only the recent efforts of Crothers and Milner and Jefferies have focused on producing GIS based three-dimensional maps of the Green River middens based on the WPA data. In an effort
8 (X (X (X (X(X (X (X (X(X $(X $(X (X (X (X $(X (X McLean County (X N (X(X (X(X$(X $(X (X (X (X(X $(X(X (X(X(X(X(X (X(X (X(X Kirkland 3 0 3 Miles (X (X r 15McL12 (X (X (X 4 0 4 Kilometers
e (X (X$(X(X (X(X (X (X v (X $ (X $ (X (X i (X(X (X (X (X $ (X(X
R Butterfield Jim(Xtown SMA Site Examined % (X(X d (X 15McL7 (X$ Hill (X (X (X n (X (X15Oh19 Rough $ SMA Site
o (X(X (X (X(X (X (X (X (X(X$(X (X P (X Ward (X (X River (X(X(X(X X( Archaic site (X(X(X(X 15McL11%(X(X (X (X(X (X(X (X (X (X $ (X(X(X (X (X (X(X (X Barrett $(X 15McL4 (X %(X C
y p (X Smallhous Ohio County r (X(X e (X $(X15Oh10 s (X s (X (X (X(X C (X (X (X (X (X(X G (X (X(X (X(X(X (X (X Big (X r (X (X r (X (X (X (X . (X (X e (X (X e (X (X (X (X (X(X (X Bend (X(X(X(X n R (X $(X (X (X iv $(X Jackson DeWeese(X(X (X$(X (X Region e (X (X (X r (X Bluff 15Bt6 (X (X $(X (X (X (X(X 15Oh12 (X (X (X Carlston (X (X(X Annis (X (X (X(X %(X (X (X (X (X Indian Knoll 15Bt5 (X (X (X (X (X 15Oh2 $(X Muhlenberg County (X(X(X(X %(X Haynes(X (X (X (X Chiggerville 15Bt11 (X (X(X (X (X (X (X 15Oh1 (X (X (X (X (X %(X (X(X (X (X(X (X (X $% Read (X (X (X (X Baker (X(X(X(X (X$(X (X (X (X (X(X 15Bt10(X(X (X (X (X (X $(X (X (X (X(X 15Mu12 (X (X Butler (X (X $ (X (X (X Bowles (X (X (X County (X(X (X 15Oh13 (X (X(X Figure 1-2. Topographic view of the Middle Green River region with the distribution of Archaic period sites. Site with Shell Mound Archaic components are highlighted in red (triangles) and the sites examined for this study are in blue (squares). 9 Table 1-2. Additional Archaic period skeletal samples from the Green River drainage excavated by the WPA. Burials Site Number Excavated Reference Kirkland (KY) 15McL12 70 Webb and Haag (1940) 1 Butterfield (KY) 15McL7 153 No WPA/UK Report Bowles (KY) 15OH13 21 No WPA/UK Report Baker (KY) 15MU12 6 No WPA/UK Report2 Jackson Bluff (KY) 15OH12 17 No WPA/UK Report Jimtown Hill (KY) 15OH10 14 No WPA/UK Report Total 281 1 Examined by Haskins (1992) but no formal report issued. Five radiocarbon dates are available (Claassen 1996b). 2 Archaeological material recently examined by David McBride (2000) of the University of Kentucky as part of his Masters Thesis.
10 to provide accurate site locations, Kentucky site file coordinates were checked against
WPA reports, WPA field notes and recent field investigations. Modern digital mapping resources were gathered to provide a drainage specific coverage of the Green River Shell
Mound Archaic Area.
The subsequent chapters detail the history of SMA archaeology within the Green
River drainage and midsouth, a summary of the radiometric determinations for the SMA sites examined, a review of quantitative and population genetics, and the description of the methods and results of this study. Chapter 2 provides a regional setting, a history of investigations of SMA sites along the Green River in Kentucky and individual site descriptions. Chapter 3 details the radiometric determinations available from the SMA sites examined and the tabulation of the temporal distance matrix used in the isolation by temporal and spatial distance model. Chapter 4 covers quantitative and population genetic theory and describes biological distance measures. Chapter 5 summarizes prior biological distance studies of SMA populations. Chapter 6 details the samples and methods employed in the current study. Chapter 7 describes the results of the analysis and discusses the implications of these findings in reference to the Green River SMA samples. Finally, Chapter 8 concludes the study with a summary of the findings and suggestions for future research on SMA skeletal samples from throughout the southeastern United States.
11
Chapter 2. Shell Mound Archaic Region and Research
Research on the numerous shell mound Archaic (SMA) sites throughout the southeastern United States has a long history. Within the upland south, Clarence B.
Moore’s (1916) investigations along the Ohio, Tennessee and Green Rivers in the early twentieth century brought these rich archaeological resources into focus. Early scholars of these enigmatic mid-Holocene burial and habitation middens recognized that the populations that produced these remains were hunter-gatherers (or fisher-gatherer-hunter
[FGH] vis-à-vis Marquardt, 1985) and not agriculturalists. Although early interpretation of some of these artifacts may not have been correct, such as Moore’s interpretation of atlatl weights as “net spacers,” these researchers acknowledged the lack of implements associated with intensive horticulture/agriculture. In this chapter, I provide a summary of the physiographic setting, theoretical perspectives on prehistoric FGH populations, and previous archaeological research in the Green River and lower Tennessee River areas. In my discussion of the physical environment and prehistoric landscape, I will primarily focus on the Green River drainage with supplemental data from the lower Tennessee
River. My discussion of previous archaeological research focuses on the Works Progress
Administration (WPA) and the Tennessee Valley Authority (TVA) impact on Archaic research in the interior Southeast and on the recent Shell Mound Archaeological Project
(SMAP) initiated by Patty Jo Watson and William Marquardt in the Big Bend Region of
Kentucky. Current research along the Green River has refocused efforts on new methods
12 and a reexamination of old data in an effort to better understand these unique archaeological resources. This chapter concludes with individual site descriptions and histories of excavation results and specific skeletal and archaeological research.
Regional Setting
The lower half of the Green River drainage basin lies exclusively in the Western
Coal Field physiographic province of Kentucky (Fenneman, 1938, Figure 2-1). The
Western Coal Field region is underlain by a variety of Pennsylvanian and to a lesser degree Mississippian sandstones, shales, and coal. Pennsylvanian era formations include the Sturgis, Conemaugh, and Tradewater (Rice, 2001). The vast alluvial floodplains of the Green and Tradewater Rivers dissect the rolling uplands and sandstone cliffs of west central Kentucky (Figure 2-2a). Alluvium deposits along the drainage systems have
Pleistocene origins resulting from the inundation of the dendritic valleys by glacial outwash (McFarlan, 1943). In some areas alluvial deposits extend to a depth of 53 m
(175 ft). Stein (1980) discusses this process in detail, and I will briefly discuss its importance to the Green River shell mound sites.
During the Pleistocene glaciation of eastern North America, the Ohio River acted as the primary conduit for glacial outflow along the southern border of the ice field. The river carried large quantities of melt water and glacial sediments. A significant percentage of these sediments began to fill the unglaciated tributaries to the Ohio River.
One such valley was the Green River drainage of central Kentucky. The mechanism of this alluvial event is believed to have been a major flood event along the lower Ohio
River. A partially blocked bedrock constriction near Caseyville, Kentucky may have
13 Physiographic Regions of Kentucky: N Eastern Coal Field Eastern Pennyrile Inner Bluegrass 25 0 25 Miles Knobs Outer Bluegrass Purchase - Alluvial Knobs
Western Coal Field #
Western Pennyrile Knobs
Inner Bluegrass
Outer Bluegrass
Eastern Coal Field Western Coal Field
#
Eastern Pennyrile
Purchase - Alluvial Western Pennyrile
Big Bend Region (A)
$ $
Green River Pond $ River $ $ $
(B) $
$ $ $ $ $ $
$ Green River $ Pond $ River $ $ $ $$ $ $ $
General Geology of Middle Green River Region: ALLUVIUM Geologic Faults N MISSISSIPPIAN Streams/Rivers 2 0 2 Miles PENNSYLVANIAN $ Archaic Sites Examined Figure 2-2. Geology of the Middle Green River. (A) Formations of the Middle Green River and (B) Geologic surface faults in the region. served as a drainage bottleneck. As a result, the Ohio River would have flooded many
miles upstream, including the Green River drainage. Sand inter-digitated with silt and
clay deposits is the geomorphological signature of such cyclical flood events. This
signature is present in the alluvial deposits along the Ohio and Green Rivers as well as
other tributaries to the Ohio (see Stein, 1980).
Periodic or seasonal flood events along the Green River inundated the river valley
from its confluence with the Ohio to Big Reedy Creek (approximately 16 km upstream
from the confluence of the Barren River). The upper third of the area represented a delta
zone feeding into Green Lake. Therefore, the deposits along the lower Green River as
well as much of the lower Ohio River Drainage above the Caseyville Gorge were
subjected to lacustrine and fluvial processes. Stein (1980) mapped the extent of the
lacustrine deposits of the Green River drainage and correlated them to similar deposits
along northern tributaries to the Ohio River. The maximum elevation of these deposits
reached 117 -119 m (385 – 390 ft), and Straw (1968, referenced in Stein 1980) suggests
that the maximum lake elevation was 133 m (435 ft). Since the Pleistocene, the lower
Green River has incised into these lake deposits. The cohesive nature of the silt and clay
deposits has significantly affected river meanders and flooding throughout much of the
lower Green River system (Stein, 1980).
One important aspect of the Pleistocene flooding events along the Green River is that the remnant sandstone floodplain features exposed during the Pliocene throughout much of the middle and lower Green River basin were completely or partially encapsulated in lacustrine deposits. These bedrock islands, or “island hills” (Shaw,
16 1911), may have influenced river meandering and shoal formation. In addition, these islands served as raw material sources for the Archaic populations exploiting the riverine environments. It is important to note that culturally modified and fire-altered sandstone cobbles are ubiquitous in the anthropogenic strata of the all Green River middens.
Crothers (1999; see also Morey and Crothers, 1998) investigated the association of shell middens, shoals and geologic features within the Green River drainage. Based on a general comparison of known shell midden sites and pre-inundation navigation maps of the Green River, Crothers (1999) found that several sites are associated with bedrock exposures, coal seams and/or geologic faults. In Figure 2-2b, the geologic faults recorded within the middle Green River drainage are shown with a select group of archaic period shell midden sites. The correspondence with faults is not consistent but several sites do lie in direct association with known faults. Crothers (1999) has shown that some faults produce sandy, bedrock shallows within the Green River. These channel characteristics are favorable to a multitude of shellfish species which were exploited by Archaic period populations.
Mid-Holocene Regional Environment
The mid-Holocene was a period of gradual environmental and cultural change in the midsouth. Pollen records from approximately 8,000 B.P. to 3,900 B.P. indicate a warmer and drier climate (King and Allen, 1977; Wilkins et al. 1991). This climatic event, known as the Hypsithermal interval, resulted in a vegetation shift which allowed mid-Holocene populations to spatially map on to these resources. The Tennessee, Ohio and Green River valleys were similarly affected although in varying degrees. Arboreal
17 communities reduced while grass and herb communities expanded. Bruce D. Smith sees
these climatic changes as favoring a shift to riverine resources as is evident in the follow
statement:
At the same time that altered seasonal stream flow patterns were causing the enhancement of aquatic habitats and increasing the level of aquatic biomass available for exploitation along segments of river floodplains, there may have also been a deterioration of upland resources due to decreased effective precipitation (Smith, 1986:24). Populations seasonally relocated to the floodplains typically on remnant terraces and
elevated areas adjacent to “spatially limited aquatic resources” (Smith, 1986:25; see also
Waselkov, 1982). Within the Green and Ohio River drainages, these elevated areas
would likely represent an encapsulated bedrock island. In addition, Crothers (1999:143)
effectively demonstrates the “mapping-in” on aquatic resources by Archaic FGH
populations at known shoals along the river. The Eva site (40BN12) in Tennessee is
located on a remnant terrace of the Tennessee River along Three Mile Slough. Although
Eva’s relationship to specific riverine resources is unknown, the site location suggests a
rich and diverse resource base similar to the Green River region.
The effect of the Hypsithermal interval has been assessed at Carlston Annis based
on maximum length of specific gastropod species (Baerreis, 1980, n.d.; Crothers, 1999).
Using regression analysis, Baerreis estimated the annual precipitation at Carlston Annis.
The shellfish were collected by excavation level as part of the SMAP project directed by
Marquardt and Watson. Using maximum lengths of Gastrocopta cantracta, Baerreis
estimated annual precipitation for levels 5-8 to be 827-857 mm; for levels 12-17, between
746-806 mm; and for levels 19-20, between 857-897 mm. The average modern 18 precipitation for the Big Bend as reported for Morgantown, Kentucky is 1219.2 mm (48
in). It is quite clear that the climate was much drier during the Hypsithermal interval, a
period of shellfish accumulation, in the Big Bend. Based on an examination of Mesodon
inflectus size, Baerreis also found that the seasonal precipitation pattern at Carlston Annis
is different than the modern pattern. M. inflectus size is primarily controlled by summer
precipitation and the prehistoric estimates (552 mm upper, 562 mm middle, and 537 mm
lower) fall only slightly below the modern values (574 mm). Combining precipitation
data from G. cantracta and M. inflectus, the estimates indicate that the amount of winter precipitation along the Green River in Kentucky was dramatically lower than modern conditions. The plant communities of this period reflected the environmental conditions
suggested by the shellfish populations. Wilkins et al.’s (1991) work at Jackson Pond in
LaRue County, Kentucky indicates a mesic deciduous woodland during the early
Holocene (10,000 B.P. to 7,300 B.P.). The middle to late Holocene (7,300 B.P. to
present) witnessed a transition to a more xeric oak-hickory forest.
Early Excavations and Public Funded WPA and TVA Archaeology
Early archaeological investigations of the Green River region and the lower
Tennessee River valley began in the late nineteenth and early twentieth centuries. Early surveys and excavations by researchers like Squier and Davis (1848) focused more on the earthworks and mound sites throughout the eastern United States. However, by the turn of the century, avocational and quasi-professional archaeologists began to investigate village sites as well as mound sites.
19 Two early researchers focused on the Green River drainage in central Kentucky.
In 1916, Nels C. Nelson investigated archaeological remains in Mammoth Cave in search
of evidence for early man in the Americas (Nelson, 1916). Nelson was associated with
the American Museum of Natural History and had conducted archaeological research
throughout the United States as well as in France. He considered caves with
archaeological deposits critical to establishing early chronological sequences. The data from the vestibule of Mammoth Cave included the recovery of at least three burials and numerous artifacts. Nelson also stressed the importance of other cave sites to produce significant archaeological deposits, artifacts, burials, or even desiccated human remains.
Although absolute dating techniques were not available, Nelson felt that the Mammoth
Cave evidence was of moderate antiquity.
Clarence B. Moore, a contemporary of Nelson, worked throughout the Southeast,
but he was not specifically associated with any museum or university. Moore financed his
own field investigations and subsequent publications through the Academy of Natural
Sciences of Philadelphia. Moore traversed the various rivers of the interior Southeast
aboard his stern-wheeler, the Gopher, from 1891 to 1918. Moore was interested in
publishing the unusual and spectacular artifacts, and often his field investigations focused
on cemeteries and mounds. The field methodologies employed by Moore must have been
quite frightening at some sites where hundreds of burials were excavated within a matter
of days or weeks.
Moore spent nine weeks during the fall and winter of 1915-16 investigating sites
along the Green River. He investigated at least five Archaic Period sites along the Green
20 River below Paradise, Kentucky including Smallhous (15OH10), Indian Knoll (15OH2),
Bluff City (15HE160), a site on the Austin Place in McLean County (15McL15) and one
near Calhoun in McLean County. Moore excavated a few burials at each of these sites, but at Indian Knoll, Moore recorded 298 burials. It is interesting to note that with a crew of eight men Moore excavated these burials in 179 work hours. Consistent with conventions current at the time, most of the burials were not completely removed. Only
66 crania and a few pathological specimens were saved from Indian Knoll and shipped to the U. S. National Museum for analysis. However, these remains do represent the first extensive human remains collected from this region. Moore’s publication detailed the archaeological resources of the region and summarizes the burials with associated artifacts or “other burials in any way noteworthy” (Moore, 1916:453).
Moore also visited several sites upstream from Indian Knoll and within the Big
Bend, but these locations are not described in detail in his report. Moore probably did not discuss these sites because no artifacts were found in association with burials. Sites investigated include the Newton Brown Place (Chiggerville, 15OH1), the Austin Place, the Rhone Place, and the DeWeese Place (15BT6). Crothers (1999) associates the Rhone
Place and Austin Place to 15BT5 and 15BT11, respectively.
Moore’s report on the Green River site was consistent with his other self-financed
publications. Maps, large format photographs and color illustrations were included to
provide full documentation of the archaeological discoveries. With the Green River sites,
Moore was particularly interested in “banner-stones” which were found in association
with several burials. Moore hypothesized that these “banner-stones” were tools utilized
21 in an elaborate fishing industry. He also suggested that that these artifacts were atlatl
parts but then dismissed this idea in favor of the fishing toolkit hypothesis. The antler hooks represented netting-needles and the banner-stones served as net-sizers or spacers
(Moore, 1916:431-433). Later during the WPA investigations, Webb (1950; Webb and
Haag, 1939) would also be enticed by these beautiful artifacts. He would argue that they represent a composite atlatl.
In 1927 the University of Kentucky established a Department of Anthropology with William S. Webb as the department head. Webb had been involved with William D.
Funkhouser in various archaeological salvage projects throughout Kentucky as well as a survey documenting prehistoric archaeological sites including open habitations, rock shelters and caves. As a result of their efforts, the University of Kentucky and Kentucky
Geological Survey published a series of books concerning Kentucky’s prehistoric past including Ancient Life in Kentucky (Funkhouser and Webb, 1928) and an archaeological survey of Kentucky (Funkhouser and Webb, 1932). Funkhouser and Webb identified numerous cultural groups throughout the state including a “Shell Mound Area.” Within the Archaeological Survey, they identified at least 15 shell middens in Ohio, Butler,
Muhlenberg and McLean counties. Several of the sites in Muhlenberg County were not specifically listed as shell middens, but “all the sites examined [by Funkhouser and
Webb] have been typical of the so-called ‘Green River Valley Culture’” (Funkhouser and
Webb, 1932:316). As part of the survey Webb and Funkhouser excavated at several sites including Chiggerville (15OH1, referred to as the Old Post Office Site). During the excavation, they were astonished at the density of shell within the midden and found
22 numerous artifacts similar to items described by Moore (1916). These early studies of the Green River sites piqued the interest of Webb and laid the foundation for future WPA investigations.
Further to the south, archaeological investigations along the Lower Tennessee
River drainage progressed according to inundation schedules dictated by the Tennessee
Valley Authority (TVA). Kentucky Dam was the seventh major dam built under the auspices of TVA, and it resulted in the largest multipurpose project of the Tennessee
River stretching 183 miles upstream from the dam location. Archaeological activities throughout Tennessee had been well established since the Norris (Webb, 1938) and
Chickamauga projects. The Pickwick and Watts Bar Reservoirs had also been investigated (Webb and DeJarnette, 1942; Nash n.d.). Following Webb’s (1938) oversight of the Norris basin excavations, Thomas M. N. Lewis of the University of
Tennessee directed most of the remaining early TVA research throughout the state. As part of the Kentucky Reservoir project in west-central Tennessee, the Eva Site (40BN12) was excavated in 1940. Based on the materials recovered, Eva is similar to the Green
River mounds of the Green River. However, several important differences are present.
First, Eva represented a well-stratified site. Several flood deposited sand strata were identified during excavation. These deposits neatly separated the cultural horizons within the midden. Second, several projectile point types identified at Eva are temporally earlier than varieties found in the Green River middens. This observation was substantiated by one radiocarbon date from Eva (Lewis and Lewis, 1961) suggesting a Middle Archaic use of the site. Several additional sites with numerous burials were excavated along the
23 lower Tennessee including Cherry, Kays Landing, and Ledbetter. Although not
examined as part of this study, these samples are considered contemporaneous and
demographically similar to the Eva skeletal sample (Magennis, 1977).
Perspectives on Prehistoric Hunter-Gatherers and SMA Research
Archaeological data gathered and reported by Webb from the Green River sites was viewed with a skeptical and critical eye by the increasing number of professionally trained archaeologists. As was customary, Webb organized tables of artifact and feature trait lists for each site. Sites were considered more similar or different based on the number of traits present or absent. Taylor (1948) acknowledges the role Webb played in development of large-scale salvage and relief archaeology of the previous decade, but he was critical of Webb’s limited interpretation of the enormous quantities of data derived from these efforts. Taylor considered the Midwest Taxonomic system at fault in that researchers were limited to a series of pigeonholes to place archaeological data.
In the years following the publication of Webb’s last site reports, the field of archaeology underwent a period of critical review. Building on the critique of Taylor
(1948), the ideas of functional versus processual interpretation were discussed and the role of time-space systematics examined. Regional settlement and subsistence strategies were developed and evaluated, and the Green River data played a critical role. Caldwell
(1958) envisioned an adaptive strategy, termed primary forest efficiency, in which
Archaic populations gradually maximized their ability to exploit resources from a diverse environment. The late Archaic Green River shell heaps represented the pinnacle of this adaptive strategy. At the same time, Fowler (1959) attempted to classify the Green River
24 sites based on the suite of functional artifacts recovered. Fowler compared eight Green
River sites with Modoc rockshelter and several other Midwestern sites and concluded that
seasonally specific activities occurred at each site. In the same time, Lewis and Kneberg
(1959) began to synthesize the Tennessee Archaic data from a variety of sites excavated
in the lower Tennessee River valley, such as Eva, Cherry, Ledbetter and Kays Landing.
With the support of radiocarbon dates these sites were placed within a long chronological sequence along with the Green River data.
Since the 1966 Man the Hunter conference at the University of Chicago, the concept of the hunter-gatherer has undergone intense scrutiny. Lee and DeVore (1968)
as well as the “new” archaeology movement of the period instigated the re-examination
and exploration of commonly held assumptions concerning hunting and gathering
populations by an untold number of archaeologists, ethnographers, and physical
anthropologists. These researchers often approached the topic from dramatically
different perspectives, but the primary goal was to provide a clearer picture of these non-
agrarian populations. The availability of computers for intricate statistical analysis
coincided with the increased interest in hunter-gatherer populations. The combination of new theoretical perspectives and new analytical tools provided the foundation for a reexamination of the Green River Archaic shell mounds and the concept of the southeastern Archaic period. In this section, I will first deal with archaeological perspectives on the prehistoric hunter-gatherer populations throughout the Southeast and specifically in Kentucky. I will then focus on research in west-central Kentucky along the Green River and examine the contributions of the Shell Mound Archaeological
25 Project (SMAP) initiated by William Marquardt and Patty Jo Watson (Marquardt and
Watson 1983; Marquardt, 1977).
Howard Winters’ (1968, 1969, 1974) perspectives on Archaic period populations
focused on the central Wabash River valley and the Green River sites. In the classic
Riverton Culture publication in 1969 and the introduction to the reprint of Webb’s Indian
Knoll volume in 1974, Winters attempted to address issues of settlement and regional interaction. He viewed the excavated Green River sites as incredible comparative data sources. Although he was working with Webb’s published data and not re-analyzing
artifacts directly, Winters attempted to classify various Archaic period sites throughout
the Midwest and Southeast into specific site types based on functional artifact
classifications. Winters envisioned a well-established transcontinental trade network
with the Green River populations as participants (Winters 1968). The extent of
participation is elusive. Were the Green River communities passive actors in the
exchange system? Or were they active initiators, commodities brokers, within the
system?
Two important dissertations bracketed Winters’ publications. The first is Martha
Rolingson’s in 1967 (see also Rolingson and Schwartz 1966) and the second is Nan
Rothschild’s in 1975 (Rothschild 1975, 1979). These studies attempted to address two
very different topics of debate concerning the Green River Archaic sites. Rolingson
(1967) focused on Webb’s interpretation of these sites as permanent villages. Rolingson
demonstrated through the examination of projectile points and ceramics that the shell
mounds represented multicomponent sites that spanned several millennia. Rolingson
26 argues that the dearth of prepared floors, post-molds and fire hearths at these sites does
not support an interpretation of year round settlements. She suggests that the simple fact
that there are non-shell middens found throughout the region (i.e. Ward, Kirkland,
Butterfield) indicates a seasonal settlement pattern. Rolingson’s work represented one of the first attempts to re-analyze the vast WPA collections to address hypotheses set forth by earlier researchers. Rothschild (1975) attempted to address issues of status within the
Green River middens through the examination of demographic data, associated grave goods, and burial placement. She compared data from various Archaic and Mississippian sites within the Midwest and Southeast. Rothschild found age and slight sex differences in the distribution of artifacts at Indian Knoll, Carlston Annis and Read. Rothschild identified two distinct “macro-clusters” at Indian Knoll based on associated grave goods;
however, the significance of these groups was illusive to Rothschild (Rothschild,
1975:136).
Although predating much of Winters’ work, Rolingson and Schwartz (1966)
presented a model for Middle to Late Archaic cultural development within the region
encompassing the Lower Tennessee /Cumberland River basin and Tradewater/Green
River basin. They hypothesized that Paleo-Indian and Early Archaic Period populations
employed very similar settlement and subsistence patterns. During the middle Archaic
these populations differentiated with several “influential centers” forming in major
drainages (Nance 1988). The Green River and the lower Tennessee River valley
represented two such “centers.” Archaeological survey work in the lower Cumberland
River valley by Nance (1987, 1988) has supported the Rolingson and Schwartz
27 hypothesis and demonstrates that this region is more similar in terms of material culture
to the southern sites such as Eva. If such a boundary existed prehistorically, the non-
metric data examined should reveal this differentiation through time from the late Middle
Archaic to Late Archaic.
Over the next decade, research into hunter-gatherers focused on settlement and subsistence, regional interaction and mortuary programs. Winterhalder and Smith (1981) and Binford (1980) began to examine foraging and collecting strategies in prehistoric hunter-gatherers using ethnographic corollaries. Numerous studies developed out of these initial publications to specifically address settlement and subsistence models within the southeastern United States (e.g. Brown and Vierra 1983; Anderson and Hanson
1988). Regional surveys and lithic technology studies play a critical role in this research
(see Carr 1991, 1994). It has been difficult to apply these models to the Green River sites given the lack of empirical evidence in the form of lithic debitage. The WPA investigators did not systematically collect lithic debitage. Investigations at Carlston
Annis have provided some evidence (Marquardt 1985). Recent work by Crothers (1999) at DeWeese and Haynes and the work of Jefferies and co-workers (Jefferies et al., 2001) at the Ward site will help extend our interpretive base.
In his review of southeastern prehistory, Smith (1986, 1992) provided an overview of mid-Holocene subsistence adaptations. In his discussion of the development of early horticulture, Smith views the mid Holocene shell and midden mounds of the
Tennessee, Ohio and Mississippi River valleys as ideal locations for the genesis of plant husbandry in eastern North America. To Smith, these settlements offered a “continually
28 disturbed anthropogenetic habitat” for wild species such as chenopod (Chenopodium sp.), sumpweed (Iva annua) and sunflower (Helianthus annuus) to propagate and evolve in a symbiotic relationship with early hunter-gatherers. The archaeological evidence from the
Green River shell mounds is tenuous at best given the small amounts of these early transitional plant species (Marquardt and Watson, 1983; Watson, 19985; Wagner, 1996).
The importance of shellfish within these Archaic populations has been a significant point of debate. Mollusk and gastropod species are small and meat contributions would have been small relative to other aquatic and terrestrial fauna
(Parmalee and Klippel 1974). Klippel and Morey (1986) demonstrated this fact in an examination of shellfish remains from the middle Archaic period Hayes site (40ML139) in central Tennessee. Klippel and Morey found that hundreds of thousands of gastropods would have contributed only about a fifth of the usable dietary meat. Klippel and Morey did acknowledge that shellfish species do provide a significant percentage of several essential vitamins and minerals. They hypothesize that “stone boiling quantities of gastropods could produce a broth or “stock” that was high in nutrients to which less nutrient dense food could be added” (Klippel and Morey, 1986:809). The larger mollusks could have been steamed, shucked, dried and stored.
Faunal remains from three Big Bend sites (Carlston Annis, Haynes, and
DeWeese) indicate a similar diversified faunal assemblage including a variety of small and large bodied mammals, reptiles, and fish (Crothers, 1999). Some species identified include white-tailed dear (Odocoileus virginianus), rabbit (Sylvilagus sp.), squirrel
(Sciurus sp.), catfish (Ictalurus sp.), Drum (Aplodinotus grunniens), wild turkey
29 (Meleagris gallopavo) and a variety of turtles. The faunal assemblages documented at these three sites are consistent with Archaic period data reported throughout the Midwest and Southeast (Styles and Klippel, 1996). Obviously the Green River Archaic FHG utilized most available resources within the surrounding environment to provide an adequate subsistence base.
Sassaman and Anderson (1996) provide a series of regionally focused papers on mid-Holocene adaptations throughout the Southeast. One common theme in these papers is regional interaction. Trade networks played a critical role in community relationships and regional politics. Marine shell, copper and lithic raw material moved throughout the
Southeast. Local and regional exchange provided a mechanism for community interaction and mate-exchange. Smith summarizes this interaction in the following statement:
The exchange networks that moved information, innovations, various raw materials, and finished artifacts around the southeast during the 5000-2500 B.P. time period apparently consisted of innumerable multidirectional, reciprocal, down-the-line exchanges between trading partners (often lineage leaders) of both nearby and distant communities (Smith, 1986:30, referencing Goad, 1980)
Jefferies (1996:228) examined the stylistic differences in bone pins from late Middle
Archaic sites as evidence of “a socially bounded area extending over several hundred kilometers of the mid-continent.” Through the exchange of information and commodities
(i.e. bone pins), “increasingly sedentary hunter-gatherers” insured their survival.
30 Coincident with the reexamination of settlement, perspectives on Middle to Late
Archaic mortuary practices needed to be revised to adjust to the new models. Charles and Buikstra (1983; Buikstra and Charles, 1999) and Hofman (1986) provide two examples of this critical re-examination using data from both old and new excavations.
Beginning with evidence from the Duck River drainage, Hofman (1986) developed a
Middle to Late Archaic mortuary model based on a series of parameters including demographic, status, resources, and settlement organization. Hofman posited that as settlement patterns shift the proportion of secondary burials should increase relative to primary interments. In addition, burial treatment will be substantially different between settlement patterns (i.e. forager versus collectors). Looking at the published Green River data, Hofman interpreted these sites as locations of seasonal aggregation where specific segments of the population received preferential burial (typically children and young adults). Hofman suggests that older adults would have been buried elsewhere during dispersed periods, possibly in upland settlements. Although not examined in the current study, the Kirkland site does appear to conform to Hofman’s model in that the population is comprised of young children and older adults (Haskins, personal communication).
However, the paucity of secondary burials and cremations at Indian Knoll and the other
Green River sites is problematic. Hensley (1994) does classify Barrett, Butterfield,
Carlston Annis, Chiggerville, Indian Knoll, Read and Ward as aggregate sites within
Hofman’s model.
Charles and Buikstra (1983; Buikstra and Charles, 1999) examined the middle to late Archaic mortuary program in the lower Illinois River valley. Charles and Buikstra
31 suggest that Archaic groups use mortuary facilities to mark the landscape and define
“corporate space” within a region. Within this model a shift in settlement patterns and resource competition resulted in a change in the definition of corporate space across the landscape. There is a shift in mortuary practices from the Helton to Titterington phases in the central Mississippi drainage. The bluff top burial facilities become more common during the Titterington phase as compared to Helton phase burial sites. Charles and
Buikstra (1983) attribute this to a possible shift in subsistence base and settlement pattern. Middle to Late Archaic populations also utilized special floodplain burial facilities. These sites are “notable for a high density of artifacts overall, for a remarkable variety of projectile points, for the presence of unassociated tools and debitage reflecting maintenance and extractive activities, for the predominance of reburials…and for a remarkable quantity of bannerstones” (Buikstra and Charles, 1999:208). Although bluff top mound burial facilities have not been identified in the Green River drainage, the burial middens are strung along the river system with the larger sites marginally separated. As mentioned above, the percentage of reburials at the Green river sites is quite low. Mensforth (2001) has documented a pattern of interpersonal violence within the Green River sample, indicating possible competition for resources.
Claassen (1991, 1992, 1996a, 1996b) also interprets the Green River middens as something different than seasonal aggregation burial locations. In a reaction to what she views as basic ecological or optimization models, she suggests that the shell mounds represent ritual mortuary facilities where people have deliberately buried their dead in association with mounded shell. Claassen suggests that the shell, dirt and artifacts were
32 transported to the site as offerings to the dead. The artifacts could represent secondarily transported materials included in the soil (Claassen 1992). Claassen stresses the need for better chronometric control at shell mounds to help in the interpretation of site formation processes and occupation duration. She has further suggested that the shell may have been transported over long distances to the midden locations. In response to Claassen’s hypothesis concerning shell transportation, Morey and Crothers (1998) found that pre-
impoundment navigation maps of the Green River indicate the presence of shoals and
shallow areas near most shell middens and these shallows would have provided ideal
shellfish beds near the site locations. Therefore, an abundant source of shellfish would
have been nearby almost all shell mound locations.
SMAP
Beginning in 1972, William Marquardt and Patty Jo Watson initiated fieldwork at
Carlston Annis (15BT5) in the Big Bend of Green River. This fieldwork focused on identifying early indigenous cultigens. The late Archaic Green River shell middens seemed to be a logical next step in the investigation of an interesting horticultural complex found in paleofecal samples recovered from Mammoth Cave. Watson and colleagues (Watson, 1969, 1974; Yarnell, 1969, 1974) identified the remains of squash
(Cucurbita pepo) and gourd (Legenaria siceraria) in the material from Mammoth and
Salts Cave in association with a suite of indigenous cultigens, such as sunflower
(Helianthus annus), sumpweed (Iva annua), goosefoot (Chenopodium sp.), and maygrass
(Phalaria caroliniana). Excavations in the vestibule in Salts Cave suggested that
cultivation of these indigenous species predated the tropical cultigens. Given the dearth
33 of known open-air habitation sites near Mammoth Cave at the time, Marquardt and
Watson (1983) viewed the Green River middens, located downstream from the park, as prime locations for the development of horticultural activities. The researchers initiated work at Carlston Annis (15BT5) with two goals: (1) obtain material for radiocarbon dates and (2) recover charred botanical material for analysis. They were quite successful in the first year but the identification the Mammoth Cave suite of indigenous cultigens proved elusive. The bulk (~90%) of the botanical remains from that initial test pit were identified as hickory nutshell. However, seven small cucurbit rinds were found in the sample. At the time, these small fragments represented the earliest cucurbits in eastern
North America. This date was quickly supplanted by evidence from other regions of the eastern woodlands including central and east Tennessee (Normandy and Tellico
Reservoirs), the lower Illinois River valley, and eastern Kentucky.
In subsequent years Marquardt and Watson returned to the Green River and expanded the Shell Mound Archaeological Project (SMAP). The researchers found that the middens at Carlston Annis and Bowles (15OH13) suggested extremely complicated depositional histories. A multidisciplinary team of researchers and students was utilized to address a multifaceted research design (Marquardt and Watson, 1983). Numerous radiocarbon dates were obtained from BT5 as well as Bowles and Peter Cave (15BT94).
Louise Robbins and Steven Ward initiated the examination of the skeletal material from
15BT5 which was completed by Mensforth (1986, 1990). Julie Stein’s work (1980,
1982) at Carlston Annis specifically addressed site formation processes and provided an
34 interpretive base for examining the artifacts, radiometric determinations and botanical
remains.
A progression of researchers has since expanded SMAP. Topics examined
include lithic and faunal studies from several sites (Hensley-Martin, 1986; White, 1990),
skeletal biology research (Herrmann, 1990; Haskins and Herrmann, 1996) and
settlement/subsistence (Hensley, 1994; Crothers, 1999). Currently, Indian Knoll (Fenton
et al., 1999; Fenton and Herrmann, 2000; Herrmann and Fenton, 2000; Morey et al.,
2002) and Chiggerville (Morey and Crothers, 1998; Crothers et al., 2002) are the focus of
several recent field investigations addressing midden stratigraphy, paleoenvironmental
reconstruction and chronology.
Individual Site Descriptions
Indian Knoll (15OH2)
The Indian Knoll site is the largest Green River shell mound excavated by the
WPA. The site is synonymous with the cultural tradition known as the Shell Mound
Archaic in the Eastern Woodlands of North America. The site is a large mound incorporating habitation debris, shell deposits, mounded earth, and human burials. Indian
Knoll encompasses the crest of a natural levee of the Green River approximately 550ft
(168m) from the riverbank. Clarence B. Moore provided the first description of “[t]he
Indian Knoll” in his classic 1916 publication on his investigations along the Green River.
Moore (1916:444) states that the “Knoll is composed of dark soil, rich with admixture of
organic matter, containing considerable shell in varying proportions scattered throughout,
but nowhere forming nearly a homogeneous deposit.” 35
Figure 2-3. WPA excavations at Indian Knoll. William S. Webb Museum of Anthropology Negative No. 4201.
Two major excavations were undertaken at the site. In the fall and winter of
1915-16, Moore excavated at Indian Knoll and recovered portions of 298 burials (Moore,
1916). Beginning in 1939, a more extensive investigation was conducted under the auspices of the WPA and directed by William S. Webb. Marion Baugh, a trained geologist, supervised the field investigations. Webb’s goal at Indian Knoll was to supplement Moore’s earlier work with controlled excavations. Webb feared that little of the site would be found undisturbed after Moore’s extensive excavations. Much to Webb and Baugh’s surprise, a majority of the site remained intact, and the WPA investigations resulted in the recovery of thousands of stone tools, worked bone objects, shell beads, and 880 human burials (Figure 2-3). WPA workers excavated a block exceeding 65,000 square feet in area (Figure 2-4). The site was excavated in ten-foot blocks along trenches by six-inch levels. In some locations cultural deposits extended to a depth of ten feet below the original ground surface. The extensive WPA excavations were summarized in two publications. Webb (Webb, 1946) described the archaeological material and mortuary practices at the site, and Charles E. Snow (Snow, 1948) reported on the skeletal remains recovered during Moore’s excavations and the WPA investigations.
The Indian Knoll skeletal series represents over 1,100 individuals. The burial sample comprises one of the largest North American hunter-gatherer skeletal collections excavated from a single site. Occupation of the site spans from the Archaic to the
Mississippian period as defined in the Eastern Woodlands of North America (Smith,
1986). However, the burial sample almost exclusively dates from the late Middle
Archaic to Late Archaic periods. The sample has been the focus of numerous
37 Indian Knoll 15Oh2
$ $ $$$ $ $ $ $ $ $ $$ $ $ $ $$ $ $ $ $ $ $ $ $$ $ $$$ $ $ $ $ $ $ $$$ $ $ $ $ $ $ $ $ $ $ $ $ $ $ 390 $ $ $ $ $ $$ $ $ $$ $ $$ $ $$ $ $$ $ $ $ $ $ $ $ $ $ $ $ $$ $ $$ $ $ $$$ $ $$$ $ $ $$ $ $$ $ $ $ $$$ $ $ $ $ $$$ $ $ $ $ $ $ $ $ $ $ $ $ $ $$ $ $ $ $$$ $ $ $ $ $ $ $$ $ $ $ $ $ $ $ $ $ $ $$ $ $ $ $$$ $ $ $ $$ $ $ $ $$ $$$ $ $ $$ $$ $$ $ $$$$$$ $ $$ $ $$ $ $ $ $ $ $ $ $ $ $$ $ $$ $ $ $$ $$$ $ $ $$$ $ $$ $ $ $$ $$ $$ $ $ $ $ $$ $ $ $ $ $ $$ $ $ $ $$ $ $ $ $ $ $ $ $ $ $ $$$ $ $ $ $ $$$ $ $ $ $ $ $ $ $ $$$$ $ $ $ $ $$ $$ $ $ $$$ $ $ $ $$ $ $$ $ $ $ $$ $ $$ $ $ $ $ $ $$ $ $ $ $ $ $$$ $$ $$ $ $ $$ $ $ $ $ $ $ $$ $ $ $$ $$ $ $ $$ $ $$$ $ $ $ $ $ $ $ $$$$ $ $ $ $$$ $ $$ $ $ $ $ $$ $$ $ $$ $ $$ $$$ $ $ $$ $$$ $$ $ $ $ $ $ $ $ $$ $ $ $ $ $$$$ $$$$ $$ $$$ $ $ $$$ $$$$$$ $ $ $ $ $$ $ $$ $$T$ $ $ $$$ $ $ $ $$$ $ $ $$$$$$ $ $ $ $$$$$$$ $ $ $ $ $ $$ $ $ $$ $ $ $ $$ $ $ $ $ $$ $$$$$ $ $$$$ $ $$ $ $ $$$$ $ $$ $ $ $ $$ $ $ $$$ $ $$$ $$ $ $ $ $$ $$ $$ $$ $ $ $ $ $ $$ $ $ $$ $ $$ 827 $ $ $ $ $ $ $ $ $ $ $ $$ $ $$ $ $$ $ $ $$$ $ $ $$$ $$ $$ $$ $ $ $$$$$ $ $$ $$ $ $ $ $$$$ $ $ $ $$$$$ $ $ $ $ $ $$$$$ $ $ $ $ $ $ $ $ $ $ $$$ $ $ $ $ $$ $$ $ $ $$$$ $ $ $ $ $ $ $ $$ $ $ $ $ $ $ $$$$$ $ $$ $ $ $ $ $$ $ $ $$$$ $ $ $$ $$$$$ $ $$ $ $ $ $ $ $ $ $ $ $$ $$ $ $ $ $$ $$$ $ $ $ $ $ $ $$ $ $ $$$T$ $$ $ $ $ $$ $ $ $ $ $ $$$ $$ $ $ $$ $ $ $ $$ $ $ $$ $ $$ $ $ $ $ $ $ $ $ $$$ $ $$$ $ $ $$$ $ $ $ $$$ $ $ $ $ $ $ $ $ $ $ $$$$ $ $ $ $$ $ $ $ $ $ 612 $ $ $ $$ $ $ $
386
N
10 0 10 Meters 381 30 0 30 Feet
$ / $T Burials / Dated Burials
One-Foot Contours 378
Excavation Block 378 Figure 2-4. Planview of Indian Knoll (15Oh2) with burials, contours and excavation block. osteological studies (see Snow, 1948; Johnston and Snow, 1961; Cassidy, 1972, 1980;
Sullivan, 1977; Wyckoff, 1977; Kelley, 1980). Moore (1916) provided basic burial data,
descriptions of the artifacts recovered during his investigations, and a summary of Aleš
Hrdlicka’s typological assessment of the Indian Knoll crania. Hrdlicka concluded that
the crania represented “typical, undeformed, Algonquin skulls” unrelated to the Shawnee
(Moore, 1916:448). Webb provided simple burial demographic data and contextual
information in the Indian Knoll monograph (Webb, 1946). The age and sex determinations provided in Webb’s publication were derived from Snow’s skeletal analysis. In Indian Knoll Skeletons, Snow (1948) provided a basic age-at-death distribution of the burial sample, described unique pathologies, tabulated metric and discrete observations, and presented a detailed typological analysis of the complete crania.
Thirteen years after the initial analysis, Francis Johnston and Snow re-evaluated the original age estimates in light of new and refined aging methods, a three component pubic symphysis aging system developed by McKern and Stewart (1957) and a standardized dental attrition technique (Johnston and Snow, 1961). The new age-at-death distribution significantly increased the number of individuals over 30 years old as compared to Snow’s original assessments, but the number of adults over 50 years old decreased from four individuals to one. Since Johnston and Snow’s paleodemographic analysis, the Indian Knoll collection has served as an excellent comparative sample for numerous researchers examining issues of subsistence change. Typically, these researchers compared the mortality profile and the pattern of pathological lesions of the
39 Indian Knoll collection to various skeletal lineages from populations that practiced
different subsistence strategies, frequently later horticulturists or maize agriculturists.
Blakely (1971) compared data from Indian Knoll to the Mississippian Dickson Mounds
sample, Cassidy (1972, 1980) contrasted Indian Knoll with the Fort Ancient Hardin
Village series, and Kelley (1980) examined the Northern Plains Mobridge and
Southwestern Grasshopper Pueblo collections relative to Indian Knoll. Nagy (2000) and
Wilczak (1998) examined musculoskeletal stress markers to address questions concerning activity patterns and social distinctions in the Indian Knoll sample as compared to other skeletal series.
Recent archaeological and skeletal investigations on the Indian Knoll collections
include field and laboratory research. The archaeological research has focused on
clarifying the excavation block boundaries both from field and laboratory data and
deriving new radiocarbon dates from burials (Herrmann and Fenton, 2000) and
stratigraphic context (Morey et al., 2002). Combined with the earlier radiocarbon dates
(Arnold, 1951; Winters, 1974), a total of ten dates are available from Indian Knoll (see
Radiometric Determinations Section). The corrected and calibrated recent AMS dates
span a 2000-year range from 3600 BC to 1700 BC. Field investigations have identified a
shell-free midden similar to that identified at 15BT5 as well as other Green River
middens. The significance of this stratigraphic zone is unclear. Crothers (1999) has
suggested that it may represent a subsistence shift away from mollusks towards a greater
dependence on other resources, possibly managed wild plant species as would be
predicted based on Smith’s (1986) model of early domesticates. The seasonal rounds
40 may have shifted toward periods when it was not advantageous to exploit shellfish as
compared to other food sources. A major problem with the wild plant hypothesis, as
Crothers (1999) and others have pointed out, is the dearth of native cultigens at the shell
middens thus far investigated. Laboratory investigation has focused on clarifying the
excavation block boundaries based on coordinate data derived from the artifact catalogue
cards which are transferred into the site GIS to identify excavated blocks (Fenton and
Herrmann, 2000; Herrmann and Fenton, 2000).
Skeletal research on the Indian Knoll collection gathered in tandem with the present study entails a re-examination of adult mortality (Herrmann 1998; Herrmann and
Konigsberg 2002). Given that all previous age-at-death distribution estimates for the
Indian Knoll skeletal collection are based on biased ageing methods (Konigsberg and
Frankenberg, 1992), this preliminary research was initiated to provide an unbiased age- at-death distribution and, if need be, individual age estimates for adults from Indian Knoll as well as the other Green River shell middens.
Read (15BT10)
The Read Site is located on the north bank of the Green River in the “Little Bend”
(Figure 2-5). The site is located on a bluff top overlooking a narrow floodplain.
Excavations at Read began in December 1937 under the supervision of Albert G.
Spaulding and continued until March 1939. In October 1938, Ralph D. Brown replaced
Spaulding as field supervisor. Webb (1950a) summarized the field investigation in a brief report that included a listing of burials and a discussion of the atlatl. The Read
41 Read $ 463 465 15Bt10 463 467 $ 469 $ $ $ $$$ 471 $ $ $ $ $ $ $$ $ 473 $ $$ $ $ $ $ $ $ $ $ $ $ $ 5 $ 86 $ $$$ 47 $ $T$ $ $ $$$ $ $ $ $$$$ $ $ $ $ $ $$$ $ 31 $ $ $ $ $$ $$ $ $ $ $T$T 15A $ $ $ $$$$ $$ $$ $$$$ $ $$ $$$ $ $ $$$$ $ $$ $ $$ $$$$ $ $$ $ $ $$$ $ $ $ $ $ $$ $ $$$ $ $ $ $$ $ $ $ $ $ $ $ $ $ $$ $ $$$ $$ $ $$$ $$$ $$$ $$ $$$$$ $ $ $$$ $ $ $ $ $ $ $ $$ $$ $ $$ $ $ $ $$$ $$ $$ $ $ $ $ $ $ 465 $$$ $ $ $ $$ $ $ $$$ $ $ $ $ $$ $ $ $ $ $ $$ $ $$$ $ $ $$$$$$$ $$ 4 $ 71 $ $$ $ $ $ $ $ $ 469 $
465 4 67 $ 463
$ Burial $T Dated Burial 30 0 30 Feet Two-foot Contours GN Excavation Block 10 0 10 Meters Areas with Shell
Figure 2-5. Planview of Read (15Bt10) with contours, burial locations and shell areas depicted. report also proved significant in that it was the last Kentucky Shell Mound report
authored by Webb.
The midden at Read consisted of two discrete shell deposits overlaying sterile
subsoil clay. Burials were placed in pits excavated into the subsoil clay and within the
midden fill. Webb reports that the entire site, an area encompassing approximately
41,000 square feet, was excavated. The site was staked in five-foot squares and
excavated in nine-inch levels. Excavations progressed across the midden from north to
south by backfilling previously excavated areas. During these investigations, Spaulding
and Brown identified 247 human burials and 63 dog skeletons (Webb, 1950a). Over one
hundred burials were not recovered due to poor preservation and as such it was “deemed
inadvisable to attempt a complete physical anthropological study of the skeletal material”
(Webb, 1950a:367). Snow examined only those burials considered of interest for the
1950 report. The tabulated burial data in Webb’s report with age and sex estimates is
based on “hasty judgment” of the field supervisor (Webb, 1950a).
Since Webb’s publication, several researchers have investigated the
archaeological and skeletal material from Read. Rolingson (1967) examined the
temporal and spatial distribution of projectile point types from the site as well as the other
shell mounds. Hensley-Martin (1986) analyzed the lithic assemblage, including the projectile points, from Read in order to gain a better understanding of the lithic production and use. In 1989, Haskins and Herrmann inventoried and analyzed the human skeletal material from Read. Herrmann (1990, 1996) reconstructed the demographic profile of the Read collection. Haskins focused on the collection of paleopathological
43 data (Haskins and Herrmann, 1996) and obtaining radiocarbon samples (Haskins, 1992).
One hundred and seventy-three individuals, most of which were extremely fragmented, were identified during the analysis.
The three radiocarbon dates available from Read cluster around 1700 BC
(calibrated). These dates are derived from bone samples from three burials, but the stone projectile points suggest site use dating back to the early Archaic (Hensley-Martin, 1986;
Milner and Jefferies, 1998). Extensive use of the site probably did not occur until the
Late Archaic. Late Archaic projectile points are far more abundant than Early and
Middle Archaic point types. Associated burial goods suggest that most interments occurred during the Late Archaic, which is consistent with the radiocarbon dates.
In 1998, Milner and Jefferies re-examined the Read collection from an archaeological and skeletal biology perspective. Their efforts addressed recent hypotheses proposing that the shell middens along the Green River represent intentional funerary monuments or corporate group markers (Claassen, 1991, 1992, 1996a, 1996b;
Charles and Buikstra, 1983). Milner and Jefferies (1998) examined both the archaeological and skeletal data from different perspectives than previous researchers, but they failed to provide any new insights into the Late Archaic. Their conclusions provide a rather dry but accurate perspective on Archaic period research in which they state:
The overall picture seems straightforward. Hunter-gatherer groups, which consisted of people of all ages and both sexes, repeatedly occupied the Read site for some part of their annual subsistence cycle. They discarded considerable amounts of debris near their encampments. When they died,
44 they often were buried on the side of the ridge that faced the valley (Milner and Jefferies, 1998:130)
Carlston Annis (15BT5)
Funkhouser and Webb (1928) initially reported the Carlston Annis mound as the
“De Weese Mound” and the site is listed as such in the Archaeological Survey of
Kentucky (Funkhouser and Webb, 1932). Webb renamed the site for Mr. Carlston Annis, the landowner at the time of the WPA excavations (Marquardt and Watson, 1983:335).
The site is located “on the northeastern side of the central section of the great double bend of the river and only some 325 feet from the present bank” (Webb, 1950b:267).
The mound itself rises approximately 1.8 m (~6 ft) above the surrounding floodplain and measures approximately 350 feet by 300 feet. The shell deposits extended up to seven and half feet below the surface. The mound represented the only topographic relief adjacent to the river and it is evident on the Cromwell USGS quadrangle as an isolated
400 ft contour line encircling the site. Given its favorable topographic setting, the mound also attracted the attention of early settlers to the area. A house, barn and well were placed on the mound in the nineteenth century. These structures were dilapidated when the WPA began work at the site in 1940. One stipulation placed on the WPA excavators at the site was the removal of the old barn which was to be replaced with a new structure
(Webb, 1950b:267).
Excavations at the site began in June 1940 under the direction of Ralph D. Brown and lasted until September 1941. Mr. James R. Greenacre replaced Brown early in the excavation effort and continued as field supervisor until investigations were completed in 45 1941. Following standard WPA procedures, the site was staked off in five-foot squares
(1.52 m) and excavated in five-foot blocks by six-inch levels. Webb described the use of controlled block excavations but the location of these blocks is unknown. An area measuring approximately 21,000 square feet (1950 m²) was cleared as part of the WPA investigations (Figure 2-6) and 390 burials were recovered. The site stands out relative to the other Green River Archaic sites in the number and density of features identified
(n=129; Webb, 1950b:271). This fact may simply reflect one excavator’s bias to identify hearths and artifact concentrations as features.
This site is by far the most studied Green River Archaic shell mound. Rolingson
(1967) examined the temporal and spatial distribution of projectile point types recovered from the site as part of her regional research of archaic middens. Rothschild (1975,
1979) included burial data from Carlston Annis in her study of prehistoric mortuary practices. In 1972, Marquardt and Watson initiated SMAP to investigate questions surrounding early plant domestication. (See the previous discussion of the Shell Mound
Archaeological Project objectives).
Watson and Marquardt began their investigations with the excavation of two units
(Operation A and B). Excavations continued in 1974 (Operation C) and 1978 (Operation
D, E, F, and G). Marquardt and Watson coordinated research by several individuals on the new excavation data as well as the enormous WPA collections. Julie Stein (1980;
1982) and Alan May (1982) examined the geoarchaeological context of the midden to address taphonomic and site formation issues at Carlston Annis. Stephen Ward (n.d.) and
46 Carlston Annis 15Bt5
$
$ $ $$ $ $$$ $$$ $ $ $$ $$$ $ $$$ $ $ $% $$ $ $$$ $ $ $ % $ $ $ $ $$ $ $ $ $ $ $ $ $ $ $ $$ $ $ $$$$ $$ $ $ $ $$ $ $ $ $ 4 $ $ 4 $$ 0 $ 0 $ $ 4 4 6 $$ 7 4 $ 4 $ 0 0 0 1 8 9 5 $$ $ 0 $$ $ $ $ $ $ $% $ $ 4 $ $ 0 $ $ $ $ 4 $ $ $ $$ $ $ $ $$$ $ $ $% $$ $ $ $ $ $ $ $ $ $% $ $ $ $ $$$$$$ $$ $ $$ $$ $ $ $$ $ $$ $% $ $ $$ $$$ $ $ $ $ $$ $ $ $$$$ $ $ $ $ $ $$ $ $ $ $ $ $ $ $ $ $ $ $% $% $%$ $ $ $$ $ $ $ $ $ $ $ % $$$% $ $$ $ $$%%% $$$$ $ $ $ $$$$$$ $$ $$$$ $$ $ $ $$$ $ $$% $ $$$$$$$%%$ $%%% $$ $ $ $$ $$ $ $ $ $ $$$ $$$ $ $ $ $ $ $$ $% $ $ $ $ $ $ $ $ $$ $ $ $ $ $ $ $ $ $$% $$$$$% $ $$ $ $ $ $ $$ $ $ $ $$$$ $ $$$ $ $ $ $ $ $$ $ $$$ $ $ $ $ $$ $ $ $ $ $ $$ $ $ $ $ $ $ $ $ $$ $ $ $ $ $$ $$ $$ $ $$ $ $ $ $ $ $% $ $ $$ $$ $$$$% $ 40 $ 1
4 0 3 $$$$$%%%%%%%%%%%%%
$ Burial (Red square indicates 5 0 5 Meters that exact location unknown) GN One-foot Contours (ft AMSL) 20 0 20 Feet Structures present in 1939 Excavation Block Figure 2-6. Planview of Carlston Annis with topographic line, structure locations and burials. Robert Mensforth (1986, 1990) conducted investigations into the skeletal biology of
15BT5. The skeletal material from 15BT5 was never washed until the 1980s when the
collection was transferred to Kent State University which further complicated
Mensforth’s task. Seventeen radiocarbon dates are available from Carlston Annis as a
result of the WPA and SMAP investigations. Five dates are based on the solid carbon
method and represent samples submitted by Webb.
Chiggerville (15OH1)
Funkhouser and Webb first described the Chiggerville site in 1928 as part of a
volume entitled Ancient Life in Kentucky. Affectionately named by the local residents of
the area, the site is located approximately 300 feet (91 m) north of the river on a wide
bottom (Figure 2-7). The river bends towards the north at this location. The Baker site
(15MU12) is located on the opposite bluff top and slightly downstream from
Chiggerville. The site measures approximately 200 feet (61 m) by 100 feet (30.5 m) and
rises 7 feet (2.1 m) above the natural terrace elevation. In the summer of 1924, Webb,
Funkhouser and Mr. W. J. Curtis visited the site and excavated a “series of exploratory
holes.” They trenched completely through the center of the site from north to south.
Webb and Funkhouser were amazed at the density of the shell at the site and provided the
following observations concerning their investigations:
The mound was found to consist of an almost solid bed of mussel-shells for a depth of six feet. Mixed with the shells were human, elk, deer, wildcat, and wild-turkey bones but no complete skeletons were found in this trench…Below the mussel-shells was found a layer of made dirt with
48 Chiggerville 376 15Oh1
8 37 $ 379 $ $
$$
$$$$ $ $ $ $ $ $ $ $ $ $$ $ $ $ $ $ $ $ $ $ $ $ $ 77 $ $ $ $ 3 $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ 377 $ $$ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ 6$ $ $ $ 37
$ $$ $
$ $ $ $ $ $ $ $ $ ° $ 375 $ 3 To the River 74
$ Burial GN One-foot Contours 3 0 3 Meters Excavation Block 10 0 10 Feet Figure 2.7. Planview map of Chiggerville (15Oh1) with one-foot contours and the burial locations. charcoal to an average of two inches and below this the natural river sand which has evidently not been disturbed (Funkhouser and Webb, 1928:157- 8).
WPA excavations began at the site in April of 1938 and continued until July 1938
under the field supervision of David B. Stout with the assistance of Marion H. Baugh
(Milner and Smith, 1986). Given the short excavation time at the site, the extent of the
archaeological investigations were limited to a small block area encompassing 8,900
square feet (827 m²). The site was excavated in ten-foot blocks by six-inch levels. Two large blocks were left unexcavated in the center of the site. These areas probably encircled the previous excavations by Funkhouser and Webb (1928). One hundred and fourteen burials were recovered from the site. These burials concentrated along the southern, or river side of the site, a pattern seen at most river bottom Archaic middens
(Figure 2-7). The Chiggerville site report published in 1939 is the first description of the
WPA investigations of an Archaic midden along Green River. The report included a basic description of the excavations, features, burials, and the artifacts recovered. Ivar
Skarland provided an extensive metric and descriptive analysis of the burials (see SMA
Biodistance Chapter). In addition to the skeletal analysis, Webb provided a thoughtful discussion of the atlatl and bannerstones, which Moore (1916) thought were “net spacers” or “sizers.”
At this time, no radiocarbon dates are available from Chiggerville. However, as with all the Green River Archaic sites, Rolingson (1967) examined the temporal distribution of projectile point types recovered from the site. A distribution plot of point
50 types by Rolingson’s (1967) clusters, which are temporally ordered, indicates that
Chiggerville should be contemporaneous with Indian Knoll and Carlston Annis (Figure
2-8). Rolingson’s Cluster V predominates the Chiggerville sample followed by Cluster II
and VIII.
In the mid-nineteen seventies, two graduate students from Western Michigan
University, Norman C. Sullivan and Larry M. Wyckoff, re-examined the Chiggerville
skeletal collection for their Master of Arts theses. Sullivan (1977) investigated the
demography and pathology, and Wyckoff (1977) looked at the biological relationship and
growth. Wyckoff’s conclusions will be examined in Chapter 5 which details previous
biodistance studies on SMA populations. Sullivan (1977:52) found that the Chiggerville
mortality profile differed slightly from other Archaic groups, but it was more similar to
later agricultural populations. The pattern in age-at-death distribution is probably related to sampling error due to preservation and the age indicators utilized by Sullivan.
Crothers and Morey (Crothers et al., 2002) initiated archaeological investigations in an effort to evaluate the site formation processes, geomorphology, and subsistence practices.
Ward (15McL11)
The Ward Site is located in McLean County on a ridge crest adjacent to Cypress
Creek south of the Green River (Figure 2-9). The site overlooks the two-mile wide flood plain. The site lacks substantial shell in the midden deposit and is considered a rock midden (Hensley, 1994). A substantial quantity of shell was located on the slope leading up the ridge from Cypress Creek, but these materials were not transported to the ridge crest. Webb and Haag (1940) reported that the surface area of the entire site was quite
51 100%
80%
VIII VII 60% V IV III 40% II
Pecentage of Points I Paleo
20%
0% Bt5 Bt10 Oh13 Oh1 Mu12 Oh2 Oh19 McL12 McL11 McL4
Green River Site
Figure 2-8. Distribution of Projectile points from Green River sites grouped by Rolingson's Clusters. Chiggerville (OH1) is very similar to Indian Knoll (OH2) and Carlston Annis (BT5).
52 $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $$ $ $ $ $ $ $$$ $ $ $ $ $ $ $ $ $ $ $$ $ $ $T421 $ $$ $$ $ $ $$ $$ $ $ $ $ $$$$$$$$$$$$$$ $$ $ $$$$$$$ $ $ $ $ $ $ $$ $$$ $$ $ $ $ $ $ $$ $ $$ $ $ $ $ $ $ $ $ $ $ $$ $$ $ $ $ $ $ $ $ $$ $$$ $ $$ $ $$$$$$$ $ $ $ $ $ $ $ $ $ $ $ $ $$ $$ $ $ $ $ $$$ $ $$$ $ $ $ $$ $ $ $ $$ $$$ $ $ $ $ $ $ $ $ $ $$ $$ $ $ $ $ $ $$ $ $ $ $$ $$$ $ $ $ $ $ $ $ $ $ $$$$ $ 224 $ $ $ $ $ $T $ $ $ $ $ $ $ $$ $ $ $ $ $ $ $T $ $ $ $ 262 $ $ $ $ $$ $$ $ $ $ $$$ $$$ $$ $ $$ $ $ $ $ $ $$ $$ $ $ $ $ $ $ $ $ $ $ $$ $ $ $ $ $ $ $ $$ $ $ $$$$ $ $ $$ $ $ $$ $ $$ $ $ $ $T $ $ $ $ $ $$ $ $$ $$ 175$ $ $$ $ $ $$ $ $$ $ $$$ $ $ $$ $ $ $ $ $ $ $ $ $ $ $$ $ $$$ $ $ $ $ $ $ $ $ $ $ $ $ $$$ $$ $ $$ $ $ $T $ $ $ $ $ $ $ 44 $ 468 $ $ $ $ $ $$$ $ $ $ $$ $ $ $ $ $ $ $ $ $ $ $ $ $
$ GN 466 $ $$ 20 0 20 Feet 4 46 $ $ Burial
2 0 8 2 0 $T 6 6 6 4 5 Dated Burial 4 5 5 5 4 4 5 4 4 4 4 8 4 One-foot Contour 4 Excavation Block
Figure 2-9. Planview of the Ward site with burials, contours and excavation block shown. large with the main axis of the site measuring approximately 600 feet (183 m).
Excavations began in February 1938 and continued until September 1938 under the field supervision of John B. Elliott. After seven months of excavation, the landowner of the
Ward Site, Godfrey Ward, abruptly increased his request for crop damages from $25 to
$1000. At that time, excavations were terminated and the crew moved to Kirkland
(15McL12).
Excavations began at the base of the south side of the ridge with a 100-foot long trench along the 200-foot line which paralleled the ridge. Excavations progressed north to the ridge crest in ten-foot blocks. Webb and Haag (1940) reported that 162 ten-foot squares were excavated down to subsoil. Numerous features, burials and moderately shallow middens were encountered during the excavation. Several ceramics were recovered and a rectangular wall-trench structure was identified in the northwest corner of the excavation block. No artifacts were found in this structure, but it probably post- dates the Archaic period occupation of the site. Several burials at Ward also probably post-date the Archaic period occupation of the site.
The WPA excavators recovered 433 individual burials from the Ward site.
However, no detailed osteological study was conducted on this collection prior to 1990, and Webb and Haag (1940) provide demographic data for only those burials with associated artifacts. Several burials at Ward probably post-date the Archaic; however, these individual could not be readily identified due to the lack of associated grave goods and skeletal fragmentation. In the 1960s, Rolingson (1967) examined the temporal and spatial distributions of projectile point types recovered from the site.
54 Recently, Robert Mensforth and colleagues (Mensforth et al., 2001, 2001; Meindl
and Russell, 1998; Meindl et al., 2001) initiated a research project examining the Ward
site collections. Mensforth coordinated efforts to inventory and analyze the skeletal
material. This research is an extension of his previous work with Carlston Annis and has
provided interesting comparative data. I utilize the age and sex estimates provided by
Mensforth in the present study. These estimates were cross-checked as material was coded. Five radiometric dates, all derived from human skeletal material, are available from the Ward site (Mensforth et al., 2001; Herrmann and Fenton, 2000). Four calibrated dates span a 2200-year period from 4700 BC to 2500 BC with a fifth date being much earlier (See the Radiocarbon Dating chapter for more detail). These dates indicate that
Ward is contemporaneous to Barrett and Kirkland and slightly earlier than Indian Knoll.
Barrett (15McL4)
The Barrett site is located on the eastern side of the Pond River about 12 miles from its confluence with the Green River. The site is located on the river bottoms; however, the northern boundary of the site slopes severely down to the river (Figure 2-
10). Excavations at the Barrett site began under the supervision of John B. Elliot in
November 1938 and continued with a three-month interruption until July 1939. Work at the site was halted in February to April 1939 due to flooding (Rolingson, 1967:82;
Jefferies, 1988:19). The WPA excavated nearly one-third of the surface area and 412 human burials were recovered from the excavation block. Numerous historic features including the foundations of a sawmill intruded into the midden, complicating excavations.
55 Barrett 15McL4 371
$ 373
$ $ 375 $ $ $ $$ 77 $ 3 $ $ $ $ $ $ $ 9 $ 7 $$ $ 3 $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ 1 $$ $$ $$ $ 38 $$ $ $ $ $$ $ $ $ $ $ $ $ $$ $ $ $ $ $ $ $ $ $$ $ $ $ $$ $ $ $ $ $ $ $$ $ $ $ $ $ $ $ $ $$ $ $ $ $ $$$ $ $ $ $ $ $ $ $ $ $ $ $ $$ $ $$ $ $ $ $ $$ $ $$ $ $ $ $ $ $$ $ $$ $$$ $$ $ $$ $ $ $ $ $$ $$$ $ $ $ $ $$$$$$ $ $$ $$ $ $ $$ $$T $ $$ $ $ $ $ $ $ $$$ 100 $ $$ $ $$$ $$$ $ $ $ $ $ $ $ $ $ $$ $ $ $ $ $ $ $ $$ $ $$ $ $ $ $ $ $ $ $ $$ $$ $ $ $ $$ $ $ $ $ $ $ $ $ $ $ $ $$ $ $ $ $$ $ $ $$ $$ $$ $$ $ $ $ $ $ $$ $ $ $$$$$$$$ $$ $ 3 $ $$$ $$$$ $$$$ $$ $ 83 $ $ $ $ $$$ $$ $ $ $ $$$ $ $$ $ $ $ $ $ $ $ $ $ $ $T$ $ $ $ $$$ $ $ 87 $ $ $ $ $$ $ $ $ $ $ $ $ $ $ $ $ $ $$$ $ $$ $ $$ $ $ $ $ $ $ $ $ $ $ $ $$ $ $$ $
$
$
GN 5 0 5 Meters 20 0 20 Feet
$ Burial $T Dated Burial One-foot Contours Excavation Block
Figure 2-10. Planview of the Barrett Site, 15McL4. Burials 100 and 87 represent the two burials that have been radiocarboin dated. Research on these collections is limited. Webb and Haag (1947) published a brief
summary of the site providing only a limited description of the burial sample. As with all
the Green River Archaic sites, Rolingson (1967) examined the temporal and spatial
distribution of projectile point types recovered from the site. Winters (1974) classified
Barrett as a base camp based on an examination of the types of artifacts found at the site
as enumerated in the WPA publication (Webb and Haag, 1947). He also classified Indian
Knoll and Carlston Annis as base camps. Hensley (1994:247) reclassified Barrett, Indian
Knoll, Carlston Annis, Read, Chiggerville and probably Ward as aggregation sites based
on the similarity of artifact types, the presence of burials, density of shell, and the
identification of numerous features.
Drs. Mary Sandford of the University of North Carolina-Greensboro and Dennis
Weaver of Wake Forest University (Sandford et al., 1998; Weaver et al., 1998) initiated work with the Barrett skeletal collection. Their studies are focused on the diagnosis and interpretation of treponematosis in New World populations. These researchers suggest
that a suite of bony changes documented in the Barrett collection is indicative of some
form of the treponemal syndrome. Currently, efforts are being made to inventory and
assess the Barrett collection by these researchers. Two radiometric determinations are
available from the Barrett site (Herrmann, in press). The dates span an 1100 year period
(uncalibrated) based on AMS analysis of bone samples from two burials (See
Radiocarbon Dating Chapter).
57 Eva (40BN12)
The Eva site is located on the western floodplain of the lower Tennessee River in
Benton County, Tennessee. The site is situated on a remnant stream-deposited natural levee in between Cypress Creek Slough and Three Mile Slough. These swales represent old stream channels that were likely swampy or active during occupation. Lewis and
Lewis (1961) suggested that Three Mile Slough represented the ancient river channel at the time of occupation. Excavations at the site began in September 1940 under the supervision of Douglas Osborne and continued until November 1940. The excavation strategy at the site was slightly different than the protocol employed by the Kentucky excavators. Initially, Osborne trenched the site using two 200-foot long three-foot wide trenches placed perpendicular and crossing at the peak of the mound. Based on the observed stratigraphy of the midden, workers excavated thirty-two ten-foot squares
(Figure 2-11).
The site presented a relatively clear-cut stratigraphic profile as compared to most of the Green River Archaic middens. Five stratigraphic units were identified at Eva.
Lewis and Lewis (1961) interpreted these as chronologically sequenced deposits with
Stratum III representing a flood or stagnant water deposit clearly dividing Strata IV and
II. These strata were divided into three cultural components based on associated artifacts.
These represent the Eva phase, Three Mile/Benton phase and the Big Sandy/Ledbetter phase. The Eva phase is the earliest and includes materials recovered from Strata IV and
V. The Three Mile phase encompasses the material recovered from Stratum II. Finally, the Big Sandy phase represents material recovered from Stratum I and the plow
58 Eva 40Bn12
3 5 3 $ .8 $ $ $ $ $ $ $ $ $ $$ $ $ $ $ $ $$$ $ $ $ $ $ 8 $ $ $$ . $ $ 9 $$ $ $$ $ $ $ 4 $ $ 3 $ $ $ $ $$ $$$ $ $ $ $ $ $ $ $ $ $ $$ $ $$$62 $ $ $T$ $ $ $ $ $$ $ $ $ $$$ $ $ $$ $$ $ $ $ $ $ $ $ $ $ $$$$ $$ $ $ $ $ $ $ $$$$ $$ $$$ $ $ $$ $ $ $ $ $$ $ $ $ $ $$ $ $ $$ $ $ $$ $ $$ $ $ $ $ $ $$ $$ $ $$$ $ $$ $ $ $$ $ $$ $ $ $ $ $$ $ $
8 .
3
4
3
8 . 9 4 3
8 N Two-foot Contours $ Dog Burials $T $/$T Human Burial / Dated 20 0 20 Feet Excavation Block
Figure 2-11. Planview of the Eva site with contours, excavation block and burials. zone. These phases span from the late middle Archaic to the late Archaic. Only one published radiocarbon date is available from Eva and this is from the Eva phase deposit
(see Radiocarbon Dating chapter). One hundred and eighty human burials were recovered during excavations, and most could be assigned a specific cultural component based on their stratigraphic position. Kneberg made every effort to reconstruct all fragmentary crania. Some of these reconstructions appear tenuous and metric data presented in Lewis and Lewis (1961) should be used with caution. Recent research on these materials includes a paleodemographic analysis by Magennis (1977). Smith examined skeletal pathology (Smith, 1996) and evidence of interpersonal violence
(Smith, 1995) at Eva as well as numerous other southeastern Archaic skeletal samples to provide a regional perspective on the incidence and pattern of pathology in these early populations. Powell (1995, see also Bedrick et al., 2000) has examined the biological relationship of Eva to other middle Holocene populations in eastern North America using dental metric and discrete data to examine questions surrounding Native American origins.
60 Chapter 3. Radiocarbon Determinations and Temporal Matrix
The number of radiocarbon dates from the Green River Archaic shell middens
varies greatly by site. Fifty-five determinations are available from nine Kentucky shell
mound sites and two dates are reported from the Eva site in central Tennessee (Lewis and
Lewis, 1961; Mensforth, 1996). Of the sites with skeletal collections examined for this
study, only 39 determinations are available. As a result of the SMAP research, Carlston
Annis provides the greatest number of dates (n=17). However, no dates are available
from the Chiggerville site. The remaining sites all have a minimum of two radiometric
determinations derived from charcoal, shell, antler, or human bone. Table 3-1 provides a list of the radiocarbon dates by site, provenience, and laboratory number. Dates were calibrated with 95% confidence intervals using OxCal 3.0 Beta (Bronk, 1994, 1995).
This chapter summarizes the available radiometric determinations by site and details the methods used to construct the temporal matrix employed in the biological distance analysis.
Several radiocarbon dates represent accelerator mass spectrometry (AMS) determinations from extracted human bone collagen (n=11). Burials from five sites
(Indian Knoll, Ward, Barrett, Eva, Kirkland) have been AMS dated. These burials were selected for a variety of reasons including stratigraphic location, associated grave goods, or observed pathological conditions. Each sample submitted to the laboratories
61 Table 3-1. Radiocarbon dates for Shell Mound Archaic Site in Kentucky and the Eva Site. Sample Lab Uncalibrated Error Calibrated Calibrated Site Number Provenience Number date (BP) (±) Lower Upper Reference Eva 40BN12 Stratum IV M 357 7150 500 7200BC 5000BC Lewis and Lewis (1961) Carlston Annis 15BT5 L 3 C-180 7375 500 7600BC 5300BC Johnson (1951); Libby (1952a) C1-L20 Uaz 5730 640 5900BC 3000BC Watson (1985) D14-2-L20 WIS-1302 5350 80 4350BC 3970BC Marquardt and Watson (1983) L 6.5 C-116 5150 300 4600BC 3300BC Johnson (1951); Libby (1952a) D14-2-L9 UGa-3393 5030 85 3990BC 3640BC Marquardt and Watson (1983) L 6.5 C-251 4900 250 4400BC 3000BC Johnson (1951); Libby (1952a) D14-2-L13 WIS-1301 4760 90 3710BC 3350BC Marquardt and Watson (1983) L 1.5-2.5* C-738 4290 300 3700BC 2000BC Libby (1952b) D14-2-L15 UGa-3391 4670 85 3650BC 3100BC Marquardt and Watson (1983) D14-3-L7 UGa-3395 4655 540 4600BC 1900BC Marquardt and Watson (1983) C13-L12 UCLA-2117I 4500 60 3370BC 2920BC Marquardt and Watson (1983) L 5.5-7.0 C-739 4335 450 4100BC 1600BC Libby (1952b) D14-2-L20 UGa-3390 4350 85 3350BC 2700BC Marquardt and Watson (1983) A1-L10 UCLA-1845A 4250 80 3080BC 2580BC Marquardt and Watson (1983) A1-L8 UCLA-1845B 4040 180 3050BC 2000BC Marquardt and Watson (1983) C3-L5* UCLA-2117B 3330 80 1880BC 1430BC Marquardt and Watson (1983) C13-L15 UCLA-2117D 2515 80 830BC 400BC Marquardt and Watson (1983) DeWeese 15BT6 B1-L4 Beta-104496 4570 80 3650BC 3000BC Crothers (1999) B2-L11 Beta-104499 4320 50 3090BC 2870BC Crothers (1999) B1-L13 Beta-104497 4650 50 3630BC 3340BC Crothers (1999) B1-L18 Beta-104498 4760 70 3660BC 3360BC Crothers (1999) Read 15BT10 Bur. 86 ISGS-2245 3470 200 2400BC 1300BC Haskins (1992) Bur. 15A ISGS-2246 3400 100 1950BC 1440BC Haskins (1992) Bur. 31 ISGS-2249 3350 70 1880BC 1450BC Haskins (1992)
62 Table 3-1 (continued). Radiocarbon dates for Shell Mound Archaic Site in Kentucky and the Eva Site. Sample Lab Uncalibrated Error Calibrated Calibrated Site Number Provenience Number date (BP) (±) Lower Upper Reference Haynes 15BT11 A1-L4 Beta-106447 4520 60 3500BC 3010BC Crothers (1999) A2-L9 Beta-102649 4650 60 3640BC 3130BC Crothers (1999) A2-L17 Beta-102650 4850 60 3760BC 3520BC Crothers (1999) A1-L19 Beta-102648 5080 90 4250BC 3650BC Crothers (1999) Indian Knoll 15OH2 L 1 C-254 5302 300 4800BC 3300BC Arnold and Libby (1951) Shell Midden 79N-2 53-60 cm TO-8792 4670 70 3640BC 3130BC Morey et al. (2002) Bur. 612 AA-31194 4570 75 3520BC 3020BC Herrmann (in press) 79N-2 36-43 cm TO-8791 4460 90 3500BC 2850BC Morey et al. (2002) 21N-3 58-63 cm TO-8794 4300 70 3100BC 2700BC Morey et al. (2002) L 1 C-740 4282 250 3700BC 2200BC Libby (1952b) 21N-3 92-100 cm TO-8793 4230 80 3020BC 2570BC Morey et al. (2002) L 4.5 C-741 3963 350 3500BC 1500BC Libby (1952b) Shell Free Midden Upper Strata NSEC 3800 80 2470BC 1970BC Fenton, personal communication Bur. 827 AA-31193 3500 60 2010BC 1660BC Herrmann (in press) Bowles 15OH13 A2-L2 UCLA-2117E 1820 300 500BC 800AD Marquardt and Watson (1983) A3-L7 UCLA-2117F 2420 200 1000BC 0AD Marquardt and Watson (1983) A3-L11 UCLA-2117G 3440 80 1950BC 1520BC Marquardt and Watson (1983) A3-L12 UAz-Oh13 4060 220 3400BC 1900BC Watson (1985) Barrett 15McL4 Bur. 100 Beta-131956 5620 40 4550BC 4350BC Herrmann (in press) Bur. 87 Beta-131957 4520 40 3370BC 3030BC Herrmann (in press) Ward 15McL11 Bur. 262 12-262 7714 50 6650BC 6440BC Mensforth (1996) Bur. 224 AA-31192 5600 100 4690BC 4240BC Herrmann (in press) Bur. 44 AA-30520 5120 90 4250BC 3700BC Herrmann (in press) Bur. 421 AA-30521 4800 65 3710BC 3370BC Herrmann (in press) Bur. 175 12-175 4134 60 2890BC 2490BC Mensforth (1996)
63 Table 3-1 (continued). Radiocarbon dates for Shell Mound Archaic Site in Kentucky and the Eva Site. Sample Lab Uncalibrated Error Calibrated Calibrated Site Number Provenience Number date (BP) (±) Lower Upper Reference Kirkland 15McL12 Bur. 18 ISGS-2297 3830 80 2490BC 2030BC Haskins (1992) Bur. 34 ISGS-2299 7320 80 6390BC 6010BC Haskins (1992) Bur. 40 Beta 82081 5680 80 4770BC 4340BC Claassen (1996b) Bur. 45c ISGS-2304 3990 160 2950BC 2000BC Haskins (1992) Bur. 45b ISGS-2304 4240 150 3350BC 2450BC Haskins (1992) Bur. 56 ISGS-2298 6600 80 5670BC 5370BC Haskins (1992)
64 consisted of 35-45 grams of rib bone fragments (Herrmann, in press). The extracted
collagen amount from all samples permitted radiometric analysis. The 13C/12C ratio
reported for the bone samples is consistent across the sites ranging from -23.1 to -20.6.
These values are comparable to other non-maize consuming middle to late Archaic populations recorded in the Eastern woodlands (Buikstra, 1992).
Radiocarbon Determination for SMA Sites
The following is a summary of radiocarbon dates for each site examined in this
study as well as radiocarbon dates derived from other Green River Archaic shell middens.
Claassen (1996b) provides a comprehensive tabulation of available radiocarbon dates for
Shell Mound archaic sites throughout the Southeast. Calibrated ranges for a select series of SMA sites excluding suspect or non-comparable dates are presented in Figures 3-1 and
3-2. A summed range for each site is also given for comparative purposes.
Eva (40BN12)
Two radiocarbon dates are available from the Eva site. The first sample
comprised of antler from Stratum IV yielded a date of 5200 ± 500 B.C. (Lewis and Lewis
1961:13). The stratum represents the Eva component near the base of the midden.
Mensforth (1996) submitted rib fragments from Burial 62 for AMS analysis because the
cranium exhibited cutmarks. Based on the pattern and distribution, Mensforth argues the
cutmarks are the result of scalping. Lewis and Lewis (1961:118) identify the individual
as an adult male and the burial was recovered from Stratum I within the Ledbetter
65 Shell Mounds
EVA - IV, M-357, 7150±500BP
B87, Beta-131956 5620±40BP
B100, Beta-131957 4520±40BP
BARRETT
B224, AA-31192 5600±100BP
B44, AA-30520 5120±90BP
B421, AA-30521 4800±65BP
B175, Mensforth 4140±60BP
WARD
WIS-1302 5350±80BP
UGa-3393 5030±85BP
WIS-1301 4760±90BP
UGa-3391 4670±85BP
UCLA-2117I 4500±60BP
UGa-3390 4350±85BP
UCLA-1845A 4250±80BP
UCLA-1845B 4040±180BP
UCLA-2117B 3330±80BP
UCLA-2117D 2515±80BP
CARLSTON ANNIS
9000CalBC 6000CalBC 3000CalBC CalBC/CalAD Calibrated date Figure 3-1. Calibrated radiocarbon dates for Eva, Barrett, Ward and Carlston Annis. Highlighted box indicates the summed calibrated range of the dates above.
66 Shell Mounds
B86, ISGS-2245 3470±200BP
B15A, ISGS-2246 3400±100BP
B31, ISGS-2249 3350±70BP
READ
79N-3, TO-8791 4670±70BP
B612, AA-31194 4570±75BP
79N-2, TO-8792 4460±90BP
21N-3, TO-8794 4300±70BP
21N-4, TO-8793 4230±80BP
INDIAN KNOLL - SHELL MIDDEN
NSEC - Webb 3800±80BP
B827, AA-31193 3500±60BP
INDIAN KNOLL - SHELL FREE MIDDEN
B34, ISGS-2299 7320±80BP
B56, ISGS-2298 6600±80BP
B40, Beta 82081 5680±80BP
B45b, ISGS-2304 4240±150BP
B45c, ISGS-2304 3990±160BP
B18, ISGS-2297 3830±80BP
KIRKLAND
9000CalBC 6000CalBC 3000CalBC CalBC/CalAD Calibrated date Figure 3-2. Calibrated radiocarbon dates for Read, Indian Knoll and Kirkland. Highlighted box indicates the summed calibrated range of the dates above. 67 component near the top of the midden deposit. The AMS date range reported for this
individual falls completely within the range of the Stratum IV determination and is
slightly earlier. Given the stratigraphic integrity at Eva, the new AMS date appears
erroneous. Therefore, the radiocarbon dates for Eva were not used for the temporal
matrix construction. Component age ranges given by Hofman (1986:189) were used in
the temporal matrix calculation with Eva component ranging from 7,300 to 6,500 B.P.,
the Benton/Three Mile component ranging from 6,200 to 4,800 B.P. and the Lebetter
Phase/Big Sandy component ranging from 4,500 to 3,200 B.P. A weighted average by
component per individual was calculated for a mean site date.
Carlston Annis (15BT5)
Seventeen dates are available from Carlston Annis. This represents the highest
number of dates from any examined site (Table 3-1). The calibrated dates from 15BT5 span a period from 7600 cal B.C. to 400 cal B.C. However, a majority of dates fall within a range from 4600 cal B.C. to 3000 cal B.C. Five dates represent early solid carbon determinations of bulk antler, shell and bone analyzed by Libby at the University of Chicago. Similar to other early radiometric assays, the counting error on these dates is usually greater than 250 years, which makes them problematic for comparing to recent radiocarbon determinations. The remaining dates represent carbon and botanical samples recovered during SMAP by Marquardt and Watson (1983; Watson, 1985). Two of these new dates are problematic due to excessive counting errors (±640 and ±540). One sample (UAZ AMS 5730±640 B.P.) represents an early AMS date of a cucurbit sample
(Watson, 1985). Another problem with many of the SMAP dates is that they are
68 stratigraphically transposed. Marquardt and Watson (1983, see Crothers, 1999)
recognized this issue. They suggested that formation processes at Carlson Annis have
dramatically altered the stratigraphic integrity of the shell midden (Stein, 1980).
Read (15BT10)
Three dates are available from the Read site (Haskins, 1992). Bone or charcoal
samples were collected from three burials (15A, 31, and 86) and submitted to the Illinois
Sate Geological Survey for analysis. All samples were subjected to long counting to
provide more precise dates (Haskins, personal communication). Burials 15A and 31
were recovered near the center of the midden at a similar depth, and their dates are nearly
identical, within 50 years (See Figure 2-5). Burial 86 was recovered on the southwestern
slope of the site. The calibrated ranges of all three dates overlap, spanning a period from
2400 cal B.C. to 1300 cal B.C.
Indian Knoll (15OH2)
Ten radiocarbon dates are available from Indian Knoll (Figure 2-4). Three dates represent solid carbon determinations submitted to Libby at the University of Chicago by
Webb in the 1940s (Arnold and Libby, 1951; Libby, 1952b). The dates are derived from bulk shell, antler or bone samples from the midden. These dates have very large counting errors, on the order of 250 to 300 years. One additional problem with these samples is that the age determinations are inverted when compared to their stratigraphic context. A single date from the early 1960s is available from the Nuclear Science & Engineering
Corporation (NSEC) of Pittsburgh, Pennsylvania. This date represents a bulk antler
69 sample recovered from six inches below surface in unit 210R14 (Fenton, personal communication).
The remaining six dates represent recent AMS dates from specific stratigraphic contexts (n=4) or from human rib bone samples from two burials, 827 and 612. These burials were selected for two reasons. First, they were located at the top (827) and bottom (612) of the midden deposit. Second, numerous artifacts were found with these individuals. The projectile points recovered from these two burials are temporally distinctive and provide a baseline for the expected age range of the midden deposit
(Herrmann, in press). WPA excavators recovered four Benton cluster projectile points
(Justice, 1987) with Burial 612 (Figure 3-3), and seven Late Archaic stemmed points were associated with Burial 827 (Figure 3-4). Webb (1946) describes these burials in detail in his Indian Knoll publication. The AMS date for Burial 612 (4570 ± 75 B.P.,
AA-31194) is consistent with other reported determinations for Middle to Late Archaic
Benton cluster points throughout Kentucky and Tennessee (Justice, 1987:111-114). This individual was found in a large pit with three other individuals (Burials 611, 613, and
614). One additional projectile point was recovered with Burial 614. This point is very similar to the four projectiles found with 612, with the exception of a broken base. The uncalibrated AMS determination of 3500 ± 60 B.P. for Burial 827 is consistent with the expected temporal range of late Archaic stemmed projectile points in central Kentucky.
The remaining four AMS determinations date charcoal samples collected from auger holes excavated in undisturbed areas outside of the excavation block at Indian
70
Figure 3-3. Projectile Points recovered with Burial 612 at Indian Knoll
Figure 3-4. Projectile points recovered with Burial 827 at Indian Knoll. Knoll. Morey et al. (2002) placed a series of auger holes along an east-west line in the
proximity of the old excavation block. Soil samples from auger probes with undisturbed
stratigraphy were collected and processed for charcoal samples. Four samples from two
different auger holes were submitted to the IsoTrace AMS Radiocarbon Laboratory at the
University of Toronto (“TO” in Table 3-1). These samples correlated to the interface between a “shell free midden”, similar to that identified by Stein (1980) at Carlston
Annis, and “shell midden.” This differentiation is based on the percentage of organic matter and carbonate in the soil samples collected from the augers (Stein, 1980; Morey et
al., 2002). The shell free midden exhibits high organic and low carbonate values,
whereas, the shell midden exhibits a high percentage of organic matter and carbonate
(Morey et al., 2002). These four dates all collected from the shell mound span an
uncalibrated range from 4670 ± 70 B.P. to 4230 ± 80 B.P. The intercept dates for the two
samples from Auger 21N are stratigraphically inverted, but the calibrated 95%
confidence intervals overlap considerably. These dates near the top of the shell zone
combined with the date from Burial 612 suggests that the shell midden at Indian Knoll may have accumulated over a fairly short period of time, potentially as little as 500 years.
Additional dates are required to test this hypothesis.
Barrett (15McL4)
The two burials sampled from Barrett (87 and 100) were recovered near the base
of the midden deposit and within the shell deposit (Figure 2-9). A three-quarter grooved
axe and various shell artifacts are associated with Burial 87, and Burial 100 exhibits an
interesting pathological condition currently being investigated by Drs. Mary K. Sandford
73 and David Weaver (Sandford et al., 1998; Weaver et al., 1998) at the University of North
Carolina-Greensboro. These two dates are the only absolute chronometric information
available from the site. The artifact suite and mortuary program is similar to the other
shell middens examined.
Ward (15McL11)
The Ward site represents one of the better-dated shell middens. The dates are all
AMS determinations derived from human rib bone collagen. Five different burials (44,
175, 224, 262 and 421) from various areas of the rock midden have produced at least four reliable dates (Figure 2-8). Mensforth (1996) selected Burials 262 and 175 given that both individuals exhibited skeletal trauma consistent with perimortem violence, scalping, or dismemberment. Herrmann (in press) provided the remaining three dates. No artifacts are associated with these individuals, but the archaeological context of burial
421 is important. This burial was recovered from a large trench burial containing at least
19 individuals. Several of the burials were extended, and it appears that numerous other individuals (or parts of numerous other individuals) were interred next to these burials possibly as reburial episodes. Burial 421 represents the uppermost individual on the northwestern end of the trench and lies on top of at least three individuals. The exact function of the trench is unknown, but similar features have been identified and excavated at Butterfield (15McL7) and Barrett. The radiometric determination from
Burial 421 provides a terminus post quem for the trench feature. Further analysis of these
individuals and additional dates from these sites may help clarify the function and the
temporal relationship of these interesting burial facilities.
74 The date for Burial 262 from Ward (Mensforth, 1996) is extremely old as
compared to the dates provided by Herrmann and to the date reported for Burial 175.
The age range of Burial 175 falls nicely at the tail end of the age range of the three Ward site dates provide by Herrmann. The determination for Burial 262 is potentially a bad date, and I have suggested that re-dating this individual is warranted (Herrmann, in press). This date is not included in the temporal matrix calculation.
Other Sites
Haskins (1992; see Claassen, 1996b) reports two dates from Burials 34 and 56 at the Kirkland site (15McL12) that fall within the range of Burial 262 from Ward (see
Table 3-1). Kirkland lies along Cypress Creek directly west of the Ward site. Claassen
(1996b) also reports a relatively early date for Burial 40 from Kirkland (Table 3-1). In addition to these three early dates, three other dates from Burials 40, 45b, and 45c are consistent with determinations from Indian Knoll and Carlston Annis. These two divergent groups of dates may represent different periods of site occupation.
The remaining dates from DeWeese, Bowles and Haynes are listed to provide a context for other shell middens located along the Green River drainage and tributaries.
These dates are based on excavations by SMAP at Bowles (Marquardt and Watson, 1983) and George Crothers’ (1999) investigations at DeWeese and Haynes. DeWeese and
Haynes appear contemporaneous, and Bowles represents one of the latest middens along the Green River.
75 Temporal Matrix Construction
As part of the biological distance analysis presented in the following chapters, a
matrix of the temporal relationships of the shell middens is required. To account for
different standard deviations, a weighted average ( t ) of available dates from each site was calculated based on the following formula (see Geyh and Schleicher, 1990:16):
n tw å ii 1 t ±s =±i=1 i nn 3.1 ååwwii ii==11 2 where wi=1/s ti
The average radiocarbon dates and temporal matrix is presented in Table 3-2. In Table 3-
3, I have identified two separate zones at Indian Knoll and Carlston Annis: (1) Shell
Midden and (2) Shell Free Midden. These distinctions will be utilized in the biological
distance analysis presented in Chapter 6. For Eva, the mean date represents a weighted
component age calculated from the coded sample with 35, 61 and 13 individuals
representing the Ledbetter/Big Sandy, Benton/Three Mile and Eva components,
respectively.
Geyh and Schleicher (1990:16) caution that the use of a weighted mean is only
valid if it can be demonstrated that the radiocarbon dates “have the same true value te and
a common standard deviation s t.” They recommend using a Chi-square test to evaluate
this hypothesis. Read produces the only non-significant Chi-square value (see Table 3-
4). This lack of uniformity highlights the critical problem of a paucity of radiometric determinations from the Kentucky shell mounds. These sites represent a composite of numerous occupations that span several millennia. The skeletal samples derived from 76 these sites truly do represent skeletal lineages, as described by Cadien et al. (1974).
Numerous additional dates are required to help parse out the burial populations. Ideally, these dates should be based on known stratigraphic contexts or human bone samples.
Although dates from some sites represent very long periods of time, the weighted average determination will be utilized in the temporal matrix as a rough approximation of temporal association. However, interpretations based on the temporal matrix should be viewed cautiously. Currently, the entire radiocarbon data set from the Green River middens is being reevaluated with BCal (http://bcal.shef.ac.uk/) and results from this analysis will be incorporated into future research.
77
Table 3-2. Temporal matrix for the Green River Archaic Middens and Eva. BT10 BT5 McL11 McL4 OH1 OH2 BN12 Weighted Carlston Chigger- Indian Site Mean Read Annis Ward Barrett ville* Knoll Eva BT10 3374 0 BT5 4317 943 0 McL11 4711 1337 394 0 McL4 5070 1696 753 359 0 OH1 4200 826 117 511 870 0 OH2* 4180 806 137 531 890 20 0 BN12 5137 1763 820 426 67 937 957 0 * No dates are available from Chiggerville. A general date of 4200BP was assigned to Chiggerville based on the artifacts recovered at the site.
78
Table 3-3. Temporal Matrix for divided Green River Archaic Middens and Eva.
15BT10 15BT5-1 15BT5-2 15McL11 15McL4 15OH1 15OH2-1 15OH2-2 40BN12-1 40BN12-2 Eva Carlston Carlston Indian Indian Eva Benton/ Weighted Annis- Annis- Chigger- Knoll- Knoll- Ledbetter/ Three Site Mean Read SFM1 SM Ward Barrett ville* SFM SM Big Sandy Mile BT10 3374 0 BT5-11 3394 20 0 BT5-2 4430 1056 1036 0 McL11 4711 1337 1317 281 0 McL4 5070 1696 1676 640 359 0 OH1* 4200 826 806 230 511 870 0 OH2-1 3608 234 214 822 1103 1462 592 0 OH2-2 4456 1082 1062 26 255 614 256 848 0 BN12-1 3850 476 456 580 861 1220 350 242 606 0 BN12-2 5500 2126 2106 1070 789 430 1300 1892 1044 1650 0 1 No specific dates are available for the shell free midden at Carlston Annis. The two stratigraphically upper dates are considered representative of the Shell Free Midden in this case (UCLA-2117B and C-738). * No dates are available from Chiggerville. A general date of 4200BP was assigned to Chiggerville based on the artifacts recovered at the site.
79
Table 3-4. Chi-Square test of the radiocarbon dates by site. Site Chi-Square Probability DFs Barrett 378.1 0.0000 1 Bowles 58.7 0.0000 3 Carlston Annis 1006.5 0.0000 16 Chiggerville - - - DeWeese 34.2 0.0000 3 Eva† - - - Haynes 33.1 0.0000 3 Indian Knoll 254.6 0.0000 9 Indian Knoll-SM 35.0 0.0001 7 Indian Knoll-SFM 9.0 0.0027 1 Kirkland 1290.5 0.0000 5 Read 0.4 0.8124 2 Ward* 194.0 0.0000 3 † Two dates stratigraphically inverted * excluding date from Burial 262
80 Chapter 4. Quantitative and Population Genetics for Non-Metric Traits
This chapter details the theoretical foundations for the analysis of non-metric traits from the Green River Archaic skeletal samples. Recent quantitative methods have been developed for estimating specific population structure parameters. The use of anthropometrical data in the study of population histories requires different mathematical approaches than analyses based on simple Mendelian traits. Metric traits are considered polygenic traits due to the simultaneous effects of many genes on trait expression. Non- metric traits differ from metric traits in that trait expression depends on the effect of numerous genes on a trait threshold. As a result these traits are non-continuous.
Researchers have developed intricate methods to deal with the assessment of population structure based on metric and non-metric traits (e.g. Falconer, 1989). Depending on the types of data involved and the amount of pedigree information available to the researcher, certain assumptions concerning dominance, epistatis and environmental interactions need to be addressed. This chapter is divided into two sections. In the first section basic concepts of quantitative genetic theory are presented including threshold traits, heritability, distance measure statistics and FST calculation. The second section deals
with methodological approaches in population genetics.
81 Quantitative Genetic Theory
A key assumption for any analysis of anthropometric data is that a trait utilized in
the study is under some degree of genetic control. In order to assess the extent of genetic
control, a basic understanding of the mathematical approach to interpreting quantitative
traits is required. The genetics of such a trait “centres round the study of its variation”
(Falconer, 1989:125). The variation evident in a trait can be divided into its contributing
2 components. Therefore, phenotypic variation, s P , of a quantitative trait can be divided into its genetic and environmental components,
222 sssPGE=+, 4.1
where the phenotypic variance is always greater than or equal to the genetic variance
(Falconer, 1989:125). As stated above, it is possible to have interaction between
genotypes and environmental factors or dominance, in which case there will be additional
2 2 variances to account for (s I or s D , see Falconer, 1989:125-126). However, this study is
not concerned with these parameters given the type of data available. Further, the
questions to be addressed do not require this information.
The next step when considering equation 4.1 is to extend it to a multivariate
example,
PGE=+, 4.2
where P, G and E represents phenotypic, genetic and environmental covariance matrices,
respectively. Cheverud (1988) found that phenotypic and genetic correlations are very
similar in studies employing large effective samples (typically over 40 individuals).
Differences in the matrices result from sampling error, and Cheverud (1988:966) 82 concludes that “when reliable genetic [correlation] estimates are unavailable, phenotypic
correlations and scaled variances may be substituted for their genetic counterparts in
evolutionary models of phenotypic evolution.” Konigsberg and Ousley (1995) supported
Cheverud’s conclusions using anthropometric traits from five Native American Tribes in
the Boas data set. Konigsberg and Ousley conclude that the results “strongly indicate”
that the genetic variance-covariance matrix is proportional to the phenotypic variance-
covariance matrix.
Critical to the relationship of variances is the concept of heritability. In this case
we will be concerned with heritability in the sense of “breeding values” (Falconer, 1989:
125). This relationship can be expressed for a single trait as the following
2 2 s A h = 2 , 4.3 s P
where the additive genetic variance is divided by the phenotypic variance. The additive
genetic variance is used in this case due to the possible effects of dominance and
2 environmental interactions. The multivariate extension of Equation 4.3 (h = VA/VP) is
often considered heritability in the narrow sense because it expresses the “extent to which phenotypes are determined by the genes transmitted from the parents” (Falconer,
1989:126). Prior to addressing the heritability of non-metric traits, a brief discussion of
threshold traits is required.
Threshold Traits
Grüneburg (1952, 1963) has described skeletal non-metric traits as quasi- continuous traits. The quasi-continuous nature of these features is dependent on the idea 83 that “one or more thresholds [are] imposed on a continuously distributed liability…causing the phenotype to be expressed in discrete categories” (Cheverud and
Buikstra, 1981a:44). In addition, non-metric traits result from the interaction of numerous genes and environmental factors and as such are considered polygenic traits
(Falconer, 1989; Cheverud and Buikstra, 1981a). The analysis of such traits requires a different approach as compared to classic Mendelian traits.
When analyzing threshold traits in biodistance studies, often the traits are converted to dichotomous values designated 0 for unaffected and 1 for affected. The percentage of individuals affected, or incidence, can be calculated by dividing the number of affected individuals by the total number of individuals. Falconer (1989) finds this value problematic because it is difficult to compare populations given that variances in trait frequencies differ. Therefore, incidences need to be converted to mean liabilities. In order to make this conversion, it is assumed that liabilities are normally distributed or can be transformed into normal distributions. Thus, “the unit of liability is its standard deviation s ” (Falconer, 1989:302).
For example, consider the two liabilities presented in Figure 4.1. In Population A, the trait incidence is quite low, only 6%. In Population B, the trait frequency is much higher at 25%. These values correspond to liabilities for Populations A and B of 1.56 and
.67, respectively. Therefore, the populations differ by .89 standard deviations of liability.
Liability differences will then be used in our calculation of a biological distance value rather than the specific incidence, or frequency of individuals affected. Not all distance measures assume a threshold model and the distance calculation of these other methods
84 p= 6%
X1 m1
p = 25%
X2 m2
Figure 4-1. Threshold model for two populations (adapted from Falconer, 1989:301).
differs substantially (for example, Smith’s Mean Measure of Divergence or Balakrishnan and Sanghvi’s B², see Distance Measures section).
Heritability of Metric and Non-Metric Traits
The concept of heritability is critical to any anthropometric study of continuous or discrete observations. Environmental factors contribute to the overall phenotypic distance measures derived from any anthropometric data set. A majority of heritability studies have focused on metric data sets (Boas, 1908; Konigsberg and Ousley, 1995;
Cheverud, 1988; Sparks, 2001), but several studies have focused on non-metric traits from human and non-human populations (Sjøvold, 1984, Cheverud and Buikstra, 1981a,
Berry, 1974, 1975; Berry, 1968; Berry and Berry, 1967). Heritability estimates can vary depending on the method of calculation and the type of trait examined (metric versus
85 non-metric) and all methods require some form of pedigree data. This second point has
been the limiting factor in most studies of human skeletal material, given that very few
skeletal collections are available with known relationship data. Metric and non-metric studies are summarized below.
Typically, heritability studies have focused on comparing metric versus non- metric traits or assessing the significance of morphometric traits for constructing evolutionary phylogenies or biological distance relationships. Most skeletal trait studies are based on non-human samples, such as mice (Berry, 1968; Leamy, 1977) or macaques
(Cheverud, 1979, 1988; Cheverud and Buikstra, 1981a, 1981b, 1982).
Metric studies have been more commonly addressed within the anthropological literature. Specifically, Cheverud and colleagues (Cheverud, 1982, 1988; Cheverud and
Buikstra, 1981a, 1981b, 1982; Cheverud et al., 1979) calculated heritabilities of cranial measurements from an extensive macerated macaque collection with the heritabilities ranging from -.04 to .87 with an average of .32 (Cheverud and Buikstra, 1982:153). A negative heritability value is possible given the calculation method employed by these authors and the fact that they were dealing with a sample not the entire population.
Studies on human samples have been less numerous and typically deal with anthropometric variables. Sjøvold (1984) assessed heritability of a series of cranial metric and non-metric traits in a sample of historic crania from Hallstatt, Austria.
2 Sjøvold calculated heritabilities by regressing offspring on parent ( hb= 2 OP , where bOP
is the regression coefficient). Several metric traits exhibited significant heritabilities but
these vary by the parent-offspring relationship, whether father or mother/son or daughter.
86 Sjøvold’s study is limited by small sample sizes of parent-offspring pairs and uncertainty
with respect to known pedigree data. Families and individuals were identified by
correlation between church death records and family-specific stylistic decorations painted
on the crania. Recent heritability studies based on metric data have typically focused on
anthropometric data given that body measurements and pedigree data are readily
available (see Black, 1982; Devor et al., 1986: Paganini-Hill et al., 1981; Konigsberg and
Ousley, 1995; Sparks, 2001).
In an examination of anthropometric data from European immigrants measured as
part of Franz Boas’ research on the influence of the American environment of body form,
Sparks’ (2001) estimated average heritabilities for head length, head breadth and
bizygomatic breadth of .54, .57 and .57, respectively. These values are slightly higher
than estimates available from Konigsberg and Ousley (1995). Konigsberg and Ousley
(1995) did not present the specific heritability estimates but these can be obtained from
their Table 3 with the assumption that the additive genetic and phenotypic variance-
covariance matrices are proportional (see Konigsberg and Ousley, 1995:489). The
heritability estimates of six cranial traits ranged from .32 to .43 with a mean of .38. In an
examination of various phenotypic and genetic-correlation matrices, Cheverud (1988) found an average heritability of .35 for morphometric traits.
Black (1982) advocates a Wright’s path analysis approach to the study of anthropometric variation. Black found that subpopulation structure greatly influences heritability estimates. In his examination of the Solomon Island populations, heritability estimates were substantially lower than values reported by other researchers. When he
87 excluded subpopulation structure within the path analysis, then heritability estimates
were similar to other studies. Therefore, the high heritability estimates found by Sparks
for the Boas immigrant data seem appropriate given that subpopulation structure
information was unavailable. Immigrants of specific ethnic affiliation and/or from large
geographic areas were grouped.
Studies of non-metric trait heritability have followed similar paths as research on
metric data. Often non-metric studies have been coordinated with metric research in an
effort to compare estimates of phenotypic and genetic correlation/covariance matrices.
Self and Leamy (1978) examined trait heritability in random-bred mice. In general,
mouse trait heritabilities were low to moderate (ranging from .17 to .20 depending on the
method used to calculate the h² value). Within the anthropological literature, Cheverud
and colleagues (Cheverud and Buikstra, 1981a, 1981b, 1982) have extensively examined trait heritability in the Cayo Santiago macaque population. Cheverud and Buikstra
(1982) found that hyperostotic non-metric trait heritabilities are on average higher than metric traits. Counter to Cheverud and Buikstra’s findings, Richtsmeier and McGrath
(Richtsmeier et al., 1984) concluded that variation in trait expression and a poor understanding of trait development limits the interpretive power of non-metric data.
Studies of human non-metric traits are limited. Often sample size plays a critical roll in these studies (e.g. Sjøvold, 1984). Small samples result in large standard errors and divergent heritability estimates. To circumvent these issues most human studies have focused on twins and have used dental morphology (see Scott and Turner, 1997 for review). In general, dental morphology appears fairly stable across populations with
88 individual trait heritability estimates falling in “a middle range value (0.40 to 0.80)”
(Scott and Turner, 1997:164).
Cheverud and Buikstra (1982) and Ossenberg (1970, 1974) found that hyperostotic traits exhibit higher heritabilities than foraminal traits. Hyperostotic traits result from “variable ossification of connective tissue which separates or encloses nerves and blood vessels” (Cheverud and Buikstra, 1982:152). On the other hand, foraminal traits represent variable numbers of bony foramina. Variation in foraminal traits results from the presence or absence of blood vessels and nerves, the degree of branching, or the
embryological positioning of the vessel or nerve relative to developing bony regions.
Heritabilities of the hyperostotic and foraminal traits within the macaque population
average .80 for four traits and .35 for five traits, respectively. Mean values for both sets
are greater than the mean metric heritability of .32 (n=56). While these heritability
values are considerably greater than those documented in other species, it does suggest
that some traits may be more valuable in population structure studies. However, traits
such as the pterygoid bridge are difficult to record accurately in fragmentary remains.
Distance Measures
Numerous distance measures are available for the analysis of discrete traits.
Several reviews of these methods have been conducted by statisticians, biologists and
anthropologists in an attempt to demonstrate the utility of one or more measures to
address research questions specific to their field of study (see Corruccini, 1974, 1976;
Finnegan and Cooprider, 1978; Ossenberg 1979; Weiner and Huizinga, 1972). Three
specific measures have found favor in the anthropological literature: Smith’s Mean
89 Measure of Divergence (MMD), Balakrishnan and Sanghvi’s B² and Mahalanobis D².
The utility of these measures is dependent on types of data available and sample size.
Several researchers (see Balakrishnan and Sanghvi, 1968; Finnegan and Cooprider, 1978)
have demonstrated that the MMD and B² provide very similar results. Tests comparing the Mahalanobis D² with the other two methods have not been published. Below I
summarize each method and detail the slight differences in their calculation.
Smith’s Mean Measure of Divergence (MMD)
Mean Measure of Divergence was developed by CAB Smith and published in
Grewal’s (1962) examination of genetic divergence of various lines of mice. Berry and
Berry (1967) utilized this method in their analysis of human crania. The MMD statistic
offers a quantitative measure of biological divergence between samples based on the
degree of phenetic similarity for the entire suite of traits examined. Higher MMD values
indicate a greater degree of phenetic dissimilarity (ands thus genetic distance) than do
lower values. Green and Suchey (1976) recommend the following formula with the
Freeman-Tukey arcsine transformation:
2 (qqji-li) -(1(nnji+1/2) ++1( li 1/2)) MMD = 4.4 å t
11 q =22arcsin(1-2k(n+1)) +arcsin(1-2(kn++11) ( )) , 4.5
where qi equals the transformed frequency of the ith trait in a population (j or l), k equals the observed number of trait occurrences, n equals the number of crania (or sides) examined for the ith trait in population (j or l), and t equals the number of traits
90 considered. Equation 4.4 represents one possible way to calculate a MMD. Several
modifications to this measure have been proposed (see Sjøvold, 1973; Finnegan and
Cooprider, 1978). To determine if MMD values are statistically significant, each MMD
value is compared to its standard deviation where:
2 2(1(nn12ii+1/2) ++1( 1/2)) SD()MMD = å . 4.6 t 2
If the MMD value is greater than two times its standard deviation then the null hypothesis
of the samples being identical is rejected at the p = .025 significance level. An
insignificant MMD value indicates that it is impossible to distinguish between the two
samples because either the samples are phenetically indistinguishable or sample sizes are
too small which can result in excessively large standard deviations.
Smith’s MMD statistic is readily calculated and can be used when only trait
frequency data is available. The main problems with the MMD statistic are that it is not a
Euclidean distance and it does not account for correlated traits. MMDs represent a
distance along a hyper-sphere. Cheverud et al. (1979) have demonstrated that a variety of cranial non-metric traits are significantly correlated. Failure to account for correlated
traits will significantly inflate MMD values when assessing biological relationships.
Balakrishnan and Sanghvi’s B²
Balakrishnan and Sanghvi (1968; Sanghvi and Balakrishnan, 1972) developed B²
to deal with attribute data and provide a Euclidean distance measure. This measure
represents an extension of the original G² statistic proposed by Sanghvi (1953).
Distances are calculated using the following formula:
91 2 t ( pp- ) B2 = å 12ii, 4.7 i=1 Vi
where pli refers to the ith trait frequency in the lth sample and Vi equals the weighted
variance-covariance matrix, or dispersion matrix as defined by Balakrishnan and Sanghvi
(1968). This measure is more appealing than Smith’s MMD in that it accounts for trait
correlation in the overall distance calculation. Cells of the dispersion matrix are obtained
by “taking the weighted mean of the elements of the dispersion matrices, per locus, of all
the various samples” (Constandse-Westermann, 1972:93). Various authors have
demonstrated that the B² values are often highly correlated with various arcsin-
transformed distance measures (Constandse-Westermann, 1972; Sanghvi and
Balakrishnan, 1972). The resulting distances are Euclidean distances that can be readily
manipulated within principal coordinate analysis (see Sneath and Sokal, 1973).
Mahalanobis D²
An adaptation of Mahalanobis=s (1936) generalized distance has been developed
by Blangero and Williams-Blangero (1991; see Williams-Blangero and Blangero, 1989)
for polygenic threshold traits. In this approach individual distances are calculated by
means of trait threshold values and a pooled within-group tetrachoric correlation matrix.
Biological distance is calculated by:
2 -1 d=(zi-zTj)¢ (zzij-), 4.8 ij
where (zi - zj) is a column vector of differences between threshold values for trait z at
sites i and j and T is a matrix of pooled within-group tetrachoric correlations between
92 traits. For all combinations of traits, tetrachoric correlations for pairs of traits are
computed within each group and then pooled incorporating sample size to determine the
weighted average correlation (Brown, 1977). The weighted average correlations are
combined to form the pooled tetrachoric correlation matrix. These represent the
minimum possible distance between groups (Blangero and Williams-Blangero, 1991).
This approach provides Euclidean squared distances that are easily manipulated.
No study has been conducted to compare the B², MMD and D². This method is computationally more complicated than the prior two methods, but does provide a similar distance measure commonly derived from metric studies. Small sample sizes or fixed traits are problematic in the calculation of the tetrachoric correlation matrix. An additional benefit to this approach is that an approximate F-test can be performed on individual distance measures to assess significance (e.g. Droessler, 1981; Konigsberg et al., 1993).
Wright’s FST
Wright (1943, 1951, 1978) divided the coefficient of inbreeding, or the F-statistic, into three distinct values. These values include FIS, FIT, and FST where FIS is the inbreeding coefficient of an individual relative to its own subpopulation, FIT is the inbreeding coefficient of an individual relative to the entire population, and FST is the average inbreeding of the sub-population relative to the whole population (Falconer,
1989:99), or the standardized gene frequency variance (Wright, 1969; Konigsberg and
Buikstra, 1995). FST is important in biological distance studies in that it provides a general measure of biological microdifferentiation of subpopulations. Relethford and
93 Blangero (1990, see also Konigsberg and Ousley, 1995) distinguished between a
minimum (phenotypic) and real (genetic) FST. By assuming that the phenotypic and genetic variance-covariance matrices are proportional and the relative census sizes across groups are equal, the minimum FST is proportional to the real FST given moderate to high trait heritabilities. Starting from Relethford and Blangero’s (1990) assumption of proportionality of the genetic and phenotypic covariance matrices, Konigsberg and
Ousley (1995) derived formula 4.9 for the relationship between the genetic FST and the phenotypic FST.
FSTP, FSTG, = 2 4.9 FST,,P+-hF(1)STP
For this study, minimum FST is calculated based on the approach described in
Relethford et al. (1997). First, the C matrix is derived from the distance matrix (Eq.
4.10). The C matrix is equal to an unscaled variance-covariance matrix.
C=-(I-1w¢¢)D2(I-1w)/2 4.10
where w is equal to a column vector of the proportion of the effective population size, I is
an identity matrix with the same dimensions as the distance matrix, and 1 is a vector of
ones equal in length to the number of populations.
Minimum FST estimates are then calculated from the C matrix (Eq. 4.11) where w
is again equal to a column vector of the proportion of the effective population size, and t
is the number of traits. I assume that living population sizes are equal for each site;
therefore, w is a vector with one over the number of sites for each element. FST is then:
w¢diag[C] F ST = 4.11 2t+w¢diag[C]
94 Finally, the R matrix can be calculated if needed from FST and the C matrix (Eq. 4.12)
following the procedures outlined in Relethford and Harpending (1994).
C(1-)F ST R= 4.12 2t
The diagonal of the R matrix represents biological or genetic distance to the centroid and
FST is the average weighted distance to the centroid as evident in equation 4.11. Thus, the
FST estimate provides important insight into population microdifferentiation. For example, if drift and gene flow are at equilibrium within an infinite island model, then
1 FST » 4.13 14+ Nme
where Ne is the effective population size and m is the immigration fraction. Several studies have utilized this relationship to address questions of increased heterogeneity within subpopulations (see Harpending and Ward, 1982; Relethford and Blangero, 1990;
Steadman, 1997).
Caution must be exercised in the examination of FST values. In a review of the literature, Jorde (1980) found that the variance of FST could become quite large when the number of subpopulations is low (see Nei and Chakravarti, 1977). Jorde (1980) suggests that the examination of a “large number of traits” will help reduce the error. Jorde
(1980:147) also identified three potential pitfalls to consider when comparing FST estimates. First, population size can greatly affect FST estimates as evident in Equation
4.13. Multiple small subpopulations will be influenced by the forces of isolation and founder effect. Second, the comparison of FST estimates derived from technologically and environmentally distinct populations is not recommended due to the potential poor
95 boundary definition between subpopulations. Finally, Jorde finds the “variability of
genetic sample size” as an important factor. Jorde recommends that estimation bias
correction methods be used when comparing divergent sample sizes. These methods
should account for divergent sample sizes and the number of subpopulations examined.
Methodological Approaches
The investigation of quantitative traits and population structure can be addressed
through a multitude of theoretical frameworks (see Relethford and Lees, 1982 for
review). These methodologies typically can be classified into two types of approaches,
often referred to as Model-Free and Model-Bound. Various analytical methods associated with these approaches are specifically tailored to address various research questions concerning population structure (e.g. Rudan et al., 1996, 1998). Each method requires the researcher to accept a series of assumptions ranging from multivariate normality to random environmental influences. Therefore, interpretations based on these methods need to be tempered depending on the number and extent of these assumptions.
Model-Bound Approaches
Model-bound approaches are concerned with the “estimation of parameters from direct application of a theoretical model to data” (Relethford and Blangero, 1990:6).
Types of methods used in this approach include the estimation of admixture, kinship and trait heritability (see Relethford and Lees, 1982). Admixture estimates relate to the overall rate of gene flow between two or more populations. Quantitative trait kinship estimates are based on methods similar to estimates from other data sources. Often
96 several assumptions concerning panmixia, nongenetic effects and trait heritability are
made in an effort to detect general patterns within the data. Frequently these studies
utilize the R-matrix method of Harpending and Jenkins (1973; see Harpending and Ward,
1982). Often these studies incorporate migration matrix models for the study of genetic
drift and isolation by distance (Bodmer and Cavalli-Sforza, 1968; Relethford and
Blangero, 1990; Konigsberg, 1990b).
Model-Free Approaches
Relethford and Lees (1982:116) describe the Model-free approach as “the indirect application of models of population structure in the assessment of biological differences among populations.” These authors divide Model-free approaches into two classes: (1) differentiation studies and (2) comparative studies. The goal of these approaches is to assess the biological similarity/difference of a population through the examination of among-group variation. Variation can be assessed through ANOVA methods or through the calculation of a distance statistic. This assessment is performed “free” of any specific genetic model but results from this analysis can be used to draw conclusions concerning specific genetic models and principles (Jantz 1973). A common example of a model-free
approach is discriminant function analysis, where individuals from all subpopulations are
classified based on individual attributes relative to the subpopulation means and a
common covariance matrix (Jantz and Owsley 2001). These types of analyses typically
provide some measure or assessment of dissimilarity. Distance matrices derived from
these approaches can be compared to geographic, temporal or linguistic distances.
97
Chapter 5. Previous Biodistance Research on SMA Populations
The study of population structure has maintained a secondary role relative to research on paleopathology and demography when dealing with Green River archaic skeletal samples. The Indian Knoll site has been the primary focus of most studies on the Green River samples. Recent paleopathological studies (e.g. Cassidy, 1980, 1984;
Nagy, 2000) have compared Indian Knoll with other non-hunter-gatherer populations.
Paleoanthropological researchers have utilized skeletal and dental data from the Indian
Knoll as representative of ancient North American Homo sapiens (e.g. Tompkins, 1991).
In the past decade the focus has shifted to the other sites within the drainage, but paleopathology and demography are still the key issues addressed by these studies (see
Herrmann, 1996; Mensforth, 1990; Meindl et al., 2001; Milner and Jefferies, 1998). The early population studies on the SMA sample entailed the assessment of typologies and classification of individuals into one of several different racial types or Neumann’s
(1952) classifications. Most of the early physical anthropologists working with the SMA collections were graduates of Harvard University and were at some point students of
Earnest Albert Hooton. Charles Snow and Ivar Skarland both studied at Harvard under
Hooton.
Early research was limited to Indian Knoll (15OH2; Snow, 1948) and
Chiggerville (15OH1; Skarland, 1939). This was primarily due to the United States declaring war on the Axis powers in 1941. All field investigation ceased and nearly all laboratory work ended. Several collections were never washed or re-examined until 98 decades later. For example the Carlston Annis (15BT5) skeletal collection needed to be washed by researchers at Kent State University prior to beginning their analysis in the early 1980s. These researchers were not specifically concerned with craniometric analyses. Therefore, no systematic attempt was made to reconstruct the fragmentary crania. The following is a summary of classification and population structure studies performed on the SMA samples. The discussion follows a chronological sequence from the WPA analysis through the advent of multivariate statistics and concludes with a discussion of SMA samples examined within a framework of modern quantitative genetics.
Early population structure research of SMA groups
One primary question of the WPA researchers dealt with the relationship of the
SMA sites relative to other SMA sites as well as sites from later periods. One way to address this issue was through the examination of the skeletal material. Skeletal analysis of the WPA and TVA sponsored excavations began with the Norris Dam report (Webb,
1938) in which William Funkhouser analyzed the skeletal material recovered. The goals of this research were two fold. The first was to “attempt to construct a anthropometric picture of the aborigines which inhabited this region.” The second was to compare the
Norris samples with “other groups in the Mississippi Valley” (Funkhouser, 1938:244).
Typically, these comparisons were based on measurement means and standard deviations that were calculated and compared between populations.
Skarland’s analysis of the Chiggerville data was the earliest detailed examination of a Green River collection. Skarland focused the craniometric analysis on the racial
99 classification of the Chiggerville material which was a common pursuit for researchers at that time. Skarland (1939:47) found that the Chiggerville sample represented “a long- headed, high-headed, short-faced population.” He found that they showed a close resemblance to a series of crania from the Green River housed at the Smithsonian
Institution. This series is probably Indian Knoll. Skarland felt the sample was most similar to an unpublished sample from McLean County, Kentucky as well as to the early
Pecos people. Based on the WPA excavation dates for the McLean County sites, the unpublished cranial data could be from Barrett (15McL4), Ward (15McL11) or Kirkland
(15McL12).
Marshall T. Newman and Snow (1942) compared the suite of metric data available from Indian Knoll and Chiggerville to skeletal remains excavated from
Pickwick Reservoir in Alabama. The Indian Knoll data represented the series recovered by C.B. Moore (1916) and analyzed by Hrdlicka (1927). Newman and Snow discussed the differences in size and dimensions of each group and determined that the Indian Knoll series was morphologically smaller than the Pickwick collections from Kogers Island
(1LU25) and Long Branch (1Lu67). Based on comparisons of craniometric means with
Chiggerville and Indian Knoll, Newman and Snow determined that individuals from
LU25 were outside the “Shell Mound” complex. In a short two-paragraph report in
Snow’s Indian Knoll Skeletons, Steele (1948:492) indicates that the low statistical variability present in the Indian Knoll sample was indicative of consanguineous marriages. Steele acknowledged that geographical isolation of these populations would
100 be conducive to inbreeding. Snow (1948) also attempted to identify relatives, or “twins,”
within the Indian Knoll sample based on similar cranial morphology.
Beginning with the advent of mainframe computers, greater interest was shown in the Green River collections, specifically Indian Knoll. In an attempt to test Neumann’s
(1952) classifications of Native American crania of the Eastern United States, Long
(1964, 1966) analyzed a series of crania from various sites throughout the eastern woodlands including Indian Knoll. Long’s research represents one of the earliest efforts to utilize multivariate discriminant analysis in the interpretation of prehistoric craniometric data. In his analysis, Long (1966) tested several hypotheses concerning the relationships of Neumann’s Iswanid, Walcolid, Otamid and Lenapid groups. Neumann
(1952) had defined the Iswanid group based on the Indian Knoll material. Long (1966) found that Archaic crania from Indian Knoll, Ward, Chiggerville and several other southeastern states were a discrete group with over 90% correct classification (albeit without cross validation).
In 1974, Vernon Hardy investigated the influence of size on the assessments of biological distance measures using the Green River Archaic materials. He specifically examined a sample of 56 male crania from Indian Knoll (Iswanid, as per Neumann, 1952 and Long, 1966) and 25 male crania from a series of Fort Ancient sites in Northern
Kentucky and Ohio (Muskogid, Neumann, 1952). Using a principal components analysis of 19 cranial measurements, Hardy (1974) found that the original data set could be summarized in five principal components. The first component represented variation in size and accounted for 40% of the total variation. Discriminant function analysis of the
101 five components revealed that size, reflected in the first principal component, plays a
critical role in the separation of these two samples.
Following Hardy’s research, a series of studies were conducted that focused on
population structure. Two researchers from Western Michigan University investigated
Chiggerville. Sullivan (1977) examined the demography and pathology of Chiggerville,
and Wyckoff (1977) addressed biological relationships and growth. These researchers
provided a good summary of the data collected thus far for the SMA series comparing
Chiggerville to Indian Knoll, Eva (40BN12), Adena, Ohio Archaic, Ohio Hopewell and
collections from Illinois.
Wyckoff (1977) examined craniometric and non-metric traits focusing primarily
on the metric data. Data for Chiggerville was derived from Skarland’s 1939 analysis.
Based on the large comparative sample of about 1500 crania, Wyckoff examined the product-moment correlation of the standardized measurements across all samples.
Wyckoff (1977:7) found significant correlation with Indian Knoll and Kentucky Adena.
The Chiggerville series was negatively correlated with the Ohio Hopewell, Red Ocher, and Old Copper samples. Eva exhibited a moderate (r = .3) but insignificant correlation with Chiggerville. Wyckoff indicates that Indian Knoll and Chiggerville are biologically similar, with Chiggerville being closely related to the Kentucky Adena and Eva.
Wyckoff interpreted this pattern as reflecting geographic isolation of the Archaic groups with a regional connection between Green River Archaic sites and the Kentucky Adena populations.
102 Wyckoff (1977) limits the non-metric analysis to comparisons with Indian Knoll and Adena samples. The Indian Knoll sample exhibits a significant mean divergence based on 14 non-metric traits (seven cranial and seven postcranial). The Adena sample divergence from Chiggerville is not significant based on 10 traits (four cranial and six postcranial). Wyckoff concludes that the Green River Archaic samples represented one population spanning several millennia although he lacks data from the other large Green
River middens (e.g. Carlston Annis, Read [15BT10], Ward and Barrett).
Indian Knoll has been compared both metrically and non-metrically to several sites. Lewis and Lewis (1961) compared craniometric data from Eva to Indian Knoll. In addition to the overall sample comparison, they divided the Eva sample by stratigraphic components (Eva, Three Mile and Big Sandy). Lewis and Lewis (1961:159) simply compared measurement variation between the sites and components. The pattern of variation was not consistent between males and females and the sample sizes for the individual components were quite small. Similarly, Burke (1981) examined craniometric variation from a series of sites including Indian Knoll. Burke compared the amount of variation between synchronic (generational/short term deposition sequence) and diachronic (long term deposition sequence) samples.
Wolf and Brooks (1979) compared Indian Knoll to a late Archaic period skeletal series from the Rosenberger site (15JF18). Rosenberger is a multi-component habitation midden located at the falls of the Ohio River near Louisville in Jefferson County,
Kentucky. Wolf and Brooks examined a series of 181 individuals recovered from 164 burial features. The biological distance study entailed the collection of metric (cranial
103 and post-cranial) and non-metric data and a discriminant function analysis of males and
females from Rosenberger and Indian Knoll. Wolf and Brooks (1979:945) found that
Rosenberger was more metrically homogeneous than Indian Knoll. Group classification
varied by sample ranging from 89% to 36%. A majority of the Rosenberger females
were classified as males indicating that the Rosenberger sample is probably less sexually
dimorphic than the Indian Knoll sample.
Although not specifically addressing biological distance measures, Perzigian
(1976, 1977) examined dental morphology at Indian Knoll. Initially, Perzigian (1976)
presented data on tooth dimensions and cusp number for Indian Knoll. Then in a later
publication, he examined fluctuating asymmetry in tooth size in several skeletal
populations including Indian Knoll. Perzigian (1976:113) found a moderate degree of
sexual dimorphism in tooth size and “the patterns of variability in both dimensions
[buccolingual and mesiodistal] were similar for males and females.”
Recent biological distance studies of SMA groups
In recent years, craniometric and non-metric studies have focused on Indian Knoll as a reference sample for comparisons to modern and American Paleo-Indian samples
(Powell and Rose, 1999; Powell and Neves, 1999). The Green River SMA collections have generally been ignored given that they are fragmentary. Powell and Neves (1999) employed Indian Knoll as a mid-Holocene reference sample to compare a series of North and South American Paleo-Indian crania. Approaching the data from a model-free perspective, Powell and Neves found that a majority of Paleo-Indian crania were atypical of modern and Native American populations as well as prehistoric Archaic samples, such
104 as Indian Knoll and Windover (8BR246). In the contentious debate concerning the
Kennewick skeleton, Powell and Rose conducted an analysis similar to that of Powell and
Neves (1999). They found that Kennewick was atypical of all analyzed modern Native
American populations based on a suite of 45 cranial measurements and radii. The
measurement series was then reduced to incorporate several mid-Holocene Archaic
samples and account for possible reconstruction and measurement bias. Based on this
analysis, Powell and Rose determined that Kennewick was most morphometrically
similar to the Indian Knoll sample followed consecutively by Ainu, a generalized
Amerindian sample, and then a Northeast Asia sample.
Although not specifically examining the Green River Archaic samples, both
Powell (1995) and Sciulli (Sciulli, 1990; see also Sciulli and Schneider, 1985; Tatarek
and Sciulli, 2000) have provided a baseline for the analysis of Archaic and Woodland
Period skeletal samples. Powell (1995) examined a series of mid-Holocene samples from the eastern United States ranging from Florida (Windover) to southern Illinois (Black
Earth). Employing a series of model-free and model-bound approaches, Powell addressed questions concerning the peopling of the New World through the documentation of mid-Holocene sample variability. Powell included data from two
Archaic Period sites in Tennessee, Eva and Anderson (40WM9). Powell found that as mid-Holocene populations became fixed on the landscape, gene flow increases and groups became more similar (at least based on the analysis of dental morphology).
Working with skeletal collections from Ohio, Sciulli and his colleagues (Sciulli,
1979; Sciulli, 1990; Sciulli et al., 1988; Sciulli and Schneider, 1985; Tatarek and Sciulli,
105 2000) have illustrated a consistent and informative analytical approach to dealing with
Archaic Period samples. Sciulli has documented biological relationships of the Ohio
Archaic and Woodland Period samples using craniometric, odontometric and non-metric data. Tatarek and Sciulli (2000:372) conclude that the Archaic Period in Ohio represented “3,000 years of relatively stable development, which saw overall population growth and gradual geographic cultural differentiation.” Tararek and Sciulli (2000:372) conclude the “general cultural stability allowed the pattern of isolation by distance to develop.”
Based on these prior studies of Archaic period populations, it is evident that the skeletal samples from the SMA region of the Green River should demonstrate a classic pattern of isolation by distance. However, the temporal span of the Green River samples may play a critical role in the biological distance pattern obtained from the analysis presented in the forthcoming chapters.
106
Chapter 6. Samples and Analytical Methods
In the following chapter, I detail trait selection and analytical methods employed
in the analysis of the Green River skeletal samples. The chapter is organized into
sections on sample selection, trait selection, exploratory methods, distance calculations, and comparative methods. Exploratory methods examine the pattern of trait frequencies by site and demographic parameters (i.e. age and sex). The distance section explains the
biological and geographic distance calculations. The calculation of temporal distances is
described in Chapter 3. The comparative method sections detail matrix comparison tests,
temporal trend assessment, covariance determinant ratio tests, and FST calculation.
Samples
Numerous factors influenced the selection of samples and specific traits when
considering the Green River Archaic skeletal series. Craniometric data is commonly
utilized for biological distance analysis and studies abound throughout the literature (e.g.
Jantz and Owsley, 2001; Relethford and Harpending, 1994; Steadman, 1997, 2001;
Stefan, 1999). Several of the Green River samples are extremely fragmentary and
reconstruction would have been laborious and time consuming. Dental metric and tooth
morphology data was also considered, but excessive adult occlusal and interproximal
wear associated with these Archaic Period populations would have considerably reduced
sample sizes. Therefore, cranial non-metric traits were selected for the unit of analysis 107 because all complete and fragmentary adult crania could be coded.
The sites selected include all the major Green River Archaic midden sites
excavated during the WPA from which over two hundred burials were identified. These
include Indian Knoll (15OH2), Carlston Annis (15BT5), Ward (15McL11), Barrett
(15McL4), and Read (15BT10). Chiggerville (15OH1) was also included given that a
prior biological distance study had been conducted on this sample (Wyckoff 1977) and
new field research was beginning at the site under the direction of George Crothers
(Crothers et al., 2002; Morey and Crothers, 1998). However, no radiocarbon dates are
available for Chiggerville and a moratorium on destructive analyses at the William S.
Webb Museum of Anthropology initiated in 2000 has prevented any dating of human
bone. The Eva site (40BN12), which is located in the lower Tennessee River drainage,
was also examined and serves as an out-group relative to the Green River sites. Table 6-
1 provides a summary of the sites examined and sample sizes.
In addition to the collection of non-metric data, individual age and sex estimates were re-assessed with the exception of the Carlson Annis and Ward samples. Mensforth
(1990, Meindl et al., 2001) has recently reassessed these series. Sex estimates included
male, female and unknown. Ages were classified into three groups: adolescent/young
adult, adult and old adult. Age estimates were assessed based on available pelvic age
indicators and dental wear.
Metric Data
Although not a focus of this study, craniometric data is available from three of the
sites: Indian Knoll, Chiggerville and Eva. These data were examined prior to the analysis
108
Table 6-1. Skeletal samples examined from Kentucky and Tennessee. Burials Burials Site Number Examined Analyzed Reference Carlston Annis (KY) 15BT5 209 2081 Webb (1950a); Mensforth (1990) Read (KY) 15BT10 82 82 Webb (1950b); Herrmann (1990) Barrett (KY) 15McL4 189 189 Webb and Haag (1947) Ward (KY) 15McL11 203 203 Webb and Haag (1940) Chiggerville (KY) 15OH1 61 61 Webb and Haag (1939); Wyckoff (1977) Indian Knoll (KY) 15OH2 506 5062 Moore (1916), Webb (1946), Snow (1948) Eva (TN) 40BN12 110 109 Lewis and Lewis (1961) Total 1360 1358 1 Including individuals recovered during SMAP (Marquardt and Watson, 1983) 2 Including individuals recovered by Moore (1916) and curated at the Smithsonian National Museum of Natural History.
109 of the non-metric data to provide a baseline relationship pattern. These data are from 386 individuals (primarily from Indian Knoll). The nine common craniometric variables include: glabello-occipital length [1], maximum breadth [2], minimum frontal breadth
[3], porion-apex height [6], upper facial height [21], left orbital breadth [29], left orbital
height [30], nasal height [31] and nasal breadth [32]. The bracketed number ([#])
represents the measurement number within Snow’s (1948) Indian Knoll data set. The
metric data from the three samples was analyzed in RMET 4.0. The program, written by
Dr. John Relethford, calculates the Mahalanobis distance matrix and R-matrix
(Relethford and Blangero, 1990; Relethford et al., 1997). In this analysis, the Indian
Knoll sample was divided into three samples according to depth below surface to reflect
the possible influence of the shell free midden (0-2.5 ft below surface, 2.5 ft or greater,
and individuals of unknown depth). Burials from Eva were divided into three groups
based on the stratigraphic unit from which they were recovered (Lewis and Lewis, 1961).
Non-Metric Trait Selection
Given the fragmentary nature of the Green River collection, a limited list of non- metric traits was selected (Table 6-2). A series of 24 traits was selected based on two prior studies of non-metric traits. The core traits were derived from Konigsberg et al.
(1993) and Herrmann and Adams (1996). This list was modified due to the low frequency of some traits and the probability that fragmentation would make it difficult to observe other traits. Three traits, recorded by Konigsberg et al. (1993), were excluded from this study: bregmatic bone, epipteric bone and an open foramen spinosum. These traits were highly susceptible to fragmentation or they occurred in such low frequencies
110
Table 6-2. Mid-line and bilateral traits coded in the SMA samples. Mid-line Bilateral Traits Traits Inca Bone Supraorbital Notch Post-Condylar Canal Apical Bone Suprorbital Foramen Divided Hypoglossal Canal Metopic Suture Infraorbital Suture Foramen of Huschke Sinus Flexure Multiple Infraorbital Foramina Auditory Exostosis Sagittal Ossicle Obelionic Foramen Mastoid Foramen Location Ossicle at Asterion Mastoid Foramen Number Parietal Notch Bone Assessory Lesser Palatine Foramen Lambdoidal Ossicle Foramen Ovale Open Masto-Occipital Ossicle Mylohyoid Bridge Location Mylohyoid Bridge Degree
as to be uninformative in the analysis. Although this is not an all-inclusive list of traits, a
reduced number of traits was specifically selected for two reasons: (1) to sample the traits
throughout the vault, base and facial regions and (2) to enable relatively expedient coding
of the crania. A majority of the time during data collection was devoted to fragment
sorting and bone identification.
Trait Coding and Side Selection
All 24 traits were coded according to a specific classification scheme presented in
Appendix A. Ten traits used a simple dichotomous present-absent (or open/closed)
coding system. For the remaining traits, anywhere from three to seven classifications
were employed. Observations of the supraorbital, masto-occipital foramen and
mylohyoid regions required two separate codes to account for form, location, frequency
and/or degree of expression. In the final analysis, these traits could be combined to express specific trait patterns.
111 Data was collected either directly into a Paradox database form (see Figure 6.1) or recorded on individual data forms. The direct data collection method reduced data recording time and allowed for quick review of coded burials. Age and sex information was also directly available within Paradox in separate databases that could be quickly accessed. Traits were then dichotomized based on the presence or degree of expression.
Appendix A provides the code reclassifications for converting all traits to binary values.
Following dichotomization of the traits, bilateral traits were reduced to single observations. This step was accomplished by randomly selecting one side, if both sides were observable. If only one side was observable, then this observation was used.
Age and Sex Effects
Age and sex effects were assessed with a series of univariate probit analyses. The effect of age and sex interaction is often difficult to interpret. Therefore, the approach presented by Konigsberg et al. (1993) was followed in which the probit can be written as:
P(t=1)=f[fx()] 6.1 f(x)=a++bb12(sex)()age where P(t=1) is the probability that an individual expresses a specific non-metric trait, f[f(x)] is the standardized normal integral from negative infinity to f(x), a is a constant, and the beta weights are regression on sex and age (Konigsberg et al., 1993:38). This approach is based on a modified FORTRAN code written by Lyle Konigsberg. Negative sex parameter scores indicate a trait is more frequent in females than in females. For age effects, a negative parameter score means the trait is age progressive (more common with advancing age).
112
Figure 6-1. Paradox data entry form for SMA non-metric traits.
113 In order to assess if the age and sex effects on the overall model are significant,
the beta parameters of each univariate probit are re-estimated by setting either the sex or
age parameter to zero. Negative two times the difference in the two log-likelihoods
provides a likelihood ratio value, which is asymptotically distributed as a chi-square statistic with one degree of freedom.
Distance Calculation
Individual Mahalanobis’s D² values are calculated by means of trait threshold values and a pooled within-group tetrachoric correlation matrix. The threshold values are estimated through the probit regression approach shown in Equation 6.1. This method assumes homogeneous age and sex effects across samples. Hence, a becomes a vector of threshold values for each population that has been corrected for common age and sex effects. Following Konigsberg et al. (1993:38-39), the tetrachoric correlations are
“estimated using bivariate probit analyses (see Ashford and Sowden, 1970) with separate age and sex effects for each pair of traits. The bivariate integrals for the bivariate probit analyses were evaluated using the FORTRAN function BIVNOR (Baughman, 1988).”
The threshold values and the tetrachoric correlations were calculated with a modified
FORTRAN program written by Lyle Konigsberg. For all combinations of traits, tetrachoric correlations for pairs of traits are computed within each group and then pooled incorporating sample size to determine the weighted average correlation. The weighted average correlations are combined to form the pooled tetrachoric correlation matrix. The associated bivariate probit values for all 24 traits were examined for extreme values and
114 any trait with a threshold greater than |3| was eliminated from the final threshold vectors
and tetrachoric correlation matrix.
The series of individual site threshold vectors and tetrachoric correlation matrix are then imported into “R” (Ihaka and Gentleman 1996, http://www.r-project.org/) and
the individual distances are calculated by the following equation:
2 -1 d=(zi-zTj)¢ (zzij-), 6.2 ij
where (zi - zj) is a column vector of differences between threshold values for trait z at
sites i and j and T is a matrix of pooled within-group tetrachoric correlations between
traits. These represent the minimum possible distances between groups given the
assumption of full heritability (Williams-Blangero and Blangero, 1989; Konigsberg,
1990b).
The analysis was performed on two sample configurations. The first analysis dealt with all sites individually and undivided. A second analysis examined the potential effect of the shell-free middens at Indian Knoll and Carlston Annis on the overall distance structure. Given that the shell-free midden at Indian Knoll and Carlston Annis is unevenly distributed across the unexcavated portions of two sites (Herrmann and Fenton,
2000; Stein, 1980), an arbitrary depth of 2.5ft was selected as representative of the shell free midden at Indian Knoll and Carlston Annis. This value may not exactly represent the depth of the shell-free midden across the two sites, but it does provide a general proxy for temporal differentiation within the burial samples. The shell free midden sample from Carlston Annis is very fragmentary and represents only 48 individuals. As a result a greater number of traits were eliminated during this analysis due to extreme threshold 115 values. Distances were based on 13 traits as compared to 16 traits for the full site
analysis.
Significance Testing
In order to test the significance of the threshold values and tetrachoric
correlations, individual standard errors were calculated on the univariate thresholds and
bivariate correlations for the individual site comparisons. Due to the smaller sample sizes
with the divided site comparisons only the univariate threshold values were calculated
and examined. The standard error is the square root of the diagonal of the inverse of the
information matrix (or the negative of the Hessian).
The biological distance measures were also assessed with the F ratio, which is calculated for each distance by the following equation
Tg--+p1 NN FD=××12 2 6.3 p(N12+-NTg)()
where F has p and Tgp--+1 degrees of freedom, T is the total number of crania, g is
the number of groups, p is the number of traits, and N1 and N2 are the sample sizes for the
two sites (Droessler, 1981:79; Powell, 1995). It should be noted that the F ratio
represents an estimate because most individuals are missing several observations.
Temporal and Geographic Distances
The geographic and temporal distance matrices are based on site locations and
radiometric determinations, respectively. The temporal data and matrix construction is
discussed in Chapter 3 and will not be repeated here. Geographic distances have been
derived based on straight-line inter-site distances and on river length separation. River 116 distances are readily calculated within ArcView using a site centroid theme and United
States stream coverage. The different geographic distance methods are illustrated in
Figure 6.2. The distance matrices are presented in Table 6-3 with the straight-line inter- site distances in the upper triangle and the river distances in the lower triangle.
Matrix Comparison
The influence of time and space on genetic distance is well documented in the genetic and anthropological literature. Isolation by geographic and temporal distance will increase genetic or, as is the case in this study, biological distances. If gene flow does occur, biological distances between the populations will gradually decrease. Mantel matrix permutation tests have been developed to examine correlation between temporal,
spatial and biological relationships (Manly, 1986, 1997; Mantel, 1967; Smouse et al.,
1986; Smouse and Long, 1992). Oden and Sokal (1992) provide an insightful review and
test of the three-matrix permutation tests discussed below.
Table 6-3. Geographic distance matrices in kilometers. Upper triangle is the straight- line distance and the lower triangle is the river distance. 15BT10 15BT5 15McL11 15McL4 15OH1 15OH2 40BN12 Calston Chigger- Indian Read Annis Ward Barrett ville Knoll Eva 15BT10 0.0 5.5 42.7 50.6 13.3 16.3 169.1 15BT5 16.1 0.0 39.1 47.8 13.4 15.2 172.6 15McL11 124.9 141.0 0.0 11.1 32.2 28.2 167.6 15McL4 115.6 131.7 38.5 0.0 38.7 34.9 160.2 15OH1 20.7 36.8 104.2 94.9 0.0 4.0 160.6 15OH2 26.1 42.2 98.8 89.5 5.4 0.0 161.3 40BN12 539.9 556.0 484.1 474.8 519.2 513.8 0.0
117 $Z15McL11 $Z15McL11 15McL4$Z 15McL4$Z $Z $Z15Bt5 $Z $Z15Bt5 15Oh2 $Z $Z15Bt10 15Oh2 $Z $Z15Bt10 15Oh1 15Oh1
Kentucky Kentucky
Tennessee Tennessee
40Bn12$Z 40Bn12$Z
Figure 6.2. Geographic distances measured by river miles and direct point-to-point distances. In a two-matrix comparison (Mantel, 1967), the products of all off-diagonal
combinations are summed (Z-value in equation 6.4). If the matrices exhibit similar
relationships then the Z-value will be large.
nn Z = ååABijij 6.4 ij==11
Often the matrix elements are standardized and the resulting statistic is an ordinary
product-moment correlation coefficient (as shown in equation 6.5).
stdABstd r = åå ijij 6.5 n -1
This approach has been extended to deal with multiple matrices which is referred
to as a Partial Mantel Test (Dow and Cheverud, 1985; Manly, 1986; Smouse et al., 1986).
In the approach proposed by Manly (1986), the solution is based on the use of multiple
regression. Manly (1997) describes the regression relationship of three matrices in the following equation:
Aij=b0++bb12BCijij+Îij 6.6
where b1 measures the relationship of A and B after allowing for the effects of C (please
note this is a different matrix than described in Equations 4.10 and 6.7), b2 measures the
relationship of A and C after allowing for the effects of B, and Î is an independent error.
The permutation aspect of the test results from reordering the elements of A (or the
biological distances) and calculating the correlation coefficient. In a Partial Mantel Test
as described by Smouse et al. (1986), the test statistic is calculated from a series of matrix
residuals (A' and B') derived from a regressing the two matrices (A and B) on a third
119 matrix (C). The statistic assesses the relationship of A and B conditional on C. Dow and
Cheverud (1985) propose a comparison of the correlation matrices of A and C to B and C
(i.e. corr(A,C) equals corr(B,C) as described by Oden and Sokal, 1992). Oden and
Sokal (1992) tested the three methods described above to examine the effect of spatial autocorrelation on the results. They found that all three methods were susceptible to error to varying degrees. Oden and Sokal recommend the Smouse et al. (1986) method with reservations.
For this study, all two-way and three-way comparisons concerning biological distance geography and time will be conducted. Both geographic distance matrices will be examined. The “Matrix Comparison Plot” package of NTSYSpc 2.1 provides a simple interface. Each test will utilize 9999 permutations and the Mantel statistic will be normalized. The package follows the approach proposed by Smouse et al. (1986) in that matrix residuals are calculated and compared. In addition, plots of the matrix residuals are generated within NTSYSpc 2.1 which allows one to examine the potential effect of individual sites on the permutation test.
Temporal Trend
In an effort to better explain the biological distance structure of the divided site matrix relative to the geographic and temporal matrices, the presence of temporal trend in trait frequencies needs to be examined. A temporal trend in trait frequencies may produce temporal autocorrelation that might mask the effect of isolation by distance examined through the three matrix comparisons. Konigsberg (1990b:57) states that “if an isolation by distance model does apply in a given region, then the expectation is that
120 genetic and spatial distance should be positively correlated and genetic and temporal
distance negatively correlated.” If the temporal correlation within a region is positive
then one possible explanation is a temporal trend in trait frequency.
Konigsberg (1990a) presents a simple method for ascertaining the extent of
temporal trend through the examination of principal components derived from the pooled
tetrachoric correlation matrix and threshold values. Principal component scores for each
site are calculated by
C= ZVL-1/2, 6.7
where C is the matrix of principal component scores (please note this is a different matrix than described in Equations 4.10 and 6.6), Z is a matrix of threshold values, V is a matrix of eigenvectors of the pooled within-group tetrachoric correlation matrix, and L is a diagonal matrix of the eigenvalues of the tetrachoric correlation matrix. The C and Z matrix are n x k dimensions with n being the number of sites and k equal to the number of traits.
Next, the Pearson product-moment correlation of each principal component with the weighted mean date for each site is calculated. I use the program CORRPERM
(Legendre 2000) to derive these correlations. The program was written by Pierre
Legendre and can be downloaded from the following address:
(http://www.fas.umontreal.ca/BIOL/Casgrain/en/labo/index.html). In addition to the
actual correlation, the program calculates correlations for a set of 999 random
permutations of the vector of time by site with each principal component. These
permutations are used to find the probability of obtaining a time and principal component
121 correlation equal to or more extreme than that observed (Legendre and Legendre 1998;
Konigsberg 1990a). Principal components with permuted probability values less than .1
are then removed from the matrix and a new distance matrix is calculated using the
reduced set of principal components and eigenvalues. The new distance matrix is then
used in the three-matrix Mantel tests of geographic and temporal correlation. Once the
temporal trend has been removed, the new matrix should provide a more accurate
correlation value for the temporal and spatial comparison.
Covariance Matrix Determinant Comparisons
In two influential papers Lane and Sublett (1972) and Spence (1974) describe
how postmarital residence patterns could be ascertained from skeletal populations
through the examination of male and female morphological variation. They concluded
that the sex with a higher migration rate would exhibit greater within-group variation and lower between-group variation. Biological distance measures by sex were used to assess the level of between-group variation. Numerous authors have pointed out that this interpretation is problematic (Kennedy, 1981; Konigsberg, 1987). The hypothesis set forth by Lane and Sublett (1972) is static in that it holds true for only the first generation of migration. Konigsberg (1988; Konigsberg and Buikstra, 1995) has demonstrated through simulations that the pattern of between and within-group variation breaks down after several generations of migration. The pattern of migration and amount of migration also influences between and within-group variation.
Konigsberg (1988) has shown that Lane and Sublett’s and Spence’s conclusions were incorrect, but they were working in the right direction. Konigsberg demonstrated
122 that they should have focused on within-group variation rather than between-group variation. Through an examination of male and female additive genetic variances,
Konigsberg (1988:476-477) shows that the sex with lower mobility within a mating network will have lower within-group trait variation relative to the more mobile sex.
To test these hypotheses, a method proposed by Konigsberg (1988) was utilized
whereby the ratio of the determinant of the male and female covariance matrix for a
select group of traits is calculated (|CM| / |CF|). Each covariance matrix was formed separately by site and sex. Therefore a determinant ratio equal to one would indicate equal mobility for males and females.
For this analysis, traits exhibiting some degree of variation were initially selected.
Covariance matrices could not be calculated for samples with fixed or invariant traits.
Next, traits with significant age and sex effects were eliminated. The final trait list
consisted of five traits: apical bone, obelionic foramen, ossicle at Asterion, parietal notch bone, and mylohyoid bridge without hiatus. Covariance matrices were calculated using the “pairwise.complete.obs” option in “R” (Ihaka and Gentleman, 1996) to allow for missing data. The covariance between each pair of variables is computed using all complete pairs of observations. Determinants of the male and female covariance matrices are calculated and the log determinants subtracted (log|CM| - log|CF|). Therefore, a log determinant ratio equal to zero would indicate equal mobility for males and females. A negative ratio would suggest greater female mobility within the network and a positive ratio indicates males move more frequently. Ratios for all sites and identified stratigraphic units were calculated and compared. To test for significance of the log
123 determinant ratios, a bootstrap approach was employed. Each male and female sample
was permuted until 999 determinant ratios were calculated. Standard 95% bootstrap
confidence intervals were calculated based on the standard deviation and mean of the 999
permuted determinant ratios (Manly, 1997). If zero fell within the 95% confidence
interval then the determinant ratio is considered insignificant. To visualize the
determinate ratios, density plots were generated by site. In order to obtain a probability
estimate for these tests, a one-tailed z-score value was calculated for the observed determinate ratio and mean bootstrap value. In order to provide a conservative probability estimate, the determinant ratio or mean bootstrap value closest to zero was
selected for this assessment. Finally, I combined the probabilities for these various full
sample bootstrap tests using the following formula: -2åln P , where P represents individual probability tests (Sokal and Rolf, 1981:779-782). The sum of the probabilities is compared to a c 2 distribution with 2k degrees of freedom.
R Matrix and Fst Statistic Calculation
Following the methods of Relethford and Blangero (1990; Relethford, 1994;
Relethford et al., 1997) detailed in Chapter 4 (equations 4.10-4.12), the R-matrix and unbiased FST values are calculated from the estimated biological distance matrices. The values along the R matrix diagonal represent the biological distance to the regional centroid. FST is the average weighted distance to the centroid and is interpreted as a useful index of microdifferentiation. These estimates will be compared with those from other hunter-gatherer populations reported in Jorde (1980), other Native North American
124 samples (Steadman, 1997; Herrmann and Adams, 1996), and FST estimates from the metric data.
125
Chapter 7. Results
In the following chapter, the results of the analysis are presented. These results are divided between basic descriptive statistics of the sample and the specific analyses.
The result from full samples and stratigraphically divided samples are presented and compared to geographic and temporal matrices. The analyses of temporal trends within the Green River samples and within-site trait variation by sex are then detailed. Finally, the implication of these results to the Green River Archaic populations and regional Shell
Mound Archaic research are discussed.
Sample and Trait Statistics
The sample analyzed for this study consisted of 1358 individuals. All individuals estimated to be over 15 years old are included. Basic trait statistics are presented by site in Tables 7-1 through 7-7. Trait frequencies are invariant in six instances within the samples. Three of these cases occur within the Chiggerville sample. Two samples exhibit no divided foramen ovale. Two sites show no evidence of metopic sutures. One sample lacks the flexure of sagittal sinus to the left, and one sample has no sagittal ossicles. Metopic sutures, inca bones and sagittal ossicles are infrequent in all samples.
The Indian Knoll sample is in excellent condition. A large percentage of the
Indian Knoll collection is complete or has been reconstructed. Madeline Kneberg had reconstructed most of the Eva crania, but several have since broken apart. Skarland
(1939) reconstructed some of the Chiggerville crania but most were fragmentary. The
126
Table 7-1. Trait frequencies for Read.
Observed
Trait Frequency Trait Present Site N Code Trait Description Unobservable 15BT10 82 INCA Inca Bone 55 3 27 0.05 Read ME Metopic Suture 60 1 22 0.02 AB Apical Bone 40 12 42 0.30 SO Saggital Ossicle 28 1 54 0.04 FX Sinus Flexure 53 11 29 0.21 SN Supraorbital Notch 64 42 18 0.66 SF Suprorbital Foramen 66 22 16 0.33 IS Infraorbital Suture 20 9 62 0.45 IFM Infraorbital Foramen 24 3 58 0.13 OF Obelionic Foramen 50 37 32 0.74 AST Ossicle at Asterion 41 6 41 0.15 PN Parietal Notch Bone 48 4 34 0.08 LO Lambdoidal Ossicle 40 20 42 0.50 MO Mastooccipital Ossicle 34 10 48 0.29 PC Post-condylar Canal 34 30 48 0.88 DH Divided Hypoglossal Canal 47 6 35 0.13 FH Foramen Huschke 60 8 22 0.13 AD Auditory Exostosis 70 15 12 0.21 MF Mastoid Foramen Location 60 41 22 0.68 MN Mastoid Foramen Number 60 14 22 0.23 Assessory Lesser Palatine AP Foramen 13 12 69 0.92 FO Foramen Ovale Open 18 0 64 0.00 Mylohyoid Bridge without MB hiatus 68 9 14 0.13 Mylohyoid Bridging MD Present 68 25 14 0.37
127
Table 7-2. Trait frequencies for Carlston Annis.
Trait Trait Site N Code Observed Present Unobservable Frequency 15BT5 208 INCA 172 2 36 0.01 Carlston ME 173 0 35 0.00 Annis AB 135 27 73 0.20 SO 97 4 111 0.04 FX 167 37 41 0.22 SN 176 104 32 0.59 SF 178 74 30 0.42 IS 79 39 129 0.49 IFM 75 12 133 0.16 OF 149 97 59 0.65 AST 146 18 62 0.12 PN 154 21 54 0.14 LO 145 71 63 0.49 MO 126 38 82 0.30 PC 123 114 85 0.93 DH 144 28 64 0.19 FH 181 49 27 0.27 AD 184 49 24 0.27 MF 165 95 43 0.58 MN 169 32 39 0.19 AP 81 46 127 0.57 FO 93 2 115 0.02 MB 164 29 44 0.18 MD 164 39 44 0.24
128
Table 7-3. Trait frequencies for Ward.
Trait Trait Site N Code Observed Present Unobservable Frequency 15McL11 203 INCA 174 7 29 0.04 Ward ME 176 1 27 0.01 AB 141 30 62 0.21 SO 112 3 91 0.03 FX 159 38 44 0.24 SN 171 112 32 0.65 SF 169 80 34 0.47 IS 59 34 144 0.58 IFM 68 8 135 0.12 OF 145 92 58 0.63 AST 147 24 56 0.16 PN 157 25 46 0.16 LO 144 71 59 0.49 MO 116 40 87 0.34 PC 99 93 104 0.94 DH 113 16 90 0.14 FH 170 59 33 0.35 AD 175 27 28 0.15 MF 155 110 48 0.71 MN 166 39 37 0.23 AP 54 39 149 0.72 FO 81 3 122 0.04 MB 157 40 46 0.25 MD 157 66 46 0.42
129
Table 7-4. Trait frequencies for Barrett.
Trait Trait Site N Code Observed Present Unobservable Frequency 15McL4 189 INCA 147 3 42 0.02 Barrett ME 159 1 30 0.01 AB 118 35 71 0.30 SO 73 5 116 0.07 FX 139 30 50 0.22 SN 161 92 28 0.57 SF 158 70 31 0.44 IS 68 38 121 0.56 IFM 63 10 126 0.16 OF 142 102 47 0.72 AST 137 28 52 0.20 PN 146 27 43 0.18 LO 122 75 67 0.61 MO 114 45 75 0.39 PC 104 95 85 0.91 DH 111 15 78 0.14 FH 164 46 25 0.28 AD 170 31 19 0.18 MF 153 102 36 0.67 MN 155 37 34 0.24 AP 67 52 122 0.78 FO 71 1 118 0.01 MB 147 42 42 0.29 MD 147 53 42 0.36
130
Table 7-5. Trait frequencies for Chiggerville.
Trait Trait Site N Code Observed Present Unobservable Frequency 15OH1 61 INCA 49 2 12 0.04 Chiggerville ME 51 0 10 0.00 AB 36 5 25 0.14 SO 28 0 33 0.00 FX 49 0 12 0.00 SN 54 34 7 0.63 SF 55 21 6 0.38 IS 17 9 44 0.53 IFM 18 0 43 0.00 OF 44 31 17 0.70 AST 39 4 22 0.10 PN 47 7 14 0.15 LO 40 10 21 0.25 MO 40 9 21 0.23 PC 23 20 38 0.87 DH 24 5 37 0.21 FH 54 9 7 0.17 AD 55 15 6 0.27 MF 51 29 10 0.57 MN 51 17 10 0.33 AP 13 10 48 0.77 FO 26 1 35 0.04 MB 52 10 9 0.19 MD 52 10 9 0.19
131
Table 7-6. Trait frequencies for Indian Knoll.
Trait Trait Site N Code Observed Present Unobservable Frequency 15OH2 506 INCA 486 10 20 0.02 Indian ME 488 3 18 0.01 Knoll AB 460 101 46 0.22 SO 389 18 117 0.05 FX 475 113 31 0.24 SN 482 301 24 0.62 SF 490 215 16 0.44 IS 385 185 121 0.48 IFM 372 46 134 0.12 OF 471 250 35 0.53 AST 487 84 19 0.17 PN 481 75 25 0.16 LO 453 218 53 0.48 MO 452 99 54 0.22 PC 430 389 76 0.90 DH 449 60 57 0.13 FH 493 135 13 0.27 AD 495 165 11 0.33 MF 462 327 44 0.71 MN 480 137 26 0.29 AP 394 232 112 0.59 FO 356 13 150 0.04 MB 484 85 22 0.18 MD 484 156 22 0.32
132
Table 7-7. Trait frequencies for Eva.
Trait Trait Site N Code Observed Present Unobservable Frequency 40BN12 109 INCA 92 5 17 0.05 Eva ME 97 1 12 0.01 AB 73 24 36 0.33 SO 73 5 36 0.07 FX 88 28 21 0.32 SN 98 51 11 0.52 SF 96 44 13 0.46 IS 56 40 53 0.71 IFM 48 4 61 0.08 OF 87 61 23 0.70 AST 85 6 24 0.07 PN 92 16 17 0.17 LO 89 50 20 0.56 MO 73 26 36 0.36 PC 56 53 54 0.95 DH 76 12 33 0.16 FH 99 30 10 0.30 AD 104 19 5 0.18 MF 82 56 27 0.68 MN 86 24 23 0.28 AP 43 21 66 0.49 FO 54 0 55 0.00 MB 93 18 16 0.19 MD 93 25 16 0.27
133 remaining four samples (Carlston Annis, Read, Barrett and Ward) have never been
reconstructed and are highly fragmented.
The extent of fragmentation varies by sample. Figure 7.1 depicts the pattern of
observable trait locations by site after the data has been reduced to single observations for
bilateral traits. Read (15BT10) is the most fragmentary sample and Indian Knoll is the
most complete sample. This pattern is substantiated when the 24 observed trait
frequencies are averaged by site (Table 7-8). Observable trait locations are 32% lower at
Read compared to Indian Knoll (OH2) and 9% less than the next lowest site, Chiggerville
(OH1). The pattern at Read is expected given that approximately 100 burials were not recovered due to poor preservation. The Read site was located on a bluff top, and the area had been cultivated for many years prior to excavation. Plowing and sheet wash greatly reduced midden depth and contributed to the poor preservation of the Read material.
Trait location frequencies are presented in Table 7-9. The least frequent observable trait locations are accessory lesser palatine foramen (AP), multiple infraorbital foramen (IFM), infraorbital suture (IS), open foramen ovale (FO), and sagittal ossicle
(SO). Individual site observation frequencies dipped as low as .18 to .25 for some traits whereas the frequency at Indian Knoll never dips below .7. Most frequencies are greater than .85 (Figure 7-1).
Age and Sex Effects
The assessment of sex and age effects was performed on the full site samples
(Table 7-10). For sex effects, negative values indicate the traits are more common in
134 Table 7-8. Comparison of the averages observed trait frequency by site. Average of Observed Trait Site Name Frequency 15BT10 Read 0.570 15OH1 Chiggerville 0.661 15McL4 Barrett 0.669 15McL11 Ward 0.670 15BT5 Carlston Annis 0.689 40BN12 Eva 0.739 15OH2 Indian Knoll 0.897
Table 7-9. Observed trait frequencies for the entire sample (n = 1358). Trait # Trait Number Frequency of Missing Observed traits 1 INCA 183 0.865 2 ME 154 0.887 3 AB 355 0.739 4 SO 558 0.589 5 FX 228 0.832 6 SN 152 0.888 7 SF 146 0.892 8 IS 674 0.504 9 IFM 690 0.492 10 OF 271 0.800 11 AST 276 0.797 12 PN 233 0.828 13 LO 325 0.761 14 MO 403 0.703 15 PC 489 0.640 16 DH 394 0.710 17 FH 137 0.899 18 AD 105 0.923 19 MF 230 0.831 20 MN 191 0.859 21 AP 693 0.490 22 FO 659 0.515 23 MB 193 0.858 24 MD 193 0.858
135 1
0.8
0.6
0.4 Observation Frequency 0.2
0 INCA ME AB SO FX SN SF IS IFM OF AST PN LO MO PC DH FH AD MF MN AP FO MB-MD
15BT10 15BT5 15McL11 15McL4 15OH1 15OH2 6BN12
Figure 7-1. Frequency of observable traits by site.
136
Table 7-10. Age and sex effects as determined through the univariate probit tests. Trait Trait Sex Age Number Code Trait Description LnLK Estimate Estimate 1 INCA Inca Bone -119.330 0.337 -0.100 2 ME Metopic Suture -28.093 0.488 0.306 3 AB Apical Bone -495.269 -0.079 0.032 4 SO Saggital Ossicle -131.865 0.124 -0.012 5 FX Sinus Flexure -541.393 -0.132 0.038 6 SN Supraorbital Notch -739.004 0.227 -0.001 7 SF Suprorbital Foramen -763.952 -0.163 -0.012 8 IS Infraorbital Suture -424.470 -0.324 0.212 9 IFM Infraorbital Foramen -229.427 0.119 0.001 10 OF Obelionic Foramen -663.515 0.082 -0.017 11 AST Ossicle at Asterion -431.959 0.018 0.022 12 PN Parietal Notch Bone -454.068 -0.132 0.050 13 LO Lambdoidal Ossicle -654.941 0.205 0.022 14 MO Mastooccipital Ossicle -508.094 -0.273 0.154 15 PC Post-condylar Canal -236.676 -0.019 -0.183 16 DH Divided Hypoglossal Canal -368.515 -0.124 0.013 17 FH Foramen Huschke -639.029 -0.379 0.003 18 AD Auditory Exostosis -584.846 0.889 -0.047 19 MF Mastoid Foramen Location -633.126 0.474 -0.057 20 MN Mastoid Foramen Number -593.894 0.468 -0.087 Assessory Lesser Palatine 21 AP Foramen -400.564 0.314 0.015 22 FO Foramen Ovale Open -81.925 -0.336 0.048 Mylohyoid Bridge without 23 MB hiatus -539.601 -0.034 -0.126 24 MD Mylohyoid Bridging Present -679.570 -0.005 -0.068 Significance shading: .1>p>.05 p<.05
137 females. For age effects, negative values indicate the trait is age progressive (more
common with advancing age). Ten of 24 traits exhibited significant sex effects with p-
values less than .05. Two additional traits exhibited sex effects between p-values of .1
and .05. Five of the twelve significant sex-associated traits are more common in females.
Female linked traits include foramen of Huschke, infraorbital suture, supraorbital foramen, mastooccipital ossicle and foramen ovale open. Most female linked traits are
hypostotic in origin or associated with foramina. The male linked traits are associated
with more sexually dimorphic regions of the cranium including the supraorbital notch,
mastoid process, and auditory exostosis.
In contrast to the sex effects, age effects are present in only two traits (infraorbital
suture and mastooccipital ossicle). Both traits are age regressive indicating that they are
more common in younger individuals. The lack of age effects in some traits is
interesting. Commonly, age effects are noted in developmental or sutural traits, such as
foramen of Huschke and various sutural ossicles. The lack of age effects may relate to
the relatively low survivorship of older adults (Herrmann, 1996; Herrmann and
Konigsberg, 2002). The high percentage of traits with significant sex effects is consistent
with conclusions by Hamilton (1975) and Allaway (1980) suggesting marked sexual
dimorphism within the Green River Archaic samples. Saunders (1989) suggests that a
high number of sex-associated traits cannot be explained strictly by differences in body
size. Other factors such as growth patterns (Richtsmeier et al., 1984) possibly contribute
to the high number of sex-associated traits. To utilize highly associated traits within the
138 distance analysis, tetrachoric correlations and threshold values are adjusted for common
age and sex effects.
Distance Structure Analysis
The biological distance analysis is divided into numerous sections to explain
large-scale regional comparisons, temporal influences, drainage specific issues and
sexual variation. Initially, I examine the limited metric data to elucidate a general
distance relationship based on the three available metric data sets from Indian Knoll, Eva
and Chiggerville. Next, the full sample including Eva is analyzed and discussed relative
to geographic and temporal relationships. The samples are then divided based on
available stratigraphic information and distances calculated. The effects of temporal
trend on the full and divided data sets are assessed and new distance matrices are
calculated based on the reduced principal components. The pattern of trait variation by
sex is then examined through covariance determinant ratios and bootstrap tests. Finally,
the results of these analyses are discussed in reference to the Green River Archaic
populations to make inferences concerning biological relationships and proposed
archaeological models.
Metric Assessment
The results from the R-met analysis of craniometric data available from Eva,
Indian Knoll and Chiggerville indicate good separation of the sites. Indian Knoll was
divided into three subsamples based on stratigraphic association. The subsamples
included individuals from the shell-free midden, individuals from the shell midden or
139 from subsoil pits and individuals of unknown provenience. The Eva data was divided
into three samples based on the stratigraphic zones identified by Lewis and Lewis (1961).
A plot of the first two eigenvectors clearly shows that the Chiggerville sample is approximately equidistant from Eva and the Indian Knoll samples on the first eigenvector
(Figure 7-2). Chiggerville is quite distinct from Indian Knoll and Eva on the second eigenvector. The three Indian Knoll samples cluster together as do the three stratigraphically defined Eva groups. This limited analysis suggests that the Green River samples and Eva are morphologically distinct. Also, substantial variation is present
within the Green River samples at least based on the limited craniometric data from
Chiggerville and Indian Knoll.
All Sites Non-metric Analysis
The initial analysis examines the relationships of full undivided samples. Seven
of the 24 traits produced extreme threshold values during the univariate and bivariate
calculations and these were excluded from the analysis. In addition, traits 23 and 24
exhibited an extremely high correlation (.99). Thus, only trait 23 was utilized in the
distance calculation. A total of sixteen traits were used in the analysis. The threshold
values range between -1.57 to 1.39 (Table 7-11) and correlations range from -.80 to .50
(Table 7-12). Distances range from .59 to 2.38 (Table 7-13). The smallest biological distance is between Ward (McL11) and Barrett (McL4). The greatest biological distance is between Chiggerville (OH1) and Eva (BN12). Following the method presented in
Droessler (1981) the distances are all significant at the p < .001. The FST calculated for the full sample distance matrix assuming equal effective population sizes equals .0194
140
0.25
OH2-UK BN12-Eva 0.06 BN12-BS OH2u OH2l BN12-TM
EV2 -0.13
-0.31
OH1
-0.50 -0.50 -0.25 0.00 0.25 0.50 EV1
Figure 7-2. Plot of the first two eigenvectors derived from the R-met analysis of nine measurements available from Indian Knoll (n=344, Snow 1948), Chiggerville (“OH1”, n=11, Skarland 1939), and Eva (“BN12”, Components: Eva=4, Three Mile [TM]=20, Big Sandy [BS]=7, Lewis and Lewis 1961). These two eigenvalues are derived from the biased R-matrix and account for 88% of the variation. The Indian Knoll sample is divided by depth. OH2-1 represent the upper 2.5 feet of the midden. OH2-2 represents individuals 2.5 feet or more below surface. OH2-UK represent individuals of unknown provenience.
141
Table 7-11. Threshold values for each site derived from the analysis of the entire sample. Trait numbers follow Table 7-10. Trait BT10 BT5 McL11 McL4 OH1 OH2 BN12 3 0.656 1.063 0.860 0.570 1.134 0.809 0.656 6 -0.489 -0.402 -0.477 -0.274 -0.474 -0.395 -0.489 7 0.395 0.312 0.111 0.227 0.402 0.205 0.395 8 0.446 0.337 -0.104 -0.054 0.009 0.204 0.446 10 -0.574 -0.431 -0.361 -0.585 -0.542 -0.127 -0.574 11 1.117 1.184 1.009 0.813 1.261 0.933 1.117 12 1.386 1.162 1.104 0.952 1.104 1.085 1.386 13 -0.101 -0.101 -0.105 -0.365 0.519 -0.064 -0.101 14 0.662 0.631 0.529 0.368 0.862 0.958 0.662 15 -1.128 -1.477 -1.568 -1.354 -1.283 -1.295 -1.128 16 1.221 0.908 1.075 1.122 0.988 1.150 1.221 17 1.296 0.871 0.626 0.750 1.180 0.802 1.296 18 0.299 0.181 0.635 0.696 0.145 0.046 0.299 19 -0.743 -0.397 -0.799 -0.697 -0.413 -0.756 -0.743 20 0.616 0.660 0.554 0.500 0.240 0.375 0.616 23 1.127 0.937 0.760 0.592 0.909 0.936 1.127
142
Table 7-12. Pooled tetrachoric correlation matrix derived for the full sample analysis. Trait numbers follow Table 7.10. Traits 3 6 7 8 10 11 12 13 14 15 16 17 18 19 20 23 1.000 -0.031 1.000 -0.012 -0.804 1.000 0.008 -0.001 0.083 1.000 -0.015 0.062 -0.077 0.014 1.000 0.316 0.002 0.028 -0.119 -0.052 1.000 0.148 0.051 0.068 0.233 0.020 0.271 1.000 0.497 0.063 -0.007 0.019 0.069 0.195 0.320 1.000 0.136 0.010 -0.016 -0.064 0.088 0.262 0.187 0.201 1.000 0.176 0.089 -0.079 0.049 -0.076 0.034 0.015 0.015 0.155 1.000 0.066 -0.021 -0.077 -0.021 0.039 0.016 0.059 0.133 -0.010 0.092 1.000 0.010 0.046 0.030 0.010 0.022 -0.088 0.144 0.115 0.007 -0.048 -0.056 1.000 0.014 0.037 -0.034 -0.082 -0.019 -0.016 -0.138 0.030 -0.131 0.123 0.096 -0.183 1.000 -0.089 -0.050 0.006 -0.035 -0.019 -0.148 -0.087 -0.057 -0.244 0.005 0.090 -0.012 0.107 1.000 -0.021 -0.002 -0.022 -0.062 -0.002 -0.108 -0.056 0.023 0.040 -0.022 0.110 0.069 0.032 0.488 1.000 0.037 -0.097 0.053 -0.115 -0.032 -0.045 -0.221 -0.041 -0.146 0.037 -0.025 -0.003 0.051 0.104 0.015 1.000
143
Table 7-13. Distance matrix calculated for the full sample analysis. Sample sizes are in parentheses. BT10 BT5 McL11 McL4 OH1 OH2 BN12 (82) (208) (203) (189) (61) (506) (109) BT10 0.000 BT5 1.102 0.000 McL11 1.587 1.073 0.000 McL4 1.481 1.236 0.592 0.000 OH1 1.853 1.182 2.213 2.203 0.000 OH2 1.113 0.697 1.056 1.294 1.217 0.000 BN12 1.979 1.963 1.345 1.068 2.381 1.892 0.000
144 (standard error (s.e.) = .0017). This estimate assumes full heritability (h²=1). When
adjusted for approximate heritability estimates for non-metric traits ( h2 » .35 ), then the
FST estimate would equal .0535 using equation 4.9 based on Konigsberg and Ousley
(1995).
A plot of the first two principal coordinates is presented in Figure 7-3. The
configuration roughly corresponds to geography with the western sites to the left side of
the plot and the eastern sites to the right. Chiggerville (15OH1) and Read (15BT10) do
conform to this general pattern but are quite divergent on the second component. These
sites also represent the two smallest samples. Chiggerville is well separated from the
other Green River sites. Read, on the other hand, is separated but falls near Indian Knoll
(OH2) and Carlston Annis (BT5). This pattern may reflect error due to small sample
sizes and missing data. The distance structure seems to support Rolingson and
Schwartz’s (1966; see also Nance 1987) contention of a marked division between the
Tennessee/Cumberland and Tradewater/Green drainages. This distinction appears more
gradual from a biological perspective as compared to the distinct division suggested by
the material culture. Data from additional sites within the Tennessee, Cumberland and
Tradewater drainages such as Parrish Village, Cherry and Ledbetter Landing would aid in
this analysis.
To examine the relationship of biological distance, mean site date and geography,
a series of eight Mantel matrix comparison tests were performed (Table 7-14). The first
four tests employ the full distance structure. These tests indicate a strong geographic and
temporal correlation. Three matrix comparisons using the temporal and geographic
145
0.80 Bt10
0.40 McL11 Bt5 Oh2 McL4 Dim-2 0.00 +
-0.40
Bn12 Oh1 -0.80 -1.20 -0.52 0.15 0.83 1.50 Dim-1
Figure 7-3. Principal coordinate plot of the full sample relationships. The ‘+’marks the centriod of the two-dimensional coordinate space.
146 matrices relative to distance matrix produce slightly lower, insignificant correlations.
The high temporal correlation is problematic. Within an infinite island model of isolation by temporal and spatial distance, biological distances when controlling for geography should be negatively correlated with time within a mating network (Konigsberg 1990b).
Because Eva is located furthest from the other sites, it was dropped from the distance structure. Two three-matrix correlation tests were run looking solely at the Green River middens. The distance structure needed to be recalculated for this comparison excluding
Eva from the threshold and correlation matrix calculation. The reduced matrix correlations are lower and less significant. However, the geographic and temporal matrices are highly correlated (r = .599; p = .010).
One possible explanation for the divergent and inconsistent distance pattern relative to time and space may be due to the temporal depth of the sites examined. These mean site dates span a period of 1700 years, and the overall span is much greater if the
Table 7-14. Mantel matrix comparisons of the full sample biological distance matrix to spatial and temporal distances. Matrix Comparison r p-value Direct Geography * Time 0.243 0.149 River Geography * Time 0.234 0.123 Bio * Direct Geography 0.461 0.106 Bio * River Geography 0.467 0.091 Bio * Time 0.497 0.045 Bio * Direct Geography (Time) 0.404 0.205 Bio * Time (Direct Geography) 0.447 0.101 Bio * Direct Geography (Time) – without Eva 0.215 0.281 Bio * Time (Direct Geography) – without Eva 0.205 0.222
147 individual radiocarbon dates are used. Indian Knoll, Carlson Annis and Eva all represent stratified sites that span several millennia. Therefore, the full sample analysis may not be parsing out the biological differences within larger sites which represent numerous generations and possibly multiple populations. To address this concern, the sample was divided into stratified samples where possible (Indian Knoll, Carlston Annis and Eva) and a second distance analysis was performed (see Divided Sites Analysis). Another possible explanation for the matrix correlation pattern seen above is that the traits are being influenced by a temporal trend in trait frequency. This issue will be addressed in the Temporal Trend section below.
Divided Sites Analysis
Following a similar approach as the full sample analysis, the univariate and bivariate thresholds as well as tetrachoric correlations were calculated. All traits with extreme threshold values were eliminated from the analysis. Smaller sample sizes associated with the divided sites resulted in a greater number of extreme threshold values.
Eleven traits were eliminated from the analysis and the distance matrix was then calculated based on thirteen traits. The samples from the upper stratigraphic zones from
Carlston Annis and Eva were fragmented and quite small consisting of 48 and 35 individuals respectively. Threshold values range from -.91 to 1.49 (Table 7-15) and tetrachoric correlations range from -.80 to .49 (Table 7-16). The correlations are similar to the full sample analysis. Distances range from .29 to 2.64 (Table 7-17).
The smallest biological distance is between the upper zone at Indian Knoll (OH2) and the lower zone from Indian Knoll. The greatest biological distance is between
148
Table 7-15. Threshold values for the divided site analysis using 13 traits. Divisions are designated by u and l identifying upper and lower stratigraphic unit, respectively. Trait BT10 BT5u BT5l McL11 McL4 OH1 OH2u OH2l BN12u BN12l 6 -0.497 -0.279 -0.443 -0.493 -0.285 -0.487 -0.519 -0.382 -0.135 -0.186 7 0.401 0.323 0.315 0.126 0.238 0.414 0.307 0.200 0.048 0.127 10 -0.563 -0.391 -0.426 -0.352 -0.576 -0.529 -0.266 -0.083 -0.906 -0.359 11 1.097 0.876 1.256 0.993 0.800 1.240 1.010 0.902 1.330 1.486 12 1.366 0.824 1.242 1.078 0.931 1.079 0.978 1.086 0.792 1.067 13 -0.112 -0.068 -0.127 -0.116 -0.375 0.505 -0.221 -0.040 -0.157 -0.292 14 0.655 0.375 0.691 0.524 0.363 0.855 1.008 0.936 0.471 0.514 16 1.225 0.743 0.961 1.084 1.129 0.999 1.326 1.148 1.331 1.116 17 1.294 1.196 0.797 0.620 0.746 1.175 0.733 0.820 1.063 0.647 18 0.299 0.016 0.232 0.636 0.697 0.146 0.054 0.093 0.322 0.863 19 -0.742 -0.653 -0.329 -0.796 -0.695 -0.409 -0.840 -0.688 -0.825 -0.825 20 0.615 0.949 0.587 0.555 0.498 0.238 0.285 0.406 0.352 0.416 23 1.141 1.160 0.893 0.772 0.603 0.926 1.206 0.891 0.793 1.236
149
Table 7-16. Pooled tetrachoric correlation matrix for the divided site analysis. Trait 6 7 10 11 12 13 14 16 17 18 19 20 23 1.000 -0.799 1.000 0.051 -0.069 1.000 0.017 0.010 -0.050 1.000 0.048 0.080 0.013 0.251 1.000 0.061 -0.007 0.072 0.210 0.301 1.000 0.010 -0.007 0.077 0.269 0.180 0.193 1.000 -0.033 -0.089 0.030 -0.022 0.045 0.112 -0.030 1.000 0.035 0.034 0.027 -0.089 0.141 0.115 -0.004 -0.087 1.000 0.054 -0.049 -0.016 0.004 -0.140 0.022 -0.120 0.109 -0.187 1.000 -0.044 -0.006 -0.029 -0.157 -0.089 -0.059 -0.262 0.078 -0.011 0.094 1.000 -0.001 -0.022 0.003 -0.114 -0.046 0.040 0.061 0.080 0.071 0.036 0.494 1.000 -0.106 0.066 -0.032 -0.051 -0.211 -0.024 -0.137 -0.005 0.001 0.050 0.121 0.037 1.000
150 Chiggerville (OH1) and the lower zone at Eva (BN12). The upper matrix triangle in
Table 7-17 represents the significance values for the divided distances. Three p-values
are larger than .05. Two non-significant p-value are associated with the upper Eva strata.
The third non-significant p-value is the distance between the upper and lower strata at
Indian Knoll. This lack of differentiation between the shell free midden and shell midden
at Indian Knoll is consistent with metric analyses indicating a homogenous sample
(Long, 1966; Herrmann, 2002). The remaining distances are significant. However,
sample sizes are small and distances may reflect sampling error. The FST calculated for
the divided sample distance matrix assuming equal effective population sizes is .0227
(s.e. = .0022). This estimate assumes full heritability (h²=1). When adjusted for
2 approximate heritability estimates for non-metric traits ( h = 0.35 ), then the FST estimate would equal .0622. This value is a similar to the full sample FST estimate.
A plot of the first two principal coordinates is presented in Figure 7-4. As compared to the full sample plot, the configuration has broken down a bit but the pattern of the larger sites remains the same. All upper stratigraphic samples fall on the margins of the component space. The two Indian Knoll samples fall close together as would be expected based on the significance test. The upper statigraphic samples from Carlston
Annis and Eva diverge from the main cluster falling on the positive end of the second principal coordinate axis. Chiggerville remains an outlier, falling on the extreme negative end of the first coordinate axis. Due to the extreme positions of several samples, the Mantel matrix comparisons are weak (Table 7-18). Geography and time are positively correlated with biological distance, but neither correlation is significant. The
151
Table 7-17. Distance matrix for divided site analysis. The lower triangle represents the distances and the upper are the significance values based on Drossler’s F-test. Shading represents non-significant differences. Sample sizes are in parentheses. BT10 BT5u BT5l McL11 McL4 OH1 OH2u OH21 BN12u BN12l (82) (48) (158) (203) (189) (61) (86) (368) (35) (61) BT10 - 0.022 0.002 0.000 0.000 0.000 0.004 0.000 0.080 0.009 BT5u 1.101 - 0.000 0.000 0.000 0.000 0.000 0.000 0.034 0.000 BT5l 0.794 1.323 - 0.000 0.000 0.002 0.000 0.000 0.007 0.000 McL11 1.052 1.576 0.864 - 0.000 0.000 0.000 0.000 0.001 0.001 McL4 1.092 1.468 0.990 0.552 - 0.000 0.000 0.000 0.061 0.000 OH1 1.643 2.141 0.973 2.169 2.182 - 0.000 0.000 0.011 0.000 OH2u 0.977 1.816 0.904 1.321 1.553 1.520 - 0.283 0.005 0.000 OH2l 1.006 1.447 0.558 0.915 1.161 1.099 0.292 - 0.000 0.000 BN12u 1.115 1.555 1.331 1.532 0.973 1.635 1.599 1.662 - 0.041 BN12l 1.061 1.867 1.123 0.969 1.076 2.641 1.507 1.618 1.376 -
152
1.00 Bt5u
Bn12u 0.57
Dim-2 0.15 Oh1 Bt10 McL4 + Bt5l -0.27 Bn12l McL11 Oh2l
Oh2u -0.70 -1.60 -0.88 -0.15 0.58 1.30 Dim-1
Figure 7-4. Plot of the first two principal coordinates derived from the divided distance matrix. The ‘+’ marks the centroid.
153
Table 7-18. Mantel matrix comparisons for the full sample divided by stratigraphic units at Indian Knoll, Eva, and Carlston Annis. Matrix comparison r p-value Direct Geography * Time 0.207 0.143 River Geography * Time 0.201 0.150 Bio * Direct Geography 0.248 0.142 Bio * River Geography 0.256 0.122 Bio * Time 0.212 0.124 Bio * Direct Geography (Time) 0.213 0.167 Bio * Time (Direct Geography) 0.170 0.169
divided stratigraphic distance pattern suggests that: (1) the divided samples may be too small/fragmentary for proper distance analysis, (2) a temporal trend in trait frequency is possibly present and (3) Chiggerville maintains a distinct separation from the main cluster. The issue of temporal trend will be discussed in the next section. The issue of
Chiggerville will be discussed at the end of the chapter.
Temporal Trend
In an effort to explain the separation and distance structure evident in the
collections examined, I performed a temporal trend analysis on the divided site data by
mean site date. Thirteen principal components of the tetrachoric correlation matrix were
extracted for the ten samples examined. The samples consist of the four undivided sites
(15McL4, 15McL11, 15OH1, 15BT10) and three sites with upper and lower strata
(40BN12, 15OH2 and 15BT5). Pearson correlation coefficients of each principal
component and mean site date are presented in Table 7-19. The probability value is based on 999 permutations of the principal component vector. Three components (PC2,
154 Table 7-19. Pearson correlation coefficients for the temporal trend analysis of the divided site sample. Shaded cells represent significant values. PC R p-Value % Variation PC1 0.141 0.728 15.1 PC2 0.740 0.040 14.1 PC3 -0.421 0.323 11.4 PC4 0.700 0.042 10.0 PC5 0.316 0.458 8.0 PC6 0.419 0.324 7.7 PC7 0.801 0.015 7.5 PC8 0.472 0.247 6.7 PC9 -0.113 0.832 6.0 PC10 -0.072 0.877 4.7 PC11 0.088 0.834 4.5 PC12 -0.258 0.535 2.9 PC13 -0.351 0.424 1.5
155 PC4 and PC7) are significantly correlated with time. These three components account for
31.6% of the total variation in the samples examined. Therefore, I dropped these
components and re-calculated the distance matrix with the remaining ten principal
components and eigenvalues (Table 7-20). A principal coordinate plot of the configuration shows that the upper unit of BT5 has moved closer to the lower unit
(Figure 7-5). The two Indian Knoll units are very close. The Eva sample has moved into
the range of the Green River samples but Eva is distinct on the third coordinate axis. The
relationship of the upper and lower strata at Carlston Annis and Indian Knoll mirror one
another. The correlation of straight-line distances and new biological distances controlling for time is .281 (p = .128) and the river distance correlation controlling for time is .271 (p = .131). The temporal correlation controlling for straight-line distances is
-.102 (p = .292). The temporal correlation controlling for river distances is -.098 (p =
.305). Although the pattern of spatial and temporal autocorrelation is as expected under an infinite island model of isolation by distance, the correlations are not significant.
The results from the temporal trend analysis of the divided site matrix should be viewed cautiously because the Eva samples significantly influence the correlation pattern.
In addition, it is very unlikely that the population inhabiting Eva participated in the migration exchange network of the Green River region. Once the two Eva samples are removed and all distances recalculated, a very different correlation pattern emerges.
First, the temporal and spatial distance matrices for the divided Green River sites are highly correlated (r = .451; p = .020). The two western sites are the earliest (15McL4 and 15McL11) and the younger samples are towards the east (15BT5u and 15BT10). By
156
Table 7-20. Distance matrix from the reduced principal components on eigenvalues after the temporal trend analysis. BT10 BT5u BT5l McL11 McL4 OH1 OH2u OH2l BN12u BN12l 0.000 0.759 0.000 0.556 0.655 0.000 0.659 0.564 0.522 0.000 1.068 0.916 0.962 0.301 0.000 0.884 1.293 0.745 1.331 1.976 0.000 0.703 0.895 0.557 0.437 1.004 0.959 0.000 0.863 0.938 0.493 0.367 0.976 0.890 0.161 0.000 0.974 1.049 1.065 0.792 0.653 1.315 1.348 1.389 0.000 0.696 0.707 0.780 0.586 0.883 1.678 0.643 0.889 0.718 0.000
157
1.00
Bn12u
0.63
Oh1 Dim-2 0.25 Bt10
Bt5u Bn12l + Bt5l McL4 -0.13 McL11
Oh2u Oh2l -0.50 -1.00 -0.44 0.13 0.69 1.25 Dim-1
Figure 7-5. Principal coordinate plot of the modified distance matrix derived from the temporal trend analysis. The “+” marks the centroid.
158 removing temporal trend one would likely remove spatial structure within the data set.
This procedure results in negligible temporal and spatial correlations.
Additional temporal, archaeological and skeletal data from other sites in the region are needed to resolve this confounding pattern. A reexamination of the stratigraphic information from Barrett may aid in dividing this sample. Ward represents a shallow rock midden. As such, temporally dividing this sample may prove impossible.
The only other dated western site within the region is Kirkland (McL12). This site may not help in the interpretation because the skeletal sample is small (n=77) and the six radiocarbon dates from the site span a 3500-year range (uncalibrated radiocarbon years)
(Haskins, 1992; Claassen, 1996b).
Determinant Ratio Bootstrap Tests
The examination of trait variation and biological distance structure by sex can provide valuable information concerning population history and postmarital residence patterns. Starting with the undivided Green River sites (n=6), the sample was divided by sex with tetrachoric correlations and threshold values calculated for the 12 groups. Once again, all traits with extreme threshold values were removed and any traits exhibiting significant sex effects were eliminated. Five traits remained and distances were calculated from the reduced threshold and correlation matrices (Table 7-21). The sex ratio in the Barrett sample appears high as compared to the other middens.
A plot of the first two principal coordinates shows that the males and females from most sites are in close proximity (Figure 7-6). Chiggerville exhibits the greatest
159
Table 7-21. Biological distance matrix of the Green River sites divided by sex. Sample sizes are in the second row of the table. BT10M BT10F BT5M BT5F McL11M McL11F McL4M McL4F OH1M OH1F OH2M OH2F 40 34 87 96 96 91 102 68 26 32 260 231 0.000 0.376 0.000 0.210 0.736 0.000 0.651 0.646 0.648 0.000 0.599 0.662 0.561 0.025 0.000 0.475 0.428 0.614 0.170 0.100 0.000 0.743 0.820 1.141 0.366 0.317 0.227 0.000 0.911 0.649 1.035 0.242 0.198 0.183 0.253 0.000 0.720 1.271 0.409 0.520 0.381 0.559 0.763 0.579 0.000 0.687 0.703 0.722 0.387 0.540 0.757 1.153 1.085 1.268 0.000 0.211 0.343 0.360 0.201 0.141 0.056 0.281 0.325 0.477 0.598 0.000 0.454 0.516 0.431 0.235 0.132 0.096 0.359 0.160 0.250 0.913 0.113 0.000
160
0.60 15Oh1M 15Bt5M
0.34
15Bt10M
15Oh2F Dim-2 0.08 15McL11M + 15Oh2M 15Bt5F 15McL11F 15Oh1F
-0.19 15McL4F 15McL4M 15Bt10F
-0.45 -0.50 -0.17 0.15 0.48 0.80 Dim-1
Figure 7-6. A principal coordinate plot of the Green River sites by sex. The “+” marks the centroid.
161 separation. The difference is probably due to sampling error. The Chiggerville male and
female samples are both the smallest. This pattern is consistent with Chiggerville’s
position in the full sample distance analysis. Carlston Annis and Read show moderate
separation. The male and female samples from Indian Knoll, Ward and Barrett are very
close. The distribution of male and female samples does not show a distinct pattern and
interpreting such a pattern would be tenuous (Konigsberg, 1987; Konigsberg and
Buikstra, 1995). The FST calculated from the Green River sex distance matrix assuming
equal effective population sizes equals 0.0230 (s.e. = 0.0040). When adjusted for
2 approximate heritability estimates for non-metric traits ( h = 0.35 ), then the FST estimate would equal 0.0629. This value is similar to the divided sample FST estimate.
The covariance matrix determinant ratios, confidence intervals and probability derived for each site and stratum are presented in Table 7-22. The covariance determinant ratios for the upper strata at Carlston Annis could not be calculated. No apical bones were present in the female sample and the males lacked mylohyoid bridging.
The determinant ratio could be calculated for the upper Eva sample. However, the small sample could not be bootstrapped. The upper strata samples exhibited too many missing cells. All the observed determinant ratios are negative except the Barrett sample. This pattern is consistent, but none of the ratios are significant based on the bootstrap distributions. The high male:female sex ratio at Barrett is intriguing. Sex estimates for
Barrett may need to be reevaluated. Sex estimation errors would influence the determinant ratio. Plots of each bootstrapped determinant ratio are provided in Figures 7-
7 through 7-11. Probabilities for the full sample analyses range from .087 to .574. The
162
Table 7-22. Covariance matrix determinant ratios and standard 95% confidence intervals for each site and stratigraphic unit. The covariance matrix could not be calculated for the upper stratigraphic unit at Carlston Annis. The male and female matrices are not positive semidefinite (i.e. each has one invariant trait). The bootstrap could not be performed for the Eva upper stratum. Lower Upper Bootstrap Bootstrap 95% 95% Site #M/#F Observed mean SD CI CI Probability* Carlston Annis (15BT5) 69/77 -1.254 -1.319 0.923 -3.129 0.491 0.087 15BT5 upper 19/14 ------15BT5 lower 50/63 -1.737 -1.834 1.086 -3.963 0.294 0.055 Read (15BT10) 27/23 -0.035 0.068 1.972 -3.796 3.933 0.493 Barrett (15McL4) 81/57 0.105 0.281 0.561 -0.819 1.381 0.574 Ward (15McL11) 76/78 -0.678 -0.844 0.823 -2.458 0.769 0.205 Chiggerville (15OH1) 18/28 -0.478 -1.293 1.602 -4.433 1.847 0.383 Indian Knoll (15OH2) 255/226 -0.389 -0.371 0.348 -1.053 0.310 0.131 15OH2 upper 40/39 -0.945 -1.282 1.337 -3.902 1.339 0.240 15OH2 lower 195/169 -0.211 -0.187 0.402 -0.975 0.600 0.300 Eva (15BN12) 34/47 -0.932 -1.455 1.319 -4.039 1.130 0.240 15BN12 upper 10/16 -1.023 - - - - - 15BN12 lower 24/31 -0.756 -1.437 1.514 -4.405 1.530 0.309 All Green River Sites 526/489 -0.276 -0.265 0.230 -0.716 0.186 0.124 * Probabilities are derived from one-tailed z-scores calculated from the difference of the observed determinant ratio or bootstrap mean from zero divided by the bootstrap standard deviation.
163
15BT5 0.4 0.3 Density 0.2 0.1 0.0
-4 -2 0 2 Log Determinant Ratio
15BT5 Lower 0.35 0.30 0.25 0.20 Density 0.15 0.10 0.05 0.00
-6 -4 -2 0 2 Log Determinant Ratio
Figure 7-7. Determinant ratio bootstrap density plots for Carlston Annis. The upper plot represents the full sample. The dashed line represent the observed ratio. Dotted lines represent the standard 95% CI.
164 15BT10 0.25 0.20 0.15 Density 0.10 0.05 0.00
-10 -5 0 5 Log Determinant Ratio
15OH1 0.30 0.25 0.20 0.15 Density 0.10 0.05 0.00
-10 -8 -6 -4 -2 0 2 4 Log Determinant Ratio
Figure 7-8. Determinant ratio bootstrap density plots for Read (15BT10) and Chiggerville (15OH1). The dashed line represent the observed ratio. Dotted lines represent the standard 95% CI.
165 15OH2 1.0 0.8 0.6 Density 0.4 0.2 0.0
-1.5 -1.0 -0.5 0.0 0.5 1.0 Log Determinant Ratio
15OH2 Upper 0.30 0.25 0.20 Density 0.15 0.10 0.05 0.00
-6 -4 -2 0 2 Log Determinant Ratio
15OH2 Lower 1.0 0.8 0.6 Density 0.4 0.2 0.0
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Log Determinant Ratio
Figure 7-9. Determinant ratio bootstrap density plots for Indian Knoll. The dashed line represent the observed ratio. Dotted lines represent the standard 95% CI.
166 15McL11 0.5 0.4 0.3 Density 0.2 0.1 0.0
-6 -4 -2 0 2 Log Determinant Ratio
15McL4 0.6 0.4 Density 0.2 0.0
-2 -1 0 1 2 3 Log Determinant Ratio
Figure 7-10. Determinant ratio bootstrap density plots for Ward (15McL11) and Barrett (15McL4). The dashed line represent the observed ratio. Dotted lines represent the standard 95% CI.
167 40BN12 0.30 0.25 0.20 Density 0.15 0.10 0.05 0.00
-8 -6 -4 -2 0 2 Log Determinant Ratio
40BN12 Lower 0.30 0.25 0.20 0.15 Density 0.10 0.05 0.00
-10 -5 0 Log Determinant Ratio
Figure 7-11. Determinant ratio bootstrap density plots for Eva. The upper plot represents the full sample. The dashed line represent the observed ratio. Dotted lines represent the standard 95% CI.
168 divided samples exhibit similar patterns with a slight probability reduction in the lower
Carlston Annis sample (p = .055). The combined probability for the full Green River
samples (without Eva) produces a c 2 value of 16.55 (df = 12) which corresponds to a p-
value of .167. This test maintains the independence of the samples. The pattern is not
significant but it is worth noting that all but one of the determinants is negative. When
the Green River sites are combined into one large sample, the determinate ratio is
negative as expected from the individual site tests and zero falls within the 95%
confidence interval with an associated one-tailed z-score probability of .124.
Discussion
While the above analyses provide a structure to the SMA skeletal populations, what are the implications of these relationships? Missing data appear to have significantly influenced biological distance measures. Chiggerville and Read have the highest levels of missing data and are the two most divergent sites within the full sample distance matrix. Based on the radiocarbon dates, however, Read potentially represents shorter and later habitation as compared to the larger middens. The skeletal sample at Read and
Chiggerville may represent shorter time capsules as compare to Carlston Annis and
Indian Knoll. As such, Read and Chiggerville may represent a smaller sampling of the genetic variation present in the Green River drainage during the late Archaic.
The full sample analysis offers a good regional perspective on the biological distance structure between the Tennessee and Green Rivers. A geographic trend is present in the large mounds with marked separation of Chiggerville and Eva. The two western Pond River/Cypress Creek sites group together and Indian Knoll, Carlston Annis 169 and Read are in close proximity. Although these relationships appear to conform to a regional pattern, the separation of Chiggerville and the position of Eva relative to all the
Green River sites significantly reduce the distance matrix correlations.
The pattern of the Green River drainage sites is consistent with a local mate exchange network where individuals were most likely moving between proximate groups. Hofman (1984; 1986) suggests that middle to late Archaic hunter-gatherers would seasonally aggregate near key resources patches, typically shoal locations along the major drainages. These groups may have maintained some form of communal “rights of exclusive access” to these resource patches, possibly claimed on the basis of kinship
(Crothers, 1999:249). Thus, these locations served as regional melting pots where groups came to live, feast, bury their dead, negotiate access rights, and exchange mates.
The groups would then disperse to smaller localities, possibly seasonal base camps or inter-riverine extractive sites (House and Ballenger 1976; Sassaman and Anderson 1995).
Whether these groups would be considered foragers or collectors within a logistical or residential mobility strategy in a Binfordian sense is beyond this dissertation.
Hensley (1994:246) views the Green River SMA sites within the perspective of
Hofman’s (1986) settlement model which describes them “as aggregation sites within a logistical settlement system maintained by a mobile population.” Therefore, biological links through proximate mate exchange and marriage would reinforce the strong geographic pattern evident in biological distance matrices.
The separation of Eva from the Green River samples appears to support
Rolingson and Schwartz’s (1966) basic model of regional Archaic period differentiation
170 along the Tennessee/Cumberland and Tradewater/Green drainages. Additional data from a variety of sites need to be incorporated into a larger midsouth Archaic biological distance study. Sites such as Black Earth (Carrier Mills) in southern Illinois (Jefferies and Butler, 1982; Jefferies and Lynch, 1983), KYANG and Rosenberger near the Falls of the Ohio River (Bader and Granger, 1989; Driskell, 1979; Granger, 1988; Janzen, 1977),
Parrish village on the Tradewater River (Rolingson and Schwartz, 1966), Cherry and
Ledbetter on the Tennessee River (Magennis, 1977) and Anderson and Robinson on the
Cumberland River (Dowd, 1989; Morse, 1967) need to be examined to gain a full regional perspective.
Insight from the biological distance structure derived from the divided site analysis is limited. The upper stratigraphic samples from Eva and Carlston Annis are small and fragmentary. However, once the temporal trend in trait frequencies was removed the overall distance pattern is consistent for the Green River sites. The geographic trend is maintained in the plot. Chiggerville maintains an extreme position relative to all the sites. The upper and lower units from Indian Knoll and Carlston Annis exhibit similar relationships. It is not known if this relationship is strictly temporal or if it represents re-sampling. The upper samples from both sites were arbitrarily defined to represent the shell-free midden capping each site. The upper and lower samples from each site may have a great deal of mixing with several individuals from later occupations being buried deep in the midden.
The distance structure by sex maintains the relationships evident in the full site analysis. In the principal coordinate analysis, female samples tend to score low on
171 coordinate two and males load high on this axis. Male and female samples from Indian
Knoll, Ward, and Barrett fall close together; whereas Read, Carlston Annis and
Chiggerville are separated. The males and females from Chiggerville fall on the extreme
ends of the coordinate space. Both sexes are contributing to the divergent position of this
sample in the full sample analysis. The determinant ratio tests are consistently negative
for all but one of the sites. This pattern would seem to suggest greater female mobility
which is a common post-marital migration pattern observed (i.e. patrilineal and/or
patrilocal) in hunter-gatherer populations throughout the world (Radcliffe-Brown, 1931,
Service, 1971; Kennedy, 1981). However, the bootstrap results clearly demonstrate that all of the ratio tests are non-significant even with substantial sample sizes. Craniometric data from Indian Knoll (Snow, 1948) produces a non-significant positive determinant ratio for a series of 19 variables. Thus, any statements on residential mobility will require additional archaeological or biological data.
172 Chapter 8. Conclusions
The large skeletal samples from the middle Green River Drainage provide a
wealth of information concerning late middle to late Archaic Period populations from
Kentucky. These samples represent a unique period in prehistory at which time hunting
and gathering groups began the long process of shifting towards food production and
more sedentary settlement patterns. The Green River groups may have in fact begun this
process later than other mid-Holocene populations found throughout the southeastern
United States. Social networks and increased social interaction envisioned by a variety of
researchers (e.g. Jefferies, 1996, 1997; Marquardt, 1985; Winters, 1969) must have had a
unifying affect on the biological structures of these groups, both on the local and regional
scale. This study specifically addresses this issue on both scales. The inclusion of the
Eva sample from the lower Tennessee River valley provides a limited regional
perspective.
The biological distance analysis described in the preceding chapters demonstrates that (1) the Green River samples are biologically distinct from the Eva sample, (2) the
Green River samples are significantly different from each other based on non-metric trait patterns, (3) the burial sample from the shell-free midden at Indian Knoll is not significantly different from the shell midden sample, and (4) the biological distance structure is consistent with a pattern of isolation by geographic distance. Isolation by temporal distance is more difficult to comment on given that the burial samples at some sites were interred over a span of several centuries or possibly millennia.
173 Powell (1995) has addressed the large regional scale issues through the
examination of a series of southeastern mid-Holocene skeletal samples. Within Powell’s analysis, the correlation of temporal and spatial distance with biological distance does indicate some level of gene flow between the spatially distant samples. I believe that
Jefferies (1996) potentially reads too much into Powell’s conclusion when he states that
“Powell’s finding provide support for long-distance migration, and such patterns appear to be correlated with exchange distances between groups” (Jefferies 1997:482). The definition of “long-distance migration” is critical to Jefferies’ statement because the samples Powell examined span an area from Florida to Southern Illinois. I suggest that group long-distance migration and direct trade at this scale would have been very low.
Down-the-line exchange of trade goods and marriage partners between proximate groups spanning the Southeast could have facilitated the level of gene flow evident within Powell’s analysis. I envision a similar level of social interaction as hypothesized by Jefferies (1997) within the Green River drainage where there is a potential for mate exchange across the entire region. The distribution of materials and mates may have
“resulted from numerous multidirectional short distance exchanges rather than from preferentially directed or long distance exchange” (Goad, 1980:7). These groups were probably small and maintained a mobile existence for a substantial portion of the year, if only for multiple short-term logistically organized extractive activities. The probability of movement between nearest neighbors was far greater than the probability of moving from the Big Bend area to the Pond River. The burial populations represent several generations, and typify the concept of the “skeletal lineage” as defined by Cadien et al.
174 (1974). With time, these samples would begin to be very similar given the level of mate
exchange within the social network across the valley. We do not know if groups were
concurrently living at each site. In addition, it is possible that the same groups occupied
several of the SMA midden sites at different times as part of a multi-year settlement schedule. Ultimately, these factors would contribute to a homogeneous regional skeletal sample. The simple fact that there is some level of geographic patterning in the biological distance structure suggests some spatial limit on group interaction.
This study has filled some gaps in our understanding of Archaic groups in
Kentucky. It has also generated additional questions. Valuable new radiocarbon dates are available from three sites (Barrett, Indian Knoll and Ward). The two dates from
Barrett represent the first absolute dates from this site. The two new dates from Indian
Knoll have provided critical information concerning the upper and lower stratigraphic units of the shell midden. Finally, the three dates from Ward have helped clarify the date range of the site and provided a terminus post quem for a pit feature with multiple burials.
The biological distance structure and non-metric data derived in this study will provide a basis for future comparisons of the Green River samples to other SMA sites throughout the Southeast. The calculated distance structure also provides archaeological researchers concerned with the Green River region a framework to build and test new hypotheses concerning social networks and settlement and subsistence models. The divided site distance structure when adjusted for temporal trend does conform to the expected correlation pattern for time and geography. However, these correlations are not significant. The confounding problem with the Green River samples examined in this
175 study is that there is a geographic temporal trend of the sites from west to east.
Additional sites need to be examined, dated and incorporated into the analysis.
The examination of the shell free midden at Indian Knoll and Carlston Annis
proved quite interesting. In both cases, the shell free midden zone was delimited based
on prior field observations from the WPA and SMAP investigations. It is obvious that
the samples did not represent a complete sampling of all burials associated with the shell free occupations but the sampling process reasonably represented the zone at each site.
The shell free burial sample from Carlston Annis is small and fragmentary, thus the distance measure for this sample is suspect. The Indian Knoll sample is quite large and well represented. The distance measure between the stratigraphically defined samples from Indian Knoll is not significant. The metric variability of the Indian Knoll sample has often been considered quite low (Steele, 1948; Long, 1966; Herrmann, 2002), and the non-metric distance results support a conclusion of lower morphological variation within the Indian Knoll sample. Does the lower level of variation reflect a very stable population during a period of greater social interaction? Or does it reflect a shorter burial interment period at Indian Knoll? Or are we simply documenting a large-scale reduction in morphological variability across the entire Southeast as a result of the greater level of social interaction?
Looking to the Future
This study represents one in a long series of osteological examinations of the
Green River material. It does however represent one of the first attempts to view the
176 Green River skeletal samples as an entire unit. By this I mean that often researchers only look at one or two of the Green River samples relative to other sites outside the region or relative to each other. This study focuses on all the major shell middens of the middle
Green River valley. As such, it provides useful data to address questions concerning the entire region.
Historically, a constant problem for all Green River researchers has been a dearth of good radiocarbon dates for a majority of the middens. Given the advances in AMS radiocarbon dating, this should only be a financial issue with minimal loss of bone, carbon or fabric. New investigations at several middens by a variety of researchers have provided invaluable chronometric data from these sites. Dated human burials are critical
to the interpretation of the WPA excavations. Unfortunately, university museums have
responded to recent NAGPRA legislation by curtailing any destructive analyses on
unaffiliated collections, and federal granting agencies are probably less likely to fund
proposals for systematic dating of numerous burials. Additional sources of carbon are
available through faunal collections, bone tools, and fabric, but these resources are
limited and vary by site. These options will need to be fully explored in an effort to tie in
the old WPA investigations with recent excavations.
Research with the skeletal material needs to be twofold: new research and
consolidating old research data. The number of research projects involving Indian Knoll
is amazing; however, a centralized database of this research does not exist. Valerie
Haskins and James Fenton have championed this effort over the years for the Kentucky
177 shell mound collections, but the concept of a unified database needs to be supported and
funded, especially if these collections are ever reburied.
New research with these collections needs to focus on the analysis of the
remaining Green River samples at Butterfield and Kirkland as well as some of the smaller
sites. These samples may be too small or too fragmentary to incorporate into the type of
analysis discussed in this study, but we need to know what is happening in these smaller
SMA skeletal samples. The fragmentary nature of most of the Kentucky SMA sites
limits types of analyses and the sample sizes. In addition, the “cemetery” areas identified
at Barrett, Butterfield and Ward need to be examined in detail and dated. These may
represent the initial interment at the site and conform to Hofman’s (1986) reburial areas.
Additional osteometric analyses need to continue with the Green River material.
Postcranial and cranial data should be combined with the non-metric data into an overall biological distance analysis employing the methods proposed by Bedrick et al. (2000).
Once additional radiocarbon dates are available and a greater number of samples are examined then time series analysis of isolation by temporal and spatial distance models will provide a powerful statistical tool in the analysis of these samples (Epperson, 1993,
1995, 2000).
A current three-dimensional morphometric study of Indian Knoll and other Green
River Archaic sites is ongoing (Herrmann, 2002), and these efforts are focused on assessing morphological variability in the Green River samples. The postcranial data from these sites may hold great promise and several researchers have already utilized these data. Although the status of DNA studies on these samples is unknown, future
178 DNA research may provide invaluable information concerning the biological relationship and internal structure of the shell mound by relating specific individuals.
Overall this study has demonstrated the utility of non-metric cranial data in the analysis of population structure. The quantitative genetic methods employed in this analysis produced valid results which will aid in future interpretations of the Green River
Archaic populations. Although several new radiocarbon dates have been reported for the
SMA sites, additional determinations are required to help elucidate complex depositional histories of the middens and burial samples. The biological distance data marginally conforms to expected patterns of isolation by distance. Osteometric and non-metric data from additional sites within the Green River region may help clarify the distance structure documented in this study. Even though some of these samples have been studied for well over fifty years, these skeletal collections continue to yield valuable clues concerning the biological and cultural history of these Archaic hunters-gatherers.
179
References Cited
180 References Cited
Allaway EB (1980) Is There Sex After Death? An Investigation into the Meaning of Skeletal Sexual Dimorphism. Master’s Thesis, Department of Anthropology, University of Kentucky, Lexington.
Anderson DG, and Hanson GT (1988) Early Archaic settlement in the southeastern United States. American Antiquity 53:262-286.
Arnold JR, and Libby WF (1951) Radiocarbon dates. Science 2927:113-114.
Ashford JR, and Sowden RR (1970) Multi-variate probit analysis. Biometrics 26:535- 546.
Bader AT, and Granger JE (1989) Recent Archaeological Investigations on the Kentucky Air National Guard Site (15Jf267), Jefferson County, Kentucky. Louisville: Granger Consultants.
Baerreis DA (1980) Habitat and climatic interpretations from terrestrial gastropods at the Cherokee Site. In DC Anderson and HS Semkens, Jr. (eds.): The Cherokee Excavations: Mid-Holocene Paleoecology and Human Adaptation in Northwest Iowa. New York: Academic Press, pp. 101-122.
Baerreis DA (n.d.) Terrestrial gastropods at the Carlston Annis Site (15Bt5): Their habitat and climatic implications. In WH Marquardt and PJ Watson (eds.): The Archaeology of the Middle Green River. In preparation.
Balakrishnan V, and Sanghvi LD (1968) Distance between populations on the basis of attribute data. Biometrics 24:859-865.
Baughman AL (1988) A FORTRAN function for the bivariate normal integral. Comput. Meth. Prog. Biomed. 27:169-174.
Bedrick EJ, Lapidus J, and Powell JF (2000) Estimating the Mahalanobis distance from mixed continuous and discrete data. Biometrics 56:394-401.
Berry AC (1974) The use of non-metrical variations of the cranium in the study of Scandinavian population movements. American Journal of Physical Anthropology 40:345-358. ADD
Berry AC (1975) Factors affecting the incidence of non-metrical skeletal variants. Journal of Anatomy 120:519-535. ADD
181 Berry AC, and Berry RJ (1967) Epigenetic variation in the human cranium. Journal of Anatomy 101:361-379.
Berry RJ (1968) The biology of non-metrical variation in mice and men. In DR Brothwell (ed.): The Skeletal Biology of Earlier Populations. Oxford: Pergamon Press, pp. 103- 133.
Binford LR (1980) Willow smoke and dog’s tails: Hunter-gatherer settlement systems and archaeological site formation. American Antiquity 45:4-20.
Black SJ (1982) Quantitative Genetics of Anthropometric Variation in the Solomon Islands. Unpublished Ph.D. Dissertation, The University of Auckland, New Zealand.
Blakely RL (1971) Comparison of the mortality profiles of the Archaic, Middle Woodland, and Middle Mississippian skeletal populations. American Journal of Physical Anthropology 34:43-54.
Blangero J, and Williams-Blangero S (1991) Estimating biological distances from dichotomous threshold traits. American Journal of Physical Anthropology Supplement 12:51-52 (Abstract).
Boas F (1908) Heredity in head form. American Anthropologist 5:530-538.
Bodmer WF, and Cavalli-Sforza LL (1968) A migration matrix model for the study of random genetic drift. Genetics 59:565-592.
Bronk RC (1994) Analysis of Chronological Information and Radiocarbon Calibration: The Program OxCal. Archaeological Computing Newsletter 41:11-16.
Bronk RC (1995) Radiocarbon Calibration and Analysis of Stratigraphy: The OxCal Program. Radiocarbon 37(2): 425-430.
Brown JA, and Vierra BK (1983) What happened in the Middle Archaic? Introduction to an ecological approach to Koster Site archaeology. In JL Phillips and JA Brown (eds.): Archaic Hunters and Gatherers in the American Midwest. New York: Academic Press, pp. 165-195.
Brown MB (1977) The tetrachoric correlation and its asymptotic standard error. Applied Statistics 26:343-351.
Buikstra JE (1976) Hopewell in the Lower Illinois Valley: A Regional Approach to the Study of Human Biological Variability and Prehistoric Behavior. Evanston: Northwestern University Archaeological Program.
182 Buikstra JE (1992) Diet and disease in late prehistory. In JW Verano and DH Ubelaker (eds.): Disease and Demography in the Americas. Washington DC: Smithsonian Institution Press, pp. 87-101.
Buikstra JE, and Charles DK (1999) Centering the ancestors: Cemeteries, mounds and sacred landscapes of the ancient North American Midcontinent. In W Ashmore and AB Knapp (eds.): Archaeologies of Landscape: Contemporary Perspectives. Oxford: Blackwell Publishers.
Burke (1981) A comparative analysis of synchronic and diachronic craniometric variation in modern homo sapiens. Ph.D. Dissertation, Department of Anthropology, University of Colorado, Boulder
Cadien JD, Harris EF, Janes WP, and Mandarino LJ (1974) Biological lineages, skeletal populations, and microevolution. Yearbook of Physical Anthropology 18:194-201.
Caldwell JR (1958) Trend and tradition in the prehistory of the eastern United States. Memoirs of the American Anthropological Association 88:1-100.
Carr PJ (1991) Organization of Technology and Lithic Analysis: Prehistoric Hunter- Gatherer Occupation of the Hayes Site (40ML139). Master’s Thesis, Department of Anthropology, The University of Tennessee, Knoxville.
Carr PJ (ed.) (1994) The organization of North American prehistoric chipped stone tool technologies. Ann Arbor: International Monograpns in Prehistory.
Cassidy CM (1972) Comparison of nutrition in pre-agricultural skeletal populations. Ph.D. Dissertation, Department of Anthropology, University of Wisconsin, Madison.
Cassidy CM (1980) Nutrition and health in agriculturalists and hunter-gatherers: A case study of two prehistoric populations. In NW Jerome, RF Kandel and GH Pelto (eds.): Nutritional Anthropology: Contemporary Approaches to Diet and Culture. New York: Redgrave Publishing Company, pp. 117-145.
Cassidy CM (1984) Skeletal evidence for prehistoric subsistence adaptation in the central Ohio River valley. In M Cohen and GJ Armelegos (eds.): Paleopathology at the Origins of Agriculture. New York: Academic Press, pp. 307-38.
Charles DK, and Buikstra JE (1983) Archaic mortuary sites in the central Mississippian drainage: Distribution, structure, and behavioral implications. In JL Phillips and JA Brown (eds.): Archaic Hunters and Gatherers in the American Midwest. New York: Academic Press, pp. 117-145.
183 Cheverud JM (1979) Genetic and environmental morphological variation among social groups of rhesus monkeys (Macaca mutatta) on Cayo Santiago. Ph.D. Dissertation, The University of Wisconsin, Madison.
Cheverud JM (1982) Phenotypic, genetic, and environmental morphological integration in the cranium. Evolution 36:499-516.
Cheverud JM (1988) A comparison of genetic and phenotypic correlations. Evolution 42:958-968.
Cheverud JM, and Buikstra JE (1981a) Quantitative genetics of non-metric traits in the Rhesus Macaques on Cayo Santiago. I. Single trait heritabilities. American Journal of Physical Anthropology 54:43-49.
Cheverud JM, and Buikstra JE (1981b) Quantitative genetics of non-metric traits in the Rhesus Macaques on Cayo Santiago. II. Phenotypic genetic and environmental correlations between traits. American Journal of Physical Anthropology 54:51-58.
Cheverud JM, and Buikstra JE (1982) Quantitative genetics of non-metric traits in the Rhesus Macaques on Cayo Santiago. III. Relative heritability of skeletal non-metric and metric traits. American Journal of Physical Anthropology 59:151-155.
Cheverud JM, Buikstra JE, and Twichell E (1979) Relationships between non-metric skeletal traits and cranial size and shape. American Journal of Physical Anthropology 50:191-198.
Claassen CP (1991) Gender, shellfishing, and the Shell Mound Archaic. In JM Gero and MW Conkey (eds.): Engendering Archaeology: Women in Prehistory. New York: Basil Blackwell, pp. 276-300.
Claassen CP (1992) Shell mounds as burial mounds: A revision of the Shell Mound Archaic. In D Pollack and AG Henderson (eds.): Current Research in Kentucky, Vol. 2. Frankfort: Kentucky Heritage Council, pp. 1-12.
Claassen CP (1996a) Research problems with shells from Green River shell matrix sites. In K Carstens and PJ Watson (eds.): Of Cave and Shell Mounds. Tuscaloosa: The University of Alabama Press, pp. 132-139.
Claassen CP (1996b) A consideration of the social organization of the Shell Mound Archaic. In KE Sassaman and DG Anderson (eds.): Archaeology of the Mid- Holocene Southeast. Gainesville: University of Florida Press, pp. 235-258.
Constandse-Westermann TS (1972) Coefficients of Biological Distance. Oosterhout, The Neatherlands: Anthropological Publications.
184
Corruccini RS (1974) An examination of the meaning of cranial discrete traits for human skeletal biological studies. American Journal of Physical Anthropology 40:425-446.
Corruccini RS (1976) The interaction between non-metric and metric cranial variation. American Journal of Physical Anthropology 44:285-294.
Crothers GM (1999) Prehistoric Hunters and Gatherers, and the Archaic Period Green River Shell Middens of Western Kentucky. Ph.D. Dissertation, Department of Anthropology, Washington University, St. Louis.
Crothers GM, Morey DF, and Stein JK (2002) The correspondence between river shoals and archaic shell midden sites in the middle Green River valley of Kentucky. Paper presented at the Joint Annual meeting of the North-Central and Southeastern Sections of Geological Society of America. St. Louis, Missouri.
Devor EJ, McGue M, Crawford MH, and Lin PM (1986) Transmissible and nontransmissible components of anthropometric variation in the Alexanderwohl Mennonites .2. Resolution by path-analysis. American Journal of Physical Anthropology 69:83-92
Dow M, and Cheverud JM (1985) Comparison of distance matrices in studies of population structure and genetic microdifferentiation. American Journal of Physical Anthropology 68:367-374.
Dowd JT (1989) The Anderson Site: Middle Archaic Adaptation in Tennessee’s Central Basin. Miscellaneous Paper No. 13. Knoxville: Tennessee Anthropological Association.
Driskell BN (1979) The Rosenberger Site. In MB Collins (ed.): Excavations at Four Archaic Sites in the Lower Ohio Valley, Jefferson County, Kentucky. Department of Anthropology, Occasional Papers in Anthropology, No. 1. Lexington: University of Kentucky, pp. 697-803.
Droessler J (1981) Craniometry and Biological Distance: Biocultural Continuity and Change at the Late Woodland-Mississippian Interface. Evanston, Illinois: Center for American Archaeology at Northwestern University.
Epperson BK (1993) Spatial and space-time correlations in systems of subpopulations with genetic drift and migration. Genetics 133:711-727.
Epperson BK (1995) Spatial distributions of genotypes under isolation by distance. Genetics 140:1431-1440.
185 Epperson BK (2000) Spatial and space-time correlations in ecological models. Ecological Modeling. 132:63-76.
Falconer DS (1989) Introduction to Quantitative Genetics. Essex, England: Longman Scientific & Technical.
Fenneman NM (1938) Physiography of Eastern United States. New York: McGraw-Hill Book Company.
Fenton JP, Herrmann NP, Rankin C, and McBride D (1999) Recent Investigations at Indian Knoll. Paper presented at the Fifty-Sixth Annual Meeting of the Southeastern Archaeological Conference, Pensacola, Florida.
Fenton JP, and Herrmann NP (2000) Finding the Boundaries: Laboratory Research on Indian Knoll. Paper presented at the Fifty-Seventh Annual Meeting of the Southeastern Archaeological Conference, Macon, Georgia.
Finnegan M, and Cooprider K (1978) Empirical comparisons of distance equations using discrete traits. American Journal of Physical Anthropology 49:39-46.
Fowler ML (1959) The Modoc Rock Shelter: A Study of the Processes of Ecological Adaptation of a Primitive Culture in the Central Mississippi Valley Area. Ph.D. Dissertation, Department of Anthropology, University of Chicago. Ann Arbor: University Microfilms.
Funkhouser WD (1938) A study of the physical anthropology and pathology of the osteological material from the Norris Basin. In WS Webb: An Archaeological Survey of the Norris Basin in Eastern Tennessee. Smithsonian Institution, Bureau of American Ethnology, Bulletin 118, pp. 225-251.
Funkhouser WD, and Webb WS (1928) Ancient Life in Kentucky. Frankfort, Kentucky: The Kentucky Geological Survey.
Funkhouser WD, and Webb WS (1932) Archaeological Survey of Kentucky. Lexington: The University of Kentucky.
Geyh MA, and Schleicher H (1990) Absolute Age Determination. Berlin: Springer- Verlag.
Goad S (1980) Patterns of Late Archaic exchange. Tennessee Anthropologist 5:1-16.
186 Granger JE (1988) Late/Terminal Archaic settlement in the Falls of the Ohio River region of Kentucky: An examination of components, phases, and clusters. In C Hockensmith, D Pollack and T Sanders (eds.): Paleoindian and Archaic Research in Kentucky. Frankfort, Kentucky Heritage Council, pp. 153-204.
Green RF, and Suchey JM (1976) The use of inverse sine transformations in the analysis of non-metric cranial data. American Journal of Physical Anthropology 45:61-68.
Grewal MS (1962) The rate of genetic divergence of sublines in the C57BL strain of mice. Genetic Research 3:226-237.
Grüneburg H (1963) Pathology of Development. New York: John Wiley.
Hamilton ME (1975) Variation Among Five Groups of AmerIndians in the Magnitude of Sexual Dimorphism of Skeletal Size. Ph.D. Dissertation, Department of Anthropology, University of Michigan, Ann Arbor.
Hardy VC (1974) The Effects of Size Variation on Statistical Assessments of Biological Distance Between Human Skeletal Series. Master’s Thesis, Department of Anthropology, University of Kentucky, Lexington.
Harpending HC, and Jenkins T (1973) Genetic distance among Southern African populations. In MH Crawford and PL Workman (eds.): Methods and Theory of Anthropological Genetics. Albuquerque: University of New Mexico Press, pp. 177- 199.
Harpending HC, and Ward RH (1982) Chemical systematics and human populations. In M Nitecki (ed.): Biochemical Aspects of Evolutionary Biology. Chicago: University of Chicago, pp. 213-256.
Haskins VA (1992) Recent dates from the Green River shell mound region. Paper presented at the 9th Annual Kentucky Heritage Council Archaeological Conference.
Haskins VA, and Herrmann NP (1996) Shell mound bioarchaeology. In K Carstens and PJ Watson (eds.): Of Cave and Shell Mounds. Tuscaloosa: The University of Alabama Press, pp. 107-118.
Hensley-Martin CK (1986) A reanalysis of the lithic industry from the Read Site, Butler County, Kentucky (15Bt10). Master’s Thesis, Department of Anthropology, Washington University, St. Louis, Missouri.
Hensley CK (1994) The Archaic Settlement System of the Middle Green River Valley, Kentucky. Ph.D. Dissertation, Department of Anthropology, Washington University, St. Louis, Missouri.
187
Herrmann NP (1990) The Paleodemography of the Read Shell Midden, 15Bt10. Master’s Thesis, Washington University, St. Louis, Missouri.
Herrmann NP (1996) Archaic shell mound paleodemography: A case study from the Read Site (15Bt10). In TN Sanders and C Stout (eds.): Current Research in Kentucky, Volume Four. Lexington: Kentucky Heritage Council., pp. 69-88.
Herrmann NP (1998) Age-at-death Distribution Estimation: an Example from Indian Knoll. Paper presented at the Ninety-Seventh Annual Meeting of the American Anthropological Association, Philadelphia, Pennsylvania.
Herrmann NP (2002) Is There Any Craniometric Variation at Indian Knoll? Paper presented at the Nineteenth Annual Kentucky Heritage Council Conference. Frankfort, Kentucky.
Herrmann NP (in press) New Radiocarbon Dates from Three Archaic Shell Middens in Western Kentucky: Indian Knoll (15Oh2), Ward (15McL11) and Barrett (15McL4). Current Research in Kentucky, Volume Eight. Lexington: Kentucky Heritage Council.
Herrmann NP, and Adams BJ (1996) Cranial nonmetric traits of Averbuch (40DV60): A study of biological distance. American Journal of Physical Anthropology Supplement 25:123-124 (Abstract).
Herrmann NP, and Fenton JP (2000) A report on new radiocarbon dates and recent investigations at three Green River Archaic sites. Paper presented at the Seventeenth Annual Kentucky Heritage Council Conference. Bowling Green, Kentucky.
Herrmann NP, and Konigsberg LW (2002) A re-examination of the age-at-death distribution of Indian Knoll. In RD Hoppa and JW Vaupel (eds): Paleodemography. New York: Cambridge University Press, pp. 243-257.
Hockensmith CD, Sander TN, and Pollack D (1985) The Green River Shell Middens of Kentucky. National Register of Historic Places thematic nomination form. On file, Kentucky Heritage Council, Frankfort.
Hofman JL (1984) Hunter-Gatherers in the Nashville Basin of Tennessee, 8,000-5,000 B.P. Tennessee Anthropologist 9:129-192.
Hofman JL (1986) Hunter-Gatherer Mortuary Variability: Towards an Explanatory Model. Ph.D. Dissertation, Department of Anthropology, University of Tennessee, Knoxville.
188 House JH, and Ballenger DL (1976) An archaeological survey of the Interstate 77 Route in the South Carolina Piedmont. Research Manuscript Series 104. Columbia: Institute of Archaeology and Anthropology, University of South Carolina.
Hrdlicka A (1927) Catalogue of human crania in the United States National Museum Collections. Proceedings (National Museum, Smithsonian Institution) 65.
Ihaka R, and R Gentleman (1996) R: A Language for Data Analysis and Graphics. Journal of Computational and Graphical Statistics. 5:299-314. http://www.r- project.org/
Jantz RL (1973) Microevolutionary change in Arikara crania: A Multivariate analysis. American Journal of Physical Anthropology 39:93-104.
Jantz RL, and Owsley DW (2001) Variation among early North American crania. American Journal of Physical Anthropology 114:146-55.
Janzen D (1977) An examination of Late Archaic development in the Falls of the Ohio River area. In C Cleland (ed.): For the Director: Papers in Honor of James B. Griffin. Anthropological Papers 61. Ann Arbor: Museum of Anthropology, University of Michigan, pp. 123-143.
Jefferies RW (1988) The Archaic in Kentucky: New Deal archaeological investigations. In D Pollack and ML Powell (eds.): New Deal Era Archaeology and Current Research in Kentucky. Frankfort: Kentucky Heritage Council, pp. 14-25.
Jefferies RW (1990) Archaic Period. In D Pollack (ed.): The Archaeology of Kentucky: Past Accomplishments and Future Directions. Frankfort: Kentucky Heritage Council, pp. 143-246.
Jefferies RW (1996) The emergence of long distance exchange networks in the southeastern United States. In KE Sassaman and DG Anderson (eds.): Archaeology of the Mid-Holocene Southeast. Gainesville: University of Florida Press, pp. 222- 234.
Jefferies RW (1997) Middle Archaic bone pins: evidence of mid-holocene regional-scale social groups in the southern Midwest. American Antiquity 62:464-87.
Jefferies RW, and Butler BM (1982) The Carrier Mills Archaeological Project: Human Adaptation in the Saline Valley, Illinois. Research Paper 33. Carbondale: Center for Archaeological Investigations, Southern Illinois University.
189 Jefferies RW, and Lynch BM (1983) Dimensions of Middle Archaic cultural adaptation at the Black Earth site, Saline County, Illinois. In JL Phillips and JA Brown (eds.): Archaic Hunters and Gatherers in the American Midwest. New York: Academic Press, pp. 299-322.
Jefferies RW, Thompson V, and Milner G (2001) Holocene Hunter-Gatherer Settlement Strategies in West Central Kentucky. Paper presented at the 58th Annual Meeting of the Southeastern Archaeological Conference, Chattanooga.
Johnson F (1951) Radiocarbon Dating: A Report on the Program to Aid in the Development of the Method of Dating. Memoirs of the Society for American Archaeology, No. 8. Salt Lake City, Utah.
Johnston FE, and Snow CE (1961) The reassessment of the age and sex of the Indian Knoll skeletal population: Demographic and methodological aspects. American Journal of Physical Anthropology 19:237-244.
Jorde LB (1980) The genetic structure of subdivided human populations. In JH Mielke and MH Crawford (eds.): Current Developments in Anthropological Genetics. Vol. 1: Theory and Methods. New York: Plenum Press, pp. 135-208.
Justice ND (1987) Stone Age Spear Point of the Midcontinental and Eastern united States: A Modern Survey and Reference. Bloomington: Indiana University.
Kelley MA (1980) Disease and Environment: A Comparative Analysis of Three Early American Indian Skeletal Collections. Ph.D. Dissertation, Case Western Reserve University, Cleveland.
Kennedy B (1981) Marriage patterns in an Archaic populations: A study of skeletal remains from Pot au Choix, Newfoundland. Mercury Series No. 104. Ottawa: National Museum of Canada.
King JE, and Allen Jr. WH (1977) A Holocene vegetation record from the Mississippi River Valley, Southeastern Missouri. Quaternary Research 8:307-323.
Knutson IP (1982) Multidimensional Contingency Table Analysis of Dental Disease in Osteological Populations. Ph.D. Dissertation, Department of Anthropology, State University of New York, Buffalo.
Klippel WE, and Morey DF (1986) Contextual and nutritional analysis of freshwater gastropods from Middle Archaic deposits at the Hayes Site, Middle Tennessee. American Antiquity 51:799-813.
190 Konigsberg LW (1987) Population genetic models for interpreting prehistoric intra- cemetery biological variation. Ph.D. Dissertation, Department of Anthropology, Northwestern University.
Konigsberg LW (1988) Migration models of prehistoric postmarital residence. American Journal of Physical Anthropology 77:471-482.
Konigsberg LW (1990a) Temporal aspects of biological distance: Serial correlation and trend in a prehistoric skeletal lineage. American Journal of Physical Anthropology 82:45-52.
Konigsberg LW (1990b) Analysis of prehistoric biological variation under a model of isolation by geographic and temporal distance. Human Biology 62:49-70.
Konigsberg LW, and Buikstra JE (1995) Regional approaches to the investigation of past human biocultural structure. In LA Beck (ed.): Regional Approaches to Mortuary Analysis. New York: Plenum Press, pp. 191-219.
Konigsberg LW, and Frankenberg SR (1992) Estimation of age structure in anthropological demography. American Journal of Physical Anthropology 89:235- 56.
Konigsberg LW, Kohn LAP, and Cheverud JM (1993) Cranial Deformation and Nonmetric Trait Variation. American Journal of Physical Anthropology 90:35-48.
Konigsberg LW, and Ousley SD (1995) Multivariate quantitative genetic of anthropometric traits from the boas data. Human Biology 67:481-498.
Lane RA, and Sublett A (1972) The osteology of social organization. American Antiquity 37:186-201.
Leamy L (1977) Genetic and environmental correlations of morphometric traits in randombred house mice. Evolution 31:357-369.
Lee RB, and DeVore I (eds.) (1968) Man the Hunter. Chicago:Aldine.
Leigh RW (1925) Dental pathology of indian tribes of varied environmental conditions. American Journal of Physical Anthropology 8:179-199.
Legendre P (2000) Corr_permute software. http://www.fas.umontreal.ca/BIOL/Casgrain/en/labo/
Legendre P, and Legendre L (1998) Numerical Ecology (2nd English edition). Elsevier Science BV, Amsterdam.
191 Lewis TMN, and Kneberg M (1959) The Archaic Culture in the middle south. American Antiquity 25:231-244.
Lewis TMN, and Lewis M (1961) Eva: An Archaic Site. Knoxville: The University of Tennessee Press.
Libby WF (1952a) Radiocarbon Dating. Chicago: University of Chicago Press.
Libby WF (1952b) Chicago radiocarbon dates. Science 116:673-681.
Long JK (1964) Multiple-discriminant analysis of Indian crania of the eastern United States. Master’s Thesis, Department of Anthropology, University of Kentucky.
Long JK (1966) A test of multiple-discriminant analysis as a means of determining evolutionary changes and intergroup relationships in physical anthropology. American Anthropologist 68:444-464.
Magennis AL (1977) Middle and Late Archaic Mortuary Patterning : An Example from the Western Tennessee Valley. Master’s Thesis, Department of Anthropology, University of Tennessee, Knoxville.
Mahalanobis PC (1936) On the generalized distance in statistics. Proc Nat Inst Sci India 2:49-55.
Manly BFJ (1986) Randomization and regression methods for testing for associations with geographical, environmental and biological distances between populations. Researches on Population Ecology 28:201-218.
Manly BFJ (1997) Randomization, Bootstrap and Monte Carlo Methods in Biology (Second Edition). London: Chapman and Hall.
Mantel N (1967) The detection of disease clustering and a generalized regression approach. Cancer Research 27:209-220.
Marquardt WH (1977) Current State Research: Kentucky: Shell Mound Archaeological Project. Southeastern Archaeological Conference Newsletter 19:4.
Marquardt WH (1985) Complexity and scale in the study of fisher-gatherer-hunters: An example from the eastern United States. In TD Price and JA Brown (eds.): Prehistoric Hunter-Gatherers: The Emergence of Cultural Complexity. Orlando: Academic Press, pp. 59-98.
192 Marquardt WH, and Watson PJ (1983) The Shell Mound Archaic of western Kentucky. In JL Phillips and JA Brown (eds.): Archaic Hunters and Gatherers in the American Midwest. New York: Academic Press, pp. 323-339.
May AJ (1982) Midden Formation Modeling Using Ethnographic and Archaeological Data: A Trend Surface Analysis of Midden Deposits at the Carlston Annis site (15Bt5), Kentucky. PhD. Dissertation, Department of Anthropology, University of Missouri, Columbia. University Microfilms International, Ann Arbor.
McBride JDE (2000) The Baker Site and the late middle Archaic period in the Green River region of west central Kentucky. Master’s Thesis, Department of Anthropology, University of Kentucky, Lexington.
McFarlan AC (1943) Geology of Kentucky. Lexington: The University of Kentucky.
McKern TW, and Stewart TD (1957) Skeletal age changes in young American males. Technical Report EP-45. Natick: Massachusetts Quartermaster Research and Development Command.
Meindl RS, and Russell KF (1998) Recent advances in method and theory in paleodemography. Annual Reviews in Anthropology 27:375-399
Meindl RS, Mensforth RP, and York HP (2001) Mortality, fertility, and growth in the Late Archaic: The paleodemography of the Ward site. In OH Prufer, SE Pedde, and RS Meindl (eds.): Archaic Transitions in Ohio and Kentucky Prehistory. Kent: Kent State University Press, pp. 87-109.
Mensforth RP (1986) Paleodemography of the Carlston Annis (BT-5) skeletal population. Ph.D. Dissertation, Kent State University, Kent, Ohio.
Mensforth RP (1990) Paleodemography of the Carlston Annis (Bt-5) Late Archaic skeletal population. American Journal of Physical Anthropology 82:81-99.
Mensforth RP (1996) Observations on the antiquity, geographic distribution, and theoretical significance of violent injuries, scalping and other trophy-taking behaviors among Archaic hunter-gatherers of the Eastern United States. American Journal of Physical Anthropology Supplement 22:166 (abstract).
Milner GR, and Jefferies RW (1998) The Read Archaic midden in Kentucky. Southeastern Archaeology 17:117-132.
Milner GR, and Smith VG (1986) New Deal archaeology in Kentucky: Excavations, collections, and research. Lexington: Program for Cultural Resource Assessment, The University of Kentucky.
193
Moore CB (1916) Some aboriginal sites on the Green River, Kentucky. Journal of the Philadelphia Academy of Natural Sciences 16:431-487.
Morey DF, and Crothers GM (1998) Clearing up clouded waters: Palaeoenvironmental analysis of freshwater mussel assemblages from the Green River shell middens, western Kentucky. Journal of Archaeological Science 25:907-926.
Morey DF, Crothers GM, Stein JK, Fenton JP, and Herrmann NP (2002) The fluvial and geomorphic context of Indian Knoll, an Archaic shell midden in west-central Kentucky. Geoarchaeology 17:521 –553.
Morse DF (1967) The Robinson Site and Shell Mound Archaic Culture in the Middle South. Ph.D. Dissertation, The Department of Anthropology, The University of Michigan, Ann Arbor.
Nagy BL (2000) The life left in bones: Evidence of habitual activity patterns in two prehistoric Kentucky populations. Ph.D. Dissertation, Department of Anthropology, Arizona State University.
Nance JD (1987) Research into the prehistory of the lower Tennessee-Cumberland-Ohio region. Southeastern Archaeology 6:93-100.
Nance JD (1988) The Archaic Period in the lower Tennessee-Cumberland-Ohio region. In C Hockensmith, D Pollack and T Sanders (eds.): Paleoindian and Archaic Research in Kentucky. Frankfort, Kentucky Heritage Council, pp. 127-152.
Nash CH (n.d.) Unpublished field notes and site survey records, Watts Bar Reservoir Survey, 1941. Originals on file, Frank H. McClung Museum, University of Tennessee, Knoxville.
Nei M, and Chakravarti A (1977) drift variances of FST and GST statistics obtained from a finite number of isolated populations. Theoretical Population Biology 11:307-325.
Nelson NC (1917) Contributions to the archaeology of Mammoth Cave and its vicinity, Kentucky. Anthropological Papers, vol. 22, part I. New York: American Museum of Anthropology.
Neumann GK (1952) Archaeology and race in the American Indian. In JB Griffin (ed.): Archaeology of Eastern United States. Chicago: University of Chicago Press, pp. 13-34.
194 Newman MT and Snow CE (1942) Preliminary report on the skeletal material from Pickwick Basin, Alabama. In WS Webb and DL DeJarnette: An Archaeological Survey of Pickwick Basin in the Adjacent Portions of the States of Alabama, Mississippi and Tennessee. Bureau of American Ethnology, Bulletin 129. Washington DC: Smithsonian Institution.
Oden NL, and Sokal RR (1992) An investigation of three-matrix permutation tests. Journal of Classification 9:275-290.
Ossenberg NS (1970) The influence of artificial cranial deformation on discontinuous morphological traits. American Journal of Physical Anthropology 38:357-371.
Ossenberg NS (1974) Origins and relationships of Woodland people: The evidence of cranial morphology. In E Johnston (ed.): Aspects of Upper Great Lakes Anthropology: Papers in Honour of Lloyd A. Wilford. St. Paul: Minnesota Historical Society, pp. 15-39.
Ossenberg NS (1979) Cranial discrete traits in population studies. American Journal of Physical Anthropology 45:701-716.
Parmalee P, and Klippel WK (1974) Mussels as a prehistoric food resource. American Antiquity 39:421-434.
Paganini-Hill A, Martin AO, and Spence MA (1981) The S-leut anthropometric traits: genetic analysis. Am J Phys Anthropol. 55:55-67.
Perzigian AJ (1976) The dentition of the Indian Knoll skeletal population: Odontometrics and cusp number. American Journal of Physical Anthropology 44:113-122.
Perzigian AJ (1977) Fluctuating dental asymmetry: variation among skeletal populations. American Journal of Physical Anthropology 47:81-8.
Powell JF (1995) Dental Variation and Biological Affinity Among Middle Holocene Human Populations in North America. Ph.D. Dissertation, Texas A&M University, College Station.
Powell JF, and Neves WA (1999) Craniofacial morphology of the first Americans: Pattern and process in the peopling of the New World. Yearbook of Physical Anthropology 42:153-188.
Powell JF, and Rose J (1999) Report on the Osteological Assessment of the "Kennewick Man" Skeleton (CENWW.97.Kennewick). Report submitted to the Department of the Interior. http://www.cr.nps.gov/aad/kennewick/powell_rose.htm
195 Radcliffe-Brown AR (1931) Social organization of Australian tribes. Oceania 1.
Relethford JH, and Blangero J (1990) Detection of differential gene flow from patterns of quantitative variation. Human Biology 62:5-25.
Relethford JH, Crawford MH, and Blangero J (1997) Genetic drift and gene flow in post- famine Ireland. Human Biology 69:443-465.
Relethford JH, and Harpending HC (1994) Craniometric variation, genetic theory, and modern human origins. American Journal of Physical Anthropology 95:249-270.
Relethford JH, and Lees FC (1982) The use of quantitative traits in the study of human population structure. American Journal of Physical Anthropology 25:113-132.
Rice CL (2001) Pennsylvanian System. Contributions To The Geology of Kentucky, USGS. http://pubs.usgs.gov/prof/p1151h/penn.html
Richtsmeier JT, Cheverud JM, and Buikstra JE (1984) The relationship between cranial metric and nonmetric traits in the rhesus macaques from Cayo Santiago. Am J Phys Anthropol. 64:213-22.
Rolingson MA (1967) Temporal perspective on the Archaic cultures of the Middle Green River region, Kentucky. Ph.D. Dissertation, Department of Anthropology, University of Michigan.
Rolingson MA, and Schwartz DW (1966) Late Paleo-Indian and Early Archaic Manifestations in Western Kentucky. Lexington: University of Kentucky Press.
Rothschild NA (1975) Age and sex, status and role, in prehistoric societies of eastern North America. Ph.D. Dissertation, Department of Anthropology, New York University.
Rothschild NA (1979) Mortuary behavior and social organization at Indian Knoll and Dickson Mounds. American Antiquity 44:658-675.
Rudan I, Rudan P, Chaventre A, Janicijevic B, Milicic J, Smolej NN, and Sujoldzic A (1998) Model-bound and model-free approach in the holistic analysis of population structure: Example from the Island of Pag, Croatia. Homo 49:201-224.
Rudan I, Rudan P, Szirovicza L, Simic D, and Bennett LA (1996) Model-bound and model-free approach in the study of population structure: Example from the Peninsula of Peljesac, Croatia. Homo 47:257-282.
196 Sandford MK, Bogdan GE, and Weaver DS (1998) Treponematosis in the prehistoric Carribbean, North Carolina coast and Kentucky: diagnostic considerations. American Journal of Physical Anthropology Supplement 26:194 (Abstract).
Sanghvi, LD (1953) Comparison of genetical and morphological methods for a study of biological differences. American Journal of Physical Anthropology 11:385-404.
Sanghvi LD, and Balakrishnan V (1972) Comparison of different measures of genetic distance between human populations. In JS Weiner and J Huizinga (eds.): The Assessment of Populations Affinities in Man. Oxford: Clarendon Press, pp. 25-36.
Sassaman KE (1994) Production for exchange in the mid-Holocene Southeast: A Savannah River valley example. Lithic Technology 19:42-51
Sassaman KE, and Anderson DG (1995) Middle and Late Archaic Archaeological Records of South Carolina: A Synthesis for Research and Resource Management (Second Edition). Savannah River Archaeological Research Papers 6. Columbia: Council of South Carolina Professional Archaeologists.
Sassaman KE, and Anderson DG (eds.) (1996) Archaeology of the Mid-Holocene Southeast. Gainesville: University of Florida Press.
Saunders SR (1989) Nonmetric skeletal variation. In MY Iscan and KAR Kennedy (eds.): Reconstruction of Life from the Skeleton. New York: Liss, pp. 95-108.
Sciulli PW (1979) Size and Morphology of the Permanent Dentition in Prehistoric Ohio Valley Amerindians. American Journal of Physical Anthropology 50:615-628.
Sciulli PW (1990) Cranial metric and discrete trait variation and biological differentiation in terminal Late Archaic Ohio: The Duff Site cemetery. American Journal of Physical Anthropology 82:19-30.
Sciulli PW, and Schneider KN (1985) Cranial variation in the terminal Late Archaic of Ohio. Am J Phys Anthropol. 66:429-43.
Sciulli PW, Janini C, and Giesen M (1988) Phenotypic selection on the dentition in a Late Archaic population of Ohio. Am J Phys Anthropol. 76:527-33.
Scott GR, and Turner CG (1997) The Anthropology of Modern Human Teeth. Cambridge: Cambridge University Press.
Self SG, and Leamy L (1978) Heritability of quasi-continuous skeletal traits in a randombred population of house mice. Genetics 88:109-120.
197 Service ER (1971) Primitive Social Organization: An Evolutionary Perspective (2nd Edition). New York: Random House.
Shaw EW (1911) Quaternary lakes in the Mississippi Basin. Journal of Geology 19:481- 491.
Sjøvold T (1973) The occurrence of minor, non-metrical variants in the skeleton and their qualitative treatment for populations comparisons. Homo 24:204-233.
Sjøvold T (1984) A Report on the Heritability of Some Measurements and Non-metric Traits. In GN van Vark and WW Howells (eds.): Multivariate Statistical Methods in Physical Anthropology. Boston: D. Reidel Publishing Company, pp. 223-246.
Skarland I (1939) The skeletal material. In WS Webb and WG Haag: The Chiggerville Site, Site 1, Ohio County, Kentucky. Department of Anthropology and Archaeology Reports, Vol. 4. Lexington: University of Kentucky, pp.28-49.
Smith BD (1986) The Archaeology of the Southeastern United States: From Dalton to de Soto, 10,500-500 B.P. Advances in World Archaeology 5:1-91.
Smith BD (1992) Rivers of Change: Essays on Early Agriculture in Eastern North America. Washington DC: Smithsonian Institution Press.
Smith MO (1982) Patterns of Association Between Oral Health Status and Subsistence: A Study of Aboriginal Skeletal Populations From the Tennessee Valley Area. Ph.D. Dissertation, The University of Tennessee, Knoxville.
Smith MO (1995) Scalping in the Archaic Period: Evidence from the Western Tennessee Valley. Southeastern Archaeology 14:60-68.
Smith MO (1996). Biocultural inquiry into Archaic Period populations of the Southeast: Trauma and occupational stress. In KE Sassaman and DG Anderson (eds.): Archaeology of the Mid-Holocene Southeast. Gainesville: The University of Florida Press, pp. 134-154.
Smouse P, and Long J (1992) Matrix correlation analysis in anthropology and genetics. Yearbook of Physical Anthropology 35:187-213.
Smouse P, Long J, and Sokal R (1986) Multiple regression and correlation extensions of the Mantel test of matrix correspondence. Systematic Zoology 35:627-632.
Sneath PHA, and Sokal RR (1973) Numerical Taxonomy. San Francisco: Freeman.
198 Snow CE (1948) Indian Knoll Skeletons of Site Oh2, Ohio County, Kentucky. Lexington: University of Kentucky.
Sokal RR, and Rolf FJ (1981) Biometry. San Francisco: Freeman.
Sparks CS (2001) Reassessment of Cranial Plasticity in Man: A Modern Critique of changes in Bodily Form of Descendants of Immigrants. Master’s Thesis, Department of Anthropology, University of Tennessee, Knoxville.
Spence MW (1974) Residential practices and the distribution of skeletal traits in Teotihuacan, Mexico. Man 9:262-273.
Squier EG, and Davis EH (1848) Ancient monuments of the Mississippi Valley. Washington : Smithsonian Institution Press (reprinted 1998).
Steadman DW (1997) Population Genetic Analysis of Regional and Interregional Prehistoric Gene Flow in West-central Illinois. Ph.D. Dissertation, Department of Anthropology, University of Chicago.
Steadman DW (2001) Mississippians in motion? A population genetic analysis of interregional gene flow in west-central Illinois. American Journal of Physical Anthropology 114:61-73.
Stefan VH (1999) Craniometric variation and homogeneity in prehistoric/protohistoric Rapa Nui (Easter Island) regional populations. American Journal of Physical Anthropology 110:407-419.
Steele DG (1948) Homogeneity at Indian Knoll. In Snow CE: Indian Knoll Skeletons of Site Oh2, Ohio County, Kentucky. Lexington: University of Kentucky, pp. 492.
Stein J (1980) Geoarchaeology of the Green River Shell Mounds, Kentucky. Ph.D., University of Minnesota, University Microfilms International, Ann Arbor, Michigan.
Stein JK (1982) Geologic Analysis of the Green River Shell Midden. Southeastern Archaeology 1:22-38.
Straw WT (1968) Geomarphology, Hydrology, and Economic Geology of the Ohio River Valley, Mauckport to Cannelton, Indiana. Unpublished Ph.D. Dissertation, Department of Geology, Indiana University.
Styles B, and Klippel WE (1996) Mid-Holocene faunal eploitation in the Southeastern United States. In KE Sassaman and DG Anderson (eds.): Archaeology of the Mid- Holocene Southeast. Gainesville: The University of Florida Press, pp. 115-133.
199 Sullivan NC (1977) The Physical Anthropology of Chiggerville: The Demography and Pathology. Thesis, Western Michigan University, Kalamazoo, Michigan.
Tatarek NE, and Sciulli PW (2000) Comparison of population structure in Ohio's late archaic and late prehistoric periods. American Journal of Physical Anthropology 112:363-76.
Taylor W (1948) A Study in Archaeology. Carbondale: Southern Illinois University Press.
Tompkins RL (1991) Relative Dental Development In Upper Pleistocene Fossil Hominids And Recent Humans. Ph.D. Dissertation, Department of Anthropology, The University of New Mexico.
Wagner G (1996) Botanizing along Green River. In K Carstens and PJ Watson (eds.): Of Cave and Shell Mounds. Tuscaloosa: The University of Alabama Press, pp. 88-93.
Ward SC (n.d.) The dental anthropology of the Middle Green River Shell Middens. In WH Marquardt, PJ Watson and MC Kennedy (eds.): Archaeology of the Middle Green River, Kentucky (In prep).
Waselkov GA (1982) Shellfish Gathering and Shell Midden Archaeology. Ph.D. Dissertation, Department of Anthropology, University of North Carolina, Chapel Hill.
Watson PJ (1969) The Prehistory of Salts Cave, Kentucky. Reports of Investigations No. 16. Springfield: Illinois State Museum.
Watson PJ (1974) Archeology of Mammoth Cave Area. New York: Academic Press.
Watson PJ (1985) The impact of early horticulture in the upland drainages of the midwest and midsouth. In RI Ford (ed.): Prehistoric Food Production in North America. Anthropological Papers No. 75. Ann Arbor: Museum of Anthropology, pp. 73-98.
Weaver DS, Sandford MK, Bogdan GE, and Kissling GE (1998) Treponemotosis in the prehistoric Caribbean, North Carolina coast and Kentucky: paleoepidemiological and evolutionary perspectives. American Journal of Physical Anthropology Supplement 26:227 (Abstract).
Webb WS (1938) An Archaeological Survey of the Norris Basin in Eastern Tennessee. Bureau of American Ethnology, Bulletin 118. Washington D.C.: Smithsonian Institution.
200 Webb WS (1946) Indian Knoll, Site Oh 2, Ohio County, Kentucky: The University of Kentucky Reports in Anthropology and Archaeology. Lexington: Department of Anthropology, pp. 111-365.
Webb WS (1950a) The Read Shell Midden, Site 10, Butler County, Kentucky. The University of Kentucky Department of Anthropology Reports in Anthropology 7:355-401.
Webb WS (1950b) The Carlson Annis Mound, Site 5, Butler County, Kentucky. The University of Kentucky Department of Anthropology Reports in Anthropology 7:265-354.
Webb WS, and DeJarnette DL (1942) An Archaeological Survey of the Pickwick Basin in the Adjacent Portions of the States of Alabama, Mississippi and Tennessee. Bureau of American Ethnology, Bulletin 129. Washington D.C.: Smithsonian Institution.
Webb WS, and Haag WG (1939) The Chiggerville Site, Site 1, Ohio County, Kentucky. University of Kentucky, Department of Anthropology and Archaeology Reports 4:1- 62.
Webb WS, and Haag WG (1940) Cypress Creek Villages, Sites 11 and 12, McLean County, Kentucky. The University of Kentucky Department of Anthropology Reports in Anthropology 4:63-100.
Webb WS, and Haag WG (1947) Archaic Sites in McLean County, Kentucky. The University of Kentucky Department of Anthropology Reports in Anthropology 7:1- 46.
Weiner, JS, and Huizinga J (1972) The Assessment of Populations Affinities in Man. Oxford: Clarendon Press.
White K (1990) An analysis of bone tools from the Carlston Annis site (15BT5), Kentucky. Unpublished Masters Thesis, Department of Anthropology, Washington University, St. Louis.
Wilczak CA (1998) A New Method for Quantifying Musculoskeletal Stress Markers (MSM): A Test of the Relationship Between Enthesis Size and Habitual Activity in Archaeological Populations. Ph. D. Dissertation, Department of Anthropology, Cornell University.
Wilkins GR, Delcourt PA, Delcourt HR, Harrison FW, and Turner MR (1991) Paleoecology of central Kentucky since the last glacial maximum. Quaternary Research 36:224-239.
201
Williams-Blangero S, and Blangero J (1989) Anthropometric variation and the genetic structure of the Jirels of Nepal. Human Biology 61:1-12.
Winterhalder B, and Smith E (eds.)(1981) Hunter-Gatherer Foraging Strategies. Chicago: University of Chicago Press.
Winters HD (1968) Value systems and trade cycles of the Late Archaic in the Midwest. In SR Binford and LR Binford (eds.): New Perspectives in Archaeology. Chicago: Adline, pp. 175-222.
Winters HD (1969) The Riverton Culture: A Second Millennium Occupation in the Central Wabash Valley. Reports of Investigations No. 13. Springfield: Illinois State Museum.
Winters HD (1974) Introduction to the New Edition: Indian Knoll. Knoxville: The University of Tennessee Press, pp. v-xxvii.
Wolf DJ, and Brooks RL (1979) The prehistoric people of the Rosenberger site. In MB Collins (ed.): Excavations at Four Archaic Sites in the Lower Ohio Valley, Jefferson County, Kentucky. Department of Anthropology, Occasional Papers in Anthropology, No. 1. Lexington: University of Kentucky, pp. 899-945.
Wright S (1943) Isolation by distance. Genetics 28:114-138.
Wright S (1951) The genetic structure of populations. Ann. Eugen. Lond. 15:323-354.
Wright S (1969) Evolution and the Genetics of Populations. Vol 2, The Theory of Gene Frequencies. Chicago: University of Chicago.
Wright S (1978) Evolution and the Genetics of Populations. Vol 4, Variability within and among Natural Populations. Chicago: University of Chicago.
Wyckoff LM (1977) The Physical Anthropology of Chiggerville: Biological Relationships and Growth. M.A. Thesis, Western Michigan University, Kalamazoo, Michigan.
Yarnell RA (1969) Contents of human paleofeces. In PJ Watson (ed.): The Prehistory of Salts Cave, Kentucky. Reports of Investigations No. 16. Springfield: Illinois State Museum, pp. 41-54.
Yarnell RA (1974) Plant food and cultivation of the Salts cavers. In PJ Watson (ed.): Archaeology of the Mammoth Cave Area. New York: Academic Press, pp. 113-122.
202
Appendix. Non-metric Trait Coding
203
Appendix. Non-metric Trait Coding
These are the traits scored and specific codes associated with each trait. Traits were then dichotomized based on simple presence of a trait or a combination of traits (e.g. mylohyoid bridging). If the trait is missing a “9” code is given. The trait were dichotomized using the “*” codes.
Midline Traits 1. Metopic Suture: 0. Absent 1. Partial 2. *Complete
2. Apical Ossicle 0. Absent 1. *Present
3. Sagittal Ossicle 0. Absent 1. *Present
4. Inca Bone 0. Absent 1. *Complete, single bone 2. *Bipartite 3. *Tripartite 4. Partial -1,2 and 3 coded as present, 0 and 4 coded as absent
5. Flexure of the Superior Sagittal Sinus 0. Right 1. *Left 2. *Bifurcate
Bilateral Traits 1. Supraorbital Notch : 0. Absent 1. *Present, < ½ occluded 2. *Present, > ½ occluded 3. *Present, degree of occlusion unknown 4. *Multiple notches 204
2. Supraorbital Foramen: 0. Absent 1. *Foramen Present 2. *Multiple Foramen
3. Infraorbital Suture: 0. Absent 1. *Partial 2. *Complete
4. Multiple Infraorbital Foramen 0. Absent 1. Internal division only 2. *Two distinct foramina present 3. *More than two distinct foramina
5. Obelionic Foramen 0. Absent 1. *Foramen present on parietal 2. *Foramen present in suture -Reduced to foramen present/absent
6. Ossicle at Asterion 0. Absent 1. *Present
7. Parietal Notch Bone 0. Present 1. *Absent
8. Lambdoidal Ossicle 0. Absent 1. *Present
9. Masto-Occipital Ossicle 0. Absent 1. *Present
10. Postcondylar Canal 0. Not patent 1. *Patent
205
11. Divided Hypoglossal Foramen 0. Absent 1. Partial, internal surface 2. Partial, within canal 3. *Complete, internal surface 4. *Complete, within canal -Dichotomized by complete division or incomplete division
12. Foramen of Huschke 0. Absent 1. *Present
13. Auditory Exostosis 0. Absent 1. *< 1/3 of canal occluded 2. *1/3 to 2/3 occluded 3. *> 2/3 of canal occluded
14. Mastoid Foramen Location 0. Absent 1. *Temporal bone 2. Sutural 3. *Occipital bone 4. *Both sutural and Temporal bone 5. *Both sutural and Occipital bone 6. *Parietal bone -Dichotomized by expressing this trait as mastoid foramen ex-sutural
15. Mastoid Foramen Number 0. Absent 1. 1 2. *2 3. *More than 2 -Multiple Mastoid Foramen
16. Accessory Lesser Palatine Foramen 0. Absent 1. *Present
17. Foramen Ovale Open 0. Not Open 1. *Open
206
18. Mylohyoid Bridge Location 0. Absent 1. Near Mandibular foramen 2. Center of Groove 3. Bridges 1 and 2 w/ hiatus 4. *Bridge 1 and 2 w/o hiatus -Dichotomized as complete bridging with no hiatus
19. Mylohyoid Bridge Degree 0. Absent 1. Partial 2. *Complete
207 Vita
Nicholas Paul Herrmann attended Washington University of St. Louis where he received a Bachelor of Art in 1988 and Masters of Art in 1990. After a short academic sabbatical in the wilds of New Mexico with Sully and Uncle Bob, he and Sherri Turner, his future wife, arrived at the University of Tennessee-Knoxville in the cold winter of
1993 to pursue his Ph.D. He has since been actively involved in bioarchaeological, forensic and archaeological research throughout the southeastern United States, Texas and Central America. He is currently employed by the Department of Anthropology at
University of Tennessee and is pursuing research in skeletal biology and archaeology.
208