Understanding Culture History Using Topographic Morphometrics Of

Home , Arrowhead, Plainview point

UNDERSTANDING CULTURE HISTORY USING TOPOGRAPHIC MORPHOMETRICS OF

LITHIC PROJECTILE POINTS: PALEOINDIAN CASE STUDIES FROM THE GREAT

PLAINS AND NORTHERN ALASKA

PHILIP ROBERT FISHER

A dissertation submitted in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

WASHINGTON STATE UNIVERSITY Department of Anthropology

MAY 2018

To the Faculty of Washington State University:

The members of the Committee appointed to examine the dissertation of PHILIP

ROBERT FISHER find it satisfactory and recommend that it be accepted.

William Andrefsky, Jr., Ph.D., Chair

Colin Grier, Ph.D.

Luke Premo, Ph.D.

Jade D'Alpoim Guedes, Ph.D.

ii ACKNOWLEDGEMENTS

There are so many people that helped make this dissertation possible, more so then can be

mentioned here. Financial support during my dissertation was provided by the Department of

Anthropology at Washington State University, Ruth Minard, Bill and Patricia Scoales, and is

based upon work supported by the National Science Foundation under Grant No. 1514455.

I need to begin by thanking all of the museums, organizations, repositories, and individuals who allowed me access to the materials scanned in this study. Specifically, this

includes David Wade at the BLM Billings Curation Center, Marybeth Tomka at the Texas

Archaeological Research Laboratory, Scott Shirar at the University of Alaska Museum of the

North, and Jane Lakeman at the NPS Alaska Region Curatorial Center. Robert Ackerman at

WSU for allowing me access to the Spein Mountain material, as well as Robin Mills and

Michael Kunz at the BLM office in Fairbanks for access to, and discussion of, the Mesa site

assemblage. Many thanks are also due to David Meltzer at Southern Methodist University. For

allowing me access to the Gerald Shelton Collection and for the insight and time you took to

discuss my research. Thank you for providing me with an office to both continue 3D scanning,

and to write this dissertation (it can be productive to get out of your pajamas and home office).

No dissertation is possible without those around you who are forced to interact and listen

to you, either in the field or the classroom. I want to thank everyone I have worked with in CRM

that made the work fun and enjoyable. To Scott Carpenter for giving me my first CRM job all those years ago, thank you. To Steve Aaberg who kept bringing me back on projects in Montana,

and to Alan Skinner for giving me the flexibility to split my time at work while allowing me to

take the time needed to finish this dissertation. Matt Boulanger at SMU for talking rocks and

points with me, your help with Momocs in R was invaluable. I made a number of friends at WSU

iii and benefited from the ability to easily find and discuss coursework and research. To the best officemate a lithicist could have, Justin Williams. Jake Adams with whom I share a mutual interest in Alaskan Archaeology, Alaska, and a car for three days driving the Alcan Highway from our homes in Montana. To Kelly Derr and Chris Kiahtipes for regularly meeting with me to discuss this project which helped enormously. In particular Kelly, for your help in thinking up a

number of “fantastic” dissertation topics before this one and for keeping it grounded during those

times it felt like it was crashing. To Kyle Bocinsky and Mark Caudell, my longest roommates

and friends at WSU. I could not ask for better friends and colleagues, thank you for all the good

times. Our ski trips were the best, yet, not frequent enough.

I am ever grateful to the members of my committee who have helped me to develop this

project. The insights, comments, and critiques that you provided made this dissertation what it is.

Thank you Colin Grier, Luke Premo, and Jade D’Alpoim Guedes. Finally, to my advisor and

friend William Andrefsky, Jr. This research would not have been possible without you. Thank

you for your guidance, support, and patience. I have learned so much from you. The best choice

of my career was coming to work with you at WSU.

I have to thank all of my family for their love and support since I was a child. To my

parents Helen Strickland and Jack Fisher for their endless support and push for me to make my

own choices in life. You taught me to love visiting new places and new experiences, even if I

was a somewhat disgruntled child at the time. To my father, Jack Fisher, who throughout this

process has provided unimaginable help and guidance, this would not have been possible without

you. I hated archaeological digs and the smell of dirt when I was a kid, I guess I never stood a

chance. Finally, to Gwen Bakke for always supporting me and having my back during all stages

of this dissertation, your laugh is contagious, thank you.

iv UNDERSTANDING CULTURE HISTORY USING TOPOGRAPHIC MORPHOMETRICS OF

LITHIC PROJECTILE POINTS: PALEOINDIAN CASE STUDIES FROM THE GREAT

PLAINS AND NORTHERN ALASKA

Abstract

by Philip Robert Fisher, Ph.D. Washington State University May 2018

Chair: William Andrefsky, Jr.

The classification of projectile points into types has long been used by archaeologists to develop regional chronologies and serves as a basis by which to explain cultural continuity or change over time. This study incorporates assumptions from social learning and culture- historical transmission of traditions to investigate questions surrounding cultural relatedness through the examination of flake scar patterns from different lithic projectile point types. The conceptual basis of this study is that flintknapping knowledge and technique in small, hunter- gatherer groups is passed from generation to generation through small numbers of master flintknappers. This should result in similar flake scar patterning on projectile points that can be identified using topographic morphometric analysis.

I created a digital topographic morphometric approach to test whether cultural relatedness between past groups or “cultures” can be determined by analyzing three-dimensional models of lithic projectile points to identify variations in flake scar patterning that result from similar or

v different flintknapping techniques. This methodology utilizes high-resolution three-dimensional

imagery to measure variation in flake scar patterns on both faces of a biface. The cross-

sectioning of projectile points at given isoheights records the morphology and patterning of flake

scars that are the result of the flintknapping knowledge and technique that goes into the

manufacture of a projectile point. If these manufacturing techniques are socially learned, and the

production of certain archaeological types of projectile points represent related groups, then

similarities in these flake scar patterns should contain the information to relate groups of people

who share this same knowledge and technique.

This method of viewing culture history is then applied to two Paleoindian case studies.

Projectile point assemblages from the Great Plains and northern Alaska that date to the

Pleistocene-Holocene Transition are analyzed to investigate the cultural relationships among

geographically and temporally similar point types. The results from this study demonstrate that similarities and differences exist in the flake scar patterning of different projectile points types

and can be successfully identified using topographic morphometrics. Results from the case

studies can help archaeologists to better understand the Paleoindian culture histories in these

different regions.

vi TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS ...... iii

ABSTRACT ...... v

LIST OF TABLES ...... ix

LIST OF FIGURES ...... x

CHAPTER

CHAPTER ONE: INTRODUCTION ...... 1

Objectives ...... 3

Conceptual Framework ...... 4

Archaeological Complexes ...... 20

Structure of the Dissertation ...... 37

CHAPTER TWO: THREE-DIMENSIONAL IMAGING AND METHODS ...... 38

NextEngine Ultra HD 3D Scanner ...... 38

Image Processing ...... 43

Image Analysis...... 46

Topographic Morphometrics and Lithic Expectations ...... 51

CHAPTER THREE: TESTING TOPOGRAPHIC MORPHOMETRICS ...... 55

Application of Topographic Morphometrics on the Assemblage ...... 63

Complete ±¼ Measurement ...... 68

Complete ±⅓ Measurement ...... 84

Haft ±¼ Measurement ...... 91

Topographic Morphometric Results and Conclusions ...... 101

vii CHAPTER FOUR: LATE PALEOINDIAN ON THE GREAT PLAINS ...... 108

Bivariate Analysis ...... 110

Great Plains Late Paleoindian Topographic Morphometrics ...... 114

Great Plains Late Paleoindian Conclusions ...... 120

CHAPTER FIVE: NORTHERN PALEOINDIAN IN ALASKA ...... 125

Paleoecological Background ...... 136

Northern Paleoindian Topographic Morphometrics ...... 157

Northern Paleoindian Conclusions ...... 167

CHAPTER SIX: SUMMARY AND CONCLUSIONS ...... 171

Future Work ...... 176

REFERENCES ...... 181

APPENDIX

A: PROJECTILE POINT INFORMATION ...... 198

B: VEGETATION CLASSES ...... 205

C: PALEOECOLOGICAL PROJECTION RESULTS ...... 207

viii LIST OF TABLES

Page

Table 1.1. Projectile Point Types Analyzed in this Study ...... 22

Table 3.1. Overview of Archaeological Sites Analyzed in the Plains Study ...... 62

Table 3.2. Plains Complete ±¼ LDA Predictions and Confusion Matrix ...... 74

Table 3.3. Plains Complete ±⅓ LDA Predictions and Confusion Matrix ...... 87

Table 3.4. Plains Haft ±¼ LDA Predictions and Confusion Matrix ...... 95

Table 3.5. LDA Reclassification Accuracy by Method and Isoheight ...... 105

Table 4.1. Plains Late Paleoindian LDA Predictions and Confusion Matrix ...... 118

Table 4.2. Chronological Range of Late Paleoindian Types ...... 120

Table 5.1. Overview of Archaeological Sites from Alaska ...... 136

Table 5.2. Projected Landscapes during the Younger Dryas and Pollen Core Signals ...... 148

Table 5.3. Mesa Type LDA Predictions and Confusion Matrix ...... 161

Table 5.4. Sluiceway Type LDA Predictions and Confusion Matrix ...... 164

Table 5.5. Northern Paleoindian LDA Predictions and Confusion Matrix ...... 165

Table 5.6. Chronological Range of Northern Paleoindian Types ...... 168

ix LIST OF FIGURES

Page

Figure 1.1. Schematic Depicting Learning Lineages and Traditions ...... 19

Figure 1.2. Distribution of Late Paleoindian Types Across the Great Plains ...... 25

Figure 1.3. Great Plains Paleoindian Projectile Points ...... 27

Figure 1.4. Variation seen in the Nebo Hill/Sedalia Projectile Points ...... 30

Figure 1.5. Generalized Distribution of Artifact Complexes in Alaska ...... 32

Figure 1.6. Northern Paleoindian Projectile Points ...... 33

Figure 2.1. Next Engine Ultra HD 3D Scanner ...... 39

Figure 2.2. The Addition of a Split Plane to a 3D Object...... 45

Figure 2.3. The Creation of a Chain Code in Elliptical Fourier Analysis...... 48

Figure 3.1. Location of Plains Late Paleoindian Archaeological Sites ...... 56

Figure 3.2. Isocontour Height Measurement ...... 64

Figure 3.3. Showing Complete Versus the “Digitally Broken” Haft Isocontours ...... 66

Figure 3.4. Sample of Haft Isocontours of the Different Projectile Point Types ...... 67

Figure 3.5. PCA Plot of Complete ±¼ Projectile Points ...... 69

Figure 3.6. Complete ±¼ Principle Component Contribution and Scree Plots ...... 71

Figure 3.7. LDA Plot of Complete ±¼ Projectile Points ...... 73

Figure 3.8. PCA Plot of the Complete ±¼ Mill Iron Goshen Projectile Points ...... 75

Figure 3.9. PCA Plot of the Complete ±¼ Plainview and Bonfire Shelter Projectile Points ...... 76

Figure 3.10. PCA Plot of the Complete ±¼ Milnesand Projectile Points ...... 77

Figure 3.11. PCA plot of the Complete ±¼ Nebo Hill/Sedalia Projectile Points ...... 77

x Figure 3.12. Depiction of the “Non-Symmetry” Isocontours in Point Faces ...... 79

Figure 3.13. Effects of Curation on the Isocontours ...... 80

Figure 3.14. The “Dog Legging” Effect of a Non-Symmetrical Biface on the Isocontours ...... 81

Figure 3.15. Revisiting the PCA Plot of Complete ±¼ Projectile Points ...... 83

Figure 3.16. Revisiting the LDA Plot of Complete ±¼ Projectile Points ...... 83

Figure 3.17. PCA Plot of Complete ±⅓ Projectile Points ...... 86

Figure 3.18. LDA Plot of Complete ±⅓ Projectile Points ...... 86

Figure 3.19. PCA Plot of the Complete ±⅓ Plainview and Bonfire Shelter Projectile Points ...... 88

Figure 3.20. PCA Plot of the Complete ±⅓ Nebo Hill/Sedalia Projectile Points ...... 89

Figure 3.21. Revisiting the PCA Plot of the Complete ±⅓ Projectile Points ...... 90

Figure 3.22. Revisiting the LDA Plot of the Complete ±⅓ Projectile Points ...... 90

Figure 3.23. PCA Plot of Haft ±¼ Projectile Points ...... 93

Figure 3.24. LDA Plot of Haft ±¼ Projectile Points ...... 94

Figure 3.25. PCA Plot of the Haft ±¼ Mill Iron Goshen Projectile Points ...... 97

Figure 3.26. PCA Plot of the Haft ±¼ Milnesand Projectile Points ...... 98

Figure 3.27. Revisiting the PCA Plot of Haft ±¼ Projectile Points ...... 99

Figure 3.28. Revisiting the PCA Plot of Haft ±¼ Projectile Points ...... 100

Figure 3.29. The Characteristics that can Affect the Isocontours ...... 104

Figure 4.1. Scatterplot of Maximum Width and Thickness of Great Plains Late Paleoindian ....111

Figure 4.2. PCA Plot of the Plains Late Paleoindian ...... 115

Figure 4.3. LDA Plot of the Plains Late Paleoindian ...... 117

Figure 4.4. Geographic and Temporal Distribution of the Goshen and Plainview Types ...... 119

Figure 5.1. Location of Northern Paleoindian Sites in Alaska ...... 130

xi Figure 5.2. Vegetation Projection of the Younger Dryas in Northern Alaska ...... 145

Figure 5.3 Vegetation Projections with Pollen Core Data ...... 147

Figure 5.4. Vegetation Projection of the Holocene Thermal Maximum ...... 150

Figure 5.5. Northern Alaska Vegetation Change from the Bølling-Allerød to Younger Dryas ..151

Figure 5.6. Northern Alaska Vegetation Change from the YD to the HTM ...... 152

Figure 5.7. Vegetation Change from the YD to the HTM with Archaeological Sites ...... 154

Figure 5.8. Scatterplot of Maximum Thickness and Width for the Mesa and Sluiceway Points 158

Figure 5.9. Mesa Type PCA Plot ...... 159

Figure 5.10. Mesa Type LDA Plot...... 160

Figure 5.11. Sluiceway Type PCA Plot PCA ...... 162

Figure 5.12. Sluiceway Type LDA Plot...... 163

Figure 5.13. PCA Plot of Northern Paleoindian Projectile Points ...... 165

xii

Dedication

for my parents Helen Strickland and Jack Fisher

and for my grandparents

Muriel and John Strickland, Dorothy and John Fisher

xiii

CHAPTER ONE: INTRODUCTION

Variation in chipped stone projectile points from the archaeological record of North

America has been used regionally by archaeologists to address questions relating to stylistic and

functional change through time and across space. However, to go further and ask what these typological changes mean culturally is much more challenging. Do chronologically sequential projectile point types in a region represent cultural adaptation by the same group of people, or do they represent population replacement? When lithics are the sole archaeological line of evidence, addressing this question can be difficult. The development of novel methodological approaches that utilize new technology is critical to address these age-old archaeological questions.

The focus of this study is to employ a relatively new methodology using three- dimensional (3D) digital imagery to morphometrically test whether cultural relatedness between past groups or “cultures” can be determined through topographic flake scar patterns that result from different flintknapping techniques on the surface of projectile points. This study incorporates assumptions from social learning, culture historical transmission of traditions, and flintknapping to investigate the question of cultural relatedness. A longstanding and central focus in the analysis of chipped stone projectile points in North American archaeology has been their typological classification as cultural and temporal markers (Bettinger and Eerkens 1999). The

classification of projectile points into types has been used to develop regional chronologies by

archaeologists and serves as the basis by which to examine cultural continuity and change over time. The research presented here extends those objectives.

Projectile point typologies are linked to lithic assemblages and the people who manufactured them to infer cultural association (Barton 1997). For example, the way in which

Clovis and Folsom cultures are defined is based largely on the distinctive morphology of the

projectile points themselves. When geographic distribution and temporal duration of projectile

points are taken into account, questions such as in-place cultural evolution, the interaction or

migration of peoples, or the role seasonality plays in lithic technological organization and

changes in projectile point types can be addressed productively.

This study examines assemblages of five Late Paleoindian unfluted, lanceolate projectile

point types from two separate geographic areas dating to the Pleistocene-Holocene Transition

(PHT): the Great Plains of North America, and northern Alaska. From the Great Plains the Late

Paleoindian assemblages include the Goshen (northern Plains), Plainview (southern Plains), and

Milnesand (southern Plains) projectile point types. In Alaska, the Northern Paleoindian projectile

point assemblage focuses on the Sluiceway (western Brooks Range) and Mesa (central Brooks

Range) types. The temporal focus of this Paleoindian research is from around 11,200 to ~ 10,000

14C BP (13,200 to ~11,000 cal yrs BP).

During the PHT the climate was characterized by three different paleoclimatic events that had different temperature and precipitation characteristics. These were the warmer and wetter

Bølling-Allerød (BA) interstadial (~14,500 to ~12,900 cal yrs BP), the cooler and drier Younger

Dryas (YD) stadial (~12,800 to 11,500 cal yrs BP), and the warmer and wetter Holocene

Thermal Maximum (HTM) (~11,000 to ~9,500 cal yrs BP) (Hoffecker and Elias 2007; Mann et al. 2002; Mason et al. 2001; Peteet 1995). The climate changes associated with each would have affected the local ecology of both the Great Plains and northern Alaska by influencing the plant communities present and their geographic distribution. In particular, the YD climatic reversal was felt strongly at northern latitudes and affected Alaska to varying degrees based on geographic location, more so than on the Great Plains of North America where the YD translated into an expansion of grasslands (Guthrie 1984; Meltzer 2009). Changes at this level would have

affected highly mobile Paleoindian hunter-gatherers in terms of the variety of different prey

species available and their population numbers. This raises the following question: are

temporally similar and/or successive projectile point types within the same or neighboring

geographic areas the result of in situ adaptation by a culture or group of people, or do they

represent a population replacement?

Objectives

This project has two objectives. The first is to test the application of 3D imagery in

developing a topographic morphometric methodology that can be used to answer questions

concerning cultural relationships based on flake scar patterns from projectile points in

assemblages from the archaeological record. To test this methodology a pilot study was

conducted on three Late Paleoindian assemblages and a single Middle Archaic assemblage of

unfluted lanceolate projectile points to see if similarities and differences were visible in the flake

scar patterns of these point types that vary in space and time. This novel methodology utilizes

high resolution 3D imagery to measure variation in flake scar patterns (like topographic lines on

a map) on both sides of a biface. These topographic measurements are used to look for

similarities and differences in flintknapping techniques from flake scar patterns that result when

a flintknapper makes a tool. The goal is to ascertain whether established projectile point types

share similar manufacturing techniques. As projectile point types are often used to define

archaeological complexes or “cultures,” these flake scar patterns that result from manufacture

can be used to assess relationships between projectile point types, and by extension “cultures,” based on similarities and differences in their flake scar patterning.

My second objective is to apply this methodology to understanding the distribution and relationships of projectile point types in two Paleoindian case studies: 1) Late Paleoindian groups on the Great Plains, and 2) Northern Paleoindians in the Brooks Range, Alaska. Topographic morphometrics can be used to analyze these projectile point assemblages to look for cultural relatedness, which in turn can be used to help infer the culture histories of these regions. This novel technique is first applied to Late Paleoindian projectile points from the Great Plains.

Topographic morphometrics are then applied to Northern Paleoindian projectile point assemblages in Alaska and compared to paleoecological reconstructions of the Brooks Range mountains where these sites are located.

Conceptual Framework

Chipped stone tool production is a socially learned process that requires knowledge, technique, skill, and practice. As humans, we live in a culturally or socially constructed environment that shapes our learning (Stout 2005). Technological knowledge of tool manufacture is an ongoing process within both the physical and cultural environments. An individual’s skill, the cognitive ability to picture the necessary steps needed to make a tool, and the motor abilities to actually complete the job are acquired over time (Bleed 2008; Stout 2005).

This is especially true for intricate technologies such as projectile point manufacture. The knowledge and skill required to flintknap a projectile point, embodied in the way by which an individual learned how to flintknap, is passed on in their social environment or culture.

The conceptual basis for this study lies in social learning and expectations as to how flintknapping knowledge and technique present themselves within a culture. This, in turn, comes both from basic assumptions of cultural transmission theory and assumptions about culture

historical transmission of traditions. In recent years anthropologists, including archaeologists, have become increasingly interested in understanding culture change through an evolutionary framework, in particular cultural transmission theory (or dual inheritance theory), that builds on the culture history theory of early 20th century anthropology and Darwinian evolution (Boyd and

Richerson 1985; Cavalli-Sforza and Feldman 1981; Eerkens and Lipo 2005).

Cultural Transmission Theory

Cultural transmission is the social reproduction of an individual’s or a group’s knowledge, and can range from social taboos to technology and tool manufacture (Mesoudi

2011; Mesoudi et al. 2004; Whiten et al. 2011). This transmission of cultural ideas, knowledge, and tools (to name a few examples) can be thought of in a genetic sense as the way in which cultural traits are passed between generations. However, unlike genetics where DNA can only be passed from parents to offspring, material culture can be transmitted in a number of ways to a variety of different and even unrelated (genetically) individuals. This transmission is subject to selective forces such as the environment and population numbers (just as in genetics) (Boyd and

Richerson 1985; Cavalli-Sforza and Feldman 1981; Hewlett & Cavalli-Sforza 1986; Mesoudi

2011). Because many studies of cultural transmission have already been produced, this section will provide a brief overview of cultural transmission theory that is based on the original works of Boyd and Richerson (1985) and Cavalli-Sforza and Feldman (1981).

Evolutionary theory is applied to archaeology to better understand variation in material culture. The focus of cultural transmission in archaeology has been not so much on understanding patterns in the variation of material culture, but rather on recognizing the processes that influence the transmission of cultural traits and how this produces variation

(Eerkens and Lipo 2005). Items of material culture, including archaeological artifacts, exhibit a

number of quantifiable traits, such as the length of a projectile point or the rim width of a

ceramic vessel. These traits, which can be modeled and tested, are the units in which cultural

transmission studies are particularly interested (Eerkens and Lipo 2005).

Take, for example, projectile points. Cultural transmission can be used to examine the

relationship between varying projectile point shapes seen in the archaeological record, and the

processes that lead to this variation. A projectile point is copied or modified whenever a new

point is manufactured within a social group or culture. If, for whatever reason (including

deliberate intent by an individual), the point is not reproduced accurately variation is introduced.

This results in the variation seen in the archaeological record as differing projectile point shapes

(Bettinger and Eerkens 1999; Lyman et al. 2008). This variation is a result of a cultural trait

coming under selective pressures during transmission (Lyman et al. 2008) or as a result of

neutral transmission (Neiman 1995).

Within a culture, individuals in a social group learn through a number of processes.

Modes of learning can occur through a parent (vertical transmission), an older member of the

group (oblique transmission), or from someone of the same age and status or anyone from

outside the group (horizontal transmission) (Boyd & Richerson 1985; Hewlett & Cavalli-Sforza

1986). In addition, the scope in which people learn varies. In a one-to-one context, a person learns from one other person in the group. In a one-to-many setting, one person teaches a large number of individuals (Mesoudi 2011). These are the ways in which cultural traits can be passed on from generation to generation. Just as in evolutionary theory, the cultural transmission of these traits are subject to many forces (Bettinger and Eerkens 1999; Cavalli-Sforza and Feldman

1981).

Forces that affect cultural evolution are natural selection, cultural drift, random variation,

guided variation, and biased transmission. Natural selection, just as in biological evolution, is the

selection of the trait best suited for the environment, which influences survival and reproduction.

Cultural drift is analogous to genetic drift in that it is random change in frequencies due to causes

such as copy error, local extinctions, or bottlenecks. Guided variation is defined as the

modification of a behavior or trait by an individual in response to current environmental

conditions. This is akin to Lamarckian evolution in that humans can change the phenotype of a

trait, which can increase variation if it is the only force at work (Boyd & Richerson 1985).

Random variation, is random innovation or error when copying a cultural behavior or trait (Boyd

& Richerson 1985). Finally, biased transmission refers to any number of factors that potentially

affect which cultural traits to copy or whom traits are copied from (Boyd & Richerson 1985).

There are three types of biased transmission: direct, indirect, and frequency dependent.

With direct biased transmission, an individual assesses a trait to decide what is best suited, or has

the best fitness, for prevailing environmental conditions. This process can result in increased or

decreased variation. Indirect biased transmission does not select traits based on the fitness of the trait but rather on the perceived fitness (Boyd & Richerson 1985). This can include similarity bias (a trait is selected because someone similar to you uses it), prestige bias (a trait is selected because it is used by someone of importance in a group), or success bias (a trait is selected because someone successful uses it) (Henrich and McElreath 2003). Indirect bias reduces variability if the trait is being used by everyone. Finally, frequency dependent biased transmission is based on the proportion of traits in a population. In conformist bias the most

frequent trait is adopted. This is most common when information on the environment is accurate and guided variation or innovation is costly. Conformist bias decreases variation. The opposite of

conformist bias is anti-conformist bias. In this situation a trait is adopted because it is not popular

(Boyd & Richerson 1985). This can lead to an increase in variation. The ways in which people

learn and transmit traits affect the amount of variation in material culture that is exhibited by a

population. Individual learning will produce different patterns of variation than that produced by

social learning, in which traits are adopted from others (Boyd and Richerson 1985)

Viewing Flake Scar Patterning through a Cultural Transmission Lens

Although this study does not use cultural transmission theory to explicitly model the

trait(s), mode of transmission, or selective force behind the variation in flake scar patterning, it is

worth considering how cultural transmission theory would view the transmission of

flintknapping knowledge and technique through flake scar patterning. In the future, it would be

beneficial to build on this study by incorporating models of cultural transmission that delve

deeper into the forces that affect cultural evolution and that could model variation in flake scar

patterning. That undertaking, however, lies beyond the scope of this project. Nevertheless,

viewing flake scar patterning through a lens of cultural transmission helps to theoretically ground

the assumptions being made in this study that are based in social learning.

This study compares the flake scar patterning of bifacial projectile points. These flake

scars are produced during the final stages of tool production. The final stages are the hardest and most complex for a flintknapper to carry out, and require the skill and training of an expert

(Andrefsky 2006; Apel 2008; Bamforth and Finlay 2008). Therefore, a comparison of flake scar reduction patterns between projectile points has the potential to reveal the most sensitive information on culturally transmitted flintknapping knowledge and technique. These results can be used to make inferences about whether different projectile point types are related or whether

different types are unrelated to one another and represent the products of technologically and culturally separate human populations.

Examples of cultural transmission studies using lithics are growing in number. Cultural transmission studies of lithics typically use metric measurements that relate to projectile point outline morphology when examining the cultural transmission of traits (Bettinger and Eerkens

1999; Lyman et al. 2008; Neiman 1995; O’Brien et al. 2001, 2014; VanPool et al. 2015).

However, flake scar patterning is more abstract and is far more complicated to model than metric dimensions. A recent study conducted by Eren et al. (2015) is of direct importance to my research. Eren et al. (2015) argue that differences in Clovis projectile point morphology are due to cultural drift. The authors held geographic area (Ohio, Indiana, and Kentucky), environment

(including target prey species), and raw material sources constant and found that drift best explains morphological differences in shape among Clovis points made from three raw material sources. The authors found that although the shape of Clovis points differed between the three sources, the flaking did not. They argue that cultural drift is the best explanation for morphological difference in shape when the environment can be held constant. This supports the notion that Clovis represents a distinct culture that shared a similar flintknapping knowledge and technique that is socially learned, even though the shape of Clovis points may differ. In a separate study of Clovis points, Sholts et al. (2012), using a 3D scanner to examine flake scar patterning in the same manner as this study, argue that Clovis represents a shared flintknapping culture that is distinguishable from other large lanceolate bifaces. The findings by Sholts et al.

(2012) are important because they indicate that flake scar patterning can be used to successfully identify similar flintknapping knowledge and technique on projectile points and that this

information is useful for studying the relatedness of different projectile point types to one

another.

It is important to consider whether traits, in this case flake scar patterning, are

homologous or analogous. Homologous traits are those that are similar due to a shared ancestry.

Analogous traits are those that share a similar function but arose through convergent evolution

and do not have a shared ancestry (Lyman 2001). It can be difficult to separate traits that are only

neutral (stylistic) from those that are only functional with projectile points. When analyzing

projectile points from the perspective of cultural transmission, it is important to identify the

differences between selective and neutral transmission that may affect traits. The selective

pressures that affect a trait will influence how that trait is transmitted. Functional traits are those

that are under environmental pressure and that influence the success of a projectile point as a

hunting weapon, usually in terms of caloric return that affects fitness and survival (Eerkens and

Lipo 2005; VanPool 2001). In the context of this study, two cultures could share similar traits in

their material culture, such as the morphology of projectile points, because these traits are the

result of selection on function in the successful killing of specific prey. Such traits are shared not because they are from the same culture, but rather because they have a similar function, which can lead to convergence and analogous traits (Lyman 2001; VanPool 2001). These are not

informative traits to examine for tracking cultural transmission as they have the potential to

display similar patterns between unrelated cultures (Eerkens and Lipo 2005). The study of

functional traits of projectile points, thus, is not ideal for achieving the goals of my study. For

this reason, I specifically do not look at functional traits of projectile points. Instead, I examine

flake scar patterning on non-fluted projectile points and I assume this characteristic is not likely

subject to functional selection. Because flake scar patterning probably is not subject to functional

selection, it is a trait that is free to vary, and as such could be argued to be a neutral trait. Neutral

or stylistic traits are those that do not influence the function of an artifact or tool. That is, they

can be defined as a non-adaptive culturally transmitted trait (Lipo and Madsen 2001).

To begin with, it is important to differentiate the facets that go into the manufacture of

lithic projectile points as this will affect how flake scar patterning is transmitted. Flintknapping is

imitation not emulation (Lycett et al. 2015). Whereas emulation is copying to match a result,

imitation is not only copying to match the result but also the behaviors that go into the

production of the object that is being copied (Lycett et al. 2015). Flintknapping knowledge (the

learned technique of how to reduce a chunk of lithic raw material in a purposeful way) differs

from the various templates of what a projectile can look like (that is, the morphological outline

shape of a projectile point). As flintknapping is a reductive process, this flintknapping

knowledge and technique manifests as the flake scar patterning on the faces of projectile points

as they represent the last stages of a complex series of manufacturing steps and meaningful flake

removal. Whereas the size and shape of a projectile point are more likely to be transmitted

horizontally than is the learned base knowledge on how to actually flintknap. In addition, size

and shape are subject to more selective pressures, such as guided variation, because size and

shape are likely to be functional traits. It is important to remember that the knowledge of how to

flintknap is different from the manufacture of a finished projectile point of a certain size and

shape, as it is flintknapping knowledge that is passed from master to apprentice within one’s own

social group. Paleoindians in North America would have consisted of small, band-sized groups

of hunter-gatherers in low population densities across the landscape. Within these small groups a

small number of master flintknappers, or perhaps just a single master, would teach younger generations through vertical or oblique transmission. This is not to say that horizontal

transmission did not occur in Paleoindian groups, but the knowledge and technique of how to

flintknap (which produces the flake scar patterning on a projectile point) would have been

transmitted vertically or obliquely through master flintknappers.

The knowledge one is taught about how to flintknap, will remain in place when

manufacturing projectile points. The expectation is that the flake scar patterning by an individual

will remain the same when producing points that have different morphologies, because the

knowledge of how to flintknap is an ingrained trait within a social group. Therefore, flake scar

patterning could be argued to be a socially transmitted trait. Social traits are those that have

social pressures that influence why they are used, and could be viewed as ethnic markers by

which groups identify themselves. In the case of projectile points, the flintknapping knowledge

and technique that produces the flake scar patterning could be used to identify a social group or

culture. How a projectile point is made could be dependent on the culture in which it is

transmitted. If flake scar patterning is not a functional or an analogous trait, it can be used to

study different cultural traditions such as projectile point types.

Flintknapping knowledge is transmitted through vertical and oblique modes from a

master to an apprentice. The social learning environment of this specific trait (flintknapping knowledge and resulting flake scar patterning) would have been subject to frequency-dependent conformist transmission or possibly indirect prestige biased transmission. If flake scar patterns undergo social transmission, in a small population of hunter-gatherers (such as Paleoindians) with master flintknappers, conformist or prestige bias (Boyd and Richerson 1985; Mesoudi

2011) would select for the most frequent trait or the trait used by someone of importance, respectively. This would lead to a decrease in variation. Conformist and prestige bias is dependent on population and it is important to distinguish how conformist biased transmission

affects traits depending on the size of the population. Flake scar patterning on these Paleoindian

projectile points occurred in small groups with the information on flintknapping knowledge

controlled by a small number of master flintknappers who teach subsequent generations.

Therefore, among small band sized Paleoindian groups, the expectation would be that the

knowledge and techniques used to manufacture these points would change at a much slower rate

and variation would be low. This low variation would enhance the archaeological visibility of

similarities and differences in flake scar patterns when comparing projectile point assemblages.

As previously discussed, the application of cultural transmission theory to the flake scar patterning on projectile points would entail the modeling of trait selection and transmission. For these reasons this study is not focusing on, nor attempting to model, the evolutionary component of cultural transmission theory. While this study draws on social learning and the transmission of flintknapping knowledge within small hunter-gatherer groups, it does not attempt to model or test the mode of transmission affecting flake scar patterning variation, the trait under selection, or the forces acting on the transmission of cultural knowledge. I do not model the cultural transmission of flake scar patterns in this study. Rather, I borrow assumptions about social learning from culture transmission theory. Assumptions pertaining to social learning from cultural transmission are combined with assumptions from culture historical transmission of traditions to make inferences about the relatedness of Paleoindian hunter-gatherer groups. The primary assumption I make is that flake scar patterning on projectile points from closely related groups will look more similar than flake scar patterning on projectile points of unrelated groups.

This assumption is based on how flintknapping knowledge and technique are socially learned in small mobile groups of Paleoindian hunter-gatherers. In this study I employ a broad, or general, level of social learning that is often used by archaeologists when building or refining the culture

history of a region. This study employs a fine-grained technique for looking at similarities and

differences in material culture among hunter-gatherers, specifically at similarities and differences

in flake scar patterning on projectile points under the assumption that flintknapping is learned

from others within a related group. It is important to stress that I do not attempt to specify how

flake scar patterning was selected in an evolutionary sense, and I am not attempting to model

rates of learning or how, specifically, flintknapping knowledge and technique were transmitted

with respect to flake scar patterns. Rather, what I have presented above is a discussion about

flake scar patterning as it might be viewed through a lens of cultural transmission.

Assumptions of Flake Scar Patterning in this Study

To become a skilled flintknapper requires substantial training. This training occurs in the context of social groups (Bamforth and Finlay 2008) to reduce risk associated with individual learning, such as the time lost unnecessarily to experimentation and to knapping hazards associated with the natural qualities of knappable raw material (Lycett et al. 2015). Developing competence in flintknapping takes considerable time, practice, and oversight from an expert

teacher or master. Stone tool production, as documented ethnographically, is done by an expert

or master who in turn passes their knowledge and skill to an apprentice (Grimm 2000; Stout

2005).

Differences in skill are apparent amongst modern day flintknappers. Those new to

flintknapping have a hard time producing even a basic Acheulian hand axe (Winton 2005). The

final stages of formal biface tool production, such as projectile points, are more complex than

hand axe production and are the hardest and most difficult to master. Although novice and

apprentice flintknappers can produce basic hand axes and biface preforms, the knowledge and

skill required to finalize a bifacial tool lies in the hand and mind of an expert (Andrefsky 2006;

Apel 2008; Bamforth and Finlay 2008; Ferguson 2008). Stone tool manufacture could potentially lead to many different techniques and styles of flintknapping that would be visible in the archaeological record (e.g. Boyd & Richerson 1985). In an experimental study examining social learning and flintknapping instructions on biface production, Putt et al. (2014) demonstrate how two different groups, one given verbal instructions and one allowed only to watch, produced bifaces in two fundamentally different ways. This demonstrates that people copying what an artifact looks like will produce stone tools in different way than those given direct instruction.

Strong correlations between flintknapping technique and flake shape have been observed

(Dibble and Rezek 2009). Experimental studies have also been conducted that examine the debitage produced by individual flintknappers manufacturing similar tools. Williams and

Andrefsky (2011) analyzed debitage produced by five different knappers making multi- directional cores and early stage bifaces. Results indicate that there are highly variable differences in the debitage produced by these individuals. These differences in debitage suggest that each individual used a different process to produce similar tools, a result of differentially learned techniques of tool manufacture by the individual knappers. A similar pattern was seen in the analysis of lithic debitage from two archaeological middens at Cox Ranch Pueblo, which led

Williams et al. (2013) to conclude that two ethnic groups co-resided at the site, each with its own flintknapping technique.

Stone tool manufacture is a reductive process. The technique used to manufacture any lithic tool is a learned tradition-whether it be reducing a core for flakes, making a scraper, or producing a bifacial projectile point. Formal bifacial tools, such as projectile points, are the most complex artifacts to manufacture, and the flintknapping technique used in their manufacture will

dictate flake scar patterns. There are a number of different manufacturing characteristics that in

any combination can lead to any number of different projectile types, all of which influence what

the flake scar patterning on a projectile point will look like. These include flaking styles and

morphological characteristics. Visible on the bifacial surfaces of projectile points are the various

flaking styles which Crabtree (1982:51) defines as regular parallel, less regular parallel,

collateral, diagonal parallel, oblique transverse, double diagonal or chevron, and random or non-

patterned. Other flaking styles than those mentioned above exist, such as horizontal transverse

and parallel transverse. Morphological characteristics of projectile points include, for example,

maximum thickness, maximum length, medial length, haft width or maximum width, location of

the maximum width along the long axis of the specimen, midpoint width, haft length, haft

constriction, base width, base thickness, basal shape and maximum indentation depth and width,

tip shape, and tang shape (O’Brien et al. 2014; VanPool et al. 2015). When manufacturing a

projectile point, a flintknapper can pull from any of the flaking styles described above, or more.

The flaking patterns, in conjunction with morphological characteristics of a projectile point, will

influence the flake scar pattern on a projectile point based on the learned techniques used to

manufacture them.

Related groups of people or cultures should have similar tools that differ from those of

other, unrelated cultural groups (Apel 2008; Eerkens and Lipo 2007). Wiessner (1983) conducted

an ethnographic survey of San (Bushman) arrows from southern Africa and was able to differentiate regionally separated groups based on attributes of their arrows. Related groups of

people should have not only similar tool kits, but also similar methods of reducing lithic material

to produce projectile points. Such reduction methods, in turn, can be identified on the artifacts

themselves. Based on the expectation that similar groups of people should share similar

knowledge of stone tool production, similarities in flake scar patterning on projectile point types can inform us about the relationships between the prehistoric people who produced them.

In stone tool manufacture, individuals learn directly from those around them in their social group. It is because of this that different generations of flintknappers, especially from small hunter-gatherer groups, can be expected to retain the same knowledge and skill in stone tool manufacture as previous generations, thus fashioning lithic projectile points in a similar manner (Apel 2008; Eerkens and Lipo 2007; Grimm 2000). Descendent populations from the same group of people should have a similar “cultural” knowledge that can be examined by looking at the flake scar patterning on projectile points resulting from production.

The terms “related groups” or “related cultures” are used here interchangeably, and are defined in this study as people who share both a common knowledge of flintknapping and an ancestral flintknapping teacher(s) or instructor(s) in their flintknapping tree. This is what

VanPool et al. (2008:77) refer to as “learning lineages” (Figure 1.1). The knowledge of how to flintknap, and the ways to produce different types of projectile points, was transmitted socially within and amongst groups of people who learned and shared the same information. “Related groups” is not meant to connote a genetic relationship, but rather a shared cultural relationship.

For this study I will refer to the transmission of similar flintknapping knowledge and technique within related groups or cultures as learning traditions. A learning lineage is used to describe the transmission of flintknapping knowledge and technique from a master flintknapper(s) to the next generation. Based on the understanding of Paleoindians as being composed of highly mobile, small groups of hunter-gatherers, related groups should have a more similar flake scar patterning on the faces of projectile points than should unrelated groups. Given that projectile point types are often used to define cultures or groups of related people in the archaeological record, the

analysis of the variation in flake scar patterning is used here to address the cultural relationships between the groups of people who produced the projectile point types used in this study.

I have, in this study, identified a novel way by which to examine the relatedness of past projectile point technologies, by looking for similarities and differences in the morphology of bifacial flake scar patterns. The more similar the flake scar patterns are, the more likely that there was social interaction and social learning between those who manufactured the projectile points.

This assumption is used to argue that the more similar the flake scar patterning, the more closely related the projectile point types are to one another. This study is based on basic assumptions of social learning from cultural transmission, and the culture historical transmission of traditions. I am not trying to test the mode in which flintknapping knowledge was transmitted, what type of selection was occurring, or the forces affecting the transmission of flintknapping knowledge and technique. The basic premise of my research is that flintknapping knowledge and technique is a socially learned process within one’s culture group, and I assume that related groups should flintknap in a similar way that is different from unrelated groups.

Figure 1.1. Schematic showing flintknapping learning lineages through master flintknappers that are part of learning traditions in small band-sized Paleoindian groups. Blue represents a learning tradition completely separate from the Red. Lighter colors equal related splits in learning traditions.

Archaeological Complexes

The primary archaeological focus of this research is on Paleoindian-period assemblages

of unfluted lanceolate bifaces from the Great Plains of North America and northern Alaska

during the Pleistocene-Holocene Transition (PHT) from ~11,200 to ~10,000 14C years BP

(13,200 to ~11,000 cal yrs BP). Paleoindian demography and mobility are usually inferred

through lithic raw material sourcing. Distance to source and the percent composition of a raw

material in assemblages are often used to estimate the number of people composing a band or

social group, the amount of raw material they could carry across a landscape, and their mobility

or how far they traveled from lithic sources during seasonal rounds (MacDonald 1999; Meltzer

2009). For example, Clovis and Folsom projectile points located between 100 and 300 km, or

more, from their geological source are common and is viewed as an indicator of high residential

mobility (Amick 1996).

Population numbers during the Early Paleoindian period are expected to have been very

low. It has been proposed that Folsom groups could have exploited territories on the order of

120,000 km2 (Amick 1996). For territories of this size the population density for Paleoindians

has been modeled at .001 to .006 persons/km2, which MacDonald (1999) uses to estimate regional Folsom populations to be between 345 and 690 people per 115,000 km2. These

demographic statistics are supported by densities of known arctic foragers, from which Seeman

(1994) inferred a population for what is now the state of Ohio to be less than 650 at the end of the Pleistocene. Paleoindian hunter-gatherer bands likely would have been composed of small family or kin groups for much of the year, with larger seasonal gatherings consisting of multiple family groups aggregating for the exchange of knowledge, goods, mates, and to conduct larger communal hunts (Meltzer 2009).

Increased populations and decreased mobility are inferred for the Late Paleoindian period

in the contiguous United States based on the regional appearance of a number of morphologically distinct unfluted lanceolate points and a decrease in the use of exotic lithic raw materials, with most raw material sources found locally. Unlike the Early Paleoindian Clovis colonizers leap-frogging across a continent devoid of other humans, Late Paleoindians could go

nowhere that had not already been discovered and inhabited. These highly mobile groups would

have carried out their seasonal movements across large areas or home ranges, but were restricted

from spreading outward, forming what might be considered territories (Meltzer 2009). The

typical size and composition of family bands likely would not have changed from the Early to

Late Paleoindian periods.

The lanceolate Paleoindian projectile point types analyzed in this study were primarily

employed as weaponry for hunting large mammals. All these types are associated with the

Paleoindian period except for the Middle Archaic period Nebo Hill/Sedalia complex from the

Missouri area. Most of the Plains Late Paleoindian assemblages studied here are from bison kill

sites or bison carcass processing sites that form after a kill event occurred.

A number of archaeological sites containing large Paleoindian-like lanceolate projectile

points have been dated to the PHT in northern Alaska. These sites have been termed the

Northern Paleoindian complex, which is comprised of the Fluted, Sluiceway, and Mesa

traditions. These archaeological traditions are identified by morphologically distinct

characteristics in the projectile points. These traditions are also associated with large mammal

hunting. However, many sites are surface scatters or shallowly buried deposits. The topographic

setting of these sites on prominent elevations with commanding views of the landscape, coupled

with similar lithic technological organization, indicate they were used as hunting lookouts and

retooling locations; no residential sites have been identified (Smith et al. 2013). Faunal remains

from Fluted complex sites indicate a focus on caribou (Goebel et al. 2013; Hedman 2010). The

two major mammals targeted by Sluiceway complex hunters in Alaska are believed by some to

be bison and caribou (Rasic 2008), while Kunz et al. (2003) argue that Mesa complex hunters were targeting bison or muskox, based on the grassland landscape at the time which favored those species over caribou. Although large mammal kill sites, such as Paleoindian bison kill sites from the Plains, have not been discovered in Alaska, analogies between projectile point

morphology and targeted resources have been made (Bever 2000, 2001; Kunz et al. 2003; Mann

et al. 2001; Rasic 2008, 2011; Smith et al. 2013). This is important, as the entire assemblage of

points analyzed in this study had the same function and are associated with large mammal

hunting.

Table 1.1 Projectile Point Types Analyzed in this Study.

Projectile Point Time Geographic Target Function of Most Type Period Region Resource Sites Late Goshen Paleoindian Northern Plains Bison Bison Kill/Processing Late Plainview Paleoindian Southern Plains Bison Bison Kill/Processing Late Milnesand Paleoindian Southern Plains Bison Bison Kill Nebo Hill/Sedalia Middle Archaic Eastern Plains Edge Unknown Campsite Large Mammal Hunting Lookout & Sluiceway Paleoindian Brooks Range, Alaska Bison or Caribou Retooling Station Large Mammal Hunting Lookout & Mesa Paleoindian Brooks Range, Alaska Likely Bison Retooling Station

Paleoindians on the Great Plains

During the Late Paleoindian period, a variety of unfluted lanceolate projectile point types appear in the Great Plains following Folsom. These include the Goshen, Plainview, and

Milnesand point types to name a few. These projectile points are associated with large mammal

hunting, in particular bison. Late Pleistocene Paleoindians in the Great Plains were faced with an ever-changing landscape. The climate was becoming warmer and drier and 32 genera of

megafauna, those ≥ 44kg (Martin 1967), became extinct. Correlations and debates about the

causation between climate change, human colonization, and megafauna extinction are still at the

center of research amongst archaeologists who study this time period in North America.

However, nine large mammal species survived into the Holocene (Meltzer 2015). Of these, Bison

antiquus (~15 to 20% larger than modern bison), and later Bison bison, would become the most

abundant and most important species to prehistoric inhabitants on the Great Plains. Climate at

the end of the Pleistocene on the Plains became warmer and drier, which altered the diversity of

plant species. In particular, grasslands shifted from cool-season to warm-season grasses, which heavily favored ruminants (Guthrie 1984). As grasslands became dominated by warm-season buffalo grass and blue-grama grass, both favorites of bison, bison populations began to explode.

The loss of megafauna during the Early Paleoindian period, followed by the increasing number of bison, led the way for a Late Paleoindian focus on bison starting with Folsom (Meltzer 2005).

It is clear bison were hunted year-round (Surovell 2009). Analyses of bone assemblages indicate that butchering was selective and that prime cuts were targeted (Todd 1987). There is little indication at Paleoindian bison kill sites of marrow extraction and bone grease extraction through the breaking and boiling of bones, nor the production of pemmican, all of which were employed by later Plains Indians during winter months (Todd 1991). All this information indicates that hunts were highly successful, likely due to large bison populations that led to high encounter rates by highly mobile Paleoindian groups. Folsom bison kill sites usually involved the slaughter of a handful to around 15 individuals (Meltzer 2009; Surovell 2009). However, within

a few centuries Late Paleoindian bison kill sites contained at least four to ten times this number

of animals. At least 84 individuals were identified at a bone bed at the Plainview site (Holliday et al. 2017a) and 190 at the Olsen-Chubbuck site (Wheat 1972). This increase in prey numbers is likely the result of larger herd sizes, correlated with the increase in more favorable grasslands and increased hunting expertise of Late Paleoindians.

After Folsom, the appearance of multiple unfluted lanceolate point types in the northern

(Goshen) and southern (Plainview and Milnesand) Plains raises questions about the cultural

evolution and cultural relationships in this region. The Plains focus of this study examines the

relationship between Goshen and Plainview assemblages from dated archaeological sites that are

roughly contemporaneous (Figure 1.2). Are Goshen and Plainview distinct and geographically separate cultures, or are they geographic expressions of the same culture (northern and southern

Plains, respectively)? Milnesand points from the southern Plains in New Mexico, which look

similar to Plainview, are included in this study. Milnesand is believed to be contemporaneous or

somewhat younger than Plainview, although Milnesand has not been securely dated. The study

of flake scar patterning on projectile points from these three Late Paleoindian entities (cultures)

presents an interesting opportunity to examine the possible emergence of multiple Late

Paleoindian projectile point types from an earlier Paleoindian learning tradition.

Figure 1.2. Distribution of Late Paleoindian Goshen, Plainview, and Milnesand Complexes across the Great Plains. Adapted from Surovell (2009) and Waters and Stafford (2014).

Goshen points were discovered in 1966 at the Hell Gap site in Goshen County, Wyoming

(Irwin-Williams et al. 1973). The Goshen assemblage was found below the Folsom layer and was separated by a sterile layer of sediment. Although strikingly similar in appearance to

Plainview points from Texas, Plainview points post-date Folsom while the Goshen assemblage at

Hell Gap was believed to pre-date Folsom, which led the excavators to believe that Goshen and

Plainview were unrelated. After further review, Irwin-Williams et al. (1973) concluded that the

Goshen points from the Hell Gap site, the Plainview points from the Plainview site in Texas, and

projectile points from the Domebo site in Oklahoma, were in fact the same and that the Goshen

assemblages should be described as Plainview. Frison (1996) cautioned against this because so

little from the Goshen assemblage at the Hell Gap site had been analyzed and compared to the

Plainview material. Instead, the term Goshen-Plainview has been adopted when discussing

northern Plainview-like projectile points, as proposed by Frison (1996).

Goshen points are lanceolate in shape and are similar in outline shape to Clovis, but they

are not fluted (Figure 1.3: A-C). They have parallel to slightly concave or convex edges with

concave bases. The points are thin and exhibit basal grinding and basal thinning accomplished

with the removal of multiple flakes (Irwin-Williams et al. 1973). In fact, Henry Irwin in his

(1967) dissertation believed Goshen were a later Clovis phase that replaced fluting with a basal

pressure flake thinning (Frison 1996). Recent reevaluation of Goshen dates by Waters and

Stafford (2014) place the Goshen phase between 10,450±15 to 10,175±40 14C years BP (~12,500 to ~11,700 cal yrs BP).

Plainview points were discovered at the type site, located in the panhandle of northern

Texas. The Plainview site consists of a large bison kill bone bed with associated projectile points

(Sellards et al. 1947). Numerous Plainview sites have been identified in the southern Plains.

Plainview is a Late Paleoindian projectile point type dating to between ~10,300 to ~ 9,900 14C

years BP (~12,100 to ~ 11,300 cal yrs BP), and its geographic distribution is centered on the

southern Plains (Holliday et al. 2017b). This date range is based on dates from the type site and

other Plainview sites as well as the stratigraphic associations of other known Plainview sites and dated components (Holliday et al. 1999; Holliday 2000). Plainview points are lanceolate in shape

with a concave base subjected to heavy basal grinding (Krieger 1947). Plainview points can be characterized as Clovis in shape but without fluting (Figure 1.3: D-F) (Justice 1987).

Milnesand points are named after the type site located in eastern New Mexico (Sellards

1955; Warnica and Williamson 1968). Although a single date of 5,730±100 14C years BP was

recorded for the Milnesand site, the sample is believed to be contaminated and the date is far too

young to be Late Paleoindian. Thus, there is not a secure date for the Milnesand site and

associated material. A major difference between Plainview and Milnesand is the flat and square

base on Milnesand points. Basal thinning flakes on Milnesand points are smaller and more

numerous than that on Plainview and result in a distinct wedge-shaped base that terminates

abruptly (Sellards 1955). The Milnesand complex is believed to be about 10,000 14C years BP

(<12,000 cal yrs BP) (Holliday et al. 2017b). While there are questions surrounding the

relationship, if any, of Milnesand points to Plainview points (just as with Goshen and Plainview),

Milnesand points are viewed in this study as a distinct type from Plainview (Figure 1.3: G-I).

Figure 1.3. Great Plains Paleoindian Projectile Points. Goshen, A-C; Plainview, D-F; Milnesand, G-I.

Middle Archaic on the eastern edge of the Great Plains

The Middle Archaic Gerald Shelton assemblage, which is included in this study, is made up of Nebo Hill and Sedalia points. These point types are both distributed across much of

Missouri and into western Illinois and southern Iowa (Justice 1987; O’Brien and Wood 1998).

The archaeological sites at which these points are found are predominantly located on hilltops and bluffs overlooking waterways. When first discovered, Nebo Hill points were believed to immediately post-date Plains Paleoindian based on morphological similarities between the two, resulting in the assumption that they were for hunting large game and that they were the successor of Plains Paleoindians (Bray 1963; Shippee 1948). It was not until radiocarbon dates were obtained that it became clear these points were not associated with Late Paleoindian groups. Nebo Hill and Sedalia points are dated between ~4,000 to ~ 2,600 14C years BP (O’Brien

and Wood 1998).

Nebo Hill points are characterized as thick, lanceolate in shape, with long narrow blades

and straight to slightly tapered bases. The bases range from slightly concave to flat to convex.

The cross section of Nebo Hill points ranges from diamond shaped to bi-convex. The length to width ratio of these points is about 4:1, with the maximum width and thickness being forward of center (Bray 1963; Justice 1987; O’Brien and Wood 1998; Shippee 1948). Sedalia points are morphologically similar to Nebo Hill. These lanceolate points are on average slightly longer, thinner with a lenticular cross section, and wider than Nebo Hill points. In addition, the maximum width of these points is usually in the distal ⅓ of the point (Bray 1963; Justice 1987;

O’Brien and Wood 1998).

However, distinguishing between Sedalia and Nebo Hill points based on these morphological characteristics is not always clear. As O’Brien and Wood (1998) note, the two

types grade into each other imperceptibly. In my examination of the Gerald Shelton collection

(housed at Southern Methodist University), what would be classified as Nebo Hill in some

instances could be argued to represent a heavily curated Sedalia point. I do not attempt to

classify the specimens in the Gerald Shelton collection as either Nebo Hill or Sedalia. This is

because of the temporal, geographic, and morphological overlap of these point types, and

because the two sites from which the Gerald Shelton collection comes from are located within

one mile of each other with points that could be classified as either. Instead, I classify the

assemblage as Nebo Hill/Sedalia and view it as a single population (Figure 1.4: A-C). I am including the Nebo Hill/Sedalia points in this study as a control sample. These points make a good control sample as they are geographically and temporally distinct from the Paleoindian point types included in this study, yet their lanceolate morphology is characteristic of

Paleoindian projectile points. Having this control sample in my study is important to examine whether differences can be distinguished between non-related groups or if similarities in flake scar patterning could be the result of overall artifact shape, in this case lanceolate.

Figure 1.4. Variation seen in the Nebo Hill/Sedalia projectile points.

Paleoindians in Alaska

In Alaska, the Northern Paleoindian complex is composed of the Fluted, Sluiceway, and

Mesa complexes. The Fluted complex was not included in this study as the removal of channel flakes for fluting introduces known variation and presents a difficulty for comparison to any non- fluted lanceolate points. The flake scar patterns will be dissimilar from the start, because removal of the channel flake changes the basal shape, thus, introducing bias that cannot be overcome

when comparing fluted to unfluted points. Sluiceway and Mesa points are lanceolate in shape, both exhibit collateral flaking producing a ridgeline running down the long axis of the point, and both are manufactured from high-quality toolstone. While similarities exist between Sluiceway and Mesa points there are differences, primarily in morphology and reduction technique. The cultural relationship between these artifact complexes is unknown (Smith et al. 2013).

The Sluiceway and Mesa complexes differ from the other Late Pleistocene to Early

Holocene complexes in Alaska (Nenana and Denali) as well as from Old World traditions in

Siberia (such as Diuktai). Lithic assemblage analysis of Mesa projectile points from five sites conducted by Bever (2001) indicates that the lithic technological organization of the Mesa complex is more closely related to Paleoindians from the mid-continent than to the other, southern, complexes in Alaska (Nenana and Denali) (Figure 1.5). Northern Paleoindian toolkits feature large bifacial projectile points and associated debitage from their production. However, other lithic tools such as bifacial cores, flake cores, formal and informal scrapers, gravers, burins, and retouched flakes are also present in these toolkits (Kunz et al. 2003; Rasic 2008, 2011).

Figure 1.5. Generalized distribution of artifact complexes in Alaska.

The Sluiceway complex dates between 13,100 to 11,500 cal yrs BP and is found in the western Brooks Range (Figure 1.5). Sluiceway points (Figure 1.6: A-C) are large, and have an elongated lanceolate shape. They display robust semiregular collateral pressure flaking that leads to a thick, biconvex cross-section. Sluiceway points were heavily ground along the proximal edges and typically have a convex base (Rasic 2008, 2011).

The Mesa complex dates between ~12,500 to ~11,100 cal yrs BP and is found in the

central Brooks Range (Figure 1.5). Mesa points (Figure 1.6: D, E) are lanceolate in outline, have

collateral flaking, and are diamond-shaped in cross-section. Most have concave bases due to basal thinning, but flat and convex bases also occur. Almost all exhibit heavy edge grinding.

Complete points are estimated to have ranged from 5-10 cm in length, 1.6-2.8 cm in width, and

0.5-1 cm in thickness (Bever 2000; Kunz and Reanier 1995; Kunz et al. 2003).

Figure 1.6. Northern Paleoindian projectile points. Sluiceway, A-C; Mesa, D-E.

Northern Paleoindians would have practiced high residential mobility, following herds of animals and moving from kill to kill. They were similar in these respects to Paleoindians from the Great Plains during the same time period (Bever 2000). While a majority of projectile points were manufactured from high quality local cherts, obsidian artifacts from the Batza Tena source have been recovered in Northern Paleoindian sites located hundreds of kilometers away (Smith et al. 2013). Although faunal remains are scarce due to preservation issues in the Arctic,

Northern Paleoindian subsistence is believed to have centered on large mammal hunting (Bever

2000, 2001; Kunz and Reanier 1995, 1996; Kunz et al.2003; Rasic 2008, 2011), based on the fact that the lithic technological organization of these complexes revolves around large, lanceolate, bifacial projectile points. This inference is also supported by the presence of Late Pleistocene megafauna in Northern Alaska and the fact that plant species make up only a small amount

(~3%) of the diet in historic Arctic cultures from the area (Hall 1961).

Steppe bison, muskox, and caribou survived the mass extinction event at the end of the

Pleistocene in Alaska. Like the Great Plains, the YD brought with it cooler and drier conditions in Northern Alaska which promoted a grassland patchwork within the tussock tundra that spread during the BA. This grassland would have been favored by bison and muskox (Kunz et al 2003;

Mann et al. 2001). The warmer and wetter BA interstadial, as well as the post-YD HTM, promoted the growth of moist acidic tundra in the Arctic Foothills (Mann et al. 2002, 2010).

Caribou are adapted to tussock tundra, allowing them to easily traverse this landscape while bison are not. The Mesa complex appears in the Arctic Foothills after the onset of the YD and continues through this paleoclimatic event when the HTM warms and becomes wetter (Kunz et al. 2003; Mann et al 2013). While not causal, the pattern is indicative of large mammal hunting, especially when considering the functions of known archaeological sites.

Northern Paleoindian sites are located on elevated features on the landscape above rivers or in river drainages with extensive views of the surrounding area. This makes them ideal lookout stations for hunters waiting for herds of large game to pass by. At the start of the HTM, the landscape became more favorable for tussock tundra as the climate became warmer and wetter. For humans this meant the landscape became harder to traverse than during the YD and their resource base would have shifted. This ecological shift would have led to the loss of bison habitat and a decrease in their numbers. The end of the YD in northern Alaska led to a landscape that was unable to support key species of prey targeted by human hunters, such as bison. This coincides with the abandonment of the area for about 2,000 years, based on archaeological evidence, until caribou populations had time to expand and the area was reoccupied (Kunz et al.

2003; Mann et al. 2001).

Central Alaska: Nenana and Denali

The Northern Paleoindian artifact complex is not the only artifact complex present in

Alaska during the Pleistocene-Holocene Transition (PHT). Two other complexes, known as the

Nenana complex (~14,000 to ~12,800 cal yrs BP) and the Denali complex (starting ~12,800 cal yrs BP) are present in central Alaska during the PHT. The cultural relationship between the people, or groups, associated with these two artifact complexes has been a topic of longstanding, and major, debate in Alaskan archaeology that remains unresolved (Graf and Goebel 2009;

Yesner and Pearson 2002). The Nenana complex is characterized by a flake and blade technology, along with bifacial projectile points. These projectile points range from straight- based to teardrop-shaped to triangular bifaces known as Chindadn points (Powers and Hoffecker

1989).

The Denali complex, originally defined by West (1967), is associated with the presence

of microblades and microblade cores. Also present are burins and burin spalls as well as

lanceolate, concave-based bifaces. Of particular interest is the fluorescence of microblades in the

Denali complex after ~12,800 cal yrs BP, at the onset of the YD. Microblade technologies,

whereby small blades ranging in width from 4-8 mm and having a length of about 2-3 cm (Clark

2001) are inserted into an osseous projectile point, are believed to be a cold weather adapted

technology in the far north (Elston and Brantingham 2002; Guthrie 1983). The increased

presence of microblades in the Denali complex after 12,800 cal yrs BP coincides with the onset of the YD, which was much more prevalent in the Arctic than at lower latitudes. Climate change associated with the YD has been identified as a likely factor for the differences seen in the lithic technological organization between these two artifact complexes. Explanations for these technological differences include: differences in the seasonal use of resources, changes in landscape use, or a complete population replacement (Graf and Goebel 2009; Graf and Bigelow

2011; Hoffecker and Elias 2003, 2007; Potter 2011; Potter and Holmes 2013; Wygal 2011).

The lithic technological organization of these two complexes and their spatial and temporal overlap in central and south-central Alaska are key to understanding how people reacted to ecological landscape change during the Pleistocene-Holocene Transition in Alaska

(Goebel et al. 2013). Archaeological research has yet to yield definitive conclusions on the nature of the cultural relationships between these two artifact complexes or the large biface- centric Northern Paleoindian complexes to the north. Unfortunately, the sample size of bifacial projectile points from the Nenana and Denali complexes was too small to conduct meaningful topographic morphometrics. As such, a comparison cannot be made to establish the cultural relatedness of these three complexes in Alaska at the Pleistocene-Holocene Transition. Only an

analysis of the Northern Paleoindian Sluiceway and Mesa points from Alaska was undertaken in this study.

Structure of the Dissertation

In Chapter Two I introduce the Next Engine™ Ultra HD three-dimensional scanner used

to produce high resolution imagery of projectile points as well as the image acquisition and

processing stages before topographic morphometrics are taken of flake scars. The topographic morphometric methodology is tested in Chapter Three with three Late Paleoindian projectile point types from the Great Plains (Goshen, Plainview, and Milnesand) and a control sample of

Middle Archaic projectile points from Missouri (Nebo Hill/Sedalia). Chapter Four is the first

Paleoindian case study where I apply this topographic morphometric analysis to the Great Plains

Late Paleoindian typologies to interpret regional cultural relationships pertaining to these projectile points that look morphologically similar, but come from different geographic areas. In

Chapter Five I present a second Paleoindian case study to test the effectiveness of topographic morphometrics analysis on a second dataset to assess the relationship between Sluiceway and

Mesa point types of the Northern Paleoindian complex in the Brooks Range, Alaska. Finally,

Chapter Six concludes by discussing the application of topographic morphometrics to projectile points, its success in the assemblages from this study, and its future in archaeology.

CHAPTER TWO: THREE-DIMENSIONAL IMAGING AND METHODS

Three-dimensional (3D) imagery has slowly made its way from healthcare and

commercial settings to use in archaeology. Initial application to archaeological studies was

hampered by the price and access. However, there are now a number of 3D scanners on the

market that meet archaeological needs. This technology is at a place where price and portability

are making it easier for archaeologists to apply high resolution 3D imagery in their studies. In

particular, the use of high resolution 3D imagery has become more prevalent in lithic studies.

Traditional geometric morphometric studies reduce a 3D object into 2D measurements. The

information that can be gained by analyzing 3D and interactive models allows for a whole range

of new lithic analyses. One such analysis is the topographic morphometrics used in this study

which requires 3D imagery of an object.

NextEngine Ultra HD 3D Scanner

This project used a NextEngine™ Ultra HD portable multi-laser scanner. It was selected

for its portability (8.8" x 3.6" footprint standing 10.9" high), quality scans, and most importantly,

low cost (compared to other 3D scanners on the market). The scanner is extremely portable,

weighing about 7 lbs., and easily fits into a carrying case. The NextEngine Ultra HD 3D scanner

is a desktop laser scanner with an accuracy of .001 cm and can record over 100,000 points per

cm2 on the maximum point density setting.

The NextEngine™ Ultra HD scanner is powered from a wall outlet and connects to a computer through a USB data cable. A revolving stand is included that plugs directly into the front of the scanner, and it auto-calibrates the distance between the two when powered on.

Rubber grippers and clay are provided to keep artifacts in place on the rotating stand while in

use. Rubber grippers were used because they hold the artifact more securely in place and it is

easier to electronically trim out the rubber gripper than clay during image processing (Figure

2.1). As added protection a 2-foot by 2-foot rubber mat was placed under the revolving stand in the unlikely event that an artifact came loose and fell from the stand.

Figure 2.1. Next Engine Ultra HD 3D Scanner.

To deal with varying characteristics of raw materials that affect the quality of laser scanning, including transparency and reflectiveness, a talcum powder was applied to the surface of artifacts to ensure accurate reflection of the laser light. The powder is not thick enough to alter

the measurements of flake scars or the accuracy of scans. It was used because when scanning

reflective and/or semitransparent raw materials (encompassing a majority of suitable raw

materials for stone tools), the laser light is partially absorbed by uncoated objects, which can

leave the scans with a bubbly or blurred surface that reduces the accuracy of surface topography

data. The talcum powder prevents the absorption of most of the laser light. Three different

talcum powder application options were used in this research: a powder pen provided with the

NextEngine™ scanner, a talcum foot spray available from any pharmacy, and billiards 100%

talcum. The powder pen provided with scanner is talcum based with a small amount of lanolin (a

lipid sheep wax) to help the powder adhere to the surface of objects. It worked very well on all of

the various lithic raw materials. The talcum foot spray uses a liquid alcoholic mixture that when dry, helps to adhere and coat the surface of an object. Finally, straight talcum powder manufactured for billiards use, while not as easily applied as the powder pen and foot spray, was effective and was used when concern over the application of alcohol or lanolin to artifacts was

voiced by collection curators. The specific talcum application for each projectile point is

presented in Appendix A.

To produce 3D models the objects are scanned using ScanStudio 2.0.2 software at fixed

intervals or divisions (out of 360°) relative to the scanner. The options range from four (every

90°) to sixteen (every 22.5°) divisions around the object. These digitized images from different

intervals are known as a “scan family,” and are subsequently fused into a single scanned object.

Three scanning qualities are available for each scan: Quick, Standard Definition (SD), and High

Definition (HD), each with three resolution settings with points/in2 ranging from 1.2k to 268k

respectively. Three targeted range settings are available at which the rotating stand can be placed

away from the scanner: Macro (7.5”-11.5”), Wide (22”-28”), and Extended (22”-40”), depending

on the size of the object being scanned. The user decides the number of divisions, resolution, and

range for each scan setting. The number of divisions is an independent choice at the start of each

scan and the choices remain constant no matter the selected settings for resolution and range.

However, the three range settings affect the resolution possibilities of each. Macro offers the

highest resolution, with a maximum of 268k points/in2, while the maximum Wide and Extended

resolution are 29k points/in2 and 3.3k points/in2 respectively.

Through trial and error, review of published literature, and personal communication with others conducting 3D scanning, an effective methodology for creating, collecting, and processing scans was developed. Because this scanning equipment was originally developed for commercial and industrial use, scanning irregular (shape, texture, reflectiveness, transparency) and sharp- edged objects such as lithic bifaces could be a challenge. The methodology discussed below, which yielded excellent results, was developed and followed unless extenuating circumstances occurred. Preferably, objects were scanned at the highest resolution of 268k points/in2 using the

Macro range. Sometimes this could not be achieved, especially for large objects, because the

capacity of the laptop memory was exceeded. At minimum scans were taken at the second

highest HD resolution of 67k points/in2. For larger objects requiring the Wide range, the highest

HD setting available of 29k points/in2 was used. Each projectile point was placed vertically on

the rotating stand, and the number of scan divisions was set to 8 (every 45°) creating a scan

family of 8 digitized images. In standing the object vertically, the very tip (~1-2 mm) and bottom

edge of the base are missing from the scan. Subsequently these missing areas are filled when the object is fused during image processing by connecting the Triangular Irregular Network (TIN) from the other data points. In the scanning software setup page, before scanning starts the user can target a specific area within the image screen by setting boundaries on where scan data are

collected. This eliminates extraneous space from the scan, which cuts down file size (~300MB-

500MB down from ~2.5-3GB) and decreases scan times to ~25-35 minutes per object depending

on its size.

Shott and Trail (2010, 2012), in scanning lithics object with a NextEngine™ scanner,

took eight vertical and eight horizontal scans and fused them together to create a 3D object.

However, they used an older version of the NextEngine scanner that had a lower resolution than

the Ultra HD model used for this study. With the Ultra HD scanner at the highest HD setting,

scans were taken at 268k points/in2 which provides much more detail than earlier scanners. Even

with a scan family of 16, scans Shott and Trail (2010, 2012) note that holes can sometimes

appear in the object, and that when this occurs the “fill mesh” tool creates a satisfactory solid 3D object by interpolating point locations.

Additionally, while developing my methodology, I initially scanned projectile points both horizontally and vertically, as Shott and Trail (2010, 2012) described. However, for larger bifaces (> 10 cm in maximum length) scanned horizontally at the macro distance (for highest resolution), the tip and base (while rotating) will pass in and out of the optimal 4-inch scanning circumference located 7.5 to 11.5 inches from the scanner. When this happens the tip and base become distorted and flake scar resolution is actually lost. Thus, when horizontal scans are combined with the vertical scans into a single fused object, the object has a lower resolution than what is produced using only vertical scans. Therefore, after numerous trial and error scans, it was decided to forego the eight horizontal scans and only collect the eight vertical scans. This procedure not only saved about 35 minutes per scanned object but actually provided higher resolution imagery right up to the very tip and base of the object.

The NextEngine™ Ultra HD scanner was well suited for the 3D goals of this project.

This study examines the topographic morphometrics from the faces of bifacial tools, and the

scanner collected those data without problem. Where the scanner had difficulties was in collecting the morphology of biface edges, which are very fine (thin) and irregular. However,

biface edges are not a focus of this study. If edge studies were the primary focus, such as edge-

wear/use-wear/grinding, another 3D scanner or photogrammetry would be required. There are

more expensive scanners that would work better for edge studies.

Image Processing

After a scan was completed the scan family was digitally trimmed to remove any minor

flaws (reflection from raw material types), noise, and extraneous material (the revolving stand

and rubber gripper) from the scanning process. The trimmed scan family was then fused into a

single 3D object using the ScanStudio 2.0.2 software which removes overlapping and redundant

data.

RapidWorks version 4.1.0 (additional NextEngine software) was then used to clean the

original Scan Studio scans and ultimately save the object as a 3D file that can be opened by a

number of software programs. The first step was to clean up the original scans using three base

tool functions. The Find Defects tool identified defective poly-faces in the mesh, which were

then deleted. The Healing Wizard tool was used to heal various defects in the mesh. Finally, the

Global Re-mesh tool re-triangulated the entire mesh and improved mesh quality while making

the object “watertight.” After the original scans had been cleaned and re-meshed the 3D model

was exported as a binary STL file (.stl) for use in 3D analysis/viewing with appropriate software.

While original scan files ranged from ~500MB to several GB, the final .stl file was usually

around 50MB. A major strength of this file type is its compact file size. However, in order to maintain a compact file size, the actual surface colors of the artifacts are not included. The absence of the original surface colors of the artifacts was not an issue for the research conducted here. All available 3D projectile point scans used in this study can be downloaded from the

Zenodo.org digital repository at CERN, and can be found using the DOI found in Appendix A.

The RapidWorks software was also used to create the isocontours of the 3D objects that are ultimately analyzed using topographic morphometrics. These “elevational” isoheight contour lines on the faces of projectile points will be referred to throughout this study as isocontours. An isocontour is defined here as a set of level curves that form a closed polyline at a contour of equal height. To split or cross-section the biface into two faces for analysis, a reference plane

(using the Add Reference Plane under the Datum Tab in the RapidWorks software) was added using the 3-point option to split the projectile point in half along the edge (Figure 3.2A). Using this tool, one point was selected on the tip, and one on each edge near its widest point to create a best fit plane, splitting the biface in half as accurately as possible. Bifaces are not perfectly symmetrical, thus, a perfect split cannot be achieved for every projectile point. The coordinate system for the object was then set off of the split plane with the 0, 0, 0 (x, y, z) origin coordinate at the center of the object. The object was then digitally cross-sectioned at a given height or elevation (z direction measurement) by inserting a reference plane (Figure 3.2B). This was done under the Datum tab of the RapidWorks software, and four additional reference planes were added at ±¼ and ±⅓ max thickness on each face of the artifact from the Split Plane, using the offset function to intersect the mesh model. A reference polyline was then added at the intersection of the offset planes and the mesh model to create a “topographic” isocontour line of the flake scar patterning at the given “contour elevation” (Figure 2.2C).

Figure 2.2. A, The addition of a split plane to a 3D object. B, The addition of two ±¼ max thickness reference planes off of the split plane. C, 3D scan of a Goshen projectile point with isocontour of flake scars at +¼ max thickness.

To add reference points (that include x, y, z point locations), the Convert Entities tool in

3D Sketch Mode was used to convert the isocontour polyline to a curve. The Rebuild tab was then used to replace the existing number of interpolation points along the curve with the desired number, which for this study was 3 points per mm (e.g., 3x length of the curve as displayed in the properties console). Depending on the size of the scanned objects, the number of points

ranged from just over 200 to over 900 points per isocontour. These steps were repeated for every

isocontour associated with the 3D object.

The rebuilt isocontours with reference points were then exported as a STEP file that

contained a list of the x, y, and z coordinates for every interpolated point which could be opened

in Microsoft Excel. In Excel, the STEP files were opened and the “Text to Columns" tool was

used to separate the comma separated values (CSV) containing the x, y, and z coordinates into

different columns. Additional data in the STEP file, such as date of scan, name of object, and file

location, were deleted. Only the x and y data were copied into a new Excel file and saved as a

.csv file. The z data are not needed, and were deleted, because the elevation was chosen and set

from the start of the process. Each biface was given a unique number for this study such as 0001,

0002, etc. with the addition of _a or _b to identify the isocontour of each face.

Not all the biface faces that I scanned were used in this study. Some faces were used for

only one analysis and were left out of others based on characteristics of the point that vary at

different isoheight measurements. If a face exhibited damage, was extremely concave, was produced on a flake blank leaving one face that was not bifacially worked, or if resharpening

extended into the isocontour that would affect the outline analysis, it was left out. The projectile

point numbering and face distinctions used in this study and their corresponding scan

information can be seen in Appendix A.

Image Analysis

Topographic morphometric analysis examines outline shape using elliptical Fourier

analysis (EFA). The Momocs package (Bonhomme et al. 2014; R core team), an add-on package in the R statistical software was chosen to conduct the EFA as it can be used to both convert the

flake scar pattern isocontours into elliptical Fourier coefficients, and to conduct multivariate

analyses on these coefficients.

Fourier analysis originated in the biological sciences as a way to quantify the visual

information pertaining to shape (biological form) that is otherwise hard to characterize. A

Fourier series is a technique to fit a series of curves to describe a shape. A Fourier series breaks

down a periodic function into the sum of sinusoidal (sine and cosine) functions. A periodic

function is a mathematical function, which at regular periods or intervals, repeats its values such

as any waveform. In this way, a Fourier series can be thought of as the approximation of a

periodic function (Bonhomme et al. 2014; Lestrel 1989, 2008).

The method of EFA described by Kuhl and Giardina (1982) builds on Fourier analysis to

fit a series of curves to a closed object or polyline shape. Elliptical Fourier descriptors are

therefore a mathematical way to describe a 2D closed curve or contour of any shape by defining

an ellipse through the sum of the sine and cosine curves. Elliptical Fourier analysis works by first

converting an outline shape into a chain code of line segments. The line segments of a closed

contour are approximated by eight line segments (a0-a7). This can be thought of as similar to

nearest neighbor analysis (with defined direction) in a raster (Figure 2.3A). Where, from a

starting point, the direction of the next point is classified based on the direction from the

previous point. Therefore, a chain code is simply a chain of numbers (e.g. 7,7,5,7,4,4,4,1,3,1,1) that describes a closed curve, such as an arrowhead, in Figure 2.3B.

Figure 2.3. A, Chain code line segment approximation. B, Example of a 2D closed contour of an arrowhead drawn using a chain code.

Elliptical Fourier analysis of a closed 2D contour produces four Fourier coefficients per harmonic. These four coefficients are the value that approximates the signal of the closed contour periodic function (ellipse) in the given coordinate system based on the number of harmonics specified. In EFA, harmonics can be thought of in the same way as they are in music and in waveform studies in physics. Harmonics are integral multiples of the fundamental frequency (1st harmonic) of a wavelength. As you increase harmonics the wavelength is decreased resulting in a higher frequency. For example, the 2nd harmonic will have a wavelength

½ of the 1st harmonic, which means the frequency of the 2nd harmonic will be twice that of the

1st. The 3rd harmonic will have a frequency three times that of the 1st harmonic and so on

(Bonhomme et al. 2014; Kuhl and Giardina 1982). The more harmonics used, the more

cumulative coefficients describe the shape, which increases the detail and equals a better fit of the Fourier series to the shape of the object, in this case flake scar patterning. More complex shapes require higher numbers of harmonics (Bonhomme et al. 2014; Kuhl and Giardina 1982;

Lestrel 1989, 2008; Rohlf and Archie 1984). For a more detailed and thorough explanation of the trigonometry behind EFA see Bonhomme et al. (2014) Kuhl and Giardina (1982) and Lestrel

(2008).

The outline function in Momocs was used to connect all of the points consisting of x and y data into a closed outline shape. All of the isocontours were aligned and scaled, thus, removing the size and orientation of the object as a factor that could influence the results. Next, the calibrate harmonic power function was used to see how many harmonics were needed to explain the variability in shape outline. Increasing the number of harmonics increases the approximate fit to the actual shape. Essentially, a lower number of harmonics will describe the overall outline of a shape (oval, round, etc.), whereas a higher number of harmonics will describe greater detail in the closed polyline contour (for example see Sholts et al. 2012 Figure 4). Generally, 99% of outline variation was explained between 14 and 16 harmonics. One-hundred percent of the variation was usually explained by 40 harmonics. For consistency, 30 harmonics was used, which explained ~99.9% of the variability on average. As each harmonic has four Fourier coefficients, each flake scar isocontour was described by 120 Fourier coefficients. The R code for this topographic morphometric analysis is available upon request.

Because each face of a projectile points was assigned its own identification number, each projectile point contributed two different isocontours per measurement elevation that could be plotted separately when conducting multivariate analyses. Two multivariate analyses, Principal

Components Analysis (PCA) and Linear Discriminant Analysis (LDA), were used to quantify

differences in isocontour shape defined by the Fourier coefficients. Principal components

analysis was carried out to identify variations in shape among the flake scar isocontours and

whether meaningful point clouds were evident in the PCA plots. A point cloud is defined and

used in this study as the total number of flake scars from either one or both faces of the projectile

points from a specific projectile point type, as points in a PCA or LDA plot. Linear Discriminant

Analysis was used to test how well the isocontours of a given projectile point type group

together, given their assigned projectile point type. While PCA determines the axes with the

most amount of variance for the data, LDA identifies the axes that best separate the point types.

The LDA reclassification in Momocs was used to assess how well test observations of isocontours from a training dataset compared to the known archaeological type, in order to

examine the test error rate of the LDA in assigning the assemblage of isocontours to a projectile

point type. Confusion matrices of the results show how misclassifications occurred by type and present a quantitative measure of how well each of the measurement sets performed and which correctly classified projectile points the best. As PCA and LDA plot the Elliptical Fourier

Coefficients that describe the outline shape of the flake scar isocontours, any discussion of point

clouds from PCA and LDA plots are qualitative evaluations of patterns in the data as there is no

datum to measure from.

When using the Momocs package in R an interesting complication arose. When plotting

the elliptical Fourier coefficients in PCA, two point clouds would plot in the morphospace with

background outlines of one cloud crossing over itself to create a bowtie like appearance in

outline. Obviously, projectile point edges cannot cross back over themselves, and it turned out this problem is due to the direction in which the outline function connects the points (clockwise or counter clockwise). The outline function in Momocs needs the connection of all of the points

to be going the same direction. Otherwise, when tracing, the x and y planes end up crossing over each other. To fix this the .csv file needs to be opened and all of the x and y points need to be reordered. The first x and y point needs to become last and last needs to become first, flipping all the other points in between. This is done by creating a third column numbered 1 through the last row of x and y data, or n. Then, the data are re-ordered (largest to smallest) based on the numbering in the third column. When this is complete, the third column can be deleted and .csv file for that point face can be saved. This reorders the points so when tracing the outline in

Momocs the program will reverse the (either clockwise or counterclockwise) connection of points.

Topographic Morphometrics and Lithic Expectations

Topographic morphometric analysis of bifaces uses Elliptical Fourier Analysis (EFA) to define the outline shape of a closed line object, in this case the isocontour of flake scar patterning at a given height. The Momocs add-on package in the R statistical software was used to conduct the EFA as both converted isocontours shape outlines into Fourier coefficients, and conducted multivariate analyses such as PCA and LDA on these Fourier coefficients to look for similarities and differences in the shape of flake scar patterning.

Studies of cultural transmission using lithics have typically employed two-dimensional

(2D) geometric morphometric analysis of stone tools to look for morphological similarities and differences in shape and size variation. For example, Bettinger and Eerkens (1999) and O’Brien et al. (2001) have taken length measurements from the haft and blade elements, weight, and ratios of these variables to look for cultural relatedness between projectile point types or

“cultures” or to explore the means by which information was socially transmitted. Other studies

have used Elliptical Fourier Analysis (EFA) to examine the morphological shape of lithic

artifacts (Gero and Mazzullo 1984; Ioviţă 2009; Ioviţă and McPherron 2011). Geometric

morphometrics have also been used to identify and distinguish projectile points as either dart or

arrow points based on size (Bradbury 1997; Hildebrandt and King 2012; Shott 1997; Thomas

1978). While 2D morphometrics have led to great advances in lithic studies, the recent

application of 3D scanners to the field has opened the door to new possibilities (Shott 2014). The

use of 3D digital imagery also removes the replicability error of linear measurements because the

imagery can be shared with digital measurements marked in the file for subsequent studies or re- evaluations of materials.

Three-dimensional digital imagery removes some of the problems associated with artifact curation (e.g., resharpening) seen in studies examining blade and haft elements (O’Brien et al.

2001) that have restricted other studies to morphometrics of hafting elements (Bettinger and

Eerkens 1999). Two-dimensional analyses produce a metric measurement of a 3D object, whereas a 3D image can provide topographic measurements pertaining to the point manufacture and subsequent episodes of curation. Topographic morphometric data of a projectile point can be thought of as features on a topographic map. Different isoheights (elevations) of flake scars can be traced, just as valleys and ridge lines on a map, to address questions focused on specific areas

(features with elevational data) on a biface. This allows for targeted analysis from parts of the blade element wherein original flake scar patterns can be identified and separated from those of re-sharpening.

Sholts et al. (2012) used this methodology to distinguish archaeological Clovis points from replicas produced by the modern flintknapper Woody Blackwell. Using 3D imagery of

Clovis points, Sholts et al. (2012) concluded that there was a common Clovis “culture” based on

similarities in the flake scar isocontours of archaeological specimens, and this differed from the

higher variability seen in modern replicas. To the authors, the “low variation in flake scar

morphology among the ancient Clovis projectile points implies that they were all crafted with a

standardized manufacturing technique…information on manufacturing technology can indeed be

extracted from flake scar contours” (Sholts et al. 2012: 3022). The authors conclude that Clovis

points across the contiguous United States had a common reduction technique that differs

statistically from replica Clovis points that even to experienced archaeologists appeared to be

genuine. Gingerich et al. (2014), utilizing the same methodology, came to similar conclusions by

examining fluted points from the American northeast as well as Clovis replicas from two modern

knappers. The application of this methodology makes it possible to test hypotheses relating to

cultural evolution and relatedness of prehistoric groups through the analysis of flake scar

patterning on stone tools from the archaeological record.

However, this methodology has only been applied to Clovis and fluted Paleoindian

assemblages. If the goal is to address questions concerning the relationships of more than one

prehistoric group, it must demonstrate the methodology has the ability to not only identify single

“cultures” based on similarities (Gingerich et al. 2014; Sholts et al. 2012) but to also distinguish

between “cultures” based on differences. The aim of this research is to expand on these previous

studies by utilizing the methodology they developed and adding geographically and temporally

distinct Paleoindian projectile point assemblages to test whether these methods can in fact

differentiate flake reduction patterns exhibited by projectile point assemblages and, thus,

differentiate archaeological “cultures,” or related groups, from one another.

The cross sectioning of projectile points at given isoheights on a face records the morphology of flake scars that are remnant of the manufacturing technique used to make the

tool. If flintknapping knowledge and technique are socially learned, and projectile point types

represent related groups, then the similarities and differences in flake scar patterning can be used

to identify groups of people who share a learning tradition. Based on assumptions of social

learning and culture historical transmission of traditions the expectation is that flake scar patterns

associated with projectile point manufacture should look similar between related types, and thus groups of people or cultures who share a learning tradition. Therefore, different projectile point

types made by groups of related people who share similar flintknapping knowledge for projectile

point manufacture should form and overlapping and indistinguishable point cloud in PCA and

LDA. On the other hand, projectile point types that were produced by unrelated groups with

differing flintknapping knowledge should each form separate point clouds, and should be

distinguishable from other clouds.

CHAPTER THREE: TESTING TOPOGRAPHIC MORPHOMETRICS

To test whether 3D topographic morphometric methodology can distinguish variation in flake scar patterning, a pilot study was conducted on assemblages of unfluted lanceolate projectile point types comprised of three Late Paleoindian complexes and a single Middle

Archaic complex. Specifically, the goal is to understand the relationship between Goshen and

Plainview points. The Late Paleoindian assemblages for this study come from the Mill Iron

(MT), Plainview (TX), Bonfire Shelter (TX), and Milnesand (NM) archaeological sites (Figure

4.1). The Mill Iron site lies in the northern Plains, and the others in the southern Plains. The selection of a Middle Archaic Missouri assemblage from the Gerald Shelton collection (MO), curated at Southern Methodist University, was based on the geographic and temporal isolation of that assemblage from the Late Paleoindian assemblages. The Late Paleoindian collections were selected based on the number of complete points that represent each of the projectile point types.

A description of each site follows and is summarized in Table 3.1. The Gerald Shelton assemblage consists of lanceolate projectile points classified as Nebo Hill/Sedalia, and were selected for two reasons; 1) to test that the methodology can successfully distinguish between flintknapping patterns separated in time and space, 2) to test the proposition that overall projectile point morphology (lanceolate) does not determine flake reduction patterning.

Paleoindian projectile points have a lanceolate shape. When Nebo Hill points were discovered in Missouri, their morphological similarities with Paleoindian points led some to believe they were the Early Archaic successors of Paleoindians from the Great Plains (Bray

1963; Shippee 1948). With radiocarbon dates from a number of sites, we now know that these points date to the Middle Archaic, some 6,000 years after the Paleoindian period (O’Brien and

Wood 1998). However, these projectile points present an excellent opportunity to test whether

topographic morphometrics can distinguish flake scar patterns on projectile points and if these differences can be associated with unrelated groups. Before, topographic morphometrics can be applied to cultural questions in the archaeological record it first has to be demonstrated that the methodology works correctly. The fact that all of the projectile points from this pilot study are unfluted lanceolate points is important. It is expected that the flake scar patterning from the

Middle Archaic assemblage should look different from the Late Paleoindian even though the

Middle Archaic assemblage looks similar morphologically. This expectation must be met, because if it is not, then overall morphology of projectile points (lanceolate in this case) could be driving the similarities seen in flake scar patterning, rather than socially learned flintknapping techniques.

Figure 3.1. Location of Plains Late Paleoindian archaeological sites used in this study: 1, Mill Iron; 2, Plainview; 3, Milnesand; 4, Bonfire Shelter. Location of the Middle Archaic site used in this study: 5, Gerald Shelton collection.

Mill Iron Site

The Mill Iron site is located in southeastern Montana near where North Dakota and South

Dakota meet, in the northwestern edge of the Black Hills (Frison 1996). The site lies along the

edge of a small isolated circular butte about 20 meters in height and 35 meters in diameter. The

Mill Iron site consists of two locales ~25 meters apart: a bison bone bed, and a

campsite/processing site. Lithic artifacts, including projectile points, were found at both locations

as well as the remains of bison. It is estimated at least 29 bison were killed and are represented

across the site. Based on molar eruption, a spring season of death was determined from both locales (Todd et al. 1996).

How the bison were trapped and killed is unknown. Arroyo traps were common features used by Paleoindians to procure bison on the High Plains (Frison 1984). Based on the local geology there could have easily been such features in the vicinity of the site, which Frison (1996) believes was where the animals were killed and later were brought to the Mill Iron site for butchery. However, after looking at the taphonomy of the bone bed, Kreutzer (1996) argues that the Mill Iron locales represent a kill location and a butchery location. The bone bed midden is located in an ancient stream channel, which geomorphologically suggests a possible drive line and impoundment or corral of some sort. There is no direct tie between the two locales at the site to indicate whether they were occupied simultaneously. Therefore, the locales could represent two separate events, or a single event with separate kill and butchery locations. The presence of

Goshen points and of B. antiquus from both locales indicates they were occupied roughly contemporaneously and by people sharing the same cultural affiliation.

Dates from both locales at the Mill Iron site are clustered and range from 10,760 ± 130 to

11,010 ± 140 14C years BP and 11,320 ± 130 to 11,570 ± 170 14C years BP (Frison 1996).

Waters and Stafford (2014) reexamined the Mill Iron material and radiocarbon dated three bison

bones to 10,450±25 14C years BP, 10,465±20 14C years BP, and 10,435±25 14C years BP, with an average age of 10,450 ±15 14C years BP (12,526 to 12,119 cal yrs BP). Fifteen complete

projectile points and one base from the Mill Iron site are included in this analysis.

Plainview Site

The Plainview site is the type site for the Plainview projectile point type. It is located outside the town of Plainview in northwestern Texas. The site was discovered in 1944 when quarry activity exposed cultural material. The bison bone bed was located in an old stream channel, and ranged in thickness from a few inches to over a foot and a half in an area of some

500 ft2 (Sellards et al. 1947). Sellards et al. (1947) originally estimated that at least 100 individual bison were represented at the site. Recently, Holliday et al. (2017a) estimated that at least 84 individual bison were killed at the site. The archaeological materials from the site indicate that this is a bison kill site. There is evidence for at least two separate kill events: tooth eruption status in juvenile specimens indicates a spring hunt and a fall hunt. Procurement took place near an active stream, and might have involved using the muddy paleochannel to help slow or trap the animals (Holliday et al. 2017a).

A total of 28 points and biface fragments were recovered at the Plainview site, 10 of which are used in this study. As with other Late Paleoindian sites on the southern Great Plains, radiocarbon dating has been an issue. A total of seven AMS dates from bison bone were obtained that range from 8,400 to 11,400 14C years. Both ages at either end are believed to be incorrect.

The younger date at 8,400 is much younger than all other dated Plainview material (~10,000 14C

years), and the older date of 11,400 would be as old or older than Clovis. Radiocarbon dates of

around 10,200 14C years BP place the occupation of Plainview at >10,000 14C years. A date of

10,170±100 14C years BP (12,373 to 11,339 cal yrs BP (OxCal 4.3)) falls in with dates of other

Plainview type sites (Holiday1999, Holiday et al. 2017a, b)

Bonfire Shelter Site

Bonfire Shelter is the second Plainview site included in this study. The site is a rockshelter along a canyon that drains into the Rio Grande River in southwestern Texas on the edge of the Edwards Plateau. Three bone beds were discovered during excavation. The bottom bed, bone bed 1, contains megafauna remains but has not been conclusively attributed to human behavior. The top bone bed, bone bed 3, dates to the Archaic and contains the remains of B. bison. The middle bed, bone bed 2, is Paleoindian in age and diagnostic Folsom and Plainview projectile points were recovered from this layer. Although the lithic assemblages is small from bone bed 2, three Plainview points were included in this study. Bison remains from this level are representative of the extinct B. antiquus. The original interpretation of bone bed 2 was that it was the result of a bison jump, whereby bison were driven over the edge to their deaths or were injured by the fall and dispatched by hunters waiting below. The formation of bone bed 2 was believed to have happened over the course of three kill episodes and, if it was a bison jump, represents the earliest known bison jump in North America (Dibble and Lorrain 1968).

Bison jumps were not a common method for bison procurement until the Late Archaic, and mostly are found on the northern Plains (Frison 1991). Some archaeologists, such as Binford

(1978) based on Nunamiut caribou hunting/processing patterns, argued that bone bed 2 represents a post-kill butchery and processing site where high-utility skeletal elements had been transported. This was based on extensive disarticulation and patterned stacking of like elements.

Recently, Byerly et al. (2005) revisited the faunal remains and came to the same conclusion as

Binford (1978). In addition, they determined that the bone assemblage represents 24 individuals,

all of about the same age at death, and that they were killed in the summer. This leads the authors

to suggest that bone bed 2 represents a single event. Radiocarbon dates from bone bed 2 date

between 10,230±160 14C years BP (12,240 to 11,695 cal yrs BP) to 9,920±150 14C years BP

(11,640 to 11,205 cal yrs BP) (Holliday et al. 2017b; Pearson et al. 1965).

Milnesand Site

The Milnesand site is located in eastern New Mexico and represents a Late Paleoindian

bison kill site (Sellards 1955; Warnica and Williamson 1968). The site lies in a broad but

shallow depression that is believed to have contained some amount of water during the year. The

depression is surrounded by dunes that might have acted as a natural trap or slowed the animals

enough to be dispatched (Holliday et al. 2017b). In total, 97 projectile points, mostly Milnesand

but also including a few Plainview, were recovered from the Milnesand site. However, most

were kept by the landowner, and only a small sample of 12 points from the original excavation are housed at the Texas Archaeological Research Laboratory in Austin, TX and are included in this study.

Hill (2002) reanalyzed the faunal assemblage of bison remains from the kill site. The recovered specimens, including male, female, and juveniles, were all considerably larger than modern bison and likely represent B. antiquus. The size of the faunal assemblage points towards

a Late Paleoindian date, which Hill (2002:331), places at around 10,000 14C (<12,00 cal yrs BP)

years ago. Curation of excavated bison remains was selective during the original excavation but

based on the curated collection, Hill (2002) identified at least 33 individuals. However, this is

likely a fraction of the number killed of which the bones were not collected. While two hearths were identified at Milnesand, no reliable radiocarbon dates were obtained. A single date from one of four hearths at the Williamson-Plainview (or Ted Williamson site) site located about 500 meters away dates to 10,282±80 14C years BP (12,180 to 11,935 cal yrs BP) was recovered. The

Williamson-Plainview site is another late Paleoindian bison kill that contained 152 projectile

points which, interestingly, are mostly Plainview with a few Milnesand. All of these points were

kept by the landowner. Apart from both sites containing Late Paleoindian unfluted lanceolate

points, a possible occupational association between the two sites is unknown.

Gerald Shelton Collection

The Nebo Hill/Sedalia points used in this study come from two unnamed sites located about a mile apart, outside the town of Knob Noster in Johnson County, west-central Missouri.

The specimens are from the collection of Gerald Shelton, a private collector who documented the location of the sites on private property before the material was donated to Dr. David Meltzer at

Southern Methodist University. Apart from the location, additional information about the sites, including site function, is not known. Only the 14 complete points were included in this study.

Both Nebo Hill and Sedalia points are dated by O’Brien and Wood (1998) at ~4,000 to ~ 2,600

BP.

Table 3.1. Overview of Archaeological Sites Analyzed in the Plains Study.

Projectile Number of Point Radiocarbon Age cal Geographic Points in Site Type Function Years BP Region Study Late Paleoindian Bison Kill and 12,500 to Northern Mill Iron Goshen Processing Camp 10,450±15 12,250 Plains 16 12,373 to Southern Plainview Plainview Bison Kill 10,170±100 11,339 Plains 10 10,230±160 to 12,240 to Southern Bonfire Shelter Plainview Bison Processing 9,920±150 11,205 Plains 3 Southern Milnesand Milnesand Bison Kill ~10,000 ~12,000 Plains 12

Middle Archaic

Nebo Hill/ Eastern Plains Gerald Shelton Sedalia Unknown None None Edge 14

The first goal of this pilot study is to determine whether variations in flake scar patterning

are discernable between projectile point types. The second goal, if the first is successful, is to address cultural questions surrounding the Late Paleoindian projectile point assemblages. The expectation is that flake patterns associated with biface production should look the same between related groups of people. If the Fourier coefficients from the flake scar patterning of Goshen,

Plainview, and Milnesand Late Paleoindian types all form an indistinguishable point cloud from one another in PCA and LDA, this would suggest that they share a learning tradition. If, however, the analysis shows that they can be distinguished from one another, it would suggest they do not. The geographically and temporally isolated Middle Archaic lanceolate assemblage should form a point cloud by themselves and separately from all the other Late Paleoindian type assemblages as it represents a separate learning tradition. If met, this would indicate that topographic morphometric analysis can distinguish flake scar patterning on projectile points even though they share a common morphological shape: lanceolate. If this methodology is validated,

analysis can then proceed to comparisons between the three Plains Late Paleoindian projectile

point types. The results of comparisons within and between these point types will be used to

infer possible cultural relationships (or the absence of relationships) between these Paleoindian

complexes.

Application of Topographic Morphometrics on the Assemblage

In their 2012 study, Sholts et al. only used a single height measurement of ±¼ max

thickness to produce isocontours. They state that ±¼ max thickness was chosen as this isoheight

captures morphological variation due to production and not the effects of resharpening closer to

the edge. However, they do not go into detail, nor do they discuss whether they explored other

isoheights before settling on the ±¼ measurement. Different isoheight contours might produce

different shape outlines from the same face of a biface. Exploring this topic is important as

different isoheights might better capture variation in outline shape that is directly related to the

knowledge and technique of manufacture. It is possible that different isoheights capture the

variation equally well, which would allow the researcher more flexibility when analyzing points

that might be partially damaged or exhibit greater amounts of retouch. For this reason, isocontours of points for topographic morphometrics were taken at both ±¼ and ±⅓ max thickness from the split plane. This was done to test whether one measurement better captured the flake scar patterning resulting from biface manufacture. As the split plane z value is 0, ±¼ max thickness will be closer to the edge of the object while ±⅓ max thickness will be closer to the center (see Figure 3.2). Another way of thinking about this would be like envisioning a topographic map. The split plane is sea level and each face represents a mountain range. If the max thickness of an object is 12mm, then ¼ = 3mm and ⅓=4mm. Therefore, from the split plane

(0, or sea level) the ±¼ will have a contour height of 3 while the ±⅓ will have a contour height of 4 (closer to the center of the face, or to the maximum height of the mountain range). As a result four isocontours were produced that are associated with each 3D object, two isocontours for each face of a given projectile point.

Figure 3.2. Isocontour height measurement. ±¼ isocontour in yellow, ±⅓ isocontour in blue.

A study population of complete projectile points provided a unique opportunity to go one step further and examine whether certain elements of projectile points are better than other

elements for analyzing topographic morphometrics. Complete artifacts from the ±¼ measurement assemblage were “digitally broken” to separate the haft element from the blade element of the projectile point. The haft element was then subjected to the same topographic morphometrics as the complete specimens from the same assemblages. This was done for two reasons: 1) Many projectile point collections include broken specimens in addition to complete points. By incorporating projectile point bases or haft elements of broken specimens, the sample size and overall applicability of this methodology to archaeological specimens can be greatly increased. 2) As the base element of projectile points usually lies within the hafted portion of the spear, dart, or arrow, it is protected from resharpening and curation events that the blade element can be subject to. While a complete point may contain more data than just the haft element of a point, measurements from haft elements might contain data of higher quality in terms of representing the original flake reduction patterns resulting from production rather than later flaking patterns resulting from rejuvenation or other curation processes. Recent morphometric studies of projectile points by O’Brien et al. (2014) and Thulman (2012) have identified the base or haft element as the segment of a projectile point that contains the most variation in object form and thus provides better discrimination between projectile point types than do other portions of a point.

The Late Paleoindian specimens in this study were examined for basal grinding, and the location of grinding, if present, was recorded. When grinding was present on a point, the location closest to the point tip where grinding terminated was used to “digitally break” the point into its haft element. However, not all specimens in this study exhibited basal grinding. Moreover, weathered specimens sometimes made assessment of basal grinding impossible. When this was the case, or if grinding simply was not present, the widest point on the biface was the location

used to “digitally break” the specimen into the haft element (Figure 3.3) using the RapidWorks software. This creates a straight-line cutoff of the polyline contour at the haft break when conducting the topographic morphometric measurements, as opposed to completed specimens which extended to the tip of the blade element. The “digital break” is present on all specimens where the haft and blade elements meet, which standardizes this shape across all specimens, thus, making comparison of these isocontours possible.

Figure 3.3. Showing complete versus the “digitally broken” haft isocontours from the same Goshen point.

Ultimately, this study directly tests two methods for collecting topographic morphometrics from the same population (complete projectile points and “digitally broken” projectile points) at two isoheights or measurements (±¼ and ±⅓ max thickness from split plane), to see whether one method and/or measurement location was better at capturing flake scar variation than others. Each isoheight measurement and method will be analyzed and addressed individually in separate sections before a final synopsis of all is given at the end of the chapter.

Figure 3.4. Sample of haft isocontours of the different projectile point types used in this analysis. A, Goshen; B, Plainview; C, Milnesand; D, Nebo Hill/Sedalia.

Because flaws are sometimes present in projectile points, such as raw material inclusions, human error in production, and resharpening (to name a few), perfectly splitting a biface in half

when bifaces are not symmetrical is not always possible. Because of this, if an isocontour did not

accurately trace the isoheight of ±¼ or ±⅓ at the center of a face and extended to the edge of a

projectile point (and episodes of curation) it was not included for later topographic

morphometrics. Another reason for not including an isocontour in an analysis is that for some

projectile points one face simply did not have flake scars to examine. This is the case for points

that were made on flake blanks that were sufficiently flat and where there was no need to

bifacially work both faces. While all the points scanned from these assemblages are included in

this study in Appendix not all were necessarily used in the analysis of the two methods or two

isoheight measurements, which will be discussed later.

Complete ¼ Measurement

The first isoheight measurement to be tested using this methodology is the ±¼ max

thickness on complete points. This was the isoheight chosen by Sholts et al. (2012), as the

location of isocontours at ±¼ max thickness are sufficiently far enough away from the edge and

from any episodes of retouch that may have occurred. Isocontours from the five sites in this

study were imported into the Momocs package in R and converted into Fourier coefficients.

These Fourier coefficients, which describe the shape of the isocontours, were then subject to

multivariate analyses. The first multivariate analysis was PCA to see: (1) how much variation in

shape was present, (2) the amount of variation explained by each principle component (PC), and

(3) whether there was separation between the projectile point types (Figure 3.5).

Figure 3.5. PCA plot of Complete ±¼ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles.

The PCA plot is informative in a number of ways. To start with, the projectile shapes that

form the backdrop of the PCA plot are generalized outlines of the flake scar isocontours. This

presents the reader with the ability to compare which characteristics of the isocontours are

affecting the formation of point clouds, what is known as the morphospace. If a point falls in the

center of a backdrop outline, then the isocontour looks essentially like this generalized shape. If a

point falls between two or even three of the backdrop shapes it will look closest to the nearest.

Examination of the PCA plot reveals that distinguishable and indistinguishable point clouds are

occurring. This is important for a number of reasons as these point clouds are meaningful and support the expectations stated earlier. First, the Middle Archaic Nebo Hill/Sedalia points used here as a control for lanceolate shape, which are separated temporally and spatially from the Late

Paleoindian assemblages, form a single and distinct point cloud amongst themselves that is separate from the Late Paleoindian assemblage. Second, the Late Paleoindian points, which

potentially could share a learning tradition, form a single point cloud and are indistinguishable

from one another. The Late Paleoindian points are being separated from the Middle Archaic points along the first component (PC 1), which explains 88% of the variability in isocontour shape. There does not appear to much meaningful separation along PC 2.

It is important to remember that the PCA presented here differs from a traditional PCA.

There are no variables that can be referenced to explain the factors that are influencing the PC loadings in the data. Rather, the input data for the PCA is simply the Fourier coefficients that describe differences in shape outline of the isocontours. Therefore, the outline shapes in Figure

3.6A show the characteristics of the flake scar patterning that are driving each of the first four components by the mean and standard deviation, as well as the overall contribution of each component in in explaining the variability.

Figure 3.6. Complete ±¼ principle component contribution and scree plots.

For example, when looking at the complete ±¼ method, PC 1 accounted for 88% of the variation while PC2 accounted for 6% (Figure 3.6B). It appears that PC 1 is describing the width of isocontours and the basal shape, while PC 2 is describing where the widest point along the length of the isocontours is and the presence of an ear (Figure 3.6A). These same isocontour characteristics can be seen in the PCA (Figure 3.5) where the backdrop of point outlines shows the possible variation in isocontour shape and where each point from the study falls. This shows how the flake scar patterning from the Nebo Hill/Sedalia points are separating from the Late

Paleoindian points. PC 3 and PC 4 explain much less of the variation, but based on the PC

contribution plot (Figure 3.6A) it appears that these components describe how rounded or flat the

base element is, the appearance of “ears” on the flake scar isocontours, and how pointed the tip

is.

Linear Discriminant Analysis

Although LDA does maximize the separation between classes (projectile point types in

this study) the results support those seen in the PCA (Figure 3.7). The formation of point clouds

by type occurred in the same means, maintaining separation between Middle Archaic and Late

Paleoindian groups, while being indistinguishable within Goshen, Plainview, and Milnesand

types. The greatest separation occurs along LD 1 which explains 92.6% of the variation and

separates the Nebo Hill/Sedalia points from the Late Paleoindian types. In addition, the

effectiveness of the projectile type assignments using LDA was examined by comparing test

observations of polyline outlines from a training dataset to the known archaeological type.

Figure 3.7. LDA plot of Complete ±¼ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles.

The percent of projectile point polyline contours that were correctly assigned their given archaeological “type” by projectile set can be seen in Table 3.2. Nebo Hill/Sedalia points are reclassified with greatest accuracy, over 92%, which is also visible by the separate point cloud in

Figure 3.7. This is important as it indicates the flake scar patterning from these points greatly differs from the others. Although the Nebo Hill/Sedalia share a lanceolate form with the Late

Paleoindian points, the flake scar patterning indicates these two groups are not similar. The accuracy of the reclassification of Late Paleoindian accuracy is much lower, ranging from 24% to 50%. Of the 27 Goshen faces, one-third are misclassified as Plainview. The same is true of

Plainview, which are misclassified as Goshen 52% of the time. While Milnesand has the highest

correct reclassification accuracy, these points are misclassified as Goshen and Plainview at 25%

and 15%, respectively. This high frequency of misclassification is to be expected if the flake scar

patterning from point manufacture is similar between the point types.

Table 3.2. Plains Complete ±¼ LDA Predictions and Confusion Matrix.

Reclassified Actual (54.5%) Goshen Milnesand Plainview Nebo Hill/Sedalia Total Goshen (48.1%) 13 5 9 0 27 Milnesand (50%) 5 10 3 2 20 Plainview (24%) 13 6 6 0 25 Nebo Hill/ Sedalia (92.9%) 1 1 0 25 27

PCA Variation Within Projectile Point Type

It would be expected that sides A and B of the same projectile point should plot close to each other because both sides almost certainly were made by the same flintknapper. There should be less variation between faces of the same point than between faces of different points.

To examine this, the Fourier coefficient scores were analyzed using PCA for each of the four projectile point types in this study (Figures 3.8-3.11). Principle component plots were used to

look at the relative distance between faces from the same projectile point, whether multiple

points and both faces were near each other, and whether large separations between faces of the

same point existed for each of the projectile types. In all PCA plots by projectile point type, the

point numbers are presented on the plots (i.e. p_0011) followed by the face, A or B, (i.e.

p_0011B). If only a point number is present and there is no face classifier, then only a single face

from that point was used in the analysis.

In examining the PCA plots of each projectile point type (Figures 3.8-3.11), points

circled in green are those whose faces are as close to, or closer to, each other than to another face

from a different projectile point. Points circled in red are considered outliers that are not only far

from their opposite face, but from any other point faces in the morphospace. As there are no measurable distances between the outline shapes of flake scars, the term outlier is used in this

study as a qualitative evaluation, not a statistical calculation, as defined above. Faces from the

same point sometimes plotted opposite from each other across the morphospace along one of the

PCs. The faces from these points are circled in blue. Point faces not circled in green, red, or blue

are those that that lie further from their opposite point faces than the next closest point face, even

if very close. If _a or _b (indicating the face of a point, A or B) does not follow the point number

(p_000X) then only one face was used in the analysis.

Figure 3.8. PCA plot of the Complete ±¼ Mill Iron Goshen projectile points to examine within type variability.

Figure 3.9. PCA plot of the Complete ±¼ Plainview and Bonfire Shelter Plainview projectile points to examine within type variability.

Figure 3.10. PCA plot of the Complete ±¼ Milnesand projectile points to examine within type variability.

Figure 3.11. PCA plot of the Complete ±¼ Nebo Hill/Sedalia projectile points to examine within type variability.

An interesting pattern became evident in the examination the PCA plots by projectile type. Upon first glance it would appear that these faces opposite each other along PC 1 or 2 in the morphospace are not very symmetrical as they plotted away from each other when they should be side by side. However, it appears that this pattern is due more to the way in which the isocontours are saved in the RapidWorks software and how they are imported into the R package

Momocs than it is a reflection of non-symmetry. Visualize a hypothetical projectile point with the isocontours on each face. If the base of a projectile point is thicker closer to one edge, the extra thickness will kick out the isocontour to one side on both faces. This might produce a left tail or “ear” when looking straight down on face A (Figure 3.12A). If the point is flipped over, the tail or ear is also to the left on face B when looking straight down (Figure 3.12B). However, if you could see through the projectile point at each isocontour, the ears extend out in opposite directions and this is how the RapidWorks software saves the isocontour information (Figure

3.12C). It is not saved from each face looking straight down, it is saved from one direction and depending on the face the isocontours will have ears in opposite directions that do not look symmetrical. In reality, the point is actually symmetrical between both faces. It is just how the software creates and imports the isocontours that creates this non-symmetry in points when plotted (Figure 3.12D). I believe that when faces of a point plot almost exactly opposite one another on a PCA plot, they should be investigated to identify any factors that might cause a situation like the one outlined above. The points from Figures 3.8-3.11 circled in blue exhibit these characteristics and are in fact symmetrical even though they do not plot near each other in morphospace. Therefore, variation seen in the faces from the same projectile point in the plots is not as large as first appears. These faces are in fact more symmetrical than they plot.

Figure 3.12. Depiction of the “non-symmetry” isocontours in point faces from those identified by blue circles and why they are in fact more symmetrical. A, Face_a straight down; B Face_b straight down; C, Looking at the isocontours through the point and how the software saves the two isocontour outlines; D, Actual depiction of their symmetry.

The expectation is that both faces of a single projectile point should plot right next to each other, because both sides were probably made by the same flintknapper. As the complete

±¼ PCA plots from the Goshen, Milnesand, Plainview, and Nebo Hill/Sedalia show (Figures

3.8-3.11), faces from the same point are paired together. This includes the point faces identified in blue which, based on how the program imports the outlines, are argued to actually be more symmetrical than plotted. However, sometimes large variation between faces of the same point was evident. When examining the plot, it is easy to see why these points are outliers, (identified by red circles) based on the closest backdrop outline the face plots next to in the PCA morphospace. Often these shapes are furthest from the general point cloud and what would be described as the mean type shape. These points, upon further analysis, demonstrate certain

characteristics that alter the morphology of the projectile point. One cause of this is simply different flaking patterns resulting from a flaw during initial manufacture, resharpening, or use

(Figure 3.13).

Figure 3.13. Effects of curation on the isocontours. Removal of flakes from the left edge of the point, as well as the impact scar and step fracture near the center of the object has altered the isocontour edge leaving it concave.

The second cause of this variation is the result of bifacial asymmetry. Bifaces are not perfectly symmetrical on both faces because of various factors, including “dog legging.” Every

attempt was made, when splitting projectile points into two faces during the processing of scan imagery, to split each specimen perfectly in half along the projectile point edge. However, when

points are non-symmetrical, this asymmetry can push the isocontours closer to the edge and to retouch when the point has been split. This “dog legging” towards one edge (Figure 3.14) is evident during the image processing stage. Because points are not perfectly symmetrical, splitting a biface in half along the edge and subsequently taking topographic morphometric measurements from this split plane can cause differences in measurements between faces.

Figure 3.14. The “dog legging” effect of a non-symmetrical biface on the isocontours. This is point p_0019 in Figure 4.9 of the Complete ±¼ Milnesand PCA plot. These isocontours are the outliers identified as p_0019A and B.

After examining the within type variation of all four projectile points, the PCA and LDA plots of complete ±¼ projectile points (Figure 3.5 and 3.7 respectively) were revisited to see how this within type variation played out in the point clouds. To gain an understanding of what the greatest variation within the flake scar patterning entails in the multivariate plots, each within type PCA plot was used to identify the projectile point whose two faces were separated the most from each other and identified on the Plains PCA and LDA plots (Figures 3.15 and 3.16). Points that were not qualitatively identified as outliers (red), and/or those with issues related to non- symmetry (blue), where chosen as they show the greatest variation in a point that is not subject to an identified, or possible, bias. Identifying the greatest flake scar variation between two faces of the same projectile point within a point type creates a baseline by which to evaluate the point clouds. This was done to make sure that greatest amount of variation within a type was not equal to the size of the point cloud when all four types are plotted. If this were the case then one could not speak about the point clouds meaningfully. As can be seen in Figures 3.15 and 3.16, this is not the case. The greatest variation within types, when analyzed together, are fairly close together and are definitely not the size of the point clouds.

Figure 3.15. Revisiting PCA plot of Complete ±¼ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles. Circled point faces are those from the same projectile point that plot the furthest from each other from the within type PCA, excluding identified outliers and non-symmetrical faces.

Figure 3.16. Revisiting LDA plot of Complete ±¼ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles. Circled point faces are those from the same projectile point that plot the furthest from each other from the within type PCA, excluding identified outliers and non-symmetrical faces.

Projectile points with isocontours affected by dog legging, damage, or resharpening to the

blade element were removed from some or all of the later methodological analyses in this study.

These effects will impact the isocontours in different ways depending on the measurement

location type (Complete and Haft) as well as isoheight (±¼ and ±⅓). For example, if a projectile

point has edge damage on one face extending into the central flake scar patterning, that face

could not be used in the Complete analysis but it could be used in the Haft analysis as long as the

damage is limited to the blade element. Thus, not every face of every projectile point was subject to each methodological analysis. As a result, the total numbers of projectile point faces for each type presented in the discussion sections vary from analysis to analysis.

Single Face Analysis

To explore how non-symmetry affects topographic morphometrics, only a single face was chosen from each point for PCA and LDA. The face for each point was not randomly chosen, rather, each point file was examined to identify the point face that was closest to perfectly split in half. This single face analysis was conducted on the Complete ±¼ and ±⅓, as well as the Haft

±¼ measurements. The results of these analyses will be discussed at the end of the chapter when a comparison between the measurements, sample, and total population can be made at the same time.

Complete ±⅓ Measurement

To determine whether different areas on the face of a projectile point (different contour

elevations on a topographic map) influence the variation in flake scar patterning, isocontours

were analyzed from a second isoheight. In this section the isocontour measurement height of ±⅓

max thickness will be examined on the complete specimen population. This was done to test

whether the information held in the flake scar patterning from certain areas on a projectile point

face are better than other areas at distinguishing point types. Because the ±⅓ max thickness

pushes isocontours closer to the centerline of a point, more problems arose with this

measurement. This is so because on thinner specimens, or those that are non-symmetrical (for

any of the reasons discussed earlier), there sometimes was not enough of the surface of the point

face to detect flake scar patterning across the entire face. This is analogous to looking at the

topographic contours of North America for only those elevations above 12,000 feet. This leaves

out much of North America. Similarly, with some projectile points using the ±⅓ max thickness

excludes important data on flake scar morphology.

The PCA (Figure 3.17) and LDA (3.18) plots the point clouds in a similar manner to the

Complete ±¼ plots. There are two distinct point clouds in the data, the Middle Archaic Nebo

Hill/Sedalia and the Late Paleoindian, and there is good seperation between them which was

expected. This seperation occurs along PC 1, which explains the most variability at 88.2%. The

fact that all of the Late Paleoindian projectile points form a single, indistinguishable point cloud

was also expected. The LDA shows simiar results with two distinct point clouds. One is the

Nebo Hill/Sedalia points, and the second is the Late Paleoindian.

Figure 3.17. PCA plot of Complete ±⅓ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles.

Figure 3.18. LDA plot of Complete ±⅓ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles.

To see how well the LDA reclassified projectile point isocontours at ±⅓ max thickness

measurement, a confusion matrix was produced (Table 3.3). Overall, the accuracy is about 10%

lower than the LDA reclassification of Complete ±¼ points. Many of the Late Paleoindian points

were misclassified as Goshen points. This is apparent with the Plainview points that were

correctly reclassified only 5.8% of the time. While it is expected that related projectile points

will form indistinguishable point clouds from one another, the drastic decrease in the

reclassification accuracy of Milnesand and Plainview points from the Complete ±¼ LDA is

cause for concern. The fact that the accuracy dropped in the Nebo Hill/Sedalia points as well,

which should be clearly separated based on temporal factors, speaks towards a problem in the

±⅓ data. These results indicate that the ±⅓ measurement is decreasing the quality of the

information collected from the flake scar patterning and is not helpful in distinguishing between types.

Table 3.3. Plains Complete ±⅓ LDA Predictions and Confusion Matrix.

Reclassified Actual (45.8%) Goshen Milnesand Plainview Nebo Hill/Sedalia Total Goshen (52%) 13 4 8 0 25 Milnesand (38.5%) 6 5 3 1 15 Plainview (5.8%) 11 3 0 0 14 Nebo Hill/ Sedalia (82.4%) 0 2 1 14 17

To ascertain whether changing the isoheight measurement influenced how similar the two

faces from the same point plotted, and whether variation between faces was decreased or

increased, PCA plots were run again by projectile point type. The variation between faces from

the same point actually increased at the ±⅓ isocontour measurement. Not only did the distance

between faces from the same points increase, fewer numbers of matching point faces occurred,

and more outliers appeared in the data. These are highlighted in two PCA plots of the Plainview

(Figure 3.19) and Nebo Hill/Sedalia (3.20) assemblages. Another pattern that was lost is the

number of points identified in blue as was discussed in the previous section, which represent

faces from the same point plotting opposite across the PC morphospace. Points identified in blue

are plotted away from each other as a result of how the software writes and imports the data. In

fact, they are most likely much more symmetrical than appears on the plot. However, the number of points with faces that plot opposite each other is lower when examining the ±⅓ isocontour measurement, just like the number of faces from the same point that are plotted next to each other.

Figure 3.19. PCA plot of the Complete ±⅓ Plainview and Bonfire Shelter Plainview projectile points to examine within type variability.

Figure 3.20. PCA plot of the Complete ±⅓ Nebo Hill/Sedalia projectile points to examine within type variability.

As in the previous section of Complete ±¼ points, the projectile point with greatest separation between faces was identified to revisit the PCA and LDA plots of all four projectile point types. While the separation between these faces that exhibit the greatest variation by type is larger in the PCA at ±⅓ (Figure 3.21), they are still nowhere near the size of the point clouds.

However, when examining the LDA plot (Figure 3.22) the separation between faces is greater.

Especially the separation of Goshen faces which are about the size of the point cloud. Compared to the separation of Complete ±¼ points, it appears that the Complete ±⅓ measurement introduces greater variation to the flake scar patterning within each projectile point type.

Figure 3.21. Revisiting the PCA plot of Complete ±⅓ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles. Circled point faces are those from the same projectile point that plot the furthest from each other from the within type PCA, excluding identified outliers and non-symmetrical faces.

Figure 3.22. Revisiting the LDA plot of Complete ±⅓ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles. Circled point faces are those from the same projectile point that plot the furthest from each other from the within type PCA, excluding identified outliers and non-symmetrical faces.

Results indicate that the Complete ±⅓ max thickness measurement is inferior at

collecting information that relates to similarities and differences in the flake scar patterning from

manufacture, as compared to the Complete ±¼ measurement. Although the PCA separates the

Middle Archaic from the Late Paleoindian lanceolate points, LDA reclassification misclassified

the projectile points by type more frequently, and the accuracy was 10% worse than the

Complete ±¼ measurement. In addition, the variation between faces from the same projectile

increased, which led to fewer faces from the same point being plotted next to each other, an

increase in outliers, and greater separation of maximum variation in faces from the same

projectile point when the PCA and LDA plots were revisited. The poorer results are the product

of the measurement. Because the ±⅓ max thickness measurement pushes the isocontour closer to

the centerline of a face, more of the similarities and differences in the flake scar patterning

resulting from manufacture are lost.

Haft ±¼ Measurement

The final method and measurement tested in this pilot study is the Haft ±¼ max thickness isocontours. After comparing the topographic morphometric analysis of the Complete ±¼ and

Complete ±⅓, the results indicated that information relating to the similarities and differences in flake scar patterning were lost at the ±⅓ max thickness measurement. This was due to the fact the isocontours were pushed too close to the centerline of the projectile face where more variability in the patterning occurred due to a loss of area holding potential information.

Therefore, only the Haft ±¼, and not ±⅓, isocontours will be analyzed here. The haft measurements were taken by “digitally breaking” the scan of complete points at the convergence

of the haft and blade element. It is expected that the haft isocontours would be better than

complete isocontours because the blade element where “dog legging” can occur due to non- symmetry, curation, and damage has been removed from analysis. It is hypothesized that the haft element of a projectile point offers a more pristine flake scar patterning resulting from point manufacture.

Multivariate analysis from PCA (Figure 3.23) and LDA (3.24) indicate that analyzing topographic morphometrics from the haft element is a viable option in discerning similarities and differences in flake scar patterning between projectile point types. There is still good seperation between the point clouds of the Middle Archaic Nebo Hill/Sedalia points and the Late

Paleoindian points in the PCA plot. While there are outliers both in the PCA and LDA (Figure

3.24), the point clouds from the Haft ±¼ LDA is as good as, if not better than, the Complete ±¼ isocontour measurements. The within type point cloud between the Nebo hill/Sedalia and the

Late Paleoindian is one of the tightest and the seperation between the two clouds is better than the Complete measurements.

Figure 3.23. PCA plot of Haft ±¼ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles.

Figure 3.24. LDA plot of Haft ±¼ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles.

The separation between point clouds of both the Nebo Hill/Sedalia and Late Paleoindian clouds is evident from the LDA reclassification confusion matrix (Table 3.4). Not only is the overall reclassification accuracy the highest of all the methods tested at 59.8%, the accuracy within the Late Paleoindian points is also the highest. This fits with the expectations for the haft method, over analyses of the complete measurements, as factors that could affect the isocontours

of the blade element are potentially removed when solely examining the haft element. These

include “dog legging,” damage, and curation to the blade element. While an accuracy of 59.8%

might seem low, it is important to think about why this might be so. If projectile point types are

part of the same learning tradition with a similar flintknapping knowledge and technique, then it

is expected their isocontour flake scar patterns should look similar. If this is the case, then LDA

would not be able to reclassify the point types with as high of an accuracy as would be expected if they did not share a learning tradition. The expectation for separate learning traditions is that the flake scar patterning from these point types would look different and LDA would reclassify them correctly with a higher accuracy. In fact, this is what the LDA results in Table 3.4 show.

The Plains Late Paleoindian point types (Goshen, Plainview, and Milnesand) have a low reclassification accuracy. The fact that the geographically and temporally (~6,000 years) separated Nebo Hill/Sedalia reclassification accuracy is over 83% indicates that this methodology can identify similarities and differences in flake scar patterning on projectile points.

Table 3.4. Plains Haft ±¼ LDA Predictions and Confusion Matrix.

Reclassified Actual (59.8%) Goshen Milnesand Plainview Nebo Hill/Sedalia Total Goshen (53.8%) 14 5 6 1 26 Milnesand (47.4%) 2 9 6 2 19 Plainview (52.2%) 8 3 12 0 23 Nebo Hill/ Sedalia (83.3%) 0 4 0 20 24

In examining the morphospace of the Haft ±¼ PCA plot (Figure 3.23), the point clouds of

the Late Paleoindian components and the Middle Archaic component are not as distinctly

separate along PC 1 as that of the Complete ±¼ plot (Figure 3.5). While the three Late

Paleoindian point clouds remain clustered indistinguishably, the distinct separation between the

Middle Archaic point cloud is not as strong as in the Complete ±¼ plot. I believe that this loss of separation between the Late Paleoindian and Middle Archaic point clouds does not indicate that the haft of projectile points contains poorer information concerning learning traditions. Rather, I believe that this loss of separation occurs because the haft has been “digitally broken” and the polyline contour is standardized across the face at the haft. While this standardizes the measurement and makes topographic morphometrics applicable at the haft location across all types, it also creates a portion of the outline that is now shared by all points regardless of point type, and thus learning traditions. This introduces a bias to the analysis that is identified in the

PCA and is visualized in the plot as the loss of space between these previously distinct point type clouds along PC 1. This hafting effect does not appear to influence the LDA plots in the same way as it does the PCA plots. Although there is a decrease in the LDA reclassification accuracy of the Nebo Hill/Sedalia points and an increase in reclassification accuracy within the Late

Paleoindian components. Nonetheless, the clear separation between the Middle Archaic component point cloud and the point clouds of the Late Paleoindian components seen in the

LDA plot at Complete ±¼ also is present in the Haft ±¼ LDA plot. In addition, the Haft ±¼

LDA shows as an overall increase in the reclassification accuracy. Although there is some overlap in the Late Paleoindian and Middle Archaic point clouds in the Haft ±¼ PCA plot, when the plot is taken into account with the LDA results, topographic morphometrics can be used successfully to identify separate learning traditions using the Haft ±¼ measurement.

To further examine how well the haft element is at collecting information on flake scar patterning information by type, PCA plots were created, as with the other measurements, to examine the variation between faces of the same projectile point. Variation, the distance between

faces from the same point, is expected to be low based on the characteristic difference between haft and blade elements and the factors that influence isocontour shape. This decrease in variation would be seen in the number of faces from the same point plotting close together

(including those identified with blue circles) as well as a decrease in outliers (circled in red).

These expectations are seen in the Haft ±¼ PCA plots by type, similar to those of the Complete

±¼ measurement. The Haft ±¼ measurement is better than the Complete ±⅓ measurement, with more faces from the same point plotting next to each other, and along with fewer outliers.

Figures 3.25 and 3.26 present the Haft ±¼ PCA plots for the Goshen and Milnesand points respectively.

Figure 3.25. PCA plot of the Haft ±¼ Mill Iron Goshen projectile points to examine within type variability.

Figure 3.26. PCA plot of the Haft ±¼ Milnesand projectile points to examine within type variability.

The greatest variation in flake scar patterning between two faces of the same projectile point were again identified from the within type PCA plots. The multivariate PCA and LDA plots for all of the Haft ±¼ projectile points (Figures 3.23 and 3.24) were revisited with the faces showing the greatest separation identified in the plots. Figure 3.27 is the PCA plot of the Haft ±¼ measurement of the four projectile point types. The separation between faces that show the greatest variation in flake scar patterning within each point cloud is small, as with the Complete

±¼ analyses. In addition, the separations are not close to the overall size of the point clouds.

Therefore, the discussion surrounding point clouds in the PCA plot and their relationship to one another is meaningful. The revisited LDA plot is also very interesting (Figure 3.28). The

separation between faces of the same type is better than that of the revisited Complete ±⅓ LDA.

Although there is separation between faces, it is not the size of the point clouds. What is especially interesting is that three of the eight faces could be qualitatively described as outliers

(Goshen, Plainview, and Nebo Hill/Sedalia). As LDA maximizes separation among groups, the fact these faces show the greatest within type variation and fall outside the point cloud as

“outliers,” indicates the remaining points in the clouds exhibit less variation.

Figure 3.27. Revisiting the PCA plot of Haft ±¼ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles. Circled point faces are those from the same projectile point that plot the furthest from each other from the within type PCA, excluding identified outliers and non-symmetrical faces.

Figure 3.28. Revisiting the LDA plot of Haft ±¼ projectile points. Goshen, red circles; Plainview, light blue diamonds; Milnesand, purple squares; Nebo Hill, green triangles. Circled point faces are those from the same projectile point that plot the furthest from each other from the within type PCA, excluding identified outliers and non-symmetrical faces.

Similar to the Complete ±¼ max thickness measurement, the Haft ±¼ element PCA plots by type (Figures 3.25 and 3.26) show good grouping of faces from the same projectile point across types. There are still some outliers, indicated in red, but that can be expected and will be discussed further in the next section. The PCA and LDA are good at separating the Middle

100

Archaic Nebo Hill/Sedalia and the Late Paleoindian types. While the Late Paleoindian points

were indistinguishable as to individual point types when examining the PCA, the LDA was very

informative. The Late Paleoindian types were separated from the Middle Archaic Nebo

Hill/Sedalia points along LD 1 which explains 88.6% of the variation. Based on the topographic

morphometric results conducted on all the measurements, especially the LDA reclassification

accuracy, the Haft ±¼ measurement is as good as, or better than, the Complete ±¼ and is better than the ±⅓ isoheight measurements at successfully identifying similarities and differences in flake scar patterning based on the expectations laid out for the Plains Late Paleoindian and

Middle Archaic assemblages. The ±¼ measurement, in general, isoheight appears to be better at capturing flake scar patterning information that is related to flintknapping knowledge and technique than the ±⅓ isoheight measurement.

Topographic Morphometric Results and Conclusions

This study expands on the novel, 3D object based, morphological approach to lithic

analysis of Sholts et al. (2012). Specifically, this study examines how flake scar patterning on

projectile points that results from manufacture can be analyzed to identify similarities and

differences in flintknapping that are, theoretically, related to flintknapping knowledge and

technique that is subsequently transmitted between generations. This study examined multiple

polyline isocontour measurements to determine which of these was most successful at

identifying similarities and differences in flake scar patterning among several projectile point

types from the archaeological record. These digital cross section measurements were taken at ±¼

max thickness and ±⅓ max thickness of the biface from the split plane on both complete

specimens and their haft elements that were “digitally broken” from the same artifact.

101

Results

Across methods and isoheights the first PC describes the basal shape and width of the isocontour. The second PC seems to describe the location of the widest point of the object, moving from closer to the base to closer to the tip. Based on the PCA results, and the LDA reclassification confusion matrices, isocontour measurements at Complete ±¼ max thickness were more successful in forming point clouds of known types and in assigning isocontours to the correct typology by site (54.5%) than the Complete ±⅓ measurement (45.8%).

The creation of a haft line by “digitally breaking” a flake scar pattern contour at the haft introduces a bias that was identified in the pilot study. This is the result of every projectile point, regardless of type, now sharing an artificially made segment of the polyline contour. This manifests itself in the morphospace of the PCA plot as a slight overlap between Middle Archaic and Late Paleoindian component point clouds along PC 1. Nonetheless, the PCA and LDA results still demonstrate that similarities and differences between flintknapping knowledge and technique can be identified at the haft measurement, and that the Haft ±¼ measurement is as

good as, and even better than (in LDA reclassification at 59.8%), the Complete ±¼

measurement. Therefore, testing topographic morphometrics from the haft element alone is as good, if not better, than analysis of complete points, as curation and resharpening of the blade element has the potential to greatly affect the outline shape of the flake scar isocontours.

If one imagines the morphological outline of a non-curated point, or one with little

resharpening, the blade element will be longer, thicker, and will come to a pointed tip. Now

imagine the blade element of the same point after it has been heavily resharpened. The curated blade element of the point will be “pushed back” towards the haft element of the biface and will be narrower due resharpening. These two shapes reflect different stages in the use-life of a single

102

projectile point. In an analytical method that looks for variation in outline shape, these two point

shapes would produce very different outlines, even though they represent a single projectile point

at different stages in its use-life. Because EFA looks at variation in outline shape, curation could

potentially have a huge impact on the results depending on the number of assemblages, the

number of specimens per assemblage, and the states of curation by specimen. Haft element

outlines, protected from curation episodes, still contain the flake scar isocontour information

resulting from manufacture, yielding an increased accuracy of identifying point types (Figure

3.29).

Another factor contributing to the increased accuracy of the haft isocontours is the

removal of “dog legging” from the blade element. This study demonstrated that the haft element

contains enough of the flake scar information to successfully apply topographic morphometrics

as do complete points. In fact, the haft element does as good a job of distinguishing point clouds by type, and in separating data from PCA and LDA into correctly reclassified projectile types as

complete points. This is promising information as including haft isocontour allows

archaeologists to look at assemblages of broken points rather than being restricted to topographic

morphometrics on complete points. Many assemblages typically include a high proportion of

broken specimens and a lower number of complete points, and analysis of just the haft element

enables researchers to expand their sample sizes.

103

Figure 3.29. The characteristics that can affect the isocontours of complete points and how they are removed by examining only the haft element only.

Results of the Single Face Analysis

Because of the possible variation between faces from the same projectile point, PCA and

LDA were run from a single face from each point for Complete ±¼ and ±⅓, and Haft ±¼

isocontours. The expectations for the single face analysis compared to the analysis of both faces is that the single face sample should increase the accuracy of LDA reclassifications for complete

points. This is because the “best” face should be a better representation of the type and should

104

therefore be easier to reclassify. For the haft element measurement, it would be expected that the

LDA classification accuracy would remain equal to or only slightly increase because analysis of only the haft eliminates effects of damage, curation, and other man-made or natural characteristics that can affect the blade element. These expectations are met at the Haft ±¼, with only a slight increase of 1.4% reclassification accuracy. However, single face LDA reclassification did not meet the expectations and actually decreased the accuracy in the complete ±¼ and Complete ±⅓ measurement by -1.3% and -2.9% respectively. The results from single face PCA and LDA were very similar to the results of the entire population (Table 3.5).

This could be because there is enough variation in the blade element that selecting a single face from the two faces does not matter. There is still variation in the blade element and the faces chosen for this analysis decreased the reclassification accuracy as the sample size was smaller.

The fact that this only had a slight impact, both positively and negatively, is important. If the decrease in reclassification accuracy was drastic it would be cause for concern. These results indicate that the Haft ±¼ remains a preferred isoheight method and measurement to collect flake scar isocontour information. The slight increase between the single face analysis and analysis of both faces, when present, does not significantly change the accuracy of the method.

Table 3.5. LDA Reclassification Accuracy by Method and Isoheight.

Complete ±¼ Complete ±⅓ Haft ±¼ Plains All 54.5% 45.8% 59.8% Plains All Single Face 53.8% 42.9% 61.2%

105

Conclusions

Although this is a pilot study with a limited sample size, the results are promising.

Results from this study indicate that topographic morphometrics using EFA of isocontours can

be successfully used to identify similarities and differences in flake scar patterning on projectile

points resulting from tool production. It is important to remember that the point clouds discussed

here are not simple outline shape morphologies of the overall projectile point shape. If this were

the case, then similar looking points (this study examined all large unfluted lanceolate points), based on outline alone, would be expected to group together because they look the same. Rather, these are highly detailed polyline representations of flakes scar patterning at defined measurements or isoheights. They can be thought of as analogous to contour elevations on a topographic map, so the formation of point clouds and distinctions discussed in the multivariate analyses are describing similarities and differences in the flintknapping technique that are a product of manufacture. This methodology demonstrated that variation in flake scar patterning occurs and that similarities and differences in flake scar patterning on lithic projectile points can be identified.

Expectations surrounding the formation of point clouds of possibly related and unrelated archaeologically types were met. The Late Paleoindian unfluted lanceolate components in question (Goshen, Plainview, and Milnesand) formed a single point cloud with some distinction from one another but were nearly completely separate from the Middle Archaic Nebo

Hill/Sedalia lanceolate point cloud. This is useful for examining the relationships between multiple groups of people, defined by projectile point types, based on the flake scar reduction patterns from manufacture. The isocontours of the Paleoindian components show a separate point cloud from the Middle Archaic component but are almost indistinguishable from each

106

other. This fits with the expectation that they shared a learning tradition with similar knowledge

in flintknapping and technique of manufacture. These patterns are expected to look more similar within a culture group that shares a learning tradition, while unrelated cultural groups, or those separated by more time, will look different. By demonstrating that this methodology can identify similarities and differences in flake scar patterning by projectile type, the relationship between projectile point types which are often used by archaeologists as a method to associate different groups of people or cultures in the archaeological record, can be made using the Haft ±¼ max thickness isoheight.

107

CHAPTER FOUR: LATE PALEOINDIAN ON THE GREAT PLAINS

This chapter takes the methodology developed in Chapter Three and applies it to the first

of two Paleoindian case studies. This case study delves further into the Late Paleoindian

projectile point types from the Great Plains introduced in Chapters One and Three. This chapter

begins by presenting some of the views and current data surrounding the relationship of these

three Late Paleoindian types: Goshen from the northern Plains, and Plainview and Milnesand

from the southern Plains. Topographic morphometrics are then applied to this dataset to provide an additional line of evidence that can be used by archaeologists to help understand the culture history on the Great Plains during the Late Paleoindian period.

Henry Irwin proposed in 1967 that the Plainview and Goshen types are in fact one and the same and should both be identified as Plainview. Recent geometric morphometric analyses by Haynes and Hill (2017) as well as Huckell and Merriman (2017) did not find any significant differences between the two types, which supports the assertion that Goshen and Plainview points are indeed the same and could be classified as Plainview. Huckell and Merriman (2017) go further and suggest that all of the Late Paleoindian unfluted lanceolate points could be thought of as a widespread Plainview culture or technology similar in distribution to that of

Folsom but slightly younger in age. Therefore, different Late Paleoindian unfluted lanceolate point types that have been subdivided into different types in the American Southwest, southern

Great Plains, and northern Plains reflect a common and widespread culture complex of local- scale groups and larger social communities comprising Plainview. These various unfluted lanceolate types represent temporal and geographic expressions of the same lithic technology or

“culture” during the Late Paleoindian period. That interpretation of these Late Paleoindian projectile point types can be tested through analysis of flake scar patterning. With the passage of

108

time and demographic shifts, variation in flake scar patterning is expected to arise in separate

geographic areas. However, by focusing at a specific point in time (Late Paleoindian)

contemporary and near-contemporary assemblages can be compared to examine the variation

between assemblages.

The ultimate purpose of this study is to examine flake scar patterning to identify the

relatedness of projectile types or “cultures” at a given point in time or space. Topographic

morphometrics can be applied in this case to answer two questions concerning these Late

Paleoindian complexes: 1) If a projectile point type is replaced by a different type in the same

geographic region, was this the result of population replacement (two distinct learning traditions

with different flintknapping styles and flake scar patterning on projectile points from

manufacture), or was it an in situ transformation in which case the flake scar patterning from

manufacture should look the same; and 2) Did two (or more) different, but contemporaneous,

projectile point types from adjacent geographic regions share a learning tradition (in which case

the flake scar patterning from manufacture should look similar), or do they represent unrelated

cultures? In which case different flintknapping styles are expected and flake scar patterning on

projectile points from manufacture should differ between the point types. This latter question is

the main question posed here about the Goshen and Plainview types. The purpose of the analyses

presented in this chapter is to determine whether these two Paleoindian complexes are indeed

related. If this is the case, the expectations is that the flake scar patterning of each point type should be similar to each other, and should form an indistinguishable point cloud. On the other hand, if Goshen and Plainview are not geographic variants of the same culture and are unrelated to one another, then the flake scar patterning of specimens within each of these types should form distinct and separate point clouds. The incorporation of the Milnesand points provides an

109

interesting, additional, line of evidence by which to test these ideas by examining the relationship

of Milnesand and Plainview points. Not as much is known about Milnesand points as is for

Plainview, but given their overlap in space and time in the southern Great Plains, it is expected

that the point clouds from these two types should overlap with each other if they are related. The

Milnesand assemblage presents and interesting case to by which to further examine the

emergence of multiple Late Paleoindian projectile point types following the earlier, Clovis and

Folsom cultures.

Bivariate Analysis

Both haft width, or maximum (max) width when edge grinding was not present, and max

thickness measurements were taken from the scanned projectile points. Blade element

characteristics and object weight change throughout the use-life of an artifact and will be

drastically different on broken specimens. However, even on broken points, when the haft

element is present this usually represents the widest point of a projectile and is close to the max

thickness center of the artifact. As all the points used in Plains Paleoindian Study were complete,

or at least contained the haft element, these two measurements can describe artifacts and their

isocontours across assemblages as these values are not subject to curation or breakage.

The bivariate plot of these two measurements is worth exploring as it can provide

additional context by which to interpret the PCA results (Figure 4.1). Multiple scenarios can be

identified that can guide inferences concerning the relationships between the Late Paleoindian

projectile point types. The first scenario would be distinct clustering of each Late Paleoindian projectile point type in the scatterplot that show similar flake scar patterning in the topographic morphometric analyses. If these point types vary morphometrically but were flintknapped in a

110

similar manner, this would support the idea that they are related. A similar conclusion could be drawn if two or more projectile point types cluster in the scatterplot indistinguishably and share a similar flake scar patterning. The opposite conclusion, that these projectile point types are unrelated, could be drawn if they vary morphologically and in their flake scar patterning, or if they did not vary morphologically but had different flake scar patterning.

Figure 4.1. Scatterplot of max width and max thickness for the Plains Late Paleoindian and Middle Archaic projectile points by type.

The results of the scatterplot show two important patterns. First, the Late Paleoindian projectile point types are clustered together indistinguishably from one another. Second, the Late

Paleoindian points are clearly separated from the later Middle Archaic points from the Gerald

111

Shelton collection. The Nebo Hill/Sedalia points also show considerable variation and are not as closely clustered with one another as are the Late Paleoindian points. This bivariate scatterplot indicates that the Late Paleoindian points are very similar to each other morphologically, which is not a surprise. To look for any significant differences, a max width to max thickness ratio was created and one-way ANOVA was conducted to compare mean ratio values between Late

Paleoindian types. No statistically significant difference was detected between projectile point types as determined by a one-way ANOVA (F (2.39) = 1.427, p = .252).

This bivariate scatterplot is only one of several ways that these projectile point types can be analyzed. Other analyses include other univariate tests, and geometric morphometrics.

Continued interest in Goshen-Plainview-Late Paleoindian unfluted lanceolate points from the

Great Plains and American Southwest have recently provided a renewed interest in this question.

Huckell and Merriman (2017) conducted univariate analyses on max width and thickness, width- thickness ratio, basal concavity, and basal taper on a number of Late Paleoindian unfluted lanceolate points. These assemblages came from New Mexico and elsewhere in the Southwest and from the Great Plains. The authors found that almost all the types they analyzed fell within the range of variation for a Late Paleoindian unfluted lanceolate point. The only points that did not fall into this range were the Goshen points from the Mill Iron site. However, Haynes and Hill

(2017) looked at a variety of morphological measurements from Goshen points at the Mill Iron site and Plainview points from the type site and concluded that it is metrically and morphologically hard to identify significant differences between the two. They argue that

Goshen and Plainview represent a single culture.

Holliday et al. (2017c), examined the relationship between Plainview from the type site and Belen points from New Mexico, which are a proposed geographic variant of Plainview. The

112

authors looked at max width and thickness, width-thickness ratio, basal concavity, basal width, basal concavity-max width ratio, and max length measurements. They concluded there was no clear difference between the two types and that Belen is a regional variant of Plainview.

Buchanan et al. (2007, 2017) used geometric morphometrics from 20 landmarks around the edge outline of points to look for variation in Late Paleoindian unfluted lanceolate points from the southern Great Plains. They concluded that while there are significant differences between

Plainview and Lubbock Lake points, they could not discern any significant difference between

Milnesand and Plainview points. They suggest that the Milnesand point type should be abandoned and that Milnesand should be classified as Plainview. However, others argue there are morphological differences, particularly in basal shape and flaking, which warrant their separation (Holliday et al. 2017b).

Huckell and Merriman (2017) do a good job in highlighting the different approaches that archaeologists have taken to the study of Late Paleoindian assemblages compared to the earlier

Paleoindian point types. Clovis, the earliest well documented Paleoindian projectile point, and the slightly later Folsom points, each are treated as a single, geographically widespread culture with Folsom lasting for a minimum of 300-400 years and up to 700 years between ~10,900 to

~10,200 14C years BP (Holliday 2000). Archaeological studies of Folsom points tend to focus on technology first (fluting), and shape second. In contrast, studies of unfluted Late Paleoindian projectile points tend to look first at projectile point shape, and secondly at technology. This approach can give rise to many more geographic types as the focus is on slight variations.

However, if the various Late Paleoindian unfluted lanceolate projectile points types are examined together as a single entity, such as is done for Folsom, a similar pattern appears. The geographic range of Folsom and of unfluted Late Paleoindians is approximately the same, the

113

age ranges are all slightly younger than Folsom (Goshen ~10,450 to ~10,175 14C years BP

(Waters and Stafford 2014), Plainview ~10,300 to ~9,900 14C years BP (Holliday et al. 2017b)),

and Folsom and Late Paleoindian last for a similar amount of time. Has the focus on the shape of

Late Paleoindian unfluted lanceolate points from the Great Plains and the Southwest caused a

splitting into multiple projectile point types that are otherwise part of a larger unfluted lanceolate

(or Plainview) tradition? Or, are these differences in shapes and the splitting into multiple types

appropriate, and do these multiple types represent the existence of multiple Late Paleoindian

groups in various geographic areas following population increase and adaptation to local environments? One way to begin addressing this question in a new light is with topographic morphometrics and how the flake scar patterning compares between these Late Paleoindian

projectile point types.

Great Plains Late Paleoindian Topographic Morphometrics

As was demonstrated in Chapter Three, topographic morphometrics of just the haft

element of projectile points were shown to be as good, or better, than topographic morphometrics

of the entire specimen. Therefore, only the haft ±¼ isocontours will be analyzed and discussed in

this chapter. The PCA and LDA are similar to those conducted in Chapter Three but now the

emphasis is solely on Goshen, Plainview, and Milnesand isocontours (Figure 4.2).

Of particular interest is how the first PC explains much less of the variability in the

Goshen, Plainview, and Milnesand points (50.7%) than it did when the Middle Archaic Nebo

Hill/Sedalia points were included (68.3%). On the other hand, PC 2 explained 33.2% and 21.9%

of the variability respectively. Less variability within these point types is explained in PC 1

114

because the isocontours of flake scars are more similar. This is expected if the projectile points were flintknapped using a traditional knowledge shared by a group of related peoples.

Figure 4.2. Plains Late Paleoindian PCA plot. Goshen, red circles; Plainview, green triangles; Milnesand, blue squares. Circled point faces are those from the same projectile point that plot the furthest from each other from the within type PCA, excluding identified outliers and non-symmetrical faces.

Apart from a few outliers, the point cloud in the PCA plot is indistinguishable by type even though possible variation from the temporally and spatially separate Middle Archaic Nebo

Hill/Sedalia points was removed. This meets the expectations set forth at the start of this chapter that these various unfluted lanceolate types represent temporal and geographic expressions of the

115

same lithic technology or “culture” during the Late Paleoindian period and strengthens the argument that Goshen, Plainview, and Milnesand all share a common learning tradition concerning flintknapping knowledge and technique. The variation seen within types, as seen upon closer inspection, is also important. The Goshen point cloud is smaller than the other two types and exhibits less variation in isocontour shape and thus, flake scar patterning in the PCA plot (Figure 4.2). The Plainview point cloud and the variation exhibited in the isocontour of this point type is larger and on par with Milnesand points. However, if the outliers are removed, then the point cloud is much tighter and the variation in shape is lower. The Milnesand point cloud exhibits more variation in shape in the PCA plot in Figure 4.2 than both Goshen and Plainview.

Again, LDA was used to examine how well the Late Paleoindian isocontours group given their assigned projectile point type (Figure 4.3). Similar relationships between the PCA and LDA can be seen. Whereas separation between projectile point types was not present in the PCA plot, there is separation between the Late Paleoindian data in the first two Linear Discriminants of the

LDA plot. Milnesand points are the most separated along the LD 1 axis, which explains 57.8% of the variation, while Goshen is separated from Plainview and Milnesand along the LD 2 axis, which explains 42.2% of the variation.

116

Figure 4.3. Plains Late Paleoindian LDA. Goshen, red circles; Plainview, green triangles; Milnesand, blue squares.

The contingency table (Table 4.1) shows how well Paleoindian points were classified, and where misclassified points fall among Goshen, Plainview, and Milnesand using LDA. The correctly reclassified percent of each type exceeds the expected percent for each type based on a random sample of the total population. In examining misclassifications, 19.2% of Goshen were classified as Plainview and 19.2% as Milnesand. For Plainview, 39.1% of the misclassified points were called Goshen while 13% were called Milnesand. Finally, in examining Milnesand,

21% of the misclassified points were called Goshen while 42.1% were called Plainview. LDA demonstrated that Goshen was reclassified most accurately, followed by Plainview, and finally

117

by Milnesand. When Plainview points were misclassified they were misclassified the most as

Goshen, and when Milnesand were misclassified they were misclassified the most as Plainview.

Table 4.1. Plains Late Paleoindian LDA Predictions and Confusion Matrix.

Classified Actual (50%) Goshen Milnesand Plainview Goshen (61.5%) 16 5 5 Milnesand (36.8%) 4 7 8 Plainview (47.9%) 9 3 11

The ways in which these misclassifications of points occurred could be meaningful when

taking into consideration with the temporal and spatial distributions of Goshen, Plainview, and

Milnesand. The Goshen complex from the northern Plains (Figure 4.4) has been dated from

10,450±15 to 10,175±40 14C years BP (12,525 to 11,700 cal yrs BP) from sites across Montana,

Wyoming, South Dakota, and Colorado (Waters and Stafford 2014). Plainview sites in the southern Plains have been dated between ~10,300 to ~9,900 14C years BP (~12,100 to ~ 11,300

cal yrs BP) from sites in Texas, Oklahoma, New Mexico, and Arizona (Holliday et al. 2017b).

Goshen points from the Mill Iron site are the oldest dated points from that complex, having an

average date of 10,450±15 14C years BP (12,525 to 12,125 cal yrs BP) (Waters and Stafford

2014). Radiocarbon dating from Plainview site produced a larger than expected range in dates.

Holliday et al. (1999) examined these dates and place the date of the Plainview site >10,000 14C

years BP (≥12,000 cal yrs BP), while Plainview points from the bison bed at Bonfire Shelter site date to between 10,230±160 to 9,920±150 14C years BP (12,240 to 11,205 cal yrs BP). Although there are no reliable dates from the Milnesand site, a single date of 10,280±80 14C years BP

(12,180 to 11,935 cal yrs BP) was recovered from the Williamson-Plainview site, located less

118

than 500 meters away containing mostly Plainview projectile points and some Milnesand points.

Further support for a circa 10,000 14C years BP date (~12,000 cal yrs BP) come from the bison remains, which are all considerably larger than modern bison and likely represent B. antiquus.

Figure 4.4. Geographic and temporal distribution of the Goshen and Plainview Late Paleoindian Complexes and the Plains Archaeological Sites from this study. 1, Mill Iron; 2, Plainview; 3, Bonfire Shelter; 4, Milnesand. Adapted from Surovell (2009) and Waters and Stafford (2014).

119

Great Plains Late Paleoindian Conclusions

The patterns previously discussed regarding flake scar morphometrics and a shared Late

Paleoindian learning tradition can be used in conjunction with spatial and chronological distributions of these sites to make inferences as to the relationships behind these projectile point types (Table 4.2 and Figure 4.4). The Goshen complex from the northern Plains is the oldest in age but overlaps temporally (younger dates) with the dates from Plainview sites on the southern

Plains. Milnesand, while not directly dated, is be placed about 10,000 14C years BP (≤12,000 cal

yrs BP) in the literature from eastern New Mexico, which is younger than the Plainview

component at the type site and at Bonfire Shelter but still within the age range of known

Plainview sites. The Goshen points are reclassified correctly more than Plainview and Milnesand

projectile points in LDA. While Plainview exhibits more variation than Goshen, as seen in the

PCA plot, it varies less than Milnesand. When Plainview is misclassified using LDA it is

misclassified more frequently as Goshen while Milnesand is misclassified more commonly as

Plainview. The topographic morphometric analysis indicates that these three Late Paleoindian

projectile points share a learning tradition based on the point cloud overlap in these types.

Table 4.2. Chronological Range of Late Paleoindian Types.

Radiocarbon Cal Age Range Type Age BP Location Source Goshen 10,450 to 10,175 12,525 to 11,700 Northern Plains Waters and Stafford 2014 Plainview ~10,300 to 9,900 ~12,100 to 11,300 Southern Plains Holliday et al. 2017a,b Milnesand ~10,000 ≤12,000 Southern Plains Holliday et al. 2017b

Taking this into account, patterns in geographic distribution and chronology are interesting for two reasons: 1) these patterns meet the expectations that isocontours of flake scar

120

patterning of the projectile point clouds would overlap and be indistinguishable amongst related

groups of people; and 2) these patterns suggest that Goshen points could possibly represent an

ancestral flintknapping knowledge and technique (learning tradition) that spread southward and

developed into what is recognized by archaeologists today as Plainview on the southern Plains.

Results of the flake scar topographic morphometrics reported here indicate that flake scar

patterning of Goshen points from the Mill Iron site are more similar to Plainview projectile

points than they are to Milnesand points. This interpretation is further supported by the

isocontours of Milnesand points, which are more variable than both Goshen and Plainview, and

are misclassified more often as Plainview than Goshen. Following these lines of reasoning,

Milnesand potentially represents a split from Plainview in New Mexico. These interpretations,

while provisional at this time, lend support to previous archaeological interpretations that have

tentatively called for Goshen and Plainview to be classified as a single complex, using other,

traditional analytical methods. While the radiocarbon record for Late Paleoindian sites on the

southern Plains leaves one wanting more, the general time line and distribution of sites

potentially indicates a southward movement of an older Goshen flintknapping knowledge and

technology, or learning tradition.

The oldest Goshen date comes from the Mill Iron site at 10,450±15 14C years BP (12,525 to 12,120 cal yrs BP), while Goshen dates to the south extend to 10,175±40 14C years BP (12,020

to11,700 cal yrs BP) (Waters and Stafford 2014). The calibrated age range for Goshen in the

northern Plains at this time is ~12,525 to ~11,700 cal yrs BP. This movement continued into the

southern Plains as “Plainview” between ~10,300 and 9,900 14C years BP (~12,100 to ~11,300 cal

yrs BP) (Holliday et al. 2017b). Spatial and temporal patterns from archaeological sites, although

limited in number, as well as the topographic morphometrics from this study that point towards a

121

shared learning tradition, can be used in conjunction to infer a possible ancestral “Goshen” flintknapping knowledge and technique that spread southward becoming what is today called

Plainview on the southern Plains.

Frison (1996) cautioned against grouping Goshen as Plainview until more was known about these Late Paleoindian complexes. The results of this study present new and unique data about previously studied artifact assemblages that can address questions concerning the cultural relationship between Goshen, Plainview, and Milnesand point types in new ways. Multivariate analyses show that the flake scar patterns from the Goshen point cloud is the smallest, and are correctly reclassified more often using LDA than either Plainview or Milnesand. Although

Plainview and Goshen are found in distinct geographic locations (southern versus northern Plains respectively), the hypothesis that they do in fact represent a geographic expression of the same

Paleoindian culture is supported by this research using topographic morphometrics.

There is variation in the temporal and geographic spread of these Late Paleoindian types, based on the radiocarbon dates. The results from this study indicate that flake scar patterning can be used to distinguish different lanceolate projectile points in time and space. In this case, there

is a definite difference in flake scar patterning in the ~6,000 years between the Late Paleoindian

and Middle Archaic Nebo Hill/Sedalia points. It is expected that Late Paleoindian unfluted lanceolate points should look similar to each other, because they presumably share a learning tradition with earlier Paleoindians. Keeping the focus on assemblages from similar points in time and/or geographic regions allows for the productive application of topographic morphometrics to examine the relationships of flintknapping learning traditions used in the manufacture of lithic projectile points. This methodology can be used to make inferences about the relationships between Goshen, Plainview, and Milnesand points as the data come from an overlapping

122

window of time from adjacent geographic regions. Patterns in the geographic location of

archaeological sites as well as radiometric data, in conjunction with the topographic

morphometric results, indicate these three point types could share a learning tradition and is used here to present a possible scenario for the culture history of the region. It is proposed that Goshen could represent an older flintknapping knowledge and technique (learning tradition), of the three

Paleoindian types analyzed in this study, that spread southward into what is today called

Plainview on the southern Plains along with Milnesand points.

Flake scar analysis could not separate Goshen, Plainview, and Milnesand points using

PCA, nor was the LDA reclassification accuracy high. The LDA did, however, show that there is

some variation between Goshen, Plainview and Milnesand. Moreover, the way in which these

point types were reclassified is informative. This lends support to some of the previously

proposed archaeological interpretations that Plainview and Goshen should be considered the same point type, along with Milnesand points.

What name to use for these three point types that appear to be geographic variants of a shared learning tradition is a question beyond this study. The issue of “splitting” and “lumping”

in archaeology is as old as the discipline itself. Whether Huckell and Merriman (2017) are

correct in discussing Late Paleoindian unfluted lanceolate points as all being encompassed under

the transition to a widespread, unfluted, culture lies outside the scope of this dissertation.

Although recent geometric morphometrics did not identify significant differences between these point types, data from this study indicate that there is some variation in flintknapping style and how it is patterned between the Goshen, Plainview, and Milnesand assemblages. The inferences proposed here from flake scar patterning and radiocarbon dates for the appearance and spread of

Late Paleoindian unfluted lanceolate points require further exploration. Whether these three point

123

types should be assigned to the Plainview type, because that name-type holds the temporal published precedence (Krieger 1947), or whether incorporating new nomenclature to acknowledge the geographic variation of Plainview, such as Milnesand-Plainview is best, remains debatable. Based on morphology and the results of Buchanan et al. (2017), Knudson

(2017) argues that Milnesand, at minimum, should be classified as a Plainview-type variety.

Arguments could be made for keeping Goshen as a separate type, if it does in fact represent the ancestral knowledge and technique, or learning tradition, as discussed here. Alternatively, the continued Goshen-Plainview nomenclature could be used to describe the northern geographic expression of Plainview.

Regardless of terminology and the name or names that are used, what is of utmost importance is a better understanding of the underlying cultural relationships between the Late

Paleoindian people who manufactured these unfluted projectile points across the Great Plains.

The use of topographic morphometric analysis as demonstrated in this chapter and Chapter Three shows how we might better discern morphological characteristics of flake scars on projectile points in the context of learning traditions and culture history. Regardless of what we call the various point types from the sites discussed in this study, the topographic morphometric flake scar pattern analysis allows archaeologists to go one step further in better explaining similarities and differences among these artifacts.

124

CHAPTER FIVE: NORTHERN PALEOINDIAN IN ALASKA

This chapter presents a second case study by which to again test the effectiveness of topographic morphometric analysis on another Paleoindian dataset. In this chapter I use projectile point data from the Mesa and Sluiceway complexes from the Brooks Range in northern

Alaska. This case study will include a brief culture history of the Northern Paleoindian period of

Alaska and a summary of the regional paleoenvironment using a new reconstructive ecological model. The chapter ends with the application of the 3D topographic morphometric methodology to the Northern Paleoindian projectile point technology in a similar manner to the Plains Late

Paleoindian case study in Chapter 4.

The Northern Paleoindian complex is comprised of the Fluted, Sluiceway, and Mesa complexes that are located in northern Alaska, particularly in the Brooks Range. The three complexes overlap temporally, as can be seen in the calibrated dates of 11 Northern Paleoindian archaeological sites in Figure 6.3 of Smith et al. (2013:114), and date from roughly 13,200 to

10,700 cal yrs BP (Smith et al. 2013). Specifically, the Sluiceway complex dates to about 13,200 to 11,000 cal yrs BP, the Mesa complex dates between 12,500 to 10,700 cal yrs BP, and the age range of the Fluted complex spans about 12,400 to 10,700 cal yrs BP (Smith et al. 2013). The morphology of these lanceolate projectile points, limited faunal remains, and site settings indicate that Northern Paleoindian hunters were focused on large mammal herd animals such as caribou, steppe bison, and muskox. While faunal remains are scarce from archaeological sites in this region and time, the Fluted tradition at the Raven Bluff and Serpentine Hot Springs sites is strongly associated with caribou remains (Goebel et al. 2013 Hedman 2010). No faunal remains have been recovered at Sluiceway sites, but caribou and steppe bison (Bison priscus) are believed to have been the primary prey. This is based on the geographic setting of sites, which

125

occur in upland areas of the Brooks Range along migration routes for caribou, as well as at lower

elevations in the Arctic Foothills that would have been summer grazing ranges for bison (Rasic

2011). This lack of faunal data, unfortunately, defines Mesa sites as well. Based on the locations

of these sites in the Arctic Foothills, steppe bison were likely targeted, as well as muskox and

caribou (Kunz et al. 2003). Faunal remains recovered from the Engigstciak site, a possible Mesa

occupation in the Yukon Territory, were composed primarily of steppe bison, along with the

remains of caribou and muskox (Cing-Mars et al. 1991).

Although all three traditions represent the Paleoindian types in Alaska during the PHT, only the unfluted Sluiceway and Mesa projectile points will be subject to topographic morphometric analysis in this study. Fluted points, as previously stated, present a challenge when trying to compare them to unfluted points because the removal of channel flakes introduces a

bias to the flake scar patterning that is not present on unfluted points. Thus, this study focuses

solely on unfluted lanceolate projectile points.

Both Sluiceway and Mesa projectile types are lanceolate in shape, made of high quality

raw materials, and exhibit a medial ridge down the long axis of the point due to collateral

flaking. Both also display proximal edge grinding of the haft segment and heavy reworking

while hafted (Smith et al. 2013). However, Sluiceway points appear to have been used for other

tasks such as cutting, as some heavily reworked points exhibit polish on the distal blade element

(Rasic 2008). Sluiceway are characteristically very large, larger than Mesa, with convex bases

while Mesa typically have a concave base. However, flat bases and slightly convex bases are

present on a low proportion of Mesa points. The reduction process on Sluiceway points involved

the comprehensive shaping of large biface blanks through percussion flaking followed by

substantial pressure flaking. Mesa points, on the other hand, were reduced from much smaller

126

biface blanks from tabular nodules mostly using percussion flaking and less pressure flaking

(Rasic 2008; Smith et al. 2013)

All of the known Sluiceway and Mesa sites from Alaska functioned as hunting lookouts and retooling stations. No residential sites have been identified or recorded. The cultural relationships between the Sluiceway and Mesa complexes are unknown (Smith et al. 2013).

Attempting to ascertain their relationship to one another is made more difficult by the fact that these complexes overlap spatially and temporally. Fluted points have been recovered at the

Tuluaq site of the Sluiceway complex, and at the Putu site of the Mesa complex. At the Mesa site, Type A points, which are described by Kunz et al. (2003), fit the description of Sluiceway points according to Smith et al. (2013). This overlap of projectile point types at sites ascribed to the other Paleoindian complex compound the issue (Smith et al. 2013). Moreover, determining possible relationships between the point types is made difficult because specimens are either surface finds or come from sites characterized by shallow site stratigraphy.

A number of explanations have been proposed to explain the differences between the three Northern Paleoindian projectile point types. These explanations include: (1) variation in hafting technique, (2) use for different prey species, (3) use during different seasons, and (4) differing culture histories for each point type (Smith et al. 2013). Rasic (2011:154) discusses the idea that Sluiceway points were hafted to spears or lances used for thrusting, while Mesa points were propelled through an atlatl and dart system. If this was the case, this could explain differences in hafting technique, target prey, and seasonality, which in turn could answer the culture history questions by explaining the seasonal targeting of certain species with different hunting technologies in different geographic settings. Thrusting spears, or lances, identified by the large size of the bifacial point indicating they were likely held rather than thrown, have been

127

identified with caribou hunting in Alaska (Ackerman 2008). Archaeological examples of caribou hunting using thrusting spears comes from the later Northern Archaic tradition in Alaska (~6,000

BP). Hunters would drive caribou into a pond and dispatch them from boats with thrusting spears

(Ackerman 2004, 2008). That setting does not match those of upland Brooks Range sites, and it is not known if boat technology was utilized with Sluiceway technology. However, Friesen

(2013) examined the use of drive lines by Dorset peoples in the Canadian Arctic (~2,100 to 650

BP) to funnel caribou towards a hunter who would use a thrusting spear or lance, or possibly even a hand thrown spear, to dispatch prey at close quarters. Archaeological and ethnographic data surrounding thrusting spears or lances, similar to those of the Sluiceway complex, could have been used in hunting caribou in the upland Brooks Range during the PHT. Thrusting spear technology would have required different hafting, and possibly different prey and seasonality utilization from Mesa points, if the land use model proposed by Rasic (2008, 2011) is correct.

Mesa sites, targeting bison in the Arctic foothills, could have utilized atlatl propelled dart technology rather than thrusting spears. This opens the possibility that the Sluiceway and Mesa points represent two different technologies adapted to seasonal hunting in different regions by the same culture group. However, this cannot not rule out the alternative that these complexes could represent different, unrelated groups of people.

As Rasic (2011:159-160) discusses, what we know about Northern Paleoindians is based on a small number of sites. To comprehensively understand a technocomplex requires a large set of sites and data. Rasic (2011) demonstrates that considerable inter-assemblage variability exists within even the geographically and temporally tightly constrained Sluiceway complex. Current data from the small number of sites makes it hard to understand the technological, let alone cultural, relationships between Sluiceway and Mesa in the Brooks Range. One way to begin

128

addressing this question with the available data is by conducting topographic morphometrics on

the large assemblage of bifacial projectile points from these complexes to look for similarities or

differences in the flake scar patterning of Sluiceway and Mesa points. The remainder of this

culture history background section provides a description and overview of the ten Northern

Paleoindian sites (Figure 5.1) that contained projectile points analyzed in this study (see Table

5.1 for a summary of site information). Because this dissertation project is focused only on the biface assemblage from each site, a detailed discussion of site stratigraphy and a breakdown of lithic artifacts by artifact types and counts is not presented. Numerous reports have been published that discuss the lithic assemblages from each site in detail. These also include

discussions about intrasite relationships that are not discussed here.

129

Figure 5.1. Location of Northern Paleoindian sites in Alaska from this study. 1, Red Dog; 2, Tuluaq; 3, Upper Kelly; 4, Caribou Crossing 1 and 2; 5, Nat Pass; 6, Mesa; 7, Hilltop; 8, Putu/Bedwell; 9, Spein Mountain. Sluiceway Sites, 1-5; Mesa Sites, 6-9.

130

Sluiceway Complex Sites from the Western Brooks Range

Caribou Crossing 1 and 2

The Caribou Crossing site is located near the headwaters of Nunaviksak Creek, a tributary of the Noatak River, just south of the Brooks Range divide. Caribou Crossing is actually two sites, Caribou Crossing 1 and 2 (MIS-376 and 377), and is considered to represent a single artifact distribution located on two hilltops separated by a small drainage about 60 meters wide that is devoid of cultural material (Rasic 2011). A source of high quality chert is located about 5 km from the site. The site produced a large assemblage of lanceolate projectile points

(n=117) from the surface and shallow subsurface (Rasic 2008). A large number of preforms and biface blanks were also recovered (n=77) along with debitage and a small number of unifaces and retouched flakes. Today, the site is located along a major caribou migration route across the

Brooks Range. Based on the lithic assemblage and geographic location of the site, Rasic (2008) proposes the site functioned as an intercept point during spring and fall caribou migrations. No cultural features were identified and no radiocarbon dates have been obtained (2011).

Tuluaq Hill

The Tuluaq Hill site is located on a high hilltop overlooking Wrench Creek, with good views both upstream and downstream in the Noatak River Basin in the southwestern Brooks

Range. Within about three km of Tuluaq Hill are two documented high quality black chert lithic quarry sites. This lithic material dominates the site assemblage. Based on the lithic assemblage at the site and at the nearby quarries, it is likely that blanks were roughed out and then transported to the site where they were further reduced. Complete projectile points, manufacturing rejects, broken (usually displaying impact damage), and reworked projectile points indicate that this site

131

served both as an area of point production and a retooling station. In addition, the presence of a number of flat areas that could have been used as campsites lie in close proximity to both

Wrench Creek and the site. Based on the topographic feature on which the Tuluaq Site rests, with viewsheds both upstream and downstream, the proximity to both lithic raw material quarries and possible campsite areas, indicate it likely functioned as a hunting lookout and retooling station. A single hearth in direct association with Sluiceway projectile points was dated to 11,200±40 14C years BP and 11,120±40 14C years BP (13,250 to 12,801 cal yrs BP) (Rasic 2008; Smith et al.

2013).

Red Dog Mine

This site is located on a large and relatively flat knoll in the DeLong Mountains near

Wrench Creek on the south side of the Brooks Range. Two surface and subsurface locales were identified and recorded. The lithic assemblage contains debitage, flake tools, early stage bifaces, and Sluiceway projectile points. Although no radiocarbon dates exist, Rasic (2008) reports these bifaces as classic Sluiceway points.

Upper Kelly

The Upper Kelly site is located on a gently sloping hillside above the Kelly River floodplain. Three lithic scatters that contain debitage, preforms, and Sluiceway projectile points were recorded. No subsurface excavations were undertaken, but as Rasic (2008) notes it appears that only a shallow and rocky sediment exist. No features were identified on the surface and no organic material was obtained for radiocarbon dating. Between the three locales, three Sluiceway points were identified, attributing the site to the Sluiceway complex (Rasic 2008).

132

Nat Pass

The Nat Pass site is located in the Anisak valley on the south side of the western Brooks

Range. The site lies on a relatively flat hilltop overlooking the Anisak River with northerly views towards the Brooks Range. The lithic assemblage was found both buried and on the surface. The buried presence of burnt bone and flakes indicate a hearth feature at some point. Charcoal was collected and dated to 10,010±40 14C years BP, and 9,910±40 14C years BP (11,709 to 11,287, and 11,595 to 11,221 cal yrs BP). Three Sluiceway projectile point fragments were recovered at the site, and along with the radiocarbon dates, indicate a PHT occupation at the site. However, the numerous lithic scatters on the surface and below ground indicate multiple occupations and make it tenuous to associate all of the assemblage to these dates (Rasic 2008; Smith et al. 2013).

Mesa Complex Sites from the Central Brooks Range

Mesa

The Mesa site, the type site of the complex, is located in the Arctic Foothills north of the

Brooks Range on a mesa-like ridge overlooking Iteriak Creek. The Mesa site provides an impressive view of the surrounding area, with the Books Range to the south and the Ivotuk hills to the north. Like other arctic sites, the sediments are shallow, ranging in depth from 5 to 35-cm.

Four localities were identified at the Mesa site. Cultural material was clustered around habitable areas and surrounding the numerous hearth features that were identified during excavation. As the hearth features are small in size and ephemeral in nature, most of them likely represent single occupation events. In total, 35 accepted radiocarbon dates cluster between 10,300 to 9,700 14C years BP (12,300 to 11,700 cal yrs BP). Two older dates, 11,660±80 14C and 11,190±70 14C years BP, from hearth features at the Mesa site, have been dismissed by some as being too old or

133

that they represent an earlier occupation of the site by almost 1,000 years. Based on several

factors, including the nature of the topographic feature that the site rests on, the location of the

site, the lithic assemblage, and the number of broken projectile points, it is classified as a hunting

lookout and retooling station (Kunz and Reanier 1994; Kunz et al. 2003; Smith et al. 2013)

Hilltop

The Hilltop site is located on top of the highest knoll on the north side of the Atigun

Gorge near the confluence of the Atigun and Saganavirktok rivers in the central Brooks Range.

As with other Mesa sites, the Hilltop site is believed to be a lookout and retooling station.

Although the viewshed is not as extensive as at the Mesa site, it does offer a wide view of the gorge and any animals that might be passing through. The assemblage was recovered from the surface and generally from within the first 10 cmbs. Excavations in 1993 recovered charcoal and the Mesa assemblage has been dated to 10,360±60 14C years BP (12,513 to 11,994 cal yrs BP)

(Bever 2000; Smith et al. 2013).

Putu/Bedwell

The Putu/Bedwell site represents two locales on a hilltop in the Arctic Foothills north of the Brooks Range above the Sagavanirktok River. They are classified as a single site because they were originally believed to be separate locales of a single site. The Bedwell locale is situated on the hilltop while the Putu locale is 30 meters lower and 100 meters away on a bench.

Soil deposits are shallow, ranging from 20-40 cmbs across the sites. A radiocarbon date at the

Putu locale of 8,810±60 14C years BP (10,158 to 9,631 cal yrs BP) is reported. The oldest

radiocarbon date at the site came from the Bedwell locale at 10,490±70 14C years BP (12,590 to

134

12,136 cal yrs BP). In common with other Mesa sites, the Putu/Bedwell site is believed to have

been used as a hunting lookout and retooling station (Reanier 1996; Smith et al. 2013).

Mesa Complex Sites in the Kuskokwim River Drainage, Southwestern Alaska

Spein Mountain

The Spein Mountain site is a Mesa complex anomaly in terms of its geographic location.

A comparison of the lithic technological organization by Bever (2000) found it to be consistent

with other Mesa sites. However, the site is located over 900 km southwest of the Mesa type site

and other sites in the complex. Spein Mountain consists of four separate areas located along a

highly weathered ridge overlooking the Kisaralik River in the Kuskokwim River Drainage. A

majority of the Spein Mountain assemblage was recovered from the surface; however, a shallow

15 to 25-cm-deep loess does exist over weathered bedrock. No technological change was seen between buried and surface assemblages, leading Ackerman (1996) to assign the material to a single component. One of the four areas at the site is believed to be a fall hunting base camp, based on its location in an area out of the wind and out of sight of migrating caribou routes that take place during the fall from Kuskokwim River lowlands. A majority of the assemblage and the diagnostic artifacts from Spein Mountain were recovered from this area that contained the only pit feature, assumed to be a hearth. Although other hearths from the site are implied, based on the recovery of burnt lithics, this pit feature produced the only radiocarbon date: 10,050±90

14C years BP (Ackerman 2001) (11,968 to 11,266 cal yrs BP using OxCal 4.3). The other three

areas are believed to be lookouts based on their viewsheds over the Kisaralik River and the

tundra uplands and the limited lithic scatter on the surface (Ackerman 1996, 2001).

135

Table 5.1. Overview of Archaeological Sites from Alaska.

Radiocarbon Geographic Number of Points Site Function Years Age cal BP Region in the Study Sluiceway Caribou Hunting Lookout & western Crossing 1 Retooling Station NA NA Brooks Range 35 Caribou Hunting Lookout & western Crossing 2 Retooling Station NA NA Brooks Range 13 Hunting Lookout & 13,250 to western Tuluaq Hill Retooling Station 11,200±40 12,801 Brooks Range 8 western Red Dog Hunting Lookout NA NA Brooks Range 4 western Upper Kelly Hunting Lookout NA NA Brooks Range 1 11,709 to western Nat Pass Hunting Lookout 10,010±40 11,287 Brooks Range 1 Mesa Hunting Lookout & 10,240±40 to 12,385 to central Brooks Mesa Retooling Station 9,780±40 11,162 Range 29 Hunting Lookout & 12,513 to central Brooks Hilltop Retooling Station 10,360±60 11,994 Range 3 10,158 to Hunting Lookout & 8,810±60/ 9,631/12,590 central Brooks Putu/Bedwell Retooling Station 10,490±70 to 12,136 Range 6 Spein Mix of Camp and 11,968 to southwest Mountain Hunting Lookout 10,050±90 11,266 Alaska 8

Paleoecological Background

Understanding the ecology of a region and changes in vegetation communities in

response to climate change is essential before inferences can be made regarding past culture

change and human adaptation to an environment. If the geographic range and diversity of plant communities shift due to climate change, so too do the resources that humans exploit. If changes are large enough these shifts in plant and animal community abundance, diversity, and geographic range can lead to changes in human technology, subsistence practices, group

interactions, and human landscape use (Anderson et al. 2003; Elias and Crocker 2008; Graf and

Bigelow 2011). During the Pleistocene-Holocene Transition three paleoclimatic events impacted

136

Alaska and Northern Paleoindians. These are the Bølling-Allerød interstadial (~14,500 to

~13,000 cal yrs BP), the Younger Dryas stadial (~12,800 to 11,500 cal yrs BP), and the

Holocene Thermal Maximum (~11,000 to ~9,500 cal yrs BP) (Hoffecker and Elias 2007; Mann et al. 2001; Mason et al. 2001; Peteet 1995). A paleoecological reconstruction for Alaska was undertaken, but not for the area of the Great Plains in the previous study, as the ecological response to climate change was far more complicated in Alaska. On the Plains the Younger

Dryas brought with it the widespread expansion of grasses and an increase in bison populations.

It is important to project the paleoecology of Alaska to ancient landscapes, using predictive modeling, to determine whether differences in cultural complexes and/or lithic technologies from the Brooks Range in northern Alaska are associated with different ecological communities or with changes in these communities during the Pleistocene-Holocene Transition

(~14,500 cal yrs BP to ~9,500 cal yrs BP). The following section uses predictive modeling to evaluate the sensitivity of northern Alaska’s paleoecology to climate change and to assess what the ecological changes during the Pleistocene-Holocene Transition entailed. This model provides a tool that archaeologists can use to assess human responses to climate change by examining the locational relationships of archaeological sites comprising the Mesa and Sluiceway artifact complexes from northern Alaska to paleoecological reconstructions. Quadratic Discriminant

Analysis (QDA) is used to determine which modern climatic variables, including monthly averaged minimum and maximum temperature and precipitation, as well as topographic factors, best define modern vegetation land cover across a raster landscape. This method is then applied to the PHT to predict the land cover for each raster cell at a given time with proxy estimates to reconstruct the boundaries of ancient plant communities on the landscape. Assessing the amount

137

of ecological change that occurred is important because shifts seen at the onset of the Younger

Dryas would have impacted the local fauna and thus prehistoric hunters living in the area.

The YD, on a broad scale, is associated with cooling temperatures and a decrease in

moisture similar to that of the Late Glacial Maximum (LGM). Fluctuations in key pollen types

recovered from sediment cores indicate shifting plant biomes in parts of Alaska in response to

these climatic changes (Anderson et al 2003; Hu et al. 2002). Therefore, identifying the amount

and distribution of change in plant communities during the Pleistocene-Holocene Transition

(PHT) can help archaeologists address questions pertaining to subsistence adaptations. This, in

turn, can contribute to clarifying the factors that underlie differences or changes through time in

lithic technology. If ecological changes occurred, how severe were they and would they have

warranted a human response that is visible in the archaeological record?

Pollen records from sediment cores represent localized ecological information as data

points on a landscape, and do not constitute a vegetative land coverage map. These sediment records are rarely located within, or adjacent to, archaeological sites for direct correlations.

However, they can be used to support paleoecological land coverage predictions generated by this model which, in turn, can be compared to the geographic distribution of different lithic technologies. Ecological and environmental studies use predictive models to statistically predict vegetation landscapes based on environmental variables (Franklin 1995; Kaplan 2003; Lowell

1991). In most cases predictive models have been employed to predict changes in future plant community distributions (Brzeziecki et al. 1993; Crumley 1993; Iverson and Prasad 1998). Here,

I apply QDA of modern climate data and land coverage maps to predict shifts in paleoecological plant community boundaries during the PHT. More specifically, this study assesses the amount of ecological change that occurred across the landscape during the PHT.

138

Younger Dryas temperature and precipitation were not uniform throughout northern

latitudes, especially across the Arctic (Anderson et al. 2002; Diefendorf et al. 2006; Yu and

Wright 2001). The YD was a climatic reversal, with conditions similar to that of the LGM, following the warmer Bølling-Allerød (BA) interstadial. The onset of the YD began ~12,800 cal yrs BP and lasted for roughly 1,300 years, ending ~11,500 cal yrs BP (Alley 1990; Fairbanks

1990; Hajdas et al. 1998). At the end of the BA in Alaska there is a shift in the pollen records. A

decrease in Betula pollen and an increase in herbaceous pollen point to a reversion to a cooler

climate (Anderson et al. 2003; Hu et al. 2002). This pollen change corresponds to the onset of the

YD as documented by δ18O values in Greenland ice cores (Cuffey et al. 1995) and gives an

indication of what the landscape and environment looked like for Alaskan hunter-gatherers. It is important to note that Beringia, especially central Beringia, was likely a patchwork of vegetation communities creating a mosaic landscape inhabited by people and prey animals (Guthrie 1982).

Regional variation of YD signals from paleoenvironmental records can be seen in Alaska

and across Siberia. Recent work suggests that the YD had strongly localized effects across

different regions of Alaska (Kaufman et al. 2010; Kokorowski et al. 2008). Pollen data compiled

by Kokorowski et al. (2008) indicate a colder Younger Dryas with increased precipitation in

southern Alaska while temperature and moisture from central and northern Alaska were only

slightly decreased relative to the previous BA (Anderson et al. 2003; Peteet et al. 1997; Edwards et al. 2001; Elias et al. 1997). The interior of central and especially northern Alaska likely experienced less cooling during the YD due to increasing summer insolation moderated by cool surface temperatures of the North Pacific and the Beringian landmass that still connected Siberia and North America (Anderson et al. 2003; Bartlein et al. 1991).

139

Ranges of plants and animals can expand or contract depending on the climate factors at

play, especially at the margins of an organism’s range (Parmesan 2006). Consequently, environmental effects on one organism can lead to effects on other organisms, all of which can cause the reconfiguration of ecological communities (Hoffman and Parsons 1997; Root et al.

2003). This is important as changes at any trophic level resulting from climate change, will

ultimately be felt by humans in the density and distribution of hunted animal species they exploit

that are reliant on particular vegetation communities. Understanding these factors and identifying

ecological changes across prehistoric landscapes allows archaeologists to begin directly

addressing differences in the archaeological record believed to be induced by climate change.

PRISM Spatial Climate Dataset

A spatial climate dataset for modern climate variables is the basis for this predictive model. This study utilized the 30-arcsec (~800 m) resolution Parameter-elevation Relationships on Independent Slopes Model (PRISM) interpolated data grids (the “LT71m” dataset) for Alaska created by the PRISM Climate Group at Oregon State University and made available by the

National Park Service (PRISM Climate Group). PRISM spatial climate datasets interpolate climate-elevation data for every raster cell on a Digital Elevation Model (DEM) from modern weather stations. Factors used in producing PRISM datasets include “location, elevation, coastal proximity, topographic facet orientation, vertical atmospheric layer, topographic position, and orographic effectiveness of the terrain” (Daly 2008). The PRISM dataset for Alaska was created using monthly mean minimum (tmin) and maximum (tmax) temperature and mean precipitation

(ppt) in millimeters from 1971-2000 for a total of 36 variables. In total, over 2.5 million raster cells with climate and vegetation data were used to predict vegetation coverage.

140

Land Cover Map

A modern USGS land cover map of vegetation in Alaska (Appendix B) was obtained

from the Alaska Geographic Data Committee (Alaska Geospatial Data Committee). Five classes

(Ocean Water, Water, 1990 Fires and Gravel Bars, 1991 Fires, and Canada/Russia) were removed (turning them NA) before model predictions were run, because they are not pertinent to a vegetation reconstruction of Alaska. However, those empty NA cells were replaced with a predicted vegetation class when the model was run to create a comprehensive map.

Quadratic Discriminant Analysis

Similar to discriminant function analysis used to predict dependent variables from one or more predictor variables, QDA is used when covariance amongst group variables is not equal.

QDA is commonly used in statistical analyses for obtaining classifiers and is used in this study to predict vegetation classes of raster cells from the PRISM spatial climate datasets. QDA examines a set of observation vectors to classify an event. Given the vectors that classify each event, QDA then determines what the predicted class would be based on a new observation vector (Gareth et al. 2013; Srivastava et al. 2007). In this project, QDA considers the climate variables for each raster cell and the vegetation class that defines them. When new values based on proxy estimates for the spatial climate data are entered, a new prediction of the vegetation classification is given to each cell based on the new observation vectors.

By classifying vegetation from modern PRISM spatial climate datasets, it is possible to use QDA to identify shifts in the boundaries between ecological communities during the PHT.

Unlike point source proxy measurements such as sediment cores, data for the entire state of

Alaska are produced and high-resolution focus on small study areas is possible. All QDA

141

predictions were conducted using the R statistical package (R Core Team. R). The R code for

this study is available upon request.

Parameter Settings

The PRISM dataset is based on modern values. Consequently, projected changes in

temperature (tmin and tmax) for the model were either increased or decreased by degrees Celsius

(°C). For this study tmin and tmax proxy estimates were increased or decreased by the same

integer value. Proxy estimates of precipitation are usually presented in publications as an

“increase” or “decrease” in precipitation, as precipitation is harder to quantify. Therefore, for this

model precipitation was modeled as either increasing or decreasing by a given percentage from

modern measurements (e.g. -20% or +10%).

PRISM data on modern temperature (tmax and tmin) are presented as 30-year averages

for those variables for each individual month of the year. The same was done for precipitation.

Predictions for this study were run with a single change from modern values in temperature and

precipitation across all 12 months (i.e. tmin -6°C, tmax -6°C, and precipitation +10%).

Decreasing or increasing the temperature by the same value across all months in the year carries

the potential to mask any seasonal differences. By decreasing or increasing all of the monthly

values evenly, this model projects the average temperature change for all 12 months across the

year. This was done to keep the model simple and avoid compounding potential errors from

overcomplicating seasonal projections (i.e. November-May -6°C and June-October -2°C) when proxy data from the time period is not fine-grained enough to be examined on a month-to-month basis. All of the QDA prediction data are available in Appendix C, which includes a suite of 147 projections ranging from -15°C to +5°C at 1°C intervals (21 possible) and for precipitation

142

values ranging from -30% to +30% at 10% intervals (7 possible). This suite of predictions was

done to cover all possible climate proxies in this study, as well as to provide data for other

researchers in Alaska focused on different time periods and proxy estimates.

The focus of this particular study is not to track ecological shifts through time during

each of the identified paleoclimatic events. Rather, it is to examine the overall effect of climate change on the paleoecology of Alaska to identify areas that would have been ecologically more sensitive and subject to higher degrees of vegetation change in the past. Proxy variables for temperature and precipitation were chosen to represent the values at the height of each paleoclimatic event which were then compared to the closest projection from the climate suite.

The first is the warmer BA interstadial of de-glaciation at the end of the Pleistocene from

~14.5k to ~12.9k cal yrs BP. Proxy estimates indicate that temperature was 1-3°C warmer than the modern temperature with slightly increased precipitation (Hoffecker and Elias 2007; Elias

2000, 2001; Elias et al. 1996; Gaglioti et al. 2014; Hu and Shemesh 2003; Mann et al. 2002,

2010). The second, which is of primary interest to this study, is the YD Stadial, a climatic reversal (~12.8-11.5k cal yrs BP) with cooler climatic conditions similar to the LGM. Research suggests that during the YD climate change in Alaska varied regionally (Kaufman et al. 2010;

Kokorowski et al. 2008). Proxy estimates range from 2-8°C colder than present, with increased precipitation in southern Alaska and drier conditions in central and northern Alaska (Bigelow and Edwards 2001; Briner et al. 2002; Hoffecker and Elias 2007; Hu and Shemesh 2003; Mann et al. 2002, 2010). The geographic separation between these two different YD expressions is roughly the foothills on the northern edge of the Alaska Range extending across the Kuskokwim

Mountains to the west. The final event is the Holocene Thermal Maximum (~11 to ~9.5k cal yrs

BP), the subsequent warming in the Holocene of Alaska following the YD. Proxy estimates of

143

temperature for the HTM ranged from 1.5-2.5°C warmer than present with precipitation slightly lower than present (Gaglioti et al. 2014; Hoffecker and Elias 2007; Kaufman et al. 2004; Mann et al. 2002, 2010; Nelson and Carter 1987).

Comparison to Pollen Cores

The interpretation of tundra vegetation from fossil pollen can be challenging due to factors such as low taxonomic resolution, broad ecological tolerances of plants representing a majority of the pollen, and the poor dispersal of minor pollen types. However, it appears that the major genera present in Alaska during the YD are sensitive to changes in climate (Anderson et al. 1994). Different tundra types in the Arctic and Subarctic contain many of the same genera.

Unlike other vegetation communities composed of distinct plant species in warmer climates, in tundra communities the differences can lie in the percent composition that a species or a genus constitutes of a particular plant community. Even though tundra types are composed of similar species it is still possible to make meaningful predictive vegetation models. The sensitivity of plant communities to climate change at the PHT in Alaska results in inverse frequencies of different tundra pollen species and genera. Ecological predications from proxy estimates seem to indicate similar patterns for where YD signals would be expected based on the comparison of fossil pollen. The best fit projections for YD proxy estimate predictions in northern Alaska is -

2°C -10% precipitation (Figure 5.2), as compared to the statewide pollen record during the PHT.

144

Figure 5.2. Vegetation projection of the YD with northern Alaska proxy conditions (-2°C - 10% precipitation). 1, Glaciers & Snow; 2, Alpine Tundra & Barrens; 3, Dwarf Shrub Tundra; 4, Tussock Sedge/Dwarf Shrub Tundra; 5, Moist Herbaceous/Shrub Tundra; 8, Low & Dwarf Shrub; 9, Tall Shrub; 10, Closed Broadleaf & Closed Mixed Forest; 12, Closed Spruce Forest; 13, Spruce Woodland/Shrub; 14, Open Spruce Forest/Shrub/Bog Mosaic; 16, Open & Closed Spruce Forest; 17, Open Spruce & Closed Mixed Forest Mosaic; 18, Closed Spruce & Hemlock Forest; 19, Tall & Low Shrub.

145

Pollen data compiled from over 50 sediment cores across Alaska were ranked by

Kokorowski et al. (2008) as either indicating a presence or absence of a YD signal. The YD

prediction of -6°C +10% precipitation (Figure 5.3A) contains the best fit with the pollen data

from the YD in southern Alaska. The increased expanse of alpine tundra in southern Alaska

matches the locations and supports the findings of (Kokorowski et al. 2008) that the YD had an

increased presence in that area. When the focus is shifted to northern Alaska, a small growth in

alpine tundra is seen in the Brooks Range. When compared to Figure 5.3B that depicts a more

suppressed climatic change during the YD in northern Alaska (-2°C -10% precipitation) this pattern is highlighted even more. This area of northern Alaska remains mostly covered in shrub tundra indicating the YD would not have been as strong as it was in southern Alaska. This pattern is also evident from fossil pollen. This indicates that certain temperature and precipitation conditions must be met in order for the predicted models to match the pollen data. These findings support the existence of a cooler and wetter YD in southern Alaska compared to a slightly more muted effect in northern Alaska.

146

Figure 5.3. Vegetation projections with pollen core data. (A) Younger Dryas southern proxy conditions (-6°C +10% precipitation), (B) Younger Dryas northern proxy condition (-2°C -10% precipitation), (C) Bølling-Allerød. 1, Glaciers & Snow; 2, Alpine Tundra & Barrens; 3, Dwarf Shrub Tundra; 4, Tussock Sedge/Dwarf Shrub Tundra; 5, Moist Herbaceous/Shrub Tundra; 8, Low & Dwarf Shrub; 9, Tall Shrub; 10, Closed Broadleaf & Closed Mixed Forest; 12, Closed Spruce Forest; 13, Spruce Woodland/Shrub; 14, Open Spruce Forest/Shrub/Bog Mosaic; 16, Open & Closed Spruce Forest; 17, Open Spruce & Closed Mixed Forest Mosaic; 18, Closed Spruce & Hemlock Forest; 19, Tall & Low Shrub.

147

Tabulated results demonstrating how well the ecological projections for the YD match

with the pollen record are presented in Table 5.2. As some of the sample size counts are low, a

Fisher’s exact test was used to test the distribution of pollen cores containing YD signals with

areas from the projections predicting the correct vegetation type. Statistical analysis (two-sided

Fisher's exact test) confirmed that there was a more significant association between YD projected vegetation types from this study and pollen cores with data indicating the presence or absence of a YD signal than would be expected by chance (P < 0.01, Fisher's exact test). The central Alaska area, bordered by the Alaska Range to the south, was likely more ecologically variable due to its location as a transitional zone between southern and northern Alaska. Based on comparisons of proxy estimates to pollen data, central Alaska would have been subject to a landscape more similar to Figure 5.3A (-6°C +10% precipitation condition) while northern Alaska would have been subject to a landscape more similar to Figures 5.2 and 5.3B (-2°C -10% precipitation condition).

Table 5.2. Projected Landscapes during the Younger Dryas and Pollen Core Signals.

Pollen Cores Pollen Cores with a YD Signal without a YD Signal In Predicted YD Signal Vegetation Types 12 7 In Predicted Non-Younger Dryas Signal Vegetation Types 4 32

In addition, the pollen data can be compared to the BA predictions (Figure 5.3C). Signals

indicating a YD should occur in areas other than herbaceous tundra (blue areas) because they

shift to herbaceous tundra during the YD. No YD signals should remain in shrub tundra

148

communities (green) where no major fluctuation in the pollen occur, indicating continuity in shrub tundra from the BA to the YD. This pattern holds true, as do the predictions of YD signals presented by (Kokorowski et al. 2008). These results strengthen the use and application of these ecological predictions to specific areas and climatic conditions in the past for interpretation and comparison to archaeological data.

Brubaker et al. (2005) note that most if not all woody plant species (Populus, Larix,

Picea, Pinus, Betula, and Alnus) dominant in Alaska today were present during the LGM, even if in minor numbers. It would be expected that they should, thus, appear in some form in the BA and YD predictions, as they do. It is not until the warmer, mesic, HTM that the ecological landscape known today in Alaska becomes evident with the widespread manifestation of spruce and broadleaf forests into central Alaska, and moist acidic tussock tundra in northern Alaska

(Figure 5.4).

149

Figure 5.4. Vegetation projection of the HTM. 1, Glaciers & Snow; 2, Alpine Tundra & Barrens; 3, Dwarf Shrub Tundra; 4, Tussock Sedge/Dwarf Shrub Tundra; 5, Moist Herbaceous/Shrub Tundra; 7, Low Shrub/Lichen Tundra; 8, Low & Dwarf Shrub; 9, Tall Shrub; 10, Closed Broadleaf & Closed Mixed Forest; 11, Closed Mixed Forest; 12, Closed Spruce Forest; 13, Spruce Woodland/Shrub; 14, Open Spruce Forest/Shrub/Bog Mosaic; 15, Spruce & Broadleaf Forest; 16, Open & Closed Spruce Forest; 17, Open Spruce & Closed Mixed Forest Mosaic; 18, Closed Spruce & Hemlock Forest; 19, Tall & Low Shrub.

150

Northern Alaska Paleoecological Discussion

To examine the extent by which the plant ecology of Alaska shifted, cells that changed vegetation class versus those that remained the same between paleoclimatic event predictions were mapped. Blue cells indicate where the vegetation class changed, and dark gray cells indicate where they remained the same. Two maps were produced for northern Alaska to assess the sensitivity of the area, one from the BA to the YD, and a second from the YD to the HTM.

In the Brooks Range, the projected ecological impact of climate change from the BA to the YD indicates about 18% of the landscape experienced ecological change, as seen in Figure

5.5. Very little change occurred at higher elevations in the mountains, but what is of more importance is the fact that very little change occurred in the Arctic Foothills to the north of the

Brooks Range. These latter areas would have supported large mammal resources important to

Northern Paleoindians hunters.

Figure 5.5. Northern Alaska digital elevation model with vegetation change from the Bølling-Allerød to Younger Dryas. Blue, vegetation change; gray, no vegetation change.

151

Northern Alaska experienced much more ecological change from the YD to the HTM than from the BA to the YD, with almost 59% of the cells changing vegetation types (Figure

5.6). The sensitivity of this area during the Holocene was likely due to the return of warmer and more mesic conditions, in conjunction with melting permafrost. Where ecological change did occur, much is due to the appearance and spread of moist acidic tussock tundra. The only areas that did not experience any ecological change were at high elevations and mountain tops, likely as a result of the vegetation class having already reached an altitudinal limit. Much of the ecological change occurred in the lowlands and river valleys, and in the Arctic Foothills. These are areas that were important to Northern Paleoindians based on the location of archaeological sites and the large mammal resources they hunted.

Figure 5.6. Northern Alaska digital elevation model with vegetation change from the Younger Dryas to the Holocene Thermal Maximum. Blue, vegetation change; gray, no vegetation change.

During the YD the colder and drier conditions would have promoted the return of grasslands in the Arctic Foothills, at least in patchwork with tussock tundra that entered during the earlier warmer and wetter BA (Mann et al. 2002, 2010). This return of grassland habitat

152

would have been favored by large mammals, including bison, but also muskox and possibly

caribou. This patchwork landscape would not have supported large numbers of bison and

muskox like earlier in the Pleistocene, but would have reverted to a condition in which these

animals would have flourished and would have been continually exploited by Northern

Paleoindian hunters of the Mesa complex (Kunz et al. 2003; Mann et al. 2001). After about 1,300 years the cooler and drier conditions of the YD ended and the climate became warmer and wetter going into the HTM, making the landscape less hospitable for these large grazers. Mann et al.

(2001) argue this climate promoted the expansion of moist acidic tussock tundra into the Arctic

Foothills, as seen today, which would have favored caribou over bison. Caribou are much more

adept at traversing the soft, wet, and uneven tussock tundra landscape due to their large hooves

and gait as opposed to the smaller footed bison that prefer even ground and would have struggled

to find enough sustenance. While tussock tundra would have been better suited for caribou, as

seen in their large numbers in the area today, it is estimated it would have taken time for their

populations to rebound. This is seen in the archaeological record by the abandonment of the area

by humans for several thousand years until ~7,500 to 8,000 14C years BP (Kunz et al. 2003).

Based on the lithic technological organization of Sluiceway sites in the western Brooks

Range, Rasic (2008, 2011) proposes a land-use model wherein Sluiceway groups would have

intercepted migrating caribou in the spring and fall in upland settings within the Brooks Range,

and hunted bison during the summer in the Arctic Foothills. Winter residence, with a presumed

reduction in mobility, would have occurred in the Noatak Basin and its major tributary valleys.

The projected ecological change after the YD going into the HTM is presented in Figure 5.7 with

the location of known Northern Paleoindian archaeological sites. Apart from high elevations, a

majority of the landscape underwent ecological turnover. The areas on the landscape that did

153

undergo ecological change are those Rasic (2008, 2011) identified in a Sluiceway Northern

Paleoindian land-use model as important to their seasonal rounds in the western Brooks Range.

If these areas underwent significant ecological change then humans would have been directly affected. Most of the Northern Paleoindian sites in the Brooks Range are hunting lookouts for migrating mammals and retooling stations. As can be seen in Figure 5.7 these hunting lookout locations are in areas where ecological change was occurring. If ecological change resulted in the loss of habitat for large mammals, especially bison, the behavior and distribution of the animals would have shifted. Northern Paleoindian hunters might have abandoned these areas in the face of resources degradation, even in the lowland areas inferred as Sluiceway winter camps. Kunz et al. (2003) and Mann et al. (2001) believe that the warmer and wetter HTM shifted the ecology to one dominated by moist acidic tussock tundra which resulted in the loss of bison habitat. This loss of habitat would have forced bison, a primary resource of Mesa complex hunters, out of the area and could have possibly contributed to its abandonment by Northern Paleoindian peoples.

Figure 5.7. Northern Alaska digital elevation model with vegetation change from the Younger Dryas to the Holocene Thermal Maximum and Northern Paleoindian archaeological sites. Green circles, sites from this study; red circles, known Northern Paleoindian sites. Blue, vegetation change; gray, no vegetation change.

154

As evident from the model’s projections, the ecological change indicates northern Alaska

became much more sensitive to climate change during the Holocene. The hypothesis presented

by Kunz et al. (2003) and Mann et al. (2001) concerning Pleistocene-Holocene climate changes and ecological changes in northern Alaska from the YD to the warmer and wetter HTM are met in the paleoecological projections presented here. From the BA to the YD a relatively small proportion of the landscape changed, which would have led to a continuation in the distribution of large mammal species that are extant today, as well as the human organization and behavior surrounding their procurement. If bison were a key species to these Mesa hunter-gatherers in the

Arctic Foothills, the YD would have been more favorable than the BA and would have allowed

for a prolongation in their numbers based on how little of the landscape changed ecologically.

Mesa and Sluiceway projectile point makers possibly shared a learning tradition. If these two

point types represent functional variations for different hunting situations in different areas of the

Brooks Range, the low amount of landscape turnover during the YD would not likely have warranted any technological or cultural adaptation. The expectations is that this would equate to a continuation of lithic technological organization with Sluiceway points in the western Brooks

Range and Mesa points in the central Brooks Range.

However, moving into the HTM the model projection sees a large turnover on the

landscape of 59%, which is mostly the result of moist acidic tussock tundra spreading into the

area. This large amount of ecological change would have affected the Northern Paleoindians in

the area and a response could be expected. This could materialize as a reorganization in the lithic

technology, or as a reorganization of Paleoindian rounds across the landscape. This model

projection supports the inferences of Kunz et al. (2003) and Mann et al. (2001), who propose that

the ecological transition from the HTM and the spread of tussock tundra would have driven

155

bison from the area. Based on the hypothesized seasonal rounds targeting bison populations in

the Arctic Foothills during summer months and other resources such as caribou in the Brooks

Range during the fall and spring (Rasic 2008, 2011), the decline and migration of bison could

have forced Northern Paleoindians from the region. This is similar to the pattern that is seen in the archaeological record. While caribou favor moist, acidic tundra, their numbers are expected to have been low during a transitional ecosystem, which would have taken time to reverse. In the meantime, the sensitivity of this area to ecological change from the YD to the HTM was great and the repercussions forced Northern Paleoindians from the region.

The projections from this study corroborate the environmental findings of Kokorowski et al. (2008) highlighting the varying ecological impact of the YD across Alaska. Paleoecological projections from this study indicate that some areas, such as central Alaska, were highly sensitive to climate changes with over 60% landscape change from the BA to YD, and 80% landscape change from the YD to HTM. In northern Alaska the effect of the YD was much more muted, as the modeling results indicate that only 18% of the landscape changed from the BA to the YD.

However, landscape change during the transition from the YD to the HTM in northern Alaska was much higher, with close to 59% of the landscape undergoing alteration. This projection of large-scale ecological change after the YD into the HTM is evident in the palynological record and the landscape of Alaska today. The modern distributions of plant communities in northern

Alaska did not exist until the early Holocene (Mann et al. 2002). This model correctly predicts the spread of spruce and broadleaf forests into central Alaska and the spread of tussock tundra and moist acidic tundra in northern Alaska after the YD, during the HTM. This was due to more mesic conditions resulting from higher temperatures and melting permafrost. In northern Alaska this meant the spread of tussock tundra which could have led to the abandonment of the area by

156

Northern Paleoindians resulting from the loss of large mammal habitat and a lower carrying capacity for key resources, such as bison (Kunz et al. 2003; Mann et al. 2001).

Northern Paleoindian Topographic Morphometrics

If the Mesa and Sluiceway lithic technologies share a learning tradition, then the expectation is that their point clouds should overlap indistinguishably. However, if they are from separate learning traditions then the expectation would be for two separate, distinguishable point clouds in the data. Because a majority of the Sluiceway and Mesa projectile points are broken or heavily reworked, topographic morphometrics analysis was only conducted on the Haft ± ¼ max thickness isocontours. The rest of this chapter will focus on comparison within point type, and a comparison of point types using PCA and LDA.

One of the biggest differences between Mesa and Sluiceway points is their size

(Sluiceway are usually much larger). A bivariate scatterplot (Figure 5.8) of max thickness and max width/haft width clearly highlights the size differences between the two projectile point types. Mesa and Sluiceway point clouds are separate from the other, and the separation at 1.09 cm max thickness and 2.55 cm max width is meaningful. Ninety-one percent of Sluiceway points are greater than 1.09 cm in thickness and 2.55 cm in width while 83% of Mesa are less than 1.09 cm in thickness and 2.55 cm in width.

157

Figure 5.8. Scatterplot of maximum thickness and width for the Mesa and Sluiceway Northern Paleoindian projectile points by type.

Within-Type Comparison

A comparison of the projectile points from sites assigned to the Mesa and Sluiceway

complexes was first undertaken to look at variability within each type, starting with the projectile

points from the Mesa complex (Figure 5.9). The PCA plot shows that the Mesa site point cloud is tightest except for a small number of outliers. There is some separation of the data. The Spein

Mountain and Putu-Bedwell points are separated from the point clouds of Mesa and Hilltop points along PC 1, which explains 72.4% of the variability. However, there are still points from the Mesa-type site within this point cloud.

158

Figure 5.9. Mesa Type PCA. Hilltop, open red circles; Mesa, open purple triangles; Putu- Bedwell, blue plus-sign; Spein Mountain, green X-sign.

The Mesa-type points were also subjected to LDA to see how well they plotted by site

(Figure 5.10). In the LDA plot three overlapping point clouds are visible. One cloud includes

Mesa, Spein Mountain, and Hilltop points, which are separated along LD 2 from a second cloud with Mesa and Hilltop points. The third cloud contains most of the Putu-Bedwell assemblage,

Mesa, and a Hilltop point, and are separated from the two other point clouds along LD 1 which explains 56.8% of the variability. Although within-cloud variability exists, all specimens fall within the scatter of points from the Mesa site.

159

Figure 5.10. Mesa Type LDA. Hilltop, open red circles; Mesa, open purple triangles; Putu- Bedwell, blue plus-sign; Spein Mountain, green X-sign.

Examination of the Mesa LDA reclassification accuracy (Table 5.3) also supports the inference that these points belong to the Mesa type. The overall reclassification accuracy is

62.1%, with Mesa site points classified correctly the most frequently. When misclassifications occur, points from other sites are classified as Mesa site points.

160

Table 5.3. Mesa Type LDA Predictions and Confusion Matrix.

Reclassified Putu- Spein Actual (62.1%) Hilltop Mesa Bedwell Mountain Total Hilltop (0%) 0 5 0 0 5 Mesa (89.7%) 0 52 4 2 58 Putu-Bedwell (9.1%) 0 8 1 2 11 Spein Mountain (7.7%) 1 10 1 1 13

The same analyses were conducted on the Sluiceway point assemblages. The results of the PCA plot show a tight point cloud between points from sites defined to the Sluiceway complex (Figure 5.11). Apart from a handful of outliers, there is an indistinguishable point cloud from the six Sluiceway sites. There is no separation of point clouds along either PC 1 or PC 2, which both explain 89.6% of the variability.

161

Figure 5.11. Sluiceway Type PCA. Caribou Crossing 1, open red circles; Caribou Crossing 2, open pink triangles; Tuluaq, open downward green triangles; Red Dog, blue X-sign; Upper Kelly, open light green square; Nat Pass, purple plus-sign.

The LDA plot results from the Sluiceway points can be seen in Figure 5.12. Caribou

Crossing 1 contains the largest sample of Sluiceway points. The point cloud of Caribou Crossing

2 is also extremely tight and overlaps the Caribou Crossing 1 cloud. Projectile points from the

Tuluaq and Red Dog sites also overlap with the Caribou Crossing point cloud. Separation is present along LD 1 between the points from Caribou Crossing 1 and 2 as well as the Tuluaq points, and the Red Dog and Upper Kelly points. There is no separation of clouds along LD 2.

162

Figure 5.12. Sluiceway Type LDA. Caribou Crossing 1, open red circles; Caribou Crossing 2, open pink triangles; Tuluaq, open downward green triangles; Red Dog, blue X-sign; Upper Kelly, open light green square; Nat Pass, purple plus-sign.

Based on the PCA and LDA plots there is overlap of point clouds amongst all of the points from the different sites attributed to the Sluiceway complex. This is made evident by projectile points from the various Sluiceway sites that fall within the range of Caribou Crossing 1 points. The reclassification accuracy of Sluiceway points using LDA is presented in Table 5.4.

163

Table 5.4. Sluiceway Type LDA Predictions and Confusion Matrix.

Reclassified Actual (53.4%) CC1 CC2 Nat Pass Red Dog Tuluaq Upper Kelly Total CC1 (81.4%) 57 5 2 4 2 0 70 CC2 (5%) 16 1 0 0 5 0 22 Nat Pass (0%) 1 0 0 1 0 0 2 Red Dog (0%) 5 0 1 0 0 0 6 Tuluaq (31.3%) 8 3 0 0 5 0 16 Upper Kelly (0%) 1 0 0 1 0 0 2

Although the total overall reclassification accuracy is 53.4%, a majority of the Caribou

Crossing 1 points are correctly classified. Caribou Crossing 1 points represents the largest

sample population of the Sluiceway sites and are correctly classified 81.4% of the time. When

misclassification of points from other sites occurred, a majority of the misclassification were

classified as Caribou Crossing 1.

Type Comparison

Following the above within-type comparisons of Mesa and Sluiceway points that was conducted to examine internal variability within these two northern Paleoindian types, a between-type comparison was done to assess the cultural relatedness of the two complexes based on their flake scar patterning. The Mesa and Sluiceway PCA plot (Figure 5.13) shows overlap between the two types. This is not what would be expected expect from two distinct learning traditions.

164

Figure 5.13. Northern Paleoindian PCA. Mesa red circles; Sluiceway, green triangles.

There are interesting nuances in the point clouds in the PCA plot. Overlap exists between the Mesa and Sluiceway points. However, a portion of the Mesa point cloud does not overlap with the Sluiceway cloud, and to a lesser extent some of the Sluiceway cloud does not overlap with the Mesa point cloud. This pattern is important because the earliest, and oldest of the Mesa complex dates, are younger than the earliest Sluiceway complex dates, and the Mesa complex date extends past the termination date for the Sluiceway complex. Does this overlap between portions of the Mesa and Sluiceway point clouds represent an offshoot learning tradition that inhabited the central Brooks Range? The overlap in data indicates a shared learning tradition, but the partial separation of the two clouds is unexpected and perhaps represents the split and the movement of peoples from the western Brooks Range to the central Brooks Range.

165

The Northern Paleoindian LDA reclassification accuracy (Table 5.5) is good at predicting

between the flake scar isocontours of Mesa and Sluiceway points (total accuracy 89.8%). The

PCA plots do not show separate point clouds that one would expect if the Mesa and Sluiceway

point types represented two separate learning traditions. However, under this assumption, the

fact that LDA correctly reclassified points at such a high rate is not expected. Although the PCA

indicates the two projectile point types share a learning tradition, variability is visible in the PCA

and LDA results.

Table 5.5. Northern Paleoindian LDA Predictions and Confusion Matrix.

Classified Actual (89.8%) Mesa Sluiceway Total Mesa (87.4%) 76 11 87 Sluiceway (91.5%) 10 108 118

This high degree of correct classification is of importance. The within-type analysis showed that there is variation between the types, but that each point cloud overlaps as would be expected if each is indeed a projectile point “type” that shares a learning tradition. The PCA results in Figure 5.13 show considerable overlap of Mesa and Sluiceway point types. This is not what would be expected if they were unrelated learning traditions. Variation might explain some of the overlap. However, the following results are important: (1) the two types differ morphologically and are clearly different in average size, (2) the point clouds from both complexes overlap with each other in PCA, and (3) the LDA reclassification can correctly predict between the two types with 90% accuracy. The first two observations indicate that even though Mesa and Sluiceway points look considerably different from each other

(morphologically), they both have very similar flake scar patterning resulting from a similar

166

flintknapping knowledge and technique that in turn is consistent with a shared learning tradition.

However, the LDA reclassification accuracy is not expected if these two types were related, as

the flake scar pattering should be indistinguishable decreasing the LDA reclassification accuracy.

The potential for the overlap in point clouds seen in PCA and LDA plots to be related to

geographic or chronological patterns was worth evaluating, such as with the Plains Late

Paleoindian points. Unfortunately, after examining PCA and LDA plots, as well as LDA

reclassification accuracy, no such patterns could be identified that could shed light on the origin

or subsequent geographic movement between the Mesa and Sluiceway point types. With a larger sample size from more sites that have radiometric data, the hope is that such patterns might

become evident. This analysis did benefit the study by providing additional support for

inferences made earlier surrounding the Spein Mountain site. The overlap of the Spein Mountain

point cloud with the points from the Mesa and Hilltop sites supports the inference that Spein

Mountain belongs in the Mesa complex. Even though the Spein Mountain site lies 900 km SW of

the Mesa site it is worth pointing out that in the Great Plains ~1,300 km separates the Goshen

Mill Iron site and the Plainview type site. A lot of this distance in Alaska has not been surveyed archaeologically, and more Mesa sites are yet to be discovered.

Northern Paleoindian Conclusions

Even though Mesa points and Sluiceway points are visibly and metrically different in outline and size, there is considerable overlap between the Mesa and Sluiceway point clouds in the PCA plots indicating a similarity in flake scar patterning. Morphologically different projectile point types are not expected to share similar flake scar patterning unless they shared a common flintknapping knowledge or learning tradition. The topographic morphometric analysis

167

support the inference that Mesa and Sluiceway share a common flintknapping knowledge and technique. The small number of radiocarbon dates from Mesa and Sluiceway complex sites in this study coupled with the fact that the overlap between Mesa and Sluiceway points in the PCA and LDA plots are not spatially or chronologically correlated hinders the understanding in which the shared learning tradition developed. The radiometric data that does exist from all dated archaeological sites of these types indicates there is temporal overlap between Sluiceway points in the western Brooks Range and Mesa points from the central Brooks Range (Table 5.6).

Table 5.6. Chronological Range of Northern Paleoindian Types.

Radiocarbon Cal Age Range Type Age BP Location Source 13,200 to Western Brooks Sluiceway 11,200 to 9,910 ~11,500 Range Smith et. al 2013 12,500 to Central Brooks Kunz et al. 2003; Mesa 10,240 to 9,800 ~11,100 Range Smith et al. 2013

Two possible scenarios surrounding the cultural relatedness of the Mesa and Sluiceway projectile point types can be addressed from the data in this study. One possibility is that these two point styles represent different bifacial tool types made and used by the same people. The larger and heavier Sluiceway points were perhaps hafted to thrusting spears focused on specific prey species in the Noatak River Basin and surrounding uplands of the western Brooks Range.

Conceivably, the same people made Mesa points for a different task, perhaps as tips on propelled projectiles such as darts, in the Arctic Foothills of the central Brooks Range. The topographic morphometric analysis partially supports this interesting possibility, particularly since there is considerable temporal overlap of the two point styles in the Brooks Range. In addition, the paleoecological projections show varying amounts of ecological change across the Brooks Range

168

from west to east during this time period which could have led to different areas supporting different prey species in varying numbers. While this is not definitive evidence by any means, the results lend support to the inference that Mesa points and Sluiceway points could represent functional variation targeting different resources in different areas of the Brooks Range, all from the same learning tradition.

The considerable overlap that exists in the PCA is not expected if the two point types were unrelated. However, they are not completely indistinguishable, and the LDA was very successful in distinguishing between the two point types (~90% accuracy). A pattern that is not expected if these two point types were part of the same culture. The topographic morphometric analysis presents a second interesting scenario. The pattern in the data could be showing the variation that occurs in a learning tradition after a split occurs, as discussed at the end of Chapter

4 on the Great Plains. The topographic morphometric data from the Mesa and Sluiceway complexes might reflect a pattern that lies somewhere between the few hundred-year, nearly indistinguishable overlap of Late Paleoindian on the Plains, and the longer separation between those types and the Middle Archaic Nebo Hill/Sedalia component ~6,000 years later.

Some of the Mesa point cloud does not overlap with the Sluiceway points. In this scenario, Mesa could represents an eastward offshoot into the Arctic Foothills of the central

Brooks Range. Although there is temporal overlap of these two point types, current radiometric data indicate that Sluiceway points in the western Brooks Range are older than Mesa points from the central Brooks Range, and that Mesa points continued for a few hundred years after

Sluiceway disappear. A picture that is painted by the current archaeological and topographic morphometric data indicate that both Sluiceway and Mesa points potentially share a learning tradition but that Mesa might represent an eastward movement. There is overlap in the PCA

169

results that support a shared learning tradition, but the accuracy in which LDA could separate

between the two point types is not expected if they indeed represent flakes scar patterning from

the learning tradition. While it cannot be ruled out that the eastward movement and appearance

of Mesa points was the result of targeting different prey species than the Sluiceway points in the

central Brooks Range area, the topographic morphometric analysis indicates that this variation

possibly represents a split in a learning tradition. The topographic morphometric data, in conjunction with the archaeological data, present two scenarios for the presence of these two projectile point types in the Brooks Range. It should be noted that these are only hypothetical scenarios at this point, and that further archaeological testing as well as the inclusion of more topographic morphometric data are needed before any stronger conclusions can be made about this patterning in the data.

This portion of the study also contributes to an understanding of the Spein Mountain archaeological site in regards to where it fits in Alaskan prehistory. The Spein Mountain site is the only Mesa type site not located in the Brooks Range. Although the Spein Mountain site is over 900 km southwest of the Mesa type site, it is considered part of the Mesa complex based on

its lithic technological organization (Ackerman 2001; Bever 2000). The topographic

morphometric analysis from this study shows that the Spein Mountain site point cloud overlaps with the Mesa-type site and Hilltop projectile points, and that LDA reclassified Spein Mountain points as Mesa-type site points 77% of the time. The data from this research indicate that the

Spein Mountain points do indeed belong to the Mesa complex based on their flake scar patterning.

170

CHAPTER SIX: SUMMARY AND CONCLUSIONS

The application of novel analyses to previously excavated assemblages provides a new and different viewpoint for interpreting patterns in the archaeological record to address core questions surrounding cultural relatedness in the past. It is important to realize that this methodology requires very specific and thought out questions with appropriate data. While traditional geometric morphometrics of artifacts continue to be of great value, they reduce 3D

objects into 2D measurements. The application of 3D laser scanning technology allows archaeologists to move beyond traditional lithic analyses into questions surrounding the “degree

and pattern of historical relationship and descent, about transmission modes and the role of drift,

about the complex patterns of allometric variation that attend resharpening and reduction.

Scanning technology…provide new data to test theory never before contemplated” (Shott and

Trail 2012:17). The methodology presented here would not have been possible without 3D scanning technology. It is through the application of new methods, such as 3D imagery and topographic morphometrics, that previously unanswerable questions can be addressed.

This study explored multiple methods for conducting topographic morphometric analysis

by looking at varying isoheight measurement locations on projectile points, and looking at

different elements of projectile points. Isocontours of flake scar patterning were taken from

isoheights of ±¼ and ±⅓ max thickness from the center plane of the projectile point on each

face. Results indicate that the ±¼ max thickness measurement better captures the flake scar

patterning information regarding manufacture technique than the ±⅓ max thickness. In addition,

two areas of the same projectile points were analyzed: the isocontour of the complete point, and

the isocontour of the haft element only which was “digitally broken” for analysis. Because both

measurement locations came from the same assemblage of projectile points a comparison could

171

be made to see if one location was better than another for conducting topographic morphometrics. This was done for two reasons: 1) the effect of varying intensities of curation on the blade element of different points even within the same type might affect the results, and 2) most projectile points in archaeological assemblages are not complete, and having the ability to run topographic morphometrics on point bases would greatly increase the applicability of this method to other assemblages. Results indicate that the ±¼ haft max thickness measurement was as good, if not better, than that of the complete ±¼ max thickness measurement, meaning that topographic morphometrics can be applied on bases that at least include the haft element. This is important as it greatly increase the number of points and assemblages archaeologists can apply topographic morphometrics to. This methodology and analytical technique was then applied to two Paleoindian case studies, the Late Paleoindians on the Great Plains and the Northern

Paleoindian complexes in Alaska, to help understand the culture history in these regions.

The results of this study demonstrate that similarities and differences in the flake scar patterning on projectile points that resulted from manufacture can be successfully identified using topographic morphometrics of isocontours on 3D models. These patterns are the result of similar knowledge and technique in the manufacture of lithic projectile points. This novel application to lithic studies was first introduced by Sholts et al. (2012) in examining the flake scar isocontours of Clovis bifaces. Based on similarities, they surmised the results to be indicative of a widespread Clovis “culture” in North America. The study presented here expanded on this and introduced multiple projectile point types attributed to different cultures.

This was done to test whether the methodology could be used to identify different archaeological

“cultures.” The conceptual basis of this study is based in social learning and archaeological assumptions surrounding culture historical transmission of traditions such as flintknapping

172

knowledge and technique. Passed from generation to generation through a learning tradition in small, highly mobile hunter-gatherer groups, the knowledge and technique for how to manufacture projectile points are expected to be more similar among closely related groups than between two or more groups that are unrelated. This produces similar flake scar patterning on projectile points that can be identified using topographic morphometric analysis. These data allow patterns to be identified and assemblages of projectile points to be examined for cultural relatedness. These “cultures” or learning traditions describe the relatedness of groups of people based on a shared flintknapping knowledge and technique that was passed down through generations of flintknappers. Groups closer in a learning tradition should produce flake scar pattern point clouds that overlap more than those groups that are further away or are not part of it at all.

In examining Late Paleoindian projectile points from the Great Plains attributed to the

Goshen, Plainview, and Milnesand types, topographic morphometric analysis indicates that these point clouds overlap indistinguishably from one another but separately from the point cloud of the later Middle Archaic lanceolate Nebo Hill/Sedalia points. The Nebo Hill/Sedalia points are temporally (~6,000 years) and geographically distinct from the three Paleoindian types. The isocontour outlines of the three Paleoindian types are nearly indistinguishable from one another within the PCA plots and LDA reclassification confusion matrix. The results of PCA plots, LDA plots, and LDA reclassification are used to infer a shared learning tradition among the Plains

Late Paleoindian projectile point types. Based on radiometric data and the geographic distribution of these three Paleoindian types, a potential southern movement of a Goshen learning tradition from the northern Plains to the southern Plains, becoming what archaeologists call Plainview is viable. The Milnesand point type in eastern New Mexico could represent a later

173

movement emanating from Plainview. Topographic morphometric analysis successfully identified similarities in flake scar patterning to address questions concerning the cultural relatedness of Late Paleoindian projectile types on the Plains based on the flake scar patterns from these different projectile point types.

In central Alaska during the PHT, the sample size of bifacial projectile points from dated

Nenana and Denali sites was too small to produce meaningful data using topographic morphometrics. As a result, topographic morphometrics could not be used help understand the culture history of the region, or their relationship to the Northern Paleoindian complex in the

Brooks Range, Alaska. However, the application of topographic morphometrics to unfluted

Northern Paleoindian bifacial projectile points from the Mesa and Sluiceway complexes was informative. Results of the PCA indicate that Mesa and Sluiceway points share a learning tradition, however, the LDA results did not fit with this expectation. Interpretations of the data provide two possible scenarios that could indicate either a functional difference in tool types from the same group of people in different geographic areas, or possibly the separation in a learning tradition from Sluiceway in the western Brooks Range to Mesa in the central Brooks

Range. In addition, this study was able to corroborate existing assumptions based on the archaeological record. The results of this study support the interpretation of the Spein Mountain site, located in the Kuskokwim River Drainage in southwest Alaska, as being a Mesa complex site even though it is around 900 km from the type site of the Mesa complex (Ackerman 2001,

Bever 2000, 2001; Kunz et al. 2003).

Topographic morphometrics is based on shape analysis of complex flake scar patterning.

Multivariate analyses, including PCA and LDA, were used to examine the relationships of the point clouds created in these plots by type to see if expectations between related and unrelated

174

learning traditions were met based on their clustering or separation from one another. However,

there were unexpected discrepancies in the PCA and LDA results from the Northern Paleoindian

case study that did not fit the expectations set forth regarding related and unrelated learning

traditions. Principle Components Analysis determines the axes with the most amount of variance,

while LDA identifies the axes that maximize the separation between defined point types. Thus,

these two multivariate analyses might be picking up on different characteristics of the flake scar patterning outlines. PCA might pick up on one characteristic while LDA picks up on another in the outline shape of the flake scar patterning, and this could explain why the Northern

Paleoindian case study did not completely fit the expectations of a shared learning traditions. At

this time, trying to understand how variation in flake scar patterning and how the various

characteristics that these different multivariate analyses might be picking up on relate to

expectations surrounding learning traditions, cannot be addressed. These multivariate analyses

might, in the future, be able to identify different approaches by which to look at learning

traditions and technical traits such as flake scar patterning. In future research on this topic, one

strategy to address this would be to conduct a lithic experiment wherein a single flintknapper,

using the same raw material, reproduces n number of biface type 1 and an equal number of

biface type 2 (such as Mesa and Sluiceway projectile points). The flintknapper would be directed

to manufacture these two morphologically distinct projectile point types using the same

flintknapping knowledge and technique. This experimental design would, hold flintknapping

technique and raw material constant, and thus, could be used to identify the characteristics PCA

and LDA pick up on. A better understanding of which characteristics of flake scar patterning

these two multivariate analyses are picking up on could be used to develop new theoretical approaches by which to examine learning traditions.

175

Future Work

Future work that builds on this study would benefit from the inclusion of a larger sample of both potentially related and unrelated projectile points that are separated by time and space.

This would help address some of the culture history and technological relatedness issues raised in this study. First and foremost would be to analyze the morphometrics of Plains Late

Paleoindian Agate Basin projectile points, to see how this Paleoindian point type fits into the picture this study presented. The similarities between Mesa points from Alaska and Agate Basin points from the Paleoindian period in the Great Plains have been noted by Kunz et al. (2003) and

Meltzer (2009). If Paleoindians from the contiguous Unites States and Alaska are related, in a learning tradition sense, how did they come to be over 5,000 kilometers apart and, more importantly, where did the Northern Paleoindian complex come from? What seems clear is that the Northern Paleoindian complex is tied to North America and not to Siberia. No fluted points have been found across the Bering Strait to the west, and no points analogous to either the

Sluiceway or Mesa complexes have been found in Siberia (Goebel et al. 2013; Meltzer 2009;

Slobodin 2011).

In Alaska, the only other archaeological sites at the PHT come from the Nenana and

Denali complexes in central Alaska. It has been argued that there could be a connection between

Mesa and Denali, as some Mesa-like bifacial projectile points have been recovered from sites south of the Brooks Range (Ackerman 2001; Bever 2000; Hoffecker 2011). The recovery of bifacial lanceolate projectile points and bases from the Denali microblade contexts at Dry Creek and Moose Creek are argued by Hoffecker (2011) to be representative of Mesa points. However,

Smith et al. (2013) do not see these specimens as characteristic of Mesa points, based on differences in the bifacial technologies. Although the Denali bifaces are morphologically

176

lanceolate, they do not exhibit the lengthy production sequence involving percussion and

pressure flaking, along with extensive edge grinding, that is present in Mesa points. In addition,

there is variability in bifacial projectile point technology from central Alaskan sites, and the

points occur only in small numbers (Holmes 2001; Powers and Hoffecker 1989). This contrasts

with Northern Paleoindian unfluted lanceolate projectile points, which display considerable

standardization and occur in large numbers. The only recognized Mesa site south of the Brooks

Range is Spein Mountain in the Kuskokwim River Drainage (Ackerman 2001; Bever 2000,

2001). The relationship between the two artifact complexes in northern and central Alaska

remains, thus, unclear.

The question surrounding the origin of Northern Paleoindian can be narrowed down to

two possible explanations. Either it represents an independent invention in Alaska, or it is the

product of a diffusion of people and/or technology from Paleoindians on the Great Plains

(Goebel et al. 2013; Meltzer 2009). About 12,500 cal yrs BP during the YD, genetic evidence

from bison points to a northward expansion from the northern Plains into the opening ice-free

corridor (Shapiro et al 2004; Wilson and Shapiro 2008). Paleoindian bison hunters on the Great

Plains could have followed these migrating herds, eventually reaching Alaska. If people followed

bison northward ~12,500 cal yrs BP, this timing would likely post-date Clovis and could

correspond to late Folsom, Goshen, and possibly Agate Basin Paleoindians in the northern

Plains.

The earliest evidence of Fluted Northern Paleoindian points in Alaska dates from

~12,400 cal BP, at the Serpentine Hot Springs site (Goebel et al. 2013). Based on the limited

chronological data on the Fluted complex in Alaska, it does not appear that the Fluted complex

in Alaska is antecedent to Clovis. The radiocarbon dates from Serpentine Hot Springs (~12,400

177

cal yrs BP) and Raven Bluff (~12,100 cal yrs BP) in Alaska, as well as Charlie Lake Cave

(~12,500 cal yrs BP) in British Columbia, support the hypothesis of a northward movement of fluting technology ~12,500 cal yrs BP (Driver 1999; Driver et al. 1996; Goebel et al. 2013,

Hedman 2010; Smith and Goebel 2018). The archaeological evidence points to either a demic diffusion of Late Paleoindian groups from the northern Plains or a transmission of fluting technology on lanceolate projectile points.

If Northern Paleoindian Mesa complex projectile points shared a learning tradition with

Agate Basin, similarities in flake scar patterning would be expected. To better understand this, the topographic morphometric analysis of Agate Basin points in the future is required to assess the proposed northward migration of Plains Paleoindians ~12,500 cal yrs BP. In addition to the topographic morphometric analysis, an additional line of evidence to test this hypothesis of a northward Paleoindian movement following bison herds would be to examine the reintroduction of bison into the Northwest Territories during the 1960’s (Larter et al. 2000). This data could be used to examine the rate of yearly migration bison display. This in turn could be used to model the northward spread of bison ~12,500 cal yrs BP (Shapiro et al. 2004; Wilson et al. 2008), focusing on expansion distance per year to predict how long it would have taken bison, and by extension Paleoindian hunter-gatherers, to make their way to Alaska. If Paleoindian hunters from the Great Plains followed bison herds north through the recently opened and vegetated ice-free corridor, they would not likely have exceeded the movement of their target prey. Depending on the modeled rate of bison expansion, topographic morphometrics of projectile points could provide an additional line of evidence for a Paleoindian movement into Alaska. The Late

Paleoindian Agate Basin type from the Great Plains is seen as morphologically similar to Mesa points in Alaska, and presents the opportunity to test the northern migration hypothesis at the end

178

of the Pleistocene using topographic morphometrics. As more technological, spatial, and

temporal data are collected, the ability to answer these questions and our understanding of the

culture history of these projectile point types will become stronger.

The incorporation of Clovis points, while difficult because of their fluting, would be an

interesting addition to this study because Clovis currently represents the likely ancestral culture group to the Late Paleoindians on the Great Plains and elsewhere across North America. This includes the Northern Paleoindian Fluted complex in Alaska (Smith and Goebel 2018), and it would be interesting to see how topographic morphometrics between fluted points in these two geographic regions compare. Another interesting avenue to explore would be an EFA of channel flakes using PCA and LDA to see if there are temporal and spatial patterns in these artifacts that

relate to the movement of Clovis, and later Folsom cultures across North America.

The focus of this study was on the analysis of larger, unfluted, lanceolate projectile

points. It would be worth exploring whether this methodology can also successfully detect

variation in flake scar patterning from notched projectile points hafted to either darts or arrows.

Before comparisons of notched and lanceolate points can be made it has to be determined that

the methodology works on notched points. The morphological differences between lanceolate

and notched points (like apples to oranges) might be too great to compare topographic

morphometrics of flake scar patterning to better understand the culture history of a region.

However, topographic morphometrics could be used to compare different types of notched points

to each other to test for cultural relatedness. One particularly interesting application of

topographic morphometrics would be a study surrounding the transition of the atlatl and dart to

the bow and arrow. An appropriate pilot study would be to apply topographic morphometrics to the assemblage of Rosegate and Elko Eared projectile points from the Great Basin to look at the

179

transmission of bow and arrow technology as discussed by Bettinger and Eerkens (1999). In sum, the application of topographic morphometric studies of the kind presented here, has the potential to provide new insights that can help archaeologists address a variety of interesting and

important questions concerning culture history among archaeologically defined projectile point complexes.

180

REFERENCES

Ackerman, Robert E. 1996 Spein Mountain. In American Beginnings: The Prehistory and Paleoecology of Beringia, edited by Frederick H. West, pp. 456-461. University of Chicago Press, Chicago. 2001 Spein Mountain: a mesa complex site in southwestern Alaska. Arctic Anthropology 38(2):81-97. 2004 The northern archaic tradition in southwestern Alaska. Arctic Anthropology 41(2):153-162. 2008 Security Cove and the Northern Archaic Tradition Revisited. Arctic Anthropology 45(2):146-168.

Alaska Geospatial Data Committee. Statewide Vegetation/Land Cover. Available from: http://agdc.usgs.gov/data/projects/fhm/index.html.

Alley, Richard B. 2000 The Younger Dryas cold interval as viewed from central Greenland. Quaternary Science Reviews 19(1):213-226.

Amick, Daniel S. 1996 Regional patterns of Folsom mobility and land use in the American Southwest. World Archaeology 27(3):411-426.

Anderson, Patricia M., Patrick J. Bartlein, and Linda B. Brubaker 1994 Late Quaternary History of Tundra Vegetation in Northwestern Alaska. Quaternary Research 41:306-315.

Anderson, Patricia M., Anatoly V. Lozhkin, and Linda B. Brubaker 2002 Implications of a 24,000-Yr Palynological record for a Younger Dryas cooling and for boreal forest development in Northeastern Siberia. Quaternary Research 57(3):325-333.

Anderson, Patricia M., Mary E. Edwards, and Linda B. Brubaker 2003 Results and Paleoclimate Implications of 35 years of Paleoecological Research in Alaska. Developments in Quaternary Sciences 1:427-440.

Andrefsky, William Jr. 2006 Experimental and Archaeological Verification of an Index of Retouch for Hafted Bifaces. American Antiquity 71:743–759.

Apel, Jan 2008 Knowledge, Know-how, and Raw Material – The Production of Late Neolithic Flint Daggers in Scandinavia. Journal of Archaeological Method and Theory 15:91-111.

Bamforth, Douglas B., Nyree Finlay 2008 Introduction: Archaeological Approaches to Lithic Production Skill and Craft Learning. Journal of Archaeological Method and Theory 15(1):1-27.

181

Bartlein, Patrick J., Patricia M. Anderson, Mary E. Edwards, and Patricia F. McDowell 1991 A Framework for Interpreting Paleoclimatic Variations in Eastern Beringia. Quaternary International 10-12:73-83.

Barton, C. Michael 1997 Stone tools, style, and social identity: an evolutionary perspective on the archaeological record. Archeological Papers of the American Anthropological Association 7(1):141-156.

Bettinger, Robert, Jelmer Eerkins 1999 Point Typologies, Cultural Transmission, and the Spread of the Bow-and-Arrow Technology in the Prehistoric Great Basin. American Antiquity 64(2):231-242.

Bever, Michael 2000 Paleoindian lithic technology and landscape use in Late Pleistocene Alaska: a study of the Mesa Complex. Unpublished Ph.D. dissertation, Department of Anthropology, Southern Methodist University, Dallas. 2001 Stone tool technology and the Mesa complex: developing a framework of Alaskan Paleoindian prehistory. Arctic anthropology (2001): 98-118.

Bigelow, Nancy H. and Mary E. Edwards 2001 A 14,000 yr paleoenvironmental record from Windmill Lake, Central Alaska: Lateglacial and Holocene vegetation in the Alaska Range. Quaternary Science Reviews 20:203-215.

Binford, Lewis R. 1978 Nunamiut Ethnoarchaeology. Academic Press, New York.

Bleed, Peter 2008 Skill Matters. Journal of Archaeological Method and Theory 15(1):154-166.

Bonhomme, Vincent, Sandrine Picq, Cedric Gaucherel, Julien Claude. 2014 Momocs: Outline Analysis Using R. Journal of Statistical Software, 56(13):1-24. URL http://www.jstatsoft.org/v56/i13/.

Boyd, Robert, and Peter Richerson. 1985 Culture and the Evolutionary Process. University of Chicago Press, Chicago.

Bradbury, Andrew P. 1997 The Bow and Arrow in the Eastern Woodlands: Evidence for an Archaic Origin. North American Archaeologist 18(3):207-233.

Bray, R.T. 1963 COMMENTS ON THE PRECERAMIC IN MISSOURI. The Plains Anthropologist 8(22):231-237.

182

Briner, Jason P., Darrell S. Kaufman, Al Werner, Marc Caffee, Laura Levy, William F. Manley, Michael R. Kaplan, and Robert C. Finkel 2002 Glacier readvance during the late glacial (Younger Dryas?) in the Ahklun Mountains, southwestern Alaska. Geology (8):679-82.

Brubaker, Linda B., Patricia M.Anderson, Mary E. Edwards, and Anatoly V. Lozhkin 2005 Beringia as a glacial refugium for boreal trees and shrubs: new perspectives from mapped pollen data. Journal of Biogeography 32:833-843.

Brzeziecki, Bogdan, Felix Kienast, and O. Wildi 1993 A simulated map of the potential natural forest vegetation of Switzerland. Journal of Vegetation Science 4(4):499-508.

Buchanan, Briggs, Eileen Johnson, Richard E. Strauss, and Patrick J. Lewis 2007 A morphometric approach to assessing late Paleoindian projectile point variability on the Southern High Plains. Plains Anthropologist 52(203): 279-299.

Buchanan, Briggs, Michael J. O’Brien, and Mark Collard 2017 A Geometric Morphometrics-Base Assessment of Point types on the Southern High Plains during Plainview Times. In Plainview: The Enigmatic Paleoindian Artifact Style of the Great Plains edited by Vance T. Holliday, Eileen Johnson, and Ruthann Knudson, pp 274-284. University of Utah Press, Salt Lake City.

Byerly, Ryan M., Judith R. Cooper, David J. Meltzer, Matthew E. Hill, and Jason M. LaBelle 2005 On Bonfire Shelter (Texas) as a Paleoindian bison jump: an assessment using GIS and zooarchaeology. American Antiquity 70(4):595-629.

Cavalli-Sforza, L. and Feldman, Marcus. 1981 Cultural Transmission and Evolution: A Quantitative Approach. Princeton University Press, Princeton.

Cinq-Mars, J., C. R. Harington, D. E. Nelson, and R. S. MacNeish 1991 Engigstciak revisited: A note on early Holocene AMS dates from the “Buffalo Pit.” In NOGAP Archaeology Project: An Integrated Archaeological Research and Management Approach, edited by J. Cinq-Mars and J. L. Pilon, pp. 33–44. Occasional Paper Number 1, Canadian Archaeological Association, Montreal.

Clark, Donald W. 2001 Microblade-Culture Systematics in the Far Interior Northwest. Arctic Anthropology 38(2):64-80.

Crabtree Don, E. 1982 An Introduction to Flintworking. Occasional Papers of the Idaho Museum of Natural His- tory number 28, Pocatello.

183

Crumley, Carole L. 1993 Analyzing historic ecotonal shifts. Ecological Applications 3(3):377-384.

Cuffey, Kurt M., Gary D. Clow, Richard B. Alley, Minze Stuiver, Edwin D. Waddington, and Richard W. Saltus 1995 Large Arctic Temperature Change at the Wisconsin-Holocene Glacial Transition. Science 270(5235):455-458.

Daly, Christopher, Michael Halbleib, Joseph I. Smith, Wayne P. Gibson, Matthew K. Doggett, George H. Taylor, Jan Curtis, and Phillip P. Pasteris 2008 Physiographically sensitive mapping of climatological temperature and precipitation across the conterminous United States. International Journal of Climatology 28(15):2031-2064.

Dibble, David S. and Dessamae Lorrain Bonfire Shelter: A Stratified Bison Kill Site, Val Verde County, Texas. Texas Memorial Museum Miscellaneous Papers 1. Austin: University of Texas, Austin.

Dibble, Harold L. and Zeljko Rezek 2009 Introducing a new experimental design for controlled studies of flake formation: results for exterior platform angle, platform depth, angle of blow, velocity, and force. Journal of Archaeological Science 36(9):1945-1954.

Diefendorf, Aaron F., William P. Patterson, Henry T. Mullins, Neil Tibert, and Anna Martini 2006 Evidence for a high-frequency late Glacial to mid-Holocene (16,800 to 5500 cal yr B.P.) climate variability from oxygen isotope values of Lough Inchiquin, Ireland. Quaternary Research 65(1):78-86.

Driver, Jonathan C. 1999 Raven skeletons from Paleoindian contexts, Charlie Lake Cave, British Columbia. American Antiquity 64(2):289-298.

Driver, Jonathan C., Martin Handly, Knut R. Fladmark, D. Erle Nelson, Gregg M. Sullivan, and Randall Preston. 1996 Stratigraphy, radiocarbon dating, and culture history of Charlie Lake Cave, British Columbia. Arctic 49(3):265-277.

Edwards, Mary E., Cary J. Mock, Bruce P. Finney, Valerie A. Barbe, and Patrick J. Bartlein 2001 Potential analogues for a paleoclimatic variations in eastern interior Alaska during the past 14,000 yr: atmospheric-circulation controls of regional temperature and moisture responses. Quaternary Science Reviews 20(1):189-202.

184

Eerkens, Jelmer, and Carl P. Lipo 2005 Cultural transmission, copying errors, and the generation of variation in material culture and the archaeological record. Journal of Anthropological Archaeoogy 24:316-334 2007 Cultural Transmission Theory and the Archaeological Record: Providing Context to Understanding Variation and Temporal Changes in Material Culture. Journal of Archaeological Research 15:239-274.

Elias, Scott A. 2000 Late Pleistocene climates of Beringia, based on analysis of fossil beetles. Quaternary Research 53(2):229-235. 2001 Mutual Climatic Range reconstructions of seasonal temperatures based on late Pleistocene fossil beetle assemblages in Eastern Beringia. Quaternary Science Reviews 20:77–91.

Elias, Scott A. and Barnaby Crocker 2008 The Bering Land Bridge: a moisture barrier to the dispersal of steppe-tundra biota? Quaternary Science Reviews 27(27):2473-2483.

Elias, Scott A., Susan K. Short, C. Hans Nelson, Hilary H. Birks 1996 Life and times of the Bering land bridge. Nature 382:60–63.

Elias, Scott A., Susan K. Short, and Hilary H. Birks 1997 Late Wisconsin environments of the Bering Land Bridge. Paleogeography, Paleoclimatology, Paleoecology 136(1-4):293-308.

Elston Robert G. and P. Jeffrey Brantingham 2002 Microlithic Technology in Northern Asia: A Risk-Minimizing Strategy of the Late Paleolithic and Early Holocene. In Thinking Small: Global Perspectives on Microlithization edited by Robert G. Elston and Steven L. Kuhn, pp 103-116. Archaeological Papers of the American Anthropological Association number 12, Arlington.

Fairbanks, Richard G. 1990 The Age and Origin of the “Younger Dryas Climate Event” in Greenland Ice Cores. Paleoceanography 5(6):937-948.

Ferguson, J. R. 2008 The When, Where, and How of Novices in Craft Production. Journal of Archaeological Method and Theory 15(1):51-67.

Franklin, Janet 1995 Predictive vegetation mapping: geographical modelling of biospatial in relation to environmental gradients. Progress in Physical Geography 19(4):474-499.

Friesen, T. Max. 2013 The impact of weapon technology on caribou drive system variability in the prehistoric Canadian Arctic. Quaternary International 297:13-23.

185

Frison, George C. 1984 The Carter/Kerr-McGee Paleoindian site: cultural resource management and archaeological research. American Antiquity 49(2):288-314. 1991 Prehistoric Hunters of the High Plains. Academic Press, San Diego. 1996 Introduction. In The Mill Iron Site edited by George C. Frison, pp. 1-14. University of New Mexico Press, Albuquerque.

Gaglioti, Benjamin V., Daniel H. Mann, Benjamin M. Jones, John W. Pohlman, Michael L. Kunz, and Matthew J. Wooller 2014 Radiocarbon age‐offsets in an arctic lake reveal the long‐term response of permafrost carbon to climate change. Journal of Geophysical Research: Biogeosciences 119(8):1630-1651.

Gareth, James, Daniela Witten, Trevor Hastie, and Robert Tibshirani 2013 An Introduction to Statistical Learning: with Applications in R. Springer, New York.

Gero, Joan, and Jim Mazzullo. 1984 Analysis of artifact shape using Fourier series in closed form. Journal of Field Archaeology 11(3):315-322.

Gingerich, Joseph AM, Sabrina B. Sholts, Sebastian KTS Wärmländer, and Dennis Stanford. 2014 Fluted point manufacture in eastern North America: an assessment of form and technology using traditional metrics and 3D digital morphometrics. World Archaeology 46(1):101-122.

Goebel, Ted, Heather L. Smith, Lyndsay DiPietro, Michael R. Waters, Brian Hockett, Kelley E. Graf, Robert Gal, et al. 2013 Serpentine Hot Springs, Alaska: results of excavations and implications for the age and significance of northern fluted points. Journal of Archaeological Science 40(12):4222-4233.

Graf, Kelly E. and Ted Goebel 2009 Upper Paleolithic Toolstone Procurement and Selection Across Beringia. In Lithic Materials and Paleolithic Societies edited by Brian Adams and Brooke S. Blades, pp. 54-77. Blackwell Publishing Ltd, Chichester.

Graf, Kelly E. and Nancy H. Bigelow 2011 Human response to climate during the Younger Dryas chronozone in central Alaska. Quaternary International 242:434-451.

Grimm, L. 2000 Apprentice flintknapping: relating material culture and social practice in the Upper Paleolithic. In Children and Material Culture edited by Joanna S. Derevenski, pp.53-71. Routledge, London.

186

Guthrie, R. Dale 1982 Mammals of the Mammoth Steppe as Paleoenvironmental Indicators. In Paleoecology of Beringia edited by David Hopkins, John V. Matthews, Jr., Charles E. Schwaeger, and Steven B. Young, pp. 307-332. Academic Press, New York. 1983 Osseous projectile points: biological considerations affecting raw material selection and design among Paleolithic and Paleoindian peoples. In Animals and Archaeology: Hunters and their Prey edited by Juliet Clutton-Brock and Caroline Grigson, pp. 273-294. British Archaeological Reports International Series 163, Oxford. 1984 Paleoecology of the site and its implications for early hunters. In Dry Creek: Archaeology and Paleoecology of a Late Pleistocene Alaskan Hunting Camp edited by Roger Powers, Dale Guthrie, and John Hoffecker. Unpublished report submitted to U.S. National Park Service, pp. 209-287.

Hajdas, Irka, Georges Bonani, Per Bodén, Dorothy M. Peteet, and Daniel H. Mann 1998 Cold reversal on Kodiak Island, Alaska, correlated with the European Younger Dryas by using variations of atmospheric 14C content. Geology 26(11):1047-1050.

Hall, E.S. 1961 Eskimo-Aleut Ethnobotany. Unpublished Master’s thesis, Department of Anthropology, Yale University, New Haven.

Haynes, C. Vance and Matthew E. Hill 2017 Plainview-Goshen-Midland Typological Problems. In Plainview: The Enigmatic Paleoindian Artifact Style of the Great Plains edited by Vance T. Holliday, Eileen Johnson, and Ruthann Knudson, pp 230-248. University of Utah Press, Salt Lake City.

Hedman, William 2010 The Raven Bluff Site: Preliminary Findings from a Late Pleistocene Site in the Alaskan Arctic. U.S. Department of the Interior, Bureau of Land Management, 7.

Henrich, Joseph and Richard McElreath 2003 The Evolution of Cultural Evolution. Evolutionary Anthropology 12:123-135.

Hewlett, Barry S., L.L. Cavalli-Sforza 1986 Cultural Transmission among Aka Pygmies. American Anthropologist 88(4):922-934.

Hildebrandt, William, and Jerome King. 2012 Distinguishing between darts and arrows in the archaeological record: implications for technological change in the American West. American Antiquity 77(4):789-799.

Hill Jr, Mathew E. 2002 The Milnesand Site: site formation study of a Paleoindian bison bonebed in Eastern New Mexico. The Plains Anthropologist 47:323-337.

187

Hoffecker, John F. 2011 Assemblage Variability in Beringia: The Mesa Factor. In From the Yenisei to the Yukon: Interpreting Lithic Assemblage Variability in Late Pleistocene/Early Holocene Beringia edited by Ted Goebel and Ian Buvit, pp. 165-178. Texas A&M University Press, College Station.

Hoffecker, John F. and Scott A. Elias 2003 Environment and archeology in Beringia. Evolutionary Anthropology: Issues, News, and Reviews 12(1): 34-49. 2007 Human Ecology of Beringia. Columbia University Press, New York.

Hoffman, Ary A. and Peter A. Parsons 1997 Extreme Environmental Change and Evolution. Cambridge University Press, Cambridge.

Holliday, Vance T. 2000 The evolution of Paleoindian geochronology and typology on the Great Plains. Geoarchaeology 15(3):227-290.

Holliday, Vance T., Eileen Johnson, and Thomas W. Stafford Jr. 1999 AMS radiocarbon dating of the type Plainview and Firstview (Paleoindian) assemblages: the agony and the ecstasy. American Antiquity 64(3):444-454.

Holliday, Vance T., Eileen Johnson, and Roberta Speer 2017a The Plainview Site: History, Geology, and Zooarchaeology. In Plainview: The Enigmatic Paleoindian Artifact Style of the Great Plains edited by Vance T. Holliday, Eileen Johnson, and Ruthann Knudson, pp 1-28. University of Utah Press, Salt Lake City.

Holliday, Vance T., Eileen Johnson, and D. Shane Miller 2017b Stratigraphic Context and Chronology of Plainview Sites on the Southern Great Plains. In Plainview: The Enigmatic Paleoindian Artifact Style of the Great Plains edited by Vance T. Holliday, Eileen Johnson, and Ruthann Knudson, pp 79-102. University of Utah Press, Salt Lake City.

Holliday, Vance T., Natalia Martínez-Tagüeña, D. Shane Miller, Ismael Sánchez-Morales, Christopher W. Merriman, Allison Harvey, Rafael Cruz, Alberto Peña, and John Seebach 2017c Plainview/Belen in the Rio Grande Basin of New Mexico, Trans-Pecos Texas, and Chihuahua. In Plainview: The Enigmatic Paleoindian Artifact Style of the Great Plains edited by Vance T. Holliday, Eileen Johnson, and Ruthann Knudson, pp 189-209. University of Utah Press, Salt Lake City.

Holmes, Charles E. 2001 Tanana River valley archaeology circa 14,000 to 9000 B.P. Arctic Anthropology 38(2):154–70.

188

Hu, Feng Shen and Aldo Shemesh 2003 A biogenic-silica δ 18 O record of climatic change during the last glacial–interglacial transition in southwestern Alaska. Quaternary Research 59(3):379-385.

Hu, F. S., B. Y. Lee, D. S. Kaufman, S. Yoneji, D. M. Nelson, and P. D. Henne 2002 Response of tundra ecosystems in southwestern Alaska to Younger-Dryas climatic oscillation. Global Change Biology 8:1156-1163.

Huckell, Bruce B. and Christopher W. Merriman 2017 Paleoindian Unfluted Lanceolate Projectile Points in the Upper Little Colorado River Valley, East-Central Arizona. In Plainview: The Enigmatic Paleoindian Artifact Style of the Great Plains edited by Vance T. Holliday, Eileen Johnson, and Ruthann Knudson, pp 208-229. University of Utah Press, Salt Lake City.

Ioviţă, Radu. 2009 Ontogenetic scaling and lithic systematics: method and application. Journal of Archaeological Science 36(7):1447-1457.

Ioviţă, Radu. and S.P. McPherron. 2011 The handaxe reloaded: A morphometric reassessment of Acheulian and Middle Paleolithic handaxes. Journal of Human Evolution 61(1):61-74.

Irwin, Henry T. 1967 The Itama: early late-Pleistocene Inhabitants of the United States and Canada and the American Southwest. Unpublished Ph. D. dissertation, Department of Anthropology, Harvard University, Cambridge.

Irwin-Williams, Cynthia, Henry Irwin, George Agogino, and C. Vance Haynes, Jr. 1973 Hell Gap: Paleo-Indian Occupation on the High Plains. Plains Anthropologist 18:40-53.

Iverson, Louis R. and Anantha M. Prasad 1998 Predicting abundance of 80 tree species following climate change in the eastern United States. Ecological Monographs 68(4):465-485.

Justice, Noel D. 1987 Stone Age Spear and Arrow Points of the Midcontinental and Eastern United States: a Modern Survey and Reference. Indiana University Press, Bloomington & Indianapolis.

Kaplan, Jed O., Nancy H. Bigelow, I. Colin Prentice, Sandy P. Harrison, Patrick J. Bartlein, T. R. Christensen, W. Cramer, et al. 2003 Climate change and Arctic ecosystems: 2. Modeling, paleodata‐model comparisons, and future projections. Journal of Geophysical Research: Atmospheres 108(D19).

189

Kaufman, Darrell S., Thomas A. Ager, N. John Anderson, Patricia M. Anderson, John T. Andrews, Pat J. Bartlein, Linda B. Brubaker, et al. 2004 Holocene thermal maximum in the western Arctic (0-180°W). Quaternary Science Reviews 23(5):529-560.

Kaufman, Darrrell S., R. Scott Anderson, Feng Sheng Hu, Edward Berg, and Al Werner 2010 Evidence for a variable and wet Younger Dryas in southern Alaska. Quaternary Science Reviews (11):1445-1452.

Knudson, Ruthann 2017 The Plainview Assemblage in Context. In Plainview: The Enigmatic Paleoindian Artifact Style of the Great Plains edited by Vance T. Holliday, Eileen Johnson, and Ruthann Knudson, pp 29-78. University of Utah Press, Salt Lake City.

Kokorowski, Heather D., Patricia M. Anderson, Cary J. Mock, and Anatoly V. Lozhkin 2008 A re-evaluation and spatial analysis of evidence for a Younger Dryas climatic reversal in Beringia. Quaternary Science Reviews 27(17):1710-1722.

Kreutzer, Lee Ann 1996 Taphonomy of the Mill Iron Site Bison Bonebed. In The Mill Iron Site edited by George C. Frison, pp. 87-101. University of New Mexico Press, Albuquerque.

Krieger 1947 Bison, Fossil, and Associated Artifacts from Plainview, Texas. Bulletin of the Geological Society of America 58:927-954.

Kuhl, Frank P., and Charles R. Giardina. 1982 Elliptic Fourier features of a closed contour. Computer Graphics and Image Processing 18(3):236-258.

Kunz, Michael L. and Richard E. Reanier 1994 Paleoindians in Beringia: Evidence from Arctic Alaska. Science. 263(5147):660-662. 1995 The Mesa Site: A Paleoindian Hunting Lookout in Arctic Alaska. Arctic Anthropology 32(1):5-30. 1996 Mesa Site, Iteriak Creek. In American Beginnings: The Prehistory and Paleoecology of Beringia, edited by Frederick H. West, pp. 497-504. University of Chicago Press, Chicago.

Kunz, Michael L., Michael R. Bever, and Constance Adkins 2003 The Mesa site: Paleoindians above the Arctic Circle. US Department of the Interior, Bureau of Land Management, Alaska State Office, 2003.

Larter, N. C., A. R. E. Sinclair, T. Ellsworth, J. Nishi, and C. C. Gates. 2000 Dynamics of reintroduction in an indigenous large ungulate: the wood bison of northern Canada. Animal Conservation Forum, 3(4):299-309.

190

Lestrel, Pete E. 1989 Method for Analyzing Complex Two-Dimensional Forms: Elliptical Fourier Functions. American Journal of Human Biology 1:149-164. 2008 Introduction and overview of Fourier descriptors. In Fourier Descriptors and Their Applications in Biology, edited by Pete E. Lestrel, pp. 22-44. Cambridge University Press, New York.

Lipo, Carl P., and Mark Madsen 2001 Neutrality, “Style,” and Drift: Building Methods for Studying Cultural Transmission in the Archaeological Record. In Style and Function: Conceptual Issues in Evolutionary Archaeology, edited by Teresa D. Hunt and Gordon F. M. Rakita, pp. 91-118. Bergin & Garvey, Westport.

Lowell, Kim 1991 Utilizing discriminant function analysis with a geographical information system to model ecological succession spatially. International Journal of Geographical Information Systems 5(2):175-191.

Lycett, Stephen J., and Noreen von Cramon-Taubadel. 2015 Toward a “quantitative genetic” approach to lithic variation. Journal of Archaeological Method and Theory 22(2):646-675.

Lyman, R. Lee 2001 Culture Historical and Biological Approaches to Identifying Homologous Traits. In Style and Function: Conceptual Issues in Evolutionary Archaeology, edited by Teresa D. Hunt and Gordon F. M. Rakita, pp. 69-89. Bergin & Garvey, Westport.

Lyman, R. Lee, Todd L. VanPool, and Michael J. O’Brien 2008 Variation in North American Dart Points and Arrow Points When One or Both are Present. Journal of Archaeological Science 35: 2805–2812.

MacDonald, Douglas H. 1999 Modeling Folsom mobility, mating strategies, and technological organization in the northern Plains. The Plains Anthropologist 44(168): 141-161.

Mann, Daniel H., Richard E. Reanier, Dorothy M. Peteet, Michael L. Kunz, and Mark Johnson 2001 Environmental change and arctic Paleoindians. Arctic Anthropology 38(2):119-138.

Mann, Daniel H., Dorothy M Peteet, Richard E. Reanier, and Michael L. Kunz 2002 Responses of an arctic landscape to Lateglacial and early Holocene climatic changes: the importance of moisture. Quaternary Science Reviews 21:997-1021.

Mann, Daniel H., Pamela Groves, Richard E. Reanier, and Michael L. Kunz 2010 Floodplains, permafrost, cottonwood trees, and peat: what happened the last time climate warmed suddenly in arctic Alaska? Quaternary Science Reviews 29:3812-3830.

191

Mann, Daniel H., Pamela Groves, Michael L. Kunz, Richard E. Reanier, and Benjamin V. Gaglioti 2013 Ice-age megafauna in Arctic Alaska: extinction, invasion, survival. Quaternary Science Reviews 70:91-108.

Martin, Paul S. 1967 Prehistoric overkill. In Pleistocene Extinctions: The Search for a Cause, edited by Paul S Martin and Herbert E, Wright Jr., pp. 75–120. Yale University Press, New Haven.

Mason, Owen K., Peter M. Bowers, and David M. Hopkins 2001 The early Holocene Milankovitch thermal maximum and humans: adverse conditions for the Denali complex of eastern Beringia. Quaternary Science Reviews 20:525-548.

Meltzer, David J. 2005 The Origin of Food-Producing Economies in the Americas: The Late Paleoindian Period. In The Human Past, edited by Christopher Scarre, pp 307-313. Thames & Hudson, New York. 2009 First Peoples in a New World: Colonizing Ice Age America. University of California Press, Berkeley. 2015 Pleistocene overkill and North American mammalian extinctions. Annual Review of Anthropology 44:33-53.

Mesoudi, Alex 2011 Cultural Evolution: How Darwinian Theory Can Explain Human Culture and Synthesize the Social Sciences. The University of Chicago Press, Chicago.

Mesoudi, Alex, Andrew Whiten, and Kevin N. Laland 2004 Perspective: Is Human Culture Evolution Darwinian? Evidence Reviewed from the Perspective of The Origin of Species. Evolution 58(1):1-11.

Neiman, Fraser D. 1995 Stylistic variation in evolutionary perspective: inferences from decorative diversity and interassemblage distance in Illinois Woodland ceramic assemblages. American Antiquity 60(1):7- 36.

Nelson, Robert E. and L. David Carter 1987 Paleoenvironmental analysis of insects and extralimital Populus from an early Holocene site on the Arctic Slope of Alaska, USA. Arctic and Alpine Research 230-241.

O'Brien, Michael John, and W. Raymond Wood 1998 The Prehistory of Missouri. University of Missouri Press, Columbia.

O’Brien, Michael J., John Darwent, and R. Lee Lyman 2001 Cladistics Is Useful for Reconstructing Archaeological Phylogenies: Paleoindian Points from the Southeastern United States. Journal of Archaeological Science 28:1115-1136.

192

O’Brien, Michael J., Matthew T. Boulanger, Briggs Buchanan, Mark Collard, R. Lee Lyman, and John Darwent. 2014 Innovation and cultural transmission in the American Paleolithic: phylogenetic analysis of eastern Paleoindian projectile-point classes. Journal of Anthropological Archaeology 34:100- 119.

Parmesan, Camille 2006 Ecological and evolutionary responses to recent climate change. Annual Review of Ecology, Evolution, and Systematics 37:637-69.

Pearson, F. J., E. Mott Davis, Murray A. Tamers, and Robert W. Johnstone. 1965 University of Texas radiocarbon dates III. Radiocarbon 7:296-314.

Peteet, Dorothy 1995 Global Younger Dryas? Quaternary International 28:93-104.

Peteet, Dorothy, Anthony Del Genio, and Kenneth K-W Lo 1997 Sensitivity of northern hemisphere air temperatures and snow expansion to North Pacific sea surface temperatures in the Goddard Institute for Space Studies general circulation model. Journal of Geophysical Research: Atmospheres 102(D20):23,781-23,791.

Potter, Ben 2011 Late Pleistocene and Early Holocene Assemblage in Central Alaska. In From the Yenisei to the Yukon: Interpreting Lithic Assemblage Variability in Late Pleistocene/Early Holocene Beringia edited by Ted Goebel and Ian Buvit, pp. 215-233. Texas A&M University Press, College Station.

Potter, Ben A., Charles E. Holmes, and David R. Yesner 2013 Technology and Economy among the Earliest Prehistoric Foragers in Interior Eastern Beringia. In Paleoamerican Odyssey edited by Kelly E. Graf, Caroline V. Ketron, and Michael R. Waters, pp. 81-103. Texas A&M University Press, College Station.

Powers, William R., and John F. Hoffecker 1989 Late Pleistocene settlement in the Nenana Valley, central Alaska. American Antiquity 54(2):263–87.

PRISM Climate Group; National Park Service. Alaska PRISM Climate Maps 1971-2000. Available from: https://irma.nps.gov/DataStore/Reference/Profile/2193235.

Putt, Shelby S., Alexander D. Woods, and Robert G. Franciscus. 2014 The role of verbal interaction during experimental bifacial stone tool manufacture. Lithic Technology 39(2):96-112.

R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna; 2015. Available from: https://www.R-project.org/.

193

Rasic, Jeffrey T. 2008 Paleoalaskan Adaptive Strategies Viewed from Northwestern Alaska. Unpublished Ph. D. dissertation, Department of Anthropology, Washington State University, Pullman. 2011 Functional Variability in the Late Pleistocene Archaeological Record of Eastern Beringia: A Model of Late Pleistocene Land Use and Technology from Northwest Alaska. In From the Yenisei to the Yukon: Interpreting Lithic Assemblage Variability in Late Pleistocene/Early Holocene Beringia edited by Ted Goebel and Ian Buvit, pp. 128-164. Texas A&M University Press, College Station.

Reanier, Richard E. 1996 Putu and Bedwell In American Beginnings: The Prehistory and Paleoecology of Beringia, edited by Frederick H. West, pp. 505-510. University of Chicago Press, Chicago.

Rohlf, James F., and James W. Archie 1983 A Comparison of Fourier Methods for the Description of Wing Shape in Mosquitos (Diptera: Culicidae). Systematic Zoology 33(3):302-317.

Root, Terry L., Jeff T. Price, Kimberely R. Hall, Stephen H. Schneider, Cynthia Rosenzweig, and J. Alan Pounds 2003 Fingerprints of global warming on wild animals and plants. Nature 421(6918):57-60.

Seeman, Mark F. 1994 Intercluster lithic patterning at Nobles Pond: a case for “disembedded” procurement among early Paleoindian societies. American Antiquity 59(2):273-288.

Sellards, Elias H. 1955 Fossil bison and associated artifacts from Milnesand, New Mexico. American Antiquity 20(4):336-344.

Sellards, Elias H., Glen L. Evans, and Grayson E. Meade. 1947 Bison, Fossil, and Associated Artifacts from Plainview, Texas. Bulletin of the Geological Society of America 58:927-954.

Shapiro, Beth, Alexei J. Drummond, Andrew Rambaut, Michael C. Wilson, Paul E. Matheus, Andrei V. Sher, Oliver G. Pybus, M. Thomas Gilbert, Ian Barnes, Jonas Binladen, Eske Willerslev, Anders J. Hansen, Gennady f. Baryshnikov, James A burns, Sergei Davydov, Jonathon C. Driver, Duane G. Froese, C. Richard Harington, Grant Keddie, Pavel Kosintev, Michael L. Kunz, Larry D. Martin, Robert O. Stephenson, John Storer, Richard Tedford, Sergei Zimov, and Alan Cooper 2004 Rise and fall of the Beringian steppe bison. Science 306(5701):1561-1565.

Shippee, J. Mett 1948 Nebo Hill, a lithic complex in western Missouri. American Antiquity 14(1):29-32.

194

Sholts, Sabrina B., Dennis J. Stanford, Louise M. Flores, and Sebastian K.T.S. Warmlander 2012 Flake scar patterns of Clovis points analyzed with a new digital morphometrics approach: evidence for direct transmission of technological knowledge across early North America. Journal of Archaeological Science 39:3018-3026.

Shott, Michael J. 1997 Stones and Shafts Redux: The Metric Discrimination of Chipped-Stone Dart and Arrow Points. American Antiquity 62(1):86-101. 2014 Digitizing archaeology: a subtle revolution in analysis. World Archaeology 46(1):1-9.

Shott, Michael J., and Brian W. Trail. 2010 Exploring new approaches to lithic analysis: laser scanning and geometric morphometrics. Lithic technology 35(2):195-220. 2012 New Developments in Lithic Analysis. SAA Archaeological Record 12(3):12-18.

Slobodin, Sergei 2011 Late Pleistocene and early Holocene cultures of Beringia: The general and the specific. In From the Yenisei to the Yukon: Interpreting Lithic Assemblage Variability in Late Pleistocene/Early Holocene Beringia, edited by Ted Goebel and Ian Buvit, pp.91–116. Texas A&M University Press, College Station.

Smith, Heather L., Jeffery T. Rasic, and Ted Goebel 2013 Biface traditions of northern Alaska and their role in the peopling of the Americas. In Paleoamerican odyssey edited by Kelly E. Graf, Caroline V. Ketron, and Michael R. Waters, pp 105-123. Texas A&M University, College Station.

Smith, Heather L., and Ted Goebel Origins and spread of fluted-point technology in the Canadian Ice-Free Corridor and eastern Beringia. Proceedings of the National Academy of Sciences: 201800312.

Srivastava, Santosh, Maya R. Gupta, and Béla A. Frigyik 2007 Bayesian Quadratic Discriminant Analysis. Journal of Machine Learning Research 8(6):1277-1305.

Stout, Dietrich. 2005 The Social and Cultural Context of Stone-knapping Skill Acquisition. In Stone Knapping: The Necessary Conditions for a Uniquely Hominin Behaviour edited by Valentine Roux, and Blandine Bril, pp. 331-340. McDonald Institute, Cambridge.

Surovell, Todd A. 2009 Toward a behavioral ecology of lithic technology: cases from Paleoindian archaeology. University of Arizona Press, Tucson.

195

Thomas, David Hurst. 1978 Arrowheads and Atlatl Darts: How the Stones Got the Shaft. American Antiquity 43(3):461-472.

Thulman, David K. 2012 Discriminating Paleoindian point types from Florida using landmark geometric morphometrics. Journal of Archaeological Science 39:1599-1607.

Todd, Lawrence C. 1987 Analysis of kill-butchery bonebeds and interpretation of Paleoindian hunting. In The evolution of human hunting, edited by Matthew H. Nitecki and Doris V. Nitecki, pp.225-266. Plenum Press, New York. 1991 Seasonality Studies and Paleoindian Subsistence Strategies. In Human Predators and Prey Mortality, edited by Mary C. Stiner, pp. 215-238. Westview Press, Boulder.

Todd, Lawrence C., David J. Rapson, and Jack L. Hofmanand 1996 Dentition Studies of the Mill Iron and Other Early Paleoindian Bison Bonebed Sites. In The Mill Iron Site edited by George C. Frison, pp. 145-176. University of New Mexico Press, Albuquerque.

VanPool, Todd L. 2001 Style, Function, and Variation: Identifying the Evolutionary Importance of Traits in the Archaeological Record. In Style and Function: Conceptual Issues in Evolutionary Archaeology, edited by Teresa D. Hunt and Gordon F. M. Rakita, pp. 119-140. Bergin & Garvey, Westport.

VanPool, Todd L., Craig T. Palmer, and Christine S. VanPool. 2008 Horned serpents, tradition, and the tapestry of culture. In Cultural Transmission and Archaeology: Some Fundamental Issues, edited by Michael J. O’Brien, pp. 77-90. Society for American Archaeology, Washington DC.

VanPool, Todd L., Michael J. O’Brien, and R. Lee Lyman 2015 Innovation and natural selection in Paleoindian projectile points from the American Southwest. In Lithic Technological Systems and Evolutionary Theory edited by Nathan Goodale and William Andrefsky, Jr., pp 61-81. Cambridge University Press, Cambridge.

Warnica, James M., and Ted Williamson. 1968 The Milnesand Site. Revisited. American Antiquity 33(1):16-24

Waters, Michael R. and Thomas W. Stafford 2014 Redating the Mill Iron site, Montana: a reexamination of Goshen complex chronology. American Antiquity 79(3):541-548.

West, Frederick H. 1967 The Donnelly Ridge Site and the Definition of an Early Core and Blade Complex in Central Alaska. American Antiquity 32(3):360-382.

196

Wheat, Joe Ben, Harold E. Malde, and Estella B. Leopold 1972 The Olsen-Chubbuck site: a paleo-Indian bison kill. Memoirs of the Society for American Archaeology 26:i-180. Whiten, Andrew, Robert A. Hinde, Kevin N. Laland, and Christopher B. Stringer 2011 Culture Evolves. Philosophical Transactions of the Royal Society B 366:938-948.

Wiessner, Polly 1983 Style and Social Information in Kalahari San Projectile Points. American Antiquity 48(2):253-276.

Williams, Justin P., and William Andrefsky, Jr. 2011 Debitage variability among multiple flint knappers. Journal of Archaeological Science 38(4):865-872.

Williams, Justin P., Andrew I. Duff, and William Andrefsky, Jr. 2013 Debitage Stylistic Variability at Cox Ranch Pueblo. Lithic Technology 38(1):3-16.

Wilson, Michael C., Leonard V. Hills, and Beth Shapiro 2008 Late Pleistocene northward-dispersing Bison antiquus from the Bighill Creek Formation, Gallelli gravel pit, Alberta, Canada, and the fate of Bison occidentalis. Canadian Journal of Earth Sciences 45(7):827-859.

Winton, Vicky 2005 An Investigation of Knapping-skill Development in the Manufacture of Paleolithic Handaxes. In Stone Knapping: The Necessary Conditions for a Uniquely Hominin Behaviour edited by Valentine Roux, and Blandine Bril, pp. 109-118. McDonald Institute, Cambridge.

Wygal, Brian T. 2011 The Microblade/Non-Microblade Dichotomy: Climate Implications, Toolkit Variability, and the Role of Tiny Tools in Eastern Beringia. In From the Yenisei to the Yukon: Interpreting Lithic Assemblage Variability in Late Pleistocene/Early Holocene Beringia edited by Ted Goebel and Ian Buvit, pp. 234-254. Texas A&M University Press, College Station.

Yesner, David R. and Georges Pearson 2002 Microblades and Migrations: Ethnic and Economic Models in the Peopling of the Americas. In Thinking Small: Global Perspectives on Microlithization edited by Robert G. Elston and Steven L. Kuhn, pp 133-161. Archaeological Papers of the American Anthropological Association number 12, Arlington.

Yu, Zicheng and H. E. Wright 2001 Response of interior North America to abrupt climate oscillations in the North Atlantic region during the last deglaciation. Earth-Science Reviews 52(4):333-369.

197

APPENDIX A: PROJECTILE POINT INFORMATION

Powder Pen (PP)/ Dissertation Collection Projectile Spray (S)/ Number Site Number Point Type Talc (T) Repository DOI 10.5281/zenodo. Point 0001 Mill Iron 24CT30-266 Goshen T BCC 1210714 10.5281/zenodo. Point 0002 Mill Iron 24CT30-268 Goshen T BCC 1210716 10.5281/zenodo. Point 0003 Mill Iron 24CT30-269 Goshen T BCC 1210718 10.5281/zenodo. Point 0004 Mill Iron 24CT30-278,279 Goshen T BCC 1210720 10.5281/zenodo. Point 0005 Mill Iron 24CT30-284 Goshen T BCC 1210722 10.5281/zenodo. Point 0006 Mill Iron 24CT30-288 Goshen - BCC 1210724 10.5281/zenodo. Point 0007 Mill Iron 24CT30-289 Goshen T BCC 1210726 10.5281/zenodo. Point 0008 Mill Iron 24CT30-1508 Goshen T BCC 1210728 10.5281/zenodo. Point 0009 Mill Iron 24CT30-1510 Goshen T BCC 1210730 10.5281/zenodo. Point 0010 Mill Iron 24CT30-1582 Goshen T BCC 1210732 10.5281/zenodo. Point 0011 Mill Iron 24CT30-1583 Goshen T BCC 1210734 10.5281/zenodo. Point 0012 Mill Iron 24CT30-1586 Goshen T BCC 1210736 10.5281/zenodo. Point 0013 Mill Iron 24CT30-1587 Goshen T BCC 1210738 10.5281/zenodo. Point 0014 Mill Iron 24CT30-1588 Goshen T BCC 1210740 10.5281/zenodo. Point 0015 Mill Iron 24CT30-1607 Goshen T BCC 1210742 10.5281/zenodo. Point 0016 Mill Iron 24CT30-1617 Goshen T BCC 1210744 10.5281/zenodo. Point 0018 Milnesand TMM 1146-2 Milnesand S TARL 1210614 10.5281/zenodo. Point 0019 Milnesand TMM 1146-4 Milnesand S TARL 1210616 10.5281/zenodo. Point 0020 Milnesand TMM 1146-9 Milnesand S TARL 1210618 10.5281/zenodo. Point 0021 Milnesand TMM 1146-10 Milnesand S TARL 1210620 10.5281/zenodo. Point 0022 Milnesand TMM 1146-11 Milnesand S TARL 1210622 10.5281/zenodo. Point 0023 Milnesand TMM 1146-12 Milnesand S TARL 1210624 10.5281/zenodo. Point 0024 Milnesand TMM 1146-13 Milnesand S TARL 1210626 10.5281/zenodo. Point 0025 Milnesand TMM 1146-16 Milnesand S TARL 1210628 10.5281/zenodo. Point 0026 Milnesand TMM 1146-23 Milnesand S TARL 1210630 10.5281/zenodo. Point 0027 Milnesand TMM 1146-39 Milnesand S TARL 1210632 10.5281/zenodo. Point 0028 Milnesand TMM 1146-55 Milnesand S TARL 1210634

199

10.5281/zenodo. Point 0029 Milnesand TMM 1146-63 Milnesand S TARL 1210636 10.5281/zenodo. Point 0030 Plainview 41HA1-1 Plainview S TARL 1210642 10.5281/zenodo. Point 0031 Plainview 41HA1-2 Plainview S TARL 1210644 10.5281/zenodo. Point 0032 Plainview 41HA1-5 Plainview S TARL 1210646 10.5281/zenodo. Point 0033 Plainview 41HA1-8 Plainview S TARL 1210648 10.5281/zenodo. Point 0034 Plainview 41HA1-9 Plainview PP TARL 1210650 10.5281/zenodo. Point 0035 Plainview 41HA1-10 Plainview PP TARL 1210652 10.5281/zenodo. Point 0036 Plainview 41HA1-19 Plainview S TARL 1210654 10.5281/zenodo. Point 0037 Plainview 41HA1-23 Plainview S TARL 1210656 10.5281/zenodo. Point 0038 Plainview 41HA1-29 Plainview PP TARL 1210658 10.5281/zenodo. Point 0039 Plainview 41HA1-X Plainview PP TARL 1210660 10.5281/zenodo. Point 0040 Bonfire Shelter 41VV218-15838 Plainview S TARL 1210663 10.5281/zenodo. Point 0041 Bonfire Shelter 41VV218-15840 Plainview S TARL 1210669 10.5281/zenodo. Point 0042 Bonfire Shelter 41VV218-15841 Plainview - TARL 1210671 10.5281/zenodo. Point 0043 Bonfire Shelter 41VV218-15843 Plainview S TARL 1210673 Nebo 10.5281/zenodo. Point 0049 Gerald Shelton 1 GS-66-5 Hill/Sedalia S SMUQL 1210676 Nebo 10.5281/zenodo. Point 0050 Gerald Shelton 1 GS-66-6 Hill/Sedalia S SMUQL 1210678 Nebo 10.5281/zenodo. Point 0051 Gerald Shelton 1 GS-66-7 Hill/Sedalia S SMUQL 1210680 Nebo 10.5281/zenodo. Point 0052 Gerald Shelton 1 GS-66-8 Hill/Sedalia S SMUQL 1210682 Nebo 10.5281/zenodo. Point 0053 Gerald Shelton 1 GS-66-9 Hill/Sedalia S SMUQL 1210684 Nebo 10.5281/zenodo. Point 0055 Gerald Shelton 2 GS-67-23 Hill/Sedalia PP SMUQL 1210686 Nebo 10.5281/zenodo. Point 0056 Gerald Shelton 2 GS-67-33 Hill/Sedalia PP SMUQL 1210688 Nebo 10.5281/zenodo. Point 0057 Gerald Shelton 2 GS-67-34 Hill/Sedalia S SMUQL 1210690 Nebo 10.5281/zenodo. Point 0058 Gerald Shelton 2 GS-67-37 Hill/Sedalia PP SMUQL 1210692 Nebo 10.5281/zenodo. Point 0059 Gerald Shelton 2 GS-67-38 Hill/Sedalia PP SMUQL 1210694 Nebo 10.5281/zenodo. Point 0060 Gerald Shelton 2 GS-67-39 Hill/Sedalia PP SMUQL 1210698 Nebo 10.5281/zenodo. Point 0061 Gerald Shelton 2 GS-67-41 Hill/Sedalia PP SMUQL 1210700 Nebo 10.5281/zenodo. Point 0062 Gerald Shelton 2 GS-67-43 Hill/Sedalia S SMUQL 1210702 Nebo 10.5281/zenodo. Point 0063 Gerald Shelton 2 GS-67-45 Hill/Sedalia S SMUQL 1210704

200

10.5281/zenodo. Point 0101 Caribou Crossing 1 MIS376-18710 Sluiceway S ARCC 1210867 10.5281/zenodo. Point 0102 Caribou Crossing 1 MIS376-18716 Sluiceway S ARCC 1210869 10.5281/zenodo. Point 0103 Caribou Crossing 1 MIS376-18717 Sluiceway S ARCC 1210871 10.5281/zenodo. Point 0104 Caribou Crossing 1 MIS376-18719 Sluiceway S ARCC 1210873 10.5281/zenodo. Point 0105 Caribou Crossing 1 MIS376-18732 Sluiceway S ARCC 1210875 10.5281/zenodo. Point 0106 Caribou Crossing 1 MIS376-18734 Sluiceway S ARCC 1210877 10.5281/zenodo. Point 0107 Caribou Crossing 1 MIS376-18735 Sluiceway S ARCC 1210879 10.5281/zenodo. Point 0108 Caribou Crossing 1 MIS376-18750 Sluiceway S ARCC 1210881 10.5281/zenodo. Point 0109 Caribou Crossing 1 MIS376-18754 Sluiceway S ARCC 1210883 10.5281/zenodo. Point 0110 Caribou Crossing 1 MIS376-18764 Sluiceway S ARCC 1210885 10.5281/zenodo. Point 0111 Caribou Crossing 1 MIS376-18772 Sluiceway S ARCC 1210887 10.5281/zenodo. Point 0112 Caribou Crossing 1 MIS376-18793 Sluiceway S ARCC 1210889 10.5281/zenodo. Point 0113 Caribou Crossing 1 MIS376-18795 Sluiceway S ARCC 1210891 10.5281/zenodo. Point 0114 Caribou Crossing 1 MIS376-18805 Sluiceway S ARCC 1210893 10.5281/zenodo. Point 0115 Caribou Crossing 1 MIS376-18806 Sluiceway S ARCC 1210895 10.5281/zenodo. Point 0116 Caribou Crossing 1 MIS376-18810 Sluiceway S ARCC 1210897 10.5281/zenodo. Point 0117 Caribou Crossing 1 MIS376-18811 Sluiceway S ARCC 1210899 10.5281/zenodo. Point 0118 Caribou Crossing 1 MIS376-18814 Sluiceway S ARCC 1210902 10.5281/zenodo. Point 0119 Caribou Crossing 1 MIS376-18819 Sluiceway S ARCC 1210904 10.5281/zenodo. Point 0120 Caribou Crossing 1 MIS376-18829 Sluiceway S ARCC 1210906 10.5281/zenodo. Point 0121 Caribou Crossing 1 MIS376-18842 Sluiceway S ARCC 1210908 10.5281/zenodo. Point 0122 Caribou Crossing 1 MIS376-18849 Sluiceway S ARCC 1210910 10.5281/zenodo. Point 0123 Caribou Crossing 1 MIS376-18859 Sluiceway S ARCC 1210912 10.5281/zenodo. Point 0124 Caribou Crossing 1 MIS376-18868 Sluiceway S ARCC 1210914 10.5281/zenodo. Point 0125 Caribou Crossing 1 MIS376-18872 Sluiceway S ARCC 1210916 10.5281/zenodo. Point 0126 Caribou Crossing 1 MIS376-18888 Sluiceway S ARCC 1210918 10.5281/zenodo. Point 0127 Caribou Crossing 1 MIS376-18889 Sluiceway S ARCC 1210920 10.5281/zenodo. Point 0128 Caribou Crossing 1 MIS376-18890 Sluiceway S ARCC 1210922 10.5281/zenodo. Point 0129 Caribou Crossing 1 MIS376-18893 Sluiceway S ARCC 1210924

201

10.5281/zenodo. Point 0130 Caribou Crossing 1 MIS376-18899 Sluiceway S ARCC 1210926 10.5281/zenodo. Point 0131 Caribou Crossing 1 MIS376-18900 Sluiceway S ARCC 1210928 10.5281/zenodo. Point 0132 Caribou Crossing 1 MIS376-18901 Sluiceway S ARCC 1210930 10.5281/zenodo. Point 0133 Caribou Crossing 1 MIS376-18902 Sluiceway S ARCC 1210932 10.5281/zenodo. Point 0134 Caribou Crossing 1 MIS376-18904 Sluiceway S ARCC 1210934 10.5281/zenodo. Point 0135 Caribou Crossing 1 MIS376-18913 Sluiceway S ARCC 1210936 10.5281/zenodo. Point 0136 Caribou Crossing 2 MIS377-19387 Sluiceway S ARCC 1210938 10.5281/zenodo. Point 0137 Caribou Crossing 2 MIS377-19388 Sluiceway S ARCC 1210942 10.5281/zenodo. Point 0138 Caribou Crossing 2 MIS377-19428 Sluiceway S ARCC 1210944 10.5281/zenodo. Point 0139 Caribou Crossing 2 MIS377-19442 Sluiceway S ARCC 1210946 10.5281/zenodo. Point 0140 Caribou Crossing 2 MIS377-19449 Sluiceway S ARCC 1210949 10.5281/zenodo. Point 0141 Caribou Crossing 2 MIS377-19452 Sluiceway S ARCC 1210951 10.5281/zenodo. Point 0142 Caribou Crossing 2 MIS377-19486 Sluiceway S ARCC 1210953 10.5281/zenodo. Point 0143 Caribou Crossing 2 MIS377-19494 Sluiceway S ARCC 1210955 10.5281/zenodo. Point 0144 Caribou Crossing 2 MIS377-19503 Sluiceway S ARCC 1210957 10.5281/zenodo. Point 0145 Caribou Crossing 2 MIS377-19519 Sluiceway S ARCC 1210959 10.5281/zenodo. Point 0146 Caribou Crossing 2 MIS377-19523 Sluiceway S ARCC 1210961 10.5281/zenodo. Point 0147 Caribou Crossing 2 MIS377-19526 Sluiceway S ARCC 1210963 10.5281/zenodo. Point 0148 Caribou Crossing 2 MIS377-19527 Sluiceway S ARCC 1210965 10.5281/zenodo. Point 0149 Tuluaq DEL360-4036 Sluiceway S ARCC 1210969 10.5281/zenodo. Point 0150 Tuluaq DEL360-4600 Sluiceway S ARCC 1210971 10.5281/zenodo. Point 0151 Tuluaq DEL360-4642 Sluiceway S ARCC 1210973 10.5281/zenodo. Point 0152 Tuluaq DEL360-4648 Sluiceway S ARCC 1210975 10.5281/zenodo. Point 0153 Tuluaq DEL360-4783 Sluiceway S ARCC 1210977 10.5281/zenodo. Point 0154 Tuluaq DEL360-4821 Sluiceway S ARCC 1210979 10.5281/zenodo. Point 0155 Tuluaq DEL360-4940 Sluiceway S ARCC 1210983 10.5281/zenodo. Point 0156 Tuluaq DEL360-13687 Sluiceway S ARCC 1210985 10.5281/zenodo. Point 0157 Nat Pass MIS495-18626 Sluiceway S ARCC 1210989 10.5281/zenodo. Point 0158 Upper Kelly DEL391-19718 Sluiceway S ARCC 1210991 Point 0159 Spein Mountain 49BTH063-14 Mesa PP WSU Contact WSU

202

Point 0160 Spein Mountain 49BTH063-17 Mesa PP WSU Contact WSU Point 0161 Spein Mountain 49BTH063-19 Mesa PP WSU Contact WSU Point 0162 Spein Mountain 49BTH063-24 Mesa PP WSU Contact WSU Point 0163 Spein Mountain 49BTH063-35 Mesa PP WSU Contact WSU Point 0164 Spein Mountain 49BTH063-35-86 Mesa PP WSU Contact WSU Point 0165 Spein Mountain 49BTH063-72 Mesa PP WSU Contact WSU Point 0166 Spein Mountain 49BTH063-75 Mesa PP WSU Contact WSU Point 0167 Mesa KIR102-98-759 Mesa - UAM Contact UAM Point 0168 Mesa UA78-76-2 Mesa - UAM Contact UAM Point 0169 Mesa UA78-76-10 Mesa - UAM Contact UAM Point 0170 Mesa UA78-76-20 Mesa - UAM Contact UAM Point 0171 Mesa UA79-160-10 Mesa - UAM Contact UAM Point 0172 Mesa UA79-160-59 Mesa - UAM Contact UAM Point 0173 Mesa UA79-160-62 Mesa - UAM Contact UAM Point 0174 Mesa UA79-160-70 Mesa - UAM Contact UAM Point 0175 Mesa UA79-160-73 Mesa - UAM Contact UAM Point 0176 Mesa UA79-160-90 Mesa - UAM Contact UAM Point 0177 Mesa UA81-56-24 Mesa - UAM Contact UAM Point 0178 Mesa UA81-56-79 Mesa - UAM Contact UAM Point 0179 Mesa UA81-56-147 Mesa - UAM Contact UAM Point 0180 Mesa UA81-56-163 Mesa - UAM Contact UAM Point 0181 Mesa UA81-56-264 Mesa - UAM Contact UAM Point 0182 Mesa UA81-56-296 Mesa - UAM Contact UAM Point 0183 Mesa UA81-56-344 Mesa - UAM Contact UAM Point 0184 Mesa UA81-56-480 Mesa - UAM Contact UAM Point 0185 Mesa UA81-56-573 Mesa - UAM Contact UAM Point 0186 Mesa UA81-56-973 Mesa - UAM Contact UAM Point 0187 Mesa UA81-56-1084 Mesa - UAM Contact UAM Point 0188 Mesa UA81-56-1456 Mesa - UAM Contact UAM Point 0189 Mesa UA81-56-1473 Mesa - UAM Contact UAM Point 0190 Mesa UA81-56-1485 Mesa - UAM Contact UAM Point 0191 Mesa UA81-56-2185 Mesa - UAM Contact UAM Point 0192 Mesa UA81-56-4566 Mesa - UAM Contact UAM Point 0193 Mesa UA81-56-4770 Mesa - UAM Contact UAM Point 0194 Mesa UA81-56-5287 Mesa - UAM Contact UAM Point 0195 Mesa UA98-77-1 Mesa - UAM Contact UAM Point 0196 Putu/Bedwell UA70-84-88-1 Mesa - UAM Contact UAM Point 0197 Putu/Bedwell UA70-84-238 Mesa - UAM Contact UAM Point 0198 Putu/Bedwell UA70-84-288 Mesa - UAM Contact UAM

203

Point 0199 Putu/Bedwell UA70-84-308 Mesa - UAM Contact UAM Point 0200 Putu/Bedwell UA70-84-309 Mesa - UAM Contact UAM Point 0201 Putu/Bedwell UA70-84-969 Mesa - UAM Contact UAM Point 0202 Red Dog UA98-60-2009 Sluiceway S UAM Contact UAM Point 0203 Red Dog UA98-60-2014 Sluiceway S UAM Contact UAM Point 0204 Red Dog UA98-60-2015 Sluiceway S UAM Contact UAM Point 0205 Red Dog UA98-60-2016 Sluiceway S UAM Contact UAM Point 0206 Hilltop UA70-126-57 Mesa - UAM Contact UAM Point 0207 Hilltop UA70-126-376 Mesa - UAM Contact UAM Point 0208 Hilltop UA73-126-60 Mesa - UAM Contact UAM

Repository Key

ARCC, National Park Service Alaska Regional Curatorial Center BCC, Bureau of Land Management Billing Curation Center SMUQL, Quest Laboratory, Department of Anthropology, Southern Methodist University TARL, Texas Archaeological Research Laboratory UAMN, University of Alaska Museum of the North, Archaeology Department WSU, Dr. Robert Ackerman, Washington State University, Department of Anthropology

204

APPENDIX B: VEGETATION CLASSES

USGS Vegetation ID QDA Prediction Map ID Description 0 NA Ocean Water 1 NA Water 2 1 Glaciers & Snow 3 2 Alpine Tundra & Barrens 4 3 Dwarf Shrub Tundra 5 4 Tussock Sedge/Dwarf Shrub Tundra 6 5 Moist Herbaceous/Shrub Tundra 7 6 Wet Sedge Tundra 8 7 Low Shrub/Lichen Tundra 9 8 Low & Dwarf Shrub 10 9 Tall Shrub 11 10 Closed Broadleaf & Closed Mixed Forest 12 11 Closed Mixed Forest 13 12 Closed Spruce Forest 14 13 Spruce Woodland/Shrub 15 14 Open Spruce Forest/Shrub/Bog Mosaic 16 15 Spruce & Broadleaf Forest 17 16 Open & Closed Spruce Forest 18 17 Open Spruce & Closed Mixed Forest Mosaic 19 18 Closed Spruce & Hemlock Forest 20 NA 1991 Fires 21 NA 1990 Fires & Gravel Bars 22 NA Canada/Russia 23 19 Tall & Low Shrub

206

APPENDIX C: PALEOECOLOGICAL PROJECTION RESULTS

(14) (17) (4) (5) (7) (10) Open Open (2) Tussock Moist (6) Low Closed Spruce (16) Spruce & (18) Alpine (3) Sedge/ Herb/ Wet Shrub/ (8) Broadleaf (11) (12) (13) Forest/ (15) Open & Closed Closed (19) (1) Tundra Dwarf Dwarf Shrub Sedge Lichen Low & (9) & Closed Closed Closed Spruce Shrub/ Spruce & Closed Mixed Spruce & Tall & Proxy Estimates (# Glaciers & Shrub Shrub Tundr Tundr Tundr Dwarf Tall Mixed Mixed Spruce Woodlan Bog Broadleaf Spruce Forest Hemlock Low Cells) & Snow Barrens Tundra Tundra a a a Shrub Shrub Forest Forest Forest d/ Shrub Mosaic Forest Forest Mosaic Forest Shrub Minus 15°C Minus 30% Precipitation 911 1281121 1219517 Minus 15°C Minus 20% Precipitation 1584 1655129 844836 Minus 15°C Minus 10% Precipitation 3127 2130474 367948 Minus 15°C Equal Precipitation 7907 2449259 44383 Minus 15°C Plus 10% Precipitation 21358 2468053 12138 Minus 15°C Plus 20% Precipitation 44136 2454824 2589 Minus 15°C Plus 30% Precipitation 82123 2418482 615 329 Minus 14°C Minus 30% Precipitation 1382 1174385 1325782 Minus 14°C Minus 20% Precipitation 2160 1559262 940127 Minus 14°C Minus 10% Precipitation 4185 2012755 484609 Minus 14°C Equal Precipitation 10407 2408491 82651 Minus 14°C Plus 10% Precipitation 27269 2456984 17296

208 Minus 14°C Plus 20% Precipitation 52954 2444689 3906

Minus 14°C Plus 30% Precipitation 101918 2396229 1024 2378 Minus 13°C Minus 30% Precipitation 2025 1075202 1424322 Minus 13°C Minus 20% Precipitation 2933 1450284 1048332 Minus 13°C Minus 10% Precipitation 5663 1871818 624068 Minus 13°C Equal Precipitation 13945 2335021 152583 Minus 13°C Plus 10% Precipitation 33791 2444215 23543 Minus 13°C Plus 20% Precipitation 63238 2430366 6080 1865 Minus 13°C Plus 30% Precipitation 131129 2362774 1553 6093 Minus 12°C Minus 30% Precipitation 2863 990371 1508315 Minus 12°C Minus 20% Precipitation 3968 1324499 1173082 Minus 12°C Minus 10% Precipitation 7782 1743981 749786 Minus 12°C Equal Precipitation 18786 2232685 250078 Minus 12°C Plus 10% Precipitation 41210 2424010 36329 Minus 12°C Plus 20% Precipitation 75008 2410510 9300 6706 25 Minus 12°C Plus 30% Precipitation 169098 2301132 2260 28758 301

Minus 11°C Minus 30% Precipitation 3949 915896 1581704 Minus 11°C Minus 20% Precipitation 5353 1186795 1309401 Minus 11°C Minus 10% Precipitation 10356 1613227 877966 Minus 11°C Equal Precipitation 24865 2105849 370835 Minus 11°C Plus 10% Precipitation 49854 2381235 68125 2320 15 Minus 11°C Plus 20% Precipitation 90727 2364990 14154 29541 2039 98 Minus 11°C Plus 30% Precipitation 208419 2210495 2982 77229 2230 194 Minus 10°C Minus 30% Precipitation 5408 850095 1646046 Minus 10°C Minus 20% Precipitation 7312 1065978 1428259 Minus 10°C Minus 10% Precipitation 13943 1459120 1028486 Minus 10°C Equal Precipitation 32120 1955541 513888 Minus 10°C Plus 10% Precipitation 60495 2291655 133777 13829 1625 168 Minus 10°C Plus 20% Precipitation 112243 2268011 27110 82904 10906 375 Minus 10°C Plus 30% Precipitation 247445 2090127 3924 148777 10769 507 Minus 9°C Minus

209 30% Precipitation 5833 1045465 1450251 Minus 9°C Minus 20% Precipitation 8137 1221386 1272026

Minus 9°C Minus 10% Precipitation 15195 1524478 961875 1 Minus 9°C Equal Precipitation 35496 1927994 537452 587 20 Minus 9°C Plus 10% Precipitation 67466 2168564 217482 45806 1764 467 Minus 9°C Plus 20% Precipitation 131758 2111567 41782 202339 13169 934 Minus 9°C Plus 30% Precipitation 273788 1972038 4673 244422 5455 1173 Minus 8°C Minus 30% Precipitation 7889 960940 1532719 1 1 Minus 8°C Minus 20% Precipitation 10986 1072580 1417981 1 Minus 8°C Minus 10% Precipitation 21142 1301046 1179307 54 Minus 8°C Equal Precipitation 44490 1650013 798460 7542 1044 Minus 8°C Plus 10% Precipitation 83077 1951970 341808 112486 10650 1558 Minus 8°C Plus 20% Precipitation 166439 1895097 100934 328 306676 29172 790 2113 Minus 8°C Plus 30% Precipitation 302863 1825123 38370 318822 14207 17 2147 Minus 7°C Minus 30% Precipitation 11253 887276 1603004 15 1 Minus 7°C Minus 20% Precipitation 15073 957738 1528614 17 107 Minus 7°C Minus 10% Precipitation 29280 1083057 1388290 11 911

Minus 7°C Equal Precipitation 55358 1352537 1069895 19642 85 4032 Minus 7°C Plus 10% Precipitation 103741 1623619 544553 1725 190718 25518 5306 6369 Minus 7°C Plus 20% Precipitation 203246 1639978 177051 9594 403190 58240 3726 6524 Minus 7°C Plus 30% Precipitation 322643 1646116 98748 268 404154 23402 108 6110 Minus 6°C Minus 30% Precipitation 16123 825384 1659862 48 132 Minus 6°C Minus 20% Precipitation 21394 862568 1616708 50 829 Minus 6°C Minus 10% Precipitation 38853 919463 1539210 45 3978 Minus 6°C Equal Precipitation 69377 1065855 1310033 391 36644 271 2352 16626 Minus 6°C Plus 10% Precipitation 130320 1153873 748456 3 83186 243079 49887 64435 15 28295 Minus 6°C Plus 20% Precipitation 235350 1177621 332093 87899 471391 111327 50672 1 35195 Minus 6°C Plus 30% Precipitation 334504 1351508 231756 3544 522865 21806 2419 1 33146 Minus 5°C Minus 30% Precipitation 24371 771828 1704244 108 998 Minus 5°C Minus 20% Precipitation 30956 775637 1691319 143 3494 Minus 5°C Minus 10% Precipitation 49630 787552 1643979 10 288 20090 Minus 5°C Equal

210 Precipitation 88442 785805 1405952 91 52075 61515 4332 29412 1 73924 Minus 5°C Plus 10% Precipitation 161418 785299 755636 141 132600 297192 102560 154515 1057 16 111115

Minus 5°C Plus 20% Precipitation 259496 714627 457246 81 116735 511686 4 179721 143863 808 22 117260 Minus 5°C Plus 30% Precipitation 335103 935570 534657 19 10360 542352 67 15907 18252 105 109157 Minus 4°C Minus 30% Precipitation 26230 640819 1831027 1636 1837 Minus 4°C Minus 20% Precipitation 33953 637379 1811379 4 4804 2 14028 Minus 4°C Minus 10% Precipitation 55768 627467 1716718 617 7762 21318 127 17 71755 Minus 4°C Equal Precipitation 103175 557131 1239224 2434 118263 173392 32503 81901 2428 474 88 190536 Minus 4°C Plus 10% Precipitation 183979 446512 582219 1452 182048 389203 1 315610 187502 13556 113 257 199097 Minus 4°C Plus 20% Precipitation 269299 413452 507130 4828 135782 501192 314 343477 120467 3201 230 1965 200212 Minus 4°C Plus 30% Precipitation 319055 535582 863026 1972 28881 528752 1606 27197 4093 12825 178560 Minus 3°C Minus 30% Precipitation 44576 608638 1834300 5643 98 8294 Minus 3°C Minus 20% Precipitation 49031 582829 1799353 1570 19811 170 48785 Minus 3°C Minus 10% Precipitation 75015 491584 1591105 16735 48686 69949 27 2574 345 596 204933 Minus 3°C Equal Precipitation 132367 366345 764351 43945 153160 275550 109 90939 271485 43456 5092 4000 350750 Minus 3°C Plus 10% Precipitation 210140 272155 366638 44258 236984 6 338523 85 5163 315760 325185 80599 3999 24608 277446 Minus 3°C Plus 20% Precipitation 257101 240799 531292 57185 143550 1 428391 313 18336 246289 191374 7403 2218 68354 308943

Minus 3°C Plus 30% Precipitation 269339 361103 1056392 10402 22978 457277 133 18908 16695 3794 42 78646 205840 Minus 2°C Minus 30% Precipitation 101669 455506 1883443 10 25442 10095 25384 Minus 2°C Minus 20% Precipitation 90551 409092 1780991 7428 4041 68149 32 19829 121436 Minus 2°C Minus 10% Precipitation 118870 291523 1165580 124267 65546 1 174441 8 2372 1178 49424 2850 2911 36363 466215 Minus 2°C Equal Precipitation 172228 187302 309841 197533 216376 8517 227019 3620 19591 101063 370699 1104 172674 63750 70905 379327 Minus 2°C Plus 10% Precipitation 212878 138330 215480 198576 261015 23552 232302 16780 329 27256 199016 402511 3355 138106 55024 99335 277704 Minus 2°C Plus 20% Precipitation 228339 174715 577380 159890 116387 16199 298860 10177 77 28176 81499 243565 4509 4937 102253 454586 Minus 2°C Plus 30% Precipitation 236957 270255 1283040 30454 21338 6114 294586 1058 20214 10591 10432 31 89799 226680 Minus 1°C Minus 30% Precipitation 121716 393239 1797628 3100 63453 221 47137 75055 Minus 1°C Minus 20% Precipitation 112625 260258 1523298 50808 15817 98 182940 90 3723 21 253 69831 281787 Minus 1°C Minus 10% Precipitation 138827 164111 409892 308761 142713 1477 25285 207901 9480 24999 28403 229853 979 48546 85772 94905 579645 Minus 1°C Equal Precipitation 169694 107162 156572 311535 236294 108350 54159 156444 30852 7002 23126 83539 206211 180780 159909 154639 114785 240496 Minus 1°C Plus 10% Precipitation 191585 109913 181245 308434 211210 88799 48950 169918 27147 10594 21510 107300 289786 89430 139870 124872 118382 262604 Minus 1°C Plus 20% Precipitation 203537 167558 688792 211596 82868 3792 28181 252584 16377 27302 28065 157470 937 6206 104380 521904 Minus 1°C Plus

211 30% Precipitation 211620 260605 1431613 29800 15903 7514 237280 854 14997 820 6666 9 84528 199340 Equal °C Minus 30% Precipitation 125486 294536 1663163 17097 922 890 171298 79 2956 14 79590 145518

Equal °C Minus 20% Precipitation 125111 169278 943850 181493 72275 1537 19721 235414 3545 22558 2229 21513 274 101656 601095 Equal °C Minus 10% Precipitation 143996 105900 201414 366991 193226 76638 48215 152987 35181 4508 25851 58212 270741 73767 116813 164930 122357 339822 Equal °C Equal Precipitation 160541 85284 131662 327601 83902 112124 242119 58860 128921 33544 17695 20105 78427 132092 280323 100137 178502 128815 200895 Modern AK Vegetation Map 171554 219746 74120 235450 211497 63189 121018 41608 230390 21221 8693 14439 84419 312761 128959 113028 88460 91894 189278 Equal °C Plus 10% Precipitation 176543 116195 221631 331287 187848 79 72713 47690 167927 32070 3029 26307 69189 272963 46459 131686 136474 114224 347285 Equal °C Plus 20% Precipitation 185313 188264 937054 191412 51947 1408 18234 246316 6663 25278 1680 57577 17 4268 91217 494901 Equal °C Plus 30% Precipitation 193308 284779 1614052 12411 2014 1077 203999 212 4351 62 1421 66112 117751 Plus 1°C Minus 30% Precipitation 132995 201181 1487314 25079 18465 8657 260071 623 12408 332 29 99753 254642 Plus 1°C Minus 20% Precipitation 133596 123573 548059 225292 132673 4810 28275 253918 16797 79 36213 20895 99602 12 25 2619 121797 753314 Plus 1°C Minus 10% Precipitation 143394 83461 156386 323652 230084 104888 50939 153138 33255 12538 25087 99049 281641 113021 143243 155022 134038 258713 Plus 1°C Equal Precipitation 155305 93481 148939 295302 258759 4 116652 55640 159301 33377 8010 27214 94007 188983 181048 180706 149279 119117 236425 Plus 1°C Plus 10% Precipitation 165161 151318 424324 275225 127065 635 24513 243388 19027 28209 22730 220972 1054 97076 63575 91192 546085 Plus 1°C Plus 20% Precipitation 170127 237464 1348747 70351 19361 307 245550 330 8339 233 16120 1480 60644 322496 Plus 1°C Plus 30% Precipitation 177847 326277 1731362 2882 88 162170 150 1 2 39873 60897 Plus 2°C Minus 30% Precipitation 146415 170645 1171795 16269 28016 6605 395596 1089 20476 2134 8527 112623 421359

Plus 2°C Minus 20% Precipitation 143501 101174 452875 152335 162845 259 15156 317580 14740 129 38355 69382 308840 961 2097 128949 592371 Plus 2°C Minus 10% Precipitation 150521 91345 177094 191674 339291 2318 21904 253645 24496 620 36170 160495 438488 3756 176977 58449 117858 256448 Plus 2°C Equal Precipitation 152089 143177 246373 186216 281854 7356 285984 15520 22 30633 95215 327112 683 251986 51154 79419 346756 Plus 2°C Plus 10% Precipitation 150263 231143 908541 131674 44860 10 290815 250 13179 2689 104038 16102 4234 39358 564393 Plus 2°C Plus 20% Precipitation 149778 311080 1660911 6146 2749 201631 152 15 2031 40 23220 143796 Plus 2°C Plus 30% Precipitation 155527 389507 1802697 795 12 112158 14590 26263 Plus 3°C Minus 30% Precipitation 201580 170730 790189 2179 19556 3 548567 69 17030 13532 23142 100984 613988 Plus 3°C Minus 20% Precipitation 200884 97775 452217 43488 103905 20 493854 186 30527 125735 415928 1848 81 81234 453867 Plus 3°C Minus 10% Precipitation 171573 168901 289640 9996 295095 8 429341 126 27580 199316 522706 15 97505 1609 24394 263744 Plus 3°C Equal Precipitation 136399 253263 634361 12707 143822 1 418985 14807 117710 223174 102707 472 3683 439458 Plus 3°C Plus 10% Precipitation 120541 330898 1497614 3977 17292 283476 1532 1342 5322 474 1136 237945 Plus 3°C Plus 20% Precipitation 121230 408225 1763688 819 784 164483 2 4 818 41496 Plus 3°C Plus 30% Precipitation 131958 478097 1802025 21 2 80314 640 8492 Plus 4°C Minus 30% Precipitation 299996 272435 637655 85 13032 682188 3464 34243 29259 28234 500958 Plus 4°C Minus

212 20% Precipitation 223095 200836 473852 325 64306 625720 8152 169494 435653 561 3841 295714 Plus 4°C Minus 10% Precipitation 141414 263076 469510 123 126358 588892 4648 221387 433152 16972 20 235987

Plus 4°C Equal Precipitation 104035 334921 1192559 727 34561 430390 513 41239 61782 4497 1 296324 Plus 4°C Plus 10% Precipitation 93608 408933 1705483 460 2616 211123 9 229 114 78974 Plus 4°C Plus 20% Precipitation 96478 480955 1807897 28 3 100116 16072 Plus 4°C Plus 30% Precipitation 106088 525784 1821255 44883 3539 Plus 5°C Minus 30% Precipitation 369482 524102 452882 6574 801501 14 45752 15705 2 1872 283631 Plus 5°C Minus 20% Precipitation 195817 393183 468591 31423 779431 24 164993 283386 97 48 184556 Plus 5°C Minus 10% Precipitation 108370 416622 748645 31123 683923 6 148438 175048 1418 187956 Plus 5°C Equal Precipitation 79252 449171 1510627 2 3753 333524 15 4152 1384 34 119635 Plus 5°C Plus 10% Precipitation 73610 499686 1772645 3 127184 19 11 28391 Plus 5°C Plus 20% Precipitation 77267 542400 1824260 51678 5944 Plus 5°C Plus 30% Precipitation 85309 563829 1830858 19560 1993