ABSTRACT POTEATE, AARON SAMUEL. Amerindian Mollusk

ABSTRACT

POTEATE, AARON SAMUEL. Amerindian Mollusk Exploitation During the Late Ceramic Age at Coconut Walk, Nevis, West Indies (ca. AD 850-1440). (Under the direction of Scott M. Fitzpatrick and John Millhauser).

Zooarchaeology is a fundamental part of investigating coastal archaeological sites around the world. Midden deposits, which typically contain the remains of shellfish, finfish, and other marine animals, provide important clues for understanding a host of issues, including prehistoric subsistence strategies, nutrition, and health. In this study, I conducted one of the most detailed analyses to date of a Pre-Columbian mollusk assemblage on the island Nevis (northern Lesser Antilles, Caribbean). The robust sample, comprising more than

58,000 individuals from the late Ceramic Age (ca. AD 850-1440) site of Coconut Walk were recovered from 25 discrete 1×1 m units within a larger 5×5 m trench that overlaid a dense midden deposit. I investigate two interrelated questions: How would interpretations of change be altered had more common sampling strategies of smaller units (e.g., several 1×1 m units versus the larger 5×5 m trench analyzed here) been used? And, how did mollusk exploitation change generally over a period of 600 years?

The results reveal a heavy dependence on only a few mollusk species with several trends apparent. As the intensity of exploitation increased through time, there was also an increase of more than 10% in the individual weight of the most commonly harvested mollusk, Nerita tessellata (tessellated nerite). Using statistical procedures, I also compared the results of the number of individual specimens (NISP), minimum number of individuals

(MNI), and weight (g) of each mollusk taxon between both units and plana. The results reveal that sampling less than 16% (4 out of 25 squares for this study) of a midden similar in type of site, site size, and diversity of activities to Coconut Walk is likely to produce results outside of a 95% confidence interval for the population composition. However, discerning change between layers may require a larger sample. In addition, a sample of this size (i.e.,

16%+) selected randomly or judgmentally will produce similar results, provided that proper spacing is given between excavation units. Overall, the data in this case suggest that sampling 16% of the midden by surface area was sufficient to determine midden composition and key species and that this sample size may be applicable to other cases worldwide.

This study contributes important new data on the subsistence strategies of Nevis’ prehistoric occupants and provides a framework for examining how sampling procedures can effect interpretation of a faunal assemblage. It also contributes to ongoing research focused on examining the underlying structure of mollusk harvesting during the Late Ceramic Age.

by Aaron S. Poteate

A thesis submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the degree of Master of Arts

Anthropology

Raleigh, North Carolina

2013

APPROVED BY:

______Dr. Scott M. Fitzpatrick Dr. John Millhauser Committee Co-Chair Committee Co-Chair

______Dr. Nora Haenn Dr. Torben Rick

DEDICATION

For Finley, whose love of biting feet has kept me on my toes.

BIOGRAPHY

Aaron Samuel Poteate grew up in the small town of Elkin, NC. He spent his early college years at North Carolina State University studying Aerospace Engineering before eventually completing his Bachelors of Art in Anthropology. He decided to continue his education with a goal to obtain a doctorate degree and become a permanent member of academia. He began this journey at North Carolina State University under the supervision of

Scott M. Fitzpatrick, where he is currently a candidate for a Masters of Arts in Anthropology and a Graduate Certificate in Geographic Information Systems. He will continue his education in the graduate program at the University of Oregon.

iii

ACKNOWLEDGMENTS

The project described in this paper was supported by Grant/Cooperative Agreement

Number G10AC00624 from the United States Geological Survey. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the USGS.

I would first like to thank my family and friends for always believing in me, even when it may have not been warranted. To Sydney Dotterer for pushing me to follow my dreams, always supporting me, and keeping me grounded. To my Mom and Dad, I appreciate your confidence and support in everything I have done.

Thank you to my co-chairs Scott M. Fitzpatrick and John Millhauser and committee members Torrey Rick and Nora Haenn for providing me with unique perspectives on my thesis topics. To Michiel Kappers, Willem Schnitger, Quetta Kaye, John Swogger, and

Calvin Emesiochel, thank you for all of the knowledge you shared with me the past two summers. It was a pleasure to work with all of you, and the experience will last a lifetime.

Thank you to Scott M. Fitzpatrick, Michiel Kappers, and Quetta Kaye – Co-Directors of the 2010 field project at Coconut walk. I also appreciate the help from students who participated in the field project and the Nevis Historical and Conservation Society, particularly Evelyn Henville, Paul Diamond, Jane Ebbit, and Cynthia Hughes, who arranged logistical support for the project. Christina Giovas, Scott M. Fitzpatrick, Jessica Stone, and

Meg Clark for helping with the initial identifications and creating the shell reference collection. I also thank the people who spent countless hours in lab sorting shells including

Jessica Stone, Meg Clark, Mary Brown, Maggie Kilcullen, and Kelsey Roepe.

Finally, I would especially like to thank Dr. Scott M. Fitzpatrick for giving me the opportunities to be where I am today. He continually provided me with the guidance and motivation to succeed. I thank him for believing in me from the start and never wavering. I would not be where I am today without him and for that I am grateful. I look forward to continuing to work with him at Oregon.

TABLE OF CONTENTS

LIST OF TABLES ...... viii LIST OF FIGURES ...... x

CHAPTER 1: INTRODUCTION ...... 1

CHAPTER 2: ENVIRONMENTAL AND CULTURAL CONTEXT ...... 3 The Caribbean ...... 3 Prehistoric Settlement of the Caribbean ...... 6 The Island of Nevis in the Lesser Antilles ...... 11 Coconut Walk ...... 13 Key Mollusk Species from Coconut Walk...... 18

CHAPTER 3: METHODOLOGICAL BACKGROUND ...... 23 Zooarchaeology ...... 23 Identification ...... 23 Weight ...... 24 Number of Identified Specimens ...... 25 Minimum Number of Individuals ...... 30 Statistical Analysis...... 32 Testing Sample Strategies ...... 34

CHAPTER 4: METHODS ...... 36 Archaeological Methods ...... 36 Zooarchaeological Methods Summary ...... 40 Identification of Taxa ...... 41 Minimum Number of Individuals ...... 42 Number of Identified Specimens ...... 47 Weight ...... 48 Species Count...... 48 Statistics ...... 49 Sample Strategies ...... 49 Testing Sampling Strategy ...... 51

CHAPTER 5: RESULTS ...... 54 Mollusk ...... 54 Trends ...... 57 Sample Selection ...... 72

CHAPTER 6: DISCUSSION: FROM REPRESENTATIVE SAMPLES TO HUMAN DIET ...... 81 Sample Size ...... 81

Mollusk Consumption ...... 84 Exploitation and Over-Exploitation of Mollusks ...... 89 N. tessellata ...... 94 C. pica...... 98 L. tuber ...... 100 Key Species...... 101 Factors for Exploitation ...... 103

CHAPTER 7: CONCLUSIONS ...... 108

REFERENCES...... 112

APPENDICES ...... 132 Appendix A ...... 133 Appendix B ...... 141

vii

LIST OF TABLES

Table 1. Summary of material excavated in 2010 at Coconut Walk ...... 15

Table 2. Top five species by percentage of weight ...... 18

Table 3. Weight, MNI, NISP, species by planum ...... 55

Table 4. Quantitative statistics for top 9 species at Coconut Walk ...... 55

Table 5. Percentage of MNI for top 9 species ...... 56

Table 6. Weight, MNI, NISP for N. tessellata, C. pica, L. tuber by planum ...... 56

Table 7. Paired t-tests for overall square composition ...... 60

Table 8. Paired t-tests for N. tessellata weight, MNI, and NISP ...... 65

Table 9. Paired t-tests for C. pica weight, MNI, and NISP...... 65

Table 10. Paired t-tests for L. tuber weight, MNI, and NISP ...... 66

Table 11. Paired t-tests for N. tessellata percentage of weight, MNI, and NISP ...... 66

Table 12. Paired t-tests for C. pica percentage of weight, MNI, and NISP ...... 66

Table 13. Paired t-tests for L. tuber percentage of weight, MNI, and NISP ...... 67

Table 14. Individual weight for N. tessellata by planum ...... 71

Table 15. Individual weight for C. pica by planum...... 71

Table 16. Fragment weight for C. pica by planum ...... 71

Table 17. Individual weight for L. tuber by planum ...... 71

Table 18. Fragment weight for L. tuber by planum...... 72

Table 19. The most abundant mollusks species at sites on Nevis ...... 87

Table 20. Top species by weight, MNI, and NISP at Coconut Walk ...... 88

Table 21. N. tessellata weight, MNI, NISP by planum ...... 94

viii

Table 22. C. pica weight, MNI, NISP by planum ...... 98

Table 23. C. pica individual and fragment weight ...... 100

Table 24. L. tuber weight, MNI, NISP by planum ...... 101

Table 25. Key species weight in kg by planum ...... 102

Table 26. Species weight, MNI, and NISP ...... 133

Table 27. Percentage of unit weight by species ...... 135

Table 28. Percentage of unit MNI by species ...... 137

Table 29. Key species ...... 139

Table 30. Weight errors ...... 141

Table 31. MNI errors ...... 141

Table 32. NISP errors ...... 141

Table 33. Species count errors ...... 142

Table 34. Sample composition with three selected species ...... 143

Table 35. Species found in each sample ...... 145

Table 36. Percentage of N. tessellata samples outside CI ...... 147

Table 37. Percentage of C. pica samples outside CI ...... 147

Table 38. Percentage of L. tuber samples outside CI ...... 147

LIST OF FIGURES

Figure 1. Map of the Caribbean ...... 6

Figure 2. Map of Nevis ...... 13

Figure 3. Trench 2273 at Coconut Walk ...... 16

Figure 4. Shell beads from excavations ...... 16

Figure 5. Incised ceramic stamp ...... 17

Figure 6. Incised shell bead ...... 17

Figure 7. Nerita tessellata ...... 19

Figure 8. Cittarium pica ...... 20

Figure 9. Lithopoma tuber ...... 21

Figure 10. Map of Coconut Walk ...... 38

Figure 11. Numbering convention of 1x1 m squares within a 5x5 m trench ...... 39

Figure 12. Parts of a bivalve shell ...... 44

Figure 13. Parts of a gastropod shell ...... 45

Figure 14. Parts of a chiton shell ...... 45

Figure 15. Hypothetical samples selected for analysis in a 5x5 m trench ...... 52

Figure 16. Weight for each square by planum...... 57

Figure 17. MNI for each square by planum ...... 58

Figure 18. NISP for each square by planum...... 58

Figure 19. Number of species per square by planum...... 59

Figure 20. N. tessellata weight by planum ...... 61

Figure 21. N. tessellata MNI by planum ...... 62

Figure 22. N. tessellata NISP by planum ...... 62

Figure 23. C. pica weight by planum ...... 63

Figure 24. C. pica MNI by planum ...... 63

Figure 25. C. pica NISP by planum ...... 64

Figure 26. N. tessellata individual weight by planum ...... 68

Figure 27. Square weight vs. MNI for N. tessellata by planum ...... 68

Figure 28. Sieved square weight vs. MNI for N. tessellata by planum ...... 69

CHAPTER 1: INTRODUCTION

As model systems, islands can provide an opportunity to elucidate the changes that occurred over time after human arrival due to their insularity and pristine ecologies (Kirch

1984, 1997, 2000; Kirch and Green 1987; Vitousek 1995, 2002, 2004). Island ecosystems around the world were often dramatically altered by humans through transforming landscapes for agriculture, introducing non-native species of plants and animals, and heavy exploitation of marine resources, among other reasons (Kirch 1982, 1994, 1997, 2000, 2007:

Burney et al. 2001; Steadman 1995; Fitzpatrick and Keegan 2008).

As expected, many of the primary resources exploited by island populations were marine in nature (Kirch 1982; Rick et al. 2005; Glassow 2003, 2005). The middens that were formed through the deposition of food remains, including mollusk shells, fish bones, and crustacean remains, were variable in size, ranging from a few square meters to hundreds or thousands of square meters. Such middens provide an important glimpse into the food harvesting strategies of past populations (Anderson 1989, 1997, 2002; Allen 1997; Martin and Steadman 1999; Grayson 2001; Rainbird 2002; Kirch 2004; Wroe et al. 2006).

In island midden contexts, prehistoric subsistence strategies and environmental interaction can be examined through the analysis of faunal remains, of which mollusks are often the most common (Fitzpatrick and Keegan 2007; Stiner and Munro 2002; Fitzpatrick and Donaldson 2007; Giovas et al. 2010). The Caribbean is no exception, particularly during the Late Ceramic Age, whereby Amerindians heavily exploited marine foods as part of their subsistence strategy. In many cases, however, zooarchaeological analysis of food remains is based on the recovery of small sub-samples within a site (see Ballbe´ 2005; Keegan et al.

2008; Newsom and Wing 2004; Wing and Wing 1995; Wing 2001). This is for good reason, given that the process of identifying and analyzing faunal remains is time intensive and challenging, even with an adequate comparative collection.

An important consideration when conducting zooarchaeological analysis is to gain a statistically valid sample with which to interpret past human behaviors (Lyman 2009). For this study, I examined a complete assemblage of mollusks from a large (5×5 meter) midden deposit at the Late Ceramic Age site of Coconut Walk on the island of Nevis in the northern

Lesser Antilles. The excavation of the site, conducted as part of the 2010 field school directed by Dr. Scott M. Fitzpatrick (formerly of North Carolina State University, now at the

University of Oregon), Dr. Quetta Kaye (Institute of Archaeology, University College

London), and Michiel Kappers (QLC, BV, The Netherlands), led to the recovery of nearly

650 kg of material, of which almost 180 kg was marine shell. Sorting and quantifying the mollusk assemblage required an additional two years of work in the lab by both graduate and undergraduate students at NC State. As a result, there was a large spatial and vertical sample with which to examine mollusk harvesting by ancient Amerindians on Nevis and determine the validity of various sampling procedures. This study adds to the current knowledge of prehistoric Caribbean subsistence strategies and environmental interaction by answering questions related to Amerindian mollusk harvesting, and provides a case study with which to explore how common zooarchaeological sampling strategies may affect our interpretations of past human behaviors and the interactions of humans with their environments.

CHAPTER 2: ENVIRONMENTAL AND CULTURAL CONTEXT

The Caribbean

The Caribbean is divided into four major island groups: the Greater Antilles, Lesser

Antilles, the Bahamian archipelago, and those islands that border the northern part of South

America, including Aruba, Bonaire, and Curacao (Figure 1). The Greater Antilles are situated along the northern part of the region, consisting of 194,000 km2 of total land area and include the larger islands of Cuba, Hispaniola, Puerto Rico, and Jamaica. The Lesser Antilles form an arc along the eastern portion of the Caribbean Sea that runs from Grenada in the south to the

Virgin Islands in the north and is comparatively smaller at total land area of 12,000 km2. The

Lesser Antilles is also sometimes subdivided into two separate groups: the Windwards and the Leewards. The Windward Islands are in the south and the Leeward Islands in the north with the division occurring between Dominica and Martinique. The division is based on where ships from Europe would often arrive (around Dominica and Martinique), creating islands that were windward (direction upwind from the reference point) and leeward

(direction downwind from reference point). The “Windwards” typically include Trinidad and

Tobago as well as Aruba, Bonaire, and Curacoa creating an overlap and inconsistencies in naming conventions throughout the Caribbean.

The environmental conditions throughout the Caribbean vary greatly between islands, with the variation in size perhaps being the most important difference (Wilson 2007). The islands of the Greater Antilles are much larger than the Lesser Antilles, with Cuba (110,860 km2) having an area 10 times larger than that of the entire Lesser Antilles. The larger islands are not only able to support larger populations, but also higher population densities due to

their large and fertile valleys (Wilson 2007). It was possible on some islands in the Greater

Antilles for inhabitants to be solely reliant on terrestrial food given their greater resource diversity and more abundant arable land. The Lesser Antilles mostly lack large, fertile valleys, which created a reliance on marine resources for their inhabitants over the duration of prehistoric settlement.

The Greater Antilles is geologically part of the cordillera mountain systems of Central

America. There are two cordillera systems that carry from Central America to the Greater

Antilles. The northern system extends from Guatemala and Belize to Cuba and northern

Hispaniola. The southern system extends from Honduras and Nicaragua to Jamaica,

Hispaniola, and Puerto Rico.

The collision of the Atlantic and Caribbean tectonic plates produced two types of islands in the Lesser Antilles (Newsom and Wing 2004; Wilson 2007). The majority of islands in the Lesser Antilles are volcanic, mountainous, and geologically young. The remainder are relatively flat as a result of its limestone sedimentary substrate composition uplifted by the plates colliding. The collision process creates a distinct difference in islands, with the western Basse Terre (inner rim) made up of a series of volcanoes and the eastern

Grande Terre (outer rim) made up of 20-million-year-old limestone. As a result, the islands in the Lesser Antilles are fairly mountainous with higher elevations that result in greater levels of precipitation than the flatter limestone islands (Watts 1987).

The Caribbean is extremely diverse ecologically, containing 2.3% of the world’s endemic plant species and 2.9% of the endemic vertebrate species, but in an area that represents only 0.15% of the world’s surface (Fitzpatrick and Keegan 2008). Marine species

are equally well-represented, with over 630 mollusk species, 1500 fish species, and 25 coral genera.

The climate varies seasonally with a wet season from June to November and a dry season from December to May (Blume 1972). Precipitation also varies among islands depending on latitude, and within islands based on elevation and the side of the island.

Temperature often only varies slightly between seasons. Palaeoclimate studies in the

Caribbean include lake core sediment analysis in Belize (Gischler et al. 2008), the

Dominican Republic (Lane et al. 2011), Haiti (Hodell et al. 1991; Curtis & Hodell 1993;

Hiquera-Gundy et al. 1999), Puerto Rico (Burney et al. 1994); Siegel et al. 1995), and St.

Martin (Bertran et al. 2004). These studies show alternating wet and dry periods during the last 10,000 years, with some localized differences, suggesting that precipitation has varied through time and space in the Caribbean. Hurricanes are another important climatic influence with frequency increasing during the last 1,000 years (Gischler et al. 2008; Lane et al. 2011;

Richey et al. 2007).

Figure 1. Map of the Caribbean (drafted by Michael Scisco, Biogeographic Creations).

Prehistoric Settlement of the Caribbean

Based on archaeological evidence, there were at least three major migrations into the

Caribbean (Archaic, Saladoid, and post-Saladoid), with there likely being many more minor events taking place over the a period of several thousand years. The structure of migratory movements (colonization) largely stems from Rouse’s (1992) work, whereby distinct styles of pottery are used to identify a population. Each style is typically named after the site where

it was first found. Local pottery styles are then grouped into subseries denoted by an –an suffix and series denoted by an –oid suffix.

The earliest dates for the prehistoric settlement of the Caribbean are found in Cuba and Hispaniola and date to around 6000 BC These groups are often referred to as Lithic, or

Casimiroid peoples (Rouse 1992) that may have originated from somewhere in Mesoamerica

(Keegan 1994; Wilson et al. 1998). Rouse (1992) classified the Casimiroid Series into the

Casimiran, Courian, and Redondan subseries. The Lesser Antilles are known to have been occupied from about 4000-2500 BP by Archaic peoples from South America (Fitzpatrick and

Keegan 2008) and are also referred to as Ortoiroid (Rouse 1986, 1992; Petersen et al. 2004).

Archaic sites are characterized by access to marine resources with an artifact assemblages including tools of chipped and ground stone and shell (Fitzpatrick and Keegan 2008). There is some evidence of cultivated of plants (Newsom and Wing 2004) and seasonal harvesting of marine fish and shellfish (Hofman et al. 2006; Hofman and Hoogland 2003). Little is known about these early groups as they were primarily mobile foragers who left only ephemeral evidence of their habitations and activities.

Archaic period sites are found on Barbados (Drewett 2000; Fitzpatrick 2011), St.

Thomas (Lundberg 1989), Nevis (Wilson 1989), Anguilla (Crock et al. 1995), Barbuda

(Watters et al. 1992), Guadeloupe (Richard 1994), Antigua (Davis 2000), and Saba (Hofman and Hoogland 2003; Hofman et al. 2006). Keegan (1994) classified the Archaic sites into three localized groups based on their artifact assemblages: the Coroso from Vieques, the

Cayo Cofresi from eastern Puerto Rico, and Krum Bay from St. Thomas.

Because of the difference in localized assemblages, the origin of Archaic peoples is debated (Fitzpatrick and Keegan 2008; see Lalueza-Fox et al. 2003; Keegan 2004, 2006;

Crock et al. 1995; Davis 2000; Lundberg 1989, 1991).Some scholars have suggested that

Belize was the place of origin of the Archaic peoples (Wilson 2007). Research links the artifact assemblages from sites in Belize to those found during the Archaic period of the

Caribbean (see MacNeish and Nelken-Terner 1983; Zeitlin and Zeitlin 1996; Wilson et al.

1998; Hester et al. 1994; Jones 1994). In contrast, investigations into Caribbean winds and currents led Callaghan (1985, 1995) to suggest that drift voyages from South America were most likely to land in the Greater Antilles and that canoe voyages between the two were possible (1992). The majority of islands in the Lesser Antilles are also intervisible (Newsom and Wing 2004). While intervisibility may seem to have been ideal in terms of island settlement, in actuality, navigation between the islands could be hazardous given the prevailing winds and currents that push westward and “bottleneck” through the intervening channels (Callaghan 1992).

The period of the second wave of migration is referred to as the Ceramic Age, so designated because well-made pottery is found coincident with intensive horticulture at settlements from this time period (Oliver 2001; Rehm and Espig 1991). The Ceramic Age begins around 2500 BP, with the migration of Saladoid groups from South American to

Puerto Rico and the Lesser Antilles (Fitzpatrick and Keegan 2008; see Keegan 2000). The term “Saladoid” comes from the type site of Saladero in Venezuela (Newsom and Wing

2004). The Ceramic age brought about the settlement of nearly every island in the Caribbean by 2200 BP (Fitzpatrick and Keegan 2008), though Fitzpatrick (2006) suggests that further

research is needed to address colonization and settlement of the Saladoid period. Rouse

(1992) suggested that the biogeographic arrangement of the islands, coupled with favorable ocean currents, would have allowed prehistoric settlers to quickly advance from the northern

South American littoral up through the Lesser Antilles. However, simulated voyaging routes conducted by Callaghan (2001) and chronological evidence (Fitzpatrick 2006) suggests that a northward “stepping-stone” migration may not have occurred. There is also a concentration of early Saladoid sites on the northern coasts of islands, suggesting a southward focus

(Keegan 2000; Callaghan 2003; Fitzpatrick 2006: Fitzpatrick et al. 2010).

Saladoid groups began transforming the Caribbean flora and fauna with the introduction of several non-indigenous plants and animals. These include the agouti, guinea pig, dog, opossum, armadillo, peccary, manioc, wild avocado, papaya, tobacco, pepper, peanut, maize, and sweet potato (Fitzpatrick and Keegan 2008; Newsom and Wing 2004).

The earliest Saladoid sites are found on the coast, with later sites at inland locations

(Fitzpatrick and Keegan 2008). Trade is evident during the Saladoid period with the discovery of rocks and pottery from the mainland and other islands (Keegan 2000, Righter

1997).

Saladoid peoples are classified into series based on their pottery styles (Keegan

2000). Keegan (2000) suggests there are three primary styles; Saladoid, post-Saladoid, and

Ostionoid. The first migration by Saladoid peoples included a jump from South America to eastern Puerto Rico, the Leeward Islands, and U.S. Virgin Islands (Keegan 2000; see

Callaghan 1995 and Keegan 1995). A similar jump is suggested by Fitzpatrick (2006) in a review of Caribbean radiocarbon dates and chronology building.

Two early pottery subseries were defined by Rouse (1992); zone-incised, crosshatched (ZIC) pottery from the Huecan subseries, and painted and modeled-incised pottery from the Cedrosan subseries. There is some mixing of pottery styles at sites

(Chanlatte Baik 1995; Petersen and Watters 1995), but Huecan is found exclusively in eastern Puerto Rico between 200 BC and AD 200 (Keegan 2000). The Huecan pottery style is characterized by ZIC decoration and a lack of painted pottery. Cedrosan pottery has been found on islands in the Lesser and Greater Antilles from 500 BC to AD 600. Cedrosan pottery is “composed of well-made vessels, including complex shapes decorated with red, black, white-on-red, and polychrome painting; zone-incised crosshatching, incision, and punctuation; modeled and incised zoomorphic adornos; and strap and loop handles” (Keegan

2000 via Alegria 1993; Petersen and Watters 1995; Rouse 1992).

Around AD 600, the Ostionoid peoples appear to represent a new cultural expansion into the Greater Antilles. Rouse (1992) divides the Ostionoid series into Ostionan, Meillacan, and Chican. Ostionan was characterized by straight-sided open bowls and boat-shaped vessels, polychrome, and red painting (Keegan 2000). Meillacan mirrors the shape of

Ostionan but with a rougher surface. Chican pottery is decorated with modeled incision

(Rouse 1992). Also beginning around AD 600, and lasting through AD 1000, was the

Troumassan sub-series of the Troumassoid style of pottery in the southern Lesser Antilles.

The Troumassan pottery represents a decrease in quality with the use of curvilinear incised lines and red, black, and white painting (Keegan 2000). The movement away from more sophisticated pieces is continued with the Suazan Troumassoid pottery (AD 1000-1450)

characterized by simple and bulky ceramics with fingernail or finger-impressed decoration around the rims of vessels and human-faced adornos, often with ear plugs (Keegan 2000).

The Taíno peoples, the probable descendents of “Ostionoid” groups, appear to have moved into Cuba, Hispaniola, Jamaica, and the Bahamas between AD 600-1200. The Taíno represent a group with increased society complexity, larger villages, and higher populations

Central plazas, ball courts, and elite chiefs are characteristic of these groups (Wilson 2008,

Keegan 2000). Notably, it was the Taíno who first encountered Columbus when he arrived in

1492.

The Island of Nevis in the Lesser Antilles

The island of Nevis is located in the northern part of the Leeward Islands and is part of the political confederation of St. Kitts and Nevis. Nevis is the smaller of the two islands with an area of 93 km2 and was known as Oualie (‘land of beautiful water’) to its original inhabitants. At its center, Nevis reaches a height of 985 m on the dormant volcano Nevis

Peak. Similar to other islands in the Caribbean, the temperature is tropical with only slight seasonal variation. The highest temperatures generally occur during the months of August,

September, and October, with an average high of 29˚ C. The lowest temperatures are in

January with an average high of 27˚ C. The average yearly rainfall on the island is around 48 inches.

Paleobotanical evidence suggests that Nevis was originally covered in tropical- subtropical dry forests, including lignum-vitae, strong bark, and torchwood (Newsom and

Wing 2004). Further evidence from post-Saladoid sites suggest a change in the environment

during the last thousand years with a migration to dry scrub vegetation such as sandbox tree, fish poison, pepper bush, and cockspur (Newsom 2004). As with most forested regions that are inhabited by humans, Beard (1949) believes that there would have been a progressive local elimination of forest species. The archaeobotanical evidence is further supported by the tools found for crop preparation, which would have required the clearing of forests and of the need for firewood by local inhabitants (Newsome and Wing 2004).

The first comprehensive archaeological surveys on Nevis were conducted by Samuel

M. Wilson in the 1980s who recorded 21 sites (Figure 2) from three major periods (1989,

2007). Wilson (1989) described these as the Archaic period from 800-1 BC, the Saladoid period from 500 BC - AD 600, and the Ostionoid (also known as post-Saladoid) period from

AD 600 to European contact. Other work on the island has included excavations at

Hichmans’ by Strutt (2003) and teams from Southampton University led by Elaine Morris, and Coconut Walk by the U.K. television show the Time Team (Bellamy 2001; Nokkert

2001). Given that very few sites on Nevis had actually been excavated in detail, archaeologists Scott Fitzpatrick, Michiel Kappers, and Quetta Kaye continued excavation at

Coconut Walk in 2010 areas to help discern the extent and intensity of occupation during the

Late Ceramic Age.

The only Archaic site excavated to date has been Hichmans’ Shell Heap which dates to ca. 790-520 BC Part of the general area was subsequently occupied by Saladoid peoples from approximately 100 BC to AD 600 and is known as the Hichmans site. Several post-

Saladoid sites have also been excavated, including Indian Castle (AD 650-880), Sulpher

Ghaut (AD 900-1200), and Coconut Walk (AD 850-1440), though all of these were limited in scope.

Figure 2. Map of Nevis with the area surveyed indicated (edited from Wilson 1989).

Coconut Walk

The Coconut Walk site, located on the east coast of Nevis, was first identified as a

Late Ceramic Age site by Samuel Wilson (1989:427-450; 2007:62-63) who noted an abundance of ceramics, food remains, and some human burials on the surface (Figure 2). The site was first excavated in 1998 as part of the British-based Time Team television show.

During these excavations, the Time Team archaeologists collected a rich sample of artifacts and faunal material and recorded evidence of household structures (Bellamy, 2001; Nokkert,

2001). However, there were no radiocarbon dates to chronologically anchor the site and the recovery methods for archaeological remains were too coarse to provide detailed examination of long-term use of the site.

In 2010, an archaeological field school co-directed by Dr. Scott M. Fitzpatrick through North Carolina State University, along with Michiel Kappers (M.Phil) and Dr.

Quetta Kaye, began work at Coconut Walk to provide a more detailed record of Pre-

Columbian activity during the later stages of settlement prior to European contact. Initially, they opened up two 5×5 m trenches, labeled 3073 and 2973, revealing a shallow (30 cm) anthropogenic layer with sterile, sandy subsoil beneath. The third trench (2273) was focused on a midden deposit slightly to the south of trenches 3073 and 2973. Trench 2273 (Figure 3) consisted of a 5cm thick layer of topsoil followed by 40 cm of midden deposits on top of a sterile, sandy subsoil. The midden deposits in Trench 2273 were dated to approximately AD

850 to 1440 based on radiocarbon dates (Fitzpatrick, personal communication).

The 2010 project at Coconut Walk produced a large assemblage of Saladoid period artifacts, ecofacts, and human remains (Table 1)(Figures 4, 5, and 6) The cultural materials included:

modeled clay adornos (small figures which would have been attached to the rims of vessels), four polishing stones (thought to be used in smoothing pots before firing), shell and stone tools (including three of polished greenstone), two polished stingray spines (possibly piercers or awls), 15 assorted shell and stone beads with three blank unpierced forms, two spindle whorls, four body stamps, numerous chert flakes, and

subsistence remains in the form of mollusks, fish, turtle and small mammal bones. (Kaye et al. 2010:141)

There was also part of a single human burial uncovered at the bottom of trench 2273 in the subsoil layer. The burial included a section of maxilla with three molars that suggests the individual was at least in their mid-teens at death (Qaye et al. 2010).

Table 1. Summary of material excavated in 2010 at Coconut Walk.

Category Weight (g) Animal Bone 6,499.20 Human Bone 606.60 Ceramics 431,259.80 Charcoal 155.43 Chert 1,237.60 Crustaceans 3.00 Glass 75.20 Plastic 8.00 Sample Animal Bone 26,554.00 Sample Ceramics 1,953.40 Shell 179,731.60 Stone 805.40 Tooth Human 13.00 Unknown 17.40 Total 648,919.63

Figure 3. Trench 2273 at Coconut Walk. Raised units are environmental squares.

Figure 4. Shell beads from the 2010 excavation at Coconut Walk.

Figure 5. Incised ceramic stamp (drawn by John Swogger).

Figure 6. Incised shell bead (drawn by John Swogger).

Key Mollusk Species from Coconut Walk

At Coconut Walk, a large variety of mollusks were recovered during excavation, with a minimum of and three classes (gastropoda, bivalvia, and polyplacophora) and 78 individual taxa. The top five species by percentage of weight are shown in Table 2. The results show that N. tessellata, C. pica, and L. tuber are the three most abundant species. These species would have lived in the same intertidal area and been prepared for human consumption in the same way, perhaps mixed together. The other less common species found at Coconut Walk also inhabited the intertidal area, with a few small land snail species. In the following sections I describe the basic attributes of the three most important species and its relative importance with regards to the mollusk harvests.

Table 2. Top five species by percentage of weight.

Species % Weight Cittarium pica (gastropod) 38.2 (1) Nerita tessellata (gastropod) 21.5 (2) Lithopoma tuber (gastropod) 10.3 (3) Strombus gigas (gastropod) 4.2 (4) Chiton tuberculatus 3.6 (5)

(polyplacophora)

Figure 7. Nerita tessellata (jaxshells.org).

Nerita tessellata

N. tessellata, commonly called the checkered nerite for its tessellated pattern (Figure

7), is the second most abundant species by weight at Coconut Walk. This species is found in large numbers in intertidal regions along rocky shorelines in up to 0.5m of water (Axelsen,

1968; Bovbjerg, 1984; Chislett, 1969; Potts, 1980) in the Gulf of Mexico from Florida to

Texas as well as through the Caribbean to Brazil (Abbott and Morris, 1995; Rehder, 1981).

The nerite is typically under 20 mm in length (Axelsen 1968; Lewis 1971). Nerites were important food sources throughout the Lesser Antilles including at XYZ (citations)

The tessellated nerite is often found with Nerita peloronta and N. versicolor (Axelsen

1968), two species that were also present at Coconut Walk. Prehistorically, N. tessellata was gathered from rocks and placed into a pot of water to be boiled where the meat could more

easily be extracted. The small size of the species makes it necessary to collect a large quantity to provide sufficient calories, though it is comparatively easy to access, harvest, and consume.

Figure 8. Cittarium pica (gastropods.com).

Cittarium pica

C. pica, commonly called the West Indian top snail is the most abundant species by weight at Coconut Walk. It is a much larger species than N. tessellata, with an average length of 32-34mm and a maximum recorded dimension of 137 mm (Robertson 2003). The shell is very thick and heavy with a color pattern of white with black zigzag spots (Figure 8)

(Robertson 2003), though its nacreous nature makes it highly susceptible to fragmentation. I have found small fragments to be easily identifiable because of the unique color pattern. C.

pica species is rarely found along the Florida coast, with primary habitats being the coasts of

Mexico south to Venezuela as through the Caribbean (Abbott and Morris, 1995; Rehder,

1981; Robertson 2003). C. pica is found in intertidal and shallow subtidal zones with a maximum recorded depth of 7 m though it tends to live in shallower water (Robertson 2003).

The taxon is still boiled and eaten in the Caribbean today (“whelk chowder”), though it is also prepared with rice or salad (Robertson 2003).

Figure 9. Lithopoma tuber (gastropods.com).

Lithopoma tuber

The third key species, L. tuber, is commonly known as the green star shell (shown in

Figure 9). It is found on rocks in the intertidal and subtidal zones with a shell size between

25-75 mm (Abbott and Morris 1995; Rehder 1981). It is recognizable through its green and white cross-hatched pattern and whorls (Abbott and Morris 1995). Like the other two key species, it would have likely been collected, boiled, and then eaten (Rehder 1981).

Conclusion

The environmental and cultural context of the excavation at Coconut Walk is critical for understanding the exploitation of mollusks by the inhabitants. Cultural evidence from the region suggests that societies were becoming more complex and permanent during the period of occupation at Coconut Walk (Rouse 1992; Keegan 2000). The shift toward permanent settlements with a heavy reliance on marine resources, especially on small islands such as

Nevis, should produce a notable change in subsistence strategies. The localization of the culture and environment create unique settlements that can be used as a proxy for local human impacts. Using the evidence of mollusks collected during the 2010 excavations, I was able to analyze changes in mollusk exploitation over the 600 year period of occupation.

CHAPTER 3: METHODOLOGICAL BACKGROUND

Zooarchaeology

Zooarchaeology is the study of animal remains from archaeological sites. It is an interdisciplinary field with methods and theories from biology, paleontology, and anthropology. The primary goal of zooarchaeology is to understand the interactions that took place between humans and their environments in the past (Reitz and Wing 2008: 1). In this chapter, I discuss a number of the most common and basic techniques used in zooarchaeology that help determine the type and quantity of animal remains found in archaeological contexts. These include taxonomic identification, weight, number of identified specimens (NISP), and minimum number of individuals (MNI). In the following sections, I describe these techniques and highlight their strengths and weaknesses in order to make clear my reasons for selecting the specific techniques I used for this study of mollusks at Coconut

Walk.

Identification

The first step to understanding the material collected is to identify animal remains to the most specific classification possible. Ideally, every fragment would be identified to species, yet complete identification is almost never the case. Often, fragments can only be identified to a higher taxonomic classification such as family or genera. This is particularly true of tropical environments where biodiversity is greater and differences between species are often difficult to discern. The identification issue is notable in the literature among tropical fish (Butler 1998).

The identification of faunal remains follows the standard biological classification scheme originally developed by Carolus Linnaeus. Having a common system has enabled scientists to assign a scientific name for each taxon based on various characteristics. It also allows for a quick comparison of species likeness; for example, two mollusks from the same genera within the same family of Neritidae (Nerita tessellata and Nerita versicolor) will be more closely related than two from different genera and families (e.g., Nerita tessellata and

Cittarium pica). Species likeness is not representative of lineage, although it is often the case.

Weight

Measuring the weight of animal remains is an easy way to compare rough differences

(amount) between taxa within and among assemblages. Weight represents the physical weight of any material collected and is often recorded after the identification process. The absence of soft tissues is assumed in all cases unless otherwise specified. Weight is the most basic quantification method and provides archaeologists with a rough proxy for the size of the living population needed to produce the archaeological sample. The usefulness of weight as a variable has been debated (see Mason et al. 1998; Glassow 2000).There have been a number of studies which have used the weight of an assemblage to infer meat weight by comparison with modern day examples (Wing and Reitz 2008; Colton 2002). These studies provide useful, albeit somewhat limited, information. The major downside to using weight in zooarchaeological analysis is that these samples are not able to take into account taphonomic processes that can lead to a reduction in bone or shell weight.

Taphonomic processes vary by site, with the most common being decomposition or alteration. Decomposition is influenced by a variety of factors, including soil chemistry.

Acidic soils which tend to rapidly break down organic material. It is the exception rather than the rule to find a set of animal remains that has not succumbed to some level of decomposition. Another process, leeching, occurs when bones or shells loose or gain minerals which can lead to fossilization (Claassen 1998). Additionally, calcification of animal remains is not uncommon in a coastal and/or limestone environment (Kent 1988), and given difficulties in cleaning extraneous soil, this can lead to inflated weights of an assemblage that limits the usefulness of this type of quantification. Therefore, in order to use weight as a meaningful variable, it is important to understand which taphonomic processes were involved in a particular site.

Number of Identified Specimens (NISP)

NISP, or specimen count, is another basic quantitative measure in zooarchaeology

(Grayson, 1984). With the measure, we can compare the relative frequencies of different taxa, genera, classes, and the like. At its core, NISP treats each fragment as if it derived from a unique living specimen, creating a count of the number of identified specimens. In practice,

NISP represents the number of fragments recovered that can be identified to a particular taxon (see Bobrowsky 1982; Chaplin 1971; Grayson 1973, 1984; O’Connor 2000; Perkins

1973; Ringrose 1993; Watson 1972; White 1953). In general, researchers agree that although

NISP is not necessarily an accurate representation of the number of individual specimens in an archaeological sample, it does provide an estimate of the quantity of specimens recovered

from an excavation. For further research on the different ways that NISP can be used, see

Cannon (2000), Gilbert and Singer (1982), Maltby (1985), Tchernov (1993), and Thomas

(1971).

The primary issue with NISP is that the fragmentation of animal remains pre- and post-deposition can produce misleading results (Reitz and Wing 2008). For instance, if a femur is broken in half to extract the marrow during butchering, it would change the NISP from one to two. If the same femur further fragments into four pieces between original deposition and recovery, the NISP is now four compared to the original one. These problems are often amplified for shellfish where a single shell can fragment into dozens or more fragments. Aside from issues associated with fragmentation, there are also concerns with identification. For most categories of animals, a significant portion of the remains can be difficult to determine if they are highly fragmented (Claassen 1998). As the size of bones required for identification increases, so will the number of unidentified fragments and the possibility that the results will be skewed.

NISP relies on the ability to link each fragment to a particular taxon (Mason et al.

1998). The identifiability of each fragment also varies between taxa. Some mollusk species require a large portion of the shell to be present to accurately identify, whereas some mammals and fish only require a small portion of bone (Claassen 1998). Likewise, the size of fragments differs between taxa. A small taxon (i.e., rat) will often create smaller fragments than a larger taxon (i.e., elephant). For example, a NISP of 1 for a small mollusk and a large mammal have significantly different meat and caloric yields, which must be accounted for

outside of NISP because this type of quantification accounts only for the number of fragments identified from each taxon.

As a result, taphonomic processes play a key role in the applicability of NISP as a variable (Reitz and Wing 2008). Animal remains found in a midden will likely undergo a number of processes that increase NISP from the time of deposition to excavation and even during transport of samples to the lab (Claassen 1998; Giovas 2007). Trampling can occur on many different levels, which causes pressure and can lead to breaking and fragmentation

(Muckle 1985). There are also natural chemical and physical processes that occur that can weaken the integrity of site constituents (Kent 1988). Physically, the ground can shift (slowly or abruptly, as with an earthquake), animals can alter the midden, and rain can weaken the physical structure of the remains. Animals can cause a variety of known alterations to middens (Carucci 1992; Goodfriend 1991; Bocek 1986). The range of alteration can be as simple as a burrow that runs through a midden, to an animal that is actively gnawing on bones. Both can cause additional fragmentation to animal remains, thereby increasing NISP.

The methods used for the recovery of faunal remains can also lead to a bias in the

NISP. Because it is nearly impossible to recover everything, there will almost always be a bias toward the recovery of larger elements. It is considered common to use, at the very least,

1/4” (6.0 mm) or 1/8” (3.0 mm) mesh screen to recover zooarchaeological material, but that means that any smaller residuals will not be recovered (Bowlder 1983; Buchanan 1985;

Shackleton 1988; Wing and Quitmyer 1985; Waselkov 1987). This bias in recovering small fragments is typically not a concern for larger animals, but smaller mammals and fish can easily pass through coarser mesh (e.g., 1/2”), and as a result, it is not used (Reitz and Wing

2008). With small fragments, a flotation sample and micro-artifact analysis can provide material and techniques for study that would normally be not be recovered (Nagoka 2005;

Muckle 1994).

In the laboratory, methods used for identification can also alter the results of NISP.

Because it is often difficult to sort certain fragments to species, they are instead attributed to the higher taxonomic categories such as genera or family (For examples, see: Boessneck

1969; Halstead et al. 2002; Loreille et al. 1997). For example, in this study, identifying a fragment as Nerita tessellata versus Nerita versicolor requires the presence of the body columella. Fragments that do not contain at least part of the identifiable element cannot be identified to species, only to genus. This is not necessarily a problem with larger sieves because they do not recover the unidentifiable fragments. But, with 1/16” sieved material, there would be an abundance of NISP attributed to the genus instead of the species (Stein

1992). Comparing the NISP of one species to another can thus lead to erroneous conclusions if one or both species is prone to having portions classified to a higher classification (Stein

1992). There are techniques to avoid bias from a higher classification, such as dividing NISP by the number of identifiable elements for each taxa though these techniques are primarily only used with fish among marine taxa (Perkins 1973, Perkins and Daly 1968). This technique can help eliminate the issue of having smaller fragments never being recovered, as well as ones associated with the variation in number of skeletal elements from different taxa.

The number of skeletal elements from each taxa will differ, though they will be similar between those species which also are very similar. For instance, a gastropod is generally composed of one shell (although it can be broken down into elements) and

operculum (rarely found with the shell) compared to a mammal that can have hundreds of skeletal elements. If both species have perfect preservation, the mammal will inherently have more fragments because it’s skeleton is naturally divided into numerous parts. If we compare an assemblage that has an NISP of 100 for both a gastropod taxon and a mammalian taxon, it might seem that the same number of specimens is present. But, accounting for the number of skeletal elements in each taxon will create a more realistic NISP for each taxa.

Once corrected for inter-species differences in fragmentation, identification, and number of skeletal elements, NISP provides archaeologists with a way to compare the relative frequencies of different taxa (see Cannon 2000; Gilbert and Singer 1982; Maltby

1985; Techernov 1993; Thomas 1971). The most common use of NISP, however, is as a relative count of skeletal frequency between taxa. For example, if one taxon comprises over half the total NISP, it can be assumed that of the specimens consumed, over half are from that taxon. There is not a set procedure for adjusting NISP to account for bias, but the above references can assist in finding the appropriate adjustments. Though it would be ideal to have a set procedure for adjusting NISP, because of the plethora of taxa and the specialization of each researcher, there is not likely to be a universal method for adjustment developed. Thus, it is left to each individual researcher to seek out literature from their specific area to find whether there have been any previous adjustments and how they were used. Essentially,

NISP is often left unadjusted so that there can be easier comparisons made between researchers and sites. In this study, I do not adjust NISP because it provides a quantitative representation of the number of fragments found per individual. If preservation was perfect,

MNI would equal NISP in gastropods and one half of NISP in bivalves.

Minimum Number of Individuals (MNI)

MNI is another important and long standing quantification technique in zooarchaeological analysis (Reitz and Wing 2008; Claasen 1998; Grayson 1984). The first uses of MNI are attributed to Stock (1929) and White (1952, 1953). MNI measures the fewest number of specimens (individuals) that it would take to create the collection of skeletal fragments. Reitz and Wing (2008: 206) state that “Critiquing MNI might be considered a growth industry among zooarchaeologists”, referencing 23 papers and books on the subject. It must be recognized that although there are various drawbacks with using MNI, it is still one of the best and most common methods used by zooarchaeologists to examine differences within and between assemblages (see Claassen 1998; Walker 1992:302). Reitz and Wing (2008) are correct when they note that zooarchaeologists are jumping on the bandwagon for critiquing MNI, though few unfortunately offer any “real” alternatives.

MNI adjusts many of the problems of NISP by accounting for skeletal differences and reduces issues such as fragmentation. Similarly, measurements of weight can vary significantly between taxa (as well as juveniles versus adults, for example); it is nearly impossible to know the number of individuals collected from each species based on weight alone. Although MNI can be quantified in various ways (Claassen 1998, 2000; Glassow

2000; Grayson 1984; Lyman 1994), it still provides a good approximation of the actual number of individuals that were collected and subsequently deposited in a midden.

The most common way to determine MNI is based on non-repetitive elements (NRE) or elements of an animal skeleton of which there are only one per individual(e.g., the spire of a gastropod). NRE-based MNI is founded on the theory that by counting a certain element

that has a set number of occurrences in an organism it is possible to determine the minimum number of individuals found within the collection. NRE MNI is also based on the fact that many species have skeletal elements that are symmetrical and can be separated into left and right sides of the body. Siding the remains reduces the number of elements by half. For some animals (such as fish), the vertebrae may be used to determine MNI.

To refine MNI estimates, Bökönyi (1970), Chaplin (1971:74), Klein and Cruz-Uribe

(1984), Flannery (1967:157), Smith (1975: 34), and Zeder and Arter (2007) suggest

“matching” or taking into account age, sex, and size when estimating MNI. Using these factors can enhance the MNI count by using obvious sexually dimorphic differences between individuals. For example, if there are 10 deer femurs in a collection, six of which derived from the left side and four from the right side, the MNI would be calculated as six since this is the highest non-repetitive element. However, if sex is taken into account, the six left femurs might produce a population of five females and one male, with the four right femurs representing three male and one female. Using both factors would alter the MNI calculation to eight, given that there are now five female and three male individuals. If the age of each femur is then taken into account, it could produce an even higher MNI. In practice, matching should be used to determine MNI.

When determining MNI, zooarchaeologists need to decide whether to look at an entire collection or separate them by layers or plana based on stratigraphy. If dividing the samples by layers, it is important to take into account the possibility that a single individual can be distributed over multiple layers due to a variety of processes, including bioturbation.

Artificial and natural layers should be treated differently, as artificial layers are more likely

to have individuals spread across the layer boundary. Simply adding the MNI from each layer will not always produce the same results as adding the counts of each element. As a total collection, adding element counts will produce a more accurate MNI than adding the layer MNI counts. For more on aggregate counts see Grayson (1973, 1984:31) and O’Connor

(2000). It should be noted that with an extremely large sample size of mollusks, such as that recovered and analyzed from Coconut Walk, that issues associated with aggregate counts are minimized. The identification of a mollusk using NRE-MNI forces each individual into one section and therefore the aggregate count is the same as the sum of the MNI count. Overall,

MNI provides a useful estimate of the minimum number of individuals, which can be compared between taxa. While it is important to remember the differences in taxa size and usefulness, MNI does provide a baseline for comparison.

Statistical Analysis

In zooarchaeology, as in other fields, statistics are used to calculate the significance of patterns, or the likelihood that observed patterns are artifacts of sampling strategies or random chance rather than reflections of past human choices or behaviors. Basic quantitative measures (sample statistics) that are determined for each collection include weight, MNI, and

NISP (Grayson 1984; Lyman 2009). Given their ubiquity, these are commonly referred to as the “known attributes” of the collection or sample. Statistics allows zooarchaeologists to take the known attributes from an excavation and use them to estimate unknown attributes.

Unknown attributes are created through sampling, which is often a necessity in archaeological research. Given time and monetary constraints, as well as the precedent for

leaving portions of sites for future excavation, sampling is a common technique in archaeology (Drennan 2009). In zooarchaeology, analyzing material is extremely time consuming and it is near impossible to analyze everything from an excavation. Thus, sampling becomes an important strategy which makes it possible to analyze a portion of a collection and statistically represent the expected results as if the entire collection was analyzed. The majority of statistics used rely on variation from a normal curve and the assumption of a normal distribution. It is in large part based on the bell curve, and the idea that every population will have a given distribution (variation) around the average of that population. Taking the variation into account, it is possible to use a sample to estimate the range or confidence interval (CI) in which a portion of the population will reside. The most commonly used range is 95%, or two standard deviations from the average.

In reality, the use of statistics allows zooarchaeologists to compare a larger array of material from different sites and contexts and still say something significant and meaningful about their similarities and variance. While time could be spent analyzing all of the material from an excavation, there are certain sizes of samples beyond which the results tend to not deviate from that which is expected in zooarchaeology. Thus, instead of knowing with certainty what a particular collection actually consists of, zooarchaeologists prefer to know the range of possibilities of numerous samples within an assemblage. Although depending on the research goals it may be worthwhile to examine the whole collection because the presence of rare species is expected to increase as the sample size increases.

Of course, there are limits to what statistics can accomplish. These limits are determined by the types of statistics used, which in itself is dependent on the number and size

of the samples used. Proportionally larger samples produce more reliable statistical results.

For example, if attempting to estimate the average height of a population of 100 people, the accuracy of the estimate increases as the number of people used for the estimate increases. A sample of 10 people is more statistically valid than a sample of 5; likewise with samples of

50 compared to 10. Eventually, a sample of 100 people would provide the actual average height of the population. The same holds true for any zooarchaeological estimate.

Testing Sampling Strategies

It is common in zooarchaeological analysis to conduct sampling procedures on large faunal assemblages for reasons that include time, labor, and expense (Lyman 2009). These considerations often prevent analysis of a complete assemblage, and the fact that statistically meaningful results can normally be achieved so long as there is sufficient sample size. In addition, only a portion of a site may be selected for excavation as a way to sample a spatially representative area. However, there has been a paucity of zooarchaeological studies that have analyzed faunal material (e.g., finfish, shellfish) from a large area (or large percentage) of a site in its entirety and tested whether more discrete units within that area would provide similar data based on a different hypothetical sampling strategy (e.g.,

Grayson, 1984; Cannon, 2001).

For example, would a sampling strategy of randomly selecting 25% of a given unit

(e.g., stratum, 1 m2, 1 m3) provide similar results as a judgmental sample where the archaeologist selects specific units based on a higher probability that something will be found? These are significant questions because the majority of zooarchaeological analyses

are based on the assumption that any random sample will be representative of what will be found in other parts of the site (or whichever unit of analysis is being used) (Grayson 1984).

In conclusion, I use the standard quantitative methods of NISP, NRE-based MNI, weight, and species count to analyze the zooarchaeological material. I also use hypothetical samples to test a variety of sampling strategies that could have been used at Coconut Walk.

In the next chapter, I present the specific parameters of the mollusk collection from Nevis, the questions I hoped to address through zooarchaeological analysis, and the adjustments I made to methods in order to suit the qualities of the sample and goals of the project.

CHAPTER 4: METHODS

In this study, the analysis is from a rich assemblage of mollusks recovered from a 5×5 m trench at the Late Ceramic Age (ca. AD 850-1440) site of Coconut Walk on the island of

Nevis. Excavation of the trench, located within a dense midden area, revealed more than

63,000 individuals (MNI) from 78 individual taxa, 58,000 of which were recovered from in situ deposits. Because the trench was organized into twenty-five 1×1 m units, and field crews recovered the material using several methods (e.g., troweling with no screening versus troweling with wet screening of different sized mesh), I am able to examine a number of issues in detail, including whether: 1) different sampling procedures were more effective in recovering various taxa; and 2) a selection of a subsample of 1×1 squares within the trench using different sampling procedures (e.g., random versus judgmental) would produce different results and henceforth, interpretation of past cultural behaviors. This research provides a means for assessing the percentage of a given area that should be sampled to produce statistically significant results in a zooarchaeological assemblage, and how different sampling methods within a larger area can affect the overall interpretation of mollusk diversity and equitability, subsistence strategies, and a host of other issues.

Archaeological Methods

Prior to excavation, the Coconut Walk site was organized into an arbitrary 5×5 m grid system which served as the basis for identifying a sample of trenches for excavation (Figure

10). Three 5×5 m trenches were excavated in 2010. Two of these (3073 and 2973) focused on re-locating one of the old Time Team trenches and finding further evidence of household

occupation, such as postholes, which had been observed by the Time Team. The third (2273) trench was opened in an area that appeared to contain a concentration of food remains and other midden constituents (Figure 10).

Excavation of Trench 2273 was subdivided it into twenty-five 1 m2 units that were numbered 1 through 25 (Figure 11). After removing a 5 cm layer of topsoil (planum 0), excavation continued in 10 cm arbitrary levels because the deposits represented continuous midden refuse deposition with few discernible strata. Excavation continued through four plana (40 cm total) until sterile soil was reached, creating approximately the same volume in each midden plana. The sterile subsoil pattern is typical of what is seen in many other

Ceramic Age sites in the Caribbean (e.g., Fitzpatrick et al. 2009). To facilitate recovery of smaller artifacts and ecofacts, four 1 m2 units (numbers 7, 9, 17, and 19) were wet sieved through 6 mm mesh using sea water. From each of these sieved square units, a 0.50×0.50 m column sample from the southwest corner was also screened, but through nested 6.0 mm and

1.6 mm sieves. The excavated material was then taken to the field laboratory where material was processed into general categories, bagged, and entered into the ArcheoLINK finds processing system (Kaye et al., 2010).

Figure 10. Map of Coconut Walk (after Kaye et al. 2010).

For the purposes of this study, only those remains recovered with 6.0 mm sieves or trowel are considered. The 6.0 mm screen and trowel materials are similar in size and material recovered, while the 1.6 mm material consists of fragments from the 6.0 mm material. To determine general mollusk consumption and test sampling strategies, it is not necessary to examine the material from the 1.6 mm screen. Though the finer sieved material would add to the overall weight, MNI, and species count, it would not significantly change the MNI or species count. In addition, the 1.6 mm screen material is only available for 4% of the total area, which prevents it from being used to analyze spatial sampling techniques.

Figure 11. Numbering convention of 1×1 m squares within a 5×5 m trench (the number 1 always represents the SW corner of a trench).

Zooarchaeological Methods Summary

All zooarchaeological material recovered through excavation was shipped to the archaeology laboratory at North Carolina State University where a crew of undergraduate and graduate student researchers conducted further identification, sorting, cataloging, and analysis. The shell recovered through excavation using trowel and 6 mm sieves produced more than 144 kg of material after sorting in the lab. Following traditional identification protocol, shells were sorted to the lowest level of taxonomic categorization possible (in most cases, down to species). Each independent taxon was counted using NRE-based MNI, NISP, and weight following standard procedures (see Reitz and Wing 2008; Ballbé 2005; Mannino and Thomas 2001; and Morrison and Cochrane 2008 for more information about counts using NRE-based MNI). I performed a quality control check of 12% of the bags (n = 19) following completion of sorting to ensure that the initial sorting was reliable. The only major issue I identified was an error in the identification of MNI and NISP of a chiton species

(Chiton tuberculatus). After the correction was made, MNI and NISP showed less than 1% difference between initial sorting and subsequent quality control. There was also a slight difference (1.3%) in the weight of species, but this was likely a result of soil continuing to exit the gastropods during handling and sorting. The majority of difference in MNI, NISP, and mollusk identifications came from identifying shell that had previously been listed as indeterminate. These differences are probably attributed to the limited experience of some sorters, especially when trying to identify smaller fragments, but this was later remedied by quality control checks using more experienced sorters.

In the following sections, I provide greater detail on the procedures for identifying taxa, determining MNI, NISP, weight, species count, and sample statistics. These provide a baseline for understanding why I emphasis some data sources over others in my interpretations of the results.

Identification of Taxa

The identification of taxa within this collection was split into two steps. The first step was identification of common taxa and creating a guide for the sorting of the rest of the collection. The second step was using the guide and other resources to identify the collection.

The purpose of splitting identification into two steps was to create a comparative collection for use in identification.

Initial identification was completed by Scott M. Fitzpatrick, Jessica Stone, Meg

Clark, and Christina Giovas. A few of the bags were sorted into individual taxa and then identified using published reference shell books and, for some taxa which were particularly difficult to identify, sending photographs of samples to Christina Giovas, Ph.D. candidate at the University of Washington who works with Fitzpatrick (Abbott 1995; Warmke 1961;

Harasewych and Maretzsohn 2010). To assist other sorters and create a list of basic taxa, a reference binder was created with photographs and descriptions of all of the common species found during initial sorting. A limited comparative collection in the lab was also helped with identification.

After the initial identifications, undergraduate students were allowed to sort bags of material under supervision. The sorting process took approximately two years, with dozens

of students participating in the process. Upon completion of sorting, taxa present in less than five bags were checked for proper identification to ensure that a misidentified specimen would not add to the species count. Identification was also checked through the quality control of the sorting process and found to be reliable. In total, 78 different mollusk taxa were identified.

Minimum Number of Individuals (MNI)

In this study, MNI was determined using a non-repetitive elements (NRE) method.

Analysis was done using each square as a sample; thus, samples were recorded to the species level and not to the individual level. In addition, the NRE-based MNI was calculated using one element that was determined to have the best representation for each type of taxon. Some researchers suggest recording all of the elements for mollusk species and then finding the largest to record for a more accurate aggregate MNI (Giovas 2009), but based on the size of our collection, I determined there would not be an improvement in accuracy significant enough to justify the enormous additional investment in time (Mason et al. 2000). NRE- based MNI solves any issues associated with aggregation of MNI across the units sampled because the NRE can only be counted once (Grayson 1984; Mason et al. 2000). Giovas

(2009) found a 14.8% discrepancy between traditional and NRE-based MNI techniques when counting C. pica (MNI=52) and a 6.9% difference with N. tessellata (MNI=175). But, if using an aggregate MNI count, the difference was reduced to 10% for C. pica and 0.5% for

N. tessellata. While 10% is significant, it comes from one sample (MNI=39), with the other six samples having the same aggregate MNI and NRE-based MNI count. The smaller

samples used by Giovas (2009), for example, could be expanded to a larger sample to test whether this phenomenon holds true.

Traditional MNI would entail recording all of the elements and would require identifying 7 or 8 elements for each shell. To accurately record this, the sorter must note the presence/absence for elements on each shell, and then calculate which element gives the highest MNI. To record this information it requires the identification and numbering of each shell, which was beyond the resources and scope of this project.

The method used in this study allows sorters to quickly count the number of shells that have one specific element and record the total count as the MNI. Since there are over

58,000 MNI in this collection, if every element were counted, it would translate to over

400,000 elements being recorded without counting the extra fragments. Because this collection is extremely large, it was not feasible to invest the time and resources on a process that has not been proven to produce significantly different results in mollusk identification.

Future research in this area could improve the understanding of the differences between

NRE-based and traditional MNI in shell middens. Furthermore, because the methods used throughout the trench are the same, comparisons within the trench will be accurate. If this sample was small or had relatively few mollusks to count, a more thorough MNI count would be required to ensure adequate results.

Bivalve MNI was counted using the beak (also known as the umbo) as the NRE (see

Figure 12), which are the most likely to survive and be recovered. There are two beaks per individual, a left and right side. To speed up processing, the total number of beaks were counted and divided by two instead of separating and counting by side. Just counting beaks is

a shortcut that may have slightly altered the results, but it reduced the time spent counting in half. During quality control, I tested two bags to see how the technique would alter results.

Counting by side produces a MNI of 12 and 44 compared to 11 and 44 for the method used in this study. The difference in MNI suggests a small difference in results. The same method was applied throughout the collection, providing a good comparison of results.

Figure 12. Parts of a bivalve shell (fao.org).

To determine MNI for gastropods, the apex was counted as the NRE (see Figure 13).

There is one apex per individual, and the apex is the most likely piece of the shell to survive

(Giovas 2009). The apex also allows for identification of species (in most cases), whereas other parts of the shell (inner spire, lip) can be impossible to identify to a particular taxon.

Counting the number of apexes for each species produces the MNI.

Figure 13. Parts of a gastropod shell (Schultz 1953).

Figure 14. Parts of a chiton shell (palaeos.com).

The third type of mollusk found in the collection is polyplacophora. There was only one species identified within this class (Chiton tuberculatus). They consist of eight plates or pieces per individual (see Figure 14), with each connected by living tissue and having distinct characteristics. As with bivalves, a shortcut was taken to speed up processing. Each plate has only one mucro or apex, creating eight per individual. The number of mucros and apexes were counted and divided by eight to create MNI. The most accurate way to calculate the MNI for chiton is to count all five elements for each plate. The plates are divided into three subgroups: anterior valves, interior valves, and posterior valves. Each individual has one anterior, one posterior, and six interior valves. After recording all five elements for each valve, the MNI is determined by the highest count. Other NRE based MNI counts include counting the anterior or posterior valves as MNI.

Using either the posterior mucro or one eighth the mucro and apexes can cause issues if there is a bias in that one valve. For example, Giovas (2009) found that using NRE-based

MNI (MNI=87) of the posterior mucro had a 13.9% discrepancy compared to traditional

MNI (MNI=101). Likewise, using her sample, the technique used for this study produced and

MNI of 86. Dividing NISP by 8 to determine MNI (MNI=103) would have produced a 2.0% discrepancy from the traditional MNI. As noted before, this was a small sample and a larger study would be necessary before adopting new methods.

By using an average of all of the valves, the NRE-based MNI used in this study produces a reliable and conservative estimate that is not biased by 1/8 of the individual. For instance, if using the anterior valve for a NRE based MNI, and 40 of these valves are along a division between squares, then there could be a large swing in MNI depending on how the

person excavating divides the squares. But, if using all of the valves, the error would be reduced by 8 to only 5 MNI. Although an exaggeration, using this NRE-based MNI produces a more conservative estimate that reduces the possibility of clumping affecting results. Due to the size of this collection and the large number of Chiton fragments (13,000+), NRE-based

MNI was deemed sufficient for analysis.

Number of Identified Specimens (NISP)

The methodology behind counting NISP is more standard than MNI. NISP is the number of fragments that can be identified to a particular taxon. In order to be included in the

NISP count for any species, a fragment must be identifiable. Whether a fragment is identifiable will depend on the features necessary to identify each species and in some cases, the skill of the sorter.

Fragments that cannot be identified to a particular species will be identified to the highest taxanomic classification possible. For example, some small nerite fragments may not be differentiated between N. tessellata and N. versicolor and thus are identified to the next highest level of the genera Nerita. If a fragment is unable to be identified to genera, an attempt is made to identify it to family or class. It may be possible to determine if large fragments of nacre are from bivalves, for example, but they may offer no clue as to the species. The same is true for pieces of gastropods such as spires or pieces of nacre. There are also pieces that cannot be identified to even higher taxonomic classifications and must be labeled as indeterminate shell. Often, shell that cannot be identified consist of tiny fragments of nacre. As such, NISP was not counted for indeterminate shell because it would have

artificially inflated the NISP count with tens of thousands of uninformative fragments.

Overall, NISP was able to be counted at the species level with only 0.9% of the total NISP being attributed to genera or family.

Weight

In this study, weight represents the total weight of shell collected in each unit and provides an approximation of the total amount of shell that was deposited, though it must be remembered that they are subject to digenesis and other natural processes (Grayson, 1984;

Zuschin et al., 2003). The weight of mollusks recovered was measured to the hundredth of a gram for each taxon within each discrete unit (plana and square). All fragments of a taxon from a given discrete unit (plana and square) were weighted together to give a total weight for each taxon in each unit. Weight was measured after sorting was completed and any remaining dirt was removed. However, it is extremely difficult to completely remove all dirt from gastropods, in particular, because the interior whorl cannot often be easily accessed without breaking. As such, the weight of the mollusk assemblage, and especially whole gastropods, is slightly inflated.

Species Count

The species count for each unique square and planum combination was calculated to determine the diversity of species found within each unit. The entire trench produced a minimum of 78 identifiable and unique species. The species count for each unit was determined by calculating the sum of the number of unique species within that unit.

Unidentifiable shell was ignored, but shell identified to family was taken into account. For example, if a unit did not have any Nerita species such as N. tessellata, but there were fragments identified to the genus Nerita, this would count as a species for that unit because it can be assumed that there is at least one Nerita species within the genus identified in in this unit. The same calculation was also applied to plana within units, plana across the entire trench, and all plana and units combined.

Statistics

The statistical treatment of the data began with tests to determine whether all variables in the different samples followed a normal distribution. I conducted all statistical analyses using IBM SPSS Statistics 20. Based on a Kolmogorov-Smirnov Test, the distributions of each variable in each plana, including both sieved and troweled, are normal

(p < 0.05). Once I had determined the normality of the data, I calculated the average and standard deviation for all variables for three key species (N. tessellata, C. pica, and L. tuber) in all plana, units, and the trench to provide a basis for comparisons of spatial and temporal patterns.

Sample Statistics

To analyze trends in the data set, a group of key sample statistics was used for comparison. Midden composition was analyzed using the total weight, MNI, and NISP of each unit. The key species (N. tessellata, C. pica, and L. tuber) had additional sample statistics that are relevant to each species. As with midden composition, species composition

includes weight, MNI, and NISP. Also included are individual weight and percentage of unit weight, MNI, and NISP. For each of these species, I calculated individual weight by dividing weight by MNI. While this measurement does not give the true weight of each individual, it provides an estimate of the weight of shell collected per individual. Theoretically, this should be similar throughout the trench unless there are differences in preservation or consumption practices.

Finally, the sample statistics of percentage of unit total provide a relative measurement for comparison across time and space. Percentage of unit total can be calculated for a variety of units, with the lowest being per square for each planum and the highest being the entire trench. The importance of measuring the percentage of the total unit is the understanding of how mollusk composition changes. For instance, if Nerita tessellata consists of 45% of the unit weight on one planum, and 65% on the adjacent planum, there is a dramatic rise in the amount of N. tessellata compared to other species. Using percentage avoids having to consider the total amount of shell collected, which can be altered by human population size and consumption. Taken with measurements of absolute weight, the percentage measurements allow for a better overall picture of the change in species composition over time and through changes in consumption. Percentage of unit weight is measured by dividing the weight of each species by the total weight of the unit, providing a relative weight comparison to everything that was collected in that unit, unlike species weight which provides an independent measurement. Percentage of unit MNI is measured by dividing the MNI of each species by the total MNI of the unit. The result is a measurement of relative species consumption that can be biased by species size. Likewise, percentage of unit

NISP is measured by dividing the NISP of each species by the total NISP of the unit. A bias can be created toward species with high fragmentation rates.

Testing Sampling Strategies

Given that the interpretation of past cultural behaviors hinges on a sample being representative of the larger population, it is critical that we evaluate the phenomena that can affect the usefulness and robustness of a sample. In this study, I analyzed how sampling strategies within a larger 5×5 meter trench could alter the population estimate. To examine sampling strategy, I compared the weight, MNI, NISP, and number of species.. A selection of hypothetical samples was selected for additional analysis (shown in Figure 15).

There are two types of techniques for sampling; random and non-random. Random sampling involves selecting samples at random (Samples 6 and 7). A non-random sample has some element of judgment involved and is typically divided into four categories; stratified random, cluster, systematic, and judgmental. A stratified random sample is performed by dividing the sampling area into smaller regions from which random samples are taken

(Samples 4 and 5). A cluster sample also divides the sampling area into smaller regions but then a region is randomly chosen and the entire region is sampled. A systematic sample is defined by taking a sample in a predefined pattern, such as every 5 meters. Judgmental samples are taken based on the judgment of the researcher (Samples 1, 2, and 3).

The sampling strategy used during excavation of Coconut Walk in 2010 was a combination of cluster and judgmental sampling. The site was divided into smaller regions

(5×5 m trenches) of which three were entirely excavated (cluster sampling). The regions

excavated were not chosen randomly but selected judgmentally based on archaeological material on the surface and previous excavation.

Figure 15. Hypothetical samples selected for analysis in a 5×5 m trench.

In conclusion, I have explained the methodologies I used to determine the quantitative measurements (MNI, NISP, weight, and species count) of mollusks at Coconut

Walk and how I tested sampling strategies. The next chapter will go through the results for the quantitative measurements and sampling strategies, answering how mollusk exploitation changed over time at Coconut Walk and how a different sampling strategy would have altered the results.

CHAPTER 5: RESULTS

In this chapter I present the results of my two sets of questions – the first looking at change over time in mollusk consumption at Coconut Walk and the second focused on determining the best sampling strategies for these particular conditions. Based on the methodologies established in the previous chapter, the results are explained in the following sections.

Mollusks

Trench 2273 at the Coconut Walk site produced a wide variety of mollusks for analysis. Table 3 shows the sample summary statistics for square composition for each planum. Table 3 is further broken down by square in Appendix A, Table 26. Using the totals from square composition; it is possible to detect changes in abundance over time and space.

It is expected that chronologically, planum 1 is the youngest and planum 4 is the oldest.

Based on evidence of pottery and the absence of discernible layers, each planum is assumed to be similar in age across the trench provided that the deposition process was symmetrical across the entire trench area (Fitzpatrick, personal communication). Each planum and square represents approximately the same soil volume.

The use of species can enable us to further analyze how square composition changed over time. Comparing the weight and MNI of each species collected from each planum shows a relative increase in consumption for each species. Table 4 shows the weight, MNI,

NISP, number of squares/planum the species is present and relative percentages of the top nine species (Appendix A, Table 27 shows all species).

Table 3. Weight, MNI, NISP, species by planum.

Planum Weight MNI NISP Species 1 68280.56 25554 35997 61 2 49894.81 20009 30628 62 3 19984.38 10708 15110 48 4 3035.38 2346 3431 32 Total 141195.13 58617 85166 79

Table 4. Quantitative statistics for top nine species at Coconut Walk.

Species Weight % MNI % NISP % Squares/ % Planum Present C. pica 54003.9 38.2 1063 1.8 6450 7.6 97 99.0 N. tessellata 30375.4 21.5 37591 64.1 38804 45.6 98 100.0 L. tuber 14588.78 10.3 973 1.7 2841 3.3 91 92.9 S. gigas 5883.625 4.2 100 0.2 464 0.5 68 69.4 C. tuberculatus 5109.005 3.6 1735 3.0 13486 15.8 96 98.0 N. versicolor 3732.33 2.6 5060 8.6 5105 6.0 98 100.0 T. excavate 3350.97 2.4 3505 6.0 4364 5.1 96 98.0 D. denticulatus 1603.52 1.1 2367 4.0 5019 5.9 96 98.0 T. muricatus 1293.01 0.9 1804 3.1 2698 3.2 91 92.9

To determine changes in the harvesting of each species based on the overall consumption found in the assemblage as a whole, I analyzed percentage of unit weight and

MNI for each species. While comparing the weight of each species collected from each planum shows a relative increase in consumption for each species, comparing the percentage provides results in changes in overall consumption based on percentage of MNI (Table 5).

The same is true for MNI. Percentage of MNI for the top nine species is shown in Table 5

(Appendix A, Table 28 shows all species).

Table 5. Percentage of MNI for top nine species.

Planum Species 1 2 3 4 Total C. pica 1.91 1.88 1.53 1.49 1.81 N. tessellata 66.39 63.27 61.53 58.70 64.13 L. tuber 1.85 1.45 1.65 1.36 1.66 S. gigas 0.11 0.13 0.11 0.00 0.11 C. tuberculatus 2.67 3.51 2.43 3.84 2.96 N. versicolor 9.40 7.65 8.74 8.18 8.63 T. excavate 5.10 6.50 6.81 7.42 5.98 D. denticulatus 1.93 4.74 6.98 7.59 4.04 T. muricatus 3.17 3.24 2.64 2.64 3.08

To further analyze changes over time within the trench, key species (N. tessellata, C. pica, L. tuber) were analyzed for each square and planum. These species provided information regarding how the three main dietary species had changed over time and space.

Table 6 shows the sample statistics for the three key species by planum (the same information is shown by square and planum in Appendix A, Table 29).

Table 6. Weight, MNI, and NISP for N. tessellata, C. pica, L. tuber by planum.

Nerita tessellata Cittarium pica Lithopoma tuber Planum Weight MNI NISP Weight MNI NISP Weight MNI NISP 1 13486 16966 17382 29132 487 3612 6457 474 1340 2 10482 12659 13196 19550 377 2046 4942 290 944 3 5377 6589 6742 4834 164 661 2923 177 472 4 1030 1377 1484 488 35 131 266 32 85 Total 30375 37591 38804 54004 1063 6450 14589 973 2841

Trends

Analyzing trends that occurred in the data through time require several different types of statistical analysis. Because the data was found to be normal (normal distribution), I was able to use a t-test to examine whether the data sets were equal or different. For the purpose of this study, the units or samples come from the 5×5 meter trench subdivided by squares and plana. Average weight, MNI, and NISP all significantly change over time (Figure 16, 17, and

18 respectively). The number of species found in each square slightly increases over time in this sample, but it is not significant enough to predict similar results for the entire population

(Figure 19).

Figure 16. Weight for each square by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Figure 17. MNI for each square by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Figure 18. NISP for each square by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Figure 19. Number of species per square by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

To enhance the statistical tests, the use of a paired t-test was used. A paired t-test is often used to see how individuals change after a treatment. For example, if studying the effects of exercise on blood pressure in individuals, a paired t-test would enable researchers to pair up each individual’s pressure before and after an exercise regimen. A normal t-test would just take the averages of the group before and after, which fail to account for differences in individuals. The trench analyzed in this study is assumed to be homogeneous, but there is likely to be some variation spatially. Because of this, we treat each square as an individual and each planum as a before and after measurement. Using this strategy, it is possible to see if there are significant changes in sample statistics over time (between layers).

Using a paired t-test, overall square composition and the key species were analyzed for change over time.

The paired t-tests reveal that overall square composition changes drastically over time

(Table 7). A downside to using a t-test is the inability to see trends in data such as a gradual increase or decrease that is within the standard deviation of the data set. These inabilities can be overcome by using the trench as the population, thus eliminating the uncertainty of not knowing the actual average and standard deviation. Using the trench as the population we can make statements about this trench, but not about any larger populations such as the site or region.

Table 7. Paired t-tests for overall square composition.

As the results show, there were significant changes in average square composition over time. Anything with significance over 0.05 can be viewed as evidence of sampling error or evidence that there is not enough evidence to suggest a difference between the samples.

The only results with significance over 0.05 were with number of species and NISP between plana 1 and 2. I expected these results as planum 1 and 2 have a total species count of 60 and

61, respectively. NISP was also similar between the two plana, although planum 1 was higher on average.

Similar to the overall square composition, the three key species show a change in weight, MNI, and NISP over time. The main quantifications are shown for N. tessellata

(Figures 20, 21, 22) and C. pica (Figures 23, 24, 25).

Figure 20. N. tessellata weight by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Figure 21. N. tessellata MNI by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Figure 22. N. tessellata NISP by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Figure 23. C. pica weight by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Figure 24. C. pica MNI by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Figure 25. C. pica NISP by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Using paired t-tests with key species for changes in weight, MNI, NISP, and percentages of all three produced similar results. For N. tessellata there is significant change for weight, MNI, and NISP between each planum (Table 8). C. pica had similar results

(Table 9), except weight (sig. 0.113) and MNI (sig. 0.063) between planum 1 and 2 produced results slightly outside the confidence level. L. tuber showed that only weight between planum 1 and 2 (sig. 0.117) was outside the confidence level (Table 10).

Testing more advanced statistics (percentage of weight, MNI, and NISP) produced mixed results. These are shown for N. tessellata (Table 11), C. pica (Table 12), and L. tuber

(Table 13). The only definite change occurs between planum 2 and planum 3 for percentage of weight, MNI, and NISP for all three species. The other measurements vary highly between species and do not show any specific trends. However, when testing this change over a larger

time period than from planum to planum, results were more significant. Aside from planum 2 to planum 3, planum 1 to 3, as well as planum 1 to 4, show significant differences. The differences suggest that there was a major shift in mollusk consumption between planum 2 and 3.

Table 8. Paired t-tests for N. tessellata weight, MNI, and NISP.

Table 9. Paired t-tests for C. pica weight, MNI, and NISP.

Table 10. Paired t-tests for L. tuber weight, MNI, and NISP.

Table 11. Paired t-tests for N. tessellata percentage of weight, MNI, and NISP.

Table 12. Paired t-tests for C. pica percentage of weight, MNI, and NISP.

Table 13. Paired t-tests for L. tuber percentage of weight, MNI, and NISP.

N. tessellata provides an interesting comparison of changing individual weight over time. There are a number of ways to estimate individual weight using a large sample. The most time consuming is to measure each individual or a sample of individuals. This study looks at methods that can be applied to most excavations without having to perform any additional shell measurements. First, a simple plot of individual weight over time was performed (Figure 26). The plot shows a slight increase in size average weight/MNI for each square over time. Plots of weight versus MNI for each square were then plotted using all of the squares (Figure 27) and only the 6.0 mm sieved squares (Figure 28).

Figure 26. N. tessellata individual weight by planum. Note: Results are sorted by recovery technique with squares sieved (ss) representing Y and squares troweled representing N.

Figure 27. Square weight vs. MNI for N. tessellata by planum.

Figure 28. Sieved square weight vs. MNI for N. tessellata by planum.

As can be seen, the plots in Figures 27 and 28 are similar. It is evident in Figure 27 that planum 4 has a lower individual weight than the other three. When analyzing the squares with more accurate collections of the species (sieved squares), a better separation appears, with an increase in size from planum 4 to 1 through time. To analyze this trend in more detail, individual weight was calculated through a number of techniques. The results are similar to those found when Giovas et al. (2013) measured the length and width of a sample of N. tessellata shells from this collection. They found around a 5% increase in length and width of shells from planum 4 to 1.

The distribution of size within living populations of N. tessellata is assumed to follow the bell curve. Thus, as more individuals are sampled, the sample average approaches the

population average. Using this theory, the sum of weight was divided by the sum on MNI for each planum to produce an individual weight. This reduces the distribution to its average, but it is assumed that the more individuals that are sampled, the closer their average moves to the population average. Thus, the average created for the entire planum should represent the population average. There are differences in the collection of smaller individuals between the two methods used in this study, necessitating that the combination of all squares, sieved

(sample) squares, and troweled (non-sample) squares must all be analyzed. The results for N. tessellata for individual weight (Table 14) show an increase in weight over time similar to those found by Giovas et al. (in press). The taxon increases in individual weight by 9.0% for all squares, 3.5% for troweled squares, and 14% for sieved squares. Part of the increase in weight over time could be attributed to taphonomy (the increased time between deposition and excavation for the older layers) though a 600 year period of time is not likely to produce these results and it is concurrent with an increase in MNI and NISP. Based on the results from Giovas et al. (2013), a 5% increase in weight is expected in the sieved squares. The results I found show a 14% increase, of which 9% is attributed to taphonomic processes.

Using the difference between the findings as a baseline for the percentage of weight lost from planum 4 to 1 due solely to taphonomy, I was able to analyze other species. C. pica shows a similar trend of increase over time in both individual weight (Table 15) and fragment weight

(Table 16). The last key species, L. tuber, shows a slightly different trend, with an increase and then decrease back to original sizes in individual weight (Table 17) and fragment weight

(Table 18).

Table 14. Individual weight for N. tessellata by planum.

Table 15. Individual weight for C. pica by planum.

Table 16. Fragment weight for C. pica by planum.

Table 17. Individual weight for L. tuber by planum.

Table 18. Fragment weight for L. tuber by planum.

Sample Selection

Determining which sampling techniques to use can depend on the questions the research is asking and the results they hope to find. I determined that there are two groups of results that need to be statistically valid for any samples. In the following sections, I provide the results for two groups of quantification: overall composition (weight, MNI, NISP, species count) and key species (N. tessellata, C. pica, L. tuber). It was determined for each quantification that 16% (4 out of 25 squares) was required to ensure a statistically significant result. Of the seven hypothetical samples, none had basic square composition factors outside of one standard deviation from the trench average and only two have a species quantification that meets this criterion.

Weight

Results for weight (Appendix B, Table 30) are based on sampling all possible combinations of certain sample types within Trench 2273. Every sample that produced an error was above the trench range. From this it is shown that 20% of the 1×1 samples produce an average weight outside of the trench average for at least one planum. If sample size is

doubled to two 1×1s, the percentage drops to 1.67%. Tripling reduces this to 0.09%, and sampling four 1×1s reduces the percentage of samples outside the range to 0%. Interestingly, samples that are clustered together (i.e., 2×1 or 3×1) produce a higher error rate. The increased rate could be due to the significantly smaller number of possible samples, but is most likely a result of the clustering of high numbers (i.e., deposition of material may not often be continuous across a larger area).

When I analyzed the selected samples, the weight averages were close to the trench average, but changed over time was significantly different for some samples. The samples in

Appendix B, Table 34, show a variation of weight that is an average of 0.22 standard deviations from the trench average. The largest difference is a 24% increase from the average of 4.937 kg to 6.124 g for the sample of corner squares 1, 5, 20, and 25. Extrapolated to represent the entire trench, it relates to an increase of 30 kg to 123 kg to 153 kg. Another issue in interpretation is the relative proportions of plana to each other. The averages suggest that weight slightly decreases from planum 1 to 2 (2.354 kg to 1.816 kg). Two of the samples create a reduction of approximately half the weight from planum 1 to 2 and one creates a significant increase. If any of these samples were selected to represent the entire population, I would interpret the change of the midden over time differently than if I had excavated the entire trench.

Minimum Number of Individuals (MNI)

The results from the theoretical samples show an overall reduction in samples that are outside the CI for MNI compared with weight (Appendix B, Table 31). As can be seen in

Table 34, there are larger variations in the results of MNI between samples than with weight.

The samples are, on average, 0.44 standard deviations away from the average, but none exceed one standard deviation. The largest error is a 39% increase in MNI from 43,033 to

59,688 when extrapolating from a sample of the corner squares 1, 5, 20, and 25. As a result, if I had based an interpretation of mollusk consumption using this sample, the MNI would be significantly higher than its actual count. Likewise, two samples create interpretation issues between planum 1 and 2, with results suggesting an increase in MNI between successive plana.

Number of Identified Specimens (NISP)

Statistical analysis of NISP (see Appendix B, Table 32) shows similarities with MNI.

There was only one sample that differed between the two quantification methods. With the samples in Appendix B, Table 34, there are similar results to MNI again. The samples are an average of 0.43 standard deviations from the trench average, with none over one standard deviation. Also, the sample of squares 1, 5, 20, and 25 is 38% above the trench average and two samples produce an error of planum 1, having a lower MNI than planum 2.

Species Count

A count of the number of species found in each sample is important because it influences numerous conclusions about the heterogeneity and equitability of taxa. For this study, I compared the number of taxa found in each sample to the average number found for each planum to estimate species diversity. While the number of taxa is not a primary source

of dietary information, it can provide insight into the prehistoric ecosystem and presence of species. Results can be seen in Appendix B, Table 33. In the study samples, the number of species found is relatively uniform; a pattern which is shown in Table 34 with the selected samples. Although the samples produced similar results, Table 22 shows that the species recovered can vary greatly between samples. A total of 78 species were found in plana 1 through 4. The samples consistently resulted in 42 to 48 species recovered (54% to 62% of the total). While the species not recovered in the samples were not significant contributors to diet, they could be valuable in environmental reconstruction and in assessing species richness within the site. A larger sample (including smaller mesh sizes) may be needed to recover a more comprehensive sample of species.

Key Species

To further analyze the effects of sample size and strategy on the results, it is important to look at the most common or ‘key’ species in those samples. For the purposes of this study, I selected three species with different characteristics that could result in errors with small sample sizes. Ideally the selection would be: 1) a high weight/high MNI species;

2) a high weight/low MNI species; and 3) a very common species that is absent in some planum/square combinations. Using this combination, it is possible to examine how large and small species are represented throughout the trench in various sampling strategies. Also, this creates a good comparison to species found at various sites around the world.

Nerita tessellata

N. tessellata contributes 64.1% of MNI (37,591), 45.6% of NISP (38,804), and 21.5% of weight (30.375 kg). It was typically found with at least 90% of the original shell in place and often had a 1:1 MNI to NISP ratio. It is the ideal species to represent the high weight/high MNI mollusks as N. tessellata weighs an average of 0.80 grams per shell and is found mostly intact with little fragmentation (MNI and NISP are almost equal). To analyze variation between samples, I used five quantification methods for N. tessellata. Weight and

MNI are basic counts, with their percentages representing the amount of the total square weight or MNI that the species represents. Weight divided by MNI provides an average weight of each individual. Because there is very low fragmentation in this species, a high completeness, and a very high MNI, it creates an accurate reflection of the average weight of each individual and can be used to estimate shell size and biomass. From my observations, almost all (95+ %) of the N. tessellata shells are at least 90% complete. Appendix B, Table

36 reveals the results of different sampling strategies for N. tessellata. There is a dramatic increase in the percentage of samples that give results outside the 95% confidence interval.

Notably, using a 1×1 sample produces an error in at least one quantification method on one or more planum in 56% of the samples. As with the basic sample quantifications, there is a drastic decrease to 8.3% as the sample size is doubled, with it decreasing to less than 1% for three 1×1s and 0% with four 1×1s.

In the case of Coconut Walk and many other Pre-Columbian sites in the Caribbean, it is important to have an accurate representation of N. tessellata in a sample given its relative ubiquity, especially if the purpose of study is to understand how this species has changed

over time due to human harvesting pressures and/or environmental fluctuations. It was difficult to determine a sample size that accurately represents the trench as a whole. The cause is unknown, but is likely attributed to variation throughout the trench or the small increase in size (~5%).

The selected samples produce varying results that would have changed the results I found. Using this collection of material, Giovas and colleagues (2013) have shown that the species is getting larger over the 600 year period of occupation. At Coconut walk, the average weight divided by MNI of N. tessellata in troweled squares for plana 1 through 4 was 0.8625, 0.8559, 0.8552, and 0.8099 grams, respectively. The sieved samples recovered a larger percentage of the species, including smaller individuals, and produced average weight results—when divided by MNI—of 0.8030, 0.7792, 0.7523 and 0.7003 g. This leads me to conclude that N. tessellata is increasing in weight over time by 6.5% and 14.7%. If a sample of squares 1, 5, 20, and 25 was used to represent the entire trench, I would have come to the conclusion that human populations caused the species to reduce in size over time. The hypothetical sample results in averages for plana 1 through 4 of 0.69, 0.87, 0.98, and 0.88 g, presenting a dramatically different interpretation. In this sample, there is a decrease in average weight of N. tessellata by 21.6% from planum 4 to 1. An opposite affect can be seen in a sample of squares 3, 11, 15, and 23, producing results from plana 1 through 4 of 0.96,

0.86, 0.83, and 0.76, suggesting a 26% increase in size over time. As this demonstrates, selecting a sample that will represent a specific population is vital. Based on which sample is used to estimate what should be found in the entire 5×5 trench, I would conclude different results in the size change of N. tessellata over time. In contrast, had I examined the complete

N. tessellata assemblage, or selected a suitable (statistically significant) sample comprised of specimens across non-adjacent 1×1 squares, I would conclude that the species actually increases in size through time, a finding mirrored by the Giovas et al. (2013) study which used a sample size of more than 2710 individuals from sieved squares, representing 4.6% of the total taxon identified in the overall mollusk assemblage.

Cittarium pica

The next species I selected was the West Indian top snail, a large gastropod that was one of the more common species exploited by peoples in the Caribbean prehistorically and that is often found fragmented given its more friable nature. This taxon accounted for a high percentage of weight (38.25%), low MNI (1.81%), and average NISP (7.57%) in the assemblage. Appendix B, Table 37 gives results for C. pica. The error rates are very similar to N. tessellata, with the exception of 3×1 and 4×1 samples. In this case, one sample (squares

23/24/25) produced a weight and NISP above the 95% confidence interval. The same sample with the addition of square 19 causes the only error with a sampled four 1×1 squares. The four squares are all connected; therefore, the error rate for any four 1×1s that are not connected is 0%.

Lithopoma tuber

The last species I examined was the green star shell, a medium sized gastropod that is the third most common species found at Coconut Walk, but was not found in every planum of every square. Using the absence allows me to look at how species which are common, but

inconsistent, are represented in a collection recovered using different sampling techniques.

Overall, L. tuber contributed 10.33% of the weight, 1.65% of the MNI, and 3.34% of the

NISP of the total mollusk collection. Appendix B, Table 38 outlines the results which shows that the error rate for L. tuber is similar to N. tessellata, in that it drops below 1% with three

1×1s and to 0% with four 1×1s.

By comparing samples to the trench average, I was able to identify statistically significant variation within the sample. Appendix B, Table 34, demonstrates the outcomes based on specific samples which were selected using a judgmental, stratified random, and random sampling procedures.

Key Species

The three species selected in this study—N. tessellata, C. pica, and L. tuber— represent 70.09% of the total weight, 67.60% of the total MNI, and 56.47% of the total NISP.

Therefore, it is critical that they make up a comparable amount of each sample, particularly relevant when one considers that these provide the bulk of mollusk (and protein) consumption. Testing for this reveals a much larger percentage of samples outside of the

95% CI than with the basic sample composition. For all three species, I see a 52% error rate with 1×1 samples, more than twice as many errors compared with basic sample quantification. The trend continues with two 1×1s, where the error rate increases from 4.7% to 10.3 - 13.7%. As the sample size increases to three adjacent 1×1s, so does the increase

0.09% to .43%, though both are relatively minuscule. The trend of connected squares being more likely to produce an error was only partially supported by species composition. L. tuber

had no support for this claim and N. tessellata had only a 2.5% increase with two squares connected. C. pica had increases when two or three squares were connected. I encountered one combination of four squares that was 2.01 standard deviations above the trench average for NISP of C. pica in planum 2. This is out of a possible 12,650 combinations of four squares, a 7.904 E-05 error rate. In addition, the four squares were connected as a 3×1 with an extra square below the middle.

Summary The results for sampling strategies suggest that a sample of at least 16% of the surface area of the trench was required for weight, MNI and NISP to be within the 95% confidence interval of the trench average. Two trends became apparent: as a sample increases in size it becomes closer to the population average and separation between samples decreases deviation from the population average. Hypothetical samples of four squares produced results that were all within one standard deviation from the trench average for weight, MNI, NISP, and the key species.

Mollusk exploitation at Coconut Walk increases in intensity over the 600 year period of occupation. In the three preferred species, size appears to be increasing over the same period of time. Thus exploitation of mollusks at Coconut Walk appears to have been sustainable for the inhabitants during the use of this midden.

CHAPTER 6: DISCUSSION: FROM REPRESENTATIVE SAMPLES TO HUMAN DIET

In this section I consider the implications of my findings for understanding human impact on the environment of Nevis and lessons of sampling strategy. The results from sampling strategy suggest that the method of sampling (random, cluster, judgmental, etc.) does not significantly change the results but sample spacing and size are influencing factors in the accuracy of a sample. Contrary to the theoretical expectations of over-exploitation, exploitation and size of preferred mollusks increases through the occupation period with no archaeological evidence for over-exploitation. In the following discussion, I consider alternative explanations for this pattern based on comparisons with other island environments and subsistence strategies.

Sample Size

By necessity, archaeological investigation often requires that the remains of the past be interpreted based on only a small sample of the available dataset or area under investigation. In zooarchaeology, there are several recovery strategies which are considered to produce a sufficient size leading to statistically significant results (Reitz and Wing 2008;

Claasen 1998; Grayson 1984; Balme 2006).

Furthermore, while there is a plethora of information on how to conduct sampling procedures (e.g., see Ames and Lyman 2004; Ambrose 1967; Bowdler 1983; Grayson 1984;

Peacock 1978; Trinkley and Adams 1994; Waselkov 1987), there is a relative dearth of information on how different sampling strategies in faunal assemblages can statistically affect the conclusions of the research. Some believe that a fixed sample size can be applied

worldwide (Bailey 1975; Bowdler 1983; Greenwood 1961; Peacock 1978), though this strategy has not been considered a viable methodology by most scholars. Several zooarchaeologists have attempted to develop a specific threshold based on species count and/or MNI (Wing and Brown 1979:120; Quitmyer 1985:34; Walker 1992). However, there are very few examples of large scale excavations where a given assemblage is not sampled, but analyzed as a whole to test the efficacy of different recovery and sampling strategies.

Although the larger 5×5 m sample selected for analysis is inherently biased given that it derives from an already known midden deposition area, the results of this study still reveal some important considerations when selecting a sampling strategy. By analyzing a mollusk assemblage in its entirety from this larger 25 m2 area, divided into individual 1×1 m squares, it was possible to take various samples (or subsamples) to test how different sampling procedures would influence our interpretation of human foraging behavior among other issues. Sample differences were divided into two main categories: square composition and species composition. The difference in square composition between different sampling techniques shows the necessity of using a reliable sample size. Overall, 24% of the 1×1 samples produced at least one error for at least one planum in weight, MNI, NISP, or species count. The percentage decreased to 4.67% with two 1×1s, 0.09% with three, and 0% with four. In other words, if we selected square 4 to represent our hypothetical unit of analysis (a

1×1 m unit of a larger area not being an uncommon strategy in archaeology), there is almost a one in four chance that it will give us a result outside of the population range. If we double that sample to squares 4 and 10, there is only a one in 20 chance. If we triple it to squares 4,

10, and 15, there is a one in a thousand chance for an error. If we add square 22, there is zero

percent chance of the number of species, weight, MNI, or NISP being outside the population range that would be produced if we had excavated the entire 25 square trench.

Surprisingly, sampling from separate squares provides a more accurate estimate of the entire trench than sampling from adjacent squares. In other words, when sampled squares were adjacent to each other, there was an increased likelihood of a higher percentage of error.

The percentages increase from 4.67% to 7.50% when two 1×1s are connected to form a 2×1 and from 0.09% to 3.33%. when three 1×1s were combined to form one 3×1. If the distribution of mollusk was uniform, I would expect the connected and separated samples to have equal error percentages. The lack of uniformity within this trench is represented by the increase in errors of squares that are spatially near one another. Variation is expected at most archaeological sites via the deposition process occurring during an extended period of discarding food remains. Thus I expect the trend of spatially separated samples providing a more accurate representation of the area from which they are sampled.

The increase could be a result of there being fewer possible samples, but based on probability, the results are still important. For example, if selecting two 1×1s, there are 300 possibilities within our trench compared to 40 possibilities for selecting a 2×1 sample.

Because of the difference in possibilities, we actually see a decrease in the total number of samples producing an error, even though the percentage of errors increases. These trends are further supported when we consider the composition of individual species.

An adequate sample size allows researchers to investigate trends in past mollusk consumption at archaeological sites. I found that the sample from Coconut Walk was sufficient to represent consumption of mollusks by the inhabitants.

Mollusk Consumption

The purpose of any zooarchaeological study is to better understand human- environmental interactions in the past (Reitz and Wing 2008). This study focuses on how ancient Amerindian peoples inhabiting Coconut Walk during the Late Ceramic Age changed their relationship with the environment based on a large sample of midden and a statistically reliable sample. The first step in this process is to determine how dietary habits evolve over time because this will help determine change in interaction.

This study focused on the analysis of mollusk remains, which, though abundant, does not take into consideration other faunal and botanical material recovered at the site such as fish. As a result, it is difficult to comment on the overall diet of the population. Yet, it is possible to explore some trends and changes in mollusk consumption that can be seen as a proxy for other food gathering behaviors. There are several possibilities for changes in mollusk consumption. The inhabitants of Coconut Walk could have relied on a more or less diverse set of species, they could have increased or decreased their dependence on certain species, they could have increased or decreased the quantity of mollusks collected, and/or there may have been independent changes in the size of species. For example, if the inhabitants of Coconut Walk overharvested certain species, there should be an increase in the frequency of selected species, perhaps coincident with a decrease in other species. There have been links between decrease in mollusk size and overharvesting in numerous archaeological sites (Allen 2012; Braje et al. 2012; Erlandson et al. 2008, 2011; Erlandson and Rick 2010; Faulkner 2009; Jeradino 1997; Lasiak 1991; Mannino and Thomas 2001,

2002; Milner et al. 2007; Morrison and Hunt 2007; Morrison and Cochrane 2008; Prummel

2005; Stager and Chen 1996). There could have also been a transition from eating more fish to an emphasis on mollusks or vice versa. While a complete dietary analysis is beyond the scope of this study, it is a topic for future research.

To define what the diet was, I analyzed a group of commonly used variables

(Claassen 1998; Grayson 1984) of weight, MNI, and NISP which combined can provide approximate contributions of each species and their respective total dietary impact. While the percentage of overall weight, MNI, and NISP produce relative contributions by species, the last variable, species count, can be used to explore the level of diversity. Biomass is a quantification technique to estimate the meat weight through the weight of remains (Claassen

1998), but I did not use it here.

To identify patterns in the contribution of mollusks to prehistoric diet, I used weight,

MNI, and NISP. As outlined earlier the relative importance of weight, MNI, and NISP in estimating the abundance of separate species are debated in the literature (To list a few:

Bobrowsky 1982; Chaplin 1971; Grayson 1973 and 1984; Mason et al. 1998; O’Connor

2000; Perkins 1973; Ringrose 1993; Watson 1972; and White 1953). Despite the disagreement between researchers, one way to approach questions of species contribution is to analyze them together paying attention to patterns instead of individually. Using weight as a variable will lead to interpretations that favor heavy and large shells (and those which are more difficult to clean and may have residual soil within the shell), while using MNI as a variable will lead to interpretations that favor abundant shells, and using NISP as a variable will lead to interpretations that favor abundant fragments of shells and taxa with multiple

NISP per MNI. However, my research shows that using the three as a group produces a more robust approximation of abundance.

I began by examining patterns in classes to identify broad trends. I then focus on inter-species patterns for a more detailed assessment. To reiterate, this study uses a very large sample of mollusk collected from a 5×5 meter trench with an assemblage weighing slightly more than 141 kg. There were a total of 58,617 MNI and 85,166 NISP. At the level of classes, it is clear that Gastropods comprise the majority of the collection. Gastropods represent 86.1% of the total weight, 92.4% of the MNI, 77.4% of the NISP, and 84.6% of the total number of species of shell found in the 5×5. The second largest class is Bivalvia, with

3.9% of the total weight, 4.6% of MNI, 6.7% of NISP, and 14.1% of the number of species.

The last class is Polyplacophora, consisting only of C. tuberculatus but representing 3.6% of the total weight, 3.0% of MNI, 15.8% of NISP, and 1.3% of the number of species. Using class as the division, Gastropods were clearly the primary component of mollusks consumed at Coconut Walk and I expect the trend to continue for regional sites.

When compared to other excavations on Nevis, the Coconut Walk site was unique in its heavy reliance on gastropods. Animal remains found in excavation suggest that each location had a different faunal composition, though evidence for seasonality has not been studied at the sites. At the Saladoid site of Hichmans (100 BC to AD 600), rice rats, land crabs, and tree snails were the most common. Of the two post-Saladoid sites, parrotfish and surgeon fish were the most common taxa at Indian Castle (AD 650-880), while juvenile grunts and donax clams were most abundant at Sulphur Ghaut (AD 900-1200) (Kozuch and

Wing 2004). A comparison of MNI between four mollusk species can be found in Table 1

along with the results from Coconut Walk (AD 850-1450) as a part of this study. It should be noted that given the limited excavation that has taken place at sites on Nevis, apart from

Coconut Walk, that comparisons are somewhat superficial, with additional data needed to provide a more robust interpretation.

Table 19. The most abundant mollusk species at sites on Nevis by the percentage of invertebrate MNI. Total size of sample of animals is listed in parenthesis next to the site name (Coconut Walk is only for mollusks) (Newsom and Wing 2004).

Saladoid Post-Saladoid

Hichmans Indian Castle Sulphur Ghaut Coconut Walk Mollusk (355+) (1024) (3988) (58617) Donax 0% 14% 78% 4%

Topsnail 7% 12% 1% 2%

Nerites 12% 22% 3% 73%

Tree snails 32% 21% 3% 1%

From Table 19, it is apparent that each site’s assemblage had a different makeup in terms of mollusks. Hichmans had a large percentage of tree snails, but it could be the result of the small sample size. The post-Saladoid sites show a marked difference in preference.

The inhabitants of Sulphur Ghaut had a strong reliance on the bivalve donax clams (78%), while the people of Coconut Walk had a similar trend for the gastropod nerites (73%).

Results from Indian Castle suggest a more balanced use of mollusk resources in the past by its inhabitants with a slightly heavier reliance on gastropods. To refine the pattern of gastropod use at Coconut Walk it is necessary to delve into the species composition.

Using species as the level of division will provide greater detail into dietary habits.

The most abundant species by weight, MNI, and NISP are shown in Table 20 with their ranking in each category. Using the top five species from each category creates a list of nine species. As is expected the majority are gastropods (n=7) with one bivalve and one polyplacophora. Determining key species based on abundance is difficult because each measure provides different results. As discussed earlier, there is no standard procedure for determining how to weigh each measurement, and so I chose weight as my determining factor.

Table 20. Top species by weight, MNI, and NISP (% of total) at Coconut Walk.

Species Weight MNI NISP C. pica (gastropod) 38.2 (1) 1.8 (7) 7.6 (3) N. tessellata (gastropod) 21.5 (2) 64.1 (1) 45.6 (1) L. tuber (gastropod) 10.3 (3) 1.7 (8) 3.3 (7) S. gigas (gastropod) 4.2 (4) 0.2 (9) 0.5 (9) C. tuberculatus (polyplacophora) 3.6 (5) 3.0 (6) 15.8 (2) N. versicolor (gastropod) 2.6 (6) 8.6 (2) 6.0 (4) T. excavate (gastropod 2.4 (7) 6.0 (3) 5.1 (6) D. denticulatus (bivalve) 1.1 (8) 4.0 (4) 5.9 (5) T. muricatus (gastropod) 0.9 (9) 3.1 (5) 3.2 (8)

Based on the three key species, as well as the other species found within the top five of each category, it is assumed that the majority of mollusks were collected prehistorically from rocks along the shore (intertidal) or in shallow waters (subtidal). Mollusks provided an easily harvested food resource.

Exploitation and Over-Exploitation of Mollusks

Based on the overall dietary preferences described, were there are any changes in dietary preference over time or changes in the mollusks collected? Possible changes that could be observed include decline in abundance, decline in richness, change in age profile, and change in size (Classen 1986, 1998; Mannino and Thomas 2002). Changes in dietary preference can be the result or cause of changes within a taxon, understanding the magnitude of consumption at a location can supplement our understanding of human impacts. A wide variety of studies are available (Broughton 1997, 1999, 2002; Butler 2001; Byers and

Broughton 2004; Cannon 2003; Erlandson and Rick 2010; Grayson and Delpech 1998;

Morrison and Hunt 2007; Nagaoka 2000, 2002a, 2002b, 2005; Wolverton et al. 2012;

Wolverton et al. 2008; Whitaker 2010). Most of these studies rely on optimal foraging theory models (Broughton 1994a, 1994b; Lupo 2007; McCollough 1996; Smith and Wishnie 2000).

The primary point of the models is that humans will harvest the largest individuals from a species first because they provide the most caloric energy. This has been shown ethnographically to occur with mollusk selection by Fenberg and Roy (2007). If humans are harvesting fully grown adults at a sustainable rate, there should not be a decline in population size or number. Historically, however, humans have often lacked the ability to enact conservation strategies (Smith and Wishnie 2000). Thus, as humans begin overharvesting a species, the species is expected to become smaller over time as more and more individuals are harvested before they fully mature.

Following the model of the over-exploitation of mollusks laid out by Classen (1986,

1999) and Mannino and Thomas (2002), the over-exploitation of mollusks should precipitate

a decline in abundance. The first quantification technique that provides a rough estimate of the total abundance is weight. The disadvantage of weight is it overlooks how individual species change in terms of average weight (i.e., two small individuals and one large individual weigh the same). The assemblage analyzed here showed an increase of twenty- three times in weight over time, from 3 kg in planum 4 to 20 kg in planum 3 to 50 kg in planum 2 and 68 kg in planum 1. The last planum recovered represents almost the same weight as the other three combined.

Clearly, at Coconut Walk the rate at which prehistoric foragers collected and disposed of mollusks in this midden increased dramatically over time. The increase suggests that at the beginning of site occupation, there were a relatively small number of inhabitants or a small, temporary camp that grew into a larger, more permanent settlement that ultimately led to increased exploitation. Though I did not perform any tests for seasonality, the temporary

(seasonal) camp hypotheses could be tested by examining the seasonality of what is found in the midden to see if it only represents a certain time of year. If it appears to represent a year- round occupation, the hypothesis of a few inhabitants would presumably be correct. The distinction between a temporary and permanent settlement is not made when analyzing over- exploitation in this study because of the evidence for permanent houses (post-hole features, post-Saladoid culture) found in other trenches at Coconut Walk. Based on the pattern of increased consumption over time I therefore expected that the most heavily exploited species should be largest in planum 4 and smallest in planum 1. The results suggest that the largest individuals are found in planum 1 and the smallest in planum 4.

The relative abundance of a species is also based on MNI and NISP, which should, and does, mirror the increase found with weight. The MNI for each planum increased from

2,346 in planum 4 to 10,708 in planum 3, 20,009 in planum 2, and 25,554 in planum 1. The measurement of abundance should also track relative frequencies (Claassen 1998) to determine if there is a switch in species preferences. I found that MNI increased at a lesser percentage than weight, when the opposite would be expected if overharvesting took place.

Similarly the NISP for each planum increased from 3,431 in planum 4 to 15,110 in planum 3,

30,628 in planum 2, and 35,997 in planum 1.

Comparing the rate of increase between MNI and NISP can serve as a proxy for the relative abundance of different species in each planum. If NISP increased at the same rate as

MNI, it would mean that the proportions of different species remained relatively similar over time because the ratio of individuals to fragments would be consistent. If MNI increased more rapidly, it would suggest a trend toward a greater proportion of mollusks that have less fragmentary shells (small gastropods). Conversely, if NISP increased more rapidly, it would suggest a trend toward a greater proportion of mollusks that produce more fragmentary shells. In this study, the MNI to NISP ratio remains relatively similar, although there is a spike in planum 2, suggesting that the lowest levels aren’t more crushed and fragmented than the upper layers.

By using weight, MNI, and NISP as variables to examine change over time, it becomes obvious that there is a significant increase in mollusk consumption over time, with the last recorded levels of consumption being the highest. If over-exploitation is occurring it is expected that as weight, MNI, and NISP increase, the ratio of weight to MNI should

decrease to reveal a smaller individual/species size on average. At Coconut Walk the opposite is true as the ratio increases from 1.29 in planum 4 to 2.67 in planum 1.

Because there is not a decrease in abundance levels at the top of the trench, it suggests that use of the midden abruptly stopped. The top of the midden was covered with 5 cm of topsoil. The topsoil layer contained some mollusks, but because of its provenience and the differing characteristics of the soil it was not included in my analysis. The ending of the midden at the highest abundance rate could represent the use of a new refuse area or the movement of the settlement. The possibility of a catastrophic event (i.e. hurricane) could have removed the top layers of midden material must be taken into account though the topsoil layer suggests that there was a natural accumulation of material following the abandonment of the midden.

The second characteristic of over-exploitation is a decrease in species diversity (Allen

2012; Grayson et al. 2001; Nagaoka 2001; Jones 2004; Mannino and Thomas 2002). I used the species count, or the number of species present in each time period, as a way to quantify the species diversity characteristic. Across the entire trench, 78 individual taxa were identified. Planum 4 had 32 species, planum 3 had 48, planum 2 had 62, and planum 1 had

61. While the species found in a midden are not totally representative of the entire population that might be available, it is a good approximation. The presence of twice as many mollusk species in planum 1 and 2 as in planum 4does not mean species diversity doubled. The increase in diversity is more likely the result of increased and expanded harvesting strategies, leading to a few less common species being added to the pile. The corresponding increases are evident through the positive correlation of MNI with species count (R=0.959, p=0.041,

Figure X). In the results, the correlation is explained through the extremely large number of species that are relatively rare within the collection. Despite the correlation between species count and MNI, species count can be used as an approximation of change over time or change in dietary preferences. As the inhabitants increase their consumption, there is also an increase in the number of species consumed. Again, the increase in number of species is attributed to larger number of mollusks collected, regardless of which species are present along the shoreline, thus including the rarer species in the collection. Initially, with the smaller harvests, certain species may have been targeted. Or, the increase in number of species could simply be a factor of sample size within the collection as seems to be the case with the correlation.

The third characteristic, a change in age profile, was not a part of my study because of the difficult to age many mollusks. Research by Giovas et al. (2013) on N. tessellata from this excavation suggests that the age profile does not change over time. For more examples of changing age profiles with over-exploitation, see Claassen (1986), Faulkner (2009), Milner et al. (2007), and Mannino and Thomas (2001, 2002).

The last characteristic is a change in size, which has been observed in a number of studies, including those by Claassen (1986, 1998), Erlandson et al. (2011), Lupo (2007), and

Mannino and Thomas (2001, 2002). Using the same characteristics for the three most abundant species, a pattern of overharvesting (as we saw with overall trends) should become more apparent if it were in fact occurring. To demonstrate this, I focus on the relative proportions of weight, MNI, and NISP. I would expect that if a species decreases in size over time, I would find that the weight would constitute a lower proportion than MNI and NISP in

later contexts and that weight would constitute a higher proportion than MNI and NISP in earlier contexts. In the following sections, I use N. tessellata, C. pica, and L. tuber to show that the key species increase in weight over time as the number harvested increases – a pattern that is not predicted by the over-exploitation hypothesis.

N. tessellata

N. tessellata, which was the most abundant species in the assemblage. By weight, it accounts for 21.5% of the total weight (30.375 kg). Although the total amount of N. tessellata deposited in each planum increased over time, there was a decrease in the proportion of mollusk weight this species represented from 33.94% in planum 4 to 19.75% in planum 1 (Table 21). The decrease in weight percentage over time suggests that the first characteristic of over-exploitation, decrease in relative abundance, is true for this species.

Table 21. N. tessellata weight, MNI, NISP by planum (percentages for all mollusks).

Planum Weight in kg MNI NISP 1 13.486 (19.75%) 16966 (66.39%) 17382 (48.23%) 2 10.487 (20.85%) 13196 (63.27%) 13196 (43.08%) 3 5.377 (26.91%) 6742 (61.53%) 6742 (44.62%) 4 1.03 (33.94%) 1377 (58.70%) 1484 (43.25%) Total 30.375 (21.50%) 37591 (64.10%) 38804 (45.60%)

The two other quantitative measurements of abundance, MNI and NISP, do not show the same trend. This suggests an increase in MNI and in percentage of MNI for N. tessellata over time. NISP shows a large increase in number and a slight increase in the percentage of total

NISP over time. For this species, the percentage of total weight, MNI, and NISP supports a decrease in size as expected with over-exploitation. Yet over-exploitation is not supported when I calculate the percentage of weight, MNI, and NISP each planum provides to the total species weight, MNI, and NISP. Planum 4 has 3.3% of the total N. tessellata weight and

3.6% of the MNI. Planum 1 has 44.4% of the weight and 45.1% of the MNI.

The third characteristic, change in age (as measured by length), was studied by

Giovas et al. (2013). They suggest that the majority of nerites exploited at Coconut Walk were reproductively immature. The documented length for reproductive maturity is 14-17 mm (Chislett 1969) with Giovas et al. (2013) finding the mean length in each planum to vary between 9.96 to 10.44 mm. Though Giovas et al. (2013) note that these nerite shells do not have any overt age indicators, and as such, we cannot be sure the individuals were not reproductively mature. Given the increase in harvest size and the exploitation of seemingly reproductively immature individuals, I would expect size to decrease over time in accordance with the fourth characteristic of over-exploitation.

Increase in size over time was also studied by Giovas et al. (2013) with a conclusion that N. tessellata did increase in length (p<0.001) and width (p<0.001) from planum 4 to planum 1. They found an increase in mean length from 9.96 mm (95% CI 9.81-10.10 mm) in planum 4 to 10.44 mm (95% CI 10.36-10.51 mm) in planum 1. Likewise, the mean width increase from 12.70 mm (95% CI 12.50-12.91 mm) to 13.27 mm (13.16-13.38 mm) (Giovas et al. 2013).

I further investigated whether N. tessellata was in fact following a trend of getting larger over time as a result of exploitation, through the use of average individual weight.

Because this was not measured during the length and width measurement, other techniques were used. There were a variety of options for calculating this trend, with the obvious option to weigh and measure a large sample from the collection. This would provide the best and most accurate data, but is extremely time consuming. The next option was to use bag measurements of weight and MNI to estimate the average weight of each individual within that bag. This option allowed for a quick estimate of weight within different sections of the trench that could be further analyzed in the future to see if these trends hold up under further scrutiny.

There are numerous ways with which to calculate the estimate of average individual weight within each planum. The first is taking the sum of weight and dividing by the sum of

MNI for each planum. The second is narrowing this down to each square and averaging the squares for each planum. This technique would show distortion, however, because a high or low square would cause a larger change in the average size for a given planum. The third technique is to divide the squares into sieved squares (screened) and non-sieved squares

(troweled) and repeat the above methods. This method creates smaller sample sizes, but allows for a separation of individuals by collection method. Because of the size of N. tessellata, smaller individuals are more likely to be missed in troweling than in screening.

Thus, the sieved squares should provide the best estimate as to the actual size of N. tessellata over time (through plana). To compare with the results from Giovas et al. (2013) on size increase, I also analyzed the same sieved squares, producing an increase of 8.18% from 0.684 g to 0.740 g. Each of these methods was tested, providing slightly different results, but with an increase in size from planum 4 to planum 1.

It is expected that the sieved squares will provide the best estimate for average N. tessellata weight. Using sieved squares, the results for the sum of weight divided by sum of

MNI by planum increased from 0.6998 g/individual in planum 4 to 0.8006 g/individual in planum 1. This is very similar to the results using the average of square weight divided by square MNI for each planum as can be seen in Figure 26 (4: 0.7003 1:0.8030). Examining non-sieved squares, as well as combining all of the squares, produced results also suggesting an increase in weight from planum 4 to planum 1. Using the average of square averages for non-sample and all squares produced a slight increase in average weight, though not as different as with the sieved squares.

The results from the sieved squares were also tested through plotting a line of best fit, based on weight compared to MNI for each planum using both sieved square and all 25 squares. The results for all squares have planum 1, 2, and 3 with a similarly projected line (R2 of 0.948, 0.993, and 0.864) and planum 4 having a lower weight for any given MNI (R2 of

0.989). As would be expected from the earlier results, the sieved squares produce a better group of best fit lines. In order from highest slope/individual weight to lowest slope/individual weight were planum 1, 2, 3, and 4 (R2 of 0.806, 0.972, 0.241, and 0.995).

Although the R2 value for planum 3 is quite low (0.241), it is in line with what would be expected based on the g/individual calculations. Consequently, it is assumed that N. tessellata is becoming larger over time within this trench based on the findings of size by

Giovas et al. (in press) and on weight here.

C. pica

The second most abundant species in this study (C. pica) was also tested for its fit with the over-exploitation expectations. Unlike N. tessellata, C. pica provides a species that is high in weight relative to its MNI and NISP. Because it is a large and more friable shell, it produces a higher weight per individual; but, it also has many more fragments and is rarely found completely intact (Giovas 2009).

C. pica is the largest by weight, with almost double that of the second highest weight species (N. tessellata at 21.5%) The percentage of weight almost triples from planum 4 to planum 1, revealing that C. pica exploitation increased exponentially over time and should provide a better example than N. tessellata for testing changes in size over time (Table 22).

The species fails the first characteristic of a decrease in abundance over time as measured through weight, MNI, and NISP. It also does not have a decrease in relative abundance, as the percentages of all three quantifications increases over time.

Table 22. C. pica weight, MNI, NISP by planum (percentages are of total for all mollusks).

Planum Weight in kg MNI NISP 1 29.13 (42.67%) 487 (1.91%) 3612 (10.03%) 2 19.55 (39.18%) 337 (1.88%) 2046 (6.68%) 3 4.83 (24.19%) 164 (1.53%) 661 (4.37%) 4 0.49 (16.08%) 35 (1.49%) 131 (3.82%) Total 54.004 (38.25%) 1063 (1.81%) 6450 (7.57%)

An interesting aspect of C. pica is its relatively low MNI count compared to weight.

The small increase in MNI compared to the large increase in weight suggests that individual

weight is growing significantly, contrary to the fourth expectation of the over-exploitation model. As with N. tessellata, it is quite possible that additional fragments are the cause for the observed increase. As a relatively large shell that is somewhat friable, particularly given its nacreous characteristics, C. pica is more prone to fragmentation. The MNI count is based on NRE, which can lead to an inaccurate conclusion without considering NISP. I do not find this to be the case, as weight increased at a higher proportion than MNI or NISP. The ratio of

NISP to MNI approximately doubles from planum 4 to planum 1 while the ratio of weight to

MNI increases seven fold.

As I did with N. tessellata, I calculated the individual weight of C. pica to estimate change in size over time. It was more difficult to determine individual weight with C. pica as almost none of the shells were completely intact, and most of the pieces recovered were quite fragmentary (representing less than 50% of the individual). As a result, the average weight for each individual and fragment were both calculated (Table 23) to determine if they are correspondingly increasing, or if extra fragments are causing the increase in weight.

Accordingly, the sieved squares provided the best information. For individual weight, there is an increase over time in sieved squares from 7.79/7.91 in planum 4 to 56.73/57.46 in planum

1.There was also an increase over time using the non-environmental squares and all squares for both methods. These are significant increases in weight over time, and are consistent whether trowels or sieves were used to collect the samples.

Table 23. C. pica individual and fragment weight.

Individual Weight (weight/MNI) Fragment Weight (weight/NISP) Planum Sum Average Sum Average 1 56.73 57.46 6.90 6.78 2 27.91 28.52 5.20 5.42 3 24.20 26.43 4.86 5.63 4 7.79 7.91 1.85 2.13

As was hypothesized earlier, extra fragments could be the cause for the increase in weight. To test this, MNI was replaced with NISP, converting from individual weight to fragment weight. It is expected that more fragments would be collected through sieving than troweling. Thus sieved squares were assumed to be the most accurate measurement of fragment weight, the results suggest that fragment weight was increasing as individual weight increased, indicating that the species was increasing in size and that extra fragments are not responsible for the increase in weight. Fragment weight increased from 1.85/2.13 in planum 4 to 6.90/6.78 in planum 1. Using sieved squares as the most accurate approximation, there was an increase in individual weight and fragment weight, suggesting an increase in individual size over time for C. pica. Again the results are opposite of the expected decrease in size with over-exploitation.

L. tuber

Unlike the previous two species, individual shells of L. tuber do not show a clear pattern of increasing size over time, but they may provide support for the hypothesis of over- exploitation. In terms of weight, the total amount of L. tuber shells is greatest in planum 1,

100

but as a percentage of all shells it is greatest in planum 3 (Table 24). As with the other key species, both the MNI and NISP of L. tuber also increase in absolute numbers between plana

4 and 1, in this case by a factor of more than ten. However, as a percentage of all individuals, the MNI for L. tuber is relatively stable, ranging between 1.86% in planum 1 and 1.36% in planum 4. Similarly, the NISP of L. tuber, as a percentage of all fragments in each planum, has a narrow range from a maximum of 3.72% in planum 1 and a minimum of 2.48% in planum 4. Finally, in terms of both individual and fragment weight, I was able to identify a pattern of increasing weight between plana 4 and 3, stability between plana 3 and 2, and decreasing size between plana 2 and 1. Because L. tuber remained relatively consistent in terms of the proportion of each planum it represented by total weight, total MNI, and total

NISP, the rise and fall of individual and fragment weight suggest that the species changed in size over time and provides some support for the model of over-exploitation.

Table 24. L. tuber weight, MNI, NISP by planum (percentages are of total for all mollusks).

Planum Weight in kg MNI NISP 1 6.46 (9.46%) 474 (1.86%) 1340 (3.72%) 2 4.94 (9.90%) 290 (1.45%) 944 (3.08%) 3 2.92 (14.63%) 177 (1.65%) 472 (3.12%) 4 0.27 (8.78%) 32 (1.36%) 85 (2.48%) Total 14.59 (10.33%) 973 (1.65%) 2841 (3.33%)

Key Species

Comparing the trends found in all three key species, the data suggests a correlation between an increase in individual size and in increase in harvest size during the occupation of

101

Coconut Walk. As such, the correlation of these two patterns provides strong evidence that human activities caused the increased size of mollusk species. N. tessellata and C. pica both show consistent increases in size over time as there is a drastic increase in harvest size. L. tuber has a similar increase in harvest size, but shows an increase in size and then a return to the original size. This suggests that there may be a human factor causing the species to become larger over time, which is contrary to what would be expected if over-exploitation was taking place (if indicated by species getting smaller over time).

It should be noted that all of these interpretations are based on the assumption that an increase in the weight of a shell indicates and increase in the overall weight of an individual animal and that, by extension, later heavier-shelled mollusks produced more meat per individual than earlier lighter-shelled mollusks (Claassen 1986). One of the primary assumptions as to the change in key species is that weight is comparable to size and that the weight of a shell is a way to predict the weight of a living mollusk. Assuming that weight is related to size, this study was able to estimate the difference in size based on weight. It is expected in mollusks that there is a proportional increase in weight as size increases.

Table 25. Key species weight in kg by planum.

Planum N. tessellata C. pica L. tuber 1 13.486 (19.75%) 29.13 (42.67%) 6.46 (9.46%) 2 10.487 (20.85%) 19.55 (39.18%) 4.94 (9.90%) 3 5.377 (26.91%) 4.83 (24.19%) 2.92 (14.63%) 4 1.03 (33.94%) 0.49 (16.08%) 0.27 (8.78%) Total 30.375 (21.50%) 54.004 (38.25%) 14.59 (10.33%)

102

Other assumptions for change in key species include the notion that MNI and NISP are representative of individuals and fragments, which are two differing ways to quantify a taxon. With the quality control test, it was found that NISP was more accurate than MNI, simply because counting fragments is easier than counting non-repetitive elements. Finally, the last assumption with key species is that weight is a significant factor for determining the most abundant species within the midden. Weight is correlated to meat weight, which should provide a ranking of importance among species.

Factors for Over-Exploitation

One of the underlying premises behind the concept of mollusk exploitation suggests that species should decrease in size as harvest size increases (Claassen 1986, 1998; Mannino and Thomas 2001, 2002; Braje et al. 2012; Campbell 2008; Erlandson et al., 2011; Fenberg and Roy, 2007, 2012; Jeradino, 1997; Lupo, 2007; Stager and Chen, 1996). It is accepted within the field that the exploitation of marine resources is common when humans move into new environments (Erlandson 1988, 2001; Erlandson and Fitzpatrick 2006; Erlandson et al.

2007; Parkington 2004; Waselkov 1987). However, as noted in this study (and Giovas et al. in press), the opposite appears to be true as there is a significant increase in size as harvests increase in size. In the following section, I consider three factors that might explain why the expectations of the over-exploitation model are not only not met, but reversed; these are: changes in the environment, cultural attributes, and population growth.

The first factor, change in the environment, could be a causal factor in the apparent growth of mollusk sizes in the Coconut Walk midden. It has been shown that nerite growth

103

rates can be affected by a number of environmental factors including air and sea temperatures

(Axelsen 1968; Chislett, 1969; Faulkner, 2009; Kolipinski, 1964; Lewis, 1971, Lewis et al.,

1969; Underwood, 1975, 1976). The midden spans a time period from AD 850-1450, which overlaps with the Medieval Warm Period (AD 950-1400) and the Little Ice Age (AD 1400-

1700) (Gischler et al. 2008; Lane et al. 2011; Richey et al., 2007). From what we know of the average temperature of the Caribbean Sea during these times, average sea temperature and hurricane frequency is likely to have increased during the occupation period of Coconut

Walk (Gischler et al. 2008; Lane et al. 2011; Richey et al., 2007). The amount of precipitation is known to have fluctuated locally based on palaeoclimate studies in the

Caribbean, including lake core sediment analysis in Belize (Gischler et al. 2008), Dominican

Republic (Lane et al. 2011), Haiti (Hodell et al. 1991; Curtis and Hodell 1993; Hiquera-

Gundy et al. 1999), Puerto Rico (Burney et al. 1994); Siegel et al. 2005), and St. Martin

(Bertran et al. 2004). There has been work to localize climate modeling as well as methods that can reproduce climate histories through botanical studies. There have been no such studies conducted on the island of Nevis. Using regional studies, it appears that temperature was increasing and precipitation decreasing in the region during occupation. I suspect the shift to a warmer, drier climate could lead to larger mollusk size and an increased reliance on marine resources due to a decrease in agriculture production. This could be tested with stable isotopes on the shellfish.

Alternatively, we might look to one or a combination of anthropogenic factors to explain the increase in the size of mollusks over time. There are four primary explanations for the dramatically increased harvests of mollusks, and indirectly, the increase in mollusk

104

size over time, at Coconut Walk. The first is an increase in population size—larger populations produce larger middens. The second explanation is that populations of a stable size changed their diets to include more mollusks, which goes along with the climate assumptions. A decline in one food resource would lead to an increased reliance of another resource. Mollusks provide an easily acquired “renewable” resource that could supplement any food shortage. The third explanation is that there was a cultural change that created an increased preference for mollusks or an expansion of catchment. Prehistorically, it may have been thought that mollusks provided some type of advantage over other foods. The last explanation is some combination of the previous three. There could have been an increase in population size that caused overharvesting of other species and resulted in a resource shift toward mollusks that was aided by cultural changes encouraging mollusk consumption.

Regardless of the cause, there was a significant increase in mollusk consumption over time at

Coconut Walk.

Factors that cause an increase in harvest size should theoretically cause a decrease in individual size based on the optimal foraging theory models (Broughton 1994a, 1994b; Lupo

2007). The opposite was found to be the case in this study. The question remains, then, what factors may have caused an increase in harvest size and an increase in individual size? The first consideration is that the inhabitants of Coconut Walk were not “overharvesting” mollusk species. They could have been harvesting enough individuals to thin competition among mollusks for resources, thus allowing some individuals to grow larger. It has been shown that nerite growth is reduced by competition for food in high densities (Underwood 1976). Other harvesting strategies could have been implemented, such as increasing the range of collection

105

for mollusks (Campbell 2008). It has also been shown that compared to adults, children collect smaller mollusks (Bird and Bird 2000; De Boer et al. 2002). Likewise, there could have been cultural knowledge about the growth cycle of mollusks. Using this knowledge, the residents of Coconut Walk could have been selectively harvesting mollusks provide optimal conditions for mollusk populations in the future. This could include harvesting the smaller adults to remove them from the breeding population, similar to selective breeding seen in domesticated species. Prehistorically there is no evidence for deliberate selection for future sizes of mollusks though there are examples of increased size with a decrease in harvest rates suggesting some management (Thakar 2011; McCoy 2008; Giovas et al. 2010).

A more likely factor would be the result of an increase in population size. If the ratio of mollusk consumption to the consumption of other foods remains similar throughout time, it is assumed that the population would have to increase to account for the larger harvest size.

With an increase in consumption of other foods, there would be an increase in any agriculture practices leading to an increase runoff of nutrients and minerals from the soil near the breeding grounds of mollusks (Giovas et al. 2010). Likewise, a larger population produces more human waste, which depending on the method of disposal could be another source of nutrients for mollusks. Thus, as the population increases, mollusk harvests increase, nutrient runoff increases creating an increase in mollusk size. Research does not support this claim.

Keegan et al. (2003) suggests that increased rainfall and agriculture can lead to unfavorable conditions for mollusks. Additionally, the harvest of mollusks could create less competition for more resources, helping spur mollusk size increase. Though these factors have been

106

speculative, they are based on probable cause, with further research required to investigate why mollusk size appears to be increasing over time.

107

CHAPTER 7: CONCLUSIONS

A primary goal of zooarchaeology is to provide information about the past interactions between humans and their environment. Other goals include looking at dies and social issues. These efforts are aimed at better understanding how we interact with our environment and how our actions can change the world we live in. Faced with the overwhelming fact that the global human population is drastically altering the environment without fully recognizing the permanent effects we can look to past examples for how to better manage our resources (Hudson et al. 2012). There is a long history of using islands as model systems (MacArthur and Wilson 1967; Kirch 1984, 1997, 2000; Kirch and Green

1987; Vitousek 1995, 2002, 2004) to present a unique opportunity to analyze local populations that are heavily reliant on marine resources (Kirch 1982, 1994, 1997, 2000,

2007: Burney et al. 2001; Steadman 1995). More importantly as sea levels continue to fluctuate and developing islands destroy countless archaeological sites it becomes ever more important to preserve and study these coastal sites before they disappear (FitzGerald et al.

2008; Gontz et al. 2011).

The issue of sampling strategies in zooarchaeological studies is important given that typically only a small percentage of a faunal assemblage is analyzed, or a smaller portion of a large site excavated (Claassen 1998; Reitz and Wing 2008). In this thesis, I was able to determine whether different sampling strategies (e.g., judgmental versus random) would provide different results across a larger area, as well as provide a guideline for other researchers interested in deciding how to most effectively sample faunal assemblages.

108

The results indicate that in the case of Coconut Walk, sampling a single 1 m2 (or 1 m3) square to represent a 25 m2 meter midden area, leads to a square composition error rate of 24% and a key species composition error rate of 52%. Sampling two non-connecting meter squares reduces those rates to 4% and ~10% respectively. Three non-connecting meter squares produce less than 1% for both error rates. Using four non-connecting meter squares produces averages that are within the 95% CI for all midden and key species composition statistics. Overall, there is a trend of increased error rates when the sample consists of two, three, and four meter squares that are adjacent to one another.

As the samples show, even a sample that is within the CI for the population can produce results between plana that differ from the assemblage. This will have an impact on conclusions drawn from the sample and further impacts on comparisons to other zooarchaeological findings. The results suggest that when looking at variation, sample size must increase as the variation decreases to validate the findings. Researchers must be cautious with the conclusions they develop as it pertains to overall faunal compositions within a site without enough information to determine if a trend is actually occurring throughout the population and not exclusively in the sample.

While it is important to remember that the faunal assemblage at Coconut Walk is not necessarily representative of what might be found at other sites in the Caribbean (or elsewhere in the world, for that matter), this exercise shows the usefulness of sampling areas within a site that are not necessarily in close proximity to one another (assuming the site is not entirely uniform), and that 16% of a given surface area (in this case, a 5×5 m trench) is a good baseline for ensuring that the sample recovered is representative of the whole

109

population. Though changes between plana may not be accurately represented in a sample of this size, it provides a baseline for the population upon which future research can be focused.

Overall, it may be prudent for archaeologists in some situations to consider conducting a complete analysis of a larger vertically and spatially representative sample of an assemblage to determine what size sample provides a statistically significant representation of the faunal population at large.

This study produced data that suggests the inhabitants of Coconut Walk, Nevis had an impact on their local mollusk species. Typically, we expect to find a decrease in the size of mollusks in response to human exploitation (and over-exploitation). In the few cases where species size increases, it seems to be accompanied by evidence for decreasing overall exploitation (Giovas et al. 2010; McCoy 2008; Thakar 2011). In my findings at Coconut

Walk, the increase in size in key species over time does not support the predicted model of over-exploitation. Furthermore, mollusks increased in size in tandem with an increase in harvest size. The well-documented discrepancies from the expectations of the over- exploitation hypothesis suggest that this is one of the first case studies of the management of mollusk populations by a prehistoric community. A similar case was noted in several fish species by Leach and Davidson (2001).

As is common, this study produces more questions than answers. Do we need to reconsider the theories of exploitation that are commonly accepted and introduce a theory that is based on local interactions with species? While much of the literature suggests that exploitation in mollusks follows a certain trend, there have recently been examples of differing trends (Giovas et al. 2010, 2013; Thakar 2011; McCoy 2008; Leach and Davidson

110

2001). With more research at Coconut Walk and other regional sites do the trends developed in this paper continue? A reexamination of mollusk size from excavations in the region can shed light on this issue. Most importantly, was the interaction between the peoples of

Coconut Walk and the environment sustainable and how did they modify their environment to create this situation? The interactions appear to have been sustainable based on the increase in species size. The interaction would need to be analyzed in a broader regional framework to see if there were local exploitations that proved unsustainable and how these shed light on the case study at Coconut Walk. An examination of mollusk size from excavations in the region would allow such a comparison. Lastly, how can we improve zooarchaeological methodologies to better represent our data? For this I suggest as a field we reconsider the methods and their justifications to form a more succinct methodology.

Likewise, we have the data available to find statistics that better represent our findings and can be applied on a universal scale.

111

REFERENCES

Abbott RT. 1995. A field guide to shells: Atlantic and Gulf coasts and the West Indies. Boston: Houghton Mifflin.

Abbott RT and Morris PA. 1995. Shells of the Atlantic and Gulf Coasts and the West Indies. Boston: Houghton Mifflin.

Alegr´ıa MP. 1993. El dise ño pintado de la cerámica Saladoide de Puerto Rico, Colección de Estudios Puertorriqueños, San Juan.

Allaire L. 1997. The Lesser Antilles before Columbus. In The Indigenous People of the Caribbean. Wilson SM, ed. Pp. 20-28. Gainesville: University Press of Florida.

Allen J. 1997. The impact of Pleistocene hunters and gatherers on the ecosystems of Australia and Melanesia: In tune with nature? In Historical ecology in the Pacific Islands: prehistoric, environmental and landscape change. Kirch PV and Hunt T, eds. Pp. 22–38. New Haven: Yale University Press.

Allen MA. 2012. Molluscan foraging efficiency and patterns of mobility amongst foraging agriculturalists: A case study from northern New Zealand. J of Archaeol Sci. 39:295-307.

Ambrose W. 1967. Archaeology and Shell Middens. Archaeology and Physical Anthropology in Oceania. 2:169-187.

Anderson AJ. 1989. Mechanics of overkill in the extinction of New Zealand moas. Journal of Archaeological Science. 16:137–51.

Anderson AJ. 1997. Prehistoric Polynesian impact on the New Zealand environment: Te Whenua Hou. In Historical ecology in the Pacific Islands: prehistoric, environmental and landscape change. Kirch PV and Hunt T, eds. Pp. 271-283. New Haven: Yale University Press.

Anderson A. 2002. Faunal collapse, landscape change and settlement history in Remote Oceania. Journal of World Archaeology. 33:375–90.

Axelsen F. 1968. Growth rate study of some tropical marine invertebrates. M. Sc. Thesis. Montreal: McGill University.

Bailey G, Deith M, and Shackleton N. 1975. Oxygen Isotope Analysis and Seasonality Determinations: Limits and Potential of a New Technique. American Antiquity. 48:390- 398.

112

Ballbé EG. 2005. Shell middens on the Caribbean coast of Nicaragua: prehistoric patterns of mollusk collection and consumption. In Archaeomalacology: Molluscs in Former Environments of Human Behaviour. Bar-Yosef Mayer, DE, ed. Pp. 40-53. Oxford: Oxbow Books.

Balme J and Paterson A. 2006. Archaeology in Practice: A student Guide to Archaeological Analyses. Malden: Blackwell Publishing.

Beard JS. 1949. The Natural Vegetation of the Windward and Leeward Islands. Oxford Forestry Memoirs. 21:1-192.

Bellamy P. 2001. Coconut Walk, Nevis, Lesser Antilles. Archaeological investigations by the Time Team, October 1998. Terrain Archaeology, Report No.5034/3.1. Unpublished report, version April 2001.

Bertran P, Bonnissent D, Imbert D, Lozouet P, Serrand N, and Stouvenot C. 2004. Pale´oclimat des petites Antilles depuis 4000 ans BP: l’enregistrement de la lagune de Grand-Case a` Saint Martin. C. R. Geoscience. 336:1501–1510.

Bird DW and Bliege Bird BR. 2000. The ethnoarchaeology of juvenile foragers: Shellfishing strategies among Meriam children. Journal of Anthropological Archaeology. 19:461–476.

Blick JP. 2007. Pre-Colombian impact on terrestrial, intertidal, and marine resources, San Salvador, Bahamas (AD 950-1500). Journal for Nature Conservation. 15(3):174-183.

Blume H. 1972. The Caribbean Islands. London: Longman Group Limited.

Bobrowsky PT. 1982. An examination of Casteel’s MNI behavior analsis: A reductionist approach. Midcontinental Journal of Archaeology. 7(2):171-184.

Bobrowsky PT and Gatus TW. 1984. Archaeomalacological significance of the Hall Shelter, Perry County, Kentucky. North American Archaeologists. 5(2):89-100.

Bocek B. 1986. Rodent Ecology and Burrowing Behavior: Predicted Effects on Archaeological Site Formation. American Antiquity. 51:589-602.

Boessneck J. 1969. Osteological differences between sheep (Ovis aries Linne) and goat (Capra hircus Linne). IN Science in archaeology: A survey of progress and research (2nd ed.). Brothwell E and Higgs E, eds. Pp. 331-358. New York: Praeger Publishing.

Bökönyi S. 1970. A new method for determination of the number of individuals in animal bone material. American Journal of Archaeology. 74:291-292.

113

Bovbjerg RV. 1984. Habitat selection in two intertidal snails, genus Nerita. Bull. Mar. Sci. 34:185-196.

Bowdler S. 1983. Sieving Seashells: Midden Analysis in Australian Archaeology. In Australian Field Archaeology: A Guide To Techniques. Connah G, ed. Pp.135-143. Canberra: Australian Institute of Aboriginal Studies.

Braje TJ, Rick TC, and Erlandson JM. 2012. A trans-Holocene historical ecological record of shellfish harvesting on Northern Channel Islands. Quaternary International. 264:109-120.

Broughton JM. 1997. Widening diet breadth, declining foraging efficiency, and prehistoric harvest pressure: Ichthyofaunal evidence from the Emeryville Shellmound, California. Antiquity. 71:845–862.

Broughton JM. 1999. Resource depression and intensification during the late Holocene, San Francisco Bay: Evidence from the Emeryville Shellmound vertebrate fauna. Berkeley: University of California Publications, Anthropological Records 32.

Broughton JM. 2002. Prey spatial structure and behavior affect archaeological tests of optimal foraging models: Examples from the Emeryville Shellmound vertebrate fauna. World Archaeol. 34:60–83.

Buchanan W. 1985. Sea Shells Ashore. M.A. Thesis, Archaeology, University of Cape Town South Africa.

Burney DA, Burney LP, and MacPhee RDE. 1994. Holocene charcoal stratigraphy from Laguna Tortuguero, Puerto Rico, and the timing of human arrival. Journal of Archaeological Science. 21:273–281.

Burney DA, James HF, Burney LP, Olson SL, Kikuchi W, Wagner WL, Burney M, McCloskey D, Kikuchi D, Grady FV, Gage R, and Nishek R. 2001. Fossil evidence for a diverse biota from Kaua’I and its transformation since human arrival. Ecological Monographs. 71(4):615-641.

Butler VL. 1998. Changin fish use on Mangaia, Southern Cook Islands: Resource depression and the prey choice model. International Journal of Osteoarchaeology. 11(1-2):88-100.

Butler VL. 2001. Changing fish use on Mangaia, Southern Cook Islands: Resource depression and the prey choice model. International Journal of Osteoarchaeology. 11:88- 100.

114

Byers DA and Broughton JM. 2004. Holocene environmental change, artiodactyl abundances, and human hunting strategies in the Great Basin. American Antiquity. 69:235-255.

Campbell G. 2008. Beyond means to meaning: using distributions of shell shapes to reconstruct past collecting strategies. Environmental Archaeology. 13:111-121.

Cannon A. 2000. Faunal remains as economic indicators on the Pacific Northwest Coast. In Animal bones, human societies. Rowley-Conwy P, ed. Pp. 49-57. Oxford: Oxbow Books.

Cannon MD. 2001. Archaeofaunal relative abundance, sample size, statistical methods. Journal of Archaeological Science. 28:185-195.

Cannon MD. 2003. A model of central place forager prey choice and an application to faunal remains from the Mimbres Valley, New Mexico. Journal of Anthropological Archaeology. 22:1–25.

Carucci J. 1992. Cultural and Natural Patterning in Prehistoric Marine Foodshell from Palau, Micronesia, Ph.D. dissertation, Anthropology, Southern Illinois University.

Casteel RW. 1977. Characterization of faunal assemblages and the minimum number of individuals determined from paired elements: Continuing problems in archaeology. Journal of Archaeological Science. 4(2):125-34.

Chanlatte Baik L. 1995. Presencia huecoide en Hacienda Grande, Lo´ıza. In Proceedings of the XV International Congress for Caribbean Archaeology. Alegr´ıa RE and Rodr´ıguez M, eds. Pp. 501-510. San Juan: Centro de Estudios Avanzados de Puerto Rico y el Caribe.

Chaplin RE. 1971. The study of Animal bones from archaeological sites. New York: Seminar Press.

Chislett GR. 1969. Comparative aspects of the ecology of three Nerita (Mollusca: Gastropoda) species from different locations in Barbados. M. Sc. Thesis. McGill University. Montreal, QC. Canada.

Claassen C. 1986. Temporal patterns in marine shellfish-species use along the Atlantic coast in the Southeastern United States. Southeast Archaeology. 5:120–37.

Claassen C. 1998. Shells. Cambridge: Cambridge University Press.

Claassen C. 2000. Quantifying shell: Comments on Mason, Peterson, and Tiffany. American Antiquity. 65(2):415–418.

115

Clason AT. 1972. Some remarks on the use and presentation of archaeozoological data. Helinium. 12(2):139.153.

Callaghan RT. 1992. A computer simiulation of routes of passage in the Caribbean and Gulf of Mexico. Paper presented at the 57th annual meeting of the Society for American Archaeology, Pittsburgh, PA.

Callaghan RT. 1995. Antillean cultural contacts with mainland region as a navigation problem. In Proceedings of the XV International Congress for Caribbean Archaeology. Alegr´ıa RE and Rodr´ıguez M, eds. Pp. 181-190. Centro de Estudios Avanzados de Puerto Rico y el Caribe, San Juan.

Callaghan RT. 2001. Ceramic Age Seafaring and Interaction Potential in the Antilles: A Computer Simulation. Current Anthropology. 42(2):308-313.

Colten RH. 2002. Prehistoric marine mammal hunting in context: Two western North American examples. Internatinoal Journal of Osteoarchaeology. 12(1):12-22.

Crock JG, Petersen JB, and Douglas N. 1995. Preceramic Anguilla: a view from the Whitehead’s Bluff site. In Proceedings of the XVth International Congress of Caribbean Archaeology (Puerto Rico). Pp. 283–94. San Juan: Centro de Estudios Avanzados de Puerto Rico y el Caribe.

Crosby AW Jr. 1972. The Columbian Exchange: Biological and Cultural Consequences of 1492. Westport: Greenwood Press.

Curtis JH And Hodell DA. 1993. An isotopic and trace element study of ostracods from Lake Miragoane, Haiti: a 10,500 year record of paleosalinity and paleotemperature changes in the Caribbean. Geophysical Monograph. 78:135–152.

Davis DD. 1988. Calibration of the Ceramic Period Chronology for Antigua, West Indies. Southeastern Archaeology. 7:52-60.

Davis DD. 1993. Archaeic Blade Production on Antigua, West Indies. American Antiquity. 58:688-697.

Davis DD. 2000. Jolly Beach and the preceramic occupation of Antigua, West Indies. Yale University Publications in Anthropology 84. New Haven: Yale University Press.

De Boer WF, Blijdenstein AF, and Longamane F. 2002. Prey choice and habitat use of people exploiting intertidal resources. Environmental Conservation. 29:238–252.

116

Drennan RD. 2009. Statistics for Archaeologists: A Common Sense Approach, 2nd ed. New York: Springer Science+Business Media.

Erlandson JM. 1988. The role of shellfish in prehistoric economies: a protein perspective. American Antiquity. 53:102–109.

Erlandson JM. 2001. The archaeology of aquatic adaptations: paradigms for a new millennium. Journal of Archaeological Research. 9:287-350.

Erlandson JM, Braje TJ, Rick TC, Jew NP, Kennett DJ, Dwyer N, Ainis AF, Vellanoweth RL, and Watts J. 2011. 10,000 years of human predation and size changes in the owl limpet (Lottia gigantea) on San Miguel Island, California. Journal of Archaeological Science. 38:1127-1134.

Erlandson JM, Graham MH, Bourque BJ, Corbett D, Estes JA, and Steneck RS. 2007. The kelp highway hypothesis: marine ecology, the coastal migration theory, and the peopling of the Americas. Journal of Island and Coastal Archaeology. 2:161-174.

Erlandson JM and Fitzpatrick SM. 2006. Oceans, islands, and coasts: current perspectives on the role of the sea in human prehistory. Journal of Island and Coastal Archaeology. 1:5- 32.

Erlandson JM and Rick TC. 2010. Archaeology meets marine ecology: the antiquity of maritime cultures and human impacts on marine fisheries and ecosystems. Annual Review of Marine Science. 2:231-251.

Erlandson JM, Rick TC, Braje TJ, Steinberg A, and Vellanoweth RL. 2008. Human impacts on ancient shellfish: a 10,000 year record from San Miguel Island, California. Journal of Archaeological Science. 35:2144-2152.

Faulkner P. 2009. Focused, intense and long-term: evidence for granular ark (Anadara granosa) exploitation from late Holocene shell mounds of Blue Mud Bay, northern Australia. Journal of Archaeological Science. 36:821-834.

Fenberg PB and Roy K. 2007. Ecological and evolutionary consequences of size-selective harvesting: how much do we know? Molecular Ecology. 17:209–220.

Fenberg PB and Roy K. 2012. Anthropogenic harvesting pressure and changes in life history: insights from a rocky intertidal limpet. The American Naturalist. 180:200-210.

Fieller NRJ and Turner A. 1982. Number estimation in vertebrate samples. Journal of Archaeological Science. 9(1):49-62.

117

FitzGerald DM, Fenster MS, Argow BA, and Buynevich IV. 2008. Annual Review of Earth and Planetary Sciences. 36:601-647.

Fitzpatrick SM. 2006. A critical approach to C-14 dating in the Caribbean: Using chronometric hygiene to evaluate chronological control and prehistoric settlement. Latin American Antiquity. 17(4):389-418.

Fitzpatrick SM. 2009. Pre-Columbian Sites on Carriacou, West Indies. Journal of Field Archaeology. 34:247-266.

Fitzpatrick SM. 2011. Verification of an archaic age occupation on Barbados, Southern Lesser Antilles. Radiocarbon. 53(4):595-604.

Fitzpatrick SM and Donaldson TJ. 2007. Anthropogenic impacts to coral reefs in Palau, western Micronesia during the Late Holocene. Coral Reefs. 26(4):915-920.

Fitzpatrick SM and Keegan WF. 2007. Human impacts and adaptations in the Caribbean Islands: an historical ecology approach. Earth and Environmental Science Transactions of the Royal Society of Edinburgh. 98:29-45.

Flannery KV. 1967. Vertebrate fauna and hunting patterns. In The prehistory of the Tehuacan Valley (Vol. 1). Byers DS, ed. Pp. 132-177. San Diego: Academic Press.

Gilbert AS and Singer BH. 1982. Reassessing zooarchaeological quantification. World Archaeology. 14(1):21-40.

Giovas CM. 2009. The shell game: analytic problems in archaeological mollusk quantification. Journal of Archaeological Science. 36:1557-1564.

Giovas CM, Clark M, Fitzpatrick SM, and Stone J. 2013. Examining Size Differences in a Nerita sp. between ca. AD 900-1150 at the Coconut Walk Site on Nevis. In review. Journal of Archaeological Science.

Giovas CM, Fitzpatrick SM, Clark M, and Abed M. 2010. Evidence for size increase in an exploited mollusc: humped conch (Strombus gibberulus at Chelechol ra Orrak, Palau from ca. 3000-0 BP. Journal of Archaeological Science. 37(11):2788-2798.

Gischler E, Shinn EA, Oschmann W, Fiebig J, and Buster NA. 2008. A 1500-Year Holocene Caribbean Climate Archive from the Blue Hole, Lighthouse Reef, Belize. Journal of Coastal Research. 246:1495-1505.

Glassow MA. 1993. Changes in subsistence on marine resources through 7,000 years of prehistory on Santa Cruz Island. In Archaeology of the Northern Channel Islands of

118

California: Studies of Subsistence, Economics, and Social Organization. Glassow MA, ed. Pp. 75-94. Salinas: Archives of California Prehistory 34, Coyote Press.

Glassow MA. 2000. Weighing vs. counting shellfish remains: a comment on Mason, Peterson, and Tiffany. American Antiquity. 65(2):407–414.

Glassow MA. 2005. Variation in marine fauna utilization by Middle Holocene occupants of Santa Cruz Island, California. In Proceedings of the Sixth California Islands Symposium. Garcelon D and Schwemm C, eds. Pp. 23-34. Arcata, California: National Park Service Technical Publication CHIS-05-01, Institute for Wildlife Studies.

Gontz AM, Maio CV, Wagenknecht EK, and Berkland EP. 2011. Assessing threatened coastal sites: Applications of ground-penetrating radar and geographic information systems. Journal of Cultural Heritage. 12(4):451-458.

Goodfriend G. 1991. Holocene Trends in 18O in Land Snail Shells from the Negev Desert and their Implications for Changes in Rainfall Source Areas. Quaternary Research. 35:417-426.

Grayson DK. 1973. On the methodology of faunal analysis. American Antiquity. 38(4):432- 9.

Grayson DK. 1978. Minimum numbers and sample size in vertebrate faunal analysis. American Antiquity. 43(1):53-65.

Grayson DK. 1979. On the quantification of vertebrate archaeofaunas. In Advances in archaeological method and theory (Vol. 2). Schiffer MB, ed. Pp. 199-237. New York: Academic Press.

Grayson DK. 1984. Quantitative Zooarchaeology: Topics in the Analysis of Archaeological Faunas. Orlando: Academic Press Inc.

Grayson DK. 2001. The archaeological record of human impact on animal populations. Journal of World Prehistory. 15:1–68.

Grayson DK and Meltzer D. 2003. A requiem for North American overkill. Journal of Archaeological Science. 30:585–93.

Grayson DK and Delpech F. 1998. Changing diet breadth in the early Upper Paleolithic of southwestern France. J Archaeol Sci. 25:119-1129.

119

Grayson DK, Delpech F, Rigaud JP, and Simek JF. 2001. Explaining the development of dietary dominance by a single ungulate taxon at Grotte XVI, Dordogne, France. Journal of Archaeological Science. 28:115–125.

Greenwood R. 1961. Quantitative Analysis of Shells From a Site in Goleta, California. American Antiquity. 26:416-420.

Halstead P, Collins P, and Isaakidou V. 2002. Sorting the sheep from the goats: Morphological distinctionons between mandibles and mandibular teeth of adult Ovis and Capra. Journal of Archaeological Science. 29:545-553.

Harasewych MG and Moretzsohn F. 2010. The book of shells: a life-size guide to identifying and classifying six hundred seashells. Chicago: University of Chicago Press.

Hedges SB. 2001. Biogeography of the West Indies: an overview. In Biogeography of the West Indies, perspectives and patterns. Woods CA and Sergile FE, eds. Pp 15-34. Boca Raton: CRC Press.

Hester TR, Shafer HJ, and Eaton JD. 1994. Continuing Archaeology at Colha, Belize. Austin: Texas Archaeological Research Lab.

Higuera-Gundy A, Brenner M, Hodell D A, Curtis JH, Leyden BW, and Binford MW. 1999. A 10,300 14C record of climate and vegetation change from Haiti. Quaternary Research. 52:159–170.

Hodell DA, Curtis JH, Jones GA, Higuera-Gundy A, Brenner M, Binford MA, and Dorsey KT. 1991. Reconstruction of Caribbean climate change over the past 10,500 years. Nature. 352:790-793.

Hofman CL, Bright A, and Hoogland MLP. 2006. Archipelagic resource procurement and mobility in the northern Lesser Antilles: the view from a 3000-year-old tropical forest campsite on Saba. Journal of Island and Coastal Archaeology. 2:145–164.

Hofman CL and Hoogland MLP. 2003. Plum Piece: evidence for archaic seasonal occupation on Saba, northern Lesser Antilles around 3300 BP. Journal of Caribbean Archaeology. 4:1–16.

Hudson MJ, Aoyama M, Hoover KC, and Uchiyama J. 2012. Prospects and challenges for an archaeology of global climate change. Wiley Interdisciplinary Reviews-Climate Change. 3(4):313-328.

120

Jeradino A. 1997. Changes in shellfish species composition and mean shell size from a Late- Holocene record of the west coast of South Africa. Journal of Archaeological Science. 24:1031–1044.

Jones EL. 2004. Dietary evenness, prey choice, and human-environment interactions. Journal of Archaeological Science. 31:307–317.

Jones JG. 1994. Pollen evidence for early settlement and agriculture in northern Belize. Palynology. 18:205-211.

Kaye Q, Fitzpatrick SM, Kappers M, and Thompson V. 2010. Beyond Time Time: Archaeological Investigations at Coconut Walk, Nevis, West Indies. Papers from the Institute of Archaeology. 20:137-147.

Keegan WF. 1994. West Indian Archaeology. 1. Overview and Foragers. Journal of Archaeological Research. 2(3):255-284.

Keegan WF. 1995. Modeling dispersal in the prehistoric West Indies. World Archaeology. 26: 400–420.

Keegan WF. 2000. West Indian archaeology. 3. ceramic age. Journal of Archaeological Research. 8:135–167.

Keegan WF. 2004. Islands of Chaos. In Late Ceramic Age Societies in the Eastern Caribbean. Delpuech A and Hoffman CL, eds. Pp. 33–44. Oxford: Archaeopress.

Keegan WF. 2006. Archaic influences in the origins and development of Taino societies. Caribbean Journal of Science. 1:1–10.

Keegan WF, Fitzpatrick SM, Sealey KS, Lefebvre MJ, and Sinelli PT. 2008. The role of small islands in marine subsistence strategies: case studies from the Caribbean. Human Ecology 36:635-654.

Keegan WF, Portell RW, and Slapcinsky J. 2003. Changes in invertebrate taxa at two pre- Columbian sites in southwestern Jamaica, AD 800–1500. Journal of Archaeological Science. 30:1607–1617.

Kent B. 1988. Making Dead Oysters Talk. St. Mary’s City: Maryland Historical Trust.

Klein RG and Cruz-Uribe K. 1984. The Analysis of Animal Bones from Archaeological Sites. Chicago: University of Chicago Press.

121

Kirch PV. 1980. The Archaeological Study of Adaptation: Theoretical and Methodological Issues. In Advances in Archaeological Method and Theory, vol.3. Shiffer MB, ed. Pp. 101-157. New York: Academic Press.

Kirch PV. 1982 The Impact of the Prehistoric Polynesians on the Hawaiian Ecosystem. Pacific Science. 36:1–14.

Kirch PV. 1984. The Evolution of the Polynesian Chiefdoms. Cambridge: Cambridge University Press.

Kirch PV. 1994. The wet and the dry: irrigation and agricultural intensiﬁcation in Polynesia. University of Chicago Press, Chicago.

Kirch PV. 1997. Microcosmic histories: island perspectives on ‘global’ change. American Anthropologist. 99:30–42.

Kirch PV. 2000. On the road of the winds: an archaeological history of the Paciﬁc islands before European contact. Berkeley: University of California Press.

Kirch PV. 2004. Oceanic islands: microcosms of ‘global change’. In The archaeology of global change: the impact of humans on their environment. Redman CL, James SR, Fish PR, and Rogers JD, eds. Pp. 13–27. Washington DC: Smithsonian Press.

Kirch PV. 2008. Hawaii as a Model System for Human Ecodynamics. American Anthropologist. 109(1): 8-26.

Kirch PV and Green RC. 1987 History, Phylogeny, and Evolution in Polynesia. Current Anthropology. 28:431–456.

Kolipinski MC. 1964. The life history, growth and ecology of four intertidal gastropods. PhD Dissertation. University of Miami. Miami, FL. USA.

Kozuch L and Wing ES. 2004. Animal Remains from Archaeological Sites on Nevis. In The Prehistory of Nevis. Wilson SM, ed. New Haven: Yale University Press.

Krantz GS. 1968. A new method of counting mammal bones. American Journal of Archaeology. 72:286-288.

Kruge MA, Stankiewicz BA, Crelling JC, Montanari A, and Bensley DF. 1994. Fossil Charcoal in Cretaceous-Tertiary Boundary Strata: Evidence for Catastrophic Firestorm and Megawave. Geochimica et Cosmochimica Acta. 58(4):1393-1397.

122

Lasiak T. 1991. The susceptibility and/or resilience of rocky littoral molluscs to stock depletion by the indigenous coastal people of Transkei, southern Africa. Biol Conserv. 56:245–264.

Lalueza-Fox C, Gilbert M, Martinez-Fuentez A, Calafell F, and Bertranpetit J. 2003. Mitochondrial DNA from pre-Columbian Ciboneys from Cuba and the prehistoric colonization of the Caribbean. American Journal of Physical Anthropology. 121:97– 108.

Lane CS, Horn SP, Orvis KH, and Thomason JM. 2011. Oxygen isotope evidence of Little Ice Age Aridity on the Caribbean slop of the Cordillera Central, Dominican Republic. Quaternary Research. 75(3):461-470.

Leach F and Davidson J. 2001. The use of size–frequency diagrams to characterize prehistoric fish catches and to assess human impact on inshore fisheries. International Journal of Osteoarchaeology. 11:150-162.

Lewis JB. 1971. Comparative respiration of some tropical intertidal Gastropods. Journal of Experimental Marine Biology and Ecology. 6:101-108.

Lewis JB, Axelsen F, Goodbody I, Page C, Chislett G, and Choudoury M. 1969. Latitudinal differences in growth rates of some intertidal marine molluscs in the Caribbean. Mar Sci Manusc Rept 12. Montreal: McGill University.

Loreille O, Vigne JD, Hardy C, Callou C, Treinen-Claustre F, Dennebouy N, and Monnerot M. 1997. First distinction of sheep and goat archaeological bones by means of their fossil mtDNA. Journal of Archaeological Science. 24(1):33-37.

Lundberg E. 1989. Preceramic procurement patterns at Krum Bay, Virgin Islands. Ph.D. Dissertation, University of Illinois, Urbana-Champagne. University microfilms, Ann Arbor, Michigan.

Lundberg E. 1991. Interrelationships among preceramic complexes of Puerto Rico and the Virgin Islands. In Proceedings of the Thirteenth International Congress for Caribbean Archaeology. Ayubi EN and Haviser JB, eds. Pp. 73–85. Reports of the Archaeological- Anthropological Institute of the Netherlands Antilles, No. 9. Curacao.

Lupo KD. 2007. Evolutionary foraging models in zooarchaeological analysis: recent applications and future challenges. Journal of Archaeological Research. 15:143-189.

Lyman RL. 1994. Quantitative units and terminology in zooarchaeology. American Antiquity. 59(1):36-71

Lyman RL. 2009. Quantitative Paleozooology. New York: Cambridge University Press.

123

MacArthur RH and Wilson EO. 1967. The Theory of Island Biogeography. Monographs in Population Biology No. 1. Princeton: Princeton University Press.

MacNeish RS and Nelken-Terner A. 1983. Final Report of the Belize Archaic Archaeological Reconnaissance. Boston: Center for Archaeological Studies, Boston University.

Maltby JM. 1985. Patterns in faunal assemblage variability. In Beyond domestication in prehistoric Europe: Investigations in subsistence archaeology and social complexity. Barker G and Gamble C, eds. Pp. 33-74. London: Academic Press.

Mannino MA, and Thomas KD. 2001. Intensive mesolithic exploitation of coastal resources? Evidence from a shell deposit on the Isle of Portland (southern England) for the impact of human foraging on populations of intertidal rocky shore molluscs. Journal of Archaeological Science. 28:1101–1114.

Mannino MA and Thomas KD. 2002. Depletion of a resource? The impact of prehistoric human foraging on intertidal mollusc communities and its significance for human settlement, mobility and dispersal. World Archaeology. 33:452–74.

Martin PS and Steadman DW. 1999. Prehistoric extinctions on islands and continents. In Extinctions in near time: causes, contexts, and consequences. MacPhee RDE and Sues HD, eds. Pp. 17–55. New York: Kluwer.

Mason RD, Peterson ML, and Tiffany JA. 1998. Weighting vs counting: Measurement reliability and the California School of Midden Analysis. American Antiquity. 63(2): 303-324.

McCollough DR. 1996. Spatially structured populations and harvest theory. Journal of Wildlife Management. 60(1):1-9.

Milner N, Barrett J, and Welsh J. 2007. Marine resource intensification in Viking Age Europe: the molluscan evidence from Quoygrew, Orkney. Journal of Archaeological Science. 34:1461–72.

Morgan GS and Woods CA. 1986. Extinction and the Zoogeography of West Indian land mammals. Biological Journal of the Linnean Society. 28:167-203.

Morrison AE, and Cochrane EE. 2008. Investigating shellfish deposition and landscape history at the Natia Beach site, Fiji. Journal of Archaeological Science. 35:2387–2399.

Morrison AE and Hunt TL. 2007. Human impacts to the near-shore environment: A case study from Nualolo Kai, Kauai. Pac Sci. 61:325–345.

124

Muckle R. 1985. Archaeological Considerations of Bivalve Shell Taphonomy, M.A. Thesis, Anthropology, Simon Fraser University.

Muckle RJ. 1994. Differential Recovery of Mollusk Shell From Archaeological Sites. Journal of Field Archaeology. 21(1):129-131

Nagaoka L. 2000. Resource Depression, Extinction, and Subsistence Change in Southern New Zealand. Ph.D. Dissertation. University of Washington, Seattle, WA. USA.

Nagaoka L. 2001. Using diversity indices to measure changes in prey choice at the Shag River Mouth site, New Zealand. International Journal of Osteoarchaeology. 11:101–111.

Nagaoka L. 2002a. The effects of resource depression on foraging efficiency, diet breadth, and patch use in southern New Zealand. Journal of Anthropological Archaeology. 21:419-442.

Nagaoka L. 2002b. Explaining subsistence change in southern New Zealand using foraging theory models. World Archaeology. 34:84-102.

Nagaoka L. 2005. Declining foraging efficiency and moa carcass exploitation in southern New Zealand. Journal of Archaeological Science. 32:1328–1338.

Newsom LA. 2004. Tubers, Fruits, and Fuel: Paleoethnobotanical Investigation of the Dynamics between Culture and the Forested Environment on Nevis, Lesser Antilles. In The prehistory of Nevis. Wilson SM, ed. New Haven: Yale University Press.

Newsom LA and Wing ES. 2004. On Land and Sea: Native American Uses of Biological Resources in the West Indies. Pp 35-57. Tuscaloosa: University of Alabama Press.

Nichol RK and Wild CJ. 1984. “Numbers of individuals” in faunal analysis: The decay of fish bone in archaeological sites. Journal of Archaeological Science. 11(1):35-51.

Nokkert M. 2001. Coconut Walk (CW98), Nevis: Worked Shell from the 1998 ‘Time Team’ Investigations. Unpublished report for the Nevis Heritage Project.

O’Connor TP. 2000. The archaeology of animal bones. London: Sutton Publishing.

Oliver JR. 2001. The Archaeology of Forest Foraging and Agricultural Production in Amazonia. In Unknown Amazon. McEwan C, Barreto C, and Neves E, ed. Pp. 50-85. London: British Museum Press.

Parkington J. 2004. Middens and moderns: Shellfishing and the Middle Stone Age of the western Cape, South Africa. S Afr J Sci. 99:243–247.

125

Payne SB. 1972. On the interpretation of bone samples from archaeological sites. In Papers in economic prehistory. Higgs ES, ed. Pp. 65-81. Cambridge: Cambridge University Press.

Peacock W. 1978. Probabilistic Sampling in Shell Middens: A Case_Study from Oronsay, Inner Hebrides. In Sampling in Contemporary British Archaeology. Cherry J, Gamble C, and Shennan S, eds. Pp. 177-190. Oxford: British Archaeological Reports.

Perkins D. 1973. A critique on the methods of quantifying faunal remains from archaeological sites. In Domestikationsforschung und Geschichte der Haustiere. Matolcsi J, ed. Pp. 367-369. Budapest: Akadémiai Kiadó.

Perkins D Jr and Daly P. 1968. A hunter’s village in Neolithic Turkey. Scientific American. 219(5):96-106.

Petersen JB, Hofman CL, and Curet AL. 2004. Time and culture: Chronology and taxonomy in the eastern Caribbean and the Guianas. In Late Ceramic HUMAN IMPACTS AND ADAPTATIONS IN THE CARIBBEAN 43 Age Societies in the Eastern Caribbean. British Archaeological Reports 1272. Delpuech A and Hofman CL, eds. Oxford: Archaeopress.

Petersen J and Watters D. 1995. A preliminary analysis of Amerindian ceramics from the Trants site, Montserrat. In Proceedings of the XV International Congress for Caribbean Archaeology. Alegr´ıa RE and Rodr´ıguez M, eds. Pp. 131-140. San Juan: Centro de Estudios Avanzados de Puerto Rico y el Caribe.

Plug C and Plug I. 1990. MNI counts as estimates of species abundance. South African Archaeolgoical Bulletin. 45(151):53-57.

Potts GW. 1980. The zonation of rocky littoral areas around Little Cayman. Atoll Res Bull. 241:23-42.

Prummel W. 2005. Molluscs from a Middle Bronze Age site and two Hellenistic sites in Thessaly, Greece. In: Bar-Yosef Mayer, D. (Ed.), Archaeomalacology: Molluscs in Former Environments of Human Behaviour, Proceedings of the 9th ICAZ Conference, Durham, 2002. Bar-Yosef Mayer D, ed. Pp. 107-121. Oxford: Oxbow Books.

Quitmyer I. 1985. Zooarchaeological Methods for the Analysis of Shell Middens at Kings Bay. In Aboriginal Subsistence and Settlement Archaeology of the Kings Bay Locality, vol.2: Zooarchaeology, Department of Anthropology Reports of Investigations No.2. Adams W, ed. Pp. 33-48. Gainesville: University of Florida.

126

Rainbird P. 2002. A message for our future? The Rapa Nui (Easter Island) ecodisaster and Pacific Island environments. World Archaeology. 33(3):436–51.

Rehder A. 1981. Field Guide to North American Seashells. New York: Knopf.

Rehm S and Espig G. 1991. The Cultivated Plants of the Tropics and Subtropics: Cultivation, Economic Value, and Utilization. IAT Institute of Agronomy in the Tropics. Weikersheim: University of Gottingen.

Reitz EJ and Wing ES. 2008. Zooarchaeology, 2nd ed. Cambridge: Cambridge University Press.

Richard G. 1994. Premier indice d’une occupation pre´ce´ramique en Guadeloupe continentale. Journal de la Socie´te´ des Ame´ricanistes. 80:241–242.

Richey JN, Poore RZ, Flower BP, and Quinn TM. 2007. 1400 yr multiproxy record of climate variability from the northern Gulf of Mexico. Geology. 35:423–426.

Rick TC, Erlandson JM, Vellanoweth RL, and Braje TJ. 2005. From Pleistocene mariners to complex hunter-gatherers: The archaeology of the California Channel Islands. Journal of World Prehistory. 19(3):169-228.

Righter E. 1997. The ceramics, art, and material culture of the early Ceramic period in the Caribbean Islands. In The indigenous people of the Caribbean. Wilson SM, ed. Pp 70-79. Gainesville: University Press of Florida

Ringrose TJ. 1993. Bone counts and statistics: A critique. Journal of Archaeological Science. 20(2):121-157.

Robertson R. 2003. The edible West Indian ‘‘whelk’’ Cittarium pica (Gastropoda: Trochidae): Natural history with new observations. Proceedings of the Academy of Natural Sciences of Philadelphia. 153(1):27–47.

Rouse I. 1986. Migrations in prehistory: inferring population movement from cultural remains. New Haven: Yale University Press . Rouse I. 1992. The Tainos: rise and decline of the people who greeted Columbus. New Haven: Yale University Press.

Shackleton J. 1988. Marine Molluscan Remains from Franchthi Cave. Bloomington: Indiana University Press.

127

Siegel PE, Jones JG, Pearsall DM, and Wagner DP. 2005. Environmental and cultural correlates in the West Indies: A view from Puerto Rico. In Ancient Borinquen: Archaeology and Ethnohistory of Native Puerto Rico. Siegel P, ed. Pp. 1–54. Tuscaloos: University of Alabama Press.

Smith B. 1975. Toward a more accurate estimation of meat yield of animal species at archaeological sites. In Archaeozoological studies. Clason A, ed. Pp. 99-108. Amsterdam: North-Holland Publishing Company.

Smith EA and Wishnie M. 2000. Conservation and subsistence in small-scale societies. Annual Review of Anthropology. 29:493-524.

Stager JC and Chen V. 1996. Fossil evidence of shell length decline in queen conch (Strombus gigas L.) at Middleton Cay, Turks and Caicos Islands, British West Indies. Caribb J Sci. 32:14-20.

Steadman DW. (1995) Prehistoric extinctions of Paciﬁc island birds: biodiversity meets zooarcheology. Science. 267:1123–1131.

Steele DG and Parama WD. 1981. Frequencies of dental anomalies and their potential effect on determining MNI counts. Plains Anthropologist. 266(91):51-54.

Stein J, ed. 1992. Deciphering a Shell Midden. San Diego: Academic Press.

Stiner MC and Munro ND. 2002. Approaches to prehistoric diet breadth, demography, and prey ranking systems in time and space. Journal of Archaeological Method and Theory. 9(2):181-214.

Stock C. 1929. A census of the Pleistocene mammals of Rancho La Brea, based on the collections of the Los Angeles Museum. Journal of Mammalogy. 10(4):281-289.

Strutt K. 2003. Hichmans’, Nevis: archaeological survey report March 2003 (SREP2003/1). Southampton: Archaeological Prospection Services, University of Southampton.

Tchernov E. 1993. Exploitation of birds during the Natufian and Early Neolithic of the southern Levant. Archaeofauna. 2:121-143.

Terrell JE. 1997. The Postponed Agenda: Archaeology and Human Biogeography in the Twenty-First Century. Human Ecology. 25(3):419-436.

Thomas DH. 1971. On distinguishing natural from cultural bone in archaeological sites. American Antiquity. 36(3):366-371.

128

Trinkley M and Adams N. 1994. Middle and Late Woodland Life at Old House Creek, Hilton Head Island, South Carolina. Colombia: Chicora Foundation Research Series 42.

Uerpmann HP. 1973. Animal bone finds and economic archaeology: A critical study of “osteoarchaeological” method. World Archaeology. 4(3):307-322.

Underwood AJ. 1975. Comparative studies on the biology of Nerita atramentosa Reeve, Bembicium nanum (Lamarck) and Cellana tramoserica (Sowerby) (Gastropoda: Prosobranchia) in S.E. Australia. Journal of Experimental Marine Biology and Ecology. 18:153-172.

Underwood AJ. 1976. Food competition between age-classes in the intertidal Nerita atramentosa Reeve (Gastropoda, Prosobranchia). Journal of Experimental Marine Biology and Ecology. 23:145-154.

Vitousek PM. 1995. The Hawaiian Islands as a Model System for Ecosystem Studies. Pacific Science. 49:2–16.

Vitousek PM. 2002. Oceanic Islands as Model Systems for Ecological Studies. Journal of Biogeography. 29:573–582.

Vitousek PM. 2004. Nutrient Cycling and Limitation: Hawai’i as a Model System. Princeton: Princeton University Press.

Walker KL. 1992. The Zooarchaeology of Charlotte harbor’s Prehistoric Maritime Adaptation Spatial and Temporal Perspectives. In Culture and Environment in the Domain of the Calusa (Monograph No. 1, Institute of Archaeology and Paleoenvironmental Studies). Marquardt W, ed. Pp. 265-366. Gainesville: University of Florida.

Warmke GLC, 1961. Caribbean seashells; a guide to the marine mollusks of Puerto Rico and other West Indian islands, Bermuda and the Lower Florida Keys. Narberth: Livingston Publishing Company.

Waselkov G. 1987. Shellfish Gathering and Shell Midden Archaeology. In Advances in Archaeological Method and Theory vol.XI. Schiffer M, ed. Pp. 93-210. San Diego: Academic Press.

Watson JPN. 1972. Fragmentation analysis of animal bone samples from archaeological sites. Archaeometry. 14(2):221-228.

Watson JPN. 1979. The estimation of the relative frequencies of mammalian species: Khirokitia 1972. Journal of Archaeological Science. 6(2):127-137.

129

Watters DR, Donahue J, and Struckenrath R. 1992. Paleoshorelines and the prehistory of Barbuda, West Indies. In Paleoshorelines and prehistory: an investigation of method. Johnson LL, ed. Pp. 15–52. Boca Raton: CRC Press.

Watts D. 1987. The West Indies: Patterns of Development, Culture and Environmental Change since 1942. Cambridge: Cambridge University Press.

Whitaker A. 2010. Prehistoric behavioral depression of cormorant (Phalacrocorax spp.) on the northern California coast. Journal of Archaeological Science. 37:2562–2571.

White TE. 1952. Observations on the butchering techniques of some aboriginal peoples: 1. American Antiquity. 17(4):337-338.

White TE. 1953. A method of calculating the dietary percentage of various food animals utilized by aboriginal peoples. American Antiquity. 18(4):396-398.

Williams EE. 1989. Old Problems and New Opportunities in West Indian Biogeography. In Biogeography of the West Indies: Past, Present, and Future. Woods CA, eds. Pp. 1-46. Gainesville: Sandhill Crane Press.

Wilson SM. 1989. The Prehistoric Settlement Pattern of Nevis, West Indies. Journal of Field Archaeology. 16(4):427-450.

Wilson SM. 2001. The Prehistory and Early Hisotry of the Caribbean. In Biogeography of the West Indies: Patterns and Perspectives. 2nd ed. Woods CA and Sergile FE, eds. Pp. 519-527. Boca Raton: CRC Press.

Wilson SM. 2007. The prehistory of Nevis, a Small Island in the Lesser Antilles. Yale University Publications in Anthropology No 87. New Haven: Yale University Press

Wilson SM, Iceland HB, and Hester TR. 1998. Preceramic connections between Yucatan and the Caribbean. Latin American Antiquity. 9(4):342-352.

Wing ES. 2001. The sustainability of resources used by native Americans on four Caribbean islands. International Journal of Osteoarchaeology. 11:112-126.

Wing E and Brown A. 1979. Paleonutrition: Method and Theory in Prehistoric Foodways. Orlando: Academic Press.

Wing E and Quitmyer I. 1985. Screen Size For Optimal Data Recovery: A Case Study IN Aboriginal Subsistence and Settlement Archaeology of the Kings Bay Locality, vol. 2: Zooarchaeology. Adams W, ed. Pp. 49-59. Gainesville: University of Florida.

130

Wing ES and Wing SR. 1995. Prehistoric Ceramic Age adaptation to varying diversity of animal resources along the West Indian Archipelago. Journal of Ethnobiology. 15:119– 148.

Wing SR and Wing ES. 2001. Prehistoric fisheries in the Caribbean. Coral Reefs. 20:1-8.

Wolverton S, Nagaoka L, Densmore J, and Fullerton B. 2008. White-tailed deer harvest pressure and within bone nutrient exploitation during the mid to late Holocene in southeast Texas, USA. Before Farming 2, article 3.

Wolverton S, Nagaoka L, Don P, and Kennedy JH. 2012. On behavioral depression in white- tailed deer. Journal of Archaeological Method and Theory. 19:462-489.

Wroe S, Field J, and Grayson D. 2006. Megafaunal extinction: Climate, humans, and assumptions. Trends in Ecology and Evolution. 21(2):61–2.

Zeder MA and Arter SA. 2007. Meat consumption and bone us in a Mississippian village. In Case studies in environmental archaeology (2nd ed.). Reitz EJ, Scarry CM, and Scudder SJ, eds. Pp. 329-347. London: Springer.

Zeder MA, Emshwiller E, Smith BD, and Bradley DG. 2006. Documenting domestication: the intersection of genetics nad archaeology. Trends in Genetics. 22(3):139-155.

Zeitlin RN and Zeitlin JF. 1996. The Paleoindian and Archaic Cultures of Mesoamerica. IN The Cambridge History of the Native Peoples of the Americas. Adams REW and MacLeod M, eds. Pp. 45-121. New York: Cambridge University Press.

Ziegler AC. 1973. Inference from prehistoric faunal remains. Reading: Addison-Wesley Module in Anthropology 43.

Zuschin M, Stachowitsch M, and Stanton Jr RJ. 2003. Patterns and processes of shell fragmentation in modern and ancient marine environments. Earth-Science Reviews. 63(1- 2):33-82.

131

APPENDICIES

132

APPENDIX A

Table 26. Species weight, MNI, and NISP.

133

Table 26. Continued.

134

Table 27. Percentage of unit weight by species.

135

Table 27. Continued.

136 Table 2: Square composition.

Table 28. Percentage of MNI by species.

137

Table 28. Continued.

138

Table 29. Key species.

139

Table 29. Continued.

140

APPENDIX B

Table 30. Weight errors. Percentage of samples that lie outside the 95% CI for the trench average weight. Note: A sample is only counted once in the sample total, even if it lies outside the CI on more than one planum.

Table 31. MNI errors. Percentage of samples that lie outside the 95% CI for the trench average MNI.

Table 32. NISP errors. Percentage of samples that lie outside the 95% CI for the trench average NISP.

141

Table 33. Species count errors. Percentage of samples that lie outside the 95% CI for the trench average species count. Note: A sample is only counted once in the sample total, even if it lies outside the CI on more than one planum.

142

Table 34. Sample composition with three selected species.

143

Table 34. Continued.

144

Table 35. Species found in each sample.

145

Table 35. Continued.

146

Table 36. Percentage of N. tessellata samples that lie outside the 95% CI for the trench average on at least one plana. Note: A sample is only counted once in the overall, even if it lies outside the CI on more than one category.

Table 37. Percentage of C. pica samples that lie outside the 95% CI for the trench average on at least one plana.

Table 38. Percentage of L. tuber samples that lie outside the 95% CI for the trench average on at least one plana.

147