Citation
Ceramic notes

Material Information

Title:
Ceramic notes occasional publications of the Ceramic Technology Laboratory, Florida State Museum
Uniform Title:
Ceramic notes (Florida State Museum. Ceramic Technology Laboratory)
Distinctive title:
Annotated bibliography of ceramic studies
Creator:
Ceramic Technology Laboratory (Florida State Museum)
Place of Publication:
Gainesville, Fla
Publisher:
Ceramic Technology Laboratory [Florida State Museum], and Florida State Museum Associates
Publication Date:
Frequency:
Irregular
completely irregular
Language:
English
Edition:
No.2
Physical Description:
v. : ; 28 cm.

Subjects

Subjects / Keywords:
Ceramics ( lcsh )
Pottery ( lcsh )
Indian pottery ( lcsh )
Ceramics ( fast )
Indian pottery ( fast )
Pottery ( fast )
Genre:
serial ( sobekcm )

Notes

Dates or Sequential Designation:
No. 1-

Record Information

Source Institution:
University of Florida
Holding Location:
Florida Museum of Natural History, Anthropology Division
Rights Management:
Copyright, Florida State Museum of the University of Florida. Permission granted to University of Florida to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
09611470 ( OCLC )
ocm09611470

UFDC Membership

Aggregations:
University of Florida Institutional Repository

Full Text
CERAMIC NOTES no. 2
OCCASIONAL PUBLICATIONS OF THE CERAMIC TECHNOLOGY
LABORATORY, FLORIDA STATE MUSEUM
CERAMIC TECHNOLOGY AT A WEEDEN ISLAND PERIOD
ARCHAEOLOGICAL SITE IN NORTH FLORIDA
ANN S, CORDELL




The James A. Ford
Library of Anthropology
Florida Museum of Natural History, SAnthropology Division
FLORIDA
Gift of.- X4 o "F(- C-TL
CERAMIC NOTES: OCCASIONAL PUBLICATIONS OF THE CERAMIC TECHNOLOGY LABORATORY, FLORIDA STATE MUSEUM, is a publication series for articles and monographs dealing with anthropological and archaeological studies of pottery and related materials. The series is edited by Prudence M. Rice, with assistance from Ann S. Cordell, Jeffrey Mitchem, and Donna Ruhl, and is produced by the Ceramic Technology Laboratory. Address all correspondence to P.?4. Rice, Florida State Museum, Gainesville, Florida 32611.
Checks for CERAMIC NOTES no. 2- should be made out in the amount of $8.00, payable to UNIVERSITY OF FLORIDA FOUNDATION. Postage is included.
Copyright Q 1984 b~y thie Florida State Museum of the University of Florida.




CERAMIC NOTES NO. 2
CERAMIC TECHNOLOGY AT A WEEDEN ISLAND PERIOD
ARCHAEOLOGICAL SITE IN NORTH FLORIDA
ANN S. CORDELL







FOREWORD
by Jerald T. Milanich
During the period A.D. 200 to 900, Weeden Island peoples flourished in northern Florida and adjacent portions of southeastern Alabama and southwestern Georgia. The magnificent Weeden Island fired clay vessels and effigy figurines, perhaps the finest ceramics manufactured by the aborigines of the eastern United States, were traded north to the fall line, west to the Mississippi River valley, and into south Florida. The importance of Weeden Island in the Southeast was considerable.
Ann Cordell, an archaeologist with the Florida State Museum, began her research on Weeden Island ceramics as part of the museum's longterm project focusing on the settlement and economic systems, social and political organization, and mortuary ritual of that culture. The research program addressed itself to questions concerning the organization of Weeden Island society. Specifically, was it a stratified and ranked system with ascribed statuses, chiefs, monumental architecture, large populations, and craftspersons? The answer is, no, Weeden Island cultures were not chiefdoms; true chiefdoms apparently did not appear in the Southeast until the Mississippian cultures after A.D. 900 or later.
In this monograph, which represents publication of Cordell's master's thesis in anthropology, she provides a detailed ceramic analysis of pottery from both village and ceremonial contexts at the McKeithen Weeden Island site in Columbia County, North Florida, and on locally available clays. Her procedures allow quantified conclusions regarding craft specialization, the dichotomy between mound and village wares, and the location of manufacture of the pottery. These data, the first of their kind for a Weeden Island ceramic assemblage, provide a greater understanding of the evolutionary position and nature of Weeden Island society.
The generous support of the A. Fillmore Wentworth Foundation of Clearwater, Florida, and its President, William M. Goza, for both the excavations at the McKeithen site and the publication of this manuscript, is gratefully acknowledged.
iii




ACKNOWLEDGMENTS
This study was made possible by funding from
the National Science Foundation and the A. Fillmore Wentworth Foundation of Clearwater, Florida, and through the generous cooperation of the Leon A. McKeithen family of Live Oak, Florida. The data collection phase of the study was carried out using the excellent facilities of the Florida State Museum Ceramic Technology Laboratory.
This manuscript benefited from constructive criticisms of Prudence Rice, Jerald Milanich, and Charles Fairbanks. These individuals, plus Brenda Sigler-Eisenberg, Timothy Kohler, James Knight, Marian Saffer, Jeffrey Mitchem, Donna Ruhl, and Sharon Hall, offered helpful comments during the various stages of planning, data collection, and interpretation.
The clay samples used in this study were collected by Prudence Rice and her students, as well as Timothy Kohler, Brenda Sigler-Eisenberg, and the late L. Jill Loucks.
The computer program to calculate Gower's
Coefficient was written by Roger Blashfield, who also lent his expertise in interpreting the cluster analyses. Marian Saffer and Kenneth Wilkins provided further assistance in solving CLUSTAN problems.
Deborah Harding inked Figure 15.
The manuscript was produced using SCRIPT with MUSIC and TCP in conjunction with the computing facilities of the Northeast Regional Data Center, University of Florida, and the laser printing facilities of the University of Chicago. Sam Wilson assisted the printing at the University of Chicago.
Publication of this manuscript was generously
funded by the A. Fillmore Wentworth Foundation. The Foundation and its President, William M. Goza, are gratefully acknowledged.
iv




TABLE OF CONTENTS
CHAPTER PAGE
FOREWORD ...................................... iii
ACKNOWLEDGMENTS .......................... iv
LIST OF TABLES ........................ x
LIST OF FIGURES .............................. xiii
ABSTRACT ......................... .. .... xv
ONE INTRODUCTION
Definition of Weeden Island and
Weeden Island Pottery ....................... 1
Role of Pottery in Weeden Island Research... 1 The Present Study .... ..... ........... ..... 3
Archaeological/Cultural and
Geographical Context...................... 4
Definition of Goals....................... 7
TWO PREVIOUS RESEARCH
Chronological Studies........... .... ....... 9
Typological Studies ............. o ...... 11
Studies of Regional or Inter-site
Ceramic Spatial Variability .................. 12
Studies of Intra-site Ceramic
Spatial Variability .......... ... o....... ... 14
Kolomoki, Georgia .......................... 14
McKeithen, Florida........ ...... .. .... 16
Summary....... .... ... ............ ... . 20
THREE METHODS
Ceramic Technology and Ceramic Ecology ...... 23
Description of Pottery Samples and
Sampling Procedures............ .......... ... 24
Description of Variables and
Methods of Data Collection.................. 35
Primary Variables ....... ........ ......... 37
Aplastic composition ................... 37
Refired porosity.................. ... .. 40
v




Refired core colors ..................... 41
Data Analysis of Primary Variables ........ 42 Secondary Variables ....................... 47
Surface decoration ...................... 48
Vessel form ............................. 48
Manufacturing technology ................ 48
Color ................................. 48
Hardness .............................. 49
Porosity ...... : ....................... 49
Refiring experiments .................. 49
Data Analysis of Secondary Variables .... 52
Description of Clay Samples and
Methods of Analysis ..................... 52
Summary ..................................... 55
FOUR NORTH FLORIDA CERAMIC ENVIRONMENT
Introduction ................................ 57
Climate ..................................... 58
Geological History .......................... 59
Topography .................................. 62
Soils ....................................... 63
Sample Clay Analyses ........................ 65
Plasticity and Handling Characteristics ... 66 Drying Behavior ........................... 69
Plastic Composition, Particle
Size, and Proportion ...................... 70
Plastic composition .................... 70
Texture: particle size and proportion ... 71 Inclusion point-counts .................. 74
Firing Behavior ........................... 74
Color and coring ........................ 75
Percent firing weight loss (%FWL) ....... 76 Porosity ................................ 77
other observations ...................... 77
Summary and Conclusions: the Effective
Ceramic Environment ......................... 77
FIVE CLUSTER ANALYSIS RESULTS
Introduction ................................ 81
Procedures .................................. 82
Definition and Interpretation
of Clusters ................................. 84
Micaceous Paste Clusters .................. 85
Sponge Spiculite Paste Clusters ........... 93
Quartz Sand-Grit Paste Clusters ........... 98
Summary and Discussion ...................... 103
vi




six SACRED-SECULAR CERAM IC DICHOTOMY
Introduction. ....................... .. .. ... .. .. .109
Variables and Procedure.......... .......... .109
Surface Decoration ... ......................... .. .111
Vessel Form ....... .. *. . .. .. ................. 113
Degree of Craftsmanship: Resource
General class of paste .............. 1
Cluster affiliation ..................... 116
M icaceous paste. ............................... 116
Sponge spiculite pas. e............119
Quartz sand-grit paste ................. 119
Summary.. .. .................... .....119
Degree of Craftsmanship: Manufacturin~g
Technology. ........................................... 120
Vessel manufacture ............................. 120
Appliqued or mod eled features ......... 121 Vessel wall thickness.................. 121
Surface finishing... ................. .. .. .123
Summary: vessel manufacture ........... 125
Original firing conditions .............. 125
Surface color and core color/degree
Hardness. .......................... ... .. .. .132
Percent firing weight loss (%FWL) ..... 137 Refired color/coring changes .......... 140 Summary: original firing conditions ... 144 Summary and Conclusions ......... ... ........ .146
SEVEN MCKEITHEN SITE POTTERY PRODUCTION
origin of Manufacture ....................... 153
Introduction.. .................. .. .. .. .. .. .. .153
Procedures. ... .. .. .... .. .. .. .. .. .. .. ..... .154
origin of manufacture: Resource Groupings.155 Origin of Manufacture: Pottery Types ...... 159
Pottery Specialization at the McKeithen,
Procedures ....... ... .. .. .. .. .. .. .. .. ..... .166
Results. ....... .. .... 6* * .. .. .. .. .. ..... .167
Vessel Form ....... .. .. .. .. .. .... .. .. ...167
Description.. .................... .. .. .. .. .167
Interpretation ............... .. ... .. ... 172
Resource Utilization. .. .. .. .. .. .. .. .. ..... 174
General class of paste ...... ... ... ... ... .174
vii




Interpretation .......................... 176
Cluster affiliation ..................... 176
Interpretation ................ ........ 177
Manufacturing Technology .................. 178
Description ............................. 178
Interpretations ......................... 188
Summary and Conclusions ..................... 193
EIGHT CONCLUSIONS AND RECOMMENDATIONS ............... 197
APPENDICES
A MCKEITHEN MIDDEN PROVENIENCES--SAMPLING
POPULATION ...................... .............. 201
B SAMPLING ELIGIBILITY ............................ 202
C TEMPORAL AND AREAL BREAKDOWN OF SAMPLING
POPULATION ACCORDING TO DECORATIVE/PASTE
SAMPLING CATEGORIES ........... ................ 202
D DATA ON PRIMARY VARIABLES ..................... 203
E DATA ON SECONDARY VARIABLES ................... 206
F SOIL TYPES IN COLUMBIA AND SUWANNEE COUNTIES..212
G FIFTEEN MOST ABUNDANT SOIL TYPES IN COLUMBIA
AND SUWANNEE COUNTIES ....................... 213
H CONTEXT OF COLLECTION FOR CLAY SAMPLES ........ 214
I WORKING CHARACTERISTICS AND PLASTICITY ........ 227
J SHRINKAGE .............................. 227
K APLASTIC CONSTITUENTS ......................... 227
L PARTICLE SIZE AND PROPORTION: SIEVING
RESULTS ....................................... 228
M USDA TEXTURE, AND APLASTIC PROPORTIONS
ACCORDING TO WENTWORTH'S SIZE NAMES ........... 229
N DATA ON PRIMARY VARIABLES .................... 230
0 COLOR CHANGE WITH INCREASING TEMPERATURE ...... 231
P PERCENT FIRING WEIGHT LOSS WITH INCREASING
TEMPERATURE .................... ........... 232
viii




Q RELATIVE STRENGTH AT DIFFERENT FIRING
TEMPERATURES.................................. 232
R VARIABILITY IN PHYSICAL PROPERTIES ACCORDING
TO PASTE.............. ........................233
S CORING VS. COLOR CHANGE AT 5000C ..............234
T SOURCE OF HARDNESS VARIABILITY................234
U ORIGINAL VS. FIRED CHARACTERISTICS............235
V DECORATIVE/FORMAL VARIETIES FOR WEEDEN ISLAND
RED AND WEEDEN ISLAND/PAPYS BAYOU PUNCTATED
POTTERY .................. .......... .. .... 236
BIBLIOGRAPHY ............................................. 237
ix




LIST OF TABLES
Table 3-1 Typological Breakdown for McKeithen Mound
Pottery ....*. .......... ........... ...... 26
Table 3-2 Typological Breakdown for McKeithen Midden
Pottery Sampling Population ................ 34
Table 3-3 Typological Breakdown for McKeithen Midden
Pottery Sample........ ...... .. .. . ... .. .. 36
Table 3-4 Broad Categories of Refired Color........... 43
Table 3-5 Categories of Original Surface Color ....... 50
Table 4-1 Relative Categories for Sample Clay
Physical Properties: Handling Characteristics, Plasticity, and Shrinkage .......... 68 Table 4-2 Relative Categories for Sample Clay
Physical Properties: USDA Texture, Relative
Paste Texture, and Fired Color ............. 73
Table 5-1 Ordinal Frequency Categories for Total
Number Quartz Inclusions, Ratio of "Fine"
to "Very Fine" Sand, Refired Percent Apparent Porosity, and Relative Paste
TalT-escrpe Statistics...for..Micaceous...Paste
Table 5-3 Descriptive Statistics for Spongeou at
piuiPseClusters. ......... ................... 89
Table 5-4 Descriptive Statistics for Quartz
Sand-Grit Paste Clusters .............. 101
Table 6-1 Sacred-Secular Variability in Surface
Tal -2 SeSc ulario Variability.. in..Vessel.For...112
Table 6-3 Sacred-Secular Variability in Vestel Form..117
Tale6- ScedSeulrVaiailt i Pst .....x1




Table 6-4 Sacred-Secular Variability in Paste,
Controlling for Decoration ................. 118
Table 6-5 Sacred-Secular Variability in Vessel
Manufacture ................................ 122
Table 6-6 Sacred-Secular Variability in Vessel
Manufacture, Controlling for Decoration .... 124
Table 6-7 Sacred-Secular Variability in Physical
Properties, Micaceous Paste Subsample:
Context vs. Variables ...................... 128
Table 6-8 Sacred-Secular Variability in Physical
Properties, Micaceous Paste Subsample: Context vs. Variables, Controlling for
Decoration ................................. 129
Table 6-9 Sacred-Secular Variability in Physical
Properties, Spiculite Paste Subsample:
Context vs. Variables ...................... 130
Table 6-10 Sacred-Secular Variability in Physical
Properties, Spiculite Paste Subsample: Context vs. Variables, Controlling for
Decoration ................................. 131
Table 6-11 Sacred-Secular Variability in Physical
Properties, Quartz Sand-Grit Paste
Subsample: Context vs. Variables ........... 133
Table 6-12 Sacred-Secular Variability in Physical
Properties, Quartz Sand-Grit Paste
Subsample: Context vs. Variables,
Controlling for Decoration ................. 134
Table 6-13 Sacred-Secular Variability in %FWL .........139
Table 6-14 Sacred-Secular Variability in Color Change
at 5000C ...................................143
Table 6-15 Summary of Trends in Sacred-Secular
Variability ............................ ..... 147
Table 7-1 Summary of Cluster Characteristics and
Clay-Cluster Matches ....................... 157
Table 7-2 Paste Variability According to Pottery
Type ....................................... 161
Table 7-3 Summary of Origin of Manufacture ........... 163
xi




Table 7-4 Vessel Form Variability According to
Table 7-5 Paste Variability According to Pottery
Table 7-6 Thickness Variability According to
Table 7-7 Variability in Surface Finishing According
to Pottery Type: Exterior Surface .......... 180 Table 7-8 Variability in Surface Finishing According
to Pottery Type: Interior Surface .......... 181 Table 7-9 Variability in Degree of Luster ............ 182
Table 7-10 Variability in Degree of Color Change at
Table 7-11 Cluster Composition in Terms of Relative
Values Representing Patterns of Technological Variability ........................ 186
Table 7-12 Variability Among Technological Clusters...187 Table 7-13 Variability in General Class of Paste and
Vessel Form Among Technological Clusters...189 Table 7-14 Variability in Physical Properties Among
Technological Clusters.......... .. ........ .0191
xii




LIST OF FIGURES
Figure 1 Known Geographical Extent of Weeden Island
and Weeden Island-Related Cultures......... 2 Figure 2 North Florida--McKeithen Weeden Island
Subregion .................................. 5
Figure 3 McKeithen Site, Columbia County............ 6
Figure 4 Mound C, McKeithen Site .................... 27
Figure 5 Mound B, McKeithen Site.................... 28
Figure 6 McKeithen Site, Columbia County............ 29
Figure 7 McKeithen Site Principal Components
Seriation.................................. 31
Figure 8 Example of a Gower's Similarity Matrix
Using Ten Entities......................... 46
Figure 9 Categories Used to Measure Core Color/
Degree of Coring........................... 51
Figure 10 Geological Exposures in Columbia and
Suwannee Counties.......................... 60
Figure 11 Approximate Location for the Clay Samples.. 67
Figure 12 Micaceous Paste Subsample--Average Linkage
Cluster Solution........................... 88
Figure 13 Spiculite Paste Subsample--Complete
Linkage Cluster Solution ................... 94
Figure 14 Quartz Sand-Grit Paste Subsample--Complete
Linkage Cluster Solution ................... 99
Figure 15 Examples of Variability in Vessel Wall
Orientation and Vessel Form................169
Figure 16 A--Simpson Collection Plate (FSM 102654),
Illustrating Vessel Form Exhibited by
Mound B Specimens; B--Mound B Plate and
Dish Fragments............................. 170
xiii




Figure 17 A--Vessel 4, Mound C--Derived Effigy with
Cut-outs; B--Vessel 15, Mound C--Derived
Effigy; C--"Wing-nut" Vessel 12, Mound C-Incurving Bowl............................. 171
Figure 18 Technological Clusters--Ward's Method
Cluster Solution........................... 185
xiv




Abstract
This study presents the results of ceramic
technological/ecological analysis of Weeden Island pottery from the McKeithen site in North Florida. Goals addressed include identification of 1) patterns of resource selection and origin of manufacture for the pottery; 2) the source of sacred-secular differences in the pottery; and 3) degree of specialization in pottery manufacture at this site. Two-hundred forty-five archaeological specimens, 33 from mound contexts and 212 from contemporaneous midden contexts, were selected for study. Their fundamental physical properties and manufacturing technology were described. Multivariate cluster analyses of these data grouped the samples into classes interpretable as distinctive clay pastes and/or resources.
At least 11 distinctive kinds of clays were consistently used in manufacture of the pottery sampled. Comparisons with 26 locally available clays suggest that at least seven of the paste groups may have been locally made, including the majority of all non-spiculite paste pottery in the sample. Northwest Florida-southwest Georgia origins are postulated for some of the Weeden Island Incised pottery at the site. Some of the Weeden Island Zoned Red pottery may also be nonlocal. Spiculite paste ceramics are thought to have east and Central Peninsular Gulf Coast origins. Sacred-secular differences in methods and materials of manufacture were found to relate to decoration, and not context of deposition. Some degree of pottery specialization was identified for six 'elite' pottery types in the Weeden Island series. The results support Rice's multi-centered model for manufacture of effigy and other Weeden Island series pottery found typically in ceremonial contexts.
xv







CHAPTER ONE
INTRODUCTION
Definition of Weeden Island and Weeden Island Pottery
'Weeden Island' refers to a prehistoric
archaeological culture found in northern Florida and adjacent parts of Georgia and Alabama. The concept of 'Weeden Island' includes as yet poorly understood patterns of sociopolitical organization associated with a mortuary-related ceremonial complex and a ceramic complex that is diverse in decoration and vessel form. These phenomena are present in several archaeologically definable cultures inhabiting different environmental zones (Fig. 1) within a broad geographical region (Milanich and Fairbanks 1980:96; Milanich 1976:1, 4-5). Weeden Island and Weeden Island-related cultures (those exhibiting only occasional features of either complex) have been under investigation in recent years by a number of Florida researchers.
Weeden Island pottery has been described as "the most outstanding of the Gulf Coast and, in many respects, of the entire eastern United States" (Willey 1949:406). It consists of a large group of decorated and undecorated pottery types. Incision and punctation are two primary decorative techniques which are used to produce a variety of stylized zoomorphic and geometric designs. Complicated stamping, check stamping, and painting are other prominent decorative modes. Vessel forms are also diverse; plate, bowl, jar, beaker or vase, and unusually-shaped effigy forms have all been recognized.
Role of Pottery in Weeden Island Research
As the most ubiquitous and often the only category of non-lithic material remains recovered by excavation, pottery has been the primary tool used to investigate the nature and evolution of, and relationships between, Weeden Island societies. Chronological, typological, and inter-site spatial variability of Weeden Island pottery have been traditional topics of study. Primary
1




Figure 1 Known Geographical Extent of Weeden Island and
Weeden Island-Related Cultures
2




research objectives have been culture history-related and include efforts to: 1) obtain temporal placement for sites yielding Weeden Island pottery; 2) provide a comparative descriptive reference of the pottery for future classificatory purposes; 3) define the geographical extent of Weeden Island culture, and delineate and describe subregions within the larger cultural region; and 4) describe patterning of Weeden Island settlements. These goals have been investigated through the creation and use of a pottery typology with temporal significance, one initially formed on the basis of decorative, formal, and technological attributes. Subsequent use of this typology, however, has generally considered only decorative and formal attributes. More recently, intra-site ceramic spatial variability has been added to the list of research topics.
Current research objectives include attempts to make inferences regarding the cultural processes involved in the origin, evolution, and interaction of Weeden Island sociopolitical, religious, and economic organization, as well as those culture history-related objectives listed above. These goals are being investigated through chronological and typological refinements and through problem-oriented ceramic analyses that place increasing emphasis on the study of the technological variability of Weeden Island pottery.
The Present Study
The nature of production of Weeden Island ceramics has been a recurring topic for supposition and investigation in studies of Weeden Island cultures. Willey implied that some classes of Weeden Island ceramics may have been the result of specialist production by describing the pottery as "the most outstanding of the Gulf Coast..." (Willey 1949:406). William Sears took this implication many steps further by suggesting criteria to differentiate Weeden Island pottery into 'sacred' and 'secular' categories (Sears 1973), inferring full-time ceramic specialization for the former. Recently, efforts to substantiate the existence of specialized production of Weeden Island pottery have been systematically carried out on pottery from the McKeithen site in northern Florida (e.g. Kohler 1978a; Rice 1980). The present study was directed toward further systematic investigation of the nature of production of Weeden Island pottery on the basis of ceramic samples furnished by this site.
3




Archaeological/Cultural and Geographical Context
The McKeithen site (8-Co-17), a Weeden Island
period mound-village complex dating within the period A.D. 150-750 (Cordell et al. 1979; Kohler 1978a; Milanich et al. 1978; Milanich and Fairbanks 1980) is located in western Columbia County within the cultural subregion of North Florida (Fig. 2). This subregion, as defined by Milanich (1976:10), lies north of the Santa Fe River and south of the Okeefenokee Swamp and wiregrass-pine barrens of southeast Georgia. The western boundary is placed at the Aucilla River; the eastern boundary is at the beginning of the Atlantic coastal flatlands just east of Lake City, Florida. Most of the area corresponds roughly to present-day Columbia and Suwannee counties.
The McKeithen site consists of three earthen mounds arranged in an isosceles triangle that is surrounded by (and in some places overlies) a horseshoe-shaped village midden (Fig. 3). This mound and village arrangement surrounds a 'plaza' area nearly devoid of cultural material (Kohler 1978a:42). Mound excavations have revealed that one mound, Mound A, functioned as a locus for charnel activities. Another, Mound C, represents the place of final interment of cleaned and bundled skeletons. The third, Mound B, contained the remains of a structure believed to have been the residence of the individual responsible for coordination of mortuary activities carried out at Mounds A and C; this individual was interred in a shallow grave pit dug into the floor of the structure. The middle phase of village occupation, A.D. 250-550 (Kohler 1978a), encompasses the period of operation of mortuary activities at these locations dated ca. A.D. 375-495 radiocarbon years (Cordell et al. 1979; Milanich et al. 1978; Milanich and Fairbanks 1980). These latter dates may more correctly date the middle phase occupation.
Evidence for some degree of ranking or status
differentiation at the site is suggested by the special treatment awarded the site's 'religious specialist' or 'mortuary director' (Milanich et al. 1978; Milanich and Fairbanks 1980) and by non-random distribution of hypothesized 'elite' and/or nonlocal items in the village (Kohler 1978a; 1980). The settlement structure, site artifact inventories, regional environments, and distribution of resource opportunities for the surrounding North Florida region suggest a segmentary lineage framework for Weeden Island social organization in this area and suggest that the development of ranking evidenced at the
4




* ...ORGIA Okefenok~ee
L OR-D.. Swamp
. . . ....
0I
Pe N
0 24 48
,73 km
Figure 2 North Florida--McKeithen Weeden Island Subregion




44
00
\ me.. \
'O'
Figure 3 McKeithen Site, Columbia County
SOURCE: Milanich and Fairbanks 1980:133
6




McKeithen site was anchored in the prevailing socioreligious institution (Sigler-Lavelle 1980a, 1980b).
Selection of the McKeithen site for the present investigation was based on two pragmatic considerations: 1) the McKeithen site is the only Weeden Island period ceremonial and village complex to have been systematically excavated since Sears' work at the Kolomoki site (Sears 1956); and 2) certain aspects of previous researches at the site have focused directly (Rice 1980) and indirectly (Kohler 1978a) upon the nature of production of McKeithen site Weeden Island pottery. The archaeological specimens examined in this study were furnished by excavation of two mounds (B and C) and selected middle phase village midden proveniences.
Definition of Goals
An initial goal of the present study was to conduct an objective analysis of Sears' 'sacredsecular' model for production of Weeden Island pottery by means of a ceramic technological analysis of pottery from temporally controlled mound or 'sacred', plus midden or 'secular', contexts. Ceramic technology involves examination and description of physical, mineralogical, and chemical properties of ceramic materials with reference to relatively objective, precise, and replicable standards of measurement and observation (Matson 1951, 1963; Rice 1976; Rye 1980; Shepard 1976), and has been used successfully in study of problems such as specialization in manufacture of pottery in Mesoamerica (e.g. Rice 1976, 1978a, 1979; Shepard 1963, 1964).
The goal of evaluation of Sears' model for
production of Weeden Island pottery provided a suitable take-off point for investigation of other related aspects of production of McKeithen site ceramics. For example, determination of local versus nonlocal manufacture for the McKeithen samples was attempted through comparison of the ceramic (sherd) data with results of similar analyses of clay samples local to the McKeithen site and North Florida subregion. This was also approached by evaluating the McKeithen technological data against Rice's (1980) trace element data-based model for production of McKeithen ceramics. Description of the nature of the McKeithen area ceramic environment (raw materials and potential for availability of ceramic resources), which permitted comparison with McKeithen ceramics in hypothesizing origin of production, also provided an environmental context for understanding the development of ceramic specialization in the area. Finally, formal
7




identification and description of products of specialized manufacture at the McKeithen site is based on a set of recommendations from Rice (1981) for archaeological recognition of pottery specialization using ceramic technological data. The broader or ultimate purpose of this intensive investigation of 'ceramic technology' at the McKeithen site is to contribute a body of comparative ceramic data to permit evaluation of questions of ceramic production on a regional level.
8




CHAPTER TWO
PREVIOUS RESEARCH
Previous Weeden Island ceramic studies have examined pottery from a variety of viewpoints-chronological, typological, and in terms of inter- and intra-site variability. The purpose of this chapter is to provide an overview of certain studies representing the above foci. In many cases the studies have suggested new questions or avenues of investigation that, together with their conclusions, form the background for the present study.
Chronological Studies
Gordon R.. Willey was the first to describe the Weeden Island archaeological complex in detail in Archeology of the Florida Gulf Coast (1949). This work includes extensive pottery typologies which were created to construct a relative chronology for the prehistoric cultures of the Gulf Coast region, and to serve as a comparative descriptive reference for subsequent research. Weeden Island pottery types recovered from stratigraphic test excavations at five midden sites in northwest Florida were seriated to produce a two-subperiod chronology for the Weeden Island culture (Willey 1949:100, Fig. 14; 396-397). Relative frequencies of incised, punctated, complicated stamped, and check stamped pottery types were the diagnostic criteria for chronological placement of Weeden Island sites within the two-subperiod scheme.
Willey's Weeden Island typology and
interpretations regarding chronology focused on the coastal distribution and manifestation of Weeden Island (see Map I in Willey 1949: after p. 2). Subsequent fieldwork, which has since widened the geographical extent of Weeden Island culture, has revealed the uneven distribution of Weeden Island village pottery. It is now generally acknowledged that separate chronologies should be constructed for the various Weeden Island subregions (Milanich 1976:1-2).
9




Percy and Brose attempted this by constructing a chronology for inland northwest Florida (1974). These researchers seriated pottery types from midden sites in the panhandle and Lower Flint and Chattahoochee river valleys in Georgia and Alabama into a six-period scheme (Percy and Brose 1974:5-6). Pottery type frequencies for their seriation were obtained through consideration of available published reports and large amounts of unpublished data pertaining to their geographical area of interest. Percy and Brose's chronology represents a refinement rather than a change of Willey's original two-subperiod sequence and may not be applicable outside their study area.
The most recent attempt to obtain temporal control from ceramic data in a restricted area is a dissertation entitled "The Social and Chronological Dimensions of Village Occupation at a North Florida Weeden Island Period Site," by T.A. Kohler (1978a). Kohler's data came from sampling excavations at the McKeithen site, outside the geographic area originally studied by Willey. To obtain temporal control for the site, Kohler seriated excavated collections from three intensively sampled village areas separately by means of a principal components factor analysis of a matrix composed of relative frequencies of selected lithic and ceramic categories (see Kohler 1978a:151-166). The three resultant seriations were tied together using ten radiocarbon dates from the village areas, plus trends of change in ceramic modes obtained through correlation of attribute states with stratigraphic placement. These data were used to divide the main occupation of the site, dating from about A.D. 150 to A.D. 750, into early, middle, and late phases (Kohler 1978a:151-177, 224-229).
The seriations revealed that certain ceramic types were found to be good temporal indicators, corroborating the findings of Willey and of Percy and Brose. Kohler's study differs, however, because decorative variants of certain established pottery types were considered separately to ascertain their potential as temporal indicators. For example, Willey's type Carrabelle Punctated (Willey 1949:425), which subsumes a variety of punctation shapes, was divided into six shape categories for analysis (Kohler 1978a:101). Kohler found, for example, that high relative frequencies of fingernail, linear, and hollow reed punctation categories of Carrabelle Punctated were indicative of late proveniences at the site, while triangular and small round punctations occurred most often in early proveniences (1978a:158, 181-182). These findings suggest that formal recognition of these variants by means of a type-variety system of
10




classification (e.g. Smith, Willey, and Gifford 1960), using Willey's types as an initial classificatory basis would be beneficial. Additionally, paste and stylistic attributes were examined for temporal significance. Attribute states or modes that seem to be useful time markers include a high incidence of extensive dark coring in sherds and a narrow range of groove width for stamped sherds, which characterize late proveniences (Kohler 1978a:176). Kohler's procedures also permitted computation and evaluation of the significance of the observed relationships.
Typological Studies
The Weeden Island pottery typology formulated by Willey for the Gulf coastal-northwest Florida area has traditionally functioned as the standard reference for ordering ceramic data for purposes of chronology (as discussed above), and for intra- and inter-site or regional comparisons. The Weeden Island pottery typology lists and describes 44 pottery types (Willey 1949:407-448). Type descriptions consist of details of a number of physical properties (such as paste color, texture, and hardness) subsumed under the label 'ware characteristics', surface decoration, rim, lip, and vessel form, and other attributes. The descriptions were formulated on the basis of sherd collections from midden excavations (see Willey and Woodbury 1942 and Willey 1945) and on consideration of pottery collections made by C.B. Moore (1901, 1902, 1903, 1907, and 1918) and J.W. Fewkes (1924). Willey's typology remains the standard reference work for initial classificatory purposes.
The results of an attribute analysis were used by Kohler to formulate type descriptions for the North Florida subregion (1978a:96-112). This analysis was carried out because in preliminary classification of McKeithen materials, Kohler observed that a significant quantity of sherds did not closely fit Willey's type categories in terms of ware characteristics or decoration (Kohler n.d.a). Kohler' s sample consisted of a judgmental sample of 366 decorated and plain rim sherds from the midden. Attributes selected for description of the McKeithen site material (characteristics of decoration, vessel form, and paste) are comparable to those used by Willey. This catalogue of formal, decorative, and technological attribute states (Kohler n.d.b) was designed to provide systematized pottery type descriptions amenable to statistical and computerized manipulation; it was also designed to be appropriate for analysis of ceramics
11




from Deptford through Weeden Island period sites in northwest and peninsular Florida in terms of the scope of variability of attribute states listed. Willey's decorative type names were used in initial classification of McKeithen ceramics, prior to formulation of the attribute catalogue, and were retained for descriptions of the McKeithen material. The actual descriptions produced by using this catalogue differ from Willey's in that they reflect the greater variability in decoration and ware characteristics observed at this site.
Studies of Regional or Inter-site Ceramic Spatial Variability
K.T. Steinen examined the spatial distribution of Weeden Island period ceramics within the Weeden Island region in a dissertation entitled "The Weeden Island Ceramic Complex: an analysis of distribution" (Steinen 1976). Steinen's principal goal was to identify subregions on the basis of differential distributions of particular ceramic series. The results were to be considered a contribution to the Weeden Island culture history data base and to provide analytical areas for future processually-oriented investigations. The analysis focused upon the component series of the Weeden Island ceramic complex: Weeden Island series (Willey 1949:409-429; Sears 1956:19), complicated stamped series (Willey 1949; Sears 1951), and Wakulla Check Stamped (Willey 1949:437-438). Only decorated members of the Weeden Island series were considered due to inconsistent classification of Weeden Island Plain (Willey 1949:409-411) sherds (by definition, applicable only to rim sherds) by subsequent researchers, and to unsystematic collection of surface materials (usually biased in favor of decorated sherds). The ceramic series data for this investigation were obtained from existing published reports pertaining to Weeden Island midden sites.
In the study, the Weeden Island region was
initially divided into eight subregions or subareas, despite the. fact that the stated goal was to discover such areas through the analytical procedures. The subregions were: Western Panhandle, Central Panhandle, Eastern Panhandle, Peninsular Coast, North-Central Florida, Three Rivers Areas, Chattahoochee River, and Alabama River (Steinen 1976:35-36). The distributional analysis proceeded by tabulating sherd frequencies for each of the three series at each site within each of the eight subareas. Raw frequencies were then standardized by calculating the percentage of each
12




series at each site within each subarea. The mean or average percentage for each series within each subarea was calculated by summing the site percentages and dividing by the number of sites. Each mean percentage (series mean percentage) was standardized a second time by computing a z-score (standard normal mean); z-scores were used for between-subarea comparisons (Steinen 1976:37-51). From comparison of z-scores, Steinen concluded that the distribution of the three ceramic series is extremely variable in seven of the eight subareas. North-Central Florida is the exception where check and complicated stamped sherds are poorly represented relative to the Weeden Island series, which would also be considered poorly represented relative to undecorated sherds, as only 0-10% of the ceramic materials in this area are decorated (Steinen 1976:52).
The a priori imposition of arbitrary boundaries upon the uncollected and unanalyzed data precluded achievement of the initially stated goal of analysis, which was to discover such boundaries, and seems to have contributed to inconclusive results. This is evident when the ranges of the series percentages (i.e. the difference between the largest and smallest percentage for each series at sites within each subarea) are considered. Ranges of greater than 90% for the three series within the subareas were the rule rather than the exception. The decision to divide the study area arbitrarily and the use of a value derived from mean percentages for comparison obscured series variability within the subareas. A more judicious choice of method was suggested by Steinen in his concluding chapter: a computerized mapping procedure such as SYMAP (Dougenik and Sheehan 1975) to plot percentages of the three series over the region as a whole (Steinen 1976:152). By using this method, the identification of trends or subareas could logically have followed analysis of the distribution. This would also be useful to map changes in the distributions through time, when such information is documented in the reports.
Despite the procedural problems in this analysis, Steinen was able to suggest certain trends for the distribution of the Weeden Island ceramic complex by adding information on the location of particular mound construction/use types (from Sears 1958:274-284) to his interpretations. Thus, when these two sources are considered, there seems to be a basic division of the Weeden Island region into northwest and peninsular Florida (Steinen 1976:87). Northwest Florida or northern mounds are generally 'patterned pottery deposit type' mounds (Sears 1958:276-277) and middens in this area generally contain 30-60% Weeden Island
13




series ceramics (Steinen 1976:162); peninsular Florida or southern mounds are generally 'continuous use type' mounds (Sears 1958:277, 279) and middens in this area generally contain 0-10% Weeden Island series ceramics (Steinen 1976:162); check stamped and complicated stamped ceramics are better represented in northern middens than southern middens; complicated stamped vessels and sherds and certain member types of the Weeden Island series are present in both northern mounds and middens (Stelnen 1976:88); and, St. Johns series ceramics (not strictly a member of the Weeden Island ceramic complex) are well represented in Weeden Island period mound and midden contexts in peninsular Florida (Steinen 1976:88).
The problem of subregional variation has been
further addressed by Milanich (Milanich and Fairbanks 1980:96), who defined seven Weeden Island and Weeden Island-related cultures: (1) the Cades Pond culture in North-Central Florida; (2) the Central Peninsular Gulf Coast culture now called the Manasota culture (Luer and Almy 1982); (3) the Northern Peninsular Gulf Coast culture; (4) the panhandle coastal Weeden Island culture; (5) the Apalachicola-Flint-Chattahoochee rivers tri-state area Weeden Island culture, also referred to as Wakulla Weeden Island; (6) the North Florida McKeithen Weeden Island culture; and, (7) the coastal plain, inland culture(s) of north Florida, Georgia, and Alabama. Definition of these cultures was based upon a combination of ceramic data (including consideration of undecorated sherds), ceremonial data, and available environmental, subsistence, and settlement data. Milanich considers cultures (4)-(7) to be 'heartland' Weeden Island cultures, while the first three are Weeden Island-related or Weeden Island period cultures (Milanich and Fairbanks 1980:96). This dichotomy maintains the panhandle or northern versus peninsula or southern division discussed by Steinen (1976).
Studies of Intra-site Ceramic Spatial Variability
Kolomoki, Georgia
William Sears has examined intra-site ceramic variability at Weeden Island sites consisting of mound(s) and associated village middens. The ceramic variability is manifest in qualitative differences which reflect functional differences for pottery recovered from ceremonial and associated midden contexts. Qualitative differences consist of one or all of the following: decoration, vessel form, and
14




quality or 'degree of craftsmanship'. These differences, recognized in Hopewellian through Mississippian period sites in the eastern United States, have been formalized by Sears as a 'sacredsecular' dichotomy in prehistoric ceramics (Sears 1973).
Sears observed that 'sacred' or mound contexts contain only or mostly decorated sherds or vessels, while 'secular' or village middens contain only or mostly undecorated ware. The Weeden Island site (Fewkes 1924; Willey 1949) and the Terra Ceia site (Bullen 1951) have been cited as examples of this pattern (Sears 1973:31-32). In addition, Sears felt that vessel forms more elaborate than simple bowls and jars "...do not belong and do not occur in midden assemblages" (Sears 1973:34). The elaborate forms to which he refers include the 'Kolomoki-style' effigy vessels (Sears 1956:22-24). He suggests that,
The forms excluded [from 'secular'
assemblages] are purely ceremonial ware,
designed and used for ceremonial function as indicated by the fact that they appear only in mounds and not in midden deposits (Sears
1973:34).
Finally, he concluded that a higher 'degree of craftsmanship' is generally considered a characteristic of 'sacred' material.
This ceramic dichotomy was extended to serve as a model for production of Weeden Island pottery and for inferring the level of sociopolitical-economic complexity achieved by Weeden Island society at the Kolomoki site in southwest Georgia. This model suggests that 'sacred' ceramics or 'mortuary ware' were products of specialized manufacture by a class of fulltime artisans (Sears 1961) who resided, presumably, at Kolomoki. This interpretation, of course, implies the existence of a complex stratified society capable of supporting such an economic class. It is now generally agreed, however, on the basis of mortuary data, that Kolomoki probably represents a chiefdom organization (Milanich 1980:12; Milanich and Fairbanks 1980: 131-132). Although the question of full-time specialization is, therefore, probably no longer at issue, the contextual/ qualitative differences observed in Kolomoki pottery used to infer the existence of specialization have yet to be adequately tested or demonstrated; only subjective criteria such as a 'feel' for craftsmanship and artistic evaluation by contemporary Western potters were used to assess the quality differences in Kolomoki pottery.
15




In a review of Sears's (1961) article, "The study of social and religious systems in North America", Fairbanks (1961:238) illustrates the erroneous conclusions that can result from the use of subjective artistic evaluation in determining the presence, kind, and degree of craft specialization. To put specialization in production (a generally accepted concomitant of some degree of social complexity) on a firmer basis, Fairbanks suggests extensive and intensive investigation of village areas to uncover possible workshop areas. If such evidence is not recovered, or if it does not exist, the ceramics themselves must, as Sears suggests, be examined for evidence of specialization in manufacture.
Contemporary standards and a 'feel' for
craftsmanship do not, however, provide an adequate basis for identifying archaeological products of specialization or for validating the inference of fulltime specialization. The fact that definitions of ceramic specialization have not been made operational for archaeological study (Rice 1981) may be one possible explanation for the lack of objective study of Weeden Island pottery production. Rice suggests that, in the absence of clear workshop areas at a site, objective measurement and description of variability in physical properties of pottery (such as color, texture, composi ion), as well as of the more traditionally observed formal and decorative characteristics, will be required to demonstrate the existence of ceramic specialization (cf. Feinman and Upham 1981, for discussion of a 'production step' measure, an ordinal index of labor input in ceramic manufacture). Until such analyses are carried out for Weeden Island pottery, it will be impossible to measure adequately the success or failure of Sears' 'sacred-secular' dichotomy as a model for production.
McKeithen, Florida
Kohler's investigation involved examination of the intra-site ceramic variability at the McKeithen site, and was undertaken to test the hypothesis that the McKeithen site represented a ranked or chiefdom level society (Kohler 1978a). The fundamental test implications of this hypothesis were that pottery indicative of ranking or status would be identified and that the differential distribution of this material could be examined to locate areas at the site which represent high-status occupation. These test implications or assumptions are analogous to, and draw from, those examined by Otto (1975) using data from an antebellum plantation context on the Georgia Coast.
16




Kohler's study was carried out in essentially four steps: identification of categories of ceramics which serve as indicators of ranking or status; calculation of diversity indices for pottery types present during each phase of occupation; SYMAP analysis of the spatial distribution of the ceramic categories and diversity indices; and computer simulations of McKeithen site population sizes.
To identify ceramic categories from which status
or ranking could be inferred, Kohler first examined the southeastern ethnohistoric literature for evidence concerning material indicators of status. For the contact period Southeastern Indians, it was found that such items were made of nonlocal or scarce materials, or evidenced great skill and/or time and effort in manufacture (i.e., products of some degree of specialization). Kohler cites the sable robes reported by the De Soto chroniclers (Elvas and Biedma in Swanton 1946:440), feather mantles of duck down (Le Page du Pratz in Swanton 1911:63), and elaborate pearl, copper, gold, silver, and shell ornaments (Swanton 1946) as examples of material status indicators. Similar evidence in the ethnohistoric literature for pottery is limited to a few references as to the great skill of, for example, Natchez and Tunica potters (Le Page du Pratz and Dumont in Swanton 1946:549-550). Possible evidence of specialized pottery manufacture in the archaeological literature for the Southeast comes from Moundville, Alabama, where an hypothesized workshop area, consisting of large fired hearths, caches of shell, clay, and fuller's earth, was uncovered (Peebles and Kus 1977:442; also see van der Leeuw 1981 and Hardin 1981).
By extending the generalizations about the nature of material status indicators recorded for contact period southeastern societies to pottery in particular, Kohler concluded that ceramic indicators of status would be made of nonlocal resources and/or will evidence skill and/or a great amount of time and effort in manufacture (Kohler 1978a:30-32). Data resulting from Kohler's attribute analysis were used to identify such ceramics.
Nonlocal or 'tradewares' were identified on the basis of presumed nonlocal paste constituents (mica, sponge spicules, and limestone inclusions) and on decorative styles (certain complicated stamped motifs--see Kohler 1978a:112-113, 187). Attributes that were identified as indicative of some degree of specialization, in the sense of skill and/or time and effort in manufacture include 'exotic' or non-bowl
17




vessel forms, exterior and interior surface burnishing, lip additions, extra fine paste textures, and slipped or painted surfaces. Many of these attribute states constitute 'production steps' (see Feinman and Upham 1981). Pottery exhibiting a higher frequency or incidence of these attributes relative to the other ceramics in his sample were designated as 'elite' ceramics (four Weeden Island series pottery types; Kohler 1978a:114, 187-188); this category actually constitutes a subset of 'tradewares', as many of the examples contain mica, but was treated separately in the analysis. In contrast to high-status-indicating tradewares and elite categories, Kohler designated a 'utilitarian' category to subsume a variety of decorated and plain wares (Weeden Island series members and others).
Diversity indices were calculated, because Kohler hypothesized that in addition to exhibiting higher percentages of elite ceramics and tradewares, and lower relative percentages of utilitarian ceramics, highstatus areas of the McKeithen site would exhibit a more diverse assemblage of ceramic types (1978a:31-34). Diversity indices for types present were measured using the Shannon-Weaver diversity index or H (Odum 1971:144), an index used in ecological and more recently, in archaeological, studies (e.g. Wing 1963; Cumbaa 1972; Kohler 1975; Rice 1981). This index measures number of entities or 'richness' and relative diversity of entities or 'evenness'.
SYMAP (Doughenik and Sheenan 1975), a computerized mapping technique, was used to analyze the spatial distribution of densities of ceramic categories and type diversities in order to locate high-status residence areas at the site. The distribution and densities of each category (elite and tradewares, plus ceramic type diversity) were plotted for each of the three phases of site occupation. This was done so that changes through time in distribution and densities could be documented and examined. The test of the hypothesis was carried out by correlating the distributions and densities of elite ceramics with those of tradewares and ceramic type diversity within each temporal phase. Tests of significance were carried out to determine the strengths of the relationship between correlated categories. Three areas in the southern, and eastern village midden were distinguished from the rest of the site using these procedures, and positive correlations between elite and tradeware categories and ceramic type diversity were found to be strongest during the final phase of site occupation (Kohler 1978a:188-207).
18




Kohler's investigation of the level of McKeithen social organization also involved the use of ceramic data in computer simulations of site population size, as cross-cultural studies have suggested that a total of about 500 persons is requisite for chiefdom level organization (e.g. Sanders and Price 1968; Birdsell 1968). A population size model was formulated using estimated weights of sherds in the midden, estimated weights of original vessels represented by the sherds, and ethnographically-derived estimates of ceramic breakage rates and size of ceramic inventory per household unit (as obtained by Foster 1960; David 1972; David and Hennig 1972; and DeBoer 1974) (see Kohler 1978a:208-223; 1978b:1-18). Results of simulations suggested that if the site had been continuously occupied, then the ceramics in the midden were probably deposited by no more than 300-400 persons, and that a resident population maximum of only 100 is not unlikely. This estimate and the distribution data were used finally to hypothesize that the site was initially occupied by a tribal group with the emergence of a 'big man' or proto-chiefdom level of organization by the end of the occupation (Kohler 1978a:223; 1978b:12-13).
Another study of Weeden Island ceramic spatial variability that can be subsumed by studies of both intra- and inter-site variability was carried out by P.M. Rice (1980). To examine variability along these dimensions, a non-probabilistic sample of 49 Weeden Island sherds was selected for neutron activation analysis (NAA), a type of physiochemical analysis that is used to characterize materials by their particular trace chemical or elemental constituents (Rice 1980: 28). Thirty-four sherds were selected from the McKeithen site to examine Sears' 'sacred-secular' contextual distinction (intra-site variability); 30 were from 'secular' or midden contexts, and four were from a 'sacred' or mound context. Fifteen other sherds (all but two from mound contexts) were selected from other Weeden Island period sites in North-Central Florida, the North Peninsular Gulf Coast, and tri-state Weeden Island areas to ascertain local versus nonlocal production for particular pottery types (inter-site variability). Eight clay samples collected from the vicinity of the McKeithen site and four from NorthCentral Florida were included in the analysis to provide a comparative body of data for hypothesizing origin of production for the sherd samples.
Trace elemental data were also applied to an
-examination of the typolological variability of the sample in order to see how the Weeden Island series and Kohler's 'elite', other 'tradewares', and 'utilitarian' categories fared in terms of their trace element
19




'fingerprints'. Relative to these typological categories, 30 sherds are Weeden Island series members, 16 of which represent elite types; six sherds represent other tradewares, and 13 sherds represent other miscellaneous stamped, incised, and plain utilitarian wares.
To achieve the above goals, Rice used a cluster analysis (a multivariate procedure for creating classification systems [Blashfield and Aldenderfer 1976]) to group the sample of sherds and clays into paste types on the basis of trace element data obtained through NAA. Five paste clusters and an 'outlier' group were interpreted from the cluster analysis (Rice 1980:32--Table 2). These data revealed that for this sample, Sears' sacred-secular contextual distinction holds up, as mound and midden sherds clustered, with only a few exceptions, into mutually exclusive paste clusters (Rice 1980:29, 33--Table 5). With respect to the typological categories, these data revealed that most sherds in Kohler's elite category differ in paste types from those representing other Weeden Island series and utilitarian types, but that some other Weeden Island series types, tradewares, and utilitarian types, do not belong to mutually exclusive paste clusters (Rice 1980:29, 33--Table 5).
To ascertain local versus nonlocal production for the sample, Rice compared the cluster data against the one-center model of production inherent in Sears' hypothesis that sacred ceramics were products of specialized manufacture at Kolomoki, the proposed organizational center for Weeden Island culture (Sears 1962), and also to one alternative (Rice 1980:33--Table 6). Results of these comparisons favored Rice's alternative model suggesting multiple centers for manufacture of Weeden Island series sacred types and local (to McKeithen) manufacture for some other sacred types and Weeden Island series 'utilitarian' types. Local status for one of the two paste clusters (containing primarily McKeithen village material) is supported by the fact the one contains a subcluster of three McKeithen-vicinity clays (Rice 1980:30).
Summary
This review has classified previous Weeden Island ceramic studies into four main categories. These include chronological studies (Willey 1949; Percy and Brose 1974; Kohler 1978a), typological studies (Willey 1949; Kohler 1978a); and studies of regional and intrasite ceramic variability (Steinen 1976; Sears 1973;
20




Kohler l978a; Rice 1980). A few of these have focused directly or indirectly upon questions regarding the nature of Weeden Island pottery production and have provided the impetus for the present study.
The sacred-secular dichotomy offered by Sears
(1973) to explain the presence of different kinds of pottery in different depositional contexts, has served as a model for Weeden Island pottery production which suggests that sacred ceramics were products of specialized manufacture. This interpretation of sacred pottery also posited a single-center locus, namely Kolomoki, for manufacture of sacred pottery. Until very recently, attempts to objectively test these interpretations were lacking.
A more recent study (Kohler 1978a) has identified a number of measurable ceramic attributes indicative of some degree of specialization in terms of skill and/or time spent in manufacture; the class of pottery exhibiting such attributes was designated as elite. Another investigation (Rice 1980) has shed light upon the locus of production question, suggesting multiple centers for manufacture of Weeden Island pottery. Both Kohler's and Rice's studies were based on rather small sample sizes, however (n=39 and n=49, respectively), and neither were controlled for time.
The present study draws from the work of Sears, Kohler, and Rice in particular, and examines these production questions in greater depth.
21







CHAPTER THREE
METHODS
This chapter consists of four sections. The first describes the methodological and theoretical approach employed in the present study. The second describes the pottery samples examined and how they were selected. The third section defines the variables selected for measurement or observation and describes the methods of analysis that were used. This section also discusses how the resultant data were manipulated for interpretation with respect to the objectives of this investigation. Finally, the last section describes the clay samples and methods used in their analysis.
Ceramic Technology and Ceramic Ecology
The present study employed the methods of ceramic technology to investigate Weeden Island pottery production. As stated in Chapter 1, ceramic technology is a body of'methods used to describe systematically the physical, mineralogical, and chemical properties of ceramic materials. Manufacturing technology (methods of vessel construction, finishing, and firing) is also the subject of ceramic technological investigations. The methods subsumed by ceramic technology include: the use of a hand lens or binocular microscope for recording observations of such phenomena as the presence of a slip, type of surface finishing, paste texture or gross mineralogical identification; the use of standard scales or methods of measurement to determine, for instance, paste color (Munsell Soil Color Charts), scratch hardness (Mohs' Mineral Hardness Scale) or size of aplastic paste constituents (Wentworth's Size Classification); and the use of more sophisticated instruments or methods such as petrographic microscopes and neutron activation analysis for obtaining precise mineralogical and trace chemical (respectively) characterizations of ceramic materials. Knowledge of the physical, mineralogical, and chemical properties of ceramic materials (clays and tempers) and of 'primitive' manufacturing technologies reported in the ethnographic literature are
23




instrumental for interpretation of the results of ceramic technological investigations.
The advantages of ceramic technology are that most of the methods: are inexpensive to implement; are precise, objective, and replicable; can be performed on small sherds rather than whole vessels; can be applied to investigation of raw ceramic materials (clays and tempers) for comparative purposes; and allow pottery to be studied as products of human behavior rather than as lists of traits or attributes.
In a study of the cultural behavior associated with pottery making, these methods are subsumed by a 'ceramic ecological' approach (Matson 1965). As defined by Matson, ceramic ecology is an aspect of Steward's cultural ecology (Steward 1955) "...which attempts to relate the raw materials and technologies that the local potter has available to the functions in his culture of the products he fashions" (Matson 1965: 203). With this approach, pottery making is viewed as a part of a culture's technological subsystem and is considered to be one point of articulation between the environment and a particular cultural system (Rice 1976).
A ceramic ecological approach-permits comparable investigation and description of the nature and variability of archaeological ceramics and available ceramic resources, or the 'ceramic environment' (geological and climatological factors). Such an approach, which integrates an assemblage of artifacts, a site, or a culture with an environmental context, is necessary for the study of production of any class of artifact. With respect to investigations of Weeden Island pottery production, direct recognition of the environmental factors involved has, with one exception (Rice 1980), been ignored. With these considerations in mind, the ceramic ecological approach and methods of ceramic technology were, therefore, not only appropriate for the present study but also essential.
Description of Samples, and Sampling Procedures
Thirty-three specimens from ceremonial contexts and 212 from midden contexts, for a total of 245 specimens, were selected for analysis. The pottery sample from ceremonial or sacred contexts was chosen from Mounds B and C at the McKeithen site. This sample consists of 19 vessels from Mound C and 13 vessels and an'effigy bird head from Mound B. The typological breakdown for these specimens is presented in Table
24




3-1, which is also an example of a catalogue form prepared for classification of McKeithen mound ceramics.
Eighteen of the 19 vessels from Mound C represent an east-side cache of vessels thought to have been deposited beside and on top of the edge of the primary mound (see Fig. 4). Vessels 1-14, 18, and 19 were recovered by L.A. McKeithen, Jr., from the mound's east side. Sherds fitting all of these vessels were obtained from spoil during excavations directed by Milanich (Milanich, et al. 1978). Vessels 15 and 16 were recovered by Milanich from undisturbed contexts (see Fig. 4). Vessel 15 probably represents the northern extent of the east-side cache. Vessel 16 was located on the Mound's west side under the edge of or right beside the primary mound and was not associated with the pottery cache. Vessel 17 was obtained from a local collector; its base was recovered from an undisturbed context during excavation. According to Kohler's categories, nine of these vessels are elite, eight are tradewares, and two are utilitarian. Pieces from four of these vessels were also used in Rice's trace element analysis (vessels 2, 5, 12, and 14 here=cases 131, 132, 129, and 130, respectively, in Rice 1980:32--Table 4).
The sample of Mound B pottery consists of portions of 13 vessels recovered from the activity floor inside and just outside the structure (see Fig. 5) and represents vessels used and broken during the occupation or use of the residence. The effigy bird head (broken off a pot not recovered during excavation) was recovered just outside of the former wall of the structure, close to and facing the individual buried in the structure floor. According to Kohler's categories, ten are elite ceramics, three are tradewares, and one is utilitarian. Materials from Mound A were not used in this study as they had not been processed (washed, catalogued, etc.) when this project was begun.
The midden ceramics considered in this
investigation came from the three McKeithen midden areas (northern, eastern, and southern--Fig. 6) identified and selected by Kohler for intensive transect- and cluster-sampling excavations (1978a). In order for differences between midden and mound ceramics to be explained in terms other than temporal change, the population of midden sherds selected for sampling came from middle phase proveniences encompassed by the temporal spans represented by the radiocarbon dates from Mounds B and C. The averages of accepted dates obtained for Mounds B and C are A.D. 418 + 53 (UM-1234, 1235, reported in Milanich et al. 1978) and A.D. 487 +
25




Table 3-1 Typological Breakdown for McKeithen Mound
Pottery
8Col7 F.S. Zone Total # ceramics 33
Context Mounds B and C Level
Square #
sponge
Type sand grit mica spicules other
plain 3
smooth plain burnished plain
single incised line
St. Johns Plain St. Johns Check St.
Papys Bayou Incised Papys Bayou Punctated 2
Papys Bayou Plain 3
Weeden Is. Plain 2
Weeden Is. Red 4 1 2
Weeden Is. Zoned Red 8 1
Weeden Is. Incised 3
Weeden Is. Punctated residual red
Carrabelle Punctated:
round to oblong
triangular fingernail
hollow-reed
linear
rect. to square
other
Carrabelle Incised Keith Incised Indian Pass Incised 1
Tucker Ridge-Pinched 1
St. Pete Incised
unid. curv. comp..st. unid. rect. comp. st. Swift Creek Comp. St. Crooked River Comp. St. New River Comp. St. Old Bay Comp. St.
Sun City Comp. St. Kolomoki Comp. St. St. Andrews Comp. St. Napier Comp. St.
check st. 4-6/inch I
check st. 7-9/inch check st. 10-12/inch
linear check st. simple stamped cross-simple st. Thomas Simple St.
eroded stamped unid. incised cord marked
total 7 1 19 6 33
26




OP
de
/ 830 sameuv
/ moundca
'B 24p 'w*ca
I.. primary mound.
mound cop--.g :
*o:-primory mound
vv8B20 -\81:I
' V16 v 15 8
BIB a Ito 2
.0
OB1 82 9
BIG ..e .. /~B4e
,t 8817a-" o- -,-w
GRID45VSE
0 3 6 1 KL
METERS 01I BURIAL AND NUMBERS
Figure 4 Mound C, McKeithen Site




11,1 - test trench -excavationT I0
I%
* burial % *
_ 0
TV%%.> 0**
_ _
_____ ___n atmp ~
MET~f%
Figure 5 on%,M~ihnSt




4McKEITHEN SITE
s- M 8-CO-17
0I IS 50 75 0
\ //
Mound C
"\O/
*0 Mound A ~v~
Figure 6 McKeithen Site, Columbia County
SOURCE: Milanich and Fairbanks 1980:133
29




49 (UM-1434, 1436), respectively (averages were obtained according to instructions by Long and Rippeteau 1974). Midden proveniences were selected by superimposing these dates (with one standard deviation) onto Kohler's seriation time-line (Kohler 1978a:176, Fig. 18 presented here as Fig. 7). This procedure bracketed groups of factor scores from each midden area which represent proveniences seriated by the principal components operation (discussed in Chapter 2). A total of 53 factored proveniences exhibiting high 'ordering efficiency' (see Kohler 1978a:144, 146) were delimited in this manner. Nine of these come from the northern midden, five from the eastern, and 39 from the southern middens. These proveniences (and corresponding Florida State Museum catalogue numbers are listed in Appendix A. A total of 1122 sherds from these proveniences (295 from the northern, 139 from the eastern, and 688 from the southern middens) formed the population of midden ceramics considered for sampling in this study. In sample selection, Kohler's elite, tradeware, and utilitarian ceramic categories operated to a certain degree.
The elite category, according to Kohler, is
composed of four Weeden Island Series pottery types: Weeden Island Incised, Weeden Island Punctated, Weeden Island Zoned Red (Willey 1949), and Weeden Island Red (Sears 1956). As discussed in Chapter 2, these types were found by Kohler to exhibit high frequencies of 'exotic' or non-bowl vessel forms, exterior and interior surface burnishing, lip additions, fine paste textures, and slipped or painted surfaces (Kohler 1978a:113-114). This category actually constitutes a special subset of Kohler's tradewares category, as many of these sherds (n=30 or 68%) also exhibit micaceous inclusions. The two categories were, however, considered separately here for purposes of sampling and analysis.
The tradeware category, as defined by Kohler, consists primarily of pottery series and types characterized by paste constituents that are considered 'nonlocal' to the North Florida area. The three classes of nonlocal paste include micaceous pastes (thought to originate in northwest Florida to southwest Georgia), sponge spiculite pastes (thought traditionally to come from eastern Florida), and crushed limestone pastes (thought to come from North and Central Peninsular Gulf Coast subregions). In terms of pottery series and/or types, the tradeware category (excluding elite types) consists of spiculite paste: St. Johns series (Goggin 1948) and Papys Bayou series (Willey 1949); crushed limestone paste: Pasco series (Goggin 1948); and micaceous paste: Crooked
30




NORTHERN EASTERN SOUTHERN
MIDDEN MIDDEN MIDDEN
A.D.
800 4.5,
-2.75
o 0 -0.13I I
I;F
600 -2
-0.21O
4 3.50
w2
S 0.24- 0.21
3 ,3 400f
-0.13-0.34E3
N -0.58N 4
200
-0.82
F
Us
0.60
1.26
0
1 -1.03
B.C.
Figure 7 McKeithen Site Principal Components Seriation
SOURCE: Kohler 1978a:176
31




River Complicated Stamped (Willey 1949; Kohler 1978a).
Other pottery types were designated tradewares by Kohler on the basis of decorative style, the known regional distributions of which suggest nonlocal (to North Florida) origins. These include: Kolomoki Complicated Stamped--southwest Georgia (Sears 1956; Steinen 1976); Napier Complicated Stamped--piedmont Georgia (Sears 1956; Wauchope 1966). Kohler also suggested that perhaps Old Bay Complicated Stamped, St. Andrews Complicated Stamped, Tucker Ridge-Pinched, and Indian Pass Incised (Willey 1949) types may be tradewares, but they occur in such low frequencies wherever they are found that it is difficult to cite possible nonlocal areas of origin (Kohler 1978a:113).
Prior to sample selection, a binocular microscope (70x magnification) was used to sort the population of sherds into paste groups on the basis of the presence/ absence of micaceous, sponge spicule, and limestone inclusions in order to distinguish hypothesized tradewares from hypothesized locally made utilitarian pottery. After this sorting, sherds were classified according to the established type categories (e.g. Willey 1949; Sears 1956; Kohler 1978a).
The crushed limestone-inclusion tradeware category was eliminated for purposes of sample selection during the classification procedure as the whitish inclusions present in some sherds (particularily those containing sponge spicules) proved not to be limestone or other carbonate substance when tested with cold and warm hydrochloric acid (see Chapter 5, discussion of spiculite paste types). Kohler's surface decorationbased tradeware category was also eliminated as a basis for sampling due to small size in the population (N=14) and because decoration- and paste-based tradeware categories were not mutually exclusive; eight of the 14 decorated sherds could be classed as tradewares on the basis of paste and were thus eligible for sampling regardless of type of decoration. The other six sherds were placed into the utilitarian category for sampling. For purposes of this investigation, Crooked River Complicated Stamped sherds were classed as tradewares only when they contained mica or sponge spicules. After elimination of the limestone inclusion and decoration-based tradeware categories, only micaceous and sponge spicule paste categories were employed in selection of the tradeware subsample; any non-elite sherd which contained nonlocal paste constituents was classified as a tradeware. It should be noted that this is a reinterpretation of the way in which Kohler used this category.
32




The utilitarian category in the present study
consists of a variety of decorated and plain pottery not containing mica or sponge spicules. This category consists of ceramics characterized by quartz sand or grit pastes. All types except elite types and examples of types or sherds containing mica or sponge spicules, were considered utilitarian for sampling purposes. Table 3-2 presents the typological breakdown of the population as a whole. According to Kohler's labels, there are 44 elite sherds (4% of total), 362 tradewares (32% of total--based on paste constituents), and 716 (64% of total) utilitarian sherds in the population.
Well-represented subsamples of elite sherds and tradewares relative to utilitarian sherds in midden were desired for comparison with the mound sample, which is characterized by relatively high proportions of the former categories. For this reason, the populations of elite pottery and tradewares were disproportionately sampled for analysis. All elite sherds, totalling only 44, were chosen for analysis regardless of sherd size (some were too small to permit all steps of analysis). Approximately 23% (84 of 362) and 12% (84 of 716) subsamples of tradewares and utilitarian sherds, respectively, were also chosen for analysis.
For selection of tradeware and utilitarian
subsamples, each population was divided arbitrarily into three groups representing the temporal range (early, overlapping, and late) established by the radiocarbon dates from Mounds B and C. Three conditions relating to the objectives and methods of analysis of this study guided the selection of tradeware and utilitarian subsamples within the three temporal segments. First, three midden sherds that were used in Rice's (1980) trace element study were selected to provide comparative data. Second, sherds diagnostic of vessel form (e.g. rims and bases, ideally, of sufficient size to indicate vessel shape and size) superseded selection of 'undiagnostic' body sherds. The third condition was size of sherds: many steps of analysis discussed in the next section required that sherds be large enough so that pieces could be broken off for analysis without destroying the whole sherd. Of the sherds remaining after the first two conditions were satisfied, only sherds greater than or equal to 4cm in one dimension (other than thickness) were considered eligible for sampling. Selection of these sherds was judgmental in most cases, in an attempt to sample the range of variability in color, texture, and decoration. Where necessary, smaller sherds were occasionally considered acceptable to increase sample sizes within particular categories.
33




Table 3-2 Typological Breakdown for McKeithen Midden Pottery Sampling Population
8Col7 F.S. Zone Total # ceramics 1122/136
Context midden Level
Square #
sponge
Type sand grit mica spicules other
plain 240/17 81/8 43/4 37/1
smooth plain 212/5 54/4 96/1 49/1
burnished plain 4 10 10/1
single incised line 7/1 1 6 1
St. Johns Plain 14/1
St. Johns Check St. 8/1
Papys Bayou Incised 1
Papys Bayou Punctated 9
Papys Bayou Plain 7/7
Weeden Is. Plain 20/20 15/15 14/14
Weeden Is. Red I 9/1
Weeden Is. Zoned Red 1
Weeden Is. Incised 3 4 10
Weeden Is. Punctated 4/1 10/6
residual red 2
Carrabelle Punctated:
round to oblong 2 1 2
triangular 2/1 1 1/1
fingernail 2 2
hollow-reed 1
linear
rect. to square 3 1/1
other 2 2 1
Carrabelle Incised 13/4 2/2 4/1 4/1
Keith Incised 3/3 1 4/1 2/1
Indian Pass Incised 4
Tucker Ridge-Pinched I I
St. Pete Incised 1/1
unid. curv. comp. st. 2 2 1 3/1
unid. rect. comp. st. 4/1 2 1 1
Swift Creek Comp. St. 2 I 2/1
Crooked River Comp. St. 2/1 1 3
New River Comp. St.
Old Bay Comp. St. I 1I
Sun City Comp. St.
Kolomoki Comp. St. I
St. Andrews Comp. St. 2 I1
Napier Comp. St. I
check st. 4-6/inch I 1 2 2/1
check st. 7-9/inch I 1/I1
check st. 10-12/inch
linear check st.
simple stamped 4/1 2 2
cross-simple st.
Thomas Simple St.
eroded stamped 3/1 1 1
unid. incised 12/I 2 4 2/1
cord marked 1/1
total 553/56 177/32 230/29 162/19 1122/136
34




Within the early temporal segment (average of
Mound B radiocarbon dates plus and minus one standard deviation, excluding years overlapping with Mound C's averaged dates; i.e., A.D. 365-438), all elite sherds (n=6), 15 tradewares, and 23 utilitarian sherds were chosen for analysis. Within the overlapping temporal segment (A.D. 438-471 from Mounds B and C averaged. dates), all elite sherds (n=ll), 19 tradewares, and 31 utilitarian sherds were selected for analysis. Within the late temporal segment (average of Mound C radiocarbon dates plus and minus one standard deviation, excluding years overlapping with Mound B's averaged dates; i.e., A.D. 471-536), all elite sherds (n=27), 50 tradewares, and 30 utilitarian sherds were chosen for analysis. The breakdown of sampling in terms of numbers of eligible sherds (trace element study, form, and size criteria) within these three temporal segments is presented in Appendix B.
The total number of midden sherds selected for analysis is 212 (44 elite, 84 tradewares, and 84 utilitarian sherds). Each temporal segment and midden area (northern, eastern, southern) is approximately proportionately represented in terms of total sample size. The temporal and areal breakdown of elite, tradeware, and utilitarian subsamples is listed in Appendix C. The typological breakdown for the midden sample is presented in Table 3-3.
Description of Variables and Methods of Data Collection
In the present study, selection of variables and methods of data collection-proceeded according to recommendations by Rice (1981) for the *study of pottery production and specialization. Rice suggests the study and definition of variability in ceramic products in terms of their physical properties (by means of ceramic technology), formal and decorative characteristics, and details of manufacture to investigate the nature of pottery production. Measurement of these variables was also necessary for testing the contextual/qualitative criteria of Sear's sacred-secular dichotomy. Rice's recommendations will be discussed in detail in Chapter 7. Data on certain physical properties of the pottery sample were also collected for comparison with the clay samples to investigate origin of production for the pottery samples.
The first analytical step in this investigation of Weeden Island pottery production involved the selection and measurement of variables that permitted definition
35




Table 3-3 Typological Breakdown for McKeithen Midden Pottery Sample
8Col7 F.S. Zone Total # ceramics 212/52
Context midden Level
Square #
sponge
Type sand grit mica spicules other
plain 19/4 12/3 5/1 7
smooth plain 15 6/1 12/1 8
burnished plain. 4 4
single incised line 1/1
St. Johns Plain I
St. Johns Check St.
Papys Bayou Incised
Papys Bayou Punctated
Papys Bayou Plain 3/3
Weeden Is. Plain 5/5 2/2 8/8
Weeden Is. Red I 9/1I
Weeden Is. Zoned Red 1I
Weeden Is. Incised 5 2 10
Weeden Is. Punctated 4/1 10/6
residual red 2
Carrabelle Punctated:
round to oblong 1
triangular 1 1/1
fingernail 1
hollow-reed
linear
rect. to square I 1/1
other
Carrabelle Incised 4/2 1/1 }/1 2/1
Keith Incised 1/ 1 1 i/1
Indian Pass Incised 3
Tucker Ridge-Pinched
St. Pete Incised
unid. curv. comp. st. 1 2/1
unid. rect. comp. st. I 1 1
Swift Creek Comp. St. 2 2/1
Crooked River Comp. St. I 1 1
New River Comp. St.
Old Bay Comp. St. I 1I
Sun City Comp. St. Kolomoki Comp. St. 1I
St. Andrews Comp. St. 2
Napier Comp. St. I
check st. 4-6/inch 2 1/I1
check st. 7-9/inch I
check st. 10-12/inch
linear check st.
simple stamped
cross-simple st.
Thomas Simple St.
eroded stamped
unid. incised I/1 I 2/I
cord marked 1/1
total 66/15 32/8 75/19 39/10 212/52
36




of paste variability in the pottery sample. The objective of this step was to determine the numbers and kinds of clay resources used in the manufacture of the pottery.
Primary Variables
In order to distinguish groupings of sherds that may have been products of the same or similar clay resources (as opposed to decorative or formal types, or groups of sherds exhibiting similar decorations or vessel forms), the variables or attributes chosen for analysis in this stage were ones that were more ecologically than culturally conditioned. For example, paste composition is conditioned more by availability of raw materials (clays and tempers) as opposed to stylistic attributes such as decoration, which are conditioned more by cultural traditions. In the present study, 'ecologically conditioned' variables are considered 'primary' and include fundamental physical properties of pottery such as aplastic composition and color. Attributes such as decoration and vessel form are considered 'secondary'. Selection of primary variables for the first step in analysis also permitted comparable analysis of the clay samples, which was necessary for legitimate comparison between sherds and clays in the investigation of origin of production for Weeden Island pottery.
The following variables were chosen for
measurement or observation for the first stage of analysis: aplastic paste composition (type, frequency or proportion, and relative size of aplastic inclusions); refired porosity; and refired core colors. Explanations of the specific ceramic technological methods used for data collection in this stage follow definition of each primary variable.
Aplastic composition. Aplastic inclusions in pottery (e.g. sand, crushed rock or shell, fibers) facilitate the working properties of clay. Aplastics function to: 1) reduce the amount of water required for preparing a workable paste, thereby decreasing plasticity or stickiness of a clay; 2) increase porosity of the clay body, which promotes uniform drying and speeds the drying process; 3) counteract drying and firing shrinkage of a clay body; 4) decrease risk of warping and cracking of a clay body during drying and firing (Shepard 1976).
The presence of aplastic material in pottery may result from tempering practices or from natural occurrence in a clay deposit, or it may result from both factors. Aplastics such as crushed shell or sherd
37




are readily identifiable as temper or added material. It is more difficult, if not impossible, to distinguish tempered from naturally present quartz inclusions, as quartz sand is one of the most common aplastics occurring naturally in clays (Shepard 1976: 161, 162). Variability in type, frequency, and size of naturally present aplastics is dependent upon the origin or source of a clay, i.e. whether a clay is residual (one formed in contact with the parent rock) or sedimentary (transported). The type of parent rock largely determines variability in inclusions present in a residual clay, while the source, conditions of deposition, and interaction with environments during transport determine the type, frequency, and size of inclusions present in sedimentary clays (Shepard 1976:11).
Types or kinds of aplastic inclusions present in the sherd sample were identified using a binocular microscope (70x magnification) during the initial classification procedure discussed in Section 3 (p. 32) of this chapter. Five classes of inclusions were recognized: 1) quartz; 2) sponge spicules; 3) mica; 4) ferruginous concretions or lumps; 5) whitish inclusions or lumps (not limestone or dolomite).
Relative frequency or proportion of sponge
spicules present in sherds was determined by examining a freshly broken edge or cross-section under a binocular microscope (70x magnification). Frequencies were subjectively ranked as follows: none, rare, occasional, common, or abundant.
Relative frequency of mica was determined from unslipped sherd surfaces. Surface frequencies were used as it was often difficult to see mica particles, when present, in sherd cross-section, even under magnification; the platy structure of mica was often obscured by the texture of the sherds. Frequencies were subjectively ranked according to the scale listed above for ranking sponge spicule frequency.
Ferruginous and whitish concretions, while not unusal constituents in the sherd sample, were distinctive features of only a few sherds. For this reason, "rrsec" was recorded only for these cases, and "absence" for the others.
Frequency of quartz sand in sherds was determined by making point-counts of inclusions within a standardized area of observation of sherd crosssections. Continuous level point-counts, rather than ordinal rankings, were made in this case to decrease the subjectivity of the data for comparison. Sherds
38




were prepared for measurement by grinding flat a broken edge of each sherd on a variable-speed thin-section polisher-grinder apparatus equipped with water lubrication and 240 Grit Carbimet abrasive paper discs (grinding medium). Point-counts were made using a binocular microscope fitted with a gridded eyepiece. The size of the grid was 2.5 x 2.5 mm, at approximately 42x magnification. The square grid was divided into one hundred .25 x .25 mmu square units. This grid size/magnification combination was chosen because it was large enough to permit counts of inclusions within an area approximately one-third the thickness of the ground cross-secton of most of the sherds in the sample and strong enough to maintain visibility of smaller inclusion sizes.
The area of observation was standardized for
measurement by positioning the ground sherd edges in the approximate center of the field of view so that the grid was positioned over the center point of the crosssections (cross-section width or thickness and length). All aplastic inclusions encompassed by the grid were counted by type and size. Relative size of inclusions was determined with reference to Wentworth's Size Classification (Shepard 1976:118), adapted here as part of Appendix D.
Although all inclusion types viewed within the grid were counted and sized, this procedure was used primarily to obtain accurate frequencies for "fine" and "very fine" quartz sand inclusions. The point-count procedure presented difficulties for obtaining frequencies of larger "grit" quartz incusions and for other inclusion types. The larger quartz inclusion sizes, when present, were not always contained within the gridded area of the field of view, and when observed, the counts, relative to the frequencies of smaller inclusions, did not appear to be representative. It was also apparent that the grinding procedure removed some of the larger quartz inclusions in some cases (in sherds characterized by soft pastes). Frequency of larger quartz inclusion sizes ("medium" through "very coarse") were instead subjectively ranked (using the scale listed on p. 38) by scanning a freshly broken edge under a microscope at 70x magnification.
Mica, when present, was rarely visible in crosssection for reasons already discussed. Sponge spicules, when present, were not clearly visible at the 42x magnification setting. Ferruginous and whitish inclusions, when present, were sometimes obscured by grinding and were not always observed within the field of view as their sizes usually ranged from "medium" to livery coarse". Only point-counts of total number of
39




quartz inclusions (all sizes), total number of "fine" quartz, and total number of "very fine" quartz were used for purposes of data analysis.
Ref ired porosity. Porosity is another fundamental physical property of pottery. It is defined as the ratio of the volume of pore space to the total volume of the piece (Shepard 1976:125). Porosity can be a useful measure of body structure as it affects such properties as density, strength, permeability, and degree of resistance to weathering, abrasion, and thermal shock (Shepard 1976:126). Clay composition, proportion, size, and shape of aplastic inclusions, and firing conditions interact to determine porosity of pottery.
Porosity of the sherd samples was measured as
Percent Apparent Porosity (Shepard 1976:127), or the relative volume of open pores. Measurements were made on pieces of the sherds that had been ref ired under oxidizing conditions for 30 minutes at a temperature of 7000C in a Thermolyne electric furnace. Sherds were refired in an attempt to remove differences in porosity 'caused' by variability in firing conditions (e.g. to eliminate unoxidized or charred organic material which clogs pores spaces, possibly skewing porosity values) and in an attempt to standardize conditions of comparisons between the pottery samples and similarily fired clay briquettes. The 7000C temperature was chosen because this temperature level (and oxidizing condition of firing) was probably higher, or at least more oxidizing, than the original firing conditions and was therefore believed sufficient for this purpose.
Dry weight, saturated or wet weight, and volume of each sherd were required for calculating porosity. Shepard's boiling method (Shepard 1976:127) was used to obtain these measurements. Dry weight was obtained by. individually weighing (in grams, using an electric balance) ref ired sherds which had been dried in a gravity convection oven set at 1100C for 24 hours to eliminate absorbed moisture. Saturated weight was obtained in the following steps: 1) soaking the test pieces in cool water until air bubbles ceased to rise to the surface; 2) then submerging the wet test pieces in a covered two quart sauce pan of boiling water for two hours (sherds were housed in a wire basket for boiling to avoid risk of abrasion in bumping the pot bottom; about 20 sherds at a time were processed); 3) placing the saturated sherds in a container of cool water after boiling until they could be reweighed; 4) individually reweighing the sherds rapidly to avoid evaporation (excess water on each sherd was blotted off with a damp sponge prior to reweighing); 5) placing
40




each reweighed sherd back into the container of water for the next measurement (volume). Volume of sherds was measured on the saturated pieces. Volume was obtained by submerging each sponged off saturated sherd (suspended by a piece of wire of known volume) into a graduated cylinder containing a known amount of water. Volume was determined as the amount of displacement in cubic cm (volume of water after submergence of a sherd minus original volume; volume of the wire suspender was taken into account for the value of original volume). Finally, Percent Apparent Porosity was calculated using the formula:
% Apparent Porosity = wet weight dry weight x 100 volume displacement
Thirty-one sherds in the sample (mostly elite sherds) were excluded from this procedure as they were not large enough to have a piece broken off for refiring.
Ref ired core colors. Color is a fundamental
physical property of pottery determined primarily by the amount of iron compounds and organic materials, the primary colorants of pottery in the clay, and by firing conditions (duration, temperature, atmosphere). Color also depends upon the proportion or frequency, size, and distribution of aplastics in the paste.
Color was measured using Munsell Soil Color Charts (Munsell 1942). Munsell color charts are pages or charts of color chips that denote hue, or position within the color spectrum (red through purple); value, or lightness or darkness of a color (white to black); and chroma, or strength or clearness of a color (amount of grey). Munsell color measurements are alphanumeric notations, e.g. 10YR 5/4 (denoting hue and value/chroma); notations have corresponding color names, e.g. yellowish brown. This system provides an accurate and replicable means of describing color. Color measurements were made on the sherds that had been ref ired to 700*C to eliminate color differences caused by variability in firing conditions, use, and post depositional modification, and to permit legitimate comparison with the fired clay samples.
The actual measurements were made by passing each sherd under hue pages (each color chip is perforated so that objects may be held adjacent to them) until a match was found. Both the alphanumeric notations and corresponding color names were recorded. All measurements were made under consistent laboratory conditions (Ceramic Technology Laboratory at the
41




Florida State Museum, equipped with overhead fluorescent lighting). Thirty-one sherds were excluded from ref ired color measurement as they were too small for refiring.
Ref ired slipped and unslipped surface colors
(exterior and interior) and core color (from a fresh break) were measured, however only core color was used for purposes of data analysis. Surface color was eliminated for the following reasons: 1) the color of refired surfaces is not as reliable for distinguishing paste differences as are ref ired core color measurements, as surfaces are more affected by use and post depositional alteration; 2) there is a great deal of homogeneity in ref ired surface color within the entire sample of sherds regardless of core color, similarity which might tend to mask important differences between sherds; 3) it was suspected in some cases that the 700*C refiring failed to eliminate surface color variation caused by the original firing (e.g. cases in which the cores refired to well oxidized colors, but whose surfaces remained greyish); and 4) so that sherds having slipped surfaces could be included in the cluster analyses (see discussion, pp. 45, 47). For purposes of data analysis, the ref ired sherds were hand-sorted into five broad categories of core color on the basis of obvious visual color differences. A numeric value label was assigned to each category for purposes of recording these data for computer analysis. Each broad category is felt intuitively to have some degree of significance in terms of resources used and encompasses a range of Munsell measurements (see Table 3-4).
Data Analysis of Primary Variables
The objective of the technological examination described in the preceding pages was to provide the data base for determining number and kind of clay resources represented by the sample. A multivariate cluster analysis procedure called CLUSTAN IC (Wishart 1978), part of a family of procedures designed to produce classifications (Blashfield and Aldenderfer 1976), was used for this purpose. This procedure considers all variables simultaneously to produce polythetic or relatively homogeneous groups of entities.
The clustering procedures used in this study proceeded on the basis of matrices of similarity coefficients computed for the sherds in the sample (see also Saffer 1979). A similarity coefficient is a measure of the value or degree of similarity or resemblance of each case or entity (sherds in this
42




Table 3-4 Broad Categories of Refired Color
REFIRED CORE COLOR
CATEGORY 1 n=23 modal Munsell color:
10YR 8/3 very pale brown
CATEGORY 2 n=64 modal Munsell color:
10YR 6.5/4 v. p. brown/lt. yell.-brown
10YR 7/4 very pale brown
CATEGORY 3 n=59 modal Munsell color:
7.5YR 5/6 reddish yellow
7.5yr 5/6-10YR 5/4 strong brown/yell.-brown
CATEGORY 4 n=38 modal Munsell color:
5YR 5/6 yellowish red
CATEGORY 5 n=30 modal Munsell color:
5Y 5/2 olive brown
REFIRED SURFACE COLOR
ext. int.
CATEGORY 1 n= 5 8 modal Munsell color:
10YR 8/3 very pale brown
ext. int.
CATEGORY 2 n= 178 155 modal Munsell color:
10YR 7/5 very pale brown-yellow
7.5YR 6/6 reddish yellow
10YR 6.5/4 v. p. brown/lt. yell.-brown
10YR 7/4 very pale brown 7.5YR 7/6 reddish yellow
7.5YR 6.5/6 reddish yellow
10YR 7/4-7.5YR 6/6 v. p. brown/red.-yellow
ext. int.
CATEGORY 3 n= 19 23 modal Munsell color:
5YR 5/6 yellowish red 5YR 6/6 reddish yellow
ext. int.
CATEGORY 4 n= 7 28 modal Munsell color:
2.5Y 5/2 greyish brown
2.5Y 6/2 light brownish grey
5Y 5/1 grey
43




case) to every rther case in the sample with respect to the variables measured or observed. The size of the similarity matrix produced for a cluster analysis depends on the number of entities to be clustered. Two hundred fourteen of the 245 sherds were used in the cluster analyses; 31 had missing values for some of the recorded measurements. The 214 sherds were divided for clustering into three subsamples on the basis of presence/absence of nonlocal paste constituents (mica and sponge spicules). The sizes of the micaceous, spiculite, and quartz sand-grit (lacking mica and sponge spicules) paste groups were 75, 43, and 96, respectively; the sizes of the similarity matrices were therefore 75 by 75, 43 by 43, and 96 by 96, respectively. The primary variables discussed above provided the basis for comparing sherds for calculation of the similarity matrices, and hence, for group or cluster formation. These include:
1) total number of quartz inclusions (point-counts);
2) total number of "fine" quartz (point-counts);
3) total number of "very fine" quartz (point-counts);
4) refired porosity;
5) relative frequency of "medium" quartz inclusions; 6) relative frequency of "coarse" quartz inclusions;
7) relative frequency of "very coarse" quartz
inclusions;
8) relative frequency of mica at sherd surface;
9) relative frequency of sponge spicules in crosssect ion;
10) refired core color;
11) presence/absence of ferruginous inclusions; and, 12) presence/absence of whitish inclusions.
This list of variables used for cluster formation varies slightly according to paste subsample. Not all of the variables were applicable for calculation of all three similarity matrices. For example, the value coding for frequency of sponge spicules was not applicable for use in cluster formation of mica and sand-grit subsamples, etc. Therefore, the value coding for an inapplicable variable was simply ignored (in the format statement) in calculating similarity matrices.
The similarity coefficient used for calculation of the similarity matrices was conditioned by the level of measurement employed in data collection. The 12 variables listed above represent three levels of measurement: continuous (variables 1-4); ordinal or qualitative multistate (variables 5-10); and binary (variables 11-12). This fact limited not only choice of the similarity coefficient but also selection of CLUSTAN procedures and options (see Chapter 5 pp.
44




81-83) for doing the actual cluster analyses. Gower's coefficient of similarity (Gower 1971; Sneath and Sokal 1973) was employed in the present study as it is the only coefficient available that works with mixed data types or measurement levels. The Gower 's similarity matrices (which formed the basis for the cluster analyses) were calculated using a separate computer program written by Roger K. Blashfield of the University of Florida Department of Psychiatry, Gainesville (see Saffer 1979; Rice and Saffer 1982; Saffer, Blashfield, and Rice n.d.).
Gower's Coefficient was calculated by comparing
two sherds at a time, counting the number of variables having the same value, and then dividing by the total number of variables. A coefficient of "l" means that the two entities are identical with respect to the values obtained for the variables chosen for measurement or observation; a coefficient of "0" means that the two entities share no values in common. more specifically, Gower 's similarity between sherds was calculated by summing the contribution to similarity made by each level of measurement and dividing this sum by the total number of variables (10-11, depending on the paste subsample). The contribution to similarity by continuous variables was obtained by dividing the absolute value of the difference in measurments between two sherds on a particular variable by the range of values exhibited by that variable (difference between the largest and smallest value with respect to the entire data sets--somewhat different for each sherd subsample). A value of "l" is the most that any continuous variable can contribute to similarity. For computing the contribution to similarity made by the ordinal and qualitative multistate variables, the number of positive and negative matches (variables with the same value) between two sherds is simply counted. For computing the contribution to similarity made by the last two (binary) variables, the number of positive matches between two sherds is counted. Figure 8 is an example of a Gower's similarity matrix for ten entities computed on the basis of the variables measured in the present study.
* The fact that mismatches between entities on noncontinuous variables, regardless of how 'similar' they seem intuitively, will contribute values of "zero" is an example of one disadvantage of the use of mixed data for calculation of similarity. This situation has the potential to 'mask' high similarity between entities on other variables. This could be reflected as clusters that are heterogeneous in terms of the data describing them. Another disadvantage is the fact that each ordinal or qualitative multistate variable carries
45




SHERD NUMBER
1 2 3 4 5 6 7 8 9 10
1
2 0.222
3 0.316 0.440
4 0.316 0.440 0.818
5 0.316 0.440 0.818 0.818
z 6 0.370 0.579 0.404 0.404 0.404
2 7 0.676 0.364 0.403 0.403 0.403 0.472
8 0.473 0.317 0.375 0.375 0.375 0.425 0.518
9 0.397 0.370 0.464 0.464 0.464 0.561 0.428 0.531
10 0.176 0.591 0.570 0.570 0.570 0.624 0.318 0.429 0.597
Figure 8 Example of a Gower's Similarity Matrix Using Ten Entities




equal weight in contribution to similarity. This could also result in heterogeneous clusters. For example, high similarity in relative frequencies of large quartz sand inclusions might cause 'white' and 'red' firing sherds to be clustered together, even though two distinct clay sources are clearly represented. This is one reason why surface color was eliminated as a basis for calculating similarity. To minimize such risks, coding schemes for recording qualitative multistate data (for purposes of calculating a similarity matrix) should be carefully formulated (see Saffer 1979:154). For color in particular, as mentioned previously, the 214 ref ired sherds were hand-sorted into broad categories that may have significance in terms of clay resources used. This was done because the original Munsell measurements (which were quite variable), while highly useful for descriptive purposes, were, in many cases, not significantly different. Their use in calculation of similarity would lead to many meaningless 'mismatches'.
Measurements recorded for ref ired color and the other variables were punched onto computer cards for calculation of the similarity matrices using Blashfield's program. The data punched for the continuous variables represent the raw or original measurements. Numeric value labels were assigned to the measurements or observations recorded for ordinal/multistate (e.g. 0-4 for none through abundant) and binary variables (e.g. 1-0 for presence/absence) for purposes of punching these data onto cards. These data are recorded in Appendix D.
Secondary Variables
'Secondary' variables are distinct from 'primary' variables in that they consist of attributes that are, for the most part, culturally, rather than ecologically, conditioned. Such variables include decorative style and vessel form. Fundamental physical properties of pottery such as color and porosity, when measured in an original, unrefired state, may also be considered secondary variables, as they permit inferences to be made about cultural decisions regarding the manufacturing process (such as type of firing conditions). It is through the analysis of secondary variables, and their covariation with cluster results from analysis of primary variables, that the nature of production can be investigated. The secondary variables measured or observed in the present study include: surface decoration, vessel form, and variables 'relating to manufacturing techology such as method of vessel construction, surface finishing, color, hardness, porosity, and firing behavior. It
47




should be understood that the secondary status assigned to these variables is only with respect to the objectives of this study, and is not meant to diminish their importance as cultural variables.
Surface decoration. The pottery typology listed in Table 3-1 was used to classify the pottery samples in the present study. The specimens were categorized according the a combination of established decoration type names (Willey 1949; Sears 1956; Kohler 1978a), and paste (micaceous versus spiculite versus quartz sandgrit).
Vessel form. Vessel form was classified when possible according to formal typologies created by Willey and Sabloff (1975) and Sears (1956). Vessel wall orientation was classified using Kohler's categories (1978a:95). A magnetic profile or contour gauge was used for illustrating vessel wall orientation. Type of vessel orifice was also noted when possible, and used Shepard's (1976:228-230) unrestricted, simple restricted, and independent restricted categories for classification. The possible values or attribute states for the vessel form variables are listed in Appendix E.
Manufacturing technology. The manufacturing
technology(ies) represented by the pottery sample was described by recording observations pertaining to methods of vessel construction, surface finishing, and conditions of firing or firing behavior.
Type of sherd or vessel fracture (e.-g. coiled
versus irregular) was noted to determine method(s) of vessel construction. Measurements of mean body thickness were recorded using sliding metric calipers.. A binocular microscope (low magnification--lox) was employed to determine type or method of surface finishing (class of implement and state of plasticity). The presence of lustrous surfaces was also noted as this can contribute to identification of surface treatment. The observed values or attribute states for variables relating to manufacture are listed in Appendix E.
To make inferences about or estimates of original firing conditions represented by the sample, measurements of original (not refired) color, hardness, and porosity were made and a number of ref iring experiments were carried out.
Color. Color is a fundamental physical property of pottery which, as stated earlier, is determined principally by the organic and iron compound content of
48




clay and firing conditions--atmosphere, duration, and temperature. Color data were used primarily to make inferences about firing conditions and resources selected or used for manufacture. Original (not refired) slipped and unslipped exterior and interior surface colors and core color of the sherd samples were measured using Munsell Soil Color Charts. Measurements were made under consistent laboratory conditions, as described earlier. Both the alphanumeric notations and Munsell color names were recorded for each sherd. After measurement, the sherds were hand-sorted into three broad categories of surface color (see Table 3-5) and four of core color/degree of coring (retention of unoxidized organic material--dark zones in sherd crosssection; see Fig. 9).
Hardness. Hardness is a physical property of pottery which is determined primarily by paste composition and time-temperature relations of firing. Method of surface finishing can also affect hardness of pottery. Hardness data may be used to evaluate the likelihood of breakage or abrasion and, other things being equal, can provide some idea of the degree of firing. Hardness values were obtained in the present study primarily for making inferences about firing conditions and surface finishing. Unslipped interior surface hardess was measured with Mohs' Mineral Hardness Scale under 30x magnification. Mohs' Scale provides an ordinal level measure of scratch hardness, or how well a surface will resist abrasion.
Porosity. Porosity, as described earlier, is a fundamental physical property of pottery that can be a useful measure of body structure and is determined primarily by clay composition, proportion, size, and shape of aplastics, and by firing conditions. Original (unrefired) porosity measurements were used in the present study principally for making inferences regarding firing conditions and resources or pastes selected for manufacture. Original porosity (not refired) of sherds was measured as Percent Apparent Porosity (relative volume of open pores) using Shepard's boiling method described on pp. 40-41 of this chapter.
Refiring experiments. Measured and observed changes in color, coring, and sherd weight (Percent Firing Weight Loss) provided the basis for estimating original firing conditions. Refirings were made on sherds that were large enough (214 of 245) to have pieces broken off for this purpose, without destroying the whole sherd. These sherds were initially refired in an electric furnace under oxidizing conditions at 3000C for 30 minutes to equalize conditions from which further
49




Table 3-5 Categories of Original Surface Color
WELL-OXIDIZED COLORS
ext. int. ext. int.
CATEGORY 1A n= 48 61 CATEGORY lB n= 25 24
modal Munsell color: modal Munsell color:
10YR 6/4 7.5YR 5/4
light yellowish brown brown
MODERATELY-OXIDIZED COLORS
ext. int. ext. int.
CATEGORY 2A n= 63 39 CATEGORY 2B n= 13 16
modal Munsell color: modal Munsell color:
2.5y 5/2 2.5y 6/2
greyish brown light brownish grey
POORLY-OXIDIZED COLORS
ext. int. ext. int.
CATEGORY 3A n= 66 67 CATEGORY 3B n= 20 24
modal Munsell color: modal Munsell color:
5Y 4/1 5Y 3/1
dark grey very dark grey
modal slip/paint color: 1OR 4/6 red
50




a- 10YR hues n=22
no coring b- 2.5YR to 7.5YR hues n=19
a light to moderate "dark" n=11
b and "medium" coring n=15
a !n=14 moderate "dark" coring
3 b n=17
ci/- heavy "medium" coring n=13
an= 79
4 b heavy "dark" coring
white zones denote "light" or oxidized colors, hues range from 10YR to 2.5YR
with values ?5
hatched zones denote "medium" greyish and brownish colors, hues range
from 10YR to 5Y with values >35
black zones denote "dark" grey to black colors, hues are mainly 2.5Y and
5Y with values <3 and chromas <2
Figure 9 Categories Used to Measure Core Color/Degree
of Coring
51




measurements or observations were made. Sherd weight and color measurements (surface and core, degree of coring) were made after this refiring. These measurements and observations were repeated after additional oxidizing refirings to temperature levels of 500"C (maintained for 30 minutes) and 700*C (maintained for 30 minutes). Scratch hardness and porosity were also measured on sherds after the 700*C refiring. Percent Firing Weight Loss (%FWL) was calculated using the formula:
%FWL = dry weight fired weight x 100 dry weight
%FWL was calculated after the 5000C and 7000C firings
and summed to obtain cumulative %FWL.
Data Analysis of Secondary Variables
Measurements of original (unrefired) color, hardness, and porosity, and measurements or observations of decoration, form and manufacturing technology were punched onto cards for computer analysis. Nominal and ordinal or qualitative multistate and binary measurements were assigned numeric values for this purpose. These data are recorded in Appendix E. SPSS FREQUENCIES program (Nie, et al. 1975) was used to obtain frequency distributions for the variables while controlling for decorative categories. The covariation of depositional context with paste, decoration, forms, and the other secondary variables, was examined in order to investigate the sacred-secular dichotomy. The covariation of decoration with paste (cluster analysis results), form, and manufacturing techniques was examined in order to investigate production. These procedures and results will be discussed in detail in Chapters 6 and 7.
Description of Clay Samples and Methods of Analysis
Twenty-six clay samples from the North Florida
subregion were examined in this study. Analyses were carried out to obtain data concerning the nature and variability of these resources as part of the investigation of the North Florida 'ceramic environment' and to provide the sherd clusters with comparative data for investigating local versus nonlocal production.
52




The clay samples were examined to describe physical properties of plasticity, workability, texture, particle size and proportion, shrinkage, and firing behavior. Clay samples were air dried and then mechanically crushed with a mortar and pestle and/or ball mill. One hundred-gram samples of each crushed clay were processed through a graduated series of U.S.A. Standard Testing Sieves to obtain particle size and proportion information on aplastics. This involved soaking each 100 g sample for a few days, then washing each through the sieves and weighing the captured sediments. Three-hundred-gram samples of each clay were measured for testing working properties of the clays, such as plasticity, working range, and texture (stickiness or leanness), and for making test bars for further analysis.
Plasticity is the property of clays which allows them to be formed under pressure when water is adsorbed, and to retain form when pressure is relaxed and water is evaporated. The working range of clay, which refers to the amount of water added to clay to produce a workable, plastic mass, can be used as a relative indicator of degree of plasticity. Water was added to each 300 g sample to form plastic clay balls for testing this property. The amount of water added was recorded for each sample and was used as an indicator of one point in the working range or plastic limits of each sample. A somewhat more accurate measure of plasticity is Water of Plasticity (WP). Water of Plasticity is calculated using wet and dry weights of clay bars formed from the clay balls.
Prior to the formation of test bars, each clay ball was aged for five days to activate any bacteria that might be present in the samples; bacterial action tends to increase plasticity. After the aging process, any subjectively noted changes in the handling characteristics of the clays were recorded. Each sample was then wedged and formed into three or four 6xlx3/8 (approximately) inch bars for further testing. Each bar was weighed on a dial-o-gram balance beam scale immediately after being formed to obtain wet weights for calculation of Water of Plasticity, and then marked with points 10 cm apart for measurement of Linear Drying Shrinkage (LDS). After bars had air dried thoroughly for several weeks, they were reweighed and the marked distances remeasured. Water of Plasticity was calculated by the formula:
53




WP = wet test bar weight dry test bar weight x 100
dry test bar weight
Linear Drying Shrinkage was computed by the formula:
%LDS = length wet length dry x 100 length dry
The dry bars were also examined for presence and degree of warping and cracking.
After the above measurements were recorded, bars
were cut into Ix3/4 (approximately) inch briquettes for firing experiments. Briquettes were fired to determine firing behavior of the clays through measured and observed changes in color, coring, and %FWL. Briquettes were fired at six different temperature levels (3000C, 4000C, 5000C, 6000C, 7000C, and 8000C) under oxidizing conditions; each temperature level was maintained for 30 minutes.
Munsell color measurements of briquette surfaces and cores were recorded after each firing to document significant changes in or loss of coring with increasing temperatures.
A briquette from each sample clay was cut for
measurement of percent firing weight loss (%FWL). The unfired briquettes were first oven dried for approximately 18 hours at a temperature of about 110*C. They were then placed into a glass dessicator to cool to room temperature and weighed individually using an electronic balance. After each firing, the briquettes were placed in the drying oven to cool slowly. They were next put into the dessicator to cool to room temperature prior to reweighing. The %FWL measurements were computed on the basis of the difference between the dry, unfired briquette weights and their weights after firing. The %FWL formula listed on p. 52 was used for calculation of this measurement.
To examine origin of production, or the
possibility that some of the clays may have been the raw materials used for manufacture of some of the vessels represented by sherds in the sample, data on aplastic inclusions, color, and porosity were obtained from briquettes for comparison with the mean, range, and modal values of variables representing sherd clusters (rather than individual sherds). Porosity and color were measured on the briquettes that had been fired to 700*C for these comparisons. Since type,
54




frequency, and size of inclusions in the sherds were determined before sherds had been refired (hence, core colors were relatively dark in many cases), these data were obtained from briquettes that had been fired to the 500 C level (also relatively dark cores), so that the conditions of measurement for sherds and briquettes would be relatively comparable. The point-count method described for obtaining aplastic data for the sherds was also those used to obtain these data for the clays.
Summary
The nature of production of Weeden Island pottery at the McKeithen site was investigated using the methods of ceramic technology within the ceramic ecological approach. This approach permits comparable study and description of the nature and variability of fired (pottery) and raw (clays and tempers) ceramic materials with respect to their physical, mineralogical, and chemical properties.
Two-hundred forty-five archaeological specimens, 33 from mound or sacred contexts, and 212 from contemporaneous midden or secular contexts, were selected for study. Binocular microscopes, Munsell Color Charts, and an electric balance and kiln were among the principal tools employed in data collection.,
Variables selected for analysis were designated as primary and secondary. Primary variables consist of ceramic attributes that are more ecologically than culturally conditioned. In the present study, these included aplastic paste composition, refired porosity, and refired color. Data on these variables were obtained to provide the basis for clustering the 245 sherds into polythetic groupings which may represent particular clay resources. A multivariate clustering procedure called CLUSTAN was employed to do this. This procedure operated on separate similarity matrices computed for micaceous, spiculite, and quartz sand-grit paste subsamples. Gower's Coefficient was selected as the similarly coefficient for calculation of the matrices.
Secondary variables, or those which are more
culturally than ecologically conditioned, were also examined. These include surface decoration, vessel form, and those relating to manufacturing technology such as methods of vessel construction, finishing, and firing. The covariation of decoration with data on these other variables, and paste clusters, constitutes the means by which production was investigated.
55




Twenty-six clay samples from the North Florida subregion were also examined in the present study. Methods of analysis and much of the data collected are comparable to those for the pottery samples (primary variables). Data obtained from the clay analyses were used to describe the nature and variability of these resources as part of the investigation of the North Florida ceramic environment and to provide the pottery paste clusters with comparative data for investigation of local versus nonlocal production.
56




CHAPTER FOUR
NORTH FLORIDA CERAMIC ENVIRONMENT
Introduction
This chapter consists of a general description of the North Florida 'ceramic environment'. Its purpose is to document the environmental context necessary for understanding the possible development of specialized production of Weeden Island pottery, and to provide the comparative data base for examining origin of production.
An examination of a region's ceramic environment from a ceramic ecological perspective involves collection of data relating to an area's potential for availability of ceramic resources (e.g. clays, tempers, and fuels--exclusively clays in the present study) and climatological variability (Arnold 1975; Rice 1976). These data are collected to document the environmental context for any study of ceramic production. Data regarding the nature, variability, and distribution of available ceramic resources furnish part of the. necessary comparative basis for investigating questions of local versus nonlocal ceramic production. With respect to studies of the development of ceramic specialization, it is essential to be aware of existing environmental constraints that might have hampered or enhanced such development. For example, given the appropriate level of socio-political complexity, fulltime ceramic specialization would probably not evolve in an area lacking in suitable raw materials or in an area lacking a marked dry season (Arnold 1975).
In the present study, description of the North Florida ceramic environment was accomplished through examination of literature pertaining to the area's climate, geological history, soils, and through study and testing of clay samples collected from restricted areas of the North Florida subregion.
57




Cl imate
Describing the climate of North Florida in order to compile evidence against the possible existence of full-time ceramic specialization would be a moot endeavor in the present study. As mentioned, other archaeological data (Sigler-Lavelle 1980a, 1980b; Kohler 1978a) suggest that the North Florida Weeden Island population was only approaching a ranked level of organization, meaning that the preconditions or concomitants of such a degree of specialization (e.g. stratified, presumably agriculturally-based society) did not exist. The following discussion of climatological factors is, rather, included here in order to describe any possible problems or constraints that the North Florida potters might have encountered with respect to this aspect of the environment.
The area considered coincides approximately with
present-day Columbia and Suwannee counties (see Fig. 2, p. 5). The climate of this region has been described as 'humid subtropical' (Meyer 1962) and supports mesic (magnolia, oak, hickory) and xeric (pine, oak) vegetation. Long, warm, and humid summers, mild winters, and abundant rainfall are the characteristic climatic features. Mean annual rainfall for the Lake City area (Columbia County), as calculated from U.S. Weather Bureau records for a 65-year period (1893-1957) is 50.50 inches and ranges from 29.5 to 82 inches (Meyer 1962:7-8). Similar data recorded at five weather stations within a 30-mile radius of Lake City for a three-year period (1955-1957) exhibit an unequal horizontal distribution of rainfall (Meyers 1962:9, Table 1). In general, nearly half of the total annual rainfall occurs during a four-month rainy season lasting from June through September. Relatively high rainfall also occurs in early spring, from late February through early April (Butson 1963). The least amount of rain occurs during the fall and winter months (November is the driest month) and late spring (may, end of April; Butson 1963).
That these climatic trends, consisting of rainy summers and relatively dry fall-winters, may have prevailed during the time of Weeden Island occupation of North Florida (ca. A.D. 200-700) is supported by studies of environmental change. Palynological studies of post-glacial Florida environments have been carried out in North-Central and southern Florida (Watts 1969, 1975). These studies suggest that little environmental-climatic change has occurred since about 5000 years B.P., as the pollen profiles recovered exhibit pollen assemblages that are essentially modern
58




in character.
Ethnographic studies of non-kiln or 'primitive' pottery making indicate that pottery making is, in general, a dry weather activity, due to the requirements of the drying and firing processes. While pottery manufacture and the process of firing in particular can take place on non-rainy days during a period of variable weather conditions, fewer precautions to lessen the risk of cracking and warping during firing would be necessary during dry, warm weather. An example of one such precaution would be to place the unfired pots progressively nearer the fire, a step which insures complete drying, as well as decreases risk of thermal shock. The climatic data for the North Florida area indicate that pottery making during Weeden Island times may have been more successful or less risky if scheduled as a late fall or late spring activity.
Geological History
The general geological sequence in the North
Florida area consists of structurally complex Paleozoic age 'basement rocks', overlain by 2800-3460 feet of sediments ranging from early Cretaceous to Recent in age. Early Cretaceous to Early Eocene deposits consist primarily of dense marine limestones, some evaporites, and clay. The early Middle Eocene through Early to Middle Miocene sediments which over lie these rocks are principally porous marine limestone's. These underlie Middle Miocene to Recent age sediments, which consist mainly of sand and clay (Meyer 1962:12).
The oldest sediments which occur at or near ground surface in the study area are the Ocala group limestones Of Late Eocene age. The Ocala group limestones generally consist of soft, or porous, cavernous and fossiliferous limestones (Meyer 1962: 15-18). Solution pipes and horizontal cavities are common and usually filled with sands and clays more recent in age. Ocala group sediments underlie the entire two-county study area and range in thickness from about 150-250 feet. These sediments occur at or near ground surface in extreme southern Columbia and Suwannee counties and are exposed along the southern and western portions of the Suwannee River and southern portion of the Sante Fe River (see Fig. 10).
The Ocala limestones are unconformably overlain by the Late Oligocene age Suwannee Limestone formation. This formation underlies nearly the entire study region
59




*..... GEORGIA
FLORIDA -. .
: .: *
, : : : : *.
) CI
SPleistocene and Recent
*h .
&.e Miocene
0 24 48
km Eocene
kmn
Figure 10 Geological Exposures in Columbia and Suwannee Counties SOURCE: Puri and Vernon 1964




except where the Ocala limestones are exposed. This formation ranges in thickness from about 40-50 feet in northern sections of the area to a few feet in southern sections. It consists of very porous to dense fragmental limestone to dense dolomitic or cherty limestone. Solution pipes filled with fine to coarse quartz or phosphatic sands and light green clay are common (Meyer 1962:18-19). This formation occurs at or near surface in portions of southern Columbia County and in a large crescent-shaped expanse of land paralleling the Suwannee River in Suwannee County (Rowland and Powell 1965); it is also exposed along the Suwannee River in northern Suwannee County (see Fig. 10).
A Miocene age sandstone and limestone unit unconformably overlies the Suwannee Limestone in northern Columbia County. This unit is up to 70 feet thick in the extreme northern part of the county and thins toward the southern half. It is generally undifferentiable in character from the underlying Suwannee formation (Meyer 1962:19-20).
The Middle Miocene age Hawthorne formation
overlies this sandstone-limestone unit in Columbia County and the Suwannee Limestone formation in central Suwannee County. The Hawthorne formation consists of clay, sandy clays, sands, marl, limestone, fuller's earth, and phosphatic material (Meyer 1962; Rowland and Powell 1965). The color of the clay ranges from dark green to black to light green to grey. In Suwannee County, this formation ranges in thickness from 100 feet to one foot where exposed or near surface. It ranges from 150 feet in thickness in northern Columbia County to 100 feet in eastern Columbia County. This formation occurs at or near ground surface in southcentral Columbia County, and in most of the eastern half of Suwannee County (see Fig. 10). It is also exposed along the Suwannee River in northern Columbia County; beds of "nearly pure light green" Hawthorne formation clay are also exposed along the valley of Olustee Creek and the Santa Fe River along the southeastern Columbia County border (Meyer 1962:20,). Trapped remnants of Hawthorne material also occur throughout areas where Suwannee Limestone is exposed. Where at or near surface in south-central Columbia County, this formation may be equivalent to the Alachua formation. The Alachua formation consists chiefly of clays 'and sandy clays. Although the exact relative age and relationship (to Hawthorne material) of this formation is disputed, it has been suggested that it represents collapsed and compacted residue of Hawthorne material and that it is of Pliocene age (Cooke 1945: 200-201). In south-central Columbia County, these
61




clayey and sandy clay sediments are oxidized to white, red, pink, brown, and buff colors (Meyer 1962:20-21).
Pleistocene terrace and Recent deposits overlie Miocene materials (see Fig. 10). They consist principally of sands and clays and represent reworked Hawthorne formation materials. The thickness of these deposits varies greatly over the study region; the thickest accumulations (up to 40 feet) are found in Columbia County.
Topography
Ease of access to clay resources (e.g. surface or near-surface exposures) was probably an important limiting factor in aboriginal utilization of clay materials. Topographic variability is important to a discussion of an area's ceramic environment, because it may, for example, make available, in the form of stream cuts or solution depressions, clay resources not accessible at or near ground surface.
The major topographic features of the North
Florida region are a series of marine terraces produced by fluctuating Pleistocene seas. These terrace deposits form the upper soil strata ranging up to about 40 feet in thickness. They are composed primarily of sands and clays which represent reworked Miocene materials. The terrace deposits overlie about 3000 feet of Marine limestone sediments ranging from early Cretaceous to middle Miocene in age (Meyer 1962).
North Florida has been divided into two
physiographic zones: the Coastal Lowlands and Central Highlands (Meyer 1962:7-12; Cooke 1945). The Coastal Lowlands are represented in southwest Columbia and southeast Suwannee counties and along the valleys of the area's major drainage systems (Suwannee and Santa Fe rivers, and Olustee Creek). The Coastal Lowland areas range in altitude from about 25 to 100 feet above mean sea level (amsl.) and were terraced by the Wicomico (100-foot), Penholoway (70-foot), Talbot (42-foot), and Pamlico (25-foot) Pleistocene shorelines (Cooke 1945; Meyer 1962). This region exhibits active karst topography, as it is underlain by low hills of silicified, cavernous limestone. Consequently, there are many collapsed sinkhole lakes and ponds. These are alligned parallel and perpendicular to major drainage features and are connected by gulleys or valleys of intermitant streams of an ancestral drainage system. A well integrated pattern of streams has, however, never developed in this area (Meyer 1962).
62




The majority of land Columbia and Suwannee counties is considered Central Highlands physiographically. The Central Highland province is distinguished from the Coastal Lowlands by the Wicomico 100-foot Pleistocene shoreline. This area is composed mainly of sands and clays terraced by early Pleistocene seas. The highest elevations in the area were deposited as the Coharie and Sunderland terraces (170 and 215 feet amsl., respectively). Remnants of these terraces form a high ridge crossing the North Florida region from west to east, passing through Wellborn, Lake City, and Olustee, Florida. Most of the counties' streams are tributaries of the Suwannee River and occur in this ridge area (Rowland and Powell 1965; Meyers 1962). Solution depressions and sinkhole lakes and ponds are also common.
The two-county area as a whole is characterized by gently undulating topography, a product of the Pleistocene terrace deposits, subsurface collapse of the underlying limestone formations, and erosion. The streams, numerous solution depressions, sinkhole lakes and ponds, and ancestral drainage features are significant in that they may, in some cases, increase the accessibility of clay resource opportunities in the region.
Soils
The potential for availability of clay resources in North Florida was investigated through examination of literature concerning clays and soils of the area.
Previous studies of clay resources in this area
were very sparse and limited to general reports on 'the clays of Florida' (e.g. Bell 1924; Matson 1909; Sellards 1918). The information recorded about the clays was used to aid evaluation of samples in terms of their potential for use in modern manufactures. Of these studies, only Bell's "A Preliminary Report on the Clays of Florida" (1924) provided any information for Columbia and Suwannee counties. This report states that sandy surface clays, which mantle a large part of Columbia County, are useful only as sand-clay road material. South of the Lake City area, "no clays of value are known" (1924:133). This observation was also said to be true of the northern and northeastern parts of the county. Suwannee County is likewise capped by mixed sand and clay which has been used as road material. Sandy clays are particularily common in the central and eastern parts of the county and were deemed
63




suitable for "poor grade common brick" (Bell 1924:207). While evaluations of clays by industrial standards as being "undesirable for any kind of manufactured clay product" (Bell 1924:133) are essentially meaningless with respect to the present study, this 1924 report does indicate that sandy clay resource opportunities in the study area are abundant and widespread.
The distribution of clays containing certain kinds of inclusions, such as mica or sponge spicules, is not well documented for Florida. For the Weeden Island region, northwest Florida and southwest Georgia have been traditionally considered as sources for micaceous paste ceramics recovered elsewhere. While clays with variable but relatively high mica content are known to occur in the northwest Florida-southwest Georgiasoutheastern Alabama areas (see White 1981:602), the geological literature yields references to the presence of mica in clays elsewhere in the state. Bell's early report states that presence of mica is not unusual in Florida clays; that few Florida clays are free of it (1924:73). Calver (1949:2; 1957:57) and Pirkle (1960) report the presence of mica in extensive deposits of sandy kaolinitic clays. These are Citronelle formation deposits which occur in the Lake Wales Ridge area of central peninsular Florida (Pirkle 1960); this distribution, which lies well outside the study area, includes portions of Clay, Putnam, Marion, Lake, and Polk counties. Pirkle also reports the occurrence of mica in Hawthorne formation deposits, which do occur in Columbia and Suwannee counties (1960:1398). Rare occurrence was noted in five of the sample clays in the present study (see p. 70, this chapter). Although the distribution of micaceous clays is wider than had been suspected, it may be reasonable to suggest that clays with abundant mica may be accessible primarily in the northwest part of the state.
Even less is known about the distribution of sponge spicule clays in Florida. East Florida is traditionally considered as the source for spiculite pottery, but Thanz and Shaak (1977) suggest that the distribution of spiculite clays may be wider ranging. At present, spiculite clays have been recovered from the Upper St. Johns area in Brevard County (Espenshade 1983) and from the Oklawaha River in Putnam County (Ceramic Technology Laboratory, Florida State Museum, Gainesville). The abundance of spiculite ceramics, that is, Papys Bayou series, in the Central Peninsular Gulf Coast (Willey 1948:211) suggests another possible source for spiculite pottery, although spiculite clays have yet to be located in the area (Mitchem and Welch 1983:149). No sponge spicules were observed in any of the sample clays in the present study.
64




The nature and distribution of, or potential for availability of, ceramic resources in the study area was determined primarily through examination of U.S. Department of Agriculture (USDA) Soil Conservation Service soil survey maps of Columbia (1976) and Suwannee (1965) counties. This focused on the generalized soil profiles which were described as to USDA texture (see Soil Survey Manual 1951:209, Fig. 38; 210-211) and USDA measured or estimated particle sizes (Soil Survey Manual 1951:207; Table 2). In the present study, potentially usable clay resources were defined as moderately fine-textured or loamy soils (clay loam, sandy clay loam, silty clay loam) and fine-textured or clayey soils (sandy clays, silty clay, clay).
The soils of Columbia and Suwannee counties were found to consist mainly of fine and loamy sands, which are, in most cases, underlain by loamy and clayey materials. Most loamy or loamy to clayey strata occur more than one meter below ground surface. Finertextured (clayey) materials are more accessible in terms of depth below surface, but are not as abundant with respect to total acreage represented. These soils are described in terms of thickness of strata, depths below surface, and particle size breakdown in Appendix F. Within a five mile radius of the McKeithen site (the McKeithen area's 'effective ceramic environment', after Arnold 1975), soils with clayey or loamy substrata are widespread but are most abundant in the northern part of the Columbia County side of this area, and the southern part on the Suwannee County side (see Appendix G).
Sample Clay Analyses
Twenty-six clay samples from the North Florida subregion were examined to gain a first-hand understanding of the nature and variability of some of the area's available clay resources, and for comparison with the McKeithen pottery samples (Chapter 7). Of the 26 sample clays, four were collected from accessible sources (e.g. stream banks and beds or roadcuts) from the vicinity of the McKeithen site by Drs. P. Rice (University of Florida and Florida State Museum, Gainesville), T. Kohler (Washington State University, Pullman, Washington), and the auther. The remaining 22 samples were gathered by Dr. B. Sigler-Lavelle (Florida State Museum, Gainesville) and the late L.J. Loucks during their respective surveys of portions of Columbia (Sigler-Lavelle 1980a) and Suwannee (Loucks 1979) counties. These samples were collected from accessible
65




sources or sources intersected by excavation of test pits. Figure 11 plots their approximate geographical locations. Information for each sample pertaining to context of collection, form, thickness, and extent of the deposits, characteristics in situ (e.g. sandiness or stickiness) and other details, is located in Appendix H.
The sample clays were examined to describe
physical properties such as plasticity, workability, texture, particle size and proportion, shrinkage, and firing behavior. Knowledge of the variability of these properties permits identification of difficulties or constraints that could be expected in terms of their suitability for pottery manufacture. The methods of measurement or observation that were used are described in Chapter 3, pp. 52-55. The results of these analyses are presented by discussing the entire group of sample clays together with respect to each of the properties measured or observed. In this way, an idea of the range of variation in these properties exhibited by the clays is obtained.
Plasticity and Handling Characteristics
The sample clays examined ranged in handling
characteristics from very short or lean and mealy to very plastic or fat (see Appendix I). The sample clays could be sorted into three categories with respect to handling characteristics ranging from "poor" (1) to "good" (3) (see Table 4-1). Category 1 consists of very short or lean clays (n=6) which generally cracked badly when squeezed, and during wedging and formation of test bars. The second category (n=8), characterized by "fair" or moderately plastic handling, also cracked when when handled or worked but not nearly to the extent of those clays in the first category. Category
3 consists of plastic to very plastic clays and characterizes the remaining 11 sample clays. The clays in this category generally cracked very little or not at all when worked. One other sample clay, C-19, was measured only for particle size and proportion (an insufficient amount was collected for more extensive analysis).
The amount of water added to standard 300 g
samples of each of the sample clays (except C-19) to produce plasticity and workability constitutes a measure of one point in the working range of a clay. It can also serve as a rough indicator of the plasticity of a clay. The amount of water added to the sample clays to achieve plasticity ranged from a low of 55 ml to a high of 115 ml (see Appendix I). The amounts closely reflect the categories of handling
66




*...... GEORGIA
Ev FLORIDA . .i i
*D *C
*B
SUWANNEE *McKeithen
COUNTY
H COLUMBIA
. COUNTY
O F
SAMPLE CLAY LOCATIONS
A- C-1A,1B,3,5
B- C-22,23 o4t, e C- C- 21, 32
'er D- C-2
E- C- 26, 27 F- C-10,11
0 24 48
, ,, G C-12, 13,14
km H- C-15,16,17,18,19.20,24,25,30,31
Figure 11 Approximate Locations for the Clay Samples




Table 4-1 Relative Categories for Sample Clay Physical
Properties: Handling Characteristics,
Plasticity, and Shrinkage
HANDLING or WORKING
CHARACTERI STICS
POOR (1) FAIR (2) GOOD (3)
n=6 n=8 n=11
2 24 1A 15 12 18 27
3 25 lB 17 13 20 30
10 32 5 22 14 21 31
11 23 16 26
AMOUNT of WATER ADDED
1 2 3 4
n=6 n=6 n=ll n=2
45-56.5 ml 60-65 ml 70-82 ml >100 ml
2 24 1A 15 5 16 23 26
3 25 1B 17 12 18 30 27
10 32 11 22 13 20 31
14 21
WATER OF PLASTICITY
1 2 3 4 5
n=6;13-17% n=6;19-21% n=6;22-24% n=5;25-30% n=2;>30%
R=15.3% R=19.5% iR=23.1% R~=25.8% R=36
2 24 1A 15 5 18 14 21 26
3 25 lB 17 12 23 16 30 27
10 32 11 22 13 31 20
LINEAR DRYING SHRINKAGE
1 2 3 4 5
n=2;<1.0% n=4;3-5% n=4;>5-<7% n=5;7-9% n=10;>9.0%
R=.% =.0 =5.8% R=8.2% R=10.0%
3 2 25 1A 11 5 15 14 23
24 10 32 lB 21 12 17 16 26
13 18 27
20 30
22 31
68




characteristics described above, but the distribution of amounts was divided into four categories. These categories are listed in Table 4-1.
A somewhat more accurate relative measure of a clay's plasticity is Water of Plasticity. To obtain this measurement for the sample clays, each clay ball was formed into three or four test bars (see Ch. 3, P. 53). Water of Plasticity (WP) was calculated using the formula on p. 54, based on wet and dry weights of the test bars. Water of Plasticity for each test bar was calculated; the range and mean WP for each clay are presented in Appendix I. Mean WP for the sample clays was 21.8%, and ranges from 13% to 35.9%. Breaks in the distribution of mean WP measurements were used to sort the sample clays into five WP categories, described in Table 4-1. As this table shows, the clays which required the least water to become plastic, and which have the 'shortest' handling characteristics, have the lowest WP measurements. Conversely, clays that were 'fatter' and which required more water to become workable, have the highest WP measurements.
Before the test bars were formed for measurement of Water of Plasticity, each clay ball was aged in a plastic bag for five days in order to detect the presence of bacteria (usually indicated by an unpleasant odor after aging). Bacteria sometimes alter the working or handling characteristics of clays by increasing plasticity. Bacteria were assumed to be present in 17 of the sample clays. The strength of the odors exhibited ranged from 'ripe' or mildly unpleasant in ten of the clays to unpleasant and very unpleasant in seven clays (see Appendix I). No changes in the handling characteristics or workability of these clays were noticed after aging, however.
Drying Behavior
The drying behavior of the sample clays was
examined through subjectively noted observations of warping and cracking, and through measurement of Linear Drying Shrinkage (LDS). These observations and/or measurements document the behavior of clays resulting from loss of mechanically combined water, primarily that which was added to the clays to produce plasticity and workability.
No warping or cracking (or further cracking of
test bars made from the very short clays) were noticed in any of the sample clay test bars.
As discussed in Chapter 3, each clay test bar was marked with 10 cm distances immediately after having
69




been formed. After the test bars had air-dried thoroughly for a number of weeks, these distances were remeasured. Linear Drying Shrinkage (LDS) measurements are based on the difference between the wet and dry measurements and were calculated by using the formula shown on p. 54. The results of these measurements are presented in Appendix J.
The mean LDS for the sample clays ranged from an extremely low 0.1% for the shortest clay (C-3), to a high of 10.5% for one of the more plastic clays (C-26). The mean measurements computed for each sample clay (from the set of three or four test bars representing each clay) were sorted into five categories based on breaks in the disribution of measurements. These categories are described in Table 4-1.
Aplastic Composition, Particle Size and Proportion
One hundred-gram samples of the 26 sample clays
were processed through graduated series of USA Standard Testing sieves to obtain aplastic particle size and proportion data. Gross identification of aplastics was determined primarily through microscopic examination (70x magnification) of the materials captured by each sieve.
Aplastic composition. All of the clays examined contain quartz inclusions ranging in size mainly from "medium" through "very fine" (see Wentworth's Scale, Appendix D). Other inclusion types observed include mica, limestone, whitish concretions or lumps, and ferruginous concretions or lumps. This information is recorded in Appendix K. Sponge spicules were not observed in any of the sample clays. Mica was observed in five of the clays (C-lA, 1B, 2, 5, 21) but in very rare frequency; presence was not conspicuous, and was not observed consistently in all fired briquettes. Presence was more noticeable in the plastic bags (adhering to the insides) containing sieved samples. Limestone concretions ("coarse" to "very coarse"~ in size) were observed in one of the clays (C-19); both the concretions and clay particles effervesced vigorously upon application of hydrochloric acid.
Whitish concretions or lumps were predominantly
"medium" to "coarse in size and often contained "very fine" and "fine" quartz sand inclusions. Some appear to be uncrushed or cemented sand-clay concretions or clay lumps not thoroughly crushed and mixed during paste preparation. They look very much like the whitish concretions or lumps that were observed in some of the pottery samples. This inclusion type is a distinctive feature (common occurrence) in three of the
70




sample clays: C-26, 27, and 31. Raw and fired samples of these clays were tested with hydrochloric acid to eliminate the possibility of their being limestone concretions. These clays did effervesce vigorously when the acid was added to the unfired samples. The reaction did not, however, come from the concretions specifically. This indicates that the clays are calcareous in composition. Fired samples of these clays also effervesced but to a less vigorous degree.
Whitish concretions were observed in occasional frequency in six of the clays (C-15, 16, 18, 22, 23, and 25). Occurrence is generally rare in the remaining 16 sample clays. Raw and fired samples of each of these 22 clays were also tested with HC1 to eliminate the possibility that the concretions are limestone. Raw samples of C-15 and 16 effervesced mildly and briefly. As with C-26, 27, and 31, the reaction did not seem to itsue from the concretions, indicating that these two clays may be slightly calcareous in composition. No effervescence was observed when HC1 was added.to the fired samples. Raw samples of the remaining clays effervesced only very mildly and briefly when treated with HC1 (again, not from the whitish inclusions specifically). The fired samples of these clays did not effervesce at all.
Ferruginous concretions or lumps were
predominantly "medium" to "coarse" in size and often contained "very fine" to "fine" quartz sand inclusions. They look very similar to the ferruginous inclusions observed in many of the pottery samples. This inclusion type is a distinctive feature in C-15 and 24. They were observed in occasional frequency in C-16, 18, 23, and 25; the remaining 20 sample clays generally exhibited these inclusions in rare frequency.
Texture: particle size and proportion. Ten of the 26 sample clays (predominantly those from Columbia County) were processed through seven sieves ranging from size 4 to 170 (sizes 4, 18, 30, 60, 70, and 170). The material that passed through sieve 170 approximates the amount of silt and clay in a sample clay. The remaining 16 clays (predominantly those from Suwannee County) were processed through 12 sieves (a new set was purchased by the Florida State Museum during the analysis) ranging from size 5 to 325 (sizes 5, 10, 18, 25, 35, 45, 60, 80, 120, 170, 230, and 325). The material that passed through sieve 325 consists mainly of clay, but still contains some silt. Sieving results for all sample clays are listed in Appendix L.
The 26 sample clays were assigned USDA texture
names on the basis of particle size and proportion data
71




obtained from sieving, with reference to the Soil Survey Manual (1951:209--Fig. 38). An understanding of the actual amount of clay-sized particles in the sample clays (not obtained through sieving), which was obtained through measurement and/or observation of other of their physical properties (e.g. handling characteristics, WP, etc.), also aided the assignment of textural categories.
The 26 sample clays were grouped into four
textural categories: 1) sandy oans (SL) to. loamy sands
(LS); 2) sandy clay boas to sandy oans (SCL-SL); 3) sandy clay oans (SCL); and 4) sandy clays to clays (SC-c), clays (C), and silty clay to clays (SiC-C) (see Table 4-2). Appendix M lists the 26 clays according to USDA texture and proportions of materials of certain particle sizes (Wentworth's Scale).
The first textural category (1) consists of sandy loams to loamy sands and includes two sample clays. The proportion of clay in these samples in probably between 10 and 20% (see Appendix M for particle size proportions). The proportion of silt and clay (material passing through sieve 170) ranges between about 18 and 20%. The predominant sand-size categories (in terms of weight) for these samples are "fine" and "medium". These two sample clays are extremely sandy and exhibited the least plastic handling characteristics and lowest WP.
Category 2 is characterized by sandy clay loam to sandy loam textures and includes four sample clays. The proportion of silt and clay exhibited by these clays ranges from about 28 to 49%. On the basis of their handling characteristics (relatively poor), WP (relatively low), and upon sieving results, the actual proportion of clay particles may be less than 25% (probably greater than 15% silt). The predominant sand-size category for these samples (in terms of weight) is "fine".
Category 3 is characterized by sandy clay loam
textures and consists of eight moderately plastic (in terms of handling characteristics and WP) sample clays. The proportion of silt and clay ranges from about 28 to 47%. On the basis of their handling characteristics (relatively good), WP (relatively high), and sieving results, the proportion of clay present in these samples probably ranges between 20 or 25% and 35% (probably less than 15% silt). "Fine" and "medium" are the predominant sand-size categories (in terms of weight) for these samples.
72




Table 4-2 Relative Categories for Sample Clay Physical
Properties: USDA Texture, Relative Paste
Texture, and Fired Color
USDA TEXTURE
1 2 3 4
n=2 n=4 n=8 n=12
SL,LS SCL-SL, SCL, CSC,C-SC,
SCL-L SCL-SC SC-CSiC-C
3 2 1A 15 12 18 26
24 10 1B 17 13 19 27
25 5 22 14 20 30
32 11 23 16 21 31
RELATIVE PASTE TEXTURE
1 2 3
FINE-MEDIUM MEDIUM MEDIUM-COARSE
n=14 n=7 n=4
1A 12 18 3 24 15
lB 13 20 10 25 21
2 14 30 22 32 26
5 16 31 23 27
11 17
FIRED CORE COLOR (7000C)
1 2 3 4
n=2 n=2 n=8 n=13
1B 1A 5 22 2 17 24
3 21 10 26 12 18 25
11 30 13 20 27
15 32 14 23 31
16
73




Eleven of the remaining 12 sample clays are sandy clay to clay materials, and the last is a silty clay to clay. These 12 clays make up texture category 4. The proportion of silt and clay ranges from about 47 to 83% (principally 50 to 70%). On the basis of knowledge of their handling characteristics, WP, and sieving results, the actual proportion of clay particles in these samples probably ranges between 40 to 60%. The predominant sand-size category (in terms of weight) for these samples is "fine".
The approximate location of each of the sample
clays was plotted on the soil survey maps for Columbia and Suwannee counties to ascertain whether or not the textural designations assigned through this analysis agreed with the soil survey textures of the soils from which the samples were taken. This comparison revealed that the textures assigned through analysis, in most cases, closely reflected ones that would have been predicted through examination of the soil survey maps; this also revealed that soils with clayey or loamy substrata may be more widespread than the soil maps indicate. Although most of the Suwannee County samples come from an area well outside the McKeithen vicinity, the soil types which occur in the southwest Suwannee County collection area are ones that also occur commonly within a five-mile radius of the site (see Appendices F, G, and H). The implication is that clay deposits similar in physical properties to those of the samples from southwest Suwannee County will also be available closer to the McKeithen site. The soil types or series that are believed to be the sources for the sample clays are discussed as a part of the individual descriptions in Appendix H.
Inclusion point-counts. The procedure for
obtaining point-counts of quartz sand inclusions from the archaeological samples, described on pp. 38-39 of Chapter 3, was also used on a fired briquette of each of the sample clays. Inclusion point-counts of the clays were carried out in order to obtain a valid basis for comparison with the sherd samples for investigating local versus nonlocal manufacture (Chapter 7). The clays were also sorted into relative paste texture categories (see Table 4-2), comparable to the ones observed for the sherd samples. Point count results and the paste texture categories are recorded in Appendix N.
Firing Behavior
The firing behavior of the sample clays was determined through measurement and observation of changes in color, coring, and weight (%FWL) of
74




briquettes fired to a number of different temperature levels. Changes in briquette strength or friability with increasing temperature were also subjectively noted. Briquettes were fired under oxidizing conditions in an electric furnace. Six temperature levels were employed (300*C through 800*C at 1000 increments) and each was maintained for 30 minutes.
Color and coring. The sample clays were grouped into four broad core color categories based on briquettes that were fired at 700'C for 30 minutes. The four categories include: 1) whites to very pale browns (lOYR hues); 2) very pale browns to pale browns or pinks (lOYR and 7.5YR hues); 3) light brown to browns to pinks and yellowish red to reddish yellows (7.5YR and 5YR hues); and 4) yellowish reds to reds (5YR and 2.5YR hues). These categories correspond roughly to categories 1-4 of refired sherd core color (see Table 3-4, p. 43 and Table 4-2). Munsell color measurements for each temperature level, and core color values (at 7000C) are recorded in Appendices 0 and N.
Two of the sample clays are white-firing at the 7000C level. Both surfaces and cores of these clays are white. The fired color of these clays indicates that they contain very low amounts of oxidized iron compounds (probably less than l%--Shepard 1976:150), relative to most of the other samples. The color of the raw, unfired clays ranged from white to light grey to grey.
Two sample clays exhibited fired core color
category 2 colors (very pale brown to pale brown to pink) at the 7000C level. Surfaces also exhibit a similar range of colors at this temperature level (fired surface color categories 1 and 2). The range of fired colors exhibited by these sample clays indicates relatively low oxidized iron content (probably less than 4%--Shepard 1976:150). The unfired colors of these clays ranges from white to light grey to pale yellow.
Seven sample clays are l ight brown to brown to yellowish red to reddish yellow firing at the 7000C level (core color category 3). A similar range of surface colors is exhibited by these clays (surface color categories 2 and 3). The range of fired colors exhibited by these sample clays indicates a higher iron content than the previous clays. The raw or unfired colors of these clays ranges from light grey to light greyish brown and greyish brown to light olive brown and olive.
75




Thirteen of the sample clays are red-firing at the 7000C temperature level (fired core color category 4); surfaces are also red-firing (fired surface color category 3). Core and surface colors range from yellowish reds to reds to reddish browns. The range of fired colors exhibited by these clays indicates that they contain more oxidized iron compounds than the others. The unfired colors of these clays are quite variable and range from light olive brown to yellowish brown to greyish brown to brown to reddish brown to reddish yellow to yellowish red.
Surfaces and cores of most of the sample clays turned grey to dark grey at the 3000 and 4000C temperature levels. Most of the samples also exhibited a noticeable or marked loss of greyness (increase in Munsell value) of core and surfaces after the 5000C firing. Loss of grey tones at briquette surface and cores appears visually to be complete or nearly so (oxidation of organics appears to be complete) after the 600'C firing for most of the sample, (after the 5000C level for some). Little to no changes in color from the 600'C through 800'C levels were observed (or perceptible). All organic materials appear to have been completely oxidized at the 6000C level.
Percent firing weight loss (%FWL). Weight loss in fired clays reflects the oxi-dation of organic materials and other impurities present,. and the loss of mechanically and chemically combined water. Mechanically combined water is driven off in firing normally below 3000C. Chemically combined water, or that which forms part of the structure of the clay minerals, is given off between about 4500 and 6000C (Rice 1976:170). Oxidation of organic materials begins at about 3000C. The duration of this process depends upon the amount of carbon present, time, temperature, and atmosphere of firing, texture of the clay (degree of fineness), and the degree to which the organics are retained within structure of the clay minerals (e.g. montmorillonites [Rice 1976:170]).
The %FWL of each sample clay was computed on the
basis of differences between the dry, unfired briquette weights and weights after each firing. The computed %FWL results for each clay are listed in Appendix P. C-24, a sandy loam, exhibited the lowest total weight loss (1.80%), while C-30, a sandy clay to clay material, exhibited the highest total loss (8.13%). The greatest weight loss at a particular temperature for most of the sample clays occurred between 50QO and 600'C, marking the range at which most of the organic materials and chemically combined water were driven off. Weight loss decreased after this range, with most
76




of the clays exhibiting nominal cumulative losses ( Porosity. Percent Apparent Porosity was measured on sample clay briquettes that had been fired at 7000C for 30 minutes to provide comparative data for the sherd cluster mean values in the investigation of locus of manufacture. Shepard's boiling method, described on pp. 40-41 of Chapter 3, was used to obtain the porosity measurements for the clays. The results of measurement are presented in Appendix N. Porosity for the clays ranged from a low of 23.6% (C-12) to a high of 42.4% (C-27). mean porosity for the entire sample is 32%; most of the clays fell between 30% and 35%.
Other observations. The affect of firing on
friability of the clays, or how easily they could be broken or crumbled, was also examined. Pliers were used to break off pieces of unfired briquettes and from those fired to each temperature level. Each break was rated 0 or "poor" (very crumbly or friable and easy to break), 1 or "fair" (less friable, but still easy to break), or 2 or "good" (more difficult to break, much less friable); a briquette could also be rated "poorfair" (0-1), etc. These observations are extremely subjective, but do provide a very general idea of the effects of firing on the strength of the sample clays. The results are presented in Appendix Q.
The sandy loams (C-3 and 24) exhibited very poor strength in the unfired state and did not exhibit any improvement upon firing. Briquettes fired to each of the temperature levels were extremely friable and crumbled under the pressure of the pliers. Most of the other clays, representing the other three textural categories, broke easily in the unfired state through the 4000C firing. Improvements were noted after the 5000 firing for most of the clays, but, in general, no improvement at higher temperatures was observed. Some seemed to exhibit slight increases in friability after the 800*C firing.
Summary and Conclusions:
the Effective Ceramic Environment
This chapter has presented a general description of the North Florida 'ceramic environment', or nature of and potential for available raw ceramic resources. The objectives were to provide an environmental context for understanding the potential for development of ceramic specialization in the area, and to provide data for comparison with the McKeithen ceramics samples to
77




ascertain origin of production. This was accomplished through investigation of the area's climate, geological history, topography, soil maps, and through analysis of 26 clay samples from the immediate vicinity of the McKeithen site, and from surrounding areas in Columbia and Suwannee counties.
The North Florida climate was categorized as humid subtropical, with rainy summers and relatively dry fall-winters. Palynological studies in Florida have suggested that this climate has existed since approximately 5000 B.P. The literature on geological history and soils indicated that the distribution of clay resource opportunities in the study area are fairly abundant and widespread; North Florida is composed mainly of sands and clays which were terraced by fluctuating Pleistocene seas. Most of the soil types within the area in general, and within a fivemile radius of the site, have clayey or loamy subsoils. Particle size and textures are variable, however, and many appear to be quite sandy and occur at least one meter below ground surface. Finer-textured materials are more accessible in terms of depth below ground surface, but are not as abundant with respect to total acreage represented.
Analysis of the 26 clay samples was carried out to describe variability in physical properties such as plasticity, shrinkage, texture, and firing behavior. For the samples as a group, working characteristics ranged from "poor" (n=6) to "good" (n=ll). Water of Plasticity ranges from "low" (n=6; R=15.3%) to "high" (n=2; i=33.6%), and Linear Drying Shrinkage ranges from "tlow" (n=2; i=.6%) to "high" (n=10; R=10%). The estimated amount of clay in the samples ranges from a low of approximately 10-20% (n=2) to a high of about 40-60% (n=12). USDA textural categories represented by the samples range from SL-LS (n=2) to SC, C and SiC-C (n=12). Fired colors (700*C) range from white and very pale brown to red and reddish brown, but high-iron clays outnumber low iron clays in the sample 21 to 4.
The modal or most frequently encountered of the sample clays can be characterized as having "good" working properties, "moderate" WP (i=21.3%), and "high" LDS (i=10%). The approximate amount of clay and silt particles in this modal clay is about 40%, with about 60% coarser aplastics. USDA texture is SC; texture with respect to the sherd samples is "fine to medium", and this clay is high in iron compounds (fires to reddish colors).
The humid-subtropical climate, depth of deposits (limiting the accessibility of the materials), and
78




sandiness of the clays (affecting their handling characteristics) represent three environmental constraints that may have been encounted by North Florida potters. Although the question of full-time specialization is not applicable given the sociopolitical and economic level for the North Florida Weeden Island society, the climatic pattern would probably not have encouraged development of formalized full-time specialization of non-kiln pottery manufacture. The second constraint, accessibility (assuming little significant change in the past 1500 years) may, however, be offset to some degree by the topographic variability in the area. The McKeithen site is located in the ridge area discussed in the second section of this chapter. As mentioned, most of the region's streams occur in this area; sinkholes and solution depressions are also common. These features have the potential to increase accessibility of clayey soils. This suggests that if access to resources was limited, social, not environmental, factors would be operating.
The third constraint--abundance of sand, which characterizes most of the available potential clay resources--may have rendered many clays unsuitable for some productions unless modified through sieving or levigation to remove a portion of the aplastics. However, the degree to which sandiness may have been considered a problem cannot be answered without committing a number of subjective value judgments. The consensus from the soils literature is that available clays were generally undesirable for use in modern ceramic industry. Many of the properties considered desirable for modern use, would undoubtedly be inappropriate for aboriginal use, given 'primitive' methods of construction and open firing. Six of the 26 samples studies were categorized as having "poor" working characteristics, and only two of these were difficult to form into 6xlx3/8 inch test bars. On a purely intuitive level, from the perspective of an individual unskilled in pottery making, it is suggested that 'sandiness' of the available clays would not in general have been considered an unmanageable problem.
79







CHAPTER FIVE
CLUSTER ANALYSES RESULTS
Introduction
Cluster analyses are heuristic devices that assist the ordering of data. Hierarchical agglomerative cluster analyses were used in this study as this type of procedure produces polythetic clustering solutions; members of each polythetic cluster will be highly similar but not necessarily identical in terms of the variables used to describe them (Blashfield and Aldenderfer 1976). Hierarchical cluster analyses produce cluster solutions in which the relationships between entities, or objects being clustered, are hierarchical; that is, the methods start with the same number of clusters as there are entities (K [number of clusters] = N [number of entities]) and join entities into larger and more inclusive groups until all entities are members of the same cluster (K=l) (Aldenderfer 1977). The results of a cluster analysis and steps in the hierarchical procedure are graphically portrayed in a 'dendogram' or tree which plots entities on the y-axis and coefficients of similarity (values at which entities join entities or clusters) on the x-axis (e.g. Figures 12-14).
The principal advantage of using a computerized cluster analysis is that it can objectively evaluate large quantities of data in a short period of time. The principal disadvantage of hierarchical agglomerative cluster analyses is the subjectivity involved in interpretation of clustering solutions. Clusters are initially defined from the clustering solutions by examining the structure of the dendrogram for breaks or gaps in the fusion of entities and/or by specifying a particular level of similarity as a cutoff point for cluster membership. There are, however, no explicit rules for selection of this cut-off value. Decisions made regarding evaluation and interpretation of clustering solutions are left up to the discretion of the researcher.
81




Procedures
The statistical package CLUSTAN IC (Wishart 1978) was used to carry out the cluster analyses in the present study. The advantage of CLUSTAN IC over other clustering packages is that it is the most versatile package in terms of the hierarchical agglomerative methods offered (Aldenderfer 1977). The analyses proceeded by analyzing the matrices of Gower's similarity coefficients computed for each of the three paste subsamples (see discussion in Chapter 3, pp. 42-45). CLUSTAN procedure HIERARCHY was used to analyze each similarity matrix in order to carry out the clustering. The matrices were input and 'read' for this purpose by CLUSTAN procedure DISTIN.
The type of linkage form or sorting strategy used in an analysis of this kind determines the way in which entities are joined to form clusters. In this study, four linkage forms were used with procedure HIERARCHY for each paste subsample: single linkage; complete linkage; average linkage; and Ward's method. These linkage forms are discussed below. More detailed explanations may be found in Aldenderfer (1977), Sneath and Sokal (1973), and Everitt (1974).
The single linkage or 'nearest neighbor' method
(Sneath 1957) proceeds by joining the two most similar entities to form a cluster; an entity may subsequently join a cluster if it has a certain level of similarity with at least one member of that cluster. The distance between two clusters is equal to the distance between the two closest members. For this reason, clustering solutions produced by this linkage form are often long and straggly (described as 'chaining') with few clear breaks in the dendrogram.
The complete linkage or 'furthest neighbor' method (Sorensen 1948) joins an entity to a cluster if it is within a specified level of similarity with all the members of a cluster; the distance between two clusters is described by the distance between the two furthest members.
Average linkage (UPGMA [unweighted pair-group
method using metric averages), Sokal and Michener 1958) clustering proceeds by computing the average of similarity between an entity and each member of a cluster; this average acts as the computed level of similarity which must be achieved by an entity in order to join the cluster. The distance between clusters is described as the average of distances between all pairs of individuals in the two groups.
82