Group Title: social and chronological dimensions of village occupation at a North Florida Weeden Island period site /
Title: The social and chronological dimensions of village occupation at a North Florida Weeden Island period site /
Full Citation
Permanent Link:
 Material Information
Title: The social and chronological dimensions of village occupation at a North Florida Weeden Island period site /
Physical Description: ix, 251 leaves : ill. ; 28 cm.
Language: English
Creator: Kohler, Timothy Alan, 1949-
Publication Date: 1978
Copyright Date: 1978
Subject: Indians of North America -- Antiquities -- Florida   ( lcsh )
Weeden Island, Fla   ( lcsh )
Antiquities -- Florida   ( lcsh )
Anthropology thesis Ph. D   ( lcsh )
Dissertations, Academic -- Anthropology -- UF   ( lcsh )
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis--University of Florida.
Bibliography: Bibliography: leaves 240-250.
Statement of Responsibility: by Timothy Alan Kohler.
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00098084
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000074577
oclc - 04695588
notis - AAH9851


This item has the following downloads:

socialchronologi00kohl ( PDF )

Full Text


Tirmothy Aian Kohler




Copyright 1978


Timothy Alan Kohler


Without the enthusiastic field assistance of many students and vol-

unteers the mapping and excavation which made this report possible could

not have been accomplished. Maryjean Arnold, Thomas Chase, Ann Cordell,

Martin Dickinson, William Jayne, Kathie Johnson, Jill Loucks, Jere Moore,

Mimi Saffer, Robin Wright, and Drew Yaros all contributed to the field-

work; Patrick Bennett, Thomas Des Jeans, and Benjamin Simpson returned

for more than one season, making their aid particularly valuable. In

her continued assistance in the laboratory classifying and analyzing

ceramics Robin Wright proved herself a model of patience and thorough-

ness. The line drawings in figures 12 14 were volunteered by Helen

Bates, and figures 11 and 15 by Jill Loucks.

Leon A. McKeithen of Live Oak kindly granted permission for the

investigations on his family property and proved to be an important source

in the location of nearby archeological sites. In our contacts with the

people of the Lake City-Live Oak region we were treated with a great deal

of cordiality. The Osaki family of Lake City, our landlords for one

field season, were particularly helpful and showed an informed interest

in the progress of the work.

During the first two seasons investigations were funded by the Went-

worth Foundation of Clearwater, Florida. Since then work has been funded

largely by the National Science Foundation.

As chairman of my doctoral committee and principal investigator

of the North Florida Archeological Project Jerald T. Milanich has shown me


every kindness and encouraged development of the research along lines

which were of interest to me. If this dissertation is of any merit, how-

ever, it is also due to the many courses and conversations, from which

I have immensely profited, with Charles H. Fairbanks. Dr. Fairbanks un-

wittingly (?) suggested this dissertation topic to me during his Seminar

in Southeastern Archeology two-and-a-half years ago when he stated that

the differences in social status among late prehistoric Southeastern

Indians which were indicated by differential burial treatment ought to

be visible in a village context as well. The other committee members--

John J. Ewel, William H. Sears, and Elizabeth S. Wing--have been of great

assistance both within their specialities as well as in areas which must

have been peripheral to their main interests. At least five other fac-

ulty members have made important contributions to particular aspects of

the research, and although they were not on the committee, patiently made

their time available to me. They are K. C. Ewel, J. Holzer, V. Hetrick,

C. S. Peebles, and P. S. Rice.

Just as the first four years of my graduate career were made more

bearable through the kind intervention of Lydia Deakin, secretary of the

Department of Anthropology, so have these last two years been enlivened

by association with the personnel of the Florida State Museum. Although

I fear he fails to appreciate the subtleties of archeological research,I

would like to thank William R. Maples, head of the museum's Social

Sciences Department, who generously made available the office space and

computer funds which were essential to the analysis.

Finally, and fortunately, no dissertation is written in isolation

from other graduate students. Fellow students who have been particularly

important in clarifying my ideas or in forming my attitudes towards


archeology include R. Crook, S. Cumbaa, N. Honerkamp, R. Marrinan, and

R. Smith. Most of all, however, I would like to thank Jill Loucks, who

has assisted in every phase of the project and without whose support I

would often have become discouraged along the way.


ACNOWLEDGEMENTS ......................... .............................. ii

ABSTRACT ............................................................. viii

ORGANIZATION IN THE SOUTHEAST ............................................ 1
Social Organization and Status ......................................4
Kinship and Residence .............................................. 5
Archeological Approaches ........................................... 8
Moundville and the Southeastern Chiefdom ..........................15
Archeology and Social Organization in the Southwest ...............18
Conclusions ................................... ................... 23
A Test of Certain Assumptions .....................................24
The McKeithen Site ................................................ 32

AND HISTORY OF EXCAVATION ........... ............................ ......35
Previous Research in North Florida ................................ 39
Summer 1976 Field Session ......................................... 42
Analysis of the Initial Probability Sample ........................ 49
Winter 1977 Field Session .......................................... 57
Analysis of the Winter Sample ..................................... 59
Summer 1977 Field Session ....................... .................... 67
Faunal Remains .................................................... 81

MATERIALS ANALYSIS .....................................................88
Ceramics at the McKeithen Site .................................... 92
Discussion ....................................................... 112
Lithics at the McKeithen Site .....................................115
Discussion ....................................................... 125

OF MATERIAL VARIATION .................................................. 130
Selection of Units for Seriation .................................131
Choice of a Method .............................. .................. 132
Principal Components Analysis, Rotation, and Factor Scores .......136
A Test of the Principal Components Seriation Technique ...........139
The Factor Score Seriations ................................ ..... 151
Chronological Implications of the Attribute Analysis ............ 167
Coordinating the Chronologies .....................................174
Discussion .......................... ............. ...... .......... 177
The Gulf Coastal Plain ........................................... 182

OF MATERIAL VARIATION ............................ ........ ......... 185

WEEDEN ISLAND VILLAGE ..................................................224
The Early Phase ..................................................224
The Middle Phase ................................................. 225
The Late Phase ................................................... 227
Some Predictions for Future Research .............................229

APPENDIX: DETAILS ON COMPUTERIZED ANALYSES ............................232
Discriminant Analysis ............................................ 232
Cluster Analysis ............................................. .... 233
Principal Components Seriation ................................... 234
Non-parametric Analyses of Ceramic Attributes Versus Depth .......234
Computerized Isopleth Maps ......................................235
Population Simulations ............................................237

LIST OF REFERENCES .................................................... 240

BIOGRAPHICAL SKETCH ............. ......................................251

Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy



Timothy Alan Kohler

August 1978

Chairman: Jerald T. Milanich
Major Department: Anthropology

The McKeithen site is a multi-mound and village complex of the

Weeden Island period in North Florida. Three seasons of mapping and

sampling excavations in the village area revealed a main occupation dating

from about A.D. 150 to A.D. 750 which was divided into three phases for

analysis. Chronological control for these divisions was provided by a

series of ten radiocarbon dates from the village, a principal components

factor score seriation of the proveniences, and an attribute analysis of

a sample of rim sherds from three areas within the village.

In addition to a basic north-south division of the village area on

the basis of the ratio of ceramics to lithics, three areas in the southern

and eastern portions of the village were distinguished from the rest of

the site on the basis of correlations among relative frequencies of

two categories of non-local ceramics, non-local lithics, and total cer-

amic type diversity. The correlations between these assumed high-status

indicators were highest during the final phase of the village occupation,

from about A.D. 550 to A.D. 750, suggesting that during this phase non-local


trade came more directly under the control of an elite group which re-

sided on and adjacent to Mound A, the largest and probably latest mound

on the site. A population model was formulated using the estimated weight

of ceramics in the midden, the estimated weight of the original vessels

which these sherds represented, and ethnographically-derived estimates

of ceramic breakage rates and size of ceramic inventory per household

unit. This model suggests that if the site was occupied continuously

rather than seasonally the ceramics in the midden could be explained by

a population of no more than 300 to 400 persons. Given this upper bound

for the estimated population of the site indicated by the model and the

emerging division of superordinate and subordinate social groups indicated

by the distributions of selected ceramic and lithic categories, it is

hypothesized that the site was initially occupied by a tribal group (a

primarily egalitarian society) with the emergence of a "Big Man" or

proto-chiefdom level of organization by the end of the occupation. Whether

this increasing centralization of authority was caused by, or resulted in,

increased control over external trade remains an important question for

research. Finally, given this hypothesized model of social organization

and change,several predictions are ventured for the results of a site

survey of the support area currently in progress.

/ Chairman


The aboriginal political system which greeted the earliest Spanish

and French explorers of sixteenth century "Florida"--as large portions

of the Southeast United States were then known--was not so dissimilar

from the explorers' Medieval roots that they were unable to interpret it

in terms of the European system (Service 1975:81-83). In the area of

present Northeastern Florida the first French chronicles treated with

"Roy" and "Royne" the Timucua leaders and their first wives; and with

"leurs sujets," the other Indians. By ties of kinship or friendship,

reported the French, as many as forty of these "kings" served as vassals

to another, more powerful lord, for example the "tres redoute Olata Ouae

Outina." Among other privileges it was the prerogative of these caciques

and their mico (as the Spanish occasionally differentiated these two

levels of political leadership) to retain more than one wife, to assemble

and direct council from a seat more elevated than the others, to convoke

groups of workers for planting and harvesting, and to conduct hostilities

against neighboring groups. The death of one of these leaders began

several days and nights of mourning, and all the chief's prized possessions

were burned, along with his dwelling. So much were the caciques the of-

ficial spokesmen for their people that their villages and the surrounding

areas took on their names. Differentiated from this "king," but of

sufficient distinction to address him in council, were the "iaruars,

c'est a dire leurs prestres et les plus anciens" (summarized from Laudon-

niere in Lussagnet 1958).

While some form of status relationships seems to be common to all

cultures, hierarchical ranking as noted in these and other reports on

the Timucua represent a common pattern of aboriginal sociopolitical

organization in the Southeast at the time of contact. Certainly more

and less elaborate systems existed, as in the case.of the highly cen-

tralized Natchez organization (Swanton 1911:100-108) or the relatively

homogeneous bands seen by the DeSoto expedition on the western margins

of the Southeast. Yet the Timucua form--village leadership centered

in one individual, usually male, with the villages integrated into a

more or less unsteady alliance under the leadership of a more powerful

ruler in a central village--seems to have been the prevalent form of

political organization in the Southeast at the time of earliest Euro-

pean contact (Hudson 1976:202-213). This level of political organiza-

tion has come to be called the chiefdom, which Service (1962:173,144)

delimits as follows:

A chiefdom is largely familistic but not
egalitarian; it has no government but does
have authority and centralized direction;
there is no private property in resources
or entrepreneurial market commerce, yet
there are rank differences but no clear
socio-economic or political classes. .
the rise of the chiefdom seems to have been
related to a total environmental situation
which was selective for specialization in
production and redistribution of produce from
a controlling center. . Chiefdoms are re-
distributional societies with a permanent cen-
tral agency of coordination.

In defining the chiefdom level of organization R, N. Adams (1975:

228, 231) stresses the primacy of total group size in the evolution of

chiefly societies:

The patron-client relationship appears between
the band and the chiefdom levels of integration.
It is the principal process whereby individuals
begin to concentrate power independent of the
allocated power granted to them by their fel-
lows.... The successful patron is he who succeeds
in accumulating more so that he may give more
away and thereby gain more clients. It is the
size of the total client body that is the fundam-
ental basis of power, since the basic source of
power for the ranking individual is still allo-
cation.... The building of support from a range
of supporters means getting tribute in kind, ac-
cumulating this, and then providing periodic re-
distributional feasts.

The chiefdom as a political phenomenon represents a first step

towards the institutionalization of power from the allocated, charis-

matic control practiced by tribal leaders and "Big Men" leading eventu-

ally, in an evolutionary framework, to the dogmatic centralization of

the primitive state. Far from implying that the chiefdom was a half-way

house between two more stable forms of political control, this balance

between allocative and coercive power present in the chiefdom resulted

in an eminently-workable adaptation to the levels of technology, pop-

ulation, and environment in the late prehistoric Southeast as demon-

strated by its widespread occurrence at the time of contact. On the

other hand, the tendency of the chiefdoms to form unstable confeder-

acies (as, for example, that of the seventeenth century Creek) reveals

an inclination towards the formation of identity or coordinative groups

(in the sense of Adams 1975:209) going beyond the political boundaries

of the central chief.

Social Organization and Status

An outstanding characteristic of the chiefdom which sets it apart

from the tribe or band is that the power of the leader is inherited by

virtue of the leader's membership in a kinship group (Service 1975:142).

Of the Timucua Laudonniere (in Lussagnet 1958:44) observed that--

Ils se marient chacun a sa femme, et est permis
aux Roys d'en avoir deux ou trois, toutesfois
i1 n'y a que la premiere honoree et recogneue
pour Royne, et n'y a aussi que les enfans de
ceste premiere qui heritent du bien et de l'autor-
ite du pere.

Swanton (1946:221-222) cites examples of such ascribed authority among

the Chitimacha, Natchez, Creek, Chickasaw, coastal Algonquians and the

piedmont eastern Siouans. While the most extreme examples are surely

found among the Natchez and other lower Mississippi Valley groups, the

Timucua provide a more typical example of a Southeastern chiefdom in


The movement from achieved status, which structures social relation-

ships in pre-chiefdom societies, to the presence of both achieved and

ascribed status in the chiefly society is potentially observable in arch-

eological context. Status carried both symbolic identifications and mat-

erial rewards. Some symbols of rank or achievement included tatoos, location

of seat in council, and respect due an individual, while the material re-

wards of the ascribed statuses included possession of exotic clothing items

such as the "Marten" or "Sable" robes reported by Elvas and Biedma (in

Swanton 1946:440); feather mantles of duck down made by Natchez women

for women of the Honored class (Le Page du Pratz in Swanton 1911:63);

and the crowns of swan feathers which, according to the same source,


could be worn exclusively by the sovereign. In reference to the Yuchi

Speck states that fans of wild turkey feathers "were the proper posses-

sion of...old men and chiefs who spend much of their time in leisure....

During ceremonies to carry the fan is a sign of leadership" (in Swanton

1946:456). Other sorts of ornamentation, while not the exclusive per-

quisites of the chief, seemed to increase in quantity and elaboration

among high status groups. Included in this category are beads, pearls,

copper gorgets, and gold or silver ornaments. It appears that objects

which were "expensive" either by dint of being manufactured from a scarce

natural resource, a non-local one, or through a great deal of'effort;

were monopolized by high status individuals. Moreover, symbols of rank

which did not fall into these categories and would, theoretically, have

been equally available to all were controlled by strong cultural sanc-

tions. Thus Adair (in Hudson 1976:203) says that bearers of unearned

tatoos would be forced to remove them, and any taking a seat in council

above his rank would be the object of public derision.

Kinshio and Residence

in the overwhelming number of ethnohistorically-known groups in

the Southeast, membership in the organized descent group was reckoned

through the mother's line (Hudson 1976:185). A matrilineage thus con-

sisted of a group of people who could trace their mutual relationship

through a common female ancestor. A larger identity group based on the

same rule of descent was in operation in the Southeast wherever ecological

conditions and level of technology permitted large, stable aggregations of

population in a localized area: this was the matrisib, an extension of

the matrilineage to include those who believed they were descended

from a common ancestress but who could not trace the actual geneo-

logical ties (Hudson 1976:191). (Some authors use clan or matriclan

in this connection, but since these terms may connote compromise kin

groups based on both a rule of residence and a rule of descent--l f.

Murdock 1949:66-67.]--the term sib has been adopted here.) Both the

lineage and the sib were exogamous groups, but while lineages were

localized units within a village, sib membership cut across town affil-

iations and served as an identity unit for particular population seg-

ments over much larger areas. It is consistent with the expectation

that status be inherited in a chiefdom that certain of these sibs

were traditionally regarded as more prestigious than the others; among

the Creek, for example, the Wind, Beaver, Bear, and Bird sibs were

"recognized as leaders in the establishment and maintenance of peace

in the nation" (Swanton 1928:113). Among the Natchez the Great Sun

was chosen from a particular sib and probably from a particular lineage

within that sib (Hudson 1976:207).

Most references to post-marital residence customs are unfortunately

rather late, but seem to indicate a general uxorilocal pattern. Female

ownership of the houses (or possibly ownership by the woman's sib) is

mentioned among the Georgia coastal Guale by Father Ore (in Larson 1977)

and might be interpreted as indirect evidence of uxorilocality. The

following synthetic sketch of Creek practices in Swanton (1928:170)

reveals a somewhat similar pattern:

The Creek towns in the times we begin to have
knowledge of them consisted of a succession of
villages or neighborhoods scattered through
the woods and along the streams, and connected
by a network of trails. The unit of such a
town consisted of a group of houses owned by
women of one clan and occupied by themselves,
their husbands, and their young children.
In practice, it worked out something like this:
a man, assisted by other members of his family
or clan, might build a house in a new situation
and clear the usual yard by hoeing up the sur-
face weeds and grass for a considerable space
around it. Now, when one of his daughters
married her husband, drawn from some other,
perhaps distant, locality, he would build an-
other house on part of the same cleared space
or in the immediate neighborhood where the
couple would set up housekeeping. As his other
girls married this process was repeated. When
the children grew up the girls would continue
to occupy the ancestral dwellings, or others
erected for them in the neighborhood, while
the boys would marry elsewhere.

Small sibs were often linked together into larger units or phratries

which operated in much the same manner as sibs. Some of these linkages

seem to have been habitual, as the Bear and Wolf were linked in 15 towns

listed by Swanton (1928:129) while other associations varied from town

to town.

Among at least the Creek, Choctaw, and Natchez the moiety offered a

further organizing principle. This division operated on both the town

level and the regional level, dividing sibs within a town into "red" and

"white," and towns themselves into one of the two possible segments.

This division is poorly understood, but among the Creek was thought by

Swanton (1928:275) to be the historical remnant of an established group

(the "Whites" whose towns are considered the oldest) merging with a

foreign group (the "Tcilokogalgi,'.' or "people of a different speech").

The minimal function of this dichotomy on the inter-town level was to

organize sides for the ball game; on the intra-town level the two moi-

eties also served as opposing teams for practice games (Swanton 1928:165).

An interpretation of the functional significance of these games as

regional regulatory devices in place of warfare is reinforced by the

traditional identification of the White moiety with peace, and the Red

with war. There is some slight evidence that the moieties may have had

an additional exogamous function largely lost during the historic period

(Swanton 1928:165). The identification of a town with a particular

moiety seems to have been dependent on the fortunes of the ball games

while the identification of a sib with a moiety was presumably permanent

(C.H. Fairbanks, personal communication). Hudson suggests that the

division originated on the intra-town level as an additional marriage

structuring device beyond the sib, and that as such it may be quite old


Archeological Approaches

The work of William H. Sears (1954, 1956, 1961, 1968, and 1973 is a

partial list) stands as one of the earliest attempts to focus attention

on the relationships between archeologically-visible patterns and socio-

political organization of archeological cultures. Without labelling the

type of sociopolitical organization indicated, Sears points to evidence

that the "strongly class-oriented social organization of the seventeenth-

century Natchez" was a common characteristic of a Gulf coastal plain

culture area extending from Tampa Bay in Florida west to East Texas

(1954:339,343). Far from being a post-contact phenomenon this type of

social organization, states Sears, seems to have had its roots in the

Hopewellian period, if not earlier, and was particularly influenced

by a more or less constant contact with Mesoamerican cultures and others

of the circum-Caribbean area. By about A.D. 1100 1300 "it was the pre-

vailing mode of social organization of. .the lower Gulf Coastal Plain."

This was, of course, not the only sociopolitical organization in the

Southeast at contact time, and in fact stands in rather strong contrast

to the "more democratic and more strongly kin-oriented systems recorded

for groups in the hinterland, such as the upper, Muskogee-speaking Creeks

and the various Siouan-speaking groups" (Sears 1954:343).

Since Sears' arguments in favor of the existence of hierarchically

organized pre-Contact aboriginal societies have been used extensively

in his more recent writings as well as by other archeologists, it is

worthwhile to list in detail what he considered to be the archeological

evidence for such stratification:

1. Special ornaments and retainer sacrifice for elite burials.

2. Deposition of trophy skulls and long bone bundles in burial


3. Mass deposit of special pottery, most made specifically for

mortuary purposes.

4. Use of litters by the elite class which were sometimes included

in the mounds.

5. Use of pole platforms or scaffolds in burials.

6. Reproduction of the temple mound in miniature as an early

construction stage in the burial mound.

7. Possible sacrifice of wives.

8. Spatial patterns of burial placement interpretable as in-

dicative of rank differences.

9. Variation of burial types.

10. Dwelling places on mounds for the elite.

11. Breakdown of organization on a clan or familial basis.

12. Multi-lineage political units of some size as well as inher-

itance of political, or politico-religious, offices by direct

inheritance along "class" lines restricted to small, familial


The salient feature of this list (adopted from Sears 1954) is that

nearly all of Sears' proposed archeological correlates of ranked-societies

are visible only in a mortuary context. One of the few distinguishing

features possibly visible in a village context is the breakdown, "at

least in part, of organization on a clan or familial basis" in the

movement from a segmentary to a stratified society (Sears 1954:343).

However the political organization and inheritance of power among

the most extreme examples of stratification in the Southeastern ethno-

historic record, the Natchez, were structured entirely along kinship

lines, although there is interestingly no evidence for totemic sibs

until after the remnants settled among the Creek and Cherokee (Swanton

1911:107-108). Rather than representing the breakdown of one set of

organizational principles and the substitution of a different set, the

appearance of more rigid social stratification seems to add another

structural dimension to those already in operation.

Several of the suggested correlates are not general character-

istics of social stratification, but rather the realization of general pro-

cesses at particular sites. In a later paper Sears (1961) considers

in more detail the classes of data.which are amenable to interpreta-

tion in terms of social and religious organization. The five promising

classes of data he proposes are settlement patterns, ceremonial structures,

burials and grave goods, specialization in artifact manufacture, and

artistic representation. As the mortuary data are essentially that

considered above we may turn to the other data categories.

Within the broad topic of settlement patterns, which includes

aspects of locational adjustment to the environment, Sears focuses atten-

tion on what he terms the "community pattern," or the "strictly social"

aspects of settlement patterning. This includes both intra-site pat-

terning and placement of sites in relation to each other:

houses and room size, room plan, type and
placement of special-purpose structures in
the site, over-all site plan, and areal
settlement patterns are all classes of evi-
dence that have been interpreted in terms of
kinship structure, social organization, and
religious and political organization. (Sears

Under the category of ceremonial structures Sears includes all

types of mounds and other earthworks, plazas, kivas, and ball courts as

possibly amenable to socio-political interpretation. Presence of fort-

ifications might also be mentioned in this category.

Specialization in artifact manufacture can be inferred, states

Sears,. from the "quality of workmanship, indicating complete mastery of

the craft." Such evidence is significant since "the extent to which

a society can afford to maintain specialized craftsmen is a reflec-

tion of its wealth and organization" (1961:22). Cited examples of

artifact classes which demonstrate this degree of mastery include the

specialized effigy forms and related wares of the Weeden Island cultures

and the effigy pipes and copper ornaments of the Illinois and Ohio

Hopewell cultures. Unfortunately, as Fairbanks (1961:238) points out,

any artistic evaluation as to whether an artifact was or was not of a

quality which could only have been produced by a full-or part-time

specialist is highly subjective; more to the point might be compositional

analyses such as neutron activation to discover artifacts which were

widely-distributed from a single source, implying specialization of a

site within a regional trade network with a possible concomitant spe-

cialization in production by particular Individuals within the site.

Finally, Sears urges that besides serving as evidence for a class

of artisans, each of the items so crafted ought to be thoroughly con-

sidered from an ethnohistoric perspective for possible analogies enabling

reconstruction of the ceremonial role played by the specialized produc-

tions. This interesting avenue of research has recently been pursued by

Flannery and Marcus (1976) in a study of Zapotec cosmology.

In a 1968 article Sears concentrates on the settlement pattern

aspects of sociopolitical organization and defines three "levels of

integration" visible in the Southeastern archeological record. The most

homogeneous of these Is the village community, where no specialized

subgroup controlling social action was present, the villages were all

about the same size, and many have their own minor ceremonial structures.

Suggested examples of this level of integration are the Deptford culture

sites in the Southeast. This reflects what would be termed a tribal or

segmentary organization by many anthropologists.

Next on a scale of increasing organizational complexity is Sears'

"priest state" which has been more-commonly termed the chiefdom level

of organization as described above. The settlement pattern correlated

with this type of organization contains a number of very similar com-

munities tied to a central ceremonial center which is a physically-

distinct community inhabited by the top levels of the sociopolitical hier-

archy as well as-the subordinate social elements. Cited examples in

the Southeastern United States are Etowah, Marksville, Mandeville, and

possibly Crystal River. These sites serve as the organizational centers

for the surrounding area; here "the ceremonialism reaches its quantita-

tive and qualitative peaks" (Sears 1968:140). The priesthood constitutes

the major portions of an "upper class or caste," representatives of which

are also found at the larger villages "where, as the local leaders of

the religion, or state cult, they control ritual and cultural patterns,

in keeping with the standards of the state cult" (1968:140).

On a third more complex level of organization Sears differentiates

between militaristic states featuring "conquest and replacement" and

those proceeding by "conquest and incorporation." Although archeolog-

ical correlates of these types of states are admittedly similar, Sears

classes societies such as those indicated for Moundville and Cahokia

with conquest and replacement states, where

urban centers for sacred and secular control
are present, as well as the residences of the
specialists who can be supported by the same
expanding economy which permits, and perhaps
causes, conquest and expansion. Similarly,
there are other centers, at a distance from the
capital, that serve these same functions in the
provinces. At the frontiers there are small,
scattered communities and a few larger fortified
settlements. (Sears 1968:147)

A warrior class is necessarily present here, and fortifications are

an obvious correlate.

Sears has divided the continuum of cultural elaboration into some-

what different compartments than most Southeastern archeologists, and it

is not the purpose of this study to quibble with the proposition that a

state level of organization was reached by selected cultures in late

pre-Contact times, as Sears suggests. Whether or not such a level of

integration was present it is probable that population losses following

European-introduced diseases (Swanton 1911:39-45) so disrupted the

aboriginal socio-political systems that maintenance of the complex organiza-

tions soon became impossible.

That the ceramics found in Hopewellian mounds and villages and

Weeden Island mounds and villages are not identical is an important clue

to the status hLerarchies in operation at those sites (Sears 1973). The

difference between the ceramics in the two contexts, which Sears formal-

izes as the sacred-secular dichotomy, varies from period to period in

the Southeast, starting from an absolute differentiation in the late

Deptford period ("Yent Complex") to little or no difference in the suc-

ceeding Santa Rosa-Swift Creek period (or Green Point complex). Dur-

ing the Weeden Island period the division seems to have been rather

strong, as it was in most Mississippian period sites.

As a particularly acute example of the sacred-secular dichotomy in

operation Sears (1973:33) cites the Crystal River site on the North

Florida Peninsular Gulf coast. Here excavations by Moore (1903, 1907)

established a sacred complex dominated by Crystal River series ceramics

(finely-executed zoned red, negative painted and incised wares, often

on unusual vessel forms). This complex is in diametric opposition to

that excavated by Bullen (1953) from the village midden areas, where

only 11 decorated sherds were recovered among several thousand Pasco

Plain ceramics, a coarse, limestone-tempered ware generally fashioned

into open bowls. Clearly the relationship between these apparently-

contemporaneous complexes speaks either for strong differentiation of

function in ceramics, or strong social differentiation between the

groups using the two complexes. We shall return to the "sacred-secular"

dichotomy shortly.

Moundville and the Southeastern Chiefdom

Recent work by Peebles (1974) and Peebles and Kus (1977) has exam-

ined the concept of the chiefdom to argue that networks of redistribution

are too general a category to be useful for defining a chiefdom, and

that, in particular, the type of redistribution between different bi-

otic zones often proposed as a causal factor in its development (cf.

Service 1975) is not a constant correlate of the chiefdom (Peebles

and Kus 1977:422-424). The complex chiefdom as it functioned in Hawaii

(Earle 1973 in Peebles and Kus 1977) maintained balanced reciprocal

exchanges of goods between kinsmen with each local unit of dispersed

residence (the ahupua'a) as well as balanced reciprocal exchanges in

unprocessed raw materials and some foodstuffs between groups of ahupua'a

united into districts under local chiefs. While such exchanges of

subsistence items were not processed through the paramount chief (who

had control over several districts) during the period of the major

agricultural festival the paramount chief received tribute of sumptuary

items such as birds' feathers, sweet-smelling woods, and edible

delicacies, some of which were eventually redistributed to other nobles

of his entourage. If the paramount chief was at the heart of a re-

distributive network, then, it was for the control of "elite" items,

not subsistence materials.

While Peebles' and Kus' argument seems more effective as a state-

ment of the probable nature of exchange in a complex chiefdom rather

than as an argument against the possible importance of redistribu-

tive networks of subsistence items controlled by a "Big Man" in a

nascent chiefdom, the authors appropriately proceed to apply the

ethnographic analogy to another probable complex chiefdom, the Mis-

sissippian period site of Moundville. Having removed ecological spec-

ialization and redistribution as necessary correlates of the chiefdom

types of organization,they suggest five other archeologically-visible

indications of the chiefdom, as follow:

1. The presence of both ascribed and achieved dimensions of status

in the mortuary record.

2. A hierarchy of settlement types and sizes.

3. The location of settlements in areas assuring a high degree

of subsistence sufficiency.

4. Activities such as the construction of large monuments which

would require planning and a large labor force; or evidence

of other activities transcending the domestic unit of pro-

duction such as the part-time specialists maintained by Ha-

waiian and Polynesian chiefs for the manufacture of elite (or

sumptuary) goods.

5. Evidence of society-wide (and thus centrally-motivated)

efforts to cope with the least predictable elements of the

social and natural environments. A conceivable example is

the construction of fortifications as a buffer against

warfare. (Summarized from Peebles and Kus 1977:431-433.)

The authors proceed to compare data from Moundville with this

set of theoretical expectations for the chiefly society, concluding

ultimately that Moundville satisfies all the above criteria. Analyses

are reviewed from Peebles (1974) illustrating the presence of 12 major

clusters of burials at the site, divisible into three major segments.

The first and most complex of these appears to represent a superor-

dinate (that is, holding ascribed status) social dimension and is

found in mounds and cemeteries near mounds with associated copper

axes, copper earspools, stone disks, shell beads, oblong copper gorgets,

bear teeth, red or white paint, galena, and accompanying fragmentary

or complete skeletons. The second and third segments are composed

of an apparent subordinate social dimension which is internally dis-

tinguished by a dimension of achieved status, the markers for which

include effigy vessels, animal bone, shall gorgets, freshwater shells,

discoidals, and projectile points. The lowest, and by far the largest,

of the two subordinate segments was buried either with no grave goods

whatsoever, or with bowls, jars, water bottles, or simply sherds.

Burials in this lowest segment are usually found in the cemeteries

near mounds and in the village areas; when in mounds they appear to

represent ritually-sacrificed retainers to superordinate burials.

As further support for these and other mortuary data which seem

to indicate the two requisite independent dimensions of ascribed and

achieved ranking, Peebles and Kus (1977:430-440) report the existence of

an elite residential area containing larger and more complex dwellings than

those in the rest of the village. Predictably, broken portions of the

artifacts interpreted as markers for ascribed status burials occur pre-

dominantly in this same area.

As evidence for organization of production the authors cite a small

activity area set apart from the residences which contained large quan-

tities of finished shell beads, unworked shell, and beadworking tools. An-

other larger area close by yielded numerous large bone awls along with the

stones used to sharpen them and may represent a hide-working area. Both

of these activities were in the same quadrant of the site as the pre-

sumed high-status residence area. On the opposite side of the site were

large fired hearths and caches of raw materials for ceramics which seem

to indicate an area of specialized pottery production. In conjunction

with the low variability in vessel form at Moundville this might be taken

to indicate that most ceramics were not produced by the domestic units.

Archeology and Social Organization in the Southwest

While most of the archeological studies of social organization in

the Southeast--especially the work of Sears--have investigated the na-

ture of the political cadre using mortuary and settlement systems data,

a very different approach to the more explicitly social aspects of or-

ganization has been evolving mainly in the Southwest. Whereas in the

complex societies which characterized the post-A.D. 1 Southeast the most

striking differentiations between individuals in a society were reflec-

tions of the hierarchical status relationships embedded in the society,

in the more egalitarian societies of the Southwest the achieved statuses,

the horizontal role and kinship differences, and a sexual division of

labor became the most archeologically-visible cultural differentiations.

Likewise the frequently obvious architectural configuration of settlement

in the Southwest has lent an apparent security to attempts at identify-

ing ceramic design elements with intra-settlement proveniences.

The often-cited studies of Deetz, Hill, and Longacre share a set of

assumptions which has been more precisely stated by Schiffer (1976:24)

than by any of the pioneers of social group residence area studies.

Without preserving Schiffer's categories, I have paraphrased his Table

2.1 below, with important additions or modifications in parentheses:

I. Assumptions of the uxorilocality inference

A. The preferred rule or custom of post-marital residence was

indeed uxorilocality.

B. Women make the pottery.

C. There is transmission of style motifs from mother to


D. The uxorilocal groups are localized within a village

(and these locations remain stable over time).

E. Ceramics made in a residential area are used in that

residential area.

F. Pottery in use in a residential unit will, when broken or

abandoned, be deposited in the vicinity of that unit.

G. (Ceramic characteristics such as surface finish, design,

and paste are not differentially-preserved in different

residential areas of the site.)

H. There is insufficient post-occupational disturbance to

significantly alter the patterns as originally deposited.

I. (The ceramics excavated by the archeologist constitute a

representative sample of those present in the record.)

J. (The archeologist is looking for, can separate, and can

consistently sort for the differences thus preserved in

different areas of the site.)

K. (Either

1. there is no change over time in the ceramics made

within a residential unit, or

2. the archeologist can differentiate between differ-

ences in design element distribution due to changes

over time and those due to manufacture of a particu-

lar style within a particular residential unit.)

II. Corollary of the uxorilocality inference

A. If uxorilocal units are equivalent to design units there

will be more sharing of designs within units than between


Deetz' (1965) study of three separate sites of the historic Arikara

of South Dakota utilized, implicitly or explicitly, the above assumptions

to hypothesize the breakdown of matrilineality and uxorilocality in the

face of declining population, a pattern of trade with Europeans which

tended to increase the importance of the males, a shortage of local

building materials necessitating frequent moves, and constant warfare

with the Dakota. Deetz used a form of row-and-column analysis to suggest

that certain design elements, such as attributes of rim form and surface

decoration, tended to become more randomly distributed in relation to

each other over time. (Given the assumptions, a high degree of non-

random association of design elements with other design elements would

be expected in the pre-Contact situation.) Deetz supplied no tests of

significance for the apparently-decreasing correlations between design

elements, and Whallon (in Binford 1968) contends that the distribution

is in fact random. Another critic (Dumond 1977:330-349) grants existence

of assymetries in the data, but argues that a more reasonable interpre-

tation of the increasing evenness in the design attribute distribution

from early to late sites can be found in the specific acculturational

process experienced by the Arikara:

the period of occupation of Medecine Crow
was a time of very drastic population de-
crease, and apparently of the aggregation
of surviving population into composite
villages. Not only would such a process
be expected to disrupt most stable social
patterns, it could be expected to result
in the sudden juxtaposition of people of
different villages who might well repre-
sent mini-gradations of ceramic making
quite above any postulated micro-mini
traditions inhering in individual families
in normal villages. In such cases one
would expect to find an initial clustering
of design attributes indicative of earlier
inter-village differences in pottery modes,
that were later relaxed in the general change
in individual patterns of communication, in-
novation, and learning that would surely re-
sult in new communities. (Dumond 1977:335)

This implies more than a criticism of Deetz' conclusions; Dumond

seems to be questioning what was termed a corollary of the uxorilocal

inference above. Without insisting that any of the assumptions of the

uxorilocal inference are wrong, Dumond denies that the result of the

normal operation of these processes in "small but multi-lineage settle-

ments of more than a few hundred people in face-to-face contact" would

be detectable in light of the vagaries of deposition and recovery.

The impressive amount of discussion also generated by the now-

classic Hill and Longacre studies is as much a tribute to their intrinsic

interest as it is a result of the number of flaws which have been noted

in their assumptions and analyses. Recalculation of the correlation

coefficients by Plog (1976:28-29) on which Longacre (1970) based his

conclusions of association between small, outlying sites and one of two

sites with great kivas suggests that the original calculations were in

error. In a later portion of the same study Longacre attempts to explain

a perceived clustering of design frequencies in the Carter Ranch site

into two groups discovered using a multiple regression analysis. While

the proposed explanation is that of matrilocal residence areas, Dumond

(1977:337) notes that the two areas are also perfectly correlated with

the early and late extremes of relative dating within the pueblo and may

thus represent chronological change rather than social variation. Plog

(1976:31) questions the appropriateness of the multiple regression model

itself on the basis that it results in correlation coefficients which

are "unstable and extremely sensitive to sampling and measurement errors"

when used for finding clusters in a matrix of high-correlated variables.

Nor has Hill's (1970) analysis of Broken K Pueblo escaped criticism.

Dumond (1977:337-344) was unable to replicate the results of Hill's

factor analytic procedures which were used to identify clusters of

pottery types and attributes and, more indirectly, clusters of rooms

from which the ceramics were excavated. Plog's (1976:35-41) examination

of Hill's analysis arrives at different clusters than those proposed by

Hill, while corroborating Hill's general conclusions of non-random

distribution of the design elements.


The archeology of social systems as it is applied to the archeolog-

ical record in North America is monolithic neither in its goals nor

its methods, and cannot be termed, as we have seen, an unqualified success.

Due to the relative poverty of chronological control in the Eastern

United States applications have centered on discovering the broad out-

lines of socio-political organization; these studies typically depend

heavily on mortuary data and area settlement systems. The assumptions

behind these studies thus include an adequate representation of the

total community in the mortuary record; an adequate control of the

chronological relationships among the sites investigated; and the correct

imputation, on the part of the archeologist, of the relative value of

the artifacts being used as indicative of status. Southwestern archeol-

ogy, on the other hand, in its recent focus on the identification of

kinship-based residence units within sites, has been forced to make an

additional set of rather restrictive assumptions which include, minimally,

those listed under "assumptions of the uxorilocality inference" above.

The most fundamental criticisms of these works question whether

these assumptions ought to be granted. A discussion of the assumptions

underlying the hypotheses is essential, as Is the presentation of justi-

fication for their acceptance. Another category of criticism lies in

the appropriateness of the particular statistical model used to demon-

strate assymetrles in the data, or in the proof that these assymetries

are significant. By now it should be obvious that the researcher is

obligated both to present as much of the basic data as possible, and to

detail the underlying assumptions and limitations of the statistical

model employed, as well as any non-standard methods used in the applica-

tion or calculation of the statistics. Finally, and probably most

difficult, a plausible connection must be demonstrated between the model

(or hypothesis) which has been proposed and the assymetries which have

been shown to exist. This is the most difficult phase of the argument

because additional alternate hypotheses to explain perceived phenomena

can always be proposed, and to the extent that these have some prior

probability of being true, they must be disproved by "deducing" from

them Implications which can be shown to be false. Unfortunately informed

investigators can disagree about the prior probability of a proposition,

what constitutes a truth-retaining Implication of a true hypothesis, and

what varieties and strengths of proofs are required to show an implica-

tion to be false.

A Test of Certain Assumptions

The approaches considered above all have in common that, as Plog

(1976:11) points out for the Longacre and Hill studies, the model derived

from ethnographic data has not been directly tested by archeologists;

rather "they have. .interpreted archeological data given the assump-

tion that the model is a valid one." A refreshing exception to this

circularity of reasoning can be found in J. S. Otto's (1975) attempt

to prove some of the assumptions habitually made by prehistoric archeol-

ogists using data from a well-documented historic context.

On Cannon's Point, St. Simons Island, Georgia, Otto excavated

portions of several slave cabins, an overseer's house, and the refuse

associated with the main planter's house from identical chronological

contexts. By evaluating the material and subsistence remains recovered

Otto was able to identify categories of material remains which best

differentiated between the three units, which are known from histor-

ical documents to have been occupied by individuals of different

ethnic, social, and economic statuses (Otto 1975:7-16). While the econ-

omic and cultural setting of the Couper Plantation limits its analogic

value for aboriginal North Florida, Otto's reflections concerning elements

shared by all stratified societies are of interest:

In stratified societies, status positions
associated with social roles or activities
are ranked in hierarchies. Upper status
individuals enjoy greater prestige and have
preferred access to the resources of the
natural and social environments. People
occupying lower status positions have less
prestige and suffer impaired access to re-
sources. . Some members of a stratified
community are relatively affluent, though
others live in relative poverty. (Otto

Otto clearly demonstrates that--at least in the context of the 19th

century plantation--it is clearly within the power of the archeologist

to differentiate status relationships on the basis of material remains.

The best indicator of difference between the free white laborers--the

overseers--and the black slaves was the housing conditions, with both

planter and overseer components far superior to those of the slaves in

terms of "available living space; number of specialized rooms; the

features available to occupants; the quality of construction materials;

and expected durability" (Otto 1975:360). Most non-ceramic artifacts

did not seem to clearly differentiate between the three groups; excep-

tions were certain bone and iron button types and pipes, all of which

appeared more frequently in the two lower status groups. Faunal remains

also reflect the status differences with the range of vertebrates in-

creasing along with social status.

For our purposes, however, the most Important category of status

indicators identified by Otto are the ceramic materials. He concludes

that the range of ceramic types increases with the status represented by

the refuse component, and notes that certain ceramic types are especially

good indicators of social status. In particular, the distribution of

banded and transfer-printed wares had a high negative correlation with

the banded wares characteristic of the slave and overseer sites and the

transfer-printed wares characteristic of the elite plantation owners.

One of the main factors conditioning this difference seems to have been

the different sources of supply available to the different social seg-

ments (Otto 1975: 187). The planter's kitchen was stocked with ceramics

supplied by a factor, who obtained his goods from Europe via New York

packet lines. The slaves and overseers, on the other hand, may have

acquired some of their ceramics from local shopkeepers who stocked the

more traditional, "folk" pattern banded wares, possibly as a response

to their customers own preferences; or may have been supplied with a lim-

ited range of inexpensive, utilitarian crockery by the planter.

Another explanation for the higher frequency of banded wares in the

lower status occupations lies in the different vessel forms and functions

represented by the banded versus the transfer-printed wares (Otto 1975:


At the slave cabin, 44% of the total table-
ware items were serving bowls, and 24% of
the tableware at the overseer's house were
serving bowls. . In contrast, only 8%
of the identified tableware items at the
planter's kitchen were serving bowls. ..
Though transfer-printed wares at the plan-
tation appeared in table, tea, and cham-
ber shapes, virtually all of the banded
ware shapes at the three sites were serving
bowls--the "common bowl" shape with foot
rings and carinated, flaring sides. . .
The high frequency of serving bowls at the
slave and overseer sites may be related to
dietary differences which existed between
the planter and subordinate classes in the

Zooarcheology and documentation confirm that the varied fare em-

anating from the famous Couper kitchen was in strong contrast to the trad-

itional "one-pot" meals of the lower status groups. Coupled with the great-

er access of the elite group to exotic, expensive forms and the greater

ability of the elite to afford specialized items with limited gener-

al utility, the more varied diet of the elite group contributed to the

wider range of vessel forms present in its refuse. The differential

decoration on different vessel forms, in turn, contributed to the differ-

ential distribution of ceramic types in the different midden contexts.

Otto then proceeds to make the potentially important generalization

that the ceramics in the higher-status contexts were more "diverse" in

terms of type classification and form (Otto 1975:161, 219). Otto's

Table 11 (pp. 175-176) to which the reader is referred in support of

the claim for differential ceramic diversity in type classification,

reveals the following:

Slave Cabin Overseer Planter's Kitchen

n of types 23 24 28

sample size 543 179 1242

The planter's kitchen refuse indeed contained a greater range of types

but this should not necessarily be considered a greater diversity, since

a measure of diversity must compensate for sample size. One such measure

used in ecological studies is simply number of species/1000 individuals;

another, which measures both diversity and evenness, is the Shannon-

Weaver diversity index, or information index, often called H (Odum

1971:144). These two indices were calculated for the data Otto reports

in Table 11 with the following results:

Slave Cabin Overseer Planter's Kitchen

n of types/1000 sherds .04 .13 .02

H 2.39 2.47 1.40

By either of these measures the planter's refuse is the least, not the

most, diverse, due to the large amounts of transfer-printed pearlware in

the kitchen refuse. It is the overseer, who is participating in elements

of both the elite and the folk culture, who leaves the most diverse refuse.

Similar diversity indices based on vessel form were also calculated

from Tables 19-21 (Otto 1975:205-217) with these results:

Slave Cabin Overseer Planter's Kitchen

sample size 126 128 309

n of forms/1000 sherds .2 .14 .08

i 1.9 1.87 2.08

Since the largest ceramic type category among the planters is found on

vessels of many different forms, while the many types among the slaves

and overseers are most often in the bowl form, the high-status component

at Cannon's Point is clearly more diverse on the basis of the H index

when ceramic diversity is computed on the basis of vessel form, as was

predicted by Otto.

In a sense it is unfortunate that Otto was unable to demonstrate

greater ceramic type diversity from the high-status occupation areas

on Cannon's Point since this is precisely what one would anticipate for

high-status residence localities within a pre-Contact Southeastern

chiefdom. Let us examine the reasons behind the low ceramic type diver-

sity of the planter's refuse and the hypothesized high diversity in the

high-status occupation area of the chiefdom.

The first and most significant difference relates to the way in

which ceramics were manufactured and acquired in the Southeastern abor-

iginal setting versus the modern economic market system in which the

planter participated. According to Otto's 1977 reconstruction the

planter regularly purchased ceramics for the big house as well as for

the overseer and slaves. This resulted in the distribution of a

special class of wares to these segments which was probably more limited

in range of decoration and form than the large sets of matching table

and teaware purchased by the planter for his own use. However, small

portions of the planter's ceramics were apparently laterally-recycled

through both the overseer and slave components; while the slaves also

apparently received some discards from the overseers. Moreover, Otto's

earlier analysis (1975:160-173) suggests that both the slaves and over-

seers may have purchased some of their own ceramics, probably purchasing

individual items as they were able from local shopkeepers.

In a chlefdom, as we have seen, the situation is entirely different.

With the possible exception of the archeological data from Moundville,

the archeological record and the ethnohistoric documents suggest that at

least utilitarian pottery was a household craft. Most ceramics were

probably produced domestically by the women of the household in the

variety of forms necessary to fulfill the functions to which they would

be put. Several references to the extreme skill of the Natchez and

Tunica potters (DuPratz and Dumont in Swanton 1946:549-550) suggest that

they were capable of either utilitarian or more recherche productions.

Le Page du Pratz, for example, commissioned of the Natchez potters a set

of dishes and plates after the model of his French earthenware, and was

well pleased with the "beautiful red color" of the result (Swanton

1946:549). As the focal point for a system of reciprocity based on (at

a minimum) sumptuary items, the elite group, and the chief especially,

controlled the flow of a large amount of goods. Some of these only

passed through his hands, since strikingly unequal accumulations of

wealth are not consistent with a chiefdom-level of organization. Some

were lost from the village midden to the mounds through mortuary rituals.

Over a certain period of time, however, it would appear inevitable that

the chief would amass, and eventually break or lose, a more diverse lot

of ceramics than those of a lower status, who normally had access only

to the ceramics they themselves produced. Moreover, a greater percentage

of the ceramics in the high-status area would be made up of sumptuary-

type items, and items received through non-local trade, than would be

the case in low-status areas.

Analogies from Otto's work in historic planter-slave systems on the

Georgia coast to aboriginal cultures in North Florida must be drawn not

through the specifics of either system but though the generalities true

of both. In both situations, refuse of high-status components can be

expected to evidence a greater percentage of goods which were deemed

exotic. We have already seen that at the time of contact such goods

included objects requiring a great amount of time and effort for their

manufacture; objects made of a scarce or non-local resource; and objects

needed for specialized functions not performed by lower status

individuals--for example, a ritual or other symbolic control functions.

Just as the differential relative frequencies of banded versus transfer-

printed ware provided an index differentiating the refuse of status

groups in Otto's study, ceramic types can be found whose relative fre-

quencies will differentiate the status-group residences at a socially

stratified aboriginal site. In this perspective those ceramics which

Sears identified as "sacred" (1973) because of their high correlation

with ranking individuals in mounds are simply elite goods which were

removed from circulation with the death of the owner to avoid "the

inflation that would result from the accumulation of these valuables in

the system" (Peebles and Kus 1977:443-444). Purely utilitarian items

fulfilling everyday functions--as the banded-ware bowls in the lower

status overseer and slave components--will have a direct counterpart in

the open, undecorated bowl form which can be expected to appear in

higher relative percentages in the lower status areas of the aboriginal


The McKeithen Site

The data on which this dissertation draws result from sampling

excavations in the village and plaza area of a multi-mound Weeden Island

period site in Columbia County, Florida. As the first stage in a larger

North Florida archeological project these excavations were meant to lay

the chronological groundwork (in a literal terra incognito) for succeed-

ing intensive excavations in the village areas and mounds of the McKeithen

site and for interpretation of materials gathered from surrounding,

supporting sites. This is done in Chapter Four, after a brief des-

cription of the site and the methods of excavation in Chapter Two,

and reflections on the materials obtained in Chapter Three. Because the

sampling nature of the excavations precluded complete excavation of

entire structures, and because neither the regional survey nor the mor-

tuary data are yet available, some classes of data normally used for

testing hypotheses of social stratification cannot be applied. Far

from being a disadvantage, this provides an opportunity to interpret

distributions of material from within the village area in ways that are

not normally attempted, since it has traditionally been felt that mor-

tuary data, size and complexity of dwellings, and relative size and

distribution of site types are more powerful indicators of social strati-


The major hypothesis to be tested is that the McKeithen site repre-

sents the remains of a socially-stratified society and that these inter-

nal social differences can be recognized from the village remains them-

selves without reference to the ceremonial mounds. The implications of

the hypothesis, which will be tested in Chapter Five, are as follows:

1. The high-status areas will be differentiable from the rest of

the site on the basis of higher concentrations of non-local ceramic

materials. The assumptions of this implication include:

a. The village head and his lineage control whatever non-local

trade is represented by the presence of the exotic ceramics.

b. A larger portion of these non-local ceramics are retained

within the high-status lineage or lineages than are redistrib-

uted randomly to all areas of the site.

2. High-status areas within the site will be differentiable from

the rest of the site by a greater diversity in ceramic vessel form. The

assumptions of this implication include:

a. The village elite have preferential access to non-utilitarian

vessel forms and to ceramics which are more elaborate than

would be required to fulfill their designated function.

b. By virtue of their different roles in the political and religious

life of the community, the village elite possess ceramic vessels

which fulfill functions not required of ceramic vessels belong-

ing to members of subordinate statuses. This might include

items such as large storage vessels or totemistic representations.

3. As a corollary of the first two implications, high-status areas

will be marked by a higher ceramic type diversity than other areas of

the site. This assumes that:

a. The type concept, which is used here in the traditional tax-

onomic manner, is not simply a construct which is imposed on

the aboriginal ceramics, as some archeologists insist (Rouse

1960:318) but rather reflects in some measure a formal and

functional ideal for the culture to which it is applied (cf.

Deetz 1967:43-65).

4. In addition to imported ceramics, materials such as bone,

shell, and lithics which are not readily available locally will be more

abundant in elite residential areas than elsewhere. This requires that

assumptions similar to those on which the first and second implications

are based be granted.

Before this problem can be effectively considered,the data-gathering

mechanisms (Chapter Two) and data-analysis procedures (Chapter Three)

must be set forth, and sufficient temporal control developed (Chapter

Four) so that any patterning observed in the data can be ascribed to

social rather than temporal variation (Chapter Five).


The casual visitor to Florida, in eager anticipation of the beaches

to the south, regards North Florida as a prolonged inconvenience. To

the motorist jaded by the heroic proportions of the Appalachians, then

lulled by the undulating Piedmont, the coastal plain of southern Georgia

and Florida seems flat and featureless. To the more attentive observer,

however, the still-rural landscapes of North Florida harbor a surprising

natural diversity. Perched on formations of limestone and other marine

deposits formed during periods of submersion as early as the Eocene, and

reworked by the rising and falling seas of the Pleistocene, the soils of

Columbia and Suwannee counties--which constitute the heart of the north

Florida region--are derived principally from the Hawthorne Formation, a

parent material rich in sand, clay, marl, limestone, fuller's earth, and

phosphates (Rowland and Powell 1965:94-98.) Towards the north-central

portion of the eastern extreme of Suwannee Country, and the western

portion of Columbia County, thick remnants of this formation underlie

the highest surface elevations in an east-west transect of the state at

the attitude of Wellborn (about 68m at 300 13' N) and form an elevated

ridge oriented north-south in the upper portion of the peninsula (see

Fig. I). Rather than flowing through a series of increasingly-large

streams into lakes or the ocean, much of the water in North Florida

percolates through the porous underlying limestone formations into the


Physiographic Provinces
Southern Georgia,
Source: Cooke 1948:9;

Fig, 1
of the Coastal Plain in Florida,
and Southeastern Alabama
Veach and Stephenson 1911:28

underground freshwater reservoir. But in the vicinity of Wellborn

originate the only significant flowing streams in Suwannee County due to

these thick Hawthorne deposits which are more impervious to percolation

than the Suwannee or Ocala limestones. The principal lakes in the two-

county area can also be found along a narrow belt about 30km long

trending northwest-southeast along a.line drawn between present-day Live

Oak and Lake City.

In the middle of this band of lakes, ideally situated to serve as

the central site for a series of communities heavily dependent on these

flowing and standing water resources, the McKeithen site is itself

bordered on the north by a small stream known as Orange Creek. Until

relatively recent times, when dams were erected upstream from the site,

Orange Creek was a small but active stream, bordered on its south by 2-

3m bluffs, and winding through a floodplain varying from 10 to 30m in

width. The meander belt of the creek and the protected bluffs support

mesic hammock and floodplain vegetation in which Magnolia spp., Carya

spp., Quercus nigra, Ilex opaca, Smilax spp., Vitls spp., and Cornus

spp. are the apparent dominants. Away from the creek, in those areas

which have not been recently cleared, the sandy, well-drained soils of

the Blanton and Lakeland association support a more xeric vegetation

dominated by various oaks and pines, especially Pinus taeda and P.

elliotii, Quercus hemispherica, Q. laurifolia, and Q. virginiana. Once

an important canopy species, Pinus palustris has largely disappeared

over the site due to recent selective lumbering. Given a regime of

regular burning, however, an association of Pinus palustris, Q. cinerea,

and Aristlda stricta could have dominated the site and much of the

surrounding area; little or no burning, on the other hand, would proba-

bly have allowed a mesic hammock vegation to develop. Other species

important on the site today include Prunus spp., Crataegus spp., Dio-

spyros virginiana, Vaccinium spp., Gaylussacia spp, Rubus spp., Rhus

radicans, Gelsemium sempervirens, and Campsis radicans.

The climate in this portion of Florida is warm but temperate, with

freezing temperatures occurring about 15 times a year. The highest mean

monthly rainfall of 190mm occurs in July, while the lowest mean monthly

precipitation of 25mm falls in November. The resultant yearly average,

as measured at a Lake City station, is about 1300mm (Butson 1965:93-95).

This exceeds the average annual potential evapotranspiration (PET) by

about 20%, a comfortable surplus in view of the fact that the months of

highest PET are also the months of greatest precipitation (Thornthwaite


The faunal component of the natural community in the North Florida

area has been somewhat truncated by recent human expansion, but origi-

nally Included the primarily carnivorous Felis concolor and Alligator

mississipiensis, the canids Canis niger and Urocyon cinereoargenteus,

the felid Lynx rufus, and Mustela vison, Ondatra zibethicus, Lutra

canadensis, and a variety of raptorial birds, snakes, and fishes. The

largest variety of species, and the species which certainly played the

most important economic role, were primarily herbivorous, including (to

name only the most prominent) a variety of birds, fishes, snakes, turtles,

and tortoises, the marsupial Didelphis virginiana, and the mammals

Sylvilagus floridanus and S. palustris, Sciurus spp., Procyon lotor,

Ursus americanus, and of course Odocoileus virginianus, the single

most important faunal species for most Southeastern aboriginal sites.

The geographic extent of the North Florida archeological region

has been defined by Milanich (n.d. b:10) as

north of the Santa Fe River and south of the Okee-
fenokee Swamp and wiregrass-pine barrens which are
found in Southeast Georgia . the western bound-
ary is usually placed at the Aucilla River and the
eastern at the beginning of the Atlantic coastal
flatlands, which begin just east of Lake City. The
Gulf coastal flatlands extend up almost to the Su-
wanee River and define the southwestern edge of
the region.

Previous Research in North Florida

As a cultural area, however, the North Florida region is defined

by default: it is that area north of North-central Florida, east of the

St. Johns River valley, northeast of the peninsular Gulf coast, and south

of the southern Georgia coastal plain in which virtually no professional

archeology has been attempted. With the exception of the interior

Georgia coastal plain, the prehistory of the cultural sub-areas surround-

ing North Florida is comparatively well known. The classic references

on the archeology of the northern half of Florida--Goggin (1952) and

Willey (1949)--bracket North Florida on the east and west. Recent short

summaries by Milanich (1976) and Schnell (1975) discuss general chronol-

ogies in North-central Florida and southern Georgia.

Much interest and excavation in all these surrounding areas has

recently focused on the cultural developments occurring during the first

millennium A.D. During this period most Southeastern peoples seem to

have made the transition from relatively egalitarian tribes to more

complex chiefdoms with obvious heirarchical status distinctions. Other

symptoms of this development are larger and more diversified settlements,

networks of trade spanning hundreds of miles, a ceramic art which reaches

a pinnacle of technical achievement, and the first tentative formations

of a pan-Southeastern symbolic system. Coincident with these changes

was the introduction of foreign cultigens, particularly maize, which can

first be documented for the Floridian peninsula by the B.C./A.D. transi-

tion (Sears and Sears 1976).

The Weeden island archeological complex defined on the basis of a

primarily ceramic trait list (Willey 1945) and conventionally dated to

between 800 and 1400 A.D. (Sears 1977) seems to hold a key position in

this development. Encompassing the coastal area and, to a less well-

known extent, the interior coastal plain between Mobile Bay and the

Little Manatee River, the complex is composed of several local cultures

which shared a unifying "sacred" ceramic series, presumably indicating

shared systems of belief and sociopolitical structures. Reflecting the

rise of interest in this period and the desire to address its still-

unresolved chronological problems, a variety of papers and publications

dealing with the sub-regions on all sides of North Florida have recently

appeared. Weeden Island-related cultures in North-central Florida have

recently been reviewed by Milanlch (n.d.a), Smith (1971) and Hemmings

(n.d.). The North Peninsular Gulf Coast area has been discussed by

Kohler (1975). Still-unreconciled reconstructions of the Northwest

Florida chronology can be found in Milanich (1974), Percy (1976) and

Sears (1963). The only extensive modern excavations in a major Weeden

Island site have been reported by Sears (1956) for the Southwest Georgia

area. Most recently Steinen (1976) has extensively reviewed the litera-

ture relating to problems of internal Weeden Island chronology, the

area distributions of the various local Weeden Island cultures, and the

ceramic types diagnostic of the period. This recent expansion of Weeden

Island studies makes a more intensive review of the literature unneces-

sary here.

For the North Florida region, however, two citations suffice to

completely review the published archeological literature, and neither is

particularly relevant to the period under investigation. In the late

1940's Goggin recovered a series of lithic materials he called the Santa

Fe complex from two sites in Alachua and Suwannee counties. These

included probable Paleo-lndian projectile points as well as other tools

which, in retrospect, look as though they may belong to a later tradition

(Goggin 1951). Collections by Goggin from a First Spanish period dump

in the Ichetucknee River, probably associated with the Santa Catalina de

Afuerica mission, yielded aboriginal ceramics of the Leon-Jefferson and

St. Johns series as well as ceramic types typical of the late prehistoric

period in North-central Florida: Alachua Cob Marked, Lochloosa Punc-

tated, and Prairie Cord Marked (Deagan 1972). A continuing series of

surveys and excavations at a nearby mission vista and associated aborig-

inal villages (8 Su 65 and others) should help clarify the nature of the

proto-historic and early historic ceramic traditions in North Florida.

The occupation of the McKeithen site is bracketed--though not very

precisely--by these temporal extremes. Professional interest was first

drawn to the site by the landowner and his son; Leon A. McKeithen Jr. en-

couraged visits by Charles H. Fairbanks and David S. Phelps during the

1960's. The site was recognized as a major Weeden Island center, and

Phelps drafted a contour map which accurately positioned the three ma-

jor mounds now known as A, B, and C. The mounds form an almost perfect

equilateral triangle with the base (defined by Mounds A and B) on an

east-west line. Mound C lies 35m south of the creek bordering the site

on the north. The mounds surround a central area which is nearly devoid

of artifacts.

In June 1976 with funding first from the Wentworth Foundation and

later from the National Science Foundation, proposal #76-82720, Jerald

T. Milanich of the Florida State Museum initiated the North Florida

Archeological Project. The initial thirty weeks of the project, under

the field direction of the author, were devoted to mapping and sampling

the extensive village area at the McKeithen site.

Summer 1976 Field Session

Today most of the eastern portion of the site (see Fig. 2) and

about a third of the western portion are in planted slash pine. These

areas were previously planted in watermelon, and before that--in the

1920's and 1930's--were in open pasture which was repeatedly burned to

halt forest succession. That portion of the site along the creek and to

the northwest which has not recently been cleared is in many places a

thick tangle of secondary growth due to the selective lumbering of


FIG. 2



longleaf and loblolly pines. The first task of the Summer 1976 field

crew was to establish a grid through these natural obstacles over a

roughly square 18 hectare area south of Orange Creek and west of State

Road 135. The grid is oriented 6.50 east of magnetic north, or about 70

east of true north, to follow as closely as possible the lines of planted

pines. North-south and east-west lines cross at 30m intervals resulting
in a cell size of 900m2. This size, while somewhat larger than would

have been ideal, was deemed the smallest practical reference unit for

the initial sampling excavations given the known size of the site.

Although it would have been desirable to make an intensive surface

collection over the entire site, time limitations forced us to settle

for a 33% sample. The 900m2 grid units were numbered consecutively from

1 to 202 and 67 of these were selected (without replacement) by means of

a random numbers table. Since surface cover and degree of surface

disturbance varied greatly among the units, and since no large-scale

attempt to remove leaf litter, pine straw, and vegetation could be

undertaken, some method to compensate for the variety of collection

conditions was necessary. Each surveyor was asked to rank the conditions

for surface collection within the selected units from 1 (good) to 4

(poor) taking into account such factors as the presence of firelane

cuts, amount of exposed ground surface, litter density, etc. Following

a complete collection of all cultural materials seen on the surface of

the selected units this score was multiplied by the number of cultural

items recovered to compute a "value" for each grid unit which would,

hopefully, reflect the Intensity of occupation. Where no artifacts were

recovered in an area scoring 4 it was not deemed that the area was

sterile, but that there was no information; hence the number of units

about which we gained information was 48, or about 24%. The resulting

richness values were collapsed into four ranges (0, 1-5, 6-15, and 16 or

more) which were shaded onto a map of the grid for easy visualization of

the pattern (see Fig. 3). One area of intense occupation along the

creek east of Mound C and arching away from the creek towards Mound A

was obvious, as was another area extending east-west just south of

mounds A and B. Since the richness values dropped off somewhat directly

east of Mound A it was thought these two midden areas might be discon-


Because we were interested in differentiating between the areas

tentatively defined as two separated middens with the third area between

and surrounding them of less interest,a stratified disproportionate

random sampling strategy was chosen. The presumed midden along the

creek was called Stratum 1, the other crescentic midden, Stratum 2, and

the presumed plaza and outlying areas, Stratum 3. Stratifying the

sample in this way helped ensure that all areas of the site would be

adequately sampled. Since we were most interested in distinguishing

between the first two strata--the areas where the most information

applicable to the hypotheses would be preserved--the sampling proportion

in these two areas was .001, while in Stratum 3 the sampled proportion

was only .0003. To achieve this disproportionate sample 10 2 x 2m

squares were excavated in each stratum. The disproportionality of the

sample was based on the research hypotheses, while the total number of

0 D Bh I

Fig. 3
Richness Values Computed from Total Surface Collection
of Randomly Selected 900m2 Grid Units

units excavated was dictated by the time and personnel at hand. Figure 4

outlines the three strata used for this sample as well as the additions

to the grid (strata 4 and 5) and the transect which are discussed later

in the chapter.

To select the actual 2 x 2m units to be excavated, the 900m2 grid

units were renumbered consecutively within each stratum and using a

random number table 10 grid units were selected from each stratum.

Then, on paper, a 900m2 grid unit was divided into 225 2 x 2m squares

which were numbered consecutively. For each 900m2 grid unit which had

previously been selected we chose one 2 x 2m square using the table of

random numbers. These squares were excavated by natural zones broken

into arbitrary 10cm levels where zones exceeded 10cm in depth and the

density of cultural materials made such a segregation desirable. All

material was passed through a V" mesh mechanical shaker screen. All

thirty squares were taken down either to sterile soil or to a level

where cultural material was extremely scarce; all but one unit contained

some cultural material. In several areas the depth of the undisturbed

midden beneath the plow zone or accumulated humic deposits was a meter

or more; the average depth of undisturbed midden was about 0.5m. In

most areas the zone of undisturbed midden displayed no obvious stratifi-

cation and was excavated in arbitrary 10cm levels. While lithics and

ceramics were generally abundant,macroscopic floral and faunal remains

were only occasionally encountered, apparently due to poor preservation

in the acid soil conditions. Nearly all such material was carbonized

calcinedd) and was very fragmentary. At least one wall of every unit

I *1
I ~ / I

II I lin i lli b Il I,

Lmits at probabtity slrata timt .f iraict sampl ---

2 2 m in coveian rnil

Fig. 4
Sampling Strata and Location of Tests in the
Probability and Transect Samples

was profiled and photographed; all features were separately recorded,

mapped, and photographed, and large flotation samples were collected for

further analysis. Thirteen samples of carbonized wood were taken for

radiocarbon dating of which four from this field season were processed

by the University of Miami laboratory (see Table 4).

Analysis of the Initial Probability Sample

After the recovered materials were washed and placed in marked

boxes a tentative classification of the ceramics on the basis of recog-

nized types was begun. The decorated ceramics were numbered and sepa-

rated from the plain during the classification stage, and the lithics

were characterized according to eight broad categories (e.g. heat-

treated coral, worked flint) and worked items were further categorized

into one of 28 analytical classes. For convenience this information was

transferred to IBM cards (three cards per provenience) and then to an

SPSS system file on direct-access disk (Nie et al., 1975). During this

process the identification of each excavation unit with a sampling

stratum was re-evaluated on the basis of the excavated materials (the

original assessment having been based on the surface collections and

spatial locations of the units). Three of the 2 x 2m squares placed in

Stratum 1 yielded much less material than expected, and therefore seemed

to fit better into Stratum 3. For the same reason three units in Stratum

2 seemed to belong in Stratum 3. On the other hand, some areas predicted

to be relatively unproductive on the basis of the surface collections

were re-assigned to the nearest midden stratum after excavations yielded

high densities of material.

Recognizing that this re-assignment of units to strata--and even

the identification of the two midden strata themselves as somehow

distinct--was up to this point intuitive, a discriminant analysis was

performed on the excavation units as characterized by the ceramic types

and lithic categories they contained to see whether this classification

could be justified solely on the basis of the amount and type of material

in each unit, rather than on the basis of the unit's location within the

site. If the classification was valid, the discriminant analysis would

show which variables were most useful in differentiating between the

three strata (see Klecka in Nie et al., 1975, Doran and Hodson 1975:209-

213, or Bennett and Bowers 1976:95-117 for further discussion of the

technique). The first phase of discriminant analysis derived a limited

number of linear discriminantt) functions which summarize the variability

between the redefined midden strata. The variables used were of both

the shape and size variety; that is, it was anticipated that the two

midden strata would differ from each other primarily in the relative

frequencies of the ceramic and lithic categories (shape variables) while

they would both differ from the plaza and outlying stratum on the basis

of raw numbers of artifacts (size variables). Table 1 summarizes the

raw frequencies of each of the redefined strata on which the discriminant

analysis was based (categories having a total site frequency of less

than five items were eliminated). Total lithic and ceramic counts for

each unit were also entered as variables. All levels in each square

were aggregated to characterize the square as an entity.


Category: Stratum 1 Stratum 2 Stratum 3

Pasco Plain 2/ 0.03 5/ 0.09 1/ 0.02
smooth sand tempered plain 653/ 8.94 641/11.05 92/ 2.24
grit tempered plain 49/ 0.67 57/ 0.98 11/ 0.27
residual plain 1227/16.81 686/11.82 180/ 4.39
St. Johns Plain 36/ 0.49 46/ 0.79 11/ 0.27
St. Johns Check Stamped 1/ 0.01 7/ 0.12 2/ 0.05
Papys Bayou Incised 2/ 0.03 O/ 0.00 0/ 0.00
Papys Bayou Punctated 1/ 0.01 19/ 0.33 1/ 0.02
Weeden Island Plain 38/ 0.52 53/ 0.91 3/ 0.07
Weeden Island Red 29/ 0.40 7/ 0.12 5/ 0.12
Weeden Island Zoned Red 1/ 0.01 12/ 0.21 0/ 0.00
residual red 31/ 0.43 10/ 0.17 5/ 0.12
Weeden Island Punctated 19/ 0.26 62/ 1.07 9/ 0.22
Weeden Island Incised 27/ 0.37 27/ 0.47 2/ 0.05
Carrabelle Punctated 42/ 0.58 26/ 0.45 4/ 0.10
unidentified punctated 7/ 0.10 5/ 0.09 0/ 0.00
Carrabelle Incised 22/ 0.30 13/ 0.22 O/ 0.00
St. Petersburg Incised 1/ 0.01 0/ 0.00 0/ 0.00
Keith Incised 10/ 0.14 6/ 0.10 1/ 0.02
Indian Pass Incised 2/ 0.03 1/ 0.02 0/ 0.00
unidentified incised 34/ 0.47 33/ 0.57 7/ 0.17
Swift Creek Complicated Stamped 26/ 0.35 5/ 0.09 2/ 0.05
Crooked River Complicated Stamped 2/ 0.03 3/ 0.05 0/ 0.00
Old Bay Complicated Stamped 1/ 0.01 5/ 0.09 O/ 0.00
Napier Complicated Stamped 4/ 0.06 1/ 0.02 O/ 0.00
Kolomoki Complicated Stamped 3/ 0.04 0/ 0.00 O/ 0.00
St. Andrews Complicated Stamped 2/ 0.03 0/ 0.00 2/ 0.05
unidentified rectilinear 9/ 0.12 6/ 0.10 0/ 0.00
unidentified curvilinear 21/ 0.29 10/ 0.17 4/ 0.10
check stamped 4 6 checks/inch 7/ 0.10 12/ 0.21 0/ 0.00
check stamped 7 9 checks/inch 14/ 0.19 29/ 0.50 5/ 0/12
check stamped 10 + checks/inch 3/ 0.04 11/ 0.19 3/ 0.07
Deptford Linear Check Stamped 2/ 0.03 O/ 0.00 0/ 0.00
Deptford Simple and Cross-Simple 2/ 0.03 0/ 0.00 O/ 0.00
Savannah (?) Cord Marked 0/ 0.00 1/ 0.02 O/ 0.00
Alachua Cob Marked 2/ 0.03 0/ 0.00 O/ 0.00
Tucker Ridge Pinched 2/ 0.03 O/ 0.00 O/ 0.00
fabric marked 2/ 0.03 0/ 0.00 O/ 0.00
other ceramics 10/ 0.14 7/ 0.12 1/ 0.02
worked flint 80/ 1.10 33/ 0.57 4/ 0.10
worked coral 6/ 0.08 5/ 0.09 1/ 0.02
heat-treated flint 513/ 7.03 284/ 4.90 29/ 0.71
heat-treated coral 27/ 0.37 8/ 0.14 3/ 0.07
other flint 1071/14.67 404/ 6.97 68/ 1.66
other coral 86/ 1.18 18/ 0.31 11/ 0.27
other worked lithic 8/ 0.11 2/ 0.03 2/ 0.05
other lithic 20/ 0.27 12/ 0.21 3/ 0.07
mica 1/ 0.01 0/ 0.00 0/ 0.00
Note: data is in raw frequencies/mean per level

In discriminant analysis the number of functions derived for sum-

marizing the dimensionality of the variability between the groups is at

least one less than the predetermined number of groups. Given the three

sampling strata the analysis phase of discriminant analysis defined two

linear functions (similar to the factors of factor analysis) which

summarized the inter-group variation. After the analysis phase, a

classification phase was entered which derived scores for each unit on

each linear function. The analytical program selected in step-wise

fashion the 15 variables which were best able, in combination with each

other,to discriminate between the three groups. These 15 variables and

the order in which they were entered are shown in Table 2. Also shown

is the lambda statistic which inversely expresses the degree of separa-

tion achieved among the groups by the variable included in that step in

combination with any other variables already included in the function.

A low lambda indicates a high degree of separation. The change in Rao's

V (a generalized distance measure) indicates the degree of improvement

in group separation achieved by the addition of the variable entered in

that step. Both the lambda and "change in Rao's V" statistics can be

tested for probability that the resultant value might be due to chance

alone. All values for both statistics listed in Table 2 are significant

at the .01 alpha level.

The standardized discriminant function coefficients for those

variables loading more than +0.30 on either of the derived functions are

also included in Table 2. These are interpreted like the loadings on

factors in factor analysis: the higher the absolute value of the coef-


Step Variable Wilks' Change in
Number Entered Lambda Rao's V

I total lithics 0.49 27.6
2 Carrabelle Punctated (%) 0.31 30.2
3 Old Bay Complicated Stamped (%) 0.22 11.6
4 smooth sand tempered plain (%) 0.17- 10.3
5 unidentified punctated (%) 0.14 10.9
6 Weeden Island Zoned Red (%) 0.10 39.2
7 residual red (%) 0.07 20.7
8 heat-treated unworked flint (%) 0.05 32.0
9 total ceramics 0.03 87.5
10 heat-treated unworked coral (%) 0.02 77.3
11 Crooked River Complicated Stamped (%) 0.01 110.1
12 other coral (not heat-treated, unworked,%)0.01 61.4
13 check stamped 4 6 checks/inch 0.01 56.2
14 Napier Complicated Stamped (%) 0.01 55.7
15 check stamped 7 9 checks/inch (%) 0.01 121.0

Standardized Discriminant Function Coefficients

Variable Function 1 Function 2

Weeden Island Zoned Red (%) 0.36
unidentified punctated (%) -0.55
Napier Complicated Stamped (%) 0.59
heat-treated unworked coral (%) 0.37
total ceramics 0.63 -0.61
total lithics 1.03
residual red (%) -0.34
check stamped 4 6 checks/inch (%) 0.31
heat-treated unworked flint (%) 0.31

ficient, the greater the contribution of that variable to that function;

the sign of the coefficient indicates whether that contribution is

positive or negative. Thus, since scores can be obtained for each unit

on each function, a unit scoring high on Function 1 can be expected to

be high in absolute amount of ceramics and in relative amount of Weeden

Island Zoned Red (both of which make positive contributions to Function

1) but to be low in relative amounts of the residual red ceramic category

(which make a negative contribution to Function 1). High scores on

Function 2, because of the overwhelming contribution of absolute amount

of lithic material to this function, are primarily indicative of large

amounts of lithlcs, with the relative importance of heat-treated unworked

coral evidently a contributing factor. A high score on this function

also indicates a small amount of ceramics, though Napier Complicated

Stamped and check stamped with 4-6 checks per linear inch are relatively

more abundant in high-scoring units than elsewhere.

In Fig. 5 each unit is placed according to its score on the two

functions. As expected, units from the third stratum (each shown in the

two-dimensional space of Fig. 5 as a "3") were differentiated from units

in the first two strata--the midden areas--primarily by the total amount

of ceramics. Function 1 is the horizontal dimension in Fig. 5, and is

primarily determined by the size variable of raw numbers of ceramics.

While the two midden strata have overlapping scores on the first

(horizontal) function, they are well-differentiated by the second, shown

as the vertical dimension in the figure. This function is a continuum

between abundant lithics and few ceramics, towards its positive pole,

,003) -;'.O'0 0.0 2.00UU 4.0 (',

1 1 *


1 1
I 3 3


I* 1 .1. 2

I.Ott0 -- I .o I .00 ..000o

Fig. 5
Excavation Units in Revised Strata 1, 2, and 3 Plotted Against Two Discriminant Functions vi
Note: denotes a group centroid. Function 1 is the horizontal axis, Function 2 the vertical axis.

and abundant lithics and few ceramics, towards its negative pole.

Somewhat surprisingly, this second linear function is also primarily

determined by size variables. We see that the midden by the creek is

typically high in lithics, and low in ceramics; the southern midden

ridge is the opposite. The shape variables are informative in that they

indicate the relation of certain ceramic and lithic categories to this

trend towards concentration of lithics in one area, and ceramics in

another. Thus Napier Complicated Stamped and check stamped with large

checks are relatively more abundant in the northern midden, in spite of

the general trend towards fewer ceramics in this midden area. Likewise,

on the first function, the relatively greater importance of Weeden

Island Zoned Red versus small amounts of residual red in the southern

midden area probably accounts for the location of the second stratum's

centroid beyond that (to the right) of the first stratum's centroid.

Where red filming on a Weeden Island paste occurs in the midden by the

creek it is usually the unzoned variant classified as Weeden Island Red

(Sears 1948).

During analysis several lacunae were noted in the distribution of

the excavated units in the summer sample. In particular it seemed

desirable to extend the grid to the southwest and to test this area, as

a decline in the density of cultural material in this direction had not

been reached. In most cases the grid stopped along the northern bluff

line of the creek which marked the northern extremity of the site; it

was decided to extend the grid to the north to include the creek meander

belt and bluff line. Finally, an area east of Mound A and the plaza

area where the two midden strata seemed to join had been inadequately

tested by the probability sample, but was a critical area for inter-

preting the relative temporal placement of the two midden strata.

Winter 1977 Field Session

The first activity during the next field season, therefore, was to

extend the grid to the southwest and to the north along the creek, and

to sample these areas at the same proportions used for the adjacent

areas during the previous summer. Fourteen 900m2 grid units were added

to the southwest of which four were selected in which to excavate 2 x 2m

tests by means of a random number table. The actual location of the

test within the grid unit was likewise determined by resort to the

random number table. The same strategy along the creek resulted in 19

grid units added to the site within which six tests were excavated.

During this time work on the contour map was completed and the elevations

were tied into the absolute sea-level datum.

Finally a transect was drawn across the eastern portion of the

site, between productive tests in the north and southern strata. Six 2

x 2m units were excavated at approximately 30m intervals along this

line. Then the first, third, and fifth tests in this transect were

duplicated 30m to the east.

The Southwestern extension

The four squares excavated in the newly-gridded southwest area were

uniformly poor in cultural materials. The extent of the occupation in

this direction was now securely bracketed.

The Creek bluffs and meander belt

Two of the tests excavated in this extension to the grid contained

very small amounts of cultural material; these were the tests on the

bottom of the floodplain of the creek. The remaining four tests were on

or near the bluff and produced abundant cultural material and features.

Apparently there was no occupation in areas of high flood potential yet

the creek was a valued resource with a high density of occupation on top

of the bluff and even on the bluffs sloping down towards the creek bed.

Near the eastern end of the site about halfway down a rather steep

section of the bluff one test intercepted a staggered row of postmolds

with extremely rich midden on one side suggestive of a structure. Two

radiocarbon samples from this feature were dated to A.D. 640 95 and

650 + 70.

The Transect

Most of the tests in the transect were quite productive of both

material and features, though the cultural material in the area of the

transect is generally less dense and the deposits less deep than along

the creek or the southern midden ridge. Towards the middle of the

transect markedly lower densities of material were encountered. Two

linear features bordered on one side by postmolds were partially exca-

vated where they coincided with the tests. One of these was radiocarbon-

dated to A.D. 785 + 75, the latest apparently valid date from the site.

As will be shown below, the types of material from this transect area

were somewhat different than those from the northern and southern middens.

Both the materials and the carbon date suggested that this midden was

occupied late in the site history.

Analysis of the Winter Sample

The initial stages of the analysis proceeded along the lines fol-

lowed for the summer sample. Ceramics and lithics were identified to

type or category and the information added to that from the previous


There was enough information now on hand to produce reliable estimates

of the relative amounts of each ceramic type in the site as a unit

and in each of its sub-areas. Artifactual densities over the site as a

whole could also be plotted. In Fig. 6 a shaded isopleth map of the

artifact densities across the site is superimposed over a gridded map of

the site with even-meter contour lines displayed. Densities are found by

dividing the total artifact count in each unit (faunal remains were

excluded from these totals) by the volume excavated. The actual densities

of artifacts per cubic meter are output on the map at the location

of the test. Dark areas indicate zones of high artifact density. Each

level of shading on this and the similar maps to be discussed later

connect areas with similar scores on the variable being mapped. The

values for areas between units are found using the standard SYMAP inter-

polation procedure which is automatically computed "based on the number

and distribution of data points, so that seven points, on the average,

will be found on the initial search" (Dougenik and Sheehan 1975:111/33).

The interfaces between shadings represent the actual contour lines. The

unshaded zones inside the gridded area represent the lowest data values--

in this case, the lowest densities of artifacts. Because units in the

transect (a non-probability sample) were included among the data points,

0 2-71 7 -112 112-302

FIG. 6

a check was run to see whether the distribution of the points was random,

or whether this addition tended to make the distribution uniform. The

resultant "point distribution coefficient" (Dougenik and Sheehan

1975:111/27) is 1.05, well within the 0.91-1.25 range which is considered

to represent a random distribution. While such a distribution is desir-

able for statistical purposes, a uniformly-distributed sample is more

accurate for mapping purposes, and minor errors in interpolation are

probably present in some areas of maps having a point distribution

coefficient of less than 1.25 (Dougenik and Sheehan 1975:111/28). The

highest artifactual densities are found in the midden along the creek

and in the area south of Mound B, indicating either a longer or a more

intense occupation of these areas (cf. Binford et al., 1970:70-71). The

area between the mounds has relatively little cultural material, as

might be expected from a plaza arrangement such as seen by early explor-

ers and documented for several Mississippian period sites (Chapman

1976:121-146). At McKeithen, however, the area west of the plaza has

never been occupied, resulting in a horseshoe-shaped distribution for

the village debris which is open to the west. Of course it can not be

assumed that this entire semicircle was occupied at one time, since all

levels of each unit have been aggregated to produce this map. A similar

plan can be found at the Kolomoki site at least during what Sears

(1956:94) believed to be the full Mississippian or "Kolomoki" period,

when a horseshoe-shaped village area open to the east partially encircled

the large platform mound on the eastern side of the site. In Chapter

Five the shifting patterns of site occupation over time at McKeithen

will be considered.

The initial stratification of the site into two very large midden

areas had worked well and had demonstrated that there were significant

negative correlations between ceramic and lithic distributions as well

as differences in the relative frequency distributions of certain ceramic

types. However, the inclusion in the discriminant analysis of the size

variables, which were the best discriminators, had somewhat masked the

importance of the differences in the distributions of minority ceramic

types. Moreover, discriminant analysis deals with a predetermined

number of groups. If there were clear-cut horizontal distributions of

groups of decorated ceramic types indicative of social residence units

within the site, a cluster analysis would be necessary to discover

groupings as opposed to a discriminant analysis used to interpret group-

ings. Cluster analysis is a tool which has seen extensive recent

use in archeology because of Its utility in forming classifications.

Examples can be cited of use in analysis of mortuary remains to find

groupings of burials indicative of status relationships (Peebles 1972),

to develop classifications of projectile points (Christenson and Read

1977) and of course for developing ceramic typology (Whallon 1971).

Cluster analysis is a family of techniques the variety of which is

explored by Blashfleld and Aldenderfer (n.d.). The particular approach

utilized here is the common heirarchic-agglomerative procedure utilizing

the complete linkage strategy performed on a matrix of distance measure-

ments utilizing a Euclidean metric (program details as always, can be

found in the appendix).

For this analysis only the 26 units having a density of 40 arti-

facts/m3 or greater were used. Distances between units were calculated

on the basis of the percentages of 32 minority plain and decorated

ceramic types. (Extremely rare types were excluded.) The optimum

configuration of the excavated units was reached at six clusters with

one group of seven units, two of four, two of three, and one unit which

did not link with the others at this level. Table 3 shows the relative

frequency of each ceramic type in the cluster in which it is most fre-

quent. More Weeden Island series types reach their highest percentages

in groups 2 and 3 than in the other groups; complicated stamping is best

represented in groups 2 and 5, while sherds of a chalky paste are most

common in groups 3 and 6. Group 1, the largest category, contains peaks

for Weeden Island Red, Carrabelle Incised, a small check stamped, and

Crooked River Complicated Stamped. The distribution of these groups

over the site is shown in Fig. 7.

In addition to a slight tendency for ceramic types from certain

series to act as identifying criteria for a single group, there is a

tendency for groups to cluster within certain site locales. Thus al-

though the trends are weak, more units identified with Group I can be

seen in the midden along the creek than in the rest of the site, while

groups 2, 3, and 4 are most strongly identified with the southern midden

ridge, and the groups 5 and 6 with the eastern midden area.

The results of this analysis certainly do not constitute a test of

the hypothesis that kinship-based residence groups are localized in

specific areas of the site and are identifiable by specific ceramic


Types by series:

Weeden Island series
Weeden Island Plain
Weeden Island Red
Weeden Island Zoned Red
Weeden Island Punctated
Weeden Island Incised
Carrabelle Punctated
Carrabelle Incised
Keith Incised
Indian Pass Incised
Tucker Ridge Pinched

Check stamped series
4 6 checks/inch
7 9 checks/inch
10 + checks/inch
Deptford Linear Check Stamped

Complicated Stamped series
Crooked River Complicated Stamped
Swift Creek Complicated Stamped
Old Bay Complicated Stamped
Napier Complicated Stamped
Kolomoki Complicated Stamped
unidentified rectilinear
unidentified curvilinear
Deptford Simple and Cross-Simple

Chalky wares
St. Johns Plain
St. Johns Check Stamped
Papys Bayou Incised
Papys Bayou Punctated

Pasco series
Pasco Plain

Note: data are in percentage form

1 2 5.


5 6



Fig. 7
Location of Intensive Excavations in Relation to
Assignement of Units to Groups by Cluster Analysis

types. They do indicate that there is a patterning to the area distri-

bution of some of the minority ceramic types. This pattern is not

clear, at this stage in the analysis, for three major reasons. The

first and most obvious distortion results from having ignored the tempo-

ral dimension of change by aggregating all levels of a unit to produce a

single score. It is probable that ceramics from different temporal

segments of the occupation are being used as defining characteristics

for individual clusters as in the case of Group 2, in which both Deptford

Linear Check Stamped and Old Bay Complicated Stamped attain peak average

relative frequencies.

A second problem is that each data point is represented by a single

2 x 2m test which may have intercepted any of a number of different

functional areas in or around a structure. Such possible functional

differences cannot be controlled for in this type of sample.

A third difficulty is that the groups selected by the cluster

analysis are identified by several ceramic types, while a single type,

or possibly even a unit of design included within the range of a single

type, may be the necessary target level in the search for design varia-

tion correlations with residence areas.

A different approach to analysis is necessary to avoid this last

difficulty. The second difficulty can be partially controlled by exca-

vation of larger areas to see whether or not there is significant micro-

areal variation in the distribution of ceramic categories which might be

assignable to functional differences. The first difficulty, that of

temporal control, is addressed in the analysis of the next season's


Summer 1977 Field Session

The final season of village area excavation reported here was

devoted to the excavation of a cluster sample from each of the three

major midden areas defined by the analyses described above. The purpose

of more intensive excavations in a particular locale was to expose whole

structures, if possible, so as to discover what important functional

distributions of ceramics might be present in and around different types

of structures. Of course, it was extremely desirable to learn more

about the structures themselves. Moreover, earlier attempts to develop

chronologies from individual, scattered 2 x 2m tests had shown that a

much larger sample from a particular locality would be necessary if

there was to be any hope of separating within-site horizontal variation

from vertical, temporal variation.

The discriminant analysis reported above, in combination with the

cluster analysis, had indicated that there were at least three major

midden areas within the site: the northern midden by the creek, the

eastern village area, and the southern midden ridge. Unless the differ-

ences between these were purely temporal, it would be hazardous to

develop a single seriation for the entire site and to assume that the

ordering so obtained was chronological. For this reason, three areas

were selected for more intensive excavation. On the basis of the radio-

carbon dates already in hand the area around 312N133E in the southern

midden area was believed to be relatively early; the vicinity of square

350N416E in the eastern midden appeared to be late; while the locale

opened around square 580N376E in the creek midden was thought to be

temporally intermediate. The three areas were thus selected to include

the maximum possible spatial and temporal variation.

The Northern Midden Area: Intensive Excavations

During the previous winter square 586N376E had yielded a staggered

row of probable postmolds bounded on the western side by a strip of rich

midden which continued into the wall beyond the bounds of the unit.

Figure 8 shows a plan of the extended excavations in this area and the

original surface contours before excavation. As is obvious the entire

structure intercepted by the original test was not preserved; the north-

ern and western portions were destroyed by erosional action, apparently

in prehistoric times, while the center and a portion of the southern

periphery were unexcavatable due to large trees.

From the oval section of postmolds that remained, however, a struc-

ture measuring about 3m north-south and 5m east-west can be extrapolated.

This is much smaller than Deptford house structures reported by Milanich

(1973) on Cumberland Island, Georgia, as ranging between 7 x 10m and 4 x

8m, and also smaller than a Weeden Island period house structure near

the Appalachicola River in Northwest Florida which was reported to be

8.9 x 6.2m (Milanich 1974:13).

From the surface contours superimposed on the plan in Fig. 8 it can

be seen that at present the ground slants down towards the creek about

90cm between the southernmost and northernmost postmolds, a slope of

about 250. There was also a slant, though of only 70cm (200), during

the time the structure was in use, raising a question as whether it

could comfortably have been used as a residence. No Fiving floor was


..... .. .. "$$




Fig. 8
Plan view of Intensive Excavations in the Northern Midden

visible in the matrix, which was pocked by Geomys burrows and root

disturbances. Feature 9 first became visible at about the level the

postmolds bottomed out, suggesting that the interior area was dug out

between 5 and 20cm below the depth of the surrounding ground. As can be

seen in Table 4, two radiocarbon dates from Feature 9 average A.D. 647 +

55 using the formula in Long and Rippeteau (1974:209) for averaging

radiocarbon dates from the same temporal context. Table 5 lists the

cultural contents of each of the three major features from this sample.

Features 9 and 9A were similar in containing numerous fragments of bone,

all of which was calcined; one of these from Feature 9 was a fragmentary

squirrel humerus, while the rest were identifiable only to mammal, or

not at all. Feature 9D, on the other hand, contained only six bone

fragments, only half of which are burnt or calcined. It likewise con-

tains a higher ratio of ceramics to lithics than the other two features.

In their contents and their placement relative to the structure the two

northern, linear features are reminiscent of features excavated inside

the periphery of a Potano house at 8 Al 100 (Milanich 1972:44). Using

ethnohistoric Timucuan parallels Milanich interpreted these as refuse

from smudges under the benches or beds lining the interior of the house

wall. The earlier radiocarbon date from Feature 90 suggests that it was

not associated with this structure at all. Excavations in the creek

floodplain to the west have shown that the creek was an active stream

which changed its course at least once during the occupation of the

site. Since excavations in this area also showed a large erosional

gully had been active during the occupation of the site (see Fig. 8),it


FS #

Area of site

63 Northern midden/
probability sample


Unit Provenience/

558N290E Feature 2
possible hearth


Processor Uncorrected Average Date
Number Date (if applicable)




115 Northern midden/ 527N431E postmold associated UM-930 A.D. 510+75
probability sample with Feature 6
263 Northern midden/ 584N380E Feature 9 UM-1092 A.D. 650+70
northern extension,
probability sample A.D. 647+55

264 Feature 9 and as- UM-1091 A.D. 640+95
sociated postmold
416 Northern midden/ 580N376E Feature 9-D UM-1257 A.D. 145+100
cluster sample
181 Southern midden/ 250N140E postmold associated UM-930 A.D. 420+95
probability sample with burnt sand
630 Southern midden/ 306N133E Feature 12 UM-1259 A.D. 245+80
cluster sample probable hearth
202 Southern midden/ 312N133E Feature 8 UM-931 30 B.C. + 95
probability sample
310 Eastern midden/ 350N416E Feature 10 UM-1093 A.D. 785+75
transect sample
505 Eastern midden/ 354N418E Feature 10-D UM-1258 A.D. 265+70
cluster sample
777 Mound B UM-1235 A.D. 465+75 A..
778A Mound B UM-1234 A.D. 370+75 A.. 4+-
782 Mound B (excluded from Mound B average using UM-1233 80 B.C. + 65
Chauvenet's criterion; Long and Rip-
peteau 1974:208)


Material category

smooth sand tempered plain
grit tempered plain
residual plain
Weeden Island Plain
Weeden Island Red
Weeden Island Zoned Red
residual red
Weeden Island Incised
Weeden Island Incised & Punctated
Carrabelle Punctated
unidentified punctated
Carrabelle Incised
Keith Incised
unidentified incised
check stamped with 4-6 checks/inch
Swift Creek Complicated Stamped
Crooked River Complicated Stamped
unidentified rect. complicated
unidentified curv. complicated
total N of ceramics

worked flint
heat-treated flint
other flint
heat-treated coral
other coral
other worked lithic
other lithic
total N of lithics

calcined bone
other bone

ratio ceramics/lithics
ratio worked/total lithics
ratio flint/coral
ratio heat-treated/total lithics

Feature number:
9A 9D

.7 20/ 50 16/39
.4 5/ 12.5 5/12
.6 10/ 25 10/24
.4 1/ 2.5 1/ 2.4
.9 1/ 2.4
1/ 2.4
.4 1/ 2.5 1/ 2.4
1/ 2.4
.3 2/ 4.9
1/ 2.5
.9 1/ 2.4




1/ 2.5


1/ 8.3
4/ 33.3
4/ 33.3

2/ 16.7
1/ 8.3

12/ 99.9

14/ 74
5/ 26


1/ 2.4
1/ 2.4



Note: data are in raw frequencies/relative frequencies

seems likely that the creek bluff was deforested and unstable. If

Feature 90 was associated with an earlier structure, the portion of that

structure to the north of the feature was destroyed by creek meandering,

flooding, or erosion before the occupation of the structure dated to

A.D. 647.

The letters on the plan in Fig. 8 indicate artifacts which were

mapped in place. This included all sherds 5 x 5cm or larger, worked

lithics, and the large faunal fragments which were rare enough in this

acidic matrix to be mapped individually. As might be expected these

bits of refuse were concentrated along the row of postmolds and help

define the original wall of the structure. One of the few mapped lithics

in the eastern half of the excavations was a fist-sized fragment of

limestone on one side of the large northeastern postmold. This apparent-

ly served as a wedge for the corner post. Judging from the poor defini-

tion of the postmolds and the absence of carbonized wood, the posts

associated with the structure seem to have rotted in place rather than

burned, making their identification more difficult. All were sectioned

along two vertical planes to differentiate them from the numerous burrows

and root traces in the matrix. Only a small proportion of the stains so

sectioned were eventually accepted as postmolds. It is possible that

the western posts, of which no trace could be found, were borrowed for

other, later uses. However, their probable former placement can be

estimated by continuing the arc of the existing postmolds through the

areas with high concentrations of mapped specimens.

The Midden East of Mound A: Intensive Excavations

The second area to be more intensively tested was the vicinity of

the square with the linear feature yielding the A.D. 785 date. This was

square 350N416E, excavated as a part of the transect of the eastern

midden area, and located 55m ESE of the top Mound A. Feature 10, from

which the carbonized wood for the date was taken, was bounded on one

side by a row of postmolds, just as Feature 9 had been; from our very

small exposure it seemed likely that this was a portion of a structure.

As in Feature 9, the dark-stained humic deposit rich in cultural material

which constituted Feature 10 and distinguished it from the light tan

sandy matrix began below the top mapping elevation of the postmolds

which delimited it, and continued about 40cm below them (see plan view

of excavation in Fig. 9). Table 6 lists the contents of each of the

major features in this area. Features 10 and 10C were very similar in

depths, general size, an abundant lithic and faunal inventory, and a

high percentage of burnt and calcined bones. More importantly, both

features contained four small fragments of a daub-like material, which

appeared to be partially-fired clay, sometimes bearing impressions of

what might be sticks or reeds. Since these were not found in the matrix

around the features they would seem to be related to the function of the

features, rather than the structure as a whole. Features 10, 10C, and

10D also contained a concentration of large rim sherds, along their

sides and bottoms, though it would an exaggeration to say that they were

sherd-lined. It is suggested that features 10 and 10C were shallow

earth ovens inside which foods were cooked with a smoldering fire.


Material category

smooth sand tempered plain
grit tempered plain
residual plain
Weeden Island Plain
Weeden Island Red
Weeden Island Punctated
Weeden Island Incised
Carrabelle Punctated
Carrabelle Incised
Keith Incised
Papys Bayou Incised
Indian Pass Incised
St. Petersburg Incised
Pasco Complicated Stamped
Kolomoki Complicated Stamped
unidentified rect. complicated
unidentified curv. complicated
total N of ceramics


Feature number:
10 1OB 10C 10D 10E 1OF

2/ 5
1/ 3
1/ 3
3/ 8


2/ 7

1/ 4
1/ 4

2/ 7
1/ 4

4/ 7
5/ 9


1/ 2

1/17 6/67
3/50 2/22

3/ 6 1/ 2
2/ 4
1/ 2

1/ 2
1/ 4
1/ 2

39/100 28/101 55/101

55/100 6/100 9/100

daub-like fragments
worked flint
heat-treated flint
other flint
worked coral
other worked lithics

total N of lithics

calcined bone
other bone


1/ 3

2/33 1/ 3
2/33 11/37
2/33 17/57

1/ 3

1/ 9

31/100 6/99 30/100 11/99


ratio ceramics/lithics-

ratio worked/total lithics

ratio flint/coral

ratio heat-treated/total lithics

40/85 21/91
4/100 7/15 2/ 9

1.3 4.7 1.8 5.0 --- 0.33

0.03 0.33 0.07 0.09

30.0 --- ---

0.52 0.33 0.37 0.45

Note: data are in raw frequencies/relative frequencies


0 27/100

4/57 15/100


--- 0.11

These three features along with 10B appear to be associated with the

structure partially outlined by the upper series of postmolds in Fig. 9.

Feature 10D, which yielded a radiocarbon date of A.D. 265, and 10E,

which had a lower mapping elevation than the upper features, may relate

to a structure hazily defined by a lower series of postmolds unlike the

upper series, which had apparently rotted in place. The posts below

appear to have burned. Neither of these structures was complete or

well-defined. A borrow pit centered 10m ENE of Feature 10F appears to

have partially destroyed the upper series of postmolds. This borrow,

apparently for the construction of Mound A, reduced the depth of the

midden at least as far away from the borrow's center as the 43.5m surface

contour indicated in Fig. 9. If the A.D. 785 date from Feature 10

adequately dates the structure defined by the upper series of postmolds,

and if the material removed from this adjacent depression was used as

fill for Mound A, then the large platform mound must in part post-date

the A.D. 785 structure from which it borrowed. This is the closest to

the mound of at least five borrows which probably provided material for

the construction of Mound A.

While neither of these structures is complete, the later one can be

roughly estimated in size. If the incomplete arc of postmolds separating

Feature 10 from 10B is extended to include 10C within the structure, as

seems reasonable, it could not have been less than 5m north-south, nor

less than 3m east-west. This is approximately the size of the structure

adjacent to the creek, but its incompleteness precludes an identification

of its function.

Fig. 9
Plan view of Intensive Excavations in the Eastern Midden


u 4
o 4 : i

0 C C -

UD m




The Southern Midden Ridge: Intensive Excavations

The third area to be more intensively sampled was the vicinity of

square 312NI33E where Feature 8, excavated in the summer of 1976, had

yielded a radiocarbon date of 30 B.C. Feature 8 was a round, symmetri-

cal pit nearly a meter deep, with straight, steep sides, which was

definitely constructed early in the occupation of this area of the site,

since the backdirt from the excavation of the pit clearly underlay, in

profile, the main midden accumulation. Despite its excellent context,

the date seems open to question in view of the ceramic inventory, pre-

sented in Table 7. Nevertheless, the relatively early occupation of

this area of the site is underscored by the A.D. 245 date from Feature

12, an apparent hearth in square 306N133E, 6m south of Feature 8.

The original plan for the excavation of this third area was to open

up as much horizontal area as possible in all directions adjacent to

square 312N133E, in order to delimit the structure (if any) to which the

feature pertained. However, no clear structural evidence was found. It

was decided to take advantage of the opportunity to excavate a transect

across what appeared to be two low ridges south of square 312NI33E.

Altogether, 27 2 x 2m tests were opened, most of them along the 133E

line. For the northernmost 16m adjacent 2 x 2m squares were excavated

along this north-south line; over the next 29m, every other possible 2 x

2 along this line was excavated; and as the southernmost extreme of the

midden was approached, tests were placed at approximate 10m intervals.

One of the southernmost tests was placed in a depression filled to a

depth of 60cm with dark grey plow zone material which elsewhere averages


Material category:

smooth sand tempered plain
grit tempered plain
residual plain
St. Johns Plain
Papys Bayou Punctated
Weeden Island Plain
residual red
Weeden Island Punctated
Weeden Island Incised
Carrabelle Punctated
Carrabelle Incised
Keith Incised
unidentified incised
check stamped 4 6 checks/inch
check stamped 7 9 checks/inch
cord marked
unidentified punctated
total ceramics

worked flint
heat-treated flint
other flint
silicified coral
other lithics
total lithics

calcined bone
other bone

ratio ceramics/lithics

ratio worked/total lithics

ratio flint/coral

ratio heat-treated/total lithics

Feature number:
8 12
52/ 38 4 33
3/ 2 2/ 17
39/ 28 3/ 25
1/ 1 1/ 8
1/ 1
7/ 5
1/ 1 1/ 8
6/ 4
4/ 3
5/ 4
1/ 1 1/ 8
1/ 1
3/ 2
4/ 3
6/ 4
1/ 1
1/ 1
136/100 12/ 99

1/ 7
9/ 60
4/ 27
1/ 7


118/ 79
32/ 21





1/ 33
1/ 33

1/ 33
3/ 99


Note: data are in raw frequencies/relative frequencies

about 25cm deep. This depression was apparently one of the three or

four borrows for the construction of Mound B, which was partially filled

during years of cultivation earlier in this century. In reality there

was only one ridge, and what had first appeared to be a second ridge to

the south was a portion of the natural terrace on which the site is

located set off by the somewhat linear borrow to the south.

It should be mentioned here that the portion of the site now planted

in pines has undergone considerable leveling of its prominences and

filling in of its depressions. Within the memory of local informants

Mound A was considerably higher and steeper in outline than it is now.

Such is also the case for the midden ridges. The northern of the two

prominences tested was, like the terrace to the south, a natural eleva-

tion which was chosen for habitation. Figure 10 shows the density of

artifacts within the area of this cluster sample; the two peaks are just

north of the northern ridge, and towards the northern part of the south-

ern terrace. These areas of peak density, which are now slightly higher

in elevation than the areas to the north and south, are less prominent

than they were during the occupation of the site. Along the long north-

south profile exposed intermittently by these tests, it can be seen that

Zone I (the plow zone) is more shallow in these two areas than to the

north and south. As will be seen in Chapter Five the northern of the

two areas with high artifact concentrations was the earlier.

Faunal Remains

The faunal materials from all three intensive excavations along

with the material from the earlier probability and transect samples were

Fig. 10
Plan view of Excavations in the Southern midden:
!sopleth map of Artifact Density

HlOE 135E 160E

I::::::::: :i::::: :. ..:::::: : :!i::: :::!; :^ ::^ 30
FT."6;?? .... ..''.'..'.''.; '.'.'.............^tt^

. . .. .. .. .. . . .. . . . . . . . . . .* .*** .. * ,
... ...................... ................,
.................... .... .... .. ........**

~...,........ I~~ ..... ... .. .***
.~.4'~.f *. '. ~'~. .... ... .... ...
2 .. ....... * *

..... ............. ....... .......... .... ...........

.+** +^*+t *^ + 1tH, +** + ,+*,,,+,^*
1 1 1 11 11 11l l l I**+ i<*t+>**ee e + *e** * 4*.* * 44

>ee8fee9 eeee~ eeoeeeee4al oeaaeeuee ee fi44 *-e eeeeoeeeeeeee eBya eeeTpga eeeeeeeeeiici eeeeeeeeeeeeeiaoin 00 26 0N
i e~e 4lae~e96eeeoe eeee46aa6aeae
i oeeeeoeeee * 49994 eoeeee 4e aa eaf4 ee ea~e ee ~~eee
ie4e6aeeoeeeeeeoeee eeeee~eeefeeeeeeee
>eeeeeeefe afe aee aeee 44,eeetik9efeeteeefe~e.
ie8 ae~e eeoeeeefa***4ee^6aaeeaeeea44eeaee
ifl464feeeoa4e4 e~a4**+++eeee(eo eee 9feelee94eaele
i~*+e9e*++4 *+--49ea eooeee~eeeeeeeeee4e4
.+***e**<, ,* + **+ +***+++ *+*++*9 ee e93, eeeee o eea6ea6
** *******++++** ***.. **+ ****+*.+*+j.++ .+ + ,e~
. -W ... ..^................... .l-*-*+**+ ,*** +rf*
...........r ....................... ,+,+
............ ............~*...... ........... ,**e
22 0
46 22
-2 0N


.. 9. . . . .
.......... .......
.............. ::: : ... :::
""" " ~ ese 8.0g ee aegaea"" 0"
::lf:a """ "' ...Oat* .....'""

Goo...- .... 0:1 = .e.0061e.. ..eee~s
Q:.a ....... ee
"See 0:00ne.... aaoa eve l:...M
'0009"leeee'o a04000'e
... ................
.......... ...................
e .......... : .................

........... ..... eeeeosee
............. .es~eeees~asee

rrrrrrr++..................... .~r~
...................... ... ... ...

IL / 230N

0 0h

analyzed by the author using the facilities of the Florida State Museum

Zooarcheology Laboratory with the assistance of Dr. Elizabeth S. Wing

and Elizabeth M. Reitz. In general, faunal preservation was very poor.

Large identifiable bone fragments were treated on excavation with a

mixture of Ethulose and Carbowax but this stabilization was often insuf-

ficient in view of the advanced decalcification of the bone brought on

by continuous leaching in the sandy acid soils. Fully 65% of the recov-

ered fragments were preserved because they had been partially burnt or

completely calcined, and these were always very fragmentary. The average

weight of each recovered bone fragments was only 0.25gm.

All identified species (excluding terrestrial snails and obviously

modern inclusions) from the midden are listed in Table 8 by number of

fragments per element. Of the more than 3000 bone fragments excavated

and sorted, 92.0% were identifiable only to phyllum (e.g. Mollusca),

class (e.g. Aves), or were completely unidentifiable. Of the few re-

maining fragments identifiable beyond class, 92% were deer. Under these

conditions of obvious differential preservation no attempt has been made

to quantify minimum number of individuals or amount of meat represented.

No seasonal information was preserved in the sample.

Nevertheless, it is interesting to note that all the positively-

identified shells and the shark could not have been obtained locally.

The freshwater mussels (Unionidae) might have been obtainable no more

than 10km distant, in the Suwanee River, but the salt-water species must

have originated on the Gulf littoral (the nearest source would be the

Steinhatchee-Deadman Bay area 70km distant) or the Atlantic coastal

TA!il.E 8

El ement :

cf. C aSS O ostra v. I )
L t .- 0o I C rO E na 4. i- E 1 -' C I.- 4-4
Srunid i ed ) 0 ) ) C C C

cf. Busyarch s si

mi ja calva 1
Sif.ri mes c ICtaluridae 3

Ict- lurus sp. 1

Micropterus salmoides 1
unidenti f ied 1 2

Kinosternon spp. 5 5
Tric h i Ls cf. mi Iber i ferox 1

ADiroa cah va ticuria 1 1
Si lui ituim s cf. Ictaluridae 2 1 3
1ct I lurus sp. 1
Micro -terus sainoidles 1 1

unidentii iedturtle 2
PTli mississiiens 1
Eiydidae i6 d I7
Kinusternmn sp|'. 5 5
Trinnyx fcrox 1
i ui -oc.helys letil aria 1 1 2
uniderf. i fie turTle IJL2 2

unidentified 16 16
cf. Sciurus spp. 1 1
Odocoileus virginiana 11 26 2 1 20 2 1 4 7 15 7 111 13 220
unidentified 1462 1462

1 6 15 1 1 17 2 27 2 1 1 20 2 i 4 8 15 7 127 13 2992 3263

areas, possibly from the area of the mouth of the St. Johns River, 100km

distant. Of the eight examples of these exotic fauna, five were obtained

from the midden directly southeast of Mound B, an area representing less

than 1/3 of the total excavated volume.

Of the remaining species, it is perhaps surprising that any fish

were identified under these conditions of preservation. Most of the

non-deer fauna which survived was calcined, while most of the identified

deer survived because it was relatively more massive. It seems likely

that fish were very under-represented in the sample, and that if the true

proportions were known, they would approximate those at a probably-

contemporaneous habitation site in North-central Florida, 8 Al 169

(Cumbaa 1972:48) where approximately 90% of the recorded individuals in

the faunal collection originated in swamp and aquatic ecosystems. A

similarly heavy reliance on aquatic resources at the McKeithen site

would explain the site location in relation to the water resources

described at the beginning of the chapter.

No analysis has yet been completed on the flotation samples extracted

from the features, although pollen samples are currently being ex-


Before the important question of chronology can be examined in

Chapter Four it is necessary to examine more closely the kinds and

amounts of cultural material recovered from the village excavations.

This is the subject of Chapter Three.


In Chapter Two the sampling phase of an anticipated multi-stage

sample of the village area at the McKeithen site was described. As a

first step, an intensive surface collection over a fraction of the gridded

area was used to stratify the site into three sub-groups for excavation.

Two of these were believed to be midden areas, the other the plaza and

outlying fringes of the village. Following the surface collection, the

first phase of the excavation was a probability sample since this elim-

inated the effects of bias in the computation of population parameters

while controlling the other undesirable source of variation in the data--

the so-called "random sampling bias" (Anderson and Zelditch 1968:195).

The site was stratified for three major reasons:

1. The variability in the computed richness value (hoped to be an

estimate of artifact density) for each gridded unit in the

surface collection was high enough to indicate that a simple

random sample would yield population estimates with very wide

confidence intervals (Asch 1975:174).

2. If strata were drawn across the site such that the two areas

with high richness values were isolated from the third area

with apparently less dense cultural material, the resulting

intra-group homogeneity should considerably narrow the eventual

confidence intervals around the population estimates.

3. Stratification before selection of grid units to be excavated

helps assure a more uniform distribution of tests across the

site by eliminating the possibility of extreme clustering in

sample location.

Given a stratified probability sampling design, the decision to

sample the strata disproportionately was justified in this manner:

1. The village areas themselves were of primary interest, not

their fringes, or the central, clear area. A proportionate

sample of each stratum without altering the total number of

units excavated would have resulted in the excavation of five

units in each of the two midden strata, rather than the 10

actually excavated.

2. As large a sample of cultural materials and features as pos-

sible was desirable in order to maximize recovery of infrequent

data categories.

3. Direct comparison of the two village areas was required by the

research hypotheses; hence these were sampled at the same

proportion, but more intensively than the intervening areas.

In addition to the three original strata, two additional probability

samples were excavated from southwestern and northern extensions to the

gridded area.

Estimation of artifact populations from the excavations used only

the probability sample and the results of the transect and cluster

samples were disregarded since they contain an unknown amount of bias.

To compute a point estimate of the total artifacts at the site, the mean


for each of the sampling strata is first computed, then weighted by the

inverse of the sampling proportion for that stratum. Multiplying this

weight factor by the mean and summing the products results in a point

estimate for the total population for that artifact at the site (Sudman

1976:126). This procedure is illustrated in Table 9 for the computa-

tion of the total number of ceramics at the site.

The reliability of the resultant estimate is found by multiplying

the variances around the means of each of the sampling strata by the

squares of the strata population sizes, and summing the products (Sudman

1976:128). The square root of this figure is the standard deviation of

the total sample estimate. The finite population correction factor

which corrects estimates for small populations sampled at a fraction of

0.05 or more can be ignored for this data, since sampling proportions

vary between 0.003 and 0.0014 (Read 1975:54).

Applying this procedure to the point estimate of 6,613,000 sherds

recoverable using V" mesh screen (and ignoring discards averaging less

than one gram in weight) a one standard deviation confidence interval

around the means results in a range of 4,774,800 8,451,200 sherds.

That is, there is a 67% probability that the true number of ceramics in

the village midden falls within this range of values. This is a sur-

prisingly wide confidence interval, resulting from the mediocre predic-

tion of artifactual densities provided by the surface collections which

were used to construct the sampling strata. In fact, in each of the

initial strata, units of both high and low artifactual densities were

excavated, considerably adding to the variance around the means for each

Excavated Sample N of Est. ceramic
Stratum Location Identification Area (m2) area (m2) proportion ceramics mean population

1 northern and village midden 29700 40 .0013 1686 168.6 1,296,923
eastern areas

2 southern and village midden 29700 40 .0013 1672 167.2 1.286,102
eastern areas

3 central, west- plaza and outlying 122400 40 .0003 1163 116.3 3,559,826
ern, and extreme

4 southwestern marginal 12600 16 .0012 18 4.5 14,175

5 northern creek village and marginal 17100 24 .0014 640 106.7 456,002
bluffs and bot-

211500 160 5179

Overall sampling proportion in probability sample = 0.00076 Overall mean number of ceramics/unit = 129.5

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs