Pluckhahn, Thomas J., and Ann S. Cordell 2011Paste Characterization of Weeden Island Pottery from...

PASTE CHARACTERIZATION OF WEEDEN ISLAND POTTERY FROM

KOLOMOKI AND ITS IMPLICATIONS FOR SPECIALIZED PRODUCTION

Thomas J. Pluckhahn and Ann S. Cordell

PASTE CHARACTERIZATION OF WEEDEN ISLAND POTTERY FROM

KOLOMOKI AND ITS IMPLICATIONS FOR SPECIALIZED PRODUCTION

Thomas J. Pluckhahn and Ann S. Cordell

We report the results of a petrographic analysis of potteryfrom Kolomoki, a Middle and Late Woodland period moundand village complex in southwestern Georgia. Thin sectionsof 65 sherds representing several prestige and utilitarianWeeden Island pottery types, from both domestic (midden)and ceremonial (mound) contexts, were obtained. Forcomparison, we also analyzed samples from a few potentialclay sources. We characterize the range of variability inpaste/resource groupings present in the Kolomoki assemblageand use these data to address patterns of manufacture andexchange of Weeden Island pottery through comparisons tothin sections of comparable types from the McKeithen siteand other Weeden Island sites in the region.

Archaeological investigations of craft specialization—like those of political economy in general—have longbeen linked with issues of social evolution. Specializedcraft production has traditionally been consideredboth a cause and marker of the development of morecomplex, hierarchically organized societies (Cobb1993:67). Within the Southeast, for example, debateconcerning craft specialization at Cahokia and Mound-ville has been permeated by broader controversiesregarding the size of these polities and the degree ofsocial stratification within them (Blitz 1993; Milner 1990;Muller 1997:342–346; Pauketat 1987; Prentice 1983, 1985;Welch 1991; Wilson 2001; Yerkes 1983, 1989).

In recent years, however, archaeologists working withnonstratified societies have endeavored to decouplespecialized craft production from social complexity.These studies are founded in a broader and moreinclusive definition of specialization. Costin (2001:276),for example, characterizes specialization as, simply, asituation in which ‘‘fewer people make a class of objectthan use it’’ (see also Cobb 1993:66). This strikes us asperhaps overly broad, potentially including even do-mestic production and consumption of a class of objectsas a form of specialization if the former is undertaken bya subset of the household. Nevertheless, we agree withthe more general point that specialization should beunderstood as multidimensional, existing along contin-uums of context (from independent to attached),concentration (from dispersed to nucleated), scale (fromsmall and kin-based to factory), and intensity (frompart to full time) (Costin 1991:1–18). So defined and

conceived, specialization is recognized as a featurecommon to small-scale societies. Explanations forspecialization in state-level societies, which have tradi-tionally emphasized political aggrandizement and eco-nomic maximization, may be a poor fit for such small-scale societies; thus motivations are increasingly soughtin the need for socially valued goods which are criticalfor ritual performance (Spielmann 1998, 2002, 2008).

This more expansive understanding of specializationhas thus far found limited application in archaeologicalstudies of the societies that were present in theSoutheast prior to the Mississippian period (forexceptions from the northern periphery of the South-east, see Spielmann [1998, 2002] and Nolan et al.[2007]). In this paper, we consider possible specializa-tion in the production of pottery among the Middle andLate Woodland societies of the Gulf Coast and adjacentinterior sections of Alabama, Florida, and Georgia(Figure 1), broadly defined by the Weeden Islandceramic series and dating from around A.D. 200 to1000 (Milanich 2002:352). Weeden Island pottery, withits diverse suite of styles and unusual vessel shapes(many of which are poorly suited to utilitarianpurposes) (Figure 2), as well as its strong associationwith mortuary deposits, has long been recognized as adistinct and specialized class of ceramics. The possi-bility that Weeden Island pottery was produced byspecialists, however, has been muddled by assump-tions regarding a linear relationship between special-ization and complexity. Briefly, archaeologists empha-sizing the seemingly precocious sociopolitical com-plexity of Weeden Island societies have argued forspecialization (Sears 1956:98, 1973:39), while those whohave asserted the more egalitarian trappings of thesesocieties have considered specialization more cautiously(Cordell 1984:195; Milanich et al. 1997:139).

As a means of assessing specialization in WeedenIsland pottery production, we focus on the intensivecharacterization of ceramic paste. Our ceramic sampleis drawn primarily from the Kolomoki site in the lowerChattahoochee Valley of southwestern Georgia. Kolo-moki is the largest Weeden Island settlement in theregion, with at least nine mounds and a U-shapedoccupation area centered on a circular plaza andextending nearly a kilometer north-south and east-west (Pluckhahn 2003; Sears 1956). The site has beendated to about cal. A.D. 350–750 (Pluckhahn 2003),although recent work suggests it likely persisted later,to about A.D. 850 or 900. Our pottery sample is derivedfrom intensive work at Kolomoki in the middle

288

twentieth century by William Sears (1956), as well asmore recent investigations by Pluckhahn (2003).

We compare paste groupings at Kolomoki to thoseidentified from a sample of sherds from the McKeithensite in north-central Florida. McKeithen, intensivelyinvestigated by Milanich and colleagues (1997), in-cludes three mounds and U-shaped habitation areacentered on a circular plaza and measuring about 600 mby 500 m. The site is dated to around cal. A.D. 200–900,although radiocarbon dates are skewed to the first halfof this range. We also compare the Kolomoki andMcKeithen paste groupings to a sample of sherds fromselected Weeden Island sites in the region.

Our analysis, while not definitive, lends supports tothe assertion that some Weeden Island pottery wasproduced by specialists, probably a subset of thepopulation at Kolomoki. We consider the nature ofthis ceramic production and distribution in light ofrecent work on specialization in small-scale societies,especially as they relate to ritual.

Considerations of Specialization in Weeden IslandPottery Production

Weeden Island pottery was recognized as a distinctand specialized class of ceramics long before the series

was formally defined archaeologically. In his work onthe Gulf Coast, C. B. Moore (1901, 1902, 1903a, 1903b,1905, 1907, 1918) noted the repeated association of theeffigy forms that we now recognize as Weeden Island,as well as other incised and punctated styles of theWeeden Island series, with caches placed on the eastsides of mortuary mounds. He frequently referred tothe effigy form as a ‘‘freak’’ or ‘‘ceremonial’’ ware (e.g.,Moore 1902:352–353).

The Weeden Island pottery series was formallydefined in the 1940s (Willey 1945; Willey and Woodbury1942). Willey (1949:406), in his landmark synthesis ofthe archaeology of the region, described WeedenIsland pottery as ‘‘the most outstanding of the GulfCoast and, in many respects, of the entire aboriginaleastern United States’’ with regard to quality ofproduction, form, and decoration. He noted thatmuch of the pottery appears to have been used forstorage, serving, or ceremonial purposes rather thancooking.

William Sears was the first to explicitly inferspecialized production from the specialized form,decoration, and function of Weeden Island pottery.In his work in Mounds D and E at Kolomoki, Sears(1951, 1953, 1956, 1973) recovered dozens of WeedenIsland effigy vessels from east-side pottery caches,prompting his classification of these as a ‘‘sacred’’

Figure 1. Location of Kolomoki and other sites mentioned in the text.

PASTE CHARACTERIZATION OF KOLOMOKI POTTERY

289

category of pottery distinct from the ‘‘secular’’complicated stamped types more common in villages.The quality of the sacred vessels at Kolomoki con-vinced Sears (1956:98) that ‘‘specialists must have beenpresent, the persons with ultra-thorough training in

the techniques and style canons of sacred potterymanufacture.’’ Sears’s argument for specialized pro-duction, however, was largely based on analogy andbound up with his interpretation of the Kolomoki asthe center of a state-level society:

Figure 2. Examples of Weeden Island pottery: (a–b) from McKeithen (reproduced courtesy of the Florida Museum of NaturalHistory), (c–e) from Kolomoki (reproduced courtesy of the Laboratory of Archaeology, Department of Anthropology,University of Georgia).

SOUTHEASTERN ARCHAEOLOGY 30(2) WINTER 2011

290

We should consider too, as another index of social complexity,the degree of craftsmanship apparent in some of the ceremonialvessels. I know of no way, except through a ‘‘feel’’ forcraftsmanship, for mastery of materials and technique, andthrough the opinion of contemporary craftsmen in the medium,to appraise this kind of thing. As an example, in the opinion ofnational prize-winning ceramicists, some of the effigy vesselsexcavated at Kolomoki in Mound D are technically perfect, andeven, in at least one instance, very nearly outside the state ofthe art as it exists today. These pots are not then the products ofpotters whose training and experience is limited to occasionalproduction of cookpots for their own use, whether these pottersbe male or female. The mastery of medium and of technique callfor complex training processes which must involve socialsupport. (Sears 1973:39)

The possibility of specialized production of WeedenIsland pottery was subject to more rigorous analysis inwork at the McKeithen site. Kohler (1978; see alsoMilanich et al. 1997:45–90) employed production stepmeasures and diversity indices to delineate threecategories of pottery at McKeithen: elite, tradewares,and utilitarian. Rice (1980) subjected pottery fromMcKeithen and other northern Florida sites to neutronactivation analysis (NAA). Her work revealed supportfor a sacred-secular ceramic dichotomy at McKeithen,with sacred wares exhibiting greater relative frequen-cies of pastes distinct from those typical of utilitariantypes and local clay sources. However, Rice argued formultiple centers of ‘‘sacred’’ Weeden Island potteryproduction rather than a model (derived from Sears)of specialized production at, and distribution from,Kolomoki.

Cordell (1984; see also Milanich et al. 1997:120–139)followed up on Rice’s (1980) study with a comparisonof mound and midden samples at McKeithen and othersites in northern Florida. Her analysis, which focusedprimarily on paste and secondarily on other produc-tion-related attributes such as color, hardness, andporosity, revealed evidence for specialized potterymanufacture for six types associated with the WeedenIsland series following criteria specified by Rice (1981).These six types were said to constitute a category of

‘‘elite’’ or ‘‘prestige’’ pottery, based on the fact that theyexhibited greater standardization of manufacture, morerestricted occurrence (mainly in mound or ceremonialcontexts), and greater standardization and elaborationin production (Cordell 1984:194–195; Milanich et al.1997:138). Cordell suggested that the manufacture ofelite pottery may have been undertaken by a restrictednumber of individuals at McKeithen, perhaps for tradefor elite types from elsewhere. However, Cordell(1984:195; see also Milanich et al. 1997:139) felt thatoccupational specialization was unlikely, ‘‘given thehypothesized socio-economic level of the McKeithenWeeden Island population.’’ Cordell’s research alsosupported Rice’s multicenter model for the manufac-ture of Weeden Island ceremonial pottery.

Identifying Paste Groups at Kolomoki

The Pottery Sample

Over 56,000 sherds and 195 restorable vessels werecatalogued from the Kolomoki excavations in the 1940sand 1950s (Sears 1956:15). More recent work at Kolo-moki has yielded more than 50,000 additional sherds(Pluckhahn 2003). From this combined assemblage, asample of 270 sherds was selected from domestic andcivic-ceremonial contexts. Kolomoki Mound D was theprimary ceremonial context sampled, but a few sherdscome from Mounds B, C, and F (Table 1).

Our sample includes a variety of cultural historicaland functional types. The cultural historical designa-tions employed here follow the formal types estab-lished primarily on the basis of temper and surfacedecoration (e.g., Goggin 1952; Sears 1956; Willey 1945,1949; Willey and Woodbury 1942). Following previouswork, we consider three categories of functional types:cult or sacred (used interchangeably in previous studies;we employ the latter term here), prestige or elite (alsomutually substitutable in prior work; we use theformer), and utilitarian (Table 2) (Cordell 1983, 1984;

Table 1. Kolomoki sample counts by pottery type and gross provenience.

Pottery Type Mound B Mound C Mound D Mound D Submound Mound F Off-mound midden Total

Weeden Island Incised - - 4 1 - 14 19Weeden Island Plain - - 2 - - 4 6Weeden Island Red - - 14 - - 18 32Weeden Island Zoned Red - - 2 - - 2 4Mercier Red on Buff - - 2 - - - 2Indian Pass Incised - - - - - 2 2Carrabelle Incised - - - - 1 1 2Carrabelle Punctated - - 3 - - 11 14Plain - - 6 1 1 85 93Swift Creek Complicated Stamped 1 1 1 - - 85 88Blakely Complicated Stamped - - - 2 - - 2Napier Complicated Stamped - - - - - 1 1Mound Field Net Marked - - - - - 1 1Lamar Complicated Stamped - - - - - 4 4Total 1 1 34 4 2 228 270


291

Kohler 1978; Milanich et al. 1997; Sears 1956, 1973). Thesacred category is comprised of effigy vessels (manywith excised geometric cut-outs), which at Kolomokiand other Weeden Island sites are invariably foundonly in mounds. Specimens assigned to this functionalcategory include a variety of cultural historical types(most commonly Weeden Island Red and WeedenIsland Incised), but distinguished by effigy forms andoccurrence in mound contexts.

The prestige category was defined at McKeithen onthe basis of production characteristics (non-effigy forms,high frequency of nonlocal pastes, high production stepmeasures, standardization in form and decoration) andprovenience (mounds and limited off- mound contexts).As previously defined, this category includes the cul-tural historical types Weeden Island Zoned Red,Weeden Island Red, Weeden Island Incised, WeedenIsland Punctated, Papys Bayou Punctated, and IndianPass Incised; to this roster we add the Mercier Red onBuff type based on its similarity in decoration anddistribution. Finally, again following the work atMcKeithen, the utilitarian category includes a varietyof other cultural historical types differentiated by widedistribution in off-mound contexts (less commonly inmounds), higher frequencies of local pastes, relativelylow production step measures, and greater variability inform and decoration.

In addition to the Woodland pottery, four Lamarperiod, Mississippian sherds were included in thesample for comparison. In addition to the ceramics, weanalyzed four clay samples from the Kolomoki vicinityfor comparison.

It is important to note several limitations of thesample. We assume that the definitions of prestige andutilitarian types made at McKeithen are applicable toKolomoki and the rest of the Weeden Island area, butthis may not be the case. Further, these functionalcategories may not reflect the full use-lives of thepots. For example, Swift Creek pottery is classifiedas utilitarian, but it was also found in mounds at

Kolomoki (Sears 1956); previous work elsewheresuggests that it was often used both for utilitarianpurposes and in mortuary-related rituals (Wallis2011).

The sample of ‘‘sacred’’ pottery is skewed to MoundD, mainly because more sherds from proveniencesrelating to this mound were available for study.Although we selected rim sherds when possible, muchof the sample is comprised of body sherds and we thuscannot rule out duplication of some vessels. We havenot controlled for variation in time; petrographicstudies by Stoltman and Snow (1998:138–139) suggestthat the size of quartz grains may increase from Middleto Late Woodland.

The sample size is small relative to the overallassemblage from Kolomoki. Nevertheless, we believe itoffers a representative picture of the variability inlocally made pottery, given the range of pottery typesthat are included. Pottery types are represented in thesample in proportions relatively similar to theiroccurrence overall. For example, plain and Swift Creeksherds make up 36.7 and 32.6 percent of the sample,respectively, generally consistent with their represen-tation in Pluckhahn’s (2003) test units and excavationblocks. The Weeden Island types that are less commonin recent excavations are represented by fewer sherdsin our samples.

Methods of Analysis

Three methods of analysis were used to characterizepaste: standard microscopy, refiring in an electricfurnace, and petrographic analysis. The two formermethods were conducted on the total sherd sample.Petrographic analysis was conducted on a subsampleof 65 sherds.

Standard microscopy entailed using a binocularstereomicroscope for gross paste analysis (i.e., toidentify predominant constituents and to distinguish

Table 2. Count (and percentage) of pottery types by functional category.

Pottery Type Prestige Sacred Utilitarian Other Total

Weeden Island Incised 19 (100) - - - 19Weeden Island Plain 1 (17) 5 (83) - 6Weeden Island Red 24 (75) 8 (25) - - 32Weeden Island Zoned Red 3 (75) 1 (25) - - 4Mercier Red on Buff 2 (100) - - - 2Indian Pass Incised - - 2 (100) - 2Carrabelle Incised - - 2 (100) - 2Carrabelle Punctated - - 14 (100) - 14Plain - 2 (2) 91 (98) - 93Swift Creek Complicated Stamped - - 88 (100) - 88Blakely Complicated Stamped - - 2 (100) - 2Napier Complicated Stamped - - 1 (100) - 1Mound Field Net Marked - - 1 (100) - 1Lamar Complicated Stamped - - - 4 (100) 4Total 48 (18) 12 (4) 206 (76) 4 (2) 270


292

gross temper or texture groupings). The microscopewas equipped with an eyepiece micrometer and fiberoptic illuminator. A magnification of 303 was usedbecause it was powerful enough for distinguishingvery fine particle sizes (0.0625mm to , 0.125 mm) butlow enough for estimation of size and relativeabundance of the larger coarse and very coarseconstituents (grit sizes, . 0.5 mm). Occasionally, highermagnifications (up to 703) were used when necessary.All initial observations were made on sherd edges thathad been freshly cut with a lapidary saw. The texturalintegrity of the pastes was remarkably well preservedin the cut edges, which also provided larger and moreuniform surface areas for examination. Size of aplasticswas estimated with reference to the Wentworth Scale(Rice 1987:38). For purposes of the gross paste analysis,particle abundance was estimated with reference to arelative abundance scale.1

Refiring was conducted to standardize color com-parisons between samples and we use this color toassess relative iron content, thus providing an addi-tional means of distinguishing clay resource differenc-es. Beck (2006) refers to this method as oxidationanalysis. The lapidary saw was used to control thedesired size of fragments for refiring in all but sixsherds in the sample that were large enough to spareremoval of pieces for analysis. Sherds were refired inan electric furnace at a temperature of 800uC for30 minutes,2 conditions that most likely exceeded thoseof the original firings. A fresh break was made afterrefiring to note color changes and Munsell colors wererecorded for core colors of a subsample of refiredsherds. Four gross refired color ranges were specified,corresponding to relative iron oxide contents rangingfrom very low to high.3

Sixty-five sherds were thin-sectioned for petro-graphic compositional and point count analyses.The petrographic analysis was conducted to evaluatecompositional homogeneity and differences within andbetween gross paste categories. Point counts weremade for quantifying relative abundance of constitu-ents. The point-counting procedure involved using apetrographic microscope with a mechanical stage andgenerally followed recommendations by Stoltman(1989, 1991, 2000).4 Point-count data were used tocalculate a Sand Size Index [SSI] for each sample,following Stoltman (2000:314).5

All gross paste and petrographic analyses werecarried out in the Florida Museum of Natural HistoryCeramic Technology Laboratory (FLMNH-CTL), wheredata recorded for all three analyses remain on file. SAS9.2 for Windows (SAS Institute 2008) was used forcomputer analysis and statistical comparisons of data.Tri-Plot, an excel spreadsheet program (Graham andMidgley 2000) was used to create ternary diagrams ofpetrographic data.

Results: Principal and Accessory Paste Constituents

The principal constituents in the sample are quartzsand (0.0625 to , 0.5mm) and quartz and quartzite grit(. 0.5mm). Quartz occurs in all sherds in the sample, invarying sizes and abundances. Its status as an addedtemper or a naturally occurring constituent, or somecombination of both, is uncertain. Quartz aplasticsfalling into silt and very fine Wentworth particle sizesare usually considered to be naturally occurring con-stituents of the clay source (Rice 1987:411; also seeStoltman 1989:149–150, 1991:109–111). Coarser particlesizes may be indicative of tempers (Rice 1987:411,Stoltman 1989:149, 1991:109–111). Polycrystalline quartzor quartzite is present in most thin-sectioned cases andis the principal constituent of grit-tempered sherds inthe thin section sample. Particles of quartz, quartzite,and other sand constituents in the present sampleare predominantly subangular, but angular and sub-rounded particles are also present. This variability isconsistent with particle shapes observed in the local claysamples (see later discussion).

Mica, primarily muscovite, is a principal constituent ofabout half the sample and is considered a naturallyoccurring constituent of the clays rather than temper.Relative frequency of mica formed the basis fordistinguishing the pottery in terms of micaceous versuslow mica resources. Grog or clay lumps, ferric concre-tions, feldspars, and mafic (ferromagnesian) mineralswere also observed in varying frequencies, in many caseswith standard microscopy. Grog particles are generallysubrounded clay lumps rather than recycled sherdtemper. They are a rare to occasional constituent ofmany sherds in the sample but are frequent in only a fewsherds. Clay lumps may have resulted from incompletemixing during paste preparation in most cases. Ferricnodules or concretions are occasional constituents ofmany sherds but probably represent naturally occurringconstituents of the clays rather than tempering materials.Grains of feldspars and dark, mafic minerals wereobserved in a few sherds during the preliminaryanalysis, but only because their frequencies weresufficient to be noticeable with standard microscopy.

Petrographic analysis shows that feldspars, mostlymicrocline and plagioclase, are occasional constituentsof most thin sections. Granitic rock fragments (poly-crystalline grains composed of feldspars and/or feld-spars and quartz) were observed in a few cases, butonly in thin section. The mafic mineral detected duringgross paste analysis was identified in thin section asamphibole, although epidote was also observed.Feldspars and mafic minerals are rare to occasionalconstituents of most of the thin-sectioned sherds andare thus interpreted as natural constituents of the claysor incidental to sand tempers. Frequent amphibole isobserved in relatively few cases in the entire sample.


293

Lesser accessory constituents include biotite mica andheavy minerals such as kyanite, tourmaline, rutile, andzircon, all of which are only detectable in thin section.An unidentified isotropic mineral, probably garnet,was observed in a few thin sections.

Siliceous microfossils, including fragmented spongespicules and opal phytoliths, were unexpected butpossibly significant accessory constituents in somesherds. Sponge spicules are biosilicate needles or rodsthat formed the skeletal support for some freshwatersponges of the class Demospongiae, family Spongilli-dae (Borremans and Shaak 1986). In contrast toFlorida’s St. Johns pottery (Goggin 1952), with itsabundant sponge spicules (Borremans and Shaak 1986),the frequency of sponge spicules in the Kolomokisamples is relatively low. Opal phytoliths are botanicalmicrofossils composed of silica (Rapp and Mulholland1992). These microfossils are considered natural con-stituents of the clay sources rather than incidentaltempers or contaminants. They have been observedtogether in other pottery assemblages from Georgiaand Florida and in a few coastal clays (Cordell 2006;Cordell and Deagan 2011; Wallis 2011; Wallis andCordell 2011). Phytoliths and fragmented spongespicules were observed only in thin section, at amagnification of 2503, but are considered significantfor distinguishing resource groupings.

Results: Pottery Groupings

The pottery samples may be divided into variousgross paste groupings on the basis of three criteria:

particle size, paste constituents, and relative iron oxidecontent. Particle size groupings may define textural,perhaps temper, groupings, whereas certain pasteconstituents and relative iron oxide content defineresource groupings.

Particle Size/Texture. The pottery sample was sortedinto four gross textural groupings on the basis ofparticle size during the gross paste analysis: very finesand, fine sand, fine-medium sand or sand/grit, andgrit (Table 3). Fine sand and sand/grit categories aremost abundant, making up 50 percent and 31 percentof the sample, respectively. Very fine sand– and grit-tempered sherds make up 13 percent and 6 percent ofthe sample, respectively. One sherd is apparentlytemperless in terms of quartz or other crystallineconstituents but does contain frequent clay lumps (ithad been categorized as limestone-tempered in theoriginal analysis). Point count data provide quantifica-tion for these textural categories (also included inTable 3), which are distinguished graphically in Figure 3.There are statistically significant differences betweentextural categories in terms of mean percentages of sandoccurring in very fine through very coarse particle sizes(see Table 3 for test statistics). The textural distinctionswithin the sample become blurred when data on bulkcomposition (percentages of matrix, sand, and otherconstituents) are considered (Figure 4).

The grit-tempered sherds and temperless sherdoccurred primarily in off-mound contexts. Grit tem-pering occurs only in the plain and complicatedstamped wares and in the four Lamar period sherds.Very fine sand and fine sand categories are modal in

Table 3. Bulk composition and sand particle abundance/size by gross paste texture for sherd samples and Kolomokiclay samples.

Pastetexture

Totalsamplen (%)

Thin-sectionsample (n)

Bulk composition

% silt

Mean % sand particle sizes SSI

% matrix % sanda % otherbVeryfine Fine Medium

Coarse–very coarse Mean Range

Very finesand

35 (13) 10 67 26 12 5 14 10 2 ,1 0.86 0.72–0.96

Fine sand 134 (50) 35 61 34 10 3 10 15 8 2 1.20 1.03–1.37Sand/grit 85 (31) 17 63 33 8 3 6 11 10 4 1.51 1.40–1.66Grit 15 (6) 2 62 34 26 2 5 8 10 12 2.00 1.98–2.03Temperless 1 (,1) 1 90 10 N/ATotal 270 65

Clay samples

9ER11 30 70 4 1 2 2 14 51 2.86 2.86

9ER21 22 78 6 1 7 38 25 9 1.52 1.52

9ER31 41 56 6 1 7 31 16 2 1.29 1.29

9ER41 72 26 3 3 4 10 8 4 1.58 1.58

a Sand counts include quartz, polycrystalline quartz, and other non-opaque minerals, excluding micas.

b Other includes all other counted constituents.

Note: t-tests comparing texture categories in terms of mean sand particle sizes yielded statistically significant values at p ranging ,0.0001 to 0.002 for the followingcomparisons: very fine sand vs. fine sand (very fine, fine, medium, coarse–very coarse); very fine sand vs. sand/grit (very fine, medium, coarse–very coarse); veryfine sand vs. grit (very fine, medium, coarse–very coarse); fine sand vs. sand/grit (very fine, fine, medium, coarse–very coarse); sand/grit vs. grit (coarse–verycoarse).


294

some prestige types, whereas fine sand and sand/gritcategories are modal in utilitarian types (Table 4).Graphical comparison of sand particle sizes (Figure 5)shows considerable overlap for sacred, prestige, andutilitarian categories, but utilitarian samples showgreater variability, as might be expected. A similarpattern is evident when the functional categories arecompared in terms of bulk composition (Figure 6).

Other Paste Constituents. Of the principle pasteconstituents, mica, or rather its relative frequency, isconsidered a principal criterion for distinguishingresource groupings in the pottery sample. Forty-onepercent of the sample (n 5 110) is made of micaceousclays, with the relative frequency of mica flecks rangingfrom frequent to common. Mica occurring in theserelative frequencies is detectable on clean, unerodedsherd surfaces with low magnification and often evenwith the unaided eye. Petrographically, the percentagesrange from 3 to 7 percent and mica particle sizes tend tobe relatively large, occurring in fine to mediumWentworth sizes.

Mica is present but only rare or occasional in most ofthe remaining 59 percent of the sample (n 5 160).Occasional mica is sometimes detectable with lowmagnification, but trace presence is most often detect-able only in thin section. The petrographically estimat-ed percentage in most cases is about 1 percent, andmica particle sizes tend to range smaller, occurring invery fine to fine Wentworth sizes. Some degree ofoverlap between micaceous and low mica groupingsmay be possible, given the probable variability in micafrequency within clay deposits.

In the present sample, there is an inverse relationshipbetween mica frequency and gross texture (Figure 7).Most sherds with very fine texture are also micaceous.Mica is generally rare in grit-tempered sherds. Finesand and sand-grit textures are not significantlydifferent, but micaceous pastes are somewhat more

common in fine sand textures, whereas low mica pastesare somewhat more common in sand/grit textures. Thedifferences between very fine texture and the othertextures in terms of mica frequency were found to bestatistically significant (X2 5 9.59 df 5 1 p 5 0.002 vs.fine sand; X2 5 12.48 df 5 1, p50.0004 vs. sand/grit;X2 5 12.02 df 5 1 p 5 0.0005 vs. grit).

For pottery types with ten or more samples, there areno statistically significant tendencies in terms of micafrequency (Table 5). Micaceous paste comprises 35–50 percent of all such types. Prestige, sacred, andutilitarian categories are also basically similar in termsof micaceous versus low mica resources (see Table 5).The relative abundance of both micaceous and lowmica sherds in the assemblage indicates both categoriesof resources are probably local to the Kolomokivicinity.

Gross compositional variability for the thin sectionsubsample is described in Table 6. Very fine texture isdistinguished from fine sand and sand/grit categoriesby lower mean percentages of monocrystalline quartzand higher mean percentages of micas. Fine sand andsand/grit textures are very similar in mean percentagesof quartz and other constituents. Grit-tempered pastediffers from the others in terms of having much highermean percentages of quartzite and other crystallinegrains (especially feldspars and amphibole), lowermean percentage of monocrystalline quartz, and lowerdiversity in accessory minerals (Table 6).

Graphical comparison of the relative percentages ofmonocrystalline quartz, polycrystalline quartz, andother crystalline grains (Figure 8) shows considerableoverlap between very fine sand and fine sand textures,and between fine sand and sand/grit textures. Thesand/grit grouping shows less relative compositionalheterogeneity than the former two textures. The grit-tempered cases are easily distinguished from the rest in

Figure 3. Ternary graph of sand particle size variation forgross textural categories (percentages within particle sizecategories were calculated with respect to total sand counts).

Figure 4. Ternary graph of variation in bulk composition forgross textural categories (‘‘other’’ constituents are based oncounts of silt, ferric concretions, clay lumps, micas, and othernonsand constituents).


295

Figure 8. The grit-tempered cases also contain frequentgranitic grains, in contrast to the other samples (seeTable 6). For these cases, granitic grains and amphibolemay indicate a rare (to Kolomoki) temper choice orperhaps that this pottery is nonlocal to Kolomoki.6 Thisis supported by our analyses of local clay samples, inwhich granitic rock fragments were not present andamphibole was only a rare constituent (see laterdiscussion). Compositional percentages for individualpottery types vary among and between functionalcategories, but samples sizes are too small to proposeany trends. Mean percentages between functional

categories are fairly similar, but comparison of relativepercentages of monocrystalline quartz, polycrystallinequartz, and other crystalline grains (Figure 9) showsthat utilitarian samples exhibit somewhat greater com-positional heterogeneity than prestige and sacredsamples.

The presence/frequency of siliceous microfossilsformed the basis for defining two gross clay resourcegroupings for the thin section sample, Clay A and B(Table 7). Clay A (n 5 29, 45 percent of cases) ischaracterized by the general absence of siliceousmicrofossils: no sponge spicules and the absence or

Figure 5. Ternary graph of sand particle size variation forfunctional categories (percentages within particle size cate-gories were calculated with respect to total sand counts).

Figure 6. Ternary graph of variation in bulk composition forfunctional categories (‘‘other’’ constituents are based oncounts of silt, ferric concretions, clay lumps, micas, and othernonsand constituents).

Table 4. Count (and percentage) gross texture by pottery type and functional category for sherd samples and Kolomokiclay samples.

Very fine sand Fine sand Sand/grit Grit Untempered Total

By pottery type (sherd samples)

Weeden Island Incised 5 (26) 12 (63) 2 (11) - - 19Weeden Island Plain 2 (33) 3 (50) 1 (17) - - 6Weeden Island Red 1 (3) 14 (44) 17 (53) - - 32Weeden Island Zoned Red 1 (25) 2 (75) - - - 4Mercier Red on Buff - 2 (100) - - - 2Indian Pass Incised - 1 (50) 1 (50) - - 2Carrabelle Incised 1 (50) 1 (50) - - - 2Carrabelle Punctated 1 (7) 7 (50) 6 (43) - - 14Plain 13 (14) 44 (47) 28 (30) 7 (8) 1 (1) 93Swift Creek Complicated Stamped 10 (11) 46 (52) 28 (32) 4 (5) - 88Blakely Complicated Stamped 1 (50) - 1 (50) - - 2Napier Complicated Stamped - - 1 (100) - - 1Mound Field Net Marked - 1 (100) - - - 1Lamar Complicated Stamped - - - 4 (100) - 4

By functional category (sherd samples)

Prestige 7 (15) 28 (58) 13 (27) - - 48Sacred 1 (8) 5 (42) 6 (50) - - 12Utilitarian 27 (13) 101 (49) 66 (32) 11 (5) 1 (,1) 206Other (Lamar) - - - 4 (100) - 4Total sherd sample 35 (13) 134 (50) 85 (32) 15 (6) 1 (,1) 270

Kolomoki clay samples

9ER1 X9ER2 X9ER3 (calcareous) X X9ER4 X


296

extreme paucity of phytoliths. Clay B (n 5 36, 55 percentof cases) is characterized by occasional but consistentoccurrence of siliceous microfossils: sponge spiculesand/or phytoliths. Their differential presence mayindicate different environments of deposition for theClays A and B, perhaps lacustrine for Clay B, although itmight be possible that naturally occurring variabilitywithin a given clay sample could explain such variabil-ity in the constituents. Clay B is only slightly morecommon in the sample than Clay A. This indicates bothclay categories were probably local to the Kolomokivicinity and were more or less equally accessible toKolomoki potters, an assumption corroborated by ouranalysis of clay samples (described below).

When compared to gross paste textures, Clay Aoccurs in all textural categories (Table 7). Clay B occursin very fine through sand/grit categories but has nocases with grit-tempered paste. Clay A is morecommon in very fine and grit textures, whereas ClayB is more common in fine sand and sand/grit textures.However, the tendencies are not statistically signifi-cant at 95 percent confidence levels (X2 5 6.69, df 5 3,p 5 0.0825).

In the present study, Clay A and Clay B occur in bothmicaceous and low mica categories, but there is asomewhat inverse relationship between mica frequen-cy and presence of siliceous microfossils (see Table 7).Fifty-seven percent (n 5 17) of Clay A samples aremicaceous, whereas 66 percent (n 5 23) of Clay Bsamples have low mica. The differences are statisticallysignificant (X2 5 6.63, df 5 1, p 5 0.01). Prestige andsacred types generally have a higher percentage ofcases made with Clay B paste, whereas utilitarian typeshave a higher percentage made with Clay A paste.However the trend is not statistically significant in thismore limited sample (with prestige/sacred combinedX2 5 1.78, df 5 1, p 5 0.1821).

Relative Iron Oxide Content. Refiring indicates that 41percent (n 5 108) of the total sample is made of claysrelatively low in iron oxides (Figure 10). Clays withmoderate and high iron oxides are also common in the

sample, whereas clays with very low iron oxide contentare relatively rare. The relative abundance of low,moderate, and high iron oxide categories in theassemblage likely indicates variability in iron oxidecontent in the resources local to the Kolomoki vicinity.This is supported by our analysis of local clays (seelater discussion). The low percentage of very low ironoxide content may indicate its relative scarceness in theregion or restricted use in pottery manufacture. Interms of clay resources, there is likely to be somedegree of overlap in relative iron oxide content withingiven clay deposits.

In terms of gross paste texture (total sample), low ironoxide clays are modal for all texture categories, but verylow iron oxide clays have higher representation in veryfine and grit pastes (Table 8). Fine sand and sand/gritcategories are very similar to each other, with lowthrough high iron oxide categories well represented.There is a slight tendency for micaceous samples to havehigher percentage with very low iron oxides, but thetendency is not statistically significant at 95 percentconfidence levels (X2 5 7.42, df 5 3, p 5 0.0597). In termsof clay groupings based on siliceous microfossils (thinsection sample), samples with very low iron oxides arerestricted to Clay A, lacking in microfossils (see Table 8).Otherwise, the groupings are not significantly different.

Low iron oxide clays are modal in all pottery typeswith greater than five cases except for Weeden IslandIncised (see Table 8). High iron oxide clays are modal

Table 5. Count (and percentage) of samples categorizes aslow mica and micaceous by pottery type and functionalcategory for sherd samples and Kolomoki clay samples.

Low mica Micaceous Total

By cultural historical type (sherd sample)

Weeden Island Incised 11 (58) 8 (42) 19Weeden Island Plain 3 (50) 3 (50) 6Weeden Island Red 21 (66) 11 (34) 32Weeden Island Zoned Red 3 (75) 1 (25) 4Mercier Red on Buff - 2 (100) 2Indian Pass Incised 2 (100) - 2Carrabelle Incised 2 (100) - 2Carrabelle Punctated 8 (57) 6 (43) 14Plain 60 (64) 33 (36) 93Swift Creek ComplicatedStamped

44 (50) 44 (50) 88

Blakely Complicated Stamped 2 (100) 2Napier Complicated Stamped 1 (100) - 1Mound Field Net Marked 1 (100) - 1Lamar Complicated Stamped 4 (100) - 4

By functional category (sherd sample)

Prestige 30 (62) 18 (38) 48Sacred 8 (67) 4 (33) 12Utilitarian 118 (57) 88 (43) 206Other (Lamar) 4 (100) - 4Total sherd sample 160 (59) 110 (41) 270


9ER1 X9ER2 X9ER3 calcareous X9ER4 X

Figure 7. Relationship between mica frequency andgross texture.


297

in this type. Very low iron oxide clays occur in a smallpercentage of Weeden Island Incised (prestige type)and utilitarian types. Very low iron oxide clays areabsent in the sacred samples, but occur in all Lamarsherds in the sample.

Results: Clay Resource Groupings

Relative iron oxide content, mica frequency, andpresence of siliceous microfossils represent attributesuseful for defining clay resource groupings in theKolomoki sample (thin section sample). Refired colorand mica frequency data alone indicate a minimum offour main potential clay resources. Low mica andmicaceous wares each occur in relatively low andmoderate to high iron oxide groupings. Particle size/textural differences within the four groupings are mostlikely explained in terms of different tempering recipesor the natural variability in sand constituents within

discrete clay resources. This estimate of four is doubledwhen siliceous microfossils are considered, resulting ina minimum of eight main clay resources (Table 9 andFigure 11). Four are characterized by the absence ofsiliceous microfossils, of which two have low micacontent and relatively low and high, respectively, ironoxide content, and two are micaceous with relativelylow and high, respectively, iron oxide content. Four arecharacterized by the presence of noticeable siliceousmicrofossils, of which two have low mica content andrelatively low and high, respectively, iron oxide content,and two are micaceous with relatively low and high,respectively, iron oxide content. The eight clay resourcegroupings are hereafter referred to as matrix categoriesand are interpreted as representing particular claysources or groups of similar clays. Variation in percent-age of silt and very fine sand in these matrix groupingsmay be explained in terms of variability within the claysources. Variation in fine through larger sand may beattributed to tempering practices or even to variation in

Table 6. Compositional variability by gross paste texture, pottery type, and functional category for sherd samples andKolomoki clay samples.

Potterycategory n % sand

% mono-crystalline

quartz

% othercrystalline

grainsaEstimatedb

% mica% polyxquartz

Estimatedb

% totalfeldspars

Estimatedb

% mafic& heavyminerals

Heavymineral

species (n)Presence (P) of

granitic rock

By texture (thin-section sample)

Very fine sand 10 26 21 6 5 1 3 2 7 P in 1/10Fine sand 35 34 29 6 3 2 4 2 8 P in 10/35Sand/grit 17 33 29 4 2 2 3 1 6 P in 6/17Grit 2 34 12 22 - 7 10 5 4 3–4%Temperless 1 - - - - - - - - -

By cultural historical type (thin-section sample)

Weeden IslandIncised

9 34 30 5 3 2 3 1 7 P in 2/9

Weeden IslandPlain

4 34 31 5 3 2 3 1 8 -

Weeden IslandRed

19 32 28 5 3 2 3 1–2 7 P in 7/19

Weeden IslandZoned Red

3 33 26 9 2 2 6 1 7 P in 1/3

Mercier Red onBuff

1 30 23 9 4 3 4 1 10

CarrabellePunctated

4 32.5 28 6 3 2 3 1 8 P in 2/4

Plain 10 29 26 4 3 1 3 1–2 6 P in 2/11Swift Creek

Comp. St.13 32 24 6, 22 c 3 3 4 1,4–5c 6 P to 4% in 5/13

Blakely Comp.St.

1 25 19 9 7 1 4 1 6 P

Napier Comp.St.

1 30 28 1 1 1 2 1 6

By functional category (thin-section sample)

Prestige 26 33 28 6 3 2 4 2 7 P in 6/26Sacred 8 32 29 4.5 2 2 3 1 6 P in 4/8Utilitarian 31 31 25 6 3 2 4 2 6 P to 4% in 10/31


9ER1 1 70 66 4 - 4 - - 19ER2 1 78 72 7 1 6 1 1 49ER3 1 56 53 4 1 3 1 P 69ER4 1 26 25 1 1 1 P 1 8

a Percentage is based on point count data and may be a bit low in some cases.

b Point count data underrepresented percentages for some constituents; estimates were made using a comparison chart of estimated percent particle abundance(Rice 1987:Figure 12.2). This accounts for sum of % of polycrystalline quartz and estimated percentages of mica, feldspars, mafic/heavies exceeding thepercentage listed for % other crystalline grains.

c Two grit-tempered cases contain significantly more mafic/heavy constituents than any other pottery in the thin-section sample.


298

naturally occurring sand. Five sherds in the thin sectionsample are outliers (Table 9).

The thin section sample is quite variable in terms ofmatrix categories (Figure 12). Matrix B, characterized bylow mica, low iron oxides, and presence of siliceousmicrofossils, is the modal grouping but accounts foronly 25 percent of the samples. Five of the eight mainmatrix groupings are more or less equally represented,each accounting for 9 to 15 percent of the cases. Matrix Cand Matrix H each account for 5 percent of the sample.

The eight matrix categories are extremely variable interms of texture (Table 10). Fine sand and sand/grittextures occur in all or most of the matrix groupings.Very fine textures occur only in micaceous matrixcategories. Grit paste is represented by only two cases,but both are outliers. Individual pottery types alsoshow considerable variability in matrix categories,despite, or perhaps owing to, small sample size (seeTable 10). There are no striking differences betweenprestige, sacred, and utilitarian categories in terms ofmatrix paste.

In sum, the results of our analysis show that a varietyof resources and perhaps tempering recipes wereacceptable and used routinely in making pottery atKolomoki. Although the differential presence of sili-ceous microfossils indicates clay resource differences,the four matrix categories with siliceous microfossilsare still outwardly very similar to the four matrixcategories without siliceous microfossils. Perhaps otheranalytical approaches, such as NAA, can be used tocorroborate the differences between matrix categories.We turn our focus now to the analysis of the four claysamples to determine the range of variability in mica,sand, siliceous microfossils, and other naturally presentaplastics, as well as relative iron oxide content, of clays

in the vicinity of Kolomoki. Comparison to the fourlocal clays may provide substantiation of implicationsof local versus nonlocal manufacture for certainaplastics and textural categories.

Sample Clay Analyses and Comparison to PotteryMatrix Categories

Four clay samples were collected from the vicinity ofKolomoki for comparison to the pottery samples. ClaySample 9ER1 was collected from the toe of a gradedand eroded ridge slope about 1.5 km east-southeast ofthe Kolomoki site. Clay Samples 9ER2 and 9ER3 wereessentially from the site itself; these clays wereretrieved from separate strata exposed in the bank ofLittle Kolomoki Creek, which adjoins the site to theeast. Finally, Clay Sample 9ER4 was collected from astream bank in the nearby town of Blakely, about 9 kmsouth of Kolomoki.

The sample clays were processed and analyzed todetermine plasticity, shrinkage, and firing behavior,particle size and proportion (grain size analysis), andcomposition. Grain size analyses were conducted andtest bars were formed and fired for gross comparison ofphysical properties (data on file, FLMNH-CTL). Firedsamples were thin-sectioned for petrographic analysis.

The physical properties of the clay samples werecompared to the excavated pottery samples. Compar-isons were made in terms the same properties used tocharacterize the matrix paste categories, that is,composition and relative iron oxide content. Compar-isons of constituent abundance are based on a pointcount data. The results are included in Tables 3–8 andsummarized in Table 9.

Figure 8. Ternary graph of compositional variation in sandconstituents for gross textural categories (percentages ofquartz, polycrystalline quartz, and other crystalline grainswere calculated with respect to total sand plus micas. ‘‘Othercrystalline grains’’ include feldspars, mafic minerals, othernon-opaque minerals, and micas).

Figure 9. Ternary graph of compositional variation in sandconstituents for functional categories (percentages of quartz,polycrystalline quartz, and other crystalline grains werecalculated with respect to total sand plus micas. ‘‘Othercrystalline grains’’ include feldspars, mafic minerals, othernon-opaque minerals, and micas).


299

Quartz sand was the predominant constituent in threeof the four samples. Polycrystalline quartz or quartzitewas also occasionally observed. Grit, or coarse and verycoarse particle sizes, were quite common in the clays butcomposed predominantly of quartz, with occasionalquartzite compositions. No granitic rock fragments wereobserved. Particle size variability is within the range ofsome sand/grit and fine sand pottery samples, as shownin Figure 3. Particle shape was predominantly suban-gular to angular and subrounded in three of the foursamples, much like particle shape in the potterysamples. Subrounded grains predominated in onesample, 9ER1. In terms of bulk composition (Figure 4),excessive sand constituents make three of the foursamples unlikely matches to pottery samples unlessmanipulated to remove some of the sand. The predom-inance of quartz in the clay samples (Table 6) is alsoreflected in Figure 8, showing relative proportions ofdifferent sand constituents.

Mica appeared to be occasional to frequent in thesieved sediments (from grain size analysis) of three ofthe clay samples but was only rare to occasional in thinsection. This may indicate variability within the claydeposits, but for these samples, mica occurrence

appears to be low. Ferric nodules, feldspars, maficminerals such as amphibole and epidote are rare tooccasional constituents, usually in very fine particlessizes, consistent with observations in the potterysamples. Siliceous microfossils, sponge spicules, andphytoliths were occasional constituents in one of thesamples, also characterized by low mica and low ironoxides. Clay Sample 9ER3 has a calcareous matrixand contains decomposed shell fragments along withquartz sand and occasional mica. Its calcareous com-position excludes it as a possible match to the potterymatrix categories. Fired colors indicate the clay samplesinclude examples of low, moderate, and high ironoxide clays (Table 8).

Most of the clay samples share some physicalproperties (color, constituents) with matrix categoriesdesignated for the pottery samples. Clay Sample 9ER1 ischaracterized by high iron oxides, low mica, and theabsence of siliceous microfossils, traits correspondingto Matrix C. However, excessive abundance of sub-rounded coarse and very coarse (grit sizes) lessen thelikelihood of the match. Clay Sample 9ER2 is character-ized by moderate iron oxide content, low mica, and theabsence of siliceous microfossils; except for excessivesand constituents, it shares traits with some members ofpottery Matrix A. Clay sample 9ER4 is characterized bylow iron oxides, low mica, and the presence of siliceousmicrofossils and sand is not excessive. It represents afeasible match to many of the Kolomoki thin sections,specifically to pottery with Matrix B paste.

Thus one of the clay samples (9ER4) matches themost abundant matrix category (Matrix B) within thecontext of the present thin-sectioned Kolomoki potterysample. The analyses confirm that constituents that areprominent in the pottery samples occur locally. Oursuggestion that micaceous and non-micaceous claysources are locally available is, therefore, at leastpartially correct. It is curious that none of the samplescould be considered ‘‘micaceous.’’ This does notnecessarily indicate the absence of local micaceous

Table 7. Percentage of clay categories by gross paste texture,mica content, and pottery type for sherd samples andKolomoki clay samples.

Clay A(siliceous

microfossilsabsent)

Clay B(siliceous

microfossilspresent) Total


Very fine sand 7 (70) 3 (30) 10Fine sand 12 (34) 23 (66) 35Sand/grit 7 (41) 10 (59) 17Grit 2 (100) - 2Temperless 1 (100) - 1

By mica grouping (thin-section sample)

Low mica (n 5 35) 12 (34) 23 (66) 35Micaceous (n 5 30) 17 (57) 13 (43) 30


Weeden Island Incised 2 (22) 7 (78) 9Weeden Island Plain 1 (25) 3 (75) 4Weeden Island Red 6 (32) 13 (68) 19Weeden Island Zoned Red 1 (33) 2 (67) 3Mercier Red on Buff 1 (100) - 1Carrabelle Punctated - 4 (100) 4Plain 7 (70) 2 (30) 10Swift Creek Complicated

Stamped 9 (69) 4 (31) 13Blakely Complicated Stamped 1 (100) - 1Napier Complicated Stamped 1 (100) - 1


Prestige 9 (35) 17 (65) 26Sacred 3 (38) 5 (62) 8Utilitarian 17 (55) 14 (45) 31Total thin-section sample 29 (45) 36 (55) 65



Figure 10. Relative iron content of pottery samples.


300

clays in the vicinity of Kolomoki. Instead, the absenceof micaceous clays is likely an artifact of the smallnumber of samples collected. Variability along hori-zontal or vertical dimensions within the clay depositsmight yield sources of more micaceous deposits.

The differential presence/absence of siliceous mi-crofossils in the clays samples supports our earliersuggestion that clays with and without these constit-uents would be locally available. The absence ofgranitic rock fragments and frequent amphibole corro-borate our earlier suggestion that these are rare(to Kolomoki) temper choices, or perhaps that theyrepresent tempers or clay sources that are nonlocalto the vicinity of Kolomoki (most likely from thePiedmont region). The grit-tempered sherds in thethin section sample, which are Swift Creek Compli-cated Stamped, may be similar in composition to somegrit-tempered Swift Creek pottery from the Hartfordsite (Stoltman and Snow 1998) or other sites with

pottery manufacturing origins closer to the Piedmont.Further investigation of clay sources in the vicinity ofKolomoki may document the occurrence of other claysthat are consistent with more of the pottery in terms ofparticle size and composition.

Weeden Island Manufacturing Origins and Exchange:Comparing Paste Groups from Kolomoki and Other

Weeden Island Sites

With our initial goal of characterizing the range ofvariability in paste/resource groupings present in theKolomoki assemblage accomplished, we turn ourattention to the question of patterns of exchange andspecialization in production of Weeden Island pot-tery. This task requires comparing Kolomoki data topetrographic data for Weeden Island pottery fromother sites in northern and northwestern Florida. The

Table 8. Count (and percentage) of samples categorized to relative iron oxide content by gross paste texture, clay category,pottery type, and functional category for sherd samples and Kolomoki clay samples.

Very low iron oxides Low iron oxides Moderate iron oxides High iron oxides Total refired/total samplea

By texture (sherd sample)

Very fine sand 8 (24) 15 (45) 4 (12) 6 (18) 33/35Fine sand 7 (5) 55 (42) 32 (24) 37 (28) 131/134Sand/grit 3 (4) 31 (37) 21 (25) 29 (34) 84/85Grit 4 (27) 7 (47) 2 (13) 2 (13) 15Temperless - - - 1 (100) 1

By mica grouping (sherd sample)

Low mica 8 (5) 61 (39) 39 (25) 49 (31) 157/160Micaceous 14 (13) 47 (44) 20 (19) 26 (24) 107/110Total (n 5 264/270) 22 (8) 108 (41) 59 (22) 75 (28) 264a /270

By clay category (thin-section sample)

Clay A (siliceous microfossils absent) 4 (14) 9 (31) 7 (24) 9 (31) 29Clay B (siliceous microfossils present) - 18 (53) 7 (21) 9 (26) 34/36Total (n 5 63)b 4 (6) 27 (43) 14 (22) 18 (29) 63/65

By pottery type (sherd sample)

Weeden Island Incised 2 (10) 4 (21) 3 (16) 10 (53) 19Weeden Island Plain - 3 (50) 1 (17) 2 (33) 6Weeden Island Red - 14 (47) 9 (30) 7 (23) 30/32Weeden Island Zoned Red - 3 (75) 1 (25) - 4Mercier Red on Buff - 2 (100) - - 2Indian Pass Incised - 2 (100) - - 2Carrabelle Incised - - 1 (50) 1 (50) 2Carrabelle Punctated 1 (10) 4 (40) 3 (30) 2 (20) 10/14Plain 6 (6) 40 (43) 23 (25) 24 (26) 93Swift Creek Complicated Stamped 8 (9) 36 (41) 18 (20) 26 (30) 88Blakely Complicated Stamped 1 (50) - - 1 (50) 2Napier Complicated Stamped - - - 1 (100) 1Mound Field Net Marked - - - 1 (100) 1Lamar Complicated Stamped 4 (100) - - 4

By functional category (sherd sample)

Prestige (n 5 46) 2 (4) 19 (41) 11 (24) 14 (30) 46/48Sacred (n 5 12) - 4 (33) 4 (33) 4 (33) 12Utilitarian (n 5 202) 16 (8) 85 (42) 44 (22) 57 (28) 202/206Other (Lamar) (n 5 4) 4 (100) - - - 4Total sherd sample 22 (8) 108 (41) 59 (22) 75 (28) 264a /270



a Six sherds too small to spare removal of pieces for refiring.

b Two cases in thin-section sample too small to spare removal of pieces for refiring.


301

data currently available for comparison are limited to22 thin sections from the McKeithen site and 20 thinsections from several other sites in the Weeden Islandregion (Cordell 2006).

The question of manufacturing origins of WeedenIsland series pottery from McKeithen has been previ-ously addressed through standard microscopic com-positional analysis, coupled with refiring to assess rela-tive iron oxide content (Cordell 1983, 1984; Milanichet al. 1997; Rice and Cordell 1986). These studiesfocused on a sample of 33 vessels (or partial vessels)from Mounds B and C and 212 sherds from thesurrounding contemporaneous midden. The moundsample was made up of prestige types primarily, ofwhich at least nine are sacred wares. The middensample included 47 sherds representing prestige typesand 165 representing utilitarian types (Cordell 1984:Tables 3-1, 3-2). Cluster analysis was used to sort the

samples into groupings on the basis of similarity in anumber of paste attributes including type, frequencyand size of temper or aplastic constituents, and relativeiron oxide content (based on refired paste colors). Localand nonlocal paste categories were identified on thebasis of direct comparison to several local clay samplesand conventional wisdom regarding recognition oftradewares in peninsular Florida. Northwestern Florida,southwestern Georgia, and southeastern Alabama makeup the presumed manufacturing source region formicaceous paste pottery (Cordell 1983:86–87; 1984:64,199; also see Bullen 1971:8; Mitchem 1986:68–69; Sears1973:33). Clusters 1 and 3, with frequent to abundantmica, were thus interpreted as nonlocal to McKeithen(Cordell 1984:156). Cluster 4, with occasional mica, wasinterpreted as having uncertain manufacturing origins(Cordell 1984:156).

Manufacturing origins of Weeden Island pottery atthe McKeithen site have also been investigated throughtrace element analysis (Rice 1980). Rice concluded thatWeeden Island Incised and Weeden Island Zoned Red

Table 9. Description of matrix categories and clay samples.

n Relative iron oxides Mica Siliceous microfossils Texture Similar categories

Matrix category (thin-section sample)

A 6 Low to moderate Low Absent Fine sand, sand/grit Otherwise like BB 16 Low to moderate Low Present Fine sand, sand/grit Otherwise like AC 3 High Low Absent Fine sand, sand/grit Otherwise like DD 7 High Low Present Very fine sand, fine sand Otherwise like CE 9 Very low to low High Absent Very fine sand, fine sand Otherwise like FF 10 Low High Present Very fine sand, fine sand, sand/grit Otherwise like EG 6 High High Absent Very fine sand, fine sand, sand/grit Otherwise like HH 3 High High Present Fine sand, sand/grit Otherwise like GI (outlier) 1 High High Absent Sand/grit Otherwise like GJ (outlier) 1 High Low Absent Temperless Otherwise like CK (outlier) 2 Low Low Absent Grit Otherwise like AL (outlier) 1 Low High Absent Silty to very fine sand More silt, vf sand E


9ER1 High Low Absent Grit Similar to matrix C9ER2 Moderate Low Absent Sand/grit Similar to matrix A/C9ER3 Low Low Absent Fsand to sand/grit Calcareous paste

precludes anymatches

9ER4 Low Low Present Sand/grit Matches matrix B

Figure 12. Relative frequencies of matrix categories.

Figure 11. Diagram illustrating constitution of matrixcategories.


302

pottery types were traded into McKeithen (Rice1980:33). At McKeithen, these types were most fre-quently made of nonlocal micaceous clays (Cordell1984:162; Milanich et al. 1997:123). Rice proposed amulticenter model of manufacture for pottery typicallyfound in ceremonial or sacred contexts. Cordell’s data,which incorporated many of the sherds and clays usedin Rice’s study, supported Rice’s multicenter model ofpottery production, but with the McKeithen site beingone of the centers (Cordell 1984:162–164).

Until recently, neither Rice’s nor Cordell’s conclu-sions had been corroborated/tested petrographically.A limited sample of thin sections was obtained fromMcKeithen pottery and pottery from several sites innorthwestern Florida (Cordell 2006). This comparativesample consisted of 42 thin sections (Table 11). Twenty-two are McKeithen samples, of which six representnonlocal (Clusters 1 and 3) or potentially nonlocal(Cluster 4) pastes. Seventeen thin sections are fromseveral northwestern Florida sites, including familiarWeeden Island period sites such as Tucker, Carrabelle,Mound Field, and Hall. Finally, three thin sections arefrom non-Florida sites: one sherd from Kolomoki(which was not a part of the present study) and twofrom the Mitchell site in southeastern Alabama.Mitchell was a Weeden Island complex similar in sizeand scale to McKeithen (Earnest 1995; Sears 1959).Prestige types such as Weeden Island series Incised,

Zoned Red, and Red are the principal decorated typesin the thin section sample.

The results of the petrographic analysis led to changesin McKeithen cluster composition and definitions.Clusters 1 and 3 proved to be very similar in terms ofmany criteria and siliceous microfossils were observedin some members of each. The two clusters weretherefore combined and the whole subdivided intothree groupings on the basis of presence/absence andtype of siliceous microfossil (Table 12). Cluster 1a ischaracterized by the presence of sponge spicules andphytoliths, whereas Cluster 1b is characterized by theabsence of siliceous microfossils. Cluster 1c has not onlysponge spicules and phytoliths but also fragmentarydiatoms, unicellular algae with ornate cell walls made ofsilica. Despite these important differences, which indi-cate that Clusters 1a–1c each represent a separateresource grouping, the three groupings are otherwiserelatively homogeneous in mica frequency, sand particlesize and frequency, and relative iron oxide content.

Several thin sections from northwestern Florida siteswere assigned to McKeithen Clusters 1a or 1b. Inaddition, one thin section from the Mitchell site inAlabama was assigned to Cluster 1b. Cluster 1c wasnot observed in the McKeithen sample. It is insteadrepresented by a northwestern Florida site.

Petrographic analysis revealed that McKeithen Clus-ter 4 may be a less micaceous version of Cluster 1a,

Table 10. Count (and percentage) of matrix categories by gross paste texture, pottery type, and functional category for sherdsamples and Kolomoki clay samples.

A B C D E F G H Outliers Total


Very fine sand - - - - 4 (40) 3 (30) 2 (20) - 1 (10) 10Fine sand 3 (9) 11 (31) 1 (3) 6 (17) 5 (14) 4 (11) 3 (9) 2 (6) - 35Sand/grit 3 (18) 5 (29) 2 (12) 1 (6) - 3 (18) 1 (6) 1 (6) 1 (6) 17Grit - - - - - - - - 2 (100) 2Temperless - - - - - - - - 1 (100) 1


Weeden Island Incised - 2 (22) - 4 (44) 1 (11) 1 (11) 1 (11) - - 9Weeden Island Plain - 1 (25) 1 (25) - - 1 (25) - 1 (25) - 4Weeden Island Red 2 (10) 5 (26) 1 (5) 2 (10) 3 (16) 4 (21) 2 (10) - 19Weeden Island Zoned

Red- 2 (67) - - - - 1 (33) - - 3

Mercier Red on Buff - - - - 1 (100) - - - 1Carrabelle Punctated - 2 (50) - 1 (25) - 1 (25) - - - 4Plain 3 (30) 2 (20) 1 (10) - 1 (10) 1 (10) - - 3 (30) 10Swift Creek Complicated

Stamped1 (8) 2 (15) - - 2 (15) 2 (15) 4 (31) - 2 (15) 13

Blakely ComplicatedStamped

- - - - 1 (100) - - - - 1

Napier ComplicatedStamped

- - 1 (100) - - - - - - 1


Prestige 2 (8) 7 (27) 1 (4) 5 (19) 4 (15) 4 (15) 2 (8) 1 (4) - 26Sacred 1 (12) 2 (25) 1 (12) 1 (12) 1 (12) 1 (12) - 1 (12) - 8Utilitarian 3 (10) 7 (23) 1 (3) 1 (3) 4 (13) 5 (16) 4 (13) 1 (3) 5 (16) 31Total (thin-section sample) 6 (9) 16 (24) 3 (5) 7 (11) 9 (14) 10 (15) 6 (9) 3 (5) 5 (8) 65


9ER1 X?9ER2 X? X?9ER3 calcareous X9ER4 X


303

with siliceous microfossils. Three thin sections from anequal number of sites in northwestern Florida wereassigned to this McKeithen cluster.

Two new micaceous groupings, reassigned Cluster 3and newly assigned Cluster 6, were also defined fromthe inclusion of thin sections from other sites. NewCluster 3 paste is characterized by common mica, lowiron oxides, and paucity of siliceous microfossils. NewCluster 6 paste is characterized by common minutemica and frequent ferric nodules.

For clusters with four or more examples (Clusters 1a,1b and 4), the groupings include samples from severaldifferent sites across the Weeden Island region. Inaddition to McKeithen, Cluster 1a includes samplesfrom four sites in northwestern Florida. Cluster 1bincludes samples from McKeithen and two sites fromnorthwestern Florida, plus the Mitchell site. Cluster 4has examples from McKeithen and three sites innorthwestern Florida.

Although sample sizes are very small, the pastesexhibit widespread geographic distribution. One possi-ble interpretation of widespread geographic distributionis that it indicates widespread manufacture of potterywith these micaceous pastes across the Weeden Islandregion. This would presuppose similarity in micaceousclays across the region. Another explanation is thatpottery with these pastes were widely distributed orexchanged across the region from one Weeden Islandcenter, or at least a very limited number of WeedenIsland centers. The latter explanation seems feasiblewhen the present Kolomoki data are considered.

The petrographic descriptions for most of theseclusters are strikingly similar to some Kolomoki matrixdata (see Table 12). Clusters 1a and 1b correspond to

Kolomoki Matrix categories H and F, respectively,characterized by relatively high iron oxides and thepresence (Cluster 1a, Matrix H) or absence (Cluster 1B,Matrix F) of siliceous microfossils. The two members ofnew Cluster 3 grouping, one of which is actually fromKolomoki, are remarkably similar to Kolomoki MatrixE. Finally, Cluster 4 has counterparts in KolomokiMatrix groupings D or H or both (mica frequency issomewhat intermediate between low mica Matrix Dand micaceous Matrix H).

This would seem to be strong evidence for Kolomokimanufacturing origins for pottery with these pastes.Cluster 1c, with fragmentary diatoms, is absent in theKolomoki sample but is similar to two clays that werecollected along the Altamaha River, Georgia, as part ofanother separate study (Wallis 2011; Wallis and Cordell2011). One sample is from Telfair County, borderingthe eastern edge of the Weeden Island region, and theother is from Wayne County, well east of the region.These clays are not necessarily matches to the potterysample with this paste, but they certainly do show thatcomparable clay sources exist. New Cluster 6, charac-terized by common minute mica and ferric nodulesalso does not have any counterpart in the Kolomokithin section sample. It may have western panhandle orsoutheastern Alabama manufacturing origins on thebasis of its presence in sites from these areas.

Summary and Interpretation

We can say with confidence that for McKeithen andother Weeden Island sites across the Gulf Coast, arelatively high proportion of the pottery previously

Table 11. Counts of pottery types represented in comparative thin-section sample from McKeithen and other sites in theWeeden Island region.

8CO17 8FR1 8FR2 8FR4 8LI1 8OK6 8OK239 8SA1 8WA4 8WA8 8WA34 8WA35 1CV32 9ER1 Total

WeedenIsland Incised

2 1 2 1 1 1 8

WeedenIsland ZonedRed

5 1 1 1 8

WeedenIsland Red

3 1 1 1 6

WeedenIsland Plain

1 1 1 1 4

Papys BayouPunctated

1 1

Indian PassIncised

1 1

CarrabelleIncised

1 1

CarrabellePunctated

3 3

Keith Incised 1 1Tucker RidgePinched

1 1

Unidentifiedincised

1 1

Plain 7 7Total 22 1 1 3 1 1 4 1 1 2 1 1 2 1 42


304

Tab

le12

.R

evis

edM

cKei

then

pas

tecl

ust

ers

and

pre

sum

edm

anu

fact

uri

ng

ori

gin

s(b

ased

on

this

anal

ysi

san

dC

ord

ell

2009

).

Rev

ised

McK

eith

enC

lust

er

Su

mm

ary

of

pet

rog

rap

hic

dat

a

Rel

ativ

eir

on

con

ten

t

Po

tter

yty

pes

rep

rese

nte

d(t

hin

-sec

tio

nsa

mp

le)

Sit

esre

pre

sen

ted

Rel

atio

nsh

ipto

oth

ercl

ust

ers

Pre

vio

usl

yas

sum

edm

anu

fact

uri

ng

ori

gin

Mat

chin

gm

atri

xp

aste

atK

olo

mo

ki

Up

dat

edm

anu

fact

uri

ng

ori

gin

Mic

a,o

ther

Sil

iceo

us

mic

rofo

ssil

s

1a(n

57)

Fre

qu

ent

toco

mm

on

mic

a(3

–5%

)

Rar

eto

occ

asio

nal

spo

ng

esp

icu

les

and

ph

yto

lith

s

Mo

der

ate

toh

igh

Wee

den

Isla

nd

Inci

sed

(n5

5)W

eed

enIs

lan

dP

lain

(n5

2)

8CO

17(n

52)

8FR

1(n

51)

8FR

2(n

51)

8FR

4(n

52)

8LI1

(n5

1)

Oth

erw

ise

sim

ilar

tom

icac

eou

scl

ust

ers

1ban

d1c

NW

Flo

rid

aM

atri

xH

Ko

lom

ok

i+

1b(n

54)

No

ne

tora

rep

hy

toli

ths

Mo

der

ate

toh

igh

Wee

den

Isla

nd

Inci

sed

(n5

1)W

eed

enIs

lan

dZ

on

edR

ed(n

51)

Wee

den

Isla

nd

Red

(n5

1)C

arra

bel

leIn

cise

d(n

51)

8CO

17(n

51)

8FR

4(n

51)

8WA

8(n

51)

1CV

32(n

51)

Oth

erw

ise

sim

ilar

tom

icac

eou

scl

ust

ers

1aan

d1c

NW

Flo

rid

a/S

EA

lab

ama

Mat

rix

GK

olo

mo

ki

+

1c(n

51)

Occ

asio

nal

spo

ng

esp

icu

les,

ph

yto

lith

san

dd

iato

ms

Mo

der

ate

toh

igh

Wee

den

Isla

nd

Zo

ned

Red

(n5

1)8W

A34

(n5

1)O

ther

wis

esi

mil

arto

mic

aceo

us

clu

ster

s1a

and

1b

NW

Flo

rid

a?N

ot

ob

serv

edN

ot

Ko

lom

ok

i;si

mil

arto

clay

sfr

om

Tel

fair

and

Way

ne

cou

nti

es,G

A3

(n5

2)C

om

mo

n(3

–5%

)m

ica

Rar

ep

hy

toli

ths

Ver

ylo

wto

low

Wee

den

Isla

nd

Zo

ned

Red

(n5

1)W

eed

enIs

lan

dR

ed(n

51)

8OK

6(n

51)

9ER

1(n

51)

Mu

chlo

wer

iro

nco

nte

nt

than

mic

aceo

us

clu

ster

s1a

–1c

WF

lori

da/

SW

Geo

rgia

Mat

rix

EK

olo

mo

ki

+

4(n

56)

1–3%

mic

aR

are

too

ccas

ion

alsp

on

ge

spic

ule

san

dp

hy

toli

ths

Mo

der

ate

toh

igh

Wee

den

Isla

nd

Zo

ned

Red

Wee

den

Isla

nd

Inci

sed

Wee

den

Isla

nd

Red

Kei

thIn

cise

dp

lain

bo

dy

sher

d

8CO

17(n

53)

8SA

1(n

51)

8WA

8(n

51)

8WA

35(n

51)

Po

ssib

lyre

late

dto

som

esh

erd

so

fsa

nd

/g

rit

1,3

NW

Flo

rid

a?M

atri

xD

or

D/

HK

olo

mo

ki

and

/o

rm

ult

iple

6(n

52)

5%v

ery

fin

em

ica;

3%fe

rric

no

du

les

Rar

esp

on

ge

spic

ule

s,p

hy

toli

ths

in1C

V32

;ra

rep

hy

toli

ths

in8O

K23

9

Lo

wto

mo

der

ate

Car

rab

elle

Pu

nct

ate

(n5

1)W

eed

enIs

lan

dP

lain

(n5

1)

8OK

239

(n5

1)1C

V32

(n5

1)U

nre

late

dto

any

oth

erp

aste

clu

ster

WF

lori

da/

SE

Ala

bam

aN

ot

ob

serv

edN

ot

Ko

lom

ok

i;w

este

rnp

anh

and

le/

SE

Ala

bam

a


305

identified as ‘‘sacred’’ or ‘‘prestige’’ based largely ontheir nonlocal pastes (Cordell 1984; Milanich et al. 1997)were produced on pastes that appear to be local toKolomoki. In our sample alone, 82 percent (n 5 18) ofthe 22 sherds belonging to these functional categoriesfrom sites other than Kolomoki exhibit pastes consis-tent with those local to Kolomoki. As compelling as theevidence is, however, until other studies are undertak-en our conclusions regarding Kolomoki manufacturingorigins of some Weeden Island pottery must beconsidered provisional. Comparative pottery and claysamples from the Weeden Island region are few innumber. Ideally, the range of paste variability inpottery at least a few other Weeden Island sites, aswell as variability in clay sources across the WeedenIsland culture area, should be documented.

Bearing in mind these caveats, the petrographic datasupport the notion that some prestige and sacredWeeden Island pottery was produced and widelydistributed or exchanged across the region from asingle Weeden Island center, namely Kolomoki. We donot argue—as Sears (1973) seemed to suggest—that allof the Weeden Island ‘‘sacred’’ pottery was producedat Kolomoki; in fact, the analysis of paste groups atMcKeithen clearly argues that a portion of such vesselsfound there were produced locally.

Nevertheless, because the production of these typesseems to have been somewhat spatially restricted, andbecause these types were widely used in communalmortuary ceremonies, we believe that the evidencerises to the level of specialization as defined by Costin(2001:276); that is, fewer people were producing sacredand prestige Weeden Island vessels than were usingthem. Specifically, we suggest that potters at Kolomoki,or perhaps a subset of the local pottery-making com-munity, produced vessels that were used in relativelydistant communities elsewhere in the Weeden Islandregion (in some cases, such as McKeithen, more than200 km straight-line distance).

Spielmann (2002:198) cites ethnographic and archae-ological evidence for specialization at the level of thecommunity among some small-scale societies. In thecase of Motu potters of the south coast of Papua NewGuinea, for example, some villages made all of thepottery for a distance of 400 km. Such a high degreeand intensity of specialization clearly does not seem tobe the case for Kolomoki and the Weeden Islandsocieties of the Gulf Coast. Instead, we believe futureanalyses should look for specialization among a subsetof the community at Kolomoki and other centers.

Given that the prestige and sacred Weeden Islandtypes are associated largely or exclusively (respective-ly) with ceremonial contexts, it would seem plausiblethat they were produced by a segment of thecommunity that had privileged access to rituals andthe esoteric knowledge associated with them, or

perhaps with artisans who were ‘‘attached’’ to suchritual specialists or other leaders. Spielmann (1998:158),based on a review of the ethnographic literature onritually related production in the native Southwest andNorthwest Coast, has posited the existence of threekinds of craft specialists in middle-range societies. First,where ritual performance is relatively open, she predictsthe presence of skilled, independent craft specialists.Where ritual knowledge and performance are importantto achieving and maintaining status, Spielmann sug-gests that craft specialists are likely to be the ritualleaders themselves, ‘‘thus greatly limiting the ability ofany uninitiated to perform ceremonies.’’ Finally, Spiel-mann argues that where ritual knowledge and perfor-mance were only one of several means for achievingstatus, skilled artisans may be embedded in contextscontrolled by ritual specialists (such as households orritual societies), ‘‘thus allowing for control by the ritualpractitioner over access to goods necessary for ritualperformance and building prestige.’’

A case could be made for each of these threepossibilities at Kolomoki. Elsewhere, Pluckhahn (2003,2010a, 2010b) has made the case that the ceremonies atKolomoki appear to have been relatively open, giventhat they took place in full view of a plaza that wasboth unrestricted by other architectural features andlarge enough to easily accommodate the site’s popula-tion (potentially hundreds of visitors in addition).Likewise, the ceramic caches on the eastern sides ofthe mounds appear to have been placed ‘‘for the deadin common’’ (Moore 1902:161; see also Willey 1949:405)rather than with specific individuals.

Finally, the themes that were represented on ‘‘sa-cred’’ vessels appear to have followed general cosmo-logical principles that are widespread on more utilitar-ian Swift Creek pottery. Following this logic, andapplying it to the first scenario within Spielmann’stripartite model, especially gifted, independent pottersat Kolomoki may have produced ‘‘sacred’’ vessels asritual paraphernalia or grave offerings—just as otherartisans produced ornaments of stone, shell, and micafor the same purposes—in addition to producing morequotidian pots used for everyday purposes. In thiscase, part-time specialists may have produced sacredand elite pottery as part of a seasonal cycle of domesticactivities, as Cobb (2003:70) suggests for the manufac-ture of chert hoes during the Mississippian period. Therelative homogeneity of pastes across functional cate-gories of pottery at Kolomoki would be consistent withthis model of specialization; so too is the fact that SwiftCreek vessels of presumed Kolomoki origin have beenfound at sites some distance away (Stoltman and Snow1998:152), suggesting the export of ceramic vessels wasnot limited to ‘‘sacred’’ and ‘‘prestige’’ wares.

On the other hand, as Sears (1973:39) noted, it isreasonable to suggest that ‘‘there were artisans as well


306

as priests, or priests who were artisans, whose ceramicefforts were primarily devoted to service of thesupernatural,’’ consistent with Spielmann’s secondmode of specialization in middle-range societies.Specifically, Pluckhahn (2010a) has pointed to elementsof Mounds D and E at Kolomoki—repeated at othermounds throughout the Weeden Island region—thatare suggestive of a ‘‘liturgical order’’ (Rappaport1999:35–36), or a formal sequence of ritual acts of thesort typically directed by a religious specialist. Forexample, Moore (1902:334) noted that the pots in east-side caches often seemed to be deliberately placed suchthat the finest effigies were positioned to the exteriorand would have been the most conspicuous. In anumber of cases, some of the pots appear to have beendeliberately broken and the fragments scattered indifferent areas of the cache (Lazarus 1979:17–18; Moore1902:130; Sears 1953:26). Extending this line of thinkingto Spielmann’s second mode of specialization, religiousspecialists at Kolomoki and other Weeden Islandceremonial centers may have produced ‘‘sacred’’vessels for use in the rituals they directed. In possiblesupport of this interpretation, it is worth noting theexistence of a subclass of Weeden Island effigy vesselsthat depict human figures. These human figures, all ofwhom appear to have been male, are portrayed inrelatively consistent dress and pose, the latter oftenresembling the stances adopted by shaman (Pluckhahn2010a). Connections with avian imagery reinforce theshamanistic theme: one of the human effigy vesselsfrom Kolomoki has a hairlock in the form of a roseatespoonbill, while the other has a wood ibis on eachshoulder.

If this was the case, however, the individuals atKolomoki who manufactured such ‘‘sacred’’ vesselsappear to have obtained their clay from the samesources as those making utilitarian vessels. Moreover,as we argued above, there is a relative paucity ofindividual markers of status in the burial mounds atKolomoki and other Weeden Island sites, as well as inthe remains of ordinary domestic activities in thevillage middens, that would seem to argue against thesuggestion that leadership was vested in individuals.This suggests that ritual authority was shared amongmembers of a more collective social formation, such asa moiety or sodality.

Consistent with Spielmann’s third mode of special-ization, skilled artisans may have been embeddedwithin these more collective social contexts. A generalanalogy might be provided by the religious sodalitiesof western Puebloan groups such as the Hopi and Zuni.Many of these societies cross-cut kin groups, thusintegrating and reinforcing communal relations withinvillages (Kantner 2004:252). Puebloan sodalities alsointegrated communities across larger regions (Adamsand Lamotta 2006; Schaafsma et al. 2002; Walker and

Skibo 2002; Ware and Blinman 2000). Prospectivemembers of ‘‘borrowed’’ sodalities were traditionallytrained and initiated in the communities where thosesodalities originated (Ware and Blinman 2000:393).Such a scenario could account for the movement of potsfrom Kolomoki to other Weeden Island ceremonialcenters, such as McKeithen.

Identifying the production system behind ‘‘elite’’Weeden Island vessels—the artisans, means of produc-tion, social relations of production, and relationships ofdistribution (Costin 2001: 277)—will require consider-ably more effort. The three possibilities consideredabove offer hypotheses for additional research. Wheth-er or not any of these possibilities ultimately provesvalid, they offer more interesting opportunities forresearch than sterile debates regarding the associationof specialization with particular forms or types of socialcomplexity.

Notes

Acknowledgments. Thin sections from Kolomoki were paid forby a Faculty Research and Development Grant, College ofArts and Sciences, University of South Florida. The collectionsfrom Kolomoki are curated at the Laboratory of Archaeologyat the University of Georgia; samples from other sitesmentioned in the paper are curated at the Florida Museumof Natural History. Earlier versions of this paper werepresented at the Meeting of the Society for AmericanArchaeology and the Southeastern Archaeological Confer-ence in 2010. We thank Editor Charlie Cobb, as well as KeithStephenson, Jim Stoltman, and one anonymous reviewer, forthoughtful comments that significantly improved the qualityof our paper.

1 The relative abundance scale is as follows: abundant,common, frequent, occasional, rare, and none.2 The kiln temperature was initially set at 275uC and held for10 minutes (with kiln door slightly open to allow for escape ofwater vapor). Then the kiln door was shut completely and thetemperature was raised to 800uC. After 20 minutes, the 800uCtemperature was achieved and was maintained for 30 min-utes. The total firing time was approximately 77 minutes.3 Very low iron oxide content is represented by white to verypale brown refired colors. Low iron oxide content isrepresented by light yellowish brown refired colors. Reddishyellow to light reddish brown refired colors representmoderate iron oxide content and yellowish red to red refiredcolors represent high relative iron oxide content.4 The counting interval was 1 mm by 0.5 mm. Each point orstop of the stage was assigned to one of the followingcategories: clay matrix, void (including channel voids, closedpores, and micropores [Rice 1987:350]), silt particles, and veryfine through very coarse quartz and other aplastics of varyingcompositions. Most of the counts were made using the 103objective, but the 253 objective (with plane-polarized light)was used to confirm the occurrence of siliceous microfossils.Size of aplastics was estimated with reference to the Went-worth Scale (Rice 1987:38). A comparison chart of estimatedpercent particle abundance (Rice 1987:Figure 12.2) was alsoused for estimating relative abundance of silt, spongespicules, and other constituents occurring in low frequency.


307

For cases in which fewer than 200 points were counted (n 510 of 65), the thin sections were rotated 180 degrees on themechanical stage and counted a second time (after Stoltman2000:306).5 Sand counts include quartz, polycrystalline quartz, feld-spars, and other non-opaque minerals (excluding micas).Counts of silt, ferric concretions, clay lumps, and other matrixconstituents were excluded from this calculation. The sand-size index listed takes into account the size differencebetween very fine and fine sand inclusions. In this index,very fine grains are given a value of 0.5, whereas fine grainshave a value of 1.6 The two grit-tempered thin sections in the sample may notbe representative of grit tempering in the larger Kolomokisample, at least in terms of the Lamar cases, which seem tocontain predominantly of monocrystalline quartz.

References Cited

Adams, Charles E., and Vincent M. Lamotta2006 New Perspectives on Ancient Religion: Katsina Ritual

and the Archaeological Record. In Religion in the Prehis-panic Southwest, edited by Christine S. VanPool and ToddL. VanPool, pp. 53–66. Altamira, Walnut Creek, CA.

Beck, Margaret E.2006 Linking Finished Ceramics to Raw Materials: Oxidized

Color Groups for Lowland Desert Clays. Kiva: The Journalof Southwestern Anthropology and History 72(1):93–118.

Blitz, John H.1993 Ancient Chiefdoms of the Tombigbee. University of

Alabama Press, Tuscaloosa.Borremans, Nina T., and Graig D. Shaak1986 Preliminary Report on Investigation of Sponge Spic-

ules in Florida ‘‘Chalky’’ Paste Pottery. Ceramic Notes 3:125–131.

Bullen, Ripley P.1971 The Sarasota County Mound, Englewood, Florida.

Florida Anthropologist 24(1):1–30.Cobb, Charles R.1993 Archaeological Approaches to the Political Economy of

Nonstratified Societies. In Archaeological Method andTheory, vol. 5, edited by M. B. Schiffer, pp. 43–100.University of Arizona Press, Tucson.

2003 Mississippian Chiefdoms: How Complex? AnnualReview of Anthropology 32:63–84.

Cordell, Ann S.1983 Patterns of Resource Selection in the Manufacture of

North Florida Weeden Island Pottery. Southeastern Ar-chaeology 2(2):84–97.

1984 Ceramic Technology at a Weeden Island Period Site inNorth Florida. Ceramic Notes, No. 2. Occasional Publica-tions of the Ceramic Technology Laboratory, Florida StateMuseum, Gainesville.

2006 Finish What You Start: Revisiting the McKeithen SitePottery Analysis. Paper presented in the symposium‘‘Jerry’s Kids: Papers in Honor of Jerald T. Milanich’’ atthe 63rd annual meeting of the Southeastern Archaeo-logical Conference, Little Rock, AR.

Cordell, Ann S., and Kathleen Deagan2011 Paste Variability and Clay Resource Utilization in 16th

Century Aboriginal Pottery from the Fountain of Youth

Site, St. Augustine, Florida. Paper presented at ‘‘LifeAmong the Tides: Recent Archaeology on the GeorgiaBight,’’ the Sixth Caldwell Conference, organized byVictor D. Thompson and David Hurst Thomas, May 20–22. St. Catherines Island, GA.

Costin, Cathy Lynne1991 Craft Specialization: Issues in Defining, Documenting,

and Explaining the Organization of Production. InArchaeological Method and Theory, vol. 3, edited by MichaelB. Schiffer, pp. 1–56. University of Arizona Press, Tucson.

2001 Craft Production Systems. In Archaeology at theMillennium: A Sourcebook, edited by Gary M. Feinmanand T. Douglas Price, pp. 273–314. Kluwer Academic/Plenum, New York.

Earnest, Tray G.1995 The History and Results of Archaeological Investiga-

tions at 1CV32, the Mitchell Site, in Covington County,Alabama. Unpublished master’s thesis, Department ofAnthropology, University of South Florida, Tampa.

Goggin, John M.1952 Space and Time Perspective in Northern St. Johns

Archaeology. Publications in Anthropology No. 40. YaleUniversity Press, New Haven.

Graham, David J., and Nicholas G. Midgley2000 Graphical Representation of Particle Shape using

Triangular Diagrams: An Excel Spreadsheet Method.Technical Communication in Earth Surface Processes andLandforms 25:473–1477.

Kantner, John2004 Ancient Puebloan Southwest. Cambridge University

Press, Cambridge.

Kohler, Timothy A.1978 The Social and Chronological Dimensions of Village

Occupation at a North Florida Weeden Island Period Site.Unpublished Ph.D. dissertation, Department of Anthro-pology, University of Florida, Gainesville.

Lazarus, Yulee W.1979 The Buck Burial Mound: A Mound of the Weeden Island

Culture. Temple Mound Museum, Fort Walton Beach, FL.

Milanich, Jerald T.2002 Weeden Island Cultures. In The Woodland Southeast,

edited by D. G. Anderson and R. C. Mainfort, Jr.,pp. 352–372. University of Alabama Press, Tuscaloosa.

Milanich, Jerald T., Ann S. Cordell, Vernon J. Knight, Jr.,Timothy A. Kohler, and Brenda J. Sigler-Lavelle

1997 Archaeology of Northern Florida A.D. 200–900: TheMcKeithen Weeden Island Culture. University of FloridaPress, Gainesville.

Milner, George R.1990 The Late Prehistoric Cahokia Cultural System of the

Mississippi River Valley: Foundations, Florescence, andFragmentation. Journal of World Prehistory 4:1–43.

Mitchem, Jeffrey M.1986 Comments on Some Ceramic Pastes of Central

Peninsular Gulf Coast. Florida Anthropologist 39(1–2):68–74.

Moore, Clarence Bloomfield1900 Certain Antiquities of the Florida West-Coast. Journal

of the Academy of Natural Sciences of Philadelphia, SecondSeries 11(3):350–94.


308

1901 Certain Aboriginal Remains of the Northwest FloridaCoast. Pt. 1. Journal of the Academy of Natural Sciences ofPhiladelphia, Second Series 11(4):421–497.

1902 Certain Aboriginal Remains of the Northwest FloridaCoast. Pt. 2. Journal of the Academy of Natural Sciences ofPhiladelphia, Second Series 12(2):127–358.

1903a Certain Aboriginal Mounds of the Central FloridaWest-Coast. Journal of the Academy of Natural Sciences ofPhiladelphia 12:361–438.

1903b Certain Aboriginal Remains of the ApalachicolaRiver. Journal of the Academy of Natural Sciences ofPhiladelphia, Second Series 12(3):440–94.

1905 Certain Aboriginal Remains of the Lower TombigbeeRiver. Journal of the Academy of Natural Sciences ofPhiladelphia, Second Series 13(2):246–78.

1907 Mounds of the Lower Chattahoochee and Lower FlintRivers. Journal of the Academy of Natural Sciences ofPhiladelphia, Second Series 13(3):426–57.

1918 The Northwestern Florida Coast Revisited. Journal ofthe Academy of Natural Sciences of Philadelphia, Second Series16(4):514–81.

Muller, Jon1997 Mississippian Political Economy. Plenum Press, New

York.Nolan, Kevin C., Mark F. Seeman, and James L. Theler2007 A Quantitative Analysis of Skill and Efficiency: Hope-

well Blade Production at the Turner Workshop, HamiltonCounty, Ohio. Midcontinental Journal of Archaeology 32(2):297–330.

Pauketat, Timothy R.1987 Mississippian Domestic Economy and Formation

Processes: A Response to Prentice. Midcontinental Journalof Archaeology 12:77–88.

Pluckhahn, Thomas J.2003 Kolomoki: Settlement, Ceremony, and Status in the Deep South,

A.D. 350 to 750. University of Alabama Press, Tuscaloosa.2010a Practicing Complexity (Past and Present) at Kolo-

moki. In Ancient Complexities: New Perspectives in Pre-columbian North America, edited by Susan M. Alt,pp. 52–72. University of Utah Press, Salt Lake City.

2010b The Sacred and the Secular Revisited: The EssentialTensions of Early Village Societies in the SoutheasternU.S. In Becoming Villagers: Comparing Early VillageSocieties, edited by Matthew Bandy and Jake Fox,pp. 100–118. University of Arizona Press, Tucson.

Prentice, Guy1983 Cottage Industries: Concepts and Implications. Mid-

continental Journal of Archaeology 8:1–16.1985 Economic Differentiation among Mississippian Farm-

steads. Midcontinental Journal of Archaeology 10:77–122.Rapp, George, Jr., and Susan C. Mulholland (editors)1992 Phytolith Systematics: Emerging Issues. Advances in

Archaeological and Museum Science, Vol. 1. PlenumPress, New York.

Rappaport, Roy A.1999 Ritual and Religion in the Making of Humanity. Cam-

bridge University Press, New York.Rice, Prudence M.1980 Trace Elemental Characterization of Weeden Island

Pottery: Implications for Specialized Production. South-eastern Archaeological Conference Bulletin 22:29–35.

1981 Evolution of Specialized Pottery Production: A TrialModel. Current Anthropology 22(3):219–240.

1987 Pottery Analysis: A Sourcebook. University of ChicagoPress, Chicago.

Rice, Prudence M., and Ann S. Cordell1986 Weeden Island Pottery: Style, Technology, and Production.

In Ceramics and Civilization, vol. 2, edited by W. D. Kingery,pp. 273–295. American Ceramic Society, Columbus, OH.

SAS Institute, Inc2008 SAS 9.2 for Windows. SAS Institute, Inc., Cary NC.Schaafsma, C. F., J. R. Cox, and D. Wolfman2002 Archaeomagnetic Dating. In The Joyce Well Site: On the

Frontier of the Casas Grandes World, edited by James M.Skibo, Eugene B. McCluney, and William H. Walker,pp. 129–148. University of Utah Press, Salt Lake City.

Sears, William H.1951 Excavations at Kolomoki: Season II 1950. University of

Georgia Press, Athens.1953 Excavations at Kolomoki: Seasons III and IV, Mound D.

University of Georgia Press, Athens.1956 Excavations at Kolomoki: Final Report. University of

Georgia Press, Athens.1959 An Investigation of Prehistoric Processes on the Gulf

Coastal Plain. NFS-G-5019, Final Report. National ScienceFoundation, Washington, D.C.

1973 The Sacred and the Secular in Prehistoric Ceramics. InVariation in Anthropology: Essays in Honor of John C.McGregor, edited by D. Lathrop and J. Douglass,pp. 31–42. Illinois Archaeological Survey, Urbana.

Spielmann, Katherine A.1998 Ritual Craft Specialists in Small-Scale Societies. In Craft

and Social Identity, edited by Cathy Lynne Costin and RitaP. Wright, pp. 153–159. Archaeological Papers 8. AmericanAnthropological Association, Arlington, VA.

2002 Feasting, Craft Specialization, and the Ritual Mode ofProduction in Small-Scale Societies. American Anthropol-ogist 104(1):195–207.

2008 Crafting the Sacred: Ritual Places and Paraphenalia inSmall-Scale Societies. In Dimensions of Ritual Economy,Research in Economic Anthropology vol. 27, pp. 37–72,edited by E. Christian Wells and Patricia A. McAnany.Emerald Group Publishing, Bingley, UK.

Stoltman, James B.1989 A Quantitative Approach to the Petrographic Analysis of

Ceramic Thin Sections. American Antiquity 54(1):147–160.1991 Ceramic Petrography as a Technique for Documenting

Cultural Interaction: An Example from the UpperMississippi Valley. American Antiquity 56(1):103–120.

2000 The Role of Petrography in the Study of ArchaeologicalCeramics. In Earth Sciences and Archaeology, edited by PaulGoldberg, Vance T. Holiday, and C. Reid Ferring,pp. 297–326. Kluwer Academic/Plenum, New York.

Stoltman, James B., and Frankie Snow1998 Cultural Interaction Within Swift Creek Society:

People, Pots, and Paddles. In A World Engraved: Archae-ology of the Swift Creek Culture, edited by Mark Williamsand Daniel T. Elliott, pp. 130–153. University of AlabamaPress, Tuscaloosa.

Walker, William H., and James M. Skibo2002 Joyce Well and the Casas Grandes Religious Interac-

tion Sphere. In The Joyce Well Site: On the Frontier of the


309

Casas Grandes World, edited by James M. Skibo, Eugene B.McCluney, and William H. Walker, pp. 167–176. Univer-sity of Utah Press, Salt Lake City.

Wallis, Neill J.2011 The Swift Creek Gift: Vessel Exchange Along the Atlantic

Coast. University of Alabama Press, Tuscaloosa.Wallis, Neill J., and Ann S. Cordell2011 Swift Creek Interactions Under the Microscope: Petro-

graphic Analysis of Pottery and Clay Samples from theAtlantic Coast. Paper presented at ‘‘Life Among theTides: Recent Archaeology on the Georgia Bight,’’ theSixth Caldwell Conference, organized by Victor D.Thompson and David Hurst Thomas, May 20–22. St.Catherines Island, GA.

Ware, John A., and Eric Blinman2000 Cultural Collapse and Reorganization: The Origin

and Spread of Pueblo Ritual Sodalities. In The Archaeol-ogy of Regional Interaction: Religion, Warfare, and Ex-change Across the American Southwest, edited by MichelleHegmon, pp. 381–409. University Press of Colorado,Boulder.

Welch, Paul D.1991 Moundville’s Economy. University of Alabama Press,

Tuscaloosa.Willey, Gordon R.1945 The Weeden Island Culture: A Preliminary Definition.

American Antiquity 10(3):225–254.1949 Archeology of the Florida Gulf Coast. Smithsonian

Miscellaneous Collections 113. Smithsonian Institution,Washington, DC.

Willey, Gordon R., and Richard B. Woodbury1942 A Chronological Outline for the Northwest Florida

Coast. American Antiquity 7:232–254.Wilson, Gregory D.2001 Crafting Control and the Control of Crafts: Rethinking

the Moundville Greenstone Industry. Southeastern Archae-ology 20:118–128.

Yerkes, Richard W.1983 Microwear, Microdrills and Mississippian Craft Spe-

cialization. American Antiquity 48:499–518.1989 Mississippian Craft Specialization on the American

Bottom. Southeastern Archaeology 8:93–106.


310

Pluckhahn, Thomas J., and Ann S. Cordell 2011Paste Characterization of Weeden Island Pottery from...

Documents

Transcript of Pluckhahn, Thomas J., and Ann S. Cordell 2011Paste Characterization of Weeden Island Pottery from...