You are on page 1of 16
274 Special Top: Ethnoarchaeological, and Ethnographic Pottery Studies 6 hic and efnearchace pres eo loi! coservatns poten” manufrs, use, and dtibaton sang neo peop The Ends, quantities, sics, and appearance of vessels in use in ind, Communities—and by extension st archaeological sites considerations that extend beyond simple fac- 9.1 Potury Classification ict kof hing: fr name, singing ter cone GEA FOU Wi tr ey snare ae ‘etenciy) The pine tat ihe smarty of ones wate ere ig pot occur chance but ree someting itherenly signa wee grouping of similar entities. Not iplines, i lies at the core of human ics in Archaeological, : Hyves of Material, technique, and style, and ther significance is intexpreted _cltrally © Different kinds of groupings—which also represent different levels of cas- cation—may be distinguished. Technically, for example, the process of "Gasification is somewhat different in creating groups for a new and previ- fusly unclassified set of materials (an activity called categorization) as op- to assigning individual objects to established classes (a process of ident- feation) previously defined by certain criteria or properties of their members. "The characteristics of the entities to be classified or identified are usually + called attributes. An attribute isa property, characteristic, feature, or variable fof an entity; in the case of pottery, attributes that are commonly of intrest include color, thickness, inclusions, hardness, form, and so on. An attribute state is usually an artfact’s specific Value or score on a particular attribute; for example, the states or values of the attribute “color” may be red, black, ‘white, of brown, and values forthe atribute “hardness” may be 1.5, 4.0, 6:5. (The terms attribute and attribute state are sometimes employed interchange- ably, however, with attribute occasionally referring to the state or value of & particular variable; see Doran and Hodson (1975, 99.) In classifying pottery, anthropologists and archaeologists commonly ap- proach their subject matter from two directions. Devised (or formal or scien titi) classification has a long history, while ethnotaxonomic or folk classfica- tion is only now beginning to receive attention. There is sotbe overlap among. the concepts and objectives of these two positions, however, so they are not mutually exclusive. 9.1.1 Devised Classifications ‘The most typical approach to archaeological data is devised, formal, or scien- tific classifications, which are created by the analyst. Although many schemes have been developed for grouping similar sites, artifacts, and cultures in con- junction with a variety of archaeological objectives, these activities are com- prehended by the more general purposes of classifications within all scientific disciplines. Formal classifications structure the domains of inguity of scien- tific disciplines by furnishing a system for describing and naming the objects of study within a science; fostering communication within a science through shared terminology and nomenclature; permitting predictions about the rela- tion of the classified items to other objects studied within the science; and serving as extensions of and empirical justification for concepts used within the body of theory of that science (Blashfield and Draguns 1976, 574). Class fication is not the final goal of any science; itis a basic procedure by which a discipline and its data are structured. Within archaeology, its long-standing importance is reflected in the Fact that major stages in the history of archacol- ‘gy in the Americas, stages spanning approximately one hundred years, have been termed the classiticatory-descriptive and classificatory-histoical periods (Willey and Sabloff 1980, 34-180) because of the emphasis on these opera- tions for describing and organizing archaeological data The object of most classificatory operations in archaeology is to create types, and variations in the definition of a type arise from different positions ram they are created (see Hill and Evans 1972; Cowgill 1982). According tributes of artifacts: thug Pires, hay be simply defined as nonrandom attribute clusters (Spaulding sees types as clusters of items (Hodson _ 9B) ad defines a type as a group or clas of items that is internally cohesive and «ai be separated from other groups by one or more discontinuities in an tribuae gates (Whallon and Brown 1982, xvii). This object-based position fy ot eeily opposed to the atribute-based definition, however, because types | (and henee indirectly on their associations) Attributes may be (and often are) employed Piersscally rather than simaltancously to create types within a collection {0 eine sition, types consist of (and are based on) 1958, 1982). An alternative view still deyend on several attributes athe: Wan on a single attribute (see Wallon 1972), Ditfieent types with different that Ceormine the group divisions. Although it sible: mmatier how many attributes are considered in any classification eye. sem, runumber that theoretically exis is infinite (se Hill and Evans 1972, 250-51. Selecting different kinds of attributes in tion beais to different kinds of types—for example, morphological, histor cal-inde, functional, and cultural (Steward 1954). Atemihtes themselves have often been classified by their role or utility in a evised lassfication, For example, attributes of archaeological pottery may be descrbed as intrinsic, including an object's composition, shape, and deco, ration extrinsic, sach as its date, provenienee, and function as assigned by the archsologist (Gardin 1980, 65~68). Attributes can also be ranked as key, essential and inessentil variables (Clarke 1968, 71) according to thet sig, nificarceand utility within a particular devised classification scheme Grovp, classes, and types are sometimes distinguished at different stages ‘of classication (see Dunnell 1971). A group consists of actual objects, such 3 potsheds, and exists in the phenomenological or empirical realm, Classes and types however, represent verbal models or descriptions of objects and are ideation A class may be further described as a “generic term refering t0 ny divison of materials into groupings based on similarities and differences” and may ie based on single attributes, whereas a type isa formal conceptual (or abstrat) unit based on “a consistent patterning of attributes ofthe mate als” Geiland Evans 1972, 233). This differentiation between the empirical (aroupd sed the conceptual (class and type) is highlighted by a distinction be- {ween clasification and typology: “A classification is no more than a set (or sets) off expirical groupings established for convenience. A typology, how. ever, Js albcoretically oriented classification that is directed toward the solu. tion of scne problem or problems” (Kluckhohn 1960, cited in Gifford 1960, 46; sow so Gardin 1980, 63). One ca distinguish different kinds of classifications by noting, for ex- ample, wether the eriteria for describing classes are equivalent and unor. sired, moequivalent and ordered, or hierarchical (ig. 9.1; see Dunnell 1971, 70). In th idcational realm, creating equivalent units is sometimes called Paradigrusc classification. An example in pottery analysis is categorizing ‘vessels Bhycertan features such as form: bowl, jar, plate, and cup are equiva. meanings arise from the particular atributes often claimed that to be LAKE cissifications should account forall attributes, ths is patently impor. ‘constructing a classifica: ‘SPECIAL Torics In Porrery Stupies 277 FieLo S pague 9.1 Model of different methods for ordering ob- ‘ pe ‘aad events by classification versus grouping proce- PHENOMENOLOGICAL] IDEATIONAL pee ates Doe (arouring) (CLASSIFICATION) | EZ] 5] seatisticn | paradigmatic giz Seiiere classification 4 Z| 213 zE 21] mumerice: | taxonomic [8] Teeny! | onasttietion Z/f ; exons aie ats of fo, andthe proce yeis “ppl Ane er ew rte tei aroha isthcl orton cent, ich peste toms as epi! pe) atest ee ym (see. 9.1.3.1), Taxonomic classification is sometimes more red ts : simply as creating types (Rouse 1960, 315—17), as opposed to analytic ood fran, ich aaacs abs to ‘bla and dec moder a sD. "Nodes re ceain abuts the nals judges weet commande sci aatacting ad tg fe ttn rc Ar etd by Rouse (1960, 313; see also Taylor 1948, 129-30), a mode is “any standard, Soran rh ge eh ows oman Stu thy hand down operate to enraton No lst thea repeant moses opel eta, Rowe or tame tis pragma mel enon ek inl woul mete costed moter, Two ge of tes ave fee cerned, both inherent in the data: conceptual modes relate to the style or form cttbe objet such mre spor pol upon, wie procera ‘lito maracas, fens, paraar nds ener the use of molds or coiling techniques (see Rouse 1960, 315). Modes repre- ‘treo tachi some bomen fetwee cours of doved cs Steam on on oft ancien aber ad unas he pte at Sept sien fo econ 9.1.2. Polk Classifications : 2 Folk clasifeations group and name ents scoring nave 8 opp tesemi, categories, Tey may ae a varie of characteris, ingly o in combination, an isnt avays cyto penta thr stra 862 in the following ancien Chinese casifcation of enim ‘Animals are divided into (a) those that belong to the Emperor, (b) embalmed ones () thos that are rained,( sucfing Ps (2) mer ‘aids, (fabulous ones, (g) stray dogs, (h) those that areineluc in this classication,() those that tremble as if they were mad, () innumerable ones, (K) those drawn with a very fine camel's hair ‘tush, (D others, (m) those that have ; se that have just broken a flower vase, an 1) those that resemble es ram adance (ergen Oak fea sons: 1937~ 1952, quoted in Aldendecfer and Blashiield 1984, 7) oie cassicions for pote Jor potery vessels, while nt as complex sth Chi ner ama clasiaton, hve long ten eefalyreoral os taaion poms chnogphie rca mo prey malate. is staan Slssitons, because archacologss hae fogsnt seed repllicite folk categories, occa S in ‘urveying the native terms for pottery vessels gathered by researchers wor kiig in pottery-making societies, one sees that the most striking cross. | calc etre i thet ems ae alton inva invaably based on projected Path ster dan big etre yates ten) deem aif by arco, sch assures tenet col, ps compose particuarise modifies referring tothe sizeof the pot or its inem tents. he Tarahumara of Metco, forexample ell pus ckovand seg ‘ui tree functional clases: pots to eat fom, pos for patching corn, and small ts (called sekori ranara, the modifier being a term usually used to refer cones children) (aston 1974, 103), AtMapumana, Papua New Gunes ‘esselsare given four names: ura or cooking pots, hodt or water vessels, nay ‘ordi, and rohe or sago storage vesels (Groves 1960, 10) : ‘As tes relatively simple classifications suggest, vessel names often com- bine sceral dimensions simultancously: size, shape, specific functions, and Content. Differently named sizes of vessels often have very different fune- tions (eg, Linares de Sapir, 1969, 8). Among the Fula ia Cameroon, for exaimph, 84% of the clasitied pots in the households studied were “jas wth Short mks and everied rims ... (that) come in five named sizes” Gin de, ‘eating order) loonde or ngiramwal, favande gaari (named for sorghum), df wt, hakoore (named fora sauce) anda minature juldude or sation Jar (Desd and Hennig 1972, 8-12). Among the Kalinga in the Philippines rice coding vessels and vegetabie/meat cooking vessels ace differentiated lexical ough sme ofthe ie ems van ter apes fe ok slassitiction there is clear differentiation of ven " Ste adconens (ble 9.0). nn Nl OPE nd names by both The: Sipbo-Conibo of Per, in contest, name vessels by general shape ‘Table 9.1 Size Function Terms for Kalinga Vessels. Use Sever Tem Soa Lbege Riss citing itayom Onan _Vepecteliclamcat cooking i tame ore Girma Lng oes) Ye in Peanogan ine nay Amato Potcowr Se-kong Chone-ctons Soe: At Longacre 1981, 3 n, pots are named primarily for general functions such as. 4 Sook, storage, or serving. These classes of vessels are commonly given ‘SPECIAL SOPICS IN POTTERY STUDIES categories, adding modifiers to denote size subsets, and though different sizes have different functions there seems to be litle specificity in naming as to their intended contents. Cooking pots or ollas ae called kénti and are distin- guished by size as ani (large, used for beet), anitami (medium, used for fneals), and vacu (small, for medicines). Similarly jars (chomo) and beer tugs (kénpo) come in three sizes, but food bowls (Kéncha) and shrania ex- hibit less standardized size ranges (DeBoer and Lathrap 1979, 105-10) ‘Many investigators who have elicited vessel terminology express some dis- may that such categories are so litle evident to (or replicable by) outsiders. ‘The attributes that anthropologists and archaeologists commonly focus on— ‘overall shape, rim and lip variations, base, decoration—may or may not co- ‘vary significantly in native classifications (see, ¢.g., Weigand 1969, 13). For ‘example, in Kathmandu a great deal of variation exists both within and be- tween classes: in terms of actual measurement the three classes (of liquid storage jis] could overlap quite considerably whereupon I was told that it ‘was a combination of shape and size; when I pressed further in at- ‘ampting to identify the precise shape variations concerned, ive., flat ‘or round base, rim forms, nothing could be selected 2s a permanent and identifying shape characteristic of one or another. However, the fact remains that all agreed instantly on names for any given pot (Birmingham 1975, 384) Potters in Kathmandu claimed that many of the minor variations of rim form ‘and decoration were efforts to cater to customers’ preferences (Birmingham 1975, 382), On & more optimistic note, precise measurements or ratios of vessel sizes, and proportions may sometimes correlate with shape classifications, and this ugurs well for traditional archaeological practice. Kalinga rice cooking pots ean be distinguished from vegetable and meat cooking pots, for example, by a relatively more restricted aperture, steeper rim angles, and a lower ratio of aperture to height (Longacre 1981, 54), Given the relative flexibility of criteria for ethnographic vessel terminol- ‘ogy—baced chiefly on size, shape, and contents or use—it is perhaps not surprising that when any ofthese features change the vessel names may change too. The names, in other words, are situation specific; in the terminology of, devised classifications, they are phenomenological, empirical, and pragmatic ‘group identifiers based on use rather than ideational, conceptual, or analytical classes based on abstract similarity relations. For example, in Papua New Guinea, cooking pots (wo) and water vessels (tohe) are given different names after the rims are accidentally broken off (Groves 1960, 1in); among the Ka- Tinga, if @ rice or meat/vegetable cooking pot becomes worn out or cracked and is transferred to use im toasting, it ceases to be an ittoyom or an oppaya and instead is called a finga. Elsewhere, if the contents or use of an intact vessel varies, so does the name: the large ollalike Huichol pot called a dale? is primarily used for brewing corn beer for ceremonial occasions, but if it is used for other purposes it may sometimes by referred to as a kaviuhéla’ (camuela; a term often interchangeable with dale), whereas if it is used for ‘water storage it becomes a ya’é, a water-storage vessel (Weigand 1969, 16, er surroundings (see Ellen 1979). The primary ao a ete from ethnosciemifc perspectives have included botanic) sg faunal lasses, disease categories, and color systems (for tevione cae Sturt: 4 vant 164; Werner 1972) snaferecent studies in pottery ethnotwxonomy examined the folk classifica ao open it highland Mexico frm diferent perspectives. Their fades Staaf understanding the nature of potery cassfcation among te sitiomalsocietes, ions ofthese classifications are relevant to sgories ef material such as pottery, where dimensions of variabli highly stuctured by biological, genetic, or social factors as are, fer ‘example, Eishep plant, and animal domains. This model holds that indvidvels bese {ty sateoris of things both by means of prototypes, which can be por See lea or ideal examples ofa panicular object, such asa fan, ond by | Sarl he Concept to other objects that ae not ideal or prototype jars eg Sil 3c basically" jars (as opposed to bows, for example). Graved oxen ioe Mowlelge that within the atsibutes or dimensions by which a proto iis, |= Peoenized, inevitable variations broaden the category beyond the ideal ihe i te or absence of certain features (sch as handle), their seen At Stags sessed shape, particularly width-o-heigh ratio (r "ster and neck sition (Kempton 1981, 36, 39-40), Ina lefinitions (both prototypes’ and extensions of definitions) varied Mth the ee; sex, occupation, and socioeconomic status ofthe informant ter Tenet (ess extension) to shape, while females gave more weight teat Te gms as handles, which are important functionally (Kempton 1981, wer, Simlaty, although one might intuitively hypothesize that pene then SERES trod have more names for vessel categories or use moe resrnnd sannzor Sinitions, this did not appear to be 50. Instead, like female infor shants; ie gave more weight to function and attachments and less to eo ‘shape in clostying the vessel drawings (Kempton 1981, 123, 138), __ SPBKINE TOPECS IN POTTERY STUDIES B85n "TIG8ORoas ener are rere an eae ere Sue Teen papery tai Sinem ietrerm Shenae nae Eocene nahn Avot uy of folk clasifetons of potery ws aed on bth fates (ates) and type names lied fom cghy-ive pots and assistants n Feta, Mexico Kaplan and Levine 1981. Chater analysand mulinen Sonal Scaling wee usd in interpreting theresa, reveling he stoctre 7 Setetonmy bath analogy wit erred dvd ca nsand by retin 4 “cognive map” ofthe eres ongaiing cultural perceptions of ‘be potery. The none mulidinensional sing procedure was bs i the late sie ofthe twenty Ave Senied categories or types of os inthis dy. The snr ofthe type i epreeced bythe potons on a {to-tmesionl lng poo dines ae nterect in ems of ‘mown cata atepris hat sree thinking and behavior i the society. Inde seating plot of te Pasha pouey (Gg, 9.2, te two pinary features ferentiated the types are axis I, restictedness or nonres ttn, ao ne, pee af mes som ee oe) Bi acs he slot wit ecogizad an are subsumed within yes: for "xample, an incised type may have a groove-incised and a fine-incised vat Foie oc tore types nay be subsumed within a seam group (et. Colon 14953), which is usually formed on the basis of similar surface treatments. for example, a red-slipped ceramic group—and may include a number of dif Fe ce ceoraivetpee such os nce, applique, an plychvome pained, ‘within red-slipped wares. Subeequent elaborations ofthe tpe-varlty approach and applications to Maya pottery (Wiley Cuber. and Adams 1967; Gifford 1976) led to devel F cpmen of higher, more inclusive levels inthis taxonomic hierarchy and acon _ awit bonder questions of intgraton and inerpretaton ese. 9.132). ha ri Cadabon iit and Levine 1981, 876), ndings thar closely fra ee oo ae and closed forms, arall oe oppositions within the potery and the rset ample types on teak eof he eae % with red colors, black-on-red designs, M if US; tine est, and femilenes. The ete ote Wich dosed forms, black Cob, nd matey ack Colors or blck-on-red decoration, and death 93 Issues in Archaeological Pottery Classification Tibi DiscuPnve ano Chonorocical. Sysreuanics | 9.1.3.2 Ane Types “REAL” on “ARTIFICIAL? In the late 1940s and early 1950s, a series of disputes erupted in the field of archaeological classification concerning whether types are inherent in the data and merely recognized by analysts or whether they are artificial units imposed ‘on the data. A key publication was Ford's (1953) “Measurements of Some Pre- historic Design Developments in the Southeastern States,” which argued that pottery types were analytical constructs created by the archaeologist as tools for chronology building and, more important, for achieving an understanding ‘of culture change. ‘The position had actually been established somewhat ear- lier (Ford 1938), and indeed a classic statement of this creationist view had been made in 1946, some years before the Ford monograph, by J.0. Brew, Pecos (ew Mexico) Conferer {Taam ma 7 Ph "ay “ean (Kiddsr'931, 21-35 siven ceramic group different from all others We must classify our material in all ways that will produce for us fect32) 8 Postion essentially aligned with what is now kong useful information. ... We need more rather than fewer classifica: tions, different classifications, always new classifications, to meet new needs. We must not be satisfied with a single classification of a group of attifacts or of a cultural development, for that way lies dogma and defeat. Brew 1946, 65) The position that archaeologists’ types were created analytical constructs was challenged impiicity in a review of Ford's monograph (Spaulding 19532), ‘but the manifesto of Spaulding’s opposing stand was published separately as “Statistical Techniques for the Discovery of Antifact Types” (Spaulding 19534). Types are said to be attribute associations inherent in a dataset that archaeologists can discover through statistical correlation techniques. This Viow of types as real and inherent in the data had been articulated earier in ‘one form or another by several scholars (e.g., Taylor 1948, 130; Krieger 1944) and continued to be popular with archaeologists in subsequent decades, particularly as quantitative techniques gained emphasis in archaeology in the 1960s. A key interpretation of types based on the discovery position is that because types ate thought to be culturally and historically real, they therefore reflect the ideas and values of the ancient people who made and used the artifacts: “Bach type should approximate as closely as possible that combination of me~ chanical and aesthetic executions which formed a definite structural pattern in the minds of a number of workers, who attained this pattern with varying de- ee aoe Pottery types (Kidder 1927, 490), analogous to the bio- ian ce, ep He ean : (¢-g., Tularosa or Jeddito) with a technically descrig ve {nuslrnas rains of sp we ge matte shred pe i ah ‘f success and interpretation” ‘ree pleuces and inerretaion” (Kreger 1944, 278). tn oti words, Sesbete ene a cee aon nh ensara g e eo ‘obsjets reflect variations in th i 4856 bur Repo 96nd, 9 eee a De Th npr Or a ee a cath atin sta ages es oo minor ceramic differences [which] were the c ich were the esa oF we tia the cones of relatively small cl proms ort fakes ters who indulged preferences the aca ae Gly mus be gure er who eres ove vent eter aee a gid MAH we nda nd sea ation”. whl the type portrays a conten ofa numberof pot tas that were scoops nek Peter but to most e Feo, 5435” MO others adhering toa given culture pattern (Giford mat Phrased (Hill and Evans 1972, 261). Any atifat class format by AIRS BE (wo or more attributes—whether these associations are iten ted ata ob eebl songs be eee ternad ‘havior on the part of preston paclbem ae Iay ac comespond to ee ° ily held nations e., mental temple) co Seine aa. I lo, a has tern Sex, ok lathe aoe auch Bid and context tae abe sccommaied wie fa of Melos dance sed intence itm, Beats teas to row the archacologist can wt devised ‘stegors canespond to nate clase. eee Atri sscitons ae enpialy rea, d ante y rl, dsconeible, and verb, i inpotar wither pes ate considered to cen a cone 7 Seer it re sa ee ae meow an an Of the attribute selection Process, ‘it is oo single classification is the best or only one { Feiner Mase rot Pitas re ae "gitention to a much wider range of attributes in artifact collections as classi “Cations began to be expected to do more than date sites, contributed to a need ric peoples. But this patterning may or zed and folk classifications of archacological pottery, this is not the only jon for the uility of a classificatory system. 3,3. STATISTICAL APPROACHES 10 CLASSIFICATION Jing their cue from Spaulding’s (1953b) work, archaeologists gradually to investigate quantitative approaches for bringing order out of their _easingly rigorous methodological focus of archacology as @ whole, plus the fot more sophisticated systems for investigating the structure of such complex cata sets. ‘Although Spaulding’s work emphasized that types were discovered by cor- relating attributes, subsequent quantitatively derived classifications were not invariably interpreted in the vein of types inherent in the data. In the late = 1960s and early 1970s, numerical taxonomic (Sokal and Sneath 1963) and ‘multivariate statistical techniques gained increasing attention among archae- ‘logis for classifying large data sets by similarities and differences in many attributes simultaneously, These classifications were usually regarded as cre~ ated rather than discovered types (¢.g., see Hodson 1982). One approach was to use factor, or principal components, analysis (Christenson and Read 1977) to extract major underlying dimensions of variability in the data, dimensions thought to be measured imperfectly and nonindependently by the original attributes, ‘More commonly, the technique of cluster analysis (Hodson 1969; Alden- derfer and Blashfield 1984) has been employed. Cluster analysis is a multi- variate statistical procedure that groups entities by their similarity on a large number of attributes. Although there are several variants ofthe technique, the ‘output is usually a treelike graph (a dendrogram; see fig. 14.1) showing suc- ‘cessive linkages of similar entities. Verifying the quality ofthe classifications resulting from these procedures— ‘determining which is “best” for the data—is problematic. The choice of at- tributes and specific procedures is critical, for different classifications can re sult fom different techniques. Although the methods replicate to some degree the archaeologist’ actual artifact sorting (in which several atributes arc con- sidered simultaneously), they do not always accurately teproduce archaeolog- ical classifications of real data sets. One procedure that did replicate archaeol- ogists’ classifications of a sample pottery collection was association analysis, (Whalion 1972), which divides the collection by individual attributes (rather than total similarity, as in cluster analysis). This and related procedures, how- ever, have been interpreted (Whallon 1982, 127) from the position that classi- fications are inherent within the data sets and can be discovered statistically Much of the appeal of statistical approaches to pottery classification is that they offer solutions to the difficult and sometimes arbitrary procedure of drawing boundaries around classes—identifying discontinuities in what ap- pear to be continuous variables. Pottery can be described with respect to a Arz= mimber of atributes, but only some are qualitative (or noncontnuods) | With eultaive variables itis easy enough to determine classes: colon exaimph, is a qualitative variable, and the boundari bet attribute states : ies between attribute sa SER white, and red por sen clear (ee chap 11), a do caeporia ot inc] 's, such as limestone, shell, voleanic ash, and quartz, sand, But there a se shades or gradation of esor and mires ine ee Sueeeing categories, continuous variables (size, diameter, thickness), ant {ther Dundary issues force the classifier to make difficult decisions shang lumping or splinting classes. Computer i e 7 Programs that permit hierarchical, ranking of atibus (Whalen 1973 orconder mixed ec ashe Guansiisive) data in groupings (Rice and Saffer 1982) may be useful in this Tegara The question is sil, however, ove of fuzzy boundaries: How mui, variably can exist within a class before it should be divided? Th i it vided? The answer ‘epenton te purse ofthe cision. At has ben emated shone encous Foo is meely one that has ret ted the seni tigatocr @enfer 1975, 246). e aoe " ‘ANGORA the stata procedues wed to det re iy eyes within tis ach clasieatin tenses of comer neat, ee nice culurlly are the inerpetions of be Opes dace Ts discovered, Ta tonal. 3pes ave ben vine at intially presente dose a fore as ual reas consiuing fondamctaateeratons sen data” Agued tan cxtene ts vewpin cn ead othe sions ae Urias ono a partclr clnsifeaay scene se only pales ey dha omc a the best and most mearingil one il and Boas een 235-38), : Ie ee aa ae Usate Pasicular problems. This view gained strength with the changes in ar- a ated has Ersanieed research goals other than chronology and oducts idence patens, and socioeconomic aon, an tans hoe ‘many of thrse have been based on vessel forms (morphological types). Curi- SEBEL TOPICS IN. CUTEMRY' SUVUREY peen used informally in archaeological ceramic studies inthe Southwestern United States and Mesoamerica, and they seem to continue as the primary ‘nit (rather than types) in Europe and the classical area. Wares have been | identified based on a large number of attributes, including function (¢.2.» fitchenware), decoration (black-figured ware), paste composition or texture {coarse ware), color (Fine Orange ware), surface treatment of color (glazed ‘ware, Phumbate ware), form (beaker ware, fring technology (earthenware), time period (Iron Age wares), and geographical location (Derbyshire wares), ‘Although in these largely informal uses the members of ware categories may share characteristics of color, fring, method of construction, composition, an 0 forth, 2s indicated by the ware name, their fundamental defining criteria are aspects of composition, manufacturing technology, or surface treatment. In the terminology discussed above (sec. 9.1.1), wares are paradigmatic classes, and they correspond to procedural modes in analytical classifications ‘Surface-tretment attributes (presence or absence ofa slip of glaze, luster, color, ete.) are closely aligned with the atribute emphasis of the traditional ‘ype-variety system classification procedures, and thus they make it relatively easy to integrate wares into the hierarchy of units. Variables of paste composi- tion (and manufacturing technology) are more dificult to place in this hicras chy, but they are also significant because they allow inferences about pottery production technique, organization, location, and trade distribution pat- terns—objectves that are aligned more closely with current archacological ‘oals than with chronology building (Rice 1976t). ‘Systematic incorporation of technological data into classification systems (2, through the ware concept) would be a significant extension of classi- ficatory procedures (see Shepard 1976, 310-14; Rice 1982). This, however, is formidable task for several reasons, chiefly beeause itis often dificult ro interpret technological propecties or identify composition without extensive laboratory analysis (see chap. 13). Other units that would have the same effect of bringing in data on manufacture include the concepts of technological style (Gee Lechiman 1977) and ceramie school (Smith 1962, 1175), though the Jattr is concerned more with decorative comparisons and similarities 9.13.5 Susmary : ‘Some of the problems in the literature on archaeological and ceramic classifi- cation are common to classifications in general. One example is boundaries: all classifications, by definition, require putting objects into one category or another. Any artificiality in archacological classifications arises not from the nature of the type concept itself or from the need to impose boundaries, but from the conflation of decisions made by the analyst. These decisions are shaped not only by the inescapable need to select a limited number of at- tributes for the classification, but also by uncertainty about whether the exca~ vated collection represents the full range of pottery at the site, adequately and without bias. In addition, types focus on attribute norms within the collec- tion, but variations—such as might be caused by changes in the mental tem- plate trough time or by new manufacturing techniques—are rarely incorpo- rated systematically (although the use of varieties in the type-variety system, has this potentia), In this vein it is also pertinent to remember that though artifacts function atributes willbe more o ess informative depending on i cutual realm, process, or function under investigation ‘Types based on Jn sam most rchacotoical al potry polos are devised li {erating te past ity xy yar pinay toca es acts develop chronologies, and help with inet compose eve zat have largely been met within the disciplined nee the develop. lata ety capi nee een of ie. empathy scat en ou, fain dan of tea ey Ho gs 8 ale oe Orr spe sr te ache. han sc opoion sar a Sass klsion cnr tape ron br ee a 9.2 Potteny Quantification + inferential, and tabulations of quantities of materi in different ceramic categories (types or forms) from various sites or parts of sites. ‘These quantitative questions in archacological pottery studies, whether rased in terms of absolute quantities or of relative amounts such as propor- tions or porcentages, are deceptively simple, however. The difficulty is that | mote often than not fragments are recovered rather than whole pots, and these fragments must somehow be translated into numbers of whole vessels. Yet this is only part of the problem, because in addition to estimating how many pots are represented, itis also important to know what proportion of the site was excavated to yield the pottery being studied. Thus archaeological pottery ‘quantification has two major components: sampling and counting vessels. 9.2.1 Sampling Methodological issues of archaeological site sampling are beyond the scope of the present work (see Redman 1974; Mueller 1975b), but the subsampling of cicavated artifact collections fora variety of purposes in pottery analysis is well within its purview. Some general issues of sampling in reconstructing pottery assemblages and site usage are considered here; more specialized issues of sampling for physicochemical analysis are addressed further in chap- ters 10 (sec. 10.3.2.2) and 13 (see, 13.1). ‘The most important consideration a the start of a ceramic analysis is how the pottery was recovered in the field, a consideration over which the analyst ‘ay or may not have control, OF specific interest are the way the excavation ‘its were selected and the techniques by which they were excavated ‘Totake the latter issue first, excavation technique is important in maintain- ing comparable conditions of recovery from one excavation to another. The Primary technique in question is sereening, for screening only some portions of an excavated site would result in differential recovery of particularly fine fragments of pottery and other materials (see Keighley 1973), and quan- titative comparisons would then be skewed. ‘An archaeological pottery collection almost never represents all the pottery {rom an entire site, because rarely isa site completely excavated, Instead, the Pottery i usually a sample ofthe total population. The excavated areas may have been chosen in any number of ways: by purposive or judgmental selec- tion to explore some specific aspect of the site, such as a particular structure zone; by haphazard or grab sampling for logistic convenience orto satisy ‘some nonscientific erterion; or by @ probabilistic sampling strategy. In this lest procedure the total area of the site is divided into equal units, and then individual units are selected for excavation by a method in which each one has 4 known probability of being chosen. Examples of such probability sampling ‘Strategies are simple random, stratified, and systematic sampling (see Mueller 1973b; Redman 1974; fora general discussion of sampling see Cochran 1963). Each of these strategies has some practical merits, depending on the overall Project goals, but they also have ramifications for potery analysis and its ob- Jestives, If all that is desined i a quantitative description of the pottery from exeavated portions of the site, sample selection is not highly significant. But iF the object is to draw some conclusions about the pottery atthe site as & Fesltm the smaller amount of excavated matral—~that i, Yo generale ome snp the populations i ce ama Gt he chances fe population ons ee ae tele, adhe mnt bane sane gs setter of thy witcha ee sepltons cane mado econ ee eat Se it saplin cee, hens oh ee Dis saving posi, tsetse sane ee ‘ecoveed from an identified population of horizontal of tie site, by rare versus abundant categor ithin the orig i " res within ts orginal bail hacen Fo Cowgill 164), of by some combination of these. o : 9.2.2 Cents, Weights, and Vessel Equivalents Athough auntiication of archaeological poutery can simply mean measuring how mucihtaterial is present, its most Common goal is to determine how traditional ue Presented in a particular deposit The siaplect nal oe _SYERINE TOPICS IN POTTERY STUDIES. ‘more complex data transformations (fora review see Deal 1983, 251-67) Counts may be made by simply tabolating the individual sherds from all ‘excavation units, bu this effectively considers each sherd to represent a differ ent vessel. In many cases two or more sherds can be joined together or cross- ‘mended; counting the joined pieces as single sherds thus more closely reflects the actual number of individual vessels than does simply counting all sherds, Counts of sherds can be reported as before- or after-mmending totals, the latter indicating a maximum number of vessels, Both measures generally overesti- ‘mate the number of vessels at a site, however. A variation of sherd counts is sherd ratios, usually calculated as number of sherds per unit of excavation volume. ‘herd counts may be misleading as estimates of numbers of whole vessels because there is no simple relation between a vessel's size or shape and the ‘number of sherds it breaks into: large vessels typically break into more picces than small vessels; thin-walled pots may break into more sherds than thick. ‘alled pots; and low-fited pieces may break into more fragments than high fired objects. The vessels that break into more sherds or are subject to mone disturbance (see Kirkby and Kirkby 1976, 237; Solheim 1984, 101; also Deal 1983, 201-2) or weathering (Reid 1984) in middens after discard will be ‘overrepresented in a particular excavation unit For these reasons, sherd weight (Baumhoff and Heizer 1959; Solheim 1960; Chase 1985) has sometimes been touted as a more accurate quantitative in, dicator. Sherd weights will effectively standardize the data for differences caused by large versus small sherd sizes or thickness, which may result from differences in disturbance and comminution in certain areas of a ste, Its fre, quently recommended that both sherd counts and weights be recorded to com. Pensate forthe inadequacies of each measure (table 9.2), herd weight totals may be modified by adjusting for thickness (Hulthén 1974). In this time-consuming procedure, the sherd collection is sorted into ‘iferent thickness categories, which are weighed separately. The gross weight Tibke9.2_ Percentages of Sends by Weight and Coun from a Test Pita Site 17, Fj Depth eta Se Se ee Clases of Shands 6 2 Phin Perenage weight 79.0 80.0810 S165 45 475 495 450 Percentage by count $9.0 90.0 91.0 $0.0 89.0 82.9 «0.0 600 GOO. Nunberofsherdsoa, 75 80 70 &0 #0 90 80 70 ‘go Incised Percentage neisht 17.0160 1401S 10038 os Pescentage by cout «91075 as S040 Numberofsherisoz. 40 30 20 30 30 40 49 — @ Relief Pereenage weight = «40-405. 13,5. 220 520 505 555, Rrmemageby cout 20225 5.0 5070 60383 dO 3D Numberofstertoz 5.0 4.0 63 50 40 70 40 40 30 ‘Source: Aes Deal 1983, able 33. The data ia Deals ele were taken rom Gifford 1976, ables 17 and 18, ‘of eaxhgroup is then multiplied or divided by the necessary factor to achieve a Stands! weight, and an estimate of the number of vessels. ‘Shuee! weight data can be converted to individual vessel estimates by at | least to procedures. One requires that the sherd collection be separated into form ortype categories (sec sec. 7.2.3) for weighing, and a whole or recon. Strucaate vessel must be present for each category. Weighing the whole vessel ‘and tes dividing the sherd weight totals by this figure gives some idca of the ‘numtsen of vessels for each category and forthe collection as a whole. Closer estimats are possible if this procedure is combined with estimates of the ‘aumiser of individual vessels based on rims, bases, handles, or other distinc, live fieaures: A sewnd weight-based procedure estimates proportions of vessel catego ries bry omparing size classes using the assumption that the weight ofa vessel 's proipotional to its surface (times thickness). Vessels are sorted into size cat-

You might also like