skibo, schiffer y kowalsky-ceramic style analysis in archaeology and ethnoarchaeology

22
JOURNAL OF ANTHROPOLOGICAL ARCHAEOLOGY 8, 388409 (1989) Ceramic Style Analysis in Archaeology and Ethnoarchaeology: Bridging the Analytical Gap JAMES M. SKIBO,MICHAEL B. SCHIFFER, AND NANCY KOWALSKI' Laboratory of Traditional Technology, Department of Anthropology, University of Arizona, Tucson, Arizona 85721 Received January 22, 1988 A refitting study with a sample of sherds from Broken K Pueblo indicates that at least some of the “patterning” in the assemblage identified by Hill (1970, Anthropological Papers of the University of Arizona 18) results from unrecog- nized whole vessels and large conjoinable sherds. It is argued that contemporary stylistic analyses are based on many of the same analytical methods employed in the Broken K study and that research findings of ethnoarchaeology are rarely applied in the analysis of prehistoric ceramics. Despite archaeologists’ increasing awareness of formation processes, spurred by ethnoarchaeological research, an- alytic methods that identify and take into account the effects of these processes on archaeological ceramic assemblages are underdeveloped. Suggestions are offered to resolve this problem, for both archaeological and ethnoarchaeological ceramic analysis, that may permit archaeologists to discover in prehistoric assemblages the same types of relationships identified in systemic assemblages. 0 1989 Aca- demic Press, Inc. INTRODUCTION Social, behavioral, and temporal inferences based on ceramic style form the core of many archaeological studies. The excavation and anal- ysis of Broken K Pueblo (Hill 1970), an early case study in the New Archaeology, introduced an innovative method for deriving social infor- mation from ceramic decoration. This paper evaluates the analytical tech- niques employed at Broken K with a reanalysis of a sample of the ceramic data. Patterning in the ceramic assemblage identified by Hill is found to be, in part, an artifact of the analytical methods and not related to past social organization. Contemporary analytical methods are also examined and it is found that, despite inspiration and insights provided especially by ethnoarchaeology, most ceramic analyses are based on assumptions and analytical techniques similar to those of the Broken K study. ’ Current address: 32 East First, Tucson, AZ 85705. 388 0278-4165i89 $3.00 copyrisht 0 1989 by Academic Press, Inc. AlllightSOfWgKOdUC~hMyh~scrvcd.

Upload: julio-cesar

Post on 25-Oct-2015

111 views

Category:

Documents


2 download

TRANSCRIPT

JOURNAL OF ANTHROPOLOGICAL ARCHAEOLOGY 8, 388409 (1989)

Ceramic Style Analysis in Archaeology and Ethnoarchaeology: Bridging the Analytical Gap

JAMES M. SKIBO,MICHAEL B. SCHIFFER, AND NANCY KOWALSKI'

Laboratory of Traditional Technology, Department of Anthropology, University of Arizona, Tucson, Arizona 85721

Received January 22, 1988

A refitting study with a sample of sherds from Broken K Pueblo indicates that at least some of the “patterning” in the assemblage identified by Hill (1970, Anthropological Papers of the University of Arizona 18) results from unrecog- nized whole vessels and large conjoinable sherds. It is argued that contemporary stylistic analyses are based on many of the same analytical methods employed in the Broken K study and that research findings of ethnoarchaeology are rarely applied in the analysis of prehistoric ceramics. Despite archaeologists’ increasing awareness of formation processes, spurred by ethnoarchaeological research, an- alytic methods that identify and take into account the effects of these processes on archaeological ceramic assemblages are underdeveloped. Suggestions are offered to resolve this problem, for both archaeological and ethnoarchaeological ceramic analysis, that may permit archaeologists to discover in prehistoric assemblages the same types of relationships identified in systemic assemblages. 0 1989 Aca-

demic Press, Inc.

INTRODUCTION

Social, behavioral, and temporal inferences based on ceramic style form the core of many archaeological studies. The excavation and anal- ysis of Broken K Pueblo (Hill 1970), an early case study in the New Archaeology, introduced an innovative method for deriving social infor- mation from ceramic decoration. This paper evaluates the analytical tech- niques employed at Broken K with a reanalysis of a sample of the ceramic data. Patterning in the ceramic assemblage identified by Hill is found to be, in part, an artifact of the analytical methods and not related to past social organization. Contemporary analytical methods are also examined and it is found that, despite inspiration and insights provided especially by ethnoarchaeology, most ceramic analyses are based on assumptions and analytical techniques similar to those of the Broken K study.

’ Current address: 32 East First, Tucson, AZ 85705.

388 0278-4165i89 $3.00 copyrisht 0 1989 by Academic Press, Inc. AlllightSOfWgKOdUC~hMyh~scrvcd.

CERAMIC STYLE ANALYSIS 389

The Problem

Stylistic analysis of prehistoric ceramics has a long history in North American archaeology and particularly in the Southwestern United States (see Rice 1987:244-272 for a review). Because prehistoric pottery ap- peared to reflect “cultural” and temporal variability, ceramics became a principal source of archaeological data. In the first half of the twentieth century a major concern was documenting synchronic and diachronic cultural variability and sherds were ideally suited to this objective.

The archaeology of the early 1960s found new ways to interpret ceramic style on sherds. Social organization and intravillage relationships were reconstructed based on variation in decorative design elements (e.g., Deetz 1965; Hill 1970; Longacre 1970; Whallon 1968). However, these initial efforts were later found to be deficient in several respects (see Graves 1985; Plog 1980 for recent reviews). Principal among these was uncertainty about the connection between ceramic style and social orga- nization (see Lathrap 1983; Stanislawski 1977). To explore the relation- ships between material and nonmaterial elements of a society (correlates) archaeologists undertook ethnographic fieldwork (Longacre 1974). Eth- noarchaeology quickly became an established subfield of archaeology and contributed to our understanding of the connections between material culture and human behavior. For instance, numerous factors have been shown to affect the intergenerational transmission of style (e.g., Graves 1985; Longacre 1981), and a beginning has been made in elucidating the causes of intra- and intersettlement ceramic design variability (e.g., Ar- nold 1983; Braithwaite 1982; Graves 1982; Hayden and Cannon 1984).

With these advances in the study of contemporary pottery decoration, one might think that archaeologists would be making important new con- tributions to the analysis of prehistoric ceramics. However, ceramic sty- listic research in ethnoarchaeology has been based exclusively on whole pots in systemic context. Although important, ethnographic information at this level cannot be directly applied by the prehistorian. Archaeologists deal primarily with sherds; these are often from various parts of a vessel and from deposits of complex, usually unknown, formation processes. These incompatibilities are mainly responsible for ethnoarchaeology’s lack of influence on stylistic analyses of prehistoric ceramics. We hasten to add that this disjunction, while serious at present, is not inevitable; it can be overcome by the development of appropriate analytic theory (sensu Sullivan 1978) with contributions from ethnoarchaeologists and prehistorians.

This paper is divided into several segments and begins with the Broken K reanalysis. A sample of sherds from the Broken K assemblage was refitted to evaluate the hypothesis proposed by Schiffer (1989), that

390 SKIBO, SCHIFFER, AND KOWALSKI

reconstructable vessels and large sherds are in part responsible for the clustering of design elements identified by Hill (1970). Second, after re- viewing a number of recent attempts at social or behavioral inference from prehistoric ceramics, we conclude that, in some respects, analytical techniques have progressed little beyond those employed by Hill on the Broken K assemblage. The analytical techniques and inferences formu- lated with prehistoric ceramics are contrasted with those in ethnoarchae- ology and the disjunction between the two is discussed. Finally, sugges- tions are offered to narrow the gap and permit the findings made with systemic collections to be incorporated more readily into analyses of prehistoric ceramics.

BROKEN K REANALYSIS

Phase 1

Broken K, excavated by James Hill in the early 1960s (Hill 1970; Martin et al. 1%7), is a 95room masonry pueblo in east-central Arizona. Despite the significance of Hill’s study and the many analyses and reanalyses of ceramic data from Broken K (e.g., Dumond 1977; Lischka 1975; Plog 1978), the formation processes of the deposits yielding those specimens have not been thoroughly investigated. In order to remedy that de& ciency, Schiffer (1989), using only the published data (i.e., Martin et al. 1966, 1967; Hill 1970), performed Phase I of the analysis of Broken K’s formation processes.

Hill (1970) carried out analyses as if he were dealing with Pompeii-like assemblages of de facto refuse; indeed, the abundant floor artifacts ana- lyzed for the room-function inferences certainly behaved like de facto refuse (Schiffer 1989). On the other hand, Hill (1970:31) claimed that 19 of 54 excavated rooms were trash-filled and only 12 restored pots were reported from all rooms-fill and floor proveniences (Martin et al. 1%7:126). These latter data suggested that Broken K contained some- thing other than Pompeii-like room assemblages (Schiffer 1989).

In order to resolve this paradox, Schiffer applied a series of indices, ranging from Reid’s (1978) measure of relative room abandonment to Kintigh’s (1984) diversity measure, to the Broken K ceramic data. These indices demonstrated that few rooms contained strong evidence for trash fdl. On the contrary, many of Hill’s “trash-fdled” rooms, especially those with arrays of de facto refuse, were probably abandoned late in the pueb- lo’s occupation. To account for the low number of restorable pots, in view of strong evidence for de facto refuse in many rooms, Schiffer (1989) proposed that many restorable pots-from floors and fills-were in fact recorded as sherds. This is the “missed pot” hypothesis.

CERAMIC STYLE ANALYSIS 391

On the basis of the missed-pot hypothesis, Schiffer (1989) suggested that the patterns in Hill’s factor analysis of ceramic design elements were determined largely by redundantly recorded design elements on individ- ual pot fragments, reconstructed vessels, and missed vessels. Sherds from the same vessel were entered into Hill’s analysis as if they were independent observations; it was these redundantly recorded vessels that created structure in the design element data.

Phase II

The present study of the Broken K assemblage was undertaken to test the missed-pot hypothesis and to explore in more detail the effects of recording design elements on sherds. In the summer of 1985, the junior author (NK) examined the Broken K artifacts, housed in the Field Mu- seum of Natural History, with the goal of refitting the decorated pottery in each room. Unfortunately, the ceramic assemblage was not complete; approximately 25-35% of the accessioned collection could not be located. Moreover, in accord with field policy at that time, unpainted sherds not part of a recognized restorable vessel were discarded after identification and tabulation (Martin et al. 1967:127). Clearly, the residual sample of Broken K decorated ceramics was not an ideal collection on which to perform refitting. Nonetheless, refitting was done on the sherd assem- blage from 16 rooms (Fig. 1).

The rooms selected for refitting were those pinpointed by Schiffer (1989) as having a high probability of containing “missed pots.” Room samples ranged from one or two bags of corrugated sherds to what ap- pears to have been the decorated ceramics used in Hill’s design study. Additional data on restored or restorable vessels were also found in the Field Museum’s accession records (Table 1).

The analysis involved refitting and matching sherds to their parent vessel. Because refitting was done with decorated ceramics, sherds from the same vessel were identified primarily by design attributes; however, other criteria such as thickness, temper, and color were also considered.

As already noted, Martin et al. (1967) list only 12 whole or partially restorable vessels from Broken K room floors and fills. The present anal- ysis (Phase II) identified 7 previously undocumented whole and partially restorable decorated vessels from floor and fill contexts (see Table 1 for vessel proveniences). The accession records proved to be an especially valuable source of additional data. The majority of Phase II refitted ves- sels are also listed in the accessions records as “restorable” or “possibly restorable.” Therefore, we were confident in adding several pots to the total vessel count that are listed as “restorable” in the accession records but which could not be located in the collection. Nonetheless, the total of 7 additional decorated pots is conservative since a number of painted

392 SKIBO, SCHIFFER, AND KOWALSKI

FIG. 1. Schematic drawing of the major roomblocks of Broken K Pueblo. Room numbers correspond to those assigned in Hill (1970); circled numbers are rooms analyzed in this study. Unexcavated rooms are denoted by cross-hatching (adapted from Hill 197O:lO).

vessels listed in the accession records were not included because they lacked provenience.

In addition to the decorated vessels, we identified 12 restorable and partially restorable plainware or corrugated pots from floor and fill con- texts (Table 1). It is safe to assume that this is also a conservative total, given that undecorated ceramics were not typically brought back from the field (Martin et al. 1967:127). Therefore, despite the incomplete collec- tions, the number of restorable or partially restorable decorated vessels from floors and fills is now 19. If corrugated and plainware pots are included, there is a conservative total of 31 whole and partially restorable pots from the excavated rooms at Broken K.

CERAMIC STYLE ANALYSIS 393

TABLE 1 RESTORABLEANDPARTIALLYRESTORABLEVESSELS LOCATEDDURINGTHEANALYSIS

Room Provenience How located Type Shape

Decorated 30 Floor RefitiAcces. BiW Bowl 39 Fill Refit B/W Jar 54 Fill Acces. B/W Jar 60 Fill RefitiAcces. B/W Jar 68 Fill Acces. B/W ? 69 Fill Refit B/W Jar 80 Fill Refit BIW Jar

Undecorated 2 Floor Acces . COIT. Jar

21 Fill Acces . COIT. Jar 28 Fill Acces. ColT. Jar 30 Floor Refit/Acces . corr. Jar 39 Fill RefitiAcces. corr. Jar 53 Floor Acces. COIT. Jar 60 Fill RetitJAcces. Corr. Jar 64 Fill RetitiAcces . con-. Jar 69 Fill RefitlAcces. ColT. Jar 69 Fill RefitlAcces. Plain Jar 78 Floor RetitlAcces. corr. Jar 80 Fill RefitlAcces. corr. Jar

Note. “Acces.” refers to Field Museum accession records, “B/W” equals Black- and-White, and “Corr.” refers to corrugated.

Figure 2 illustrates the locations of whole and partially reconstructable decorated vessels and Figure 3 shows the locations of all whole and partially reconstructable vessels, both decorated and undecorated, recov- ered at Broken K. Pots tend to be evenly dispersed throughout the pueblo, with the exception of the isolated roomblock. There is only one reconstructable vessel from this roomblock despite the fact that refitting was attempted on a large sample of sherds from several rooms. It appears that Hill’s (1970:31) “trash-tilled” inference for some of these rooms could be correct.

The general impression that emerges from the reanalysis, in agreement with Schiffer (1989), is that there is a good deal more de facto refuse in the Broken K assemblage than the previous vessel counts suggested. Unfortunately, because of the limitations imposed by the sample, many specific hypotheses about room formation processes remain untested. However, one method to pinpoint vessels proposed by Schiffer (1989) was very effective. This measure is important because it permitted us to illustrate how conjoinable sherds and reconstructable vessels affected patterns in Hill’s (1970) design element study.

394 SKIBO, SCHIFFER, AND KOWALSKI

FIG. 2. The location and number of decorated whole vessels from Broken K. Numbers in parentheses represent whole vessels discovered in this study.

The design elements at Broken K are thinly distributed throughout much of the pueblo (Dumond 1977). Schiffer (1989) proposed that high counts of a design element in a floor or fill assemblage could be caused by sherds from an unrestored whole pot. Rooms identified by Schiffer (1989) as having a high probability of containing missed pots (because of high frequencies of particular design elements) and which were also in- cluded in our analysis are listed in Table 2. (Table 2 also presents the specific design elements that exhibit anomalously high frequencies in each room.) Of the 32 high-frequency design elements (Table 2), 25 result from reconstructable vessels (5) or conjoinable sherds (20). For example, five instances of design element 146 (Fig. 4) occur on the floor of Room 30, and all could easily be from one reconstructable vessel (Fig. 5). It is

CERAMIC STYLE ANALYSIS 395

FIG. 3. Locations and total number of decorated and undecorated vessels from Broken K (all sources).

apparent that high frequencies of particular design elements most often resulted from conjoinable sherds rather than whole pots. For example, Fig. 5 illustrates the conjoinable sherds making up the cluster of design element 110 (Fig. 4) on the floor of room 64; this conjoinable rim seems to be responsible for all six occurrences of design element 110.

Because Hill’s factor analysis relied on Pearson’s r, results will be strongly influenced by high frequency occurrences (caused by restorable pots and conjoinable sherds). Much of the patterning in the factor matri- ces, which Hill attributed to social processes, may have resulted from redundant entries of the same vessel. For example, Factor 4 (fill) consists of design elements 7, 9-11, 45, 46-49, 71, 77-81, 127, 135, 136, 147, and 160. Rooms 68 and 69 had high frequencies of this factor (Hill 1970:61-

396 SKIBO, SCHIFFER, AND KOWALSKI

TABLE 2 RELATIONSHIP BETWEEN DESIGN CLUSTERING AND CONJOINABILITY

Room Provenience Design el. cluster Refits

2 Floor 2 Fill

45 Fill 415 Fill 415 Fill 415 Fill 415 Fill 415 Fill 415 Fill

7 Floor 8 Floor

11 Floor 30 Floor 30 Fill 33 Floor 33 Floor 33 Floor 33 Floor 33 Floor 39 Fill 60 Fill 64 Floor 64 Fill 69 Fill 69 Fill 69 Fill 69 Fill 69 Fill 69 Fill 69 Fill 69 Fill 80 Fill

82 Yes 31 No 15 Yes 18 Yes 20 No 43 Yes

146 Yes 158R No 173 Yes 153 Yes

5&51 Yes 153 No 146 Yes

69 Yes 45 Yes 46 Yes

146 Yes 155 Yes 158R No 13 No

155 Yes 110 Yes 139 Yes

7 Yes 45 Yes 46 Yes 49 Yes

9 Yes 77-a 1 Yes

135 No 136 Yes 141 Yes

62). Table 3 illustrates that in the fill of room 69, all but two of the design elements in Factor 4 are parts of conjoinable sherds or reconstructable vessels. Hill suggests that the clustering of Factor 4 in rooms 68 and 69 “may indicate . . . that these two rooms should be considered as a sub- division of (residence) Area IA” (Hill 1970:61). However, given that the

FIG. 4. Design elements 110, at left, and 146, on right.

CERAMIC STYLE ANALYSIS 397

FIG. 5. Conjoinable sherds each with design element 110 (Top). Partially reconstructable vessel composed of sherds with individual occurrences of design element 146 (Bottom).

vast majority of design element clusters are caused by conjoinable sherds or reconstructable vessels, the patterning appears to be an artifact of the analysis, and not marital residence.

Although the Phase II analysis was carried out on an incomplete as- semblage, two hypotheses about the Broken K ceramics were confirmed. First, support was found for the “missed pot” hypothesis; reanalysis of the Broken K material revealed 19 additional restorable or partiahy re- storable vessels. Second, it was found that the patterning in Hill’s factor analysis of design elements on floors and in fills was influenced by re- dundant design elements from the same vessel. The latter finding is es- pecially noteworthy because it underscores our claim that analytic tech- niques used in the study of prehistoric remains are at odds with the implications of ethnoarchaeology.

398 SKIBO, SCHIFFER, AND KOWALSKI

TABLE 3 THE EFFECT OF CONJOINABLE SHERDS AND RECONSTRUCTABLE VESSELS ON DESIGN

ELEMENTS IN FACTOR 4

Occurrences

Design element Rm 68 Rrn 69

I 9-11 45 46 49 71

77-81 127 135 136 147 160

2 0

4 0 3

9 Reconstructable vessel 15 Reconstructable vessel 5 Reconstructable vessel

12 Reconstructable vessel 9 Reconstructable vessel 3 Three conjoinable sherds 8 Reconstructable vessel

13 none found 3 none found

14 5 conjoinable sherds 0 NA 4 Three conjoinable sherds

Room 69 Refits

CONTEMPORARY ANALYTICAL METHODS

Much ethnoarchaeological research was a direct reaction to hypotheses proposed by Deetz (1%5), Hill (1970), and Longacre (1970). For example, Stanislawski (1973, 1977) attempted to demonstrate that the learning framework is not exclusively a mother/daughter interaction, and Lathrap (1983) outlined the many factors that can influence pottery designs. Re- sults of these studies and others indicated that the learning framework is much more complex than originally assumed. Only recently, however, has ethnoarchaeology begun to seek further causes of ceramic decorative variability.

Graves (1985) looked at the learning framework of the Kalinga to isolate which factors influenced the transmission of design style. He concluded that birth cohort of the potter and vessel size had the strongest influences on design (see also Longacre 1981). The effect of the individual on ce- ramic style has also been studied (Hardin 1977; Hill and Gunn 1977). This work, for example, focuses on how the motor habits of individuals influ- ence ceramic decoration (Hill 1977).

Hodder (1978, 1982a) and others have emphasized that ceramic deco- ration can play an active role in information exchange (see also, papers in Hodder 1982b; Graves 1982; Pollock 1983; Wobst 1977). For example, Miller (1982) describes how the Indians of Dangwara, central India, try to raise their social position by emulating design styles of the higher caste. Similarly, Braithwaite (1982) illustrates how ceramic decoration and ves- sel distribution “symbolize” male and female roles among the Azande.

Clearly, significant progress has been made in ascertaining the social

CERAMIC STYLE ANALYSIS 399

causes of ceramic decorative variation, and also how those social factors can be discerned using ceramic vessels in systemic context. However, with few exceptions (e.g., Graves 1982) these advances have not been implemented in prehistoric studies. Archaeologists continue to count and type sherds and do analyses that are identical to those performed 25 years ago (see also Rice 1982, 1984:249). Why is there a disjunction between ceramic studies in ethnoarchaeology and prehistoric archaeology? The two areas of decorative ceramic research differ because of presently in- compatible units of observation and units of analysis.

An analytic unit has both behavioral and material components and has been defined, for archaeological deposits, as the material produced by specified cultural and noncultural formation processes (Schiffer and House 1977:251). Observational units are that part of the material culture used by the researcher as evidence. Tables 4 and 5 list the analytic and observational units of a sample of ethnoarchaeological and archaeological studies. Although not exhaustive, these cases are thought to represent studies that use ceramic data to make social, organizational, and behav- ioral inferences. Tables 4 and 5 demonstrate that analytic and observa- tional units in ethnoarchaeology and prehistory are often very different.

TABLE 4 THE ANALYTIC AND OBSERVATIONAL UNITS USED IN A NUMBER OF

ETHNOARCHAEOLOGICAL STUDIES

Author

Arnold (1984)

Braithwaite (1982)

Deal (1983)

DeBoer (1984)

Graves (1985)

Hardill (1983)

Hodder (1982a:37-48)

Lathrap (1983)

Longacre (1981)

Miller (1982)

Analytic unit

Vessels produced by a community of potters

Pottery within the homestead compound

Household pottery inventory

Vessels produced by the potter, compound, and village

Vessels produced by the potter within kin and work groups

Vessels produced and used by the village of San Jose

Material inventory of Baring0 District compounds

Vessels produced by the extended family compound and by the village

Vessels produced by the potter and kin and work groups

Pottery produced by the pottery caste in the village of Dangwara

Observational unit

Whole vessels

Whole vessels

Whole vessels

Whole vessels (design borders)

Whole vessels

Whole vessels

Whole vessels

Whole vessels

Whole vessels

Whole vessels

400 SKIBO, SCHIFFER, AND KOWALSKI

TABLE 5 THEANALYTICANDOBSERVATIONAL UNITS USED INANUMBEROF

ARCHAEOLOGICAL STUDIES

Author

Braun (1985)

Brunson (1985)

Clemen (1976)

Hill (1970) Longacre (1970)

Kintigh (1985)

Plog (1980)

Washburn (1977)

Washburn and Matson (1985)

Analytic units

The frequency of ceramic attributes deposited by the former inhabitants of the site

The frequency of surface ceramic types deposited by the former inhabitants of the site

Ceramic design elements from floor and fill contexts deposited by the room’s former occupants

Ceramic design elements from floor and till contexts deposited by the room’s former occupants

Ceramic attributes from floor, fill, and extramural contexts deposited by the former inhabitants of three sites in the El Morro Valley

The frequency of surface ceramic attributes deposited by the former inhabitants of the site

The ceramic design structure of excavated ceramics deposited by the former inhabitants of a site

The ceramic design structure of surface and excavated ceramics deposited by the former inhabitants of a region

Observation units

Rim sherds

Sherds

Sherds

Sherds

Sherds

Sherds

Whole vessels and sherds

Sherds

Units of Observation

The unit of observation in ethnoarchaeology is the whole vessel in systemic context-vessels currently in use (Table 4). For example, to identify design correlations in the Quinua region, Peru, Arnold (1984) employed four utilitarian wares made and used in the community. Graves’s (1985) study of design variation in the Kalinga village of Dang- talan was based on vessels manufactured by different kin and work groups in the village over a short period of time. Similarly, Braithwaite (1982) based her study on vessels in use in one Azande community.

The units of observation for the archaeologist, however, are often sherds (Table 5). Based on ethnoarchaeological ceramic studies, sherds may not be appropriate for some types of ceramic decorative analyses. First, important stylistic data may be only partially represented on sherds. This idea is best expressed by Arnold (1984:147):

CERAMIC STYLE ANALYSIS 401

This study suggests that the use of pottery design to reconstruct ancient patterns of social interaction must be done within the behavioral domain of particular vessel shapes, layout types, and design zones. Sherds, of course, represent only random and arbitrary subdivisions of the vessel shapes which do not represent discrete units of cultural behavior. They should not be used as ad hoc boundaries for defining design elements. Furthermore, she& do not provide the structural data necessary for adequately describing designs.

Second, myriad formation processes affect the differential use, break- age, deposition, and recovery of ceramic vessels. A collection of sherds used uncritically in an analysis cannot inform on these processes and can lead to interpretive errors. The reanalysis of the Broken K Pueblo ce- ramic collection demonstrates how sherds, as units of observation, can create patterning unrelated to the distribution of past social groups. In principle, the whole vessel must serve as the unit of observation in ce- ramic design studies even if only sherds are available, and the processes that reduce vessels to sherds (Sullivan 1978208) must be investigated. This is best accomplished by a program of sherd refitting and matching to parent vessels (discussed below in more detail).

Units of Analysis

In the ethnoarchaeological cases listed in Table 4, both elements of the analytic unit (cultural and material) are easily identified. The material component is the whole pot in systemic context and the behavioral or organizational component is the potter, the household, or the settlement. However, in most archaeological examples (Table 5), the analytic units are vague and difficult to identify. For example, each case in Table 5 is followed by a generic statement, “deposited by the former inhabitants of the site, room, or region,” which highlights the imprecision of analytic units in prehistory. Although some information about the past can be gained by using such analytic units, if the objective is to identify and explain ceramic variability at the level of the individual potter, the house- hold, or groups of households (as in ethnoarchaeology), the vague ana- lytic units of Table 5 would be inappropriate. To infer intra- or intercom- munity social relations from ceramic design there must be precise knowl- edge about how the deposits formed. It makes a great difference whether the material was deposited, for example, on a room floor, in a room fill, or on a trash mound; the frequencies of ceramic types and forms from each context will vary because floors, fills, and extramural trash deposits are created by different processes.

The reason for the disjunction between ceramic studies in prehistory and ethnoarchaeology is that little effort has been devoted to devising models, within analytic theory (sensu Clarke 1973; Schiffer 1988; Sulli-

402 SKIBO, SCHIFFER, AND KOWALSKI

van 1978), that take into account formation processes. For the units of observation and analysis in ethnoarchaeology to approximate those in prehistory, the deposit’s formation processes must be inferred. The fol- lowing section discusses how archaeological and ethnoarchaeological ce- ramic analyses can make contributions to analytic theory.

MODEL BUILDING IN ANALYTIC THEORY

Analytic theory provides the framework within which the archaeologist can seek to identify formation processes and segregate items deposited by different processes to make inferences about the past (Sullivan 1978:206; Schitfer 1983, 1987:263-302). The following discussion focuses on the identification and interpretation of formation processes as they relate to ceramic design analysis in both archaeological and ethnoarchaeological studies. The analytic methods and models described can help to narrow the disjunction between units of observation and units of analysis in eth- noarchaeological and prehistoric archaeological ceramic design studies.

The Broken K reanalysis demonstrates that, conceptually, the unit of observation in ceramic design studies should be the whole vessel even if only sherds are recovered. Moreover, it is suggested that the unit of analysis needs to be clearly delineated by identifying the deposit’s for- mation processes. In archaeological ceramic analysis, both of the above can be fulfilled by what has been referred to as “pottery reassembly,” “refitting,” “conjoining,” or “cross-mending” (Schiffer 1987:285). Re- fitting and matching have been employed occasionally in studies attempt- ing to estimate the frequency of ceramic types or wares (e.g., McPherron 1967; Newell and Krieger 1949; Smith 1971), although factors such as vessel use-life and household depositional behavior complicate such es- timates (see Deal 1983:255-264). More recently, a number of investigators (e.g., Hally 1983; Nelson 1985; Sullivan 1983, 1988; Sullivan et al. 1986) have employed refitting and matching to identify and interpret a site’s formation processes. These analyses segregated various types of refuse, laying a foundation for a variety of inferences on, for example, the use of space, pottery manufacture, discard activities, and the employment of sherds as tools.

Generally, ceramic refitting is important for two reasons. First, it helps the archaeologist avoid using sherds from the same vessel as independent units in a design study. Moreover, this type of analysis will provide, in many cases, reconstructed large sherds that can serve as minimal units of observation in many design studies (e.g., Jernigan 1986; Washburn 1977). Second, conjoining and matching sherds permit the archaeologist to as- sess the appropriateness of the assemblage for making specific inferences. Not all data are appropriate for specific social, behavioral, or temporal

CERAMIC STYLE ANALYSIS 403

inferences (Sullivan 1978), but conjoining and matching can provide an effective means for making this type of assessment.

Ethnoarchaeology can also do more to remove the disjunction between archaeological and ethnoarchaeological ceramic design studies. Some ethnoarchaeological research has focused on disposal and abandonment behavior (e.g., Deal 1985; Foster 1960; Hayden and Cannon 1983), but few studies discuss the units of observation and analysis that can be immediately transferred to archaeological ceramic analyses (see also Deal 1983:252-253). The development of analytic methods and models would progress more rapidly if ethnoarchaeologists were to create simulated sherd assemblages, based on realistic formation processes, and then seek in those data patterns that were evident on pottery in systemic context. Deal’s (1983) ethnoarchaeological study among the Tzeltal Maya is an exemplary effort along these lines. Although he did not create simulated assemblages, he did document changes in the ceramic assemblages through a household’s life cycle (Deal 1983: 185-309). Deal concludes that ceramic diversity is the best method to document variability between households.

Although the isolated efforts by those working with archaeological ce- ramic assemblages and ethnographic collections make contributions to analytic theory, a project that combines the analytic techniques of ar- chaeology and ethnoarchaeology would provide the greatest insights. To test models of archaeological interpretation, Deal (1983:33&331) suggests that more research should be focused on recently abandoned sites where previous occupants can be interviewed. Bonnichsen (1973; see also Lange and Rydberg 1972; Longacre and Ayres 1968) performed such a study but within a traditional analytic framework. The project, unsuccessful in pre- dicting the activities that occurred at the site, served primarily to caution those performing traditional activity area research and to call for a re- evaluation of archaeological interpretive techniques.

A more productive approach today would be to excavate an abandoned room or household residence in a village that had been investigated by an ethnoarchaeologist (see also Chang 1988; Gorecki 1985). The archaeolog- ical analysis would include refitting and would be sensitive to the sundry processes that contributed to the formation of the deposit. Ideally, the excavation should be carried out unaware of the ethnoarchaeological re- sults and blind tests could be made of archaeological methods and infer- ences. For example, the ability of refitting and matching to segregate various processes (such as discard, abandonment, and loss) could be examined along with various sampling and recovery strategies. In addi- tion, archaeological inferences about the household, such as length of occupation, activities, and various social and economic elements, could be tested against the reconstruction based on the ethnoarchaeological

404 SKIBO, SCHIFFER, AND KOWALSKI

study. A project like this would make contributions to archaeological method and theory unattainable by any other means including interview- ing previous occupants of an archaeological site.

SUMMARY AND DISCUSSION

The analysis of ceramic decoration has been and continues to be an important avenue for documenting and interpreting temporal and spatial variability. Since the 1960s and the “New Archaeology,” ceramic deco- ration has taken on a new emphasis. Stimulated by ethnoarchaeological research, new information on social phenomena can now be extracted from ceramic designs. However, the advances made in design analysis in ethnoarchaeology are based on whole pots whereas archaeologists are often limited to the study of sherds. The Broken K reanalysis demon- strated that the use of sherds as the unit of observation can create pat- terning unrelated to the past behaviors of interest. Regrettably, the un- critical use of sherds in stylistic analyses is a problem that still plagues ceramic studies.

It was demonstrated that ethnoarchaeological studies, initially stimu- lated by the “ceramic sociologists,” have had little effect on contempo- rary ceramic stylistic analysis in prehistory. In an ethnographic setting, the archaeologist can observe whole pots and the unit of analysis (e.g., household, Workgroup, village) can be easily identified and studied. How- ever, prehistorians usually have only sherds as the unit of observation and behaviorally relevant units of analysis are much more difficult to isolate in the archaeological record. Many of the findings of ethnoarchaeology have not been implemented in archaeological ceramic analysis because of these differences in units of analysis and observation. In the case of Broken K, for example, the units of observation should have been designs on indi- vidual vessels or large sherds. The unit of analysis that should have been isolated is the refuse (of various types) deposited by individual house- holds. The only way to obtain these more appropriate analytic and ob- servational units is to identify the deposit’s formation processes. It was suggested that the way to begin this process is to (1) employ an intensive program of sherd conjoining and matching, (2) continue to document ethnoarchaeologically the processes that contribute to the formation of an archaeological deposit, and (3) test the efftcacy of analytic techniques such as refitting by excavating an ethnoarchaeologically known house or compound.

If progress is to be made in archaeological ceramic style analysis, in- ferences must be based on a clearly defined and understood data base. Ceramic variability is an important line of evidence for reconstructing past human behavior, but the advances made in understanding “why”

CERAMIC STYLE ANALYSIS 405

and “how” ceramic decoration varies are of little use archaeologically if we do not find ways to better interpret our sherd data.

ACKNOWLEDGMENTS

The research was supported by a grant from the National Science Foundation (BNS- 84-19935) to M. B. Schitfer. The Field Museum of Natural History permitted access to the Broken K collection and we are grateful to the museum staff, especially Phyllis Rabineau and Ronald Weber, for their assistance. Numerous individuals provided advice regarding the location of the Broken K material, among whom are David Gregory, Christine Gross, James Hill, William Longacre, Fred Plog, and John Rinaldo. We are grateful to Michael Deal, Christian Downum, Carol Kramer, William Longacre, Stephen Plog, J. Jefferson Reid, and especially Alan P. Sullivan for their comments and advice, although some ques- tions could not be resolved. Our writer’s group also offered support and good advice. Finally, we thank Barbara Montgomery for drawing the figures.

REFERENCES CITED

Arnold, Dean E. 1983 Design structure and community organization in Quinua, Peru. In Structure and

cognition ofart, edited by Dorothy Washburn, pp. 56-73. Cambridge University Press, Cambridge.

1984 Social interaction and ceramic design: community-wide correlates in Quinua, Peru. In Pots and potters: current approaches in ceramic archaeology, edited by Prudence Rice, pp. 133-161. Institute of Archaeology, University of California, Monograph 24.

Bonnichsen, Robson 1973 Millie’s camp: an experiment in archaeology. World Archaeology 4277-291.

Braithwaite, Mary 1982 Decoration as a ritual symbol: a theoretical proposal and an ethnographic study

in southern Sudan. In Symbolic and structural archaeology, edited by Ian Hod- der, pp. 80-88. Cambridge University Press, Cambridge.

Braun, David P. 1985 Absolute seriation: a times-series approach. In For concordance in archaeo-

logical analysis: bridging data structures, quantitative technique, and theory, edited by Christopher Carr, pp. 509-539. Westport, Kansas City.

Brunson, Judy L. 1985 Corrugated ceramics as indicators of interaction spheres. In Decoding prehis-

toric ceramics, edited by Ben Nelson, pp. 102-127. Southern Illinois University Press, Carbondale.

Chang, Claudia 1988 Nauyalik fish camp: an etlmoarchaeological study in activity-area formation.

American Antiquity 53:145-157. Clarke, David L.

1973 Archaeology: the loss of innocence. Antiquity 47:6-18. Clemen, R. T.

1976 Aspects of prehistoric social organization on Black Mesa. In Papers on the archaeology of Black Mesa, Arizona, edited by G. J. Gumerman and R. C. Euler, pp. 113-135. Southern Illinois University Press, Carbondale.

406 SKIBO, SCHIFFER, AND KOWALSKI

Deal, Michael 1983 Pottery ethnoarchaeology among the Tzeltal Maya. Ph.D. dissertation, Simon

Fraser University, Bumaby, British Columbia. 1985 Household pottery disposal in the Maya Highlands: an ethnoarchaeological in-

terpretation. Journal of Anthropological Archaeology 4~243-291. DeBoer, Warren R.

1984 The last pottery show: systems and sense in ceramic studies. In The many dimensions ofpottery, edited by S. E. van der Leeuw and A. C. Pritchard, pp. 527-568. University of Amsterdam Press, Amsterdam.

Deetz, James 1%5 The dynamics of stylistic change in Arikara ceramics. University of Illinois

Press, Chicago. Dumond, Don E.

1977 Science in archaeology: the saints go marching in. American Antiquity 42:330- 349.

Foster, George M. 1968 Life expectancy of utilitarian pottery in Tzintzuntzan, Michoacan, Mexico,

American Antiquity 25506-609. Gorecki, P.

1985 Ethnoarchaeology: the need for a post-mortem enquiry. World Archaeology 17(2):175-191.

Graves, Michael 1982 Breaking down ceramic variation: testing models of White Mountain Redware

design style development. Journal of Anthropological Archaeology 1:305-354. 1985 Ceramic design variation within a Kalinga village: temporal and spatial pro-

cesses. In Decoding prehistoric ceramics, edited by Ben Nelson, pp. 5-34. Southern Illinois University Press, Carbondale.

Hally, David J. 1983 The interpretive potential of pottery from domestic contexts. Midcontinental

Journal of Archaeology 8:163-l%. Hardin, Margaret Ann

1977 Individual style in San Jose pottery painting: the role of deliberate choice. In The individual in prehistory, edited by James N. Hill and Joel Gunn, pp. 109-136. Academic Press, New York.

1983 The structure of Tarascan pottery painting. In Structure and cognition in art, edited by Dorothy Washburn, pp. 8-24. Cambridge University Press, Cam- bridge.

Hayden, Brian, and Aubrey Cannon 1983 Where the garbage goes: refuse disposal in the Maya Highlands. Journal of

Anthropological Archaeology 2117-163. 1984 The structure of material systems: ethnoarchaeology in the Maya Highlands.

Society for American Archaeology Papers 3. Hill, James N.

1970 Broken K Pueblo: prehistoric social organization in the American Southwest. Anthropological Papers of the University of Arizona 18.

1977 Individual variability in ceramics and the study of prehistoric social organiza- tion. In The individual in prehistory, edited by James N. Hill and Joel Gum, pp. 55-108. Academic Press, New York.

Hill, James N., and Joel Gunn (editors) 1977 The individual in prehistory. Academic Press, New York.

CERAMIC STYLE ANALYSIS 407

Hodder, Ian 1978 The maintenance of group identities in the Baring0 District, West Kenya. In

Social organization and settlement, edited by D. Green, C. Haselgrove, and M. Spriggs, pp. 47-73. BAR International Series (Suppl.) 41.

1982a Symbols in action. Cambridge University Press, Cambridge. 1982b (editor) Symbolic and structural archaeology. Cambridge University Press,

New York. Jemigan, E. W.

1986 A non-hierarchical approach to ceramic decoration analysis: a Southwestern example. American Antiquity 51:3-20.

Kintigh, Keith 1984 Measuring archaeological diversity by comparison with simulated assemblages.

American Antiquity 49:44-54. 1985 Stylistic variation and social structure analysis of El Morro Valley decorated

ceramics. In Decoding prehistoric ceramics, edited by Ben Nelson, pp. 35-74. Southern Illinois University Press, Carbondale.

Lange, F. W., and C. R. Rydberg 1972 Abandonment and post-abandonment behavior at a rural Central American

house site. American Antiquity 37341w32. Lathrap, Donald W.

1983 Recent Shipibo-Combo ceramics and their implications for archaeological inter- pretations. In Structure and cognition in art, edited by Dorothy Washburn, pp. 25-39. Cambridge University Press, Cambridge.

Lischka, Joseph J. 1975 Broken K revisited: a short discussion of factor analysis. American Antiquity

40:220-227. Longacre, William A.

1970 Archaeology as anthropology: a case study. Anthropological Papers of the Uni- versity of Arizona 17. University of Arizona Press, Tucson.

1974 Kalinga pottery-making: the evolution of a research design. In Frontiers of unthropology, edited by Murray J. Leaf, pp. 51-67. D. Van Nostrand, New York.

1981 Kalinga pottery: an ethnoarchaeological study. In Patterns of the past: studies in honour of David Clarke, edited by I. Hodder, G. Isaac, and N. Hammond, pp. 49-66. Cambridge University Press, New York.

1985 Pottery use-life among the Kalinga, Northern Luzon, the Philippines. In De- coding prehistoric ceramics, edited by Ben Nelson, pp. 334-346. University of Illinois Press, Carbondale.

Longacre, William A., and James E. Ayres 1968 Archaeological lessons from an Apache wickiup. In New perspectives in arche-

ology, edited by Sally R. and Lewis R. Binford, pp. 151-159. Aldine, Chicago. Martin, Paul S., James N. Hill, and William A. Longacre

1966 Documentation for chapters in the prehistory of Eastern Arizona, III. Society for American Archaeology, Archives of Archaeology 27.

Martin, Paul S., William A. Longacre, and James N. Hill 1%7 Chapters in the prehistory of Eastern Arizona, III. Field Museum of Natural

History, Chicago, Fieldiuna: Anthropolgy 57. McPherron, Alan

1967 The Juntunen site and the Late Woodland prehistory of the Upper Great Lakes area. Museum ofAnthropology, University of Michigan, Anthropological Paper 30.

408 SKIBO, SCHIFFER, AND KOWALSKI

Miller, Daniel 1982 Structures and strategies: an aspect of the relationship between social hierarchy

and cultural change. In Symbolic and structural archaeology, edited by Ian Hodder, pp. 89-98. Cambridge University Press, Cambridge.

Nelson, Ben A. 1985 Reconstructing ceramic vessels and their systemic contexts. In Decoding pre-

historic ceramics, edited by Ben Nelson, pp. 310-328. Southern Illinois Univer- sity Press, Carbondale.

Newell, H. P., and A. D. Krieger 1949 The George C. Davis site, Cherokee County, Texas. Society for American Ar-

chaeology, Memoirs 5. Plog, Stephen

1978 Social interaction and stylistic similarity: reanalysis. In Advances in archaeo- logical method and theory (Vol I), edited by Michael B. Schiffer, pp. 143-182. Academic Press, New York.

1980 Stylistic variation in prehistoric ceramics. Cambridge University Press, New York.

Pollock, Susan 1983 Style and information: an analysis of Susiana ceramics. Journal of Field Ar-

chaeology 2:354-390. Reid, J. Jefferson

1978 Response to stress at Grasshopper Pueblo, Arizona. In Discovering past behav- ior: experiments in the archaeology of the American Southwest, edited by Paul F. Grebinger, pp. 195-213. Gordon and Breach, New York.

Rice, Prudence M. 1982 Pottery production, classification, and the role of physicochemical analysis. In

Archaeological ceramics, edited by J. S. Olin and A. D. Franklin, pp. 47-56. Smithsonian Institution, Washington, DC.

1984 Overview and prospect. In Pots and potters: current approaches in ceramic archaeology, edited by Prudence Rice, pp. 245-255. Institute of Archaeology, University of California, Monograph 24.

1987 Pottery analysis: a sourcebook. University of Chicago Press, Chicago. Schiffer, Michael B.

1983 Toward the identification of formation processes. American Antiquity 48:675- 706.

1985 Is there a “Pompeii premise” in archaeology? Journal of Anthropological Re- search 41:1&41.

1987 Formation processes of the archaeological record. University of New Mexico Press, Albuquerque.

1988 The structure of archaeological theory. American Antiquity 53:461-5. 1989 Formation processes of Broken K Pueblo: some hypotheses. In Quantifying

diversity in archaeology, edited by R. D. Leonard and G. T. Jones, pp. 37-58. Cambridge University Press, Cambridge.

Schiffer, Michael B., and John H. House. 1977 An approach to assessing scientific significance. In Conservation archaeology:

a guide for cultural resource management studies, edited by M. B. Schiier and G. J. Gumerman, pp. 249-257. Academic Press, New York.

Smith, Watson 1971 Painted ceramics of the western mound at Awatovi. Papers of the Peabody

Museum of Archaeology and Ethnology 38.

CERAMIC STYLE ANALYSIS 409

Stanislawski, Michael B. 1973 Review of “Archaeology as anthropology: a case study.” American Antiquity

38117-122. 1977 Ethnoarchaeology of Hopi and Hopi-Tewa pottery making: Styles of learning.

In Experimental Archaeology, edited by D. Ingersoll, J. Yellen, and W. Mac- donald, pp. 378408. Columbia University Press, New York.

Sullivan, Alan P., III 1978 Inference and evidence in archaeology: a discussion of the conceptual problems.

In Advances in archaeological methodand theory (Vol. l), edited by Michael B. Schiffer, pp. 183-222. Academic Press, New York.

1983 Storage, nonedible resource processing, and the interpretation of lithic scatters in Sonoran Desert lowlands. Journal of Field Archaeology 10~309-323.

1988 Prehistoric Southwestern ceramic manufacture: the limitations of current evi- dence. American Antiquity 53:23-35.

Sullivan, Alan P., III, James M. Skibo, and Mary Van Buren 1986 Sherds as tools: the role of ceramics at an agave processing site in the Northern

Tucson Basin. Paper presented at the Tucson Basin Conference, Tucson. Washburn, Dorothy K.

1977 A symmetry analysis of Upper Gila area ceramic design. Papers of the Peabody Museum of Archaeology and Ethnology 68.

Washburn, Dorothy K., and R. G. Matson 1985 Use of multidimensional scaling to display sensitivity of symmetry analysis of

patterned design to spatial and chronological change: Examples from Anasazi prehistory. In Decoding prehistoric ceramics, edited by Ben Nelson, pp. 75- 101. Southern Illinois University Press, Carbondale.

Whallon, Robert 1968 Investigations of late prehistoric social organization in New York State. In New

perspectives in archeology, edited by Sally R. and Lewis R. Binford, pp. 22E 244. Aldine, Chicago.

Wobst, H. Martin 1977 Stylistic behavior and information exchange. In papers for the director: Re-

search essays in honor of James B. Griffii, edited by Charles E. Cleland, pp. 317-334. University of Michigan, Museum of Anthropology, Anthropological Papers 61.