CHAPTER 4: RESEARCH DESIGN: IMPLEMENTING AN EVOLUTIONARY APPROACH TO THE CORRUGATION PROBLEM
Many different frames of reference may be causally relevant to the problem of the development and spread of corrugated pottery in the American Southwest. The pottery itself may be both replicator and interactor competing with other technological options known to the pottery makers and users for accomplishing some task. The behaviors and ideas required to make and use corrugated pottery may be the replicators while the pottery itself is the interactor in the particular human cognitive and behavioral environment present at the time. Plain and corrugated pottery may be traits of human organisms or populations, which affect their biological replicative success. The alternative pottery technologies may also be traits of larger social systems, potentially affecting the replicative success of these systems, or possibly just swept along as a consequence of differential replicative success at these larger scales.
Determining which of these frames of reference were causally involved in the rise of corrugated pottery requires that we know: (1) how corrugation developed out of plain pottery and then spread across the northern Southwest; (2) how the different plain and corrugated pottery vessels were made and used; (3) if there were any differences in the cost or performance of plain and corrugated vessels during manufacture and use; and (3) the relevant environmental contexts in which the various changes occurred. In the next section, I will discuss each of these data requirements in turn, and describe the analytical approaches I employ to generate the required data. These include analyses of selected pottery collections, and a series of experiments performed with replicas of plain and corrugated vessels. In the next section of this chapter, I discuss how I will use these data and the evolutionary approach developed in the previous chapter to formulate and begin testing hypotheses to account for corrugated pottery from different frames of reference. The final section presents information on the particular pottery assemblages analyzed for this study including their selection criteria, dating, sample selection, and potential biases.
Data Requirements
History of Corrugation
What was the sequence of technological changes that led from completely plain vessels to full- body corrugation? Where did these changes first take place, and how did they spread across the northern Southwest? We must answer these questions if we are to identify relevant frames of reference and construct testable hypotheses. Currently, most descriptions of pottery in the American Southwest involve assigning individual pottery specimens to types. These types were created to distinguish the temporal order of assemblages, not to track technological changes. Consequently, the types tend to be defined by an uneven and, often, complex mixture of attributes that do not allow one to isolate particular technological features. This is especially true for utility wares, which are generally given short shrift because of their perceived lack of stylistic content needed to make the best temporal distinctions. Although the data on various utility ware types have allowed the reconstruction of the general outline of change discussed in Chapter 2, they do not allow one to construct the kind of detailed technological histories required for an evolutionary account. To generate the needed technological descriptions, I must make new observations of samples of utility ware pottery that more fully document how the pottery was manufactured at different points in time, and both complement and enhance the existing published data. These new observations should be made on attributes of ancient utility ware pottery from selected assemblages that are dated accurately to different times during the shift from plain to corrugated vessels. Measuring attributes, rather than simply making typological distinctions, allows me to document the history of particular innovations as well as the association of technological features through the creation of classes by the paradigmatic or mutually exclusive formation of attribute combinations. Being able to distinguish individual features as well as combinations of features makes it possible to document the scale or scales at which replication may have taken place. Details on the pottery assemblages selected for these analyses, and the selection criteria employed are provided later in this chapter. The methods and techniques used in the specific analyses, and the technological history constructed from the various observations are presented in Chapter 5.
Once we have a better understanding of the technological changes involved in the development of corrugation, we can reexamine published data to construct a more complete picture of the spread of corrugation across the Southwest. This involves compiling information on the location and timing of the first appearance of particular corrugation traits, and tracking the spread of these traits to different areas of the Southwest. This is easier to do for some traits than for others because the pottery types often obscure particular traits by mixing them in unknown ways. Consequently, I focus on a few traits that can be reliably distinguished in the published data on utility ware types. This reconstruction of the spread of corrugation throughout the Southwest is presented in Chapter 5.
Pottery Production and Use
Documenting the production and use of utility pottery provides important information on the interaction dimension of the Darwinian process. Normally, when we think of how a pottery vessel may have interacted with its environment, we only consider how the pot was used. However, the process by which the pottery comes into being is just as important a part of a pot's environment as the conditions of use. Not only must a vessel survive the production process before it can be used, but pottery production can have very different functional and semiotic requirements than use. These differences between production and use can result in features of vessels that are puzzling when thought of from the use stand point alone. Did corrugation develop in a homogeneous or multifaceted production and use environments? What kinds of conditions were the utility vessels subjected to during production and use? Were there any changes in the production and use of utility wares during the transition from plain to fully corrugated vessels? These are the kinds of questions that must be answered for an understanding of Darwinian interaction that occurred during the development and spread of corrugation.
Archaeological analyses of the processes involved in pottery production come in many forms, and have been carried out with increasing sophistication during the last decade ( Bey and Pool 1992; Costin 1991; Mills and Crown 1995a; Pierce et al. 1999). Although the production of utility ware pottery in the Southwest has been the focus of very few studies, some published information does exist. This information includes both direct and indirect evidence of pottery production. Relying heavily on these published sources and contributing original data, when available, from my analyses of pottery collections, I organize and synthesize our current knowledge of the production of Southwestern utility wares. The results of this synthesis are presented in Chapter 6.
Data on how plain and corrugated utility ware vessels were used will be derived from the morphology of the ancient vessels and patterns of use-related alterations they display. Again, existing typological data on Southwestern utility wares provide very little information on vessel use. Data on utility ware types do furnish some information on the shape and, occasionally, size of vessels, but data on possible use-related alterations, the most important source of information on use, are extremely rare and usually anecdotal at best. To generate more detailed information on the use of utility ware vessels during the development of corrugation, I record data on the shape and size of vessels, and possible use-related alterations from pottery sherds analyzed from the same assemblages selected for the technological analyses discussed above. I use these data in combination with published information to develop as complete a picture as possible of vessel use during the transition form plain to corrugated utility ware pottery. The methods and results of these analyses are also presented in Chapter 6.
Engineering Properties
Evolutionary explanation involving selection as a mechanism necessitates that interaction with the environment differentially bias replication. Interactions that are more efficient or effective tend to result in greater replication. Consequently, being able to distinguish selection from other sorting mechanisms requires that we know something of the engineering properties of the interactors. Ever since Darwin, evolutionary explanations have involved a component of reverse engineering to document the efficiency and effectiveness of design features ( Burian 1983; Dennett 1995). One of the significant challenges in conducting reverse engineering studies is determining what entity should be studied. Given our focus on changes in pottery technology, the engineering properties of plain and corrugated vessels are a logical place to begin. However, if multiple frames of reference were causally relevant to the development and spread of corrugation, we may also need to know the engineering properties of other entities such as the cognitive architecture of the people involved, or the larger-scale sociocultural systems in which the people were immersed.
To document the engineering properties of the pottery, I conducted a set of experiments designed to assess and document the relative cost and performance values of plain and corrugated vessels. I performed these experiments with replicas of ancient plain and full-body corrugated vessels. The design of these replicas is based on knowledge gained from analyses of ancient pottery fragments, with the goal of making the replicas resemble the ancient pottery as closely as possible. The procedures used to manufacture these replicas are described in Chapter 7. I designed the experiments to measure specific engineering properties that previous research and speculation have identified as potentially different between plain and corrugated. These include manufacture costs, heating effectiveness, ease of handling, and vessel use-life. In addition to these experiments, I draw on physical theory, the results of earlier experiments, and certain characteristics of ancient pottery assemblages when they are available and appropriate to the problem. Chapter 7 presents the methods and results of these various engineering studies. I address the properties of other entities that may be relevant to explaining corrugation in Chapter 8 in the context of formulating hypotheses regarding the development and spread of corrugated pottery.
Contexts of Change
As discussed in the previous chapter, the environment, or context of change, is the third part of a logical triad of conceptual units in Darwinian evolutionary theory. Defined as the entities interactors interact with, the environment is a critical factor in determining the kinds of sorting mechanisms that may be involved in any particular example of evolutionary change. Given the multiple frames of reference approach I have adopted in this study, there may be several different environments pertinent to the corrugation problem, one for each causally relevant frame of reference. In addition, the environment plays an important role in stimulating potential non-Darwinian processes of change including the behavioral plasticity of biological organisms, and, possibly, the orthogenetic growth of societies and institutions. Consequently, identifying and describing the relevant environments can only occur in the context of explanation. Hypotheses formulated to account for the rise of corrugation must specify the relevant environment(s) to be complete, and generic environmental descriptions are of little value. Therefore, I rely almost exclusively on data in the primary and secondary literature of the prehistoric Southwest and human cognition to characterize the environment in the various hypotheses for explaining corrugation presented in Chapter 8.
Formulating and Testing Hypotheses
In the normal notion of the scientific process, one begins a body of research with hypotheses, which clearly state the possible explanations for some phenomenon, and then proceeds to test these hypotheses with a battery of data. Although the true significance of this view of the scientific method is often over stated ( Bauer 1992), it still represents an admirable goal. However, archaeologists rarely find themselves in a position where they can proceed in this manner. Usually, we lack sufficient understanding of the empirical problem we hope to explain to formulate detailed, testable hypotheses. This is compounded by the poor state of theory development in archaeology resulting in hypotheses that often take the form of intuitively appealing rationalizations for which empirical testing is nearly impossible.
Consequently, the goal of my research on the problem of corrugation is to present a set of useful hypotheses after we have a better understanding of the problem through generating the data described above. Attempting to formulate testable evolutionary hypotheses regarding the rise of corrugation would be foolish without first having a clear understanding of what it is we are trying to explain. Our understanding of the requirements for evolutionary explanation and our general knowledge of the corrugation problem are sufficient to specify the data needed to formulate, but not necessarily test, viable evolutionary explanations. Once the necessary data on the history, environment and interaction of Southwestern utility wares are available, then we can turn to the problem of formulating and, possibly, testing hypotheses with a much greater likelihood of success.
Commonly, evolutionary accounts take the form of historical narratives that specify the nature and possible causal relationships of the various replicators, interactors, and environments relevant to a particular problem ( Hull 1981; O'Hara 1992; Ruse 1971). I take this approach in presenting my explanations for corrugation in Chapter 8. In that chapter, I divide the corrugation problem into four parts corresponding to the initial adoption of neck banding, the introduction of variation in neck banding, the adoption of full-body corrugation, and the return to plain utility wares. For each part of the corrugation problem, I present a historical narrative that integrates the knowledge of different parts of the corrugation problem developed in Chapters 5 through 7, and presents possible explanations for changes seen in each period. To the extent possible with current evidence, I attempt to test the various alternative explanations formulated for each part.
The Pottery Collections: Selection, Sampling and Potential Biases
Ideally, to document the history of the development, spread and use of corrugated pottery, I should analyze samples of ancient pottery representative of the entire temporal and spatial range of the transition from entirely plain to fully corrugated vessels. Unfortunately, given the vast number of pottery collections involved, the ideal approach is not practical in this case. Consequently, I was forced to make several decisions to determine which pottery assemblages I would analyze for this study. Four levels of decisions needed to be made in designing the detailed analysis of utility ware pottery assemblages. First, I needed to select the region from which the assemblages would be drawn. Second, I had to frame the criteria I would use for selecting individual assemblages, and then make the selections based on data available for a larger number of sites. Third, I made sampling decisions regarding what pottery would be analyzed in each assemblage. Finally, I chose the observations I would make on pottery in these samples. Each of these decisions affects the kind and quality of the resulting data. Consequently, this section describes how and why I made these decisions, and assesses their impacts on data reliability and validity.
Selection of the Region
Corrugated pottery was used over a large area of the Southwestern U.S. and parts of northern Mexico. Rather than analyze pottery from widely separated areas, I chose to document the development of corrugation in one region well, and rely on published data to extrapolate these findings to other areas. Few regions in the Southwest contain a sufficiently large number of well-dated pottery assemblages spanning the period of change from plain to corrugated pottery. Of those possessing a substantial number of dated assemblages, the Mesa Verde region of southwestern Colorado stands out for several reasons. The region has a long history of archaeological and paleoenvironmental research including some of the earliest studies to address the corrugation problem. Nearly 1/3 of all tree-ring-dated sites in the entire American Southwest are located in southwestern Colorado ( Robinson and Cameron 1991). In addition, several of these sites date to the period of interest for this study, and have yielded substantial pottery assemblages from recent, well- controlled excavations. Finally, most of these pottery collections are available for study at easily accessible repositories located within the region. Given all of these qualities, the Mesa Verde or northern San Juan region seemed the most likely area to contain an adequate number of assemblages meeting the selection criteria described in the next section.
Selection of the Assemblages
Recent archaeological research in the Mesa Verde region, particularly the Dolores Archaeological Program and other work associated with the development of McPhee reservoir and related irrigation canals, has generated numerous collections dating to the period of change from completely plain to full-body corrugated vessels. In selecting among these collections for inclusion in this study, I strove for collections containing assemblages generated by well-dated, functionally equivalent, short duration occupations that provide even coverage of the period from AD 750 to 1050. Collections meeting the dating criterion include those with multiple absolute dates, either tree-ring dates, archaeomagnetic dates, or both (radiocarbon dates are rarely obtained in this area of the Southwest because of the dominance of tree-ring dating). In addition, these absolute dates must occur in an understandable and reliable association with the material to be analyzed, and the relative frequencies of painted pottery types in the selected assemblages must produce a properly ordered seriation. To control for functional variation in activities that might bias the formation and composition of the assemblages, I restricted consideration to collections from habitation sites with architectural features and recognizable trash or midden deposits. Finally, I distinguished relatively short duration occupations by a lack of substantial architectural remodeling, and a reasonable clustering of dating evidence. In addition, because assemblages form the unit of comparison for this study, selected collections must be relatively large, and collected in such a way that they are reasonably representative of the material deposited during the occupation. I used an arbitrary assemblage size cut-off of at least 300 utility ware fragments. Although probably a higher cut-off than necessary, I wanted to avoid, as much as possible, the disruptive effects of small samples. This is one of the possible biases I examine at the end of this section and in Appendix A. Finally, to minimize biases introduced during the collection of the material in the field, selected collections must have been generated by systematic surface collection or screening of excavated deposits sampled in a representative manner.
Once I identified collections for inclusion in this study, I further sampled these collections for the actual specimens to be analyzed. To secure samples representative of the occupation as a whole, I selected only material from trash deposits. I focus on trash middens because the chance that individual sherds in the same collection unit derived from the same vessel is greatly reduced relative to assemblages collected from indoor and outdoor use or living surfaces. Thus, trash deposits have a better chance of yielding the maximum amount of data on variation for a given sample size. In addition, the composition of refuse assemblages is less susceptible to distortion by abandonment-related formation processes than exposed activity surfaces ( Schiffer 1987). A comparison of variance in three collections under consideration for selection showed that trash deposits contain considerably less within sample variance in the abundance of pottery types than samples from activity surfaces. Finally, I chose to analyze only the utility ware pottery. In the American Southwest, pottery has been traditionally divided into groups based on the presence or absence of slip, polish, and paint. The utility wares consist of those vessels lacking these surface modifications. In the Mesa Verde region, these utility wares normally have a dull gray appearance produced by a neutral to reducing firing atmosphere, and are thus usually identified as gray wares. This gray, utility ware category includes the vast majority of pottery displaying corrugation in the northern Southwest.
Assemblage Descriptions
Application of these collection and assemblages selection criteria resulted in the inclusion of six assemblages from the Mesa Verde region in the current study. The following descriptions present very general information on the sites and the utility ware assemblages selected for analysis from these sites. More detailed information on each of these sites can be found the reports cited in each section.
Dos Casas Hamlet (5MT2193)
Dos Casas Hamlet is a residential site consisting of two, sequentially occupied pit structures with six associated post and adobe surface structures and several outdoor pit features, and appears to have been occupied by a single household ( Brisbin et al. 1986). The site is located in the Sagehen Flats area approximately 18 km north of Cortez, Colorado, and 2.5 km west of the old course of the Dolores River in the area now submerged by McPhee Reservoir (Figure 4). Members of the Dolores Archaeological Program carried out surface collections and excavations at the site prior to inundation by the reservoir, but after the site had been ploughed during a short period of cultivation. Artifacts over the entire surface of the site were collected in contiguous 4 by 4-meter squares while excavation focused on the areas with structural remains. All of the surface structures and the two pit structures were completely excavated either by hand or with heavy machinery. Material recovered from floor contexts and dense midden deposits was screened dry through 1/4-inch mesh. These excavations revealed that pit structure 2 was built after and slightly superimposed over pit structure 1. Tree-ring dates on 23 charred roof beams recovered from pit structure 2 indicate that the structure was constructed between AD 769 and 771. Two dates on roof beams from pit structure 1 suggest that it was built around AD 760. Characteristics of the architecture, site layout, and painted pottery (Figure 5) match those found at other sites dated to the latter half of the 8th century in the Mesa Verde region.
Figure 4. Map of the Mesa Verde region sowing the location of the six sites from which utility
ware sherds were analyzed for this study.
Figure 5. Seriation of painted pottery types recovered from the six sites from which utility ware
sherds were analyzed for this study. Error bars = 90% confidence intervals.
A total of 315 gray ware sherds were selected for analysis from Dos Casas Hamlet. Almost half of these sherds (n=151) were recovered from trash used to fill pit structure 1 during and after the construction of pit structure 2. The remaining sherds were collected from the surface of the trash dump located in subarea 2 south of the pit structures. This trash dump appears to have accumulated throughout the use of both pit structures. Consequently, the sherds analyzed from Dos Casas Hamlet constitute a mixture of material deposited from the middle to late eighth century AD.
Periman Hamlet (5MT4671), Area 1
Area 1 of Periman Hamlet is a residential site consisting of one pit structure and 19 post and adobe surface rooms with associated outdoor features, and appears to have been occupied by three contemporary households ( Wilshusen 1986). The site is located approximately 22 km north of Cortez, Colorado, on an alluvial fan immediately east of the Dolores River where the canyon narrows north of Sagehen Flats (Figure 4). The site was investigated by the Dolores Archaeological Program in advance of inundation by McPhee Reservoir. Investigations included surface collection of the entire site in contiguous 4 by 4-meter units, intensive excavation of the surface room block, plaza and pit structure, and test excavation of the trash dump located south of the pit structure. A combination of hand removal and the use of heavy machinery accomplished these excavations. Material excavated from floor/use surface contexts and dense midden deposits were screened dry through 1/4-inch mesh. A coherent site layout and lack of evidence of substantial renovation or abandonment of structures suggest that only one occupation occurred in Area 1 of Periman Hamlet. Seriation of painted pottery (Figure 5) and three archaeomagnetic dates indicate that the occupation dates to the first half of the 9th century AD.
The sample of gray ware pottery selected for analysis from Periman Hamlet consists of 785 sherds recovered from two trash dumps at the site. Complete excavation of two trash filled borrow pits located north (Feature 25) and east (Feature 144) of the surface room block yielded 323 sherds. Excavation of four systematically selected 2 by 2 meter test pits in the trash dump south of the pit structure produced the remaining 462 sherds included in the sample. The eastern borrow pit (Feature 144) was truncated by construction of a surface room suggesting that the trash deposited in this pit may date to the early part of the occupation at Periman Hamlet. However, most of the material analyzed accumulated during the entire occupation of the site, which likely lasted between 20 and 50 years.
Duckfoot Site (5MT3868)
This residential site consists of four pit structures, 20 post and adobe surface rooms north of the pit structures, and a trash dump located south of the pit structures, and appears to have been occupied by at least three contemporary households ( Lightfoot 1994; Lightfoot and Etzkorn 1993). The site is located on the mesa top 5 km west of Cortez, Colorado (Figure 4), and was excavated from 1983 to 1987 by Crow Canyon Archaeological Center. The architectural, courtyard and midden areas of the site were completely excavated by hand and all excavated material was screened dry through 1/4-inch mesh. A total of 375 tree- ring dates was obtained on construction beams recovered mainly from the pit structures with a few samples from other parts of the site. Based on these tree-ring dates, seriation of painted pottery (Figure 5), and construction evidence, the site appears to have been continuously occupied from AD 855 to 880.
I selected a sample of 1504 gray ware sherds for analysis from the Duckfoot Site. All of these sherds were recovered from 2 by 2-meter pits excavated into the trash dump or midden south of the pit structures. Because the complete excavation of the trash dump produced almost 60,000 gray ware sherds, I used a probabilistic sampling approach to select the sherds included in the analysis. Unfortunately, existing information on the collection made it impossible to identify individual pottery sherds that would allow a simple random sample of gray ware sherds, which would produce the most representative sample from the midden. Consequently, I performed a cluster sampling based on provenience units similar to that prescribed by Cowgill (1964). Restricting the sampling to the trash dump should minimize the adverse impacts on sample variance commonly introduced by cluster sampling since sherds from the same vessel are less likely spatially associated than in other depositional contexts. I drew the analyzed sherd sample by randomly selecting three of the 100 pits excavated into the trash dump. The grid coordinates of the three 2 by 2-meter units selected are 66N/93E, 56N/111E, and 62N/101E. The relative frequencies of gray ware pottery types in the random cluster sample closely match those in the total midden collection indicating that the small sample analyzed for this study adequately represents the larger trash dump population.
5MT8371
Site 5MT8371 consists of a single, shallow pit structure, adjacent outdoor pit features and small trash dump, but apparently lacks any associated surface rooms ( Dykeman 1986). The presence of a pit structure with a rich assortment of floor features and a discrete trash dump indicate that this was a single household residential site. However, the lack of surface architecture suggests that the occupation was for a short duration (probably less than 20 years), or possibly on a seasonal basis. The site is located on the mesa top near the head of Negro Canyon approximately 30 km northwest of Cortez, Colorado (Figure 4). Complete hand excavation of the pit structure, adjacent outdoor use surface, and trash midden by a team from the Division of Conservation Archaeology, a private consulting firm, occurred in 1984 after the site was exposed in a pipeline trench. Dates on five different charred beams from the collapsed roof of the pit structure yielded two cutting dates of AD 935 and a noncutting date of AD 940. Four radiocarbon dates on small charred wood pieces from features at the sites and seriation of painted pottery types (Figure 5) add further support for a middle 10th century date for the occupation of this site.
All of the gray ware pottery recovered from the midden, consisting of 326 sherds, was included in the sample analyzed for this study. Ten square meters of the midden were excavated primarily in contiguous 2 by 2-meter pits, and all of the material removed from these units were screened dry through 1/4-inch mesh to recover artifacts. The midden continued east of the excavated area for an unknown distance so it is impossible to determine the proportion the analyzed sample constitutes of the entire trash deposit. However, because the site appears to have been occupied only once for a relatively brief period, it is very likely that the analyzed sample adequately represents the pottery discarded during that occupation.
Gnatsville (5MT1786)
Gnatsville is a small residential site consisting of one pit structure, an unknown number of post and adobe surface rooms, various outdoor features, and two small trash dumps, and appears to have been occupied by a single household ( Kent 1989, 1991). The site is located approximately 20 km northwest of Cortez, Colorado, on the broad upland between McElmo and Yellowjacket Canyons (Figure 4). Investigations at the site were conducted as part of a university field school, and involved systematic surface collections and excavations in the areas containing structural remains and the trash dumps. All excavations proceeded by hand and the excavated material was screened dry through 1/4-inch mesh. Nine tree-ring dates (five from the burned and collapsed pit structure roof, two from the surface rooms, and two from midden 2) produced three cutting dates that indicate construction of the site in AD 1035. A single archaeomagnetic date from the surface room block and seriation of painted pottery (Figure 5) support this construction date estimate. Lack of architectural and feature remodeling and the shallow depth and small size of the trash dumps suggest a single, short duration occupation of the site probably lasting less than 20 years.
A sample of 918 gray ware sherds was selected for analysis from the Gnatsville site. All of the sherds were recovered through either surface collection or excavation in Midden 1, the largest and most dense trash dump at the site, located south of the pit structure and room block. Midden 2, a small, low density dump adjacent to the pit structure, appears to have been used for a very short time prior to the construction of the pit structure. I restricted my analyses to all of the sherds recovered from Midden 1 because it likely accumulated during the main occupation of the site.
Dobbins Stockade (5MT8827)
Dobbins Stockade is a residential site consisting of two sequentially occupied pit structures, four post and adobe surface rooms, numerous outdoor features, and a stockade that surrounded most of the structures and features ( Kuckelman 1988a). The site is located in a modern farm field on the broad upland between the heads of Hovenweep and Ruin Canyons approximately 32 km northwest of Cortez, Colorado (Figure 4). Investigations at the site in 1986 by a team from Complete Archaeological Service Associates included intensive surface collection, blading with heavy equipment, and hand excavation of the central portion of the site containing architecture. All of the hand-excavated material was screened dry through ?- inch mesh. Thirty-one tree-ring dates obtained on wood samples recovered from various structures and features at the site indicate that initial construction occurred around AD 1030 and the site was probably abandoned around AD 1060.
The only distinct trash dump documented at Dobbins Stockade was encountered during excavation of one of the pit structures (Pit Structure 2). Above a layer of roof fall and a thin deposit of wind and water transported sediment, a roughly 50 cm thick layer of trash filled almost half of the pit structure. After initially encountering the trash deposit in a backhoe trench, the layer was completely excavated by hand and screened dry through ?-inch mesh producing the 326 gray ware sherds included in the analysis for this study. The roof fall deposit below the trash produced three tree-ring dates the latest of which is a noncutting date of AD 1034. It is likely that the initiation of trash dumping in the pit structure dates to sometime after AD 1040 because a 10 to 15 cm thick deposit of naturally transported sediment covered the roof fall layer before the deposition of trash began. The trash probably accumulated more or less constantly until the site was abandoned around AD 1060.
Potential Assemblage Biases
Although I have addressed the conceptual validity or purpose of the various descriptions I am trying to produce, it is also important to evaluate the reliability of the units and particular observations used to generate these descriptions ( Ramenofsky and Steffen 1998). As I noted earlier, potentially disruptive biases can be introduced at several points in this analysis ranging from the choice of region to the size of the samples and the individual sherds analyzed. The actual impact of some of these potential biases on the accuracy and precision of the analyses performed for this study can only be speculated on at this point while others can be assessed quantitatively. In this section, I evaluate what we can currently know about the biases that potentially affect the data derived from the collection analyses.
Restricting the analyzed collections to the Mesa Verde region has both positive and negative effects. On the positive side, it is unlikely that an adequate number of collections would have been available in most other regions of the northern Southwest given the selection criteria I employed. However, as one might expect, the Mesa Verde region is probably not representative of changes in pottery technology that occurred in other areas of the Southwest. Existing data indicate that at least the timing and rate of change differ among the various regions of the northern Southwest. These differences become a significant issue in Chapter 8 as I attempt to explain the change from plain to corrugated pottery, and the importance of processes working over very large spatial scales becomes clear. Carefully examining published data from other regions helps minimize the impacts of the Mesa Verde region bias, but ultimately, comparable analyses will be needed from other areas to obtain reliable descriptions at the appropriate scales.
The problem with the representativeness of the region also extends to the six assemblages within the Mesa Verde region selected for study. There are two issues here. How thoroughly do these assemblages document the period of change from plain to corrugated pottery, and how well do individual assemblages represent the particular span of time they cover? The late 8th, 9th and early 11th centuries are reasonably well covered by the selected assemblages, but the 10th century is represented by only one collection which accumulated during the early to middle part of that century. This paucity of 10th century assemblages is in large part a reflection of the decrease in population that occurred in the Mesa Verde region during that time. This resulted in few documented assemblages, and only one meeting the selection criteria. Assessing the representativeness of these six assemblages of the pottery made and used at a given time in the Mesa Verde region is a more difficult problem. Existing typological descriptions suggest that the pottery in the Mesa Verde region changed in similar ways at more or less the same time across the region. However, these typologies tend to obscure and suppress the kinds of technological details of interest for this study.
Consequently, it remains an open question whether this typological uniformity can be reasonably extrapolated to the technological aspects of utility wares. In describing the history of corrugation in the next chapter, I attempt to assess the representativeness of various technological trends documented in the six assemblages by comparisons with other published information from the region. However, only comparable analyses performed on multiple collections from the same period of time will produce a completely reliable answer to this question.
Deciding to focus my analyses on the gray, utility ware pottery in the selected assemblages also introduces some potential biases. As noted earlier, the gray ware category in the Mesa Verde region does include most of the pottery with corrugation, but not all. Some slipped and painted pottery is corrugated to a very limited extent. Although I am familiar with the occurrence of corrugation on non-gray ware vessels, the decision not to analyze white and red ware pottery means that I will have to present this information anecdotally rather than quantitatively. In addition, as discussed in Chapter 6, the gray ware category includes pottery used in a wide variety of ways. This is particularly true for the earliest assemblages when most of the pottery produced was unslipped and unpainted. Later, as more and more of the different vessel forms were manufactured as white or red wares, the gray ware category came to include cooking and possibly storage vessels almost exclusively. These changes in the range of vessel forms analyzed for this study could bias some aspects of our understanding of the manufacture and use of vessels during the transition to full-body corrugation. Consequently, I try to control for the potential biases where their effects might be most disruptive when I present and analyze the data in Chapters 5 and 6.
The particular assemblages of sherds selected for analysis in this study could also present a biased picture of utility ware technology for two reasons. If the samples drawn from each of the assemblages are not large enough, sampling error could affect the accuracy and reliability of the resulting descriptions. If significant differences exist among assemblages in the size of sherds, these differences could also affect the documentation of certain characteristics. Appendix A presents a detailed analysis of these potential sample and sherd size biases for the six assemblages. Although sample size varies considerably among assemblages, all of the samples appear to be large enough to avoid disruptive sample size effects on assemblage richness and the relative frequencies of most attributes. However, the assemblages differ, sometimes substantially, in terms of sherd size. Most of this sherd size variation derives from differences in the formation histories of the assemblages, particularly the context of deposition and resulting differences in post-depositional breakage. The effects of these differences in sherd size and occasional incidents of inadequate sample size will be considered in Chapters 5 and 6 as data potentially affected by these biases are presented.
Return to Table of Contents
Go to Chapter 5
