Chapter 15: Artifacts, The Sand Canyon Archaeological Project: Site Testing

Back

15

Artifacts

Christopher Pierce, Mark D. Varien, Jonathan C. Driver, G. Timothy Gross, and Joseph W. Keleher

Contents

Introduction

Pottery Production
Pottery Exchange
Sherd Tools and Ornaments and Other Pottery Artifacts

Stone Tools and Manufacturing Debris

Introduction
Methods

Artifact and Assemblage Descriptions Chipped-Stone Tools Projectile Points Paleoindian Stemmed Large Side-Notched Large Corner-Notched Small Corner-Notched Medium Side-Notched Bull Creek Small Side-Notched Unusual Projectile Points Bifaces Probable Projectile Point Preforms and Fragments Stage and Finished Bifaces Drills Modified Flakes Modified Cores Other Chipped-Stone Tools Ground-Stone Tools Manos Metates Abraders Pestle Stone Disks Indeterminate Ground Stone Battered and Polished Stones Battered Stones Axes and Mauls Polishing Stones Polished Igneous Slabs Other Stone Artifacts Ornaments Other Modified Stones and Minerals Polished Ground Edge-Shaped Slabs Flaked/Battered Pigment Fire-Altered Rocks Other Unmodified Stones and Minerals Cores and Debitage Attribute Analysis of Debitage Results

Bone Artifacts: Identification, Classification, and Recording
Bone Artifact Frequency by Site
Selection of Raw Material
Awl Shape and Wear
Conclusions

Shell: Methods
Results
Summary and Discussion

Introduction
Christopher Pierce and Mark D. Varien

This chapter presents information on the portable artifacts recovered and analyzed from the 13 sites investigated during the Site Testing Program. These artifacts include pottery, stone tools and manufacturing debris, modified bone, and shell. Analytical methods and results for each of these general categories of artifacts are discussed in separate sections of this chapter. Data on unmodified plant and animal remains recovered during the Site Testing Program are presented in Chapters 16 and 18, respectively.

Pottery
Christopher Pierce and Mark D. Varien

Pottery is the most abundant type of artifact in each of the tested-site assemblages. Together, the 13 sites yielded a total of 81,144 sherds (459,991 grams), including five complete vessels. This section outlines the methods used to analyze the pottery and presents the results of the analyses, including a summary of the basic pottery assemblage data. Additional discussions focus on the quantification of pottery, the use of pottery types to date site components, an analysis of vessel form, and evidence of pottery production and exchange. Finally, the sherd tools, ornaments, and "other pottery artifacts" in the tested-site assemblages are described.

Methods

All pottery recovered from the tested sites was analyzed using a system that involved typological identifications and observation of selected attributes, including vessel form, vessel part, and surface finish. Pottery was submitted from the field in lots according to their specific proveniences. A single lot may have consisted of one sherd or many hundreds of sherds, depending on the number found in a given context. After the materials were cleaned by scrubbing with toothbrushes in tap water and assigned lot catalog numbers (referred to as FS numbers in the Crow Canyon system), all sherds were sorted into groups of the same type, form, part, and finish. The sherds in these groups were then counted and weighed.

The pottery was analyzed the Crow Canyon laboratory staff. Crow Canyon participants conducted basic sorting; all sorting by participants was subsequently checked by a Crow Canyon analyst. The pottery from the tested sites was analyzed in accordance with the system developed during the winter of 1989. The criteria for the typological analysis are recorded in detail in the Crow Canyon pottery analysis manual (Crow Canyon Archaeological Center 1990). Minor, but important, changes in this system since 1989 have been documented in the Crow Canyon laboratory archives and are discussed in this chapter when appropriate. The criteria for the typological analysis of the tested-sites pottery, and the significant changes made in this system while the analysis was ongoing, are summarized in Appendix A(1).

However, two aspects of the Crow Canyon pottery analysis system that differ from other analysis systems used in the Mesa Verde region should be emphasized. First, a number of grouped types are used in the Crow Canyon analysis system. Grouped types were created so that sherds did not have to be forced into a traditional type or left unclassified. Grouped types are primarily used for small sherds or sherds with ambiguous attributes or attribute combinations. Definitions of these grouped types are presented in Appendix A(1) as well as a detailed discussion of changes in the definitions of two grouped types, Late White Painted and Pueblo III White Painted, that occurred during the analysis of pottery from the tested sites. Second, some of the criteria used at Crow Canyon to identify traditional types may be more exclusive than the definitions used by other researchers. Particularly important for this project are the criteria used to categorize McElmo Black-on-white and Mesa Verde Black-on-white, which are discussed in detail in Appendix A(1).

The Crow Canyon pottery analysis system is sherd-based, that is, sherds are the basic unit of observation. In this kind of analysis, sherds from the same vessel are analyzed separately and may be identified as different types. For example, a single reconstructible vessel might contain sherds classified as Late White Unpainted, Pueblo III White Painted, and Mesa Verde Black-on-white because the painted design covered only a part of the vessel's surface; thus, some sherds were unpainted or contained only a small part of a design. The advantage of a sherd-based analysis is that it ensures comparability between proveniences that yielded reconstructed and/or whole vessels and proveniences that did not.

The next section presents a brief discussion of the analytic criteria employed in the Crow Canyon analysis system. This includes definitions of wares, criteria for vessel form and part identifications, distinction of paint type on painted white wares, and a description of how materials were counted, weighed, and curated. Criteria used to assign sherds to specific types are presented in Appendix A(1) and draw heavily on the Crow Canyon pottery analysis manual (Crow Canyon Archaeological Center 1990). Date ranges for each type are taken from a synthesis of Mesa Verde-region pottery by Wilson and Blinman (1991).

Analytic Criteria

Assigning sherds to a distinct tradition is the first step in our pottery analysis. The local wares and types are a part of the northern San Juan (Mesa Verde) pottery tradition. Pottery traditions are defined for the most part by material and technological attributes of temper, paint, paste, and slip clay, and to a lesser degree by stylistic elements. A more comprehensive discussion of the northern San Juan (Mesa Verde) tradition can be found in Wilson and Blinman (1991).

The next step is to assign sherds to ware categories. These assignments are based on the material resources employed, the production technology, and surface manipulation (that is, corrugated versus smooth). The characteristics of gray wares, white wares, and red wares are presented below.

Finally, pottery was assigned to types. Type identifications, in most cases, are based on aspects of surface manipulation and painted-design style that changed through time. Explicit criteria for typological classification were developed by the Crow Canyon research staff in consultation with C. Dean Wilson. The analysis system used by Wilson was more complicated than the one eventually adopted by Crow Canyon, but his suggestions greatly improved our typological classification.

Criteria for typological analysis were developed to minimize subjectivity, promote consistency between analysts conducting the pottery analysis, and ensure comparability between sites. The types we use generally follow Breternitz et al. (1974), but we also define several new and broader grouped types, which allow us to categorize sherds that do not fit easily into any previously recognized categories. Thus, fewer sherds are assigned to traditional types (for example, Mesa Verde Black-on-white), but the sherds assigned to these traditional types are so designated because they meet a set of well-defined criteria. Most decorated white wares have been assigned to the grouped types (for example, Pueblo III White Painted).

The pottery types employed are based on polythetic combinations of attributes. The descriptions in Appendix A(1) cannot list every attribute of every type. Rather, we emphasize those attributes that are most important in assigning sherds to typological categories in the Crow Canyon analysis system. Additional information presented includes the common vessel forms for each type, the current date range for each type, and analogous types in other pottery traditions.

Local Wares and Types

Almost all Mesa Verde-region pottery can be assigned to gray, white, or red ware categories. The single exception in the tested-site assemblages is Basketmaker Mudware. Ware distinctions are based on broad divisions in decorative and technological treatment. The different wares appear to represent differences in the production and use of pottery (Wilson and Blinman 1991).

Pottery types are defined within wares and traditions. Traditions are groups of pottery types and wares with great temporal continuity within a limited geographic area. In the Southwest, assignment of pottery to a particular tradition is usually based on attributes of material and technology, including temper, paint, clay, and firing characteristics. Local wares are a part of the northern San Juan (Mesa Verde) tradition. Other Puebloan pottery traditions include the upper San Juan (Gobernador), Chaco (Cibola), Chuska, and Kayenta (Tusayan). Appendix A(1) first presents a definition of Basketmaker Mudware, and then discussions of all the gray, white, and red ware types recognized in the Crow Canyon analysis system. A sample of sherds from most of the painted white ware types used in the Crow Canyon system is illustrated to give the reader a better idea of the distinctions being made, particularly between the grouped and traditional types (Figure 15.1, Figure 15.2, Figure 15.3, Figure 15.4, Figure 15.5, Figure 15.6, and Figure 15.7).

Determining Vessel Form

The Crow Canyon pottery analysis system employs four vessel form categories in the analysis of all sherds--jars, bowls, other, and unknown. The criteria for identifying these form classes are presented below. An additional and more detailed analysis of vessel form was performed on the rim sherds. This analysis is discussed later in this chapter in the section on vessel form.

Jar. All gray ware sherds, plain or corrugated, are assigned to the jar category, unless a sherd is from the rim of the vessel and is curved such that it could only be from a bowl or some vessel form other than a jar. Gray ware forms other than jars were produced primarily before A.D. 900; after that time, jars constitute almost 100 percent of gray ware assemblages. White ware and red ware sherds are assigned to the jar category on the basis of curvature and the presence of paint, polish, and/or slip on the exterior surface, and the absence of these surface treatments on the intact interior surface.

Bowl. A gray ware sherd is assigned to this category if it is a rim that is curved such that it could only come from a bowl. White ware and red ware sherds are recorded as bowls on the basis of rim form, sherd curvature, and the presence of paint, polish, and/or slip on at least their interior surfaces.

Other. A sherd is assigned to this vessel form category if, based on curvature and/or rim form, the analyst is certain it is from neither a typical jar nor a bowl and is able to specify an alternative vessel form. Alternative vessel forms include dippers/ladles, mugs, cups, kiva jars, kiva jar lids, canteens, rectangular boxes, seed jars, gourd-shaped vessels, pitchers, effigy vessels, and miniature vessels. Coils, appliques, and handles that are no longer attached to vessels are also recorded as "other" during analysis of vessel form. When a sherd was classified as "other," the analyst described that sherd in the comments column on the analysis form.

Unknown. A sherd is classified as "unknown" if the analyst cannot tell from which of the more specific vessel forms the sherd originated. A variety of factors can make reliable vessel form distinctions difficult or impossible. For instance, a white ware sherd with a slipped, polished, or painted exterior but an eroded or missing interior surface would also be classified as an unknown form.

Determining Vessel Part

Only two readily identifiable vessel part categories were used in this analysis--rim and body. More elaborate classifications of vessel part are possible, but the subtle nature of many vessel part distinctions often results in inconsistencies among analysts. Thus, in the interests of assuring analytical consistency, we chose a more conservative route that still furnishes much of the data needed for sampling for and conducting studies involving vessel part.

Rim. The finished lip of the vessel must be present to call a sherd a rim. A kiva jar lid fragment with a finished edge is recorded as a rim sherd. Care was taken so that the ground edge of a modified sherd and sherds broken on coil junctures were not recorded as rims.

Body. Any sherd not from the finished lip or rim of a vessel is classified as a body sherd. This includes detached coils, appliques, and handles.

Determining Finish

Recording finish includes determining paint type on white ware sherds and the presence or absence of slip on red ware sherds. Paint type was recorded as either carbon or mineral paint. Many sherds have a mixture of carbon and mineral paint. In the Crow Canyon system, the presence of any carbon paint resulted in the sherd being classified as having carbon paint, and the fact that there was a mix of carbon and mineral paint was noted in comments.

Carbon Paint. Carbon paint, often referred to as organic paint, is presumed to have been made from boiled-down beeweed (Cleome surrulata) or tansy-mustard (Descurainia richardsonii). Carbon paint can be recognized by the following characteristics: (1) it often fires to a bluish black color; (2) it soaks into the sherd, frequently resulting in a fuzzy boundary between the painted and unpainted areas; (3) it sometimes allows the background to show through the paint; and (4) it tends to be shinier than mineral paint, resembling the finish on a glossy photo. Sherds classified as having carbon paint include those with paint that is believed to be exclusively carbon paint, as well as sherds with paint that is believed to be a mixture of carbon and mineral paint.

Mineral Paint. Mineral paint, often referred to as inorganic paint, is made from ground iron, manganese, or copper-rich rock. The following characteristics are used to recognize mineral paint: (1) it fires to a wider range of colors than carbon paint, often a brownish black color; (2) it does not soak into the sherd, but rests on top of the surface, frequently resulting in a sharp boundary between the painted and unpainted areas; (3) it usually does not allow the background to show through--the paint provides a solid cover over the surface of the sherd; and (4) it tends to be duller than carbon paint. The sherds identified as having mineral paint are sherds that the analyst believes do not have a combination of carbon and mineral paint.

Slip. The presence or absence of a slip is recorded for red ware sherds only. The presence or absence and quality of the slip are important in distinguishing Bluff Black-on-red from Deadmans Black-on-red.

Counting, Weighing, and Curating Sherds

Sherds are counted, weighed, and bagged by the smallest analytic unit from each provenience. For example, Mesa Verde Black-on-white bowl body sherds with mineral paint would be counted, weighed, and bagged separately from Mesa Verde Black-on-white bowl body sherds with carbon paint. When recognized, sherds broken during excavation or laboratory processing were counted as a single sherd.

The bags containing sherds of the same type, form, part, and finish were then grouped together into larger bags by ware. White wares and red wares from a single provenience are bagged separately. Within gray wares, corrugated gray wares and plain gray wares are bagged separately. Finally, all unknown and nonlocal wares are bagged separately. Each of these larger bag units (for example, white wares) were boxed together in order by provenience (PD and FS)⁽¹⁾ for each individual site. All of the artifacts bagged and boxed in this way and the associated analytical records and data files are curated at the Anasazi Heritage Center in Dolores, Colorado.

The Site Assemblages

Table 15.1 lists the counts and weights of all pottery (including sherds, complete vessels, and modified and shaped sherds) by pottery type and vessel form for each of the tested sites. Data on the five complete or nearly complete vessels are presented in Table 15.2 and the four vessels that are not funerary objects are illustrated in Figure 15.8, Figure 15.9, Figure 15.10, and Figure 15.11. Although 35 different pottery types were identified from the tested sites, examination of Table 15.1 shows that most of the pottery in all assemblages belongs to a small number of gray and white ware types associated with the Pueblo III period. This is the expected pattern because sites were selected for testing based on the presence of abundant Pueblo III pottery on the surface (see Chapter 1 for more information on site selection and sampling strategy). However, some assemblages possess significant amounts of non-Pueblo III pottery indicative of earlier use of these locations. Data from pottery, architecture, and other dating evidence are used to identify these earlier occupational components (see Identifying and Dating Tested Site Components with Pottery Type Assemblage Data, below).

In terms of vessel forms, the assemblages are also fairly consistent. Roughly three-quarters of the sherds come from jars, one-fifth from bowls, and the remaining small amounts came from other and unknown forms. The variation in vessel forms that does exist among assemblages is the subject of additional analyses later in this chapter.

Comparisons of the relative frequencies of different types and forms derived from counts and weights indicate sometimes substantial discrepancies. This underscores an issue of quantifying pottery that must be addressed before further discussion of the analytical results.

Quantifying Pottery

The need to quantify pottery from archaeological contexts stems from the desire to compare different pottery assemblages or collections. Since a primary goal of the Sand Canyon Archaeological Project and the Site Testing Program involves quantitative and qualitative comparisons of assemblages to ascertain temporal, functional, and social relations, the methods used to quantify pottery for comparison must be considered. The task of quantifying pottery is complicated by the fact that most pottery recovered from archaeological contexts consists of broken pieces, or sherds, rather than the complete vessels--the unit of manufacture and use. This discussion begins with an assessment of the characteristics of sherds and vessels as units of comparison of archaeological assemblages and finds sherds to be the most appropriate unit in most cases. The two measures of sherd abundance used in the Crow Canyon analysis, count and weight, are then examined, and weight is found to be the best measure for most cases because of the effects that pottery technology, use, deposition, and postdepositional history (including excavation and lab treatment) have on sherd counts.

Sherds vs. Vessels. Although a variety of methods have been developed for estimating the numbers of vessels represented by a particular group of sherds (Orton 1993), the issue remains as to whether vessels or sherds offer the most valid and reliable units for quantification. A number of problems arise from the use of vessels as the unit of quantification. The most prominent problem concerns the practical issues involved in refitting sherds into vessels or estimating vessels from particular sherd characteristics. Most refitting efforts incorporate only a small proportion of the sherds recovered, and these sherds are usually from restricted contexts that introduce additional biases.

Methods for estimating numbers of vessels from sherds, usually minimum numbers, also introduce practical and quantitative problems. Vessel estimation methods either count unique parts of vessels such as bases, or measure some part of vessels for which a known quantity exists in a single vessel. For example, circular rims always cover 360 degrees. Thus, if you sum the degrees covered by sherds from vessels of the same size and divide this number by 360, you obtain an estimate of the minimum number of vessels represented by those sherds. The accuracy, precision, and validity of these estimation methods are affected by the degree of fragmentation of sherds and by approaches used for grouping or aggregating sherds for measurement and estimation. The smaller the sherd, the more difficult it is to identify specific vessel parts and make accurate and precise measurements.

Grayson (1984) explores the problem of aggregation in the context of estimating minimum numbers of individuals from faunal remains. For pottery, aggregation affects minimum vessel number estimates in two ways. First, in cases where different pieces of a specific vessel part--rims for instance--are used to calculate vessel estimates, these estimates will vary with the scale of the provenience units used to group sherds for estimation. If the entire assemblage of rim sherds is used, the smallest vessel estimate will result. As smaller and smaller provenience groupings are employed, the minimum vessel estimate for the whole assemblage will increase because in each provenience group the estimate is derived by rounding up to the nearest whole vessel (for example, sherds totaling 380 degrees of arc equals two vessels). The most extreme case would be provenience groups that contain individual rim sherds or single sherds from different vessel forms. In this case, the minimum vessel estimate equals the number of rim sherds.

A second aggregation issue arises when rim sherds are used to estimate numbers of vessels. The process of summing degrees of arc for sherds from vessels of the same size also involves aggregation. Grouping sherds from vessels of the same size usually involves estimating the diameter of the vessel opening and creating groups through arbitrary divisions of the total distribution of sizes, or using modes in the distribution of diameter values to make "natural" groups. Once again, the minimum vessel estimate will vary with the number of size groups employed.

Although the significant practical and quantitative problems inherent in estimating numbers of vessels from sherds support the use of sherds as the unit of quantification, using sherds has been criticized because their abundance is affected by both the number and size of vessels (Egloff 1973; Orton 1993). However, as Feathers (1990) argues, this criticism is valid only if vessels are the unit of comparison, and in most cases it is sherd assemblages and not vessels that are compared. When comparing assemblages, the relationship of sherd abundance to vessel count and size is no longer a bias, but a potential source of meaningful variation among sherd assemblages. However, in cases where vessels are the unit of comparison, there is little choice but to work with the less reliable vessel estimates.

Sherd Count vs. Sherd Weight. Although sherd abundance is a more reliable unit of quantification, important differences exist between count and weight as measures of sherd abundance. The most significant difference between counts and weights is that counts vary dramatically with the degree of fragmentation whereas weights do not (Egloff 1973; Evans 1973; Solheim 1960). The degree of fragmentation of pottery in a given assemblage is a function of a variety of factors, including manufacturing technology, uses, discard processes, and postdepositional history. Patterns of variation between counts and weights of pottery can, therefore, yield important information about the pottery itself and potentially unique histories of individual assemblages that can affect comparisons. We examine the relationships between counts and weights of sherds along several dimensions to assess the relative importance of these factors on the tested-site assemblages and their potential impact on comparative analyses.

The effects of manufacturing technology and vessel use on sherd fragmentation can be determined by comparing counts and weights of sherds for different classes of pottery. The most dramatic difference between counts and weights occurs at the ware level, particularly between gray wares and white wares. At each of the tested sites, the abundance of gray wares relative to white wares increases when using counts and decreases when using weights. This pattern derives from a consistent difference in the average size of gray and white ware sherds in all assemblages (Figure 15.12). This pattern almost certainly results from differences between gray wares and white wares in their manufacture and use.

White wares recovered from the tested sites are predominantly bowls for serving and jars for storage, whereas corrugated gray wares are primarily cooking vessels. White wares tend to have thicker walls and finer temper, and they may have been fired at higher temperatures than gray wares, all of which promotes their mechanical strength. Gray wares, on the other hand, have larger temper composing a greater proportion of the paste (Wilson and Blinman 1991), which decreases their mechanical strength but increases their resistance to thermal stress (Steponaitis 1983; West 1992). In addition, the exposure of gray wares to thermal stress during use further weakens the pottery through thermal fatigue. Consequently, the larger size of white ware sherds mostly reflects their greater mechanical strength and less rigorous use relative to gray wares. These conditions also account for the differences apparent in Table 15.1 in frequencies of bowl forms relative to other forms when using counts and weights.

In addition to the differences between gray wares and white wares, there is also a consistent tendency for the more specific traditional types to compose a larger percentage of the site assemblage when using weights (Table 15.1). This reflects the fact that sherds assigned to traditional types are larger because larger sherds are more likely to have the attributes necessary to be assigned to a traditional type. Examining the percentage of Mesa Verde Black-on-white in the site assemblages illustrates the difference between counts and weights (Table 15.1). At every site, Mesa Verde Black-on-white makes up a greater percentage of the assemblage when weights are used. The difference between counts and weights can be dramatic, as illustrated by Stanton's Site, where Mesa Verde Black-on-white makes up 1.8 percent of the site assemblage by count but 8.1 percent of the assemblage by weight. The opposite pattern often holds for grouped types because smaller sherds are more likely to be assigned to these types.

The degree of sherd fragmentation also varies between sites located on the mesa tops in upper Sand Canyon, where sherds are smaller, and sites located on the talus slopes or in lower Sand Canyon, where the sherds are larger (Figure 15.12). This systematic difference in sherd size most likely results from differences in historic land use. The mesa-top sites have received greater impact in historic times than the sites on the talus slopes and the sites in lower Sand Canyon. The sites with the smallest sherds, Shorlene's (5MT3918), Kenzie Dawn Hamlet (5MT5152), and Roy's Ruin (5MT3930), have been repeatedly plowed. The reduction of artifact size as the result of plowing has been discussed in detail elsewhere (Dunnell and Simek 1995; Pierce 1994). The unplowed mesa-top sites (Troy's Tower, G and G Hamlet, and Lillian's Site) have all been chained and used for cattle grazing. The remaining tested sites, all of which have larger-than-average sherds, can be divided into those located on the talus slopes or in lower Sand Canyon; these sites have not been plowed or chained, although cattle have grazed these areas.

Finally, there is a consistent difference between counts and weights when the pottery from early components is considered. This is especially true for Basketmaker III pottery. This early pottery makes up a greater percentage of the site assemblage when counts are considered; the percentage decreases when weights are considered, which indicates that the average size of these early sherds is smaller.

This discussion indicates that a great deal can be learned about different kinds of pottery and entire pottery assemblages by comparing counts and weights. However, as a general measure of abundance, weight is superior because it is less affected by sherd size. This is particularly true for the tested-site assemblages because of the large degree to which postdepositional factors have differentially altered sherd size. Consequently, most of the pottery assemblage comparisons discussed in this volume are based on sherd weights, although data tables normally present both counts and weights. In a few cases, sherd counts or other units and measures are used, and the rationale for the decisions to use these alternative measures are discussed in those sections.

Identifying and Dating Tested-Site Components with Pottery Type Assemblage Data

Several methods have been employed to use pottery to date archaeological sites. Use of the presence or absence of a specific traditional type whose occurrence has been bracketed in time by associated tree-ring dates is common (Blinman 1988b; Breternitz et al. 1974). Working with pottery from the Dolores Project, Blinman developed a method that relies on the examination of the entire pottery assemblage from a site or provenience to make dating inferences (Blinman 1988b). Wilson and Blinman (1991) revised this assemblage-based dating scheme for the Mesa Verde region. In this chapter, pottery assemblages are used to identify and date the components at each of the tested sites. In Chapter 20, dating with pottery is expanded to more precisely date the Pueblo III components. This is done using the ratio of McElmo Black-on-white to Mesa Verde Black-on-white at each site and by examining the attribute-based analysis of decorated white wares developed by Hegmon (1991).

Blinman (1988b) and Wilson and Blinman (1991) argue that assessment of entire pottery assemblages results in more consistent and accurate date estimates than either specific pottery attributes or statistical manipulation. Their approach is similar to Colton's use of pottery periods. They characterize an assemblage content for broad time periods, based on the pottery assemblages from tree-ring dated deposits within those periods. This results in an "idealized" assemblage for each time period, and the content of any pottery assemblage can then be compared to the idealized assemblage. The best fit on all types is the most logical date, while points of difference are keys in determining whether or not that date is likely to be accurate.

Wilson and Blinman (1991:44) acknowledge several limitations to their method. First, change in pottery is continuous; therefore, no single idealized assemblage can perfectly bracket the beginning and end of a given period. Likewise, pottery is not homogeneous throughout the entire Mesa Verde region, and variation in frequencies occur from one place to another. Also, not all periods are equal in terms of the quality of the dating and description of the pottery assemblages. Finally, adequate sample size is an important consideration. With smaller sample sizes, the absence of a type might not be real but, rather, the result of chance in archaeological recovery. The distinctiveness of each period influences how large an assemblage is needed, but Wilson and Blinman estimate that between 200 and 400 sherds are needed to be confident that the absence of a type is real and not the result of chance.

Wilson and Blinman (1991:45-47) provide idealized assemblage content for the following periods: A.D. 300 to 575, A.D. 575 to 725, A.D. 725 to 775, A.D. 775 to 825, A.D. 825 to 860, A.D. 860 to 880, A.D. 880 to 910 (930), A.D. 910 (930) to 980, A.D. 980 to 1025, A.D. 1025 to 1100, A.D. 1100 to 1140, A.D. 1140 to 1180, A.D. 1180 to 1225, and A.D. 1225 to 1300. The description of each of these idealized assemblages is presented in Appendix A(2).

The tested sites were selected because pottery on the modern ground surface indicated that the main component at each site dated between A.D. 1150 and 1300. Testing revealed earlier components as well. At some sites, stratigraphically earlier architecture was found, often directly beneath the Pueblo III architecture. As detailed in Chapter 20, these stratigraphically earlier deposits are often dated with tree rings. At other sites, no earlier architecture was found but earlier strata were present (for example, a stratum of fill beneath the Pueblo III masonry roomblock). At these sites, earlier components are dated solely on the basis of the pottery assemblages. There are also sites where there was no stratigraphic evidence for earlier use of the site, but where traces of earlier pottery are present. These different kinds of evidence provide a relative scale for the intensity (that is, the number of people and the duration of time) of occupation during each component. Components with identifiable architecture are interpreted as having the greatest intensity; components with distinct strata, but no architecture, have a moderate intensity; and sites with only traces of earlier pottery, but no architecture or strata associated with an occupation, are interpreted as having the lowest intensity of occupation.

The earliest components date to the Basketmaker III period (A.D. 575 to 725). These components are present at Kenzie Dawn Hamlet and Shorlene's Site, where Basketmaker III habitations are inferred based on the presence of the architectural facilities. Basketmaker III pottery indicates less intensive Basketmaker III components at Lillian's Site and G and G Hamlet. Pueblo I pottery (the pottery periods spanning the A.D. 725 to 930 period) is found at several sites, but at no site is enough pottery present to argue for a Pueblo I habitation. Less intensive use of these sites may have occurred during the Pueblo I period. Early Pueblo II pottery (the A.D. 930 to 1025 pottery periods) is almost entirely absent; the small amounts present may be material deposited during the middle and late Pueblo II occupation of the sites. Middle (A.D. 1025 to 1100) and late Pueblo II (A.D. 1100 to 1140) components are present at G and G Hamlet, Lillian's Site, Roy's Ruin, Kenzie Dawn Hamlet, Shorlene's Site, Catherine's Site, and possibly Castle Rock Pueblo.

Finally, every tested site has a Pueblo III component. Wilson and Blinman's pottery periods can be used to place these components into one of three time periods, early Pueblo III (A.D. 1140 to 1180), middle Pueblo III (A.D. 1180 to 1225), and late Pueblo III (A.D. 1225 to 1300). Each component, from Basketmaker III through Pueblo III is described in greater detail below.

Basketmaker III (A.D. 575-725)

Basketmaker III architecture was found at Shorlene's Site and at Kenzie Dawn Hamlet. Substantial amounts of Basketmaker III pottery were found on both sites. Lesser amounts of Basketmaker III pottery were also found at Lillian's Site and G and G Hamlet, indicating some occupation of these sites during the Basketmaker III period.

Traces of either Chapin Gray or Chapin Black-on-white were also found at Roy's Ruin, Saddlehorn Hamlet, and Castle Rock Pueblo, and Indeterminate Plain Gray is also present at these sites (Table 15.1). The Indeterminate Plain Gray is not necessarily Basketmaker III pottery (for example, it could come from Moccasin, Mancos, or Mummy Lake gray ware). The numbers of Chapin Gray and Chapin Black-on-white sherds are so small from these three sites that they are not interpreted as indicating a Basketmaker III occupation. The four sites from which Basketmaker pottery was recovered in greater quantities, indicating a Basketmaker III occupation, are listed in Table 15.3.

The table shows that Basketmaker III pottery was abundant at the sites where Basketmaker III architectural facilities were identified--Shorlene's Site and Kenzie Dawn Hamlet. These two sites are located within .5 km of each other. Shorlene's Site has one burned Basketmaker III pit structure; tree-ring dates indicate that construction of this pit structure took place around A.D. 655. Two additional unburned pit structures at Shorlene's Site might also date to the Basketmaker III period (see Chapter 6). Two structures at Kenzie Dawn Hamlet were identified as dating to the Basketmaker III period; tree-ring dates indicate that this occupation dates to sometime after A.D. 665 (see Chapter 5). The Basketmaker III occupations on these two sites were therefore either contemporaneous or immediately sequent.

The Indeterminate Plain Gray, Early White Painted, and Early White Unpainted are all types that are probably associated with these Basketmaker III occupations, but could be associated with later occupations. This is especially true for the Plain Gray, as a single broken Mummy Lake Gray vessel associated with the Pueblo III occupation at a site could account for a large percentage of the Plain Gray present. However, the high percentage of Plain Gray to white wares at Shorlene's Site and Kenzie Dawn Hamlet is typical of Basketmaker III sites elsewhere (Wilson 1991:762).

G and G Hamlet and Lillian's Site have the next-highest amounts of potential Basketmaker III pottery; however, these assemblages are much smaller than the Basketmaker assemblages at Shorlene's Site and Kenzie Dawn Hamlet. Basketmaker III habitation sites are located adjacent to both Lillian's Site and G and G Hamlet (Crow Canyon site number 103 is adjacent to Lillian's Site, and site number 121 is close to G and G Hamlet [Van West et al. 1987; Adler 1988]), so the presence of small amounts of Basketmaker III pottery at these two sites is not surprising.

Three tree-ring samples from a burned post-and-adobe room found beneath the masonry roomblock at Lillian's Site yielded noncutting dates of A.D. 499, A.D. 543, and A.D. 560. These may be from a Basketmaker III structure; however, a noncutting date of A.D. 1075 was present as well. Basketmaker III, Pueblo I, and Pueblo II pottery was present in the fill beneath the masonry roomblock, but the Pueblo II pottery was the most numerous, so the favored interpretation is that the burned room dates to the Pueblo II period. Thus, it is not clear if Basketmaker III structures are present at Lillian's Site.

The remaining sites contain only traces of Basketmaker III pottery. If there was use of these sites during the Basketmaker III period, it appears to have been minimal. It is also possible that the Basketmaker III pottery was incorporated into the pottery assemblages at these sites by either natural or cultural postabandonment processes; Basketmaker III sites are located relatively close to each of these sites.

In the Basketmaker III assemblages from the tested sites, gray wares make up over 90 percent of all pottery. Although the percentages may be overestimated because of the incorporation of later plain gray wares, the values are consistent with those documented elsewhere in the Mesa Verde region (Blinman 1986b:73; Wilson 1991:762). There is a dramatic increase in the percentage of white wares in the Pueblo II and III assemblages, as discussed in the following sections. Wilson (1991:762) argues that this change in the proportion of gray wares to white wares takes place during the Pueblo II period and not before.

Pueblo I (A.D. 725-930)

The Pueblo I pottery period is here defined as A.D. 725 to 930. The terminal date is later than most previously published dates for the Pueblo I period and is based on work by Blinman (1994), who argues that Pueblo I pottery continues to dominate assemblages until A.D. 930 to 940. Pottery dating for this period is more accurate than for any other period in the Mesa Verde region, largely due to the work of the Dolores Archaeological Program (Blinman 1988b:501). Wilson and Blinman distinguish five different idealized pottery assemblages within the A.D. 725 to 930 period. These pottery periods are grouped together in this report because so little Pueblo I period pottery was recovered from the tested sites.

Table 15.4 lists the pottery that is most confidently dated to Pueblo I occupations. Gray wares that dominate at Pueblo I sites include Moccasin Gray, Mancos Gray, and Indeterminate Neckbanded Gray. The most important decorated pottery types found in Pueblo I sites include Piedra Black-on-white, Abajo Red-on-orange, Bluff Black-on-red, and Deadmans Black-on-red. The San Juan Red Ware types make up approximately 8 to 10 percent of the pottery assemblages in the Dolores River valley during the Pueblo I period (Wilson and Blinman 1991:36).

Table 15.4 indicates that neckbanded gray wares and decorated white wares and red wares that should be present in large numbers in Pueblo I period components are represented by only a handful of sherds from the Testing Program pottery assemblages. This is especially true for the San Juan Red Ware types, which make up no more than .2 percent of any of the tested-site collections. In sum, the low numbers of Pueblo I decorated pottery indicate that use of the tested sites was minimal during this period.

Table 15.4 does not include Chapin Gray, Indeterminate Plain Gray, Early White Painted, and Early White Unpainted. These types occur in largest numbers on the sites interpreted as having Basketmaker III occupations, and they are listed in Table 15.3. These types occur in small numbers on tested sites not listed in Table 15.3, and the totals can be found in Table 15.1. It is possible that some of this pottery was deposited as a part of Pueblo I occupations, but because the Pueblo I decorated pottery is so sparse, most of the plain gray is believed to be associated with Basketmaker III occupations.

Early Pueblo II (A.D. 930-1025)

Cortez Black-on-white is the dominate white ware during this period. Five Cortez Black-on-white sherds were recovered from the tested sites. These low numbers indicate that there was no significant occupation at the tested sites during this time period. The small numbers of Cortez Black-on-white that are present probably result from the middle Pueblo II components identified at the tested sites. The pottery for these components is summarized below.

Middle Pueblo II (A.D. 1025-1100)

The middle Pueblo II is one of Wilson and Blinman's (1991:46) idealized time periods. In terms of the traditional temporal breaks between Pecos periods, this would be the late Pueblo II period (Cordell 1984:103). Recent discussions of Mesa Verde-region culture history have extended the Pueblo II period beyond A.D. 1100 to the middle A.D. 1100s (Varien et al. 1996). For the purpose of describing the pottery assemblages, we will refer to this period as the middle Pueblo II and the next pottery period as late Pueblo II. A summary of the temporal frameworks used by a variety of researchers working in the northern Mesa Verde region is presented in Chapter 21.

The idealized assemblages from this period are based largely on several tree-ring-dated sites excavated as a part of the South Canal Project (Kuckelman and Morris 1988). The South Canal sites are located approximately 20 km (12 mi) north of the Sand Canyon locality. The South Canal tree-ring-dated sites are small habitations with post-and-adobe surface rooms and earth-walled pit structures. The architecture at these sites is similar to the architecture found at G and G Hamlet, Kenzie Dawn Hamlet, and Lillian's Site.

In Wilson and Blinman's idealized assemblages from this period, the dominant white ware is mineral-painted Mancos Black-on-white, and the dominant gray ware is Mancos Corrugated. Cortez Black-on-white, Indeterminate Plain Gray, and Mummy Lake Gray can be present in smaller quantities. Red wares can include both San Juan Red Ware and traces of Tsegi Orange Ware. Sites that date toward the end of this period will have higher frequencies of Mancos Black-on-white and Tsegi Orange Ware (and Dolores Corrugated--a type not used in the Crow Canyon analysis system).

Several tested sites have components that date to this time period. Architectural facilities that are tree-ring dated to this period occur at G and G Hamlet and Kenzie Dawn Hamlet (see Chapters 2 and 5, respectively). Architectural facilities are also present at Lillian's Site, but these are less confidently dated (Chapter 3). In addition, Roy's Ruin, Shorlene's Site, and Catherine's Site have moderate amounts of pottery that date to this period; traces of pottery that date to this period are present at Troy's Tower, Stanton's Site, Saddlehorn Hamlet, Mad Dog Tower, Castle Rock Pueblo, Lester's Site, and Lookout House.

Table 15.5 presents data on the pottery from this period, including Mancos Corrugated and the following decorated white wares: Mancos Black-on-white, Cortez Black-on-white, Pueblo II Black-on-white, and Late White Painted sherds with mineral paint. Late White Painted is a grouped type for sherds that could be either Pueblo II or Pueblo III pottery (that is, Cortez, Mancos, McElmo, or Mesa Verde black-on-white), but which cannot be assigned to a specific traditional type. Paint type was recorded for these sherds. It is believed that the sherds with mineral paint are the most likely to be Pueblo II sherds, whereas the sherds with carbon paint are most likely to be Pueblo III in age. This is only an approximation, because Pueblo III decorated white wares in the Hovenweep area are decorated predominantly with mineral paint. Thus, some of the Late White Painted sherds with mineral paint could be from Pueblo III-period vessels produced in the Hovenweep area and exchanged into the Sand Canyon locality. Other types could also be associated with this component, including Indeterminate Plain Gray, Mummy Lake Gray, Late White Unpainted, Indeterminate Local White Painted, Indeterminate Local White Unpainted, Deadmans Black-on-red, Indeterminate Local Corrugated Gray, and Other Red Nonlocal. The number and weight of sherds in each of these categories can be found in Table 15.1. These categories are not included in Table 15.5, because it is believed that the majority of the sherds in these categories are associated with earlier or later components at the sites. However, some of the sherds in these categories, especially the Indeterminate Corrugated and the pottery in the grouped types (such as Late White Unpainted), were almost certainly deposited during occupations that date to this period.

The three sites where middle Pueblo II architectural facilities were discovered have the highest number of Pueblo II-period decorated white wares. At G and G Hamlet and Kenzie Dawn Hamlet, the type of architectural facilities and the amount of pottery present indicate that the sites functioned as habitations during these periods. The function of Lillian's Site during the middle Pueblo II period is unclear; it, too, may have functioned as a habitation during this period, but no pit structure was found. When the Pueblo II decorated white wares from these three sites are compared, Lillian's Site has less middle Pueblo II pottery than either G and G Hamlet or Kenzie Dawn Hamlet. Thus, the middle Pueblo II occupation at Lillian's Site may not represent use of the site as a year-round habitation. Instead, Lillian's Site may have functioned as a seasonal field house during that period.

Roy's Ruin, Shorlene's Site, and Catherine's Site have the next-highest totals of middle Pueblo II pottery, yet these totals are substantially less than the totals for the sites discussed above. Testing did not uncover any Pueblo II architecture at these sites. The testing may have missed small structures, but it is unlikely that middle Pueblo II roomblocks, if present, escaped detection. The low number of pottery sherds from the middle Pueblo II period suggests that the use of these sites during that time was most likely for a limited set of activities, rather than for year-round habitation.

Only a few middle Pueblo II pottery sherds were found at Troy's Tower, Stanton's Site, Saddlehorn Hamlet, Mad Dog Tower, Castle Rock Pueblo, Lester's Site, and Lookout House. There are so few middle Pueblo II sherds on these sites that occupations were either of extremely limited use, or the earlier sherds were incorporated as a part of the later Pueblo III occupations.

Late Pueblo II (A.D. 1100-1140)

Wilson and Blinman (1991) recognize this pottery period, which we label the late Pueblo II period. This pottery period is difficult to isolate using the Crow Canyon analysis system. Wilson and Blinman report that the major change is that Dolores Corrugated becomes the dominant gray ware vessel form, although both Mancos and Mesa Verde corrugated can be present. Because our system does not recognize Dolores Corrugated as a type, the contribution of this type to the overall assemblage cannot be evaluated. Mancos Black-on-white is still the dominate white ware during this period, but carbon-paint decoration becomes more common. (Mancos Black-on-white with carbon paint is what Hayes [1964:63] labels Wetherill Black-on-white.) Carbon-painted Mancos Black-on-white and Pueblo II White Painted sherds were tallied to see how well this period might be represented in the assemblages from the tested sites. These sherds were present at G and G Hamlet (N = 12), Shorlene's Site (N = 6), Roy's Ruin (N = 4), Lillian's Site (N = 2), Catherine's Site (N = 8), and Kenzie Dawn Hamlet (N = 15). The components at G and G Hamlet and Kenzie Dawn Hamlet that began in the previous period may have therefore extended into this period. At the other sites, the intensity of occupation during this period is interpreted as having been minimal.

Early Pueblo III (A.D. 1140-1180)

This is an interesting period for two reasons. First, large-scale great house construction at Chaco Canyon terminates, and this area appears to no longer exert influence in the Mesa Verde region. Second, some of the worst droughts recorded in the Mesa Verde region occurred during these decades. We had hoped to locate sites that date to this period to compare them to sites and contexts that date to the A.D. 1270s, when another major drought occurred. However, dating components to this period has proved difficult.

The presence of rim eversion on corrugated vessels is important in recognizing this pottery period as defined by Wilson and Blinman (1991:47). Dolores Corrugated is the dominant corrugated vessel form during this period, Mancos Corrugated is virtually absent, and Mesa Verde Corrugated is present, but less frequent than Dolores Corrugated. Again, this cannot be evaluated, because the Crow Canyon system does not recognize Dolores Corrugated. The other major change in pottery assemblages is that McElmo Black-on-white becomes the most common decorated white ware during this period, to the near exclusion of Mancos Black-on-white.

The Crow Canyon analysis used for the Site Testing Program has a restrictive definition of McElmo Black-on-white, so it is not identified in great numbers at any site. G and G Hamlet and Kenzie Dawn Hamlet have the largest amounts of McElmo Black-on-white and they are the sites most likely to have components that date to this period. Unfortunately, the near absence of tree-ring dates from the Pueblo III contexts on these sites makes precise dating of these occupations impossible. At Kenzie Dawn Hamlet, tree-ring dates indicate that the early Pueblo III occupation dates to sometime after A.D. 1142 (Chapter 5). G and G Hamlet and Kenzie Dawn Hamlet are multiple-component sites; if components dating to the early Pueblo III are present, they are probably mixed with earlier (in the case of both G and G and Kenzie Dawn) and later (in the case of Kenzie Dawn) occupations. The pottery assemblages from these sites also contain Mesa Verde Black-on-white, indicating that at least a portion of their Pueblo III occupations postdates A.D. 1180. Because of this, we will discuss the Pueblo III pottery from these sites with the next period.

Middle Pueblo III (A.D. 1180-1225)

Wilson and Blinman (1991) characterize this period as the time when Mesa Verde Black-on-white first appears and when Mesa Verde and McElmo black-on-white occur in roughly equal numbers. Mesa Verde Corrugated and Dolores Corrugated are both abundant. Red wares are scarce but, when present, are almost exclusively White Mountain Red Ware.

The middle Pueblo III components at the mesa-top sites in the area surrounding Sand Canyon Pueblo are believed to date predominantly to the end of this period. Lillian's Site has tree-ring cutting dates of A.D. 1207, 1211, and 1212 (N = 2) and a noncutting date of A.D. 1214. Roy's Ruin has late cutting dates of A.D. 1213 (N = 2) and a noncutting date of A.D. 1223. Mesa Verde Black-on-white outnumbers McElmo Black-on-white by approximately four to one at these two sites, and the same is true for Shorlene's Site. These ratios of Mesa Verde to McElmo black-on-white indicate that these sites date to the late end of this period.

G and G Hamlet has almost equal numbers of Mesa Verde and McElmo black-on-white, which indicates that the Pueblo III component on this site probably dates to earlier in this period. The Pueblo III occupation at Kenzie Dawn Hamlet appears to be extremely long-lived, based on several distinct building episodes, and this occupation could span most of this pottery period. Frequencies of the Pueblo III decorated white wares and the Mesa Verde Corrugated from these sites are presented in Table 15.6. Indeterminate Local Corrugated and Late White Unpainted are omitted from this table because they could have been deposited during either the Pueblo II or III components at these sites.

Late Pueblo III (A.D. 1225-1300)

In Wilson and Blinman's system, Mesa Verde Black-on-white becomes the most abundant decorated white ware and Mesa Verde Corrugated the most abundant gray ware during this period. Mesa Verde Black-on-white increases in abundance throughout this period. Red wares are scarce but, when present, are likely to be White Mountain Red Ware.

Sites that are tree-ring dated to this period include Troy's Tower (noncutting date of A.D. 1271), Lester's Site (noncutting date of A.D. 1271), Saddlehorn Hamlet (noncutting date of A.D. 1256), Lookout House (cutting date of A.D. 1257), and Castle Rock Pueblo (noncutting date of A.D. 1274). Except for the dates for Saddlehorn Hamlet and Lookout House, these dates indicate that there was occupation at each of these sites after A.D. 1270. Archaeomagnetic dates from Troy's Tower, Catherine's Site, Lookout House, and Lester's Site indicate that these sites were occupied sometime after A.D. 1225 and as late as A.D. 1300. Thus, these sites are dated to this pottery period with absolute dates as well. The decorated white wares and the Mesa Verde Corrugated from these sites are summarized in Table 15.7.

With the exception of Mesa Verde Corrugated at Troy's Tower, these sites all have a higher percentage of Mesa Verde Black-on-white and Mesa Verde Corrugated than the sites listed above in Table 15.6. This fits the expectation that these pottery types increased in frequency during this period. In Chapter 20, the ratio of McElmo Black-on-white to Mesa Verde Black-on-white and an attribute-based analysis of Pueblo III white wares are used to date the Pueblo III components more accurately.

All of the sites occupied during this period contain small amounts of early pottery. Catherine's Site, where an earlier limited-use occupation is probably present, contains slightly more early pottery. At the remaining sites, only a trace of earlier material is present; mixing as a result of multiple occupations is believed to be minimal.

Vessel Forms

As described in the methods section of this chapter, the Crow Canyon pottery analysis system includes four vessel form categories--bowls, jars, other, and unknown. Serving bowls and other forms are almost entirely white wares (99 percent and 96 percent, respectively, by weight); however, jars include significant amounts of both white and gray wares (Table 15.8). Aspects of form and use-related alterations indicate that gray ware jars were used primarily for cooking, whereas the white ware jars were used for wet and dry storage (Pierce 1998). Thus, by distinguishing between white and gray ware jars, a more complete analysis of functionally distinct vessel forms is possible.

A chi square analysis of the weights of four different vessel forms (gray ware jars, white ware jars, bowls, and other) recovered from each of the 13 tested sites indicates that variation exists among the sites beyond that expected due to normal random sampling error. An analysis of standardized differences between observed and expected weights shows that differences in the abundance of bowl sherds accounts for most of the variation. Castle Rock Pueblo, Stanton's Site, and Kenzie Dawn Hamlet each possess significantly (p < < .05) more bowl sherds than expected based on marginal and grand totals, and Lookout House and Catherine's Site also show relatively high amounts of bowl sherds (.1 < p < .2). Castle Rock Pueblo also has fewer white ware jar sherds than expected, and Stanton's Site has fewer gray ware jar sherds (for both cases .05 < p < .1). The weights of other vessel form sherds do not vary significantly from the expected in any of the assemblages.

Since the weight of sherds in different assemblages can vary due to differences in the number and size of parent vessels, distinguishing the cause of the vessel form variation identified above requires estimates of either the size or number of vessels. One crude estimate of vessel size is the ratio of rim to body sherds for different vessel forms. All else being equal, a larger vessel will have a smaller ratio of rim to body sherds (rim sherds/body sherds). Table 15.9 presents these ratios for bowls, white ware jars, and corrugated gray ware jars from the tested sites. Of the five assemblages identified as having abnormally high amounts of bowl sherds (Castle Rock Pueblo, Stanton's Site, Kenzie Dawn Hamlet, Lookout House, and Catherine's Site), four have among the highest ratios and only one, Kenzie Dawn Hamlet, has a relatively small ratio that may result from unusually large bowls. The ratio of rim to body sherds for white ware jars at Castle Rock Pueblo is moderate, suggesting normal vessel sizes. Finally, the ratio for corrugated jars from Stanton's Site is among the highest, indicating that small vessel size could be a contributing factor in the lower amounts of gray ware jar sherds. Since the rim/body sherd ratio produced ambiguous results, an analysis of rim sherds was conducted to provide estimates of the number and size of vessels

Analysis of Vessel Rims

The size and shape of rim sherds collected during the Site Testing Program were analyzed to provide more detailed form identifications and estimates of the size and amount represented of the parent vessel. Rim shape was observed to place the sherd into one of 10 vessel form categories--bowl, ladle, wide-mouth jar, narrow-mouth jar, seed jar, kiva jar, canteen, mug/pitcher, special form, and unknown. The bowl category is the same as that presented above, and the wide-mouth jar and narrow-mouth jar conform well to the gray and white ware jar categories, respectively. Consequently, most of the additional vessel forms recognized in this analysis expand on the "other form" category used in the initial analysis.

Vessel size and amount of vessel represented were estimated by matching rims to a polar coordinate template. After refitting conjoinable rim pieces from the same provenience and properly orienting the sherd(s) with respect to the horizontal plane of the rim, the curvature of the rim was matched to precisely drawn arcs to the nearest .5-cm radius. Once the radius match was obtained, the degrees of arc covered by the individual sherd or refitted sherds were measured along the appropriate arc. If a confident radius match could not be obtained because of warping or small length of rim relative to the size of the original rim opening, the length of the rim was measured and no other size estimate was attempted.

Radius and arc estimates were made on a little less than half (47.6 percent) of the rim sherds recovered from the tested sites. The rest were eliminated because they were too small or distorted to yield reliable estimates. The proportion of sherds for which estimates were made varies among different sites, vessel forms, and wares due mainly to differences in the degree of fragmentation. Table 15.10, Table 15.11, and Table 15.12 show the frequencies of measured and unmeasured rims by site, vessel form, and ware, respectively. Lillian's Site, Stanton's Site, and Catherine's Site all have significantly (p > .05) more measured rims, whereas G and G and Kenzie Dawn have significantly fewer measured rims than the other sites when comparing all rims using chi square analysis. This same pattern persists when comparing only bowl rims, but when comparing corrugated jar rims, only Lillian's Site has significantly more measured rims and Kenzie Dawn Hamlet has significantly fewer than expected. Vessel forms with small openings (ladles, mug/pitchers, narrow-mouth jars, and seed jars) have a significantly greater proportion of measured rims, whereas wide-mouth jars and unknown forms have significantly fewer measured rims because they tend to be smaller. Of the pottery wares, only plain gray has significantly fewer measured rims than expected. This is probably also because they are frequently smaller in size.

This measurement procedure can be biased against larger vessel sizes because rim sherds from larger vessels are probably measured in lower frequency than rim sherds from smaller vessels. This bias will increase in magnitude with increasing fragmentation of an assemblage. Consequently, given the variation in degree of fragmentation among different sites, vessel forms, and pottery wares documented above, the extent of bias against large vessels probably differs among assemblages. Fortunately, the effects of these different biases can be controlled somewhat by considering the degree of fragmentation and proportion of measured rims when comparing different assemblages.

Table 15.13 shows two measures of the frequency of the 10 different vessel forms identified from rim sherds. The first measure is a simple count of rim sherds in each form category. As a measure of the abundance of different vessel forms as vessels, rather than as sherds from vessels, counts are biased against vessel forms with small openings because they tend to produce fewer rim sherds than do vessels with large openings. As an alternative to counts, a total of the degrees of arc covered by each rim in a form category was also used. However, this measure of abundance contains a bias against vessels with large openings that have been broken into small sherds because the degrees of arc were generally not measured for these sherds. Although the relative frequencies of vessel forms derived from these two measures vary, sometimes dramatically, the rank order of vessel forms does not differ significantly.

In almost all assemblages, bowls constitute the most common vessel form, followed closely by wide-mouth jars. At Mad Dog Tower, wide-mouth jars are the most abundant by count, followed by bowls, but this order is switched when measured by degrees of arc. The reliability of both measures is probably compromised by the small sample size from Mad Dog, but the degrees-of-arc measure is the most problematic because only one-quarter of the wide-mouth jar sherds were suitable for arc measurements. Narrow-mouth jars and ladles alternate between being the third and fourth most abundant forms in the tested site assemblages (leaving out the unknown forms). Mugs, kiva jars, seed jars, and canteens all occur in low frequencies, and consequently their rank orders often vary among assemblages. The only clear pattern in these fluctuations is that seed jars tend to be more common at sites with abundant evidence of earlier components.

The role that vessel size plays in determining the abundance of sherds from different vessel forms can be evaluated for some forms by examining the distributions of rim radius estimates. For bowls, ladles, and wide-mouth jars, the relationship between vessel size and rim radius or diameter is linear with little variance (Blinman 1988a; Mills 1989a). Vessel forms with restricted openings, such as narrow-mouth jars, seed jars, kiva jars, and canteens, display a poor correlation between vessel size and rim radius. Table 15.14, Table 15.15, and Table 15.16 present the abundance of rim sherds in 2-cm rim-diameter size intervals for bowls, wide-mouth jars, and ladles, respectively. Both sherd counts and total degrees of arc are presented for each size interval. Degrees of arc offers a better measure for comparing abundance in different size intervals; it compensates for the underrepresentation of small vessels when quantified by count due to the fact that small vessels generally produce fewer rim sherds. The 2-cm size interval was chosen because replicate measurements of the same rim sherds indicate that measurement precision is usually slightly less than 2 cm, and thus it is difficult to distinguish reliably among vessel diameters less than 2 cm.

The size of ladles varies little among the 13 tested-site assemblages. The size distribution of all measured ladle sherds from the tested sites is unimodal, with the mode at 11 cm, and most (82 percent) fall between 8 and 14 cm in diameter. This matches well with the sizes of complete ladles in the Mesa Verde region documented by Mills (1989a:188, 251). The few ladle rim diameters that are much smaller or much larger than the normal range may represent unusually small or large ladles, or these values may be due to analytical error.

The data on bowl rim diameters display a multimodal distribution for all sherds and considerable variation among assemblages. Figure 15.13 shows the distribution of rim diameters standardized across all assemblages by summing the percentages for all assemblages in each size interval and using these values to calculate the proportions presented. This was done to reduce the influence on the total distribution that the few very large assemblages would have. The figure shows strong modes at diameters of 19 and 27 cm and a less pronounced mode at 11 cm. These modes match almost exactly those identified by Mills (1989a:176) in her study of complete Mesa Verde-region bowls, and also closely approximate size modes found in bowls from the Tusayan region (Mills 1989a:176; Turner and Lofgren 1966).

This close match between data derived from measurements of sherds and complete vessels has two important implications. First, although the precision of rim radius estimates from sherds can be assessed through multiple measurements of the same rim sherds, accuracy is difficult to determine without data on the size of the parent vessels. Since none of the measured sherds from the tested sites were refit into complete vessels, comparisons with data generated by Mills for other vessels from the same region provide the only check on accuracy. The correspondence of modes in the two data sets indicates that the accuracy of the sherd-based estimates of bowl diameters is probably fairly good. Second, the prominence of the mode at 27 cm suggests that the bias against larger vessels inherent in the sherd measurement process does not appear to adversely affect the tested sites data set. Perhaps the bias has a significant effect only on vessels larger than the largest consistent mode in Mesa Verde-region bowls, or perhaps enough of the sherds from large vessels at the tested sites are of sufficient size to yield reliable estimates. In either case, the size measurements from bowl rim sherds can be used as a reasonably unbiased estimate of vessel size distributions at the tested sites, except to the extent that sample size affects these distributions.

Figure 15.14 uses the relative frequency of sherds weighted by degrees of arc in three size classes constructed around the bowl diameter modes to compare bowl size distributions for the tested sites. The five assemblages containing unexpectedly large amounts of bowl sherds are shown as solid-filled symbols in this figure. Although Lookout House and Kenzie Dawn Hamlet have unusually large amounts of medium-size bowls, the other three assemblages (Stanton's Site, Catherine's Site, and Castle Rock Pueblo) look very similar to most of the other assemblages in terms of the distribution of bowl sizes. Thus, the greater amounts of bowl sherds in these assemblages appear to be due mainly to more bowls being consumed at these sites rather than larger bowls. Another intriguing pattern is evident in this figure. Two of the three assemblages with the most bowls in the large class come from the tower sites. Although small sample size could be a factor, particularly at Mad Dog Tower, it is unlikely that sampling error would bias both tower assemblages toward large vessels. Perhaps the use of these isolated tower-kiva sites involved the consumption of proportionately more large bowls than at the other residential sites.

Figure 15.15 indicates that sizes of corrugated wide-mouth jars are distributed unimodally, with the mode at 19 cm. This figure was generated in the same fashion as Figure 15.13. Although Mills (1989a:160) identified size modes in complete corrugated jars from the Mesa Verde region, the modes were based on vessel height rather than rim diameter. Since these two dimensions are correlated, modes should also appear in rim diameter data, but they may not be widely separated or pronounced enough to show clearly in data derived from rim sherd measurements. Several of the individual assemblages, such as Lookout House and Castle Rock Pueblo, are multimodal. But the position of these modes varies among the assemblages and in many cases may be an artifact of small sample size. The distribution of corrugated jar sizes at Stanton's Site, the only assemblage with significantly fewer gray ware jar sherds than expected, does not indicate that the use of smaller jars is the source of this variation. Consequently, as with the bowls, it appears that less consumption of gray ware jars is the most parsimonious explanation for the paucity of gray ware sherds.

The data presented above indicate that most sites have relatively similar pottery assemblages when wares and vessel forms are compared. The general similarity of these pottery assemblages indicates that most sites are probably functionally similar, at least in terms of the use of pottery. The general similarity of the pottery assemblages also suggests that the variables that influence the formation of pottery assemblages, including the size of the systemic inventory (that is, the number of different types of vessels in use) and the use life of different vessel classes (that is, the rate at which vessels break and are discarded) must be roughly similar at each of the tested sites. Differences in the size of the systemic inventory and vessel use life, combined with different formation processes (for example, the length of site occupation), would affect the formation of pottery assemblages and the frequencies of different categories of wares and vessel forms in assemblages (Mills 1989b). The general similarity in the pottery assemblages indicates that duration of occupation at most sites must have been sufficiently long for the frequencies of different vessel forms to stabilize (Mills 1989b).

Within the pattern of general similarity of the pottery assemblages, however, there is also interesting variation in the abundance of different forms. The analysis of vessel size presented above suggests that most of this variation results from differences in the consumption and accumulation of particular forms at some sites, rather than a consistent use of vessels of different sizes. The possible causes of this variation remain to be explored.

Pottery Production

Evidence of pottery production comes in two forms--direct and indirect (Blinman and Wilson 1993; Costin 1991). Direct evidence includes the features and debris resulting from pottery production, such as kilns, manufacturing tools, and raw materials. Indirect evidence usually consists of compositional data on the pottery itself, which can be used to link pottery to specific production areas based on similarities with spatially restricted sources of raw material. Although a study of pottery and raw-material composition is underway, insufficient data are currently available to evaluate indirect production evidence. Consequently, this discussion focuses on direct evidence recovered from the 13 tested sites.

Pottery-manufacturing tools, unfired pottery, and manipulated raw clay constitute the possible direct evidence of pottery production at the Sand Canyon locality sites. Manufacturing tools readily preserved in archaeological settings include base molds (or pukis), scrapers used to shape and smooth vessel surfaces, and polishing stones used to polish surfaces of white ware pottery (Blinman and Wilson 1993; Hill 1985, 1994; Waterworth and Blinman 1986).

Four of the five sherd containers recovered from G and G Hamlet, Lillian's Site, Stanton's Site, and Castle Rock Pueblo are possible base molds. (Two sherd containers are illustrated in the section on sherd tools.) Three of these items are made from the bases of corrugated jars, and the fourth is an unpainted white ware. Fragments of base molds may also exist among the modified sherds, but were not identified as such.

All of the modified sherds (defined later in this chapter) were examined for evidence of wear described by Waterworth and Blinman (1986) as diagnostic of tools used to scrape wet clay. No items display the drumlin-like wear features illustrated by Waterworth and Blinman (1986), although numerous modified sherds possess striations which run perpendicular to the length of the sherd edge and are indicative of a scraping motion. Further experimentation is required to determine the characteristics of the material worked by these sherds, but until then, no modified sherds from the tested sites have been identified as pottery scrapers.

All smooth stones and pebbles recovered from the tested sites were examined for evidence of abrasive wear, and those possessing such wear were identified as polishing stones. Most of the tested sites produced at least one polishing stone. (Polishing stones are discussed in more detail in a later section on battered and polished stone artifacts.) A small number of these polishing stones retain clay adhering to their surfaces, providing further evidence for their use as pottery polishers.

Small amounts of unfired pottery were recovered from three of the tested sites--Kenzie Dawn Hamlet, Lookout House, and Castle Rock Pueblo. All of the unfired pottery consists of fragments of white ware vessels of unknown form. Unfired pottery was recovered from a variety of contexts, including roof fall, trash fill, and mixed postabandonment fill.

Raw clay in various states was recovered from most of the tested sites. Most of the raw clay consists of unaltered chunks of shale or lumps of clay lacking temper or any other clear signs of manipulation. Small amounts of the raw clay had temper added, or were shaped into balls or coils, or were manipulated into other forms. Most of the raw clay is gray, although small amounts are white and reddish brown. Raw clay was usually recovered from trash and mixed postabandonment deposits. However, in some cases, large amounts of unmanipulated clay were recovered from structure floors and other possible use surfaces.

Table 15.17 shows the presence or absence of these possible different indicators of pottery production at the different tested sites. Only two sites, Saddlehorn Hamlet and Mad Dog Tower, yielded no direct evidence of pottery production. However, these assemblages are also among the smallest of the tested sites, and there is a significant positive linear relationship (r = .67, p = .04) between sample size and the richness of production evidence as shown in Figure 15.16. Consequently, without increasing the sample size from Saddlehorn Hamlet and Mad Dog Tower, it is impossible to argue convincingly that no pottery production took place at these sites. It is intriguing, however, that both sites are located in the lower Sand Canyon area. If borne out in future research, this pattern may indicate variation in the organization of pottery production between the upper and lower Sand Canyon areas during the late Pueblo III period.

Another fascinating aspect of the pottery production evidence from the tested sites is that it all appears to relate to white ware production only. No unfired gray ware or coarse-tempered raw clays were recovered, and the possible base molds and polishing stones were probably used exclusively in white ware manufacture. There are three possible explanations for this pattern: (1) very little or no gray ware pottery was produced at these sites; (2) something in the way gray ware pottery was produced in the Sand Canyon locality led to their raw materials and manufacturing tools not entering the archaeological record at settlements; or (3) the kinds of evidence we see are not actually clear indicators of pottery production, but result from other activities that involve unfired pottery, polishing stones, sherd containers, and raw clay. More work with indirect indicators of production should help evaluate these alternatives.

Pottery Exchange

Archaeologists use two approaches to study pottery exchange, one focused on identifying exotic or nonlocal styles, and the other employing compositional analyses of pottery and raw materials. Studies of the elemental and mineralogical composition of pottery recovered from sites in the Sand Canyon locality are currently underway, but were not yet ready to present at the time this report was written. Consequently, this discussion of pottery exchange relies entirely on the occurrence of readily identifiable nonlocal pottery styles which, in the Mesa Verde region, consist mainly of polychrome and red wares.

Table 15.18 presents the frequencies of nonlocal pottery types recovered from the tested sites. These nonlocal types cover a wide time range, with 65 percent of the nonlocal sherds coming from types made before A.D. 1150--most of which are San Juan Red Ware. Although many of these earlier types were recovered from sites with Basketmaker III and Pueblo II occupations, they were also found at sites with no other indications of occupations earlier than Pueblo III. This suggests that sherds were being scavenged off of earlier sites by the people occupying the sites during Pueblo III times. This pattern of scavenging may also account for sherds from types with time ranges that extend back to the middle A.D. 1000s and early 1100s, such as most of the White Mountain Red Ware. Consequently, much of the nonlocal pottery recovered from the tested sites may not indicate trade by the Pueblo III inhabitants of the sites but, instead, scavenging behavior that resulted in the accumulation of exotic materials at these sites. However, regardless of the mechanisms of accumulation, the occurrence of these nonlocal types does indicate that some trade occurred by people probably living in the Sand Canyon locality during the twelfth and thirteenth centuries. The particular types represented also indicate a western and southwestern orientation of this trade.

Sherd Tools and Ornaments and Other Pottery Artifacts

Numerous broken pieces of pottery recovered from the tested sites show evidence of recycling or reuse as tools, ornaments, and containers. These items are included in the discussions of pottery types and forms above, but are also treated separately here. In addition, a few fragments of coils and other nonvessel pottery artifacts were recovered and are discussed in this section. The pottery type identifications for the items identified as Late White Painted were checked in 1995, resulting in the reassignment of most of these pieces to other categories, mostly Pueblo III White Painted.

Perforated sherds with shaped edges are identified as pendants in the Crow Canyon analytical system (Figure 15.17a-f). Similarly shaped sherds, usually in some recognizable form such as disks, triangles, rectangles, etc., lacking a drilled hole or perforation, were identified as shaped sherds. These are probably ornament blanks, gaming pieces, and other nonutilitarian items. Large sherds with flaked or ground edges and sufficient depth to serve as containers are identified as sherd containers (Figure 15.18). This category may also include large sherds used as pottery molding trays or pukis. Irregularly shaped pottery fragments with wear on one or more broken edges are identified as modified sherds. Most modified sherds were probably used as tools and display a variety of wear patterns, but this category may also include fragments of sherd ornaments (shaped sherds and pendants) and containers.

Table 15.19 and Table 15.20 respectively show the frequencies of sherd tools and ornaments at the different tested sites and the pottery types from which these items were manufactured. The modified sherd category is the most abundant, with shaped sherds, sherd pendants, sherd containers, and "other pottery artifacts" being relatively rare or absent in the assemblages. The types used for sherd tools and ornaments include a wide variety of white, gray, and red wares. The white wares are dominated by Pueblo III types but also include some Pueblo II pottery, and both corrugated and plain gray ware are represented. Red wares account for a disproportionately large number of the shaped sherds and pendants and include types covering a wide temporal range and obtained from great distances. Most of the sherd containers were manufactured from the bases of corrugated jars.

Stone Tools and Manufacturing Debris
Christopher Pierce

Introduction

A wide variety of stone tools and manufacturing debris was recovered from each of the sites. The goals of this section are to briefly describe how these objects were analyzed, characterize the assemblages from each site, and briefly compare the different assemblages. Most of the space is devoted to describing the tool assemblages so that others interested in these collections can use the data for their own comparisons or for design of further research with the collections. Detailed descriptions of the laboratory procedures, analytic criteria, and quality-control measures can be found in the Crow Canyon laboratory manual (Schwab and Bradley 1987).

Methods

Two and sometimes three different levels of analysis were conducted on all of the stone materials recovered from the tested sites. The first level of analysis occurred coincident with the fieldwork, which began in the summer of 1988 and ended in the spring of 1992. After the artifacts were washed (with toothbrushes in tap water), this first analysis involved sorting the artifacts into categories and assigning catalog numbers to items in each category. Stone tools, cores, and other heavily modified objects were cataloged individually; debitage, minerals, pebbles, and certain unmodified objects were cataloged in groups by provenience. For items cataloged individually, the first analysis involved recording the raw material and condition, or completeness, of each artifact, in addition to identifying the artifact category. For items cataloged in groups, only the artifact category and comments were recorded in the first analysis.

This cataloging analysis was conducted primarily by Crow Canyon participants under the direct supervision of members of the lab staff. Of the 35 supervisors participating in this analysis, including full-time staff, part-time staff, and interns, the largest amount supervised by any single individual totaled 22.5 percent of the stone artifacts analyzed. Consequently, all of the stone tools were reanalyzed by Christopher Pierce and Joseph Keleher during the summer and fall of 1992 to reduce the impact of analytical bias on the collection descriptions. The analytic criteria used in the second level of analysis were identical to the initial analysis except for some changes in raw-material distinctions, discussed in more detail below. Finally, a third level of more-detailed analysis was conducted on selected artifact categories to generate data for addressing particular research problems. The methods and analytic criteria employed in these problem-oriented studies will be discussed, along with the presentation of the results.

Tools were identified in a variety of different general categories, including chipped stone, ground stone, battered and polished stone, pottery, ornament, "other modified stone," core, chipping debris, and probable manuport (unmodified stones and minerals that were probably carried into the sites by people). The criteria used to identify specific tool types within these categories will be discussed briefly, along with the assemblage descriptions. Stone raw-material categories used in this analysis consist of rock type identifications frequently combined with the geologic formation or member in which the material occurs naturally. Etzkorn (1993) presents descriptions of the raw-material categories and the identifying criteria used by Crow Canyon Archaeological Center. The raw-material categories used in this report differ from Etzkorn's in that some materials identified as nonlocal chert/siltstone by Etzkorn are grouped with the unknown chert/siltstones here. Our current knowledge of locally available raw materials indicates that these unknown cherts are probably nonlocal; thus, they are considered as such in this report. However, more intensive investigation of local raw-material sources could alter this assessment.

Artifact and Assemblage Descriptions

Chipped-Stone Tools

The chipped-stone tool category includes artifacts that were manufactured by removal of flakes either through percussion or pressure techniques. The Crow Canyon analytical system distinguishes six general categories of chipped-stone tool--projectile points, bifaces, drills, modified flakes, modified cores, and "other chipped-stone tools." Table 15.21 and Table 15.22 show the frequencies of these different tools and of the chipped-stone-tool raw materials, respectively. More details on the characteristics of artifacts identified in these six categories are presented below.

Projectile Points. Testing at 13 sites in the Sand Canyon locality produced 107 artifacts identified as projectile points. To be identified as a projectile point in the Crow Canyon system, the bifacially flaked items must have a pointed end opposed by a hafting element (either notches, stem, or flute) at the other end. Table 15.23 shows the frequency of projectile points from each tested site in nine form categories, and Table 15.24 presents the raw materials from which these different point forms were made. In the following discussion, descriptions are presented of the form, temporal span, condition, raw material, and recovery context of projectile points in each of the nine categories.

Paleoindian: One basal fragment of a projectile point (Figure 15.19a) attributable to the Cody Complex was recovered from Stratum 1 in the midden (grid square 92N/113E) at Kenzie Dawn Hamlet. The fragment shows roughly parallel pressure flaking, a slightly concave base, a lenticular cross section, and ground edges at the base. The raw material is mottled and speckled orange and pink chert reminiscent of material available in the Kaibab Plateau region.

Stemmed: Stemmed projectile points of probable Archaic age (four fragmentary and one complete) were recovered from the tested sites (Figure 15.19b-f). The four fragments are bases of stemmed points, but they are too fragmentary to be identified to a specific type. Each of these stem fragments is broken at the point where the stem meets the blade, and none shows evidence of edge grinding. However, these bases do vary in raw material and aspects of form. Two from Lillian's Site are made of locally available Dakota Formation orthoquartzite (Figure 15.19b,c) and are relatively thick in cross section. One has parallel sides; the other has a contracting stem. The stemmed base from Troy's Tower is made of jasper (Figure 15.19d), is relatively thin, has parallel sides, and may have been heat treated, as indicated by small thermal spalls on one surface. The base from a midden at Castle Rock Pueblo is made of a clear chalcedony, is relatively thin, and has a slightly contracting stem (Figure 15.19e). The complete stemmed point (Figure 15.19f) is identifiable to the Armijo type in the Oshara tradition (Irwin-Williams 1973) and shows evidence of recycling. The Armijo-style point is made of a reddish chalcedony or translucent chert and was recovered from the midden at Stanton's Site. A small projection was fashioned at the tip, and polish on this projection indicates that the artifact was last used as a drill.

Large Side-Notched: A single small fragment of the base of a large side-notched projectile point was recovered from the courtyard at Stanton's Site. The point, made of Dakota orthoquartzite, is too fragmentary to assign to a specific type.

Large Corner-Notched: Seven complete and seven fragmentary large corner-notched projectile points were recovered from eight of the tested sites (Lester's Site, Lookout House, Stanton's Site, Shorlene's Site, Roy's Ruin, Lillian's Site, Troy's Tower, and Kenzie Dawn Hamlet) (Figure 15.19g-q). Ten of the large corner-notched points conform to the expanding-stem and convex- or straight-based En Medio and Basketmaker II styles (Figure 15.19g-j) and are found at all of the sites with large corner-notched points except Troy's Tower. Of the other four large corner-notched points, two match the narrow, slightly contracting stem Basketmaker III/Pueblo I style (Figure 15.19o,q), and two fall into a broad-bladed, almost basally notched, form. A large, En Medio-style corner-notched point (Figure 15.19k) from Shorlene's Site (5MT3918) shows extremely heavy wear on the tip and edges of the blade, suggesting final use as a scraper. In addition, one of the Pueblo I-style points from Stanton's Site (Figure 15.19o) had been unifacially resharpened, forming beveled edges, with the tip used as a drill. Seven of the larger corner-notched points (one from each site) are made of Dakota orthoquartzite, one (from Stanton's Site) is made of chalcedony, and four (two from Lookout House, one from Stanton's Site, and one from Roy's Ruin) are made of unknown and presumably nonlocal cherts. Eight of the large corner-notched points were recovered from, or directly downslope of, midden areas; the remaining six points were recovered from fill in pit structures (N = 2), inner periphery areas (N = 2), a courtyard (N = 1), and an outer periphery area (N = 1). Complete large corner-notched points occur with equal abundance in midden and nonmidden areas.

Small Corner-Notched: Eight complete and 18 fragmentary small corner-notched points (Figure 15.20) of both the straight and expanding stem varieties were recovered from nine of the tested sites (Lester's Site, Stanton's Site, G and G Hamlet, Mad Dog Tower, Castle Rock Pueblo, Lillian's Site, Troy's Tower, Catherine's Site, and Kenzie Dawn Hamlet). These points match styles recovered from Basketmaker III, Pueblo I, and early Pueblo II contexts elsewhere in the region (Etzkorn 1993; Hayes and Lancaster 1975:144-146; Lekson 1987; Phagan 1988). Fifty percent of these small corner-notched points are made of locally available materials, with 10 made out of Dakota orthoquartzite (Figure 15.20a-e) and three made of Morrison chert (Figure 15.20f). The other 13 points are made of materials whose natural sources probably lie outside of southwestern Colorado, with 10 made of cherts from unknown sources (Figure 15.20g-k), two made of chalcedony (possibly from southeast Utah), and one made of Washington Pass chert (Figure 15.20m). There are no clear differences in the occurrence of small corner-notched points of local and nonlocal materials among the sites. About 50 percent of the small corner-notched points were recovered from midden deposits; the rest came from the fill of pit structures (N = 6), courtyard areas (N = 3), inner peripheries (N = 3), and outer peripheries (N = 2). Complete points are equally distributed between midden and nonmidden areas.

Medium Side-Notched: Medium side-notched points (two complete and one fragmentary, Figure 15.21a-c) were recovered from three separate tested sites (Lookout House, Saddlehorn Hamlet, and Kenzie Dawn Hamlet). This style of point is frequently associated with middle to late Pueblo II contexts in the Mesa Verde region and elsewhere in the northern Southwest (Bradley 1988b; Lekson 1987). The medium side-notched points from Lookout House and Saddlehorn Hamlet are made of Dakota orthoquartzite (Figure 15.21a,b), and the example from Kenzie Dawn Hamlet (Figure 15.21c) is made of chalcedony that does not appear to be available in southwestern Colorado. All three of the points were recovered from nontrash deposits, one from the fill of a pit structure (Structure 12 at Kenzie Dawn Hamlet), one from the surface room area of Lookout House, and one from the inner periphery at Saddlehorn Hamlet.

Bull Creek: Four complete points identified as the Bull Creek type (Figure 15.21d-g) were recovered from four different sites (Stanton's Site, G and G Hamlet, Castle Rock Pueblo, and Catherine's Site). These points are long and narrow with slightly to moderately indented bases and no notches. They occur in moderate to high frequencies in the Fremont area of southeastern Utah, where they have been dated to between A.D. 1050 and 1300 (Holmer and Weder 1980). All of the Bull Creek points are made of apparently nonlocal raw materials, with one made of agate/chalcedony (Figure 15.21d), two of unknown cherts (Figure 15.21e, f), and one of jasper (Figure 15.21g). Three of the points were recovered from midden areas; the example from Castle Rock Pueblo was recovered from the upper fill of a pit structure (Structure 204) constructed in the mid-A.D. 1200s. This appears to be the first documented occurrence of Bull Creek points in southwestern Colorado.

Small Side-Notched: Thirty-five complete and 11 fragmentary small side-notched points were recovered from the tested sites (Figure 15.22). This is the most abundant form in the projectile point assemblage, and only one site, Mad Dog Tower, failed to produce at least one example. These small side-notched points are the dominant form in Pueblo III contexts in the Mesa Verde region (Cattanach 1980; Hayes and Lancaster 1975; Rohn 1971). About 50 percent of the small side-notched points are made of locally available Dakota orthoquartzite (N = 20), Morrison chert (N = 2), and Morrison orthoquartzite (N = 1). The other half of the small side-notched points are made of a variety of nonlocal materials, including unknown cherts (N = 14), chalcedony (N = 5), petrified wood (N = 3), and obsidian (N = 1). Twenty-five of the small side-notched points were recovered from midden areas, 12 came from the fill of kivas, six came from inner periphery areas, one came from the lower fill of a tower (Structure 401 at Castle Rock), and one came from an outer periphery area.

Unusual Projectile Points: Five complete and two fragmentary projectile points of unusual form (Figure 15.21h-n) were recovered from five of the tested sites (Stanton's Site, Mad Dog Tower, Castle Rock Pueblo, Catherine's Site, and Kenzie Dawn Hamlet). These unusual projectile points fall into four categories. There are (1) three small, corner-notched forms with notches on only one side (Figure 15.21h-j); (2) two small, triangular forms with indented bases (Figure 15.21k,l); (3) a single small, narrow form with parallel sides and no notches (Figure 15.21m); and (4) a single pointed flake with only marginal retouch on the blade and two shallow notches (Figure 15.21n). Four of these points are made of Dakota orthoquartzite, two are made of unknown (probably nonlocal) chert, and one is made of obsidian (with small white inclusions) similar to obsidian that comes from the Jemez Mountains. Five of the projectile points come from midden areas, and the other two come from pit structure fill at Kenzie Dawn Hamlet (Structures 4 and 14).

Bifaces. Bifaces are items from which flakes were removed on two sides from a common edge. Flake-removal scars cover most or all of the surface of the object. In the Crow Canyon analytical system, this category consists of relatively thin items usually having pointed ends, including probable projectile point tips, preforms, and possible knives, but excluding larger, thicker bifacially worked items which are classified as cores. The tested sites produced 91 artifacts identified as bifaces (Table 15.25 and Table 15.26).

Probable Projectile Point Preforms and Fragments: Thirteen small, thin bifaces lacking notches and interpreted to be projectile point preforms (Figure 15.23) were recovered from six of the tested sites (Lester's Site, Lookout House, Stanton's Site, Mad Dog Tower, Catherine's Site, and Kenzie Dawn Hamlet). All of the preforms are small. Eight are complete; the rest are basal fragments. Three forms were identified: triangular (N = 8) (Figure 15.23), convex base (N = 4) (Figure 15.23), and concave base (N = 1). Nine of the preforms are made of Dakota orthoquartzite, and the other four are made of nonlocal materials (two chalcedony, one Washington Pass chert, and one unknown chert). Eight of the preforms were recovered from midden areas, four from the fill of pit structures (Structures 6 and 16 at Kenzie Dawn Hamlet and Structure 2 at Mad Dog Tower), and one from the outer periphery of Kenzie Dawn Hamlet.

The biface category also includes fragments of projectile points that lack hafting elements. When compared to other bifaces in Puebloan assemblages, projectile point fragments stand out because of their small size, thin cross section, and fine marginal retouch (Figure 15.23). Thus, in many cases, it is possible to distinguish projectile points even when only blade and tip fragments are present. Fifteen projectile point fragments spread fairly evenly across eight different tested sites (Table 15.25) were identified in the biface category. Of these, eight are made of Dakota orthoquartzite and the rest are made of probable nonlocal raw materials, including chalcedony, unknown chert, unknown quartzite, and petrified wood. Only four of the projectile point fragments were recovered from midden areas. Nine occurred in structure fill, courtyard, and inner periphery areas, and the remaining two came from outer peripheries.

Stage and Finished Bifaces: Most of the bifaces recovered from the tested sites consist of items in various stages of reduction from cores or large flakes (Figure 15.24). All of the sites except Saddlehorn Hamlet produced bifaces of this type. These bifaces were separated into four categories indicative of reduction stage--early, middle, late/finished, and unknown. Of the 63 artifacts assigned to these categories, most (67 percent) are late-stage or finished bifaces (Table 15.25), are made of locally available raw materials (81 percent; Table 15.26), and are fragmentary (84 percent). Over half (N = 34) of these bifaces were recovered from structure fill (mostly pit structures), courtyard, and inner periphery areas, but 20 came from midden areas, and the remainder were found in outer peripheries.

Drills. Artifacts identified as drills include three types: (1) formally shaped tools (usually shaped through bifacial reduction) with pointed projections; (2) flakes with ad hoc projections which show distinctive rotational wear on the tips; and (3) projectile points that were recycled into drills and show distinctive drill wear on their tips (Figure 15.25). Although drills are relatively rare in the collections from the tested sites, making up only 4 percent of chipped-stone tools, they were recovered from almost every site (Table 15.21). Drills are made mostly of locally available raw materials and consist of more fine grained materials than do other chipped-stone tools (Table 15.22). Most of the artifacts identified as drills are complete (67 percent), and the fragments are dominated by tip portions and a few formal drill bases. Almost 60 percent of the drills (16 items) were recovered from midden areas, and the rest came from structure fill, courtyard, and inner periphery areas.

Modified Flakes. Flakes from which other flakes were removed, either unifacially or bifacially (usually to form a tool edge), were identified as modified flakes. A total of 375 modified flakes was recovered from the tested sites. Although modified flakes account for over 55 percent of all chipped-stone tools recovered from the tested sites, their relative frequencies vary considerably among the different sites, ranging from 17 to 72 percent of assemblages (Table 15.21). More than 90 percent of modified flakes are made of locally available materials, and, unlike other categories of chipped-stone tools, Morrison Formation cherts dominate (Table 15.22). Most (75.5 percent) of the modified flakes are complete. Structure, courtyard, and inner periphery areas produced 48 percent of the modified flakes, whereas 43.5 percent came from midden areas and 8.5 percent were found in outer peripheries.

Modified Cores. A modified core is a chunk of rock from which large flakes were removed; the core was subsequently modified through retouch or as a result of use. If the modification is the result of battering, the item is classified as a peckingstone rather than as a modified core. Only two modified cores were recovered from the tested sites, one from G and G Hamlet and one from Kenzie Dawn Hamlet. Both modified cores are complete, made of material available in the Morrison Formation, and were recovered from mixed refuse and constructional deposits in or around structures.

Other Chipped-Stone Tools. The category called "other chipped-stone tools" contains artifacts that have had one or more edges modified by retouch or use, but the artifact is not a flake. In addition, two flaked items classified as modified cobbles (stream-worn stones made into tools) are grouped here with other chipped-stone tools. The tested sites produced 74 other chipped-stone tools, with Troy's Tower and Saddlehorn Hamlet being the only sites yielding none (Table 15.21). Like the modified flakes, over 90 percent of the other chipped-stone tools are made of local materials (Table 15.22), and most (62.5 percent) of the items are complete. A slight majority (47.2 percent) of other chipped-stone tools were recovered from structure, courtyard, and inner periphery areas, but midden areas produced 43.1 percent, and outer peripheries yielded the remaining 9.7 percent.

Ground-Stone Tools

Ground-stone tools include artifacts made of relatively coarse-grained materials; they show wear produced by abrasion, usually over broad surfaces of the artifact. Ten categories of ground stone were recovered from the tested sites, including three kinds of manos, three kinds of metates, abraders, a pestle, stone disks, and indeterminate pieces of ground stone. Table 15.27, Table 15.28, and Table 15.29 present the frequencies in the different ground-stone categories, the raw materials used, and the condition of the ground-stone tools, respectively.

Manos. Manos are the active, or hand-held, element of a grinding system that includes manos and metates. Two kinds of manos were recovered from the tested sites, two-hand and one-hand (Figure 15.26). Two-hand manos are large enough to require two hands to manipulate, whereas one-hand manos fit comfortably into a single hand. A third category, referred to as "indeterminate mano," includes items too fragmentary or unusually shaped to identify to either of the more specific mano categories. Two-hand manos make up 66 percent of the 151 manos recovered from the tested sites. Most (30 percent) of the remaining manos were assigned to the indeterminate mano category and only seven one-hand manos were recovered. Almost all manos (96 percent) are made of locally available raw materials, with sandstone dominating all mano categories. The Testing Program recovered mostly fragmentary manos, with only 19.9 percent complete or almost complete (called incomplete on Table 15.29). Considerable variation exists in the relative frequencies of manos recovered from the tested sites. Structure, courtyard, and inner periphery areas produced most of the manos (55.6 percent) from the tested sites; midden areas and trash deposits within structures accounted for 39.7 percent; and outer peripheries contributed 4.6 percent. The intrasite distributions of two-hand and indeterminate manos roughly match one another, but one-hand manos were restricted to structure and inner periphery areas.

Metates. Metates were separated into two types, slab and trough. Slab metates (Figure 15.27) are flat pieces of stone of variable thickness with grinding on one or both sides; frequently they are shaped along the edges by flaking and pecking. Trough metates possess a deep trough within which the grinding took place. Ground-stone artifacts that are too fragmentary or irregularly shaped to be assigned to a specific metate type, but that still show characteristics of metates (flat or concave grinding surface, relatively thick cross section, etc.), are classified as "indeterminate metates." Ninety-eight pieces identified as metates, most of which are fragments (see Table 15.29), were recovered from the tested sites, with 84 (85.7 percent) classified as indeterminate, 13 (13.3 percent) as slab, and one (1.0 percent) as trough. All of the metates are made of local materials and, like the manos, sandstone dominates. Most (90.8 percent) of the metate pieces collected are fragmentary. The single trough metate fragment came from Kenzie Dawn Hamlet, one of the two sites with Basketmaker III components. Relative frequencies of slab and indeterminate metates vary considerably among the different assemblages. Within settlements, 49 percent of the metates came from structure, courtyard, and inner periphery areas. Midden areas produced 46 percent of the metate pieces, and the remaining 5 percent came from outer peripheries.

Abraders. The abrader category includes both active and passive grinding elements that are too small to be part of a mano-metate grinding system or that display wear (such as ground grooves and small, shallow depressions) that distinguishes them from the other large, relatively flat-surfaced grinding tools (Figure 15.28). Thirty-five abraders were recovered from 11 of the tested sites. The two tower sites were the only assemblages lacking abraders. All but one of the abraders are made of local sandstone, with the single exception composed of conglomerate. Abraders are mostly complete or of unknown condition. This pattern results from the definition of the category. Items had to be complete enough to determine if the object, when whole, was a small, hand- or lap-size tool. Small pieces of abraders were likely identified as indeterminate manos, metates, or indeterminate ground-stone fragments. Most (65.7 percent) of the abraders were recovered from midden areas. Structure and inner periphery areas produced 31.5 percent, and outer peripheries yielded only 2.9 percent of the abraders. No abraders were recovered from use surfaces.

Pestle. Pestles are cylindrically shaped stones with grinding and battering wear on at least one end. A single complete pestle made of igneous rock was recovered from the fill of a surface room (Structure 9) at Kenzie Dawn Hamlet.

Stone Disks. Stone disks are tabular pieces of stone shaped into a circle by flaking and grinding. They probably served as lids for pots and other containers. One fragmentary and four complete stone disks were recovered from two of the tested sites, Kenzie Dawn Hamlet and Castle Rock Pueblo. Of the three from Kenzie Dawn Hamlet, two are complete sandstone disks from the fill of a pit structure (Structure 4), and one is a fragment made of unknown chert recovered from trash located in an earth-walled pit structure (Structure 13). The two stone disks from Castle Rock Pueblo are both complete and were recovered from midden areas. One is made of sandstone, and the other is a dark, indurated shale. The final stone disk, from Lester's Site, is a complete item made of indurated shale.

Indeterminate Ground Stone. Indeterminate ground stone includes objects made of abrasive material that have at least one ground surface, but the items are too fragmentary to be identified to more specific types. The tested sites yielded a total of 416 pieces of indeterminate ground stone, making this the most abundant (58.8 percent) of the ground-stone categories. All of the indeterminate ground-stone artifacts are fragments, and almost all are made of local materials, with sandstone far outnumbering all other materials. Indeterminate ground-stone fragments were recovered from all areas of the tested sites, but midden areas and trash fill in structures produced almost half of these artifacts. Structure, courtyard, and inner periphery areas yielded 38.4 percent; outer peripheries produced 13 percent.

Battered and Polished Stones

Stones made of materials that fracture conchoidally and are battered or polished as part of manufacture or use make up the 10 categories in this group, including three kinds of battered stones, five kinds of axes and mauls, polishing stones, and polished igneous slabs. Table 15.30 and Table 15.31 show the frequencies and raw materials, respectively, of the different battered and polished tools recovered from the tested sites.

Battered Stones. Three kinds of battered stones were distinguished--peckingstones, hammerstones, and polishing/hammerstones. Peckingstones (Figure 15.29) are tools that show battering on acute-angled edges and rounded surfaces; edges were usually produced by flaking. This category commonly includes recycled cores. Hammerstones (Figure 15.30e-g) are tools with battering on rounded edges; they are often cobbles or heavily worn items. Polishing/hammerstones are hammerstones that have at least one polished surface (Figure 15.30a-d). Battered stones make up almost 80 percent of all battered and polished stones from the tested sites. Peckingstones and hammerstones are the most abundant and occur at all sites, whereas the polishing/hammerstones are rare and were recovered from only four sites (Table 15.30). Most (greater than 80 percent) of the battered stones are complete and show varying degrees and kinds of wear. The battered stones are made primarily of local Morrison and Dakota Formation cherts and orthoquartzites, with a few made of other local and possibly nonlocal raw materials (Table 15.31). A majority (51.5 percent) of the battered stones were recovered from midden areas and trash deposits located in and around structures. Battered stones also occurred in relatively high frequencies (43.9 percent) in the fill and occasionally on use surfaces of structures, but they were rarely recovered from courtyard (2.9 percent), inner periphery (8.4 percent), and outer periphery (4.6 percent) areas. Peckingstones were most common in midden areas and other trash deposits, whereas hammerstones occurred most often in the fill of structures.

Axes and Mauls. Axes are polished and sometimes flaked stones with a sharpened bit on one (Figure 15.31) or both (Figure 15.32a,b) ends and a notch or groove for attachment to a handle. Mauls have similar hafting elements, but have blunt, battered ends rather than sharpened bits (Figure 15.32c-e). Two kinds of axes were identified--single- and double-bitted. Single-bitted axes possess a sharpened bit on one end only; double-bitted axes have bits at both ends and a hafting element in the middle. Pieces of axes missing one end are identified as indeterminate axes because it is not possible to determine if they had one or two bits. Fragments missing both ends and without sufficient characteristics to distinguish between an axe or a maul were placed into an axe/maul category. Axes and mauls are rare at all sites, but they appear to be slightly more common in the later, post-A.D. 1250 assemblages (Table 15.30). Only three assemblages, Lillian's Site, Troy's Tower, and Mad Dog Tower, lack axes and mauls altogether. Among the axes that could be identified to a specific type, single-bitted axes are the most common. Almost all axes and mauls are made of Morrison Formation orthoquartzite; a few items are made of Morrison chert, Dakota orthoquartzite, and unknown chert and quartzites. Most of the items identified as mauls and specific axe types are complete; the indeterminate axes and axe/maul categories are dominated by fragments. Structure fills, courtyard, and inner periphery areas produced most (53 percent) of the axes and mauls from the tested sites. Midden areas and trash deposits in and around structures yielded 38.2 percent of axes and mauls. Only 8.8 percent came from outer peripheries.

Polishing Stones. Polishing stones are smooth pebbles with one or more surfaces abraded by rubbing the stones against another surface. The degree of wear varies from slight striations to heavy attrition of the abraded surface. These stones are commonly thought to have been used to polish the surfaces of pottery vessels, although other uses are possible. Twenty-nine polishing stones (62.1 percent complete and 37.9 percent fragmentary) were recovered from eight of the tested sites (Table 15.30). For rare items such as polishing stones, recovery and abundance are likely determined to a great extent by sample size; thus, sampling error can play a large role in intrasite and intersite distributions. Polishing stone raw materials are dominated by unknown quartz-rich rocks and include dinosaur gastroliths and stream-worn pebbles probably derived mainly from the conglomeritic layer at the base of the Dakota Formation (sometimes referred to as the Burro Canyon Formation). Most (53.6 percent) of the polishing stones came from the fill of structures, 39.3 percent were recovered from midden areas and other trash deposits, and 7.1 percent came from inner periphery areas. Outer peripheries produced no polishing stones.

Polished Igneous Slabs. Polished igneous slabs are chunks of water-worn porphyritic igneous rock (probably grano-diorite) usually with one highly polished surface. Four (one complete and three fragmentary) polished igneous slabs were recovered from three of the tested sites (Table 15.30). The occurrence of two polished igneous slabs at Castle Rock Pueblo is probably related to its proximity to the source of the raw material. Three of the slabs were recovered from midden areas, and the fourth came from the fill of a pit structure.

Other Stone Artifacts

This general group of stone tools includes ornaments and a variety of "other modified stones and minerals" that do not fit easily into the categories discussed above. Table 15.32 and Table 15.33 show the frequencies at the tested sites and the raw materials of ornaments and other modified stones and minerals, respectively.

Ornaments. Ornaments include shaped pieces of stone or mineral that were perforated, usually by drilling. Small circular ornaments are identified as beads (Figure 15.33); larger, elongated specimens are pendants (Figure 15.17g-q). Fragments of ornaments lacking the perforated portion were identified as other modified stones and minerals and are discussed below. Small numbers of ornaments were recovered from 10 of the 13 tested sites, and pendants outnumber beads by more than three to one (Table 15.32). Most of the ornaments are made of unknown cherts and other fine-grained stone materials probably of local origin (Table 15.33). Only one piece of turquoise was recovered. Ornaments made of bone, shell, and pottery are discussed elsewhere in this chapter. Of the 32 stone or mineral ornaments recovered from the tested sites, most (56.3 percent) are complete or missing only small portions, and the rest are fragments that still retain evidence of a perforation. Nontrash fill of structure, courtyard, and inner periphery areas produced 15 (46.9 percent) of the ornaments, and three other ornaments came from structure or courtyard surfaces. Midden areas and trash fills yielded most of the remaining ornaments (12 pieces), with only one ornament coming from an outer periphery area and one from a looter's backdirt pile at Kenzie Dawn Hamlet.

Other Modified Stones and Minerals. A wide variety of stones and minerals modified through polishing, grinding, flaking, battering, and/or fire-alteration are included in this category. The tested sites yielded 200 items identified as "other modified stones and minerals." In a subsequent analysis, these artifacts were divided into seven subcategories (polished, ground, edge-shaped, flaked/battered, pigment, fire-altered, and other) based on the kind of modification and, in some cases, the raw material (Table 15.32).

Polished: The subcategory "polished other modified stones and minerals," includes mainly ornament fragments and blanks lacking signs of a perforation. In addition, larger fragments of polished shale, including one possible tchamahia fragment, were grouped into this subcategory. These polished items make up most (58.9 percent, 118 items) of the other modified stones and minerals from the tested sites. Each tested site produced at least one polished item, with abundance at any particular site increasing with increased sample size (Table 15.32). As with the ornaments, polished other modified stones and minerals are made principally of local and unknown cherts and other fine-grained materials, including jet (five pieces) (Table 15.33). Midden areas and other trash fills accounted for 44.4 percent of the polished items from the tested sites. Structure, courtyard, and inner periphery fills and surfaces produced most (47.0 percent) of the other polished artifacts, with outer peripheries, looted areas, and general site contexts yielding only 8.6 percent.

Ground: Ground other modified stones include mano and metate blanks, anvil stones, recycled ground-stone tools, and other miscellaneous pieces of ground, abrasive stone that did not fit into the other ground-stone tool categories. After the polished items, the 38 ground pieces make up the next-most-common subcategory of other modified stones and minerals, and Saddlehorn Hamlet and Lester's Site were the only tested sites lacking such items. Relative frequencies at the different sites vary considerably and do not appear to correlate directly with sample size. The ground items consist mainly of fragments (65.8 percent) made of sandstone and other abrasive raw materials, with just a few fine-grained materials represented (Table 15.33). Most of the ground other modified stones came from midden areas and other trash fills (60.5 percent); structure, courtyard, and inner periphery fills and surfaces produced all of the remaining items (39.5 percent).

Edge-Shaped Slabs: Edge-shaped slabs are pieces of sandstone and orthoquartzite with shaped edges--shaped mainly by flaking and occasionally grinding. These items were probably used as covers for structure openings. Eight fragments and two complete specimens were recovered from five of the tested sites (Table 15.32). Half of the items were recovered from the fill of pit structures, 30 percent from trash deposits, and the remaining 20 percent from inner periphery deposits.

Flaked/Battered: This subcategory includes mainly large angular chunks and subangular stream cobbles that lacked much initial shaping but from which flakes had been removed through use as choppers, scrapers, and crushers. Nine mostly complete flaked/battered items were recovered from seven of the tested sites (Table 15.32). All are made of local or unknown materials, including porphyritic igneous rock, sandstone, orthoquartzite, chert, and conglomerate. Three of these artifacts came from midden and trash fill deposits, three came from nontrash structure fills, one came from courtyard deposits, and two came from outer peripheries.

Pigment: Iron-rich raw materials with abraded surfaces were placed into the pigment category. Twelve such pigment stones were recovered from eight of the tested sites, with most sites producing only one (Table 15.32). Nontrash structure fills and inner periphery deposits yielded eight (66.7 percent) of the pigment stones, with midden areas and trash fills in structures accounting for the remaining 33.3 percent.

Fire-Altered Rocks: Stones fractured or otherwise altered by exposure to heat but lacking any other modification were identified as fire-altered rocks. Four such items were recovered from two of the tested sites (Table 15.32). Three of them are angular chunks; one is a sandstone slab that may have served as a griddle or hearth liner. They were recovered mostly from structure fill, one from a trash deposit and two from nontrash deposits, and one came from inner periphery deposits.

Other: The "other" category includes pebbles, gastroliths, and other rounded stones (which were either produced or altered through some form of modification) as well as shaped building stones. One of the building stones shows a petroglyph pecked into one face (Figure 15.34). Nine items, most of them fragmentary, were recovered from seven of the tested sites (Table 15.32). All are made of local or unknown raw materials, including sandstone, shale, orthoquartzite, and porphyritic igneous rock. All of these other modified stones were recovered from pit structure fills (66.7 percent) or midden areas and other trash deposits (33.3 percent).

Unmodified Stones and Minerals

Unmodified stones and minerals were collected during the Testing Program when, in the excavator's opinion, the material did not occur naturally at the site. Given the wide variety of experience levels of field personnel and the difficulty that even experienced field workers can have in distinguishing natural sediment from manuports, it is likely that there was considerable variation within and between sites in the recovery of these kinds of artifacts. Consequently, this presentation focuses on the presence or absence of certain kinds of unmodified stones and minerals in the tested-site collections and considers abundance only when substantial differences exist.

Unmodified stones and minerals were grouped into six categories for presentation in this report--gizzard stones, pebbles, igneous rock, concretions, crystals, and fossils. Gizzard stones are small pebbles consumed by birds to aid digestion. While in the bird's gizzard, sharp edges become rounded and polished, creating stones with heavily worn edges surrounding relatively unworn surface facets. Stones matching this description were recovered from each of the tested sites, and only one site, Kenzie Dawn Hamlet, produced a relatively large number when weighted by the frequency of debitage.

Pebbles are naturally rounded stones that can be collected from streams and outcrops of the conglomeritic Burro Canyon Formation, which lies between the Dakota and Morrison Formations. Pebbles were recovered from each site, and although there is variation in abundance, no assemblages stand out as substantially different.

Igneous rock consists of angular chunks of diorite that may be the residue from crushing these rocks for use as temper in pottery. However, studies designed to determine if the composition of these rocks matches that of the igneous temper have not yet been conducted. Igneous rock was collected from all sites except Troy's Tower, and Castle Rock Pueblo yielded considerably more than the other sites, probably due to its proximity to the source at Ute Mountain.

Concretions are iron-rich nodules formed in the sedimentary rocks exposed in Sand Canyon that may have been collected for grinding to make pigment or because of the odd forms they sometimes take. G and G Hamlet, Lester's Site, Mad Dog Tower, and Saddlehorn Hamlet were the only sites that did not yield any concretions. Both Stanton's Site and Catherine's Site produced a large number of concretions, possibly because of their natural inclusion in the sediments eroding from the Dakota Formation escarpment below which the sites are located.

Crystals of a variety of minerals were recovered from the tested sites. Carbonate minerals, mostly calcite and aragonite, are the most common, but quartz and other unidentified crystals were also recovered in small numbers. There are several sites from which no crystals were collected, including Mad Dog Tower, Shorlene's Site, Roy's Ruin, Lillian's Site, Troy's Tower, and Catherine's Site. However, given the small numbers recovered from other sites, this may be due entirely to the combined effects of sampling error and collection bias.

Fossil shell, bone, and wood were collected in very low frequencies from all of the tested sites except Mad Dog Tower, Troy's Tower, and Lillian's Site. About half of the fossils are unmodified rounded pieces of petrified wood, and most of the others are shell fragments. It is difficult to determine at this time if these fossils were transported to the sites by people.

Cores and Debitage
Christopher Pierce and Joseph W. Keleher

The process of flaking stone to manufacture tools results in the production of cores and debitage, or chipping debris. Cores are nodules of stone from which relatively large flakes have been removed, but these nodules were not then used as tools. The analysis of cores occurred at the cataloging stage and involved recording the condition and raw material of individual items. These identifications were checked and adjusted if necessary during the reanalysis of tested site materials conducted by Pierce and Keleher discussed earlier. Different types of cores were not distinguished in either analysis. Debitage includes flakes and shatter, or angular debris, removed from cores and chipped-stone tools.

Pieces of debitage, including possibly utilized items lacking intentional modification, were separated from other stone artifacts during cataloging. All debitage was then subjected to a second level of analysis in which debitage types and raw materials were recorded for items grouped by provenience. Four debitage types were distinguished--flakes, flake fragments, edge-damaged pieces, and angular debris. Flakes possess a clear ventral, or inside, surface and a striking platform remnant. Flake fragments also have a clear ventral face, but they lack a platform remnant. Angular debris is conchoidally fractured material lacking both a clear ventral face and a platform remnant. Edge-damaged items can be any of the debitage types (though usually flakes and flake fragments) that possess wear along one or more edges. This wear could result from use, or it may have been produced by natural attritional processes, such as trampling. Flakes from tools such as peckingstones, hammerstones, and axes were also included in the general debitage category and were not distinguished in the analyses performed. An additional small sample of debitage from four of the tested sites was analyzed to obtain more data on raw-material selection and use. The methods and results of this analysis are discussed in a separate section below.

Table 15.34 presents the frequencies of debitage types and cores, and Table 15.35 shows the raw materials of cores and debitage from each of the tested sites. The chert and orthoquartzite from the Morrison Formation that are reported separately in other tables are combined into one category in this table because of inconsistencies in the criteria used to distinguish them during the debitage analysis. Although flakes and flake fragments dominate all assemblages, some variation exists in the relative frequencies of debitage types and debitage/core ratios among the different sites. However, this variation does not appear to be related to geographic position, age, site type, or historic land use. At the level of local vs. nonlocal raw materials, little variation exists among assemblages, with over 90 percent of the debitage at all sites being of local material. However, within the local and nonlocal groups, some variation exists among assemblages. G and G Hamlet (on the mesa top) and sites located in the east side of Sand Canyon (Stanton's Site, Catherine's Site, and Mad Dog Tower) yielded more Dakota Formation orthoquartzite than did other sites. This pattern may be a result of variation in source proximity, but insufficient documentation of material sources in the locality makes it impossible to test this possibility. Among the nonlocal materials, obsidian shows the most pronounced pattern. Two sites, G and G Hamlet and Stanton's Site, produced 87 percent of the obsidian debitage from all of the tested sites, a pattern not seen in the chipped-stone tools.

Attribute Analysis of Debitage. To obtain additional information on the selection and use of chipped-stone raw material, samples of debitage were analyzed from four of the tested sites. G and G Hamlet and Roy's Ruin are located on the mesa top north of Sand Canyon and were occupied primarily between A.D. 1180 and 1240. Lester's Site and Stanton's Site were built on the talus slopes in upper Sand Canyon and date between A.D. 1250 and 1280. All four sites lie within 2 km of each other and residents of the sites probably had similar access to stone raw materials.

For this analysis, only the chipped-stone debris from middens was examined. Artifacts in middens accumulated through refuse disposal during the occupation of the sites and therefore are more likely to be representative of the kinds and quantities of materials used than artifacts from nontrash deposits. There are two habitation components at G and G Hamlet, and each component has a spatially discrete midden. Only debitage from the later midden was selected for this analysis. The sample analyzed from the four sites totals 3,512 pieces of chipped-stone debris, including 588 pieces from G and G Hamlet, 253 pieces from Roy's Ruin, 463 pieces from Lester's Site, and 2,208 pieces from Stanton's Site.

Five attributes were recorded for each item--geologic formation, color, graininess, use wear, and weight. Geologic formation was identified by comparing the rock type, color, and grain size of the archaeological specimens with the reference collection at Crow Canyon. Six categories were used: Morrison Formation, Dakota Formation (including material from the Burro Canyon Formation), local sandstone, local igneous, nonlocal, and unknown. Additional information was recorded under comments. For example, when a nonlocal material could be identified to a known source, this was noted. The unknown category includes mostly cherts that probably derive from geologic formations exposed in Sand Canyon but whose sources have not yet been located.

The color of archaeological specimens was recorded by matching to color groups represented by collections of paint chips of similar color. Eight color groups were used: green, brown, purple, red, yellow, gray, white, and black. If a piece was a mix of different colors, it was recorded as mixed, and if the color could not be determined easily, it was coded as other. Variation among assemblages in color of materials from the same geologic formation may indicate selection for particular colors or use of a specific source with distinctive color characteristics.

Graininess was observed on freshly broken edges and faces and was classified as fine, medium, coarse, mixed, or indeterminate. Materials identified as fine have sharp, clean edges and faces that are smooth and glossy. Medium-grain materials have slightly less sharp edges and duller, grainier surfaces when compared with fine material. The edges and surfaces of coarse material are dull and rough when compared with the fine and medium materials. If an individual piece varied in graininess enough to crosscut the grain categories, it was recorded as mixed. Finally, unknown pieces were too small or weathered to identify easily to a particular grain class. Graininess recorded in this fashion provides a measure of material quality.

Macroscopic use wear was noted as being present or absent. A conservative approach was taken in recording use wear, and only distinct examples of edge grinding, flaking, and polish were used to identify pieces as worn. Consequently, it is possible, perhaps likely, that many utilized items were not identified as such in this analysis.

Weight was recorded to the nearest tenth of a gram using an electronic balance. In cases in which an artifact weighed less than 0.1 grams, the item was recorded as weighing 0.1 grams. Weight was recorded for use as a measure of artifact size.

Results: Table 15.36 shows the grain and color characteristics of debitage from the six geologic source categories for the entire analyzed sample. The Morrison Formation material consists mostly of medium-grained gray, purple, and green stone. The Dakota Formation material is evenly divided between medium- and coarse-grained white stone with some gray and mixed colors. The nonlocal and unknown materials are predominantly fine grained, with white, black, gray, brown, and mixed colors most common. The relationship between grain and color varies among the different geologic sources. In the Morrison Formation material, a chi square analysis indicates a highly significant (Pearson chi square = 215.9, p < .00000) association between grain and color, with the mixed and white materials being finer grained than expected and purple being coarser grained. The Dakota Formation, nonlocal, and unknown materials show no clear association between color and grain.

A comparison of the characteristics of the two most abundant and locally available sources, Morrison and Dakota Formation, in the four assemblages provides insight into patterns of raw-material selection. Table 15.37 and Table 15.38 present the frequencies of graininess and color for Morrison and Dakota Formation materials, respectively, in the four assemblages analyzed for this study. A chi square analysis indicates that significant (Pearson chi square = 233.3, p < .00000) variation exists among the assemblages in the color of Morrison Formation materials. G and G Hamlet shows significantly higher frequencies of gray and mixed color materials and lower frequencies of purple and green materials than expected from marginal and grand totals. Roy's Ruin has significantly higher frequencies of brown and mixed colors and lower frequencies of gray. Stanton's Site has more purple than expected and less green, brown, and mixed colors. Lester's Site is higher than expected in green Morrison materials and lower in purple. A similar pattern of association between color and debitage assemblage does not exist with the Dakota Formation material (Pearson chi square = 10.6, p = .10).

The graininess of materials varies significantly among assemblages for both Morrison (Pearson chi square = 82.2, p < .00000) and Dakota (Pearson chi square = 71.0, p < .00000) Formation materials. For the Morrison Formation materials, G and G Hamlet contains more fine- and coarse-grained and fewer medium-grained materials than expected. Roy's Ruin does not differ significantly from expectations in the distribution of graininess. Stanton's Site and Lester's Site both possess significantly less fine-grained Morrison material than expected. In the Dakota Formation materials, G and G Hamlet and Roy's Ruin have significantly higher frequencies of fine and medium material and lower frequencies of coarse material than expected. Stanton's Site contains significantly more coarse-grained Dakota material than expected, and Lester's Site does not differ from that expected from marginal and grand totals.

These patterns of variation among the assemblages in local raw-material characteristics could have resulted from use of different discrete sources with distinct qualities, selection of particular raw materials for specific uses whose frequencies of execution varied among the settlements, or stylistic variation in raw-material selection with no implications for, or constraints from, differential use or availability. The lack of adequate data on the specific sources of these materials in the Sand Canyon locality and their particular use and performance characteristics makes it difficult to evaluate these different hypotheses. However, patterns in the presence of use wear and the size of debitage provide some data on the possible differential use of raw materials.

Table 15.39 summarizes the color, graininess, and source characteristics of used and unused debitage from the four assemblages grouped together. Chi square analyses indicate that no significant association exists between the use of flakes as expedient tools and their color (Pearson chi square = 2.7, p = .84) or geologic source (Pearson chi square = 3.8, p = .15). There is a weak association between use and graininess (Pearson chi square = 6.5, p = .03). Fine-grained material shows a significant increase in the amount of utilization. No clear pattern of variation exists among the assemblages in terms of the graininess of utilized flakes; however, the low frequency of utilized flakes in most of the assemblages makes it difficult to search for meaningful patterns using chi square analysis. The size of debitage as measured by weight varies with graininess and geologic source (Figure 15.35). Controlling for these factors in comparing the different assemblages indicates little or no difference in debitage size among assemblages. Although the average weight of used debitage is about five times greater than that of unused debitage, there is no significant difference in the weight of utilized pieces among the settlements. Thus, short of conducting a more intensive use-wear analysis, there does not appear to be any difference in the use of debitage among the four settlements. Given that there probably are measurable differences in fracture toughness and edge quality among the different kinds of raw materials used in the Sand Canyon locality, differences in the characteristics of discrete sources used by the different settlements appear to be the most likely explanation for variation in raw-material selection exemplified in the four assemblages.

Bone Artifacts
Jonathan C. Driver

Small collections of bone artifacts, such as those from the Site Testing Program, provide relatively few opportunities for analysis beyond description. Three hundred six bones from 13 sites excavated in the Sand Canyon locality were identified as deliberately modified pieces (as opposed to pieces modified accidentally by burning, butchery, etc.). These deliberately modified specimens include complete and fragmentary bone and antler tools, as well as waste material produced during manufacture. The purpose of this section is to describe and classify the artifacts, to examine methods of manufacture, and to identify the frequency of different wear patterns on some artifact classes.

Identification, Classification, and Recording

Artifacts made of bone and antler were often identified during excavation and bagged separately. Other artifacts were identified later in bags of unmodified nonhuman bone sent to zooarchaeologists for analysis. These were separated from the remainder of the zooarchaeological assemblages after analysis and returned to storage as separate items. All of the bone artifacts were subsequently sent to the author for study.

All specimens were examined by eye and with a low-powered (up to 40×) microscope. A number of specimens were rejected as bone artifacts, and a separate list was kept of these; these items were returned to the collections manager. Criteria for acceptance of a specimen as deliberately modified are somewhat subjective and also depend to some extent on the size of the specimen. The following criteria were used: (1) overall shape, (2) presence of cut marks forming a pattern which would not be created by butchery, (3) presence of numerous striations aligned in parallel sets, and (4) well-defined polish.

For each item, the following information was recorded in the database: provenience, type of artifact (using Kidder's [1932] system for awls, and categories defined by Bullock [1992] for other artifacts), taxon from which the bone originally came, skeletal element from which the artifact was made, and length of the fragment. In addition, information was recorded on the shape of tip and wear patterns on bone awls.

Awls are artifacts with a single pointed end and a blunt head formed in a variety of ways. Awls were frequently produced on long bones, and Kidder defined five varieties which in this study are defined as awl A through awl E. Another type was made on a rib (awl F). Finally, three types (awls G, H, and I) define fragments. The types are as follows: awl A, long bone with intact unmodified epiphysis forming the head; awl B, split long bone with epiphysis forming head, otherwise unmodified; awl C, long bone awl with head formed of partly worked epiphysis; awl D, long bone awl with head formed of fully worked epiphysis; awl E, long bone shaft fragment awl; awl F, awl made on a rib; awl G, any awl fragment lacking the pointed tip but retaining the head; awl H, any awl fragment lacking the head but retaining the tip; and awl I, any awl fragment lacking tip and head.

Tube artifacts were produced by cutting grooves around long bones (generally turkey) and snapping the bone along the groove to remove the unwanted proximal and distal ends. This allowed the production of a range of artifacts. Tubes are long bones which have been grooved and snapped at two ends and which are longer than 3 cm. Tube beads were made in the same way as tubes, but are 3 cm or less in length (Bullock 1992). Offcuts are the waste pieces produced in the production of tubes; they have been grooved and snapped at one end and display the articular end at the other. Offcut/tubes are pieces of long bone which have been grooved and snapped at one end but are broken at the other end. It is not possible to tell whether these are broken tubes or whether they are offcuts which lack the articular end. Whistles are tubes which have one or more holes cut through the cortical bone into the medullary cavity.

A range of miscellaneous artifact types was also recorded. The artifact type "scapula" refers to turkey scapulae which have been broken across the blade and have had a beveled edge created along the break. These tools have been identified at other sites in the region (Bertram 1991) but seem rare elsewhere. Scrapers are usually formed of mammal long bones (typically a humerus) and have a broad, beveled worked end. Flakers are antlers or bones with a blunt point, assumed to have been used for flaking stone. Beads are small, circular discs of bone with a hole in the middle. Drilled bones are usually small mammal long bones in which a hole has been drilled into the proximal end, piercing the epiphysis and entering the marrow cavity. Walker (1990a) describes these as tinklers. "Other" refers to fragments of worked bone artifacts which cannot be matched to one of the formal types defined above. They are usually quite small. It is likely that many are small fragments of awls.

Following the criteria defined by Bullock (1992) in her study of bone artifacts from a number of sites in the Sand Canyon locality, the shape of awl tips was defined by the plan view and cross section. For the plan view, Bullock (1992:Figure 2.2) defined three shapes: triangular (tri), rounded/convex triangular (rct), and plano-concave (pcc). For the cross section of the tip, she defined three types: angular (an), circular/oval (co), and flat (fl) (Bullock 1992:Figure 2.2). Each awl tip was classified by assigning a shape and a cross section type (for example, rctco, trian, etc.). In addition, the presence of striations and other wear patterns was recorded using Bullock's (1992:Figure 2.3) system. Striations were defined as longitudinal (lo), transverse (tr), rotational (ro), crisscross (cc), and diagonal (di). In addition, "weaving grooves" (we) were also recorded when deep, polished grooves perpendicular to the long axis of the awl shaft were noted.

Notes were also kept concerning any modification to the bone artifact (for example, burning or rodent damage). Pieces which could be refitted were combined to produce a single artifact.

Bone Artifact Frequency by Site

Artifacts are listed by site in Table 15.40. There is a strong correlation between the total number of animal bone fragments and the number of bone artifacts (Figure 15.36). This suggests that there was little difference in the rate of bone tool production from one site to another.

For the purposes of intersite comparison, artifacts not included in the "other" category were grouped into three types: awls (awls A through I), tubes (tubes, tube beads, offcuts, offcuts/tubes, whistles), and miscellaneous (all other recognizable types). Most sites have too few artifacts to make meaningful comparisons. The three largest assemblages are from Castle Rock Pueblo, Kenzie Dawn Hamlet, and Stanton's Site (Figure 15.37). Bone tool assemblages at these three sites were dominated by awls (62 percent to 79 percent), followed by tubes (11 percent to 30 percent). Given the small number of specimens at most sites, there is little point in speculating on any variation in these figures.

Selection of Raw Material

Table 15.41 lists bone artifacts by taxa. Of 306 artifacts and artifact fragments, 193 (63 percent) were not identified to taxon. One large mammal and one medium mammal were identified. Of the remaining 111, 89 (80 percent) were identified as turkey or large bird. Of the larger mammals, artiodactyl bones were used for nine tools, and a bobcat ulna was used for one. Small mammal bones included five jackrabbit, four cottontail, two sciurids, and a prairie dog. Three of the jackrabbit bones were drilled tibiae, but apart from this, there was no obvious association of species with tool type, except for the turkey scapula tools. The high percentage of turkey and large bird is consistent with its frequency in the unmodified faunal assemblages (see Chapter 18) and with its large size relative to other common fauna, such as cottontail rabbits. At Sand Canyon Pueblo, where large mammals are more common, a correspondingly higher frequency of artifacts was made of large mammal bone (Walker 1990a). It would appear that most bone tools were produced expediently from locally available raw materials.

Only the turkey and large bird assemblage is large enough to warrant analysis of element frequency. Four elements--ulna (N = 16), radius (N = 10), tibiotarsus (N = 38), and tarsometatarsus (N = 19)--were used for 93 percent of the artifacts made from this taxon (Table 15.42). There is also clear selectivity of elements for artifact types. Seventy-seven percent of modified ulnae and radii were made into tube-type artifacts (tubes, tube beads, offcuts, and offcuts/tubes), and 77 percent of all turkey bone tubes were made of these elements. Seventy percent of modified tibiotarsus and tarsometatarsus were used for awls, and 91 percent of turkey bone awls were made from these elements. This probably reflects the fact that different bone morphologies are suited to different artifacts.

Awl Shape and Wear

The plan view and cross section of awl tips were described using criteria established by Bullock (1992). For the purposes of this analysis, I distinguish between complete awls (types A through F) and broken tips (type H). The complete specimens are tools which were either lost or discarded before breakage. The broken specimens are tools broken during use or after discard. Table 15.43 and Table 15.44 provide data on the frequency of plan view and cross section types. Table 15.45 and Table 15.46 provide data on the relationship between tip plan view and cross section.

For complete tools, rounded/convex triangular tip forms are the least common. There is no clear dominance of any cross section. For broken tips, triangular forms are by far the most common, and flat cross sections dominate the broken tips. This suggests that the combination of triangular plan and flat cross section was particularly susceptible to breakage. This combination occurs in 38 out of 87 awls considered here (44 percent). However, it occurs only seven times in the sample of 32 complete awls (22 percent) and 31 times in the sample of 55 broken tips (56 percent). If this differential breakage is the result of use, rather than of postdiscard breakage, it would suggest that flat awls were used in such a way as to put force on the flat surface, thus straining the weakest part of the artifact.

In order to see if there might be a functional difference between awls with this configuration, wear patterns were examined, again using criteria set out by Bullock (1992). As Table 15.47 shows, there is very little difference between the frequency of different wear patterns on the "trifl" awls and other awls. The most obvious difference is the lack of rotational striations on the "trifl" awls, which is not surprising in view of their flat cross sections. This suggests that awls had similar functions. In view of the fact that turkey was the most easily available raw material, it is possible that the use of this species predetermined the shape of the awl tip, even though such tips were more susceptible to breakage. Turkey long bones are relatively thin, and any awl produced from the tibiotarsus (the most commonly used bone) would probably have a fairly flat cross section.

Conclusions

Data in this chapter have been provided mainly for descriptive purposes, and there are relatively few conclusions which can be drawn from the information presented. The inventory of bone tools from site to site is similar in terms of artifact types and species used as raw material. Turkey bones were the most common raw material, reflecting the importance of turkey as evidenced in the faunal assemblages. The most common artifacts were awls, followed by tubular artifacts. The same pattern of artifact types has been observed at other sites in the locality, including Sand Canyon Pueblo (Walker 1990a) and the Green Lizard site (Bullock 1992). At Sand Canyon Pueblo, the higher frequency of larger mammals in the faunal assemblages is reflected in the more frequent use of those species for bone tools (Walker 1990a). It would appear that each hamlet or village relied on locally produced bone tools.

Certain elements were preferentially selected from the available unmodified bone. This is best seen in the assemblages reported here for turkey, where a few long bone elements predominate, and where different long bones were selected for different artifact types.

The most common artifacts are awls. Analysis of the morphology of the awl tips shows no clear association between awl type and tip type, suggesting that different types of awls do not correspond to different functional categories. For complete awls, the most common shape of the tip is triangular, and no single cross section predominates. However, broken awl tips were much more commonly triangular in plan and flat in cross section, suggesting that awls with this combination of tip attributes were broken most frequently. It is possible that the use of turkey bones (which have relatively thin cortical bone) predetermined the shape of the awl tip, because there seems to be little difference in wear patterns between triangular-flat tips and any of the other tips.

Shell
G. Timothy Gross

Methods

The tested sites yielded 22 shell items, including eight worked pieces. Crow Canyon Archaeological Center personnel cleaned and cataloged each piece of shell and submitted the collection for analysis. The analyst used procedures that were employed in the analysis of the shell from Crow Canyon's Duckfoot site excavations (Gross 1993), which were patterned after those developed at the Dolores Archaeological Program (DAP) (Neusius and Canaday 1985; Phagan and Hruby 1984). The taxonomic identification of each piece of shell was recorded to the finest level possible, along with the number of items in each provenience lot and the weight of those items. If any identifiable shell structures were present, these were noted, along with an estimate of the amount of the original shell present. The condition of the shell (fresh-looking, weathered, abraded), any indication of burning, and whether or not the shell was modified were also recorded for all items. Taxonomic identification was aided by reference to standard works (Keen 1963, 1971; Morris 1966; Pennak 1978; Pilsbry 1939, 1948) and by comparison with reference specimens in the analyst's collection or in the collection of the Department of Marine Invertebrates at the San Diego Museum of Natural History.

All of the pieces of shell in the collection were examined for evidence of modification. When necessary, a 10× hand lens or a variable-power binocular microscope was used for examination of the items. The shell artifact typology used in this analysis was that used for the analysis of the Duckfoot site collection (Gross 1993) and was adapted from Haury's (1937:137) description of shell from Snaketown. For modified shell items, the analyst recorded the length, width, and thickness of each piece (measured with sliding calipers), an estimate of the proportion of the finished shell item represented by the specimen, and observations on a series of technological variables. These variables included evidence of specialized manufacturing procedures (such as drilling or grooving), production stage evaluation (a ranked evaluation of the amount of effort that went into manufacture of the item), and the type of blank from which the artifact was manufactured (for example, whole shell, whole valve). The analytic data are available at the Anasazi Heritage Center in Dolores, Colorado.

Results

In general, the shell from the tested sites was in fairly good condition. Over half the shells (N = 12) looked fresh or only slightly weathered, with six that were weathered to the point of looking chalky, and one that had an abraded look with rounded edges, suggesting that it had been wave-rolled. This latter piece was recovered from Castle Rock Pueblo.

The taxonomic composition of the collection is presented in Table 15.48. Table 15.49 lists the taxa recovered at each of the sites and modification for the eight worked pieces. The level of taxonomic identification possible varied considerably because of breakage, weathering, and modification. Each taxonomic class identified in the collection is discussed briefly below and is correlated with artifact type.

Three kinds of marine gastropods are present in the collection, and all of the specimens have been modified. Olivella and Agaronia were made into whole-shell beads, and the Haliotis specimen is a ground pendant. The three examples of Olivella were recovered from three different sites--Mad Dog Tower, Castle Rock Pueblo, and Stanton's Site--and from each specimen the spire had been deliberately removed to create a hole through which a cord or string could be passed. This modification, along with wear and breakage in the aperture area, makes a positive identification difficult, but the beads appear to be Olivella dama. O. dama is available in the Gulf of California (especially at the head of the gulf), but it is not found along the Pacific coast of California (Keen 1971:628; Morris 1966:192).

The single specimen of Agaronia testacea, which resembles Olivella (and which is in the same family, the Olividae), was found at Stanton's Site. This species is found in the Gulf of California (Keen 1971:625; Morris 1966:191), but not along the Pacific coast. The Stanton's Site specimen had been formed into a bead by the removal of its spire.

The final marine gastropod is Haliotis sp., the abalone. Castle Rock Pueblo produced the only example of abalone shell--a ground pendant. Several species of Haliotis are found along the Pacific coast of North America (Morris 1966:52), but they are not found in the Gulf of California (Keen 1971). Abalone are edible and are a common midden constituent in some parts of California. The colorful, pearly shell of the abalone was often made into ornaments (Gifford 1947).

Marine pelecypods are represented by a single modified piece of Glycymeris found at Kenzie Dawn Hamlet. This item is not identifiable to species, but it is probably G. maculata or G. gigantea, species identified from Hohokam sites in Arizona (Haury 1976:307; Craig 1982). These species are found in the Gulf of California (Keen 1971:55). The Pacific coast Glycymeris (G. subobsoleta) (Morris 1966:6) is too small to serve as raw material for bracelet manufacture and is too small to have produced the worked piece found at Kenzie Dawn Hamlet.

Three items were identifiable as marine mollusks of some type, based on their structure and thickness, but could be identified no further. One of these pieces was a fragment of worked shell from Stanton's Site that, based on the presence of a drilled hole, appears to have been some kind of pendant. Some characteristics of the shell suggest that it is Glycymeris sp., but that cannot be determined with any certainty. The other pieces of shell identifiable only to marine mollusk preserved no evidence of having been worked, although they may well have been parts of broken ornaments.

The shell collection includes two different types of nonmarine gastropods--one that appears to be Succinea sp. and one that is probably Oreohelix strigosa depressa. Collectors have recovered both these taxa from their natural habitats in southwestern Colorado.

The genus Succinea includes a number of widely distributed species that occur in a variety of habitats Pilsbry 1948). Drake (1959:150) says that members of the family Succineidae in the Southwest are "almost amphibious and are usually considered so because they have generally been found alive in the damp, wet zone on the substratum at the edge of bodies of water." Other sources indicate that a wider range of conditions is suitable for members of the genus Succinea, including drier areas (Baker 1939:121-127; Pilsbry 1948:839).

The type locality for Oreohelix strigosa depressa is a canyon near Durango, Colorado (Pilsbry 1939:431). Oreohelix occurs in talus or duff in areas that are rich in lime, where they are found near the surface, concealed by rock, bark, or leaves (Pilsbry 1939:415). Although O. strigosa can withstand a wide range of environmental conditions, Brandauer (1988:Figure 13) reports no specimens having been collected from Montezuma County.

Concentrations of what appear to be beads made from Oreohelix shells have been recovered from the Duckfoot site (Gross 1993) and from Mug House (Rohn 1971:128, Figure 161), a site in Mesa Verde National Park. None of the specimens from the tested sites were complete enough to indicate whether or not they may have been beads.

Two specimens are identifiable only as shell, but whether they are from marine, fresh water, or terrestrial species cannot be determined. One of these, recovered from Troy's Tower, is a fragment that preserves some evidence of having been shaped, but it is too fragmentary to allow a determination of what type of item it may have been. The other piece, found at Lillian's Site, preserved no evidence of having been worked.

Finally, one piece of fossil shell was found at Mad Dog Tower. It is probably a piece of fossil oyster, but the identification is not certain. The piece was not modified.

Summary and Discussion

Shell was not abundant at any of the sites in the sample. This fits the pattern for sites in the Mesa Verde region, especially when those sites are compared with sites in Chaco Canyon (Jernigan 1978:153; Morris 1939) or with Chacoan outliers such as the Dominguez Ruin (5MT2148) (Nickens and Hull 1982:211; Reed 1979:53). For example, excavations at the sites that make up the Badger House community at Mesa Verde National Park produced only two shell beads (Hayes and Lancaster 1975:73).

The collection of shell from the tested sites is consistent with that found at the Duckfoot site and is also within the range of types found at sites investigated by the DAP. At the Duckfoot site, 36 pieces of shell were recovered, 18 of which were worked. Both Olivella dama and Glycymeris sp. were found at Duckfoot, as were nonmarine gastropods that appeared to be Succinea sp. and Oreohelix strigosa depressa; however, Haliotis sp. and Agaronia testacea were absent from the Duckfoot collection. The types of modified items found at the tested sites are similar to those found at the Duckfoot site (Gross 1993).

The DAP collections have been summarized by Blinman (1986a) and Gross (1993). The marine taxa (again, with the exception of Agaronia testacea) and the artifact types found at the tested sites are also found at DAP sites. Three DAP sites--Tres Bobos Hamlet (5MT4545), Grass Mesa Village (5MT23), and McPhee Pueblo (5MT4475)--yielded considerably more shell than did any of the tested sites, but many of the other DAP sites had shell in amounts similar to those at the tested sites. Differences between the sites discussed here and the DAP sites, as well as the Duckfoot site, may be due to differences in intensity of excavation or intensity of occupation.

Although the sample size is small, all but one of the pieces of marine shell from the tested sites would have been obtained from the Gulf of California. Only the Haliotis pendant would have come from the Pacific Ocean. This trend is seen throughout much of the Southwest. Although sites in the Mesa Verde region have yielded specimens of Oreohelix strigosa depressa that may have been made into beads, the nonmarine shell in the tested-site sample is probably in site deposits as a result of natural, rather than cultural, processes.

1. The abbreviations PD and FS are part of a two- or three-part code which identifies the location of an artifact at a site. The first number is a provenience indicator (PD), the second number is a laboratory catalog number (FS), and the third, when assigned, is a point-location number (PL).

The Sand Canyon Archaeological Project: Site Testing

The Sand Canyon Archaeological Project

Site Testing

Edited by Mark D. Varien

15

Artifacts