The Archaeology of Castle Rock Pueblo

Chapter Contents

Introduction

¹
This report synthesizes information on portable artifacts collected during excavations at Castle Rock Pueblo. Several analyses of Castle Rock artifacts and comparisons with other Pueblo III Mesa Verde–tradition sites in southwestern Colorado are presented. Artifacts from the test excavations conducted as part of the Sand Canyon Project Site Testing Program (Kleidon 1999*1; Pierce et al. 1999*1; Varien 1999*2) and from the intensive excavations of 1992–1994 are considered together.

²
Many of the tables and figures presented in this report were produced using the artifact databases as they existed in June 1998. Since that time a few minor provenience changes have been made, so there may be slight discrepancies between the data discussed here and those contained in the current database. It is unlikely that these changes affect any of the conclusions presented in this report on the basis of the June 1998 data.

Processing of Artifacts in the Laboratory

³
All objects collected during the excavations at Castle Rock Pueblo were processed according to Crow Canyon's standard laboratory procedures, which are described in the on-line laboratory manual.

Definitions of Analytic Categories

⁴
All objects were classified into various stone, bone, pottery, vegetal, and other categories, as defined in the on-line laboratory manual.

Disposition of Materials
Curation

⁵
All artifacts, ecofacts, and other samples collected from Castle Rock Pueblo, with the exception of wood samples submitted for tree-ring dating, are currently curated at the Anasazi Heritage Center, 27501 Hwy. 184, Dolores, Colorado, USA. The collections are indexed to the artifact databases accessible through this report, and all curated objects are available for future study through the Heritage Center. Tree-ring samples that produced dates, along with samples that might potentially be datable in the future, are curated at the Laboratory of Tree-Ring Research, University of Arizona, Tucson, Arizona, USA.

Repatriation

⁶
As of this writing, human remains and associated funerary objects collected during excavations at Castle Rock Pueblo are in the process of being repatriated according to the Native American Graves Protection and Repatriation Act (NAGPRA). The Anasazi Heritage Center is curating these items during the repatriation process. Objects falling under the jurisdiction of NAGPRA are not currently available for study, and their future disposition has not yet been decided.

Destructive Analysis

⁷
A few artifacts have been subjected to destructive analysis, including small portions of certain white ware bowl rim sherds that were included in Glowacki's studies of pottery production and exchange using instrumental neutron activation analysis (Glowacki 1995*1; Glowacki et al. 1995*1, 1997*1, 1998*1). These sherds are identified in the "comments" field of the pottery data tables. Small portions of numerous sherds were also removed to facilitate temper identifications. Small portions of selected human remains from Castle Rock and other Sand Canyon locality sites were used for stable isotope analysis (Katzenberg 1999*1). Samples submitted for tree-ring dating that possessed little dating potential have been discarded by the Laboratory of Tree-Ring Research.

Additional Studies of Castle Rock Pueblo Artifacts

⁸
In addition to the analyses reported here, numerous other studies of artifacts from Castle Rock Pueblo have been conducted or are in progress. Pierce et al. (1999*1) reported on artifacts collected during the Site Testing Program portion of the excavations at Castle Rock Pueblo and at 12 other sites in the Sand Canyon locality (Varien 1999*2). Studies of white ware pottery designs that used sherds from Castle Rock include those by Hegmon (1991*1) and Ortman (1998*1, 2000*1). Variation in the manufacturing techniques used in the production of corrugated gray ware jars was examined by Dobshuetz (1999*1). Corrugated gray wares also figured prominently in Varien's (1997*1, 1999*1) study of sedentism and mobility in the Sand Canyon locality. Glowacki and others included sherds from Castle Rock in their studies of local pottery production and exchange (Glowacki 1995*1; Glowacki et al. 1995*1, 1997*1, 1998*1; Thurs et al. 1996*1). Fratt (1997*1) analyzed manos from Castle Rock and several other sites in southwestern Colorado. Driver (1996*1, 1997*1) and Munro (1994*1) studied faunal remains from Castle Rock and other sites in the Sand Canyon locality.

Organization and Use of This Report

⁹
This report is organized into sections and subsections, a list of which can be accessed by clicking on the expanded table of contents at the top of the chapter. Clicking on a selection in the table of contents will allow you to go directly to the section you are interested in without having to scroll through the entire chapter. When you click on a link to a table, figure, or reference in the text, a new browser window will open in which the selected information will be displayed. You can move back and forth between the chapter text and the data window by keeping both windows open, overlapping them (i.e., not viewing them full screen), and selecting one or the other window. The data window will be updated each time a link for a table, figure, or reference is selected in the narrative text window; the text window will maintain your place in the longer document. Database Maps open in the same window as this report, and you will need to use your back button to return to this chapter. In many subsections, contextual information taken from the field context database is given along with analysis information for selected artifacts. Explanations of field context information can be found in the on-line field manual.

Pottery
Unmodified Sherds

¹⁰
More than 40,000 pottery sherds, weighing more than 250 kg, were collected during excavations at Castle Rock Pueblo. All of these were analyzed according to Crow Canyon's standard analysis procedures, which are described in the on-line laboratory manual. The following pages present several summaries of the basic sherd data and discuss several issues related to the comparison of sherd data from Castle Rock with data from other sites.

Total Inventory
By Ware and Type

¹¹
The sherds collected from Castle Rock are tabulated in Table 1 according to pottery type (for type definitions, see the on-line laboratory manual). The list of pottery types is arranged according to the more general ware to which each type belongs. Unknown white, gray, and red ware sherds are listed separately because such sherds may or may not represent local wares. Results are given by count and by weight in grams, and the percentage of each pottery type in the collection, by count and weight, is also given. Pierce and Varien (1999*1) discuss the relative merits of counts vs. weights as measures of abundance.

¹²
The table shows that percentages of various pottery types can vary depending on whether counts or weights are used. This effect is especially clear for Mesa Verde Black-on-white, which is much more abundant by weight than by count. In contrast, the relative abundance of Pueblo III White Painted, a more general type used for sherds that do not exhibit diagnostic attributes of either McElmo or Mesa Verde black-on-white, is approximately equal by count and weight. Consistency in the relative frequency of a type for both count and weight probably indicates that sherds assigned to that type tended to be of average size for the collection overall. Greater relative frequency by count indicates that sherds assigned to that type were smaller than average, and greater frequency by weight indicates that sherds assigned to that type were larger than average. That sherds identified as Mesa Verde Black-on-white tend to be larger than average is expected, since the classification of local white ware sherds to traditional types relies heavily on the identification of specific painted designs, which are often difficult to identify on small sherds.

Dating the Occupation of Castle Rock Pueblo Using Assemblage Type Data

¹³
Wilson and Blinman (1999*1) defined pottery assemblage profiles that characterize Mesa Verde-region ancestral Puebloan sites dating to various time periods between A.D. 575 and 1300. The pottery from Castle Rock Pueblo suggests that the ancestral Puebloan occupation at this site falls squarely into their A.D. 1225–1300 period. During this period, Mesa Verde Black-on-white was the dominant decorated white ware type, with only trace amounts of McElmo Black-on-white and Mancos Black-on-white. Mesa Verde Corrugated was the dominant gray ware type, and red wares were scarce. The pottery assemblage from Castle Rock follows this idealized assemblage exactly, regardless of the measure of abundance used, and it corresponds with the period of occupation suggested by tree-ring and architectural data. Thus, the pottery assemblage from Castle Rock is indicative of a single occupation dating between A.D. 1225 and 1300. There is no pottery evidence of an earlier occupation.

By Ware and Form

¹⁴
All sherds collected from Castle Rock were assigned to one of four basic form categories: bowl, jar, other, and unknown. Total counts, weights in grams, and percentages by count and weight for these four form categories are presented in Table 2 by ware category. Unlike the type percentages, the percentages of various ware-form combinations are fairly consistent for both counts and weights. This pattern suggests that sherd size does not significantly affect the ability of analysts to assign sherds to wares and forms. Consistency in the percentages of ware-form combinations by count and weight is most likely due to relatively consistent sherd sizes across wares and forms in the Castle Rock sherd assemblage. As a result, intrasite comparisons can be made using either counts or weights. Notice, however, that weight is the superior measure for comparing the relative frequencies of wares and forms across sites, because sherd sizes can vary systematically across depositional contexts, owing to a number of depositional and postdepositional processes.

¹⁵
These ware-form combinations are also found in roughly the same proportions in other Pueblo III sites in the Sand Canyon locality that have been interpreted as permanent, year-round habitations (Pierce and Varien 1999*1). This suggests that the ware-form characteristics of the Castle Rock sherd assemblage resulted from a set of domestic activities that produced sherds of various wares and forms at a relatively consistent rate across sites. This inference is supported by the fact that nonhabitation sites do not possess the same proportions of these ware-form categories in their sherd assemblages. For example, the sherd assemblage from Site 5MT12086, a reservoir in Woods Canyon approximately 20 km from Castle Rock, is dominated by white ware jars and contains few corrugated jars or white ware bowls (Wilshusen et al. 1997*1:Table 1). Obviously, the activities that occurred at the Woods Canyon reservoir led to different patterns of sherd deposition than are typical of habitation sites, including Castle Rock Pueblo.

By Type and Finish

¹⁶
Two kinds of paint are identifiable on decorated Mesa Verde White Ware. Mineral paint derives from ground iron, manganese, or copper-rich rock that is held in liquid suspension. Carbon paint is believed to derive from the condensed extract of certain plants, such as Rocky Mountain beeweed (Cleome serrulata) and tansy mustard (Descurainia richardsonii). In the Sand Canyon locality, mineral paint is most common in sherd collections dating to before A.D. 1150, whereas carbon paint dominates in later periods. Mineral-painted white ware, however, continues to be common in thirteenth-century sites located northwest of the Sand Canyon locality, in the bean-field and canyon country along the Utah-Colorado border west of Pleasant View, Colorado (Wilson 1991*1).

¹⁷
Table 3 presents counts and weights of white ware sherds assigned to various type and finish categories. The row percentages show the relative abundance of carbon and mineral paint within each pottery type, and the column percentages show the relative abundance of each white ware type among all white wares, regardless of paint type. The table shows that very few mineral-painted sherds were collected from Castle Rock Pueblo. Only five of the 38 mineral-painted sherds in the Castle Rock assemblage were classified as Mesa Verde Black-on-white, a definite late Pueblo III type. Whether the vessels that produced these sherds were imported from areas where potters more commonly used mineral paints, or whether a few Castle Rock potters used mineral paint, is unknown. The remaining mineral-painted sherds were all classified as potentially earlier pottery types (Pueblo III White Painted and Late White Painted) or definitely earlier ones (McElmo Black-on-white, Mancos Black-on-white, Pueblo II White Painted, and Early White Painted). Although their presence could be explained by one of the scenarios described above, it is also possible that these sherds are from "heirloom vessels" that lasted an especially long time before breaking.

Rim Sherds
By Ware and Type

¹⁸
Rim sherds may provide a better indication of type frequencies among the vessels used during an occupation because rim sherds usually preserve more diagnostic attributes of pottery types than do body sherds, and therefore they tend to be classified more precisely than body sherds. Table 4 presents counts and weights of rim sherds in the Castle Rock sherd assemblage by ware and type. The relative frequency of rim sherds assigned to each type is given as a percentage of all rim sherds by count and weight. The relative frequency of specific, named types is clearly much higher among the rim sherds alone than in the entire sherd assemblage, but the basic typological assemblage profile, with Mesa Verde Corrugated the dominant gray ware type, Mesa Verde Black-on-white the dominant white ware type, and red wares quite rare, is still apparent.

¹⁹
As was the case for the overall sherd assemblage, significant differences in the relative frequencies of different types by count and weight probably relate to the average sizes of rim sherds assigned to each type. As examples, Mesa Verde Black-on-white and Mesa Verde Corrugated are nearly twice as common by weight as by count, whereas Pueblo III White Painted and Indeterminate Local Corrugated Gray are nearly twice as common by count as by weight. These patterns indicate that rim sherds assigned to specific traditional types tend to be larger than average, whereas rim sherds assigned to generic types tend to be smaller than average. Nevertheless, the higher frequencies of specific types among the rim sherds indicates that rims were assigned to these specific types more often than body sherds were.

By Ware and Form

²⁰
Rim sherds can often be assigned to more specific form classes than can body sherds, and when it was apparent during analysis that a rim sherd came from a ladle, canteen, mug, or kiva/seed jar, this was recorded in a "comments" field. Ladle rims curve more tightly than bowl rims and possess distinctive use wear on the outside edge of the rim or evidence of a handle attachment. Canteen rims are small jar rims with very tight curvature. Mug rims are square and upright, are seldom everted, usually possess intricate painted decorations on their exteriors, and sometimes preserve evidence of a handle attachment near the rim. Kiva and seed jars are slightly larger than canteens, do not have necks, and, in the case of kiva jars, have a distinctive lip that is designed to hold a lid in place.

²¹
Table 5 summarizes the wares and forms of rim sherds in the Castle Rock assemblage by count and weight. The more specific vessel forms of kiva jar, seed jar, ladle, and mug are split out in this table on the basis of information recorded in the comments field of the pottery data file. It is assumed in this table that white ware jar rims for which no additional comments were recorded in the file are from large storage jars, or ollas. As was the case for the entire sherd assemblage, the three most common vessel forms represented among the rim sherds are corrugated jars, white ware jars, and white ware bowls. The relative frequencies of these three forms, however, are strikingly different when rim sherds alone are considered. White ware bowls are by far the most common ware-form combination among rim sherds only, whereas corrugated jars are by far the most common among all sherds.

²²
These differences relate to the typical circumferences of rims in the original vessels of these various ware-form combinations and to differences in the relative numbers of rim and body sherds produced by vessels of different sizes. White ware bowls are open forms with large rim circumferences; when they break, they produce numerous rim sherds and a relatively high ratio of rim to body sherds. Corrugated and white ware jars are closed forms with small rim circumferences that produce far fewer rim sherds per vessel than do white ware bowls. As a result, the best way to estimate the relative number of vessels of different ware-form classes in a pottery assemblage is to compare the total degrees of arc subtended by the rim sherds of various ware-form classes. Such data were considered by Pierce and Varien (1999*1) in their study of the Sand Canyon locality Site Testing Program assemblages, including the testing sample from Castle Rock. They found that raw counts of rim sherds, though less precise than degree-of-arc measurements, nevertheless gave a closer approximation of the relative numbers of vessel ware-form classes than did raw counts of all sherds. Judging from this finding, it appears that white ware bowls were the most common vessel form used at Castle Rock, followed by corrugated jars and then white ware jars and white ware ladles. Canteens, mugs, and kiva/seed jars were all relatively rare.

²³
As is the case for the overall assemblage, the rim sherds show relatively little variation between the percentages based on counts and those based on weights when classified in terms of ware-form combinations. This suggests that sherd size does not significantly affect the ability of analysts to assign rim sherds to wares and forms.

Counts by White Ware Type and Form

²⁴
Table 6 summarizes the forms of rim sherds assigned to various white ware types. Counts alone are considered in the upper register of this table, and the lower register gives relative frequencies across types for each form class. The total and relative frequency by count of each type for the overall white ware rim sherd assemblage is given in the far right column of the table. The table illustrates that significant variation exists in type frequencies across form classes. As examples, Late White Unpainted sherds are much more common among jar rims than among bowl and ladle rims, and McElmo Black-on-white, a type common in pottery assemblages dating between A.D. 1140 and 1225 (Wilson and Blinman 1999*1), is much more common among ladle rims than among bowl rims.

²⁵
The elevated frequency of Late White Unpainted sherds among jar rims probably relates to differences in the decorative treatment of jars vs. bowls and ladles. The primary decorative field of white ware jars was the upper half of the jar body, and the necks and rims of many such jars were left unpainted. Although some white ware bowls and ladles, too, were left undecorated, in most cases the rims and interiors of such vessels were intricately painted, leading to a lower frequency of unpainted bowl and ladle rim sherds.

²⁶
One possible explanation for the elevated frequency of McElmo Black-on-white among ladle rims derives from ethnographic studies which indicate that vessels of different forms and uses have varying use lives (Varien and Mills 1997*1). If the vessel forms used by the inhabitants of Castle Rock also had varying use lives, then vessel forms with longer use lives might have tended to be older by the time they were broken and discarded than were vessel forms with shorter use lives, and this might be reflected in a higher percentage of earlier types among rim sherds of longer-lived vessel forms. Since the cup of a ladle is much smaller and sturdier than that of a bowl, it is likely that ladles had longer use lives than bowls. If this was the case, then a higher frequency of McElmo Black-on-white ladle rims might relate to differences in the use lives of these two form classes.

²⁷
A second possible explanation relates to the details of pottery type definitions. McElmo Black-on-white designs are usually seen as being simpler than Mesa Verde Black-on-white designs, and since ladle interiors are much smaller than bowl interiors, it might have been difficult to execute a more complex Mesa Verde Black-on-white design on a ladle interior. It is therefore possible that some ladle rims were classified into an earlier style, not because ladles lasted longer before breaking, but because their designs were necessarily simpler than those of larger bowls.

Weights by White Ware Type and Form

²⁸
Table 7 presents the total weight of white ware rim sherds assigned to various types and forms. Percentages by weight of white ware types are given by form, as are the mean weights of rim sherds assigned to various type-form combinations. The total and relative frequency by weight of each type within the white ware rim sherd assemblage is given in the far right column of the table.

²⁹
The first two registers present the sum of weights and the column percentages by weight for rim sherds assigned to each form category. They illustrate some of the same patterns discussed for the type-form data by count. There is an elevated frequency of Late White Unpainted sherds among the jar rims and an elevated frequency of McElmo Black-on-white among the ladle rims. And as is the case for the overall sherd assemblage, there is significant variation in type frequencies depending on whether counts or weights are used (compare this table with Table 6). It is likely that the effects of sherd size on the typing of rim sherds are responsible for these differences, as was discussed for the entire sherd assemblage. The third register presents the mean weights of white ware rim sherds assigned to various type-form combinations.

Size Distributions for White Ware Bowl Rim Sherds

³⁰
Table 8 summarizes the number of white ware bowl rim sherds assigned to different types, along with means and standard deviations of their weights in grams. These data support the notion that sherd size significantly affects the ability of analysts to assign sherds to specific named types. It has already been shown that Mesa Verde Black-on-white sherds are consistently more abundant by weight than by count. This table shows that for white ware bowls, rim sherds classified as Mesa Verde Black-on-white are larger on average than rim sherds assigned to more generic types, such as Pueblo III White Painted and Late White Painted. These data indicate that larger sherds are generally easier to type than smaller ones.

³¹
It is also interesting that rim sherds assigned to the earlier McElmo Black-on-white and Mancos Black-on-white types are smaller on average than rim sherds classified as Mesa Verde Black-on-white. It is possible that rim sherds assigned to earlier types are smaller because they really are older, were deposited earlier in the occupation of Castle Rock, and had more opportunity to be fragmented before being safely buried. Alternatively, this relationship might represent an analytical bias in which smaller sherds had a greater likelihood of being classified as earlier types than did larger sherds. It is also possible that earlier types are easier to identify from small sherds than are later types.

By Form and Finish

³²
Table 9 presents total counts and weights of white ware rim sherds assigned to various form-finish combinations. Percentages of paint types within each form class are given in the second register, along with the percentages of forms across all decorated white ware rim sherds in the assemblage. As is apparent in the entire sherd assemblage, mineral paint is rare among the decorated rim sherds at Castle Rock.

Summary of Pottery Sherd Analysis

³³
The pottery sherd assemblage from Castle Rock Pueblo mirrors the tree-ring data in arguing that Castle Rock was inhabited between A.D. 1225 and 1300. There is no evidence in the pottery assemblage of an earlier occupation. The decorated sherds in the assemblage are dominated by carbon paint, to the near exclusion of mineral paint. The most common vessels used by the inhabitants of Castle Rock were white ware bowls, followed by corrugated jars, white ware jars, and white ware ladles. Canteens, mugs, and kiva/seed jars were all relatively rare. The wares and forms represented in the sherd assemblage occur in roughly the same proportions as in other Pueblo III habitation sites in the Sand Canyon locality.

³⁴
The summaries of Castle Rock pottery sherd data presented in paragraphs 10–32 have raised several important issues regarding the effects of ancient activities, of depositional and postdepositional processes, and of analytical biases on pottery sherd data. These issues have implications for comparisons of sherd data across sites. Sherd size does not significantly affect pottery ware or form data, making weight a suitable measure of abundance for ware-form classes across assemblages with different inherent sherd size distributions. Sherd size, however, does exert a significant effect on pottery type data, and this is a cause for concern. Because larger white ware sherds are more easily and accurately assigned to specific traditional types, it may be necessary to restrict comparisons of assemblages from different depositional environments to sherds that are larger than a certain minimal size. The 1/4-inch screens used in the field to collect the Castle Rock assemblage appear to have been too fine to control for sherd size effects in typing.

³⁵
Type frequency data can also vary widely owing to variation in the inherent mix of activities that occurred at a site. Different vessel forms were decorated differently, and as a result, an assemblage dominated by white ware jars will produce many more Late White Unpainted sherds than an assemblage dominated by white ware bowls, for example. Therefore, pottery type data for individual form classes should be considered separately. Comparing type data for each form class separately also controls for the possibility of varying use lives for different vessel forms, which could also have an effect on type frequencies.

Modified and Shaped Sherds
Inventory by Type

³⁶
A number of sherds that had been modified or shaped after their parent vessels broke were collected during the Castle Rock excavations. Table 10 summarizes the types to which such sherds were assigned, by count and by weight, along with relative frequencies of different types by count and weight. Modified sherds possess at least one abraded edge. Shaped sherds have edges that were flaked, ground, or both to make a specific shape. Some larger shaped sherds may be pottery fragments that were used as containers or as pottery molding trays called pukis. Perforated sherds with shaped edges were classified as sherd pendants and are discussed in paragraph 132. Sherds with shaped edges but lacking a perforation, such as disks, triangles, and rectangles, were classified as shaped sherds and are included here. These shaped sherds may have been pendant blanks, gaming pieces, or other nonutilitarian items.

³⁷
The table shows that relative to the overall sherd assemblage, modified corrugated sherds are underrepresented but tend to be larger than the other modified or shaped sherds. Corrugated sherds are not well suited for use as pottery scrapers because they have rough surfaces, coarse paste, and large temper inclusions that make it difficult to create a smooth scraping surface. Several complete examples of corrugated sherd containers, however, have been found in excavations at Sand Canyon Pueblo. White ware and red ware sherds are overrepresented relative to the entire sherd assemblage. Red ware is also overrepresented among the sherd pendants, suggesting that the modified red ware sherds were pendant blanks and that the inhabitants of Castle Rock preferred to make pendants using sherds from rare or unusual pottery vessels. Most modified and shaped sherds are of white ware and probably represent portions of pottery scrapers, gaming pieces, or pendant blanks.

Pottery Vessels
Inventory by Form and Ware

³⁸
Nineteen whole, partial, or reconstructible vessels were collected from various contexts at Castle Rock. Table 11 summarizes the forms and wares of these vessels. Most of the collected vessels were local white wares, and no nonlocal vessels were found.

Vessel Analysis Data

³⁹
The type, form, condition, and metric data for each collected vessel are listed in Table 12. If the vessel was reconstructed, you can click on the vessel's photo number to see a photograph of it. In this table, multiple entries for a given vessel number indicate that sherds belonging to that vessel were recovered from multiple proveniences.

Vessel Provenience Data

⁴⁰
Type, form, condition, and context are listed in Table 13 for each vessel. If the vessel was reconstructed, you can click on the vessel's photo number to see a photograph of it.

Functional Analysis

⁴¹
This section compares data on the sizes of pottery vessels used for preparing and serving food at Castle Rock Pueblo, Sand Canyon Pueblo, and several small hamlets in the Sand Canyon locality. These comparisons are made to explore possible changes in foodways and food consumption groups as villages formed during the thirteenth century A.D. (Adler 1990*1, 1992*3, 1994*1, 1996*3)—that is, to examine whether different kinds of meals were being made and served to different social groups in hamlets and villages of varying sizes during the final period of Puebloan occupation. Such differences might be expected for several reasons. First, it is possible that village formation led to the elaboration of community rituals (see Adams 1989*1), and because such rituals are often associated with meals in the historic pueblos, one might expect ritual elaboration in villages to have affected food preparation and vessel assemblages in these larger settlements. Second, it is possible that larger villages, with their sizable populations, supported more ritual specialists than did smaller villages. If this was the case, one might expect more public ceremonies and more associated communal meals to have occurred in larger villages than in smaller ones. Third, it is possible that the size of the typical food consumption group changed along with village formation, as more relatives and other social groups came to live closer together in a single settlement.

⁴²
Sand Canyon Pueblo, a very large village approximately 10 km north of Castle Rock (Bradley 1993*1, 1996*1), was the largest thirteenth-century village in the Mesa Verde region (Adler and Johnson 1996*1). It housed a population at least six times larger than that of Castle Rock and was probably a more significant place in the regional social landscape than Castle Rock. One might therefore expect more communal rituals and associated meals to have taken place at Sand Canyon Pueblo than at Castle Rock. Also, since open-air plazas were created in both Sand Canyon and Castle Rock pueblos, and since plazas have not been found in earlier sites or in contemporaneous hamlets, one might expect that more communal rituals and associated meals occurred in these two villages than in earlier or contemporaneous hamlets in the Sand Canyon locality.

Vessel Size Classes

⁴³
There is a long tradition of research that attempts to relate the relative sizes of different vessel forms to variation in the size and composition of food consumption groups (Blinman 1988*1, 1989*1; Mills 1999*1; Nelson 1981*1, 1985*1; Turner and Lofgren 1966*1). The relative sizes of vessels of various forms, the distributions of vessel sizes within a form class, and the volumes of material that the vessels could hold provide important clues to the sizes of meals that were produced and how meals were eaten. Although the number of complete, partial, and reconstructible vessels collected from Castle Rock was insufficient for such an analysis, the very large collection from Sand Canyon Pueblo can be used to define size classes for vessels of various forms. If there is a reasonably strong relationship between vessel volume and rim diameter for vessel form-size classes in the Sand Canyon Pueblo collection, then rim-arc data from Castle Rock would be a reasonable surrogate for examining the sizes of vessels used in this site.

Sand Canyon Pueblo Vessels

⁴⁴
Table 14 presents means and standard deviations for the total volumes, estimated total weights in grams, and rim diameters of form-size classes in the Sand Canyon Pueblo vessel assemblage. Total volume was measured by completely filling each vessel with small birdseed. For vessels that were incomplete, the total weight of each vessel was estimated by dividing the weight of the portion present by the estimated portion present, expressed as a percentage. Size classes within each vessel form class were defined on the basis of k-means cluster analysis of metric variables recorded for that form (Kintigh and Ammerman 1982*1). The most efficient cluster solution—the one that produced the most significant drop in the total distance of individual vessel data points from cluster centroids relative to a solution of one more or fewer clusters—was taken as an indication of natural disjunctions in the distribution of size classes for a particular vessel form. The k-means analysis suggested two size classes for bowls and three size classes each for corrugated jars and white ware jars. A single size class apiece was suggested for all other vessel forms. The vessel size classes identified in this analysis mirror those identified in previous analyses of vessels from thirteenth-century Mesa Verde-region sites (Mills 1989*1:Chapter 5; Rohn 1971*1:Chapter 8).

Maximum Diameter vs. Volume for White Ware Bowls

⁴⁵
Figure 1 illustrates the relationship between maximum diameter and total volume for white ware bowls in the Sand Canyon Pueblo vessel assemblage. The relationship is best approximated by the shown quadratic regression line. Two size classes are clearly distinguishable on the basis of these two variables, with a gap in between. Random variation in bowl size around a single mean could not produce such a distribution. Also, direct evidence of pottery manufacture from Castle Rock (paragraphs 68–75) and the Sand Canyon locality tested sites (Pierce and Varien 1999*1:Table 15.17; Thurs et al. 1996*1) suggests that white ware bowls with this bimodal size distribution were made by a large number of nonspecialist potters, strengthening the case that small and large bowls were conceptual and functional categories for the inhabitants of Sand Canyon Pueblo. Finally, the fact that the fit line has a moderate slope suggests that the volume of a white ware bowl can be reasonably estimated on the basis of its maximum diameter, which in most cases occurs at the rim. This suggests that diameter estimates drawn from rim sherds should be sufficient for identifying large and small white ware bowls.

Rim Diameter vs. Volume for Corrugated Jars

⁴⁶
Figure 2 illustrates the relationship between rim diameter and total volume for corrugated jars in the Sand Canyon Pueblo vessel assemblage. Although the reason for it is unclear, rim diameter appears to produce a better correlation with volume than does orifice diameter for corrugated jars. The relationship between these variables is best approximated by a quadratic fit line, but no clustering of vessels into distinct groups is evident in this chart, as is apparent for white ware bowls. Although the k-means analysis suggests that breaking up this distribution into three groups efficiently captures the inherent variation, it is less likely that the inhabitants of Sand Canyon Pueblo thought of corrugated jars as having distinct size classes. Corrugated jar size may have related more to the domestic needs of households than to social or cultural convention. Also, since the slope of the fit line is generally shallower for corrugated jars than for white ware bowls, the volume of a corrugated jar cannot be estimated very precisely on the basis of its rim diameter. The relationship seems positive enough, however, to suggest that, in general, corrugated jars with larger rim diameters tended to have larger volumes. Thus, rim diameter estimates drawn from rim sherds should generally reflect the sizes of corrugated jars in the assemblage.

Volumetric Relationships among Vessel Size Classes

⁴⁷
Table 15 uses data from the Sand Canyon Pueblo vessel assemblage to examine volumetric relationships among size-form classes of vessels that are generally believed to have been used in the preparation and serving of food. Food was cooked inside corrugated jars, was scooped out of the larger jars using ladles, and was emptied into bowls for serving. Mean orifice diameters in millimeters, mean volumes in milliliters, and mean volumes of the larger size-form classes expressed in units of the mean volumes of ladles and small bowls are presented for these various vessel size-form classes. A mean volume in ladles is not given for small corrugated jars, because their mean orifice diameter was too small to accommodate an average-sized ladle.

⁴⁸
This table illustrates how many ladle scoops of food could be contained within corrugated jars of various sizes, how many scoops it would take to fill up large and small bowls, how many small bowls' worth of food were contained in a large bowl, and so forth. If a large bowl was sized to contain a meal for an entire family, and if one large bowl contained enough food to fill five small bowls, then it seems possible that small bowls were used for individual servings. If this was the case, then large and small bowls would have contained the same foods, but the number of people eating out of large and small bowls would have varied. Alternatively, small bowls might have contained side dishes, in which case the same number of people would have been served, but different foods would have been eaten out of large and small bowls. These possibilities are discussed further in paragraphs 62–66.

Rim-Arc Data
White Ware Bowls, White Ware Ladles, and Corrugated Jars, Castle Rock Pueblo

⁴⁹
Figure 3 summarizes rim-radius estimates drawn from rim sherds of white ware bowls, white ware ladles, and corrugated jars in the Castle Rock rim sherd assemblage. Pierce and Varien (1999*1) have discussed the methods used to collect these data, along with several possible sources of analytical bias in rim-arc analysis. Several procedures have been used to help control for these biases. First, comparisons of rim-arc diameter estimates with vessel diameter measurements using sherds from reconstructed vessels suggest that rim-arc radius estimates are within 2 cm of the true radius of the parent vessel approximately 80 percent of the time for sherds that encompass at least 20 degrees of arc. Thus, only sherds encompassing 20 degrees of arc or more are considered in the chart, and the radius estimates for these sherds have been grouped into 2-cm-radius classes. Second, the total degrees of arc assigned to each radius class has been used as the measure of abundance, rather than the count or weight of sherds assigned to each radius class. This was done to compensate for the tendency for smaller-diameter vessels to break into fewer rim sherds encompassing more degrees of arc relative to larger-diameter vessels.

⁵⁰
In the Sand Canyon Pueblo vessel assemblage, ladles have a mean rim radius of 5.7 cm, small bowls a mean rim radius of 7.95 cm, and large bowls a mean rim radius of 14.2 cm. These means are precisely mirrored in the rim-arc data from Castle Rock Pueblo, indicating that, in general, the inhabitants of Castle Rock used the same kinds and sizes of bowls and ladles as the inhabitants of Sand Canyon Pueblo.

⁵¹
The rim-radius distribution for corrugated jars from Castle Rock does not closely mirror the size classes derived from k-means analysis of Sand Canyon Pueblo vessel data, which suggests that corrugated jars in these two villages might have had different size distributions.

⁵²
It is difficult to tell from a comparison of the Sand Canyon vessel data and the Castle Rock rim-arc data whether bowl size distributions in the systemic assemblages of these two sites—the numbers and sizes of vessels in use at any one time—were similar. The Sand Canyon (and Castle Rock) vessel assemblage probably underestimates the number of small bowls in the systemic assemblage, because the vessel assemblage was collected primarily from abandonment contexts, and smaller bowls were lighter and more easily carried away from the site at abandonment. The Castle Rock rim-arc data, on the other hand, may overestimate the number of small bowls in the systemic assemblage because of the bias against large-diameter vessels inherent in rim-arc analysis (see Pierce and Varien 1999*1). Comparison of rim-arc data from the two sites, however, may give some indication of the relative importance of large and small bowls in these sites. This comparison is made in paragraphs 53–54 below.

White Ware Bowls, Sand Canyon Locality Sites

⁵³
Figure 4 compares rim-arc data for white ware bowls from Castle Rock and Sand Canyon pueblos, along with a number of smaller, tree-ring-dated thirteenth-century sites in the Sand Canyon locality, using the same data conventions described in paragraph 49. A bimodal distribution of bowl sizes is clearly evident for Sand Canyon and Castle Rock Pueblos, both of which are late-thirteenth-century villages. The modes, at 8 cm and 14 cm, correspond closely with the 7.95-cm and 14.2-cm mean rim radii for small and large bowls in the Sand Canyon Pueblo vessel assemblage (see paragraphs 43–48). The large size mode is slightly more pronounced at Sand Canyon Pueblo.

⁵⁴
The early-thirteenth-century small sites (Kenzie Dawn Hamlet, Lillian's Site, Shorlene's Site, and Roy's Ruin) and late-thirteenth-century small sites (Saddlehorn Hamlet, Troy's Tower, Lester's Site, and Lookout House), on the other hand, have less well defined size modes in their rim-radius distributions. The large size mode is less developed, and smaller bowls are more variable in size than is the case for the larger villages. These data may suggest that distinct white ware bowl size categories developed as villages formed in the Sand Canyon locality. Since white ware bowls were serving vessels, it is possible that the formalization of these size modes and the increased frequency of large bowls in villages were related to changes in the way food was served and consumed in villages relative to hamlets. This and other possible explanations are discussed in paragraphs 59–66, and possible shortcomings in the available data are addressed in paragraphs 55–56.

White Ware Bowl Rim Sherd Weight Distributions, Sand Canyon Locality Sites

⁵⁵
Figure 5 presents box plots of weights in grams for white ware bowl rim sherds for the same groups of Sand Canyon locality sites discussed in paragraphs 53–54. In these plots, the box represents the midspread (middle 50 percent of cases) for each weight distribution, the thick line inside the box represents the median, and the tails illustrate the range of weights for each distribution. For sherds of roughly equal thickness, weight is proportional to the overall size of a sherd, and thus sherd weight distributions can be taken as proxies for sherd size distributions. The chart shows that bowl rim sherds tend to be largest at Sand Canyon Pueblo, followed by Castle Rock, then by the late-thirteenth-century small sites, and finally by the early-thirteenth-century small sites. Comparison of these results with those presented in Figure 4 shows that the degree of large size mode development in bowl rim radius estimates correlates with sherd size among the Sand Canyon locality sites. That is, Sand Canyon Pueblo has the most pronounced large bowl size mode and the largest sherds overall, whereas the early-thirteenth-century small sites have the least pronounced large bowl size mode and the smallest sherds overall. Since rim-arc analysis is biased against large-diameter vessels, and this negative bias becomes more pronounced as sherd size decreases, the differences in bowl radius distributions shown in Figure 4 may relate in part to sherd size; however, intrinsic differences in bowl sizes across these four groups of sites may still exist, because the data in Figure 4 partly control for sherd size by excluding rim sherds preserving fewer than 20 degrees of arc. In addition, the average thickness of white ware bowl sherds is known to have increased over time, making it difficult to control for changing sherd thickness with these data. What we really want to know is whether the mean degrees of arc encompassed by sherds selected for Figure 4 varies across the four site groups. These data are considered in paragraph 56 below.

White Ware Bowl Rim Sherd Size Distributions, Sand Canyon Locality Sites

⁵⁶
Figure 6 presents the mean degrees of arc encompassed by the white ware bowl rim sherds assigned to each radius class in paragraphs 53–54, using the same four site groups as in that section. If sherd size is primarily responsible for the variation in large size mode development in Figure 4, then one would expect Castle Rock Pueblo sherds to be significantly larger on average than sherds from hamlets for the 14-cm radius class, the class at which the large size mode occurs in Figure 4. Although Sand Canyon Pueblo sherds do tend to be larger than those from the other sites at the radius of the large size mode, there are no significant differences between rim sherds from Castle Rock and those from the small Sand Canyon locality sites, which suggests that the pattern illustrated in Figure 4 relates to real differences in bowl size distributions among these site groups.

Corrugated Jars, Sand Canyon Locality Sites

⁵⁷
Figure 7 presents rim-radius estimate distributions for corrugated jar rim sherds among the four groups of sites discussed in paragraphs 53–54, using the same conventions described in that section and in paragraph 49. Although the relationship between rim diameter and volume is weaker for corrugated jars than for white ware bowls, it is still generally positive, suggesting that corrugated jars with larger rim diameters tended to be larger overall (see paragraph 46). The effects of sherd size notwithstanding, this chart suggests that inhabitants of Sand Canyon Pueblo used more large corrugated jars than did inhabitants of other Sand Canyon locality sites. The common occurrence of sooting on the outsides of corrugated jars indicates that such vessels were typically used for cooking. If more large-volume corrugated jars were in fact used at Sand Canyon Pueblo, it would indicate that more large meals were prepared there than at other Sand Canyon locality sites. Since household organization appears to have been similar at Sand Canyon and Castle Rock, the preparation of more large meals at Sand Canyon Pueblo may be an indication that more communal feasts occurred there. Driver (1996*1) has developed such a hypothesis on the basis of higher proportions of artiodactyl (deer and elk) remains at Sand Canyon Pueblo than at other Sand Canyon locality sites. Unfortunately, the sherd size effects discussed in paragraph 58 below cannot be ruled out for the corrugated jar rim sherd data, and as a result these data offer only equivocal support for Driver's hypothesis.

Corrugated Jar Sherd Size Distributions, Sand Canyon Locality Sites

⁵⁸
Figure 8 presents the mean degrees of arc encompassed by corrugated rim sherds assigned to each radius class shown in Figure 7, using the same four site groups as in that section and in paragraphs 53–54. For the 12-cm-radius class—the interval for which differences among the site groups are most marked in Figure 7—the percentage of the total degrees of arc assigned to the interval is correlated with the mean degrees of arc encompassed by sherds assigned to that interval across the four site groups. Thus, the effects of sherd size may be partly responsible for the apparent emphasis on larger vessels in the Sand Canyon Pueblo collection.

Exterior Paint on White Ware Bowls, Sand Canyon Locality Sites

⁵⁹
One possible explanation for the development of distinct size modes and the increased frequency of large vessels among white ware bowls in thirteenth-century Sand Canyon locality villages is that village formation led to changes in the way food was presented and consumed. These changes in presentation and consumption patterns might also have affected the characteristic ways in which these vessels were viewed, leading to changes in the way they were decorated. Prior to the thirteenth century, most inhabitants of the Sand Canyon locality lived in small hamlets containing one or, at most, a few households (Adler 1992*3). Ethnographic accounts (e.g., Stevenson 1904*1:369) suggest that traditional Pueblo meals involved small groups of people sitting around one or more communal serving bowls, scooping food out of them with their fingers. As villages formed during the thirteenth century, an increasing number of meals were probably consumed in contexts that exposed bowl exteriors to view by more distant social relations. In historic and modern Pueblo villages, plazas are settings for community events including dances, ceremonies, feasts, and the redistribution of food. Informal plazas were created inside the enclosing walls of both Sand Canyon and Castle Rock Pueblos, suggesting that analogous events might have taken place in these villages. If so, prepared food would have been carried into the plaza by participants in the event, giving spectators an opportunity to view vessels from the side. Also, everyday household meals eaten outdoors during mild weather would have exposed bowl exteriors to viewing by neighbors. Thus, bowl exteriors might have been viewed much more often in villages than in hamlets.

⁶⁰
Ancient Pueblo pottery vessels tended to be decorated most intensively on areas that had relatively high contextual visibility (for a cross-cultural perspective on this phenomenon, see Carr [1995*1:185-215]). The decorative field on white ware jars was usually the upper half of the spherical body, which would have been more visible than the lower half when the vessel was sitting on the ground, a floor, or a bench. The interiors of white ware bowls were also more intensively decorated and during meals would have been viewed from above as food was eaten. Given this correlation between contextual visibility and intensity of decoration, one might expect the exterior surfaces of white ware bowls to have been decorated more intensively in villages than in earlier and contemporaneous hamlets. Table 16 presents these data for Sand Canyon Pueblo, Castle Rock Pueblo, and several tree-ring-dated thirteenth-century hamlets in the Sand Canyon locality. Only continuous exterior paint, in the form of circumferential lines or band designs, is tabulated here, because the proportion of sherds with continuous exterior paint mirrors the proportion of vessels with such designs, whereas the frequency of bowls with isolated exterior designs would be significantly underestimated by sherds.

⁶¹
These data show, first, that a higher frequency of white ware bowls with continuous exterior designs was broken during the occupation of Sand Canyon Pueblo than that of Castle Rock; second, that the frequency of painted bowl exteriors increased significantly during the period of village formation; and, third, that for the most part, more such vessels were broken at the villages than at contemporaneous hamlets. Also note that among the hamlets contemporaneous with Sand Canyon and Castle Rock pueblos, the two sites with the highest proportion of exterior painted bowls—Lookout House and Lester's Site—were located within 300 m of Sand Canyon Pueblo. The other two contemporaneous hamlets, Troy's Tower and Saddlehorn Hamlet, lay more than a kilometer away from the nearest village (Varien 1999*1) and had far fewer bowls with painted exteriors in their assemblages. These data are consistent with the notion that the overall relative contextual visibility of white ware bowl exteriors was increased in village contexts and that potters responded to this increased visibility by decorating bowl exteriors more often and more elaborately. Possible relationships between this changing decorative treatment and the formalization of small and large serving bowls are discussed in paragraphs 62–66.

Summary of Functional Analysis

⁶²
Data on the sizes and uses of vessel forms in Sand Canyon locality sites suggest that white ware bowls, white ware ladles, and corrugated jars formed a functionally interrelated complex. Meals were prepared in corrugated jars and were scooped out of these jars and into serving bowls using ladles. Rim-arc data suggest that inhabitants of Sand Canyon and Castle Rock pueblos tended to make bowls of two distinct sizes. Such distinct size modes were not emphasized in the pottery assemblages of contemporaneous and earlier thirteenth-century hamlets in the Sand Canyon locality. Relatively more large bowls also appear to have been broken in thirteenth-century villages than in earlier and contemporaneous hamlets. More large corrugated jars may also have been used at Sand Canyon Pueblo than at other Sand Canyon locality sites, but sherd size effects that could contribute to this pattern in the available data cannot be ruled out. Finally, it appears that bowl exteriors were decorated in accordance with their relative visibility in various settlement contexts. Bowl exteriors were decorated most often at Sand Canyon Pueblo, followed by Castle Rock and contemporaneous hamlets close to villages, then by contemporaneous hamlets farther away from villages, and finally by hamlets dating prior to the era of village formation, where bowl exteriors were only rarely decorated.

⁶³
One possible explanation for these related changes in serving bowls is that village formation was associated with an intensification of communal ritual (Driver 1996*1; Muir 1999*1), and this intensification was associated with changes in the ways certain meals were served and consumed in villages compared with how meals were served and consumed in earlier and contemporaneous hamlets. Communal rituals probably did not originate in thirteenth-century villages, but when such ceremonies occurred in village contexts, it would have been relatively easy for inhabitants of the village and other sites nearby to prepare a meal to be taken to the plaza for consumption during the event, following the lines of Blinman's (1989*1) "potluck" model of communal feasting. Serving and consuming food in the context of communal ceremonies in plazas would have changed two aspects of meals. First, if food was eaten by spectators watching a ceremony, it would have been difficult for them to eat around a communal serving bowl and watch the event at the same time. Second, if food was presented to participants in ceremonies or was served to spectators, then the exteriors of the bowls in which it was served would have been visible to other, more distant social relations attending the event—much more so than in hamlet-based meals. The development of distinct bowl size modes may relate to the first change, and the increased intensity of exterior decoration to the second.

⁶⁴
If large serving bowls could contain enough food to feed an entire family, and if small bowls were one-fifth the size of large bowls, then it may be that distinct bowl sizes developed along with villages to facilitate the serving and consumption of meals at communal feasts. That is, small bowls would have been formalized as individual serving vessels, and large bowls as communal serving vessels. It may be unjustified, however, to assume that food was actually consumed by spectators watching public ceremonies. In some modern pueblos, food presented or redistributed during a ceremony is not eaten in public view but is taken back to individual households for consumption later. If public ceremonies in the Sand Canyon locality took this form, then individual serving vessels would have been unnecessary and perhaps even contrary to a communal concept. This scenario would not necessarily affect the visibility argument of the hypothesis, but it would undermine the explanation of bimodal bowl size distributions. On the other hand, the modern Pueblo practice may simply indicate that the character of communal meals has changed since the thirteenth century.

⁶⁵
A second possible explanation for these changes in pottery vessel sizes and decoration is that food consumption groups grew larger as villages formed and as more distant relatives came to live in a single settlement. The average number of rooms per kiva in kiva suites and the average number of mealing bins in corn-grinding areas did not change appreciably between A.D. 1150 and 1280 (Ortman 1998*2:Table 9.7), suggesting that household composition itself did not change significantly as villages formed. Whether inhabitants of neighboring kiva suites in villages ate together more often than did the inhabitants of adjacent hamlets in earlier times is unknown.

⁶⁶
A third possible explanation for the bimodal bowl size distribution, suggested by members of Crow Canyon's Native American Advisory Group, is that different kinds of food might have been served in vessels of these two sizes. If the sizes of bowls were related to the weight and richness of the foods served in them, it remains to be seen what kinds of changes in cuisine might have coincided with both village formation and the development of distinct bowl sizes. As of yet there is little evidence in cooking technology or in the botanical and faunal records of dramatic changes in food-preparation techniques or in the types of foods prepared as villages formed in the Sand Canyon locality. Also, if a range of foods was prepared, and if the sizes of serving bowls were tailored to the characteristics of different foods, then one might expect a range of bowl sizes to have been made to accommodate them, especially given the evidence that most households produced their own pottery (see paragraphs 68–69). Obviously, firm explanations for the patterns identified in this analysis require further research.

Pottery Production and Exchange

⁶⁷
This section summarizes direct and indirect evidence of pottery production at Castle Rock Pueblo and examines the nature of the local and interregional pottery exchange networks in which Castle Rock participated. Direct evidence of pottery production at Castle Rock includes manufacturing tools (polishing stones), raw materials (potting clay), and unfinished vessels (unfired sherds). A fourth potential item of direct evidence is pottery scrapers, but although such artifacts have been collected from other sites in southwestern Colorado (e.g., Wilson 1988*2:Table A.6), no definite examples have been identified in the Castle Rock modified sherd assemblage. Indirect evidence of pottery production and exchange consists of temper and compositional data from white wares found at Castle Rock and several other villages and hamlets in southwestern Colorado. Evidence of long-distance pottery exchange consists of identifiable nonlocal sherds from Castle Rock and other sites in southwestern Colorado.

Direct Evidence of Pottery Making

⁶⁸
The amount and distribution of direct evidence of pottery making can be used to assess the nature of pottery production at Castle Rock. If pottery making was an unspecialized, household-level industry, then raw materials and tools associated with it should occur occasionally throughout the site. On the other hand, if pottery production was specialized, such that relatively few people made most of the pottery used in the village, then direct evidence should be relatively abundant in a few locations and absent in most others.

⁶⁹
The evidence from Castle Rock suggests that white ware production was unspecialized. Direct evidence of white ware manufacture is widely distributed at the site, despite the fact that most structures received only limited testing. Because Castle Rock Pueblo was not completely excavated, it remains possible that concentrations of direct evidence remain to be found in areas that were not excavated. Nevertheless, the fact that direct evidence was found in so many of the tested structures suggests that white ware production was a household-based, part-time activity. This pattern has been noted at numerous other sites in southwestern Colorado (Errickson 1993*1; Wilson 1988*2, 1991*1). In contrast, little direct evidence of gray ware production was found. This is partly due to the fact that polishing stones, base molds, and pottery scrapers are not required for gray ware manufacture. Also, despite the fact that raw igneous rock, which could have been crushed for use as white ware or gray ware temper, was widespread at Castle Rock, it is impossible to determine which rock samples were used for this purpose. Nevertheless, no unfired corrugated sherds or coarse-tempered raw clay samples were identified in the Castle Rock collection, raising the possibility that gray ware production was organized quite differently from white ware production.

Polishing Stones

⁷⁰
Polishing stones are small, very smooth, and very hard stones or pebbles that exhibit evidence of abrasive wear. Traces of clay were found adhering to the surfaces of a few such stones from Sand Canyon locality sites, indicating that at least some of these stones were used for polishing the surfaces of white ware vessels (Pierce and Varien 1999*1). It is unknown whether polishing stones had additional uses.

Inventory

⁷¹
Table 17 gives the distribution of polishing stones from the excavations at Castle Rock Pueblo by study unit and by vertical position. It shows that most polishing stones were found in the lower fill and on floor surfaces of structures. The locations of study units in which polishing stones were found indicate that polishing stones were widely distributed across the site (Database Map 509).

Raw Materials

⁷²
Table 18 lists the types of stone out of which polishing stones were made. It shows that polishing stones were made of high-quality, fine-grained stone. Even if some of these stones were found locally, they were rare and required some effort to procure.

Potting Clay and Unfired Sherds

⁷³
The strongest direct evidence of pottery making consists of sherds from vessels that had not yet been fired when the site was abandoned. Several such sherds were found in various locations at Castle Rock Pueblo. A few of these are fine-tempered and have the smooth surfaces characteristic of white ware vessels; none possesses the corrugated surface and coarse temper of gray ware vessels. Raw clay suitable for use in pottery making was also found in several locations at Castle Rock. Most of these samples are of raw, untempered clay, although some were tempered or had been molded into balls or other shapes. No samples of raw clay were recognized as containing the coarse temper characteristic of gray ware pastes.

⁷⁴
Table 19 summarizes the distribution of potting clay and unfired sherds by study unit and vertical position. Most samples were found in the lower fills, on floors, and in features of structures, indicating that pottery making likely occurred within those structures. Importantly, these artifacts were found in at least one room or kiva in 10 of the 16 kiva suites defined for Castle Rock Pueblo. If we were to add polishing stones as a less certain indicator to this distribution, then 12 of the 16 kiva suites would contain direct evidence for pottery making. Direct evidence has not been identified from Kiva Suites 101, 102, 204, and 405, but these suites received very limited testing (kiva suites are numbered according to the structure number of the kiva around which the above-ground rooms are arranged). Direct evidence of pottery making was found in nearly every kiva suite in which significant excavations took place, suggesting that nearly every kiva suite had a resident potter living in it.

⁷⁵
Raw clay and unfired sherds were found in seven of the 16 kivas at Castle Rock, including the square kiva, Structure 103. They occurred on the floors and inside features of three kivas—Structures 104, 406, and 125. If polishing stones are included as a less certain indicator of pottery production, then nine of the 16 kivas at Castle Rock contain direct evidence of pottery manufacture, including the potentially oversized kiva, Structure 105. In all historic pueblos, pottery was made primarily by women, and kivas functioned primarily as ceremonial structures used by men. That so much direct evidence of pottery making should occur in kivas suggests that either the use of kivas or the sexual division of labor has changed since the thirteenth century. Since corn-grinding areas are also found in a significant proportion of Mesa Verde–region small kivas (Cater and Chenault 1988*1; Ortman 1998*2), it appears more likely that the functions of kivas have changed and that small kivas were a component of residential architecture during the occupation of Castle Rock (see also Lipe 1989*1).

Other Clay Artifacts

⁷⁶
Table 20 lists a small number of unusual fired and unfired clay objects from Castle Rock that might or might not have been parts of pottery vessels. Most of these objects were found in secondary refuse or midden deposits, suggesting that they represent waste products of pottery making.

Local Pottery Exchange

⁷⁷
In this section, temper data for white wares at Castle Rock and several other late Pueblo III (A.D. 1250–1280) hamlets and villages in southwestern Colorado are used to examine the question of local pottery exchange. Temper data have not been collected for gray wares. This analysis builds on previous studies of local pottery exchange in the Sand Canyon locality (Glowacki 1995*1; Glowacki et al. 1995*1, 1998*1; Thurs et al. 1996*1) and other areas in southwestern Colorado (Glowacki et al. 1997*1) using instrumental neutron activation analysis (INAA) data as well as temper data. These studies have identified distinct white ware manufacturing tracts and have documented modest levels of vessel movement between sites. The temper data discussed in this section generally support and amplify these conclusions.

Temper Data

⁷⁸
A sample of white ware bowl rim sherds from several Sand Canyon locality sites and additional villages and hamlets in southwestern Colorado was analyzed for temper using a binocular microscope (Table 21). All the analyzed sites are associated with tree-ring dates or architectural features indicative of habitation during the late Pueblo III period, A.D. 1250–1280. Each sherd was classified on the basis of the most abundant type of nonplastic inclusion mixed with the clay during paste preparation. The four categories identified were crushed sandstone, quartz sand, crushed igneous rock, and crushed sherd. The results are tabulated by count and by the proportion of each category within the sample from each site. Sample sizes vary in this data set from 15 to 391 sherds.

⁷⁹
Two of the four temper types—crushed sandstone and crushed sherd—were readily available to potters in all these sites. The distributions of quartz sand and igneous rock, on the other hand, were much more circumscribed. Of the natural distributions of these two tempering agents, that of igneous rock is the better known. Igneous rock originates in the intrusive volcanic mountains of the Four Corners area, including Sleeping Ute Mountain and the San Juan Mountains in Colorado, the Abajo Mountains in Utah, and the Carrizo and Chuska mountains in Arizona and New Mexico. Weathered igneous cobbles suitable for use as pottery temper can be found on terraces along the watercourses that drain these mountains. The closest known sources of igneous rock for the sites discussed here are Ute Mountain and McElmo Creek, located just south of Castle Rock Pueblo. The sites in the table are arranged according to the proportion of sherds in each sample tempered with igneous rock, in descending order. This order closely mirrors the distance of each site from the center of Ute Mountain. The site farthest from Ute Mountain, Hedley Main Ruin, is approximately equidistant from Ute Mountain and the Abajo Mountains in southeastern Utah.

⁸⁰
Cross-cultural data compiled by Arnold (1985*1:51–56) suggest that potters in small-scale societies tend to travel no more than 6 to 9 km to obtain temper for pottery making. The top seven sites listed in the table, including sites from the Cowboy Wash area (5MT9943) and the Sand Canyon locality, lie close enough to Ute Mountain for their inhabitants to have procured igneous temper directly and incorporated it into their pottery. Since other temper sources were readily available to potters in these areas, the proximity of these sites to Ute Mountain does not necessarily mean that their inhabitants made all their vessels with igneous temper. Also, since the analyzed sherds are from broken vessels that could have been exchanged several times during their use lives, igneous-tempered sherds found at a site could have been made by a potter at another site within the area of direct igneous temper procurement. The cross-cultural limit on the distance that potters are willing to travel to obtain temper, however, does suggest that nearly all the igneous-tempered sherds in the samples from these 11 sites were made within a zone extending 6 to 9 km around the edge of Ute Mountain. Igneous-tempered sherds from sites outside the Ute Mountain source area could have come from a different source area, but if so, they would have come from even farther away. (The Dolores River valley is not a likely source for the single igneous-tempered sherd from Yellow Jacket Pueblo included in this table, because the Dolores River is more than 9 km from the site, and thirteenth-century settlement along the river was minimal to nonexistent [Kane 1986*1].) There may also be igneous-tempered sherds at sites within the Ute Mountain source area that were made using rock from a different source, but owing to the rarity of igneous-tempered sherds at sites outside the Ute Mountain source area, this seems unlikely. So it seems reasonable to assume that nearly all the igneous-tempered sherds found in these sites were made within the Ute Mountain source area.

Igneous-Tempered White Wares in Late Pueblo III Sites

⁸¹
Figure 9 presents the proportion of white ware sherds containing igneous temper in the sites listed in Table 21. The x-axis value is the distance in kilometers of each site from the geographical center of Ute Mountain. The hamlet with the highest frequency of igneous temper is Site 5MT9943, located in the Cowboy Wash area on the Southern Ute Piedmont. The hamlet closest to Castle Rock Pueblo is Saddlehorn Hamlet (5MT262), and the hamlets that cluster close to Sand Canyon Pueblo include Lookout House (5MT10459), Lester's Site (5MT10246), and Troy's Tower (5MT3951). Two interesting patterns are apparent in this chart. First, as the distance of a site from Ute Mountain increases, the frequency of igneous-tempered pottery generally decreases. Second, at any given distance from the most likely source, smaller hamlets tend to have higher proportions of igneous-tempered white wares than larger villages. A possible explanation for the second pattern is discussed in this section. The first pattern is considered further in paragraph 83.

⁸²
One possible interpretation for the elevated frequencies of igneous-tempered pottery in hamlets relative to villages is that village inhabitants, taken as a whole, exchanged pottery more widely than did inhabitants of hamlets. All of the hamlets, as well as Sand Canyon and Castle Rock pueblos, lie within the Ute Mountain source area, and therefore their inhabitants could have made igneous-tempered vessels. But since many more people lived in Sand Canyon locality villages than in each nearby hamlet, it is likely that the social relationships of village inhabitants, taken as a whole, defined larger social networks that encompassed greater geographical areas than did the networks defined by the social relationships of hamlet inhabitants. If there was no marked directionality to these networks, then the larger social networks of villages would have been more likely to include more potters who lived outside the Ute Mountain source area than would the social networks of hamlets. Effectively random movement of vessels through social networks of these differing sizes would have resulted in a net depletion of igneous-tempered vessels in Sand Canyon locality village assemblages, because more igneous-tempered vessels would have moved out of them than into them, relative to hamlets.

Fall-off Pattern of Igneous-Tempered White Wares, Late Pueblo III Sites

⁸³
Figure 10 presents the same data as those in Figure 9 on differently scaled axes: the x-axis presents the squared distance of each site from Ute Mountain in kilometers, and the y-axis presents the log of the proportion of igneous-tempered sherds in the sample from each site. To accommodate a logarithmic scale, the proportion of igneous-tempered sherds from Hedley Main Ruin was changed from 0 percent to .001 percent. This scaling produces a strong correlation between the data and a linear regression line. Thus, the fall-off pattern of igneous-tempered pottery with distance from the most likely source closely approximates a linear log/square relationship. This pattern of decay is typical of down-the-line exchange patterns, in which artifacts diffuse across a social landscape through informal exchange networks. Hodder and Orton (1976*1:Chapter 3) showed that linear log/square decay patterns are closely approximated by random-walk computer models, in which a commodity moves a certain distance in a random direction and is exchanged a certain number of times before being discarded and entering the archaeological record. That the dispersal of igneous-tempered white ware vessels is consistent with a random-walk process suggests that thirteenth-century villages did not function as redistribution centers for white ware vessels. It is also consistent with the interpretation offered in paragraphs 81–82 and with the evidence for widespread, unspecialized, household-level production of white ware pottery. Market redistribution is unnecessary for such a commodity. Given the distance-decay data, a more likely mechanism for the dispersal of white wares was gift exchange among relatives and friends who lived in nearby habitations.

Long-Distance Pottery Exchange
Catalog of Nonlocal Pottery Sherds

⁸⁴
A small percentage of the pottery sherds found at Castle Rock was of nonlocal manufacture, judging from paste, temper, and color characteristics. Table 22 catalogs these items by provenience designation (PD). Most of the nonlocal sherds found at Castle Rock are San Juan Red Ware, the production of which appears to have been centered in southeastern Utah (but see Glowacki et al. [1997*1] for possible evidence of red ware production using alluvial clays from McElmo Creek). The term Indeterminate Local Red refers to indeterminate San Juan Red Ware, with local being used to distinguish such sherds from red wares made even farther away from the Mesa Verde region. San Juan Red Ware is not known to have been made after A.D. 1150. Pierce and Varien (1999*1) argue that the presence of San Juan Red Ware sherds at Pueblo III sites reflects the scavenging of red ware sherds from earlier middens by thirteenth-century inhabitants. Other nonlocal sherds include Tsegi Orange Ware (including Tusayan Black-on-red) and White Mountain Red Ware (including St. Johns Polychrome). The identified nonlocal sherds suggest trade connections to the west, southwest, and south.

⁸⁵
No nonlocal white wares or gray wares were identified. Tusayan and Chuska white wares and Chuska Gray Ware are easily distinguishable from local sherds, so their absence in the Castle Rock collection is probably real. There are also no Fremont, Pueblo IV, or historic period pottery types in the Castle Rock collection. It is difficult to distinguish between Tusayan, Cibola, and Mesa Verde gray wares, because many of the same tempers were used in all three. It is also difficult to distinguish between Pueblo III Cibola and Mesa Verde white wares, owing to overlapping temper, paste, paint, and design characteristics. In general, gray wares and white wares were assumed to be local Mesa Verde wares if no distinctive characteristics of other pottery traditions were identifiable. There could, therefore, be a few nonlocal gray wares and white wares in the Castle Rock assemblage that were indistinguishable from local sherds. Compositional analysis of sherd pastes might distinguish such sherds, if they indeed exist in the collection, but because many of the same geological strata are exposed in the San Juan basin as in the Mesa Verde region, even compositional data might prove inconclusive.

Distribution of Nonlocal Sherds by Study Unit

⁸⁶
Table 23 summarizes the distribution of demonstrably nonlocal sherds (San Juan Red Ware and other nonlocal red wares) at Castle Rock by study unit. For an explanation of the study unit codes, see the on-line field manual. Nonstructure 9 is a surface with a pit feature underlying a portion of Nonstructure 1 south of Block 100. All sherds, both local and nonlocal, from each study unit are included in the column headed "Total Sherds." All but two of the 32 nonlocal sherds (those in PDs 739 and 978 in Table 22, which are from Arbitrary Unit 300 and Structure 402, respectively) were found in excavation units south of the central butte. This pattern probably reflects the fact that most excavations occurred on the larger, southern side of the site. Nonlocal sherds were most common in secondary refuse or midden deposits, where general artifact densities were high. Fewer than 1 percent of the sherds found at Castle Rock were of nonlocal origin, indicating that long-distance pottery exchange was exceedingly rare.

Nonlocal Sherds from Other Sites in Southwestern Colorado

⁸⁷
Table 24 compares the proportions of nonlocal sherds collected from Castle Rock and other sites in southwestern Colorado. The sites are arranged first according to location, second according to core occupation date as determined by the excavators of each site, and third according to the frequency of nonlocal sherds in each assemblage by count. Although weight would be the better measure of abundance for comparing these assemblages, weight figures are unavailable for many of the sites, necessitating the use of counts. As a result, sherd size effects and analytical biases cannot be ruled out for these data.

⁸⁸
All of the listed sites other than Castle Rock are small hamlets, with the exception of Escalante Ruin, which has been labeled a "Chacoan great house." The piedmont area south of Ute Mountain is the only part of southwestern Colorado in which nonlocal sherds occur as more than 5 percent of total pottery assemblages. This area is on the periphery of the Mesa Verde culture area and lies closest to other centers of ancient Puebloan culture. Most of the nonlocal sherds from Site 5MT10206, for example, are from the Chuska Mountains area in northwestern New Mexico. In other areas of southwestern Colorado, the frequency of imported pottery tends to decrease through time. Within the Sand Canyon locality, imported pottery appears to be no more common at Castle Rock than at contemporaneous hamlets, suggesting that villages did not function as nodes for long-distance exchange networks, similar to findings for local pottery exchange in paragraph 83.

⁸⁹
The relative contribution of San Juan Red Ware and other nonlocal sherds to nonlocal sherd totals varies along geographical lines. The western Montezuma Valley is closest to centers of San Juan Red Ware production, and imported pottery in these collections is dominated by San Juan Red Ware. In contrast, the eastern Montezuma Valley and Southern Ute Piedmont are farther removed from San Juan Red Ware production centers, and their nonlocal assemblages are dominated by sherds from other areas. Nonlocal sherds of all kinds are rarest in the Sand Canyon locality, which lies in the center of the Mesa Verde region and is most distant from adjacent regions.

⁹⁰
The extreme rarity of imported sherds in late Pueblo III sites suggests that inhabitants of the Mesa Verde region had very few interactions with Pueblo people in other parts of the Southwest during the final decades of the A.D. 1200s. Implications of this finding are further discussed in paragraphs 129–131.

Chipped-Stone Tools and Manufacturing Debris
Definitions of Raw Material Categories

⁹¹
Although knowledge of lithic procurement sites and raw material availability in southwestern Colorado is limited, the raw materials out of which Castle Rock Pueblo chipped-stone tools were made can be grouped into local, semilocal, nonlocal, and unknown stone types. Each group is discussed briefly in this section.

Local Raw Materials

⁹²
Local raw materials are of average to poor quality, occur within the geological strata exposed around Castle Rock and lower Sand Canyon, and were likely available within easy walking distance of the site. The closest known source of Dakota quartzite is in upper Sand Canyon near Stanton's Site, approximately 6 km north of Castle Rock. Morrison quartzite and chert/siltstone are also available within Sand Canyon, as are fine-grained and conglomerate sandstones. Igneous rock occurs on Ute Mountain, immediately south of McElmo Creek. Finally, slates and shales are available in the Mancos Formation, which outcrops throughout the uplands of southwestern Colorado.

Semilocal Raw Materials

⁹³
Semilocal raw materials are of relatively good quality and probably occur less widely in their geological strata of origin than do local raw materials. As a result, such materials were probably more difficult to obtain, possibly requiring special collecting trips. Agate/chalcedony and petrified wood occasionally occur within the Morrison and Dakota Formations, as well as in other formations that outcrop only farther away. Jet occasionally occurs within shale and coal-bearing deposits in the Dakota, Mancos, and Menefee Formations. The closest known source of Burro Canyon chert occurs at a lithic procurement site on Cannonball Mesa, approximately 15 km west of Castle Rock. Additional known sources occur in the Dolores River valley. Known sources of Brushy Basin chert occur around the San Juan River near the Four Corners monument, approximately 45 km from Castle Rock (Green 1985*1:71–72).

Nonlocal Raw Materials

⁹⁴
These are high-quality materials that definitely do not occur within easy walking distance of Castle Rock and must have been acquired through either trade or special collecting trips. Red jasper comes from Triassic and Permian formations of the Monument Upwarp and Elk Ridge Uplift in southeastern Utah, west of Cottonwood Wash. Obsidian is likely to have come from either the Jemez Mountains of New Mexico or the San Francisco Peaks in Arizona, where sources of widely exchanged obsidian are known (Shackley 1988*1, 1995*1). Washington Pass chert occurs only in the Chuska Mountains of northeastern Arizona (Warren 1967*1).

Artifact Type vs. Raw Material
By Count

⁹⁵
Table 25 summarizes the number of chipped-stone artifacts in the Castle Rock Pueblo assemblage made from various raw materials (for definitions of the artifact types used, see the on-line laboratory manual). The full suite of raw material categories was considered for projectile points, bifaces, and drills, but since Brushy Basin chert, Burro Canyon chert, and red jasper were recorded only in comments for other chipped-stone artifact categories, their presence may be underrepresented in those categories. If these materials were not consistently identified, then red jasper would probably have been classified as nonlocal chert/siltstone, Brushy Basin chert as Morrison chert/siltstone or unknown chert/siltstone, and Burro Canyon chert as unknown chert/siltstone.

By Percentage

⁹⁶
Table 26 summarizes the proportion of objects in each chipped-stone artifact category that were made of various raw materials. Even if possibly inconsistently identified raw material categories are excluded, it appears that semilocal and nonlocal raw materials were used most often for making formal tools, including projectile points, some drills, and bifaces. On the other hand, informal, expedient tools, including peckingstones and modified flakes, were most often made of local, readily available materials. Also, no cores of demonstrably nonlocal materials were identified. This suggests that nonlocal materials were not procured directly and that primary reduction of nonlocal materials did not occur at Castle Rock. Nonlocal raw materials may have been procured in the form of formal tool blanks or as finished tools, which could have been used as is or modified into other tools.

Raw Materials in Tools and Manufacturing Debris
By Count

⁹⁷
Table 27 summarizes the raw materials of all chipped-stone tools and a sample of the manufacturing debris from the Castle Rock excavations. Chipped-stone objects are grouped into the following categories: chipped-stone debris, cores and core tools (cores, modified cores, and peckingstones), expedient tools (modified flakes and other chipped-stone tools), and formal tools (projectile points, bifaces, and drills). The column for chipped-stone debris includes only materials from the probability sample at Castle Rock (see paragraphs 134–154); the other three columns consider all such objects collected during the excavations. Chipped-stone debris is tabulated only for the probability sample because raw material identification for chipped-stone debris was incomplete for the intensive excavations, whereas all but nine pieces in the probability sample were analyzed for material type. A simplified set of raw material categories has been used to account for changes in the analytical criteria and raw material categories used by Crow Canyon laboratory staff over the course of the Castle Rock excavations. Dakota quartzite was simplified as Dakota Formation material, Morrison quartzite and chert/siltstone were simplified as Morrison Formation material, Burro Canyon chert and Brushy Basin chert were added to unknown chert/siltstone, and red jasper was added to nonlocal chert/siltstone.

By Percentage

⁹⁸
Table 28 presents column percentages for the data in Table 27, so that the relative abundance of different raw materials can be compared across chipped-stone artifact categories. The general correspondence between percentages of raw materials in chipped-stone debris, cores and core tools, and expedient tools probably relates to the relative amount of knapping of these various materials. The relative frequencies of raw materials across these categories appear roughly proportional to their availability in the local environment. In contrast, raw materials occur in very different proportions among formal tools. The most readily available materials from the Morrison Formation were seldom used for formal tools, whereas Dakota Formation materials, agate/chalcedony, nonlocal chert/siltstone, and unknown chert/siltstone are all overrepresented relative to other chipped-stone artifact categories. These data suggest that formal tools were made from certain preferred raw materials regardless of their availability in the local environment, whereas expedient tools, including peckingstones and modified flakes, were made from whatever material was at hand or easily obtainable.

Analysis of Bifaces, Points, and Drills
Catalog, Analysis Data, and Provenience

⁹⁹
Table 29 presents a catalog of all bifaces, projectile points, and drills collected from Castle Rock, along with the original use, condition, material, production stage, size, and context of each object. The point-classification scheme used follows Pierce (1999*1), Hayes and Lancaster (1975*1), Geib (1996*1), and Jennings (1986*1). A Bull Creek point (PD 97, FS 5) conforms to the variety found in Kayenta- and Mesa Verde-tradition sites in southeastern Utah south of the Colorado River and as far east as Montezuma Canyon (Geib 1996*1; Matheny 1962*1). This point was made of red jasper, which occurs widely in this area west of Cottonwood Wash. The combination of nonlocal material and an exotic point style suggests that this point was made in southeastern Utah. A second point of red jasper (PD 371, FS 4) may also be nonlocal, but since it appears to have been informally made on a small flake, it could also have been made locally using a nonlocal material.

¹⁰⁰
There is also a Desert Side-Notched point (PD 781, FS 2) made of petrified wood, a high-quality, semilocal material. Such points are believed to have been made by Numic peoples who migrated into southwestern Colorado several centuries after the departure of Puebloan peoples. This point was found 10 cm below the modern ground surface mixed with wall fall from structures on top of the butte and therefore was probably not deposited during the Puebloan occupation of Castle Rock. It seems more likely that this point is evidence of occasional visitation by Numic peoples several centuries after Puebloan peoples had left the region.

Form vs. Raw Material

¹⁰¹
Table 30 summarizes the raw materials out of which formal tools of various types were made. Because projectile points could have been used as weapons, and because point forms can vary across cultures, the presence of exotic projectile points might give some insight into the conflict that apparently ended the Puebloan occupation of Castle Rock (see "The Final Days of Castle Rock Pueblo"). Unfortunately, the only definite exotic projectile point dating to the Puebloan occupation that can be sourced to a particular area is the jasper Bull Creek point, which was probably made in southeastern Utah west of Cottonwood Wash. There are several points of petrified wood and agate/chalcedony that might be exotic, and a few additional points were at least made of nonlocal materials (obsidian, nonlocal chert/siltstone). But none of these points is also stylistically distinctive. The Bull Creek point indicates some interaction with Pueblo peoples of southeastern Utah, but not necessarily warfare. Among the local raw materials, Dakota quartzite was clearly favored over Morrison chert/siltstone for projectile points, despite the fact that Morrison Formation materials clearly dominate the overall chipped-stone assemblage.

¹⁰²
A number of the projectile points collected from Castle Rock are of styles that date from the Basketmaker II, Basketmaker III, Pueblo I, and Pueblo II periods. Since there is no additional evidence of occupation at Castle Rock prior to A.D. 1250, the older projectile point styles may represent heirlooms or points found by inhabitants of Castle Rock at previously abandoned sites and kept for nonfunctional purposes. Another possibility is that the large corner-notched points indicate continued used of atlatl darts during Pueblo III times. Further study of production evidence for different projectile point styles might clarify this issue.

Production Stage vs. Raw Material

¹⁰³
Several of the formal tools collected from Castle Rock are interpreted as projectile points in various stages of production (Table 31). These unfinished projectile points were classified according to Whittaker's (1994*1:199–206) scheme. Stage 1 refers to edged blanks, Stage 2 to preforms, and Stage 3 to refined but unfinished points. These objects indicate the range of raw materials that were made into projectile points at Castle Rock. The Stage 2 agate/chalcedony preform suggests that at least some of this material might have been directly procured by the inhabitants of Castle Rock. The Stage 3 obsidian preform is also interesting. It is clearly of a nonlocal material and might indicate that obsidian preforms were imported into Castle Rock to be finished by local knappers.

Ground-Stone Tools
Inventory
Artifact Category vs. Raw Material

¹⁰⁴
Table 32 summarizes the ground-stone artifacts collected from Castle Rock Pueblo according to the kind of stone out of which each was made (for definitions of the artifact categories used, see the on-line laboratory manual). The table shows that most ground-stone tools were made of locally available, relatively coarse grained material. The two manos of unknown chert/siltstone might have been used for polishing rather than grinding.

Artifact Category vs. Condition

¹⁰⁵
Table 33 summarizes the ground-stone artifacts from Castle Rock according to their condition. Relatively few fragmentary ground-stone artifacts were classified as abraders or one-hand manos, because these are difficult to distinguish from other ground-stone artifact categories on the basis of fragments.

Functional Analysis of Two-Hand Manos

¹⁰⁶
In ethnographic accounts (e.g., Cushing 1974*1:378–394; Kidder 1932*1:67; Lange 1968*1:116–117; Mindeleff 1989*1:211), two-hand manos were used in conjunction with slab metates and peckingstones for grinding dried corn (maize) kernels into flour. The mano was held in the hands parallel to the shoulders and moved back and forth across a flat metate surface by a kneeling person. The metates were often set into fixed, slab-lined mealing bins to keep the resultant flour contained. To maintain their effectiveness, the grinding surfaces of both manos and metates had to be periodically roughened, or "sharpened," using peckingstones. Corn-grinding tools appear to have been used the same way by inhabitants of Castle Rock, but the transverse cross sections of two-hand manos from the site are quite variable, suggesting that a number of slightly different techniques might have been used. Three possible sources of variation in two-hand mano cross sections are examined in this section. Data collected on two-hand manos from Castle Rock suggest that all three sources may have been involved to some extent.

Grinding Strokes

¹⁰⁷
Bartlett (1933*1) observed that, historically, Hopi women used two different grinding strokes, and these strokes produced different wear patterns. In one stroke, the mano was held flat against the metate and moved back and forth. Since there was more pressure from the palm of the hand on the downstroke than on the upstroke, this stroke resulted in more wear on the proximal edge of the mano (that is, the edge closest to the grinder's body). Without rotation, this stroke would eventually produce a grinding surface that ran at an angle through the natural strata of the stone. In the second stroke, a rocking motion was used and the mano was held at an angle, creating two adjacent, beveled grinding surfaces. Again, greater pressure was exerted during the downstroke, creating more wear on the proximal surface. Without rotation, this stroke would produce a larger grinding surface on the proximal edge than on the distal edge. The proximal surface would also eventually cut at an angle through the natural strata of the stone. In both strokes, proximal edge wear could be evened out by periodically rotating the mano 180 degrees.

Wear Management

¹⁰⁸
Adams (1993*3) argued that wear was managed on two-hand manos through a periodic, patterned flipping and rotating of the grinding surfaces. Two-hand manos are flipped by keeping the ends stationary and turning the mano over, and they are rotated by moving the ends 180 degrees. Consistent, periodic flipping and rotating in conjunction with a flat grinding stroke would ideally produce a mano with two opposite grinding surfaces, one on either side of the mano, running parallel to the natural strata of the stone. Flipping and rotating with a rocking stroke would ideally produce a mano with four grinding surfaces of relatively even width that cut at an angle through the natural strata of the stone.

Reduction Sequence

¹⁰⁹
Members of Crow Canyon's Native American advisory group rotate their manos to keep the grinding surface from getting too hot and melting into the flour. In addition to rotating, an alternative method for keeping grinding surfaces cool is to switch manos periodically, beginning with coarse-grained stones for heavy crushing and moving to finer-grained stones for fine grinding. This practice may also have occurred in the more distant past, because several manos are often found for each metate bin in grinding areas in Pueblo II and III sites in southwestern Colorado (Mobley-Tanaka 1997*1). If the use of multiple manos obviated the need for wear-management techniques, then a third possible source of variation in mano cross sections would be that manos went through a characteristic reduction sequence. In this scenario, variation in mano cross sections would result from two-hand manos breaking or becoming unusable at varying stages in the sequence. The reduction sequence model is presented in Figure 11.

Material Grade vs. Number of Grinding Surfaces

¹¹⁰
Table 34 summarizes the number of grinding surfaces that occur on two-hand manos from Castle Rock Pueblo according to the grade of the stone out of which each was made. The ethnographic literature on the Pueblos suggests that manos of these varying grades were used in different stages of the grinding process. Manos of the coarsest material, conglomerate sandstone, were used first to crush the corn kernels, followed by manos of increasingly fine-grained stone. There is a possible tendency for manos of conglomerate sandstone, a coarse material, to have fewer grinding surfaces than manos of finer-grade stone.

Condition vs. Number of Grinding Surfaces

¹¹¹
Table 35 summarizes the condition of two-hand manos with varying numbers of grinding surfaces. Most of the two-hand manos were broken. Nine of the 12 complete manos were found in collapsed structures or on structure floors and were probably in use at the time Castle Rock Pueblo was abandoned. Most of the complete manos have one or two opposing grinding surfaces, and very few have adjacent surfaces that could have emerged from a rocking stroke or periodic rotation. It is unclear from these data whether routine flipping or a reduction sequence created manos with two opposing surfaces. There are no complete manos with four grinding surfaces. If flipping was used in conjunction with rotation as a means of wear management, one would expect to see at least some complete manos with four surfaces in early stages of wear.

Condition vs. Material Grade

¹¹²
Table 36 summarizes the condition of two-hand manos of varying raw material grades. It shows that there is no relationship between mano grade and condition at the time of discard. Broken and complete manos occur in similar proportions across raw material grades, suggesting that manos of all grades were equally susceptible to breakage. That different grades occur in the same proportions among both complete and broken manos suggests that these proportions reflect the relative numbers of manos of varying grades in use at any one time at Castle Rock. If so, many more manos of medium-grained sandstone were used than of either coarser or finer grained material. The implications of this pattern for the corn-grinding process are unclear. It may indicate that medium-grained manos wore down more quickly than manos of other materials, and therefore more were needed at any one time for alternating use. It could also indicate that coarse- and fine-grained manos were used less intensively during the grinding process. Both possibilities would have implications for the deposition rate of medium-grained manos.

Number of Surfaces Parallel to Grain of Stone

¹¹³
Table 37 summarizes the number of surfaces on two-hand manos and their relationships to the grain of the stone from which each was made. Manos were made of sedimentary rock containing laminated layers. The rock tends to fracture along these layers, creating stones that naturally have two parallel, relatively flat surfaces. Manos were made by flaking the edges of such pieces to the appropriate shape, so that the naturally occurring flat surfaces could be used as grinding surfaces. In most cases the grinding surfaces on manos with one or two opposite surfaces run parallel to the laminated layers, or "grain," of the original stone. A flat stroke must have been used with these manos. It is impossible to determine whether manos with two opposite surfaces parallel to the grain were periodically flipped or whether the two surfaces were created in a sequence. It also is impossible to determine whether manos with one or more surfaces parallel to the grain were periodically rotated to keep the grinding surfaces parallel, although this seems likely.

¹¹⁴
Manos with two adjacent, three, or four surfaces all have adjacent grinding surfaces on one or both sides of the mano. Two adjacent grinding surfaces could both come to cut across the grain of the original mano if rotation occurred consistently or if a rocking stroke was used. There are several manos from Castle Rock with two adjacent grinding surfaces on which one of the surfaces runs parallel to the grain of the stone and the other cuts across the grain. This wear pattern could have emerged through a rocking motion without rotation, such that the surface proximal to the grinder wore through the layers faster than did the distal surface. Alternatively, the observed pattern could be the product of a reduction sequence in which the surface parallel to the grain developed first, as the result of a flat stroke, and the surface that cut across the grain developed second, as the result of the thinner, worn-down mano being held at an angle to the metate. If the mano broke early in this reduction sequence, it would exhibit one adjacent surface running parallel to the grain of the stone and another cutting across the grain. Two adjacent surfaces that both cut across the grain of the original mano would develop only if a mano lasted long enough before breaking.

Widths of Adjacent Grinding Surfaces

¹¹⁵
Figure 12 summarizes differences in width between adjacent grinding surfaces on two-hand manos from Castle Rock Pueblo. Twelve of the 25 cases of adjacent grinding surfaces exhibit width differences equal to or greater than 2 cm. These differences seem relatively large for a wear-management regime in which manos were consistently rotated to even out use wear. The differences are more consistent with rocking grinding strokes without rotation or sequential development using a flat stroke.

Summary of Functional Analysis

¹¹⁶
The functional analysis indicates that broken manos tend to have more grinding surfaces than complete manos, that often one of each pair of adjacent grinding surfaces runs parallel to the original grain of the mano, and that there are differences in the widths of adjacent grinding surfaces. These observations suggest that two-hand manos at Castle Rock Pueblo were not consistently flipped and rotated as part of wear management. The other two models are better supported by the available data, although these data are insufficient to enable us to distinguish between the two models. More detailed study of mano "sharpening" might help distinguish mano cross sections resulting from a reduction sequence from those resulting from different grinding strokes. If manos followed a characteristic reduction sequence, then one might expect that only the surfaces in use during the final stages of reduction would have been kept sharpened. Comparison of two-hand mano assemblages from Castle Rock and other sites in the Sand Canyon locality might provide evidence of intensified food preparation associated with village formation and increasing levels of communal ritual (see paragraphs 41–66).

Pecked and Polished Stone Tools
Polished Igneous Stones and Polishing/Hammerstones

¹¹⁷
Table 38 catalogs all the polished igneous stones and polishing/hammerstones collected from Castle Rock (for definitions of these artifact categories, see the on-line laboratory manual). It is intriguing that three polished igneous stones were found on the floor of the square kiva, Structure 103. Although the uses of polished igneous stones are unknown, the fact that several were found in the same structure might indicate that one of the inhabitants of that structure specialized in making them, or that activities requiring polished igneous stones occurred there with unusually high frequency.

Axes and Mauls

¹¹⁸
Hafted stone axes and mauls are important artifact categories for the interpretation of Castle Rock Pueblo because, in addition to being used for activities such as wood chopping and stone quarrying and shaping, they might also have been used as weapons during the conflict associated with the abandonment of the village. Table 39 lists these items along with their condition, material, and form, measurements and use-wear data, and assessments of their likely uses. For definitions of the ax and maul categories, see the on-line laboratory manual. Minimum measurements recorded under "Comments" indicate the minimum possible measurement for a given dimension based on a fragment.

¹¹⁹
Woodbury (1954*1) reviewed ethnographic data on the traditional uses of axes and mauls among the Pueblos and found that the types of axes used as weapons were relatively small, lightweight, and well balanced. Morris (1924*1) found a few such axes at Aztec Ruin in association with a thirteenth-century burial that clearly was that of a warrior. The individual was a large male, over six feet tall, who was accompanied by a basketry shield, several knives, a flintknapping tool kit, a wooden sword, and several small, single-bitted axes.

¹²⁰
Unfortunately for archaeologists, most of the axes and mauls discarded at Castle Rock appear to have been worn out by use, and few were found in their original, undamaged state. Of those whose likely uses could reasonably be inferred on the basis of form, size, and evidence of use wear, only one or two might have been made specifically as weapons. Given the evidence for a violent conflict at Castle Rock, it is somewhat surprising that so few were found. If any surviving inhabitants of Castle Rock felt threatened by additional violence, however, they would likely have scavenged all usable axes and carried them with them for protection. Alternatively, victorious attackers might have taken usable axes with them after the battle. It would be interesting to examine the raw materials of axes found in early-fourteenth-century sites in the Rio Grande valley and on the Hopi mesas to see whether any were made of materials that could be traced to the Mesa Verde region.

¹²¹
Most axes and mauls were made of Morrison Formation materials. The evidence of battering on maul heads and of large flakes removed from them suggests that mauls were used for quarrying and shaping stone. Large flakes were also removed from the bit ends of axes, but battering damage, such as occurs from stone-on-stone contact, was rare, suggesting that heavier axes were used for chopping and splitting wood. On the basis of replication experiments, Mills (1987*1) inferred that axes might also have been used for chopping sagebrush at ground level, possibly as a step in field clearance.

>Other Modified Stones and Minerals
Modification vs. Raw Material

¹²²
A wide variety of stones and minerals modified through polishing, grinding, flaking, battering, and/or fire alteration were found at Castle Rock. Data for these objects are summarized in Table 40 according to the kind of modification present (classification scheme after Pierce [1999*1]) and the raw material out of which each was made.

¹²³
Most edge-shaped slabs were made of sandstone and may have been architectural elements, portions of grinding bins, hatch covers, and so forth. Fire-altered rocks are most likely spalled fragments of burned building stones rather than evidence of cooking inside watertight baskets. Flaked/battered stones and minerals are primarily large, angular chunks and subangular stream cobbles that were initially unshaped but from which flakes had been removed through use as choppers, scrapers, and crushers. The "ground" category refers to miscellaneous pieces of ground, abrasive stone that could not be classified into the other ground-stone tool categories. Pigment stones are of iron-rich material and have one or more abraded surfaces resulting from grinding to obtain the pigment. Since little mineral paint was used in pottery painting at Castle Rock, it is unclear how this pigment was used. The "polished" category represents the most common type found at Castle Rock. It includes objects that are probably ornament fragments and blanks lacking signs of perforation, as well as several larger fragments of polished shale. Stone disks were probably used as lids for pottery vessels.

Unmodified Stones and Minerals
Inventory by Raw Material

¹²⁴
Unmodified stones and minerals were collected when, in the excavator's opinion, they were objects that did not occur naturally at Castle Rock Pueblo and therefore must have been collected and carried to the site by its inhabitants (Table 41). In many cases unmodified stones and minerals may represent unused raw materials for pottery temper (igneous rock), chipped-stone tools (agate/chalcedony, Morrison chert/siltstone, Morrison quartzite, unknown chert/siltstone), or pigment (concretions or pigment). Fossils, concretions, and petrified wood might have been collected for their spiritual value. Concretions reminiscent of real-world forms, especially of animals, are known to have been kept for spiritual reasons among the historic Pueblos (Jeancon 1923*1:65–68). Also, since no marine molluscan species occur naturally in the Pueblo area, and people's only knowledge of such creatures would have been through imported objects made of shell, fossil shells of marine species probably had special significance.

Animal Remains
Distribution of Animal Remains by Study Unit

¹²⁵
Table 42 summarizes the study units in which animal bones, gizzard stones, and eggshells were found. Animal bones were collected from nearly every excavation unit at Castle Rock and clearly indicate the butchering of animal carcasses for meat. Animal bones are discussed more fully in "Faunal Remains." There is ample evidence that domestic turkeys were raised in the Sand Canyon locality (Munro 1994*1). Eggshells probably derived from turkey raising or from the collection of eggs of other bird species for food. Identifiable gizzard stones probably derived from turkeys. Since gizzard stones are not passed by turkeys until they are too small to be captured by a 1/4-inch screen, the presence of identifiable gizzard stones is evidence of turkey butchering at Castle Rock. Also, the fact that several gizzard stones appear to be worn-down pieces of debitage from chipped-stone tool making suggests that turkeys were raised at the site and may have fed themselves on food remains thrown onto middens.

Worked Bone Tools
Inventory by Type and Condition

¹²⁶
Table 43 summarizes the worked bone tools collected from Castle Rock Pueblo (for definitions of the categories used, see the on-line laboratory manual). Driver (1999*2) discussed the morphology of bone awls in the Site Testing Program sample from Castle Rock more fully.

¹²⁷
Several authors have discussed a distinctive type of wear on bone awls consisting of transverse notches or grooves, sometimes called "weaving grooves" (Kidder 1932*1). No examples of such grooves were found among the bone awls from Castle Rock. Unfortunately, ethnographic information about the causes of transverse grooves is sketchy, and replication experiments have not yet determined which weaving techniques, if any, produce these grooves (Bullock 1992*1). So it is not yet possible to determine the specific activities represented by weaving grooves, if any, and whether the absence of weaving grooves at Castle Rock is significant.

Objects of Nonlocal Materials
Catalog of Stone and Shell Objects

¹²⁸
Table 44 catalogs all stone and shell artifacts made of raw materials that definitely do not occur in southwestern Colorado. Pottery sherds of nonlocal manufacture are discussed in paragraphs 84–90. Shell species were identified by G. Timothy Gross (see Gross [1999*1] for a more complete discussion of the shell objects from the Site Testing Program sample from Castle Rock). The material out of which each artifact was made, the closest possible source of each material, and summaries of the provenience of each item are given in the table. Half of all the objects of nonlocal material were found within Roomblock 300, on the north side of the village. Possible interpretations of this concentration are discussed in paragraph 168.

Provenance of Nonlocal Stone, Shell, and Pottery Objects

¹²⁹
For most archaeologists, provenance refers to the place where an artifact is inferred to have been made, based on characteristics of the material out of which it was created. This is in contrast to provenience, which refers to the place where an artifact became incorporated into the archaeological record (Blinman and Wilson 1993*1). Table 45 summarizes the raw materials and closest possible provenances of stone, shell, and pottery objects made of nonlocal materials found at Castle Rock. Most of the nonlocal objects appear to have come from other parts of the Puebloan world to the south, east, and west. In addition, items made of shell species that occur no closer than the Gulf of California and the Pacific coast suggest that Castle Rock Pueblo participated in indirect trade networks that went far beyond the Pueblo area. There is no evidence, however, of exchange with more northerly peoples such as the Fremont of Utah and Colorado.

¹³⁰
It is also important to recognize that only 43 of the more than 44,000 identified artifacts from Castle Rock—less than 0.1 percent—were made of definite nonlocal material. This suggests that the overall intensity of exchange relationships with people outside of southwestern Colorado was quite low. That such low levels of long-distance exchange occurred immediately prior to the final migrations of Pueblo people out of southwestern Colorado may have implications for the nature of this migration. It has recently been proposed that the final emigration from the Mesa Verde region took the form of a "migration stream," with small family groups moving to new areas over several generations (Wilshusen and Duff 1999*1). Studies of migration streams in the modern world suggest that small family groups do not usually immigrate to places about which they have limited information. In most cases, friends and relatives who have already moved provide information to potential migrants about attractive destination areas (see Duff 1998*1:32–34). In his review of recent migration literature, Duff (1998*1:33) concluded that long-distance trade relationships were probably a primary source of information about potential destination areas in the past. So if the final emigration took the form of a migration stream, one would expect to find increasing levels of long-distance exchange between the Mesa Verde region and destination areas during the final periods of Puebloan occupation.

¹³¹
The evidence for long-distance exchange reviewed for Castle Rock Pueblo and other southwestern Colorado sites in paragraphs 84–90 suggests that in fact the opposite happened. Over time, inhabitants of the Mesa Verde region seem to have had fewer and fewer contacts with Pueblo people living in other parts of the Southwest. If this pattern is also evident in other late-thirteenth-century villages in southwestern Colorado (studies are currently in progress), it might provide contrary evidence for the migration stream model, or at least evidence that needs to be accounted for. Alternatively, if other late-thirteenth-century villages provide more abundant evidence of long-distance exchange relations, then the rare evidence for such contacts at Castle Rock might somehow relate to the ultimate fate of its inhabitants.

Objects of Personal Adornment
Catalog of Beads, Pendants, and Tubes

¹³²
Table 46 summarizes analysis and provenience information for objects of personal adornment found at Castle Rock (for definitions of the artifact categories used, see the on-line laboratory manual). The majority of beads and pendants were complete, although it is likely that fragmentary pendants were classified as shaped sherds (see paragraphs 36–37). Most of the objects of personal adornment were found in contexts that suggest accidental loss rather than discard in middens. Several objects of personal adornment were found in the fill of Structure 302, and a large number were found in the fill of Structure 304, all in deposits that also contained informally deposited human remains. It is possible that some of these items were worn by these individuals at their time of death. Overall artifact densities, however, were anomalously high in the fill of Structure 304 (see paragraphs 166–167), raising the possibility that this apparent concentration was simply a function of sample size.

Summary of Raw Materials

¹³³
Objects of personal adornment (Table 47) were made of many different raw materials, many of which were unusual, precious, and/or nonlocal to southwestern Colorado (see paragraphs 129–131). Unusual or rare materials are selectively used for personal adornment in many cultures throughout the world.

Probability Sample
Prior vs. Posterior Stratification

¹³⁴
Castle Rock Pueblo was first investigated as part of the Sand Canyon Project Site Testing Program. At that time, the site was divided into six sampling strata, and a stratified random sample of 1-by-1-m units was excavated (Kleidon 1999*1). This probability sample was intended to collect a representative sample of artifacts from the site and to estimate total numbers of discarded artifacts as a step in estimating the length of occupation (see Varien 1999*1). The six sampling strata were defined on the basis of surface expressions at the time the original site map was made. Owing to the setting of Castle Rock Pueblo on talus slopes around a large butte, however, surface expressions were variable from place to place. As a result, many probability squares did not uncover what the sampling stratum to which each was assigned suggested they would (compare site maps Database Map 510 [Topography and Surface Remains], Database Map 511 [Probability Sampling Plan], and Database Map 509 [Major Cultural Units]). For example, many units in the surface room sampling stratum did not encounter rooms, and many units in the midden sampling stratum did not encounter midden deposits. An efficient stratified sample is one that produces greater homogeneity in artifact totals for sampling units within each sampling stratum than across all sampling units. This becomes less likely when units within a sampling stratum contain mixtures of different kinds of deposits, such as pit structure deposits and peripheral, nonstructural deposits. Because this often occurred in the sampling strata defined for Castle Rock, the probability sample was not ideal as originally defined.

¹³⁵
Poststratification, or stratification after selection of the sampling units, is one method for dealing with probability samples for which the stratifying variable (i.e., cultural feature type) is poorly known prior to sample selection (Kish 1965*1:90). However, it is statistically sound to poststratify only a simple random sample, and the Castle Rock Pueblo probability sample was designed as a stratified sample. So before poststratification can be used, a judgment must be made that the original stratified sample, taken as a whole, could reasonably have been selected as a simple random sample if this strategy had been used in choosing the units to be excavated.

Comparison of Sampling Intensities

¹³⁶
A simple random sample is one in which each sampling unit has an equal chance of being selected for investigation. One way to assess the degree to which the prior stratified random sample for Castle Rock approaches a simple random sample is to consider the sampling intensities for each stratum in the prior sample. These intensities represent the fraction of each stratum that was excavated. Because all units within each stratum had an equal probability of being selected, relatively even sampling intensities across strata—that is, a stratified sample in which relatively even proportions of each stratum were investigated—would produce a sample that could reasonably have been selected using simple random sampling. Sampling intensities for the prior strata are given in Table 48; they vary between 0.005 and 0.025. On the whole, slightly more than 1 percent of the site was excavated as part of the probability sample, and although units in the surface room stratum were five times more likely to be chosen than units in the inner periphery stratum, no unit in any stratum was more than 2.5 percent likely to be chosen. If units in some strata were, say, 10 percent likely to be chosen, while units in other strata were only 0.1 percent likely, it would be unreasonable to consider the prior stratified sample comparable to a simple random sample. But because no one unit was significantly more likely to be chosen than any other in the sample we have to work with, it is reasonable to conclude that the prior stratified sample could have been chosen as a simple random sample. As a result, poststratification of the site and the sample, based on knowledge of the site at the conclusion of all excavations (both testing and intensive excavations), is reasonable. Revised sampling intensities for the posterior stratification are included in this table. The poststratification procedure is described in paragraph 137.

Definition of Posterior Stratification

¹³⁷
In an attempt to improve the statistical results obtained from the Castle Rock Pueblo probability sample, the boundaries of the six sampling strata used in the prior stratification were redrawn on the site map using the results of all excavations (testing and intensive excavations) as well as surface expressions. Each 1-by-1-m unit excavated for the prior sample was reassigned to one of these six sampling strata on the basis of their new boundaries, which incorporated the field archaeologists' interpretations for each excavated unit. Database Map 539 illustrates the refined boundaries of the six sampling strata. Resulting sampling intensities for these strata are presented in Table 48.

Correspondence Table for Probability Excavation Units

¹³⁸
Table 49 lists the prior and posterior stratum to which each unit in the probability sample was assigned. The coordinates of the southwestern corner of each probability unit on the site grid are also given.

Comparison of Excavation Unit Assignments

¹³⁹
Table 50 compares the number of probability units assigned to each stratum in the prior and posterior stratifications. If there were no changes in assignment between the prior and posterior stratifications, all sampling units would be accounted for along the diagonal cells of the table. That this does not occur shows how ineffective the prior stratification was at predicting the kinds of features that would be found under the modern ground surface. Prior Stratum 1 units, for example, ended up exposing pit structures, middens, and courtyards in addition to surface rooms. Since different depositional settings tend to exert a strong influence on artifact totals, the prior stratification scheme is likely to be less efficient than the posterior stratification.

Comparison of Sherd Weights by Stratum

¹⁴⁰
Table 51 presents means and standard deviations for corrugated jar, white ware bowl, and white ware jar sherd weights for each of the six prior and posterior sampling strata. Sherd size clearly varies across sampling strata, probably owing to depositional and postdepositional processes. As a result, the better estimator of pottery deposition is weight, not count. Because stone artifacts do not fragment the same way pottery does, count is a more interpretable measure of stone artifact deposition.

Probability Samples

¹⁴¹
Table 52 presents the prior and posterior probability samples used in the following sections. In order to create categories that contain a relatively large number of artifacts, certain categories have been grouped together. Formal chipped-stone tools include bifaces and projectile points, and expedient chipped-stone tools include modified flakes and other chipped-stone tools. All ground-stone tools and all worked bone tools, respectively, are also considered together. The values for pottery categories are total weights in grams; values for all other artifact categories are counts. Notice that the total numbers of artifacts in the prior and posterior probability samples are identical. This is because material from the same 64 probability units is used for both samples. The only difference between the prior and posterior probability samples is the allocation of excavation units to the six sampling strata. Also notice that the sampling unit is the excavation unit. The depths of archaeological deposits in different areas of Castle Rock are not taken into account.

Comparison of Stratum Standard Deviations for Common Artifact Categories

¹⁴²
An effective stratification procedure is one that produces smaller standard deviations for a measure across sampling units within each stratum than would be obtained from a simple random sample of the site as a whole. The upper register of Table 53 shows that in most cases, standard deviations are lower in the posterior stratification. The only artifact category for which this is not the case is chipped-stone debris, and the only stratum for which posterior standard deviations are consistently greater is Stratum 5, midden deposits.

¹⁴³
Why midden deposits have greater posterior standard deviations is unclear. One possibility derives from the fact that midden deposits generally contain the highest artifact densities of any archaeological deposits. Given this, variation in the depths of midden deposits in different sampling units may have a more significant effect on artifact totals than variation in the depths of other site deposits. This could account for greater standard deviations for midden deposits than for all other sampling units in the posterior stratification. In this scenario, prior Stratum 5 standard deviations would be lower simply because many of these units did not actually uncover midden deposits, making the prior Stratum 5 sample closer to a simple random sample.

¹⁴⁴
The lower register of Table 53 shows that stratum standard deviations are greater than the whole site standard deviation in 11 cases in the prior stratification but in only five cases in the posterior stratification. This suggests that the posterior stratification is more efficient and will produce correspondingly more precise estimates of total discard.

Comparison of Variances for Common Artifact Categories

¹⁴⁵
The efficiency of a stratification scheme is measured as the ratio of the stratified variance to the variance obtained by taking the entire sample as a simple random sample (SRS). The lower the ratio, the more efficient the scheme. Table 54 presents prior and posterior stratified and SRS variances for four common artifact categories and computes the efficiency of the prior and posterior stratification schemes. These figures show that the posterior stratification is significantly more efficient than the original stratification scheme used at Castle Rock Pueblo. The ratio of variances for the prior and posterior stratifications also shows that the posterior stratification consistently produces lower variances than the prior stratification. The increased efficiency of the posterior stratification should lead to more precise estimates of population totals.

Estimation of Artifact Population Totals
Comparison of Population Total Estimates, Prior vs. Posterior Stratification

¹⁴⁶
Table 55 presents point estimates and confidence intervals for the total discard of common artifact categories at Castle Rock Pueblo, on the basis of prior- and poststratified random samples. As an example, using the poststratified sample, we can be 95 percent confident that between 1,231 and 2,213 peckingstones were discarded during the occupation of Castle Rock. For most artifact categories, the posterior stratification produces slightly lower point estimates of total discard than does the prior stratification. In addition, the posterior confidence intervals are significantly smaller in most cases, indicating that the posterior estimates are indeed more precise, as the analysis of variances in paragraph 145 suggested.

Estimating the Occupation Span of Castle Rock Pueblo

¹⁴⁷
More precise estimates of artifact population totals should enable more precise estimates of site occupation span. Varien (1999*1:Chapter 4) and Varien and Mills (1997*1) developed and justified procedures for estimating site occupation span from the total discard of cooking pots in the form of corrugated jar sherds. Varien's method is to divide estimates of the total discard of corrugated jar sherds by an estimate of the accumulation rate for corrugated jar sherds derived from strong archaeological cases; the estimates are expressed in grams of corrugated pottery per household per year. This results in an estimate of the total household-years of occupation at a site, which is then divided by an estimate of the number of households that resided at the site, based on architectural evidence, to yield a point estimate and confidence interval for the occupation span of the site, expressed in years.

¹⁴⁸
These methods are used in Table 56 to estimate the total household years of occupation and the occupation span of Castle Rock Pueblo. The prior and posterior probability sample estimates are based on the Castle Rock Pueblo artifact database as of June 1998, whereas Varien's 1999 estimate (after Varien 1997*1) is based on the database as it existed in 1995. Varien considered 13 of the 16 kiva suites at Castle Rock to have been residential, but judging from the analysis of midden assemblages in paragraph 161, it appears more likely that all 16 kivas were components of residential household architecture. All three estimates use the accumulation rate developed by Varien and Mills (1997*1) on the basis of the Duckfoot site (Lightfoot 1994*1). Table 56 presents a simplistic model of site occupation span based on the assumption that all kiva suites were built and initially occupied during a single year and that all kiva suites were occupied until the entire site was abandoned. Note that the 95 percent confidence interval derived from Varien's data encompasses the posterior point estimate and confidence interval ranges. This is an expected result, since all these estimates are based on the same probability sample from the same site. The poststratification procedure therefore does not invalidate previous estimates but merely refines them, producing a more precise estimate. For example, the 95 percent confidence interval for the posterior estimate encompasses the same number of years as the 80 percent confidence interval for Varien's data.

Estimating the Year of Abandonment

¹⁴⁹
The site occupation spans shown in Table 56 are based on two simplifying assumptions—first, that all kiva suites were built and occupied during the same year, and second, that all kiva suites continued to be occupied until the site was abandoned. These assumptions are reasonable for small hamlets, which contained only one or two kiva suites, but for a larger village such as Castle Rock they are less realistic. The settlement history of Castle Rock, and especially the construction and abandonment of individual kiva suites, could have had a significant effect on the rate at which corrugated pottery accumulated during the occupation.

¹⁵⁰
Fortunately, the construction history and abandonment of Castle Rock are relatively well defined for a site of its size, and this enables a slightly more realistic assessment of its occupation span. The abandonment of Castle Rock appears to have followed on the heels of a violent conflict. All of the kivas at Castle Rock have been tested, and all but Structure 304 appear to have been in use until the site was abandoned (see "The Final Days of Castle Rock Pueblo"). Structure 304 appears to have fallen out of use before the attack, but since it could not have been built before A.D. 1274 and the latest tree-ring date from any site in the Mesa Verde region is A.D. 1280, it was probably abandoned no more than a few years before the battle. So it is reasonable, even if a bit imprecise, to model the abandonment of Castle Rock as having occurred during a single year.

¹⁵¹
Construction dates for eight of the 16 kiva suites at Castle Rock have been estimated on the basis of tree-ring data from one or more structures within each suite (see "Chronology"). These dated kiva suites appear to have been built over approximately an 18-year period, between A.D. 1256 and 1274, suggesting that the population of Castle Rock grew accretionally during this time. If so, it is unrealistic to use a single start date for all kiva suites at the site. What we really want to do is allocate household-years of trash deposition to various kiva suites on the basis of the pattern of growth suggested by tree-ring data. In an attempt to provide a slightly more realistic model, the growth pattern suggested by the dated kiva suites has been used to model the settlement history of Castle Rock.

¹⁵²
In Table 57, the cumulative number of kiva suites inferred to have been built and occupied by a given date is listed by year, beginning with A.D. 1256, and this figure is multiplied by two, so that by A.D. 1274 all 16 kiva suites have been accounted for. Using this settlement model, the cumulative household-years of occupation have been added year by year until the estimates based on pottery accumulations are reached.

¹⁵³
The results presented in the table show that, on the basis of this settlement model, the posterior point estimate was reached in A.D. 1293, and the 80 percent confidence interval spans the period from 1288 to 1299. If Castle Rock grew more quickly than the dated kiva suites suggest, then these dates would be pushed a little earlier in time; if the village grew more slowly, the dates would be pushed slightly later. These data suggest that the abandonment of Castle Rock occurred during the late A.D. 1280s or 1290s, approximately 15 to 20 years after the latest tree-ring date from the site, A.D. 1274vv. Pottery accumulations suggest a somewhat larger gap between the latest tree-ring date and abandonment than many previous studies suggest is typical (e.g., Ahlstrom 1985*2). Also, the latest tree-ring-dated construction timbers from Puebloan sites in the Mesa Verde region overall (from cliff dwellings in Mesa Verde National Park) date only to A.D. 1280. The late date suggested by pottery accumulations at Castle Rock Pueblo is therefore somewhat out of line with the regional tree-ring record. Varien (1999*1:Chapter 4) discusses several possible sources of imprecision when using pottery accumulations to estimate site occupation span. Possible explanations for the discrepancy identified in this study include (1) that the accumulation rate used is slightly inaccurate; (2) that Castle Rock Pueblo grew more quickly than the model used here suggests; (3) that little new construction took place during the final 20 years of occupation; and (4) that some of the cooking pots deposited at the site were left by periodic or temporary visitors who were not permanent residents (paragraphs 41–66 present evidence that Castle Rock Pueblo functioned as a central place for communal feasts).

¹⁵⁴
The fourth possibility does not seem especially likely, because ratios of cooking-pot sherds to other classes of material culture, such as chipped-stone tool manufacturing debris, are consistent across hamlets and villages in the Sand Canyon locality. This suggests that most artifact deposition in Sand Canyon locality sites occurred in the context of daily activities performed by permanent residents (Varien 1999*1:80–85). If deposition by nonresidents in the context of communal feasts made a significant contribution to the total accumulation of cooking-pot sherds at Castle Rock, then one would expect the ratio of cooking-pot sherds to chipped-stone debris to have been affected. The second and third possibilities argue that our knowledge of the settlement history of Castle Rock remains too limited to produce an occupation span estimate as precise as is possible for smaller and simpler hamlets. This seems possible, but regardless of the reasons for the discrepancy, the fact that pottery accumulations do provide occupation span estimates that are within the realm of possibility, even for such a complex site as Castle Rock, is a testament to the basic validity of accumulations research.

Intrasite Analyses
Midden Composition

¹⁵⁵
One of the primary research domains specified in the Sand Canyon Project research design was intracommunity differentiation in terms of status, function, and so forth (Lipe 1992*3:3–5). The smallest social unit below the community that is consistently expressed in Mesa Verde Puebloan archaeological sites is the household, which can be defined architecturally as a kiva suite with a small kiva and associated above-ground rooms and trash (Lightfoot 1994*1:Chapter 7; Lipe 1989*1; Ortman 1998*2; Varien 1999*1:Chapter 1). Kiva suite (or unit pueblo) architecture is quite variable at Castle Rock: Suite 103 contains a rare square kiva; Suite 105 contains what may be an unusually large kiva; Suite 402 contains a kiva inside a D-shaped enclosure; and the number of surface rooms in kiva suites appears to have varied significantly across the site. Whether this architectural differentiation correlates with social or functional differentiation across kiva suites is an important question. This possibility is examined in this section by comparing the trash deposits associated with kiva suites. If social and/or functional differentiation existed across kiva suites at Castle Rock, then one might expect different mixes of activities to have occurred in different areas of the site and for these different activities to be reflected in the trash they generated.

Assignment of Midden Deposits to Kiva Suites

¹⁵⁶
Midden deposits were identified during the Castle Rock excavations as areas of high artifact and ecofact density, usually in a matrix of gray, ashy soil. In Table 58, all excavation units containing such deposits are assigned to kiva suites on the basis of spatial proximity and topography. Each suite is named after the structure number of its kiva. The basic assumption in making these assignments is that trash generated by activities in kiva suites would generally have been tossed from the kiva courtyard southward, downslope, or away from the associated roomblock and kiva. It is also assumed that representative samples of all durable artifactual trash generated by activities in kiva suites ended up in midden deposits. That is, trash of all sorts was unreflectively deposited in the trash mound associated with each kiva suite. The table presents these assignments by kiva suite and excavation unit, along with the depth in meters and density per cubic meter of common artifact types (i.e., artifact categories described in paragraph 141) in these deposits. There are no excavated midden deposits that can be associated with Kiva Suites 102, 107, 204, and 406. These suites most likely produced midden deposits that were not specifically identified or sampled in the Castle Rock Pueblo excavations.

Depth vs. Density of Midden Deposits by Excavation Unit

¹⁵⁷
Figure 13 compares the relationship between the depth and artifact density of midden deposits at Castle Rock by excavation unit; it shows that there is only a modest correlation between them. The total number of artifacts deposited in a midden area depends on the duration and intensity of occupation, but denser deposits are not necessarily deeper. Thus, the abundance of various artifacts by volume in a midden deposit may relate more to depositional environments than to differences in artifact deposition per se. A measure that should be less influenced by depositional and postdepositional processes is the relative frequency of different artifact categories in a deposit. If different mixes of activities occurred in different places, this should be reflected in different mixes of artifacts in midden assemblages.

Kiva Suite Midden Assemblages

¹⁵⁸
Table 59 presents counts of common artifacts by category in midden deposits associated with various kiva suites at Castle Rock Pueblo. The artifact categories are the same as those used in paragraph 141—that is, formal chipped-stone tools include bifaces and projectile points, and expedient chipped-stone tools include modified flakes and other chipped-stone tools. All ground-stone tools and all worked bone tools, respectively, are also considered together. Counts rather than weights of pottery sherds were used for this analysis in order to increase the interpretability of relative frequencies across all artifact categories in paragraph 159. The use of counts rather than weights is also reasonable because sherd size is unlikely to vary as significantly across midden deposits as it does across archaeological contexts, as was discussed in paragraph 140.

Relative Frequencies of Common Artifact Categories in Midden Assemblages

¹⁵⁹
Table 60 presents relative frequencies of common artifacts by category in midden deposits associated with kiva suites at Castle Rock Pueblo, expressed as a percentage of the artifacts tabulated for each kiva suite in Table 59.

Box Plots of Artifact Frequencies across Kiva Suites

¹⁶⁰
Figure 14 examines the relative frequencies of common artifacts by category across kiva suites at Castle Rock Pueblo. The percentages of these artifacts were converted to Z-scores across kiva suites to facilitate comparison, because some categories are much more common than others. Z-scores rescale the values of a distribution in such a way that the mean value equals 0 and the standard deviation equals 1. The boxes represent the midspread (middle 50 percent of cases) of the distribution for each artifact category; the solid line represents the median value; and the tails represent the range of cases, excluding outliers. Outliers are values for a given artifact category that fall more than 1.5 box lengths away from the boundaries of the box. In other words, outliers represent trash deposits with unusually high relative frequencies of a particular artifact category. Notice that all the outlying cases identified in this analysis are at the high ends of the distributions. There are no kiva suites for which associated trash deposits contain anomalously low numbers of artifacts in any of these categories. The same set of outliers is identified in box plots of raw frequencies as in these plots of Z-scores.

¹⁶¹
There is obviously some variation in midden composition at Castle Rock, but there is no clear, interpretable relationship between architectural variation and midden artifact variation. The unusually high frequency of chipped-stone debris associated with Suite 402 might indicate an unusual amount of stoneworking in the area of the D-shaped enclosure, and one might be tempted to interpret this pattern as evidence of male-dominated activities in this area. However, both men and women made and used chipped-stone tools for various activities, and no stone tool types are unusually common in this area, making any gender identification tenuous. Also, chipped-stone debris is not strongly associated with Suite 402 in the multivariate analysis presented in paragraphs 162–165. In addition, Suite 103, containing a square kiva, and Suite 105, containing a possibly "oversized" kiva, also appear to be associated with generally unremarkable trash. The lack of correspondence between architectural variation and midden composition makes it difficult to rule out random sampling error as a cause of the observed variation in artifact frequency, and it suggests that kiva suites with distinctive architecture were not necessarily functionally distinctive.

Correspondence Analysis of Artifact Counts by Kiva Suite

¹⁶²
Correspondence analysis (CA) is a multivariate analytical technique that produces the best possible projection of multivariate data onto two axes, so that the degree of relationship between cases and variables can be examined visually (Baxter 1994*1:Chapter 5). Figure 15 presents results of a CA of the data in Table 59. Counts are appropriate input data for CA because the technique takes sample size into account in such a way that larger collections and more common categories exert a greater effect on the placement of variables and cases on the resultant axes. The first two axes produced account for more than 80 percent of the total variation or inertia in the input data.

¹⁶³
Two of the outliers identified in Figure 14 are also expressed in the CA results as correspondences in the placement of certain artifact categories and kiva suites. First, high frequencies of turkey gizzard stones are clearly associated with trash downslope from Suite 304. Turkey gizzard stones are also associated with Suite 302, but this suite does not also appear as an outlier in Figure 14. These data suggest that turkeys were often butchered and possibly also raised on the north slope of the butte in the center of Castle Rock. Second, white ware jar sherds are clearly associated with trash near Suite 405. The trash in this area is relatively near an ancient reservoir that lies approximately 50 m to the north. Excavations at ancient reservoirs on Woods Mesa, approximately 15 km north of Castle Rock (Wilshusen et al. 1997*1), and at Mummy Lake in Mesa Verde National Park (Breternitz 1999*1:25) produced pottery assemblages that were also dominated by white ware jar sherds, presumably because such vessels were used for collecting and storing water. Inhabitants of Suite 405 may have traveled to the reservoir more often than other people in the village, or people from other kiva suites may have deposited white ware jars broken in transit between the reservoir and the village in the trash mound of Suite 405. No special jar forms were identified among the white ware jar rim sherds found in this midden.

¹⁶⁴
A third relationship is not apparent in the box plots but may have parallels at other sites. Fratt (1997*1:248) found that two-hand manos from the D-shaped structure at Sand Canyon Pueblo exhibited more intensive use wear than manos from other parts of that village and interpreted this pattern as evidence that inhabitants of the D-shaped structure produced more food than the average household, perhaps for ceremonial consumption. At Castle Rock, the CA results suggest that ground-stone tools were also strongly associated with Suite 402, which includes a kiva inside a D-shaped enclosure. The fact that discarded food-preparation tools were unusually abundant in trash associated with this kiva suite suggests that its inhabitants processed more food than other households at Castle Rock, as Fratt argued for the potentially analogous structure at Sand Canyon Pueblo. This may be evidence of organizational continuities across communities in the Sand Canyon locality.

¹⁶⁵
Although there is some evidence for differences in the mix of activities across kiva suites at Castle Rock, there is little evidence that qualitatively different activities occurred in one place or another. In other words, activities that led to deposition of the analyzed artifacts occurred in every kiva suite at Castle Rock Pueblo, and the mix of these activities appears to have been relatively consistent across kiva suites. This finding suggests that every kiva suite was primarily residential. It does not rule out the possibility, however, that certain activities that do not leave significant artifactual traces, especially ritual activities, occurred only in certain kiva suites. Implications of this analysis of midden composition are discussed further in paragraphs 169–185.

Abandonment of Structures
Artifact Densities in Kiva Fills

¹⁶⁶
In another portion of this report ("The Final Days of Castle Rock Pueblo"), Kuckelman provides an overview of the abandonment of Castle Rock, and more intensive studies of this abandonment are currently under way. This section considers only one line of evidence related to abandonment processes: artifact densities in the fills of kivas. Some artifacts will naturally wash into collapsing kivas as they fill with postabandonment sediment, but in some cases kiva depressions were used as trash dumps or were intentionally filled. Trash-filled kiva depressions can be interpreted as evidence of continuing habitation at a site for some time after those kivas were abandoned. Analysis of artifact densities in kiva fills may therefore shed some light on the abandonment of Castle Rock.

¹⁶⁷
One excavation unit was chosen from each excavated kiva at Castle Rock, and the density of all artifacts in strata interpreted as representing structural collapse was calculated. The results are presented in Figure 16. The artifact densities of most kivas were below 300 artifacts per cubic meter, in the range that Varien (1999*1:Chapter 6) interpreted as typical of naturally collapsing and filling pit structures. Also, none of the kivas at Castle Rock was filled with midden deposits, suggesting that the abandonment was not a gradual process. However, the fill of one kiva—Structure 304 on the north side of the village at the base of the butte—did contain a surprising number of artifacts in roof-fall strata immediately underlying informally deposited human remains. The artifact density of these deposits seems too high to be accounted for by slopewash from structures above this area on the butte. It is also clear from the sediment matrix that this structure was not gradually filled with an ashy midden deposit after the structure was abandoned. The most likely interpretation of this deposit, then, is that the structure was deliberately filled after the roof collapsed and before the human remains were deposited, and this fill just happened to have a high artifact density. Other structures at Castle Rock could also have been deliberately filled, but if so, the parent material used did not contain a large number of artifacts. This interpretation, if correct, suggests that some human activity, which may have included intentional or accidental deposition of human remains, did occur after the conflict associated with the abandonment of Castle Rock.

Abandonment Deposits Relating to Warfare

¹⁶⁸
There was an unusually high concentration of projectile points, bifaces, and objects of personal adornment in Structures 302 and 304. Both structures also contained concentrations of informally buried human remains. High overall artifact density (see paragraphs 166–167) hinders interpretation of this concentration in Structure 304, but in Structure 302 the concentration of these artifacts was clearly anomalously high. These two structures are located at the base of the north side of the butte in the center of Castle Rock Pueblo, and the only way to gain access to the top of the butte today is by walking through this area. The top of the butte would presumably have been the last bastion for defenders of the site, so it makes some sense that fighting would have been especially intense along the north side of the butte, leading to the observed concentration of projectile points, bifaces, and objects of personal adornment. Alternatively, this concentration could have resulted from the deposition of victims who died on top of the butte, along with their personal belongings, inside these structures. Regardless of the specific explanation, it is clear that some sort of cultural deposition was involved in creating these deposits.

Consideration of Results in Light of the Sand Canyon Project Research Design

¹⁶⁹
The Castle Rock Pueblo excavations took place in the context of the Sand Canyon Archaeological Project, a 10-year, interdisciplinary effort that focused on the Pueblo III occupation of the Sand Canyon locality. Fieldwork conducted as part of this project included full-coverage survey within the locality (Adler 1992*3; Adler and Metcalf 1991*1; Gleichman and Gleichman 1992*1), test excavations at 13 small sites (Varien 1999*2), and intensive excavations at Sand Canyon Pueblo (Bradley 1992*2, 1993*1, 1996*1) and the Green Lizard site (Huber and Lipe 1992*1) in addition to Castle Rock Pueblo. The research design for the Sand Canyon Archaeological Project was presented by Lipe (1992*3). This section considers how the results presented throughout this report relate to the major research domains identified in this article.

Community Organization and Change

¹⁷⁰
The concept of community adopted for the Sand Canyon Project follows from the cross-cultural research of George P. Murdock (Murdock 1949*1; Murdock and Wilson 1972*1), who defined the community as the largest social unit within which members interact face-to-face on a regular basis. In middle-range agricultural societies lacking beasts of burden and modern communication and transportation technologies, such communities are necessarily small in both geographic and demographic extent (Varien 1999*1). Adler (1990*1, 1994*1, 1996*3) and Adler and Varien (1994*1) added that such communities also tend to have a decision-making capacity above the level of their primary economic units, especially for the adjudication of resource access rights within community territories. Archaeological correlates of such communities in the Mesa Verde region include the spatial clustering of settlements around good springs and farmland and the presence of public architecture centrally located within settlement clusters (Adler 1990*1; Lipe 1970*1, 1992*2; Rohn 1965*1, 1977*1; Varien 1999*1:Chapter 1).

¹⁷¹
Lipe (1992*3), building on the work of Blanton and his colleagues (Blanton et al. 1981*1), identified four dimensions of organizational variation in communities. These dimensions are scale, differentiation (horizontal and vertical), integration, and intensity. These dimensions and the relevant findings of this report that address each one are considered in the following sections.

Scale

¹⁷²
Scale is defined as the size of the geographic area occupied by a community and its population. A related concept, the reach of a community, is the distance from which exotic and imported goods were obtained. Castle Rock Pueblo was probably the central place of a community that has been designated the Lower Sand Canyon community (Lipe 1992*1). This settlement cluster includes a number of small Pueblo III sites (including the tested sites Mad Dog Tower [Kleidon 1999*2] and Saddlehorn Hamlet [Kleidon 1999*3]) in lower Sand and East Rock canyons and adjacent portions of McElmo Canyon. Because only portions of this area have been systematically surveyed, the overall scale of the Lower Sand Canyon community is not yet clear. For Castle Rock itself, however, 16 households, one for each kiva at the site, is a reasonable estimate of the resident population. This estimate is based on the evidence for residential activity in every kiva suite at which associated middens were sampled (see paragraphs 155–165) and on the widespread occurrence of direct evidence of pottery making (see paragraphs 68–69). Also, at population estimates of fewer than 16 households, it becomes difficult to account for the total accumulation of corrugated jar sherds and the span of tree-ring-dated structures at Castle Rock. The reach of Castle Rock Pueblo did not extend beyond the boundaries of the Mesa Verde culture area to the north. To the east, south, and west, however, there was a small amount of indirect interaction through the exchange of material originating outside the region. Objects came from as far away as the Pacific Ocean and the Gulf of California to the west, possibly the Jemez Mountains to the east, and the Mogollon Rim country to the south (see paragraphs 129–131).

Differentiation
Horizontal

¹⁷³
Horizontal differentiation refers to functional specialization among parts of equivalent rank within a community. There is some evidence of possible horizontal differentiation within Castle Rock Pueblo (paragraphs 162–165). Gizzard stones indicative of turkey butchering are especially common in the midden area directly northeast of Kiva Suites 302 and 304. Whether this indicates that turkeys were also raised in this area is unclear. Trash deposited in the midden of Kiva Suite 405 is unusually rich in white ware jar sherds and is relatively close to a domestic water source for the village. It is unclear whether these data reflect an emphasis in the activities of this particular household or whether broken jars from other households were more often deposited in this trash mound as people came and went from the reservoir.

¹⁷⁴
Despite this equivocal evidence of horizontal differentiation, the more striking pattern that emerges from intrasite analysis of trash (paragraphs 155–165) is that a wide range of activities appears to have taken place in nearly every kiva suite. Trash deposits throughout Castle Rock Pueblo contained food-preparation tools, hunting tools, serving vessels, cooking pots, sewing tools, and waste products of chipped-stone tool manufacture. Unequivocal direct evidence of pottery making was collected from 10 of 16 kiva suites, and possible evidence was collected from another two, despite the fact that most kiva suites received only limited testing. These data suggest that pottery making continued to be a household-level industry even after village formation and that nearly every household included at least one resident potter. It is clear that a basic set of productive and maintenance activities took place in every kiva suite, strengthening the argument that kiva suites represent the domestic architecture of Mesa Verde Puebloan culture.

Vertical

¹⁷⁵
Vertical differentiation is defined as rank differences among functionally diverse parts within a community. There is no overwhelming evidence of qualitative rank differences between households on the basis of trash deposits associated with kiva suites (paragraphs 155–165). Kiva suites that include a square kiva (Suite 103), a kiva inside a D-shaped enclosure (Suite 402), and a potentially oversized kiva (Suite 105) all have associated trash that reflects an essentially consistent set of domestic activities. The fact that nondomestic trash cannot be distinguished suggests that all kiva suites were primarily residential. The fact that kiva suite architecture, however, is so variable at Castle Rock Pueblo, Sand Canyon Pueblo (Bradley 1993*1), and other late Pueblo III villages in southwestern Colorado (Lipe and Ortman 1999*1) suggests that vertical differentiation might have existed among households but that such distinctions were not of a kind that produced qualitative differences in basic household trash.

¹⁷⁶
Quantitative differences in trash across kiva suites include the relatively large number of ground-stone tools in the midden associated with Suite 402, which includes a kiva in a D-shaped enclosure. This concentration might indicate that inhabitants of this household prepared or processed more food than did other households. This quantitative difference might have resulted from social obligations that went along with whatever status was signified by the D-shaped enclosure. An unusually large household could also be responsible, but this seems unlikely, judging from the absence of a large surface roomblock in this kiva suite. Two-hand manos from the D-shaped structure at Sand Canyon Pueblo also exhibit more use wear than manos from other households (Fratt 1997*1), suggesting continuities in community organization in the Sand Canyon and Castle Rock communities.

¹⁷⁷
The only evidence bearing on vertical differentiation between individuals lies in objects of personal adornment. The materials out of which these were made tended to be rare and exotic (paragraph 133), suggesting that there was some concern for distinguishing between people on the basis of their attire. Change in items of personal adornment over time has not been examined.

Integration

¹⁷⁸
Integration is defined as the interdependence of structural units within a community and the manner in which the interdependence is accomplished. In general, households at Castle Rock appear to have been self-sufficient. At least one member of each household appears to have made chipped-stone tools, prepared food, and made pottery (paragraphs 68–75). On the other hand, it is possible that certain households specialized in turkey raising, that others were responsible for provisioning the village with water from the adjacent reservoir (paragraphs 155–165), and that another household specialized in the manufacture or use of polished igneous stones (paragraph 117). So in some respects households were relatively autonomous, and in others, they were potentially interdependent.

¹⁷⁹
The sizes and decoration of serving bowls also suggest that food was more often presented and consumed in public contexts in thirteenth-century villages than in earlier communities with dispersed settlement patterns (paragraphs 62–66). Cooking-pot sizes, serving-bowl sizes, and frequencies of exterior decoration on serving bowls all suggest that communal feasting was more common or intense at Sand Canyon Pueblo than at Castle Rock Pueblo. At Castle Rock, such public gatherings probably took place in the plaza, since no single structure in the village was large enough to accommodate the estimated village population. This increase in public presentation and consumption of food suggests that interhousehold integration increased along with village formation.

Intensity

¹⁸⁰
Intensity is defined as measures of population, material, information, or energy use per unit area or per capita. Spatial proximity alone indicates that residents of Castle Rock interacted more intensely than they had prior to moving into the village, but it is impossible to determine with present data whether village formation affected the intensity of interaction between residents of adjacent communities. Depressed frequencies of igneous-tempered pottery in village assemblages relative to those of adjacent, contemporaneous hamlets suggest that the larger populations of villages obtained pottery vessels across larger areas that extended farther away from igneous temper sources (paragraphs 77–83). It is unclear, however, whether this pattern reflects more-intense interaction per village resident or is simply a statistical effect of assemblages produced by larger population aggregates. In other words, even if the intensity of interaction per person was constant across hamlets and villages, village pottery samples would still exhibit more evidence of interaction than hamlet samples simply because more people contributed to the village samples. But even if interaction intensity per person was consistent in villages and hamlets, information gained through interaction with people in other sites would have been shared much more effectively within thirteenth-century villages than between earlier and contemporaneous hamlets.

¹⁸¹
Use-wear patterns on two-hand manos appear to relate more to variation in grinding strokes and to wear-reduction sequences than to wear management (paragraph 116). If this is the case, mano use wear appears unrelated to a possible intensification of food processing that one might expect to have occurred along with village formation. Possible evidence of a high frequency of large cooking pots at Sand Canyon Pueblo may indicate, however, that more large meals were prepared for larger consumption groups in this very large village than at other Sand Canyon locality sites, including Castle Rock. Serving vessel size and decoration suggest an increase in communal ceremonialism and feasting associated with village formation (paragraphs 62–66).

Abandonment

¹⁸²
Pottery accumulations, taken together with population estimates and tree-ring data, suggest that Castle Rock was abandoned sometime during the A.D. 1280s or early 1290s (paragraphs 146–154).

¹⁸³
Other results presented in this report improve our picture of social conditions immediately prior to the final migrations of Pueblo people out of the Mesa Verde region. It is apparent from frequencies of nonlocal pottery in Castle Rock and other southwestern Colorado sites that levels of interregional interaction declined significantly during the thirteenth century, especially in the central Mesa Verde region (paragraphs 87–90). This pattern is mirrored in the extreme rarity of other objects made of nonlocal materials at Castle Rock. These data suggest that very little interaction took place between Castle Rock and possible destination areas for the final migrations. This pattern seems contrary to the expectations of migration-stream models recently proposed to characterize this migration (paragraphs 129–131).

¹⁸⁴
A violent conflict clearly occurred during the final days of habitation at Castle Rock, and the concentration of projectile points and objects of personal adornment, along with the discovery of human remains in the fills of kivas on the north side of Castle Rock butte, suggests that the fighting was particularly intense and prolonged in this area (paragraphs 166–168). Unfortunately, artifacts collected from Castle Rock shed little light on the identity of the attackers. There is no clear artifactual evidence that the attackers originated outside the Mesa Verde region. The only possible evidence of nonlocal attackers comes from projectile points (paragraphs 99–103). The one point clearly of non-Puebloan origin is a Desert Side-Notched point that was found in a postabandonment context and thus probably originated from relatively recent Ute visitors to the site. The form and material of a Bull Creek point make it consistent with the type that was made in Mesa Verde-tradition sites in southeastern Utah west of Comb Ridge. Whether this point was brought to Castle Rock by the attackers or through peaceful means is unknown.

¹⁸⁵
Although Castle Rock Pueblo was built in a defensible setting, contained defensive architectural features, and was clearly the scene of a major battle, only one or two hafted axes and mauls that could have been made specifically as weapons were found there (paragraphs 118–121). Even if usable axes were carried away from the site by survivors of the battle, it is surprising that so few remained to be found during the excavations. The relationship between the abundance and elaboration of weaponry and the actual frequency of warfare is an important topic for future research on violence in agricultural village societies.

Acknowledgments
The Castle Rock Pueblo excavations took place in the context of Crow Canyon's public archaeology programs, and innumerable participants in these programs spent countless hours washing, cataloging, and analyzing the artifacts considered in this report. This project would have been impossible without their support. All research that is accomplished at Crow Canyon results from a group effort by the research and publication staffs. In addition to myself, current and former Crow Canyon staff members Louise Schmidlap, Angela Schwab, Mary Etzkorn, Melissa Churchill, Melita Romasco, Jamie Merewether, Michelle Hegmon, Chris Pierce, Joe Keleher, and Maggie Thurs either supervised data collection or directly collected the data for Castle Rock artifacts. Maggie Thurs, Adele Bigler, and Robin Lyle collected the temper data from Sand Canyon locality white wares. The rim-arc data were collected by Molly Duncan. Jamie Merewether assisted in collecting data for points, knives, and drills. Chris Pierce classified most of the other modified stones and minerals and identified most of the direct evidence of pottery production. G. Timothy Gross identified shell species. Tim Kohler offered useful advice on poststratification of the probability sample. Lee Gripp and Melita Romasco offered critically needed technical computer support. I also thank Crow Canyon Native American Advisory Group members Jane Polingyouma, Ernie Vallo, and Peter Pino for drawing on their own experience in offering many interesting avenues for interpretation of Castle Rock Pueblo artifacts. Kristin Kuckelman, Donna Glowacki, Mark Varien, and Robin Lyle all provided many useful comments on an earlier draft. Jane Kepp gave the text an excellent copy edit, and Louise Schmidlap and Ginnie Dunlop accomplished the arduous task of translating the original print version of this report into electronic format.

References cited | To borrow, cite, or request permission
Permissions obtained: See captions appended to individual figures.
Copyright © 2000 by Crow Canyon Archaeological Center. All rights reserved.