DONATE TODAY
22
Summary and Conclusions
Mark D. Varien and Kristin A. Kuckelman
Contents
- Instrumental Studies
- Chronology
Length and Season of Site Occupation - Length of Site Occupation
Season of Site Occupation
Summary: Length and Season of Site Occupation - Site-Formation Processes
Paleoenvironmental Reconstruction
- Abandonment
- Structure Abandonment and Occupation of the Locality
Site Abandonment and Occupation of the Locality
Regional Abandonment
Introduction
During four years of test excavation, 13 sites with primary occupations dating to sometime between A.D. 1180 and 1280 were sampled. Ten of these sites are located in the Sand Canyon community, defined as the settlement cluster within a 2-km radius around the head of Sand Canyon; of these, six are located on mesa tops and four are in the upper canyon. The remaining three sites are located in the Castle Rock community, in lower Sand Canyon. Two of these are situated on benches within the canyon, but above the inner gorge. The last site is just above the McElmo Creek floodplain.
The Testing Program was designed not only to provide a better understanding of each individual site but, more importantly, to provide data that could be used in comparative studies, the results of which would improve our understanding of community organization and change during the Pueblo III period. These studies involved comparing the tested sites with one another, comparing the tested sites with Sand Canyon Pueblo, and comparing the Sand Canyon locality sites (including both tested and intensively excavated sites) with other sites in the Mesa Verde region and the greater Southwest. Because representative, unbiased samples in which the degree of sampling error could be measured were critical to these comparative studies, stratified random sampling was used to obtain the samples for analysis.
Testing Program investigations were guided by the Sand Canyon Project Research design (Lipe and Bradley 1986, 1988; Lipe 1992a), and this chapter is organized according to the objectives outlined in that design. In the first part of the chapter, the results of a series of instrumental studies are presented. These studies, part of a body of middle-range research designed to provide data critical to addressing higher-level research questions, include the following: refining chronology, estimating the length and season of site occupation, understanding site-formation processes, and reconstructing paleoenvironments. The instrumental studies were designed to address two broad research domains, which are the subject of the remainder of this chapter: (1) community organization and change during the Pueblo III period and (2) the abandonment of the Mesa Verde region. Four dimensions of community organization--scale, differentiation, integration, and intensity (Lipe 1992a:4-5)--are evaluated in light of the Testing Program data, and finally, the question of abandonment is addressed at the structure, site, community, and regional levels.
Instrumental Studies
A number of instrumental studies were identified by Lipe (1992a:5) as critical to addressing higher-level research domains. These studies are supported by a body of middle-range theory that is sometimes well developed (see, for example, Dean [1978a, 1978b] and Ahlstrom [1985a] for the theory that underlies the interpretation of tree-ring dates) and sometimes not. It is hoped that the instrumental studies developed during the Testing Program will not only make a contribution to our understanding of the Sand Canyon locality, but will further develop middle-range method and theory useful to archaeologists working in other parts of the world.
Chronology
A primary objective of the Testing Program was to refine the chronology of sites where Mesa Verde Black-on-white pottery was present (the period between A.D. 1180 and 1300). This refinement was needed to determine if Sand Canyon Pueblo was contemporaneous with any of the smaller, surrounding sites. A related objective was reconstructing the history of occupation at each of the tested sites. The Testing Program chronological studies used stratigraphic, archaeomagnetic, pottery, and tree-ring data to achieve these objectives.
Pottery was used to identify the number of separate components at multiple-component sites and to estimate the intensity of occupation for each component. When pottery dating to a particular period totaled more than 1 percent by weight of the total assemblage, a separate component was identified; using this method, 23 components were recognized at the tested sites. These include four Basketmaker III components, two of which appear to have been habitations (at Shorlene's Site and Kenzie Dawn Hamlet). Tree-ring dates from structures at these sites indicate that construction took place sometime between A.D. 660 and 670. The use life of these types of structures is assumed on the basis of cross-cultural analogy to have been between five and 15 years (Ahlstrom 1985a:82-90, 638; Cameron 1990; Schlanger 1987:586-589). Therefore, these habitation sites were probably occupied between A.D. 660 and 685. The remaining Basketmaker III components (at G and G Hamlet and Lillian's Site) probably represent more-limited use of the sites.
Six Pueblo II components were identified as dating primarily to the A.D. 1025-1100 period. Two of these are interpreted as habitations (G and G Hamlet and Kenzie Dawn Hamlet), and the remaining four (Lillian's Site, Roy's Ruin, Shorlene's Site, and Catherine's Site) are interpreted as having had more-limited use. In addition, traces (1 percent or less) of Pueblo II pottery were found at the remaining seven sites. Tree-ring dates from the habitation components indicate that they date to between A.D. 1060 and 1100. The pottery dating supports this interpretation but suggests that these occupations may have also spilled over into the early A.D. 1100s.
The remaining 13 components date to the Pueblo III period, between A.D. 1150 and 1280, and they constitute the main occupations at each of the 13 tested sites. At Kenzie Dawn Hamlet, there was a long, relatively continuous Pueblo III occupation with several distinct building episodes. One of the primary goals of the chronometric instrumental studies was to refine our understanding of these Pueblo III components, and our success in this endeavor has been an important achievement of the Testing Program. Refinement of the tested sites chronology was accomplished by combining the results of tree-ring dating, archaeomagnetic dating, stratigraphic analysis, and three types of pottery analysis: assemblage-based dating, attribute-based dating, and relative dating using the ratio of McElmo to Mesa Verde black-on-white pottery (see Chapter 20). On the basis of these data, a core date range for each site was identified, which represents the period within which the most intensive use of the site occurred (Table 22.1).
When these dates are considered in terms of site location, a clear pattern emerges. With the exception of Troy's Tower, all of the mesa-top sites date predominantly to before A.D. 1250, whereas the sites located within upper and lower Sand Canyon date to the late A.D. 1200s (the possible exception to this is Mad Dog Tower, which is poorly dated). These settlement patterns will be examined in greater detail in the discussion of community organization and change.
Crow Canyon's work on pottery chronologies is continuing. Hegmon's study indicates that there is great promise for attribute-based dating (Hegmon 1991). The ability to identify components dating to the middle to late A.D. 1100s is particularly important for understanding the relationship between the Chaco and post-Chaco periods in the Mesa Verde region and for understanding the formation of the large villages that date to the A.D. 1200s. Unfortunately, this interval has been notoriously difficult to date using pottery (Varien et al. 1996). In a comparison of typological- and attribute-based pottery dating, Ortman (1995) found that the latter, when applied to white wares, appears to result in more accurate assignment of occupations dating to the second half of the twelfth century. In contrast, typological-based dating of white wares could result in an underestimation of settlement across the region during this period.
It should be noted that Wilson and Blinman's (1991) assemblage-based dating scheme relying on traditional types acknowledges the difficulty of using white wares to date sites in the second half of the twelfth century. They argue, however, that consideration of full assemblages, including gray wares and red wares, can overcome the problems associated with looking only at white wares. Gray ware frequencies, particularly the increasing numbers of Dolores Corrugated, are critical to discriminating sites dating to the A.D. 1140-1180 period in Wilson and Blinman's system. The corrugated types (Mancos, Dolores, and Mesa Verde) are distinguished from one another on the basis of a single attribute, rim eversion. Currently, Crow Canyon analysts use rim eversion to identify sherds as either Mancos Corrugated or Mesa Verde Corrugated; Dolores Corrugated is not recognized as a separate typological category. Incorporating rim eversion into attribute-based dating schemes may further improve our ability to recognize sites that date to the A.D. 1150-1200 period.
Length and Season of Site Occupation
Determining length of site occupation and whether occupation was continuous or seasonal was another important goal of the Testing Program. The lack of reliable estimates of the length and season of site occupation undermines archaeologists' efforts to accurately estimate the number of people who occupied a given area in the past (Powell 1988). The inability to accurately estimate the size and density of human populations in turn undermines our efforts to address higher-level research questions (Nelson et al. 1994), including--in the case of the Testing Program--the organizational complexity of the Sand Canyon communities and the reasons why the Puebloan people abandoned the Mesa Verde region.
Our research into estimating the length and season of occupation has been reported in detail elsewhere (Varien 1997; Varien and Potter 1997; Varien and Mills 1997) and will only be summarized below.
Length of Site Occupation
Several authors have demonstrated that perhaps the weakest link in paleodemographic estimation is our inability to specify the length of occupation of individual sites (Nelson et al. 1994; Powell 1988; Santley 1990). Other authors have pointed out that it is necessary to control for length of occupation before comparing assemblages from different sites to address questions such as site function (Schlanger 1990). This is so because the length of site occupation affects the relative frequencies of different artifact classes (Clarke 1972; David 1972; David and Hennig 1972; de Barros 1982; DeBoer 1974, 1985; Mills 1989b), a phenomenon known as the "Clarke Effect" (Schiffer 1975:275). Clearly, estimating the length of site occupation is key to addressing many of the basic questions asked by archaeologists. Two sources of data are used in our assessment of the length of occupation at the tested sites: (1) the number and weight of cooking-pot sherds that accumulated at the sites and (2) cross-cultural analyses that examine how architecture and building materials constrain the use life of buildings.
The first step in estimating the length of site occupation was to model the rate at which artifacts accumulate at sites (Varien 1990a; Varien et al. 1992), a process referred to as "accumulations research" (Varien 1997; Varien and Mills 1997). Accumulations research examines the relationship between time, population, and the accumulation of discarded artifacts. Archaeologists conducting accumulations research assume that the amount of material discarded is a function of the length of site occupation, the number of people living at a site, and the rate at which the specific type of material was discarded.
Archaeologists since the turn of century have used the assumed relationship between these variables to estimate both the length of site occupation (Nelson 1909; Gifford 1916; Cook 1946) and the number of people living at sites (Cook and Treganza 1950). Others have used the population, or site use-life estimates derived from accumulations research, to address a diverse set of problems, such as evaluating the degree of social and political complexity of past societies (Kohler 1978; Pauketat 1989), determining whether site occupation was seasonal or year-round (Kohler and Blinman 1987), and evaluating the effect of site abandonment on de facto refuse assemblages (Lightfoot 1993, 1994). The relationship between the variables of time, population, and the accumulation of artifacts has been stated as a formula known as the discard equation (Schiffer 1975:840, 1976:59, 1987:53), which has been used in many recent accumulations studies (Mills 1989b; Lightfoot 1993; Pauketat 1989).
The posited relationship between these variables is a type of middle-range theory. Both ethnoarchaeological and archaeological research suggest that the common-sense assumption on which this theory is based is basically correct: more time and more people result in the accumulation of more material. This relationship is clear in Yellen's (1977) research among the Dobe !Kung, for example. Nelson et al. (1994) examined archaeological assemblages from sites in the Dolores River valley and found strong correlations between the accumulation of functionally unrelated artifacts, such as pottery sherds and chipped-stone debitage. They believe that the variable most responsible for the strong relationship is the person-years of occupation, which combines the average momentary population with the length of site occupation. These authors evaluated the potential of accumulations research by examining how well the estimates of pottery populations predict the length of occupation (Nelson et al. 1994). They used a number of sites where excavation had exposed virtually all of the architectural features, allowing them to estimate the number of households present at sites, and they estimated site use life on the basis of cross-cultural regularities in the use life of earthen architecture (Nelson et al. 1994; cf. Ahlstrom 1985a; Cameron 1990; Schlanger 1987). Together, the estimates of the number of households and the site use life provided an estimate of the number of household-years a site was occupied. They examined the relationship between household-years of site occupation and the estimated total pottery discard at sites and found a strong correlation (r² = .76), which is interpreted as further evidence that there is a positive relationship between the number of people at a site, the amount of time spent at the site, and the amount of material that accumulates at the site. Nelson and his colleagues concluded by saying that the weakest link in paleodemographic estimation is our appreciation of site use life and that we can best evaluate site use life by building the middle-range theory of artifact accumulation rates (Nelson et al. 1994:137).
Schlanger's study of artifact accumulations concludes that, if it is possible to estimate site size, then the best way to measure occupation duration is to count the types of artifacts that accumulate the fastest (Schlanger 1990:119). In the case of the tested sites, corrugated cooking-pot sherds are one of the fastest-accumulating artifact types, and the accumulation of these sherds is used to estimate the length of site occupation. Cooking pots are ideal for modeling accumulations because recent experimental and ethnoarchaeological studies have focused on their use and discard. Experimental research demonstrates that thermal shock induced by the repeated heating and cooling of vessels ultimately results in their breakage (Bronitsky 1986:249-259; Bronitsky and Hamer 1986; Rice 1987:105, 226-231; Rye 1976, 1981:27; Steponaitis 1983:36-45; West 1992), and ethnographic and ethnoarchaeological studies indicate that pots used for everyday cooking have some of the shortest use lives of any vessel form (Varien and Mills 1997). Finally, Crow Canyon Center archaeologists have investigated two sites, Sand Canyon Pueblo and the Duckfoot site, where abundant tree-ring dates and thorough excavation of architectural features allow estimates of the number of households present and the length of occupation, thereby allowing estimates of annual rates of cooking-pot accumulations per household (Varien 1997). A comparison of these archaeologically derived accumulation rates with the cross-cultural sample indicates that the Mesa Verde-region accumulation rates are within the range of variation of societies around the world.
The data and methods used to estimate the length of occupation of the tested sites are presented in detail elsewhere (Varien 1997, 1998). A brief summary of these data and methods are presented here, along with the results.
To calculate occupation span, the following data were used: (1) statistical point estimates and confidence intervals for the total weight of corrugated cooking-pot sherds from each tested site; (2) an annual discard rate per household for cooking-pot sherds; (3) an estimate of the number of households at each site, made on the basis of the number of pit structures and kivas present; and (4) contextual information that includes the architectural, stratigraphic, and chronometric data for each site (Varien 1997, 1998).
The total weight of corrugated cooking-pot sherds was divided by the annual discard rate per household to calculate the total household-years of occupation at each site. This figure was then divided by the number of households at each site to obtain the length of occupation per household. Finally, contextual information from sites with more than one household was evaluated to determine whether multiple households were occupied simultaneously, sequentially, or partly sequentially. Using these data, a best estimate and a range for occupation span was calculated (Varien 1997, 1998). These estimates are presented in Table 22.2 and Figure 22.1.
A striking feature of the total length of occupation estimates is their variability. This variation is not due entirely to sampling error, and it would be present even if the annual accumulation rate and the associated absolute estimates of occupation were in error. There simply are large differences in the number of cooking-pot sherds that accumulated per household at the 13 tested sites. The accumulations research allows us, for the first time, to quantify that variability, and it appears that the occupation spans ranged between approximately one and three generations at the tested sites.
Single-generation occupations at year-round habitations include the Pueblo II components at G and G Hamlet and Kenzie Dawn Hamlet and the Pueblo III habitation components at G and G Hamlet, Saddlehorn Hamlet, and Castle Rock Pueblo. Single-generation occupation is also possible for many other sites if the low end of the range is correct. The best estimates for the remaining sites indicate that they were probably occupied for more than a single generation. Occupation for more than one generation is virtually certain at sites where multiple kiva suites were occupied sequentially (the Pueblo III habitations at Kenzie Dawn Hamlet and Lookout House) and is strongly indicated at sites where there was partly sequential occupation of kiva suites (Lester's Site and Catherine's Site). Occupation longer than one generation is also likely at many of the residential sites with a single kiva suite. At these sites, the same residential facilities appear to have been used by multiple generations. In general, the mesa-top unit pueblos had the longest occupation spans, but all of the talus sites, especially Stanton's Site, were probably occupied for more than one generation.
The factors that affect occupation span--including climatic change, resource depletion, and social factors--have been examined elsewhere. It appears that no single factor can account for the variation in occupation span estimates, but the considerable variation suggests that social and demographic factors play an important role (Varien 1997, 1998). Residential sites occupied for a single generation may relate to the domestic cycle, with these sites being established as new households form at marriage. In addition, the demographic viability of small sites may set upper limits on their occupation span (Gaines and Gaines 1997).
Cross-cultural research that examines how different types of architecture constrain the length of site occupation can clarify the interpretation of occupation span estimates at the tested sites. These studies are particularly strong in terms of the use-life constraints on earthen architecture. Cross-culturally, these buildings typically last between six and 12 years, rarely last longer than 15 years without extensive remodeling, and have an upper limit of 30 years even with remodeling (Ahlstrom 1985a:83-84, 638; Cameron 1990; Diehl 1992; McIntosh 1974; Schlanger 1987:586). The use-life constraints on masonry architecture are much less severe than those documented for earthen architecture. Diehl's (1992) cross-cultural analysis indicates that earthen buildings typically have shorter use lives than masonry buildings. Ahlstrom (1985a:642) found that a use life of 60 years for masonry buildings in the ancient Southwest is not unreasonable.
The occupation-span estimates calculated on the basis of the accumulation of cooking-pot sherds and the use life of earthen vs. masonry buildings indicate that there was an important change in the length of site occupation between the Pueblo II and Pueblo III periods. Pueblo II residential sites, characterized by earthen buildings, were limited to single-generation occupations, but the Pueblo III residential sites with masonry architecture were often occupied for multiple generations (Varien 1997, 1998). This change in occupation span is viewed in light of the larger social context of which these residential sites were a part and includes an increase in population density and greater competition for resources. The change is interpreted as indicating a change in agricultural land-tenure systems, from land tenure claimed on the basis of usufruct rights, during the Pueblo II period, to land tenure claimed on the basis of some form of heritable property rights, during the Pueblo III period (Varien 1997, 1998).
Season of Site Occupation
This section addresses the question of whether the tested sites were occupied year round or seasonally, and if the latter, the season of occupation. Macrobotanical, pollen, and faunal remains are important seasonality indicators, as are the types of architectural facilities present and, to a lesser degree, the types of tools found at the tested sites. In addition, the sites are evaluated in light of ethnoarchaeological research on how the intensity of site use affects site layout and the types of architectural facilities present. These data and the subsequent interpretations refer only to the Pueblo III components at the sites, because these were the most intensively sampled.
Macrobotanical remains were recovered from the tested sites in two ways: through flotation sampling and through the collection of individual specimens during routine excavation and screening. In all, 183 flotation samples were processed. Sampling focused on secondary refuse in middens and primary refuse in hearths. Most of the hearths were located in kivas, but hearths located in other contexts, including surface rooms, were also sampled when they were encountered. A mix of domestic and wild plants was found at the majority of the tested sites, and the taxa identified indicate occupation from late spring through fall, basically throughout the growing season. A wide variety of maize plant parts were recovered, suggesting that agricultural fields were located relatively close to the sites (Adams, Chapter 16). The wild-plant assemblage is dominated by species that thrive in disturbed habitats like agricultural fields, which suggests that wild plants, too, may have been harvested fairly close to the sites.
Economic inferences can be made using pollen samples collected from sealed contexts on kiva floors. The pollen record supports the interpretations made on the basis of macrobotanical remains: both agricultural and wild plants were harvested, and the taxa represented indicate that people were in the area from late spring to early fall (Gish, Chapter 17). In general, fewer taxa are represented by pollen than by macrobotanical remains, but this is probably because more flotation samples than pollen samples were analyzed. There are, however, certain taxa--such as beeweed--that are commonly found in pollen samples but rarely in macrobotanical samples. Similarly, squash pollen was found in one pollen sample, but there was no evidence of this plant in the macrobotanical record. Cheno-am pollen is present in pollen samples from both floor and roof fall contexts, which suggests that it was an important economic resource while the kiva was being used and that it was growing in disturbed habitats around the site after the kiva was abandoned. Pollen interpretation is discussed in greater in the section on paleoenvironmental reconstruction.
One research topic that we had hoped to address through analysis of plant remains was the possibility that there was seasonal mobility between upper Sand Canyon and lower Sand Canyon. McElmo Creek, in lower Sand Canyon, is approximately 1,400 ft lower in elevation than the mesa tops in upper Sand Canyon, and it is possible that the lower canyon would have been more attractive during the winter months. However, there is no support for this type of seasonal residential mobility in the macrobotanical and pollen records. Sites in both areas contain similar plant remains, with squash being the only cultigen identified in the lower Sand Canyon sites that was not present in the upper Sand Canyon tested sites (although squash was found at Sand Canyon Pueblo).
The results of macrobotanical and pollen studies conclusively demonstrate that sites were occupied during the growing season, but they do not indicate whether sites were occupied during the winter. Although there was not enough faunal material to assess season of occupation using data such as tooth eruption and incremental studies, the faunal remains still can be used to comment on the continuity of occupation within the year. Turkey bone and eggshell were recovered from every site, and Munro (1994) used a variety of data to infer that turkeys were domesticated and were raised on most sites. Maintaining turkeys requires daily activity, especially in the winter, and this implies that sites were occupied year round.
The interpretation that most sites were occupied year round is further supported by the type of architectural facilities present. Eleven of the 13 tested sites have masonry roomblocks. These roomblocks contain a number of rooms, and some are interpreted as having been used for long-term storage of food remains. The abundance of maize in hearths, middens, and pollen floor scrapes at most of the sites supports the interpretation that it was being stored and consumed well after it was harvested--an inference further supported by the presence of food-grinding tools at the sites. Thus, the presence of long-term storage facilities, the abundance of maize, and the recovery of tools used to process that maize, all suggest that groups were living at the sites during the winter.
Two possible exceptions to this pattern are Troy's Tower and Mad Dog Tower. These sites do not have surface roomblocks. A single surface room at Mad Dog Tower might have been a storage facility, and the same may be true of large pit features at Troy's Tower. However, there does not appear to have been as heavy an investment in long-term storage facilities at these sites when compared with the other tested sites. The absence of long-term storage facilities may indicate seasonal use of these sites, with site abandonment during the winter.
Several ethnoarchaeological and archival cross-cultural studies have been undertaken to determine how architecture, site layout, and refuse disposal correlate with season and intensity of occupation. Many of these studies deal with hunting and gathering societies (e.g., Kent 1990, 1991b, 1992; Yellen 1977); others deal with societies with mixed and agricultural economies (e.g., Arnold 1990; Deal 1985; Gilman 1987; Hayden and Cannon 1983; Killion 1990). The studies are diverse, but a number of points relevant to the tested sites can be generalized from them. Cross-cultural generalizations of this sort run the risk of being reductionistic or overly simplistic (Kent 1990:151), but in this case, they are viewed as an important body of middle-range research that provides support for a number of interpretations about the continuity of occupation at the tested sites. The generalizations are as follows:
1. Increasing length of occupation results in increasing amounts of material, both artifacts and architecture, resulting in increasing site size (Yellen 1977). Kent (1990, 1991b, 1992) would take issue with Yellen and argue that it is the anticipated length of occupation rather than the actual length of occupation that does more to condition site size and the amount of material at sites.
2. Increasing continuity and length of occupation correlate with the appearance of formal, as opposed to informal, storage areas, and increasing investment in the construction of storage areas indicates increasing sedentism (Kent 1991b; Gilman 1983, 1987).
3. Increasing sedentism correlates with increasing investment in architectural facilities designed as habitations (Kent 1991b; Diehl 1992).
4. Increasing the continuity and length of occupation results in increasing specialization and segmentation of space at sites (Arnold 1990; Kent 1990).
5. Increasing continuity and length of occupation correlate with increasingly formal refuse disposal (Arnold 1990; Deal 1985; Hayden and Cannon 1983; Kent 1991b, 1992; Killion 1990).
When the tested sites are evaluated in light of these generalizations, their architectural, site layout, and refuse-disposal patterns suggest that they were occupied continuously for relatively long periods.
A substantial amount of material accumulated on these sites, as quantified in the estimates of corrugated cooking-pot sherds (Varien 1997). For this much material to have accumulated on seasonally occupied sites, the sites would had to have been occupied for most of the year and returned to year after year for many years. Distinguishing this type of planned reoccupation from year-round occupation is difficult. The presence in midden stratigraphy of periodic natural deposition might be one indicator of seasonal abandonment; however, no such stratigraphy was noted at the tested sites, with the possible exception of Mad Dog Tower. Much less material had accumulated at Troy's Tower and Mad Dog Tower, indicating less intensive use of these two sites.
The presence of probable storage areas in masonry roomblocks has already been noted as evidence for intensive year-round occupation at most of the sites. The possible use of storage pits at Troy's Tower and the presence of only a single room at Mad Dog Tower indicate that these sites may have been occupied less intensively than the other sites.
The Testing Program excavations revealed a heavy investment in the construction of architectural facilities--all of the sites have masonry-lined or partly masonry lined kivas, as well as double-stone surface rooms and/or towers. The Pueblo III components stand in contrast to the Pueblo II components, where surface architecture was primarily post-and-earth construction and pit structures were earth-walled. The masonry construction of the Pueblo III architecture implies lengthy anticipated use of these buildings. Thus, even at Troy's Tower and Mad Dog Tower, where artifact accumulations and less-substantial storage facilities suggest the possibility of less-intensive use, this use may have been anticipated over many years.
There was not enough excavation at the tested sites to allow an evaluation of the internal organization of architectural space, but intensive excavations at Sand Canyon Pueblo and the Green Lizard site, as well as excavations at other sites in the Mesa Verde region, can be used to address this question. These excavations demonstrate that specialized rooms for mealing activities became common during the Pueblo II period. Towers, which first appeared during the late Pueblo II period, are another type of distinct, segmented space within sites, although their use is poorly understood. Finally, Pueblo II pit structures and kivas are located closer to roomblocks compared with their counterparts in earlier periods. Thus, excavation indicates that even small habitation sites occupied during the Pueblo II and III periods are characterized by increasing organization and segmentation of space.
Finally, there is a formal area for refuse at most of the tested sites. By formal refuse areas, we mean mounds with distinct boundaries--in contrast to the sheet trash with more diffuse boundaries present at earlier Basketmaker III and Pueblo I habitations. In all periods, middens served as the locations for burials, as well as areas for refuse disposal (Prudden 1903:12). The symbolic meaning that may have been accorded to refuse disposal and the relationship between ash discard, artifact discard, and human interment are topics that remain to be investigated (e.g., Hodder 1987). The only tested site without a formal trash mound is Mad Dog Tower, where the refuse consists of dispersed sheet trash. And at Troy's Tower, the midden is small and almost abuts the tower.
Another aspect of refuse deposition characteristic of more intensively occupied sites is the presence of "toft zones" (Arnold 1990:918; Deal 1983:262; Hayden and Cannon 1983:126). These are areas of refuse surrounding the courtyards; the refuse accumulates in these locations because courtyards are high-activity areas that are routinely maintained to keep them free of debris. Toft zones of higher refuse accumulation, which often appear as low berms, enclose the courtyards on all the mesa-top sites in upper Sand Canyon except Troy's Tower.
A final consideration with regard to the length and season of site occupation is whether year-round sites continued to be seasonally used after they were abandoned as permanent habitations. Artifact density is greater in the fill of mesa-top kivas that were abandoned earlier in the A.D. 1200s, which may indicate that these sites continued in use as field houses after they were abandoned as year-round habitations (Varien 1997; Varien and Kuckelman 1993). Microartifacts, as measured in a detailed geoarchaeological study of pit structure fill sequences (Kilby 1998), were also more common in the fill of the mesa-top pit structures, supporting the interpretation that these sites continued to be used after the pit structures were abandoned. In their analysis of the intrasite distribution of faunal remains, Driver et al. (Chapter 18) examined the three sites with the largest assemblages: Castle Rock Pueblo, Stanton's Site, and Kenzie Dawn Hamlet. They found that, at Kenzie Dawn, turkey bone was more common in exterior contexts, whereas cottontail was more common in structure fills. Turkey bone may have been more common in exterior contexts because these birds were raised when the site was occupied as a year-round habitation, and their bones were discarded in exterior middens. In contrast, hunting cottontails may have continued in the fields surrounding Kenzie Dawn after the site was abandoned as a habitation, but while it continued to be used as a field house; the bones of these animals were discarded in the abandoned kivas. The same pattern may hold true for the remainder of the mesa-top sites.
Summary: Length and Season of Site Occupation
The data presented above indicate that all or most of the tested sites were occupied year round for relatively lengthy periods. The possible exceptions are Mad Dog Tower and Troy's Tower, which may have been occupied only seasonally. Even at these sites, extended use over many years is suggested by the energy investment in the construction of architectural facilities and the amount of accumulated material.
Architectural data indicate that there may have been an increase in the duration of occupation between the Pueblo II habitation components, which consist primarily of earthen buildings, and the Pueblo III components, which include masonry buildings. This interpretation is supported by the accumulation of cooking-pot sherds, which indicates that more material accumulated during the Pueblo III occupations. Accumulations data and cross-cultural research on the use life of earthen vs. masonry buildings indicate that Pueblo II occupations were limited to a single generation, but later Pueblo III occupations typically spanned multiple generations.
Site-Formation Processes
Studies of site-formation processes must address two broad issues. The first is how sites are altered by historic land-use practices. The second is how cultural and natural processes affect sites during occupation and after abandonment, but before historic land use begins.
Historic land use includes a variety of activities that have impacted the tested sites, primarily farming, chaining, grazing, and artifact collecting. Analysis of the artifact assemblages from the tested sites makes it clear that these activities have had differential effects on the various sites. Sites that have been cleared, plowed, and cultivated have the smallest sherds. Sites that have been chained and intensively grazed have the next-smallest sherds, and sites in the canyons, where there has been no plowing or chaining and only limited grazing, have the largest sherds. This variability means that weights are better than counts when comparing pottery assemblages from the different sites.
The accumulations research described in the section on length of site occupation is a type of site-formation-process research. In conducting this research, one must consider both the cultural process of refuse disposal and the various natural processes that deplete artifact populations during site use and after site abandonment. The research on modeling the accumulation of cooking-pot sherds has been described above and will not be repeated here. The cooking-pot accumulation rates can be used as a means of calibrating accumulation rates for other types of artifacts and provide a better understanding of assemblage formation (Blinman 1988a; Mills 1989b; Nelson et al. 1994; Schlanger 1990).
Much of the effort in understanding site-formation processes has focused on pit structure and kiva fills. The results of three types of analysis constitute the data sets used to interpret the fill sequences at the tested sites: (1) analysis of stratigraphic profiles recorded when the structures were excavated, (2) analysis of sediment samples collected from those profiles, and (3) analysis of artifact size and density in the different strata that fill each of the structures. These fill sequences include cultural, natural, and mixed deposits. Two types of cultural deposit are present. The first consists of the remains of the roof of the structure, which was either burned or unburned. The second type of cultural deposit is secondary refuse that accumulated when abandoned pit structures were occasionally used for refuse disposal. Natural deposits result from wind, water, and colluvial deposition. Mixed deposits contain materials deposited by both natural and cultural agents.
An important question addressed by this research is the treatment of the roof at the time of structure abandonment. In a related study (Varien 1997, 1998), the stratigraphy of 34 kivas or pit structures that were partly excavated during the Testing Program was examined. On the basis of this study, roof treatment at the time of abandonment was placed in one of the following categories: burned, with timbers not salvaged; burned, with the larger timbers salvaged before burning; unburned, with the timbers unsalvaged and left to rot in place; unburned, with timbers salvaged; and unburned, with unknown treatment of the roof timbers.
Sediment analysis of 10 kiva and pit structure fill sequences was conducted by Kilby (1998). This study included particle-size analysis, organic and calcium carbonate content analysis, and microartifact analysis. Kilby's analysis provides empirical support for the categorization of fills into the cultural, natural, and mixed categories and for the interpretation that timbers were salvaged from most of the unburned structures. (Kilby's analysis is discussed in greater detail in the section on abandonment.)
Artifact density in the fill of abandoned kivas and pit structures (Varien 1997, 1998) was studied to complement the sediment analyses. These analyses show that, as expected, natural fills have the lowest artifact densities, mixed fills have higher densities, and cultural fills have the highest densities. The mixed fills are of particular interest. These occur most often in abandoned kivas at mesa-top sites. The higher artifact density is interpreted as resulting from periodic, but intentional, deposition of artifacts in these structures (Varien 1997, 1998). This deposition is interpreted as having occurred after sites were abandoned as habitations but while they continued to be used as field houses (Varien 1997, 1998).
These three data sets--stratigraphic profiles, the results of sediment analysis, and the size and density of artifacts in kiva and pit structure fills--will be combined to address the larger research domain of abandonment later in this chapter.
Paleoenvironmental Reconstruction
In this section, macrobotanical and pollen data are used to reconstruct environmental conditions in the Sand Canyon locality during the time the tested sites were occupied. The primary focus of this summary is to assess the extent of the human impact on the environment during the Pueblo III period. Pollen samples are used to address the question of deforestation of the Sand Canyon locality. The Testing Program pollen samples were collected from three broad contexts: modern ground surface, roof fall strata, and sealed floors in structures. The modern ground surface samples provide a baseline for evaluating the ancient samples. The roof fall samples provide information about the environment surrounding the sites in the period immediately following structure and, in some cases, site abandonment. They also provide a baseline for interpreting floor context samples. The samples taken from sealed floor contexts provide the best information on pollen deposition inside structures during occupation. With the exception of a sample collected from one tower, all the floor pollen samples are from kiva floors (see Gish, Chapter 17). Four taxa--juniper, pinyon pine, cheno-am, and sage--dominate each of the contexts sampled (Table 22.3).
Pollen data are notoriously difficult to interpret because the deposition of pollen is an ongoing process complicated by many factors, some of which cannot be known to the individual interpreting the sample. In addition, if the frequency of one taxon goes up or down dramatically, it affects the frequencies of the other taxa, even if there was no absolute change in the rate of deposition of pollen of the other taxa. For example, if the value of cheno-ams rises dramatically, the pinyon values concomitantly drop, but this does not necessarily mean there were fewer pinyon trees in the environment.
Taking control samples from the modern ground surface and from the roof fall above floors improves the interpretability of samples from structure floors. All the modern samples were taken in what is today a pinyon-juniper woodland, although the area around Lillian's Site had been chained during the 1960s and therefore is not a mature pinyon-juniper forest. Lillian's Site and Castle Rock Pueblo are located relatively near modern agricultural fields; Catherine's Site and Saddlehorn Hamlet are at greater distances from modern fields. The modern samples from these four sites produced relatively consistent results: juniper pollen dominates each sample, with pinyon pine contributing a strong secondary value. Sage and cheno-am pollen make the next-highest contributions. The presence of sage indicates a somewhat open understory for the pinyon-juniper woodlands. Plants in the cheno-am category flourish in disturbed environments.
The roof fall samples are all from unburned roof fall strata that cover structure floors, except for the samples from Lester's Site and Saddlehorn Hamlet, which are from burned roof fall. The pollen found in these samples is believed to have been deposited just after the structures were abandoned. In the case of G and G Hamlet and Lillian's Site, this would have been the early to middle A.D. 1200s; for the remaining structures, abandonment is inferred to have taken place in the middle to late 1200s (the sampled structures at Lester's Site, Saddlehorn Hamlet, and Castle Rock Pueblo were probably abandoned in the A.D. 1270s or 1280s). The pollen in the roof fall samples is believed to have been deposited primarily as the result of "pollen rain" in the area around the site at the time the structures were abandoned, but some economic pollen related to human behavior (for example, roof construction, activities conducted on the roof, or the storage of materials on or near the roof) may also be present.
A consistent pattern emerges when the roof fall and floor samples are compared with the modern control samples. Together, the juniper, pinyon, cheno-am, and sage pollen from roof fall and floors make up a smaller percentage of the total pollen sample when compared with the modern controls. Other taxa, including many that could be interpreted as economic pollen, constitute a larger proportion of the roof fall and floor samples relative to the modern samples (see Chapter 17).
Specifically, juniper and pinyon pollen percentages drop in the roof fall samples, although they are still among the most dominant taxa in roof fall. The juniper and pinyon pollen drops in part because cheno-am pollen increases. The rise in cheno-am pollen in roof fall samples (relative to modern samples) indicates that the area surrounding the sites was more disturbed immediately after abandonment than it is today. The ancient disturbance was probably largely the result of two factors: (1) clearing necessitated by the construction of the site itself and (2) clearing of fields for agricultural purposes. The results of analysis of the roof fall samples indicate that there probably was some reduction of the pinyon-juniper woodland in the areas surrounding the sites immediately following abandonment, but the fact that these taxa are still well represented in the samples indicates that there was not necessarily large-scale deforestation. This conclusion is supported by Hovezak's study of the amount of timber required for the construction of Sand Canyon Pueblo (Hovezak 1992). He found that timbers needed for the largest construction elements may have become scarce, but that deforestation of the locality as the result of construction was unlikely.
The floor samples were collected from the contexts most protected from the ambient pollen rain. The pollen in these samples was protected by the roof of the structure during the occupation of the building and by a thick layer of roof fall after abandonment. Most importantly, almost all of these samples were collected from sediments beneath stones that lay directly on the structure floor. Thus, it is highly likely that the pollen from these samples was deposited when the structure was occupied. The samples from Lillian's Site, Roy's Ruin, and Kenzie Dawn Hamlet almost certainly date to the early A.D. 1200s; the samples from Troy's Tower and Catherine's Site date to the middle to late A.D. 1200s; and the samples from Lester's Site, Saddlehorn Hamlet, and Castle Rock Pueblo are well dated to the A.D. 1260s and 1270s.
The floor-associated pollen samples show trends similar to those displayed by the roof fall samples. Together, juniper, pinyon, cheno-am, and sage in the floor samples make up a smaller percentage of the overall pollen record than in the modern control and roof fall samples, which is due in large part to the decreasing percentage of juniper pollen and the increasing percentage of a variety of economic taxa (for example, the Saddlehorn floor sample is 36 percent maize pollen; see Chapter 17). The mean percentages of pinyon and cheno-am pollen in the floor samples are similar to the values in the upper-control samples, but the floor samples are characterized by slightly greater variability from sample to sample. This is particularly evident in the four floor samples from Castle Rock Pueblo, which have exceptionally low percentages of pinyon and juniper pollen and exceptionally high values of cheno-am pollen; the mean value for cheno-am pollen if these four Castle Rock samples are removed from consideration is 22 percent, considerably below the mean value for the roof fall samples. The high Castle Rock values indicate that cheno-ams may have been processed within the sampled structures, whereas the lower values of cheno-am pollen in the remainder of the floor samples suggest that these contexts were protected from the surrounding ambient pollen rain, as measured by the roof fall samples. Pinyon pine also has a mean value that is actually higher in the floor samples than in the roof fall samples (when the four low values from Castle Rock are excluded, the mean is even higher--16.4 percent), and pinyon pine values are particularly high in certain samples at Lillian's, Troy's, Lester's, Catherine's, and Saddlehorn. The samples from Lester's and Saddlehorn were from structures with burned roofs, and all 151 tree-ring samples from these structures were identified as juniper. Thus, the high percentage of pinyon pine pollen does not appear to be related to the structure roofs, but instead may be related to the processing of pinyon inside the structures or to the nearby presence of intact stands of pinyon, which could have resulted in the deposition of pollen on the floors during the use of the structures.
There are lower values of juniper and pinyon pollen and higher values of sage and cheno-am pollen in the roof fall and floor samples when compared with the modern control samples. This suggests more open and disturbed environments in the areas surrounding the sites in the A.D. 1200s. The arboreal pollen is lower in part because of the increasing contribution of economic taxa in roof fall and floor samples. It is difficult, if not impossible, to determine whether the smaller percentage of pinyon and juniper pollen is the result of the increasing percentage of taxa from economic and disturbance species or is the result of local deforestation. But the fact that arboreal pollen is present in relatively high percentages in many of the sealed floor contexts suggests that the local pinyon-juniper woodland was not heavily deforested.
This interpretation is supported by the macrobotanical data. There is substantial diversity in the taxa identified in flotation samples, but pinyon and juniper are the most commonly recovered specimens from both hearths and middens at all of the sampled sites (Adams, Chapter 16). Throughout the A.D. 1200s, pinyon and juniper appear to be the most common fuel resources. In addition to the pinyon and juniper wood charcoal identified in the flotation samples, a wide variety of other plant parts from these two taxa were recovered. The diversity of plant parts suggests that the pinyon-juniper woodland grew within a relatively short distance of the tested sites. Pinus wood is especially common in samples collected from the upper canyon sites, which were occupied in the middle to late A.D. 1200s. It appears that Pinus wood, probably pinyon, was particularly abundant in the upper canyons, even in the last decades of occupation of the Sand Canyon locality.
Other macrobotanical remains indicate that both weedy taxa from disturbed environments and wild plants from undisturbed pinyon-juniper woodlands were being exploited by the inhabitants of the tested sites throughout the A.D. 1200s. There is a slight shift in the post-A.D. 1250 sites toward gathering a wider diversity of weedy resources, a practice which apparently resulted in slightly less reliance on plants of established vegetation. In summary, it appears that the post-A.D. 1250 environment was more disturbed than before, but that pinyon-juniper woodlands survived near the tested sites up to the time of the abandonment of the Mesa Verde region.
Community Organization and Change
One of the main research goals of the Sand Canyon Archaeological Project is to understand the social, political, and economic organization of the Mesa Verde-region communities and how that organization changed through time. The framework for studying the community is structured by four separate, but related, dimensions of organizational variation: scale, intensity, differentiation, and integration.
Scale
Lipe (1992a:4, 1992b:123) defines scale as the size of a community's population and the geographic area it occupies, as well as the "reach" of the community, which is the distance traveled by imported and exported goods. Defining the population and geographic size of the community requires the identification of community boundaries. Adler and Varien (1994) point out that archaeologists in the Mesa Verde region have defined community boundaries in a variety of ways (e.g., Eddy 1977:3; Neily 1983; Rohn 1977). Adler and Varien (1994:84-90) argue that communities in the Sand Canyon locality are best defined on the basis of the clustering of domestic architecture and the distribution of associated public architecture.
Using these criteria, Adler and Varien (1994:89) identified two communities in the upper Sand Canyon survey area: the Sand Canyon community and the Goodman Point community. They argue that the geographic scale of these communities was relatively constant between A.D. 1000 and 1300, with the main cluster of domestic sites located within a 2-km radius of the public architecture. The population of these communities, however, increased over time within their fixed geographic boundaries. Adler (1992c:12-13) calculated momentary population estimates for each community for three post-A.D. 1000 time periods: A.D. 1000-1075, 1075-1150, and 1150-1300. In his calculations, he assumed that sites were occupied for either 25 or 50 years, that there were 1.5 people per room, and that a room is equivalent to 10 m² of surface rubble. On the basis of his calculations, it is estimated that there were between 100 and 200 people living in each community between A.D. 1000 and 1075. Between A.D. 1075 and 1150, an estimated 150-300 people were living in each community, and this number grew to between 200 and 600 people in the A.D. 1150-1300 period.
Adler (1992c) has examined the geographic scale of communities in the Mesa Verde region outside the Sand Canyon locality and found them to range between 40 and 150 km², much larger than the 10- to 15-km² area of the Sand Canyon and Goodman Point communities. These differences in scale are due in part to the different types of data used in the comparison--intensive survey in the Sand Canyon locality and the distribution of large sites and public architecture in the Mesa Verde region as a whole (Varien 1997). The data, however, do indicate that there may have been considerable range in the geographic scale of Mesa Verde-region communities.
The reach of these communities can be examined by looking at nonlocal pottery and nonlocal stone in the tested site assemblages. Nonlocal pottery was rare at the tested sites, making up less than 1 percent of the total pottery assemblages (see Chapter 15). Nonlocal chipped-stone debris (debitage) and nonlocal stone tools were slightly more abundant than nonlocal pottery, but still made up only a small percent of the site assemblages.
Figure 22.2(A) and Figure 22.2(B) show scatter plots of nonlocal pottery by nonlocal debitage. One scatter plot uses the weights of nonlocal pottery and counts of nonlocal debitage; the other uses the percentage of nonlocal pottery (by weight) in the total site pottery assemblages and the percentage of nonlocal debitage (by count) in the total debitage assemblages. The scatter plot created from weights and counts indicates the absolute frequency of these nonlocal materials, whereas the plot created from percentages takes sample-size differences between the sites into account.
These plots show that there is a poor relationship between nonlocal pottery and nonlocal debitage (Rsq = 0.38 and 0.34) but that there is a general pattern with regard to which sites have more nonlocal pottery and which have more nonlocal debitage. Sites above the regression line are those with greater amounts of nonlocal pottery relative to nonlocal debitage, and these are almost exclusively the mesa-top sites. Sites below the line are those at which nonlocal debitage is relatively more abundant, and these are sites within the canyon in both upper and lower Sand Canyon.
G and G Hamlet stands out as the only site with relatively large amounts of both nonlocal pottery and nonlocal debitage, even when sample size is considered. The mesa-top sites, where nonlocal pottery is relatively more abundant, have earlier components than the sites located within the canyon; in fact, with the exception of Catherine's Site, earlier components are virtually absent at the latter sites. The use of G and G Hamlet as a habitation during the Pueblo II period lasted as long as its later Pueblo III occupation, also as a habitation. So time may be a factor affecting the presence of nonlocal pottery, with fewer nonlocal sherds in the latest sites.
It should be kept in mind that nonlocal pottery is rare at all the tested sites, with only 171 red ware (including San Juan, White Mountain, and Tusayan red wares), 11 polychrome, and three nonlocal white ware sherds (see Chapter 15). While rare, nonlocal pottery is found in the highest frequencies at sites with Pueblo II occupations. G and G Hamlet and Kenzie Dawn Hamlet have the most nonlocal pottery (37 and 50 sherds, respectively), and these are the only two sites with Pueblo II habitation components. There is less nonlocal pottery at the mesa-top sites, where Pueblo II occupation was more limited, but even at these sites there is more nonlocal pottery compared with the tested sites that are located within the canyon. The fact that nonlocal pottery is most common at sites with Pueblo II components suggests that the reach of the Sand Canyon community may have become more restricted after A.D. 1150 and that the reach continued to shrink during the A.D. 1200s.
Sites below the regression line are those with greater amounts of nonlocal debitage relative to nonlocal pottery. These are predominantly the sites located within the canyon, and they are single-component sites (with the exception of Catherine's) dating to the middle to late A.D. 1200s. When sample size is considered, the three sites with the highest relative frequencies of nonlocal debitage are Mad Dog Tower, Troy's Tower, and Lester's Site. As previously discussed, Mad Dog and Troy's differ from the remaining sites in many respects. Lester's Site is the tested site located closest to Sand Canyon Pueblo. It currently is not clear why these three sites have the highest percentage of nonlocal debitage. It does appear, however, that the pattern of nonlocal material acquisition shifted during the A.D. 1200s, such that acquisition of nonlocal pottery was exceedingly uncommon, while the acquisition of nonlocal stone became slightly more common.
Neily (1983) studied interaction and exchange between communities in an area 25 km northwest of the upper Sand Canyon locality. His analysis of surface assemblages showed that the pottery and lithic materials at sites became increasingly distinct, emphasizing locally available materials, during the Pueblo III period. On the basis of this observation, he suggested that exchange networks and interaction between communities contracted during this time.
Several ongoing projects are designed to provide greater understanding of exchange in the Sand Canyon locality. Glowacki (1995) is sourcing pottery-production areas by conducting neutron-activation analysis of pottery from Castle Rock Pueblo, Sand Canyon Pueblo, and sites on Mesa Verde. Initial results indicate that distinct production locales can be identified and used to track exchange and interaction. Glowacki's research does document some pottery exchange between Castle Rock Pueblo, Sand Canyon Pueblo, and sites on Mesa Verde in the middle to late A.D. 1200s. In addition, a pilot study was conducted in which the temper in Mesa Verde Black-on-white pottery from each of the tested sites was examined (Thurs 1995; Thurs et al. 1996). Differences noted in the tempers used at each of the tested sites suggest that there is excellent potential for tracking exchange and interaction within the Sand Canyon locality through even more extensive examination of temper.
Intensity
Accurately determining a community's demographic and geographic scale provides the foundation for understanding the dimension of intensity. Lipe (1992a:5) defines intensity as the "amounts of population, material, information, or energy use per unit area or per capita." Intensity as a per-unit-area measure is a common means of distinguishing subsistence intensification (e.g., Boserup 1965), whereas the per capita measure can be used to characterize standard of living. Therefore, intensity is an important variable in the study of sociocultural evolution and the development of social inequality.
Adler's reconstruction of Sand Canyon locality demography (Adler 1992c; Adler and Varien 1994) indicates that there was steady population growth within fixed geographic boundaries between A.D. 1000 and 1300. Lipe (1992b:128) points out that this, by definition, constitutes intensification, because humans were extracting more energy per unit area through time. Faunal data and agricultural features at the tested sites can be used to examine the question of intensification in greater detail.
Munro (1994) analyzed turkey bone from the tested sites and reviewed faunal data from sites throughout the Mesa Verde region, including from the tested sites. First, she addressed the question of whether the turkey remains found at sites were the bones of wild or domesticated birds. She found the osteological evidence to be equivocal: not only was it difficult to distinguish morphological changes that were the result of domestication, but her data also did not support the proposed division of turkeys into three distinct breeds (cf. McKusick 1980, 1986). On the other hand, the contextual evidence, such as the presence of eggshell, gizzard stones, droppings, age structures, and pens, was clear evidence that domesticated turkeys were present at Sand Canyon locality sites (Munro 1994:142).
Next, Munro examined whether turkeys were used in the Sand Canyon locality predominantly for ritual or for food. Direct evidence of the ritual use of turkeys was not found in the tested site sample, although archaeological evidence from elsewhere in the Mesa Verde region and ethnographic evidence from Southwestern Pueblo societies indicate that ritual use was likely. The criteria used for documenting food utilization include cut marks, breakage patterns, burning, ossified tendons, dispersion of bone, and age/sex profiles (Munro 1994:105-122, 145). The tested sites data clearly demonstrate that turkeys were being used for food (Munro 1994:145-149).
Having established that turkeys were an economic resource, Munro addressed the question of whether turkey production was intensified through time. To do this, she examined the proportion of turkey bone relative to other economic fauna (Munro 1994:25). Turkeys were used throughout the Basketmaker III-Pueblo III periods in the Mesa Verde region, but their use intensified dramatically during the Pueblo II period. Munro examined sites from throughout the region and found that in Basketmaker III and Pueblo I sites turkey bone makes up approximately 2 percent of the economic fauna. This figure increases to 12 percent during the Pueblo II period (Munro 1994:125-126), and by the Pueblo III period, turkey bone makes up an average of 49 percent of the economic fauna at the sites examined (Munro 1994:122-126). Faunal assemblages from the tested sites are similar to this regional sample; an average of 46 percent of the economic fauna from the Sand Canyon locality tested sites is turkey.
Examination of intersite variation in faunal assemblages at the Sand Canyon locality tested sites reveals distinct patterns. First, intensification of turkey use relative to early periods had occurred by the early A.D. 1200s. Second, turkey use continued to be intensified throughout the 1200s. Turkey bone at the mesa-top sites in upper Sand Canyon make up an average of 36 percent of the total economic faunal assemblage (Munro 1994:131). Frequencies at sites dating to the second half of the thirteenth century are even higher, but these vary depending on site location. The percentage of turkey bone in the faunal assemblages at Castle Rock Pueblo and Saddlehorn Hamlet in lower Sand Canyon are 45 and 44 percent, respectively. At the four sites located on the talus slopes in upper Sand Canyon (Catherine's, Stanton's, Lester's, and Lookout House), turkey bone ranges between 66 and 70 percent of the total assemblage (Munro 1994:131). This contrasts with Sand Canyon Pueblo, which also dates to the late A.D. 1200s, but where only 37 percent of the faunal assemblage is turkey bone.
Munro interprets these data as indicating that turkey production was definitely intensified relative to earlier periods at all of the Sand Canyon locality sites after A.D. 1150. This intensification is interpreted as having been a response to the increased scarcity of wild game due to over-hunting (Munro 1994:156). The intersite variation within the Sand Canyon locality during the Pueblo III period is viewed as evidence of further intensification of turkey production during the A.D. 1200s, particularly among the residents of the small sites located on the talus slopes in upper Sand Canyon, which were occupied in the middle to late A.D. 1200s. Munro suggests that the occupants of these talus-slope sites were late migrants to the Sand Canyon locality and that they received land that was marginal for farming. She believes that turkey production was intensified as a dietary supplement, a response that was stimulated by decreased crop yields due to the marginal agricultural land (Munro 1994:162).
There is also evidence of agricultural intensification at the tested sites. Contour terraces are present at two of the sites located on the talus slopes in upper Sand Canyon--Catherine's Site and Lester's Site. These terraces are interpreted as agricultural features that would have caused sediment to accumulate to greater depths, thereby improving the moisture-retaining capacity of the soil. Their location on the talus slopes, close to small intermittent drainages, indicates that they may have been one component of a larger system of runoff irrigation. At Catherine's Site, some of these terraces are downslope from what may have been an active spring, and this area may have been the location of gardens that were irrigated from this spring (see Chapter 11).
The contour terraces are simple stone alignments, and they do not appear to have great depth. Nevertheless, they are evidence that the talus slopes were being used for agriculture. This is tentative support for Munro's model, but much remains to be learned about the significance of these talus-slope contour terraces. We need to know if the fields associated with these terraces provided the majority of the agricultural land used by the inhabitants of Catherine's and Lester's, or if the contour terraces mark the locations of smaller gardens which supplemented agricultural production in larger fields located in other areas. Another empirical question that remains to be addressed is whether intensively cultivated areas on the talus slopes are in fact marginal when compared with mesa-top fields. If these fields were watered and manured (for example, with turkey dung), could their production have exceeded that of the dryland fields on the mesa tops? Regardless of how these contour terraces were used, they are evidence of agricultural intensification.
There is also the important question of when these features began to be used. Agricultural features are notoriously difficult to date, but elsewhere in the Mesa Verde region, there is evidence that these features were first used during the Pueblo II period (Greubel 1991:124; Hayes 1964:79; Rohn 1977:262) and that their use intensified during the Pueblo III period (Adler 1990b:318; Fetterman and Honeycutt 1987:104; Greubel 1991:124, 140; Winter 1976:289-290, 1977:188-208). Catherine's Site is the only site located on the talus slope in upper Sand Canyon that has enough Pueblo II pottery to suggest that there was limited use of the site at this time (that is, between A.D. 1025 and 1150). Therefore, the agricultural use of the area around Catherine's Site, which may have involved irrigation from a spring, could have begun during the Pueblo II period.
Together, Adler's paleodemographic reconstruction, Munro's analysis of turkey bone, and the evidence for agriculture on the talus slopes are strong evidence for intensification during the Pueblo II and III periods when compared with earlier times. Intensification appears to have continued throughout the Pueblo III period up until abandonment in the late A.D. 1200s. This can be measured in terms of increased energy inputs both per unit area and per capita.
Differentiation
Lipe (1992a:4, 1992b:124) examines community differentiation as a means of evaluating social complexity. Following Blanton et al. (1981:21), he distinguishes two types of differentiation: horizontal differentiation refers to the functional specialization of parts of equal rank; vertical differentiation refers to ranking among functionally diverse parts. The Testing Program was designed to address the question of community differentiation through comparative analyses of site architecture, site size, site assemblage composition, and human remains.
In general, there is a great deal of similarity among all the tested sites. Each contains above-ground and subterranean architecture. The pottery on each site consists of gray ware cooking jars, white ware storage jars, and white ware serving bowls. Stone assemblages consist of abundant chipped-stone debris, similar types of chipped-stone tools, and similar kinds of ground-stone tools. All sites have faunal assemblages dominated by rabbit and turkey bone, with smaller amounts of deer, carnivore, and bird bone other than turkey; the assemblage of rodent bone on all sites is dominated by sciurids and packrats. Macrobotanical and pollen remains likewise display similar taxa from one site to another. The variability between sites, which is the basis for interpreting community differentiation, is a matter of degree rather than kind.
Differences in site size and in the type of architecture present at sites, which were noted previously in the section on length and season of site occupation, suggest that there may have been some functional differentiation within the Sand Canyon Pueblo III community. Mad Dog Tower and Troy's Tower are the smallest sites tested. A kiva and tower connected by a tunnel are the main architectural features at these sites; surface roomblocks, which are present at other tested sites, are absent. As discussed in the regional overview in Chapter 21, there are fewer tower-kiva sites than unit pueblos, but the former are not uncommon. Tower-kiva sites have been documented throughout Mesa Verde National Park (Hayes 1964:112; Rohn 1977:116-117; Smith 1987:36, 39, 44), and an examination of Bureau of Land Management files indicates that sites consisting of only a tower and an associated kiva have been recorded during many of the smaller survey projects in the region. In addition, isolated towers have been reported (Fetterman and Honeycutt 1987), and these may be associated with kivas that are not visible on the modern ground surface.
The fact that Mad Dog Tower and Troy's Tower are small and have similar architecture, and that less material accumulated at these two sites than at any of the other tested sites, was cited earlier to suggest that they may have been seasonally occupied. Beyond these observations, however, the similarities begin to break down. For example, when the details of their respective artifact assemblages are compared, the only similarity between Mad Dog and Troy's is the unusually high frequency of bowls (see Chapter 15). Mad Dog Tower has the lowest frequency of white ware jars (14.3 percent of the total pottery assemblage) of any of the tested sites, whereas Troy's Tower has the second-highest frequency (18.7 percent). Mad Dog Tower has high frequencies of projectile points and bifaces relative to the total chipped-stone assemblage; Troy's Tower does not.
The relatively few white ware jars at Mad Dog Tower might be explained by a shorter occupation span at this site. White ware jars are believed to have been used for storage, and cross-cultural research indicates that storage vessels have some of the longest use lives of any functional category of pottery vessel (Mills 1989b:137; Varien and Mills 1997). Sherds from these vessels would accumulate at a much slower rate, and their frequency relative to other artifacts would not stabilize until the site had been occupied for a sufficient length of time (Mills 1989b). A short occupation, however, does not account for why Mad Dog Tower has high frequencies of projectile points and bifaces. In addition, the hearth in the kiva at Mad Dog Tower had been remodeled, suggesting that the occupation was not excessively short (see Chapter 13).
Troy's Tower stands out from all of the other tested sites in terms of macrobotanical remains because it yielded the greatest diversity of reproductive plant parts (N = 16; see Chapter 16 for an inventory of the identified taxa). This is so despite the fact that fewer contexts from this site were analyzed. In contrast, a single unidentified seed was the only reproductive plant part found in the flotation samples from Mad Dog Tower.
Together, the artifact and macrobotanical data from Troy's Tower and Mad Dog Tower support the interpretation that these two sites were functionally distinct compared with the other tested sites. However, the data also suggest that Troy's Tower and Mad Dog Tower were functionally distinct from each other, something not readily evident when only architecture and site size are considered.
It is difficult to specify why these sites have such distinct assemblages. One hypothesis is that both sites were in fact habitations, but that each was occupied by a household of atypical composition (for example, unmarried males or females, or an elderly couple with no children). Another possibility is that they were habitations that were abandoned before the completion of a full domestic cycle (for example, before a couple had children). Finally, it may be that Troy's and Mad Dog each had their own specialized functions but that each was used periodically within an annual cycle repeated over many years or decades.
Castle Rock Pueblo was the largest site tested and the largest known site in the lower Sand Canyon settlement cluster. An important question that relates to community differentiation is how Castle Rock was related to the surrounding settlement cluster. The final report on the intensive excavations at Castle Rock will address this question more fully, but the results of the Testing Program can be used to offer preliminary interpretations here. The Castle Rock pottery assemblage has the highest percentage of corrugated jar sherds and the lowest percentage of white ware jar sherds of any of the tested sites. A high percentage of corrugated jar sherds is something that characterizes all three lower Sand Canyon sites. White ware jar sherds, on the other hand, were found in low frequencies at Castle Rock Pueblo and Mad Dog Tower, but in higher frequencies at Saddlehorn Hamlet.
The corrugated jar assemblage at Castle Rock Pueblo is dominated by medium-size vessels. In this regard, Castle Rock is similar to the other tested sites (see Chapter 15) but different from Sand Canyon Pueblo, where there are many more large jars (Christopher Pierce, personal communication 1995). Larger vessels may be linked to feasting and ceremony at Sand Canyon Pueblo (cf. Blinman 1988a; Driver 1996) or to the presence of larger coresidential groups, which are also food-preparation and consumption groups (cf. Nelson 1981; Tani 1994; Turner and Lofgren 1966). The layout of Sand Canyon Pueblo and Castle Rock Pueblo differ. Ricky Lightfoot (personal communication 1995) has noted that the Castle Rock site layout consists of spatially discrete kiva units, which contrasts with Sand Canyon Pueblo, where multiple kiva units are grouped into larger architectural blocks. It may be that food-preparation and consumption groups were smaller at Castle Rock than at Sand Canyon Pueblo. Nicklaw (1995) argues that there was functional differentiation among kiva units at Sand Canyon Pueblo, with some units being more specialized for ritual activity.
Potter (1994, 1997) has argued that bird (excluding turkey) and carnivore bone is best interpreted as having had a ritual use. Castle Rock does have relatively large amounts of bird bone, although this may be simply a result of the larger sample size from this site. Driver (Chapter 18; Driver 1996) has noted that the overall faunal assemblage at Castle Rock is most similar to the assemblage at Saddlehorn Hamlet, which was occupied at the same time, and to the assemblages at the mesa-top sites in upper Sand Canyon, which were earlier. The mesa-top sites are dominated by cottontail bone, in contrast to the talus-slope sites in upper Sand Canyon, where turkey bone is most common. Collectively, the tested site assemblages differ from the Sand Canyon Pueblo assemblage, which has a higher frequency of artiodactyl bone. As discussed in the next section on integration, Driver et al. (Chapter 18) interpret the higher percentage of deer bone at Sand Canyon Pueblo as having resulted from increased ritual, ceremony, and associated feasting at that site.
Thus, the pottery and faunal assemblages from Castle Rock Pueblo differ from those at the other tested sites in many respects, but they are also different than the pottery and faunal assemblages from Sand Canyon Pueblo. Castle Rock Pueblo may have functioned as a center for a surrounding community in lower Sand Canyon, but it appears to have played a much different role than Sand Canyon Pueblo.
The remaining tested sites have similar architectural layouts: all are unit pueblos with masonry roomblocks, masonry-lined kivas, and formal middens; some have masonry towers as well. On the basis of these features and the artifact and ecofact assemblages present, these sites are interpreted as year-round habitations. Detailed analysis, however, reveals considerable intersite variation in their artifact assemblages.
In terms of pottery assemblages and the vessel forms present, Stanton's Site stands out for having an unusually high frequency of white ware bowls (35.8 percent of the total assemblage), with no other site approaching this frequency (the next-highest percentage is at Lester's Site, with 26.0 percent). In addition to having the highest relative frequencies of white ware bowls, Stanton's Site has the largest number of projectile points and bifaces and the largest amount of obsidian. And Stanton's Site has the greatest amount of turkey bone at any of the tested sites.
Castle Rock Pueblo and Kenzie Dawn Hamlet also have significantly more bowls than expected. Lookout House and Catherine's Site have relatively high frequencies of bowls, but their frequencies do not represent a statistically significant departure from the expected. The size of the bowls at each of the tested sites is generally similar; however, Lookout House and Kenzie Dawn have unusually high frequencies of medium-size bowls, whereas the two tower-kiva sites--Mad Dog Tower and Troy's Tower--have the highest frequencies of large bowls. Lookout House and Kenzie Dawn have the highest frequencies of white ware jars (21.8 and 21.2 percent, respectively, of the total assemblage, when the Basketmaker III plain gray jars are excluded). Kenzie Dawn also has the greatest number of manos and slab metates.
It is possible that at least some of the intersite variation in assemblage composition is the result of economic specialization within the community. Descriptive models of specialized production typically view economic specialization as a multidimensional continuum along which the organization of production ranges from relatively simple to more complex (Costin 1991; Rice 1987, 1991). The intensification of household production--when items are produced not only for use but also for exchange--represents a type of economic specialization in which the organization of production is relatively simple. Household specialization is therefore often seen as the beginning of more complex types of specialized production.
Specialized household production of white ware bowls may account for the high relative frequency of these sherds at Stanton's Site. Ethnoarchaeological studies indicate that households with potters tend to have larger systemic inventories, and having more pots could lead to the greater accumulation of pottery at the site (Arnold 1988; Deal 1988; Nelson 1991). There has been no systematic study of grave offerings in the Mesa Verde region, but available data suggest that white ware bowls are the most common pottery vessels found as grave offerings (Rohn 1977:93; Cattanach 1980:142-145). Therefore, bowls may have had considerable exchange value in addition to their use value, especially when compared with other types of pottery. Greater exchange value may have stimulated the intensified production of bowls. Along similar lines, Abbott (1994) documents specialized production of red wares by the Phoenix-basin Hohokam and argues that this may have been stimulated by red wares having greater exchange value as a result of their use in mortuary contexts.
Distinct from the descriptive models that provide a framework for discussing specialized production are explanatory models that attempt to specify the factors that stimulate the development of specialized production. Some of these explanatory models link the development of specialized production to the existence of a subset of a society that is forced onto agriculturally marginal land (Arnold 1985; Graves 1991). In these models, people on agriculturally marginal land turn to craft specialization as means of buffering shortfalls in food production. Ethnographic studies support the notion that specialized pottery production is conducted by people who do not have access to agriculturally productive land; pottery production is seen as an undesirable alternative to agriculture, and when the opportunity to farm becomes available, pottery producers typically abandon their craft (Arnold 1985). At its most rudimentary stages, craft specialization that was stimulated by households being agriculturally marginalized would result in small-scale, part-time, independent specialists who produced sporadically to supplement a household income (Costin 1991).
As discussed previously, Munro postulates that the inhabitants of the talus-slope sites in upper Sand Canyon, such as Stanton's, intensified turkey production because the only agricultural land to which they had access was of marginal quality. Turkeys are a food resource that could have buffered shortfalls in agricultural production, and their feathers may have had exchange value. It may be that the inhabitants of Stanton's Site also specialized in the production of nonfood items, such as white ware bowls, which could have been exchanged for food during difficult years.
This scenario, however, is not supported by other lines of evidence. For example, the assemblage data can be used to argue that life at Stanton's Site was anything but marginal. For example, the inhabitants apparently had more bowls, projectile points, and bifaces than did their neighbors in the mid A.D. 1200s, and they had more of these items than did the inhabitants of the mesa-top sites that were occupied slightly earlier. The residents of Stanton's Site also had more obsidian, presumably a highly prized nonlocal resource. The higher percentage of turkey bone and the results of isotopic analysis, which suggest that the turkeys were eating corn (Munro 1994:154), also indicate that the residents of Stanton's Site may have been better off than most (it is unlikely that agriculturally marginal groups could have produced a surplus of corn to feed to their turkeys). Finally, the estimates of the total amount of corrugated pottery discarded at Stanton's Site indicate that this site was occupied as long as, or longer than, any of the other tested sites (Figure 22.1). A long occupation also seems to contradict the interpretation that the inhabitants of Stanton's Site were economically marginalized.
The case for economic specialization can be further examined by looking at the evidence of pottery production. There is direct evidence for pottery production at most of the tested sites. Possible base molds were found at four of the tested sites (G and G Hamlet, Lillian's Site, Castle Rock Pueblo, and Stanton's Site), unfired pottery was found at three sites (Kenzie Dawn Hamlet, Lookout House, and Castle Rock Pueblo), and raw clay and polishing stones were found at most sites. Only two sites, Mad Dog Tower and Saddlehorn Hamlet, lack direct evidence for pottery production, and in both cases, the lack of evidence may be the result of relatively small sample size. All of the direct evidence for pottery production relates to the manufacture of white wares; the absence of direct evidence for the production of gray wares is puzzling. Additional study of indirect indicators of production (for example, compositional analysis) is needed to examine gray ware production at the tested sites.
Studies of stone and bone artifacts also suggest local production of these items by the households at each site. Chipped-stone debris from four of the tested sites was examined in detail, and considerable variation was noted in the color and graininess of the Morrison Formation materials recovered. This suggests that the occupants of each site produced tools from different sources of the Morrison Formation materials, probably the materials that were easiest to obtain. It also appears that the raw materials available for the production of bone tools varied from site to site, and the bone tools reflect this variation. This suggests that bone tools were likewise produced at each site from locally available materials.
Although it seems plausible that variation in site assemblages is linked to specialized production by households, future research should also examine the possibility that more mundane phenomena account for this variation. For example, sampling error and differences in formation processes, such as variation in length of site occupation, might also result in, or at least contribute to, assemblage variation.
Driver et al. (Chapter 18) do not emphasize economic specialization in their explanation of the variation in relative frequencies of faunal remains at the tested sites. Instead, they focus on the organization of ritual and ceremony within the community. This interpretation is discussed in greater detail in the next section on community integration.
Integration
In studies of community differentiation, researchers attempt to identify the structural parts that make up a community, identify how they functioned, and explain the interdependence between them (Lipe 1992a:5, 1992b:127). Lipe (1992a:5) acknowledges that community integration can be achieved in various ways, ranging from consensual decision making to coercive control. Ideology and ritual are also seen as important means of achieving community integration (Hegmon 1989). Theoretical perspectives differ as to the role played by ideology and ritual; some view ideology and ritual as integrating societies through a set of shared beliefs (e.g., Durkheim 1965), whereas others view ideology as a means by which those who have power manipulate those who lack power (Wolf 1984). The Testing Program results will not resolve the question of the role played by ideology and ritual within society, but these analyses bring the question of community integration into sharper focus.
Kivas have traditionally been interpreted as architectural facilities that promoted integration, functioning in a manner analogous to historic-period kivas. Lipe and Hegmon (1989) have recently reviewed the historical and analytical perspectives that underlie this interpretation, with particular attention to its application in the northern part of the Pueblo Southwest, and they discuss recent challenges to this traditional interpretation (e.g., Lekson 1988; Cater and Chenault 1988). A point highlighted by Lipe and Hegmon (1989:16) is that the interpretation of kiva function is typically stated in either-or terms--kivas functioned either as specialized ceremonial facilities or as pit structures for domestic use.
Papers in the volume edited by Lipe and Hegmon (eds. 1989) present a reassessment of this traditional interpretation. The general conclusion of this reassessment is that either-or interpretations are too simplistic. There is general support for Lekson's (1988) proposition that kivas commonly had domiciliary functions through the Pueblo III period, but pit structures and kivas also housed ritual activities that may have integrated small segments of the community (Lipe and Hegmon 1989:17). The interpretation that most Mesa Verde-region kivas had both domestic and ritual uses and that they integrated relatively small segments of communities is supported by Adler's (1989) cross-cultural research. Adler found a distinction between larger integrative facilities, which serve entire communities and are functionally specialized for ceremony, and smaller facilities, which serve fewer social segments and have a more generalized function that includes both ritual and domestic use.
Lipe (1989:55) argues that the formal architectural properties of Mesa Verde kivas likely reflect some consistent functional or cognitive differences between kivas and other types of buildings. Formal differences include ritual features like sipapus and floor vaults (cf. Wilshusen 1989b) and the subterranean character and roof entry of the buildings themselves (Lipe 1989:64). Lipe (1989:64) believes that specific architectural attributes of kiva architecture are likely related to a myth of original emergence from the underworld or underworlds that is universally important in historic Pueblo belief systems, and was likely widespread in the past. He argues that kivas and pit structures, which had domestic use but which were also heavily invested with religious symbolism and which were the locus for ritual activities organized by households, may have actually worked against strong community-level integration (Lipe 1989, 1992b:127).
The size of the social unit that used a kiva is commonly inferred on the basis of the ratio of kivas to surface rooms (Lekson 1988; Lipe 1989; Steward 1937). Over 30 pit structures dating to either the Pueblo II or Pueblo III periods were tested as a part of the Site Testing Program. Of these, 24 are masonry-lined kivas, one is partly lined with masonry, and nine are earth-walled structures dating to the late A.D. 1000s or early 1100s. The number of surface rooms at sites was estimated by excavating shallow trenches around walls and by measuring the size of the rubble mounds. The number of surface rooms associated with a single kiva ranges from zero at Troy's Tower to perhaps as many as 10 at Lillian's Site. Most mesa-top sites have five to seven rooms per kiva, whereas the sites located on the talus slopes in upper Sand Canyon appear to have fewer than five rooms associated with each kiva. On the basis of these estimates, kivas at the tested sites are inferred to have been used by small social groups, probably single large households.
Elsewhere in the Southwest during the A.D. 1200s, kivas do not appear to have been used by such small social segments. A ratio of 1:23 "kivas to secular rooms" is reported for the Pettit site, which is located in the upper Zuni drainage and which dates to the early A.D. 1200s (Saitta and Keene 1990:217). Certain surface rooms, labeled "limited activity rooms," may have had nonresidential functions; this interpretation is based on the presence of architectural features similar to those found in kivas (Saitta and Keene 1990:217-219). The ratio of nonresidential to residential structures drops to 1:6 when these unusual rooms are counted as integrative, rather than as domestic, architecture. Regardless of problematic interpretation of these surface rooms, the kivas themselves were integrative at a larger social scale at this site, when compared with kivas at the Sand Canyon locality tested sites. Sites in the Mesa Verde and Zuni regions became even more different in terms of site size and layout during the late A.D. 1200s, with small kivas continuing to proliferate in the Mesa Verde region but even fewer kivas being constructed in the Zuni region.
As discussed above, some tested sites have kivas that may not have been used by households as domiciles. These include the kivas at the smallest sites, Troy's Tower and Mad Dog Tower. In addition, kiva use may have been more variable at the largest sites, Castle Rock Pueblo and Sand Canyon Pueblo. In contrast to the situation at the Pettit site discussed above, the trend at these late, aggregated, Mesa Verde-region sites is toward greater numbers of kivas with even lower kiva-to-room ratios. Bradley (1993) argues that some kivas at Sand Canyon Pueblo were used for domestic purposes and were part of the household architecture, whereas others had more specialized uses and were probably not associated with individual households. The same may be true at Castle Rock Pueblo. Structure 402 is a stone enclosure with an associated tower located to the north and a single isolated room located to the south. Just to the east of the enclosure are Structures 405 and 406--two kivas located adjacent to one another, each with only a single associated room. These architectural units do not appear to be standard kiva suites, and they may have been used for nonresidential purposes. Ongoing research at Castle Rock will examine this problem in greater detail.
The analysis of faunal remains from the Sand Canyon locality sites also contributes to our understanding of community integration (Driver et al., Chapter 18; Driver 1996). Intersite variation in faunal assemblages indicates that Sand Canyon Pueblo has much higher relative frequencies of deer than do the tested sites and higher cottontail-to-turkey ratios when compared with the contemporaneous tested sites on the talus slopes in upper Sand Canyon. Small sites located on the talus slopes in upper Sand Canyon have higher frequencies of turkey. The remaining mesa-top sites (in upper Sand Canyon) and Castle Rock and Saddlehorn (in lower Sand Canyon) have higher frequencies of rabbits.
Driver et al. (Chapter 18) reject hypotheses that explain these differences in relative frequencies as being the result of environmental change or economic specialization. Instead, they believe that the social and ritual importance of faunal resources is the key to understanding variation in the Sand Canyon locality faunal assemblages. Faunal assemblages at the tested sites are interpreted as reflecting strictly subsistence needs, whereas the content of the Sand Canyon Pueblo faunal assemblage is viewed as having been more heavily influenced by hunting for ritual and social purposes.
The social and ritual importance of hunting is documented for ethnographically known Southwestern societies (Potter 1997:356-359). Deer hunting was particularly important in these societies and was often associated with ritual activity. In addition, the relatively high cottontail-to-turkey ratio at Sand Canyon Pueblo, when compared with the nearby contemporaneously occupied sites on the talus slopes, may be evidence for communal rabbit hunting. These data suggest that feasting associated with ritual and ceremony was more common at Sand Canyon Pueblo than at any of the tested sites. The interpretation that ritual and ceremony were intensified at Sand Canyon Pueblo is supported by the presence of public architecture at this site, including a great kiva and a D-shaped bi-wall structure (Bradley and Churchill 1995). Preliminary analysis of the distribution of faunal remains at Sand Canyon Pueblo indicates that deer bone occurred in the highest frequencies in roomblocks with the highest kiva-to-room ratios, suggesting that certain kivas at the site may have been focal points for feasting on hunted animals and feeding large groups of people (Driver et al., Chapter 18).
The lower Sand Canyon sites of Castle Rock Pueblo and Saddlehorn Hamlet, which were contemporaneous with Sand Canyon Pueblo, have frequencies of deer bone similar to those at the mesa-top sites in upper Sand Canyon, which were occupied earlier than Sand Canyon Pueblo. Driver et al. suggest that the inhabitants of Castle Rock, Saddlehorn, and the upper Sand Canyon mesa-top sites had access to deer, but not to the same quantity of deer available to the residents of Sand Canyon Pueblo. The upper Sand Canyon talus-slope sites have lower frequencies of both deer and cottontail bone when compared with the other tested sites and with Sand Canyon Pueblo. Driver et al. (Chapter 18) interpret this as indicating that there was differential access to the procurement and/or the consumption of deer and cottontails.
Driver et al. (Chapter 18) propose that the inhabitants of Sand Canyon Pueblo may have controlled the procurement and/or consumption of deer and cottontails. They develop a number of hypotheses to explain how this control may have been accomplished. In one scenario, the inhabitants of Sand Canyon Pueblo denied access to hunting territories to the residents of the contemporaneously occupied talus-slope sites. In this model, the residents of the talus-slope sites were economically deprived relative to the inhabitants of Sand Canyon Pueblo. A second model sees the inhabitants of the smaller talus-slope sites as having been fully integrated into both the hunting of large game and the ceremonial events where this meat was consumed; the faunal assemblages differ only because the ceremonial events and associated feasting were held exclusively at Sand Canyon Pueblo rather than at the smaller sites. In this hypothesis, inhabitants of the small sites and Sand Canyon Pueblo are not economically differentiated. Finally, it is possible that deer meat was not obtained through hunting but, rather, through exchange with hunter-gatherers living on the margins of the Mesa Verde region (see, for example, Spielmann [1991] for essays that describe mutualistic exchange relations between agriculturalists and hunter-gatherers, and Fetterman and Honeycutt [1990] and Reed [1995], who raise the possibility that hunter-gatherers were living on the margins of the Mesa Verde region). The issue of whether or not Sand Canyon Pueblo residents controlled this exchange would be a separate problem, distinct from documenting that deer meat was obtained through exchange rather than through direct procurement. A detailed study that examines the intrasite patterning of faunal remains at Sand Canyon Pueblo is in progress and will attempt to determine which of these hypotheses is correct.
Abandonment
The abandonment of structures, sites, localities, and regions is addressed by examining four sources of data: (1) stratigraphic observations that allow evaluation of how pit structure/kiva roofs were treated at the time structures were abandoned, (2) artifact densities in the fill of these structures, (3) sediment analyses of pit structure/kiva fills, and (4) tree-ring dates from the Sand Canyon locality and the Mesa Verde region. These analyses build on the work of numerous archaeologists who have studied issues related to abandonment, including Cameron (1990, 1991), Lightfoot (1993, 1994), Reid (1985), Schlanger and Wilshusen (1993), Wilshusen (1988a, 1988b), and Wilshusen and Schlanger (1993).
Structure Abandonment and Occupation of the Locality
Ancient kiva roofs were massive constructions. Using observations of intact kiva roofs in Mesa Verde-region cliff dwellings, Hovezak (1992:68) estimates that each roof required between 96 and 192 large timbers. The fact that these roofs were so massive is the basis for a number of assumptions that underlie the following analyses.
First, it is assumed that traces of unburned timbers would still be present if unburned roofs had been left behind when structures were abandoned. Unburned timbers are found in some abandoned structures, but as the data reported below indicate, this is a rare occurrence.
The second assumption addresses why the remains of unburned roofs are so rare. The acquisition of timbers for roofs is assumed to have required a substantial investment in energy. Roof timbers were therefore an important economic resource that, whenever possible, would have been salvaged and recycled. Recycling would include the reuse of timbers as construction elements and reuse as fuelwood.
Finally, accidental, catastrophic fires that destroyed entire roofs are believed to have been rare occurrences because roof timbers were so tightly packed and because the timbers were covered with a thick layer of earth. Support for this final assumption is provided by experimental burnings of aboriginal-style dwellings; in these cases, the structures burned only after having been intentionally set on fire and then only after several hours of trying to get the fire to ignite (Wilshusen 1986:247; also see Cameron 1990). Therefore, burned roofs are interpreted as having been the result of intentionally set fires unless there is compelling evidence to believe otherwise.
With these assumptions in mind, the treatment of kiva roofs in the Sand Canyon locality can be analyzed. Forty-five structures are examined, including all of the Pueblo II and III pit structures and kivas investigated during the Testing Program, 10 kivas at Sand Canyon Pueblo, and the single excavated kiva at Green Lizard (Huber 1993). This analysis excludes the three tested-site pit structures for which excavation was too limited to determine construction methods. Four general types of roof treatment were identified: (1) roofs that were burned intact, (2) roofs that were burned after the large timbers had been salvaged, (3) roofs that were unburned and abandoned with all roof timbers left to rot in place, and (4) roofs that were unburned but whose timbers had been salvaged and recycled. In addition, an "unburned-unknown" category was used. The frequencies of these four types of roof treatment are illustrated in Figure 22.3.
Structures in which the entire roof burned are easy to identify because a thick stratum of burned roof fall containing hundreds of burned timbers is present. Entire burned roofs were found in eight of 45 pit structures (18 percent) examined, including five cases at Sand Canyon Pueblo and three cases at the tested sites (Lester's Site, Saddlehorn Hamlet, and Castle Rock Pueblo). All eight cases are structures with kiva-style architecture. In seven of the eight cases, the burned roof fall was found directly on the floors, indicating that the roofs burned near or at the time of kiva abandonment. The single exception is a structure in which a thin stratum of naturally deposited sediment lay between the floor and the burned roof fall, indicating that the kiva had been abandoned for a time before the roof was burned.
Structures that appear to have burned after the large roof timbers were salvaged are not reported extensively in the literature. Although in the sites under consideration it was clear that there were cases in which the large roof timbers had been salvaged, the inference that smaller construction elements had been left behind and burned at the time of structure abandonment was more problematic. Evidence that large timbers were salvaged and smaller roofing elements were subsequently burned includes (1) roofing sediments lying directly above the floor; (2) the presence of abundant charcoal in the fill above the floor, with individual pieces measuring no larger than 7 cm in diameter; and (3) the presence of oxidation on walls and/or small patches of ash on floor and bench surfaces. Using these criteria, this type of roof treatment (burned after the removal of large timbers) is inferred as having occurred in nine (20 percent) of the structures considered.
The next category of structure abandonment--unburned roof timbers left to rot in place--is recognized during excavation by the presence of rotting timbers in the fill above floor. This situation occurred in five structures (11 percent) at two sites.
Salvaging timbers from unburned structures is the most common type of roof treatment, occurring in 22 (49 percent) of all cases examined. The inference that timbers were salvaged is made on the basis of the following observations of the stratum covering the floor of an abandoned structure: (1) the sediments were culturally deposited, (2) the sediments include unburned roof fall, and (3) roof timbers are absent and there is no evidence to suggest that they were ever a part of the stratum (that is, it does not appear that timbers were present originally but later completely decayed). The details of the stratigraphic arguments that support these interpretations are presented in the descriptions of the individual structures in this report, but the presence of unburned adobe, some with beam impressions, was one important reason why these strata were interpreted as unburned roof fall, and sediment analyses support this interpretation (Kilby 1998).
The distribution through time of each of these types of roof treatment was plotted for four partly overlapping time periods: A.D. 1060-1090, A.D. 1140-1240 (includes structures built before 1225 and abandoned by 1240), A.D. 1225-1260 (includes structures built after 1225 and abandoned by 1260), and A.D. 1250-1280 (includes structures built after 1250) (Figure 22.4). Structures were dated using tree-ring samples when available; otherwise, dates were inferred on the basis of stratigraphy, pottery, and archaeomagnetic dating analysis. There were only two structures that dated to the earliest period, A.D. 1060-1090, and both had unburned roofs with salvaged timbers. During the next period, A.D. 1140-1240, unburned structures with salvaged roof timbers again is the most common type of roof treatment. In addition, there are two cases of burned structures with salvaged timbers and three cases of unburned roofs with unsalvaged timbers left to rot in place. The latter occurred at three small, earth-walled pit structures at Kenzie Dawn Hamlet; it may be that the timbers were not salvaged because the structures were smaller and therefore had smaller timbers of less value. The next period, A.D. 1225-1260, sees a continuation of the trend, with 10 unburned and three burned structures, all with salvaged timbers. The final period, A.D. 1250-1280, exhibits the greatest diversity in roof treatment at the time of structure abandonment. It is the only period in which unburned structures with salvaged timbers (five cases) is not the most common mode of abandonment. It is also the period when burned structures with salvaged timbers are the most common (four cases). Finally, all structures with entirely burned roofs date to this period.
To complement these studies, a detailed geoarchaeological analysis of the fill sequences in 10 of the tested-site pit structures was conducted (Kilby 1998), including structures from the range of site locations investigated during the Testing Program. The strata from pit structure fill were evaluated using the following analyses: (1) particle-size analysis, including calculating the median particle size and the sorting coefficient; (2) organic content; (3) calcium carbonate content; and (4) microartifact content. Particle-size analysis was conducted to determine the type and strength of transport agents; organic content and calcium carbonate content were evaluated to help determine the degree of soil formation and the type and extent of any cultural contribution to the deposits; and the microartifact content was examined to allow further evaluation of the cultural influence on the deposition of each stratum.
On the basis of these analyses, Kilby (1998) developed a general model for how the pit structures filled. Early in their depositional history, the pit structures were deep and steep-sided, making them efficient sediment traps that filled relatively quickly with alluvial and colluvial deposits. Later in their depositional history, as the pit structures became shallower, deposition was slower and dominated by the aeolian transport of sediment.
There is patterned variation to this general model that depends on pit structure location. In pit structures located on the mesa top, aeolian deposition dominated the fill sequence, interrupted only occasionally by episodes of alluvial deposition; in these locations, pit structures filled gradually. More rapid deposition characterized the pit structures located at the base of the cliff and top of the talus slope, and their fill sequences are primarily the result of alluvial and colluvial processes. On canyon benches, pit structure fill is dominated by alluvium, with some colluvium related to structure collapse; soil formation in these cases is weak to absent, indicating that these pit structures filled rapidly.
Culturally influenced deposits could be recognized as massive deposits that lacked laminae, were poorly sorted, and had higher median particle size. One type of culturally influenced deposit, roof fall, was distinguished by its high calcium carbonate content. Microartifact and organic content support the interpretation that beams were robbed from unburned roof fall deposits; there was no evidence that beams had rotted in place. Finally, microartifacts were more than twice as abundant in all strata from the mesa-top structures, even though the chance for their redeposition was lower in this environment. The higher-than-average microartifact content in the fill sequences of pit structures located on the mesa tops supports the interpretation that these mesa-top sites continued to be used after the pit structures were abandoned (Kilby 1998). We have argued that these sites were used as field houses after they had been abandoned as habitations.
To summarize, stratigraphic analysis indicates that structures were abandoned in a number of ways. The way that a specific structure was abandoned appears to have been conditioned in part by the distance of the move to the next site (cf. Lightfoot 1993; Stevenson 1982) and also by how the structure was used. The salvaging of timbers implies that the wood was reused at locations not too far distant, probably within the Sand Canyon locality, which, if true, indicates that the locality continued to be occupied after some structures were abandoned. Although salvaging timbers was most common during the earlier periods, it was practiced throughout the entire time sequence examined in this study, which suggests that the locality was relatively continuously occupied from at least A.D. 1140 to 1280, and possibly between A.D. 1060 and 1280.
Structures whose large timbers were salvaged before burning may represent cases in which it was important to burn the structures as a part of ritual abandonment, but only after the most valuable timbers were removed. This occurs in masonry-lined pit structures, commonly referred to as kivas. Cases in which timbers were left to rot in place are the most rare. It may be that these structures were abandoned during a time when many structures were being abandoned; under such circumstances, there may have been a surplus of wood and therefore no need to salvage and reuse every available beam.
In several recent studies, researchers have concluded that when roofs burned, it was as the result of fires set intentionally as part of ritual abandonments (Cameron 1990; Lightfoot 1993; Lightfoot and Etzkorn 1993; Walker 1995; Wilshusen 1986). Abandonment ritual may account for the burning of structures when timbers are salvaged and when timbers are burned in place, and this burning may indicate that these structures were used differently (cf. Cameron 1990; Walker 1995; Wilshusen 1986). In Lightfoot's (1993) model, entire roofs are intentionally burned when structure abandonment coincides with regional abandonment, that is, when the move to the next site covers a great distance. Following Stevenson (1982), Lightfoot notes that the distance to the next site is an important factor that conditions whether materials will be salvaged from abandoned sites. The Sand Canyon locality kivas with complete, burned roofs follow the model proposed by Lightfoot; all were burned in the period just before the abandonment of the Mesa Verde region. It is also clear that masonry-lined pit structures, or kivas, were the structures that burned most frequently. Burning kivas at abandonment may relate to the ritual and ideology associated with these structures.
Site Abandonment and Occupation of the Locality
The density of artifacts in pit structure and kiva fills was calculated to further examine the question of structure abandonment and to determine how structure abandonment corresponds to the abandonment of sites and localities. The primary goal of this analysis was to determine if sites continued to be used after the pit structures and kivas were abandoned. Twenty-eight pit structures sampled during the Testing Program were examined. The fill in these structures was categorized into three general types of deposits: cultural, natural, and mixed deposits. Cultural deposits include both ashy refuse and construction debris. Natural deposits include any sediments deposited by wind, water, and gravity. Mixed deposits are some combination of natural and cultural deposits or are deposits that have been altered by postdepositional disturbance so that cultural and natural deposition cannot be distinguished. The assignment of fill to one of these depositional categories is an interpretation made during excavation and when profiles are recorded.
Artifact density was evaluated by calculating the mean number of pottery sherds and stone artifacts per cubic meter of fill for each of the three general fill-type categories (Figure 22.5). The mean number of artifacts in culturally deposited fill is 609 per cubic meter, whereas there were 128 artifacts per cubic meter of naturally deposited fill, and 446 per cubic meter of mixed fill.
Ashy refuse with a high artifact content is interpreted as trash generated by a household on an ongoing basis (for example, the cleaning out of hearths). The presence of ashy refuse in the fill of abandoned structures indicates that the site continued to be occupied after the structures were abandoned. Twenty-five percent of the pit structures examined by Cameron (1990) had trash fill. At the Test Program sites, however, ashy refuse was found in only three structures (11 percent) on two sites. Both sites (Lookout House and Castle Rock) have more than one pit structure, and the refuse in the abandoned pit structures was probably deposited while the site continued to be occupied year round by the inhabitants of other structures.
Because trash deposits were so rare at the tested sites, we examined natural deposits and mixed deposits in greater detail. Most artifacts in natural fills are too heavy to have been deposited by wind, so if the artifacts were transported by natural processes, water or gravity must have been the responsible agents. In all cases, pit structures and kivas are upslope from the middens, they are on the highest portions of the site, and they lie on nearly level ground. The main sources for naturally deposited fill are the roomblock located north of the pit structures and the courtyard surrounding the abandoned pit structures. Given these limited catchments, natural deposits were expected to have the lowest artifact densities, and this turned out to be the case. But there is an interesting difference when sites abandoned before A.D. 1260 and those abandoned after 1260 are compared: the earlier structures have an average of 145 artifacts per cubic meter of fill, whereas the later fills averaged only 71 artifacts per cubic meter. The higher artifact densities in the earlier structures are interpreted as tentative evidence for continued use of these sites after the pit structures were abandoned. Most of the earlier sites are on the mesa tops on what are today the most productive agricultural soils. These sites may have continued to be used as field houses after they were abandoned as habitations.
Mixed fills are those that lack evidence of either natural deposition (for example, sorting by particle size and inclusions oriented to bedding planes) or cultural deposition (for example, ash, large inclusions oriented randomly, or construction debris). Mixed fills occur in 12 structures at seven sites. These fills are believed to consist of sediment that accumulated gradually, predominantly through wind deposition, but the relatively high artifact content may also indicate a cultural component to the deposition of these strata. Rather than representing a situation in which household refuse was continuously dumped into these structures, the mixed fills may represent less-frequent artifact deposition. Three of the seven sites where mixed fills are found have more than one pit structure; in these cases, the mixed fills may represent continued use of sites as year-round habitations after one of the pit structures was abandoned. Four of the sites with mixed fills have only one pit structure. In these cases, the mixed fills may indicate continued use of the site on a seasonal or intermittent basis after it was abandoned as a year-round habitation. Again, these structures and sites are located on the fertile mesa tops in upper Sand Canyon, and they may have continued in use as field houses, an interpretation supported by Kilby's (1998) detailed analysis of pit structure fill sequences.
Regional Abandonment
Abandonment of the Mesa Verde region is examined by focusing on the set of kivas abandoned after A.D. 1250. Thirteen kivas are interpreted as having been constructed and used after A.D. 1250. These are among the latest dated structures in the Mesa Verde region. Twenty-one additional sites in southwestern Colorado and eight sites in southeastern Utah have dates from the A.D. 1260s and 1270s (Robinson and Cameron 1991). The latest date from the Mesa Verde region is a single noncutting date of A.D. 1288 for Square Tower House; the next-latest are cutting dates of A.D. 1280 and A.D. 1281 for Square Tower House and an A.D. 1279 cutting date for Long House on Wetherill Mesa (Nichols and Harlan 1967:17). The two latest dates outside Mesa Verde National Park are a cutting date of A.D. 1277 at Hovenweep Castle and a noncutting date of A.D. 1277 from Sand Canyon Pueblo. These dates indicate that the Sand Canyon locality was occupied as late as any part of the Mesa Verde region.
Hantman (1983) has argued that wood use was relatively continuous at occupied sites, and therefore site abandonment can be inferred to have occurred at about the same time as the latest tree-ring date, but this is true only if the sample of tree-ring-dated specimens is representative of the entire population of timber used by the inhabitants of the site (Ahlstrom 1985a:79; Lightfoot 1992:226). The principles suggested by Hantman, Ahlstrom, and Lightfoot can be used for dating the abandonment of the region.
It would be difficult to determine if the excavated sites in the Mesa Verde region are representative of the entire population of sites in the region, much less to determine if the tree-ring samples are representative of all timber used by the inhabitants of the region. Nevertheless, the dated material from the Mesa Verde region includes samples from many sites and almost 10,000 tree-ring dates. This large sample exhibits a nearly continuous distribution of dates throughout the A.D. 1260s and 1270s. Dates are especially numerous through the early A.D. 1270s. They fall off gradually in number in the mid-1270s, and by the late 1270s, few dates are recorded (Schlanger et al. 1993; Varien 1997). Applying Hantman's principle, the abandonment of the Mesa Verde region took place at approximately A.D. 1290.
To evaluate this interpretation, the Sand Canyon locality kivas that date to between A.D. 1250 and 1290 are examined in greater detail. As noted above, the construction and abandonment of 13 kivas are tree-ring dated to this period. Seven additional structures are assigned to this period on the basis of pottery, stratigraphy, and in some cases, archaeomagnetic dates. Figure 22.6 shows the roof treatment at abandonment of these 20 structures.
What is interesting is that roof timbers were salvaged from nine of the 20 structures abandoned during this period. Indeed, some structures interpreted as having been constructed in the A.D. 1270s had their roof timbers salvaged at the time they were abandoned. In earlier periods, there is ample evidence for reuse of salvaged beams in the construction of new buildings, but the salvaged wood typically was used along with freshly cut timbers. It may be that the late 1270s was a time of declining population and that this resulted in a surplus of abandoned structures. Salvaged timber may have been so plentiful that new roofs were constructed entirely with reused beams (it is also possible that some timbers salvaged from structures abandoned in the A.D. 1270s were used for firewood rather than for construction). If roofs were indeed constructed completely with salvaged beams, it may be that construction in the Mesa Verde region extended beyond A.D. 1280, as suggested by the latest tree-ring dates. But it is unlikely that occupation of the Mesa Verde region could extend beyond A.D. 1290 without some new timbers being harvested. Thus, any continued occupation after A.D. 1280 must have been one in which regional population was rapidly declining. The near absence of new construction in the late A.D. 1270s and 1280s suggests that migration from the region was a process, but a rapid process that accelerated during the decade of the A.D. 1280s, with the region being largely abandoned for residential settlement by A.D. 1290.
Further Research
The Site Testing Program analyses presented in this volume link three analytic scales--structures, sites, and localities--and flesh out the history of each as a component in an active settlement system. This research provided the framework for additional studies of settlement patterns and for the development of a new model of population movement within the Mesa Verde region, which is articulated in the companion volume to this report, to be published by the University of Arizona Press (Varien 1998). In this study, the movement of both households and communities is evaluated in terms of changes in the social landscape of the Mesa Verde region between A.D. 950 and 1300. Population movement is interpreted as a social process whereby households negotiated their residential moves and their access to natural and social resources in a social landscape structured by the historical development of the many communities found throughout the region. The publication of this in-depth study will provide an important supplement to the analyses presented in this report.