Measuring Community Interaction: Pueblo III Pottery Production and Distribution in the Central Mesa Verde Region


Christopher Pierce, Donna M. Glowacki and Margaret M. Thurs


The scale, intensity, and character of interaction among Pueblo people during the 13th century A.D. likely played an important role in the processes and events leading to the abandonment of the Northern San Juan region in the 1280s. Characteristics of pottery production and distribution in the Sand Canyon locality provide one means of investigating these interactions. Variation among Pueblo III settlements in the use of temper and available raw clay sources, and the distribution of pottery production tools demonstrate the existence of at least two production areas within the locality. The nature of the boundary between these production areas indicates a complex pattern of settlement and community interaction that challenges models based on settlement proximity. Further, an almost complete lack of extra-regional pottery at Pueblo III settlements suggests that the Northern San Juan region may have been economically isolated from other regions inhabited by Pueblo people.


An edited version of this paper was publiahsed as:


Pierce, C., D. M. Glowacki, and M. M. Thurs
2002 Measuring Community Interaction: Pueblo III Pottery Production and Distribution in the Central Mesa Verde Region. In Seeking the Center Place: Archaeology and Community in the Mesa Verde Region, edited by M.D. Varien and R. Wilshusen, pp. 185-202. The University of Utah, Salt Lake City.


In the late 13th century A.D., Pueblo people left the Mesa Verde region as well as other parts of the northern Southwest and moved south to the areas now occupied by the modern Pueblo people. This migration by a substantial population from a large area has attracted considerable attention from archaeologists and the public since it was first documented in the latter half of the nineteenth century. Although a variety of explanations have been proposed for this large-scale migration, most have focused primarily on the debilitating effects of the severe drought that engulfed the region from A.D. 1276 to 1299 (Douglas 1929; Petersen 1988). However, recent research on potential agricultural productivity in the Mesa Verde region indicate that even during periods of severe drought the region remained sufficiently productive to support a viable population (Van West and Lipe 1992). In addition, the environmental challenges of the late thirteenth century do not appear to have been any more severe than those that were faced during the late ninth and twelfth centuries. Although ancestral Pueblo populations in the Mesa Verde region responded to both of these earlier climatic downturns by migrating south, small populations remained in the area during both episodes, and substantial populations returned shortly after conditions began to improve. Consequently, we can refine our questions regarding the late-thirteenth-century depopulation and migration in two ways. First, why did the entire Pueblo population leave the Four Corners area in the thirteenth century, when complete emigration does not appear to have been the case during earlier droughts? Second, why did Pueblo populations not return to the Mesa Verde region after the thirteenth century drought ended, as they had done after previous droughts?


Answers to both these questions may involve changes in the ways Pueblo people interacted. If Pueblo III communities in the northern San Juan region became increasingly autonomous, or even competitively hostile, this would have resulted in limited mobility outside community territories (Kuckelman et al. 2000; Lightfoot and Kuckelman 1994; Neily 1983; Varien 1997a). A lack of cooperative interaction and short distance movement and a rise in competition and conflict between communities would have significantly limited the ability of populations to compensate for local shortfalls in agricultural productivity. As a consequence, the degree of subsistence stree experienced by members of individual communities during periods of drought may have increased, possibly forcing a simultaneous long-distance migration. On the other hand, if Pueblo III communities had become increasingly integrated and interdependent, it may have been extremely difficult for small populations to remain behind while others migrated in the face of drought-induced agricultural disruption. Consequently, determining if small settlements and larger centers of central Mesa Verde region communities were autonomous and competative, or integrated into larger interdependent  local and regional systems, has important implicaitons for explaining the thirteenth-century depopulaiton of the northern San Juan region.


Traditionally, studies of the scale and intensity of interaction within the Mesa Verde region have focused on the spatial arrangement of settlements. Clusters of contemporary settlements are used to identify communities particularly if some settlements within these clusters contain potentially integrative public or civic architecture such as great kivas (Adler 1990; Adler and Varien 1991; Lipe 1992; Rohn 1977). Communities identified in this way are then assumed to represent areas of intense social and economic interaction with less interaction having occurred between such communities.


Application of this approach to the Sand Canyon locality (Figure 1), a 200 square kilometer area northwest of Mesa Verde, has resulted in the identification of three communities—upper Sand Canyon, lower Sand Canyon, and Goodman Point (Adler and Varien 1994; Lipe 1992). During the late Pueblo III period (A.D. 1225-1280), each of these communities consisted of clusters of settlements focused around a relatively large pueblo with public architecture. In the upper Sand Canyon community, the principal settlement is Sand Canyon Pueblo, a site of more than 500 rooms, including a multiwalled, multistoried D-shaped structure, a great kiva, and an enclosed plaza (Bradley 1992). Castle Rock Pueblo, with more than 60 rooms, a D-shaped structure and an enclosed plaza, is the largest settlement in the lower Sand Canyon community (Lightfoot and Kuckelman 1992;1994). The central site in the Goodman Point community is Goodman Point Pueblo, which consists of over 400 rooms, including kivas, a bi-wall structure, and a great kiva, (Adler 1990). Fifteen large and small Pueblo III settlements in the upper and lower Sand Canyon communities have been the subjects of intense investigation by Crow Canyon Archaeological Center since 1984.


Figure 1
Figure 1. Map of the locations of San Canyon locality sites from which materials were analyzed for this study.


The movement of pottery between contemporary settlements offers another means of documenting economic and social interaction in addition to settlement proximity. We have begun to document evidence of pottery production and movement at late Pueblo III settlements in the upper and lower Sand Canyon communities and on Mesa Verde that were occupied just prior to the 13th century regional abandonment. This evidence consists of direct indicators of pottery production (including the tools, raw materials and by-products of pottery manufacture), and the differential use of raw materials identified through microscopic analysis of temper and the chemical characterization of pottery and raw clay using neutron activation analysis. Analyses completed to date indicate the presence of at least two, and possibly three pottery production zones within the Sand Canyon locality. Two of these production zones correspond with the upper and lower Sand Canyon communities, but the possible third zone, at this point marked primarily by differential temper use, cross-cuts these communities and may indicate interaction among settlements not predicted by the community models based on settlement proximity.


Methods and Materials


Measuring the movement or distribution of pottery between settlements involves delineating areas in which relatively homogeneous pottery was produced, and then identifying pottery from these production areas in settlements outside the production area boundary. Commonly, attributes of pottery such as design styles, technological features, and raw material composition have been used to distinguish production areas and establish the criteria for identifying pottery from these areas. In this study, we employ two kinds of evidence for pottery production and distribution in the Sand Canyon locality of the central Mesa Verde region: (1) direct indicators of pottery production including the tools, raw materials and by-products of pottery manufacture, and (2) compositional variation identified through microscopic analysis of temper and chemical characterization of pottery and raw clay using instrumental neutron activation analysis. We chose not to examine stylistic and technological aspects of pottery variation because of the small spatial scale over which we are trying to measure the movement of pottery.

Our study focuses on data from 15 Pueblo III settlements in the upper and lower Sand Canyon communities (Figure 1) intensively investigated by Crow Canyon Archaeological Center between 1984 and 1994 (Lipe 1992a; Varien 1997, 1999b). In addition, compositional analyses were performed on a small number of sherds from two Pueblo III settlements on Mesa Verde, Mug House (Rohn 1971) and Long House (Cattanach 1980). The Sand Canyon and Mesa Verde settings offer excellent opportunities for compositional studies because of the geological variation within and between these areas. Raw clay sources available include outcrops of ancient marine and alluvial formations and recent alluvial deposits. Potential temper materials include distinctive sandstone and intrusive igneous formations (Ekren and Houser 1965; Griffitts 1990).


Direct Evidence of Pottery Manufacture


If inhabitants of a settlement were manufacturing pottery, the various activities involved in making pottery should have left material traces in the form of tools, facilities, raw materials, and other by-products. Identifying these traces in the archaeological record can provide direct evidence of not only the presence of pottery production at that location, but also other aspects of the production process such as its organizational context, spatial concentration, scale, and intensity (Costin 1991; Mills and Crown 1995). Pottery manufacture tools, unfired pottery, and manipulated raw clay and temper materials constitute the range of direct evidence for pottery production commonly found at sites in the central Mesa Verde region and elsewhere in the northern Southwest. Manufacturing tools include base molds or pukis, scrapers used to shape and smooth vessel surfaces, and polishing stones used to polish the surfaces of white ware pottery (Blinman and Wilson 1993; Hill 1985, 1994; Waterworth and Blinman 1986). Raw materials include lumps of raw clay, chunks of temper materials, and examples of clay and temper mixed together. Pottery kilns and their associated production debris have also been found in the region, but they are usually located a considerable distance from the nearest settlement, and are difficult to link to production at individual sites or even communities (Blinman and Swink 1997; Bernadini 2000; Brisbin 1996; Fuller 1984; Purcell 1993).


We documented direct evidence for pottery production at 14 of the sites excavated in the Sand Canyon locality. One additional site, Green Lizard Hamlet, excavated in the San Canyon locality, was not included in the study of direct production evidence because the artifact collection from the site was not available at the time this study was performed. During excavation, workers in the field carefully collected all possible examples of pottery raw materials including clay lumps that were different than the mortar used in wall construction, chunks of broken igneous rock of the type used as temper in some vessels, and unfired pottery vessels and fragments. In the lab, these possible raw materials were further examined for evidence of manipulation, and only those specimens that were clearly manipulated by shaping, crushing, or mixing temper and clay, or occurred in large quantities on use surfaces were considered evidence of pottery manufacture. We examined the pieces of unfired pottery to determine the type of pottery and raw materials used. Possible tools used in pottery manufacture were also analyzed. All modified sherds, pottery fragments with one or more edge showing wear from use, were examined for traces of the drumlin-like wear pattern (Waterworth and Blinman 1986) produced by scraping wet clay. We examined all rounded, smooth pebbles and stones for striations and adhering clay, which result from their use in polishing pottery. Only smooth stones possessing signs of abrasive wear or adhering clay were classified as polishing stones. Finally, all possible base molds, shallow bowls, and vessel base fragments were examined for indications that the shallow form was produced or enhanced deliberately.


Compositional Analyses


Analyses of the composition of pottery provide indirect evidence for both the production and exchange of pottery, thereby allowing an assessment of social interaction. By linking composition patterns to the locations of available raw materials we can identify likely pottery production areas. Exchange is evaluated by identifying pottery produced in one area and recovered from another. We generated compositional data by examining and classifying the temper and measuring the bulk elemental composition (clay and temper) using instrumental neutron activation analysis. For both analyses, we selected a sample of rim fragments from Mesa Verde Black-on-white bowls and Mesa Verde Corrugated jars, the two most common stylistic types and vessel forms made during the thirteenth century in the Mesa Verde region (Breternitz et al. 1974). Individual specimens were selected randomly from the population of rim sherds recovered from each site. Once the sample of 30 was drawn, the rim sherds were compared to one-another to determine if any came from the same vessel. If so, only one sherd was retained and an additional specimen selected until all sherds in the sample represented different vessels. Consequently, the materials analyzed from each assemblage constitute a sample of whole vessels, not sherds. Although the total population of whole vessels represented by the individual rim sherds in each assemblage is not currently known, the proportion of vessels analyzed by our samples is considerably higher than if sherds were the population being sampled. This means that even though the absolute number of specimens characterized in each analysis is small, the proportion of vessels characterized is fairly high for most assemblages, and in some cases included all bowls represented in the excavated sample. The specific sample sizes and procedures for each analysis are presented below. 


Temper Analysis

Temper was identified in Mesa Verde Black-on-white bowls from all 15 of the Sand Canyon locality sites excavated by Crow Canyon Archaeological Center. Although we examined bowls rim sherds for the temper analysis, they are representative of unique vessels because the sherds were selected so that they displayed unique combinations of morphological and design features, which would not be present in a single vessel. We analyzed sherds from 30 bowls at all but two sites, Mad Dog Tower (N=13) and Troy's Tower (N=20), for which we analyzed all available and bowl rims from different vessels. We also identified temper in 30 Mesa Verde Corrugated jars from three of the upper Sand Canyon sites (Sand Canyon Pueblo, Stanton's site, and Green Lizard Hamlet) and two lower Sand Canyon sites (Castle Rock Pueblo and Saddlehorn Hamlet).

We analyzed each specimen by first breaking off a small corner of the sherd with pliers to expose an unweathered edge. This fresh edge was examined using a binocular microscope with magnification up to 60X. We recorded the relative abundance and type of all common sand-sized grains in each specimen. These data were then used to construct three temper classes: crushed sherd, crushed sandstone, and crushed igneous rock. We identified each specimen to a single temper class based upon its dominant temper ingredient.


Instrumental Neutron Activation Analysis

We performed this bulk elemental characterization on 30 Mesa Verde Black-on-white bowls each from Castle Rock Pueblo in lower Sand Canyon, Sand Canyon Pueblo in upper Sand Canyon, and Long House and Mug House on Wetherill Mesa in Mesa Verde National Park (Glowacki 1995; Glowacki et al. 1995, 1998). We also analyzed 30 Mesa Verde Corrugated jars each from the two sites on Mesa Verde. In addition, raw clay samples from 19 sources in Sand Canyon and the McElmo Drainage, and 35 sources in Mesa Verde National Park were characterized for this study.


Instrumental neutron activation analysis (INAA) was undertaken at the Missouri University Research Reactor (MURR) using their established standard-comparator procedure (Glascock 1992). Individual pottery samples were prepared by removing the exterior surfaces using a tungsten-carbide drill to avoid possible contamination from paint, slip, and other material adhering to the vessel surface. The clay samples were fired to 700?C. Both pottery and clay samples were crushed into a fine powder and weighed into two 200mg samples.


INAA of pottery and raw clay samples at MURR consists of two irradiations and three counts on 200mg samples prepared in polyvials and high-purity quartz vials (Glascock 1992). A pneumatic tube system transports samples in polyvials to the reactor core where it is exposed to a neutron flux of 8 x 1013 n/cm2/s for five-second irradiations. After irradiation, each sample decays for 25 minutes followed by a 12-minute count with an HPGe detector to measure concentrations of the short-lived elements Al, Ba, Ca, Dy, K, Mn, Na, Ti, and V in each sample. A long irradiation of 24 hours is carried out on the samples sealed in quartz vials using a neutron flux of 5 x 1013 n/cm2/s to measure the concentrations of elements with longer half-lives. A middle count for 2,000 seconds with an HPGe detector coupled to an automatic sample changer is performed after samples decay for seven days to determine concentrations of As, La, Lu, Nd, Sm, U, and Yb. After an additional three to four week decay, a final count of 10,000 seconds measures the long-lived elements (Ce, Co, Cr, Cs, Eu, Fe, Hf, Ni, Rb, Sb, Sc, Sr, Ta, Tb, Th, Zn, and Zr).


Resulting data are analyzed using multivariate statistical procedures through a series of programs written in GAUSS language (Neff 1990). The analysis objectives are to identify pottery compositional groups and to link pottery of unknown provenance to known clay sources and previously established groups, or to determine that a group assignment is not possible (Glascock 1992). First, element concentration data are transformed to log base 10 values, scaling the data to account for large magnitude differences between major trace elements. Principal components analysis (PCA) provides a view of subgroup structure. Using eigenvector extraction, PCA finds the orientations and lengths of axes of greatest variance in the data and organizes these axes by decreasing variance (Baxter 1993; Davis 1986). Using PCA as a simultaneous RQ mode allows the plotting of both elements and objects to evaluate the relationship of elements or groups of elements contributing to group separation (Baxter 1992; Neff 1994). Simultaneous RQ mode PCA of the variance-covariance matrix was used to analyze these data. To evaluate the coherence of each group, Mahalanobis distances were used to calculate multivariate probabilities of group membership (Bieber et al.1976; Bishop and Neff 1989; Harbottle 1976; Sayre 1975).




Production Evidence


Table 1 shows the presence or absence of four different kinds of direct evidence of pottery production at 14 of the sites in the Sand Canyon locality. We recognized base molds only when a complete or nearly complete example was recovered. In most cases, base molds consist of broken bases of corrugated and white ware jars that had been modified by grinding and shaping the broken edge. Similar modification appears on numerous fragments identified as modified sherds, and consequently, the modified sherd category probably contains fragments of base molds not identified as such (Pierce et al. 1999). Modified sherds were examined for the presence of "pot scraper" wear, but no examples of this distinctive wear pattern were found. Although sherd scrapers used to work wet clay are fairly common in Basketmaker III through Pueblo II assemblages, some other tool or technique appears to have been used to shape pottery vessels during Pueblo III (see also Wilson 1991). All of the examples of unfired pottery examined for this study that could be identified to pottery ware are slipped and sometimes painted white wares. No examples of corrugated or plain gray ware unfired pottery are present in the Sand Canyon collections. This may be a bias produced by the very small sample of unfired pottery. However, it may also indicate that a selected subset of vessels, white wares, were used in an unfired state leading to their differential preservation (Bradley 1992b:38-39). If the occurrence of unfired pottery is related to the use of these vessels rather than the chance preservation of pottery before it was fired, then the recovery of unfired pottery may not be a reliable direct indicator of pottery production. Angular igneous rock chunks are almost certainly the by-products of temper production because very few of the stone tools at these sites were manufactured from this kind of raw material (Pierce et al. 1999). In addition, the material closely matches the crushed igneous rock used as pottery temper (Hegmon 1995).


Table 1. Presence (+) or absence (-) of different kinds of direct evidence of pottery production recovered from 14 sites in the Sand Canyon locality
 Pottery Production Evidence
 AssemblagesBase Molds
Polishing Stones
Unfired Pottery Raw Clay Igneous Rock Chunks
 Upper Canyon Sites
Lillian’s Site++-++
G and G Hamlet+--++
Roy’s Ruin---++
Kenzie Dawn Hamlet-++++
Sand Canyon Pueblo+++++
Shorlene’s Site-+-++
Troy’s Tower-+-+-
Lester’s Site-+-++
Lookout House-++++
Stanton’s Site++-++
Catherine’s Site---++
Lower Canyon Sites 
Saddlehorn Hamlet----+
Mad Dog Tower----+
Castle Rock Pueblo+++++


Although the 14 sites show considerable variation in the richness of pottery production evidence, a significant positive correlation (r2 = 0.60, p = 0.001) exists between the log10 value of the size of the sample from each site (measured by the total number of pieces of pottery recovered) and production evidence richness (Figure 2). This correlation suggests that variation in production evidence may be due to differences in sample size among the sites rather than an indication of the differential presence or intensity of pottery manufacture. However, sample size also correlates with the size and complexity of the site; this could also cause the strong richness-sample size relationship if people living in the larger settlements were more heavily involved in pottery manufacture. It is impossible to argue convincingly that little or no pottery production took place at sites with small samples and meager evidence of pottery production, such as Saddlehorn Hamlet and Mad Dog Tower without increasing the sample size from these sites. However, the overall pattern of direct production evidence indicates that most, if not all Pueblo III settlements in the Sand Canyon locality were involved in pottery manufacture to some degree suggesting a fairly dispersed, small-scale organization of production.


Figure 2
Figure 2. Scatter plot showing the correlation between the richness of direct evidence of pottery production and a measure of the recovered sample size from 14 sites in the Sand Canyon locality. 1=LillianÕs Site, 2=G and G Hamlet, 3=RoyÕs Ruin, 4=Kenzie Dawn Hamlet, 5=Sand Canyon Pueblo, 6=ShorleneÕs Site, 7=TroyÕs Tower, 8=LesterÕs Site, 9=Lookout House, 10=StantonÕs Site, 11=CatherineÕs Site, 12=Saddlehorn Hamlet, 13=Mad Dog Tower, 14=Castle Rock Pueblo.


Pottery Temper


Figure 3 shows the relative frequencies of the three main temper classes (crushed igneous rock, crushed sandstone, and crushed pottery sherds) used in the production of Mesa Verde Black-on-white bowls in the 15 Sand Canyon locality assemblages arranged vertically from north to south. Results from five assemblages for which we identified the temper of corrugated jars indicate that crushed igneous rock was used almost exclusively throughout the locality. Of the three temper categories, igneous rock is the only one with a restricted source within the locality. Sleeping Ute Mountain and the alluvial terraces of McElmo Creek at the mouth of Sand Canyon constitute the nearest sources of igneous rock to the Sand Canyon locality sites. Only one of the sites, Castle Rock Pueblo, lies within the zone of natural igneous rock availability. We were unable to distinguish easily different kinds of crushed sandstone temper even though very distinctive sandstone deposits exist within the locality. A more detailed petrographic analysis would probably yield better results. Despite these limitations, sandstone for temper was likely readily available to the inhabitants of all 15 sites. Clearly, the inhabitants of all sites also had easy and almost unlimited access to potsherds for grinding into temper. Consequently, our discussion of the temper data focuses on the differential use of igneous rock in Mesa Verde Black-on-white bowls throughout the locality.


Figure 3
Figure 3. Bar graph of the relative frequencies of the three main temper classes in Mesa Verde Black-on-white bowls from 15 sites in the Sand Canyon locality arranged vertically from north (top) to south (bottom).

Among the assemblages from upper Sand Canyon, three different patterns of igneous rock use can be seen (Figure 4). Igneous rock temper occurs in small amounts (less than 20%) in all of the small settlements located on the mesa top north of Sand Canyon Pueblo. These sites were occupied early in the thirteenth century before Sand Canyon Pueblo was constructed (Varien 1999b). Sand Canyon Pueblo and the two contemporary small sites located near it (Lester's Site and Lookout House) show almost no use of igneous rock temper in white ware bowls although almost all corrugated jars analyzed were tempered with igneous rock. Finally, igneous rock temper occurs in relatively high proportions, close to half of all bowls analyzed, among the three settlements located within Sand Canyon, a short distance south of Sand Canyon Pueblo (Green Lizard Hamlet, Stanton's Site, and Catherine's Site). These three settlements were also occupied at the same time as Sand Canyon Pueblo. These sites show the greatest use of igneous rock temper within the entire Sand Canyon locality even though the nearest igneous source is 6.5 km to the south and 300m lower in elevation. The possibility that the igneous rock in these white ware bowls came from ground up corrugated jars is unlikely because the igneous-tempered bowls rarely contained any crushed sherd.


Figure 4
Figure 4. The percentage of Mesa Verde Black-on-white bowls with predominantly crushed igneous rock temper in 15 Sand Canyon locality assemblages plotted against UTMG north coordinate for each site.

All of the assemblages from the lower Sand Canyon community have Mesa Verde Black-on-white bowls with igneous rock temper, but at Saddlehorn Hamlet, igneous rock temper is abundant (43%). Despite their close proximity to the igneous rock source, Castle Rock Pueblo and Mad Dog Tower yielded white ware bowls with only 20 and 15 percent igneous tempered-bowls. However, all corrugated jars analyzed from Castle Rock and Saddlehorn contained igneous rock temper. All three of these lower Sand Canyon sites appear to have been occupied during the latter half of the thirteenth century, and were contemporary with Sand Canyon Pueblo (Varien 1999b).


Instrumental Neutron Activation Analysis


Most of the 120 Mesa Verde Black-on-white bowls and 60 Mesa Verde Corrugated jars in the INAA data set separated easily into two distinct reference groups, MV1 (N = 80) and MV2 (N = 65) (Figure 5). Thirty-five sherds were unassigned to a compositional group by the principal components analysis. Temper differences between the two distinct INAA groups are not contributing to group separation on this level because the analyzed sherds in both reference groups contain heterogeneous tempers. Diagenesis is also not a contributing factor because non-mobile trace elements are included in the characterization of each reference group. Thus, MV1 and MV2 represent the use of distinct raw clay sources within the region. For a more complete discussion of the INAA results than can be presented here, see Glowacki, Neff, and Glascock (1995, 1998).


Figure 5
Figure 5. Principle components plot showing the separation of the four INAA pottery compositional reference groups MV1 and MV2.


Each of these reference groups can be broken down further into two subgroups using PCA, and these subgroups represent separate compositional groups that reflect either a shared source or manufacturing technique (Figure 6). Based on the criterion of abundance in which the assemblage contributing most of the samples to a group is used to identify the group and its possible production location, we have made the following determinations. The bowls in Subgroup A (N=36) appear to derive from the Sand Canyon Pueblo area, which contributed 58 percent of the vessels to this group. Subgroup B (N=43) consists mainly (95%) of bowls from the Mesa Verde sites. Subgroup C (N= 18) is assigned a Castle Rock/McElmo Canyon provenance with 72 percent of the bowls in this group coming from Castle Rock Pueblo. Subgroup D (N=43) is comprised solely of the Mesa Verde Corrugated jars from the two Mesa Verde sites. 


Figure 6
Figure 6. Principle components plot showing the separation of the four INAA pottery compositional subgroups, and their relationships to potential raw clay sources.


Figure 6 shows the compositional similarity in multivariate space between the pottery composition groups and potential raw clay sources. Compositional associations between raw clays and pottery compositional groups were evaluated using Mahalanobis distances based on 96.4% variance in the data set. The Mesa Verde Black-on-white bowls in the Sand Canyon Pueblo Subgroup A closely match the composition of Dakota Formation clays exposed at the site. The Subgroup B bowls match well with clay from a Menefee Formation exposure below Mug House. The bowls inferred to have been produced at Castle Rock Pueblo (Subgroup C) were most likely made from McElmo Creek alluvial clays. The corrugated jars in Subgroup D may have been made with clay from the Mancos Formation, but this association is not yet well established. However, the raw clay associations made with the three Mesa Verde Black-on-white subgroups support the production area assignments made using the abundance criterion.


Table 2 shows the number of Mesa Verde Black-on-white bowls from each of the four pottery assemblages that were assigned to different compositional groups and those unassigned. These data indicate that a significant proportion of bowls found in one location were produced elsewhere, particularly if we view the unassigned vessels as representatives of as yet poorly characterized production areas rather than statistical outliers of the current compositional groups. Regardless of how one handles the unassigned vessels, Castle Rock Pueblo had the greatest reliance on nonlocally produced bowls at around 50 percent of vessels followed by Mug House (~30-40%), Sand Canyon Pueblo (~15-30%), and Long House (~10-20%).


Table 2. Summary of INAA data on the movement of Mesa Verde Black-on-white bowls between possible areas of pottery production identified by compositional subgroups.
  Compositional Subgroups
Pottery Assemblage A

Percent Nonlocal 

 Sand Canyon Pueblo 211
 Castle Rock Pueblo
 Mug House
 Long House
 * Computed from subgroups only/computed from subgroups and unassigned.




This initial foray into documenting the patterns of production and distribution of pottery among late twelfth and thirteenth century settlements in the Mesa Verde region as a measure of community interaction has yielded intriguing, though still preliminary, results. Both the direct production and indirect compositional evidence suggest that most, if not all, settlements manufactured pottery. The lack of direct evidence of white ware manufacture at the two small sites in the lower Sand Canyon community may indicate some variation among settlements and communities in the organization and intensity of pottery production, but may also result from the relatively small samples recovered from these sites. Compositional data indicate the potters tended to rely heavily on locally available temper and clay raw materials, and many of these different materials can be distinguished through chemical and temper analyses. Both of these qualities bode well for being able to identify pottery production areas unambiguously.


However, there may be exceptions to this rule of local raw material use as exemplified by the relatively high proportions of igneous rock-tempered bowls at some sites in the upper Sand Canyon community where igneous rock was not locally available. The presence of igneous rock chunks and corrugated jars with igneous rock temper at all of the sites in the upper Sand Canyon community indicates that the inhabitants of all these settlements probably had easy access to the igneous raw material sources. The use of relatively coarse igneous rock temper in corrugated cooking pots probably improved the durability of these vessels (West 1992), but it is unclear that the finely crushed igneous rock present in the white ware bowls would have had a beneficial effect on the cost or performance of these vessels. Consequently, it is enigmatic that some upper Sand Canyon site hade relatively high levels of igneous rock-tempered bowls (~50%) while other sites in the same community had very low levels (<20%).


We can imagine two possible explanations for this variable pattern of igneous rock temper in the upper Sand Canyon community. First, the variation could have resulted from different choices made by potters based on very local traditions of raw material acquisition and use. Much of the white ware pottery made earlier during the Pueblo I and Pueblo II periods over most of the Mesa Verde region contain crushed igneous rock as temper (Blinman and Wilson 1988, 1992; Breternitz et al. 1974; Hegmon 1993, 1995). The use of igneous rock temper in white ware vessels began to decline in the twelfth and thirteenth centuries as crushed sherds became a much more common tempering additive. However, it is possible that this change was more gradual and less comprehensive than is commonly portrayed in Mesa Verde region pottery type descriptions. Under this scenario, the moderate to low percentages of igneous-tempered bowls in the earlier Pueblo III sites on the mesa top may represent the tail end of a gradual trend away from the use of igneous rock for white ware temper, while the moderate to high percentages of igneous temper in bowls from the later pueblo III sites located within Sand Canyon may indicate the local persistence of the igneous rock temper tradition resulting in a distinct production area.


The second possible explanation is that the igneous rock-tempered white ware bowls were not made at the upper Sand Canyon sites, but were acquired from other production areas where igneous rock is immediately available such as the lower Sand Canyon area. If the bowls with igneous temper were imported, then the patterns of interaction implied by this movement of pottery differ substantially from what we would expect based on the proximity of settlements. The very low frequencies of igneous-tempered bowls at Sand Canyon Pueblo and the two contemporary sites located immediately adjacent to it may indicate that these settlements were more isolated than those sites in the same community, but located a little farther south within Sand Canyon whose inhabitants may have interacted heavily with people in the lower Sand Canyon community.


Since our study was performed, Ortman (2000) conducted temper analyses of white ware bowl sherds, though not restricted to Mesa Verde Black-on-white as in the current study, from several sites in the central Mesa Verde region including six of the sites from the Sand Canyon locality used in our study. Ortman found that over a larger area than the Sand Canyon locality, the distribution of white ware bowls tempered with crushed igneous rock forms a linear log/square relationship with the distance to the Ute Mountain source of igneous rock. Ortman points out that this distance decay pattern is likely to have been produced by down-the-line exchange of pottery, which can be modeled by random-walk processes. This suggests another possible explanation of the variation in the relative abundance of bowls with igneous rock temper at sites in the Sand Canyon locality. The variation may be noise created by random fluctuations in the patterns of pottery manufacture, movement, and discard over short distances. We must keep in mind that random processes can produce statistically significant variation. This is particularly true when the processes occur rarely during the time interval of the measurement, which in this case is the duration of occupation of a settlement. We do not yet know if pottery exchange was a rare event over the duration of occupation of settlements dating to the thirteenth century in the Sand Canyon locality. If it was, then random fluctuations may play a large role in producing the variation in temper materials in the Sand Canyon locality.


The INAA data do not yet allow us to distinguish between these possible explanations of temper variation, but they do give us another glimpse at possible patterns of vessel movement. The data on bowls analyzed from Sand Canyon Pueblo shows a relatively low proportion of imported vessels with some coming from the Castle Rock production area (Compositional Subgroup C). The data from Castle Rock Pueblo indicate a much higher level of vessel movement into the settlement with most coming from the Sand Canyon production area (Compositional Group A). Very few vessels recovered from Sand Canyon or Castle Rock Pueblos appear to have originated on Mesa Verde (Compositional Subgroup B). However, both of the sites on Mesa Verde contain bowls that may have been produced in the Sand Canyon locality (see Table 2). Mug House contains a relatively high proportion of bowls from the Sand Canyon Pueblo production area, while Long House has a small percentage from the Castle Rock production area. Based on these data, it appears the Sand Canyon Pueblo may have been more involved in exporting than importing pottery. If the presence of bowls with igneous rock temper in settlements of the upper Sand Canyon community is indicative of the movement of vessels into the community, then the greater emphasis on export by Sand Canyon Pueblo may account for its very low proportion of igneous rock-tempered bowls. The high proportion of igneous-tempered bowls in the sites from the southern part of the upper Sand Canyon community may also indicate a greater reliance on imported pottery, particularly from the Castle Rock production area. However, the relatively close proximity of all of these sites and our small sample of INAA data make it impossible at this time to preclude the differential use of shared resource areas rather than the distribution of completed vessels.

In an attempt to distinguish among some of these different possibilities, we have initiated another round of 300 INAA assays including: Mesa Verde Black-on-white bowls and corrugated jars from Stanton's Site and Saddlehorn Hamlet in the Sand Canyon locality, and Yellow Jacket Pueblo located outside the locality to the north; Mesa Verde corrugated jars from Sand Canyon and Castle Rock pueblos; and numerous additional potential clay and temper raw material sources. By analyzing bowls and jars from Stanton's Site and Saddlehorn Hamlet and corrugated jars from Sand Canyon Pueblo, we hope to determine if the igneous-tempered vessels also contain McElmo Creek alluvial clays. If so, then the movement of completed vessel rather than raw materials would be much more likely. The inclusion of pottery from Yellow Jacket Pueblo, which is located in a very similar geological setting to Sand Canyon Pueblo, should allow us to determine if chemical variation within the Dakota Formation clays is sufficient to enable the identification of different production areas within the geologically homogeneous central Mesa Verde region.



We had two questions when we began this research, one methodological and one substantive. The methodological question focused on whether documenting direct evidence of pottery manufacture and conducting compositional analyses in the central Mesa Verde region would yield a sufficiently fine-grained view of pottery production and distribution to use these data to investigate some aspects of the interaction among ancient people living in different settlements and communities. Some scholars working in the region doubted that the available clays would have sufficiently robust chemical signatures to allow the differentiation of sources and identification of production areas at the small scale needed for tracking vessel movement over the short distances between adjacent communities. Our results should begin to dispel this notion. Although the work presented here is just a first step, it clearly demonstrates that continued studies of the evidence of pottery production and pottery composition will provide a richer, more complete understanding of the social, political, and economic interactions that existed during the thirteenth century in the Mesa Verde region.


Our substantive question involved understanding the nature of interaction among the people inhabiting settlements and communities during the thirteenth century in the Mesa Verde region. Settlement analysis has been used in the Mesa Verde region to identify communities many of which may have persisted through the last two to three hundred years of Pueblo occupation and substantial changes in community organization (Adler and Varien 1994, Varien 1999a). In this approach, the spatially clustered settlements used to identify communities are also thought to indicate greater levels of interaction among the people inhabiting these sites than with those in other communities. However, a more thorough understanding of the level and kind of interaction among communities in the Mesa Verde region could contribute important insights into the processes and mechanisms that resulted in the complete depopulation of the region in the thirteenth century. We have begun to build this more complete picture by examining evidence for the production and movement of pottery between settlements of three different communities; two adjacent communities at either end of Sand Canyon, and one much farther away on Wetherill Mesa in Mesa Verde National Park. Although our data are currently inadequate to track clearly the patterns of vessel production and movement within and among these communities, they do suggest that most, if not all, settlements were involved in making pottery, and that sometimes substantial amounts of either raw materials or finished products moved between different communities. In addition, some settlements may have produced pottery for export while others relied more heavily on imported materials or vessels for their pottery inventories. Whether it was people, vessels, or raw materials that were moving between the different settlement clusters or communities, the degree of interaction suggested by these data far exceeds what we would expect if the Pueblo III communities were autonomous and self-sufficient, and challenges our current understanding of community organization and boundaries.




The Colorado Historical Society, NSF Grant DBS 91-02016, the 67th Pecos Conference Kiln Auction, and Crow Canyon Archaeological Center provided funding for this research. We received assistance for various parts of this project from Hector Neff, Mike Glascock, the late Lee Lacey, the Mesa Verde National Park Research Center staff, and the Anasazi Heritage Center staff. Materials analyzed for this study were collected under permits from the Bureau of Land Management, San Juan Resource Area, and Mesa Verde National Park, and with permission from private landowners. This chapter has also benefited from the comments of Barbara Mills, David Breternitz, Mark Varien, and Scott Ortman. We sincerely appreciate all of this support.