Regional Context: Architecture, Settlement Patterns, and Abandonment
Mark D. Varien



Architectural Change
Regional Data
Earthen Architecture
Masonry Architecture
Towers, Tunnels, and Kivas
The Pithouse-to-Kiva Transition
Sand Canyon Project Data
Settlement Patterns
Settlement Patterns and Site Location
Regional Survey Data
Sand Canyon Project Data
Settlement Patterns and Site Size
Regional Survey Data
Sand Canyon Project Data
Dispersed and Aggregated Settlement Patterns
Regional Survey Data
Sand Canyon Project Data
Summary and Conclusions: Settlement Pattern Data
Regional Data
Sand Canyon Project Data


This chapter places the results of the Site Testing Program in a regional context. Testing Program data relevant to the study of architectural change, settlement patterns, and abandonment are summarized and compared with survey and excavation data from elsewhere in the Mesa Verde region. The various survey and excavation projects drawn upon in this chapter include almost a century of research, and they address a wide range of questions using a variety of field and analytic methods. Problems resulting from methodological differences between research projects are resolved in part by the level of generality in the following review. Figure 21.1 illustrates the area considered in this context review, and Figure 21.2 shows the locations of the largest block surveys in the McElmo district.

The Pueblo II and Pueblo III periods are the focus of this discussion; earlier components identified at the tested sites are not discussed. When survey and excavation data are used, the site is the basic spatial unit of analysis. For the most part, the study is confined to sites interpreted as year-round habitations; less intensively used, seasonally occupied sites are discussed only briefly.

In general, the original Pecos classification (Kidder 1927) provides the temporal framework for discussing patterns and changes in patterns, but in virtually every report reviewed for this study, the researchers subdivided the Pecos periods into smaller units as well. These units either have their own phase names or are subdivisions of the Pecos periods (for example, early, middle, late). The various temporal frameworks used in each of the surveys considered in this chapter are summarized in Figure 21.3 and used throughout the remainder of the report.

According to the Pecos classification, the Pueblo II period lasted from A.D. 900 to 1100, and the Pueblo III period extended from A.D. 1100 to 1300. Subdivisions of these two phases are recognized on the basis of changes in pottery, architecture, and settlement patterns. A few researchers have considered regional patterns of organizational change in developing their temporal frameworks.

The Pueblo II period is usually subdivided into two or, less frequently, three temporal units. Most researchers recognize an early Pueblo II period, beginning around A.D. 900; ending dates for the early Pueblo II vary between A.D. 980 and 1050. The early Pueblo II period is traditionally seen as a time when corrugated gray ware replaced plain gray ware pottery, when Cortez Black-on-white was the dominant decorated white ware type, and before the appearance of pecked-block masonry. In every survey considered in this chapter, sites dating to the early Pueblo II period are reported. The date range for this early period, however, is called into question by a recent revision of pottery dating, which indicates that corrugated gray ware and Cortez Black-on-white did not become prominent until the late A.D. 900s (Blinman 1994). Several lines of evidence suggest that there was a substantial abandonment of the Mesa Verde region during the early 900s and that the region did not begin to become heavily settled again until the late A.D. 900s or early 1000s (Varien 1997; Wilshusen 1994).

Mancos Black-on-white was the most dominant decorated white ware type during the next period, which is sometimes referred to as the middle or late Pueblo II period. This period is seen as beginning sometime around A.D. 1000 and ending between A.D. 1060 and 1100. Traditionally, this is viewed as the period during which masonry architecture became more sophisticated (Hayes 1964:94), but in the discussion below, it is clear that earthen architecture actually was common during this period.

Adler (1988, 1990a) recognizes a late Pueblo II period beginning about A.D. 1060 and lasting until approximately 1150. The Chacoan system reached its peak during this period, and a number of sites dating to this interval in the Mesa Verde region have been interpreted as Chacoan outliers (Marshall et al. 1979; Powers et al. 1983). This period straddles the traditionally defined boundary between Pueblo II and Pueblo III. It also straddles a change in pottery, with Mancos Black-on-white being dominant earlier and McElmo Black-on-white becoming dominant by the end of the period.

The Pueblo III period is usually subdivided into early and late periods as well. McElmo Black-on-white was the most common decorated white ware type early in the period, and Mesa Verde Black-on-white was most common late in the period. Important changes in settlement patterns also took place during this time, with a shift from the mesa tops to cliff and canyon locations. In addition, larger, aggregated sites became more common through time.

Architectural Change

Close to 200 Pueblo II and III sites have been at least partly excavated in the northern Mesa Verde region, and more than 100 of these have produced tree-ring dates. Excavation methods and the level of site reporting vary considerably, but this large body of excavation data provides a basis for a discussion of architectural change. Syntheses of architectural change have been done before (Roys 1936:115-142; Morris 1939; Brew 1946:203-226; Lancaster et al. 1954:53-61; Lister 1966:65-87; Hayes and Lancaster 1975:73-97; Gillespie 1976; Wilshusen 1988b:599-634). In this section, I draw on these syntheses to describe general trends in architectural change, focusing primarily on the Pueblo II and III periods. Recent excavation data are used to illustrate the variability associated with these general patterns, providing a context for the discussion of the architecture of the tested sites.

Regional Data

Synthetic studies of architectural change from the Pueblo II to Pueblo III periods have focused on the adoption of masonry and subsequent changes in masonry style, as well as on the transformation of the pithouse as an architectural form into a kiva (often called the pithouse-to-kiva transition). In these studies, researchers have synthesized the results of fieldwork conducted over a number of years and in a variety of locations: Mesa Verde, Mancos Canyon, the Dolores River valley, the Lowry area, and Alkali Ridge (in southeastern Utah). Studies of architectural change in the Dolores River valley deal primarily with architectural change between the Basketmaker III and Pueblo I periods (A.D. 575-900), although Wilshusen (1988b:627-628) touches on Pueblo II and III architecture as well. Gillespie (1976) focuses primarily on Pueblo I to early Pueblo II architectural change. Synthetic studies of late Pueblo II and Pueblo III architectural change come primarily from the Lowry area (Roys 1936), from Mesa Verde (Lancaster et al. 1954; Hayes 1964; Lister 1966; Hayes and Lancaster 1975; Rohn 1977), and from Alkali Ridge in southeastern Utah (Brew 1946). Brew described the architecture uncovered during the Alkali Ridge Project in the 1940s and compared it to what was known about Basketmaker II through Pueblo III architecture in the entire region (Brew 1946:215-226).

The synthetic studies focus primarily on change through time, and they thoroughly document the general trends that characterize architectural change in each study area. The studies are linear because the researchers viewed one architectural style as evolving into another, and because they attempted to define the origins of particular architectural attributes.

The best of these studies acknowledge architectural variation across space at a given time, but they do not explore this variation as thoroughly as they document changing trends over time. Brew in particular takes note of the variation in architectural details present and points out that "there is no place for them in the pigeon-holes of taxonomy. Yet they are the very life-blood of 'progress'" (Brew 1946:216). Although Brew describes examples that do not conform to the general trends, his summary of architectural change is still largely typological and linear, with the variation of one period stabilizing into the norm of the next. Wilshusen's work also documents and quantifies the variation present in architectural features, but his sample sizes are small for the Pueblo II and III periods.

The following four sections address various aspects of architectural change from a regional perspective. The section on earthen architecture focuses on surface rooms and roomblocks constructed primarily of earth. Next, a section on masonry architecture discusses the use of masonry in surface rooms and roomblocks, and the subsequent changes in masonry styles. This is followed by a section on towers, tunnels, and kivas. Finally, a section on the pithouse-to-kiva transition focuses on the shift from earth-walled pithouses to masonry-lined kivas.

Earthen Architecture

Surface rooms constructed primarily of earth include a wide variety of building styles. Five such varieties are defined by Brew at Site 13 on Alkali Ridge in southeastern Utah, which dates to the late A.D. 700s (Brew 1946:191). Three of these types incorporate some stone into the lower portions of earthen walls--upright slabs, upright slabs with horizontally coursed masonry on top of the slabs, and horizontally coursed masonry foundations. Upper walls are not preserved, but smaller stones often found in the fill are interpreted as having once been incorporated into the earthen walls. Similar varieties of earthen surface rooms continued to be built through the A.D. 800s throughout the northern Mesa Verde region as documented at Ackmen-Lowry (Martin 1939), in the Dolores River valley (Varien 1988:81-143; Morris 1988c:671-712), near the modern town of Cortez (Lightfoot and Etzkorn 1993), at Mesa Verde (Hayes and Lancaster 1975; Lancaster et al. 1954; O'Bryan 1950), in the area south and east of Mesa Verde (Morris 1919, 1939), and even farther east in the Piedra River valley (Roberts 1930). At most sites where there are earthen surface rooms, the rooms are found in association with earth-walled pit structures. Wilshusen (1988b:593-634, 1989a:626) demonstrates an important link between the construction of surface rooms and pit structures: the backdirt from pit structure excavation probably was used in the construction of surface rooms.

As a result of these studies, especially the work at Ackmen-Lowry (Martin 1939) and early work at Mesa Verde (O'Bryan 1950), earthen architecture became associated with the Basketmaker III and Pueblo I periods, whereas masonry architecture was seen as typifying the Pueblo II and III periods (Cordell 1984:103). The transition between the two styles was documented as having occurred during the early Pueblo II period (Hayes 1964:91; Lancaster et al. 1954).

Much of the early work took place on multiple-component sites where different types of architectural remains were stratigraphically superimposed. One type of earthen architecture documented in these studies consists of what are often called jacal roomblocks. I prefer the term "post-and-adobe" because the remains of these walls found at the tested sites consist of closely spaced post holes and traces of earthen walls, but show no evidence of a jacal framework on the interior of the walls. The term "post-and-adobe" as used here should not be confused with the post-and-adobe construction of the southern deserts.

At Mesa Verde, these post-and-adobe rooms were overlain by masonry rooms at Site 1 (O'Bryan 1950:44) and at Site 16 (Lancaster et al. 1954:32), whereas at Badger House, post-and-adobe rooms were covered by the midden from a later Pueblo II occupation of the site (Hayes and Lancaster 1975:73). At Site 16 and at Badger House, the post-and-adobe rooms are interpreted as dating to the early A.D. 900s on the basis of the types of pottery found in the overlying deposits; at Site 1, the post-and-adobe rooms date to A.D. 859. The masonry roomblock covering the post-and-adobe rooms at Site 16 is interpreted as dating to between A.D. 975 and 1000 (Lancaster et al. 1954:37, 78), and the masonry overlying the post-and-adobe rooms at Site 1 is tree-ring dated to A.D. 1064 (Robinson and Harrill 1974:78-79). Masonry foundations dating to the A.D. 960s are present at Site 1676 on Wetherill Mesa, but the earliest full-height masonry walls at Badger House are interpreted as dating to the Mancos phase, A.D. 975-1075 (Hayes and Lancaster 1975:79; see Hayes [1964:94] for the date range of the Mancos phase). On the basis of the Wetherill Mesa survey data, Hayes argued that masonry replaced post-and-adobe construction at the end of the preceding Ackmen phase, A.D. 900-975 (Hayes 1964:91).

Martin found a sequence of superimposed architectural units at Little Dog Ruins (Martin 1930). The earliest building was a post-and-adobe house, overlain by a masonry house with walls one stone wide, which in turn was overlain by another masonry house. Martin (1939) also found post-and-adobe rooms adjacent to rooms with lower walls of upright slabs at Site 2 in the Ackmen-Lowry area; these rooms are tree-ring dated to the A.D. 760s.

Farther west, Brew found post-and-adobe rooms on Alkali Ridge. Brew provides the most thorough description of the variation in earthen architecture, including Pueblo I and Pueblo II varieties of post-and-adobe walls, and he explicitly discusses the differences between the two (Brew 1946:191, 220). The Pueblo II varieties were found at sites where, as at Mesa Verde, they were stratigraphically earlier than masonry rooms. Both the post-and-adobe roomblocks and their masonry replacements are interpreted as dating to the Pueblo II period.

Brew discusses the distribution of known post-and-adobe surface rooms and argues that in considering them we deal "not with isolated local variations but with a widespread fashion" (Brew 1946:220). Subsequent excavation has proven Brew to be correct and has demonstrated that the post-and-adobe roomblocks were constructed later than suggested by the excavations at Mesa Verde and Ackmen-Lowry. This is best illustrated by recent excavations on the South Canal Project, located just west of Pleasant View, Colorado. A number of sites with earthen architecture dating to throughout the A.D. 1000s were excavated as part of this project (Kuckelman and Morris 1988). These sites contain post-and-adobe surface rooms and earth-walled pit structures, which are sometimes surrounded by what are interpreted as stockades. The sites include Chameleon House, which is tree-ring dated to approximately A.D. 1020; Dobbins Stockade, tree-ring dated to the A.D. 1030s; Casa Bisecada, tree-ring dated to sometime after A.D. 1051; and Bindweed House, tree-ring dated to sometime around A.D. 1071.

To the southwest of the South Canal Project, excavation on the Hovenweep Laterals Project also uncovered post-and-adobe roomblocks, earth-walled pit structures, and stockades (Morris 1991a). These features are present at Dripping Springs Stockade, which was tree-ring dated to approximately A.D. 1043, and at the Pueblo II component at Roundtree Pueblo, tree-ring dated to approximately A.D. 1078.

A site with surface rooms constructed predominantly with adobe was excavated within the Sand Canyon locality, approximately 5 km west of the upper Sand Canyon mesa-top tested sites. This site, Gnatsville, is tree-ring dated to the A.D. 1030-1050 period (Kent 1991a:59). Even at Mesa Verde, where a number of excavations indicate that post-and-adobe architecture dates to the A.D. 900s, there is a post-and-adobe component at Big Juniper House dating to the A.D. 1050-1080 period (Swannack 1969:179). Wattle-and-daub rooms are abutted to masonry rooms in several of the cliff dwellings constructed in the late A.D. 1200s, including Long House (Cattanach 1980:12).

Excavations therefore establish that the use of earthen architecture continued through the A.D. 1000s. At many recently excavated sites, post-and-adobe roomblocks have been found beneath masonry roomblocks--a sequence noted at many of the early excavations discussed at the beginning of this section. This is true at Big Juniper House, where the post-and-adobe structures are overlain by a masonry roomblock that is interpreted as dating to the A.D. 1080-1130 period. At the Ewing Site, there are four separate roomblocks where post-and-adobe rooms lie beneath masonry rooms. The post-and-adobe occupations at the Ewing Site are tree-ring dated to approximately A.D. 1040 (Hill 1985:20). The superposition of Pueblo II post-and-adobe and Pueblo III masonry structures is also documented at Site 5MT3, a site excavated under the direction of Joe Ben Wheat and located near Yellow Jacket Pueblo (Cater 1989).

Summarizing the regional data on earthen architecture, post-and-adobe roomblocks and earth-walled pit structures continued in widespread use throughout the A.D. 1000s in southwestern Colorado and southeastern Utah. The Pueblo II post-and-adobe architecture differs from the earth construction found in Pueblo I sites, and these differences have been summarized by Brew (1946:191, 220). Finally, sites with post-and-adobe roomblocks and earth-walled pit structures were frequently reoccupied by groups using masonry architecture.

Masonry Architecture

In discussions of masonry architecture, the distinction needs to be made between (1) masonry incorporated into the lower portions of walls whose upper portions are predominantly adobe or adobe over a vegetal framework and (2) full-height masonry walls. Related to this is the distinction between non-load-bearing walls, which do not support the roof, and load-bearing walls, which do. In general, earthen walls and composite masonry-and-earth walls are non-load-bearing walls; full-height masonry walls are load-bearing walls.

As noted above, several examples of Pueblo I-style walls incorporate stone into their construction (e.g., Brew 1946:191), but these walls are predominantly earth above the masonry foundations. Morris (1988c) describes how the typical Pueblo I wall does not have enough standing masonry, or rubble in the collapsed wall, to indicate full-height masonry. He offers a reconstruction of how walls with masonry at the base and adobe over a vegetal framework on top might have looked (Morris 1988c:680).

Another type of composite wall, megalithic slab architecture, is most common during the Mancos phase (Hayes 1964:37, 94), although Rohn (1977:253) dates it to between the late Pueblo II and late Pueblo III periods. This type of wall has large upright slabs at its base and presumably adobe and randomly added spalls of stone in the remainder. Excavated sites with this type of construction include Site 5MT8899 (Hammack et al. 1992), Hindmarsh One (Flint 1994), and several sites at Mesa Verde (Fewkes 1923; Hayes 1964:38; Lister 1968:9, 35, 38; Rohn 1977:254).

Although most Pueblo I walls that incorporate stone into their construction are predominantly earthen, there may be exceptions. McPhee Pueblo, particularly Rooms 66 and 67, is interpreted as having full-height walls, one stone wide (referred to in the remainder of this chapter as "single-stone walls"), that were built in the late A.D. 800s (Brisbin et al. 1988:104-110). Martin (1939:324) argued that the same kind of walls were present at Site 1, which tree-ring dates to the middle to late A.D. 800s.

The construction of single-stone walls continued into the Pueblo II period, and some of these were full-height masonry walls. These single-stone masonry walls are sometimes referred to as chipped edge (Rohn 1977:47) or scabbled masonry (Hayes 1964:37) walls because of the distinctive shaping on the edges of the stone. In fact, Rohn places this type of masonry, which he distinguishes from "crude masonry," in the late Pueblo II and early Pueblo III periods (Rohn 1977:253). At Badger House, scabbled masonry characterizes the Mancos-phase rooms, which date between A.D. 1000 and 1075 (Hayes and Lancaster 1975:184). These single-stone masonry walls are similar to Type I-style masonry at great houses in Chaco Canyon (Lekson 1986:17), and such walls may have stood two stories high at Wallace Ruin, a Chacoan outlier north of Mesa Verde (Bradley 1988b:8). Thus, the widespread use of post-and-adobe architecture, which was the only building style at some sites throughout the A.D. 1000s, continued, despite the fact that the first known full-height masonry walls were constructed 200 years earlier. Composite masonry-and-adobe walls, including walls with megalithic slabs, were also used during this period.

Post-and-adobe construction, single-stone masonry, and megalithic slab masonry eventually became less common, and walls built two stones wide (referred to in the remainder of this chapter as "double-stone walls") became more common. There is a difference between double-stone walls, in which no stone spans the full width of the wall, and compound walls, which are mostly two stones wide, but which occasionally have stones spanning the full width of the wall. There are also double-stone walls with rubble cores, which many authors refer to as core-veneer masonry. All of these walls were almost certainly load bearing, that is, they were full-height masonry walls that supported the weight of the roof. Pecked-face masonry also became more common with the construction of double-stone and double-stone walls with rubble cores (Hayes 1964:99; Rohn 1977:56).

The date of the change from single- to double-stone walls varies from one area to the next. An important question, one which is outside the scope of this report, is whether this variation is the result of imprecise dating or whether the change really did occur at different times in different places. In this chapter, I will use the dating arguments presented by the individual authors and note cases in which pottery dating seems to contradict the results of tree-ring dating.

At Site 875 on Mesa Verde, the stratigraphically earlier building has single-stone walls and is interpreted as having been occupied between A.D. 975 and 1000. The superimposed building with double-stone walls is interpreted as dating to the A.D. 1025-1075 period (Lister 1965:109). Given the amounts of McElmo Black-on-white pottery at this site, an A.D. 1100-1140 date seems more likely for the roomblock with double-stone walls. At Site 866 on Mesa Verde, both single- and double-stone walls are present. This site has an A.D. 1060 cutting date, which is consistent with the pottery assemblage, in which Mancos Black-on-white is by far the most common decorated white ware type (Lister 1966:61). Unit II at Site 16 on Chapin Mesa has double-stone masonry and an A.D. 1074 cutting date for the associated kiva. The pottery assemblage from this site is also predominantly Mancos Black-on-white, with no Pueblo III decorated white ware types (Lancaster et al. 1954:78). Site 499 (Lister 1964), Site 1253, and the late component at Big Juniper House (Swannack 1969) all display double-stone masonry that is tree-ring dated to the A.D. 1123-1130 period.

Hayes argues that double-stone walls were first built during the Mancos phase, became the rule during the McElmo phase, and that after A.D. 1150 on Mesa Verde there was little change in the masonry styles except for greater shaping of the stones, which included grinding as well as pecking (Hayes 1964:94, 97, 100).

The shift to double-stone walls at sites not located on Mesa Verde appears to have been sporadic. As noted above, full-height, single-stone masonry walls may have been present as early as the late A.D. 800s at McPhee Pueblo in the Dolores River valley. In addition, a few double-stone walls were found in rooms constructed predominantly with single-stone masonry in the next occupation of McPhee Pueblo, which dates to A.D. 910-980 (Brisbin et al. 1988:251-317).

The earliest consistent use of double-stone, compound, and double-stone-with-rubble-core walls occurs at sites interpreted as Chacoan outliers. All walls at Lowry, with the exception of one, are double-stone with rubble core (Martin 1936:26), and the earliest construction at this site dates to approximately A.D. 1090 (Martin 1936:194; Robinson and Harrill 1974:17). Bradley (1988b:15) documents double-stone-with-rubble-core construction at Wallace Ruin during the first two decades of the A.D. 1100s. At Escalante Ruin, this same type of construction is tree-ring dated to A.D. 1129, although the types of pottery found suggest that an earlier date is possible (Hallasi 1979).

Double-stone walls are also documented for small sites dating to the early A.D. 1100s, as evidenced by the Dominguez Ruin, which, on the basis of tree-ring dates, is inferred to have been constructed in about A.D. 1123. One of the small units at Cahone Ruin (also called Ansel Hall Ruin--a possible Chacoan outlier with a great kiva and numerous tightly clustered small sites) appears to have double-stone walls and pecked-face masonry, although the wall description is not entirely clear (Guthe 1949:144-154). Noncutting tree-ring dates indicate that this small site at Cahone Ruin was constructed and used after A.D. 1074 (Robinson and Harrill 1974:19), and the near absence of McElmo Black-on-white pottery argues for abandonment by A.D. 1100. Within the Sand Canyon locality, the earliest tree-ring-dated double-stone walls were built at the Mustoe Site in A.D. 1174 (Gould 1982).

Towers, Tunnels, and Kivas

People began building towers at the same time that they developed masonry architecture. Greubel (1991:84) argues that towers first appeared as circular structures attached to the ends of roomblocks during the late Pueblo II period. Freestanding towers detached from the roomblock are found at the Mancos-phase sites on Wetherill Mesa (Hayes 1964:94) and at sites dating to the late Pueblo II period on Chapin Mesa (Rohn 1977:239). In the Hovenweep area and on Mockingbird Mesa, freestanding towers did not appear until the early Pueblo III period (Greubel 1991:95; Fetterman and Honeycutt 1987:86). Neily argues that all towers on Cow Mesa and Squaw Point postdate A.D. 1150 (Neily 1983:105). As with many of the general trends reported above, change appears to have occurred on Mesa Verde before it occurred in other parts of the region. In all survey areas, the frequency of towers increased through the Pueblo III period.

Towers at habitation sites are most often spatially associated with kivas, and excavation data from throughout the region indicate that the two structures are commonly linked by a tunnel (Fetterman and Honeycutt 1987:93; Gould 1982:95; Luebben 1982:66; Luebben and Nickens 1982:69; Martin 1929:6, 24; Lister 1964:43; Lancaster et al. 1954:90; Fewkes 1921:89). Tunnels also commonly link kivas to roomblocks (Prudden 1914:45, 1918:9, 18, 29; Martin 1929:6, 24; Cater 1989:18, 87), kivas to subterranean rooms (Cater 1989:86; Luebben 1982:66, 71; Luebben and Nickens 1982:69; Martin 1929:6), and kivas to kivas (Luebben and Nickens 1982:69; Cater 1989:87; Lister 1964:11). Although tunnels are known in the Mesa Verde region as early as the Pueblo I period (Roberts 1930:42), the elaboration noted above occurred predominantly during the Pueblo III period.

A specific type of site documented by both survey and excavation consists of sites where towers and kivas are the main architectural features and roomblocks are absent. Hayes (1964:112), Rohn (1977:116-117), and Smith (1987:36, 39, 44) all document numerous sites on Mesa Verde that consist of a tower and a kiva. Excavated examples on Mesa Verde include Far View Tower and Cedar Tree Tower (Fewkes 1921, 1922, 1923). Off the mesa, tower-kiva sites are not specifically noted in the major survey reports, but isolated towers are reported (Fetterman and Honeycutt 1987), and these may have associated kivas not visible on the surface. Sites that consist of a tower and associated kiva depression are also documented in the Bureau of Land Management site files, for example, Lightning Tree Tower (Site 5MT1691), located near Burro Point, and Site 5MT4878, located approximately 2 km south of Moqui Lake.

The Pithouse-to-Kiva Transition

The changes in architectural form referred to as the pithouse-to-kiva transition have been capably described by a number of authors (Lancaster et al. 1954:55-61; Lister 1966:77-87; Gillespie 1976). The details of these studies are not repeated here; however, it needs to be emphasized that I view this transition as an important change in architectural form, but not necessarily a change in the use of the buildings. Changing use of these structures needs to be evaluated by other data sets in conjunction with the changes in architectural form, as has been attempted by Gillespie and numerous authors who have recently addressed the question of kiva use (cf. Lipe and Hegmon, eds. 1989). These studies indicate that pit structures with specialized ritual functions may have existed before the architectural changes normally equated with kivas took place (Wilshusen 1989b), and that many pit structures with all the architectural attributes of kivas may have functioned primarily as domiciles (Cater and Chenault 1988; Lekson 1988, 1989; Lipe 1989).

The changes in architectural form that mark the pithouse-to-kiva transition do mirror the changes that took place in surface room architecture in one important respect: they reflect a shift from earthen to masonry buildings. Perhaps most importantly, there is a change from roofs supported by upright posts to roofs supported either by load-bearing masonry walls or by load-bearing masonry pilasters. The earliest dated kiva with stone pilasters at Mesa Verde is the kiva at Site 1, which is tree-ring dated to A.D. 1064 (O'Bryan 1950:49; Robinson and Harrill 1974:78). The earliest kiva fully lined with masonry also appears to date to the second half of the A.D. 1000s (Lister 1966). As with surface architecture, variability characterizes pit structure construction. Clearly, sites with masonry kivas were contemporaneous with sites with earth-walled pit structures.

Sand Canyon Project Data

Architectural remains dating to the Basketmaker III, Pueblo II, and Pueblo III periods were found during Testing Program excavations. Pueblo II architecture includes post-and-adobe surface rooms at G and G Hamlet, Lillian's Site, and Kenzie Dawn Hamlet, and earth-walled pit structures at G and G Hamlet and Kenzie Dawn Hamlet. At each of these sites, the Pueblo II earthen architecture is tree-ring dated to between A.D. 1050 and 1100, which is consistent with the widespread occurrence of earthen architecture in the late Pueblo II period documented elsewhere in the region.

Post-and-adobe rooms were also documented at the Green Lizard site. Although a Pueblo II component is present at this site, the post-and-adobe rooms appear to date to the A.D. 1200s (Huber 1993; Huber and Lipe 1992). A row of posts at Shorlene's Site is probably part of a post-and-adobe room that dates to the late Pueblo III occupation. Although post-and-adobe architecture is noted in many of the cliff dwellings that date to the A.D. 1200s, it has not been associated with open sites dating to the late Pueblo III period at Mesa Verde.

Masonry roomblocks are superimposed over post-and-adobe roomblocks at G and G Hamlet, Lillian's Site, and Kenzie Dawn Hamlet. The roomblock masonry at each of these sites consists of double-stone or compound walls. These masonry roomblocks date to either the late A.D. 1100s or early A.D. 1200s.

Masonry roomblocks dating to the Pueblo III period are present at 11 of the 13 tested sites. Troy's Tower has no masonry surface rooms. Mad Dog Tower has one surface room with single-stone walls; collapsed wall debris was minimal, and the preserved portion of the wall stood only 23 cm high. Therefore, the room walls at Mad Dog Tower may have been of composite construction, with the upper walls built of adobe or adobe covering a vegetal framework. The masonry walls at the remaining tested sites are all double-stone, compound, or double-stone with rubble core.

Twenty-eight of the tested pit structures date to the Pueblo III period, five date to the late Pueblo II or early Pueblo III periods, two date to the Basketmaker III period, and two pit structures were difficult to date but were probably constructed during the late Basketmaker III or early Pueblo I periods. Little can be said about the late Pueblo II/early Pueblo III pit structures because only small portions were exposed during excavation, but all are earth-walled structures without benches, and they are not as deep as the Pueblo III pit structures.

Of the 28 Pueblo III pit structures, 24 are masonry lined. One of the tested pit structures is earth walled, but it has a bench and masonry pilasters. The presence or absence of masonry cannot be determined for three of the pit structures, because the test pits did not expose the structure walls. Thus, at the Testing Program sites, the transition from earthen architecture to masonry architecture was perhaps never completed, with some earthen buildings present in late components. However, the majority of the earthen buildings exposed by testing date to before A.D. 1100, and the majority of masonry buildings date to after A.D. 1100. Even the earth-walled pit structures that date to the post-A.D. 1100 period have masonry pilasters, indicating that the roof was no longer supported by upright posts set in the ground.

The pit structure at Green Lizard is also predominantly earth walled, but it has masonry pilasters, a masonry-lined southern recess, and some masonry reinforcement above a tunnel opening in the otherwise earthen bench face (Huber 1993). The 10 kivas that were excavated at Sand Canyon Pueblo are all masonry lined (Bradley 1993). The Sand Canyon Pueblo kivas differ from the kivas at the tested sites because, for the most part, they are not subterranean. Instead, most of the Sand Canyon Pueblo kivas are above-ground or partly subterranean structures enclosed in square rooms; the spaces created between the curved kiva walls and the right-angle walls of the square enclosing structures often were used as small rooms, called "corner rooms." This type of construction may have been partly necessitated by the topography at Sand Canyon Pueblo--in many parts of the site, bedrock was exposed or present just below the ground surface, making the excavation of a subterranean kiva impossible.

The transition from earthen architecture to masonry architecture has important implications for the mobility strategies of the groups that use the buildings and for the length of occupation of individual sites (Varien 1997). A number of ethnographic studies have addressed the question of structure use life, especially the use life of earthen architecture (McIntosh 1974; Weltfish 1965; Wilson 1934; cf. Cameron 1991). Other studies of structure use life have combined ethnographic data with archaeological data from the prehistoric Puebloan region (Ahlstrom 1985b; Schlanger 1987). These studies place the use life of earthen buildings at six to 12 years, or, with extensive remodeling and rebuilding, up to 20 or perhaps 30 years.

The life span of earthen buildings is limited by the life span of the posts that support the roof (other factors such as vermin infestation and insufficient maintenance can also limit structure use life). Because the construction of full-height, load-bearing masonry walls eliminates the need for roof-support posts, the use life of masonry structures is potentially greater than that of earthen structures. Ethnographic studies on the use life of masonry buildings are not as developed as those for earthen buildings, although Ahlstrom (1985b:642) examines archaeological data and suggests that the use life of masonry-lined kivas exceeds the 20-year use life he proposes for earthen pithouses. Use lives of masonry buildings in historic-period pueblos can be 80 to 100 years or more (Ferguson and Mills 1988; Ahlstrom et al. 1991). Greater use life of masonry buildings has also been demonstrated with a cross-cultural data set (Diehl 1990, 1992; Diehl and Gilman 1996).

Analyses have been completed in which artifact assemblages are used to evaluate site use life independently of the type of architecture present (Varien 1997). These demonstrate that the shift from earthen to masonry architecture was accompanied by longer site use life. Thus, residential moves became less frequent after A.D. 1100.

Settlement Patterns

The following discussion of settlement patterns addresses changes in site location and size, as well as the shift from dispersed to aggregated settlement. Through time, site location shifted from mesa tops to canyons, site size increased and became more variable, and settlements became more aggregated. Settlement patterns are also used to define communities and community boundaries. In each section, the regional survey data are summarized to provide a context for these discussions, and the Sand Canyon Project results are then compared with the larger regional patterns.

Settlement Patterns and Site Location

Regional Survey Data

The shift in site location from mesa tops to canyons has been noted in every major survey conducted in the Mesa Verde region. At Mesa Verde, Hayes (1964), Rohn (1977), and Smith (1987) all noticed increased use of the canyons for site locations through time. Hayes (1964:109) argues that talus-slope sites on Wetherill Mesa were present as early as the Ackmen phase, and the number of talus-slope sites increased during the Mancos phase. Hayes notes that by the McElmo phase "the most popular type of settlement was a unit pueblo on the talus or in the canyon bottom. Most of these were on broad benches at the canyon heads or were built at the top of the talus against the cliff" (Hayes 1964:109). During the Mesa Verde phase, the final period of occupation on Wetherill Mesa, the number of sites located on both the mesa top and the talus slopes decreased, as alcoves in the cliffs became the most common locations for sites (Hayes 1964:110, Table 7).

Rohn argues that talus-slope sites on Chapin Mesa occurred first in the Mancos phase but were more common in the subsequent McElmo phase (Rohn 1977:70, 241). Like Hayes, Rohn (1977:243) argues that Mesa Verde-phase occupation was confined mostly to the cliff dwellings, with some remnant occupation of the mesa-top sites continuing in this late period. Although Rohn does not mention talus-slope sites in his summary of the Mesa Verde phase, the table that summarizes the information about these sites indicates that most have Mesa Verde Black-on-white pottery, so they are assumed to have had Mesa Verde-phase occupations (Rohn 1977:72-73, Table 9). Talus-slope sites are clearly less common on Chapin Mesa than on Wetherill Mesa. On Wetherill Mesa, these sites were usually associated with check dams and terracing, whereas on Chapin Mesa, the talus slopes were steeper, had fewer benches, and were subject to less terracing. These differences are believed to account for the different frequency of talus-slope sites on the two mesas (Alden Hayes, personal communication 1991). Rohn also documents a shift in settlement from the north part of Chapin Mesa, which was heavily occupied during the McElmo phase, to the south end of the mesa, which saw an increase in population during the Mesa Verde phase; this corresponds to the increased occupation of alcoves, which are rare on the north end of the canyons (Rohn 1977:288, Figure 99).

Smith (1987) summarizes the survey data for the remainder of Mesa Verde National Park. He does not identify talus-slope sites as a specific site type in his report, but he does mention that McElmo-phase sites are located mainly on the mesa tops and talus benches. Cliff shelters are reported as the main site locations during the Mesa Verde phase, but mesa tops and talus-slope benches were still occupied (Smith 1987:67). Smith also reports that the east side of Mesa Verde National Park was abandoned earlier than the west half of the park (Smith 1987:66).

Although all of the above authors agree that the cliff shelters became the focus of occupation in the A.D. 1200s, tree-ring dates from mesa-top sites indicate that at least some of these were also occupied late into the A.D. 1200s (Robinson and Harrill 1974:45, 93, 59). For example, on Chapin Mesa, there are noncutting dates of A.D. 1201 for Site 820, A.D. 1214 for Pipe Shrine House, and A.D. 1243 for Far View House. On Wetherill Mesa, Badger House has a cutting date cluster that terminates at A.D. 1258.

Dates recently obtained for Oak Tree House indicate that construction of this cliff dwelling began between A.D. 1190 and 1207 (Tom Windes, personal communication 1993); at Balcony House (Fairchild-Parks and Dean 1993) and Spruce Tree House (Dean, personal communication 1993), the earliest construction appears to date to the early 1200s. Construction at these sites continued through the A.D. 1270s. The Johnson Canyon cliff dwellings, located just south of the Mancos River in the Ute Mountain Tribal Park, are interpreted as having even earlier construction, with one episode dating to the A.D. 1130s and 1140s and a second construction episode dating to the late A.D. 1100s and early 1200s (Nickens 1981b:17). Here, construction appears to have continued into the A.D. 1240s.

Outside Mesa Verde National Park, a number of survey projects have been completed in the northern Mesa Verde region. The survey located farthest north was conducted on Squaw and Cow Mesas (Neily 1983). Neily's survey was essentially a mesa-top survey, and use of the talus slopes is not documented. Neily's survey does document a shift in site location in the late Pueblo III period (A.D. 1230-1300) away from the central portion of the mesa tops to the canyon rims at the edges of the mesas. The canyon rim sites are typically located at the heads of canyons and near springs (Neily 1983:122).

Just south of Neily's survey, the areas in and around Hovenweep National Monument have been surveyed (Greubel 1991; Hammett and Olsen 1984; Riley 1948; Schroeder 1968; Winter 1975, 1976, 1977). These surveys tend to complement one another. Riley's survey focused on the five monument units. Schroeder's survey covered the mesa top between Square Tower and Holly House. The survey supervised by Winter combined small block surveys associated with the Cutthroat Castle, Hackberry-Horseshoe, Holly House, Square Tower, and Cajon Ruin groups with four 40-acre quadrats designed to sample the environmental-physiographic diversity in the Hovenweep area. Greubel's survey covered a 4,090-acre area surrounding, or adjacent to, four of the Hovenweep units: Square Tower, Cutthroat Castle, Hackberry-Horseshoe, and Holly House. Greubel's survey did not include the area inside the monuments themselves.

Each of these surveys documented substantial Pueblo II and III settlement in the area. Winter (1976:286) argues that there was an enormous increase in population during the late Pueblo II period. Greubel (1991:80-89) documents a similar phenomenon and attributes this to the expansion of the dry-farming belt during the A.D. 1000-1100 period proposed by Petersen (Greubel 1991:135; see Petersen 1986:Figure 58). Winter (1976:287) reports the greatest number of sites in the late Pueblo II period, but Greubel (1991:89) finds the greatest number of sites in the early Pueblo III period. The largest sites in both Winter's and Greubel's surveys date to late Pueblo III.

Winter (1976:287-290) argues that the shift from mesa tops to canyon rims took place during the Pueblo III period. Greubel argues that long-term habitation sites were predominantly located on the mesa tops during the Pueblo II period (Greubel 1991:86-87), but he also documents use of the canyon rims and talus slopes for site locations during this same time (Greubel 1991:86-89). Greubel reports a decrease in mesa-top site locations in the Pueblo III period and an accompanying increase in rimrock site locations; he notes that this "trend is most apparent for long-term habitations, suggesting that populations, at least at the household level, were selecting rimrock locales for permanent habitation more frequently than ever before" (Greubel 1991:96).

A mesa-top survey of Mockingbird Mesa was conducted just east of the Hovenweep surveys (Fetterman and Honeycutt 1987). This survey, too, documented a population increase from the Pueblo II to Pueblo III periods. Like other researchers, Fetterman and Honeycutt report a shift in site location from the central portion of the mesa to the mesa margins. They argue that this change began during the late Pueblo II period, intensified through the Pueblo III period, and was most pronounced during the late Pueblo III period (Fetterman and Honeycutt 1987:71, 90).

The Dolores River valley was the focus of another large block survey. This survey documented peak population during the Pueblo I period, with very low population levels during the Pueblo II and early Pueblo III periods, and virtually no population in the valley during the late Pueblo III period (Schlanger 1985, 1988). Thus, Pueblo II-III settlement-pattern shifts are not as applicable to the Dolores valley.

The Sacred Mountain Planning Unit survey was designed to provide a statistically reliable projection of the cultural resources within a 702,000-acre area covering a portion of the McElmo drainage system (Chandler et al. 1980). Ecologically defined areas within the planning unit served as the basis for a stratified random sample in which 80-acre quadrats were defined as the sampling units. One question addressed by the survey was whether there was movement from the mesa tops to the canyons through time. The results are somewhat ambiguous, in part because mesa tops and canyon rims were combined into a single "mesa" category, and other site locations were combined into a single "canyonlands" category (Chandler et al. 1980:92). Using these groupings, the researchers found no statistically significant shift in site location through time. When habitations were examined, however, they found that canyonland habitations increased from 33 percent during the Mancos phase to 72 percent during the Mesa Verde phase (Chandler et al. 1980:92).

In southeastern Utah, the largest survey undertaken to date was a quadrat-based survey on Cedar Mesa (Matson et al. 1988). Occupation on Cedar Mesa was characterized by more clearly defined periods of abandonment and reoccupation when compared with the areas discussed above. Pueblo II and III occupation began in A.D. 1060 and ended at approximately A.D. 1270, with a possible hiatus between A.D. 1150 and 1165. Increased use of the canyons was documented for the last phase of occupation, the Red House phase (A.D. 1210-1270) (Matson et al. 1988:254). In addition to sites recorded during this systematic survey, several large sites associated with canyon heads, canyon rims, and mesitas overlooking canyons have been reported elsewhere in southeastern Utah (Varien et al. 1996). The same is true in southwestern Colorado, where numerous large canyon-rim sites have been reported that date to the late Pueblo III period, just before regional abandonment (Varien et al. 1996).

The survey farthest west is on the Red Rock Plateau (Lipe 1970). Periods of abandonment were longer and periods of occupation were shorter in this area. The Kletha phase (A.D. 1100-1150) (Lipe 1970:87) is characterized by material remains that are most similar to those found in the Kayenta region. Use of the Red Rock Plateau during the Kletha phase focused on areas where canyons were broad, shallow, and accessible and where the best agricultural soils were located (Lipe 1970:114). After a 60-year abandonment, there was another occupation during the Horsefly Hollow phase (A.D. 1210-1260). This occupation includes sites in the south and west portions of the survey area that have predominantly Kayenta-region traits and sites in the north and west that have Mesa Verde-region traits. During this phase of occupation, the same areas used during the Kletha phase were settled, but there was intensified use of the narrower, deeper, and less-accessible canyons. This is interpreted as being a result of increasing population density and the need to make the fullest possible use of the land resources (Lipe 1970:125).

In summary, the shift from mesa tops to canyon rims was documented in every large survey area where Pueblo II and III settlements were present. Survey results differ as to when this shift began; some researchers believe that it started as early as the beginning of the Pueblo II period (the A.D. 900s), while others argue that the shift took place entirely within the Pueblo III period (A.D. 1150-1300). In all project areas that included talus slopes and canyon bottoms, use of these locations during the Pueblo II period (A.D. 900 to 1100) is documented. In all survey areas, however, use of the talus slope increased during the Pueblo III period, especially with regard to use of these locations for long-term habitation sites.

Sand Canyon Project Data

Upper Sand Canyon survey data were not reported with regard to physiography, but the location of sites in relation to arable soils was recorded. Soils are deepest and most arable on the mesa tops and thinner on the canyon rims and talus slopes. Adler found that, through time, more sites were associated with less arable soil, a result of more sites being located on the canyon rims and talus slopes. This trend began in the early Pueblo II period and continued through the Pueblo III period (Adler 1990a:239).

The lower Sand Canyon surveys (Gleichman and Gleichman 1989, 1992; Adler and Metcalf 1991) are located in canyon environments just above McElmo Creek. The McElmo Creek floodplain is broad and relatively well watered compared with many Mesa Verde-region canyon bottoms, which are narrow and contain little arable soil. The lower Sand Canyon survey area is remarkable for the scarcity of Pueblo II remains. There are only four habitation sites and four limited-activity sites in the Gleichman and Gleichman (1989) and Adler and Metcalf (1991) surveys. This is the lowest density of Pueblo II remains in any of the survey areas.

Late Pueblo II habitation sites are slightly more numerous, and in both survey areas, these sites are located in the extreme southern end of the survey areas, near the McElmo Creek floodplain. In addition, in a recent study of the alluvial history of McElmo Creek, Force and Howell (1997) recorded several sites in the McElmo floodplain that date to the late Pueblo II period. Sites in the McElmo floodplain are covered by postabandonment alluvial deposits, the result of McElmo Creek aggrading during the Pueblo II and III periods (Force and Howell 1997). The Pueblo II sites are visible only where they have been exposed by recent arroyo cutting, so it is difficult to estimate the size of the Pueblo II community. It is clear, however, that the canyons and talus slopes in lower Sand Canyon were not intensively occupied during the Pueblo II period.

During the Pueblo III period, the number of sites in the lower Sand Canyon survey areas increased dramatically. Seventeen habitation sites and 19 limited-activity sites were recorded in the two survey areas (Gleichman and Gleichman 1989:42, 1992; Adler and Metcalf 1991:47). These are small sites with only one or two kivas, except for Castle Rock Pueblo, which has 12 to 15 kivas. Survey indicates that occupation of the canyon and talus slopes in lower Sand Canyon seems to have been largely restricted to the Pueblo III period.

The Testing Program excavations can be used to refine our understanding of the shift in settlement from the mesa tops to the canyons. Test excavations in upper Sand Canyon took place at six mesa-top sites (G and G Hamlet, Lillian's Site, Roy's Ruin, Kenzie Dawn Hamlet, Shorlene's Site, and Troy's Tower) and four talus-slope sites (Lester's Site, Lookout House, Stanton's Site, and Catherine's Site). Sites tested in the lower Sand Canyon settlement cluster include one (Castle Rock Pueblo) constructed on the talus slope surrounding a prominent butte just above the McElmo floodplain and two sites (Mad Dog Tower and Saddlehorn Hamlet) located on terraces above the deeply incised inner gorge of Sand Canyon. Each of the tested sites has Pueblo III components that date between A.D. 1150 and 1200. Pueblo II components that date primarily to the A.D. 1025-1100 period are also present on the mesa-top sites (see Table 20.3, this volume). On the basis of the type of architectural remains and the amount of Pueblo II pottery found at G and G Hamlet and Kenzie Dawn Hamlet, it is believed that both sites were used as year-round habitations during the Pueblo II period. Lillian's Site also had abundant Pueblo II pottery and the remains of an architectural facility, but the Pueblo II use of this site appears to have been less intensive than that of G and G Hamlet and Kenzie Dawn Hamlet. The remainder of the mesa-top sites have even less Pueblo II pottery, suggesting limited use of these sites during this period. Therefore, the mesa-top sites in the upper Sand Canyon settlement cluster have the longest occupational histories of any of the tested sites.

Catherine's Site is the only talus-slope site to have a Pueblo II component. Catherine's Site is located on a "bench" that slopes much more gradually than the talus above and below. This setting is similar to that of the Green Lizard site (Huber 1993; Huber and Lipe 1992). The Pueblo II component at Catherine's Site is interpreted as a limited, seasonal use of the site, and like the mesa-top sites, this component appears to date to the late Pueblo II period. Green Lizard also has a Pueblo II occupation that probably represents limited, seasonal use of the site. Both of these sites are located adjacent to springs. Stone terraces and check dams that may be agricultural features have also been found near the sites. The use of these two sites during the Pueblo II period may therefore relate to the use of the springs and the agricultural features. The remainder of the upper Sand Canyon talus-slope sites are located at the base of the cliff, just below the canyon rim. These sites, like the lower Sand Canyon sites, have no substantial Pueblo II components.

The Pueblo III components at the tested sites date to the A.D. 1200s, with the possible exceptions of G and G Hamlet and Kenzie Dawn Hamlet, where occupation probably began in the late A.D. 1100s. The mesa-top unit pueblos were occupied and abandoned before the main period of occupation of the upper Sand Canyon talus-slope sites (see Chapter 20, this volume, for the details of these dating arguments). These mesa-top sites appear to have been abandoned as year-round habitations by about A.D. 1250. The only mesa-top site that clearly dates to after A.D. 1250 is Troy's Tower, where tree-ring dates for a roasting pit place the use of that feature in the A.D. 1270s.

The talus-slope and lower Sand Canyon sites, with the possible exception of Mad Dog Tower, were occupied in the middle to late A.D. 1200s. Tree-ring dates for Lester's Site, Lookout House, Saddlehorn Hamlet, and Castle Rock Pueblo indicate that each of these sites was primarily occupied after A.D. 1250 and that occupation of Lester's and Castle Rock continued into the A.D. 1270s. Pottery assemblages at the remainder of the canyon-oriented sites--Catherine's Site, Stanton's Site, and Mad Dog Tower--indicate that these sites were occupied later than the mesa-top sites, with occupations that probably postdate A.D. 1250.

In summary, the Sand Canyon locality data indicate that there was use of the canyons during the Pueblo II period, but this was largely restricted to seasonal use of the upper canyon. There may have been use of the McElmo floodplain in the lower canyon, but heavy alluviation during the Pueblo II period buried any sites that were present (Force and Howell 1997). Use of the canyons for year-round habitation became more common in the A.D. 1200s, and in upper Sand Canyon, this occurred primarily in the middle to late 1200s.

Settlement Patterns and Site Size

The construction of the large pueblos led early researchers in the Southwest to characterize Pueblo III as the "grand period" of "urban development," in contrast to Pueblo II, which was characterized as the "small house" period in the original Pecos classification (Martin 1927:181; Kidder 1927:489-491; Cordell 1984:56). Similarly, Roberts (1935:32) designated the Great Pueblo period in his classification scheme as the time when the large pueblos were constructed.

Regional Survey Data

Every major survey found that larger sites became more common during the Pueblo III period. A recent reconnaissance-level survey and literature review of southeastern Utah and southwestern Colorado documented over 100 sites with 50 or more rooms that date between A.D. 1150 and 1300 (Varien 1997; Varien et al. 1996). Regional surveys also document the presence of small sites that date to throughout the Pueblo III period. Thus, the trend is not just toward larger sites, but toward an increasing range of site size.

It is difficult to compare site size when survey data are used because some researchers report size in terms of the number of structures present, while others report it in terms of the number of households represented. The method for converting archaeological remains into an estimate of the number of households is not always consistent. Definition of the number of households represented depends in part on the functional interpretation of the kiva. Most authors interpret kivas as structures that integrate more than one household, and they infer that the primary residence of the household was located in the surface roomblock. Thus, a site with one kiva would have two or more households. Lekson, however, has argued that the kiva is the primary residential facility of the household (Lekson 1988). From this perspective, one kiva would represent one household. It is beyond the scope of this summary to specify the analytic criteria used in every study; however, in most of the studies summarized here, the researchers interpret the kiva as an integrative facility for two or three households.

Occasionally when discussing the largest aggregated sites, or clusters of smaller sites, authors equate the archaeological remains with a larger social group, the community. The organization of sites into communities is discussed in greater detail in the section on dispersed and aggregated settlement. In the discussion that follows, I have tried to preserve the language used by the individual authors to discuss site size. The reader should consult the individual reports for the definitions of these terms.

At Mesa Verde, Hayes (1964:99) and Rohn (1977:280) document a trend toward larger sites beginning during the early Pueblo III period. Neily argues that on Squaw Point and Cow Mesa this trend began during the late Pueblo II period, sometime between A.D. 1050 and 1150, but that it became especially noticeable in the Pueblo III period, after A.D. 1150 (Neily 1983:173, 190). By the late Pueblo III period, habitation sites varied widely in size; Neily (1983:209) documented as few as one and as many as seven kiva depressions in the Squaw Point-Cow Mesa area. In the Hovenweep area, Greubel (1991:130-132) documents the increasing frequency of what he calls "multiple dwelling sites" beginning in late Pueblo II and continuing through late Pueblo III. However, the number of pit structure/kiva depressions at habitation sites ranges from zero to six in each of the periods he examined, from late Pueblo II to late Pueblo III (Greubel 1991:Table 4-61). The canyon-head sites documented by Riley (1948) and Winter (1975, 1976, 1977) are the largest sites in the Hovenweep area, and here each site is often regarded as representing a community. This collective research on Hovenweep-area archaeology has been summarized by Thompson (1993).

Fetterman and Honeycutt (1987) argue that, on Mockingbird Mesa, increased household aggregation began during the late Pueblo II period and continued until the area was abandoned. Seventy-two percent of the recorded sites that date to the early Pueblo II period were interpreted as single-household sites; this drops to 35 percent by late Pueblo III (Fetterman and Honeycutt 1987:129). Habitation sites dating to the late Pueblo III period vary in size from those with only one kiva depression to one site with 10 kiva depressions (Fetterman and Honeycutt 1987:159). Seven Towers Ruin, located on the canyon rim and talus slope below the rim on Mockingbird Mesa, has an estimated 43 kivas and 150 to 200 rooms.

In survey areas on Cedar Mesa, the trend toward larger sites is not as pronounced. Single-household sites are by far the most common size site throughout the Pueblo occupation (A.D. 900-1300), and the largest sites recorded during the Cedar Mesa survey were occupied by no more than four households (Matson et al. 1988:252). On the Red Rock Plateau, sites are even smaller, rarely exceeding one kiva and its associated rooms (Lipe 1970).

Much larger sites that date to the Pueblo III period have been documented in southeastern Utah outside of the intensively surveyed areas on Cedar Mesa (Varien 1997; Varien et al. 1996). The same holds true for many large sites in southwestern Colorado, which have been individually recorded but which are not in areas that have been broadly surveyed (Varien 1997; Varien et al. 1996).

Sand Canyon Project Data

Adler interprets the beginning of the trend toward larger sites and greater diversity in site size as dating to between A.D. 1060 and 1150 in the upper Sand Canyon area; this trend continued in the subsequent periods and culminated during the late Pueblo III period with the construction of Goodman Point Pueblo and Sand Canyon Pueblo (Adler 1990a:264-267). Each of these sites has approximately 90 kivas, several hundred surface rooms, towers, a great kiva, and a multiple-walled structure, all surrounded by a site-enclosing wall (Adler 1990a; Bradley 1993). Just as the lower Sand Canyon survey area differs from the upper Sand Canyon survey area in terms of the timing of the move into the canyons, the lower survey area does not exhibit the variation in site size found in the upper survey area. In the lower survey area, the largest habitation sites recorded have two kiva depressions, except for Castle Rock Pueblo, which has 12 to 15 kivas (Adler and Metcalf 1991; Gleichman and Gleichman 1989, 1992).

The Testing Program was not designed to sample the range of site size documented by the surveys. Most of the sites investigated during the Testing Program are small, with either one or two kivas. These smaller sites were selected because survey crews evaluated them as the most likely to be single-component sites and because estimating the population at smaller sites was believed to be easier than at larger sites. With 12 to 15 kivas, Castle Rock Pueblo is the largest site examined during the Testing Program. Testing Program results do indicate that, although larger sites became more common from Pueblo II to Pueblo III times, small sites, including single-kiva habitations, were still occupied after A.D. 1250 and probably up to the time when the region was abandoned.

Dispersed and Aggregated Settlement Patterns

Related to increasing site size is the change from dispersed to aggregated settlement patterns. As larger sites became more common, the settlement pattern changed from dispersed to aggregated. The processes of aggregation, however, include more than just the construction of large sites, and dispersed and aggregated settlement patterns need to be viewed as parts of a continuum, not as a simple dichotomy of small vs. large sites. The data summarized below indicate that clustering of smaller sites represents a type of aggregation distinct from the construction of the large settlements like Goodman Point Pueblo and Sand Canyon Pueblo.

Examining the processes that result in aggregation often includes an attempt to identify communities (Neily 1983; Adler 1990a). Large aggregated sites are often equated with a single community, but defining communities in areas with a dispersed settlement pattern is more difficult. Many researchers have used the clustering of small sites to try to identify communities, but the boundaries of the settlement clusters, and the number of sites included in the clusters, vary considerably in each of the studies (Adler and Varien 1994; Lipe 1970; Rohn 1977; Eddy 1977). Sand Canyon Project survey data indicate that clustering occurs at a number of spatial scales. There are clusters of households within isolated sites, there are clusters of sites located close to one another, and there are even larger settlement clusters that encompass these smaller site clusters. The Sand Canyon survey data make it clear that identifying settlement clusters, and interpreting these clusters as communities, depends in part on the spatial scale of analysis.

In the discussion below, aggregation is examined both in terms of clusters of small sites in an otherwise dispersed settlement pattern and in terms of the construction of large, nucleated sites. Sand Canyon Project researchers have argued that processes of aggregation are best understood by making the community the social scale of the research (Adler 1990a; Adler and Varien 1994; Varien et al. 1996). These researchers use cross-cultural data, the distribution of habitation sites, and the distribution of public architecture to determine which level of settlement clustering represents community boundaries.

Regional Survey Data

Regional survey data illustrate that site distribution was never totally dispersed in the sense of equal spacing between all sites; at some geographic scale, the distribution of sites will always reveal clustering. During the Pueblo II and III periods, these clusters were made up of individual roomblocks with associated kiva depressions located close to other roomblock-kiva complexes. Over time, these clusters of sites in an otherwise dispersed settlement pattern changed in the following ways: (1) the size of the individual sites within the cluster became larger, (2) an increasing number of sites were included in the cluster, and (3) the individual sites became clustered more closely together. This type of clustering occurred earlier during the Pueblo I period. In describing this earlier aggregation, Orcutt (1987:617) uses the term "household aggregation" to refer to the formation of larger roomblock-pit structure complexes and the term "spatial aggregation" to refer to the clustering of roomblock-pit structure complexes closer together. The following discussion is concerned both with the development of site clusters within what might otherwise be characterized as a dispersed settlement pattern and with the subsequent construction of large sites, which are more typically equated with an aggregated settlement pattern.

Hayes argues that the clustering began during the McElmo phase on Wetherill Mesa (Hayes 1964:109). Rohn identifies clusters of numerous small sites on Chapin Mesa beginning in the Pueblo II period, and he was one of the first to interpret these clusters as communities. Following Linton (1936), the basis for this inference was the spatial proximity of the sites, which would have brought the occupants of the sites into regular face-to-face contact (Rohn 1977:266). Rohn also argues that the early clusters are composed of small sites and that individual sites within the clusters do not get larger until the Pueblo III period (Rohn 1977:237-246). The occurrence of large sites consisting of more than five kivas and 50 rooms appears to date to the A.D. 1150-1300 period; this development includes sites on the mesa top (for example, Far View Ruin), on the talus slopes (for example, Site 34), and in the alcoves (for example, Cliff Palace).

Neily argues that slight clustering accompanies the increased variation in site size that begins in the A.D. 1050-1150 period, but distinct clustering of sites is not apparent in the subsequent A.D. 1150-1230 period (Neily 1983:172-206). Neily reports a great reduction in the number of occupied sites and an increase in the sizes of the sites that continued to be occupied in the final period of occupation in the Squaw Point and Cow Mesa survey areas. Neily's work follows the traditional interpretation of settlement pattern changes from dispersed to aggregated, with numerous smaller sites rapidly replaced by a few big sites.

Greubel does not specifically discuss the clustering of sites relative to one another in the Hovenweep area, although examination of his site-distribution figures shows that, beginning in the late Pueblo II period, habitation sites apparently started to cluster near other habitation sites (Greubel 1991:Figures 4-14, 4-16, 4-17). The large canyon-head complexes found in Hovenweep National Monument also represent the aggregation of households, and this took place largely in the A.D. 1200s (Winter 1976:287-289).

Fetterman and Honeycutt (1987) specifically address the question of clustering in their Mockingbird Mesa survey by doing a cluster analysis and conducting nearest-neighbor analysis. The cluster analysis identified groups of sites that were distinctly clustered, but these were interpreted as potentially meaningful only if they contained more than five households. Site clusters were identified for all periods, but in all periods there were habitation sites outside of the site cluster boundaries. This indicates that clustering was present above the level of a single site or household throughout the occupation of Mockingbird Mesa, if the sites within the clusters were contemporaneously, not sequentially, occupied. Fetterman and Honeycutt found that the mean distance between sites when all sites were considered was least during the early Pueblo II period and greatest during the late Pueblo II through late Pueblo III periods. In addition, the results of the nearest-neighbor analysis indicate that settlement during the Basketmaker III, Pueblo I, and early Pueblo II periods was slightly clustered, whereas settlement during the early Pueblo I, late Pueblo II, and early and late Pueblo III tended to be more evenly spaced (Fetterman and Honeycutt 1987:129).

In southeastern Utah, the settlement pattern remained dispersed throughout the Pueblo occupation both in the Cedar Mesa and Red Rock Plateau survey areas (Matson et al. 1988; Lipe 1970). There was, however, a slight tendency toward larger sites during the latest period on Cedar Mesa. Outside of these survey areas, the presence of large sites dating to the Pueblo III period indicates that increased household aggregation did occur there during this period (Varien et al. 1996). For the most part, these large sites occur east of Cedar Mesa and Comb Ridge. It appears that the pattern in the area east of Comb Ridge more closely resembles the patterns documented for the core of the Mesa Verde region in southwestern Colorado. In the area west of Comb Ridge, these patterns are present but much less pronounced.

Sand Canyon Project Data

Despite the fact that the upper Sand Canyon survey documents wide dispersal of sites in the A.D. 930-980 (early Pueblo II) period, the beginning of one site cluster may also date to this period (Adler 1990a:247). This pattern continued in the subsequent middle and late Pueblo II periods, when additional loose clusters of sites appeared (Adler 1990a:250-256).

It was during the A.D. 1150-1300 period when the first multiple-roomblock sites occurred. Adler uses this term to refer to "sites which contain two to four roomblocks, usually placed one after another, 10-30 meters apart on ridge crests" (Adler 1990a:267). In a sense, the definition of site boundaries has changed. Before A.D. 1150, every site was a separate roomblock-kiva unit. After A.D. 1150, some of these roomblock-kiva units were so close together that they are considered single multiple-roomblock sites.

The multiple-roomblock site represents an important phase of aggregation--part of the continuum often missed when aggregation is viewed simply as a shift from small to large sites. The average distance between nearest neighbors decreased during the Pueblo III period compared with the previous periods, and this decrease would have been even greater had the individual roomblocks in the multiple-roomblock habitation sites been recorded as separate sites (Adler 1990a:269). Adler interprets these multiple-roomblock sites as having been occupied just before the occupation of the large sites of Sand Canyon Pueblo and Goodman Point Pueblo (Adler 1990a). The multiple-roomblock sites may represent the architectural, and possibly the social, equivalent of the spatially discrete roomblock groups at Sand Canyon and Goodman Point Pueblos (Adler 1994).

Within 1 km north of Goodman Point Pueblo, there are at least 25 habitation sites, some as large as 30 rooms and totaling an estimated 500 rooms. This group of 25 sites includes several of the multiple-roomblock sites described above. Although tree-ring dates would be needed to fully resolve the chronological relationships between these sites, the types of pottery observed indicate that the sites were occupied contemporaneously in the late A.D. 1100 to early A.D. 1200 period, probably just before the construction of Goodman Point Pueblo. As a group, these 25 sites have approximately the same number of rooms and kivas as Goodman Point Pueblo.

As noted above, Adler argues that these multiple-roomblock sites are the forerunners of the spatially discrete roomblocks seen at Sand Canyon and Goodman Point Pueblos (Adler 1994). At these large pueblos, the roomblocks are much more closely clustered than at the multiple-roomblock sites. The majority of the roomblocks at Sand Canyon Pueblo and Goodman Point Pueblo are surrounded by an enclosing wall. The construction of large sites like Sand Canyon and Goodman Point is often equated with the beginning of aggregated settlement, but Adler has clearly shown that aggregation was taking place before the construction of these large pueblos.

Multiple-roomblock sites are not present in the lower Sand Canyon survey area. The lower Sand Canyon sites are more evenly spaced, and there is greater distance between sites (Adler and Metcalf 1991; Gleichman and Gleichman 1989, 1992). The exception to this is Castle Rock Pueblo, where 10 to 20 households aggregated into a single site. Although Castle Rock Pueblo is much smaller than either Sand Canyon Pueblo or Goodman Point Pueblo, it is much larger than the surrounding sites in lower Sand Canyon.

The Testing Program did not include excavation at any of the multiple-roomblock sites identified during survey. Stanton's Site and Catherine's Site are the two tested sites closest to each other. Tree-ring dating for these sites is poor or absent, but large pottery assemblages from a variety of subsurface contexts indicate that these two sites were largely contemporaneous, and not sequential occupations (see Chapter 20).

Survey indicates that multiple-roomblock sites became more common during the Pueblo II and III periods. But the majority of the tested sites are not tightly clustered or parts of multiple-roomblock sites. As with site size, there is variability in terms of whether or not individual roomblock-kiva units are parts of larger multiple-roomblock clusters.

Sand Canyon Project survey and testing has therefore demonstrated that settlement clustering occurred at several spatial scales, that there is variability in the size of these clusters, and that the clustering became more pronounced through time. At the smallest spatial scale, there was an increase in the size of individual roomblock-kiva units, indicating that more than one household was present at a distinct site. This is household aggregation in Orcutt's (1987) terminology. At a larger spatial scale, clustering occurred when these roomblock-kiva units were located close to one another, forming what Adler calls multiple-roomblock sites. This is spatial aggregation in Orcutt's terminology.

There are settlement clusters in the upper Sand Canyon survey area that are more inclusive than the multiple-roomblock sites described above. One cluster is found in the western portion of the survey area, and the other is in the eastern portion (Adler and Varien 1994). Site density is lower in the area between these two clusters. The western cluster is located around the head of Sand Canyon, the location where Sand Canyon Pueblo was eventually constructed; the eastern cluster lies around the head of Goodman Canyon, the eventual location of Goodman Point Pueblo. These two settlement clusters began in the Pueblo II period and continued throughout the Pueblo III period, and they are referred to as the Sand Canyon settlement cluster and the Goodman Point settlement cluster (Adler and Varien 1994). An ancient road that probably dates to sometime in the Pueblo II-III period connects these two first-order communities, providing physical evidence for the interaction between them. Sand Canyon Pueblo and Goodman Point Pueblo were eventually constructed within the area encircled by each of these settlement clusters, and these two large pueblos represent the most extreme examples of both household and spatial aggregation in the Sand Canyon locality and the Mesa Verde region.

This discussion illustrates how the spatial scale of the analysis influences our understanding of the processes of aggregation and the definition of communities. Sand Canyon Project survey and excavation data indicate that clusters are present at several spatial scales, and these represent a continuum of aggregation rather than a simple dichotomy of small- vs. large-site settlement pattern. In terms of social organization, the question is, which cluster represents a community? Rohn's (1977) Chapin Mesa communities were defined on the basis of the clustering of small sites, and the communities he identifies range in size from six to 30 habitation facilities. These are similar in size to some of the multiple-roomblock sites in the Sand Canyon survey area, but they are considerably smaller than the Sand Canyon and Goodman Point settlement clusters (Adler and Varien 1994).

Adler used cross-cultural research to examine community function and to resolve which level of settlement cluster might represent the community. He demonstrated that there is an upper limit on population levels in nonstratified communities, that these communities typically have a public structure used by the entire community, and that this public structure has the largest floor area of any facility in the community (Adler 1989, 1990a; Adler and Wilshusen 1990; Adler and Varien 1994). Adler and Varien (1994) used the population limits and the patterns of use in public architecture established by the cross-cultural research to help identify communities in the Sand Canyon locality. They combined these data with the evidence for settlement clustering in the upper Sand Canyon survey to argue that the Sand Canyon and Goodman Point settlement clusters represent two first-order communities (Adler and Varien 1994).

These are much larger than the communities identified by other Mesa Verde-region researchers; however, it is these settlement clusters that have structures interpreted as community-level public architecture. This public architecture includes a Pueblo II great kiva found in each settlement cluster and a Pueblo II great house in the Sand Canyon settlement cluster. In the subsequent Pueblo III period, public architecture is found at Sand Canyon Pueblo and Goodman Point Pueblo. Each site contains a great kiva and a bi-wall structure. The results of the Site Testing Program indicate that some of the small sites within the Sand Canyon settlement cluster were at least partly contemporaneous with Sand Canyon Pueblo. The degree to which these small sites were integrated into the larger Sand Canyon community is discussed in Chapter 22. In lower Sand Canyon, a similar community may have been focused around Castle Rock Pueblo.

Summary and Conclusions: Settlement Pattern Data

Changes in settlement patterns discussed in this section include the shift in site location from the mesa tops to the canyon rims and talus slopes, changes in site size, and the change from dispersed to aggregated settlement patterns. The results of large survey projects in southwestern Colorado and southeastern Utah were used to examine possible region-wide patterns and the variability found in those patterns. The results of the Sand Canyon Project survey and testing programs were then compared with the regional survey data.

In terms of site location, regional survey results indicate the following: (1) mesa tops were occupied during both the Pueblo II and Pueblo III periods; (2) occupation of the canyon rims and canyons increased through time; (3) the timing of this shift from the mesa tops to the canyons varied in the different areas surveyed; and (4) the intensity of the use of the canyons also varied in the different areas surveyed. When survey projects included canyons in their coverage, they documented a wide range of site types dating to both the Pueblo II and III periods. Survey results from Cedar Mesa, for example, indicate that the Pueblo II occupation of the canyons consisted primarily of limited, seasonal use, but that construction of habitations in the alcoves began in the early A.D. 1100s (Haase 1983). All surveys indicated that the shift to increased use of canyon rims and canyons for habitation sites occurred primarily during the Pueblo III period.

The Sand Canyon Project provides one of the most detailed studies of the timing of the shift in settlement location. Survey and testing data indicate that the canyon environments were occupied during both the Pueblo II and Pueblo III periods in the upper Sand Canyon area. In the lower Sand Canyon area, occupation of the canyon environments seems to have been largely restricted to the Pueblo III period. Upper Sand Canyon Pueblo II occupation of the canyons appears to have been predominantly seasonal. The Testing Program excavations demonstrate that the shift in the location of year-round habitation sites from the mesa tops to the canyons took place late in the Pueblo III period, probably in the middle A.D. 1200s. Eddy et al. (1984:43) recognized this settlement shift and suggested that this last phase of occupation in the Mesa Verde region be called the Hovenweep phase, which they would date to A.D. 1225-1300. These authors further suggested that population decline began about A.D. 1150 and that the Hovenweep locality and lower McElmo canyon were the last bastions of Puebloan settlement within the McElmo drainage unit. Sand Canyon Project data support the idea of a late Pueblo III phase characterized by a move to the canyon rims and talus slopes, but these data suggest that the shift may have been even more restricted in time to the A.D. 1240-1290 period. On Mesa Verde, the occupation of the mesa tops apparently continued until the middle 1200s, but building in the alcoves began in the late A.D. 1100s and early A.D. 1200s and continued through the A.D. 1270s. Therefore, the move to the canyons appears to have been relatively sudden and relatively late in the Sand Canyon locality, but more gradual and somewhat earlier on Mesa Verde.

Evidence of this shift to the canyons is seen throughout the region, and the interpretation that the Hovenweep phase was limited to a remnant population in two small areas seems incorrect. Rather, this review of the regional survey literature suggests that population levels remained high in most localities into the middle A.D. 1200s. An important question for future research is whether the beginning of the shift to the canyons was also the beginning of the depopulation of the Mesa Verde region, or whether this depopulation occurred in the A.D. 1260s and 1270s, after the settlement shift to the canyons.

The settlement shift from the mesa tops to the canyons was spatially widespread and relatively synchronous, occurring within a 50-year period. This suggests that the causes of this population movement were similar in the different localities across the region. Archaeologists, however, have offered a wide range of explanations for why this settlement shift occurred.

Lister argued that intensive use of the mesa tops for centuries resulted in the depletion of mesa-top forests and erosion of mesa-top soils, which in turn led to the shift in settlement to the canyons (Lister 1966:71). Similarly, Stiger (1979) argued that resource depletion stimulated population movement on Mesa Verde. Fetterman and Honeycutt, who argue that the shift from mesa tops to canyon rims was gradual, interpret the change in site location as the result of the increased use of masonry architecture and the need to move closer to building-stone resources (Fetterman and Honeycutt 1987:123-125).

Working in the Hovenweep area, Winter argued that the shift from mesa top to canyon rim took place largely during the A.D. 1200s and that the move was a result of changing farming practices. Canyon-rim and talus-slope locations were adopted when mesa-top dry-farming fields were abandoned, possibly due to drought, in favor of spring- and flood-watered canyon-head, talus-slope, and canyon-bottom floodplain farms (Winter 1976:287-290). The results of Greubel's survey tend to support these findings, although Greubel (1991:101) documented a longer, more gradual shift in site location that gained momentum after A.D. 1200.

Changing agricultural strategies do not seem to explain changes in settlement pattern on Cedar Mesa; here, many of the late sites are located in canyons with little or no arable soil. Instead, the pull toward the canyons in the late Pueblo III period on Cedar Mesa is explained by a number of factors, including the following: access to reliable domestic water; access to locations where sheltered structures--especially storage facilities--could be built; and the ability to secure the storage facilities, as evidenced by the defensive arrangements observed at many of the sites (Matson et al. 1988:255). Rohn also argued that the shift from mesa tops to cliff dwellings was the result of a desire to occupy defensible locations (Rohn 1977:292).

Explanations of the shift in settlement location that stress resource procurement may be undermined by a simple observation: the distance of the move from the mesa tops to the canyon rims is typically not very far. In the upper Sand Canyon survey area, most mesa-top sites are within 1 or 2 km of the canyon rim and talus slopes. The distances required to transport materials or to walk to fields between the canyons and the mesa tops would not be prohibitive. Depletion of mesa-top resources is possible, but recent studies suggest that this, too, is unlikely. Hovezak (1992) studied the depletion of timber as a result of its use in construction in the Sand Canyon locality. Although the results of his study indicate that construction required large amounts of wood, complete deforestation was seen as unlikely. Further, it seems unlikely that the shift was entirely the result of increasing dependence on masonry. As noted above, the shift to masonry architecture on the mesa tops occurred long before the shift to canyon settlements.

A change in farming practices from mesa-top dry farming to flood- and spring-watered canyon farming may have been a factor in the change in settlement location. Use of the canyons for runoff agriculture appears to have begun at least as early as the Pueblo II period, and an agricultural strategy that relied on many different techniques may have been in place at this time. It is unknown to what degree agriculture in the canyons intensified during the Pueblo III period. Intensification could have resulted in larger areas being cultivated and greater reliance on the production from irrigated canyon fields when compared with the mesa-top dryland fields.

A microeconomic argument links intensification of canyon agriculture and a shift in settlement patterns: as people devoted more energy to runoff and spring irrigation they would have sought to minimize transportation costs to and from the fields. This would be especially true if people had been pot-irrigating from springs--in which case, it would have been essential to have people, pots, water, and fields closer to one another. It may also have been critical for people to live near the fields so that when it rained they could take full advantage of the runoff for watering fields. Finally, intensification of runoff agriculture on the talus slopes might have required greater labor organization, which would have promoted both aggregation and the shift in settlement location. These are all empirical questions which require further research to resolve.

The final explanation for the shift in settlement location is that it was undertaken for reasons of security, that is, the desire to move to more-defensible locations. Evidence of violence was found at Castle Rock Pueblo, Troy's Tower, and Lookout House. Site-enclosing walls that may have served a defensive purpose are also present at most of the canyon rim sites (Kenzle 1997). The fact that many of the late sites enclose a spring, thus ensuring access to a reliable water source, may reflect a concern over conflict. And aggregation, which provides safety in numbers, may have been stimulated by a social environment in which conflict, or the threat of conflict, was heightened. Like the explanations that stress agricultural intensification, this model deserves more careful attention and additional research.

In terms of site size, most regional survey data document that larger sites became more common through time during the Pueblo II and III periods (the one exception is Cedar Mesa, where this trend is discernible, but not pronounced). The trend toward larger sites during the Pueblo III period did not result in the development of large sites all of the same size. The lower Sand Canyon survey area is more like Cedar Mesa in southeastern Utah. Settlement remains dispersed, and the sites are uniformly small. The single exception to this in the lower Sand Canyon survey area is Castle Rock Pueblo, which represents a medium-size aggregated site. Numerous large sites are present in southeastern Utah, but most are located east of Comb Ridge. Sand Canyon Project testing data indicate that sites with as few as one and as many as 15 kivas were occupied in the period just before abandonment. These are in addition to the extremely large sites of Sand Canyon Pueblo and Goodman Point Ruin, each with more than 90 kivas and several hundred rooms.

The shift from smaller to larger sites is one aspect of the change from dispersed to aggregated settlement patterns. Another dimension of this change is the formation of site clusters within a settlement pattern that might be traditionally referred to as dispersed. This clustering was not specifically measured by most surveys, and when it was measured, the results varied. Certain surveys, as on Cedar Mesa, documented dispersed settlement throughout the Pueblo II and III occupations; however, the quadrat-based survey that was used in this area was not well suited to the documentation of site clusters. In other areas, such as Mockingbird Mesa, researchers have identified clusters that existed during both the Pueblo II and III periods, but the clustering does not appear to have become more developed through time. In still other areas, such as Chapin Mesa on Mesa Verde, clusters are identified for the Pueblo II period, but these clusters became more numerous, larger, and more compact through the Pueblo II and III periods.

As noted above, settlement clustering occurred in the Sand Canyon locality at a variety of spatial scales. Adler (1990a, 1992b) examined cross-cultural data and found a relationship between the size of the coresidential group and the intensity of subsistence-related production. This relationship is labeled the "curvilinear hypothesis." Small, independent households (which may equate to sites with a small roomblock and a single kiva) are associated with both extensive and highly intensified subsistence economies. Larger coresidential groups, represented by multiple-roomblock clusters, are associated with moderate levels of intensification. Adler argues that it is this gradual development of moderate levels of subsistence intensification--which is associated with population growth, increasing scarcity of resources, and the development of more formalized land-tenure systems--that stimulated the aggregation of households into larger coresidential groups in the Sand Canyon locality.

Testing in upper Sand Canyon demonstrated that two sites located close to each other--Catherine's and Stanton's--were probably occupied at the same time. Thus, site clusters identified during survey may indeed represent larger groups of people living close to one another, rather than sequentially occupied sites. During testing we also examined sites that did not appear to be part of larger, well-defined clusters. These sites were occupied throughout the A.D. 1200s, indicating that there was an increase in variability in the size and composition of the site clusters.

This variability is illustrated in the lower Sand Canyon survey area, which again differs from the upper Sand Canyon survey area. Clusters of sites are not apparent. The only site where it appears that large numbers of households were spatially aggregated is Castle Rock Pueblo. In terms of site location, site size, and the clustering of individual sites into larger groups, the patterns seen in the lower Sand Canyon area differ from those documented for the upper Sand Canyon area and for most of the region. It seems likely that the processes that produced the upper Sand Canyon settlement system differed from those that produced the lower Sand Canyon settlement system.


In this section, I briefly review the archaeological evidence for the timing and duration of the abandonment of the Mesa Verde region. First, I examine data collected throughout the region, then I evaluate those data in light of the Sand Canyon locality results. This discussion provides the foundation for the interpretive summary of abandonment that appears in Chapter 22.

Regional Data

The abandonment of the Mesa Verde region remains one of the central questions of Southwest archaeology. This discussion summarizes the dating of the abandonment of southwestern Colorado and southeastern Utah.

Archaeologists do not agree on when the depopulation of the Mesa Verde area began. Smith (1987:67) argues that population decline at Mesa Verde may have begun at the end of the Mancos phase (A.D. 1075) and continued gradually through the subsequent McElmo and Mesa Verde phases. Hayes also argues that depopulation of Wetherill Mesa began during the Mancos phase, despite the fact that the number of rooms is greatest in the later Mesa Verde phase (Hayes 1964:109). Hayes believes that, because earlier occupations are covered by later occupations, the size of the earlier occupations has been significantly underestimated. Thus, although he documents a large population still present during the last period of occupation, he believes that a gradual decline in population had been taking place for 300 years before the final abandonment. Eddy et al. (1984:43) argue that the abandonment of the McElmo drainage began by A.D. 1150. In a regional overview, Lipe (1983) also argued that the population decline began in the A.D. 1100s, although he has since revised this view (Varien et al. 1996).

Rohn, on the other hand, has argued that population in the northern San Juan area grew continuously from the Basketmaker III to Pueblo III periods and that it peaked just before abandonment (Rohn 1977:267, 1989:166). This position is supported by survey data from Mockingbird Mesa (Fetterman and Honeycutt 1987:122) and from the Hovenweep area, if data from both the large sites and the areas surrounding the large sites are considered (Winter 1975, 1976, 1977; Greubel 1991). The same is probably true for the Cow Mesa and Squaw Point survey areas, although Neily does not specifically address the question of population (Neily 1983). It is clear from Neily's work that population grew at least into the early Pueblo III period (Neily 1983:184). On Cedar Mesa in southeastern Utah, the population levels were stable throughout the Pueblo III period until abandonment (Matson et al. 1988).

There is a clear trend toward fewer but larger sites in the period just before abandonment, but this is largely the result of the aggregation that occurred at that time (Hayes 1964:110; Rohn 1977:287; Neily 1983:228; Fetterman and Honeycutt 1987:129; Winter 1975, 1976, 1977; Greubel 1991:131). If only sites are counted, there was an apparent decrease in population, but if rooms are counted, population appears to have increased through the late Pueblo III period. Hayes's concern that the sizes of earlier populations are underrepresented because evidence of those occupations is covered by the remains of later occupations still needs to be addressed by studies of regional population, as does the question of how many rooms were in use at a given time and for how long.

The final phase of occupation has been called the Hovenweep phase, which dates to between A.D. 1225 and 1300 (Eddy et al. 1984:43). Eddy and his colleagues distinguish the Hovenweep phase, which is characterized by canyon-head pueblos, from the earlier Yellow Jacket phase, which is characterized by large mesa-top communities and is dated to A.D. 1050-1250 (Eddy et al. 1984:43). These researchers further propose that "the cliffs and canyons of lower McElmo canyon and the Hovenweep locality apparently were the last bastion of Anasazi settlement within the unit with the exception of the Mud Springs complex" (Eddy et al. 1984:40). A recent review of Mesa Verde-region settlement supports the distinction between the Yellow Jacket and Hovenweep phases, but it demonstrates that late canyon-head and canyon-rim pueblos have a widespread distribution in southwestern Colorado and in southeastern Utah, primarily east of Comb Ridge (Varien et al.1996).

Tree-ring dates in the A.D. 1260s and 1270s are abundant in the northern Mesa Verde region, with eight sites in southeastern Utah and 21 sites in southwestern Colorado providing tree-ring dates in these decades (Robinson and Cameron 1991). The latest date for southeastern Utah is a noncutting date of A.D. 1268 from Moon House (Bloomer 1989). The latest dates from southwestern Colorado are a single A.D. 1288 cutting date and many 1280 and 1281 dates from Square Tower House on Mesa Verde. There are also several A.D. 1279 cutting dates from cliff dwellings on Mesa Verde, including Cliff Palace (Robinson and Harrill 1974), Balcony House (Fairchild-Parks and Dean 1993), and Long House (an A.D. 1280 cutting date is reported for Long House in Cattanach [1980:410], but Robinson and Cameron [1991] and Nichols and Harlan [1967:15-17] report A.D. 1279 as the latest date for Long House). The latest dates outside Mesa Verde National Park are an A.D. 1277 cutting date from Hovenweep Castle and a noncutting date of A.D. 1277 from Sand Canyon Pueblo (Winter 1976:288; Kleidon and Bradley 1989; Robinson and Cameron 1991).

Sand Canyon Project Data

An important question is whether abandonment of the Mesa Verde region began at the same time that settlement shifted from the mesa tops to the canyon rims, or whether population levels were sustained into the A.D. 1270s. Sand Canyon Project survey data support the interpretation that population grew from the Pueblo II through the Pueblo III periods and that the highest population levels were reached in the period just before abandonment (Adler 1990a:272). Thus, in the Sand Canyon locality, the population increase appears to have been sustained beyond the period when settlement locations shifted from the mesa tops to the canyons (Adler 1990a:314). Relatively steady growth characterizes the upper Sand Canyon area (Adler 1990a:314), whereas the lower Sand Canyon survey area is characterized by rapid population increases during the Pueblo III period (Gleichman and Gleichman 1989:36; Adler and Metcalf 1991:40).

By the end of the seventh season of fieldwork, Sand Canyon Pueblo had produced a total of 358 tree-ring dates, with 181 cutting dates (Bradley 1993). Cutting dates for the site peak in the A.D. 1240s (N = 40), are still numerous in the A.D. 1250s (N = 25) and 1260s (N = 30), and fall off in number in the A.D. 1270s (N = 5). Five structures have well-documented construction dates: A.D. 1252 for Structure 501, A.D. 1260 and 1262 for Structures 1205 and 1206, A.D. 1265 for Structure 1004, and A.D. 1274 for Structure 102 (Bradley 1993). A.D. 1274 is the latest cutting date, and a noncutting date of A.D. 1277 is the latest date from the site (Kleidon and Bradley 1989).

Tree-ring dates for the individual tested sites were summarized in Chapter 20. All tree-ring dates for each of the tested sites are presented as a stem-and-leaf plot in Figure 21.4; cutting dates only are presented in Figure 21.5 ("v" dates are included as cutting dates). An explanation of how to read stem-and-leaf plots is provided in Chapter 20.

The stem-and-leaf plots presented in Figure 21.4 and Figure 21.5 are intended to identify periods of construction within the Sand Canyon locality; they also indicate when construction within the locality, but outside of Sand Canyon Pueblo, ceased. Several studies of wood use indicate that reuse of salvaged beams was substantial (Ahlstrom 1985b:629; Ferguson and Mills 1988; Ahlstrom et al. 1991). By comparing the tree-ring dates for the tested sites with those for Sand Canyon Pueblo, we can look at beam reuse at the locality and community levels.

The assemblage of dates from the tested sites cannot be said to be representative of all the sites in the Sand Canyon locality; in fact, we know that the assemblage is biased toward late (Mesa Verde phase) sites because this was our focus when selecting sites for testing. Nonetheless, the dates can be used to evaluate the timing of construction events and--because site selection was biased toward the latest sites--the timing of the abandonment of the locality.

Figure 21.4 illustrates a "peak" in the distribution of dates in the A.D. 1000s, with the largest number of dates in the A.D. 1060s. This peak is largely the result of the Pueblo II occupations at G and G Hamlet and Kenzie Dawn Hamlet. Another peak occurs in the early A.D. 1200s, with the decade beginning at A.D. 1200 having the greatest number of dates of any decade. It was argued in Chapter 20 that the mesa-top unit pueblos, with the possible exception of G and G Hamlet, date to the early A.D. 1200s. These mesa-top unit pueblos account for many of the dates in the late A.D. 1100s and early A.D. 1200s. Although the number of dates decreases in the later decades of the A.D. 1200s, there are still numerous dates from each decade until the A.D. 1270s. The post-A.D. 1223 dates (N = 102) are for samples collected from canyon-oriented sites, except for nine late dates from Troy's Tower. The late dates come from Lester's Site (N = 18) and Lookout House (N = 2), in the upper Sand Canyon settlement cluster, and Saddlehorn Hamlet (N = 22) and Castle Rock Pueblo (N = 51), in the lower Sand Canyon settlement cluster. Late construction outside of Sand Canyon Pueblo is therefore interpreted as having taken place primarily on the talus slopes and in lower Sand Canyon.

The same pattern, although less robust, is apparent when only the cutting dates are examined (Figure 21.5). Cutting dates in the last half of the A.D. 1000s reflect the Pueblo II occupations at G and G Hamlet and Kenzie Dawn Hamlet. There is a gap in the dates during the early A.D. 1100s, but two dates from the decades of the middle and late A.D. 1100s suggest that there was construction within the locality during these decades. As in Figure 21.4, Figure 21.5 shows a peak in the A.D. 1210s and 1230s, reflecting construction in these decades. There is only one date in the A.D. 1240s, the decade with the most dates for Sand Canyon Pueblo, which may indicate that the first construction at Sand Canyon Pueblo took place in that decade. The cutting dates from the A.D. 1250s to 1270s are for Troy's Tower, Lester's Site, Lookout House, Saddlehorn Hamlet, and Castle Rock Pueblo.

With the exception of the A.D. 1240s, the distribution of dates for the tested sites resembles the distribution of dates for Sand Canyon Pueblo. These dates indicate that construction was taking place outside this large site up to the time of locality abandonment, which may have begun in the late A.D. 1270s and early A.D. 1280s. This proposition is examined in greater detail in the final chapter of this report.