Early Warning Signals of Social Transformation: A Case Study from the US Southwest

Recent research in ecology suggests that generic indicators, referred to as early warning signals (EWS), may occur before significant transformations, both critical and non-critical, in complex systems. Up to this point, research on EWS has largely focused on simple models and controlled experiments in ecology and climate science. When humans are considered in these arenas they are invariably seen as external sources of disturbance or management. In this article we explore ways to include societal components of socio-ecological systems directly in EWS analysis. Given the growing archaeological literature on ‘collapses,’ or transformations, in social systems, we investigate whether any early warning signals are apparent in the archaeological records of the build-up to two contemporaneous cases of social transformation in the prehistoric US Southwest, Mesa Verde and Zuni. The social transformations in these two cases differ in scope and severity, thus allowing us to explore the contexts under which warning signals may (or may not) emerge. In both cases our results show increasing variance in settlement size before the transformation, but increasing variance in social institutions only before the critical transformation in Mesa Verde. In the Zuni case, social institutions appear to have managed the process of significant social change. We conclude that variance is of broad relevance in anticipating social change, and the capacity of social institutions to mitigate transformation is critical to consider in EWS research on socio-ecological systems.


Introduction
It is difficult to predict when complex systems are approaching thresholds where a small change (forcing) can result in a significant transformation. Recent research in ecology suggests that generic indicators, referred to as early warning signals, may occur before such transformations. Although Scheffer and co-authors [1] have suggested that the social sciences are an important arena in which to expand early warning signals analysis, such analyses are rarely undertaken. Indeed, within both modelling and empirical studies of socio-ecological systems (SES) humans are invariably seen as external sources of disturbance or management. Our goal is to explore ways to include societal components of SES within the systems under investigation. Given the growing archaeological literature on 'collapses' in social systems (e.g., [2,3,4,5]), and our long-term participation in collaborative and comparative research concerning social transformation in the US Southwest and the North Atlantic [6], we undertook a series of analyses to determine whether early warning signals are apparent in the archaeological records of the build-up to two contemporaneous cases of social transformation in the pre-Hispanic US Southwest. As numerous researchers have noted (e.g., [7] and references therein), the well-researched and highly detailed paleoclimatic and archaeological records of the US Southwest make it a particularly appropriate region in which to investigate SES and, by implication, to search for possible early warning signals in cultural data. Our two case studies, Mesa Verde and Zuni (Fig 1), differ significantly in the nature and degree of social transformation, thus allowing us to explore the contexts under which warning signals may (or may not) emerge.
Transformation in socio-ecologicalsystems involves reorganization to a distinctly different state. Thus, social transformations entail processes such as notable shifts in the nature and strengths of social relationships, changing community location and structure, institutional change, and changes in identity. Transformations may be either non-critical (reorganization with some continuity) due to resilience within the system [8] or critical (catastrophic), reflecting a regime change to an alternative stable state [9]. Early warning signals reflect the increasing sensitivity of a complex system to external or internal perturbation, and they may occur before both critical [1,10,11] and non-critical transformations [12]. In this paper we present analyses that examine the build up to reorganization using data from archaeological research on a non-critical transformation in the Zuni area and a critical transformation in Mesa Verde. We selected these two regions because both form the core of major population concentrations in the late pre-Hispanic period. Each concentration was characterized by strong similarities in material culture and economy indicating frequent interactions at the regional scale. Both cases contain high resolution spatial data on thousands of archaeological sites in closely spaced time steps that can be explored for potential harbingers of social transformation.
In the following we briefly discuss the indicators related to the phenomenon of critical slowing down [13] that have been highlighted as signs that a system is close to transformation, and identify increasing variance as the metric most appropriate for our purposes. We then introduce the case studies and discuss our hypotheses related to critical slowing down, our methods of acquiring the variance data, and our results and interpretations. We conclude the paper with a discussion of the strengths of using the "completed experiments" of the past to identify ways that change might be anticipated. We also advocate the explicit inclusion of human dimensions in the study of early warning signals, in what has up to now been largely an ecological examination of socio-ecological systems.
Funding: Funded by National Science Foundation (http://www.nsf.org/) BCS-9043134. MAP is co-PI on this grant with Keith Kintigh. The Zuni settlement data used for the analyses in this paper were acquired in part through this funding. The Wenner-Gren Foundation (http://www.wennergren. org), 09094295. MAP is the PI on this grant. The Zuni settlement data used for the analyses in this paper were acquired in part through this funding. Research funded by the National Science Foundation (http://www.nsf.org/) BCS-0119981 to Timothy Kohler (lead PI) in part acquired the Mesa Verde settlement data used in the analyses in this manuscript. The specific analyses for and preparation of this manuscript were not supported by external funding.
consequences of critical slowing down that are amenable to data collection and analysis. These consequences include increasing autocorrelation and increasing variance and skewness in state variables [9,[16][17][18][19]. Temporal autocorrelation increases in the fluctuations between states as recovery slows because the system remains longer in its perturbed state, thus rates of change decrease. Variance is inversely proportional to the return rate of the system and reflects the amplification of small shocks as a dynamic system approaches a critical transition [16,20].
The settlement data used here include locations and size information for over 2,500 individual habitation structures from over 90,000 acres of full-coverage archaeological survey. These data have a relatively limited number of periods in the time series and contain distributions of values rather than a single metric for our state variables. Therefore we use the coefficient of variation (CoV) to measure variance in system state in order to evaluate whether or not critical slowing down occurred in our case studies prior to social transformation. We did not use other commonly employed analyses of early warning signals in ecological systems because they were either less useful for our purposes or they were inappropriate. For example, autocorrelation is assessed in terms of recovery time, and our data do not have the kinds of signals or sufficient temporal resolution necessary for this evaluation. Skewness is another metric of system instability, but is related to variance [1]. We did evaluate our aggregate site size data in terms of skewness; all time periods for both cases show statistically significant skewness, but, as expected, skewness and CoV are highly correlated (see Table A in S1 File).
Archaeological data provide an enduring record of completed cycles of social transformation, many of which may have relevance for transformations we are experiencing today. Archaeology is thus an ideal domain in which to undertake in-depth comparative analyses of the lead-up to both critical and non-critical social transformations and to assess what factors lead to different outcomes. We recognize that archaeological data differ from the detailed time series ecological data acquired from modeled or extant ecosystems that typify much of the published early warning signals research. Holding all researchers to the same kinds and resolutions of data that ecologists use, however, excludes the important understanding of past human societies that is possible using relevant data available from the archaeological record. Indeed while many environmental records extend over very long time scales, the archaeological record provides the only way to assess key aspects of social change over similarly long time scales. Archaeology is also the only source of direct data on societal change through transformations that occurred beyond the range of oral history or written sources, making it critical to explore ways in which the archaeological record can be used to identify early warning signals. The analyses discussed in this paper illustrate the knowledge that can be gained from data with spatial detail but limited time steps and comparatively coarse temporal resolution. We found the early warning signals approach extremely useful in the comparative analysis of the two case studies introduced below and propose that it be more widely adopted in social science research on transformation.

The Zuni and Mesa Verde Socio-Ecological Systems
Zuni and Mesa Verde are ideal cases for an early warning signals analysis given the wealth of existing environmental, climatic, and archaeological data from more than a century of research (e.g., [21,22,23]). In both regions, the emphasis on archaeological survey and analysis of settlement patterns chronologically anchored by tree-ring dated pottery type distributions allow for inter-regional comparison and an assessment of change over time (e.g., [22,24,25]). The period on which we focus, ca. 1020-1375 C.E., includes the lead-up to marked social transformations, which in both cases occurred during the 1200s.
In both cases climate change was a key directional driver over decadal to century timescales, to which people responded in terms of their choice of where and how to live. When people relocated they often created new communities and communal social institutions. In the U.S. Southwest, communal social institutions fused belief, power, and action, rather than separating what today we distinguish as the religious and the political. In the remainder of the paper we thus refer to these communal social institutions as religio-political institutions. Between ca. 1020-1375 C.E., marked periods of drought are one of the overarching external drivers of change across the US Southwest. Although drought is a perennial concern in the US Southwest, this period was complicated by severe and prolonged droughts in the middle 1100s and during the 1200s C.E. [26,27]. Almost the entire twelfth century, in fact, was dry. While that drought might be considered a sufficient shock in itself to cause a catastrophic shift, neither region experienced radical social or demographic change in the 1100s. Both areas, however, experienced relocation of population to better watered portions of their respective regions in the late 1100s-early 1200s (e.g., [28][29][30]). This population relocation became an internal driver of social change, as we discuss below. Although the driver is the same, population relocation led to different outcomes in each region.

Zuni
The Zuni region is one of only a few places in the Southwestern US that was densely occupied by sedentary small-scale farmers continuously from at least the last millennium B.C.E., through Spanish conquest, and into the present [31,32]. Our period of interest spans several transformations in Zuni settlement organization and distribution that are clearly recorded in the archaeology of the region. The 11th century was marked by the greatest extent of settlement with both hamlets and a smaller number of large settlement clusters, which are found across a range of elevations and physiographic settings. Several of these settlement clusters were centered on great houses, which are distinctive, multi-storied buildings associated with the expansion of religious and political ideologies from Chaco Canyon. Chaco Canyon (Fig 1) was the prominent Pueblo ritual, political, and economic center in the eleventh and twelfth centuries. Over the course of the 13 th century drought, population relocation resulted in the occupation of new areas in the uplands of the region that became viable for cultivation as a result of higher temperatures and drier conditions. Thus in Zuni, climate change opened new places for farming such that per capita arable land was not reduced and may have been enhanced [30,33]. Through the 13th century, people increasingly clustered in a smaller number of locations, surrounded by largely unoccupied areas. The distances between clusters were relatively small, however, and people remained closely interlinked, regularly exchanging everyday materials like cooking pots [34].
A few of the largest communities included as many as 500 rooms distributed among closely spaced residential structures, typically ranging from about 2 to 20 rooms. Many of these clusters contained features that harkened back to the earlier Chaco-style great houses, but these late post-Chaco structures included many local innovations in form [35]. The last quarter of the 13th century represented a qualitative shift in settlement organization [25,34]. Within a single generation, between about 1260 and 1300 C.E., the entire population of the Zuni region reorganized. People went from living in widely distributed small residential structures clustered into communities to massive, planned, nucleated towns containing from about 200 to as many as 1500 rooms within single structures (Fig 2). Room blocks in these large, planned villages were often built around one or more open plazas, and represented a dramatically different scale and form of community. Thus, in the Zuni region, there was continuity of settlement but a transformation in its organization [36]. By 1300 C.E. people lived in a qualitatively different settlement and social system. There is no direct evidence of violence or indirect indications of conflict (e.g., towers and defensive walls) preceding this marked reorganization.

Mesa Verde
The Mesa Verde region was occupied by Pueblo farmers from the last millennium B.C.E. to the end of the thirteenth century. As with the Zuni case, there were several significant but non-critical transformations in settlement prior to the complete Pueblo depopulation of the region by the late 1200s, which is considered a critical transformation (or regime change) from one state to an alternative. During the mid-to-late 1000s, population increased, settlement distribution expanded, and many communities became focused on Chaco-style great houses. Although population declined somewhat during the severe mid-1100s drought, many who remained in the region relocated to more defensible canyon-oriented settings that were closer to water sources and lived in increasingly aggregated settlements (up to 400-600 rooms) [28,37,38]. Population concentrated in the core of the Mesa Verde region, perhaps in part motivated by the fact that it was better watered than many other areas of the northern Southwest. Population growth, concentration, and climate change reduced per capita arable land in the Mesa Verde core as people moved into areas that were already densely occupied. With this increase in population and settlement density, particularly by the mid-1200s, there was increasing diversity in village organization including the use of different religious buildings and village configurations, suggesting experimentation with new social and religious structures. These developments also suggest there were differing ideas about the role of Chaco and its ideologies. The mid-to-late 1200s was a turbulent period with a marked increase in defensive structures, including enclosing walls and especially towers [28,37], and direct evidence of violence at a number of villages (e.g., [39,40]). By the end of the thirteenth century all the ancestral Pueblo residents, numbering perhaps as many as 20,000 people in the core area [37], and likely well over 25,000 people in the region [41], emigrated from Mesa Verde. These emigrants and their descendants never returned to live year-round in the Mesa Verde region even though it would have been environmentally possible to farm there.
In sum, both of our cases are examples of marked social transformation, but one entailed continuity in settlement with local organizational changes while the other involved a dramatic regime shift. In Zuni, people remained in place while in Mesa Verde, despite millennia of occupation, an entire population of tens of thousands of people emigrated. Environmental changes occurred in both regions and, yet, although continued settlement in Mesa Verde was possible for at least a part of the population [7,42], in the end everyone moved out. The explanation for this cannot be purely environmental.  Table 1 provides information on the external and internal drivers and responses of key internal variables to these drivers. The external driver is climate change. In response to climate change people relocated within the Zuni and Mesa Verde regions. Internal, mutually interacting variables that were affected by both the external driver of climate change and the internal driver of population relocation include the per capita quantity of arable land, settlement size, conflict, and local religio-political institutions and the power and leadership they entailed. Certain settlements grew in size partly due to the availability of arable land but also for protection against conflict and to enhance people's ability to compete for increasingly scarce patches of agriculturally productive land or permanent sources of water. The relocation of people on both landscapes led to experimentation with new forms of community and leadership. Religio-political leaders created new institutions that are manifest in changes to public architecture in both regions [30,37,43]. Some settlements likely grew because these leaders were actively recruiting followers, thereby enhancing their prominence. Leaders and members of these institutions have the potential to mitigate negative factors or to contribute to factionalism, conflict, and ultimately system collapse.

Hypotheses
The overarching hypothesis driving this study was that early warning signals are detectable prior to social transformation. Further, we hypothesized that there would be differences  between critical and non-critical cases of social transformation in early warning signals and in potential responses to those warnings. Fig 4 summarizes our hypotheses by case study. We began the project with the simple hypothesis that the non-critical transformation, Zuni, would show less variance in settlement size than the critical transformation, Mesa Verde. We then examined social factors that we expected to have played a role in the criticality of the social transformation. Thus our second hypothesis pertained to local religio-political institutions. Recent literature on early warning signals raises the question as to whether signals can be detected early enough to act upon and prevent regime shift [1,9,10,44,45]. Although we cannot identify "detection" in the archaeological record, and thus cannot say whether actions that coincide with the early warning signals identified in this analysis are "responses, " we can compare the specific contexts and actions taken with the degree to which a transformation was critical or not.
Some ecological research on SES has emphasized the importance of human management in maintaining system resilience (e.g., [8,46]). Walker and co-authors [8] note that the "capacity and intent" of human managers strongly influence SES resilience. In the pre-Hispanic US Southwest, religio-political institutions and their leaders would have been a primary locus of community-scale management of social transformation, or a source of social factionalism and cleavage. Factionalism would emerge if a social system became increasingly unstable and existing public social institutions and belief systems were challenged. Past practices and institutions may be revived, existing ones kept but less popular, and new ones developed. Therefore, increasing diversity in religio-political institutions is a potential indicator of factionalism and thus instability in social institutions as people question the status quo and pursue alternatives. Resilience, however, could be enhanced if religio-political institutions were used to dampen social instability. Thus we hypothesised that religio-political institutions were more stable in the Zuni region, than in Mesa Verde, and consequently able to manage social transformation in a way that was not possible in Mesa Verde. Size of public architecture, discussed further in the methods section below, was our proxy for religio-political institutional scale. An increase in variation in public architecture size would indicate increasing diversity in institutional scale and possibly instability in religio-political institutions, while no change or a decrease in CoV would signal stability. We expected that the CoV in Mesa Verde communal structure sizes would be greater than that for similar structures in the Zuni region.
Our third hypothesis also addressed the role of religio-political institutions in social transformation though an assessment of control over ritual practice. Ritual knowledge confers power in Pueblo society [47][48][49] and secrecy regarding ritual knowledge is a means of controlling and consolidating power [48,50,51]. We thus assessed the degree to which ritual leaders sought to control knowledge and thus power by analyzing the types of public architecture found at Zuni and Mesa Verde in terms of public accessibility. Greater conflict, indicative of internal divisions, in the Mesa Verde region led us to expect that it would exhibit more restricted, less inclusive public architecture than Zuni.

Settlement Size
We assessed settlement size using the number of rooms in habitation structures from several large (> 5 square km) blocks of full-coverage archaeological survey in both the Zuni and Mesa Verde regions representing more than 1,000 habitation structures within each region. The individual sites within each survey were further assigned to specific temporal intervals ("binned") based on absolute dates and/or the frequencies of temporally sensitive artifacts or features [52,53]. As site definitions varied across our study areas based on in-field analysts' decisions to either separate or join individual structures, we developed an additional procedure using Arc-GIS to define spatially explicit communities as our units of analysis (see S1 File). Thus, we were able to track change through time in settlement sizes across our case study areas.

Religio-Political Institutions
In the pre-Hispanic US Southwest public architecture is a primary source of information on social institutions above the level of the household [34,54,55]. Most such architecture at least in part is devoted to the preparation for and/or performance of communal rituals. For this project we first relied on the size of public architecture as an indicator of the scale of the organization of social institutions. Different sized public architecture is assumed to indicate different types of ritual organizations in the US Southwest [55,56]. These buildings include great houses, great kivas, unroofed great kivas, plazas, and in Mesa Verde, multi-walled structures (Fig 5). These features could be assigned temporal intervals of potential use based on material culture and absolute dates. Our public architecture data come from the broader Zuni and Mesa Verde regions, including areas outside of the full-coverage survey blocks (see Tables B and C in S1 File).
The available data allowed us to investigate the CoV of the sizes of great kivas, unroofed great kivas and plazas in the Zuni and Mesa Verde areas to determine whether the social scale of ritual communities increased in variance prior to the transformations in each case. We divided the analysis between communal ritual buildings (great kivas and unroofed great kivas) on the one hand, and communal spaces (plazas) on the other.
Our second source of information on religio-political institutions was the form of public architecture. Following previous analyses [28,37], we grouped public architecture forms by type: group assembly (great kivas and plazas) and restricted use (great houses, multi-walled structures, D-shaped structures, and oversized pit structures). Group assembly structures could host large numbers of people and are inferred to promote inclusivity. Restricted use structures were used by smaller numbers of people and suggest some degree of social exclusivity and control of ritual knowledge. Sample sizes for these types of public architecture were not large enough in all periods to undertake a CoV analysis leading us to rely on raw counts by period.

Climate
Archaeologists in the US Southwest have not considered drought to be the primary driver of socio-cultural change for quite some time. It is clear that demographic and socio-cultural context are critical in understanding the impacts of climate change on the peoples whom we study. Our hypotheses have to do with early warning signals, and not climate per se, but climate is the primary external driver in our two systems. Thus to provide information on the climatic context of the early warning signals we are investigating, we used the high-quality 2,129-year-long El Malpais precipitation reconstruction data from northwestern New Mexico [57] to represent major trends in climate through time across our case studies. This precipitation reconstruction provides annual estimates of rainfall in inches. We converted these data to Z-scores to illustrate deviations above and below the mean and further calculated a 9-year running mean to capture longer-term trends in rainfall. Further information is provided in the S1 File.  Malpais data illustrate that the external driver of climate delivered repeated stresses to these regions of the US southwest in the late pre-Hispanic period. Settlement changes, however, are varied and do not always occur during or immediately following dry periods.

Settlement Size
Both plots show that the relative amount of variation in aggregate size increased dramatically prior to each transformation; however, the peak CoV for Mesa Verde is greater than that for Zuni. In the Zuni region, the peak in CoV in the early-to-mid 1200s is due to a small number of population aggregates increasing dramatically in size, the very first nucleated villages that predate 1275 C.E., while most villages remained small. In the Mesa Verde region, two periods of increasing variance prior to major settlement transformations are captured: 1) a rapid increase in aggregate size in the late 1000s due to immigration [7,38], in situ population growth, and the establishment of great house communities; and 2) the peak in CoV in the mid-1200s (similar in timing to the Zuni region) due to marked population growth, internal population movement, and increased site density, which included the development of new types of village configuration and increased variation in settlement sizes [28,37,55]. After 1250 C.E., the rate of emigration from the Mesa Verde region increased significantly, and other behavioral changes discussed below indicate the mid-1200s was the tipping point for Mesa Verde. Further comparative analyses of variance prior to threshold changes will be necessary before we understand whether there is a relationship between absolute differences in the magnitude of the CoVs and the degree of criticality in the transformation.

Religio-Political Institutions
Fig 7 shows that the coefficient of variation in size of great kivas and unroofed great kivas in the Zuni region was relatively stable over time, although it did increase slightly in the early 1100s to early 1200s C.E. when roofed and unroofed forms co-existed. The CoV for Mesa Verde great kivas is perhaps even more stable than Zuni, though this likely relates to the longevity of occupation of these settlements. Plaza size CoVs, however, differ markedly between the two areas. As Fig 7 indicates, Zuni plaza size variation was minimal, while that of late period Mesa Verde plazas was much higher than any form of Zuni public architecture and increased markedly in the mid-1200s, even as emigration from the region was increasing. Formal plazas developed later in time and are much more diverse in the Mesa Verde area than in Zuni (where villages typically had only one or two central plazas). Mesa Verde plazas are more often associated with specific religious buildings or room blocks, suggesting that their size is a reasonable index of variation in internal religio-political organization size. As with the settlement size data, variation in the size of communal ritual structures does not track broad patterns of variation in climate.
With respect to institutional accessibility and control of ritual knowledge, both the Zuni ( Table 2) and Mesa Verde (Table 3) regions exhibit a trajectory toward greater inclusivity in the size and visual accessibility (roofed great kivas, to unroofed great kivas, to large plazas) of group assembly public architecture over time [34,37]. In addition in Zuni, after the establishment of large planned settlements, restricted-use public architecture declined dramatically [34] likely indicating a reduction in the emphasis on exclusivity in ritual practice and participation. In the Mesa Verde core area, however, restricted-use structures continued to be used until abandonment. Rather than a reduction, there was a reorientation and revitalization of restricted-use architecture. Great houses continued to be used, but in different ways, and multi-walled structures were introduced and elaborated upon [37,58]. In addition, the distributions of particular forms of multi-walled structures (circular versus D-shaped) were mutually exclusive, possibly representing the development of distinct religious factions that contributed to larger rifts among the Mesa Verde centers [37].

Discussion
Our analyses support our overarching position that early warning signals can be detected prior to social transformation. In both the critical and non-critical transformation, variance in settlement size markedly increased prior to transformation. Our three hypotheses concerning differences in early warning signals between a critical and a non-critical case of social transformation were also upheld. The CoV of settlement size in Mesa Verde was substantially greater than that for Zuni; the CoV for plaza size was also substantially greater for Mesa Verde. While Zuni people dispensed with restricted public architecture and increasingly built inclusive ritual  spaces, Mesa Verde ritual organizations maintained some degree of emphasis on exclusive ritual space and presumably ritual action. Moreover, Glowacki [37] has documented the differential distribution of different types of multi-walled structures, indicating a growth in factionalism in religio-political institutions in 13 th century Mesa Verde. Our study implicates such factionalism and an emphasis on hierarchical organization in the catastrophic regime change experienced by Mesa Verde peoples. Overall, the critical transformation was characterized by social fragmentation and inter-personal violence and the non-critical one by social stability and inclusivity. A recent paper by Bocinsky and colleagues [59] documents that marked demographic and cultural (ritual, architectural, material cultural) changes are associated with transitions across cultural phases in the US Southwest (e.g., Pueblo III to Pueblo IV) that were defined by archaeologists almost a century ago. Our study shows that these changes are not just the hallmarks of cultural transition but that demographic and cultural variance can in fact be early warning signals that social transformation is about to take place.
Long-term history clearly plays a role in understanding the context in which critical transformations take place. As Fig 6 indicates, the Mesa Verde area experienced a marked increase in variance in community size in the mid-1000s C.E. at a time of dramatic population influx, and variance remained high even when immigration declined. Thus, coming into the droughts of the mid 1100s and 1200s, this region was already less stable than Zuni. In social systems, then, it is clear that historical processes affect how societies respond to early warnings. This paper captures the primary trends in settlement and religio-political institution change that pertain to early warning signals prior to the social transformations at Zuni and Mesa Verde. In the course of analyzing these changes we have come to understand that finer-grained processes were at work as well, such as differences in the relevance of Chaco to our two cases, which we will pursue in further research. Chaco may have been an important external driver for Mesa Verde but was not for Zuni, amplifying the differences in religio-political institutions that may be implicated in differences in social stability between the two regions.

Conclusions
A wealth of well-dated archaeological settlement and architectural data from the US Southwest has allowed us to investigate whether the concept of early warning signals for transformations is applicable to social systems. The results of our analyses suggest that this approach is useful in understanding social transformation. Moreover, given that the archaeological record contains information on completed cycles of critical and non-critical threshold change, it provides important data for new comparative analyses of the contexts of and apparent responses to these signals that can illuminate the study of early warning signals in contemporary social systems and ways in which to manage threshold change. In both archaeological cases, directional drivers of increasing stress (climate change and population relocation) were followed by greater variation in the size of settlements people chose to live in, with some settlements increasing dramatically in size. That variance presaged transformation in both cases. There were large differences, however, in the absolute scale of variation between the cases. These differences raise the issue of whether there is an absolute value of variance above which a transformation becomes critical. Biggs and co-authors [10], for example, suggest that research should ultimately be able to identify such critical indicator levels. Although Brock and co-authors [17] have cautioned that no single threshold level will apply to all cases of a system, they encourage extensive comparative analyses so that the system of interest may be appropriately divided into subsets relevant to driving variables. Our research begins that kind of comparative analysis for late pre-Hispanic SES in the US Southwest.
Unlike the responses of ecological systems to drivers of system change, social systems contain institutions that can both be used to manage responses (stabilize), and to introduce instability. By taking a socio-ecologicalperspective and including analyses of different aspects of social institutions, we have identified marked differences in how each society handled increasing instability. Communal social institutions in the non-critical transformation (Zuni) appear to have managed regime change by promoting inclusive communal action and reducing exclusionary actions and overt hierarchy. In contrast, in the critical transformation (Mesa Verde), hierarchy, instability and conflict in social institutions appear to have added to the stress already experienced with drought and population packing.
This emphasis on the capacity of social institutions to mitigate transformation brings us to our final point. The current early warning signals literature is heavily focused on ecological systems and humans are generally perceived either as external sources of disturbance or external forces for management. Some ecologists express concern that society may not have the traditions, policies, and institutions necessary to take appropriate action in time to prevent critical ecological transformations [10,17]. Including the societal component of socio-ecological systems will be critical in deducing what actions in fact are considered appropriate and what social and political processes and institutions are available to take those actions. The people who formed the communal social institutions that we analyzed for this paper were deeply embedded in the Zuni and Mesa Verde socio-ecological systems and thus knowledgeable of the effects of long-term drought on their crops, where people were living and why, and the degree to which their social relations were fractured or coherent. A strong integration of ecological and social science research is critical for effectively using the early warning signals research that ecologists have pioneered.