1We are both fortunate and burdened to be discussing here one of the most famous archaeological districts of the world: the central Mesa Verde region of southwestern Colorado, USA. Fortunate, because generally good preservation and visibility of the archaeological record, the possibility of tree-ring dating and climate reconstruction based on tree rings and pollen, and a long history of intense investigation combine to provide a well understood, finely resolved dataset. Burdened, because so much has been written that any brief account will ignore some previous important research. Doubly burdened, in fact, since the archaeological data with which models can be compared are so rich that the inevitable shortcomings of any simple model will be all too obvious.

2Despite these challenges, in this chapter we attempt to introduce the most recent and relevant literature, data, and models helping to explain TransMonDyn’s Transition 3, which we see as encompassing two linked changes. First, the transition from farmers living in dispersed farmsteads loosely associated in spatially extended communities, to villagers living in relatively compact settlements of a dozen or more households. This change takes place between the first arrival of farmers in significant numbers in our study area ca. 6001 and the apex in size of Pueblo I villages in the late 800s, by which time a few of the largest villages had some one hundred households. At the same time, though they are more visible later, there is the emergence of systems of obligations linking these villages (and the communities they serve) into larger entities we’ll call «polities». These organizations very likely had meanings for their participants that included dimensions of identity, ritual practice, economic exchange, political coordination, and military action, as we discuss below.

Background: The Village Ecodynamics Project

3Data and models relevant to these transitions (from relatively autonomous hamlets to villages, and from relatively autonomous villages to regional systems of villages) come from numerous sources. Here we emphasize findings of the Village Ecodynamics Project (VEP), running alongside TransMonDyn but funded separately, that for the past decade has been describing and modeling subsistence and settlement processes in the areas shown in figure 1. Our focus here will be on the northern study area (VEPIIN). This area encompasses most of the central Mesa Verde region, which in turn is a key component of the larger northern San Juan region. The VEP also studies another area to south, in the northern Rio Grande region in New Mexico, to which we allude only briefly2. There is a strong historical link between these two areas. Many lines of evidence suggest that most people leaving the central Mesa Verde region when it was depopulated in the 1200s migrated to the northern Rio Grande3 4. Here we’ll only briefly mention the causes of this migration as they are currently understood. Timothy Kohler and Mark Varien5 provide an overview of the history and focus of the VEP; Scott Ortman and colleagues6 describe the environment and provide an introduction to the culture history of the northern study area. Richard Wilshusen and colleagues7 detail the early portions of the occupation and its precursors, as does Donna Glowacki8 for its final century of occupation; and Aaron Wright9 discusses regional paleoclimates, emphasizing low-frequency variability reconstructed from pollen data.

4Most VEP analyses focused on one or both of these two study areas, but we also assembled data on the dates and ages at death of human remains excavated over the last century from throughout the greater Southwest. Using an approach pioneered by Jean-Pierre Bocquet-Appel10 we were able to demonstrate very high birth rates that from about 400 - 1100 probably exceeded the highest birth rates in the world today (in Africa11). Infants are not counted in this method; birth rates are estimated from the ratio of children between 5 and 19, and adults, so higher infant survival also increases the inferred birth rate. These high birth rates likely resulted from improvements in nutrition that enabled women to birth healthy children more frequently and help those children survive infancy. It is difficult to know the most important contributions to these improvements, since many things happened in the mid-first-millennium A.D. The introduction of the bow and arrow, widespread cultivation of beans, the appearance of more productive varieties of maize, the introduction or development of efficient ceramic containers for boiling foods, and increased residential stability all likely contributed to high birth rates and, we assume, high population growth rates.

Figure 1: VEPII study areas (Southwestern U.S.).

5Over the course of the VEP and in conjunction with TransMonDyn we have begun to realize that such « Neolithic Demographic Transitions » are a key force giving shape to post-Pleistocene prehistory in many parts of the world. In our area, and likely elsewhere as well, the immediate effect is simply rapid population growth, something these societies had never before experienced. Over the course of decades and centuries, however, the larger and denser populations posed problems that, if not novel, were at least more severe than similar problems ancestors of Pueblo peoples faced earlier. The first was local depletion of key resources (especially deer and fuelwood12 13). Secondly, as local and regional populations increased14, the traditional response to local depletion—residential mobility—became more difficult. Decreased movement entrained additional local subsistence intensification, including subdividing what had been a commons into household (or more likely lineage) property by the 800s if not earlier, and increased use of domesticated turkey for food, especially after the mid-1000s15.

6Using sherd (ceramic) accumulation rates from a series of excavated and tree-ring-dated sites, Varien and colleagues16 estimate that house uselives17 in small sites increased from an average of about 8 years in the 600s to about 18 years in the 800s and 900s, eventually peaking around 45 years in the early 1200s. Houses in villages were occupied for about a decade longer than households in contemporaneous hamlets until the early 1100s, when house uselives in the two site types converged. Longer house uselives imply greater probability of inter-generational transmission of wealth (inheritance), both through time in this sequence, and also more strongly for households in villages than in hamlets for most of the sequence. Borgerhoff Mulder and colleagues18 have argued that increased inheritance allows disparities among households to accumulate. This seems to imply that villages (with their typically longer house uselives) would be more likely to evidence disparities in wealth or income than hamlets. To examine this, we calculated Gini coefficients for household dwelling and storage areas to determine the extent to which increased house uselife (as a proxy for efficiency of inheritance) contributed to increasing inequality across households. For an initial sample dating from about AD 660 to 925, households in (small) hamlets were slightly more unequal than households in (larger) villages19. Perhaps the levelling effects of rituals, conducted mostly in villages, accounts for this surprising result.

From Hamlets to Villages: Explaining Aggregation in Early Pueblo Societies

7It is in this context that small villages of a dozen or so households first appeared in our area in the mid-to-late 700s, followed within a century by relatively compact villages (in this area often called community centers) of a hundred households or so). What caused this aggregation? Why did these populations not accommodate more people by simply building more isolated farmsteads or small hamlets of a few households, as their ancestors had done?

8One possible answer -protection from predation (violence) -does not seem to be a principle motive. Although violence was by no means absent in the larger Northern San Juan region in the 700s and 800s20 the VEPIIN area itself seems relatively pacific prior to the 1000s21. Aggregation may have been initially driven in part by cooperation stemming from inter-village competition, but overt conflict is not evident.

9We use theory, simulation, and the archaeological record itself to explain this aggregation22. Considering that these populations obtained 70% or more of their dietary biomass from maize23 and that maize productivity is spatially patchy and temporally variable, we began by constructing a simple microeconomic model (from the perspective of the household) to explain variability in payoffs for maize exchange among households, as explained in detail by Kohler and Carla Van West24. This model suggests that households would maximize their utility by exchanging maize with other households in periods when production is relatively high. Such exchanges should break down, on the other hand, when production is relatively low, since in these conditions households do better on average when they hoard their production. If we assume that aggregation greatly facilitates exchange of bulky materials like maize, then we may predict that high maize productivity, high densities of exchange, and aggregation should co-occur.

10Indeed, the period from about 760-790 in which the first wave of village formation takes place was slightly above average in the amount of land within the local potential maize dry-farming niche (figure 2, top panel. The size of the niche for dry-farmed maize is not a direct measure of maize productivity, but we assume they are positively correlated.) The period from 840-880 in which the Pueblo I villages reached their maximum size was uniformly well above the long-term average in the area that is productive for maize agriculture. These villages fell apart, as theory suggests they should if a primary reason for aggregation is to improve efficiencies in interhousehold exchange, as productive conditions collapsed in the early 900s.

Figure 2: The difference in the potential maize dry-farming niche size between the study areas.

11These predictions are based solely on a very simple model with no simulation, but our simulations using the «Village» platform25 make some of the same predictions. In that model, exchange helps families survive periods of low productivity, and also encourages aggregation of dispersed households into more compact settlements26. By comparing output from runs with no exchange, and runs in which exchange was enabled with various alternative parameters, Stefani Crabtree was able to determine what combination of parameters resulted in household dispersions most closely corresponding to those known from the archaeological record. Simulations in which households exchanged food, preferentially moved closer to reliable exchange partners, and learned with whom to exchange from their parents, produced household distributions that were more similar to those in the archaeological record in the degree of aggregation than did simulations with different parameter combinations for exchange, or simulations with no inter-household exchange.

12Crabtree also examined the social networks that formed among the exchanging households. The percent of exchanging households (around 50%) was surprisingly constant both through time and among multiple simulations, showing a stable strategy (figure 3). Even when landscape productivity decreased (when the potential maize niche shrank in size), common network statistics remained stable, showing that, despite variability in population size and environmental conditions, there is a stable structure to the systems of reciprocal exchange in the simulation that we believe can be extrapolated, with caution, to the archaeological record.

Figure 3: Percent of agents exchanging in three runs (From Crabtree, 2015).

13Finally, the archaeological record itself provides important clues about how successful villages forged a common identity and unity of purpose among residents who were at most only distantly related. Some early villages formed around large, round, semi-subterranean structures called great kivas that had no apparent domestic uses. They did, however, provide a locus for community activities and ceremonies. Other early villages were built to partially enclose plazas in which oversized pit structures either sheltered emerging elite households who would sponsor ceremonies, or (if these pit structures were not partially residential, which is not easy to determine) provided a locus for group-level activities.

14Decisions to move into and stay in villages were thus influenced by both group-identity-building participation in shared rituals27 —which we acknowledge but did not model— and practical economic reasons28 29, including exchange, which we do model in «Village». Both practical reasons and cultural reasons appear to be necessary—one kind of reason cannot make up for the other; they do not seem to be additive in their effects.

From Villages to Polities: the «Polity» model

15Allen Kane30 has demonstrated that a three-tiered hierarchy of settlement size (small hamlets, large hamlets/small villages, and large villages) was in place in portions of the VEPIIN area by the mid-to-late 800s. Such distributions are frequently interpreted as asserting the presence of supra-village organizations, since they seem to call for the presence of emergent leaders in the largest settlements to coordinate the actions not only of residents of the large villages, but also the smaller villages and hamlets «depending» on each large village31. As before, we are faced with the problem of explaining an emergence: here, of polities spanning several sites, and not just one village. This task is all the more important because by the late 1000s the polities in the VEPIIN area were probably considerably larger than any we envision for the Pueblo I period32. In fact, in the late 1000s our study area became enmeshed in the history and operation of the « Chaco Phenomenon »33, a spatially extensive entity that is variously interpreted as primarily religious, primarily political, or (more likely) combining both logics, which in these societies would almost certainly not have been distinguished by the participants.

16Simulation has been important in our efforts to understand how polities incorporating a number of settlements might operate and grow. Here we begin to reference a more complicated model that is built on top of Village described above, retaining as substructure all the interactions between households and their environments, and the inter-household exchanges built into Village. Since in our case polities are forming and expanding in the context of high population growth, we orient the model on which the simulation is based around this fact. The model does not however depend on the logic of scalar stress34, which for several decades has been a main source of theory on political change in the small-scale societies of the prehispanic Southwest35. Johnson’s model employs research on information flows and decision making in small groups to suggest that groups keeping the number of decision makers below about six (at any organizational level) will not exceed acceptable levels of «scalar stress». To avoid such stresses, in his model, additional hierarchical levels are added to decision making in a sequential fashion. Our model instead focuses on competition between groups that first formed in order to achieve benefits from public goods which require scales beyond those achievable by individual households or kin-based corporate groups. Here we provide a brief overview of this model—«Polity»36—.

17Consider a group of households adjacent to a productive set of fields. They routinely carry out a number of joint activities, likely including deer hunting, keeping predators out of fields, helping each other build and repair domestic and shared structures, and so forth. What little monitoring of each others’ activities is needed is easy to accomplish given the small size of the settlement. The probable high degree of relatedness, and shared interests and identity within the settlement make monitoring largely unnecessary in any case.

18As years turn into decades their numbers grow and their kinship becomes more tenuous. For whatever public (or common) good we might imagine—defending claims to fields, conducting long-distant hunts, providing one’s fair share of food and labor for ceremonies—it’s getting harder to tell who is doing her part, and who isn’t. In fact, in the larger population of which our group is one part, those groups who consider it worthwhile to support specialists to keep track of who is doing their fair share, and punishing those who do not (if only by making fun of them, or holding them up to ridicule) likely do better on average than groups of equal size with no such specialist, because the advantages they get from their close cooperation outweighs the expense of supporting a specialist. Let’s lay our cards on the table and call these specialists «leaders» even though a visiting ethnographer might consider them to hold (for example) «ritual offices». It is also possible this leadership was a role divided among several «offices»; that’s immaterial to the argument.

19All is well until our focal group or a neighboring group grows beyond the limits of its traditional agricultural territory. There seem to be two main options at that point: they can merge peacefully into a single group (albeit one in which the dominance of the larger is honored), or one (almost always the larger) can subjugate the other by force. In either case there is a flow of tribute from a subdominant to a dominant group. Theoretically this process could repeat itself indefinitely until the entire study area becomes a single polity. Practically, though, the ebbs and flows of production through time either accelerate or slow the process down. The spatial relationship among groups is also important, since polities with large buffers between them will be unlikely to come into conflict—if at all—until the process is well advanced. Moreover, some areas are so unproductive that they support little or no population growth, in which case this process never proceeds very far.

20This quick sketch conveys the main dynamics we observe in our current simulation work on social evolution. We start with the basic structure that we call «Village» that models household demography, subsistence, reciprocal exchanges, and movement in an environment that changes from year to year both because of climate forcing (derived from tree rings sensitive to annual variability in precipitation or temperature) and human impacts.

21The simulation begins with 200 households placed randomly on the landscape and told to move to the «best» cell in their vicinity, i.e., that which allows them to farm enough maize, gather enough wood and water, and hunt successfully for any of three species of game, with the lowest energy outlay. Households however cannot move into an area controlled by a group other than their own; groups are corporate and defend their claims to arable land. Individuals in the population are made up of 11 different agent types with differing inheritable propensities to work under a leader, to pay taxes to support a leader, to contribute to the public good, or to be a leader. We have somewhat arbitrarily chosen to consider these to be matrilineages, so the wife in each of the 200 founding households forms the apical ancestor for a matrilineage and her sons and daughters receive their lineage ID from her. Households act according to the social strategy of the mother if she is alive, else those of the father. Groups doing well grow in size.

22Two key features are shared by all groups. All groups control and defend their core agricultural land, and every year households in each group play a public goods (PG) game within the group37. In such games each household gets a positive return on its contributions if everyone else is cooperating, but a temptation to defect results in declining cooperation through time in many experiments with human subjects38. Groups in the simulation may be either non-hierarchical (with no leader) or hierarchical (with a leader), as determined by the preference of the majority of their households (figure 4). Non-hierarchical groups may have one or more households playing as «mutual monitors». Such players spend some time monitoring for non-cooperation in the PG game, and punishing non-cooperators (after which they cooperate). Both monitoring and punishing are costly however and so mutual monitors do poorly (and therefore have fewer offspring) as groups get large, since their costs for monitoring and punishing increase more rapidly than does their share of the returns from the PG game. So non-hierarchical groups beyond a certain size frequently get little or no benefit from playing the PG game because their mutual monitors disappear, allowing most households to defect.

23Hierarchical groups though elect a leader (selecting from the potential leaders the one offering the lowest tax rate) who gets supported by a tax to monitor and punish defection from the PG game. Leaders also enforce collection of the tax that supports them. Hence the rate of household cooperation in the PG game in hierarchical groups is very high, and returns from the game are correspondingly high. However, such groups will do relatively poorly when they are small, because the advantages of scale from the PG game may not outweigh the additional costs of supporting a leader.

Figure 4: Conceptual Flowchart for the top-level dynamics in «Polity».

24In general, in fact, «small» groups will do better if they are non-hierarchical, and «larger» groups will do better if they are hierarchical. (We put these terms in quotes because the group sizes to which they refer depend on other parameter choices.) We call both hierarchical and non-hierarchical groups «simple» groups to indicate their independence from any other group. Soon however most simple groups come into conflict over access to agricultural land and may form groups of groups through conquest or merging. We call groups that incorporate two or more simple groups «complex groups».

25The rather complicated processes of merging and inter-group conflict are described in detail elsewhere39; figure 4 provides a simplified overview. There is no need ask the fundamental question as to whether conflict is «realistic» for these societies, since the evidence for it is abundant after about year 100040. By that time, population growth in the model has also reached the point where groups are beginning to come in conflict over territories on a regular basis. In our northern study area, conflict as reconstructed from violent trauma to human bone (i.e., empirical data) was generally higher in periods with lower per capita maize production and higher variance in potential maize production. Violence peaked when the Chacoan system was disintegrating, in the mid-1100s, strongly suggesting that some violence had motivations and ramifications that were at least in part political41.

26In the model, although outcomes are probabilistic, not deterministic, in general larger groups will prevail over smaller groups in conflict over land, and subsume them either with or without a fight. Subordinate groups then pay an annual tax to the dominant group whose amount is a parameterized proportion of their net benefit from the PG game. Taxes are channeled from each subordinate group (if there is more than one) towards the group at the top of the hierarchy, possibly passing through one or more groups at intermediate levels in the case of large complex groups. Intermediate groups retain a parameterized proportion of such pass-throughs.

27We conducted a small exploratory sweep of 37 runs with fighting and merging in which we varied four parameters. These control the probability that a casualty in fighting will die; how big simple groups may become before fissioning; the size of the tax (tribute) that subordinate groups pay to the dominant group on their PG game benefits; and the amount of that tax that is consumed by any group it passes through. We explored the effects of two or three different values for each parameter, chosen initially around those we expected to be most likely, based either on the archaeological record or on ethnography42.

28By the end of these runs (each beginning at «year 600» and ending in «1299») the processes of fighting and merging we simulate tended to produce just a few large complex groups with many constituent simple groups. The mean result across all parameter combinations is that by the end of the runs, about 2 large complex groups each composed of about 120 simple groups had formed. The simple groups within these complex groups may be either hierarchical or non-hierarchical. In contrast to our expectations, the fighting/merging dynamic did not increase the proportions of «prosocial» types (those who willingly pay taxes and contribute to the public good). Kohler et al.43 report many additional results.

29Here though is our «model’s short answer» to the TransMonDyn question as to why polities emerge in this record—the second part of Transition 3. It is due to a process of competition between groups that are growing in size in a general context of increasing regional population size towards the end of the local Neolithic Demographic Transition. These competitive processes on average favored larger groups and permitted ever-larger and more politically complex groups to emerge.

30We are certain that this model is likely incorrect in many details, and we are addressing several known or suspected inadequacies. For example, what are the limits for expansion of complex groups? We need to allow for dissatisfied or «ambitious» simple groups, or segments of complex groups, to try to break away from their polity, presumably motivated by a perception of probable success given the landscape distributions of group sizes. Once we make this and a few other more minor changes we will attempt to determine whether the general thrust of the model is correct.

31There are after all only a few ways one may assess any assertion about what causes change through time we see in prehistory. The record itself is silent on causation—a set of dynamic processes—except to the extent that humans posit models of change that imply different observations that can be made on the static record. Models that can be simulated, in which specific sets of causes are embedded, generate implications that may potentially be examined at the highest temporal or spatial resolution and along whatever dimensions (variables) the reference record allows. We have barely begun to cross-examine the record and our models in this way, but we strongly believe that this is by far the most powerful way to adjudicate among competing models that may only differ in relatively subtle ways, thus giving us a chance to infer the existence of processes that are not themselves obvious in the record.

The Depopulation of the Northern Southwest

32Although it is not one of the transitions we were asked to study as a part of TransMonDyn, it’s not possible to completely ignore the famous episode in which farmers decided to leave this area entirely in the 1200s. This is particularly true since the population growth that underlay the two transitions described above eventually collided with the ability of the landscape to support ever-larger numbers of people in the context of climatic variability affecting success of the extremely dominant maize dry-farming strategy.

33The depopulation of the northern Pueblo area—including not only the VEPIIN study area but all the farmers in Utah, southwestern Colorado, and many of those in northern New Mexico and Arizona—was not a single event, but a process composed of a large number of events that led to complete abandonment by 128544. The causes for the abandonment are linked to both climatic variability affecting maize production, and social dynamics.

34Crabtree et al.45 have been developing models examining the environmental causes for abandonment, specifically looking at Pueblo foodwebs. In foodweb studies one examines how links of trophic relations (who eats whom) create large interconnected networks. Crabtree and colleagues created a matrix of every known species in the central Mesa Verde region, linking them through their feeding patterns (figure 5), then analyzing large macrobotanical and faunal assemblages from three archaeological sites collectively spanning the period from 800 to 1280. Statistics calculated on these networks show, for example, that in Pueblo III times humans increasingly relied on species that also preyed on each other (visible in higher clustering coefficients for these late assemblages). This potentially leads to greater vulnerability, since when humans eat species that also are food sources of the primary prey of humans, humans are also decreasing their own potential food sources. When turkeys became a prime food source for people after about 1060, since humans fed maize to their turkeys, low maize yields affected humans both directly, and also indirectly through decreases in their primary protein source.

Figure 5: Trophic web of every species living in Southwestern Colorado between AD 800 and 1300.

35These foodweb studies have also suggested increasing aridification of the landscape prior to depopulation. Elk and snowshoe hare—dependent on moist woodlands—disappear in the 1200s and sandhill crane and scaled quail appear; both are dependent on sparse grassland.46 These results show how humans directly changed the environment, and how these changes would have affected human vulnerability, leading to decisions to depopulate the area.

36Bocinsky and Kohler47 show that for the entire century between the mid-1100s and the mid-1200s, during which migration streams were established from the north to the south, the relative size of the dry-farming niche was larger in the south than in the north (figure 2, middle panel). Two subareas in our northern study area that had developed dense populations when maize-growing conditions were better were hard hit by declines in the sizes of their maize niche in the mid-1200s48. The violence plaguing the final years of the northern occupation49 50seems likely to stem originally from conflicts between disadvantaged groups in those two subareas and those in the other subareas whose lands were still relatively productive.

37All this was taking place in the context of extremely high dependence on maize, which was also fed to turkeys. Given prior depression of deer populations, turkey was by that time a principle protein source; further declines in maize production beginning in the 1270s were damaging to both protein and carbohydrate supply. We now have reasons to believe that the long-discounted notion that hostile nomads (ancestors of the Navajo and Apache) were beginning to appear from the north at this same time can no longer be completely ruled out51. Tethered to the confines of their pueblos by hostilities, the relatively small number of Pueblo peoples still in the central Mesa Verde area by 1270 were increasingly unable to range widely for hunting or gathering. Facing failing crops, declining rafters of turkey, and increasing violence, yet beckoned by the relative success of their kinsfolk in the northern Rio Grande, those who could made the journey south.

Conclusions

38The research we present here as part of TransMonDyn’s efforts to understand worldwide transitions helps advance understanding of how autonomous households would choose to group together to form larger villages, and how those villages in turn might seek—or be unable to avoid—immersion in larger polities. These themes occurred in other transitions as well. The centrality of population growth to the Neolithic Bantu expansion (Transition 2) and subsequent village formation cannot be doubted, and there were likely negative demographic consequences for the hunter-gatherers they displaced. Perhaps some aspects of Transition 3, with its high-resolution record, can serve help us understand what may have been happening among the Bantu, for whom the archaeological record is very patchy. Transition 5 (600 B.C. – 400 B.C.), when dispersed Gaulish hamlets grouped together to form highly aggregated oppida may also have been driven in part by factors shared with Transition 3. While it is clear that the arrival of the Etruscans and Greeks along the shores of southern Gaul had social and economic impacts on native groups, it remains less clear how individual decisions may have influenced the decision to aggregate. Perhaps an understanding of oppida as providing public goods to their residents, as well as modeling exchange with merchant settlers, would help advance understanding of the transition from dispersed hamlets to these highly aggregated towns.

39More generally, studying how Ancestral Pueblo people transitioned from living in dispersed hamlets to large aggregated villages embedded in polities has important implications for understanding post-Pleistocene change in societies worldwide. Wherever domesticated plants and animals were more productive than their wild predecessors they drove a cycle of increased sedentism, increased natality, increases in group size and inter-group competition, and increases in sociopolitical complexity. The main outlines of the results of these processes have been evident for a long time, in the Southwest and elsewhere. One thing the research discussed here adds is more attention, including via modeling, to the processes that caused those patterns. Of course, these processes will not unfold in identical ways in different historical, cultural, and environmental settings, nor are these processes monotonically increasing in any record. The social, political, ritual, and economic structures of the post-1300 Pueblo world, which we do not have space to explore here, were radically different than earlier, in ways that reflect the final surrender of the domestic mode of production52 to centripetal forces in all these domains53 54. But that is a different transition.