Timothy A. Kohler, Carla R. Van West, Eric P. Carr, and Christopher G. Langton
Archaeologists have long been interested in why ancient peoples located their settlements where they did and why these settlements were abandoned. These questions have always been approached by comparing known site distributions with resource distributions of various types, possibly (but not often) taking into account how resource distributions might have been different in prehistory. It has never been possible, with these techniques, to also model the effect of humans on the landscapes they occupy. Recently, these inductive settlement pattern studies have been greatly aided by GIS and a growing battery of statistical techniques. Here we report the first stages of a long-term project to begin understanding settlement processes by departing from household-level decision-making rather than from the archaeological record. We simulate the placement of residences and of population growth and decline, as they respond to changing maize production opportunities and local human impact on the environment, in southwestern Colorado between A.D. 900 and 1300. The effort combines GIS data planes and agent-based modeling in a way that seems promising for many observational sciences.
See also:
http://www.santafe.edu/~carr/model/village.html
http://www.santafe.edu/projects/swarm
Despite their many differences, sciences that share a strong observational component also share the serious problem of having to infer process from pattern. Some traditionally observational sciences, including ecology, have been able to introduce experimentation as one way to help overcome this difficulty. Our ability to do this in archaeology, however, is severely limited. There are even problems with applying the weaker procedure of building analogies, for many of the societies in which we are interested may have no very applicable living or historically documented analogs. Finally, we have to cope with the fact that many things that we would like to know about prehistoric societies Ñfor example, their kinship systems, religious practices, and political and economic traditionsÑare not directly preserved in the archaeological record.
In the spirit of trying to start with what can be observed in the archaeological record and working towards what can not, many archaeologists have invested a great deal of effort in attempting to understand prehistoric settlement patterns. Except where the tempo of human movement across the landscape was rapid and non-repetitive in pattern, archaeologists have a good chance of finding and dating human settlements. Archaeologists have been particularly successful in building accurate and complete maps of human settlement for the late prehistoric period in the U.S. Southwest, where relatively sedentary farmers left frequently substantial remains that can often be dated with relative precision. The resultant settlement maps can be, and have been, compared with maps of known resource distributions to describe, through statistical inference, the relative strengths of various resource distributions in affecting human settlement (Kohler and Parker 1986:402-431 provide an introduction to such work). Similar work around the world has been greatly aided by GIS in the last decade.
The comparative completeness and accuracy of these maps, however, has not magically dissipated various problems of inferring the processes that produced these settlement patterns Ñif anything, it has made the difficulty of carrying out this inferential procedure in a satisfactory way even more embarrassing. It gets harder and harder to blame the foibles of the archaeological record for our inability to agree on the processes that produced it as that record becomes increasingly complete! Key problems for the traditional inferential procedure have included:
In short, traditional inferential approaches fail to capture the dynamic and coevolutionary nature of human settlement decisions as they respond to shifting resources and the presence and actions of other people. As a result, to take as example the Anasazi region where the project we are reporting is located, there remain large disagreements among archaeologists concerning the relative importance of warfare, political competition, climate change, and human impact on the environment in effecting the settlement patterns we observe. If only we had a laboratory to study the outcomes of various processes as they might play themselves out through hundreds of years on realistic landscapes! This project reports our current progress towards building just such a virtual laboratory, and outlines our eventual goals.
Although archaeologists since the 1960s have made some use of simulation, most such work has been at the systems level. Well known examples include the attempt to apply systems-dynamic simulation in the spirit of Forrester (1968) to the problem of the Classic Maya collapse (Hosler et al 1977). By their very nature, such simulations failed to capture the importance of space in conditioning type and frequency of social interaction, and, by default, considered societies to be either internally homogeneous or characterizable by reference to a very few discrete internal components (such as elites and commoners).
A few years ago a handful of researchers in anthropology began the difficult task of modeling societies on the computer from the bottom up, beginning with the individual or the household; a notable example is the work by Jim Doran et al. (e.g., 1994) on Upper Paleolithic societies (see also Biskowski [1992] and Renfrew [1977] for more general programmatic statements). British researchers often discuss such models as examples of "distributed artificial intelligence," emphasizing that they represent cognitive processes at the level of the individual. Similar approaches are called individual-based models in the ecological literature and agent-based models by many U.S. social researchers. Regardless of what they are called, these models share an emphasis on modeling behavior at the lowest practical level, and an interest in studying the emergence of spatial arrangements and interactions among agents, and the nature and evolution of strategies for agent interaction with the environment and with other agents. Whereas systems-level models often focused on finding parameter ranges that would permit system homeostasis, agent-based models tend to focus on the problem of morphogenesis (change in structure of the macro-system) as a result of interactions at the micro level.
The immediate motivation for the present project was a desire to understand why, during certain times in prehistory, most Puebloans lived in relatively compact villages, while at other times, they lived in dispersed hamlets (Cordell et al. 1994). Our chosen approach has roots extending at least 15 years back into the early 1980s, when a dissertation from the University of Arizona by Barney Burns (1983) showed that it was possible to retrodict potential prehistoric maize y ields in a portion of southwest Colorado by combining prehistoric tree-ring records with historic crop-production records of local farmers. A few years later, Kohler et al (1986; see also Orcutt et al. 1990) simulated agricultural catchment size and shape in a northern portion of the present study area, to arrive at the suggestion that avoiding violent confrontation over access to superior agricultural land was a major force in forming the villages that appeared in this area in the late A.D. 700s and again in the mid 800s. Shortly after that, Carla Van West, in a 1990 dissertation (published 1994) used a different and larger set of tree-ring data to produce spatialized Palmer Drought Severity Indexes in 1,070 GIS data planes, one for each year from A.D. 900-1970, for the portion of Southwestern Colorado, at a spatial resolution of 4 ha. Construction of these landscapes is described briefly below. Van West was the first to use a series of local weather stations and specific soil types to reconstruct PDSI in a way that made these measures respond to very local conditions. Finally, Kohler and Van West (1996) examined these production landscapes against the known record of aggregation in this area and suggested that microeconomic processesÑat the level of the householdÑcould successfully explain whether settlement was dispersed or aggregated at any time. Specifically, we suggested that villages formed during periods when it was in the best interests of households to share food with other households, and dissolved when it was in the best interests of households to hoard their production. Our arguments were based on comparing the relative payoffs to households of sharing vs. hoarding under various production regimes, using sigmoid-shaped utility curves. To test this model more rigorously, however, we needed to simulate household placement, maize production, consumption, and exchange with other households in considerable detail.
The basic structure of the model is a rectangular lattice representing the area described above. Each "cell" in this model consists of a 4-ha square which keeps track of relevant information from the paleoproduction database. Upon this landscape exist numerous households.
Each household makes planting decisions based upon past expectations of harvests and fertility of land gathered from local cells around their central location. Dependent upon the past success of the agent's strategy in the local environment, the household might opt to search a wider area and possibly relocate if the internal storage of maize is dangerously low. The household also has probabilistic natality and mortality rules specifying when members are born and die. New household formation or "marriage" provides a dynamic element to planting and location considerations since neighbors will influence how much available land there is for planting in any local environment. Efforts are now focused on providing a plausible model of how decision-making concerning planting and residence location on this landscape. Onto this model will be added exchange relationships, more complex learning rules, additional household activities, etc. over the course of the coming year.
The primary area of research will be the effect of exchange relationships upon the formation of larger social groups. Since agricultural yields varied greatly from year to year, farmers needed to adapt mechanisms to reduce their uncertainty of future yields. One such mechanism thought to be important is reciprocity between households. After a reasonable model of agent planting is constructed, we will endow agents with balanced reciprocity behaviors and adaptive encodings of exchange, placing the households into a social and an economic network or other (related and unrelated) households. This network will be flexible enough to evolve according to agent interactions and changes in the world environment.
It is not the goal of this project, however, to generate settlement patterns identical to those found by archeologists in the Four Corners area of the United States, although the extent to which our simulated settlement patterns approach those in fact found should help us understand the processes that generated those patterns. Just as important to us is the process of exploration and discovery. What agent-based behaviors can be used to explain emergent aggregate social structure? How can we model the behavior of emergent entities such as clans or villages where some decisions at the higher level displace or limit decisions made at the local level? We are entering into relatively uncharted waters where archaeology indeed becomes anthropology, but with the benefit of a data base accumulated over hundreds of years.
This work is partially supported by Grant No. MT-0424-5-NC-026 to Washington State University from the National Center for Preservation Training and Technology of the National Park Service, USA, and by the Santa Fe Institute, Santa Fe, New Mexico.
Biskowski, Martin (1992) Cultural Change, the Prehistoric Mind, and Archaeological Simulations. In Archaeology and the Information Age: A Global Perspective, edited by Paul Reilly and Sebastian Rahtz, pp. 212-229. Routledge, London.
Burns, B. T. (1983) Simulated Anasazi Storage Behavior using Crop Yields Reconstructed from Tree-Ring Records, A.D. 652-1968. 2 vols. Ph.D. Dissertation, University of Arizona. University Microfilms, Ann Arbor.
Cordell, Linda S., David E. Doyel, and Keith W. Kintigh (1994) Processes of Aggregation in the Prehistoric Southwest. In Themes in Southwest Prehistory, edited by George J. Gumerman, pp. 109-133. School of American Research Press, Santa Fe, NM.
Doran, Jim, Mike Palmer, Nigel Gilbert, and Paul Mellars (1994) The EOS Project: Modelling Upper Paleolithic Social Change. In Simulating Societies: the Computer Simulation of Social Phenomena, edited by Nigel Gilbert and Jim Doran, pp. 195-221. UCL Press, London.
Forrester, J. (1968) Principles of Systems. Wright-Allen Press, Cambridge, MA.
Hosler, D., J. A. Sabloff, and D. Runge (1977) Simulation Model Development: A Case Study of the Classic Maya Collapse. In Social Processes in Maya Prehistory, edited by N. Hammond, pp. 552-590. Academic Press, London.
Kohler, Timothy A., and Sandra C. Parker (1986) Predictive Models for Archaeological Resource Location. In Advances in Archaeological Method and Theory 9, edited by M. B. Schiffer, pp. 397-452. Academic Press, Orlando.
Kohler, Timothy A., Janet D. Orcutt, Kenneth L. Petersen, and Eric Blinman (1986) Anasazi Spreadsheets: The Cost of Doing Agricultural Business in Prehistoric Dolores. In Dolores Archaeological Program: Final Synthetic Report, compiled by D. A. Breternitz, C. K. Robinson, and G. T. Gross, pp. 525-538. Bureau of Reclamation, Engineering and Research Center, Denver.
Kohler, Timothy A., and Carla R. Van West (1996) The Calculus of Self Interest in the Development of Cooperation: Sociopolitical Development and Risk among the Northern Anasazi. In Evolving Complexity and Environment: Risk in the Prehistoric Southwest, edited by Joseph A. and Bonnie B. Tainter, pp. 169-196. Santa Fe Institute Studies in the Sciences of Complexity, Proceedings Vol. XXIV. Addison-Wesley, Reading, MA.
Orcutt, J. D., E. Blinman, and T. A. Kohler (1990) Explanations of Population Aggregation in the Mesa Verde Region prior to A.D. 900. In Perspectives on Southwestern Prehistory, edited by Charles Redman and Paul Minnis, pp. 196-212. Westview Press, Boulder.
Renfrew, A. C. (1987) Problems in Modelling Socio-Cultural Systems. European Journal of Operational Research 30:179-192.
Reynolds, R. G. (1986) An Adaptive Computer Model for the Evolution of Plant Collecting and Early Agriculture in the Eastern Valley of Oaxaca, Mexico. In Guila Naquitz: Archaic foraging and early agriculture in Oaxaca, Mexico, edited by Kent V. Flannery, pp. 439-500. Academic Press, Orlando.
Van West, Carla (1994) Modeling Prehistoric Agricultural Productivity in Southwestern Colorado: A GIS Approach. Reports of Investigations 67. Department of Anthropology, Washington State University, Pullman.
Timothy A. Kohler
Department of Anthropology
Washington State University
Pullman, WA 99164-4910 USA
ph: (509) 335-2698
e-mail: tako@wsu.edu
Carla R. Van West
Statistical Research, Inc.
P.O. Box 31865
Tucson, AZ 85751 USA
e-mail: sristats@aol.com
Eric P. Carr
Carleton College
Northfield, MI 55057 USA
e-mail: carr@santafe.edu
Christopher G. Langton
Santa Fe Institute
1399 Hyde Park Rd.
Santa Fe, NM 87501 USA
e-mail: cgl@santafe.edu