Non-coastal Mississippian Period cultures (A.D. 900 to 1700?) have traditionally been thought to be ranked societies supported by intensive maize agriculture (Brose and Percy 1978; Brown et al. 1978; Fowler 1969; Gibson 1974; Griffin 1967; Larson 1972; Overstreet 1978; Peebles 1978; Smith 1978; Ward 1965; and others). While considerable research has produced lengthy statements on the proposed sociopolitical aspects of these precontact populations (Autry 1983; Brown 1971; Goldstein 1980; Larson 1971; Peebles 1971, 1983; Peebles and Kus 1977; and others), little attention has been focused on generating models of the economic system which served to support these societies. Yet, some archaeologists associate changes in the prehistoric record with cultural adjustments precipitated by intensive agricultural practices:

The settlement pattern of Mississippian populations in some floodplain situations might also change through time if soil depletion necessitated shifting the location of homesteads, and perhaps even local centers [Smith 1978],

The reasons for the abandonment of hamlets were probably varied but may have centered on both the depletion of natural food resources and on soil fatigue by unrestricted crop-growing [Harn 1978].

Such concerns are warranted given the ethnohistoric accounts of the shifting nature of aboriginal agricultural systems :

As the Indians never manure their ground and do not even let it lie fallow, it is soon exhausted (and worn out). Then they are forced to move their villages elsewhere and make new fields in new lands [Lafitau 1977:69-70],
The land, as they do not cultivate it, produces for only ten or twelve years at most; and when the ten years have expired, they are obliged to remove their village to another place [Thwaites 1896-1901:15:153],

. . .on the 7th of November, 1715, Monsieur Begon wrote that Father Cholenec, the missionary of these savages, represented in 1714 to Monsieur The Marquis de Vaudreuil and to him that these savages could no longer remain in their village, because the soil was exhausted and the woods too far away; and that it was absolutely necessary for them to settle elsewhere [Thwaites 1896-1901:67:25].

Recognition of the effects of agrarian practices on cultural systems is not new. Cowgill's (1961, 1962; Cowgill and Hutchinson 1963) Mayan research examined various aspects of Guatemalan soil productivity, concluding that soil depletion did not account for the Mayan collapse in the region (cf. Reina 1967; Street 1969). Heidenreich (1971:159-198) produced a preliminary examination of the Huron's soil needs which suggested that by the early seventeenth century this Lower Great Lakes population was living close to its maximum carrying capacity. A detailed study by Parry (1975, 1978) examined the drastic effect of soil depletion and climatic change on land tenure in medieval Scotland. Additional studies and reviews substantiate the importance of measuring the cultural effects of man's interaction with the environment in an agricultural setting (Bennett 1973; Green 1980a, 1980b; Hosler et al. 1977; Meadows and Meadows 1973; Moylan 1973; among others).

This study is an extension of the current trend towards redefining the relationships of theory and methodology. Its purpose is to augment the current format of archaeological inquiry while delineating the structure, stability, and changes in prehistoric agricultural systems for the Mississippian Period (A.D. 900 to 1700?) in eastern North America. This will be accomplished by presenting a model of Missis-


sippian agricultural productivity potential based on the recognized boundary conditions of such systems as defined by Stability Theory (Nicolis and Prigogine 1977:71). By attempting to model culture change as a byproduct of dynamic systems, I will address whether we have the ability to associate the observed fluctuations in Mississippian cultures with a steady decrease in agricultural potential; not as a singular cause and effect relationship but as one component of an overall system instability requiring social and cultural adjustment. In so doing, I explore potential weaknesses in the current empirical approach to data recovery and examine the explanatory power of well-defined models developed from concepts of stability theory.

Recent discussions (Friedman 1975, 1982; Friedman and Rowlands 1977; Peebles 1978; and Renfrew 1979) provide us with several theoretical relationships, some of which Schroedl (1986) suggests may be capable of qualitatively describing change in terms of a systemic reaction to nonspecific fluctuating conditions. Are there ways to take what we know about specific cultural phase states and use this information to predict trajectories in phase space (i.e. the topological coordinate system mapping cultural characteristics against time)? My goal is to model the dynamic relationships incorporated in shifting agricultural systems, creating a methodology suitable for quantitatively testing the effects of one set of specific fluctuations (agricultural impacts) on the stability of Mississippian systems. It will differ from the purely theoretical discussions by presenting the model in actual Mississippian phase space. That is, the model will predict Mississippian trajectories along a real time line; not solely in terms of


abstract relationships. I will not try to "predict" that Mississippian cultures will fail. We know that to be the case. I will show one reason why the failure occurred and, more importantly, reproduce its rate of occurrence.

A global or regionally independent definition of Mississippian variability recognizes growth in social complexity and elaboration through various discrete stages beginning with an emergent phase that transcends to a climax followed by a precontact decline into historic societies (Peebles 1983). For Mississippian studies the historical precedent has been the presentation of distinctive associations of elaborate material and structural remains as indicators of changing, complex social organization along a space-time continuum. Current systemic approaches focus on the boundary conditions of the system as an "... adaptation to a specific habitat situation" with "a particular level of sociocultural integration" (Smith 1978:480). The resulting abstract models are derived from optimization theories for resource allocation and redistribution of prestige goods. Such a paradigm makes agriculture an unbounded resource, important only in terms of its labor (i.e. organizational) requirements. Such a perspective never requires a detailed study of agriculture's long term impact on influencing Mississippian cultural evolution (cf. Peebles 1978). Agriculture has been ignored in favor of the more observable aspects of the archaeological record. Nevertheless, as an integral part of the total subsistence base it must be included in any examination of overall system structure.


This examination starts with the premise that the response of any agricultural system will fluctuate over time. How will these fluctuations affect the stability of the parent cultural system? Traditional approaches to archaeological inquiry leave this question unanswered. This serves as the initial impetus for this study, but it also leads to another more basic question. How can we measure or observe stability? To interpret Mississippian phase shifts (morphogenesis), we must develop some means to address this problem. At this point in the discussion only a general definition of stability is needed. After developing the model, a more lengthy examination will be appropriate.

All systems can be defined by a finite set of interacting variables. If a characteristic definition is developed that expresses this interaction in a way that facilitates predicting system response, then the fluctuating conditions can be isolated. In terms of macroscopic detail, stable systems maintain (within finite limits) their initial reference state despite changes in the values of internal variables. This is referred to as structural stability (Nicolis and Prigogine 1977:69). As demonstrated from evolutionary theory, stability is maintained within environmental and behavioral limits (Rindos 1984:264). Their respective parameters constitute the system's control (a) and a behavior (x) spaces. The rules that define the behavioral limits form a x potential, V(a,x). Stability is defined when the rate of change in V(a x) potential is zero (i.e. dV/dx = 0). Stability Theory concentrates on V(x) identifying these absolute limits.


As a simple example, the response of an agricultural system may be defined to be energy production per unit catchment area (quintals per hectare) and the potential to be the amount of available arable land (ha). The system itself is composed of four internal parts (x): botanical resources, an extractive technology, some level of horticultural knowledge, and a work force. Externally, it is affected by weather conditions and the nutrient level of the soil (a). In the absence of a optimal stability, a system will respond to errors (e.g. insufficient production levels) by changing internal behavioral parameters in ways that minimize undesirable effects. At some point the system's response may not be maintainable because these internal adjustments trigger an eventual system collapse necessitating a redefinition of the operating rules (dV/dx not equal 0). This "triggering" is caused by the forced acceptance of behavioral rules, like planting larger fields, that are outside the current behavioral limits which preserve stability. We view the occurrence of such a change as a shift in phase space or, archaeologically speaking, a Mississippian transition.

Following Green's (1980b:337) approach the agricultural process is necessarily an interaction between cultural and environmental systems. This linkage involves management, impact, response, and feedback mechanisms. Horticultural practices, the primary behavioral input, form the management portion of the process. The initial effects of these decisions produce an environmental impact on the ecosystem. The ecosystem's response to the impact, varying from one environment to another, provides feedback to the behavioral system. It is the interaction of


these two systems, cultural and biological, that define the larger agricultural process.

Given such interactions, does the response of the system in this simple example remain constant over time? The historical record documents change in such systems. To demonstrate this, the internal and external variables must be examined by mapping the fluctuations in agricultural response over time and defining the parameter space appropriately to account for established agricultural processes. In dealing only with agricultural subsistence strategies, the questions related to the origin of agriculture can be effectively ignored. These topics are best left to other approaches concerned with adaptive potentials and not the kind of stability questions examined here (Rindos 1984:275). Given that the agricultural choice has been made by Woodland and Mississippian populations, then, it is necessary to measure the impact of that choice.

How does this conceptualization of the problem differ from that of the past and promote alternative trends in analysis? Addressing the complex question of stability requires a certain theoretical understanding of systems interpretation and data acquisition techniques. Our interpretation of archaeological cultures can tend to be overly simplistic from an explanatory perspective. We are often content to isolate singular cause and effect relationships to explain transformations without quantifying their rate functions. This concentration on certain avenues to the exclusion of others is largely a result of the developmental process of enhancing archaeological science.

At any point in time a discipline consists of a finite set of approaches, some of which may compete with others for supremacy (e.g.


Binford 1985; Gould 1985). Archaeology concentrates on describing, defining, and explaining changing residual patterns in cultural refuse using varying techniques of observation and generalization to link the patterns with hypothesized behavioral models. By the 1960's recognition of the archaeological implications of culture as a dynamic, if unobservable, phenomenon led to the replacement of the more static material dependent theories of culture change with what has become known as the processual approach (Willey and Sabloff 1974:209).

Although this New Archaeology has been characterized as a methodological binge (Moore and Keene 1983), its earliest application depended more on qualitatively invoking an unobservable Processual Being as the ultimate explanation of most material patterns in the archaeological record. As a complex adaptive guidance system, "process" was seen as the underlying causation for the morphogenesis of one discrete archaeological unit into another. Methodologies were selected that presented artifact patterns in ways that were assumed to be quantitatively scientific and logically valid in hopes of extracting process from static assemblages. Yet, strong arguments linking the process with the patterns could not be made largely because the methodologies were not adequately linked to an archaeological theory relating objects, context, and morphogenesis to a single system. In the end, culture was reduced to a byproduct of the Processual Being who, like Laplace's all-knowing Demon (Prigogine and Stengers 1984:75-77), orchestrates the dynamics without revealing any of the mechanics.

This initial failure to fully explain culture process resulted from an inability to recognize that, while thinking we were explaining


change, we were really addressing philosophical problems of system identification (see Maciejowski 1978:19). After developing some analytical sophistication, we have begun to challenge the unobservable nature of the Being, realizing that process and context must at least be described, if not explained, as archaeological systems. We should be able to assign these systems operational parameters and recognize boundary conditions that trigger change. We recognize that to define process effectively, we need to develop new tools of observation and analysis that are derived from archaeological theory and capable of relating qualitative expectations with quantifiable data. Whether we call this theory Middle Range (Binford 1977) or not (Moore and Keene 1983), its development is beginning to replace the strictly methodological emphasis of earlier studies. The research presented here serves as an extension to this trend.

Most examinations into the workings of agricultural systems have failed to adequately address Green's (1980b:337) four mechanisms cited above. In particular, carrying capacity measurements have ignored the effect of environmental degradation caused by agricultural practices. This is largely because most studies begin by assuming that their focus is on a stable process (see Street 1969). It is a mistake to assume that the ethnographic present results from stable conditions. Similarly, we should not assume that archaeological phases are associated with stability. Indeed, processual instabilities direct cultural trajectories. Modeling the entire agricultural system (aspects of management, impact, response, and feedback) produces a more useful representation of cultural dynamics that recognizes inherent fluctuations. But


more importantly, when done in the manner described here, such models display the ability to predict archaeological phenomena, such as the timing of phase shifts. This is what makes this study unique and important.


Clarke's (1972) presentation and volume (1972, edited) on the subject of modeling demonstrates that modeling has been accepted as a valid analytical approach. Archaeologists interpret their observations on the basis of a set of conceptual models (paradigms) of one sort or another. However, the degree to which our conclusions are testable is dependent on some measure of the explicitness or refutability, in Popper's (1959:86) sense, of our constructs. Here modeling deals with predicting system behavior under specified conditions based on observed, past behavior (Maciejowski 1978:12). Because they require rigid specification of relationships, such abstract summaries serve as the most specific, manipulative framework for organizing observations along integrative and interpretive lines. In this sense they can become powerful tools of analysis and description.

In exchange for precision, modeling requires an acceptance of approximation. Any real phenomenon being analyzed must be defined under some rule of closure whereby the interrelationships within a finite set of essential parameters (i.e. sufficient to approximate observed responses) are examined (Bellman 1968:7). We cannot expect to account for all the interactions or all the variables. Despite this limitation, quantitative models do provide a means of addressing questions that expose the


continuum of processual dynamics, otherwise unapproachable by conventional methods.

Our applications have, however, often ignored the conceptual differences between model identification and realization in Maciejowski's sense (1978:1-22). The identification of the system involves coming to terms with the larger philosophical questions surrounding the nature of the scientific method. How is it that we observe and interpret the archaeological record? Is there a singularly "scientific" way this should be done? What should be observed? These questions were partially addressed during the development of the New Archaeology, although not explicitly in terms of modeling. System identification involves placing experimental observations within a larger conceptual framework which serves as an abstract summary of the data. Archaeologists have assumed this involves the almost impossible task of taking static material observations and reconstructing cultural processes. In retrospect, this was the cause of the New Archaeology's analytical failure (Binford 1982). Today's Middle Range applications try to reverse the order by observing cultural processes in an attempt to define the resultant archaeological record. By restructuring our concepts of scientific archaeology in this way, we are in a better position to construct refutable models of the archaeological record.

For this study the identification process involves developing a way to quantify aboriginal behaviors in terms of their ecological impact. We must duplicate the rules governing Mississippian agriculture by extrapolating modern analogues from historical records which describe Mississippian-like lifeways. Given these behavioral options, we can use


observations from agronomy to predict the ecological response to Mississippian conditions.

System realization involves generating the means by which inputoutput relationships can be calculated. Selection of a model, whether statistical or dynamic, is largely based on one's assumptions about the structure of the phenomena under study. To accomplish both one must first present a system definition based on observed relationships specified under well defined research restraints. To improve accuracy, the observations used to construct the model should not exhaust those available for testing its validity. In other words, a model which only reiterates its initializing observations lacks predictive credibility. Using data independent of the specific archaeological application or choosing a different set of situations to test a model is the only appropriate way to utilize the modeling approach free of tautologies.

Our level of resolution is largely dependent on our goals. Detailed specificity in model design sacrifices the global application of the results. For example, a model designed around a single site loses its general validity at the coarser scaled regional level. Likewise, the more general and theoretical the model the less useful it is in addressing specific micro-level variability. A model's application must be clearly presented in terms of some level of specificity and is constrained at a particular level of resolution in terms of space and time. Just as the modeler must not overstate the conclusions, the reader must be cautious not to misapply the results to an unsuitable situation (or at a scale inappropriate for the model's assumptions and parameters).


The problem orientation of this research is somewhat complex, having been derived from several studies on the nature and effects of change during and after the Mississippian period. Any model used to examine a critical part of such a system should produce equally discontinuous responses at equivalent intervals along a temporal dimension. These intervals make up the frequency of change for the system.

Recent studies (Autry 1983; Peebles 1983; Schroedl 1986; among others) have benefited from theoretical discussions of culture as an information processing system (Johnson 1978, 1982), a means of social reproduction (Friedman 1975, 1982; Friedman and Rowlands 1977), and a topological manifestation (Renfrew 1978, 1979; Renfrew and Poston 1979; Poston and Stewart 1978:412-413; Zeeman 1982). Each of these theoretical approaches offers a processual perspective suitable for addressing Mississippian change. Unfortunately, initial applications have been limited to qualitative discussions of their processual elegance (in an explanatory sense) and not directed towards quantification of the frequency of change. This has largely been due to the difficulties inherent in quantifying information, social reproduction, and relevant topological parameters along a time line.

This study recognizes the usefulness of these approaches as a first step towards realigning theory with observation and methodology. My conclusions are not intended to replace the intuitive elegance of their arguments. But, to add substance (i.e. a level of refutability) to the application of theories such as these, they should be restated in terms that are directly testable and that facilitate dynamic modeling of specific situations. That is, it is necessary to identify, through


observations on working systems, the parameters that serve to influence the direction of change and create a computational model to plot the fluctuations of relevant response variables over time.

My intent is to supplement their intuitive conclusions with such an analysis of the, as yet unexplored, agricultural subsystem of the Mississippian Period. The subsystem is worth studying because of the historical references to its instability, its assumed importance in the literature, and the lack of any substantive description of its prehistoric form.

The Hypothesis

Because the focus of this study is on the agricultural limitations of Mississippian cultures, the model's definition of crop production should be in terms of the needs and capabilities of the practitioners. This involves determining what is grown, how it is grown, and how much is produced within a specific cultural context. This observational stage involves comparing similar populations in like environmental and technical settings. Unfortunately, finding extant data on maize agriculturalists lacking domesticated livestock and living in temperate climates is difficult (see Nye and Greenland 1960). Therefore, these parameters must be extracted from ethnohistoric accounts of contact groups in North America.

Resorting to analogy by enumeration assumes that the observed situations in the sixteenth through eighteenth centuries approximate those of the previous 700 years. Although the specifics will change over this time period, the maximum technological knowledge should


approach that of the contact period. The material technology of hoe agriculture prior to the introduction of European tools can be safely assumed to be a constant. Maize varieties, if grossly different between A.D. 900 and 1700, will certainly be more productive at the end of the period. Thus, inaccuracies will tend towards understating the negative effects of aboriginal agriculture over the entire period. The minimum caloric needs of each individual will be a constant even though surplus requirements may vary. Given the discipline's use of ethnographic analogy, data extracted from ethnohistoric accounts should be acceptable as a basis for describing the macroscopic behavior of hoe agriculturalists.

The observational base will provide the structural framework of the system in terms of behavioral and technological inputs and their resultant productive output. By identifying the options available to traditional aboriginal cultures, we can address the long term maintenance implications of such a man-plant-soil relationship. This involves understanding the physiological needs of maize and the productive response of the agrarian ecosystem to nutrient extraction and replenishment. Such independent empirical information is readily available from botanical, agricultural, and ecological sources. Understanding the agrarian ecosystem makes it possible to select computational models of specific responses to traditional practices. These responses will be compared to the dynamic fluctuations indirectly observable in the actual archaeological systems.

Once a model is constructed on the basis of a set of observed system responses, its application must be archaeologically tested. As


his study represents an initial attempt to model such a system within an archaeological context, its application must be at a level of specificity suitable for data acquisition and presentation. This would involve a single catchment area with well defined environmental and cultural data. I have selected data from the lower Little Tennessee River valley of east Tennessee as the subject of the study. The geographical boundaries include the counties of Monroe, Blount, and Loudon (see Figure 1). These arbitrary limits serve to contain the agricultural catchment around known sites in the Little Tennessee and Tellico River Valleys and provide manageable soil data relevant to this study.

The availability of extensive archaeological data produced by late nineteenth century investigations, the Works Progress Administration period of Southeastern archaeology, and the recent Tellico Archaeological Project (Riggs and Chapman 1983) provide sufficient cultural data for model testing. Ethnohistoric accounts of the Overhill Cherokee settlements in east Tennessee (see Baden 1983; Schroedl and Russ 1986) contribute regionally specific estimates of model parameters. The region also benefits from palynological research (Cridlebaugh 1984) and published soils data relevant to agricultural modeling. The combination of model identification, realization, and testing following the guidelines discussed above should provide a suitable demonstration of the power of system modeling for archaeological applications.


Figure 1. Study area.