3.0. Method and Theory

3.1 Overview

Egypt is one example of an apparently obvious case of a highly complex culture that differs in striking ways from other complex early civilizations such as Mesopotamia, yet is still regarded as a classic example of a complex preindustrial civilization. Wilson (1960), for example, noted that Egypt's towns and villages, as then known, did not appear to exhibit the degree of size, density of occupation, and range of non-agricultural activities as their Mesopotamian counterparts, and therefore did not qualify as true 'cities' as in the Mesopotamian model. Sjoberg (1960) however, argued that even though Egypt's city structures differed in many respects from the Mesopotamian model they still were cities in their own right. In any case, since Egypt exhibited all the other hallmarks of complexity -- monumental architecture, administrative hierarchies, specialized craft production, etc. -- the question of whether Egyptian cities were functionally equivalent to the Mesopotamian variety is somewhat moot.

Whether Egypt really was a 'civilization without cities' reflects deeper theoretical issues involving explanation of the initial development of cultural complexity. It also highlights, as Wenke (1997) notes, the typological nature of the comparative method: without a set of theoretically-derived necessary and sufficient conditions to define what "cities", "city-states", and "nation-states" are, comparison becomes a matter of simple (though in many ways useful) description rather than explanation.

Nevertheless, comparing the nature of Egypt's formation into a complex state can be useful for identifying those factors common to the formation of cultural complexity worldwide and those that are highly variable or unique to specific regions or cultures. Trigger, for example, in his review of several complex societies found that there was much less variation in some cultural elements than expected given different ecological settings.

I expected to discover that, because of ecological constraints, the differences in economic systems would be limited, and there would be more variation in sociopolitical organization, religious beliefs, and art styles. In fact, I have found that a wider variety of economic behavior was associated with early civilizations, the one constant being the production of surpluses that the upper classes appropriated through a tributary relationship. Yet I have been able to discover only one basic form of class hierarchy, two general forms of political organization, and a single basic religious paradigm. . .I have documented significant variation from one early civilization to another only in terms of art styles and cultural values. (Trigger 1993:110)

He interpreted this lack of variation as indicating that certain structures were more efficient than other structures and were thus optimized to particular forms.

Trigger's use of the comparative method to highlight particular traits that are potentially adaptive with regards to entire civilizations points the way to the use of Darwinian evolutionary theory to explain such similarities scientifically. However, the atheoretical character of the comparative method limits its utility to simple description and ad hoc generalizations of what constitutes "significant differences" rather than providing the necessary framework for any sort of scientific explanation of why those common traits arise. In other words, it frames the general question of cross-cultural similarities and differences but does not provide the appropriate theory (or the units derived from theory) to explain those similarities and differences in a scientific fashion.

Dunnell and Wenke (Dunnell 1978a, 1996; Dunnell and Wenke 1979) have argued that the advent of complex societies represents a change in the scale at which selection operates. In simple societies, the individual or kin-group is the basic functional and reproductive unit and it is on this unit that selection (as differential reproductive success) operates. With occupational specialization and centralized decision-making, the scale moves to that of aggregates of individuals: groups of people who are functionally differentiated but integrated into a larger whole. The functionally integrated group becomes the reproductive unit.

Within this framework, differences between cultures are explained by delineating the specific evolutionary trajectories each has taken. This is accomplished by tracking changing frequencies of functional traits over time and determining what environmental factor(s) is responsible for the changes. In many cases, different cultures will have certain traits and selective factors in common, much as similar bodily structures evolve in different species that have common functional requirements (e.g., wings). It is important to recognize, however, that such similarities are only incidental to an evolutionary explanation, they are not what drives explanation. Since the traits under consideration and the mechanism(s) by which they reproduce themselves are specified by the underlying theory, cross-cultural similarities are explained by default. With the inductive approach used by Trigger (and the comparative method generally), similarities and differences amount to empirical generalizations that must be further explained by some other (usually unspecified) theory. Formal theory is, however, strict about the kinds of units which it can explain, a point clearly made by Lewontin:

If one simply cannot measure the state variables or the parameters with which the theory is constructed, or if their measurement is so laden with error that no discrimination between alternative hypotheses is possible, the theory becomes a vacuous exercise in formal logic. . .(1974:11)

Consequently, if one is to construct an evolutionary explanation of the development of complexity one must develop units which the theory can explain. Key to a Darwinian explanation is the use of units that affect the Darwinian fitness of the organism, in this case the society. Those forms that affect the Darwinian fitness (i.e., the differential reproductive success) of the population in which they occur are functional (Dunnell 1978c:199), contrasted with those forms that have no detectable selective value termed stylistic. The contrast between style and function is basic to evolutionary explanation since the factors that affect their transmission and distribution are very different.

Stylistic forms are not affected by selection and consequently their spatial and temporal distributions are controlled by the degree and frequency of interaction. Since their distributions are determined by direct interaction, similarities produced are homologous and are explained by processes involving interaction such as descent, trade, invasion, and so forth. Because of this interaction they are the basic units by which "cultures" are defined and their interactions with other "cultures" explained. The creation of stylistic types has been a fairly rigorous process: types are formulated and reformulated until they display the monotonic distributions characteristic of temporal-spatial interactions. The directional characteristics of these distributions over time give rise to the formal seriations and other chronological schemes such as Petrie's Sequence Dating (Petrie 1900). The formulation of the types themselves tends to be intuitive, but the distributions they are compared against -- the test of "historical significance" in American parlance (Krieger 1944; Ford 1954a, 1954b) -- provide an empirical performance test of their utility. Larger units based on these types also tend to retain the spatial-temporal contiguities of the original types, giving rise to spatially and temporally distinct units such as phases and cultures (Dunnell 1978c).

Functional units are and have been created differently. The notion of "function" is generally not a specific archaeological concept like historical significance, but a common sense one. Functional classes are defined most often by the similarity of archaeological specimens to modern or ethnographic forms whose use is known. The identification of function is thus a formal process of comparing archaeological to modern or ethnographic forms; no additional testing is generally done and the "functionality" of the resulting type designation is judged by the procedure by which the modern analog was determined. Thus, while stylistic types have been informally defined and then formally tested, the reverse is the case for most functional types. It is only fairly recently that more rigorous methods of determining function (e.g., wear) have been employed, but even so, the traditional designations (e.g., axe, knife) are still widely employed, especially in Egypt.

The distribution of functional types also has a different set of controlling factors from stylistic forms. Since functional forms are controlled by selection and their interaction with the environment, similar forms can arise without any sort of direct interaction among the populations giving rise to analogous similarities which are ostensibly the kinds of similarities of interest to Trigger and others. The environmental interaction of functional types also provides a means for testing their utility: functional classes ought to correlate with other evidence of activities or with environmental variables (Dunnell 1978b:60). This may take the form of covariation among different functional types (e.g., bread molds and oven structures) or the correlation of individual types or suites of types with micro-environmental variables (e.g., sickle blades and agricultural fields). Functional artifact types should translate to larger units much as their stylistic counterparts do: the "site" becomes a set of co-occurring functional types representing intrasocietal functional differentiation. In this light, describing the functional complexity of an individual town or village and its history becomes part of the process of explaining, in evolutionary terms, the history and selective environment of the organism under selection, i.e., the society.

Such an analysis does not necessarily entail simple functionalist explanations since the unique ecological and cultural sequences of each civilization condition and constrain the trajectory that each society follows. This can be appreciated by comparing the basic economic units of Egypt and Mesopotamia. Among early Mesopotamian territorial states and empires, the city constitutes the building block, the 'basic economic unit and primary producer of the culture' (Joffe 1991:35). In contrast, the basic economic unit of Egypt was the nome, a distinct province administered to one degree or other by local officials and recognized as individual units by the central government. The nome boundaries apparently did not change significantly from Predynastic times until the medieval period and seem to correspond in large part to natural flood basins (Baines and Malek 1980). This preexisting structure may have contributed to the largely non-urban character of most Egyptian settlements since the relative ecological uniformity of Valley and Delta environments, along with the efficient transportation link that the Nile provided, would afford few advantages for the concentrated and highly diverse urban centers that predominate in Mesopotamia. That Egypt eventually developed more urban settlement patterns along with site placements that bear no apparent relation to ecological factors (e.g., Amarna) serves to militate against simple ecological determinism.

The differing structure of the basic economic unit between the two also forms the basis of their respective settlement pattern structure. In Mesopotamia, rank-size plots of settlement sizes tend towards a log-normal distribution with several large cities surrounded by outlying towns and villages of varying size. Egypt however, until at least well into the Middle Kingdom, seems to have had a settlement distribution that approached a primate distribution with a single large city (Memphis) and numerous smaller towns and villages scattered up and down the Nile and throughout the Delta. This apparent difference is a result of the more rural, non-urban character of Egypt's settlements which, as argued above, may reflect the particular ecological characteristics of the Nile and is the essence for Trigger's (1983, 1993) contrast between the city states of Mesopotamia and the 'territorial state' model of Egypt's state structure.

However, rank-size plots are difficult to interpret in the absence of functional data on individual settlements. Key to such an analysis is defining the appropriate variables in such a way that they are archaeologically measurable and applicable to different regions and cultural contexts. Traditionally, one of the primary goals of urban archaeology is to identify the overall functional complexity of the site as a whole; that is, to determine the range and spatial distribution of functions that were carried out within the settlement. In other settings (particularly open-air sites) intrasite functional analysis has concentrated on various statistical and mapping procedures to locate and define artifact clusters thought to represent sets of specific activities -- the discovery of 'activity areas' (e.g., Carr 1984, 1985; Dacey 1973; Whallon 1973, 1974). In architectural settings where definable structures are present this analysis usually takes the form of identifying the activities carried out within specific architectural features (e.g., Hill 1966, 1970; Lightfoot and Feinman 1982; Longacre 1964).

The major problem in all aspects of intrasite spatial analysis is the dynamic nature of the sedimentation process. Few sites fulfill the Pompeii Premise (Binford 1981) of all artifacts remaining in their original positions of use. That their occurrence is so rare has to do with both the distorting factors of the sedimentation process and with the activities of the human inhabitants: artifacts can be used in more than one location, can be used in different ways in various locations, and the activities performed in one location may change over time. Thus, even if the Pompeii Premise were fulfilled at a particular site, what is eventually discovered by the archaeologist is a synchronic manifestation of a diachronic process of change -- an aspect of the material record that is the main contribution of archaeology to the study of culture.

A great deal of research has taken place to sort out the various factors, both natural and cultural, that distort the original 'pristine' archaeological record and shape the distribution of artifacts into its modern configuration (e.g., Butzer 1982; Schiffer 1972, 1983, 1987; Stein 1987; Stein and Rapp 1985). Generally subsumed under the rubric of 'formation processes' these studies have had as their goal tracing individual artifacts or classes of artifacts through the cultural system to their eventual deposition. The rationale behind this strategy is that if one knows the processes by which an artifact was moved from its primary location of use (or minimally to know if its present location is its primary location of use) then one can reconstruct to some degree the original use-position of the artifact, thereby fulfilling the Pompeii Premise analytically rather than physically.

This research program has been invaluable in giving archaeologists a sense of the numerous factors that may be responsible for the spatial distribution of artifacts across a site. Because the distribution of artifacts is one of the primary ways in which the functional configuration of a site is determined, a knowledge of the factors that move artifacts around is of utmost importance. Nevertheless, these studies have been hampered by their reliance on ethnographic analogy and their inconsistency in separating the processes that brought together a deposit as a unit from the life histories of the artifacts contained therein. Since aggregates of artifacts are the basic analytic unit by which the spatial distribution of artifacts is determined (through counts or frequencies of artifact types and the co-occurrence of same) this is not a trivial matter.

The most critical step in this endeavor is the definition of the deposit itself. Particularly in urban settings defining deposits has been a two step process of using visual characteristics to identify depositional units and then interpreting the units in common sense terms. These interpretations have tended to be ad hoc and based on colloquial notions of what a particular type of deposit should look like as opposed to a set of mutually exclusive and exhaustive conditions defining in theoretical terms how a deposit was created. For example, trash dumps should look something like modern trash dumps, occupation surfaces should look something like modern (generally ethnographic) occupation surfaces, and so forth.

In architectural sites the critical issue has been one of associating particular activities with structures by analyzing the contents of occupation surfaces or living floors that are thought to contain objects and features in their primary use contexts. Other deposits, such as dumps, are assumed to contain remains from a variety of locations and thus are not generally considered to contain information relating to the use location of the artifacts contained therein. However, as several authors (e.g., Binford 1982; Butzer 1982:98-100; Gifford 1978; Hayden and Cannon 1983; Kent 1984, 1999; Rosen 1986, 1989; Schiffer 1987) have noted, numerous processes can move artifacts from their location of use to their eventual point of deposition such as cleaning, trash disposal, reuse of surfaces, and scavenging, along with natural processes such as mass wasting and water flow.

Diehl (1998), for example, has divided these cultural processes into three general classes of behaviors that affect floor deposits: caching, scavenging, and trash disposal. These processes operate to remove items from their original use locations (de facto assemblages), and place them into their final archaeological locations. In order to reconstruct the original de facto assemblages from the resulting 'disturbed' assemblages one must identify the processes that have affected the original distributions and work backwards to find the original distributions. However, as Diehl notes "there is no unitary body of theory that calibrates the frequencies of different kinds of artifacts with the intensity of curation behavior or the manner of disposal. One must rely on a strategy of generating, from theory a set of general conditions that one ought to observe given the assumptions that certain processes affect artifact distributions"(1998:620; emphasis added).

The 'theory' to which Diehl refers has usually been based on ethnoarchaeological studies aimed at determining the ways in which classes of artifacts work their way through the cultural processes of procurement, manufacture, use, maintenance, discard, and incorporation into the archaeological record (Schiffer 1987). The ethnographic observations are then used to determine what particular deposits should look like archaeologically when a set of behaviors acted upon the artifacts contained therein. The expected configuration of artifacts is measured on the diversity of the artifact inventory, its completeness (in terms of whole or reconstructable objects such as ceramic vessels), the nature of the artifacts in terms of their value to the occupants, and their physical characteristics (such as size or weight).

The degree to which the artifacts in a deposit can be reassembled into whole objects, for example, has been used as an indicator that the objects in a deposit have not moved very far from their location of use and presumably use-related breakage (e.g., Cressey, et al. 1982; Nelson 1985). This presumably indicates that there was a short and simple waste stream between use and discard locations in a particular site (Schiffer 1987:286).

Diehl also argues that artifacts that require high manufacturing or maintenance costs will generally not be left behind upon abandonment because of the high cost to replace them. When several of these types of artifacts are found together it indicates either rapid catastrophic abandonment and deposition, or caching behavior. Indeed, Schiffer (1987:89-98) posited numerous factors that determine curate behavior, or the proclivity of the inhabitants to carry off artifacts upon abandonment. These include the rapidity of the abandonment process, whether it was planned or unplanned, ritual abandonment, whether or not a return was anticipated, the length of the habitation, the nature of the artifacts themselves, and the distance to the next habitation.

That these observations have some heuristic value is not in question here; certainly on a broad scale these sorts of analogies have analytic value especially where they can be tested against other sorts of data (for example, textual descriptions or the house models of Middle Kingdom Egypt [Winlock 1955]). However, as with all analogies these definitions have their limitations (Ascher 1961; Gould and Watson 1982; Wylie 1982, 1985), and analogical arguments should be treated as statements of plausibility that require testing by alternate means.

Clearly, one must explain the association of artifacts within a deposit and their location in space with reference to theoretical principles that are not dependent on analogic reasoning. Also, one must separate the interpretations of the life histories of artifacts contained within a deposit from the determination of the depositional event that brought the artifacts together archaeologically. In order to proceed, one must specify the kind of unit one is creating with some sort of theoretical justification. Fortunately, a body of theory has been developed in the geological sciences that deals specifically with the formation of deposits. This research will adapt the geological principles of deposit formation to a cultural context.

3.2 Aggregate units (deposits).

Archaeologists have borrowed concepts from the geological sciences since the two disciplines share many goals (e.g., the study of stratigraphic sequences). However, archaeologists have reinterpreted geological principles to suit the essentially anthropogenic nature of many archaeological deposits. Stein (1987, 1990) has reviewed the similarities and differences between geological and archaeological uses of the concept of the deposit and found several key concepts that have been misunderstood or inappropriately used by archaeologists.

Geologists deal with several different types of units. A basic distinction is that between observable and inferential units. Observable units are based on a description of material referents. These include lithostratigraphic units which are defined on the basis of lithologic characters such as texture, color, and rock type, and biostratigraphic units which are defined on the basis of fossil remains contemporaneous with the formation of the deposit. Since biostratigraphic units are based on different criteria than litho-stratigraphic units, the boundaries between the two need not be coincident. Several biostratigraphic units may be contained within a single lithostratigraphic unit and vice versa. Both of these units are largely descriptive; they only characterize the deposit based on its structure and composition.

Other depositional units, such as those dealing with the chronology of deposition (Stein 1987:345-346), such as chronostratigraphic units (all deposits formed in a given span of time), are based on these observable units and are inferential. These units are not descriptive but based on certain geologic principles relating to the process of deposition (e.g., superposition). Other interpretive units may relate to the kind of environment represented by a deposit (e.g., fluvial, marine, etc.).

Several authors have attempted to standardize the process of creating definitions of deposits that are relevant to archaeological analysis. Fedele (1976, 1984), for example, proposed the Elemental Sediment Unit which would be defined by its boundaries and lithologic homogeneity, essentially a lithostratigraphic unit. Harris (1977, 1979) and Schiffer (1987) both propose to define deposits on the basis of the mode of deposition of a given stratum (source, transport mechanism, depositional environment, and post-

depositional alterations), with Schiffer tending to rely on the inferred histories of the artifacts within the deposit as the basis of classification. The obvious difficulty, as Schiffer (1987) himself notes, is that a single deposit may contain numerous artifacts with widely different life histories and that this strategy is more properly applied to individual artifacts. Doing so, however, merely reduces the scale at which a deposit is defined (individual artifacts rather than aggregates of artifacts) and does not really address the reasons why the artifacts in a particular deposit came together as they did.

Clearly there are numerous ways in which a deposit can be defined. For the purposes of this research, a deposit is defined following Harris (1977, 1979) as the mode of deposition, in geological terms, that brought the set of sedimentary particles (both natural and artifactual) together. The mode of deposition can be defined broadly as the source, transport mechanism, depositional environment, and post-depositional alterations experienced by the particles making up the deposit. Together these four factors describe the process by which any set of sedimentary particles formed a deposit. This type of definition differs from the sort of behavioral formation processes favored by archaeologists in that it only describes the last set of processes that acted on all of the particles in a deposit rather than the entire use-lives of artifacts before final deposition.

The rationale behind this is that the current positions of all artifacts in any site are a product of the last depositional event that moved and deposited them, be it the natural flow of water or people depositing trash into a pit. In analytic terms human and natural depositional agents are not qualitatively differentiated since, at base, they both act to move particles from one location to another. Hence, with respect to their spatial location, all particles within a given deposit can be expected to owe their location to a single set of factors. These factors, based as they are on the physical properties of particles and their movement, are grounded in a firm body of theory. Because the location of all of the particles in a deposit defined this way can be assumed to owe their position to the same set of factors, the spatial variation due to the mode of deposition is thus controlled. Consequently, spatial variation within a given set of deposit types can be assumed to result from something other than the mode of deposition.

When deposits as analytic units are defined strictly by sedimentary principles, the question then becomes one of explaining non-random variation within a given type of deposit. When the deposit is defined in behavioral terms only, the interpretation is implicit in the definition. For example, the contents of a habitation floor are interpreted by the artifactual content as the 'function' or activities performed in the room. But when the deposit is defined in sedimentary terms interpretation of non-random variation must proceed by analyzing other aspects of the deposit's contents in addition to the characteristics that defined its mode of deposition initially. This does not involve tracing the life histories of artifacts contained therein or attempting to define de facto assemblages that act as a yardstick for determining processes that have affected individual artifacts since they were used in their 'primary' position, á la Diehl (1998). Rather, behavioral inferences are made only as they relate to the final depositional process that brought the artifacts together as a unit. In most cases, the interpretation may closely approximate traditional interpretations of how a deposit was formed (e.g., 'dumps'). However, since each deposit must be defined in terms of sediment movement nebulous terms such as 'fill' deposits will be absent. Thus, each deposit will have a clearly defined set of attributes upon which the interpretation of its genesis was based.

In this context, many of the characteristics that have been proposed to identify the behavioral factors of deposit formation have some value when expressed as processes that affect the movement of sedimentary particles. Gifford (1978:82-83), for example, related the likelihood of artifacts becoming incorporated into floor sediments to the size of the artifacts (<3 cm) and the permeability and median grain size of the floor sediment itself. Rosen (1986, 1989) expanded these ideas and suggested that 'primary' debris in occupation surfaces may be identified by the grain size distribution of certain sediment categories (such as charcoal). Still, as Diehl (1998) and Schiffer (1987) note, caching of artifacts may enrich the artifact inventories of occupation surfaces, though this should not affect the distribution of microarchaeological remains which are incorporated into the sediment base largely through trampling (also Stein and Teltser 1989).

The strategy of this research is to define deposits on characteristics which identify the source, transport mechanism, depositional environment, and post-depositional alterations that brought together all of the sediment particles as a unit. Thus, each class of deposits can be assumed to have been formed by analytically identical processes, even though the artifacts contained in them may have had radically different life histories. Each class of deposits will then be analyzed for its distribution over space and its association with other deposit classes and architectural features. In this way a general model of the site's depositional history will be constructed.

Following this analysis I investigate the distribution of artifact types within each deposit class. The basic procedure for doing so is basically a statistical one: I will be testing the null hypothesis that a particular class of artifacts is randomly distributed across deposits in a single class. If the null hypothesis is rejected, the significant variation is explained by processes other than those involved in the formation of the deposit. The specific source of this variation is determined in part by the types of artifacts that are varying non-randomly and their contextual associations. It is at this stage where traditional formation process studies will be most useful in interpreting the existing variation.

3.3 Artifact classes

As Dunnell (1983:124) notes, there are two largely contrastive methods of spatial analysis depending on how the assemblages have been described. Analysis of the spatial configuration of artifacts described in functional terms (Dunnell 1978b, 1978c) broadly describes activity patterns, while the same analysis applied to artifacts described in stylistic terms describes patterns of interaction. The latter has often been interpreted as representing 'social structure' (e.g., Hill 1966, 1970; Longacre 1964).

Because of long-standing chronological concerns some notion of 'style' has played a central role in much of 20th century archaeology (e.g., Holmes 1886; Kroeber 1919; Phillips, et al. 1951). Much of the discussion involving style centered around why types created for chronological purposes behaved the way they did. Without any theoretical justification, the monotonic distributions that chronological types display were explained based on common sense notions of popularity. The types used in the construction of chronologies tended to be defined using a mixture of many different attributes -- for example, ceramic decoration, temper, and paste -- and tinkered with until they displayed the kind of distributions necessary for seriation. Only recently has attention been paid to the theoretical basis for the behavior of stylistic traits and the mechanisms by which they are transmitted (Dunnell 1986, 1996; Dunnell and Beck 1978; Lipo 2000; Lipo and Madsen 2000).

The identification of function has been treated in much the same way. Objects are assigned a functional meaning based on a form of 'functional uniformitarianism' (Dunnell and Beck 1978) in which the relationship between form and function are constant through time. Using this principle, archaeologists have interpreted the function of specific items with reference to ethnographic items of similar form. Though certain rules apply to the linking of ethnographic and archaeological objects (Fritz and Plog 1970), the basic procedure remains the same: applying an English (French, German, etc.) name whose meaning is commonly known to an object or set of similar objects. In this way, the function of an object is 'explained' through common sense notions of the activities a particular set of named objects are involved in.

The concept of function changed somewhat with the work of Semenov (1964) and later workers (Feathers 1985, 1990; Frison 1968; Gould, et al. 1971; White 1969; White and Thomas 1972; Wilmsen 1968). Within this paradigm, function is defined more narrowly as the "relationship that obtains between and object. . .and its environment both natural and artificial" (Dunnell 1978b: 51). Function is placed in an empirical realm by focusing on such attributes as edge damage and wear patterns in addition to gross morphology. Much of this research has focused on tools at a scale lower than the individual object (such as individual edges and instances of wear on stone tools), thus eliminating the problems of alternate uses, multiple uses, and reuse.

In a Darwinian evolutionary context, 'function' has a theoretically defined meaning: forms that have identifiable selective value (Dunnell 1978c). This is contrasted with 'style' which consists of forms that have no identifiable selective value. Functional traits confer a selective advantage on their bearers while stylistic traits are neutral in this respect, either through no detectable change in fitness or through equivalent fitness contributions across their distribution. Ideally, one could distinguish functional traits through engineering analyses of the performance characteristics of alternative configurations (e.g., Feathers 1990). The differential transmission of certain traits through time is explained with reference to variables that describe the selective environment in which the traits are differentially selected for or against. Archaeologists have generally used stochastic processes to model the way neutral traits are distributed through time (Teltser 1995, and recently to test a model of neutral distributions against archaeological samples, Lipo 2000).

Nevertheless, most assemblages, especially in Egypt, are still described in standard generalized type systems originally constructed in the late 19th and early 20th centuries. Often these generalized typologies were created for one use (usually chronological) but put to another when the disciplinary focus changes. This leaves the relationship between the attributes used to define the types and the mechanisms their distributions describe unchanged. Thus, one can easily be left with a set of data in the form of type descriptions that are unsuited for dealing with the proposed problem.

Since the main focus of this research is concerned with functional variation over space and the bulk of the data consists largely of standard Egyptian type descriptions, the central question becomes: to what extent can one expect spatial variation in artifact type frequencies to reflect functional variation? The answer to this question will vary depending on the overall class of artifact (lithics, ceramics, bone, etc.). In some cases (e.g., bone, which can be tied directly to diet) the question is largely moot; in others, it is more complicated.

Ceramic types have been used often to create types sensitive to chronological variation (e.g., Ford 1954a). As noted earlier, these types often contain a variety of attributes that may have either functional (or functionally neutral) or stylistic significance, depending on the nature of the problem for which the types were created. In Egypt, Predynastic pottery was especially important for dating purposes, primarily in tomb and temple contexts though it was also used for deriving later Dynastic Sequence Dates (Petrie 1900). It has long been noted, however, that Dynastic pottery, especially in Early and Old Kingdom times, is much less diverse in both form and decoration than during Predynastic and earlier times. In addition, textual and artistic remains (especially king lists) assumed relatively more importance for constructing chronologies than did ceramics or other types of artifacts, thus lessening the importance of chronological concerns when constructing ceramic typologies. For these and other reasons, as I argue in more detail in Chapter 6, I believe that the ceramic types used at Kom el-Hisn (a typology based largely on the standard typology used for this period) may be assumed in a general sense to be functional rather than stylistic.

Lithics and other stone objects are probably more prone to this type of attribute-mixing than ceramics since the same type descriptions are applied to Dynastic specimens as were used for earlier periods when chronological concerns were paramount. This is not, however, particularly important at Kom el-Hisn as the vast bulk of chipped-stone tools are sickle blades whose function is fairly straightforward to determine given the obvious wear patterns involved and their discovery (elsewhere) in association with whole tools (sickle handles). Other kinds of chipped-stone tools (such as knives and scrapers) are rare and do not provide a large enough sample for distributional studies. Ground stone is more difficult to assess as most of the objects found at Kom el-Hisn are fragmentary and wear patterns have not been studied in any detail. Certain categories, such as manos and metates, are often recognizable in their fragmentary form and have clear functional interpretations (though multiple use and re-use is a problem that can only be addressed through wear studies). Many other objects (e.g., 'pounders') have more obscure functional meaning; as with chipped-stone tools, however, most of these are relatively rare and not used for the quantitative distributional studies presented here.

Thus, even with the relatively generalized typologies employed in the description of Kom el-Hisn artifacts, it appears likely that much of the variation will be functional in nature. My strategy for examining this variation is threefold. First, I will determine if any types are differentially distributed among different deposit types. These will be used to interpret patterns of use and discard. Second, I will use clustering techniques to detect patterns of co-occurrence both among artifact types and within particular structures. Third, I will apply diversity measures to room assemblages and test them against distributions derived from a repeat-sampling simulation. The goal is to detect non-random distributions of artifacts that describe the functional layout of the site. The patterns described for Kom el-Hisn can then be applied in much the same way to other sites and thus begin to generate more general patterns of functional complexity for this period in Egypt's history.