h 7

7.0. Spatial distributions.

The previous chapter examined all artifact classes individually and only dealt with the distribution of different artifact types (or plant/animal taxa) over deposit types. This chapter examines how the different kinds of remains are spatially distributed and how/if any types covary with each other and with architectural units and features.

7.1 Ceramic patterns

I have chosen to employ cluster analysis to investigate the spatial patterns in the Level 3 structures. The deposits themselves can be treated as purely formal units without inherent spatial coordinates; the spatial configuration of the clustered units can be investigated further once they have been grouped. The main problem with cluster analysis when used in an inductive mode is the identification of a 'correct' solution since any cluster analysis will always produce a solution. The strategy I have employed follows Dunnell (1983) by using several different methods. If the basic structure of the solution stays the same over different clustering algorithms it is assumed to be accurately portraying the structure of the excavated units. The number of clusters chosen for a solution is, of course, a matter of judgment.

The units of measurement are the percentage of each ceramic type (only Old Kingdom types). Only dumps, floors, and pits were used giving a total sample size of 39 deposits. The final solution provided here uses the squared Euclidean distance measure and Ward's method of agglomeration. The resulting dendrogram is shown in Figure 7.1. I have chosen to further analyze the 4-cluster solution.

Immediately apparent are two sets of deposits (Cluster 1) that cluster tightly together: three deposits from Room 17 and five from unit 1192/1035. The Room 17 deposits consist of a floor (DU-5), dump (DU-4) and a pit/hearth (DU-3) in a habitation structure; the 1192/1035 deposits are all dumps. Viewing the bar graph in Figure 7.2 that compares the relative frequencies of ceramic types in each of the clusters produced shows that these deposits are dominated by Type C jars with a lower percentage (but still higher than any other cluster) of Type E bread molds, and Type H (Meidum) bowls. These are the only units where Types C and E occur in any abundance. An obvious hypothesis is that debris from Room 17 was removed and deposited within the basin containing the 1192/1035 dump deposits.

A second distinctive cluster is Cluster 4 which consists of three DU's, two each from Room 5, DU-8 and DU-9 (both pit structures) and the floor surface from Room 9, DU-7. The sherds from these deposits are isolated finds. The pits in Room 5 each contained a single diagnostic sherd each (Type O) and few other sherds. Three diagnostic sherds were found in Room 9 DU-7 (two of Type O and a single Type G) and few other sherds. Both of the Room 5 deposits resemble small dumps deposited within an abandoned storage pit and are probably not related to the occupation of the room, or perhaps a short final occupation before it was abandoned.

Cluster 3 contains deposits that have no diagnostic Old Kingdom ceramics and few sherds at all. There is little that ties all of these deposits together. The deposits from Room 2 consist of a floor (DU-4) and three thin layers from the bottom of the pit structure in that room. The pit deposits from Room 2 (DU-6, 7, and 8) were excavated as sediment samples and artifact counts are not available, though none had more than a few bits of ceramics within them (Cagle 1991). The brick pit structure from Room 5 (DU-11) shows no evidence of use for refuse dumping and also contains few ceramics, similar to the pit in Room 2. Room 13 also contains a pit structure (DU-3) that was not clustered with these, but the floor associated with that pit structure, DU-4, was. Other than these deposits having something to do with brick pit structures, the remainder represent a mix of deposit types and contexts: Room 1 DU-2 is a shallow excavated pit, Room 18 DU-3 is a laminated floor, and unit 1220/1072 DU-4 is a thin layer of redeposited dump material. The only real pattern one can infer from these deposits is that, without the later small dumping events in the structures in Rooms 5 and 13, the pit structures would be devoid of any ceramic remains, and probably any other types of remains as well. Thus, the rooms with brick pit structures tend to have few ceramics associated with them.

The last cluster, 2, contains a mix of deposits with varying percentages of different ceramic types. I have thus chosen to conduct a second cluster analysis on only these deposits to see if there are any other patterns present within this cluster. Again, I used a variety of clustering routines and similarity measures to ensure an accurate representation of any patterning present. The final dendrogram using squared Euclidean distance and Ward's method is shown in Figure 7.3.

This second routine produced three and perhaps four distinct clusters. Figure 7.4 displays the bar graph showing the type distributions among using three clusters. Cluster 4 is probably the most distinct group with only two types -- B (a jar) and D (bread platters) -- present at all in Room 5 DU-6, a floor. This results from only two identified specimens. The deposits from Cluster 1 are more similar to Room 5 DU-6 than to the others as a result of containing relatively high proportions of Types B and D, but they also contain very abundant Type G Bent-sided bowls. Type A spouted jars are relatively more abundant in this cluster but are represented by two specimens in the unit 1166/1066 DU-4 floor. Specifically, the two units making up Cluster 1 are similar largely because of their relatively high proportions of Type G bowls and Type D trays. Otherwise, the Room 13 DU-3 pit is dominated by Type D bread platters, which occur as two diagnostic sherds and several kilograms of unidentifiable (but presumably Type D) fabric type C sherds.

Cluster 3 may reveal a pattern similar to that of Room 17 and unit 1192/1035: all of the unit 1204/1060 dumps are clustered together and are associated with the floor deposits of Room 8 DU-4, Room 6 DU-5, Room 1 DU-3, and Room 5 DU-12. These rooms are not all contiguous, and not particularly near to the dumps (actually redeposited dumps) of unit 1204/1060. Their main common types are Type H (Meidum bowls) and Type I (bowls with molded rims). All of these floor deposits contain significant blackened sediment with abundant charcoal and presumably represent cooking debris, though identifiable hearth features are not evident.

The remaining deposits making up Cluster 2 are a mix of floors and two dumps (Room 23 DU-1 and Room 12 DU-1). Overall these deposits contain a fairly even mix of several types with the accent on Type O bowls. There is no discernible spatial pattern in these units, except that all of the deposits from the units in the far western edge of the excavated area, Rooms 22 and 23, are clustered together.

Interestingly, while some deposits from the same rooms cluster together (e.g., Rooms 12 and 23) others do not. For example, the two floors in Room 2, DU-4 and DU-5, do not cluster together largely due to the absence of ceramics from DU-4. The several floors from Room 5 (DU's 6, 7, 10, and 12) mostly cluster differently. The lower two floors, DU-10 and DU-12, are associated with two of the brick pit structures (DU-8 and DU-9), and the upper two floors, DU-6 and DU-7, are associated with a third brick pit structure, DU-11. The pit structures associated with both sets of floors suggest that the basic function of the room did not change significantly over time though the different ceramic assemblages suggest that some change did occur.

These two sets of cluster analyses suggest that some sort of functional patterning is discernible within the excavated areas. Room 17 and unit 1192/1035 have similar ceramic distributions and may represent the habitation source (Room 17) and refuse destination (1192/1035) for this material. That two ceramic types (C and E) are by far the dominant types suggests a degree of specialization. A similar relationship may exist between three floors in Rooms 5, 6, and 8 and the redeposited dump deposits in unit 1204/1060. These latter deposits have a wider array of ceramic types but are dominated by types H and I (type H is also common in Room 17 and unit 1192/1035).

Several deposits have very few ceramics associated with them. Many of these are brick pit deposits (Room 2 DU-6, 7 and 8 and Room 5 DU-11) which were not (as opposed to several other brick pits) used as later dumps. Two of the floors from rooms containing brick pits are also similarly devoid of ceramics (Room 2 DU-4 and Room 13 DU-4), but this pattern is not consistent (also, the pit structure in Room 13 is probably a baking facility rather than a storage facility). Two other brick pit structures from Room 5 (DU-8 and DU-9) are clustered differently but have only one sherd apiece from an apparently later dump deposit. Similarly, Room 9 DU-7 contained only three sherds which may be intrusive.

The remaining deposits show no readily interpretable patterning. Several deposits (most from the same room structures) cluster together and are characterized by a relatively even distribution of most types, but with an emphasis on Type O bowls. The brick pit structure in Room 13 (DU-3) is clustered together with a floor in unit 1166/1066, but seem to be similar only in the proportion of type G bowls and type D trays. These two units are not structurally similar nor are they close together spatially and may be a fortuitous grouping.

Thus, there does seem to be some coherent patterning in room function in terms of ceramic distributions. Rather than attempt to explain this patterning in terms of the ceramic functions themselves I will investigate whether similar patterns are detectable in other types of remains before positing specific functional interpretations of each room.

7.2 Faunal patterns

For these analyses only cattle, sheep-goat, pig, fish, and birds were used for the quantitative analyses. These were added together to create a new NISP value for only these species and percentages for each group were calculated from this. The group percentages were entered into a cluster analysis for Level 3 dumps, floors, and pits. The same criteria were used to accept an accurate cluster solution as for the ceramics. The first clustering solution is shown in Figure 7.5. A 4-cluster solution is used for the remainder of this section.

The results are similar in a number of ways to those resulting from clustering by ceramic types. Once again, the deposits from Room 17 are clustered together, but this time with only one deposit from unit 1192/1035, DU-7. Also included within this cluster are the contents from the brick pit structure from Room 13 (DU-3) and a floor in Room 5 (DU-12). Referring to the graph in Figure 7.6 comparing the proportions of these five species, it is apparent that these deposits are dominated by ovicaprid remains. In fact, the only reason Room 13 DU-3 and Room 5 DU-12 are clustered together with these other deposits is because they both have a single identifiable bone, both ovicaprids, in relatively small deposits.

I expected more deposits from unit 1192/1035 to be clustered with the Room 17 deposits. That they are not seems to be more due to the presence of other kinds of species besides the ovicaprids in several of the 1192/1035 deposits that so fully dominate the Room 17 deposits. For example, if one calculates the ratio of ovicaprids to pig, only seven deposits have a ratio higher than 1.0. Of these, six are from either Room 17 or unit 1192/1035; the other is Room 22 DU-2 which has only three identifiable ovicaprid and pig remains. Thus, with this one exception, only Room 17 and unit 1192/1035 contain more ovicaprid than pig.

Fish dominate in Cluster 4. Several of these units are from brick pit structures which seem to have functioned as dumps upon abandonment, primarily for fish remains, particularly cranial and vertebral specimens (Cagle 1991). Three of the four floors from Room 5 also clustered together with these pits, along with the dump and floor from Room 12. For the most part, the Room 5 deposits contained only a small number of fish remains and nothing else. The same is true of Room 9 DU-7 which contains a single identifiable bone, a Tilapia. Only DU-10 from Room 5 contained other species making up more than half of the total number. The other deposits in this cluster all contain significant fish remains but occur in context with several other taxa and in greater numbers. I would argue that most of Room 5 and Room 9 DU-7 can be treated as isolated specimens and are not reflective of room function or subsistence activities. With that in mind, only Rooms 12 and 9 and DU-9 in Room 5 contain evidence of substantial fish processing. The unit 1204/1060 DU-5 dump also contains a large component of fish remains. The bulk of the remains from these four deposits are Tilapia, with a small component of Synodontis, Bagrus, and Lates (Room 5 only) in Room 5 and unit 1204/1060, and Clarias in Room 12.

Cluster 3 is characterized by abundant pig, a smaller component of ovicaprids, and small numbers of both fish and birds. In some respects, this cluster is similar to Cluster 2 derived from the second ceramic clustering routine. The floors from Room 6 (DU-5) and Room 8 (DU-4) and Room 1 (DU-3) are again clustered together with two of the dumps from unit 1204/1060 (DU-2 and DU-4), as are all of the units from Rooms 22 and 23 on the far western edge of the excavated area. Several deposits from unit 1192/1035 are also in this cluster, but as noted earlier may be more appropriately placed with the Room 17 deposits. Two floor deposits from Room 2 are also contained in this cluster: DU-4 contained no diagnostic ceramics, and DU-5 contained a mix of several types.

Several of the same deposits that had no ceramics also had no identifiable faunal remains. None of the small deposits from the interior of the Room 2 pit structure (DU-6, 7, and 8) had any ceramics or identifiable faunal remains, as did the Room 13 DU-4 floor, the Room 18 DU-3 floor, or the shallow pit in Room 1 (DU-2).

In sum, several recurrent patterns are seen when the ceramic and faunal cluster analyses are considered together. First, a clear relationship seems to exist between Room 17 and the dumps in unit 1192/1035. They both have distinct ceramic distributions emphasizing Types C and E, and are the only units (with one exception) where ovicaprids outnumber pigs. This implies that much of the refuse from Room 17 was removed to the topographic depression in which 1192/1035 is situated. A similar pattern may exist between three floors in Rooms 1, 6, and 8, and the dump deposits in unit 1204/1060. The faunal and ceramic composition for all of these deposits is different from that of Room 17 and 1192/1035 emphasizing pigs over ovicaprids, and ceramic types H and I. Second, the two rooms (22 and 23) located in the far western edge of the excavated area consistently clustered together, though not generally with the same deposits from other areas of the site.

Third, the pits located in Room 5 probably contain small amounts of dump material unrelated to the occupation of the room. They tend to be small deposits with few identifiable ceramics or faunal remains and this leads to their being clustered with a variety of other deposits. Most of the faunal remains from the floors of Room 5 (except for DU-10) seem to be isolated specimens, probably not related to subsistence activities within the room. There does not seem to be any consistent patterning in rooms with brick pit structures. The pits from Rooms 5 and 2 would probably be similar except for the minor dumping that occurred in several of the Room 5 pits, but their floor assemblages are quite distinct. The brick pit structure in Room 13 (DU-3) has a much different set of remains from other pits, being high in ceramic types D and G.

7.3 Stone object patterns

Stone objects include chipped stone tools, debitage, and ground stone objects. Most types are rare and not well suited for quantitative analysis. Those that can be found in sufficient quantities are sickle blades, unretouched flint/chert blades, and ground stone chunks, flakes, and other small fragments. Consequently, the following analyses will use only these types.

Lithics as a whole tend to be sparsely distributed -- over half of the Level 3 deposits have none or only a single object -- and are not well suited to quantitative analyses. Clustering routines using all deposits tended to group together deposits that were dominated by a single type which would obviously give greater weight to those deposits with small samples. Tests of linear association were also unduly affected by outliers. Chi-square tests were marginally more informative. Using all of the deposits, sickle blades were shown to occur more often with ground stone flakes (p=0.002), ground stone chunks (p=0.000), and unretouched blades (p=0.000) than would be expected by chance. This may be more reflective of the relative ubiquity of sickle blades throughout the deposits; when only those with five or more specimens were used (N=11 deposits), this association disappeared. Ground stone may occur less often with sickle blades than chance would allow, but this pattern only presented itself using the small sample of eleven deposits with more than five specimens.

An examination of individual room contents is more informative. Three deposits from Room 17 had large numbers of stone objects (the hearth, DU-3, contained no stone objects). The DU-5 floor had 37, the majority of which were ground stone flakes (N=26); it also had 4 sickle blade fragments and 7 unretouched blades. The Room 17 DU-4 dump had a similar distribution but was dominated by ground stone chunks rather than flakes. The deposits from unit 1192/1035 had somewhat similar distributions as well, but tending to emphasize sickle blades and unretouched blades more than ground stone flakes. Interestingly, none of these units contained any ground stone implements, either whole or fragmentary. The large number of unretouched blades and sickles suggest that some production of sickle blades occurred here and in fact, this was one of the few rooms where in an earlier study I found small amounts of debitage (in a flotation sample from SU-3/DU-4; Cagle 1991:45). Related to this may be the presence of several chipped-stone artifacts in the deposits from these two units. Of the 9 chipped-stone tools recovered from all Level 3 deposits, 5 are from Room 17 or unit 1192/1035 (the former occurring in a wall collapse deposit covering the floor). Together with the large number of ground stone chunks and flakes (presumably from remanufacture or maintenance work) one is left with the inference that some small scale stone-working was concentrated in this room.

Taken together, the dump and floor deposits from Room 23 are similar to those from Room 17 in having a large number of ground stone chunks and several (N=7) unretouched blades. However, the faunal and ceramic distributions in these two rooms are entirely different.

In contrast, Room 12 is almost entirely dominated by ground stone fragments and whole objects (N=23), only a small number of ground stone chunks (N=3), and a single unretouched blade fragment. Most (N=21) of the grinding stone fragments were of sandstone while the two whole objects were small limestone metates (both circular and 12-13 cm in diameter). Two of the ground stone chunks were limestone and one was sandstone. Most of these were found in the southern 2/3 of the room beyond the small partition that divided the structure, and a number of them were clustered in the far southern end. A large portion of a turtle (Trionyx triunguis) was also found in this immediate vicinity. This room also contained a wide variety of ceramics and faunal remains, particularly fish.

Two deposits from Room 5 also deserve attention, DU-7 and DU-10. Both of these have several grinding stone fragments (N=4 and N=6, respectively) and ground stone chunks and flakes. All of the ground stone fragments are from sandstone metates, though the chunks and flakes are from a variety of materials including limestone, chert, silicified wood, and sandstone. DU-10 also contained two globular pieces of chert. It is thus doubtful that the ground stone debris is a result of modifying the grinding stones. Since both DU-6 and DU-10 floors contained rather unique globular chert nodules with rubbing wear on several faces, at least one activity or set of activities was probably carried out during these two separate occupations. Also, these deposits are dominated by fish and pig rather than the ovicaprids in Room 17, and the ceramic distributions are quite distinct as well.

The remaining deposits have small numbers of stone objects associated with them. Sickle blades tend to be common, though unretouched blades are found in more deposits overall. Ground stone chunks and flakes also occur sporadically. Grinding stone fragments are only found in two other deposits not previously discussed the DU-9 pit deposit in Room 5 and the DU-5 floor deposit in Room 2.

7.4 Floral remains

A total of 23 Level 3 deposits were analyzed for their plant material. Of these, 13 are either floor, dump, or pit deposits. In the following discussion the same aggregate groups -- Cereal, Chaff, Weeds, Reeds, Fodder, Other-Identified and Other-Unidentified -- as in the preceding chapter are used. In this section I use density values rather than proportions as this measure can directly reflect certain depositional processes (mostly burning) as well as differential use.

Of the 13 Level 3 deposits analyzed, 7 are from floor deposits, 3 are from dumps, and 3 are from pits. Six deposits showed evidence of burning (4 floors, 1 dump, and 1 pit). As expected, burned deposits had a much higher overall density of plant remains, and the density of each category was also higher in burned deposits (except for Chaff and Other-identified) but the differences were not statistically significant. Dumps also contain higher densities of all categories except Other-Identified and Other-Unidentified but these differences are also not statistically significant.

Table 7.1 shows the categorical densities for all deposits in Level 3. Room 1 DU-3 has the highest density of plant remains in this sample largely due to an extraordinarily high density of Other-Unidentified plants. The relatively high overall density in Room 8 DU-4 is also due largely to the Other-Unidentified category. Both of these floor deposits were described as dark black, laminated structures with abundant burned bone and ceramics, though Room 1 DU-3 tended to have more discrete pockets of ash and other burned material than those in Room 8.

In contrast, the floor and pit from Room 17 (DU-5 and DU-3, respectively), which, based on other data outlined earlier, is presumably the source for the 1192/1035 material, are somewhat different from those in Room 1 and Room 8. The pit/hearth in Room 17 is sharply defined by a wall of burned, hardened clay lining its edges as opposed to the more diffuse nature of the burned deposits in the other two rooms. This suggests more of a well-defined function for Room 17, perhaps a specialized cooking facility.

The main differences in plant remains between these two sets of deposits is the high density of unidentified plants in Rooms 1 and 8, and the higher density of chaff in the 1192/1035 samples. Clover (Trifolium alexandriunum), a high-quality animal feed, is especially dense in the 1192/1035 samples compared to the others. Since much of this material probably entered the archaeological record as fuel for fires (either directly or incorporated within animal dung), the possibility exists that the differential distributions are a result of different fuel mixtures used in these rooms. One interpretation is that Room 17 specialized as a cooking facility for a particular segment of the Kom el-Hisn community while the other rooms were used for more generalized food preparation. Room 17 is defined by a large number of Type C jars, while Room 1 DU-3 and Room 8 DU-4 have more even distributions of different types (particularly bowls, such as Types H, G, I and O). The plant material may indicate some kind of differential access to fuel types, either for functional or social reasons. For example, a variety of plant material may have been used as fuel in Rooms 1 and 8 resulting in a large amount of unidentified remains, while Room 17 had access to a more homogeneous supply of dung fuel.

Comparing these Level 3 floor deposits with those from earlier levels reveals some important similarities. Room 18 DU-8 (Level 4) has a very similar distribution of remains to Room 1 DU-3: very dense cereal grains and unidentified plants, and moderate amounts of other taxa. Clover especially is of similarly low density. The structure of Room 18 DU-8 is also similar to the floors of Rooms 1 and 8: a series of (3) dark patches rich in burned organic material (the flotation sample came from SU-17). These patches, like those in Room 1 and Room 8, are not as well-defined as the hearth structure in Room 17. Room 18 only contained a single (identified) pottery type, H, also similar to Room 8.

The DU-13 floor from unit 1235/1056 (Level 4) is more similar in plant distribution to the dumps from 1192/1035 than to the other floors: it has a moderate density of cereal grains, higher chaff and somewhat higher fodder densities, and a low density of unidentified specimens. Structurally, it is somewhat distinct from the floors in rooms 1, 8, and 18 in that, while no well-defined hearths are present as in Room 17, the lenses of dark ashy material are more sharply separated from other areas of clean sand than the indistinct patches of the other rooms. The ceramic content is also somewhat unique: it contains several of the same bowls (types F, G, and H) found in the room 1, 8, and 18 floors, but relatively more Type C jars that are abundant in Room 17.

In most other rooms from all levels, plant densities are much lower. In some cases this is what one would expect. For example, Room 9, which contained numerous 'column bases' and no apparent connection to any kind of cooking activity, contained very little plant material of any kind. Room 12, however, contained numerous grinding stone fragments and a great deal of faunal material and seemed to be directly related to food preparation. However, DU-1 from Room 12 (a dump or possibly an intact floor assemblage) contained no cereal grains, and very little other plant material. The only burned material within these two upper layers (the DU-1 dump and the DU-2 floor) was a thin layer of blackened material within DU-2 and it is possible that while some form of food preparation may have occurred in this room little opportunity in the form of charring was afforded for the preservation of plant material.

Those rooms with brick storage pits -- 2, 5, and 13 -- contain relatively little plant material, but the generally moist conditions at the site and the absence of burning in any of these deposits (except Room 13 DU-3) precludes extensive preservation of organic matter. The burning evident in the pit structure in Room 13 (DU-3) seems conducive to plant preservation. The ceramic contents, primarily Type D bread platters, suggests this room's use as a bakery. If this interpretation is correct, the absence of plant remains is puzzling as the fuel used would certainly leave some charred plant remains.

The overall impression one gains from the evidence provided is of an area of mixed-use, much of it related to food storage, processing, and cooking. Different forms of cooking seem to have been practiced which may be distinguished by the types of fuel used. Room 17 (as derived from the unit 1192/1035 dumps) contained a distinct hearth structure, a specialized ceramic form (Type C jars), and a great deal of chaff and clover which one might expect if cattle dung were the primary fuel. Other rooms (1, 8, and 18) contained less-distinct areas of burning, more generalized ceramic forms, and high densities of unidentified plant types which might result from the use of various plant stems and brush for fuel.

Other rooms contained predictably low levels of plant material and probably had little to do with food processing. The storage facilities (rooms 2, 5, and 12) also contained little plant material, probably due to preservation factors. Lack of preservation may also be the cause for low plant remains in Room 12 which contained abundant other evidence of plant processing (grinding stones).

7.5 Summary of artifact patterns

This section will examine patterns of co-occurrence between different artifact types and the implications such patterns have for individual room function and the overall functional complexity of the site.

Linear correlations between pairs of variables are difficult to assess. The distribution of artifacts often results in statistically significant correlations, but when plotted against one another the correlations are most often seen to be a function of a few outliers. For example, ceramic types C and E can be shown to be significantly correlated with ovicaprid remains (r=0.633 and 0.636, respectively; p=0.000). When these two pairs of variables are plotted against one another (e.g., Figure 7.7), one sees that most of the relationship is defined by the deposits from Room 17 and unit 1192/1035; most other deposits have few of either type. In many other cases a significant correlation coefficient is generated but the distributions are too scattered to suggest a truly linear relationship.

The cluster analyses above do show, however, that there is consistent patterning measured across different artifact categories in that several sets of deposits consistently clustered together (e.g., those from Room 17 and unit 1192/1035). This suggests that at least some artifacts co-occur discretely in space rather than continuously. The obvious implication of this observation is that, since the artifact classes are assumed to be primarily functional in nature, some degree of functional specialization is being observed.

An examination of artifact diversity is useful in explaining this patterning. Generally, discussion of artifact diversity is couched in the language of 'activities', the assumption being that co-occurring sets of artifacts are describing distinct activity sets. As with all other analyses conducted thus far, this lack of theoretical sophistication in the creation of units will result in some interpretive ambiguity (Dunnell 1989).

I argued earlier that the ceramic typology used herein is largely derived from functional concerns. Thus, spatial variation in ceramic distributions is likely reflecting different functions or suites of functions carried out in different structures. I further posit that rooms with a greater diversity of artifact types represent more functionally diverse structures. In this case, Room 17, which contains a few abundant ceramic types, would be considered more specialized than others containing a more even distribution of several types. The question remains as to whether variation in diversity is non-random. That is, do diversity indices vary more than one would expect if the underlying distribution of artifacts were distributed randomly across the site?

To test this, I created a series of simulated random samples using the observed proportions of ceramic types found at Kom el-Hisn. I determined the probability density function for each ceramic type (i.e., its proportion in the entire assemblage), and drew a series of 50 samples for a range of sample sizes from 1 to 300 based on these density functions. I then calculated the average richness and standard deviation for each sample size. In this case, richness was computed as the number of classes. I also calculated the average and standard deviation for evenness, which was measured using S pi2, where pI is the percentage of each type in the sample. This provided an expected range of values for richness and evenness for a randomly distributed set of ceramics based on the observed proportions. Consequently, any excavated unit whose richness or evenness index falls outside of two standard deviations can, with 95% confidence, be considered to be non-random.

Plots of the simulated random samples are shown in Figures 7.8ab and 7.9ab. The plot in Figure 7.8a, sample size against mean richness, shows that average richness increases fairly rapidly as sample size increases until the sample size reaches approximately 50-75 where it begins to level off. Figure 7.8b, sample size against standard deviation, shows a general trend downwards in the standard deviation which one would expect with increasing sample sizes, approaching 0.00 as the sample size nears the population size. Again, the standard deviation tends to stabilize when the sample size reaches approximately 50-75 sherds. A minimum sample of about 50 sherds is thus required before significance can be reliably assigned.

The behavior of the evenness measure (Figures 7.9a and b) shows that as sample size increases the evenness index (Figure 7.9a) approaches that of the population (0.1365). Very small samples, of course, have very high indices as they restrict the number of classes that can be represented. The standard deviation also decreases rapidly with sample size, again approaching 0.00 for sample sizes near the population size. Evenness measures are much less variable than richness, probably as a result of the inherently uneven distribution of the underlying population. The inflection point where evenness stabilizes is somewhat lower than for richness at approximately 25-50 sherds. For both measures, a conservative estimate for minimum reliable sample size is about 50 sherds.

Overlaying the actual richness values for Level 3 deposits on top of the simulated richness values for the same sample sizes reveals the variation from randomness that the actual distributions contain, shown in Figure 7.10a. The plots of evenness and sample size are shown in Figure 7.10b. It is readily apparent that several actual samples deviate substantially from their random counterparts. A total of 12 deposits deviate by more than two standard deviations (95% confidence interval) from the average of the simulated samples for richness values; 16 deposits deviate by more than two standard deviations from the simulated samples for evenness values. The two sets of deposits are compared in Table 7.2.

Note that the deposits are the same in each category except for Room 5 DU-12, unit 1166/1066 DU-4, and several deposits in unit 1192/1035. The Room 5 and unit 1166/1066 deposits tended to be included or excluded depending on the particular simulation as they are close to the average cutoff for the 95% confidence interval. Also, all of the richness values that are significantly different from the simulated average are negative and all of the significant evenness values are positive. Thus, all of these deposits have fewer classes and are more uneven than expected compared to random distributions.

To determine which ceramic types were causing these discrepancies I ran the same simulation of random samples for each ceramic type individually. The same set of sample sizes (300) and trials (50) were used as for the richness and evenness values. For each sample size I generated an expected proportion (the average, which should and did approximate the percentage of the total assemblage for each type) and standard deviation. The percentage of each type in a deposit is compared to the average for that type and any percentage which falls outside of two standard deviations is considered significant.

The results (shown in Table 7.3) largely confirm the results gained earlier from cluster analyses. Taken together, the Room 17 deposits have higher than expected proportions of Types C and E and lower than expected proportions of Types B, D, G, H, and O. The dump deposits from unit 1192/1035 results are similar except that Types G and H have higher than expected frequencies in several of the units and Type K is higher than expected in DU-6. The deposits from other units are varied, but most contain lower than expected frequencies of type C. Three types of bowls (types G, H, and O) are more frequent in several of these deposits along with Type B jars and Type D platters.

An obvious question regards the distribution of deviation from normality for both of these measures: no deposits have significantly more classes or significantly more even distributions than expected. This richness pattern is similar to that observed at Broken K Pueblo by Schiffer (1989) who found depressed richness values in both floor and "fill" deposits. Schiffer argued that one plausible interpretation of this pattern results from the sherds of reconstructable (whole) vessels being counted as individual sherds rather than as a single unit. The presence of restorable vessels would tend to increase the number of sherds of that type giving the deposit a relatively lower overall richness value for a given sample size. Reconstructable pots are then taken as evidence that the floor deposits in which these pots occur are "primary" floor assemblages, indicating that the pots are in their locations of use.

I suspect that a similar pattern may be represented at Kom el-Hisn albeit with different implications for site formation processes. Previous analyses have shown that Room 17 and unit 1192/1035 have similar distributions of ceramics, faunal remains, and plant taxa. Similar, though not as distinctive, patterns also exist between the floor deposits of Rooms 1, 6, and 8 and the dumps making up unit 1204/1060. All of these deposits (except for the floor in Room 6) have depressed richness indices and are more uneven than expected. If the "missed pot" hypothesis were correct, it would imply that some form of functional differentiation is evident within these room structures. Room 17 and its associated disposal area, unit 1192/1035 would be specialized in the use of types C and E, with the result that more whole or nearly whole pots would have been disposed of in these two areas. Similarly, the other rooms would be more specialized in the use of different sets of ceramic types. Rather than being indicators of "primary" or "secondary" refuse as Schiffer suggests, these patterns seem to describe patterns of use and discard in both contexts.

I believe the data presented supports this hypothesis. Room 17 contains a distinct hearth structure not found in other rooms with evidence of in situ cooking activities (e.g., the diffuse areas of burning and abundant charcoal in rooms 1, 6, and 8). Further, the two most abundant types in Room 17 -- bread molds and coarse jars -- exhibit characteristics suitable for cooking (i.e., resistance to thermal shock). Since both of these vessels were relatively cheap to manufacture it is not surprising that upon breakage they were readily curated to a nearby dump (unit 1192/1035). These deposits also contained quite distinctive faunal (mostly ovicaprids) and floral remains, the latter indicative of a specialized firing regime using animal (probably cattle) dung as fuel.

While very few reconstructable pots were found, the distribution of vessel elements of Type C jars militates in favor of large numbers of whole or mostly whole vessels in Room 17 and unit 1192/1035. Type C jars are composed of rim sherds originally designated as Type 11, and bases originally assigned to Type 1A. Type 11 rim sherds have a fairly wide distribution in the Level 3 excavation units and rooms, occurring in 12 different rooms or units, of which Room 17 and unit 1192/1035 make up approximately 65% of the total number of sherds. Type 1A bases are only found in 4 total rooms or units, of which Room 17 and unit 1192/1035 make up 93% of the total. Clearly, bases are much more concentrated in these two units than in others suggesting that upon failure the entire vessel was discarded.

In contrast, Rooms 1 and 8 (and also 6) have a very different array of ceramics, concentrating on more finely made bowls which may have served as occasional cooking vessels, but more probably as serving or preparation vessels (see Chapter 6). The plant taxa from these rooms also suggests a more diverse set of fuels used in the cooking process. The faunal assemblages for these rooms also differs from Room 17, emphasizing pig, some ovicaprids, and minor components of fish and birds. Due to their depressed richness and uneven distributions, these rooms may also be considered as more specialized around a different set of functions compared to Room 17. The data suggest that some form of food preparation was carried on in these rooms, but perhaps of a more generalized nature than that found in Room 17. The dominant ceramic types, however, do not allow a more detailed analysis of the whole-pot content of these rooms.

Nothing in particular unites the remaining deposits with significantly deviating diversity values. Based on all other evidence they have varied functions. Room 12 seems to have been used for processing a variety of animals, and the large numbers of grinding stones indicates plant processing as well, though little plant material was found in this deposit (probably due to the absence of any firing regime to preserve them). The Room 5 deposit is associated with brick pit structures that presumably functioned as grain storage facilities. Its small ceramic assemblage (n=12) is composed mostly of three types of bowls, H, I and O. Room 22 contains a much wider array of ceramic types than the others, its unevenness largely accounted for by relatively abundant types D (bread platters) and H (Meidum bowls) and a generalized set of faunal remains (primarily pig) much like those in Rooms 8 and 1.

It is thus possible that different aspects of function have played a role in producing these distributions. The unusually large numbers of reconstructable vessels in Room 17 and unit 1192/1035 certainly contribute to the deviant richness and evenness values for these deposits, which is probably a function of both specialized activities and the stressful environment in which the vessels were used. Given the burning evident in the Room 1 and 3 deposits a similar functional environment may have contributed to the in situ breakage of a different suite of vessel types, though data on reconstructable vessels for these rooms is currently unavailable. The ceramic and other data from the remaining rooms indicate these also contained different sets of functions though without more detailed analysis of the artifacts themselves (e.g., wear) the exact nature of these functions is unclear.