F. Deblauwe: Melebiya Spatial Analysis

Space Speaks Volumes:
Quantifying Spatial Characteristics of 3rd Millennium B.C. Houses
at Tell Melebiya, Syria

Francis Deblauwe

(University of Missouri-Kansas City)


Text
End notes
Figures
How to contact the author


From 1984 through 1988, a mainly Belgian team of archaeologists under the direction of Dr. Marc Lebeau excavated Tell Melebiya (1). This site is located on the Khabur river, a little south of Hassake, in northeast Syria. The excavated remains of this provincial town date back to the Early Dynastic III and medieval Ayyubid periods. In this paper, I will focus on the houses from the Early Dynastic III period (ca. 2500-2300 B.C.) in their final phase. I am especially interested in the spatial characteristics of these buildings and how they relate to room functions and building categories (2).

First, I analyzed the circulation and access patterns of the floorplans as found in the final report. How does one access the different spaces of a building, how are they connected and what is their place in the complex as a whole? When one constructs a building, choices are made, be they conscious or unconscious. Passage through and access to certain rooms can be made difficult or easy. Consequently, a certain hierarchy is evident in the distribution of and relationship between the spaces contained in a building. A building can itself also be more or less integrated.

The first set of variables was obtained through a real relative asymmetry analysis. This is a technique developed by Bill Hillier which quantifies how well-connected different spaces in a building are (3). Let us have a look at figure 1. First, one draws a justified diagram of a building by tracing the passage from one space (kitchen, courtyard, etc.) to another, starting from the exterior, this is the "outside" from which one enters the building. The exterior is indicated by a circle with a cross in it, and the other spaces are represented by simple circles. Attested doorways are indicated by a solid line connecting spaces; reconstructed doorways by a broken line.

In figure 3 we see the second step: the calculation of the relative asymmetry parameter. The smallest amount of steps possible to go from every space to every other space (including the exterior) is noted down and the sum is made for every space. The next line contains the values per space for the mean depth. All spatial analysis formulas are given in figure 5. Then the relative asymmetry is calculated. The smaller the value of this parameter, the better the space is structurally integrated in the building or in other words the easier it is to get to it. However, to be able to compare different buildings with different numbers of spaces, we need to compute the real relative asymmetry or RRA. This still leaves us only with values for individual spaces, while we naturally would want to compare complete buildings too. This can easily be done by calculating the mean RRA value.

Let us now return to the imaginary example. Notice in figures 2 and 4 how the RRA value of space 2 and the mean RRA increase when one blocks the doorway between spaces 1 and 2. A higher RRA value equals less accessibility, a lower RRA value stands for higher accessibility. The floor plans of the most complete Tell Melebiya houses and the circulation diagrams based on Dr. Lebeau's understanding of them are illustrated in figure 7. I computed the RRA parameters for the individual spaces within the structures as well as for the houses in general (building RRA).

There remains one problem however: due to the nature of this RRA analysis it can only be performed reliably on buildings of which we have at least a nearly complete floorplan. Therefore, I would like to suggest a second way of measuring circulation within buildings, based on doorways per space. One simply counts how many doorways each space has. Then, one calculates the mean doorways per space or DpS for the building as a whole.

In the two imaginary examples, the DpS of space 2 changes from 2 to 1, the DpS of space 1 from 3 to 1, and the mean DpS decreases accordingly. The lower the mean DpS, the more access to spaces is controlled within the building. In the case of the Tell Melebiya houses, an obvious problem arose because quite a lot of doorways are hypothetical. Therefore, I decided to differentiate the DpS parameter. I first counted the doorways for which we have unmistakable evidence and called this parameter the certain DpS. Then I counted the number of doorways per space again but included the doorways reconstructed by the excavator. The latter value is plainly called DpS. In both cases, I also noted down the mean value for each building.

Furthermore, we have a closer look at some dimensional aspects. Four types of "spacings" are distinguished: door width, longest and shortest side of a space, and surface area. Door width is abbreviated "DW," longest side "LS" and shortest side "SS." The surface area of a space as reconstructed by Dr. Lebeau is abbreviated "area." This approach is inspired by dimensional research done by Roland Fletcher (4). In order to obtain only one value for each space, the door widths are averaged per space (average DW). Overall mean values for each house are also obtained.

All these purely architectural data are thought to be indicative of what went on inside the spaces and may highlight differences between the spaces and houses. I correlated these spatial variable values with the function of the spaces. Does a pattern emerge? For instance, are kitchens easily accessible? Moreover, the function attributed by Dr. Lebeau to some spaces leaves some doubt. I separated these spaces from the rest and treated them each as a separate function category. I looked at the patterning for variables according to these "special functions."

I present the data under the form of notched box plots. What are notched box plots? Figure 9 starts out with an imaginary example. Without going into statistical detail, this plot can be explained as follows: the box contains 50% of all values, the line inside the box is the median value. The median value is the middle one in a ranked list of values. It is therefore a real value unlike the mean which is only a mathematical average. The confidence interval notches result in a wasp-like silhouette with the waist at the median.

I made notched box plots for all the spatial variables. In figure 9, the box graphs of each spatial variable are plotted for the functions and the problematic spaces. The problematic spaces are represented by a single vertical line: naturally, all they have is one value. Figure 8 lists the abbreviations of the functions. The first question is: are the functions differentiated adequately by the spatial variables? As far as the RRA values are concerned, the workshops, reception and living spaces, and kitchens are related in that they are less accessible. As was to be expected, the entrance and the transitional spaces are very accessible. The doorways per space box plot correctly shows that the entrance spaces and to a lesser extent the transitional spaces are well connected. Least connected, most secluded in a way, are the kitchens and workshops. A similar picture is presented by the certain DpS plot. The LS values show that the transitional spaces stand out: they are often corridors which indeed have a very long side. The variation within the transitional spaces is very significant, however: their shape is not uniform. Entrance spaces tend to be rather small. In the shortest side box plot there is also a lot of variation for the living spaces. The entrance spaces again show up to be rather small in dimensions. What jumps out of the average DW plot is that the living spaces have a large variation in door widths, some of them are rather large. The entrance spaces also have large DWs which makes good sense. The kitchens on the other hand don't need easy access. Finally, the area values show that living spaces are again a varied lot ranging from small to large size. Kitchens tend to be tiny just like entrance spaces and workshops.

Let us now turn our attention to the so-called problematic spaces which were also included on the box plots as their own special, one-member function categories. These spaces are identified on the plans and on the RRA diagrams in figures 6 and 7.

Space 186 is a possible side entrance space for house B1 but was probably primarily used as a workshop. Its workshop status is strongly confirmed by the certain DpS box plot and supported by the DpS and SS values. What evidence is there relating to space 186 being an entrance space? Only the area and LS box plots support this. However, the DpS values show a perfect match for a living space function.

Space 955 in the same house is thought to have been a transitional central space which also served partially as a workshop. Evidence for this being a transitional space can be found only a little in the average DW box plot. Maybe it served more as a workshop? It doesn't really seem to fit that pattern either: only the average DW values might possibly confirm this. There is no clear alternative though.

House B1 contains one more interesting space with was assigned the number 1052. Here too, Dr. Lebeau sees a secondary entrance to the house, but not from a street or alley. He nevertheless throws out this function nearly totally in favor of a role as reception space. The entrance function finds support in the certain DpS box plots and to a lesser extent in the average DW box plot. Its role as reception area receives some support only from the RRA and LS values.

Let us now turn to the large house B7. It has a rather ill-attested central part: later intrusions have erased much archaeological information. Major parts of spaces 1213, 2150 and 2151 are pure reconstruction. They are all labeled transitional by the excavator. Space 1213 is identified by the DpS box plot as a transitional space with some added support from the certain DpS and LS values. A strong alternative is not offered, although there is support for a living space function in the LS and RRA box plots.

The transitional character of space 2150 is supported by the RRA, DpS and area box plots. An alternative is offered by the DpS box plot: a living space maybe? The certain DpS values for workshops and kitchens are also in the same general range. Space 2151 is identified as a transitional space in the certain DpS and area box plots. The RRA box plot supports this somewhat. Finally, space 2511 in house G1 is considered transitional but also seems to have contained kitchen facilities. Only the certain DpS and average DW values support the transitional aspect. The other role as a kitchen is contradicted by the box plots. The bread oven or tannur is obviously located in a space which does not have the spatial characteristics of a kitchen.

We can get some more information concerning the so-called problematic spaces in another way. Only two spatial variables have a value for each space: certain DpS and area. In figure 10, I plotted the values for these variables against the values for the other spatial variables. In this case, the "problematic spaces" are indicated by capital letters which are explained in figure 11. I will summarize the results. The position of the letter "A," which is space 186 of house B1, points very strongly at a workshop function but also at an entrance function. Space 955 in the same house or the letter "B" does not seem to fit in well with any function. House B1's space 1052 gets equal support for an entrance and a reception role. Going to the next building, house B7, neither space 1213 or space 2150 (capital letters "D" and "E") can be categorized easily on the plots. Only "F," space 2151, seems to fit in a little with the transitional spaces. The evidence for house G1's space 2511 represented by the letter "G" is unequivocal: it ought to be primarily a transitional space and not a kitchen.

There is one more thing to consider. The excavator of Tell Melebiya proposed to subdivide the most complete houses according to 5 building categories. Category A is made up of small houses with a square plan and an oblong main space. Houses B6, C2, G2 and probably C6 belong to this category. Category B consists of the central-space houses, differentiated according to size. The small houses B2 and B3 are in category B1. Houses B4, B5, C1, G1 and probably C5 are of medium size: category B2. Finally, category B3 contains the large houses B1 and B7 with multiple central spaces. The abbreviations for all buildings, complete and incomplete, are explained in figure 13. Let us see if the incomplete buildings can be fitted into these categories by using notched box plots of the spatial variables. Also, maybe we can test the tentative membership of house C6 in category A and house C5 in category B2.

In figure 12, the building RRA box plot confirms the fact that the category B is different from category A. The same holds true for the mean DpS box plot. The mean certain DpS values don't show such a split however. The one category A house does appear different from the B categories. Category B2 is separated from B1 and B3. The house north of house B6 seems to belong to the B category while the area B northwestern quarter does not fit in with anything. The latter building might be a member of category B2 or B3 according to the mean average DW box plot. Finally, the mean area plot shows that category A houses tend to have smaller spaces than the B houses. House C6 leans toward the B category and house C5 toward the A category which contradicts Dr. Lebeau's proposal. Plotting twosomes of spatial variables did not yield much extra information because of the too small number of available values for houses C5 and C6.

In conclusion, I think that this spatial analysis approach has not proven entirely successful in the Tell Melebiya case study. The sample most probably was too small to allow for more consistent correlations and patterns to emerge. Nevertheless, spatial analysis is an extremely interesting addition to the tools we have at our disposal in the study of excavated architectural remains.


End notes

(1) M. Lebeau, Tell Melebiya. Cinq campagnes de recherches sur le Moyen-Khabour (1984-1988) (Akkadica Supplementa, 9), Leuven, 1993; the plans in figure 6 were adapted from this publication.

(2) This paper is part of my ongoing research involving spatial analysis of buildings:

- "A Study of Accessibility and Circulation Patterns in the Sin Temple of Hafagi from the Third Millennium B.C.," in Mesopotamia, 27 (1992), pp. 89-118
- A Spatial Analysis of Mesopotamian Buildings from the Late Bronze Age till the Parthian Period, Ph.D. dissertation, University of California, Los Angeles, 1994
- "'Spacings' and Statistics, or a Different Method to Analyze Buildings. A Test with Mesopotamian Houses from the Late Bronze and Iron Ages," in Akkadica, 89-90 (September-December 1994), pp. 1-8
- "A Test Study of Circulation and Access Patterns in Assyrian Architecture," in H. Waetzoldt and H. Hauptmann (eds.), Assyrien im Wandel der Zeiten. XXXIXe Rencontre Assyriologique Internationale, Heidelberg 6.-10. Juli 1992 (Heidelberger Studien zum Alten Orient, 6), Heidelberg, 1997, pp. 239-246
- "Discriminant Analysis by Means of Selected Spatial Variables Derived from Mesopotamian Buildings of the Late Bronze Age till the Parthian Period," forthcoming in Mesopotamia.

(3) See esp. B. Hillier and J. Hanson, The Social Logic of Space, Cambridge and Sydney, 1984.

(4) See among others: "Space and Community Behaviour: A Discussion of the Form and Function of Spatial Order in Settlements," in B. Lloyd and J. Gay (eds.), Universals of Human Thought. Some African Evidence, Cambridge and Sydney, 1981, pp. 71-110; "Identifying Spatial Disorder," in A. McNicoll, Taskun Kale. Keban Rescue Excavations Eastern Anatolia (British Institute of Archaeology at Ankara Monographs, 6 = British Archaeological Reports. International Series, 168), Oxford, 1983, pp. 193-241.


Figures

1 - Example floor plan and diagram
2 - Example floor plan and diagram after the change
3 - Example RA table
4 - Example of RA table after the change
5 - Spatial analysis formulas and abbreviations
6 - Floor plans of the most complete houses

House B1
House B2
House B4
House B6
House B7
House C2
House G1
House G2
7 - RRA diagrams of the most complete houses
House B1
House B2
House B4
House B6
House B7
House C2
House G1
House G2
8 - Abbreviations of the functions
9 - Distribution of the spatial variables for the spaces by function
Example of a box plot
Real relative asymmetry
Doorways per space
Certain doorways per space
Longest side
Shortest side
Average door width
Area
10 - Plots of selected spatial variable values for the spaces by function
Area & real relative asymmetry
Area & doorways per space
Area & certain doorways per space
Area & longest side
Area & shortest side
Area & average door width
Certain doorways per space & real relative asymmetry
Certain doorways per space & longest side
Certain doorways per space & shortest side
Certain doorways per space & average door width
11 - Abbreviations of the "Problematic Spaces"
12 - Distribution of the spatial variables for the buildings by building category
Building real relative asymmetry
Mean doorways per space
Mean certain doorways per space
Mean longest side
Mean shortest side
Mean average door width
Mean area
13 - Abbreviations of buildings

Figure 1 - Example floor plan and diagram

Figure 2 - Example floor plan and diagram after the change

Figure 3 - Example RA table

Figure 4 - Example of RA table after the change

Figure 5 - Spatial analysis formulas and abbreviations

Figure 6 - Floor plans of the most complete houses

House B1 (level 2)

House B2 (level 2)

House B4 (level 2)

House B6 (level 2)

House B7 (level 2)

House C2 (level 3)

House G1 (level 2)

House G2 (level 2)

Figure 7 - RRA diagrams of the most complete houses

House B1 (level 2)

House B2 (level 2)

House B4 (level 2)

House B6 (level 2)

House B7 (level 2)

House C2 (level 3)

House G1 (level 2)

House G2 (level 2)

Figure 8 - Abbreviations of the functions

Figure 9 - Distribution of the spatial variables for the spaces by function

Example of a box plot

Real relative asymmetry

Doorways per space

Certain doorways per space

Longest side

Shortest side

Average door width

Area

Figure 10 - Plots of selected spatial variable values for the spaces by function

Area & real relative asymmetry

Area & doorways per space

Area & certain doorways per space

Area & longest side

Area & shortest side

Area & average door width

Certain doorways per space & real relative asymmetry

Certain doorways per space & longest side

Certain doorways per space & shortest side

Certain doorways per space & average door width

Figure 11 - Abbreviations of the "Problematic Spaces"

Figure 12 - Distribution of the spatial variables for the buildings by building category

Building real relative asymmetry

Mean doorways per space

Mean certain doorways per space

Mean longest side

Mean shortest side

Mean average door width

Mean area

Figure 13 - Abbreviations of buildings


You can pay a visit to the author's home page

You can also contact him by e-mail

This web page was created by the author on a Macintosh computer.


[icon1] Symposium Index
St. Louis Aia Information Page
Comments to: Dr. Michael Fuller-mfuller@artsci.wustl.edu