When we began transferring the mounds of Temple Mount material from the Kidron Valley dump to the Tzurim Valley National Park, we divided the dump into various areas. We suspected that the order of the removal of the earth from the Temple Mount and the location of its dumping may correlate somehow to the way it was excavated. We also separated the marginal areas of the material from the internal areas that had not been disturbed by the other illicit dumps in the Kidron Valley. The Temple Mount material was eventually divided to 11 areas that were removed separately.
Early in the Sifting Project, we already noticed that there were differences in the frequencies of certain types of finds from different areas. Moreover, similar finds, and sometimes fragments of the same object, were discovered within short periods of time. This suggested that these similar objects were originally next to each other. But the full significance and value of dividing the material at the dump into different areas was discovered only last summer, during the processing of quantitative data for the Third Preliminary Report which we recently published. We found that artifacts which we assume to be from the same context were also distributed in a similar manner. Another example is that we found that artifacts which can be identified with the Horses of the Crusader era Templar Knights were distributed in a similar way among the dump areas.
We concluded from this that we can define a statistical distribution “fingerprint” for each artifact type. Artifacts that have a similar “fingerprint” may have originated from the same context. The statistical technique for finding such relationships and verifying their statistical significance is called Cluster Analysis.We will not go into a detailed explanation of this technique, but we can foresee that at the completion of the classification and sorting process of all the different types of finds that we have, we will be able to apply this technique on a unified data table of all the finds. The results of this analysis will show clusters of finds having similar distributions. These clusters may also represent a similar context of the finds within them. Currently, we are still investigating the application and implication of this method. Only after finishing the classification and sorting process, will we be able to create a full data table that will be adequate for such an analysis, and then we will be capable of fully estimating the value of this method. If we are able to achieve valuable information from this type of analysis, it will be a substantial innovation in archaeological method and theory research which could also be applied by other archaeologists who focus their research on excavations of fillings or site surveys.
We can illustrate this idea using the following example:
Suppose we prepare a salad using four vegetables and two cutting boards. On one board we cut cucumbers and tomatoes, and on the other carrots and onions. The vegetables on each cutting board are thoroughly mixed and placed in a large bowl. They are then lightly tossed in the bowl. Such a mixture will result in the vegetables being scattered unevenly throughout the salad. It can be assumed that the distribution of vegetables that we cut and mixed on each board will show a similar distribution within each of the various areas of the salad. Let’s further illustrate this with the following table:
Board 1: 21 cucumber pieces and 11 tomato pieces (32 total pieces)
Board 2: 6 onion pieces and 12 carrot pieces (18 total pieces)
Mix the cut vegetables well on each cutting board and then combine them together in a large bowl. The vegetables in the large bowl are lightly tossed and then its contents are divided equally into 4 smaller bowls. This procedure may yield the following data table:
It can be seen in the table that the distributions of the carrots and the onions within each bowl are similar, but differ from the distributions of the cucumber and the tomatoes in the corresponding bowl and vice versa.
This is the value of “cluster analysis.” By observing the percentages of various types of finds within each area, we may be able to determine which types of finds originated from the same context.