Why did I decide that? A first reason is that the colour is not informative of real differences. To go back to the previous example, the colour of the faience can change with time, thus the colour seen on the beads is not always the original one.
Another reason is that measurements can also change or cannot be immediately known. For example, a weapon can break, or a blade can become smaller because it has been sharpened time and again (which can also change the shape slightly). And finally, techniques, as well as the measures, are often not reported in the extant secondary literature.
Once I decided that I wanted to focus on the shape and the material of the objects, I had to decide how I wanted to define the shapes. That depended very much on the objects themselves. I did not considered objects that were too fragmentary to reconstruct.
For example, for the seal impressions, I considered the main feature of the design (e.g. floral, with spirals, with a frame). For the stone vessels I considered the shape of the rim and the neck, of the main body, and of the base: I regarded the vessels as similar only when they shared all of these features. Some objects these criteria were simpler: weapons I considered similar when they were of the same type (e.g. axes, swords, spear and javelin points, knives), shape, and material.
Another important decision that I had to make concerned the quantitative aspect of my analysis. Did I want to consider how many objects were found in each context (technically called “abundance”). For example, would it be useful to know how many stone vessels of a specific type were found in a single tomb?
Alternatively, did I want to consider in how many contexts – e.g. how many tombs – the objects were found at each site? Or, instead, did I want to take into consideration only if a type was present at a site or not (a simple binary choice), without counting either the abundance or the number of contexts?
I ended up choosing the third option, because of the data that I had to work with. The publications available did not mention consistently the abundance nor the exact number of contexts. Therefore, taking that into consideration would have skewed my analysis, giving more importance to the sites for which the abundance and the contexts are reported.
Moreover, the finds and the contexts known in the archaeological record do not represent all that was once present at a site. To go back to the example of the beads, it can happen that during the excavations some of them are overlooked. A few are also small enough to pass through sieves. Hence, the number of beads that are reported almost certainly are incomplete.
Another example is perhaps also illustrative. When excavating, it does not necessarily follow that what is unearthed is complete. The tombs that may have been dug up do not necessarily reflect the total number of graves that are actually present at the site. Therefore, counting the contexts would have again skewed my analysis, making the sites where more has been discovered – or where excavations were more intensive – appear more important than they actually may have been.
The next step was deciding how to internally divide the period under investigation. The period is dated in absolute terms from ca. 1850 to 1550 BC. It includes the later part of the Middle Kingdom (up to ca. 1775 BC), when there were precursors to the phenomena visible in the Second Intermediate Period, and the Second Intermediate Period itself (ca. 1775-1550 BC).
The separation between the Late Middle Kingdom and Second Intermediate Period is self-explanatory. However, the Second Intermediate Period also needs to be separated into an early part (pre-Hyksos rule) and a late part (contemporary with the Hyksos), because these two parts are characterized by different politics and by different relations between the various regions in Egypt.
After dividing the sites into the three main periods mentioned, I counted how many types of objects – that is to say objects of a similar shape and material – the sites of each period share with each other. This led to the creation of matrices, where the sites and the amount of types shared are reported: each row corresponds to a site, and each column corresponds to a type of object.
There is also another, more complex, type of analysis that I carried out, which considered the entire range of objects, not only the shared types. But this will be a subject for a future article.
To conclude, you can see that there are many factors that I had to think about when I started my study, and many choices to make that had profound consequences for my analysis and my results. I made my choices based on the data available – which I was collecting from publications – and by evaluating which data would actually be significant with regards to determining differences and similarities between sites. All these decisions were made to facilitate answering my main question.
With a different set of data, my choices would probably have been different. For example, if I had more data about the dimensions of the objects – and if that information were useful to my research – I would have included those too as part of the criteria for what constitutes a similar type. In that case, objects found in two sites would have been considered similar if their shape, material and their size were similar.
Before I leave you, I would like to say one more thing: though I didn’t include some of the data into the analysis, I put it into my database anyway. So, for example, when the information was available, I recorded in which contexts the objects were found, and what colour the objects were, or the type of scarabs to which some of the seal impressions belonged.
Writing the data down is useful for the sake of completeness. It will serve as a basis when in the future we will be able to include more data. Moreover, it gives a larger picture, which helps in interpreting the results of network analysis. Future researchers may also find it useful when the database is eventually released in Open Access.
The next article in this series will be about what happens when the matrix is imported in the software used for network analysis.