Appendix I: The Methodology of the 1992 SARCP Pottery Analysis Workshop.

One of the most striking features of the excavations in rural settlement sites, conducted as part of the SARCP program, has been the very large amount of pottery retrieved. Not only has this created problems of storage, but also of data process­ ing.

As a means of handling these very large amounts of pot­ tery in a convenient way and yet getting the most out of them information-wise, a method of pottery analysis founded on the distribution of wares was applied.

The method was tested at a special pottery workshop or­ ganized within the SARCP framework, held at the National Museum in Colombo, March 1-10, 1992, and continued less intensively at another location until April 30th. We are most obliged to the management of the museum for putting a room at our disposal during the first intensive days.

At this workshop, the total pottery assemblage from the two large scale settlement excavations at Tammannagala 1990 and Ibbankatuva-Polvatta 1991 was analysed, in addition to the less bulky ceramic assemblage from the Mapagala excava­ tion of 1990 and the excavations at Dehigaha-ala-kanda in 1990 and 1991. Examples of diagrams are published with the accounts of Ibbankatuva and Mapagala. Here it will be suffi­ cient to give a description of the methodology applied.

A pottery analysis of the kind applied here is carried out under one absolute prerequisite: the excavation must be car­ ried out using a strict stratigraphical method, where the pottery is collected and registered according to stratigraphical units. If this is not done, a matrix chart according to the Hanis method cannot be put together, and without a reliable stratigraphic sequence analysis is impossible. Hence, the pottery from ear­ lier excavations at Ibbankatuva-Polvatta cannot be processed in this manner, since those excavators used the less accurate planum method for excavation. This means that the separate excavations at this same site can never be compared.

The objectives of the workshop were:

1) To contribute towards working out an operative descrip­ tive code for wares. We were working from the awareness that it is important to avoid dividing wares into too many varieties; the sub-division has to be operational for dating purposes. Attention was paid to earlier attempts at classifying wares in Sri Lanka.

2) To use the stratified pottery finds for establishing rough relative chronologies based on wares (which later can be elaborated by working with rim-forms). Those relative chronologies are subsequently being correlated with 14C datings.

3) To compare analyses of total pottery assemblages with analyses of separate sub-square assemblages (this was done on the Ibbankatuva-Polvatta pottery only) and to compare analyses based on weight with analyses based on the number of sherds. This was done mainly for the inherent pedagogical value.

4) To experiment with different ways of displaying results, using tables, graphs etc.

The advantage of the four chosen sites, besides their all having been excavated and recorded stratigraphically, is that work could be done with pottery assemblages that had a rela­ tively high ware variability, i.e. there was EHP1/EHP2 or even protohistoric pottery, preceeding the pottery of the Kasyapan period and later Anuradhapura period. In those later periods the ware variability appears to have been slight over a long period of time.

The pottery bags were sorted context-wise and in the order of stratigraphical sequence. Then each single context was processed from the bottom up. Sherds were sorted according to wares. Special forms were filled in with weight in grams and the number of sherds for each ware type. The completed forms constitute the data base that has been used in various ways:

1) The frequency of undifferentiated wares in each context was calculated: the total number/weight of potsherds from an excavation = 100%. The number/weight of each context = X% of the total. By doing this, we acquired knowledge of the nature of deposits in different layers, which might indicate differences in the intensity of occupation over time, as well as an indication of which layers could be suspected to consist of redeposited fill. In the latter case, a correlation to the ware analysis must be made. By this method, something about the stratigraphical sequence that would otherwise be invisible, can be learnt.

2) The proportion of different wares in the different con­ texts was calculated: the total number/weight from one context = 100%. The number/weight of each variety of wares = X % of the total. By doing this, an idea of how the uses of different wares have changed over time can be gleaned. We may find the whole set of wares in most of the contexts, but the impor­ tant thing is the change in emphasis for the different wares.

The data was then processed in a computer, using a pro­ gram for graphic representation. Two different types of graphs were produced: the calculation of the frequency of undifferen­ tiated pottery per context, was displayed in a simple bar-chart. Only contexts containing pottery were included. On the horizontal axis we show weight, or number of sherds, and on the vertical, the contexts starting with the stratigraphically ear­ liest at the bottom, according to the matrix chart.

The proportion of different wares in each context was displayed in another type of graph: also, in this only the con­ texts containing pottery are included. The display gives a graphic representation of the stratigraphical record along the vertical axis. Thus the earliest contexts are represented by a lying bar at the bottom of the graph, etc. The horizontal axis displayes the percentages of different wares. One bar here = 100%. The various wares are hatched or screened differently. On completion of all this, yet another advantage of the method becomes evident. It is now possible to check on the details in the matrix and its division into phases. Finally, it is possible to work out a second set of graphs, displaying phases instead of contexts. The reason for not being satisfied with a display of the contexts is, of course, the results we can obtain from the first calculation: the frequency of undifferentiated potsherds in each context. Redeposited fill can perhaps be sorted out if required, alluvial layers with very few potsherds may be merged with adjacent cultural deposits, to avoid too fragmented a picture, or if they represent a hiatus in settlement, sorted out or enhanced according to what is required, etc.

The method implies a lot of tedious work, but it is only hard data of this type that will move research forward, and computer technology is now making it possible to concentrate weeks of sorting, counting and weighing into one or two graphs which, within a second, will give the reader of a report a visual picture of the often very complicated stratigraphy of an archaeological site. That is what makes all the work worthwhile.