QUANTIFICATION METHODS.

Several methods of quantififying pottery assemblages exist and all give rise to some problems. It has therefore been suggested that a minimum of two methods should be used in order to counteract the biases of the individual methods (Young, 1980). The methods are:

• Weight
• Sherd Count
• Volume
• Minimum vessel count
• Estimated vessel equivalents (EVEs).

Weighing is the simplest method and was used for certain assemblages in Gloucester and Berkshire. However, a set of portable scales was not available for use in the fieldwork and therefore weighing could not be used as a standard measure. The method is biased in favour of heavy types, such as storage jars and against light types such as cups. Since it is usual for a particular ware consistently to produce the same type of vessels this gives rise to a systematic error, which can therefore be corrected by introducing a weighting constant based on the average weight of a complete example of the type. In practice this could not have been done for more than a handful of types in the region due to the absence of complete vessels but it would have been possible to estimate the constant by dividing the weight of a group of sherds of one type by their

Sherd counts are subject to much more variablility than weight since sherd size depends on many factors. In practice the discrepancy between relative frequencies based on sherd count and those based on weight are not as great as might be expected (Hinton, 1977, 232-235). This is probably due to the fact that many factors (such as whether the pottery is crushed under foot or thrown into a feature) apply equally to all of the pottery in an assemblage.

The use of the volume displaced by sherds to quantify pottery assemblages was an attempt by Hinton to allow for differences in density between wares. It is unlikely to have any practical application and proved to have no advantages over weight and sherd count methods (Hinton, 1977, 233).

The minimum number of vessels in an assemblage is calculated by laying out all of the sherds and attempting to group them into individual vessels. The results of this exercise vary between groups and depend on a number of factors which affect the dispersal of sherds of a vessel after breakage. This method was used on some Gloucester assemblages and it is clear that where the sherds come from a small number of vessels the groups have to be larger than normal to produce 'reasonable' results. A systematic error will occur if the minimum vessel count is based on analysis of sherds by context when the results are later grouped together into assemblages. This is because vessels found wholly in one context will be counted once whilst those present in more than one context will be counted once for each context. Another source of error with this method is that it is possible to identify sherds of distinctive, decorated vessel when it is impossible to recognise sherds of the same coarseware vessel. Occasionally the difference between vessel counts and sherd counts is given in the gazetteer (Appendix 2). The presence of complete or smashed vessels in the Newbury Bartholomew Street sequence resulted in some anomalies in the frequency of wares, calculated by weight (fig. 1.2).

Estimated Vessel Equivalents are calculated by counting those parts of a vessel whose frequency on the vessel is known. This could be rim percentage, each vessel having 100% of rim, base percentage or perhaps applied features, such as bung-holes, feet, spouts or handles. The use of basal percentages is difficult with medieval material since in many cases the basal angle is indistinct and the percentage present is difficult to measure. Similarly, the use of applied features is only useful for specific tasks, for example to check whether a vessel type always has three feet one could calculate the rim EVEs and the number of feet. The number of feet could then be divided by three to give the foot EVEs and the two figures should then be of the same order of magnitude. However, no groups of sufficient size to use this method of analysis were encountered in the fieldwork. The use of EVEs avoids the bias introduced by using the minimum vessel count but in the process information is lost. As with weight and sherd counts it is not possible to say how many vessels are represented by the figures. This should make no difference when only relative frequencies are being compared but the minimum vessel count does give a measure of the reliablility of the data. Another method of ensuring the reliability of the data is used in the Department of Urban Archaeology, Museum of London, where only groups of over 25 EVEs or 10kg are accepted as sufficiently large for quantification. This test would exclude virtually all of the data in this thesis and has therefore been ignored. It does however point out just how little reliance should be placed on minor differences between assemblages.

RESIDUALITY.

One constantly recurring problem with the interpretation of stratified assemblages is that of residuality and intrusion.

A sherd is said to be residual if it is present in a context after having been previously buried. Other examples of old vessels being present in a context are excluded from this category and are often known collectively as 'heirlooms' although there may be many reasons for their great age at burial. There is always a certain lapse in time between manufacture and burial, even if it is too short to be detectable by archaeological methods. A calculation of the average age at burial of Roman samian ware vessels has been published by Orton and Orton (1975). Their estimates for two sites in the London area are c.18 years and c.29 years. The difference between the two dates is thought to reflect the speed of deposition. Similar data are available for post-medieval artefacts from the City of London (Vince, 1982a). Both estimates are probably much too high for the coarse pottery of the medieval period and lead one to suspect that if a difference in date between vessels in a group can be detected it is more likely to be due to residuality than to the use of the pottery over a long period.

An intrusive sherd is one which is present in an assemblage predating the manufacture of the sherd. Intrusion into sealed deposits can only be the result of poor excavation technique whereas there are many ways of residual pottery being incorporated into a perfectly excavated assemblage.

Most urban sites consist of a mixture of horizontal layers and features, mainly robber trenches and pits. The earth displaced by digging these features usually contains potsherds and if it is deposited on the site will result in that pottery being 'out of sequence'. Pit fills normally contain a mixture of contemporary rubbish, soil derived from the decomposition of organic refuse and earth thrown into the pit to cover the deposit. This earth will probably contain residual pottery. Large scale earth-moving, for example cellaring and terracing result in large quantities of residual pottery being dumped somewhere, usually at some distance from the site where the pottery was originally discarded.

On a smaller scale the techniques of excavation can give rise to residual pottery. For example, excavation of a horizontal layer often causes sherds in the exposed surface below to be dislodged. These are incorporated into the later assemblage. When excavating pits it is almost impossible not to over-excavate the edges of the feature, causing pottery to be dislodged from the strata through which the pit was dug. It is quite common to find stratified sites where the apparent sequence through time consists of more and more types being present but little decline in the relative frequency of earlier types. On some sites the relative frequency curve of a type can have two peaks, one for the actual lifespan of the type and another for the residual appearance of the type. Detailed analysis of residuality can be very useful in the analysis of excavation sequences but has hopefully been recognised and excluded from discussion in this study. However, when a site is the only one available in an area, or is the first site to be analysed then residuality may not be noticed, for example highly mixed groups are published from one of the first excavations in Abingdon (Parrington and Balkwill, 1976).

Sometimes there are objective differences between residual and contemporary pottery (see Ch.6) but often residual potsherds can be as large and unworn as the contemporary pottery in a group. In these cases it is only possible to divide a group into its constituents by reference to other assemblages, where it can be shown that certain types cannot be contemporary. Unless this can be proven the groups cannot be used for quantitative analysis.

Intrusive pottery normally enters an assemblage by one of three processes: intrusive features are not noticed during excavation; pottery falls back into the excavation area from a section and is mistakenly placed in the finds tray or there is a mistake in the recording of the finds context number. All these processes are symptoms of poor excavation technique but examples occur in most excavations. Urban excavations rarely proceed at a slow enough pace to ensure completely reliable results. Rural excavations suffer from having no stratified groups at all, due to the mixture of pottery through soil formation processes and agriculture. It will be clear that selection of contemporary assemblages for analysis is a complicated process and that suitable assemblages are rare.