DATING.

A rigorous method was adopted to produce the pottery chronology (see Ch. 6). The sequence of introduction of types was independently derived from several sites where one or more stratified sequence was available. Close attention was paid to both the quality of the archaeological evidence and to the quantity of pottery being examined. In this way a series of chronological markers was found for each site (so that a group could be given a relative date of the form "after the introduction of type x but before the introduction of type y"). Further precision was given by quantitative analysis of stratified groups. The difference in the nature of assemblages (for example the time-span represented, the amount of residual pottery and random variation) means that too much emphasis on differences in relative frequency is misleading.

Nevertheless, it is possible with most wares to identify a point in time at which they first appear, a period in which they are most common and, more tentatively, a point at which the ware ceases to be used. The number of wares found in a sequence and the amount of variation between successive groups will determine how precisely the pottery found can be dated in other, stratigraphically unconnected contexts.

The same frequency curve used to describe the date of a ware can be used to describe its marketing at that site. One can illustrate the growth and decline of productions centres and even sometimes marketing policy by comparing frequency curves for the same ware at different sites. For example, where a growth in one area is marked by decline in another it cannot be explained as a variation in the output of the industry and must therefore be due to a change in the marketing of the ware.

Sequences were available for study at: Hereford (Shoesmith, 1980, 1982 and forthcoming a), Worcester, Sidbury (Carver, 1980), Droitwich (Hunt, unpublished), Gloucester (Hurst, H., 1972, 1974; Heighway et al., 1979, forthcoming), Winchcombe (Saville, forthcoming), Chepstow (Shoesmith, forthcoming b), Cheddar Palace (Rahtz, 1979), Bath, Citizen House (Greene, 1979), Trowbridge Castle (Smith, forthcoming), Newbury, Bartholomew Street (Vince, forthcoming b), Cirencester (Leech, forthcoming, and Wilkinson, forthcoming).

The crucial sequences at Bristol and Ludgershall Castle, with the exception of one group sealed below the motte of Bristol Castle (Ponsford, 1974), were not available for study. These sites form an interconnecting framework into which isolated assemblages can be slotted (see fig.1.1), assuming that they contain wares present in one or more of the sequences. There are nevertheless areas where no stratigraphic control is possible, notably N. Herefordshire/Shropshire and East Gloucestershire.

Detailed comparisons between the sequences found at these sites can be made, although they are too few and far between to give an accurate indication of distribution patterns. Since most of the extensively excavated sites are towns and thus the centres of marketing and transport networks it is possible that they may have received pottery from further afield than sites in the surrounding countryside. For this reason, data from poorly or completely unstratified sites have been used to fill in the gaps. Obviously in these cases the dangers of misinterpretation are greater than in the well-stratified sequences.