5.2.3 Representativeness in the sample

The collection strategy was aimed at creating a representative and reliable sample. It was intended that the collection of data be 'even' in terms of geographical, chronological and site type coverage. Data from sites in Wales and Scotland was also sought (cf. Appendix 2.1), for comparison and completeness. Collecting representative samples is not necessarily straight forward in archaeology (!), since there are biases at various levels, such as where interventions via excavation and fieldwork have occurred and upon what types of site these have tended to occur (cf. Section 5.1.1.). Such factors impact upon the nature of the sample that can be a available to study. In terms of the Roman period archaeological investigation and reporting are not evenly spread but rather have a tendency to occur in 'drifts' geographically and with regard to site type. Paul Tyers' comprehensive map showing the distribution of sites with records of Roman pottery in the United Kingdom (1996, Fig. 49), for instance, whilst not definitive when produced and inevitably now a little out-of-date, shows a very marked density of 'sites' in lowland Britain, in contrast to Wales, the South-West peninsula and northern Britain, where such sites are more sporadic. On the other hand Jerry Evans has demonstrated that prior to the advent of PPG 16 the investigation of sites of the Roman period via fieldwork was characterized by a marked bias towards military and urban sites, while of those rural sites explored there was a disproportionate attention to those of apparent higher status, particularly villas (Evans 1995a). A number of factors therefore lie behind what samples may be available to a study such as the present one, and determines the content of the collected sample, which, despite the endeavour for even coverage, is subject to some skewing (as can be seen below in Table 1). Key determinants of the nature of the available sample include:

• the character of settlement patterns and military deployment in the Roman era
• the focus of older excavations and how the forthcoming pottery was reported
• the differential pace of modern development between areas
• the methods used to report pottery and samian, as well as publication procedures followed by Units

In practice these factors and others configure such that some site types are better represented than others and some areas have more information available on remains of the Roman era than others. Hence in terms of this Project there is more suitable data available from some areas (eg. Essex, Hampshire, Oxfordshire) than others (eg. Nottinghamshire and Derbyshire, the northern frontier and Wales (cf. Appendix 2.1)). Data was sought for all types of site, and in terms of chronological criteria it was considered that an approximate 'cut-off' date for samples should normally be c. AD 260, after which samian present in deposits was becoming residual (cf. Section 5.7). To obtain samples where full lists of samian forms represented occur, over three hundred publications were trawled during Phase 2 of the Project alone (this total includes 72 monographs publishing Roman sites, 220 national and county/area journals, plus unpublished project archive, forthcoming and 'Developers' reports).

5.2.4 Database: Its composition considered

The database includes 179 dated groups of samian. In all 7000 vessels are included from these groups, constituting a robust sample.

The version completed for Phase 1 of the project incorporated data for 51 phased groups, from 33 sites. This was subsequently enhanced with the inclusion of several more groups, particularly of the second century, in advance of the article published in The Archaeological Journal so that it extended to 59 dated phase groups from 39 sites, with values for 2050 vessels included. The enhancement of the database during Phase 2 should have resulted in the construction of a subtler tool with which to address refined questions.

Table 1 summarizes the constitution of the groups now forming the sample.

Table 1: The composition of the data: showing the nature of the available sample

It is apparent from Table 1 that a fair number of dated groups of samian are listed in the literature for most of the main site types. There are 41 groups from military sites and 38 from major civil centres (plus a further 3 more if one includes early coloniae). Similarly 33 groups from smaller civil centres were identified, and 27 from rural sites. Somewhat surprisingly though only 4 samples were available from villas. Despite the fact that this category of site received much attention in the 20th century, clearly few full lists of samian from such sites have been published. Various factors lie behind this, one being that many of the villas investigated date to the late Roman period, after the floruit of samian; in the case of Lulllingstone villa in Kent a full list is published but as a site collection, not in dated groups (Meates 1987; Simpson 1987). Other factors relating to villas are discussed below (cf. Sections 7 and 13); the only remaining aspect to note is that the total of vessels represented in the case of villas is skewed by the large number of vessels reported from Fishbourne during its Phase II. The striking aspect of the sample is that there is a trend in terms of the size of the groups coming from the main types of site. From military sites there are just three more samples than there are from major civil centres, though in terms of number of vessels represented the aggregate for military sites is double that from the major civil centres. One military group (from the fort at Castleford) includes some 529 vessels, but even if this unusual case were excluded the difference in numbers of vessels represented is very substantial. The average number of vessels in the groups from military sites is over 50, as it is for extra-mural settlements outside forts, and with the three groups from the early coloniae. The equivalent figures for the major civil centres are 35, for smaller civil centres the average is 24.5, while for rural sites it is 12. Hence despite the large set of variables that may impact upon the size of the groups and assemblages of samian recovered at sites, there is a clear pattern: the size of the samian samples recovered from sites descends as one moves down the settlement hierarchy. This reflects the varying frequency of samian amongst pottery assemblages from these sites (Section 7).

There are 17 useful groups from military sites associated with the northern frontier.

5.2.5 Database: Its uses

The database is designed so that subsets of data by, for instance, region and site type, can be selected for examination/interrogation, in addition to working with the aggregate figures for the whole sample. Hence the database has a strong degree of flexibility in so far as it is possible to select, filter and sort data by a variety of criteria such as: date, site type, samian source or form, and county/district, English Heritage (EH) region. It was anticipated during the design that (i) these criteria would be key variables, and (ii) users of the database might have samian assemblages that they would wish to compare with the general pattern of samian from within a region or with other sites of similar type (go to the ADS archive for downloading). Hence it is possible within this body of data to arrange sub-sets in order to compare 'like with like'. Hence any excavated and surface collected groups might be compared by project officers, specialists, etc., against both the overall 'normal pattern' of samian occurrence and supply for any given period and with the regional and site type 'norms'. This is desirable in so far as it enables, in principal, like to be compared with like. Observed similarities and differences between specific groups, and regional and site type patterns are thereby highlighted and can be evaluated. As demonstrated below there are systematic differences in the incidence of samian at different types of site, hence the initial comparison of a site group might be most appropriately made with groups from similar types of site of similar date, and perhaps from within the same region, rather than with the overall provincial pattern.

© Internet Archaeology URL: http://intarch.ac.uk/journal/issue17/1/5.2.3_5.html
Last updated: Mon Mar 7 2005