This series of blog posts is intended to give readers an overview of the contents, structure, strengths and biases of the Portable Antiquities Scheme database (PASd) of archaeological small finds. Further guides and advice can be found on the PAS website. Building on these, later posts will examine the PAS medieval finds records with particular reference to medieval markets, and to economic and commercial developments in England and Wales during the Middle Ages.
Arguably the most important field for all finds in the PASd, certainly for the purposes of searching and using the data, is the Object Type field. This is not a freeform field but rather limited by the guidelines set out in the Controlled Vocabulary section of the PAS website, itself based on the MDA thesaurus of archaeological object terms.
As of 1st June 2017 there were 1,266 different object types among 718,782 finds records (here we do not include the over 90,000 Iron Age and Roman coins and coin-related records added from the Celtic Coin Index and the Iron Age and Roman Coins of Wales databases in 2010, as they will have been subject to different recovery and recording conditions). From a user perspective this multiplicity of different object types presents both opportunities and challenges. One the one hand it potentially allows for a great deal of specificity in mining the database; on the other, closely related object types are sometimes recorded under sub-types that may not be immediately apparent to the untrained user. Finds related to Spurs, for example, may also be found under Rowels, Rowel Spurs, Spur Rowel and Prick Spurs. The search engine does, however, suggests similar finds entries on the search results page – on the right-hand side, under Object Type – which is always worth checking in case some of the finds you would be interested in are lurking under a different object type entry.
There is an ongoing project by PASt Explorers to group PAS object types related by their function and historic use into a handful of broad categories, which will be invaluable for users investigating the database for broad categories of finds such as dress accessories, weaponry or items related to commercial exchange. In our research project we are adopting a similar approach tailored for the medieval dataset, which will be discussed more in depth in a future post.
Here is a brief overview of the most common object types in the PAS database. Though at the time of the writing there are 1,266 object types in the PASd, in terms of finds numbers a small handful dominates. The below pie chart is a breakdown of PAS object types that represent more than 1.0 % of the entries in the database. Coins, at 44 %, are overwhelming the largest category. This no doubt reflects the interests metal-detectorists have in recovering and recording coin finds, as well as the ease by which coins can be identified as items of archaeological value relative to many other metalwork objects. This may be the case for tokens and jettons as well. Buckles, finger rings, brooches, strap ends, strap fittings and some mounts are similarly of interest as ornamental or dress objects. Weights and especially vessels are a large and rather miscellaneous category into which common metalwork objects fall. The small but significant number of lithic implements shows that though the vast majority of the PAS objects are found via metal-detecting, the database does contain important information going back to the Stone Age.
The fact that 1,253 object types account for the remaining 26 % of the finds has implications for the structure of the database that are important to keep in mind. The majority of these are in very small object type groups. There are 1,056 object types that have less than 100 entries, together amounting to only 12,300 (or about 1.7 %) of the finds. The drop-off in terms of entries per object types is quite sharp: 779, or the clear majority, of all PAS object types are associated with less than just 10 records. Depending on the nature of the data mining and analysis that one is conducting using the PASd these entries may be either dismissed as a minuscule rounding errors or hunted down and consolidated into larger object type datasets one by one.
Records in the PAS database are also divided into broad chronological periods. The broad period Medieval (defined as AD 1066-1540) captures the clear majority of the records that interest us in the PAS and Markets-project, but it should be noted that there are also several thousand finds the dating of which straddles the period boundaries from Early Medieval (410-1066) to Post Medieval (1540-1900). Nevertheless, if we concentrate for now on the bp: Medieval records, object types containing more than 1.0 % of the entries break down as follows:
There are broad similarities between the most common medieval object types and those of the full database – not surprising since at 185,290 records (and counting) the medieval dataset is the second largest after the Roman. At 33%, coins remain the single largest category by some distance. Then there are other object types associated in particular with the medieval period: seal matrices and harness pendants may possess a heraldic or personal pictorial device marking their original owners’ political, religious and social affiliation, and spindle whorls and thimbles relate to medieval textile manufacture. Altogether, in June 2017 there were 612 object types within the bp: Medieval chronological category, of which 74 contained more than 100 records and 22 more than 1,000.
PAS record entries contain a free-form description field with a considerably more detailed interpretation of the find and (usually) a high-resolution picture of the object. In terms of examining assemblages of limited size this means that the controlled vocabulary fields serve to provide general breakdowns of the material while object-specific information is also made available. But for coarse-grained regional and national-level analysis of multiple finds scatters the relatively small number of numerically important objects types means that there are challenges in differentiating various types of ‘productive’ sites or regions from one another. Does a particular scatter represent a market site, a manufacturing site, or is it an artificial spike in the finds pattern created by the fact that metal-detectorists enjoy high level of access to that area? One of our aims over the course of this project is to establish methodologies for investigating the PAS data for statistically meaningful patterns both spatially and temporally.