Spatial Methods for Analysing Large-Scale Artefact Inventories

Article excerpt


Archaeology is experiencing its own modest version of a wider 'big data' phenomenon, with the arrival of broadly disseminated, rapidly escalating quantities of geo-referenced digital information, gathered at spatial scales ranging from the microscopic to the truly continental. Sticking solely with national or trans-national inventories, some good archaeological examples are various large-scale databases for radiocarbon dates (e.g. Steele & Shennan 2000), archaeobotanical or faunal samples (e.g. Sherman & Conolly 2007) and cultural heritage sites (e.g. national monument records), to name but a few. Metal artefacts are slowly also receiving such treatment and three important and inter-related UK examples are discussed in more detail below. Part of this emphasis also reflects an increasing priority on 'open' data across all sections of academic research, commerce and the public sector, particularly with commitments to less restrictive copyright, interoperable formats and joined-up approaches to digital archiving, all again part of a wider trend in several countries (e.g. Bevan in press).

Archaeological datasets certainly do not represent the same order of computational and data-mining challenge raised by those gathered in other research sectors, where the flow of digital information from mobile devices, social networking sites, web browsers, medical records, remote sensing, etc. require innovative storage and management solutions even before they are analysed (this being the proper meaning of 'big data' as used elsewhere). However, the wide-ranging, geo-referenced and/or progressively licensed inventories that are now increasingly common in archaeology should still shift our goalposts and enlarge our interpretative ambitions. Small- and medium-grain studies will rightly continue to be a core archaeological focus, but there are now some new opportunities if we re-engage with the large-scale in a systematic way.

One major challenge that remains, however, is to find a set of robust methods that not only grapple with the inevitable issues of sample bias, but also go beyond such problems to achieve traction on more interesting questions relating to human behaviour in the past. This paper explores some ways to assess such data via spatial statistical methods. As case studies, it considers three important and related efforts at recording metal finds across England and Wales: the Portable Antiquities Scheme (PAS), the Celtic Coin Index (CCI) and the Corpus of Early Medieval Coin Finds (EMC). It begins by characterising some well-known biases present in the PAS dataset, but argues strongly that we can now work very productively with such large-scale inventories. Thereafter, the second and third sections of this paper consider two coin databases that are either almost wholly integrated into the PAS (CCI) or complementary to it (EMC), and emphasises: a) the importance of comparing artefact inventories with measures of contemporary regional demography, and b) the utility of spatial methods that explicitly account for differential recovery effects via 'relative risk' surfaces. A final section draws some general conclusions about where future analytical emphasis might be placed.

A national perspective on portable artefacts

The Portable Antiquities Scheme (PAS) is a national experiment in the voluntary recording of archaeological artefacts that have been found by members of the public. It is a highly innovative response to the UK's unusual historical and legislative framework for dealing with newly-found precious (and now base) metal antiquities, especially those discovered via the use of metal detectors (the use of the latter being prohibited in many other countries, see Bland 2005). The scheme has involved an advisory body of artefact specialists, a technical infrastructure for widespread digital dissemination and the work of 35-40 liaison officers in different parts of the country. …