An analysis of the Roman coins lost on different types of site within Roman Britain shows strong chronological and geographical trends, as well as providing a critique for the current classification of sites.
Finds in Roman archaeology have a hierarchy. For many years the only relevant finds were inscriptions. These were collected and published in the Corpus Inscriptionum Latinarum and, as a result, any new finds could be evaluated against a background of general knowledge. Coins are also inscribed and they have therefore had some attention in the past, but it has only been in the last 30 years that John Casey and I have been able to get across the idea that the coins found on any site in a region will reflect first the coin supply to that region, and only secondly, and less strongly, the coin loss on the individual site. The British pattern of coin finds is now reasonably well known (Reece 1987) but, in a sense, knowledge of that pattern has hindered attempts at more detailed interpretation by suggesting that the provincial pattern will always mask the coin-using events on any individual site.
This idea of a coin pattern was used by Geoff Marsh for pottery, and, running true to hierarchy, was applied to the 'best' pottery, samian ware (Marsh 1981). He was able to show that fine pottery occurred on many sites in similar chronological patterns of plenty and scarcity, and properly suggested that this reflected the pattern of import into Britain rather than an astonishing historical congruence between phases of success and depression on very different sites. Finally the depths have been plumbed by Chris Going (1992) who has demonstrated to the boredom of most Romanists, and the discomfort of a very few specialists, that coarse pottery in Britain was produced in waves or cycles of activity. I find his paper one of the most exciting published for many years because it tried to relate pottery production, the economy of Roman Britain, and the place of Britain in the Empire, to general ideas of economic cycles and the real world of interacting units of production and consumption. In 1991 I published a provisional summary of coins from 140 assorted sites in Britain (Reece 1991, referred to as 140 sites) so that the material which had appeared from excavation, but had not been published, should be available for general use. I base my ideas in this note on those 140 sites and no others simply so that anyone who wishes to follow the numbers further has only one printed source to find. There are a number of other sites whose coins have been published in very varied sources and these can be used to test the ideas formulated here. To the basic numbers of the 140 sites were added a few pages of 'work in progress' and it is a surprisingly comprehensible fragment of further work in progress that forms the substance of this note. I hope that it moves the process of the interpretation of material forward, and that the ideas may once again trickle down the finds hierarchy.
The 140 sites were published in five main groups -- large towns, villas, temples, military sites and other rural settlements not further classified. The large towns included London, the certain Civitas or Tribal capitals, two possible capitals, Ilchester and Water Newton, and the Coloniae. Villas, temples and military sites were usually clearly differentiated, and the rural group ranged from settlements of over 100 acres (40 ha) down to small groups of huts. Because some of the rural settlements were larger than some of the tribal capitals the dividing line taken was that of government hierarchy. Well before the publication of 140 sites it had become clear (Reece 1987: 88-97) that not all of those five groups were homogeneous. Most town sites followed a set pattern, and were therefore called Good Towns, others, notably such as Cirencester, did not. Because they were clearly towns from the administrative point of view. but did not follow the majority pattern of coin finds, these have become known as Bad Towns. …