FLAME has gathered datasets from several different sources, all of which are described below. The composition and objectives of these datasets determines the results of any query of the database. Therefore, critical users the FLAME database and application must take into account these different biases as they are built into FLAME’s structure. Although the biases are unavoidable, users can mitigate them to an extent by excluding or including specific datasets from their analyses.

FLAME’s core data

This dataset was compiled by FLAME’s participant members, in particular at the beginning of the project’s Circulation phase. Members were (and are) mostly junior scholars ranging from junior faculty members through postdocs and to graduate students. Each member received a territory that corresponded with their expertise; territories were defined ad hoc and based on informal communications between members. Within these boundaries, members were asked to collect as much information as possible about coin circulation in late antiquity. The results varied based on member engagement, knowledge of past studies, and extant (and accessible) publications. FLAME’s leadership made accessible to the project members a bibliographic list of c. 1000 items complied from relevant references from key numismatic and economic publications. The extent to which members made use of this more standardized compilation depended on their personal judgements.

The dataset was produced through participants inputting their data through digital forms designed for this purpose. These digital forms required specific types of input in certain fields, forcing a greater level of internal standardization than in some individual projects. On the other hand, the forms were also less tolerant of the ambiguity that characterizes many numismatic publications.

PAS UK – Portable Antiquities Scheme – UK

This dataset was received through the PAS, a British initiative that allows individuals who find portable antiquities, chiefly coins, to keep them and provides governmental support in recording these objects. The PAS has been operating for a few decades and has cataloged well over a million objects. All the project’s data is made available to the public, with the exception of the precise find locations, which are masked to keep the privacy of the finders and protect archaeological sites. The PAS is the largest online open access repository of ancient and medieval numismatics, with hundreds of thousands of coins already cataloged, with new ones appearing daily.

The major spatial bias of the PAS (the project is limited to Britain alone) is immediately visible in FLAME’s app – the number of coin finds and the number of coins found are extremely high in Britain compared to the rest of western Eurasia. This gives the impression that Britain was an economic center although in reality it was a far periphery of the Roman center, and was considerably poorer than the empire’s core territories. Because its method of composition is essentially a random sampling of finds in Britain into the modern period, the PAS does not contain a chronological bias.

To import this dataset FLAME director Lee Mordechai wrote a script that converts the data from the PAS format into the form used by FLAME. The PAS data was broader and more detailed; superfluous fields of data were removed. Different spellings for mints, denominations and rulers were standardized to FLAME’s version. Potential years for minting the coins were calculated based on ruler regnal years where possible, or within broader limits based on the available data. FLAME also converted the PAS system for locating finds to coordinates, albeit with some loss of precision.

The script for converting PAS data into FLAME’s form can be accessed here.

CHRE – Coin Hoards of the Roman Empire

The CHRE project aims to catalog all coin hoards from the Roman Empire, defined as between Augustus and the late fifth century CE. The project was created through partnerships in which partners at the national level in their respective countries contributed data about coin finds in their territories. Using this method, the CHRE was able to compile a respectable dataset covering hoards across western Eurasia.

Due to these reasons, the CHRE imported dataset into FLAME contains multiple important biases following its mission. First, it includes only hoards – and not excavation reports or individual finds. The emphasis on hoards creates an additional bias in favor of precious coins, as these were more likely to be preserved in hoards. Another bias is the spatial distribution of the coin finds – which likely reflects the biases in excavation, publication, data accessibility and levels of engagement among the project’s partners at the local level. Finally, the project’s chronological limit (i.e. late fifth century) is another bias that gives more weight to coin circulation in the fourth and fifth centuries.

To import this dataset, FLAME director Lee Mordechai wrote a script to convert the data from the CHRE format to one that fits FLAME. Superfluous fields of data were removed. Different spellings for mints, denominations and rulers were standardized to FLAME’s version. Potential years for minting the coins were calculated based on ruler regnal years where possible, or within broader limits based on the available data.

The script for converting the CHRE data into FLAME’s format can be found here.

AG THS: Andrei Gandila’s Thessloniki Dataset

This dataset was privately compiled by FLAME’s member Andrei Gandila, who generously agreed to share his personal data with FLAME. Gandila compiled the dataset for an article that covered the circulation of the mint of Thessaloniki over late antiquity.

The Thessaloniki dataset is unique in that it is near-comprehensive based on published scholarship, with a few additional museum collections. It is likely more balanced with regards to eastern Europe as many of Gandila’s coin finds come from countries in the region and were published in languages that are not traditionally used for international research. At the same time, the dataset’s emphasis on Thessaloniki introduces an important spatial bias in the Balkans. The particular intervals in which the mint of Thessaloniki operated determine the chronological bias of the dataset. Finally, it is important to note that Gandila surveyed well over a thousand different coin finds, but cataloged only those coins that were minted in Thessaloniki, creating a major bias when looking at many of these finds independently since they are not completely cataloged.

To import this dataset, FLAME’s director Lee Mordechai wrote a script to convert the data from Gandila’s idiosyncratic format to one that fits FLAME. Non-standard forms were harmonized while working together with Gandila. A particular issue in importing this dataset was the simplification of some of the spellings of place names and references, converting special characters to their simpler equivalent in the Latin alphabet (e.g. ‘i’ instead of ‘î’). The script for converting the Thessaloniki dataset into FLAME’s format can be found here.

PP EI: Peter Philps’ Early Islamic dataset

This dataset was privately compiled by Peter Philps, who generously contributed it to FLAME after hearing about it as a project.

The Early Islamic dataset covers the Middle East in particular. It is spatially biased towards finds in Israel, where there are more detailed publications, while other areas where there were less archaeological excavations of the period and/or detailed publications such as Syria, Jordan, Lebanon and Egypt are underrepresented. The dataset also introduces a chronological bias – almost all coins in it are from the seventh and eighth centuries. Finally, the mints of many coins in the dataset is vague, hence the many references to uncertain mints in different locations.

FLAME’s director Lee Mordechai worked with Peter Philps to standardize the dataset as a first step; as a result, FLAME added several new mints that were not represented in its earlier versions, and introduced a distinction between different Arab-Byzantine coin types. In addition, a substantial amount of early Islamic coins could not be identified with a single minting year, leading to more vague estimates in that regard as well.

The script for converting the Early Islamic dataset into FLAME’s format can be found here.