The nature of the transformation of Western Eurasia from the “ancient world” to the beginning of the medieval one has long been debated. With the development of scholarship, the types of the evidence used to attempt to answer this question have continuously expanded. The classic studies on this subject, from [Gibbon 1776] to [Dopsch 1923] and [Pirenne 1922], based their arguments to a great extent on the sparse literary and legal sources that survive from the fourth through the eighth centuries CE. At the beginning of the twenty-first century, two major works brought material culture into the conversation: [McCormick 2001] Origins of the European Economy: Communications and Commerce, A.D. 300-900 and [Wickham 2005] Framing the Early Middle Ages: Europe and the Mediterranean, 400-800. McCormick cited the importation of Arabic and Byzantine coinage into Europe c. 750-900 as one of the catalysts of the development of a distinctive European economy, but otherwise ignored earlier monetary developments and was more anecdotal than systematic in his use of the evidence of coin finds to illustrate commercial relations.[1] Wickham avoided McCormick’s Eurocentric approach and used a meticulous study of ceramic manufacture and distribution to establish patterns of continuity and rupture of exchange around the Mediterranean. He was explicit in his intention to eschew numismatic analysis, explaining his decision in a footnote that expressed hope that future studies would examine coinage more carefully.[2]

The outlines of the changes in minting from antiquity to the Middle Ages have long been clear. In the fourth century CE, the Roman Empire, which controlled the Mediterranean and beyond in antiquity, had been using a trimetallic system of gold, silver, and bronze in the Mediterranean world, while its Sasanian (Persian) rival to the East used a currency that consisted almost entirely of silver. By the eighth century, monetary economies in the region morphed into new forms and standards. A silver-based monetary economy developed in Europe, a bimetallic one of gold and bronze in Byzantium in the northeast Mediterranean, and a trimetallic system in the extensive lands of the Umayyads in the greater MENA (Middle East and North Africa) region. The extent and volume of the circulation of the coins that these economies used can be determined only from a careful compilation and study of modern coin finds. Although digital tools facilitate the analysis and visualization of the resulting patterns in the literature, an understanding of the underlying forces shaping this data is necessary. Only a critical reflection on the processes that led to the deposition, excavation, study, and compilation of late antique coins can help overcome the biases inherent in these processes and contribute to the historical narrative of the period.

FLAME was formed in 2013, in the context of the annual Stone Lectures at Princeton which Chris Wickham delivered. It began as an explicit attempt to take Wickham’s aforementioned footnote to illuminate the economic history of western Eurasia from the fourth to eighth century by assembling and presenting numismatic data. From the beginning, its creators conceived it as a resource database, that is, making the data available to others online but also meeting and managing the expectation of a broader audience than its creators.[3] In retrospect, the project’s inception influenced several subsequent biases.[4] Although FLAME began with the necessary academic expertise in numismatics, represented by Princeton’s numismatist, Alan Stahl, who came up with the idea for the project, the early stages of the project did not include any digital technical expertise and included rudimentary and unstandardized Excel files. Then-graduate student Lee Mordechai “grew” into the role of technical director over time. As a result, FLAME developed organically and benefitted from formal expertise in the digital humanities only after a couple of years. Most initial members were postdocs and graduate students affiliated with Princeton at the time. Project development was inconsistent until the hiring of Mark Pyzyk as full-time project manager in 2020. The subsequent regular discussions in the project’s weekly meetings were crucial in thinking about and fleshing out the biases underlying FLAME, as well as attempts to resolve them and how to communicate them to FLAME’s audience. FLAME’s leadership defined the temporal delimitation of the project to c. 325-750 CE and divided the project into two consecutive stages that would examine the simpler question of coin minting and the far more complex problem of coin circulation.

As FLAME operated for several years without funding, its leadership prioritized rapid data collection which would both serve as a proof-of-concept for subsequent funding, as well as producing results that its many non-tenured members could use in their research. In retrospect, this decision introduced and amplified existing (but then unrecognized) biases. For example, each member of FLAME was given responsibility for a defined region of western Eurasia. They worked independently of each other, coordinating with the leadership. This accelerated data gathering but probably also sharpened differences between regions, compounding the problems in regional biases (see below). Although many additional contributors have joined FLAME over the years, the rationale underlying the project’s spatial division remains an organizing principle and the project’s temporal limits have not moved. From an early stage, the project decided to focus on a few key variables relevant to its driving questions and ignore others. For example, FLAME records a coin’s metal but ignores its weight, as well as descriptions of text or images it might bear. This was another decision that expedited data gathering, but it also resulted in fewer variables for cross-checking and made it more difficult to address some of the project’s subsequent biases. A key subsequent decision, taken a few years later, was to import data from other digital sources - again greatly increasing the amount of data in FLAME but introducing its own biases (discussed below).

Today, FLAME’s application provides comprehensive data to its users while also highlighting the many gaps in current scholarly knowledge. It encourages historians and other users to be more cognizant of these unknowns and their implications, fostering a more methodological and transparent approach to the study of the economic aspects of the transition between Late Antiquity and the Middle Ages. This paper focuses on the methodological and theoretical contribution of FLAME to the broader discourse in the digital humanities.

FLAME experimented with several different data structures before settling on its current format. FLAME’s core is a relational database (SQL Server). Project members add to and edit the database through an Admin Panel website, developed in JavaScript. The public version of this website includes a list of possible values for FLAME’s key variables such as metal, denomination and mint.[5] Automatic tests within the Admin Panel highlight potential issues with existing entries — for example, coins with dates outside the project’s range — which require additional attention from project contributors and administrators.

FLAME’s users interact with a graphic user interface (GUI). The GUI is based on a map, developed in ArcGIS and JavaScript, on which the project’s data is displayed. A side panel allows users to run advanced searches and place filters on the data, with most results displayed on the map. Additional results may appear in tabular or graph formats. FLAME’s design encourages exploration of the results through customization. For example, when visualizing the coinage output of a mint over time in a histogram, users can color-code different elements based on the metallic content of coins minted in that location; the data behind the histogram can then be exported in several formats (e.g. CSV, XLS).

FLAME has imported data from other projects and datasets. Some of these projects are openly accessible online. Others consist of data collected by individuals, who then decided to donate it to FLAME. The decision to import data resulted in an increase in the size of FLAME’s database by an order of magnitude but also led to multiple challenges in using and maintaining these data (see below for biases). Each of these datasets is idiosyncratic in its coverage and data structure, so FLAME converts the data to its own format using custom-built scripts. Since these other projects and datasets are continuously updated but have no API (Application Programming Interface), the process is undertaken on an ad hoc basis, using custom-written scrapers and data transformations, leaving the resultant data unlinked except for a reference URL on FLAME’s side for users wishing to follow entries back to their source. For this reason, FLAME keeps a record of the dataset, from which the information is derived. Since the data is not linked according to a wider linked open data standard, the same data points could appear in two different datasets with different names. As digital numismatic projects operate autonomously, two or more databases could and often do end up containing the same coin finds but digitized and classified them according to different criteria, depending on the idiosyncratic goals or structure of these projects. To evaluate potential duplicate data, FLAME developed an additional program that suggests possible duplicates, which are then checked by project staff.[6]

Beginning in parallel to FLAME’s public launch in 2021, the project has begun to place more emphasis on User Experience (UX) to make the site and web applications more user-friendly. FLAME’s leadership has periodically observed and interviewed potential users — specialist and non-specialist academics, as well as some students. These interviews have resulted in changes to the public-facing user interface (UI), as well as the development of new features. The emphasis on UX developed further through collaboration with the University of Houston’s SYRIOS project, a National Endowment for the Humanities-funded project that illustrates the history of ancient Syria through coinage.[7]

Within the typology of digital humanities projects, FLAME shares some of its characteristics with other projects that work with material culture, and more specifically projects that focus on numismatics.[8] As such, FLAME’s development coincides with that of several other digital projects concerned with ancient numismatics. These projects were generally developed organically as digital versions of traditional reference publications that include as much data as possible that conform to certain criteria. Thus, for example, the Online Coins of the Roman Empire (OCRE) represents a catalog of all coin types minted in the Roman Empire until 491 CE and was based on the Roman Imperial Coinage (RIC) series of reference works. Other projects include the Coin Hoards of the Roman Empire (CHRE) project, which aims to collect all the information about coin hoards in the Roman Empire and beyond between c. 30 BCE and 400 CE, and the Portable Antiquities Scheme (PAS), which includes archaeological objects found in England and Wales — mostly finds made and reported by the public, including coins. Both these projects collected new material and had no published precedent, but were nonetheless perceived as digital versions of printed catalogs.

FLAME shares some of its core features with these projects. For example, like FLAME, both the CHRE and PAS projects include a GUI, but unlike FLAME, both CHRE and PAS place more emphasis on their data than their visual representation. Both are also similar to FLAME in their data structure, separating data at the level of individual coins and finds of coins found within a common archaeological context. These approaches emerged organically in different projects, resulting in different architectures. A subsequent collaborative ontology, nomisma.org, was later established in an attempt to provide stable digital representations of numismatic concepts between these and other numismatic projects. Projects such as OCRE and PAS have organized at least some of their data in a Nomisma-compliant manner, although the many technical challenges have slowed collaboration, including with FLAME.[9]

Other digital numismatic projects focus almost always on a single political entity (e.g., the Roman Empire) and/or a limited region (e.g., France). FLAME, however, attempts to push the discussion forward by aggregating and harmonizing data from different sources during a time of major change in its database. Other numismatic projects facilitate answering numismatic research questions, such as classifying coins into types, attempting to provide digital catalogs of coins that share some variable, or finding patterns across coin finds.[10] FLAME, on the other hand, has as its core a historical question — the nature of the Late Roman to Medieval transition — that is primarily concerned with change over time whether through coin production (i.e. how many coins were produced) or coin circulation (i.e. which coins circulated where). This has heightened our sensitivity to data biases — regional, temporal, and compositional with respect to coin find contents — at the level of project design. While our digital numismatic colleagues are all well aware of these topics (we are under no illusions here), FLAME has been incentivized by the nature of its starting questions to dig deeper into these factors: how, otherwise, could we claim to accurately pursue our main question?

Much of the bias in FLAME’s data emerges from regional differences rooted in modern national or institutional factors, such as law, cultural organization, national historical memory, or political circumstance — to say nothing of economic reality. These national differences strongly influence the discovery, preservation, and publication of archaeological objects. The absence of a substantive rule of law can be ruinous to the cultural heritage of a region. The most common form that this takes is the looting and selling of artifacts. Such trade in looted goods has been regulated internationally since 1970 but with unequal implementation.[13] Those artifacts that are not effectively controlled make their way through various channels to an enormous market of collectors. In a global context, the vast majority of countries are significantly affected by artifact looting. Even countries like France, the United Kingdom, or the United States, which more effectively regulate objects found within their borders, serve as an endpoint for artifacts illicitly exported from other nations.[14] Ongoing warfare can also degrade or destroy a country’s cultural heritage. Thus, regions such as Syria, Libya, and more recently Ukraine have suffered serious loss of heritage objects and structures.[15]

The circumstances particular to each region can be easily observed on FLAME’s map of coin finds. Thus, FLAME’s users might be surprised to see that Italy and Turkey at present contain considerably fewer coin finds in FLAME than England and France. Italy and Turkey were far more central components in the late antique monetized macro-economy and therefore we would expect them to be better represented. That this is not the case is probably the result of modern regional biases, rather than any ancient pattern of circulation and deposition. FLAME has attempted to lessen such disparities, for example, through the establishment of coordinated teams in such under-reported areas.

Numismatists with knowledge of these regions will be mostly unsurprised by this. Most are aware that archaeological and scholarly circumstances are more favorable in some places than in others. The question is why. Though a tempting guess, it is not always a matter of wealth. Italy, for example, is a wealthy country with a dynamic academic culture and a strong interest in excavating and publishing archaeological materials on its territory. And yet we find comparatively little published material there compared to some other countries. On the other hand, Romania–significantly poorer today than Italy, but even more so under communism–is among the best-excavated and published numismatic contexts in Europe for late Roman coins. This was because its communist government was strongly invested in establishing Romania’s “Roman” roots and devoted enormous resources to supporting archaeological work, including coin evidence, that furthered this ideological objective.

FLAME has approached this problem by investigating why a given region looks the way that it does in FLAME. Answers are rarely simple and often require considering the legal, political, and historical contexts that have shaped the disciplinary histories and coin publication records of the region. These factors have remained largely invisible to non-experts and seldom discussed in publication, even among specialists.

One of the best examples of such regional bias can be found in the United Kingdom, which shows the influence of politics and law on the shape of modern scholarship not only externally, in relation to other countries, but internally as well. This bias explains the stark differences in coin finds reported in England and Wales versus Scotland and Northern Ireland (discussed further below).

The main force behind this distortion has been the Portable Antiquities Scheme, a government program operating since 1997 within the parameters of several national heritage laws but in particular the Treasure Act — which deals with archaeological finds of precious metal objects.[16] The PAS regulates and (sometimes controversially) incentivizes the discovery and reporting of archaeological objects found by private individuals. While individuals are required by law to report significant discoveries of more than a single coin if gold or silver, or more than ten base metal (i.e. bronze) coins, the PAS framework facilitates the process, preventing some of the more important archaeological objects from being funneled into gray and black markets. The objects are first inspected by local authorities — Field Liaison Officers (FLOs) — who analyze, categorize, photograph, and record the objects in an online database .[17] Under most circumstances, the reporters will receive the objects back, at which point they are free to do with them what they want - and many do sell their finds to museums or on the market. This process is arguably a great incentive for them to report their finds or at least removes significant disincentives to doing so. The PAS is therefore a point of contrast to the many countries where found artifacts are considered the property of the state, with no (or limited) compensation offered to finders. The data gleaned from the process is published by PAS online, freely and openly. This has meant huge gains in coin find data for England and Wales in the years since its adoption.

This bias is obvious when comparing England, a peripheral and minor part of the Atlantic-Mediterranean macro-economy for much of the period FLAME covers, with Mediterranean countries. Thus, England and Wales’s 1355 coin finds far outnumber those of core economic regions in antiquity like Italy (467), Turkey (265), and Egypt (74). This has much to do with the structure of reporting in the UK, as well as the resources — and the legal enforcement efforts — devoted to making the system work. As we mentioned earlier, because Scotland and Northern Ireland do not participate in the PAS, the publication of coin finds discovered on their territory lags significantly behind England and Wales. This leads to a significant distortion in the British data encoded in FLAME; at the moment, FLAME records almost nothing about coin circulation in the United Kingdom outside of England and Wales. The British case study therefore demonstrates the extent of some of the biases underlying intra- and international variations within FLAME’s data, biases that would remain unknown by most of the project’s users.

The regional biases discussed above are a modern phenomenon. But other biases lie much closer to the materials themselves and to the ancient economic activity of which they were a part. What we define as primary biases derive from lived experience in Late Antiquity and the early medieval period. Since we cannot fully reconstruct lived experience in the past, many of these biases are theoretical. Therefore, we must use proxies to answer questions about, e.g., the differential loss of coins. The few scholarly attempts made to understand the effects of primary biases on the modern numismatic landscape (i.e., the coins that scholars are left to work with in the present) have generally based their studies on theoretical models derived from modern economic phenomena.[18] The vast differences between societies over time, however, obstruct such an analogy. For example, there are obvious issues inherent in comparing a monetary economy, where transactions and taxes were paid in precious metal coins, to one where payments are primarily in credit and most cash denominations are bills.

In the pre-modern world, coins served as means of exchange and holders of wealth; they were issued in three metals (alloys based on gold, silver, and copper, respectively) and in various denominations (standards of fineness and weight holding a fixed value in the monetary system, e.g., in the contemporary United States, coins of 1, 5, 10, 25, or 50 cents). At various times and places in antiquity, certain metals and denominations were available, corresponding to different sectors and levels of economic activity. In general, gold-based coinages were used for extraordinary, costly transactions (e.g., purchases of land) and constituted the principal medium for storing wealth. Copper-based coins were used for everyday transactions. Silver-based coinages, between gold and copper in value, were used for salaries and middle-range purchases, and could also facilitate day-to-day exchanges and larger purchases.

Virtually all ancient and medieval coins available for collecting and study have spent most of the time since their minting underground. They may have been buried intentionally or accidentally. The chief reason for intentionally burying coins in antiquity was that there were few alternatives to save wealth securely. Pre-modern societies rarely had any form of deposit banking, and so individuals or institutions usually put saved coins along with other precious objects into a vessel, burying them somewhere they could be dug up when needed. Such “hoards” frequently comprised the highest-value coins available to whoever hid them. Hoarding was relatively common in the pre-modern world, though perhaps more so in times of perceived insecurity (e.g., war). If the burier failed to retrieve the coins, they might come down to us as a hoard. Other intentional contexts for coin deposits would be the burial of individuals or other ritual offerings. Such cultural practices lead to biases in the coins that survive up to our times — for example, the clear bias towards high-value gold and silver coins in most hoard finds.

Other coins ended up in the ground through a different mode — accidental loss. The vast majority of coins found in excavations and single finds — which represent random accidental losses — are copper. The simple act of accidentally dropping a coin during a transaction or travel would have been the first step in this process, but the likelihood that the person would have recognized the loss and retrieved the coin is far more likely in the case of high-value coins, especially in antiquity where a single gold coin could be worth hundreds of bronze coins. Also important is the fact that copper is often less bright in color than gold or silver. As artificial lighting was rare, many transactions took place outdoors in dusty or muddy locations that were close to the color of copper. Such coins were less likely than gold or silver to attract the attention of passersby. Another factor in the predominance of copper coins in archaeological contexts is their generally fiduciary nature; that is, their value was dependent on the state’s commitment to convertibility into precious metal coinage. In cases of demonetization, for example as part of a coinage reform, such coins would be worthless and might be casually discarded. Within this context, silver coins issued within a tri-metallic system might be significantly under-represented in modern samples since they were noticeable enough to be observed and retrieved when lost (as opposed to copper), but not valuable enough to be well-represented in hoards (as opposed to gold).

The second class of bias focuses on the archaeological context in which coins are found, as well as the various natural and technical processes at play in a coin’s survival over time and its subsequent discovery. Excavations are the most common legal context from which buried coins are intentionally retrieved. Within excavations, coins are valued as chronological indicators for a particular archaeological layer. Most excavation coins are presumed to be the result of unintentional loss and so, as sketched out above, are usually copper-based and small. They are often in bad physical shape owing to wear (since they were used daily), direct contact with corrosive environments (e.g., soil), and the greater susceptibility of copper to corrosion compared to silver or gold. These factors tend to coincide with each other, often corroding coins to the point that they are unreadable. A case in point is the minimi of the fifth and sixth centuries CE, a designation for coins of about 10 mm in diameter, often poorly struck in the first place. The categorization of such coins as unattributed or unidentified, — if they are recognized at all in the course of excavation — is a source of bias on sites, especially when dealing with strata associated with Late Antiquity. On the other hand, prominent features on certain denominations might allow for partial identification. For example, the sixth- and seventh-century Byzantine 40-nummi bronze piece is quite large and heavy and carries a large M which is easily identifiable even if other features are illegible.

A further source of bias among coins from excavations is the process of site selection by archaeologists. Some of these are limited by definition - for example, built-over areas in modern cities generally allow for only limited archaeological excavations. Moreover, Late Antiquity, the period FLAME investigates, has traditionally attracted less scholarly and public interest than e.g. Greco-Roman antiquity, a period that remains the focus of many excavated sites, or sectors within larger excavations. In FLAME’s case, this probably contributes to the sharp reduction in coin finds between the beginning and end of the period, with which the project is concerned. This imbalance in the results of excavations often stems from ideological interests (for example, demonstrating the level of civilization or identity of certain site inhabitants). This might operate at the level of the archaeologists, those granting funding or permits for excavations, or both. For instance, it is common for excavations in Greece or Turkey for much of the 20th century to bulldoze (often literally) through levels associated with medieval or late antique remains, usually to reach Hellenistic and Classical layers, which scholars, funders, and governments considered more prestigious. This affected not only what came out of the ground, but where it went (into a bucket or a museum showcase), as well as how it was preserved, cataloged, studied, and published.

Besides excavation contexts, coins are often discovered in hoard contexts. Hoards are typically defined as two or more coins thought to have been buried together, often in a vessel or other container, which exhibit a radically different pattern of coin preservation. As hoards were usually buried in secret, they are often found outside of habitation areas excavated by archaeologists. In the nineteenth century, hoards were frequently discovered by farmers as the result of adopting deeper plows or through exploitation of areas that had been abandoned since the Middle Ages. The expansion of urban areas – with their deeper digging for infrastructure (e.g. building foundations, sewers, utilities, etc.) – as well as transportation routes (e.g. roads and railways) resulted in more hoard finds as well. In more recent times, the use of metal detectors has greatly increased the rate at which such finds are made (hence the phenomenon of “detectorists” in the United Kingdom, who are a major constituent of the aforementioned Treasure Act and PAS policies). Whether such coins are brought to the attention of government authorities or local numismatists varies greatly depending on legislation and policy, and is a major source of bias affecting FLAME’s database and circulation map, as discussed above.

A third class of bias derives from the scholarship upon which FLAME draws, in particular its unevenness in key variables such as the number of publications (e.g. articles, archaeological reports) or the ratio of excavated-to-published coin finds. These biases are also largely intractable as they stem from broader trends in scholarship and publication practices (e.g., who studies what? Who publishes, where and for what reasons?).

In light of modern academia’s pressures on scholars’ time, researchers must decide how to allocate limited time and attention. This touches upon, but is not the same as, the archaeological biases discussed above. Scholarly bias may take the form of prioritizing certain academic activities over others. Archaeologists, for example, often prioritize excavating sites over publishing the results. Because of this, many sites that are excavated are not — and likely will not be — published. A 2001 survey of archaeological excavations in Israel revealed that fewer than 20 percent of excavations carried out in the seasons between 1970–1989 had published their results.[19] Similarly, the major salvage excavations in Beirut’s city center over the 1990s and 2000s had at least 15 different teams excavating over 200 sites. Only a handful of these have been published to date.[20]

Similar decisions are made within excavations themselves. For example, as discussed above, researchers decide to excavate certain strata and features and not others based on ad hoc criteria. In the case of Beirut, the published excavation sites bulldozed the layers above the destruction layer of 551, making it extremely difficult to understand post-551 Beirut. Even after the excavation of a site, the numismatic material is often challenging for the average archaeologist to deal with. Although the ideal solution is to bring in a numismatic specialist, in reality many coins remain unpublished because such a specialist is unavailable.

Relatively few hoard finds of coins are published after their discovery and many of the publications lack key details like the mint of origin and denomination of the coins. Coins are often deposited with regional archaeological services or local museums; their contents can be reconstructed only by an inspection of such an institution’s accession records or its physical storage facilities. Many, perhaps a majority in some countries, of discovered hoards are immediately dispersed on the collectors market, with no record made of their contents.

A further bias derives from the notorious decentralization of numismatic publications, an existing problem exacerbated by FLAME’s scope. Reports of finds can appear in local and relatively inaccessible journals, excavation reports, edited volumes, festschrifts, and journal articles, among others. Many of these are not systematically used for disseminating numismatic research. This phenomenon already makes it difficult to locate all these reports, so that FLAME’s information is more representative of the most accessible publications rather than the whole of numismatic scholarship. It also favors English or other standard research languages (e.g., German over Armenian), often published by Western-oriented presses.

There is, in general, no comprehensive centralized resource to track coin finds, even at the national level. Such attempts have been attempted in the past with various levels of success — for example Kropotkin’s survey of Roman and Byzantine coin finds in the Soviet Union or Militky’s equivalent inventory in the Czech Republic.[21] The latter example indicates that additional biases intrude at the sub-national level, with certain counties or regions being better covered than others (Prague, for instance, is hugely overrepresented among Czech coin finds).[22] Even the PAS reports of British finds include only material discovered since its inception in 1997; earlier finds have to be added from traditional sources. While there are a few aggregated numismatic publications that aim to report on all recent numismatic papers,[23] they are rarely comprehensive and tend to briefly survey an enormous number of diverse publications, leading to relatively low resolution.

Finally, the unequal products of scholarship resulting from the biases above contribute to biases in reference works – the catalogs and other methodological apparatus most scholars use to identify a coin or better understand a numismatic process. As a general rule of thumb, the more coins are available for study (through publication, ideally), the better the resulting product. There are other factors as well — for example, the popularity of certain topics over others. While good catalogs for Roman imperial coins (e.g. aforementioned Roman Imperial Coinage series) have appeared from the 1920s, their Byzantine equivalents date only to the 1970s. Comprehensive catalogs of Sasanid Persian coins date to the 2000s and are still incomplete today.[24] Other types of coinage — for example that of central Asia — are barely covered by any scholarship.

The biases listed above mean that FLAME better represents — in both quality and quantity of information — areas and periods with good reference works such as the Roman Mediterranean compared to areas that are less well covered such as central Asia.

Previous sections have described biases in FLAME's data at the level of coinage itself, as well as regional contexts affecting the discovery, preservation, and publication of coin finds. These are, in essence, exogenous to FLAME. There are, however, biases inherent to FLAME itself that involve how the project is structured, institutionally and operationally.

Most FLAME researchers are volunteers who provide the numismatic data for regions about which they have some expertise. In general, they work alone, coordinating their activity through project representatives. The participation of such individuals does not, generally, arise in a vacuum, and is structured by pre-existing individual and disciplinary contexts. Many early participants, for example, were Princeton University graduate alumni. Subsequently, volunteers might have seen a conference presentation on the project given by a FLAME representative, or they might have been put in touch by someone with knowledge of the project. Such weak ties follow familiar paths, generally through North American and Western European universities.[25] Many contributors were first visiting scholars at Princeton or had passed through programs such as Dumbarton Oaks' numismatic summer seminar (co-taught by Stahl) or that of the American Numismatic Society. Disciplinary and institutional networks play a critical role in the shaping of such work, which favors some regions over others. For example, the authors of this article, who have constituted the leadership of FLAME, all specialize in aspects of northern Mediterranean and European scholarship. When FLAME has attempted research on coin finds in regions where its leadership has fewer ties such as North Africa, the Indian subcontinent, or East Asia, it has run into problems recruiting individuals who might help. These are limitations we have sought to address, with varying degrees of success.

In certain cases, enough researchers studied a particular region that it made sense to organize them into clusters working under a regional coordinator. Up to this point, five regional clusters have come together to work on FLAME. This has included groups focused on the Anatolian peninsula, the Iberian peninsula, Southern Italy, Ukraine, as well as a group working at the University of Hamburg focusing on early Islamic coinage. That the last group departs from the typical regional focus is a result of scholarly specialization — most FLAME scholars focus their research on pre-Islamic (e.g., Roman, Byzantine, etc.) or Islamic coins but not both categories.

At both the individual and cluster levels, biases affect the data structurally and, occasionally, more or less at random. Structurally, FLAME’s biases intersect with the modern secondary biases that we identified above. Thus, some countries are better able to support individuals and groups of scholars or have particular national priorities that favor the publication and reporting of coins, such as the Romanian example cited above. These factors affect the potential for collaboration with scholars from these countries.

There are likewise biases that FLAME imported from other projects (with biases of their own) through their datasets. Examples include the Coin Hoards of the Roman Empire (CHRE), the Portable Antiquities Schema (PAS), Antike Fundmünzen in Europa (AFE), the Early Islamic dataset (EI), and the Thessaloniki dataset (THS). [26] It is safe to assume that all these projects suffer from many of the same biases we have described above. The extent to which they identify or publish information on these biases varies but is often minimal, and so their specific project biases are largely outside of our capacity to address.