Uneasy bedfellows? Roman archaeology and quantifications (early 20^th cent – ca. 1960)

In the first half of the twentieth century, data-based methodologies were unsurprisingly not widely integrated within Roman archaeology. Nonetheless, a tendency to systematically collect and collate information can be observed. In Germany, between 1894 and 1937, the Reichslimeskomission published 15 sections (550 km in total) of the Uppergermanic–Rhaetian Limes (ORL) (Moschek 2011: 30). In France, the first edition of the Carte Archéologie de la Gaule was published in 1931 (Couisson 1931). In Britain, from the 1920s onwards, Osbert Guy Stanhope Crawford pioneered quantitative methods in the shape of distribution maps and aerial photographs (Bowden 2001: 38–43). In the same period, Cyril Fox (2010) published his doctoral thesis on the archaeology of the Cambridge region. The Roman period listed eight Roman Towns, Earthworks, etc. and canals as well as 15 rural sites (Fox 2010: 159–236).

After the middle of the twentieth century, a possible paradigm shift can be observed in Roman archaeology in Britain and Europe. In his assessment of post-war Romano-British archaeology, Martin Millett names the growth of archaeology outside of academia as the defining characteristic of this period (2015: 27–33). The increase in excavations, both in quantity and in size, will have led to an exponential increase in archaeological data.

A similar assessment is made by Paul-Marie Duval in his overview of Gallo-Roman archaeology in post-war France, naming the creation of the CNRS (Centre national de la recherche scientifique), founded in 1939, and the institutionalisation of archaeology along with aerial photography and subaquatic archaeology as two new fields which contributed substantially to the understanding of Roman Gaul (Duval 1989). Aerial archaeology contributed substantially to a growth in the dataset for Iron Age and Roman rural settlements in northern France. A growing interest in pottery studies—especially terra sigillata—not only from an art-historical perspective but also as a mass-produced and widely traded commodity, can also be registered. Both the archaeology of rural settlements and pottery studies were topics which would become synonymous with Roman archaeology in western Europe.

In a chapter titled ‘Short History of the Beginnings of Mathematics in Archaeology’, the French prehistorian François Djindjian notes especially the years between 1950 and 1965 as the period when quantification truly began to permeate archaeological methods, listing a dozen papers which contributed to this movement (2015: 66). From a theoretical point of view, the emergence of the Annales school of thought in France in the second quarter of the twentieth century, with its focus on interdisciplinarity, long-term change, and overarching social phenomena, can be seen as further anticipation of data-driven research—as many of the questions posed by the l’école des Annales could be operationalised by systematic data collection and analyses (Bintliff 1991: 1–32; Schnapp 1981).

New Roman archaeology? (ca. 1960 to the turn of the millennium)

In archaeology in general, the 1960s saw the publication of landmark essays and volumes which would leave permanent marks on the discipline. The rallying call of New Archaeology (Binford 1962; Clarke 1973) and the subsequent responses have been discussed in depth elsewhere (e.g. Malone and Stoddart 1998).

In France, the rebranding of the Centre d’Analyse Documentaire pour l’Archéologie in 1958 by Jean-Claude Gardin saw computational archaeology make a quantum leap in terms of methodology and theoretical underpinning (Moscati 2016). In the subsequent years, the centre would focus on objects, iconography, and texts and would be at the forefront of the discussion on the integration of computing methods into more classical archaeological topics.

In Germany, the discussion revolved less around the quantifying of archaeological source material and gravitated more towards the interpretation. Hans Jürgen Eggers, the chair of prehistory in Hamburg, published his Introduction to Prehistory in 1959, which rallied against the narrow interpretation of material culture as a marker for ethnicity (Eggers 1959: 255–297; von Carnap-Bornheim 2001: 179–197) and made the case for stronger methodological reflection (Theune 2006: 328).

The impetus for New Archaeology was slow to arrive in Roman Archaeology. In an article titled ‘A Classical Archaeologist’s Response to the “New Archaeology”’, Stephen Dyson (1981) reflects on why ‘New Archaeology’ penetrated so superficially into classical archaeology. One reason he offers is the primacy or even dependency of the subject on literary texts and canonised works: ‘Living with Virgil, Homer, or the Parthenon is rather like having a dominant senior partner. One has trouble speaking one’s piece’ (Dyson 1981: 242). In his review of the role of classical archaeology in the journal Antiquity, Nicola Terrenato strikes a similar chord: the mastering of ancient languages and a knowledge of key works is the—sometimes arduous—prerequisite before one’s own original research questions and methods can be developed (Terrenato 2002: 1109). The same can be said for Roman archaeology north of the Alps: the discussion of the accounts of events by ancient authors such as Caesar and Tacitus continued to dominate the discourse, even if the archaeological body of evidence for the periods and regions in question has continued to grow in quantity and quality.

How did Roman archaeology deal with these ever-increasing large datasets? A look at the early conferences organised within the scope of the CAA (Computer Applications in Archaeology) shows that with very few exceptions, Roman themes were not really treated until well into the 1980s (Laflin 1982). The creation of the Italian journal Archeologia e Calcolatori in 1990 (Cristofani and Francovich 1990)¹ was a further step in this direction.

In many European countries, the anchoring of archaeology within a legal framework and the subsequent growth of excavations in size and frequency meant that many scholars were taught archaeology not only in the lecture hall but also in the field (for England, see Fulford and Holbrook 2011; for Germany, see Heising Unpubl.). The hands-on aspect of excavation archaeology shifted the interest towards managing the sheer quantities of varying archaeological evidence (Laflin 2013: 7–30), and by 1993, Richard Reece was not so much criticising the lack of data-driven research in Roman archaeology, but more specifically the lack of theory with which the data was to be interpreted (Reece 1993).

The creation of the Archaeology Data Service (ADS) at the University of York in 1996 aimed to ‘collect, describe, catalogue, preserve, and provide user support for digital resources that are created as a product of archaeological research’ (ADS 2017). Almost 25 years later, it has become an indispensable research and archiving facility for archaeology in the United Kingdom (Richards 2017).

Finally, in his stimulating essay from 2004, Greg Woolf wrote of an ‘information explosion’ in the Roman archaeology of the past few decades (2004: 417–418). Almost 20 years on, the shockwaves of this explosion show no sign of abating, and the data is becoming richer, more varied, and—yes—bigger.

Digital Roman archaeologies: from the turn of the millennium to the present and beyond

In the last 20 to 25 years, archaeology and cultural heritage have embraced the digital turn.

Personal computers have become faster and more powerful but also smaller and more mobile, opening up new opportunities to gather and disseminate gigabytes of data in the field, in museums, archives, libraries, and (home) offices. Likewise, technological advances have created a surge in possibilities for connectivity, leading to speedier rates of data transfer.

As a consequence, new pathways have opened up for digital archaeology to explore and expand on: 3D scanning and reconstruction, neural networks and artificial intelligence, and remote sensing are only some of the fields which make up the world of digital archaeology of today (Figure 1).

Figure 1

Artifically generated image created with “midjourney” and using this paper’s title as the prompt (Source: Author).

The creation of research frameworks and infrastructure enabling the linking and exchange of (meta)data is another hallmark of the last decades: the ARIADNE project, part of the EU-funded DARIAH network ran from 2013 to 2017 (and to 2022 under the name of ARIADNEPlus) and ‹aims to bring together and integrate the existing archaeological research data infrastructures so that researchers can use the various distributed datasets› (DARIAH-EU 2012).

The launching of new data infrastructure tools and portals shows no sign of slowing, with databases for Roman-period sites and objects moving online at a speedy pace. The list of open databases on the Project Mercury homepage lists 42 datasets offering information for Roman-period material (Project Mercury 2018). The website of the Oxford Roman Economy Project (OxREP) links to seven datasets which were created between 2005 and 2010. Likewise, image databases and digital journal repositories are also expanding at a phenomenal rate, yielding access to information in hitherto unwitnessed quantities (OxREP 2012).

The establishment of university courses, research labs, and even departments for digital or computational archaeology further underlines how integral the digital has become to the discipline. From a ‘Romanist’ perspective, the courses of digital archaeology offered by universities such as Cologne² and York³ are especially interesting, as these archaeological institutes/departments also offer courses on classical and Roman-provincial archaeology, which could lead to innovative synergies.

The literature on ‘digital archaeology’ in its broadest sense has thus multiplied exponentially in the last few years. Multi-authored volumes, conference proceedings, e-publications, and journals all contribute to the publication landscape of this dynamic subject.

In 2017, the journal Archéologies numérqiues⁴ was established in France. A year later, the Journal of Computer Applications in Archaeology was launched.

Despite the growth of publications of digital archaeology, papers covering case studies from the Roman Empire or methodological innovations working with Roman-period datasets are still in the minority. This marginalisation seems almost mutual. A brief look at the tables of contents of the journals for Roman archaeology in Europe and the United States of America (Germania, Berichte der Römisch-Germanischen Komission, Mitteilungen des Deutschen Archäoloigschen Instituts, Gallia, Journal of Roman Archaeology, American Journal of Archaeology) over the past decade reveals that data-driven research has not truly taken off in these periodicals and their respective research traditions.

Finally, similar to the transformations taking place in digital humanities (Klein and Gold 2019), a salient call has been made for ‘a form of introspective or more self-aware Digital Archaeology, one which consciously seeks to understand the underlying processes and behaviours that sit behind the tools, technologies, and methodologies applied’ (Huggett 2015: 89) or to be more critical of the processes which lead to the structuration of archaeological source material (Hacıgüzeller, Taylor, and Perry 2021: 1711–1714).

Pottery studies and Samian Ware

Data—i.e. the systematic quantification and measuring of fragments and vessels—has always played a significant role in Roman pottery studies,⁵ even if it was not always designated by this specific term. Due to its standardised repertoire, chronological sensitivity, and popularity within the Roman world and beyond, the study of Samian ware is closely intertwined with quantitative methodologies. Although early forays in Roman pottery studies in the eighteenth and nineteenth centuries were concerned with the motifs on the decorated bowls and the names on the stamps in the late nineteenth century (Pinon 2015), the potential of the quantitative study of terra sigillata began to be realised, and data-driven methods were quickly adopted.

The introduction of Hans Dragendorff’s seminal essay on terra sigillata from 1895 in the Bonner Jahrbücher states that ‘every Roman shard found outside Italy is very directly a document of cultural and commercial history’ (Dragendorff 1895: 2). Without specifically naming it, Dragendorff was demanding and possibly anticipating some kind of database.

In 1904 ‘Les vases céramiques ornés de la Gaule romaine (Narbonnaise, Aquitaine et Lyonnaise)’,⁶ Joseph Déchelette (1904) repeatedly uses tabular overviews to illustrate quantities of various motifs and of production centres. At the 1969 Fautores Conference in Speyer (D), a much-discussed paper was submitted with the title ‘Computer-aided identification of fragmentary potters’ stamps’ (Zabehlicky-Scheffenegger 2013).

In 1971, Elisabeth Ettlinger’s review of Christian Goudineau’s analysis of the pottery from Bolsena (I) underlined the link between the quantity of black gloss ware (‘Campana’) and terra sigillata on one hand and the chronological development of Late Republican/Early Augustan sites in the western Roman Empire on the other. Ettlinger’s (1971) tables show how, in the space of around 15 years, in one of the excavated zones, the ratio of Campana to terra sigillata changes from 20:1 to 0.3:1.

From a methodological point of view, Clive R. Orton’s work is of great importance. His papers and publications address not only how to carry out statistical analyses in detail but also why and to what end (e.g. Orton 1980: 15–24). Inspired by Orton’s work, Armand Desbat published a succinct yet important paper in the 1990 proceedings of the Société Française d’Étude de la Céramique Antique en Gaule (Desbat 1990). Here, again, the focus is on the method—i.e. how pottery can be counted and what this means for statistical analyses. A conference on Gallo-Roman pottery held at Bibracte and published in 1998 went one step further and collected 18 papers by different authors on pottery quantification. The final paper is a ‘protocol’ of the results gleaned from the discussion and shows in no uncertain terms how Roman-period pottery can be not only quantified but also published (Arcelin and Tuffreau-Libre 1998).

It is not only pottery forms or types which have been systematically collected. Even before the First World War, Auguste Oxé started to collect all potters’ stamps of Arretine ware; it would take several decades until this undertaking was finally published by Howard Comfort in 1968 (Oxé and Comfort 1968). Philipp Kenrick’s revised and enlarged edition of this opus magnum, consisting of 36,000 entries, advanced the state of research on early terra sigillata considerably (Oxé, Comfort and Kenrick 2000). Although it is a pity that the database has not been migrated to a new platform (Kenrick 2014). the dataset has been integrated into the ‘Samian Research’ portal of the Römsich-Germanisch Zentralmuseum of Mayence (Samian Research 2016).

The acknowledgement of pottery’s role in the dating of archaeological structures, in answering economic questions, and in addressing topics concerning food and drink consumption led to a growth in pottery studies. The 2018 edited volume by Penelope Allison, Martin Pitts, and Sarah Colley titled ‘Big Data on the Roman Table’ sets the scene for further studies in this field by implementing data-driven analyses to explore trade and social networks associated with terra sigillata. Future studies on the social behaviour of terra sigillata and on Roman eating and drinking practices would do well to consult this methodologically innovative and theoretically grounded collection of papers on the subject.

Most recently, the Römisch-Germanisch Zentralmuseum im Mayence has been creating a Linked Open Data infrastructure for terra sigillata stamps incorporating data from reference works and connecting them to dated military sites on the frontier in order to establish a more robust stamp chronology (Thiery, Mees, and Gottwald 2020).

Zooarchaeology

In its early years, the investigation of faunal remains concentrated on ahistorical periods, mostly the Paleo-, Meso-, and Neolithic (Gifford–Gonzalez 2018: 19–50). In Roman archaeology, it was again in the decades after the middle of the twentieth century when bone assemblages from Roman sites began to be collected, measured, analysed, and published. In Germany, the faunal remains from military sites, such as the extramural settlement of the fort of Butzbach and the Augustan supply base of Rödgen were published in 1960 (Habermehl 1960) and 1962 (Habermehl 1962), respectively. In Austria, at roughly the same time, a report on animal bones from the Magdalensberg was published (Ehret 1964). In France, early zooarchaeological reports can be found in Marseille (Jourdan 1976). In Britain, a couple of years later, assemblages from London and Exeter were published (Maltby 2019: 11).

The research questions tied to the data were primarily concerned with the diet of the social groupings studied (Davies 1971). Other aspects, such as bone-working (Crummy 1981) and ritual (King 2005: 329) gained traction quickly. The development of the Ossobook database first at the University of Basel (Schibler 1998) in 1990 using dBASE and from 2007 at the Ludwig-Maximilians University of Munich (Kaltenthaler and Lohrer 2018) with its own open-source framework (‘xbook’) allows for standardised, international collaboration based on common criteria of animal bones. This programme lays the foundation for biometrical analyses, which further our knowledge of the size variations of species before, during, and after the Roman Empire.

It is exactly this type of standardisation which allows for large-scale studies. For example, Anthony King’s paper from 1999—’Diet in the Roman World: A Regional Inter-Site Comparison of the Mammal Bones’—shows, using an impressive dataset collected from published site reports, how meat consumption within the imperium romanum could vary from region to region and raises questions regarding what extent Roman imperialism may have impacted foodways and agricultural practices (King 1999).

A more recent database project, consisting of 200 rural sites mainly from northern Gaul, shows how cattle gained importance between the second and the first centuries BC (Méniel et al. 2009). The past decades have seen zooarchaeology going beyond working with single databases, with the compilation and comparison of data from different sources and databases now being pursued systematically. For example, the ZooMWest project is tracing trajectories in animal husbandry in different regions in the western Mediterranean (ZooMWest 2017). With the project’s wide geographical and chronological scope, its integration and publishing of large datasets and statistical analyses, long-term (i.e. from the Bronze Age to Late Antiquity) trends in the development of livestock and their interactions with ancient societies can be reconstructed (see for example Nieto-Espinet et al. 2021).

In theory, the integration of multiple datasets seems like a fantastic opportunity to enlarge the sample size of a study and to broaden its scope either regionally and/or chronologically; however, in recent years several methodological issues have come to light regarding the comparability of source material coming from different sources or analysts. Training, experience, access to resources for comparison, and financial and time constraints are all factors which can influence the identification of specimens within bone assemblages, and which can generate biases within the published data. While such distortions cannot be completely nullified, it should be at least possible to pinpoint possible sources of errors: Explicit explanations of the methodologies chosen, of the criteria implemented for the identification of the bones as well as the publishing of images of uncertain identifications, have been put forward as viable recommendations to render datasets more transparent, reusable, and sustainable (Lau and Kansa 2018: 37–38; see also Atici et al. 2013 for a case study and recommendations on working with legacy data). A further step for the interconnecting of databases is the creation and implementation of common ontologies as exemplified by the tDar initiative of the Digital Antiquity organization (tDar 2011).

Quantitative zooarchaeology is a good example as regards to the potential and challenges of the interoperability of multiple databases and the required standardization of the processes involved in recording and publishing. Inevitably, in recent years, critical voices have rightly been raised. In his introductory chapter for the edited volume Zooarchaeology, Umberto Albarella acknowledges that while data comparability is paramount, a certain amount of leeway should be tolerated when it comes to the analysis itself. ‘The study of a zooarchaeological assemblage represents an intellectual undertaking rather than a mechanical collection of data’ (Albarella 2017: 9). One immediate reaction would be to state that there should not be a binary choice between standardised recording and creative interpretation. A minimal amount of normalized data can be expected in the shape of a database; other, more specialised measurements and their registration can be optional.

Numismatics

Due to their standardised production, Roman and, more recently, Iron Age coinages have lent themselves to quantification and statistics, paving the way for their use in big-data projects. An early example of coin quantification for chronological discussion and socio-political significance can be seen in the Swedish numismatist Sture Bolin’s visionary thesis from 1926 on coin finds in Germania libera, in which he painstakingly collected several thousand coins. (Bolin 1926; Bolin 1929). Konrad Kraft’s essay from 1956, ‘zum Enddatum des Lagers von Haltern’, continued in a similar vein (Kraft 1955/56): his method, determining the ratio of issues of types Lyon Altar I to Lyon Altar II in Augustan sites, is still used today (Wigg-Wolf 2018: Fig. 2). ‘Coins and the Archaeologist’ is an early showcase for implementing statistical analyses and data visualisation on coins (Reece and Casey 1974). Pie charts, bar graphs, distribution maps, and cluster analyses all feature in the papers.⁷

From the 1970s onwards, the concept of cliometrics began to permeate Roman studies. Cliometrics can be described as ‘the application of economic theory and quantitative methods to the study of economic history’ (Diebolt and Haupert 2021: 1) and was developed in the 1950s by US economic historians (Williamson and Whaples 2003). For the Roman world, the impetus came from ancient history, with scholars such as Keith Hopkins (e.g. 1980), Elio Lo Cascio (e.g. Lo Cascio and Malanmina 2014), and Willem Jongman (e.g. 2014) using data-driven methods to answer questions on economic growth, trade networks, and demography. A debate can be observed on how ancient cliometrics can be integrated into the wider discipline of ancient history (Diebolt 2011). Despite many reservations (e.g. Greene 2006), archaeology could help to bridge this gap, as it is a discipline which has a long tradition of working with fragmented data and models.

Moving away from large-scale statistical models and focusing more on archaeological approaches, a methodology which has only been received sporadically outside its original publication is the analysis of the empirical probability of coin sequences in the Upper Germanic-Rhaetian frontier zone, as devised by Klaus Kortüm (1998). In this approach, Kortüm compares the coins from military sites situated in the frontier zone with a standard cumulative coin curve consisting of various sites from the left bank of the Rhine. The calculated deviations—especially the first peaks of the graphs—point to intensity in site occupation. While the method has been criticised in some quarters based on a lack of representativity of the samples used, the study is still impressive in its innovative use of large coin datasets.

It is not only coins from sites which have been subject to big data analyses. A recent conference proceeding from 2019 on coin hoards shows the complexity and challenges which come with the processing and analysis of thousands of objects belonging to the same archaeological event. In the conference’s final contribution, Jean-Marc Doyen asks, ‘If we really have to study mega-deposits, how far can we realistically go?’ (Doyen 2019: 321), naming financial and temporal constraints as possible criteria which can influence the study of these huge datasets. Doyen makes the case that the analysis of mega-hoards will yield new information, as opposed to more of the same, only if the right questions are asked: for example, on the chemical composition of the coins, information on the dies, the possible owner(s) of such hoards, or the reasons for these kinds of massive accumulations. It is interesting to note that the ‘discovery’ of new emperors as issuers of hitherto unknown coin types does not occur from the analysis of mega-hoards but rather from small depots or even from coin finds from sites (Doyen 2019: 316).

It is, however, in the shape of databases and linked open data that the combination of numismatics and large datasets really comes into its own (DARIAH-EU 2018). Early coin databases started in the late 1980s and have been expanding in scale and complexity since the turn of the millennium. The main current trend is the linking of different collections (e.g. in museums) and existing databases to build a networked data collection which can be queried.

The collaborative project of Nomisma.org (NOMISMA 2022) was initiated by the American Numismatic Society in 2010 to facilitate the online presentation of numismatic concepts and employ the methods of the Semantic Web and Linked Open Data. It is thus now possible to jump from a single coin in the collection of Berlin State Museums, to information on its place of minting, to other coins minted from the same site. Thanks to the OCRE project (Gruber et al. 2014), the collation and visualisation of a series of coins belonging to the same type can now be achieved in minutes. The premise for these interlocking databases is the utilisation of the now widely used Research Description Framework (RDF) and a common ontology for all participating databases (RDF Working Group 2014). A further resource which showcases the power of big numismatics datasets is the coin hoards of the Roman Empire database (Lockyear and Gruber 2013). At the time of writing, 13,170 hoards are in the database. A short query of hoards discovered in the last 20 years (100 BC – 400 AD) shows that only 8.1% of the registered hoards date to decades between 80 and 140 AD. It is these types of preliminary results which can lead to deeper investigations on hoards, hoarding practices, and motives in the Roman world (CHREP 2022).

Landscape archaeology

In Roman archaeology, the study of the built environment has traditionally focused on cities, towns, villas, and farmsteads. While these categories are undoubtedly still relevant for understanding any region, since the 1950s, tentatively at first then at an increased intensity, the emergence of landscapes as an additional unit of analysis can be observed (Trigger 1967).

A case can be made for linking the rise of the prevalence of landscape archaeology to the increasing use of databases and geographical information systems on one hand, as both have facilitated the shift away from the study of a single site towards larger multi-site, regional studies, but also to the more frequent integration of scientific methods such as palynology, paleoecology, geoarchaeology, archaeobotany, and zooarchaeology on the other. According to John Lloyd, one of the strengths of landscape or settlement archaeology is its ability to generate data with which to test accepted ways of looking at past societies and to suggest alternative models (Lloyd 1991: 233). Ian Hodder and Mark Hassal’s (1971) famous case study of Romano-British walled towns shows exactly how this could be achieved.

In German-speaking archaeology, the debate on the differentiation between ‘Sieldungsarchäologie’ and ‘Landschaftasarchäologie’ is more explicit. Nowadays, according to Sebastian Brather, ‘Siedlungsarchäologie’, in its widest definition, can be divided into three categories: Environment, Settlements, and Settlement. Environment takes ecological factors into account; Settlements focus on specific sites, and Settlement (singular) on habitation patterns (Brather 2011: 134, Fig. 5).

In France, landscape archaeology (archéologie du peuplement/archéologie du paysage) really took off with the development of aerial photography. In his many publications on the landscapes of Northern France, Roger Agache (1978), the pioneer of French aerial archaeology, was able to show how dense the hitherto unknown Iron Age and Roman period rural landscape for this area was.

The approaches and methods of francophone archéologie du paysage seem to have been better integrated into mainstream archaeological research (Robert 2017). Philip Leveau’s work on Caesarea Mauretania from 1984 seamlessly integrates urban analysis, landscape archaeological investigation, and epigraphy into one coherent study. The Archaeomedes Project, financed by the European Commission and running from 1992–1994 with the aim of understanding soil degradation and desertification in Mediterranean landscapes, focused, in one sub-project, on the settlement in the Rhone valley, thus impressively demonstrating how settlement networks can change and persist for several centuries. By integrating geological and environmental data, it also made an explicit connection to contemporary anthropogenic activity and raised awareness of the gradual deterioration of the cultural heritage within the areas studied (van der Leeuw 1998).

If landscape archaeology is understood in its most positivist way, the understanding of the built environment of a landscape should rest on a dataset of all features pertaining to human activity for a specific period in a specific region. The tool du jour of this type of landscape archaeology—Geographical Information Systems—thus relies heavily on the conversion of archaeological source material into tabular data (Verhagen 2018: 21). Here is an interface in which the ‘translation’ of archaeological source material into machine-readable tables is at its most apparent. For Roman archaeology, this could entail defining and classifying sites and, at best, dating the beginning and abandonment thereof with the aim of reconstructing the development of the built environment.

The creation of such a database is not without issues: imprecise coordinates, indistinct chronologies, and a lack of information on the sites themselves (size, internal layout, phasing, etc.) are stumbling blocks which must be overcome. In recent years, however, several projects have been completed showing not only the potential of large-scale and rich datasets but also their relevance for our understanding of the Roman Empire.

In 2016, the publication of the database of the Roman Rural Settlement of Britain, consisting of around 3,650 sites, shows how datasets for landscape archaeology could be constructed (Smith et al. 2016). The immanent publication of the Roman Hinterland Project Database (RHPdb) and the accompanying publications will show the potential and challenges of merged datasets from three different project regions (Attema et al. 2021; 2022). The combined dataset should yield in-depth and broad-ranging insights into urban-rural relations, demography, and economic integration between Rome and its Hinterland. It will be interesting to see if the collective value of the RHPdb is ‘greater than the sum of its parts’ (Attema 2021: 15) as the authors claim. While not resulting in an actual database, M. Reddé’s (2017) double volume Gallia Rustica collected over 15 case studies from Northern Gaul and beyond. The project focused on the development and economic systems of the rural settlement in this area and showed how multidimensional and asynchronous the chronological trajectories of the Roman provincial rural settlement may have been.

Future directions for landscape archaeology are being cyclically debated. Ten years ago, Piraye Hacıgüzeller made the salient point that ‘[n]either walking across the landscape nor collecting the most precise spatial data will reveal a past reality that exists independently from us, the archaeologists’ (Hacıgüzeller 2012: 257). Bigger and better datasets cannot and will not compensate for epistemological gaps, but they do help us design and render more complex models, which in turn can open up new research perspectives.