Early investigations

Late antique defences were identified early in the development of archaeology in the region. These antiquarian investigations were overwhelmingly rooted in the Roman historical sources and took inspiration from Gibbon’s Decline and Fall of the Roman Empire. Footnote ²¹ These painted a picture of a ‘crisis’ in the late third century and imperial collapse in the late fourth/early fifth century.Footnote ²² Phases of fortification were quickly assigned to central policy based on the historical references.Footnote ²³ These early excavations established our basic model for different types of site and their temporal spread. Despite a clear focus on sites associated with the ‘military’, typologies and patterns taking in civilian structures were also developed: a belt of hilltop sites, the refuges or Höhensiedlungen, burgus-like features, urban defences and large fortresses were identified.Footnote ²⁴ Naturally, these excavations influenced the wider narrative, feeding into a story of decline, unrest and instability across the north-western provinces. National research traditions developed during this formative period; for example, German archaeology focused on military installations. This investigation drove further analysis of fortified settlements, uncovering more defences as the research tradition developed.

Post-war approaches

By the end of the Second World War, this phase had given way to more standardised investigation. New excavations exposed the long-term development of some sites and led to works of synthesis in the 1970s and 1980s.Footnote ²⁵ This momentum continued, with work examining shifting patterns of defence as well as architectural changes and a re-evaluation of key problems in our understanding of the period.Footnote ²⁶ Concurrent with this, new work established an earlier phase of activity associated with the Gallic Empire and the late-third century.Footnote ²⁷ Work on the Höhensiedlungen developed a basis for future work on hilltop sites and excavations at the Belgian refuges have given us an important dataset for the study of these sites.Footnote ²⁸

The publication of the controversial Grand Strategy of the Roman Empire in 1976 influenced the developing narrative.Footnote ²⁹ Although decline and unrest remained a key part of the story, broad concepts of ‘defence-in-depth’ began to be applied to the region.Footnote ³⁰ The emerging pattern appeared to be one of defended highways and elastic defence through a partially depopulated landscape.Footnote ³¹

Modern analysis

By the end of the twentieth century, the defence of north-western Gaul was increasingly considered ‘solved’, with new data envisaged as filling the gaps in a pattern of defended hilltops and fortified river and road networks. In recent years, however, this model has become increasingly fragile. Examinations have moved beyond an isolated approach to north Gallic defence, viewing the region as part of a wider defended whole rather than a unique example of a heavily militarised zone.Footnote ³² This analytical expansion has impacted on the study of all types of defences. Urban fortifications have undergone a revolution in theory, with important contextual work identifying trends, policies and development, both on the level of individual sites as well as groups of defences.Footnote ³³ The re-evaluation of the Höhensiedlungen has driven re-engagement with their function — either for defence or display as well as their form and distribution across the north-western provinces.Footnote ³⁴ Systematic excavation and aerial photography from the 1970s onwards in the German loess belt has seen the number of defended sites increase significantly.Footnote ³⁵ This has been mirrored by developments elsewhere with unexpected new sites appearing in Picardy.Footnote ³⁶

The last 20 years have seen an improved understanding of fortified settlements. Despite the recent flurry of studies, however, there have been few statistical or intra-regional examinations of the area. Work has primarily focused on individual sites or thematically linked groups of installations.Footnote ³⁷ Recently, synthetic investigations have begun to move towards comparable and replicable typologies for defence with further work underway to apply this on a wide scale.Footnote ³⁸ By examining different facets of the study of fortified settlements, we can attempt to establish biases and investigation distributions inherent in the dataset.

Defining fortification

A first step in the process of assessment is establishing definitions of the interrelated terms ‘defence’ and ‘fortification’. In the nineteenth and early twentieth centuries, fortifications were identified as defensive based on thematic morphology, with installations divided into regular (military), irregular (rural) and urban.Footnote ⁴¹ By the twenty-first century, further development had identified that this division is problematic, in part due to the blurring of civilian and military in later Roman Gaul and the wide morphological variation of sites.Footnote ⁴² A vast range of Latin terms have been identified, both in inscriptions and the historical sources, that appear to apply to late Roman defended sites.Footnote ⁴³ These terms, perhaps unsurprisingly, do not provide a firm basis for clear definitions of different types of late Roman defences. This confusion has spilled over into modern languages and there is a vast web of different and often inter-related and contradictory terms and phrases used to describe defended installations in the late antique world.Footnote ⁴⁴

Rather than deferring to the Latin terms and their modern counterparts, this analysis follows recent empirical classification systems based on the morphology of settlements, although these have, until now, been primarily limited to ‘private’ or small-scale installations.Footnote ⁴⁵ In this article, all fortified settlements are examined, regardless of their presumed function. Sites were defined as fortified if there was a demonstrable presence of architectural elements that could be used for a defensive purpose. Naturally, a vast range of different features and site types, some of which may have been multi-purpose, fall into this category (Fig. 3). This includes everything from the small earthwork defences identified in the Rheinische Lößbörden and the hilltop sites identified across the region to large urban wall circuits and military fortresses on the Rhine.Footnote ⁴⁶ A significant proportion of these sites could be considered multi-purpose, or include structures that, when combined with other features, appear to be defensive in nature. A good example of this is Seclin-Les Haut des Clauwiers (Fig. 3), a villa located in northern France.Footnote ⁴⁷ The site underwent an early fourth-century transformation with the villa largely converted to artisan production and grain storage. From the early/mid-fourth century onwards, a new grain storage structure with reinforced buttresses and thick walls, interpreted as a tour-silo, was constructed.Footnote ⁴⁸ On its own, this would not necessarily be enough to ascribe a defended function; however, the site appeared to be contemporary with a series of palisades and ditches regulating access to the building. This points towards a multi-purpose defensive element at the site. In light of instances such as this and the probable multi-function or unclear function of many of these fortifications, the sites in this paper have been graded at three levels according to their certainty based on the availability of evidence from published reports (Table 1). These methods allowed the establishment of a reasonably coherent model for examining the temporal and spatial density of fortified settlements over the later Roman period and reflects the degree of uncertainty inherent in the dataset.

Fig. 3. A variety of defences: selected fortified installations showing the diversity of types in the region (author).

Table 1. A breakdown of fortifications by data grade (n=357) (author).

Moving beyond the morphological dataset, several different groups of finds are associated with defensive or state activity in late Roman north-western Gaul. Some artefact groups, long associated with the late Roman army or state, have been recorded at potential fortified installations and include crossbow brooches and chip-carved belt sets.Footnote ⁴⁹ Reinforcing this is the presence of weaponry, or at the very least, multi-functional tools that could have acted as weapons.Footnote ⁵⁰ Dating these objects has, in the past, been problematic; however, we are now aware of clear typologies from the late third to fifth century for belt sets and crossbow brooches and this allows a level of chronological resolution for material culture findspots.Footnote ⁵¹ This has seen the identification of a wider military community in northern Gaul based on the material culture and includes finds from rural settlements, cemeteries and isolated findspots and allows us to infer the presence of defences close by.Footnote ⁵² In some cases, this is relatively straightforward: there is limited evidence of a defensive morphology beyond terracing at Ben-Ahin-Lovegnée, a hilltop site on the Meuse; however, the presence of an associated cemetery at Sarte-à-Ben, complete with weapons and belt buckles and late Roman ceramics and coins on the hill itself, is suggestive of a fortification on the spur above the river.Footnote ⁵³ Data on these sites were collected in a separate database and has been integrated here to highlight the temporal spread and distribution of activity associated with defence.

Investigation type

Historically, fortifications have been identified through a variety of means. Excavation, both limited and extensive, has been the primary vehicle for study; however other methods have been used to identify and categorise late antique defensive architecture. Aerial photography has a long tradition in parts of the study area, notably northern France and the Rhineland and has proved useful in spotting potential late Roman installations.Footnote ⁵⁴ Field survey and geophysics have also taken important roles in confirming late antique activity, especially in combination with aerial photos.Footnote ⁵⁵ Although the chronological and morphological classification of these sites is often less clear, in many cases they can be easily identified as late Roman defences.Footnote ⁵⁶ This extends to sites identified through finds. The majority of findspots however are less secure and there is significant variation in their location, relationship to a fortification and chronology.

The data presented here are overwhelmingly the result of excavation (Table 2). This reflects both the complex nature of some of these fortifications and in some places, notably the Eifel-Ardennes zone, the difficulties surrounding aerial photographical analysis. There is, however, some regional variation. Aerial photography has been repeatedly used in Nordrhein-Westfalen, and to a lesser extent Hauts-de-France, to identify defended sites, while field survey has commonly been employed as a strategy in Luxembourg and Rheinland-Pfalz. This is the result of the difficult nature of invasive fieldwork at hilltop installations, which are often forested or hard to define. This is also reflected in the large number of Höhensiedlungen in Rheinland-Pfalz that have been identified based on their finds alone.Footnote ⁵⁷ This group of sites, paralleled in Wallonia, has generally been linked with finds of military equipment on sites with defensive topography.Footnote ⁵⁸ Linking finds and topography has a long history and accounts for a significant number of sites defined as fortifications in the dataset. The differences in archaeological investigations demonstrate that our analysis of the fortified landscape rests on a very varied base and this is an important element to bear in mind as research into the topic advances.

Table 2. Breakdown of activity types at fortifications in the countries under study, where evidence is currently available (n=357) (author)

¹ This includes the limited number of sites in Hessen (3) and Baden-Württemberg (2).

² This includes the département of Seine-Maritime.

Investigations by period

Assessments of the data by excavation period highlight the biases and assumptions inherent in the data landscape. This breakdown, including sites that were repeatedly examined in different periods, demonstrates the variability of activity across the previous two centuries. Excavation at these sites has a very long history, with some of the first notable antiquarian studies of the Roman period in north-western Europe focusing on fortifications, especially along the Rhine limes.Footnote ⁵⁹ The temporal spread of excavation reflects this (Fig. 4) — there is significant nineteenth-century or earlier activity across all countries. These early activities, dominated by France and Germany, both reflect the high numbers of sites within their borders but also hints at the early focus on the military fortresses on the Rhine and the urban wall networks of French towns, sometimes to the detriment of other site types.Footnote ⁶⁰ In other countries, especially the Netherlands, villas held more attention with relatively limited interest in fortifications.Footnote ⁶¹ The situation in Belgium is slightly different. The investigation of defended installations began in the early nineteenth century; however, this antiquarian work generally focused on early medieval cemeteries and the study of the defended hilltop sites associated with them was somewhat secondary.Footnote ⁶²

Fig. 4. The temporal distribution of archaeological investigation at fortified sites, based on the totals set out in Table 2, where evidence is currently available, broken down by country and administrative region (n=548) (author).

Although there are minor variations between each country, driven in part by different national research traditions and priorities, there is a broad increase in archaeological activity at defended sites in the twentieth century. This activity peaks between 1950 and 1980, coinciding with the wider post-war construction and the subsequent economic boom.Footnote ⁶³

Geographical distributions

Our current understanding of the geographical spread of late antique fortifications is highly regionalised. Research has, by and large, examined thematic groups of similar sites, such as the Höhensiedlungen, or focused on specific zones, for example the Rhine frontier and the Via Belgica.Footnote ⁶⁴ This scholarship focus has been supplemented by excavation biases. The large-scale excavations in Nordrhein-Westfalen prior to lignite mining have exposed large numbers of defended sites in the Rheinische Lößbörden, whilst the expansion of the rail network and Zones d’Activité Économique (ZAE) in France have generated similar data.Footnote ⁶⁵ Other regions such as Belgium have similar excavation biases. The focus has been on specific sites, such as Braives and Liberchies, and some areas, such as Flanders, are poorly understood.Footnote ⁶⁶ This is equally true in regions where other site types have been the focus of study. In France-Le Grand Est and Luxembourg, scholarship has been concerned more with rural settlements and the first major surveys of defended installations date to the late 1960s and 1970s.Footnote ⁶⁷

With these biases in mind, it is clear that while in some areas we have a detailed understanding, our clarity on the distribution of sites in other regions is poor. Overall, the distribution is fragmented, reflecting a series of modern and natural geographical factors. Overwhelmingly, fortifications are located in the east of the study region and broadly, the closer an area is to the Rhine, the greater the density of fortifications (Fig. 5). Dense belts exist along the key transportation arteries; the main roads, notably the Via Belgica, as well at key junctions, some of which, such as Saarbrücken, were located on earlier agglomerations.Footnote ⁶⁸ Supplementing this is the concentration of hilltop defences in the Eifel-Ardennes-Meuse zone and we would expect further sites in modern day Luxembourg to complete this concentration.Footnote ⁶⁹ Instead of continuing southwards through the Vosges and Lorraine Plateau, the dense belt peters out, giving way to a range of limited Höhensiedlungen, burgi and other forms of defensive sites, including ex novo military installations such as Saint-Laurent-sur-Othain-Le Châtelet or urban defences like Metz.

Fig. 5. Spatial distribution of fortified sites (250–450) across the study region and beyond. Underlying map from the EAA Copernicus DEM (n=357) (author).

This distribution confirms the lack of fortifications in the depopulated zones of the Dutch and Belgian Kempenland and highlights our limited evidence for late Roman defensive installations along the North Sea coast.Footnote ⁷⁰ We see a slow increase of sites towards the south along the coast, beginning with the third-century forts at Aardenburg and Oudenburg, eventually forming a complex cluster of morphologically identified installations in the Pas-de-Calais, reinforced by a range of sites with artefacts traditionally associated with militarising communities.Footnote ⁷¹ This is complemented by an expected and partially identified drop-off in defended sites further south and west, where the majority of defended settlements are walled towns such as Paris, Melun and Reims and there is a less diverse range of other fortifications.

To some extent these patterns are mirrored across the channel in Britannia. Key defended installations are located on the main arteries, some of which have late antique circuits.Footnote ⁷² Where this differs, however, is that the majority of major urban centres in Britannia were defended by the mid-third century and received fourth-century augmentation, primarily in the form of towers, for example at London.Footnote ⁷³ Important differences are present between Britannia and the continental provinces: rural fortifications are significantly rarer, with few examples of burgi or hilltop sites.Footnote ⁷⁴ The military footprint is also more static in comparison to northern Gaul: there are limited examples of new fortifications, with the Saxon Shore and Yorkshire towers being exceptional examples.Footnote ⁷⁵

The geographical pattern reinforces the fragmented distribution identified in previous thematic studies.Footnote ⁷⁶ Bringing the data from these sources together allows us to move our geographical understanding forward (Fig. 5). Firstly, we can move towards analysis of location, orientation and interconnectivity and secondly, we can identify lacunae in the wider pattern. These lacunae present interesting problems, especially in the highland zone between the Mosel and Maas rivers and further research may indicate that these zones were simply too far from movement corridors in the landscape to require fortification (see Fig. 1 for locations of rivers).

Temporal distribution

The chronological distribution of fortification presents a more complex picture. Applying a basic classification method to the data allows for the analysis of the temporal intensity of fortified settlements over the period and reflects the degree of uncertainty caused by chronological problems in the data.

To achieve a realistic assessment of certainty of the temporal distribution, three different levels of a graduated scale of occupation certainty were applied to each site (certain, probable, possible). Assessing this employed a variety of different chronological indicators, including ceramic, artefactual and numismatic records and where available, radiocarbon dates or other scientific methods.Footnote ⁷⁷ It is important to note that the data primarily derive from published sources and given the historicising biases present, this can lead to errors in interpretation. Chronologies based on the ceramics and coin sequences are subject to problems. Ceramic studies have highlighted the problems with fifth-century coarse wares and the longevity of Argonne sigillata in the third centuryFootnote ⁷⁸. Coin data hold similar problems; the huge number of third-century radiates, the issue of production of the mint at Trier and the longevity of coin use beyond the early to mid-fifth century have made it difficult to accurately pin down abandonment dates for sites.Footnote ⁷⁹ To offset some of the chronological issues, 50-year blocks were chosen to represent broad resolution changes. Intensity of use was not necessarily identifiable in this broad approach to our current data and further work will be needed to assess this archaeologically.

With these issues in mind, the chronological spread illustrates an unexpected spread of data (Fig. 6). Peaks of activity in the late third to early fourth century and the late fourth to early fifth century were anticipated. These periods have generally been characterised as times of rapid change. The late third century saw the depopulation or étiage of rural settlements and urban centres while broader Roman political control broke down in the fifth century. The instability inherent in these periods has traditionally been seen as a driver for fortification construction and use.Footnote ⁸⁰ Instead, we see a steady increase in fortified activity over the course of the late third and early fourth century, beginning with limited examples, mostly in the form of military installations, in the first period (200–250).Footnote ⁸¹ This peaks in the second half of the fourth century (the period 350–400) and is followed by a sharp drop-off in sites into the fifth century, perhaps associated with the slow withdrawal of Roman control.

Fig. 6. The chronological range of activity at fortified settlements across the region under study, highlighting the certainty of temporal identification (n=248) (author).

The single peak of activity indicates that several other factors are at play and worth exploring — clearly it demonstrates we are not dealing with the cyclical use of instability-linked fortifications, nor are they short-term knee-jerk examples of ‘crisis architecture’ associated with limited periods of occupation. Instead, the evidence suggests many of these installations were developed, expanded and used over the course of half a century or more.Footnote ⁸² Vireux-Molhain-Le Mont Vireaux for example, was occupied from the late third to mid-fifth centuries and included several phases of reconstruction, expansion and rebuilding.Footnote ⁸³ Although a small number of sites did have the rapid lifespans associated with ‘crisis architecture’, for example the refuge at Nismes-Roche de Sainte-Anne, turnover was relatively low with most sites occupied continuously from the late third/early fourth century onwards.Footnote ⁸⁴ Given that this peak is visible across both the macro-region and in all sub-regions, it suggests that the same priorities — a desire for defendable sites — was of key importance to groups across the north-west, both on a regional and local level. This priority of construction and use was maintained until at least the mid- to late fourth century, when the number of sites in use begins to decline, although it is worth stating that our chronological resolution for the early fifth century is still unclear and relying on the ceramic and coinage data remains problematic.Footnote ⁸⁵

This increase in activity, rising from about 40 early third-century military sites focused on the limes and coastlines to nearly 140 active installations in the mid-fourth century, indicates that various interest groups felt a requirement for defence. Equally, it hints that the local population, the state and other agents were motivated and, more importantly, able to marshal the labour and materials needed to construct and defend installationsFootnote ⁸⁶. This motivation is not limited to one region, but occurs as a single wave of activity across the fourth century.

The temporal overlay of probable and possible sites presents a different story. Both these groups are less chronologically anchored phases and show the expected peaks in the late third and early fifth centuries, broadly coinciding with pre-conceived periods of ‘crisis’ in northern Gaul. This is likely to represent the biases within the dataset inherent from an older, historicising approach. Traditional scholarship, especially on French town walls, has sometimes assumed that fortifications should have been built during the late third-century crisis, as a reaction to barbarian raiding.Footnote ⁸⁷ Although this may be true, and there is some evidence that some of these defences did have a third-century phase, for example at Amiens, where coins recovered from the wall foundations dated to the reign of Probus, the evidence is ambiguous at many sites.Footnote ⁸⁸ Future work may indeed identify such phases beyond doubt; however, as it currently stands, the evidence does not point towards significant peaks of third- or fifth-century fortified activity.Footnote ⁸⁹

The chronological sequence paints a picture of a region with similar defensive priorities, transcending local or regional activity. Although what we see on the ground cannot in any way be called integrated at a supra-provincial level, it is clear that fortified activity is driven by factors at a level above local concern. This suggests that the mid-fourth century was the point at which the state, local populations and other actors were most concerned with defence, and perhaps, by extension, the perception of instability across the entire region was most important.