The fifth century is often considered a turning point in European history, yet when we try to characterise this pivotal century in the archaeological record, we often fall short and can only provide a limited amount of evidence, which is frequently unclear in nature or interpretation. In other words, archaeologically speaking, for many parts of the (former) Roman Empire, the fifth century is an obscure one. The general issues of the fifth century in Northern Gaul can roughly be summarised as follows: it lacks a clear understanding of how people lived, what materials they used, who they were, and where they came from. Today, pressing issues are no longer necessarily the total absence of evidence (Heidinga et al. 1992: 121–122; Dierkens and Périn 2003: 174–175) but rather the identification and interpretation of the Late Roman fifth century sites, features, and artefacts. The main problems currently include poor chronological resolution of material culture, the blind labelling of features and finds as Germanic based on an uncritical application of the traditional narrative of the end of the Roman West, and the use of the fifth century as a convenient chronological end-phase for the Roman occupation of settlements that in itself is not considered in sufficient detail.
This paper aims to investigate the factors involved in obscuring the ‘Late Roman fifth century’ in Northern Gaul. Starting from the state of research in Flanders (northern Belgium, Figure 1), certain issues have been observed that hinder the identification and interpretation of the Late Roman period and fifth century in the archaeological record. Causes for these issues are first sought in the particular research tradition from Belgium and Late Roman archaeology as a discipline, followed by a review of the role of theory and scholarly biases in the state of research of Late Roman and fifth century archaeology in Northern Gaul. To move beyond the traditional debate, the impact of basic scientific methods such as periodisation and classification are also explored. Finally, a number of suggestions for new approaches are provided to address the existing issues as it is necessary to ameliorate the current situation caused by the shift from scholarly to developer-led archaeological excavations and research.
The Case for an Archaeologically Obscure Fifth Century in Northern Gaul Based on Evidence from Flanders, Belgium
The notion that archaeological practice is heavily influencing our understanding of the fifth century (or lack thereof) emerged from the study of the Late Roman period in the region of Flanders in Belgium (traditionally AD 270–410). Similar to the widespread perception of the fifth century, it was assumed that there was not much Late Roman archaeology to speak of. Yet, a multitude of sites and finds emerged from a literature study (Van Thienen 2016: 48–125) in which it became apparent that the Late Roman period is not absent from the archaeological record (Figure 2). It only appears absent in comparison with the archaeological record of the Middle Roman period, which can be seen as exceptionally rich. To address how material evidence and archaeological practice obscure the (Late Roman) fifth century in the archaeological record, it is necessary to first discuss the existing problems regarding identification, dating and interpreting Late Roman archaeology in general.
State of research on Late Roman archaeology in Flanders, Belgium
In most regions in Northern Gaul, a settlement regression is observed from the late second and early third century, continuing into the fourth and fifth centuries. This settlement regression is well observed in Flanders based on the data from the Central Archaeological Inventory (CAI). This database records all archaeological excavations, field studies, and finds within the territory of Flanders.1 When the excavated Roman sites are subdivided into Early, Middle, and Late Roman, the demographic peak of the Middle Roman period and subsequent regression is quite obvious with a significant settlement reduction (Table 1, Figure 3). This demographic drop has recently been confirmed by the study of woodland recovery during the Roman to Medieval transition in Flanders (Deforce et al. 2020).
|Total Roman sites||Early Roman sites (40 BC–AD 70)||Middle Roman sites (AD 70–270)||Late Roman sites (AD 270–410/450)||Not determined|
|Civitas Tungrorum – All sites||1465||262||545||223||735|
|Civitas Tungrorum – Habitation||580||103||200||98||288|
|Lys-Scheldt valley (Vermeulen 1992)||80||28||66||9||20|
|Lys-Scheldt valley (CAI 2018)||273||31||101||6||135|
A comparison between the known Middle Roman (c. AD 70–270) and Late Roman (c. AD 270–450) sites within the civitas Tungrorum also provides a clear image of settlement regression (Table 1, Figure 4). The civitas Tungrorum, with Tongeren (Atuatuca Tungrorum) as its capital, spans most of east Belgium (Flanders and Wallonia), part of the southeast Netherlands and smaller segments of Luxembourg, Germany, and France. The data presented here is based on the recently constructed database for the civitas Tungrorum (Van Thienen et al. 2019). The difference in reduction rates between the CAI data from Flanders and the civitas Tungrorum data can be explained as a difference between micro-regions in which many factors have to be taken into account, such as the degree of arable land, the presence/absence of or proximity to urban and military centres, access to interregional networks through major roads, river systems, and Roman administrative borders. The distinction in the scale, nature, and chronology of landscape abandonment and settlement regression between neighbouring micro-regions has been noted in various studies in the northern continental Roman provinces: Louis (2004) for the Scarpe valley, Heeren (2015) for the Lower Rhine, and Van Thienen (2016: 126–144) for Flanders. These studies create a more nuanced image of the Late Roman demographic reduction and illustrate that the process of abandonment happened more organically rather than in one massive depopulation phase.
While the number of central places, villas, and rural settlements reduced heavily by the end of the third century, a number of sites with a clear military and central or even urban character were present in the Late Roman period in Belgium. In Flanders, the best documented sites include the Oudenburg Saxon Shore fort (e.g. Mertens 1978; 1987; Vanhoutte 2007; 2018), the city of Tongeren (e.g. Vanderhoeven 2012; 2017), and the central place of Kortrijk with potentially a Late Roman military fort (Cortoriacum, e.g. Rogge 1988; Despriet 1990; 2012). In Wallonia, the most important sites include the city of Tournai (Tornacum, e.g. Brulet 1984; 1990; 2012) and military forts on the major road known as the Limes Belgicus such as at Liberchies, Braives, Montaigle, Furfooz, and Éprave (Brulet 1990). A certain continuity of Gallo-Roman rural settlements from the third to the fourth century can also be observed. These sites mainly cluster along the major rivers and roads, such as the rural settlements at Donk (Van Impe 1983), Kruishoutem (Vermeulen et al. 1993: 58–74; 172–174), Nazareth (Dyselinck 2017), and Turnhout (De Smaele et al. 2012). In addition, new settlements and features dating to the Late Roman period have been uncovered, with an increase in the second half of the fourth century and the beginning of the fifth century. These new settlements often contain structures and material culture with Germanic traits (i.e. north of the Rhine) and are found in larger numbers compared to earlier Roman phases. These are generally interpreted as evidence for Germanic settlers, such as at Neerharen-Rekem (De Boe 1983), Sint-Martens-Latem (Vermeulen 1989), Zele (De Clercq and Taayke 2004), Meldert (Smeets and Steenhoudt 2012), but also at Donk (Van Impe 1983), and Baelen in Wallonia (Hanut et al. 2012).
The Lys-Scheldt valley
The identification, dating, and interpretation of Late Roman and fifth century sites remains problematic despite the growing number of excavated sites that date to the periods in question. The micro-region of the Lys-Scheldt valley provides an excellent example to illustrate the problems with the identification of Late Roman rural sites. In 1992, Vermeulen published an archaeological overview of the Flemish part of the Lys-Scheldt valley. Between these two rivers, Vermeulen counted a total of 76 definite and probable Roman sites by means of excavation and survey. Of these, a Late Roman chronology was noted for nine locations: three excavated rural sites with clear settlement evidence – Asper (Vermeulen 1986: 111–115), Kruishoutem (Vermeulen et al. 1993: 58–74; 172–174), and Sint-Martens-Latem (Vermeulen 1989: 71–76); three coin finds dating to the fourth or fifth century; and three field surveys yielding surface finds consisting of Eifelkeramik and handmade pottery with quartz temper that are generally dated between the second half of the third century and the fifth century (Tables 1, 2).
|Site||Type||Structures and finds||Reported date|
|Asper – Jolleveld (ASP3)||Excavated settlement||Postholes and pits containing Eifelkeramik, Chenet 320 Terra Sigillata bowl, Chenet 342 foot-vessel and handmade pottery tempered with chamotte (grog), plant or coarse quartz minerals. Additional Merovingian cemetery and pottery, possible continuity suggested.||Second half 4th century (AD 360–390)|
|Deurle – Oude Pontweg (D3)||Surface find||Handmade pottery and possible late tradition East Gaulish Terra Sigillata.||Iron Age, Middle or Late Roman|
|Huize – Lozer Noord (HU1)||Surface find||Coin of Constantine I.||AD 307–377|
|Kruishoutem – Kappelekouter (KR5)||Excavated settlement||Well containing Chenet 342 foot-vessel. A large collection of 4th century coins. Postholes and pits indicating an early Germanic settlement.||4th–5th century|
|Sint-Denijs-Westrem (SDW1)||Historic find||Coin hoard found in 1787 containing coins from Valentinian I, Valens, Theodosius, Honorius, and Constantinus III.||AD 408–410|
|Sint-Denijs-Westrem/Rosdam (SDW3)||Surface find||Eifelkeramik Niederbieber 89 and handmade pottery tempered with coarse quartz minerals.||3rd–4th century|
|Sint-Martens-Latem – Brakel (SML3)||Excavated settlement||Partial farmstead and two sunken hut features from early Germanic settlement containing Eifelkeramik and a Theodosius coin.||Late 4th–first half 5th century|
|Sint-Martens-Latem (SML15)||Historic find||19th century coin find containing coins from Hadrian, Septimius Severan, Gallienus, and a Theodosian solidus.||Middle to Late Roman|
|Zulte-Donk (ZU2)||Surface find||Mayen Eifelkeramik and handmade pottery tempered with coarse quartz minerals.||Late Roman?|
By 2018, a total of 99 excavated Roman sites between the rivers Lys and Scheldt were recorded in the CAI database, with an additional 174 sites located at the tributary streams in the wider Lys-Scheldt region (Figure 5, Table 1). Closer consideration, however, shows that only six locations in the extended Lys-Scheldt region contain a Late Roman chronology: three excavated sites with settlement evidence from Kerkhove, Kortrijk, and Sint-Martens-Latem; three burials from Kerkhove, Velzeke, and Zwevegem (although no reason has been recorded in the database for the Late Roman date for the burials from Velzeke and Zwevegem); and one unstratified find from Destelbergen. The settlements in Asper and Kruishoutem by Vermeulen are not clearly indicated in the CAI as containing a Late Roman chronology because the abundance of Middle Roman evidence dictates the primary chronology for these sites, overshadowing the less numerous Late Roman finds. Therefore, the Late Roman chronology of these two sites is easily missed when excavators consult the CAI database.
A review of recent excavation reports indicates that also other sites can easily be overlooked (Figure 5, Table 3). In addition to excavated evidence from Asper, Kruishoutem, Sint-Martens-Latem, Kortrijk, and Kerkhove, at least three more sites should be added to the list of Late Roman sites. The first is a small settlement in Nazareth where excavations in 2014 uncovered a Wijster-type house and recorded radiocarbon dates2 from food crust and soot from a pot that was deposited in one of the postholes, pointing to a third to early fourth century date (Dyselinck 2017). Wijster-type houses have only recently become known to the Flemish developer-led archaeology as potential Late Roman structures (Figure 6). While some features from the excavated large Roman settlement complex in the harbour of Ghent (Kluizendok, excavated in 2006–2007, see Laloo et al. 2008) were thought to show similarities to structures found at Wijster in the Netherlands (Van Es 1967), it was not until the excavations in 2013 at the Rode Rokstraat in Hasselt (Hazen 2014: 103–106) that this type of house-plan was identified as a Late Roman house-type in Flanders. The second site is Bachte-Maria-Leerne, where excavations in 2016 revealed a large quantity of possible Germanic pottery. The ceramic assemblage consists of handmade pottery containing minerals from glacial deposits from the northern Netherlands and Germany (Van Thienen et al. 2020: 135–146), similar to finds from multiple fourth to fifth century sites in Belgium, the Netherlands, and northern France (De Paepe and Van Impe 1991; Bouquillon et al 1994). Three 14C dates3 from cremated bones support a third to fourth century chronology. A third site is the burgus in Velzeke located by excavations and geophysical survey, which has been dated to the third quarter of the third century (Lamarcq, Rogge 1996: 89–91; Deschieter 2016). A series of radiocarbon dates4 from surrounding sites had already hinted at a possible extension of activities into the fourth and possibly fifth century. Finally, two early medieval settlements were recently excavated in Merelbeke (Heynssens and Hoorne 2018) and Lemberge (Beke and van den Dorpel 2017). Both sites yielded radiocarbon dates ranging from the early fifth century to the middle of the sixth century (Table 3). While these sites are undoubtedly early medieval settlements, a possible link with the remnants of the Late Roman communities in the early and mid-fifth century has to be considered.
|Site||Type||Date||Structures and finds|
|1||Kortrijk||Central place||300–450||Hypocaust, concrete, mortar and an abundance of ceramic building material suggesting the presence of a bath. Based on associated ceramics and coins, this bath structure could be dated to the 4th–5th century. Two structures with dug out features have been encountered, together with ‘Germanic’ handmade pottery. Also evidence of artisanal or production activities is given by kilns and metal waste. Additionally, an abundance of 4th century pottery is frequently found in this location.|
|2||Kruishoutem||Central place||300–450||Germanic brooches, coins, pottery (Chenet 342), stratigraphy (post-3rd century well deposits).|
|3||Bachte-Maria-Leerne||Rural settlement||250–390||14C: AD 210–390 (95.4%)|
|4||Sint-Martens-Latem||Rural settlement||370–450||Sunken huts, pottery (handmade Germanic).|
|5||Nazareth||Rural settlement||250–340||14C: AD 140–340 (95.4%), Wijster-type house, handmade pottery.|
|6||Asper||Rural settlement||360–390||Pottery: Argonne TS (Chenet 320) Late Roman terra nigra (Chenet 342).|
|7||Kerkhove||Rural settlement||270–400||One-aisled timber building dug into 3rd century stone construction associated with the former road-side building. Dated post-AD 260/270. Additional Eifelkeramik (Alzei 32/33: 4th–5th century), follis of Constantine II or Constans (post-337); and a female inhumation in a filled 3rd century ditch, interpreted as Germanic based on wooden bracelet and silver ring, dated to end of 3rd to 4th century based on a Postumus coin.|
|8||Merelbeke||Rural settlement||410–550||14C: AD 410–550 (95,4%)|
|9||Lemberge||Rural settlement||425–550||14C: AD 130–350 (95,4%) + AD 427–548 (95,4%)|
|10||Velzeke||Burgus||250–275/300||Small finds (unspecified). Dated based on (assumed) Germanic raids in the third quarter of the 3rd century.|
|11||Tournai||Urban/Central place||270–480||Construction of Late roman town wall is dated to the late 3rd to mid-4th century and burial site is dated to 4th and 5th–7th century.|
Thus, in nearly 30 years, only four sites with a Late Roman chronology can be added to the initial list by Vermeulen (1992) for the entire Lys-Scheldt region. Despite the established settlement regression, this remains a very poor number after three decades of intensive archaeological fieldwork. It would be expected that the higher number and random spatial locations of excavations by developer-led archaeology would have led to at least a small increase in the frequency of Late Roman and fifth century finds. Here, it has to be stressed how little chronological markers have been used to identify the Late Roman and fifth century phases: mainly radiocarbon dates, Germanic handmade pottery, Wijster-type houses, and sunken hut features. Without reliable material to submit for absolute dating, the mineralogical study of handmade pottery, or the knowledge of Wijster-type houses, the rural settlements at Nazareth and Bachte-Maria-Leerne would have been overlooked as potential Late Roman settlements. Similarly, the Early Medieval focus in the cases of Merelbeke and Lemberge neglects the possibility of a Late Roman component altogether despite the possibility of early fifth century dates and the documented Late Roman sites in the near vicinity. This implies that in a richly occupied Roman provincial landscape that was relatively well-connected to (inter)regional economic networks—as recently demonstrated by network analysis of the river Scheldt in the Roman period (Bongers 2020: 47)—the Late Roman evidence is easily overlooked.
Interpretation and chronology issues in Late Roman Flanders
The observation that only a handful of material culture is used as a chronological indicator for the Late Roman period is not only visible in the Lys-Scheldt region, but for Flanders as a whole. From the review of various publications of 40 sites in Flanders (after Van Thienen 2016: 423–479), 73 references were found of chronological markers used to identify features, layers or structures to the Late Roman period and/or first half of the fifth century AD (Supplementary Table 1). Those markers can be divided into 14 categories (Figure 7), of which only seven have been noted more than three times (Table 4). Among the most frequently cited markers are: numismatic evidence ranging from Postumus to Constans II, Argonne Terra Sigillata (mainly roller-stamped decoration and bowls of the type Chenet 320), Eifelkeramik, sunken hut features, Wijster-type house-plans, Late Roman terra nigra foot-vessels (specifically type Chenet 342), and radiocarbon dates.
|Coins||From Postumus to Constans II.||Post-AD 260 to post-AD 411|
|Argonne Terra Sigillata||Roller-stamped decoration.||Late 3rd and 4th century|
|Bowl type Chenet 320.||AD 320–410|
|Eifelkeramik||Alzei typology and Mayen production.||4th–5th century and Early Medieval|
|Sunken hut feature||Early Germanic settlement feature.||4th–5th century and Early Medieval|
|Wijster-type house||Germanic settlement feature.||3rd–4th century (2nd–3rd century north of the Rhine)|
|Foot-vessel type Chenet 342||Related to Germanic Fuβschale, but sometimes made in Roman pottery traditions.||Traditionally dated to the 4th century, revised ca. AD 350–450|
|Radiocarbon dates||Calibrated dates between AD 260/270 and 410/450|
On top of the very limited array of material sources, further complications arise from how these materials are used. First, the use of imported ceramic and numismatic evidence as a primary dating method cannot be approached in the same manner for the Late Roman as in the Middle Roman period. The entire economic structure, trade mechanisms, and flow of materials had changed from the third to the fourth century. This resulted in a reduced quantity, variety, and distance of imported goods to the northernmost regions of the Roman Empire. Esmonde Cleary (2013: 311–312) observes that economic activities closely tied to the imperial system and complex organisation of the Early Empire all show a similar pattern of increasing in the middle or later second century and then going into decline. Furthermore, while the fourth century was still a highly monetised economy, it is characterised by a large number of imitations and emergency money, as well as a high insecurity towards the time span and regularity with which coins were struck (Van Heesch 1998; Stroobants 2013). Moreover, in the numismatic study of Holtum, Kemmers (2014: 165–170) demonstrates that bronze coin deposits on Germanic rural settlements from the later fourth and fifth centuries could have served as caches of raw materials for metal production, rather than as money.
Second, handmade pottery is the most frequently found ceramic group on rural settlements between the fourth and sixth centuries (Rogge and Van Doorselaer 1990: 14–17; Vermeulen 1992: 113–116), yet it remains undervalued and often incorrectly dated and interpreted (Van Thienen 2017: 120–121). Handmade pottery (Figure 8) is mainly assigned to the Late Roman period when it is associated with typical imports such as roller-stamped Argonne bowls or Mayen Eifelkeramik, or when its fabric contains large white minerals (quartz) that can be indicative of a provenance from the Elbe-Weser region (but not necessarily). While pottery with a Germanic provenance has been attested on Belgian sites, it is a faulty assumption that all Late Roman handmade ceramics are made by Germanic people or produced in Germanic territories (De Paepe and Van Impe 1991). Moreover, Germanic handmade pottery does not necessarily have to be dated to the fourth and fifth century, as is indicated by Germanic ceramics dated to the second and third centuries in assemblages from Elewijt (Van Impe and De Buyser 1990; Van Impe et al. 2005), Zele (De Clercq and Taayke 2004) and Kontich (Opsteyn 2003).
Third, a similar Germanic-bias exists regarding building traditions, specifically the Wijster-type houses and sunken hut features. The Wijster-type house is named after the houses documented by van Es (1967: 49–76) from the Germanic settlement at Wijster (Netherlands). In Flanders, these houses are generally dated to the fourth century, while north of the Rhine this house type is generally dated to the second and third century (Hiddink 1999: 137–141). The dating discrepancy seems solely based on the assumption of Germanic settlers not arriving in the northern Roman provinces before the fourth century. A more critical assessment and a comparison of contemporaneous houses in Northern Gaul and Germania is necessary to confirm or clarify this difference in chronology. Sunken huts are even more difficult to date, as they are also a common feature in Early Medieval rural settlements. The lack of sufficient datable material makes it hard to assign sunken huts to one of both chronological phases. Overall, it remains quite difficult to correctly date Late Roman and fifth century rural settlements without the use of absolute dating techniques or the presence of strong stratigraphic evidence.
The lack of sufficient and varied material culture to date these settlements also has had an impact on how the Late Roman landscape is perceived. In excavation reports and literature, features and finds are often dated very generally as ‘Late Roman’ (traditionally c. AD 270–410, more recently c. AD 250–450) or ‘Late Roman to Early Medieval’ (i.e. AD 300–600) (see Tables 2, 3, 4 for examples).5 As a result, these two or three centuries are often lumped together in a single chronological phase that represents this period as stagnant and non-dynamic. Despite a reduced demography, the landscape in Northern Gaul is not abandoned and still contains important Roman urban and military centres such as Tongeren, Maastricht, Nijmegen, Oudenburg, Tournai, or Cologne. The lack of clear evidence and the lingering image of mass abandonment as the result of barbarian invasions and Germanic settlers still dictates the archaeological narrative in Flanders and Belgium at the expense of more nuanced interpretations. This is not surprising given the relatively late development of Late Roman (provincial) archaeology as a separate discipline within Roman archaeology in Belgium.
A historiography of Late Roman and fifth century archaeology in Flanders, Belgium
In the early twentieth century the concept of Romanisation internationally dominated the mind-set of Roman archaeologists and historians. Many Belgian scholars, however, did not partake in the debate in light of the ideological conflicts that coloured this era in Belgian history. The association of the Belgian archaeologist De Maeyer with the Nazi-regime resulted in topics such as Romanisation and Germanic incursions being avoided altogether after World War II (De Clercq 2009: 47–53). In general, the dominant view of the Roman culture was ‘romantic’, as it stood for civilisation, a unified state, and a sophisticated high society. Thus, much of the initial focus in Roman archaeology concentrated mainly on elites and Mediterranean material culture, which stood in contrast with the ‘indigenous primitive tribes’ and ‘barbaric Germanic warriors’. It stands to reason that, in the decades following WWII, the knowledge of the Late Roman period suffered from this rhetoric, since the third to fifth century was considered a period in which Germanic tribes continuously raided and settled in the Western Roman Empire. The fifth century migrations or Völkerwanderung were largely accepted as a historical fact from the antique sources and were seen as a large factor, if not the most important factor, in the downfall of the Roman West.
In the mid-twentieth century, Late Roman archaeology in Belgium mainly consisted of small find reports and brief observations in excavation publications. It was not until the 1970s and 1980s that the first Late Roman studies emerged from a number of excavated settlements in Flanders. The first major study was that of the excavated fourth century cemetery in Oudenburg (Mertens and Van Impe 1971), and the historical and archaeological study of the Belgian coastal plain by Thoen (1978; 1987). Besides Oudenburg, the coastal plain did not reveal much material evidence that could be placed beyond the end of the third century. An explanation was sought in military and political factors alongside what was thought to have been ecological changes in the coastal dynamics which led to great flooding and land-loss (i.e. the now-outdated model of the Dunkirk transgressions). Unfortunately, the interpretation of a deserted coastal plain was extrapolated to the hinterland and had a negative impact on the identification of Late Roman archaeological features and material culture in the rest of Flanders. Other studies initially remained quite close to the excavated sites, such as the Germanic rural settlements at Donk (Van Impe 1983), Neerharen-Rekem (De Boe 1983), Wange (Lodewijckx 1996: 214–220; Opsteyn and Lodewijckx 2004), or the Late Roman burials at the cemeteries of Tongeren (Vanvinckenroye 1982; 1984). At this time, the main interpretative framework tried to connect the archaeological evidence with the historic sources and labelled these sites as new Frankish settlements. This ethnic-centred framework is illustrated by the ‘Germanisation’ of the Lys-Scheldt valley (Vermeulen 1992: 256–257) or the tracking of the Frankish migration in Flanders (Opsteyn 2003). Gradually, the first Late Roman ceramic studies appeared such as the initial study of the production of Argonne Terra Sigillata (Brulet et al. 1994; Bakker et al. 1996) or the petrographic analysis on Germanic handmade pottery by De Paepe and Van Impe (1991). It was not until the overview of Late Roman sites in Northern Gaul by Brulet (1990) that a pan-regional synthesis study was carried out. In general, in the second half of the twentieth century, archaeologists stayed close to the general narrative revolving around a series of Germanic invasions resulting in the abandonment of most Gallo-Roman settlements and arable lands, to which the Roman authorities responded with a reorganisation of, and investment in, the military. As a result, the focus of Late Roman archaeology remained fixed on military infrastructure and Germanic invaders/settlers at the expense of the continuity of urban and rural populations. The study by Van Ossel (1979; 1992) on rural settlements in the southern part of the civitas Tungrorum demonstrates that the rural landscape of the fourth and fifth centuries was not completely unknown, but rather that more visible and better documented Roman archaeology remained the preferred topic of study.
In the twentieth century, the interests of individual scholars proved to be the largest drive for the research and identification of Late Roman archaeology in Belgium. Since the start of the twenty-first century, the organisation of archaeological fieldwork has changed drastically in Flanders through developer-led (commercial) archaeology. Research has become more spatially randomised and less programme-led, due to the decreasing impact of individual scholars on the focus of archaeological research. Now, most excavations take place on locations where heritage is threatened, which has made these issues with identification and chronology highly problematic and urgently need to be addressed. Given the poor state of research for the Late Roman period and the fifth century, only a handful of material markers are known to the excavators—who are not necessarily specialised in this time period—which leaves them with little knowledge of how to successfully identify this more elusive part of the archaeological record.
The Impact of Theory and the Persistence of Biases
To fully understand how we ended up with only a few material markers to date the Late Roman period and fifth century and why it seems so difficult to make any significant progress in interpretation, we need to acknowledge the impact of theory and narratives on the collection of archaeological data. It can be argued that the current perception of the Late Roman Empire goes back to Gibbon’s work in the eighteenth century (1776–1789) on the decline and fall of Rome and the end of the Classical world in Western Europe. From Gibbon’s view, two main perspectives emerged and were described by Ward-Perkins as the catastrophic view and the accommodating view (2006: 1–10). The catastrophic view presents the Classical civilisation as a peak in human achievement that was destroyed by barbarian invasions which led to the destruction of the Roman political structure and lifestyle. The accommodating view, however, argues a more peaceful transformation by which the Germanic migration resulted in altered Western societies. These new societies are labelled by Pohl as barbarian successor states and he ascribes to them a raw strength and courage that replaced a state that had ‘lost their ambition and military virtues and abandoned the values of public service under the influence of the Church’ (Pohl 1997: 33). Often, both views have been placed in opposition to each other where scholars choose to focus on either collapsing political and economic structures or social and cultural-historical changes (Ando 2008: 31–32). As it is easier to grasp economic processes with material culture, the decline model strongly embedded itself in the archaeological interpretations of the Late Roman period in the twentieth century.
Initially, the decline in Roman structures provided a historical context to understand the lack of material evidence. Coincidentally, information from written sources was used to date, categorise, and interpret finds and settlements from the third to fifth century. Unfortunately, outside scholarly debates, it became commonplace to apply the decline model uncritically to explain (without understanding) the end phase of Roman sites. This resulted in the creation of two unyielding chronological markers in Northern Gaul. In AD 260, the Limesfall event served as a readily available end-date for Roman vici, villas and rural settlements. In AD 410, the withdrawal of Roman military forces from the northern provinces became synonymous with the end of the Gallo-Roman society, providing another abrupt end-date for Late Roman sites. While these events and their implications can be debated, they have nevertheless been transformed into fixed chronological barriers, excluding much of the potential evidence for the continuity of local rural and urban communities into the ‘post-Roman’ world of the fifth century.
Equally problematic is the long persistence of ethnic interpretations with the application of a static monolithic concept of culture and an uncritical use of written sources. For a long time, studies focused on mapping tribes by specific artefacts and linking these with historically attested migrations (Heeren 2017: 149–153). Despite often good analytical work, the focus on ethnicity as an interpretative framework tended to create dichotomies that placed overgeneralised cultural labels in opposition of each other, e.g. Roman vs Germanic (Esmonde Cleary 2013: 386–394; Roymans and Heeren 2017: 4–5). Despite the generally changed attitude in provincial Roman archaeology after the Romanisation debate leading us away from the culture historical discourse (e.g. Woolf 1997; 2000; 2014), the practice of mapping Germanic ethnicities in the fourth and fifth centuries (e.g. Franks, Saxons, Alemanni) had hardly diminished by the twenty-first century. It can be argued that the influences and insights generated by the discussions following the Romanisation debate—such as a renewed focus on local agency and the impact of globalisation on local communities (Pitts and Versluys 2015) or sociocultural differentiation expressed as merged or redefined communities (Theuws 2009; Collins 2012)—have not yet become commonplace in Late Roman archaeology. For instance, Late Roman material culture is often still considered as ‘less Roman’ based on sentiments related to the aesthetics and sophistication of artefacts and architecture. In addition to a change of perspective, it is equally important to challenge the singular interpretations in the archaeological narratives of the fourth and fifth centuries, e.g. new town walls only serve as defensive structures against Germanic raids, Mediterranean imports and coins are the only measures for economic inclusiveness in the global Roman economy, handmade pottery has to be Germanic because it does not match the idea of a rich and varied Roman material culture, etc.
Gradually more attention has arisen for the slower processes of transformation and ways in which aspects of the Roman world survived after the fall of the Roman Empire (Halsall 2007: 19). It was necessary to broaden the chronological scope to capture such transformation. This led to the development of Late Antiquity as a separate discipline as a means to understand social and cultural change, but also to act as a counterweight to the decline and fall tradition as a holistic explanation for the end of the Roman West (Brown 1971: 7–8). Its development has inspired interests in processes that cannot fully be contained by Roman or Medieval history and archaeology such as the rise and development of Christianity or the transformations of socio-political systems across Europe. At this point, Late Antiquity has become synonymous with the transformation perspective. Although it offers us the benefits of a longue durée approach and a more positive attitude towards the end of the Roman era, its application requires large datasets and a firm understanding of the multitude of factors that play a role in the development of complex societies.
Classification, Periodisation, and Time Perception
The impact and influence of theoretical discourses on archaeological practice and the creation of scholarly biases have been well established. Despite this understanding, it has remained difficult to change our approach to fifth century archaeology. Rarely considered to have played a role are basic scientific practices, more specifically how a theoretical framework is transferred into practical application, how complex information is processed, and how time is dealt with. Because it is not possible to deal with infinite amounts of information in a practical manner, the complex variations of material culture and continuous changes through time and space are compartmentalised into more manageable pieces of information. This compartmentalisation is better known as the classification of data.
If first the necessity of classification to process data is considered, an insight into the longstanding success of the decline model and ethnic discourse can be suggested: both provide a straightforward narrative that can be converted in actual chronologies and simple explanations that can be related to the observed changes in the material record, making it very attractive to apply in data-driven archaeological methods. In contrast, transformation narratives are often more complex and generally cover a larger time span. In addition to vague chronological boundaries, they usually offer a variety of possibilities to explain the observations from the material record, making it less straightforward to apply to a material culture study or in the post-excavation process. The inability to convert the transformation perspective into easily applicable archaeological models might also be explained by the need for classification.
A second influence of classification is the periodisation practice. Periodisation is used to convert an abstract concept of time into everyday use by creating artificial chronological classes of perceived cultures with a clear start, middle, and end (Collingwood 1927: 324). The middle part of a culture usually corresponds to the societal peak for which the largest quantity of sources and information is available. In contrast, the start and end phases of a culture are usually less clearly defined. This indicates that chronological boundaries have been drawn up in transitional periods, between one clearly defined culture and the next. Thus, by organising our understanding of the past into a linear progression of cultures, these transitions have been marked as less attractive periods in the gaps between the historically more interesting periods. Collingwood (1927: 445) phrased it as: ‘… we see history split up into disconnected episodes, each episode forming a relative intelligible whole, separated from its neighbours by dark ages’. In this view, Lucas (2004: 100) has argued that the obscurity of the fifth century is a fictitious problem caused by the incapacity to move beyond the constraints of our chronological systems. While this assessment is highly debatable, it can be observed that the fifth century in Northern Gaul has indeed become the victim of our periodisation practices. While it should mark the transition between the Late Roman and the Early Medieval period, it is generally not included as the end of the Late Roman phase—which is actually placed in AD 410 with the withdrawal of Roman military forces—nor is it fully included in the beginning of Merovingian history, which starts with Clovis c. AD 480. This means that, in the traditional organisation of the archaeological discipline, the bulk of the fifth century is left unaccounted for. Despite many studies and debates, there are still few systematic approaches to refurbish our understanding of the fifth century mainly due to a general scarcity of sources. As a result, the fifth century in Northern Gaul lacks a clear identity, culture, and structure that could be studied systematically. Therefore, the fifth century remains not fully Roman, Gallo-Roman, Germanic, Frankish, Merovingian, or Christian. It does not fit in any of the predetermined categories, time periods, or research traditions.
An additional factor that can be considered regarding the influence of time on archaeological practice are the concepts of time perception and estimation, i.e. how we deal with time on a cultural (perception) and biological/psychological (estimation) level (Allan 1979; Graham 1981; Bailey 2007). Our time perception and periodisation systems are linear, which inspires teleological thinking and cultural sequences in which one culture is replaced by another. Time estimation is more related to how we experience time and remember the past, which is linked to the events taking place during that time. Gilliland et al. (1946: 173) stated that: ‘…[Time] estimation depends on the memory of events occurring within any interval. If the interval is filled with many events it will seem long in recall. If it is uneventful it will be remembered as short’. If the notion of history serving as a societal memory is entertained, it offers a possible explanation to the observation that the first half of the Roman imperial period (c. 27 BC– AD 250) appears longer than the second half (c. AD 250–476) since more source material or ‘memories’ are available from the Early Empire than from the Late Empire. This indicates that our periodisation practices are already biased by how we perceive and process time. Acknowledging this might provide an opportunity to change certain practices and methods.
Increasing the Visibility of the Late Roman Fifth Century
The fate of the fifth century is closely related to that of Late Roman archaeology. It stands to reason that if we are unable to successfully identify and provide a proper interpretation for the archaeological record of the late third and fourth centuries, then progress in fifth century archaeology will also be hindered. As already advocated for Roman Britain by various scholars such as Esmonde Cleary (2013: 466–482), Gerrard (2013: 1–14), and Cool (2014), an extension of the Late Roman chronology of Northern Gaul to c. AD 450 would already make a difference. Since local communities and their material culture rarely show drastic changes that could be traced to a singular specific historical event, including the fifth century in both the Late Roman and Early Medieval period would yield better results. A good example is Childeric’s burial assemblage, traditionally dated to AD 481 (Kazanski and Périn 1988: 20–26). The combination of high class Roman and Germanic objects in the burial assemblage can be considered as clear evidence that Roman culture persisted to some degree (Halsall 2007: 269), at least among the fifth century elites residing in the former Roman West. Some recent studies have already begun to apply this extended chronology and successfully integrated the fifth century in their archaeological narrative, e.g. Vanhoutte (2018) for the Roman fort occupation in Oudenburg and Vanderhoeven (2017) for the Roman town of Tongeren.
The continuity of Gallo-Roman culture into the fifth century is of course not a new idea. The influential work of medievalist Henri Pirenne (1937) already argued for the survival of Roman culture into the Early Medieval period and Herman Fischer (1925) made a case for continuity of the Roman Empire into the fifth and sixth centuries. However, the idea of continuity gained little traction in Roman provincial archaeology. At times, it is difficult to find Gallo-Roman communities past the third century, as many typologies and settlement occupations end in correspondence with the (assumed) barbarian invasions in the third quarter of the third century. This first hiatus created by the end of the third century already decreases our chances of positively identifying the fifth century. If we do not consider the possibility of a continuity of Gallo-Roman culture into the Late Roman period and apply a differentiated approach towards the Germanic communities settling into the Roman provinces, how would it be possible to form an accurate picture of the Late Roman fifth century?
Nonetheless, it remains very difficult to date sites to the fifth century and only a handful of sites in Belgium are well documented. According to Dierkens and Périn (2003: 174), the fifth century has on occasion been referred to as the ‘archaeological hiatus’ in their overview of the twentieth century state of research for the fifth century in Belgica Secunda. They listed burial evidence from the cemeteries at Haillot (Breuer and Roosens 1956) and Vron (Seillier 1989), the study of weapon graves by Böhme (1974), as well as regional studies by Dasnoy (1955) on Namur, Legoux (1998) on Picardy, and Vallet (1997) on Northern Gaul. However, for a territory the size of Belgica Secunda (stretching from the North Sea to the Rhine and Somme), this list consists of very little evidence. Furthermore, most of these sites and studies go back to the mid-twentieth century excavations and interpretations and should be revisited with more recent developments in the understanding of Late Antiquity, the Late Roman Empire, and Merovingian society. Dierkens and Périn (2003: 174–175) did acknowledge that most knowledge of the fifth century is based on a limited amount of funerary evidence, lacks general studies and distribution maps for different artefact types, and that not all regions have been adequately identified and excavated. Unfortunately, this is still the case. Only a few studies with a fifth century focus can be added to this list, such as the ‘Frankish’ rural settlement at Wange (Opsteyn and Lodewijckx 2004: 125–155), the ‘Saxon’ cemetery at Sint-Gillis-bij-Dendermonde (Van Doorselaer and Opsteyn 1999), and the study of fifth century ‘Frankish’ sites between the Rhine and Meuse (Heidinga et al. 1992). Despite an increase in fifth century radiocarbon dates from recent excavations, such as at Lemberge and Merelbeke (Table 3), studies regarding the overall settlement image and material fingerprint of the fifth century remain sparse.
Constructing New Approaches
A final crucial factor that needs to be addressed is the scholarly nature of theoretical debates. The longstanding discussion between ‘romanists’ and ‘germanists’—as described by Pohl (2008: 94)—on the end of Roman rule in the West, or the validity of the decline and transformation models, primarily take place in an academic setting. While many studies are being written from a theoretical point of view, there is a sharp contrast with the limited amount of reliable material evidence that is available for the fifth century. This can be stressed by the use of only seven recurrent chronological indicators in Flanders to identify a find or site as Late Roman, of which only five extend into the fifth century (Table 4, Figure 7). As ‘one-model-fits-all’ scenarios are not applicable in a practical way, it reinforces undifferentiated interpretations of archaeological data and results in circular reasoning. Since all large societal changes are too complex to be contained in one model and these models cannot be directly applied to interpret a specific archaeological site or context, it is necessary to address our biases and flaws in historical and archaeological practices and develop new tools to study the fifth century based on material evidence by investing in detailed studies of well datable contexts and assemblages.
A number of examples from recent research have attempted to create new approaches such as the study of dark earth and post-Roman layers with micro-morphological approaches (e.g. Macphail et al. 2003), or using new technological approaches such as Flückiger’s (2019) combination of 3D-registration, GIS, stratigraphy, and material culture on a context from Augusta Raurica (Augst/Kaisaraugst) in which Late Roman layers were screened using precisely located and dated coin finds to clarify unclear settlement stratigraphies. In her conclusions, Flückiger (2019: 76–77) calls for refining chronologies through secure settlement contexts that eventually can be combined to paint a more detailed picture of the fifth century. In extension of that conclusion, it can be proposed to support typological studies of Late Roman and Early Medieval material culture with archaeometrical techniques to better characterise and date them. Such as by petrographic analysis to determine the actual provenance of Germanic handmade pottery (e.g. De Paepe and Van Impe 1991), or metal analyses of brooches in combination with their socio-historical development to differentiate between various meanings in different archaeological contexts (e.g. Van Thienen and Lycke 2017).
Furthermore, we need to focus on novel ways to approach the interpretation of the fifth century material record and make the invisible visible. For instance, by focusing on social and cultural change like Theuws (2009) did to reinterpret the assumed Germanic weapon graves as the manifestation of new local elites, or how Collins (2012) constructed a new narrative for understanding changes in local communities at Hadrian’s Wall. Finally, customised approaches are needed to investigate the role of invisible activities in rural communities operating on subsistence-based systems, such as increased pastoralism, short-lived farmsteads, and the exchange of organic commodities and livestock.
To conclude, how did we get to an obscure fifth century? In the case of Northern Gaul, an uncritical application of the decline model resulted too often in the use of barbarians as a simplistic deus ex machina to create a one-model-fits-all scenario, which has limited our capacity to identify and study Late Roman and fifth century archaeology. Indirectly, the persistence of the catastrophic view generated hard chronological boundaries that shaped typo-chronologies for material culture resulting in fixed end-dates based on their correspondence to historical events such as the Limesfall in AD 260 and the withdrawal of Roman military forces in AD 410. In turn, these end-dates were then used to establish settlement chronologies that again supported the abandonment narrative. Although the discipline of Late Antiquity and the transformation approach facilitate the study of the Roman to Medieval transition, its broad chronological range and complexity make it difficult to create practical applications that could help identify and date the material culture of the fourth and fifth centuries. Despite many studies and decades of scholarly discussions, much of the fifth century still remains invisible not only because of the persistence of biases and a poor research history, but also because of more basic archaeological practices concerning the need for classification, the use of periodisation, and our time perception. The combination of all these factors makes it difficult to make significant progress in identifying and interpreting the fifth century.
The points addressed in this paper are not intended as a lament of the problems in fifth century archaeology, nor as a criticism of the people who study or excavate it, but primarily to raise awareness of certain issues to counteract current assumptions that limit the identification of new archaeological evidence for this elusive century. Excavators in developer-led archaeology are not necessarily specialised in this time period or aware of all scholarly debates on this matter. If we do not address these biases and limitations where the archaeological information is collected—during excavations and material culture processing—we might never be certain if the fifth century is really absent from Northern Gaul or if we, as archaeologists, are just missing the evidence.