Journal of Archaeology in the Low Countries 2-1 (May 2010)W.A. van Es; W.J.H. Verwers: Early Medieval settlements along the Rhine: precursors and contemporaries of Dorestad

2 Settlements in the Dorestad micro-region

2.1 De Geer near Wijk bij Duurstede

next section

2.1.1 Location

The site is situated to the northwest of the fourteenth-century town centre of Wijk bij Duurstede (fig. 2). Its name De Geer refers to the tapering shape of a residual plot of land in a medieval field system. The excavation site lies on the western edge of a broad river basin immediately beyond the bifurcation of the Rhine in which the meandering Kromme Rijn, after branching northward, has gradually shifted its bed from west to east. An early part of the riverbed survived until the early Roman Period, when it filled up with sediment. By the time Dorestad was built, the Kromme Rijn had shifted some 500 metres eastward. A natural levee formed along its left riverbank and this must have been habitable from at least the early phase of Dorestad, around AD 650/675. Maybe the main stream had been following roughly the same course from as early as the Late Roman Period, but there are no indications that Dorestad’s levee was occupied before the seventh century. As long as this levee was not too high, habitation would have remained concentrated on the older levee of De Geer. The zone between the two levees was unoccupied, this low lying, wet zone is where the Kromme Rijn had twisted its way in intervening centuries.

Before the rivers were flanked by dikes, first built in the Wijk bij Duurstede area in the twelfth century, only the relatively high lying parts of the natural landscape were habitable. Apart from the De Geer levee, these included De Horden, a site just south of the levee. This field name is thought to refer to a ‘corner’ or ‘angle’. The subsoil at De Horden contains remnants of an older river system, predating that of the Kromme Rijn, which after the Early Middle Ages was entirely covered by clay deposited by the river Lek. Here too, long habitation has been attested, but this settlement, in contrast with that at De Geer, failed to survive into the Middle Ages. Some Carolingian finds and the imprints of many cows’ hooves, well preserved by the Lek sediments, tell us that during Dorestad’s heyday the area was used as pasture. Yet the actual settlement here had been abandoned in the late 2nd or early 3rd century AD. For this reason, De Horden will not be discussed further in this article.


Fig. 2 The excavations at Wijk bij Duurstede. 1. phosphate concentration; 2. excavated areas: De Geer, Dorestad, De Horden; 3. prehistoric riverbed of the Rhine; 4. Carolingian period riverbed of the Rhine; 5. modern riverbed.

2.1.2 Occupation phases

The excavations at De Geer, which took place between 1989 and 1994, produced a vast quantity of data. These have not yet been studied in every detail but the outlines of the settlement history are clear. The use of metal detectors meant the recovery of a considerable number of metal finds. These include about 175 brooches, which have been dated by J. van der Roest. Table 1 shows their chronological distribution and offers a first impression of the periods during which De Geer was occupied. The High Middle Ages, which at De Geer constitute the final occupation phase (not counting modern times), are not represented by brooches.


Table 1. De Geer: percentual distribution of the brooches over the various occupation phases (N = c. 175); after J. van der Roest.

2.1.3 Prehistory

The Bronze Age and Iron Age in the Netherlands are poor in brooches and the oldest brooches at De Geer are indistinct fragments. The first settlement traces at De Geer and De Horden date to the Middle Bronze Age (1500-1300 BC) (Arnoldussen 2008). No finds are known from the ensuing half millennium. Settlement at De Horden resumed in the Early Iron Age (700-500 BC) (Hessing 1989, 1991; Hessing & Steenbeek 1990).

An initial survey of the handmade pottery suggests that people returned to De Geer in the Middle Iron Age (from +/- 400 BC) (Linnemeyer 1995). By then, De Horden seems to be deserted once more, though there are thought to be burials dating to the Middle and Late Iron Ages. This site also shows how parts of the landscape changed through time. From the Late Iron Age on, both De Horden and De Geer were inhabited again and the amount of brooches increased steadily.

2.1.4 The Roman Period

An important feature shown in table 1 is that there is roughly an equal amount of brooches from the Middle and Late Roman periods. As the two periods were of roughly equal duration and both rich in brooches, this must mean that De Geer was a substantial settlement in the Late Roman Period. This conclusion is supported by a comparison with De Horden. There 270 brooches came to light, dated by Van der Roest the final decades BC to the final quarter of the 2nd century AD (Van der Roest 1988). Late Roman brooches are absent from De Horden as the local settlement had by then been deserted. The Roman coins complete this picture. From a total of 240 coins at De Horden, 95% predate AD 260. Late Roman coins there are no more than incidental finds, reflecting occasional activities at a deserted dwelling site. At De Geer the ratio is quite different, c. 25% of the 280 coins predate AD 260 and 75% date from between 260 and 388 (Aarts 2000, 284, 291; see also Vos 2009, notes 136 and 222). Moreover, De Geer yielded numerous other metal finds from the so-called Foederatenhorizont, such as hairpins with faceted ornamentation and belt fittings (fig. 3). A very unusual find is a Byzantine coin weight dating from around AD 400 (Van Es 1991). All in all, the Late Roman period turns out to be particularly well represented at De Geer, making this a rather special site.

De Geer and De Horden were both inhabited during the first half of the Roman period. In his doctoral thesis, W.K. Vos recently devoted an extensive and most illuminating discussion to these settlements (Vos 2009, 59-116). He convincingly argues that since Flavian times these two, together other settlements, were united in a single system of field division which had been laid out by some higher authority (fig. 4). Within it, De Horden formed the principal local element. Here an estate developed comprising one or more farmsteads, which must have belonged in the top echelon of rural settlements in the Batavian Kromme Rijn region. At De Geer, the second century saw the construction of an ‘enclosure’, smaller than that at De Horden, which probably was a continuation of Iron Age and Early Roman Period habitation. Around AD 200 De Horden had to be abandoned because of increasing drainage problems but habitation at De Geer continued into the third century.


Fig. 3 De Geer: Late Roman pins and belt fittings.


Fig. 4 Field systems at De Geer and De Horden; after Vos 2008, fig. 3.32.

Vos reckons that habitation in this area may have been interrupted between AD 250 and 350. There certainly was a discontinuity in this period due to the first Frankish incursions into the frontier zone of Germania Inferior. Around 250/275, and often even earlier, all known Batavian settlements sank below the horizon of archaeological visibility. Were they completely abandoned or did a remnant population stay on? This answer is not crucial to the current discussion. Even if De Geer was temporarily abandoned at this time, we may still safely speak of continuity (the briefer the abandonment, the better, of course). Order was restored by Constantine (the Great) in the early fourth century. Any settlement abandonment would not have lasted for more than a generation and traces of the Middle Roman-period field system would still have been clearly visible when the site was re-occupied.

2.1.5 Frankish pottery: Frankish colonisation

Pottery is among the most important finds from De Geer. The archaeological potential of this material is evident from E. Taayke’s authoritative study of the local ware from De Horden (Taayke 2002). From this study it emerged that De Horden initially maintained contacts with the north-western coast of Frisia. Soon, however, the ‘Batavian’ component in the pottery eclipsed the Frisian ware.[1] This reflects the development of the civitas Batavorum as a socio-economic entity. The limes increasingly became a cultural border.

After the mid-third century, Germanic handmade pottery once more found its way south of the Rhine and even into Belgium through breaches in the frontier line (De Paepe & Van Impe 1991; Rogge & Van Doorselaer 1990). De Geer yielded many potsherds of the Rhine-Weser Germanic (RWG) style group. The coarse ware category is especially well represented mostly via large pots (cauldrons) tempered with crushed shell (the shell now has mostly disintegrated, leaving just cavities) and globular neckless bowls or pots with an S-shaped profile whose rim and/or shoulder may be decorated with (fingertip) impressions (fig. 5). By contrast, fine wares, especially the Von Uslar II situla, are very rare (Von Uslar 1938). There are some fragments in a comparatively thin and smooth walled, fine sandy fabric but these are more reminiscent of Anglo-Saxon Schalenurnen, or they are sherds of situlae of the types Wijster IA (especially IC and ID) (Van Es 1967).

Good parallels for the coarser ware have been found at Bennekom, a Free Germanic settlement on the far side of the Rhine a small distance east of De Geer (Van Es et al. 1985). There are just a few Wijster-type situlae there. These point to links with the coastal region of Germany and the northern Netherlands. Possibly the suppliers of this ware at De Geer came from Drenthe, where the pottery styles of the RWG group and the North Sea coast are found together. However, the provenance of the Late Roman handmade ware at De Geer may be anywhere along the western flank of the RWG region on the sandy soils north of the Rhine from Gelderland up to Drenthe. It is impossible to date this handmade pottery at De Geer more closely than to the third-fifth centuries AD.

This kind of handmade pottery is associated with settlements in which rectangular longhouses were the principal components. Unfortunately, there are no distinct house plans, a problem that affects all settlement periods at De Geer as a result of long habitation in a fairly confined area and medieval and later tilling. So far, it is only at Tiel-Passewaaij, not far from De Geer on the Roman side of the Rhine, that two distinct plans of Germanic houses of the Later Roman period have been uncovered (Heeren 2009, 71-74). One of these two is a short structure of the special type Wijster IIa, which is known from the sandy soils from Drenthe all the way down to the Rhine and thus is a positive indication of settlers from those parts (Van Es & Taayke 2001).


Fig. 5 De Geer: Late Roman handmade pottery. Scale 1:4.

The plans at Tiel are attributed to Phase 7, i.e. between AD 270/290 and 350. Similar houses would also have been introduced by Germanic settlers at De Geer. There, farmstead sites rather than house plans can be recognised, but the size of the settlement cannot be ascertained. It appears to be centred on an older Roman ‘enclosure’.

The conditions under which Germanic settlers recolonised the frontier zone are unclear. Did they settle on their own terms or under Roman supervision? Probably it was a bit of both. Even in the case of autonomous colonisation, some form of accommodation with the Roman authorities would have been inevitable, certainly during episodes of revived Roman authority. In any case, people were happy to make use of what remained of the Roman infrastructure. Evidence of this is the presence of Late Roman coinage and ceramics. Not only did fourth-century cooking pots of Mayen ware reach De Geer, but also rouletted sigillata ware, painted ware in ‘d’ technique and terra nigra-like drinking cups of the type Chenet 342, although the latter can no longer be confidently traced to truly Roman production centres (Bakker 1997).

Near the Kromme Rijn and upstream, in the eastern river region, fourth-century settlements like that at De Geer were much less numerous than their predecessors from the Middle Roman period, but still far from rare. As in Batavian times, they would have had an agrarian basis but many of their young men are likely to have opted for a warrior’s life in the Roman legions or elsewhere. In this respect too, very little had in fact changed. However, the inhabitants of the frontier zone now no longer called themselves Batavians, but Franks, and they were largely made up of (possibly) expressly invited Germanic colonists from beyond the Rhine. Not all of these Late Roman period settlements withstood the ravages of time. At Tiel-Passewaaij, for instance, habitation abruptly ceased in the first half of the fifth century but at De Geer the sequence of archaeological finds continues unbroken right into the days of Dorestad. Although archaeological finds never offer total certainty, we can rule out long periods without any occupation between the fifth and ninth centuries (fig. 6). Why continuity of habitation is clearly evident at De Geer and not throughout the region is a question that the current state of research cannot yet answer. Its location in the principal corridor of the Rhine system and close to a (former) Roman frontier fort may have played a part.


Fig. 6 De Geer: Merovingian rough-walled wheel thrown pottery (Wölbwandtöpfe). Scale 1:4.

2.1.6 Early Middle Ages

The origin of the Early Medieval settlement at De Geer is as yet unclear. No clear-cut plans of dwellings have been identified. They must have been rectangular, timber-built farmhouses, probably quite like those of the types Odoorn A to C in Drenthe (Waterbolk 2009). Similar houses are known elsewhere in the Rhine delta, for instance at Rijnsburg and Oegstgeest (Van Es 1972; Hamburg & Hemminga 2007). Comparisons are also provided by Dorestad itself, where colossal timber houses were built (fig. 7) (Van Es & Verwers 1995).


Fig. 7 Excavation plans of 1. Rijnsburg, 2. Oegstgeest, 3. Dorestad. Scale 1:200.


Fig. 8 Dorestad: plan of a timber farmhouse with pits. Scale 1:200. 1. house 1 phase 2; 2. house 1 phase 1; 3. house X; 4. wattle fencing?; 5. house Y phase 2; 6. house Y phase 1; 7. house Z; 8. Well; 9. spicarium, granary.

The sites of the actual houses at De Geer are often recognisable by rows of pits running parallel to the gable ends, a phenomenon also known from Dorestad where multiple rows of pits may accompany the same house (fig. 8). These pits may have served a variety of functions, such as rubbish pits or latrines. When the gable ends were moved in the event of refurbishment or extension, it seems that the pits were also relocated. Moreover wells, lined with wine casks or hollowed tree trunks, were commonly used at De Geer (Verwers & Botman 1999).


Fig. 9 De Geer: topography. 1. western levee along riverbed remnant; 2. Roman complex; 3. Carolingian complex; 4. moated dwelling site; 5. limits of excavated area; 6. phosphate concentration.

Their position relative to the farms is, however, difficult to ascertain. The latter seem to focus on the site of the Roman ‘enclosure’, which included the highest part of the site. Here, clusters of pits point to the presence of ten, possibly twenty, houses (the picture is quite uncertain), which of course need not have existed simultaneously. However, the settlement was not limited to this part of the site and has not been fully excavated. The settlement’s size therefore remains uncertain but it must have been a fairly large complex comprising at least five, and probably more farmsteads. In as far as can be ascertained, the farmsteads were oriented roughly north-south, parallel to the axis of the levee, and were built on the levee and the sediment-filled riverbed remnant beside it.

Major changes occurred at De Geer towards the end of the Dorestad’s lifespan or perhaps somewhat later. An elongated ‘enclosure’ measuring 90 x 380 m was laid out partly on the levee and partly on the former riverbed beside it (fig. 9). The surrounding ditches were mostly over 2 m wide and it seems likely that the excavated soil was made into a bank.


Fig. 10 De Geer and Dorestad: plans of timber-farmhouses. Scale 1:200. 1. De Geer; 2. Dorestad ( Van Es & Verwers 1995 , no. 18).

No traces of any bank could be identified, however, since the excavation level lay below the Early Medieval surface. The ‘enclosure’ consisted of at least two parts which connected at a slight angle. The northern part was c. 220 m long and the southern part, c. 160 m. Possibly the two parts were not constructed at the same time and the fact that the join coincides with the old Roman ‘enclosure’ can hardly be coincidental. A ditch running parallel to the complex on the west side and disappearing beneath associated ditches hints at an even more complex site biography. The ditches of the large ‘enclosure’ cut across Early Medieval habitation features in several places. One of the very few distinct house plans at De Geer possibly belongs to the ‘enclosure’. It is a timber structure of 9 to 10 x 25 m and these dimensions match the large and comparatively late farmhouses at Dorestad (fig. 10) (Van Es & Verwers 1995, e.g. fig. 8). The distinctive plan at De Geer also indicates a late date. That this building, most probably a farmhouse, is coeval with the system of drainage ditches is suggested by its position parallel to the longitudinal axis of the ditch system and in line with the kink where the two parts of the system link up. This does not mean that the house was the only one within the ditch system but no other clearly associated features were distinguished.

The dating and interpretation of the large ‘enclosure’ remain problematic. The earliest pottery in the ditches is Carolingian but many or all of these sherds may have been brought to the surface from older settlement features as the ditches were dug. Among this pottery the youngest varieties are late Badorf/early Pingsdorf ware dating from the latter half of the ninth century, possibly even the early tenth. There doesn’t seem to be very much of this late material. Younger pottery is absent from the ‘enclosure’s’ ditches. All of this suggests that the ditch system dates from the ninth century, possibly contemporary with late Dorestad or even somewhat later.

After the Carolingian occupation phase, De Geer seems to have been abandoned for about three centuries. This was followed by a final occupation phase between c. 1250 and 1400. Then a moated dwelling site (begraven hofstad) arose at De Geer: a small, brick-built castle with a farmhouse surrounded by a system of moats which, in contrast to the older system, was laid out in a neatly rectangular fashion (Van Doesburg 1994). The east side of the dwelling site exactly coincides with the centre of the older enclosure.

Just as may have happened in the Middle Roman period, after the Carolingian period De Geer temporarily changes from a ‘site’ into an ‘off-site’ without there necessarily being a break in the chain of owners or users. The founder of the thirteenth-century moated dwelling may have been a descendant or some other legal successor of the owner of the large ditch complex. The seigneurial (knightly) status of the High Medieval owner is beyond doubt. Might his predecessor also have belonged to the elite of his day? The huge array of ditches (and possibly banks) seems to suggest this. Thus far we have regarded this complex as a fortified refuge coeval with the (final) phase of Dorestad (Van Es & Hessing 1994, fig. 193). We now consider this explanation a little fanciful. The defensive potential of the ditches and possible banks cannot have been very great. We are now more inclined to regard it as a demesne, the seat of a rich landowner or his representative. At any rate it is evident that at the end of the Dorestad epoch or shortly after, a far-reaching reorganisation took place at the settlement of De Geer but it is hard to archaeologically determine what happened with a degree of precision. Unfortunately, we lack any documentary evidence that might shed light on the course of events.

2.2 Rijswijk and Leut near Wijk bij Duurstede

2.2.1 The castellum Levefanum

It had long been assumed that the name Levefanum on the Peutinger Map referred to a Roman frontier castellum in the vicinity of Rijswijk (province of Gelderland). This hypothesis was corroborated in 1979, when dredging in the floodplain of the Lower Rhine at Rijswijk brought up large numbers of Roman finds among the sand and gravel. In conjunction with our excavations at Dorestad, archaeologists aboard the dredgers over a period of about six months collected samples of these finds. However voluminous, the gathered material represents just a fraction of what must have been dredged up in the way of artefacts (Van Es 1984). The vast majority were lost and others were dispersed in poorly accessible private collections. Our findings relating to Levefanum are largely based on the material from the sample. This consists mainly of potsherds but also includes parts of Roman military equipment, especially fragments of helmets, which demonstrate the military character of the site. Moreover, the composition of the ceramic assemblage points to a fully Roman rather than a Batavian settlement (Table 2).

Thanks to the dredging finds, the castellum Levefanum has now been approximately located. In all probability, the actual fort was eroded away by the Rhine as the finds came from a deep sandy deposit covered by three metres of clay. The strategic location of Levefanum, where the rivers Lek and Rhine diverge, is comparable to that of Vechten and Utrecht, the next frontier forts downstream which guarded the forking of the Kromme Rijn and Vecht. On this basis we may assume that Levefanum covered about the same area as these two castella, between 1 and 2 hectares. The history of Levefanum will in the main have matched that of the other forts in the Dutch frontier zone. Hence this castellum would also have been rebuilt in brick and stone during the third century. The defensive wall and the (main) buildings within are likely to have been kept in a more or less serviceable state right into the early fifth century.

A standard feature of every frontier fort was the vicus, the ‘civilian’ service settlement adjoining the fort on two or three sides and covering an area at least equal to it. The name Levefanum has been emended by Stolte to ‘Haevae Fanum’, the sanctuary of the otherwise unknown and possibly regional goddess Haeva (Stolte 1963). If this is correct, her temple may have stood within the vicus of the castellum or in its immediate vicinity. In that case, this area was of more than local significance, maybe even from before the arrival of the Romans. The dredged-up helmets may have belonged to a deposit of votive gifts.

The third century must have also resulted in great changes at Levefanum. Our Roman dredging finds mainly date to the Middle Roman period. Late Roman finds are absent from our collection. We are unaware of any fourth-century potsherds or other finds, such as coins, in private collections.

Given the conditions under which the archaeological documentation of Levefanum took place, a lack of certain finds does not justify the conclusion that the site was deserted. Hence it remains an open question whether the castellum Levefanum cum annexis ever accommodated a Late Roman or Frankish settlement, like those so clearly evident at De Geer and arguably present at the nearby castellum of Traiectum (Utrecht) (Van Lith de Jeude 1993).

There are no other dredging finds for subsequent periods until the Merovingian period. It is difficult to tell exactly when, but in the course of the seventh century at the latest there definitely was habitation at the site of Levefanum and sherds indicate that this occupation phase continued beyond the days of Dorestad.[2] A comparison with the pottery excavated at Dorestad suggests that the castellum site was reoccupied before the middle of the seventh century. Whether the site was unoccupied throughout the intervening centuries remains a moot point because of the scarcity of the available evidence. The name Levefanum at any rate was lost, as at the end of the seventh century the written sources speak of ‘castrum Dorestad’, which must refer to the former Roman frontier fort.


Table 2. Rijswijk/Levefanum: sample of Roman pottery (dredging finds 1979); after Van Es 1984.

2.2.2 Rijswijk / villa Risuuic

In a List of Landed Property of the diocese of Utrecht, Rijswijk features as ‘villa Risuuic’ (Henderikx 1987, especially 95, 122). This list was compiled in the late ninth and first half of the tenth century in an attempt to reclaim lost possessions of the Chapter of St Martin’s, and hence reflects an earlier situation. It relates to the Utrecht diocese’s landed property acquired between the first quarter of the eighth century and the years of Viking domination (c. 860-885). At Rijswijk this included part of a villa, comprising a church with its land and three other mansa. The list tells us nothing about the full size and the layout of the settlement. The presence of a church suggests a fair-sized habitation, which we might envisage as an ordered complex of farms, comparable to that of De Geer, but with a church. If we presume that the latter stood at roughly the same spot as the present-day village church of Rijswijk, the villa would have been at a distance of no more than a few hundred metres from the castellum. Maybe it was even located within the vicus. We might indeed speculate that the first (wooden?) church Christianised the old pagan cult site of Haeva. Such a (private) chapel is unlikely to have been founded before the late eighth century. The founder is unknown. It is generally assumed that the territories of the Roman castella became royal estates. Might the villa have been part of this?

There are two main possibilities for the origins of the villa. Either it was a new foundation in Merovingian times or it developed from an older habitation, going back to at least the Late Roman period. In the latter case, it may have evolved from the Roman vicus, possibly with a connecting Frankish phase. The parallels to De Geer would in that case be very strong, but without further evidence this must remain mere conjecture. Nor do we have any indication as regards the donor’s identity.

2.2.3 Leut / Lote villa

Archaeologically nothing is known about the Lote villa. Not a single find can be attributed to it with certainty.[3] Even the location of Leut has never been pinpointed as it ceased to exist as a separate settlement when it was incorporated into the town of Wijk bij Duurstede. It is believed to have occupied the opposite (right-hand) bank of the Kromme Rijn but survives only in the field name ‘Leuterveld’ (Dekker 1983). There is no doubt about its existence as a contemporary of Dorestad as it is mentioned in the Utrecht List of Landed Property. Therefore it merits consideration as one of the possible roots of Dorestad. At the Lote villa, the chapter of St Martins possessed a church with its land and seven other mansa, but not the royal tithe (Henderikx 1987, especially 96, 122).[4] Much of what was said about Risuuic may also apply to Lote. In any case, the presumably private chapel also points to high-ranking donors.