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

3 Other Early Medieval settlements along the Rhine

3.1 Roomburg / Matilo

next section

Close to present-day Roomburg lay the Roman frontier fort Matilo which guarded the Fossa Corbulonis where it joined the Rhine. Thus the site is a parallel to Rijswijk/Levefanum. Recent excavations have shed light on Matilo. This research focused mainly on the canal banks and on the vicus to the west of the fort.[5] We still know very little about the castellum itself and since what remained of the site after the building of a housing estate has been declared an archaeological monument, there is little prospect of new evidence emerging in the near future[6]. The occupation history of Matilo will on the whole be quite similar to that of other castella along the limes in the Netherlands. This means a transition from timber construction to building in stone and brick and regular use well into the third century.

So far, finds from the fourth and fifth centuries are rare at Roomburg. Only one brooch dates to the fifth century and the fourth is represented by one coin, a few brooches and some pottery. W.A.M. Hessing, the excavator at Roomburg, considers this insufficient evidence to prove continuous habitation at or close to the castellum. He believes the finds to reflect the occasional presence of small numbers of ‘visitors’, but in fact it is rather premature to rule out more intensive activity. The excavations have so far been quite small-scale and the fifth century in particular is notoriously hard to detect in this part of the world. We are dependent on handmade ‘Frankish’ or ‘Anglo-Saxon’ pottery, which is difficult to date and interpret.[7] As regards Matilo, it would be prudent to suspend judgment for the time being.

The same problem occurs with other frontier forts that have yielded fourth- and/or fifth-century finds. The 1986 survey by W.J.H. Willems remains an excellent overview (Willems 1986, 452-456). Evidence of Roman military use in the fourth century at the castella is mostly limited to the stretches of the limes upstream from Utrecht. One of these is the fort Maurik/Mannaricium, situated near Wijk bij Duurstede not far east of Rijswijk. This also supports the likelihood of a Late Roman phase at Levefanum. Downstream from Utrecht, only the mouth of the Rhine seems to have remained fortified at the castella of Valkenburg (province of Zuid-Holland) and presumably also Brittenburg, which was subsequently lost to the sea. The rich river region east of the Utrecht-Rossum line remained more markedly Roman during the fourth century, being more closely linked to the Belgian-German hinterland.

There are only two castella where fifth-century and possibly also Merovingian finds might point to continuous habitation, Utrecht/Traiectum, and Meinerswijk/Castra Herculis near Arnhem. In both cases, the evidence results from comparatively small excavations. Therefore full continuity can be warranted in neither case or, as Willems concludes in the case of Meinerswijk, the finds are ‘evidence for either continued, renewed or intermittent occupation’. But he is hopeful: ‘... future research will undoubtedly produce evidence for several Early and Late Medieval phases’ (Willems 1986, 352-353).[8] Given its position not far from the mouth of the Rhine, there may yet be hope for Matilo in this respect, but any verification will require more digging there. At Nijmegen, which can boast a huge volume of excavations, there is no longer any doubt as to the continuous transition from the local Late Roman fort, via a Frankish and a Merovingian phase to the Carolingian imperial palace (Willems & Van Enckevoort 2009, 95-105).

From the sixth century onwards, habitation at Roomburg is no longer a matter of debate. Its start is impossible to pinpoint closely. A pseudo-coin brooch found close to the castellum dates from the end of the sixth century at the earliest, but habitation is likely to have started before then. The number of Early Medieval pottery finds is quite large, suggesting a settlement of some significance. These were mainly recovered from confined areas on both banks of the Fossa Corbulonis immediately west of the castellum. Clearly the former vicus site was in use during Merovingian and Carolingian days (possibly without interruption). This definitely applies for seventh and early eighth centuries, as new revetments are shown to have been put in along the Fossa between 620 and 625, between 680 and 690, and between 714 and 716. Evidently the settlement was oriented on the waterway. One of the definite attractions of castellum sites was that they offered good access to the Rhine. The old harbour facilities apparently were still quite serviceable after a few repairs.

There is not enough evidence to determine the nature and the extent of the Early Medieval settlement at Roomburg. Neither do we know whether it included the actual fort. Early Medieval reuse of a former castellum site on the limes has been attested at Utrecht and at Valkenburg (province of Zuid-Holland), not coincidentally two of the very few Dutch castella where excavations have taken place.[9] Utrecht became the seat of the Frisian missionary and Valkenburg may have been a manorial centre. This may also have been the case with the Roomburg castellum. The pits at the former vicus site contained evidence of artisanal activity but this does not necessarily mean that the settlement specialised in trade and industry. Artisans were also active at mainly agrarian villae. The harbour precinct on the Rhine and the Fossa Corbulonis would have belonged a larger complex which undoubtedly also included a number of farms and maybe the actual castellum. Its location and facilities made it possible for Early Medieval Roomburg to operate as one of the smaller trading settlements in the Dutch river delta that coexisted with Dorestad. Other examples are Utrecht, where a commercial district evolved beside the ecclesiastical precinct, and Meinerswijk near Arnhem, possibly the vicus where, according to a written source, Frisian traders vainly sought refuge against raiding Norsemen (Lebecq 1983, 2 (corpus), 314 & 336).


Fig. 11 Reconstruction of the Carolingian settlement at Koudekerk (after Van Grinsven & Dijkstra 2006 ).

All of this of course is strongly reminiscent of Rijswijk/Levefanum. Maybe we should also visualize Rijswijk/Levefanum as a complex encompassing a castellum, a revitalised vicus and a number of farmsteads in the background. Hessing (Brandenburgh & Hessing 2005) makes a similar argument for Roomburg. He surmised that in the eighth to tenth centuries Roomburg included ‘several dispersed farms under the supervision of a steward, who was a member of the local elite, and whose home was on the site of the former castellum’. At Levefanum such farms may have occupied the site of today’s Rijswijk (the Risuuic villa). It is hoped that the location of the farms at Roomburg will be revealed in the future. Maybe they were not all that dispersed after all. A clearer picture of an Early Medieval is offered at Koudekerk.

One final note about Roomburg, it is said that original crown land there had been donated to the church at Utrecht only to have been embezzled by the count of Holland during or in the wake of the Viking raids. Rodenburg castle was built here in the thirteenth century and it is reminiscent of the moated dwelling site at De Geer. However, the developments that prompted its construction fall outside the scope of this article.

3.2 Koudekerk

The archaeological evidence relating to Koudekerk is comparatively plentiful, mainly thanks to an excellent recent publication (Van Grinsven & Dijkstra 2006). Excavations in 1978-1979 covered an area of 1.7 hectares. The earliest habitation was found to date from around the beginning of the Christian era. This settlement is thought to have been abandoned as a result of the establishment of the Rhine limes. It lay on the opposite side of the river but probably still fell under the prata legionis, a fiscal zone where people might visit but not settle. On this basis, the chain of habitation and use may be extended into at least the third century and possibly the zone’s Roman imperial status cast an even longer shadow. The thread of continuous use, however, is a thin one, as the next Merovingian-Carolingian habitation phase is thought not to have started until around AD 500. Maybe this starting date, given the presence of some fifth-century brooches and mortar fragments, could be brought forward a little but it still leaves a gap of two centuries or more. Continuous occupation from Roman times onwards is ruled out, at any rate in the excavated area. Interestingly, the Early Medieval habitation occupied roughly the same site as that of the beginning of the first century AD.

The Early Medieval settlement lay midway between two former limes forts, 6 km upstream from Roomburg (fig. 11). Thus there was no direct link with a castellum site. Neither was it located on the bank of the Rhine, with which it had no navigable connection. An ancient crevasse gully connected to the Rhine bisected the habitation sit, but by Merovingian times most of it had already become filled with sediment and was merely used for drainage. The distance to the river was just a few hundred meters. This geography is reminiscent of De Geer and parallels are also found in the nature of the settlements. Merovingian-Carolingian Koudekerk was a complex of farms laid out on a rectangular grid system on either side of the rudimentary gully (fig. 10). The complex consisted of six farmsteads, each with a farmhouse, a well and the usual outhouses. There were six farms and maybe more since the excavation did not cover the entire settlement. No house plans were recorded but the evidence of their clearly recognizable sites indicates that they were rectangular longhouses measuring about 6 x 20 meters. All six or more are thought to have been occupied contemporaneously.

This evidence pertains to the Merovingian period. It is, however, doubtful that the excavations fully cover this period. No habitation traces of Carolingian times were found, though there were finds of the period. Could it be that the Carolingian settlement lay just outside the excavated area or were its traces obliterated by subsequent land use such as clay digging? Unfortunately, Koudekerk resembles De Geer even in the poor preservation of its archaeological features. Therefore we can only assume that the settlement pattern remained essentially unchanged. It is the finds which make it plausible that occupation at Koudekerk was continuous between AD 500 and 800/850.

3.3 Oegstgeest, Rijnsburg, The Hague-Frankenslag, Katwijk-zanderij Westerbaan and Valkenburg-De Woerd

Parts of similar settlements have been uncovered near the mouth of the Oude Rijn (Hamburg & Hemminga 2006; 2007). Oegstgeest-Rijnfront and Rijnsburg are fine examples. Oegstgeest is dated between 525/550 and 800 and Rijnsburg from the seventh to the tenth centuries. No Roman precursors were found, probably because these Early Medieval settlements, like Koudekerk, lay on the opposite side of the Rhine in the former Roman military zone. Both had direct or indirect access to the river. It is clear that the excavations revealed parts of farmstead clusters laid out on a grid. Rijnsburg is thought to have been much larger than the excavated area. The same structure was observed at The Hague-Frankenslag. The excavations at Katwijk-zanderij Westerbaan were fairly extensive. Still, the recently published detailed excavation report does not manage to answer all the questions (Van der Velde 2008).

The sand quarry (zanderij) Westerbaan lies on the Roman side of the Rhine and here an earlier occupation phase did precede the Early Medieval settlement. The situation parallels that of De Geer in several respects. The Roman occupation phase ended in the second half of the third century and here too was followed by an episode marked by a lack of evidence either of occupation or of abandonment. This quandary is due in part to the difficulty of dating imported pottery from the late third and early fourth centuries and the fact that its production went into crisis in the second half of the fourth century (Steures 2009).[10] At any rate, the Early Medieval settlement in the north of the excavated area immediately adjoins the Roman settlement.

The beginning of the Early Medieval phase at this site is dated to about 450/475. From then on it was intensively occupied right into Carolingian times. Its heyday is thought to have been between AD 550 and 700. The layout of the settlement is hard to reconstruct. There seem to have been several habitation nuclei in a methodically parcelled landscape. The largest excavated settlement fragment is a complex of three or four adjoining farmsteads presumably laid out along a road. The full extent of this settlement site is unknown. The area as a whole may have accommodated some 16 to 22 farms, allowing 4.5 to 6 hectares of arable and grassland for each. Apart from farmhouses – longhouses – of normal length, there also were short houses with little byre space, which may indicate a differentation in terms of wealth and/or occupation among the occupants. The number of inhabitants of Katwijk-zanderij Westerbaan is estimated to have been about 100 to 165. There is no doubt as to their (mainly) agrarian way of life.[11] According to Blok, the Early Medieval name for Katwijk was Houerathorp, which is believed to mean ‘village of farmhouse dwellers’ and maybe this name referred to one or more of the habitation nuclei uncovered at the sand quarry (Van der Velde 2008, 409-410).

It remains unclear whether the Early Medieval farms of Katwijk-zanderij Westerbaan were part of any villa or villae. The settlements were of a fairly dynamic nature and in the course of the Merovingian-Carolingian era at least partially shifted their location. The origin of the settlers is uncertain. Those who exploited the third-century farms are called newcomers, arriving either from north of the limes or from elsewhere within the region. Given the problem of continuity in the Late Roman period, the question arises whether the fifth-century inhabitants can also be branded as newcomers. On the basis of his analysis of the metal finds, Knol’s conclusion is that initially Frankish connections prevailed, followed by strong Frisian links (Knol 2008).

Thus far, all excavations around the mouth of the Rhine have only revealed settlements made up of multiple dwellings. There seem to have been no scattered, isolated farmsteads.

Imports and traces of artisanal production have been found in all of these settlements. This implies that they were part of a (supra-)regional, indeed international socio-economic exchange network (Van Es 1990) but not that they were specialised trading or production sites. Artisanal activities up to a certain point were an everyday part of the farming economy. There is no evidence that the settlements in question exceeded the agrarian ‘standard’. The excavated part of Oegstgeest-Rijnfront does not appear to warrant the conclusion that that it ‘occupied an important position in the region and functioned as a trading and (production) site of regional significance’ (Hamburg & Hemminga 2006, 307). Such places without doubt did exist in this region and, in our opinion, are sooner to be expected at the former vicus sites near castella, as at Roomburg. Another good candidate is the (also very partially excavated) Early Medieval site of Valkenburg-De Woerd.

The history of Valkenburg-De Woerd begins at the establishment of the Roman limes. In the mid-first century AD a military entrepot harbour was laid out here, which must have been part of the vicus of castellum Valkenburg. The distance between the two is just over half a kilometre. After studying the recovered terra sigillata stamps, J.H.F. Bloemers and H. Sarfatij believe that the larger settlement which in the second century replaced the port, became more ‘civilian’ in character (Bloemers & Sarfatij 1976). Yet this is unlikely to mean that De Woerd ceased to be fiscal territory and that we are now dealing with a purely civilian vicus. Roman occupation ceased around AD 230. Between that year and the seventh century there is a lack of finds. If we want to propose any kind of continuity between the Roman and the Early Medieval occupation phases, this must again be on the strength of Early Medieval princes harking back to Roman imperial prerogatives.

Our image of Early Medieval Valkenburg-De Woerd can only be based on the very provisional evidence published in 1986 and 1988, immediately after the excavations (Bult & Hallewas 1987; Bult et al. 1990). The results were not very clear-cut, the reconstructed house plans in particular being quite different from what might be expected. In our opinion the settlement is best characterised as a miniature Dorestad. Its physiography perfectly matches the part of Dorestad excavated along the Hoogstraat. In both cases the settlement was laid out along the inner curve of a Rhine meander, on a natural levee originating in or directly after the Roman period in a river basin in which the meandering river shifted its bed from west to east. This natural process continued after the Early Middle Ages. At Valkenburg-De Woerd, the habitation might have reoccupied part of the site of the Roman settlement behind it (i.e., west of it). The layout of Early Medieval Valkenburg-De Woerd also seems to parallel that of Dorestad. The shore was divided into fairly narrow plots at right angles to the river. The width of the plots it still difficult to ascertain but a riverside abutment is thought to have been about 12 metres wide. According to the excavators, the houses were rectangular, measured 5.5 x 10 to 11 metres, and in some cases were aligned one behind the other, all at right angles to the river, like the plots. Their plans are so unusual, however, that one wonders whether these posthole patterns are not more likely to represent substructures of houses. Just as at Dorestad, wells are remarkably numerous and often placed in rows. There is evidence of bone and antler working and of livestock rearing or, at any rate, the butchering of fairly young livestock.

If anywhere, it is at Valkenburg-De Woerd that archaeology has hit upon one of those small, secondary trading settlements of which there must have been many in the delta and which are occasionally referred to in written sources, e.g. at Meinerswijk. It is an interesting possibility that De Woerd and the castellum were still part of a single estate even in the Early Middle Ages. The church in the castellum is believed to have been one of the earliest in the western Netherlands and according to the List of Landed Property it belonged to St Martin’s at Utrecht. The donor must have been none less than the king. It is possible that De Woerd was also crown land. The List of Landed Property expressly mentions the church cum omnibus appendiciis ... totum et integrum, but it is unclear whether this included De Woerd (Bult et al. 1990, 165). For the early church of Oegstgeest, a foundation date ‘in the first quarter of the eighth century or a little later’ has been suggested as part of a donation to St. Willibrord (Bult & Hallewas 1990, 86; Halbertsma 2000, 175-179, 203-205). That Oegstgeest-Rijnfront was anything other than a rural settlement is unlikely, if only because of its location on the Rhine across from Valkenburg-De Woerd but on the bank that started to be eroded in the Early Middle Ages. It is imaginable that De Woerd and Rijnfront were complementary settlements, which would present an interesting parallel to Leut situated across the river from Dorestad.

The farming settlement at Rijnsburg is one of those rare cases where the archaeological evidence is supplemented by the names and social background of its owners. In an important study, Sarfatij demonstrated that the excavated remains belong to the Rothulfuashem villa mentioned in the List of Landed Property, the precursor of today’s Rijnsburg (Sarfatij 1977).[12] According to Henderikx it was an early donation dating from the early years or the first half of the eighth century. The donors were probably relatives, among whom Rothulf, presumably a descendant of the villa’s original name-giver, seems to have been the most important. Also mentioned are Aldberga (Rothulf’s wife?) and one Erulf. The family possessed more landed property in the region than just Rothulfuashem and donated other properties to Utrecht. They clearly were major landowners. The information about Rijnsburg in the document is exceptionally elaborate. It says that Rothulf and Aldberga lived locally at their villa. Did they move out after donating it or was their family line dying out? Of course we remain ignorant of most of the story. However, we do learn that the actual settlement (ofstedi) lay at Rijnsburg south of the river Vliet and the associated farmland (a total of 26 mansa) on the opposite side. This covered an area of about 157 hectares. Of the ofstedi Rothulfuashem, three houses (i.e. three farmsteads?) have been excavated. There must have been far more indeed, in order to farm such an expanse of land. We do not know whether the family’s possessions included harbourage on the Rhine. If so, it was not immediately connected with Rothulfuashem as the distance to the river is about half a kilometre. The nearest harbour presumably was at Valkenburg-De Woerd.

There is a possibility that the villa excavated at Koudekerk belonged to a larger complex. It has been suggested that Koudekerk was one of the toll ports of the Carolingian realm (Verkerk 1992). If this was the case, it cannot have been at the excavated site, which lacked access to the Rhine. On the riverbank today we do find the church and it cannot be ruled out that this church had an Early Medieval predecessor. Harbour facilities might then be sought in the vicinity of the church, together with any dwelling for the owner or the steward of the estate. In the Early Middle Ages Koudekerk was known by a different name, now lost. It has been suggested that Koudekerk should be identified with a place called Holtlant, or with yet another still enigmatic place name mentioned in the Utrecht List of Landed Property (Van der Linden 1998). This would mean that Koudekerk had been wholly or in part donated to the diocese by wealthy landowners. Thus there are several pointers to suggest that in the Early Middle Ages Koudekerk was an estate comprising various elements, the excavated farmhouses representing the agrarian component.

3.4 Upstream from Rijswijk / Levefanum

To date, no Early Medieval farming settlements have been excavated upstream from Rijswijk/Levefanum. However, some cemeteries in the area attest to a fourth- and fifth-century Frankish phase, followed by Merovingian and (early) Carolingian phases.[13] The principal cemeteries are those of Rhenen, Wageningen and Elst, all located on the northern bank of the Rhine. Rhenen and Wageningen are at strategic locations where land routes from Free Germanic territory crossed the river (Rhenen: Ypey 1972; Wageningen: Van Es 1964; Hulst & Van Es 2007; Elst: Verwers & Van Tent in prep.). South of the Rhine, continuous development from Roman into Merovingian and Carolingian times may be expected at several sites, such as the castellum at Meinerswijk. Yet such developments did not occur without a hitch and problems are encountered in the fifth century. Just north of the Lower Rhine, the farming settlement at Bennekom in Free Germany but with close links to the neighbouring Roman province, was deserted (or at any rate vanishes from sight archaeologically) (Van Es et al. 1985). In the cemetery of Rhenen we see a shift, though not a break, in the abandonment of the old Frankish part (Van Es & Wagner 2000). Such changes are not a local phenomenon but occur throughout the sandy regions of the north-eastern Netherlands and adjacent Westphalia, where settlements ‘disappear’ (relocated) and funerary rites are altered.

No definite explanation has yet been proposed but there might be a link with the early stirrings of Merovingian ascendancy. The Franks from the eastern part of the Rhine corridor were trying their luck elsewhere, supporting Childeric and Clovis and their campaigns in northern Gaul. In the sixth century the Merovingian elite remembered the importance of the Lower Rhine and moved in from Ripuarian Cologne to put things straight here. From c. 530 until the early seventh century are represented in the cemetery of Rhenen by three generations of rich graves, probably belonging to a single family (Wagner 1994). Putting things straight meant that the (family) estates along the Lower Rhine were newly laid out, staffed and in some cases (temporarily?) reoccupied by their owners. After a dip in the fifth century, this may be what started the development of the agricultural estates which in later written sources appear as villae. In the eastern part of the river delta, many, but not necessarily all of these, might have had a Late Roman and fifth-century origin. Indeed, at De Geer some gold finds date to the sixth century.

At the same time, the Frisian newcomers in the western Rhine delta behaved in a similar fashion. They too set about organising their newly acquired estates. Ambitious leaders gained political power. Gradually the two power blocks came to oppose each other as Franks and Frisians started to vie for the Rhine delta. These power politics provided the backdrop against which we see the emergence of Dorestad in the mid-seventh century.