4. Historical Overview

It would be misguided to attempt an overarching historical narrative for roundhouses in Wales on present evidence, but there are certain trends and issues relevant to particular periods. These rough chronological divisions correspond with changes in the development of the house form as presented in 2.2.

4.1 Early roundhouses: 3000-1000 BC

The third and second millennia BC are characterised by a division between the more insubstantial and ambiguous evidence for structures and occupation in the earlier period and the appearance of more familiar roundhouse structures in the later part of the period. The earliest circular structures recorded in the project were of Neolithic to Early Bronze Age date and were often found under later funerary monuments. They occur in radiocarbon-dated contexts at Stackpole (Pembrokeshire, Benson et al. 1990), Brenig 6 (Denbighshire, Lynch 1993), Redberth-Sageston (Pembrokeshire, Page 2002), Treylstan and Upper Ninepence (Powys, Gibson 1999), and without radiocarbon dating at Llanelwedd Mound (Powys, Spurgeon 1967), Cefn Cilsanws (Rhondda Cynon Taff, Webley 1958), Rhos-y-Clegryn (Pembrokeshire, Lewis 1974), Sant-y-Nyll (Glamorgan, Savory 1960) and Newton (Oystermouth, Swansea, Savory 1972). Variously interpreted as temporary shelters or structures involved in funerary rites and other ritual activities, those included in the project have ambiguous evidence for use, including in some cases hearths and evidence of food preparation.

Apart from these structures, whose characterisation as houses is equivocal, the earliest roundhouses seem to be in south Wales. Those at Stackpole Warren site A, Pembrokeshire (Benson et al. 1990) and the Atlantic Trading Estate, Glamorgan (Sell 1998) could date to the Early Bronze Age. Of Middle Bronze Age date are the wooden structures found in coastal locations on the Gwent Levels at Chapeltump (Whittle et al. 1989), Redwick (Bell 2001) and Rumney Great Wharf (Allen 1996). Unfortunately, these structures are poorly preserved and occupation evidence is lacking. The undated structure at Collister Pill could be a similar date (Bell et al. 2000, 303).

Roughly contemporary with these is the site at Meyllteyrn Uchaf, Gwynedd (Ward and Smith 2001). Although earlier, it is a 'concentric circular' enclosure and its similarities with Erw-wen and Moel y Gerddi are remarkable. Meyllteyrn Uchaf features three houses, the earliest with a clay wall and internal wall slot (for a timber or wattle revetment), replaced by a later house with clay wall and revetted with both stone and wattles. There is a smaller, contemporary structure with a clay wall and wall slot.

Other northern, Middle to Late Bronze Age sites are Glanfeinion, Powys (Britnell et al. 1997) and Graig Fechan, Conwy (Manley 1991). The latter is notable for its use of stone, a construction technique shared by the site at Llyn Morwynion, Gwynedd (not excavated but dated to the Bronze Age; Caseldine et al. 2001). The Graig Fechan house is a small stone structure in a larger stone enclosure. Unfortunately, evidence is quite limited from this site. The structure seems unusually small with a maximum internal diameter of 2.5m; it is possible this is some sort of entrance feature with the main building formed from the larger, 11m diameter, circular enclosure. At Glanfeinion, only an internal post ring survives, suggesting possibly a turf outer wall. There is evidence of occupation and repairs to this structure.

Bronze Age dates have also been obtained at Trostrey, Monmouthshire, (Mein 1992; A. Mein pers. comm.) although full details of associated structures have not been published. Sadly, no dates were available for the series of circular structures under the Roman fortress at Old Market Street, Usk (Marvell 1988a). A Bronze Age date could be suggested due to the nature of finds in the area, including many Bronze Age flints and a gold earring – interpreted as Roman by Manning et al. (1995, 101 no. 30) but now thought to be Bronze Age (Marvell 1988b ; cf. Peterson and Pollard 2004, 68). The structure being excavated at Llanmaes, Glamorgan, appears to be of later Bronze Age to Iron Age date (A. Gwilt pers. comm.), and those at Llandegai, Gwynedd (undated) (Lynch and Musson 2004) could also be this early, although finds from similar structures recently excavated a short distance away suggest an Iron Age to Roman date.

Even within this small group there cannot be said to be a 'typical' Bronze Age architecture: the house at Stackpole appears to have an outer wall of posts (probably associated with wattle and daub) and an internal timber wall slot. It has definite evidence of occupation in the form of a hearth, but this may not have been primarily 'domestic' in nature (Benson et al. 1990, 238). There was associated Beaker pottery, human bone and cereal grains. The later standing stone on this site suggests that this structure could be considered with others found under Bronze Age funerary monuments, but on the other hand such a practice could equally commemorate a location of habitation. This structure can be compared to that from the Atlantic Trading Estate development (Sell 1998). The structure had an internal wall slot and may also have had a wattle and daub outer wall (daub was found in this part of the structure). One side of the structure appears to have had a timber wall and the other a stake wall; these were possibly supports for the daub wall. However this structure is much larger; it seems to have had internal roof supports and internal stake partitions.

Site visibility is a significant issue in this period; the prevalence of wooden forms does indeed suggest a reason for the relative invisibility of roundhouse sites of this date. The Llŷn Crop Marks Project (Ward and Smith 2001) equated the relative absence of Bronze Age structures in the region not only to the use of construction materials other than stone but also to modern arable land exploitation (for example the dense distribution of Bronze Age funerary monuments and stray finds in the Llŷn peninsula was contrasted with the lack of occupation evidence). Visibility might also be affected by a lack of enclosures around settlements. The presence of pre-rampart occupation on a number of sites, such as Moel y Gaer, Thornwell Farm, Castell Odo, Broadway (Llawhaden) and Coygan Camp, demonstrates that unenclosed roundhouses were a feature of at least the early first millennium BC landscape. A further explanation may be that Bronze Age roundhouses were genuinely rare, and occupation practices were instead typified by less substantial structures of which only the occasional hearth or pit now survives. This may account for the common presence of third and second millennia BC activity on later sites, something reflected in the distribution of the radiocarbon dates. A parallel might easily be drawn with the attempts to characterise Early Bronze Age settlement in southern England (Brück 1999).

4.2 A landscape of roundhouses: 1000 BC-AD 400

About one-third of the sites in the study have their origins in the Iron Age and the majority of these appear to continue through to the Roman period. The distribution of radiocarbon dates would support this argument for continuity between periods: while there is a large group of dates spanning the later Iron Age only (56 dates in the range 400-100 BC) there is also a significant group between 200 BC-AD 150 and a group between 100 BC-AD 250. The increase in the number of sites with Late Bronze Age origins reaches an overall peak in this period. However, it is the north-west which has the only notable increase in the number of new sites, with a decline in the number of new sites beginning in the north-east (and continuing into the Roman period). It should be stressed, however, that the numbers are too small to state confidently a regional trend. These new sites tend to be nucleated.

The majority of Iron Age houses in the sample were timber-built structures. However, stone roundhouses do occur in the Iron Age, notably in the north-west region. In fact, all periods show a diversity of construction types, although stone is indeed more common in the later periods, as are roundhouses in general. The sites of Erw-wen and Moel y Gerddi contain exceptionally early stone houses, in both cases following from timber phases, and house T1 at Holyhead Mountain also seems to pre-date the Roman period (Smith 1987, 24). The earliest structure at Cefn Graeanog II, Gwynedd (Mason and Fasham 1998), was a late Middle Iron Age (early 2nd century BC) stone construction, built as an isolated structure. It is, however, probable that most houses with stone walls and an Iron Age date extend into the Roman period. For example, at Cefn Du, Anglesey (Cuttler 2004) there was a house with a stone revetment of an earth and rubble wall and an inner post ring, of mid to late Iron Age date into the 2nd century AD. Also on the site were later clay-walled and stone-walled structures.

Dating is a particular issue for the Iron Age to Roman period. It is possible that there are further pre-Roman stone houses but without better dating this remains uncertain. The sites of Ynys Leurad, Plas Meilw, Penmon and Parc Dinmor on Anglesey all lack Roman pottery. It is not known if they are of Roman date and aceramic, or earlier, which may be suggested by the other finds. An aceramic Roman date is often assumed, but, for example, no Roman pottery was found at the radiocarbon dated site of Rhydydefaid (spanning the 2nd centuries BC to AD) (Davidson 1998). Other roundhouse sites on Anglesey have produced Roman pottery and finds (Parc Salmon, Pen y Bonc, Bodafon Mountain and Caer Leb), so there is no doubt about the availability of Roman pottery. However, there have been few modern excavations of this type of site, and Roman finds have tended to lack a recorded stratigraphic context. At Porth Dafarch (O'Neil 1940) there was a combination of Roman and earlier finds, suggesting continuity or Roman period reuse of an Iron Age site and a similar interpretation can be given to the late phases of Pant y Saer. The more recent excavation at Holyhead Mountain (Tŷ Mawr) suggests a similar picture (Smith 1987).

Stone structures on northern hillforts such as Dinas (Conwy), Garn Boduan and Conwy Mountain are also undated (Hogg 1960; Griffiths and Hogg 1956). A focus on hillfort defences has done little to elucidate the chronology of internal structures. There is Roman pottery from Braich y Ddinas, Conwy (Hughes 1912), but the use of saddle querns, other finds such as briquetage and a structure with similarities to Bryn y Castell suggest an earlier date for use of the site (Crew 1984, 42). Garn Boduan also produced briquetage, suggestive of an Iron Age date. The problem is even more marked for the upland sites of south Wales, where there has been a lack of recent excavation. Poorly recorded excavations were undertaken by the Rhondda Naturalists Society in 1901 at Maendy Camp (Williams 1902), Blaenrhondda (Wheeler 1921, 70; Savory 1984, 455) and Hen Dre'r Gelli in 1903-6 (Griffith 1906). The site at Blaenrhondda (Garreg Lwyd, Rhondda Cynon Taff) is undated and may be pre-Roman. Maendy Camp produced Roman pottery, as did the nearby site at Hen Dre'r Gelli, but it has been suggested that in the latter instance this could represent reuse of an Iron Age site.

In this period the roundhouse is a common architectural form and its use for domestic occupation is not disputed. There appears to be a trend towards nucleation. It is notable that over 70 radiocarbon dates fall within the range 200 BC-AD 250, which can be contrasted with the 39 dates wholly within the Roman period. This suggests that there is an argument for considering the later Iron Age to early Roman period as one distinct period in Wales, something supported by the nature of the archaeological evidence.

There are relatively few sites in the study that appear to be founded in the Roman period; the majority of these are in the north-west. This is partly a reflection of the fact that where the earliest date of the site is not known it has often been assumed to be in the Iron Age. A larger proportion of sites (42%) have occupation thought to end during the Roman period. About half of these are in the north-west of Wales. The remaining sites are mostly in the south. From the Iron Age, there is a slight decrease in the number of individual houses radiocarbon-dated to the Roman period, with no dated houses from the north-east. This is reflected in broader estimated dates, with a complete absence of new houses in the north-east in the Roman period. When compared to the number of houses with an estimated later Iron Age foundation date, there were comparable numbers founded in the Roman period in the north-west, more Roman period houses being founded in the south-east and fewer in the south-west.

Again, the period is characterised by diversity in roundhouse construction, and the use of stone cannot be said to be typical, even in the northern area of Wales. Recent work on the A55 project on Anglesey (Maynard et al. 1999) has shown the presence of clay-walled structures in the Roman period at Melin y Plas, Cefn Du and Cefn Cwmwd. This form of construction has a thick clay (or possibly turf) wall bounded by narrow gullies (possibly to hold a wattle or timber revetment). They were also present at Pant, Bryncroes, Gwynedd (Ward and Smith 2001) and Prestatyn further east. These sites all appear to have had Iron Age origins but show continuity during the Roman period. Structures A and B at Bryn Eryr, Anglesey (Longley et al. 1998), were of clay-walled construction with a timber internal lining and dated to the late Iron Age to Roman period. Here, as at Cefn Du and Cefn Cwmwd, a later, Roman, house was constructed with a stone facing to a rubble and earth core, possibly with a timber lining.

In the southern area, the use of timber circular structures continued well into the Roman period. Post-built structures appear throughout the Roman period at Thornwell Farm, Gwent (Hughes 1996) and Biglis, Glamorgan (Parkhouse 1988). At Whitton, Glamorgan (Jarrett and Wrathmell 1981), timber structures were constructed in the early Roman period (1st to 2nd century), before being replaced by rectangular buildings in the later Roman period. At Woodside, Pembrokeshire (Williams 1998d), Roman period roundhouses survive as wall gullies, with some evidence for internal posts and wattle and daub construction. These seem to date up to the 2nd century AD. The site of Pencoedtre in Glamorgan (Hughes and Bashford 1999) illustrates the poor survival of what may have been clay-walled roundhouses, surviving as mounds containing daub and charcoal but without other structural features.

The apparent increase in roundhouse building in the south-east of Wales is of interest in the context of the relative integration of this region into the Roman administrative network. It seems that, in most cases, non-stone roundhouses in south Wales date before the 2nd century AD and are on sites with earlier Iron Age occupation. However, the presence of later structures, as at Thornwell Farm, should not be discounted. The variety of and co-existence of different building styles and both round and rectangular structures should be stressed, although it seems that there are few large post-built structures comparable to those seen in the earlier Iron Age.

4.3 Post-Roman roundhouse settlement

The number of sites with early medieval radiocarbon dates represents a significant reduction from the Iron Age and Roman period. This may reflect the difficulty in identifying early medieval settlements in Wales more genereally. There is some overlap between the Roman and early medieval period in the radiocarbon date ranges, with 12 dates falling between AD 200-600 but this is slight compared to the continuity between the Iron Age and Roman period. The existence of 16 dates in the range AD 400-800 and 25 dates in the range AD 600-1000 is, however, a clear demonstration of the existence of early medieval roundhouse occupation.

Discussion of early medieval settlement has tended to focus on higher status sites, including hillforts, occupied in this period and containing imported goods. Lower status and lowland settlement sites have been more difficult to identify, but it has been suggested that this included wattle and daub roundhouses (Edwards 1997, 4). It is clear that there is a variety of building forms being used (Arnold 2000, 162). The nature of the evidence from Dinas Powys (not analysed here) illustrates the difficulty of interpretation of this later occupation. There, sub-rectangular gullies were interpreted as relating to dwelling structures (Alcock 1963). But excavation techniques only revealed rock-cut features, and shallower features relating to, for example, a wattle and daub roundhouse, or a rectangular building of sill-beam construction, may not have survived. At Longbury Bank, a more recent excavation, dwelling structures also proved hard to recognise (Campbell and Lane 1993).

Nevertheless, there is some evidence to suggest that roundhouses do occur in early medieval Wales. The strongest evidence comes from the Viking Age and earlier settlement site at Y Glyn, Llanbedrgoch (Redknap 2001; 2004). This survived as a drip gully located underneath the later rectangular building 1 and produced a 7th to 9th century radiocarbon date. Associated stake-holes and daub suggest a wattle and daub structure (Redknap pers. comm.). Other features on the site date to this period and it is possible that the settlement enclosure was contemporary. Post-holes and a beam slot from a structure at Drim Camp (Mytum 1998b) have been associated with a 7th to 9th century radiocarbon date. A double-walled wattle house has been suggested (Mytum 1998b, 63) and an early medieval pennanular brooch comes from the area.

Early medieval radiocarbon dates have also been associated with wattle and daub oval structures at Trostrey, Gwent (Mein 1992). The site has not yet been fully published and consequently it is not possible to comment on the significance of these discoveries. Many phases of occupation have been recorded and similar structures on the site were associated with Bronze Age pottery. Ambiguous evidence comes from Burry Holms Church site (Hague 1967), where dating evidence is lacking: an early medieval circular structure could also be of Iron Age date. However, the sequence of development of the later church site does suggest a significant continuity of use over a number of centuries.

There are two early medieval radiocarbon dates from the site of Tŷ Mawr at Holyhead Mountain, Anglesey (Smith 1987). The hearth in the circular stone structure T3 gave a date of AD 429-768. Little of the floor survived in this structure and it is possible that the hearth is a later intrusion; structure T1 appears to pre-date structure S (thought to be Iron Age) but its floor layers produced a later radiocarbon date (cal AD 405-870) probably relating to 6th century use of the structure. However, the non-circular structure T4 has been given a 6th-century date.

Early medieval activity has been recorded for a number of earlier sites in the study, in the form of non-structural features or radiocarbon dates. At Arddleen (Powys) (Grant 2006) there appears to have been early medieval recutting of the enclosure ditch, with associated ditches and post-holes but no definite structural evidence. Here it appears that the activity represents reuse rather than continuation of occupation, but it is possible that low-level activity continued on these sites and that places persisted in communal memory. A structure at Pant (Bryncroes, Gwynedd) (Ward and Smith 2001) appears to have been used for metalworking in the 5th to 7th century, and there is no evidence for prior abandonment. At Graeanog, Gwynedd, there is a 6th century date for the last firing of the hearth in the sub-rectangular structure 'A', suggesting genuine occupation. However, it could be argued that this relates to use of existing structures rather than a new phase of construction.


© Internet Archaeology URL:
Last updated: Mon Nov 26 2007