A key aspect of the data collection policy for the project was to examine the quality of the dating evidence from individual structures and sites. This emphasis on good chronological control at the core of the methodology reflects a trend, in prehistoric studies particularly, towards evaluating and refining chronological resolution. In this case, our concern was with tracing long-term changes in house construction and use, and therefore re-evaluating long-held views about the historical development of roundhouse architecture in Wales.
Figure 3: The distribution of pre-1960 and post-1960 excavated settlements highlights the division between north-west and south-east Wales. The emphasis on visible archaeology and traditions of survey furthered interest in settlement patterns in north-west Wales. (Map data: Crown Copyright/database right 2007; an Ordnance Survey/EDINA supplied service.)
The earliest efforts to synthesise the settlement evidence for prehistoric and early historic Wales were significantly biased by the predominance of Roman pottery and associated finds from antiquarian excavations (Radford 1937). This was, understandably, considered indicative of a Roman date for all so-called hut circles. Even by the second half of the 20th century, there was still a poverty of settlements other than hillforts that could be unequivocally attributed to the pre-Roman period; Gresham wrote of north-west Wales (the region that dominated much of the early discussion on settlement patterns, see Figure 3) that there are 'no dwelling sites in the region chosen that have definitely been proved to be earlier than the period of the Roman Occupation' (Gresham 1963, 266).
Taking the artefactual evidence in isolation, the ubiquity of Roman pottery at excavated roundhouse settlements certainly seemed to support these interpretations (e.g. Tŷ Mawr, Cors y Gedol and Din Lligwy, all in north-west Wales). Nonetheless, there were clear differences in the morphology of the settlements, particularly in the presence, absence or degree of regularity of enclosing banks and ditches. This was acknowledged by Griffiths when he observed that the small houses associated with irregularly shaped field plots found in the uplands of north-west Wales were similar to excavated sites on Dartmoor, which were dated to the Bronze Age (Griffiths 1951, 69). Roman sites, by comparison, revealed their origins not only because of their regular, straight-sided enclosures, but also through their association with terraced fields which were perceived to be a result of the introduction of technology of Mediterranean origin and their proximity to Roman forts such as Segontium (Caernarfon). Gresham thought the changes in settlement morphology were suggestive of Roman influence on a native tradition (Gresham 1963, 266).
The basic typological distinctions between unenclosed, curvilinear and regular enclosures have remained as one of the principal means through which roundhouse settlements are chronologically organised. While this allowed for a prehistoric phase of roundhouse settlement, the much longer chronology was encouraged by the widespread recognition of first timber and then turf- and clay-built structures. Bersu's excavations in the late 1930s at Little Woodbury, Wiltshire, marked one of the earliest identifications of such structures in Britain (Evans 1989). At Llwyn-du Bach, it was Bersu again, together with Griffiths, who excavated post-holes that he believed to be associated with the stone phases of a concentric enclosed settlement. It is possible that the timber settings were the remains of an earlier, unrecognised phase, as at Moel y Gerddi and Erw-wen (Kelly 1988b). The discovery of timber and clay/cob walled buildings became commonplace from the late 1960s onwards, most notably with the major excavations at the Breiddin and Moel y Gaer, where 24-25 timber walled and 42 clay or cob walled structures were uncovered. The relatively recent excavations across Anglesey at Cefn Cwmwd, Cefn Du and Melin y Plas, have emphasised the importance of the cob- and turf-built structures (Maynard et al. 1999).
The longer chronology for settlements that was suggested by revised morphological schemes and the identification of timber phases was confirmed by the use of radiocarbon dating. It is of particular value to the understanding of prehistoric settlement in Wales because artefact assemblages for much of the first millennium BC are predominantly aceramic. Only a very few settlements have produced pottery that can be attributed to the Late Bronze Age and Iron Age, for example the Breiddin, Castell Odo and Rhuddlan. The post-Roman period is also aceramic apart from a few imports from high-status settlement sites.
The dataset assembled for the Welsh Roundhouse Project placed particular emphasis on radiocarbon-dated sites in order to offer a perspective on settlement archaeology that was not allied to morphological categories or reliant upon the presence of diagnostic ceramics. The following analysis uses the radiocarbon and estimated dates from settlements to identify long-term changes in house construction and use.
Figure 4: Histogram showing the frequency of the earliest (n=145) and latest (n=126) estimated dates of settlements in the database organised within 500 year intervals.
Starting at the coarsest, most unrefined level of dating evidence, the estimated dates of settlements range from the later Neolithic to the early medieval period. When sorted into 500-year intervals, the period with the largest number of dated settlements with excavated roundhouses is from 500 BC to AD 500 (Figure 4). These are the earliest and latest known dates for the main phases of occupation. The main period of earliest settlement dates ranges from 1500 BC to AD 500 (over 90% of the settlements are within this range), whereas the range of latest settlement dates is more restricted to the period AD 1 to 500. This may be due to the use of Roman pottery as a means of identifying a terminus ante quem and terminus post quem for the occupation of a settlement.
Figure 5: Histogram showing the frequency of the earliest (n=188) and latest (n=223) estimated and calibrated radiocarbon dates from individual roundhouses within one hundred year intervals.
A more nuanced picture is achieved by focusing on the chronology of individual house structures (Figure 5). The frequency of earliest and latest dates from houses within 100-year intervals during 3000 BC-AD 1200 supports the general trends identified above. But rather than a continuous rise in the numbers of houses dated after 1500 BC up until AD 500, it indicates (1) a significant increase in the numbers of earliest dates from houses during the 9th century BC and a corresponding increase in the frequency of latest dates during the 3rd century BC; and (2) the majority of earliest and latest dates for roundhouses occur during the 6th century BC to the 4th century AD.
Figure 6: Cumulative frequency curves for earliest (n=188) and latest (n=223) estimated and calibrated radiocarbon dates from individual roundhouses based on numbers of settlements represented in each 100 year period.
The notable increase in the number of earliest dates from the 9th century BC is in part a consequence of the over-representation of a few settlements where large numbers of houses have been excavated and dated, particularly Moel y Gaer which has produced 42 of the 60 dates lying in this range. Yet even when allowance is made for this bias by plotting the frequency of settlements represented by the earliest dates in each period (Figure 6), there is still a significant change in the 9th century, with ten settlements producing earliest house dates during this period compared with 1-3 settlements otherwise represented in each of the two centuries either side of the 9th century. An obvious explanation for this pattern is the plateau in the radiocarbon calibration curve which is identified for the period c. 700-400 BC (calibrated). This has the effect of extending the range of any radiocarbon dates estimated to lie within this range. Since radiocarbon dating is the only method of establishing any form of chronological control for this period, the lacuna in earliest and latest dates during 800-600 BC is perhaps to be expected. This interpretation is supported by the distribution of uncalibrated dates (Figure 7), which does not show an isolated peak in the number of dates after 500 bc uncalibrated (c. 950-200 BC calibrated).
Figure 7: Histogram and cumulative frequency curve of uncalibrated radiocarbon dates (n=432) from excavated settlements.
While the plateau in the radiocarbon calibration curve may offer an explanation for the peak in the earliest dates for houses in the 9th century, it also serves to mask socio-historic processes. Despite the lack of precision in the chronology, there is a substantial increase in the number of houses from the 9th to 5th centuries BC onwards. Importantly, this is represented by a relatively continuous and stable presence of this architectural form up until the end of the Roman period (Figure 6). Based on this, it seems reasonable to conclude that roundhouses were a substantially more common element in the Welsh landscape from the Late Bronze Age to the late Roman period. There still remains a 'head and tail' of structures dating from the Bronze Age and the early medieval period respectively, which, although few in number, may contribute to an understanding of the development and decline in roundhouse architecture.
Settlements with dated roundhouses form roughly one-third of the overall dataset of excavated settlements. Despite the obvious limitations of attempting to interpret spatial patterns, some broad regional patterning can be seen in the dates of houses and settlements (Figures 8-11). What is striking is the relative lack of new Roman-period settlements or early medieval dates for roundhouses in the north-east, whereas in the south-east and north-west there is an increase in the number of houses during the Roman period. The houses in the south-west last from the Iron Age to the Roman period but, unlike the north-east, they do not appear common before the Iron Age. There appear to be more early medieval settlements in the north-west than the other regions, although the numbers are too small to make a meaningful distinction.
Figure 8 (left): Distribution of settlements with roundhouses dated earlier than 900 BC, based on estimated or calibrated radiocarbon dates
Figure 9 (right): Distribution of settlements with houses whose earliest estimated or radiocarbon dates lie during 1000-600 BC, based on estimated or calibrated radiocarbon dates.
Figure 10 (left): Distribution of settlements with houses whose earliest estimated or radiocarbon dates lie during 600 BC-AD 200, based on estimated or calibrated radiocarbon dates
Figure 11(right): Distribution of settlements with houses whose earliest estimated or radiocarbon dates lie after AD 50, based on estimated or calibrated radiocarbon dates.
A clearer set of observations can be drawn from a comparison of settlement morphology and chronology (Figures 12 and 13). This broadly supports the trends identified in earlier syntheses of the evidence (discussed in 2.1). For early sites, there is a slight tendency towards scattered settlement in the second millennium BC and most of the single huts appear in the third millennium, mainly represented by the circular structures found under funerary monuments. The circular/concentric settlements appear restricted in time to the later Bronze Age and early Iron Age, while nucleated or enclosed settlements predominate in the later Iron Age and Roman period. There is, however, a potential weakness in this data owing to the incomplete excavation of settlements (for 48% of sites the extent of excavation is not known, 27% have had 0-25% excavation, 9% have had 25-75% excavation and 16% have had 75-100% excavation).
Figure 12: Plans of settlements representative of the morphological categories adopted in the project database (after Crew 1998; Kelly 1988b; Phillips 1934).
Figure 13 Histogram showing the percentage frequency of the earliest estimated dates of settlements of different morphological types in the database organised within 500-year intervals.
The project has highlighted the importance of a radiocarbon dating programme for Welsh settlement archaeology. Radiocarbon dating has made a significant contribution to the identification of prehistoric structures, which generally contain few diagnostic finds. Although this may in turn help to identify characteristic finds assemblages, the improved characterisation of the roundhouse relies on the continuation and expansion of such a dating programme. The data from this project have shown just how slight are the chronological foundations for discussion of the roundhouse in Wales. There is also some regional variation in the availability of, or prioritisation of, dating. In general, the study appears to illustrate the need for many more radiocarbon dates per settlement: some roundhouses were subject to multiple episodes of rebuilding and alteration that cannot always be captured by one date from the hearth. The shifting location of houses around a settlement and the tendency for houses to respect the locations of earlier houses necessitate dating of different structures within a settlement.
On the other hand, radiocarbon dating proves to be a blunt tool for the interpretation of subtle modifications to roundhouse structures. Based on the radiocarbon dates, the approximate average date span of a house is 450 years (taking the average of radiocarbon latest date minus radiocarbon earliest date for each house). This is self-evidently not a good estimation of use-life, but it does offer an indication of the typical chronological range obtained using this dating method. This is particularly problematic for the later periods. The use of other dating methods is not common but has produced good results – for example at the settlement of Bryn y Castell, where archaeomagnetic dating has given greater subtlety to interpretation of structures with overlaying hearths and furnaces (Crew 1987).
Despite these limitations in the data, there is an interesting pattern in the dates for the widespread adoption of roundhouses; equally there are regional variations that require explanation and testing through future research. The dating evidence analysed for this project places the widespread adoption of roundhouses in Wales firmly in the Late Bronze Age and Early Iron Age, with relatively few examples of structures dated prior to the late second millennium BC. This contrasts with south-west Scotland where a number of roundhouse sites now date to the later third millennium, and much of England where there are a great many sites dating from the mid-second millennium BC. The rarity of roundhouses in Wales during the second millennium BC and, in some areas, the first millennium AD, raises the issue of what it meant to construct roundhouses. This is a question of relevance to the Bronze Age and early medieval period and, arguably, given the regional analysis presented in 2.3, also for the late Iron Age and earlier first millennium AD in north-east Wales. The next section tackles this question by considering what inhabiting roundhouses might have meant.
© Internet Archaeology
URL: http://intarch.ac.uk/journal/issue23/1/2.html
Last updated: Mon Nov 26 2007