A Norse Settlement in the Outer Hebrides: Excavations on Mounds 2 and 2A, Bornais, South Uist

By Niall Sharples

The settlement at Bornais in the Western Isles of Scotland is one of the largest rural settlements known from the Norse period in Britain. It spans the period from the fifth to the fifteenth century AD when the Atlantic seaboard was subject to drastic changes. The islands were systematically ravaged by Viking raiders and then colonised by Norse settlers. In the following centuries the islanders were central to the emergence of the Kingdom of Man and the Isles, played a crucial role in the development of the Lordship of the Isles and were finally assimilated into the Kingdom of Scotland.

This volume explores the stratigraphic sequence uncovered by the excavation of Bornais mounds 2 and 2A. The excavation of mound 2 revealed a sequence of high status buildings that span the Norse occupation of the settlement. One of these houses, constructed at the end of the eleventh century AD, was a well preserved bow-walled longhouse and the careful excavation and detailed recording of the floor layers has revealed a wealth of finds that provides invaluable insight into the activities taking place in this building. The final house in this sequence is very different in form and use, and clearly indicates the increasing Scottish influence on the region at the beginning of the thirteenth century.

The excavation of mound 2A provides an insight into the less prestigious areas of the settlement and contributes a significant amount of evidence on the settlement economy. The area was initially cultivated before it became a settlement local and throughout its life a focus on agricultural activities, such as grain drying and processing, appears to have been important. In the thirteenth century the mound was occupied by a craftsman who produced composite combs, gaming pieces and simple tools.

The evidence presented in this volume makes a major contribution to the understanding of Norse Scotland and the colonisation of the North Atlantic in a period of dramatic transformations.

• Language: English
• Publisher: Oxbow Books
• Release date: Dec 19, 2019
• ISBN: 9781789250473

Book preview

1 Bornais and the Norse settlement of the North Atlantic

Introduction – N Sharples

The settlement at Bornais consists of a complex of mounds that protrude from the relatively flat machair plain in the township of Bornais on the island of South Uist in the Western Isles of Scotland (Figure 1). The machair plain forms the west coast of the island and comprises a sand deposit made up of shell and siliceous minerals, which provides a calcareous environment that is markedly different from the peat covered moorlands to the east of the coastal plain (Angus 1997; 2001). Ritchie (1979; 1985; Ritchie et al. 2001) has argued that the sand has formed since the last Ice Age as a result of the submergence of the extensive shallow coastal plain that lies off the west coast of the Uists.

The sand is comprised of relic glacial material and large quantities of comminuted shell from the ancient storm-washed shoreline. It is assumed that there was a fairly rapid rise in sea level up until about 5000–7000 years ago, when sea level was probably only about a couple of metres below the current position (Ritchie 1979, Ritchie et al. 2001). There was then a massive influx of sand onto the land around 4050–3890 cal BC that created a coastal plain somewhat similar to that of today (Sharples 2009). This influx probably preceded the final rise in sea level that divided the islands of South Uist, Benbecula and North Uists.

These geomorphological developments created a west-ward extension to the solid geology of the island that proved an attractive settlement location from the Beaker period onwards and the machair appears to have been continuously and intensively occupied from the Late Bronze Age to the end of the Norse period (Sharples et al. 2004). In recent years, although the machair has not been occupied, it has proved a valuable and extensively exploited agricultural landscape; its attraction is that it provides a relatively well-drained area of flat land that is one of the few parts of the island where cereals can be cultivated. The relatively large area of machair present on the Uists has meant that these islands have been described as one of the most agriculturally productive regions of the west coast of Scotland (Dodgshon 1998, table 3.1).

The reasons behind the excavations at Bornais have been described in detail in the first volume of the Bornais report (Sharples 2005b) and it is not intended to re-tell the general history here. Some introduction is necessary, however, as the mounds described in this volume are different in significant ways to mound 3, described in volume 1 (Sharples 2005b) and mound 1, described in volume 2 (Sharples 2012).

The history of the excavations – N Sharples

During the 1994 field season, the excavation of the broch at Dun Vulan (Parker Pearson and Sharples 1999) largely drew to a close and the SEARCH team were looking to develop further their understanding of the settlement sequence on the island of South Uist by broadening their research to encompass the early medieval, medieval and post-medieval settlement patterns up to the ‘Clearances’ of the early nineteenth century. A group of prominent mounds at the centre of the Bornais machair was targeted for investigation, as the discovery of distinctive platter wares dating to the Norse period (Lane 1990; Parker Pearson 2012a) suggested these mounds had been occupied sometime in the ninth to twelfth centuries AD.¹ The mounds were only 1.6 km from Dun Vulan and the evidence recovered from Bornais could potentially provide a direct comparison to the evidence from this broch (Parker Pearson and Sharples 1999).

The dateable pottery was originally recovered from the surface of mound 2, the most prominent mound at the centre of what then appeared to be a complex of three mounds (Parker Pearson and Webster 1994). This mound was targeted for excavation and a trench 20 m by 2 m was excavated across its highest point on an orientation of northwest to southeast (Figure 2). This orientation was deliberately chosen as it was expected that any Norse houses present would be oriented east–west and the trench would therefore cut diagonally across the houses and make them easier to identify. The excavations were directed in the field by Jane Webster and Mike Parker Pearson and Niall Sharples made a short visit.

The aims of the excavation were outlined in the sub–sequent interim report (Parker Pearson and Webster 1994, 3) and were identified as:

•to identify the latest structural phase from the Norse period;

•to locate buildings within the mound;

•to establish the depth of the stratigraphic sequence and, as far as possible, to examine its formation;

•to recover artefact assemblages that would allow us to characterise and date the sequence of occupation;

•to find out if the apparent discontinuity of occupation noted at the Udal for the beginning of the Viking period was also identifiable here;

•to recover bone and carbonised crop assemblages to enable comparisons with the earlier assemblages from Middle Iron Age Dun Vulan and from the Late Bronze settlement at Cladh Hallan (Parker Pearson et al. 2004a)

Figure 1. A plan of the mounds at Bornais showing the areas excavated and the location of the site on the Western Isles

The excavations proved remarkably successful. After the removal of a nineteenth-century ‘field wall’, a sequence of Norse deposits was revealed that included thick midden layers and structural evidence for several buildings, including the corner of a large, well-built stone-walled house, later numbered as House 2 (Figure 3). Amongst the material recovered from this house was a number of significant finds: a complete comb and pin suggested a date in the tenth to eleventh century AD. An earlier structure was also identified, which it was suggested could be a figure-of-eight structure, possibly of Late Iron Age (Pictish) date.

The 1994 excavations demonstrated that Bornais was an important site that could answer many of the aims set out in the research design, i.e. to fill in the settlement sequence for the area and to provide comparable data to that acquired from Dun Vulan. Niall Sharples was persuaded to direct the project and started work on the site in 1995, though supervision of the latter part of that season’s excavation was completed by Jane Webster (Sharples et al. 1995).

Figure 2. A plan of the areas excavated on mound 2 from 1994 through to 2004

Figure 3. A view of the long exploratory trench across mound 2 in 1994 from the northwest; the walls of House 2 are clearly visible in the foreground

Figure 4. A view of the mound 2 excavations in 1996 from the west

In 1995 and 1996 the work on mound 2 was limited (Figure 2), as the main aim of those seasons was to characterise the archaeological activity on mound 3 (Sharples 2005b) and mound 1 (Sharples 2012). In 1995 a trench approximately 4 m by 4 m was excavated at the east end of the original 1994 trench where the putative Pictish building had been identified. The evidence from 1995 indicated that this was in fact a rectangular building of Norse date. Work continued in this area in 1996 when a trench 5 m by 5.5 m was excavated (Figure 4). Another trench 2.8 m square, was opened in order to try to define the extent of the large eleventh-century house discovered in 1994. These excavations were supervised by Jerry Bond.

Figure 5. A view of the excavation of House 2 mound 2 in 1997 from the east

Neither of these small trenches was particularly successful in achieving their objectives and it was realised that small-scale trenching was not helping to clarify the complex structural remains present. The inadequacy of the trenching strategy was emphasised by the completion of a geophysical survey of the mounds, which revealed a wealth of additional data. It was now clear that the settlement was much larger than had initially been thought, with two additional settlement foci, known as mounds 2A and 2B, as well as isolated structures scattered around the edge of the mounds (Hamilton et al. in Sharples 2005b).

As a consequence of these discoveries, it was decided to increase the student labour force recruited from Cardiff University for the 1997 season, and this enabled us to increase the area excavated. In 1997 the main focus of the work continued to be mounds 1 and 3, where complete buildings were exposed (Sharples 2005b; 2012). Nevertheless, a larger trench, 7.8 m by 5 m, was opened up over the west end of the original trench on mound 2, directed by Jerry Bond. This clarified the nature and extent of the large house, and emphasised the quality of the internal revetment wall and the likely status of the structure (Figures 2 and 5; Sharples 1997). It is striking that the area excavated in 1997 coincided precisely with the only part of House 2 that had not been ruthlessly robbed by later activity. The wall of the east half of the house seldom survived to more than one course in height and much of it had been completely removed!

No excavation took place at Bornais in 1998 due to a lack of financial support from Historic Scotland. Work recommenced in 1999, when the excavations ambitiously encompassed work on mounds 1, 2, 2A and 3 (Sharples 1999). The excavation season was expanded to eight weeks and an increased number of Cardiff University students were recruited.

Figure 6. A view of the excavation of House 2 in 1999 from the northwest

Figure 7. A plan of the areas excavated on mounds 2A in 1999, 2000, 2003 and 2004

On mound 2 in 1999, a trench 14 m by 7 m was excavated (Figure 2), with the continued principal objective of defining the extent of the large eleventh-century house (Figure 6), and excavating floor levels that would confirm its date. On mound 2A, a new trench approximately 9 m by 7 m was excavated (Figure 7); this was in response to continued ploughing of the mound, which appeared to be damaging a structure indicated by the geophysical survey and eroding midden deposits. Supervision of the work on mound 2 was undertaken by Suzi Reeve and Katinka Stentoft, and that on mound 2A was undertaken by Joe Danks and Rachel Jackson.

The excavation of mound 2 suggested we were dealing with a large bow-sided hall, at least 18 m long, and the material recovered from limited excavation of the floor confirmed that this contained a substantial and very significant assemblage of material culture. The excavations on mound 2A revealed a house (Figure 8) comparable to the Late Norse houses on mound 3 and at Cille Pheadair (Brennand et al. 1998). This house had been very badly damaged by recent ploughing and rabbit-burrowing, but it was clear that it sat on top of a settlement mound similar to the more prominent mounds 1 and 3.

Figure 8. The initial cleaning of House 15 on mound 2A in 1999 from the north

Figure 9. A view from the west of House 2 during the excavation in 2000

In 2000 the excavations were restricted to mounds 2 and 2A, as the excavation of mounds 1 and 3 had reached a satisfactory conclusion (Sharples 2000). The goal for the excavation of mound 2 was to examine the large eleventh-century house (Figure 9), and an area 25 m by 7 m was opened up (Figure 2). The complete excavation of this house would meet the principal objectives of the overall project, to identify and explore the use of domestic space throughout the occupation of the island of South Uist. The excavation of mound 2A was considerably expanded, to an area roughly 12 m by 9 m, which was designed to explore the deposits underlying the Late Norse house discovered in 1999, and to clarify the chronological span of these settlement mounds (Figures 7 and 10). Two trench extensions were excavated to the east (9 m by 2 m) and to the north (9.5 m by 2 m), to define the extent of the settlement on mound 2A and to explore the middens deposited around the edge. The excavations on mound 2 were supervised by Mark Brennand and Suzi Reeve and mound 2A was supervised by Katinka Stentoft.

Figure 10. A view across the excavated area of mound 2A from the southwest in 2000

The 2000 season was very productive and, by the end of the excavation, it was possible to create a narrative for the development of the settlement at Bornais (Sharples 2000). It was clear that activity on mound 2 began in the Late Iron Age: a distinct shift in the location of settlement at Bornais occurred in the middle of the first millennium AD, with occupation moving from mound 1 to mound 2.

The excavations in 2000 finally defined the full extent of the large eleventh-century house on mound 2, but also revealed that it was the second in a sequence of three large houses built at the centre of this mound, which will henceforth be referred to as Houses 1, 2 and 3. These houses were significantly larger and better built than the houses identified on the other mounds and suggested that mound 2 was a high-status focus for a settlement that had expanded considerably in the eleventh century AD (this interpretation proved to be wrong), when the subsidiary mounds were created. In the last fortnight of the excavation, the opportunity was taken to excavate two small trenches to the south and north of the east end of the main trench, in order to define the exact size of House 3, the last of the three houses constructed on mound 2.

The excavation in 2000 of mound 2A began to identify better preserved structures and revealed a sequence of deposits, roughly 1 m deep, which included a thick cultivation soil beneath a sequence of occupation soils and midden layers. The structures on mound 2A appeared to be associated with craft activity, including metalworking (wrong again) and composite comb-making.

We originally intended to return and excavate in 2001, but unfortunately the outbreak of foot and mouth disease meant that permission for the excavations was not forthcoming and work in 2001 was thus focused on postexcavation processing of the large number of environmental samples stored in Cardiff. The importance of this work was obvious and a second post-excavation season was organised for 2002; excavations only recommenced in 2003. This excavation schedule fortuitously corresponded with the agricultural cycle on the machair. In 2001 and 2002 the area around the settlement mounds was cultivated and this would have made access difficult. In 2003 and 2004 the area was fallow and the inconvenience of cattle-grazing could be remedied by the use of an electric fence.

The 2003 work on mound 2 was designed to excavate the final house (House 3; Figure 11), which covered the east end of House 2, and to explore the west end of the first house (House 1; Figure 12), which extended beyond the end of House 2 (Sharples 2003a). Consequently two separate trenches were opened at either end of the area excavated in 2000 (Figure 2). The western trench, 12 m by 7.6 m, was designed to examine House 1 and to confirm the presence or absence of pre-Viking deposits cut by House 1. The eastern trench, 14.5 m by 7.5 m, was designed to complete the excavation of House 3, and a small extension to the east was designed to examine the entrance.

The trench on mound 2A was extended in 2003 to the south, to try to define the extent of the mound in this area (Figures 7 and 13). This was done by excavating a series of three small trenches that demonstrated the existence of important structural remains some distance to the south of the previously excavated area. The trenches on mound 2 were supervised by Katherine Stronach, Suzi Reeve and Oliver Davis and on mound 2A by Katinka Stentoft with the help of Kate Waddington.

Figure 11. The initial cleaning of House 3 on mound 2 in 2003 from the south

The excavations were very successful. On mound 2, House 3 was almost completely excavated and the floor of House 2 was exposed ready for excavation in the following season. The excavation of the west end of House 1 recovered an important assemblage of pottery and steatite that was different to that in the later House 2 floors. Distinctive ceramics, combs and pins were recovered here and an earlier occupation, dating to the seventh to eighth century, was thus identified.

The excavation of mound 2A began to make sense of the spatial arrangement of the settlement on that mound. It was apparent that the main structural focus of the mound lay to the south of the area that was being examined and that a sequence of at least three north-south oriented houses were present in this area. The poorly preserved and difficult to interpret structures that survived in the main excavated area were badly-built ancillary buildings.

The 2004 season was designed to be the final season of excavation at Bornais. On mound 2, a large L-shaped trench, up to 30 m by 12.5 m, was opened up over Houses 2 and 3 (Figure 2). The principal goal was to complete the excavation of the floor of House 2 (Figure 14); any further information on the nature of House 1 would be a bonus. The trench outside the entrance to House 3 was extended by 2 m and its excavation exposed a deep stratigraphic sequence on the edge of the settlement mound.

On mound 2A excavation was extended to the south, creating a main area roughly 15.75 m by 11.25 m (Figure 7). The principal goal was to excavate as much as possible of the basal deposits in the northern half of the mound. It was realised that the basal cultivation soil (Figure 15) was rich in artefacts and animal bone that would provide important data to interpret the economy of the settlement in its earliest phase of occupation. The southern extension was designed to expose, but not excavate, a sequence of houses in this area. An 8 m by 2 m extension to the trench was also excavated to define the western extent of the settlement mound. Work on mound 2 was supervised by Suzi Reeve and Oliver Davis and on mound 2A by Kate Waddington.

The 2004 excavations proved to be successful and it was possible to claim that we had taken the excavations to a satisfactory conclusion. On mound 2, a series of three substantial and prestigious houses had been identified. Houses 2 and 3 were completely excavated and House 1 was defined and partially explored. The presence of Late Iron Age settlement preceding the Norse occupation had been confirmed and a large assemblage of material culture had been recovered. On mound 2A, a sequence of deposits spanning the Norse occupation had been excavated, the domestic focus of the mound had been identified and a series of ancillary buildings, including a comb-makers’ workshop, had been excavated. Substantial assemblages of artefacts and ecofacts were recovered that provide important information on the economy of the Norse settlement and this included an important assemblage of Early Norse material.

The Norse settlement of Atlantic Scotland – N Sharples

The Norse settlement of the North Atlantic is a topic of considerable interest to a wide range of scholars with interests in history, linguistics, literature, place-names and archaeology. Scholarly interest in this topic ranges across a number of countries and though research is dominated by scholars from the Scandinavian countries and from the separate nations of Britain and Ireland, important work is also carried out by scholars in Germany, France, Canada and the United States of America, amongst others. The international nature of the research was recognised by the creation of the Viking Congress in 1950. This meets every four years in Britain and Scandinavia and involves an intellectually diverse range of scholars from a number of countries (Smith 2016). The literature on the settlement of the North Atlantic is consequently vast and varied, and there are numerous strands and aspects of research that could be explored and debated, but not in this introduction. Instead I will limit myself to a short summary of the history of ‘Viking’ settlement exploration in Scotland, which identifies some of the problems with the archaeological record and discusses some of the key settlements in Atlantic Scotland.

Figure 12. A view from the east of the excavation of the west end of House 1 on mound 2 in 2003

Figure 13. A view looking south across mound 2A during the excavations in 2003

Figure 14. A view from the east looking across House 2 on mound 2 in 2004

Figure 15. A view of the excavation of the basal cultivation marks on mound 2A

History

In the early twentieth century three sites were excavated that transformed our understanding of the archaeological record (Figure 16): Freswick, Caithness (Curle 1939), Birsay, Orkney (Curle 1982) and Jarlshof, Shetland (Hamilton 1956). These sites are still of considerable significance today and reinterpretation of the remains at Freswick and Birsay (Morris 1989; 1996) was undertaken in the 1970s and 1980s. However, Jarlshof, perhaps the most important of these sites, has not been systematically reconsidered, though it is acknowledged that this needs to happen (Graham-Campbell and Batey 1998, 156).

Excavations on Scandinavian settlements were limited in the period between the Second World War and the 1980s, comprising a couple of sites in the Western Isles, at Drimore (MacLaren 1974) and the Udal (Crawford 1975, 1981, 1986, 1988; Selkirk 1996); the settlements at Skaill, Deerness (Buteux 1997) and Buckquoy (Ritchie 1977) in Orkney; and the farm at Underhoull, Unst, Shetland (Small 1966).² The most important of these excavations was at the Udal, where a large area with numerous structures and a wealth of artefactual and environmental material was examined. Unfortunately the Udal excavations have never been fully published (Ballin Smith 2018).

At the end of the 1970s a new phase of interest in the Scandinavian archaeology of the Northern Isles began. This was pioneered by John Hunter and Chris Morris who were originally based at the University of Durham. The work began with their joint excavations on the Brough of Birsay (Morris 1989) and continued with a range of projects that included major excavations by Morris at Freswick, Caithness (Morris et al. 1995), Orphir, Orkney (Batey and Morris 1992) and Beachview, Birsay, Orkney (Morris 1996) and by Hunter at Birsay (Hunter 1986) and Pool, Sanday, Orkney (Hunter 2007). Offshoots of this programme were ‘tapestry’ excavations at Tuquoy, Westray (Owen 1993) and St Boniface, Papa Westray (Lowe 1998). A separate but contemporary project was the excavation of a Late Norse settlement at Sandwick, Unst, Shetland (Bigelow 1985).

Figure 16. A map of Scotland showing the location of the principal Scandinavian settlements mentioned in the text

In recent years the work on Orcadian settlements has continued with the important excavations at Snusgar on the Bay of Skaill (Griffiths 2015; Griffiths and Harrison 2011) and Quoygrew in Westray (Barrett 2012). In Shet-land a project on the island of Unst resulted in the excavation of three houses, which were part of a large group of houses that survived as field monuments (Bond 2013; Larsen 2013).

In the Western Isles the excavations at Bornais, reported here, and at the smaller settlement at Cille Pheadair (Parker Pearson et al. 2004b; 2018) transformed our understanding of the western fringe of Atlantic Scotland. Small-scale rescue excavations have also occurred at Barvas, Lewis (Cowie and MacLeod 2015) and Bostadh, Lewis (Neighbour and Burgess 1997), which are directly relevant to the current report. Unfortunately there have been no excavations of Norse settlements on the Inner Hebrides, or the adjacent areas of mainland Scotland (Graham-Campbell and Batey 2017; Raven 2017), though these areas have important evidence for Viking settlement preserved in their place-names (MacNiven 2015).

Problems

This long list of excavations appears to provide a comprehensive record of the archaeology, but closer examination reveals a very patchy understanding of the Norse settlement record in Atlantic Scotland. For example, a number of houses have been explored but various systemic problems have restricted our understanding of the occupation of these structures. The problems include:

•poor preservation;

•complex building histories;

•incomplete excavation;

•inadequate publication.

The poor preservation of material is often due to the location of such houses on the acidic moorland soils that are so common in Atlantic Scotland. Many phases of activity and rebuilding are often represented by conflated stratigraphy that does not fully document the complex histories that occurred. This is a particular problem with the sites excavated in the Viking Unst project: even at a relatively straightforward site such as Hamar (Bond 2013), numerous phases of activity were present but only partially survived later activity and erosion. A complex site such as Belmont (Larsen 2013) is very difficult to interpret and it is impossible to fully understand the nature of the occupation at any one period of time. Similar problems are present in Orkney: at the western settlement on the Brough of Birsay, the stratigraphy is very shallow (Morris 1996).

The lack of stratigraphy would not be a major problem if it were not for the nature of the settlement record: Norse settlements are characteristically long-lived. In Orkney it seems likely that most of the settlements established at the beginning of the Norse period continued to be occupied up to the recent past. Quoygrew (Barrett 2012) and Skaill (Buteux 1997) are examples of such Orcadian settlements that arguably have complete sequences that ended only in the recent past. This is also the case at Jarlshof, Shetland (Hamilton 1956), though here the settlement sequence extends back to early prehistory.

The sequences are made complex by the relatively short lifespan of the built structures. It seems to have been culturally important for houses to be systematically rebuilt every couple of generations. The nature of the rebuilding varies dramatically. Sometimes the earlier structure was demolished and rebuilt, with only an apparently ‘casual’ incorporation of older structural features in the new build (e.g. Birsay Building C; Morris 1996). In other instances, the building survives largely intact with additions and extensions added (e.g. Quoygrew; Barrett 2012). These rebuildings can result in a displacement of the settlement, and many excavations have only been successful in exploring sequences because the current farm lies to one side of the original settlement mound (e.g. Pool; Hunter 2007).

The complex sequences of activity cause major problems to the archaeological investigation of these sites because unless there is a serious threat of destruction, archaeologists seldom feel justified in removing upstanding, well-preserved, stone-built house walls. Good examples of the problems caused by the preservation of structural features are Birsay Area II (Morris 1996) and Jarlshof (Hamilton 1956), where the confused agglomerations of buildings makes it almost impossible to get a clear understanding of any phase of the settlement. Later structures cover and obscure the remains of the original settlement and make it impossible to appreciate fully the earliest phases of building.

Finally a recurrent feature of the archaeological work on the Norse settlement of the Scottish islands is the poor record of publication. A detailed report on the work undertaken on the central area of the Brough of Birsay, Orkney in the 1930s was only published by Morris in 1996 and we still await reports on the work undertaken in the western settlement (Area III). The extensive and long-lasting excavations at the Udal on North Uist have never been published and only a very partial understanding of the remains is accessible in the interim reports (Crawford 1975; 1981; 1986; Selkirk 1996). Complete excavation reports of substantial settlements with significant chronological sequences are restricted to Jarlshof (Hamilton 1956), Pool (Hunter 2007) and Quoygrew (Barrett 2012).³

Viking houses in Scotland

As a result of these problems our understanding of the archaeology of Scandinavian settlement in Scotland is more limited than might first appear. Complete plans of houses dating to the ninth or tenth centuries are very rare and problematic.

The most famous early house is that identified at Jarlshof (Figure 17 C), and this is frequently used to illustrate the nature of Norse settlement in the colonisation phase (i.e. Graham-Campbell 1980). However, this structure was substantially rebuilt in later periods and our understanding of the earliest occupation at Jarlshof is very problematic. Hamilton (1956) argues that the principal house was a rectangular structure approximately 21 m long by 6 m wide, oriented east–southeast and west–northwest. It had a pair of opposed entrances in the western half of the house and a subsidiary entrance in the east gable. The main entrances divide the interior into two; a living area, occupying approximately two-thirds of the house at the east end, and a kitchen in the remaining area at the west end. The living area was divided into three aisles indicated by two lines of post holes and the presence of small upright stones suggests the side areas had benches. No hearth was identified in the living room but a spread of red ash ran down the central aisle. The kitchen was slightly raised and had a fireplace and an oven, which was located in the centre of the room but ran up to and abutted the back wall. This house underwent considerable modification in later phases when a byre was created in the east end, but it is important to note that there was no evidence for cattle-stalling in the original house. It was also not classically bow-walled: the south wall is very definitely straight.

The evidence from Pool is very important and two major phases of relevant activity have been recognised. Neither phase is particularly well preserved from an Orcadian standpoint. The first phase (phase 7) spans the transition from a Pictish to Viking settlement. Hunter (2007, 138–45) suggests that this was a period of cultural assimilation when a resident local population was transformed by contact with Scandinavia. Phase 7 is split in two, with the earlier part (7.1) comprising reuse of existing Late Iron Age buildings, the construction of a series of (badly preserved) curvilinear structures on the southern slopes, and an unusual rectangular building, Structure 25, on the northern half of the mound.

Structure 25 is a rectangular building, 16 m by 5 m, with slightly bow-shaped walls (Figure 17 I). It is assumed to indicate Scandinavian influence, but its width, the presence of an axial line of posts, and an arrangement of upright packing stones that suggest a series of transverse partitions projecting from the side walls (Hunter 2007, 141), were not typical features in Norse houses. Rectangular buildings of Late Iron Age date were identified at the Howe (Ballin Smith 1994) and this appears to be a well-established tradition in Caithness, where the structures were known as wags.

In phase 7.2A at Pool, Structure 27 is also a rectangular building, with a central stone-lined hearth and the entrance at the south end of the east wall (Figure 17 H). It is argued that this is a house, measuring 9.5 m by 3.5 m, but the plan of this structure is drawn from a composite of detached sections of badly preserved walls, including a curvilinear wall that belongs to an earlier Iron Age structure. Structure 27 is an unusually squat structure that has none of the normal drains and partitions that characterise Early Norse houses in the Northern Isles (see in particular the early house at Quoygrew [Figure 17 E] and the structures at Snusgar [Griffiths 2015]). The hearth of Structure 27 certainly suggests the presence of a Norse house but it is unclear if the precise dimensions of the structure are accurate or whether it is indeed a house rather than an ancillary building with a specialist function.

A relatively well-preserved bow-walled house identified at Skaill (Buteux 1997; Figure 17 F) was estimated to be 18.9 m long and 4.6 m wide. It was divided in two by an entrance in the centre of the west wall. The northern half of the house was identified as a living area; benches defined by edge-set slabs were identified on both sides of a central area. The south side was defined by a large rectangular hearth with stone paving and a kerb. A line of paving stones leading south from this might have led to an entrance through the south gable. This house was built over a sequence of two earlier rectangular structures. The primary structure was thought to be Pictish in date by Gelling but, in the final publication, it was argued that it could be Norse (Buteux 1997). Both the preceding structures were relatively small and do not appear to have been major domestic buildings.

Three substantial houses were excavated on Unst (Turner et al. 2013): Belmont, Hamar and Underhoull (the latter two are illustrated in Figure 17 G and B). These were large bow-shaped halls; Belmont and Hamar had an internal space roughly 20 m long and 5 m wide and Underhoull was roughly 18.4 m long by 4.5 m wide. The construction of the latter two houses has been securely dated to the middle of the eleventh century. Belmont is argued to begin earlier, but the radiocarbon dates from secondary contexts cluster in the twelfth century AD. These houses are important in demonstrating the relatively late use of bow-walled longhouses and do not provide any secure evidence for an early date in the ninth or tenth centuries AD.

The house at Underhoull had several distinct features worth noting (Figure 17 B). It had two entrances; the main entrance was at the east end of the south wall and there was a subsidiary entrance through the east gable. Several structures were built to abut the house walls, but none had direct access to the interior of the house. The interior was split into three rooms. At the west end was a room with a paved stone floor that was interpreted as a cold store, and contained fragments of steatite and ceramics. The central room was interpreted as a living area with a wooden floor. Although some ash deposits were found in this area, they are not sufficient to suggest an in situ hearth, and it was argued that this would have sat on the wooden floor. The east end of the building was tentatively interpreted as a barn (Bond 2013, 159). The evidence for use of the house interior at Hamar and Belmont is more difficult to interpret.

Viking houses in the North Atlantic

This is a very imperfect understanding of the Early Norse settlement in Atlantic Scotland and to gain a more accurate picture of the character of substantial bow-walled longhouses, it is necessary to look to other areas of the North Atlantic where the archaeological record is more informative. Several houses have been excavated in the Faroes, notably Toftanes (Stummann Hansen 1991) and Kvívík (Dahl 1951; Matras 2005). After his excavations at Toftanes, Stummann Hansen went prospecting in Unst, Shetland, where he identified a number of Norse longhouses (including the example at Hamar discussed above) that exhibited distinctive architectural characteristics that he dates to the ninth to tenth centuries AD: the houses ‘…have curved walls of approximately one metre in thickness. The houses may have sunken floors with benches along the inner walls in the upper parts of the houses. The upper part of the structure has probably been the sleeping room, while the central part has been what might be termed the living room with the long-fire. The lower-lying end has been the byre with a drain running through an opening in the gable’ (Stummann Hansen 2000, 99). Such houses were typically aligned downslope to allow drainage out of the house.

Figure 17. Simplified plans of early Scandinavian houses in Atlantic Scotland. A Drimore, South Uist, Western Isles; B Underhoull, Unst, Shetland; C Jarlshof, Dunrossness, Shetland; D Brough of Birsay, Mainland Orkney; E Quoygrew, Westray, Orkney; F Skaill, Deerness, Orkney; G Hamar, Unst, Shetland; H Pool, Sanday, Orkney; I Pool, Sanday, Orkney

Stummann Hansen’s interpretation is based on the houses at Toftanes, Jarlshof and Birsay, but only Toftanes had a byre; evidence for byres was not definitely present at any of the other sites. The discussion also specifically excludes the evidence from Iceland, which is acknowledged to be different but dismissed as misleading and too late to be relevant. The recent work in Unst (Turner et al. 2013) does not support the argument that these houses were early and had an internal byre.

In a recent synthesis of the evidence from Iceland, Milek (2006) has argued that Early Norse houses have a distinctive form and spatial organisation. Most of the Icelandic houses were classic bow-shaped longhouses, but there are examples that show little expansion in the central area, e.g. Snjáleifartóttir and Herjólfsdalur V can be compared to the early house at Jarlshof. The main entrance to the house was located close to the end of one of the long walls, but other subsidiary entrances were common, often on the same wall as the main entrance. There were attached annexes, small rooms built onto the sides or ends of the house that were accessed from the main house; annexes with direct access to the outside are much rarer. Separate ‘pit houses’ are also common.

The interior of the Icelandic house was normally divided into three separate rooms: the living room, which was a large room at the centre of the house, and two gable rooms, which were much smaller spaces at either end of the house (Milek 2006, 98). These rooms were probably separated by partition walls made from timber but this is often difficult to identify. The living room was divided into three aisles; the middle aisle contained a large hearth, which had a stone kerb and paving. This aisle frequently accumulated ash that appears to have been deliberately spread to create a dry surface. The side aisles, in contrast, tend to have very thin insignificant deposits, which suggests they were protected by wooden floors; these could have been low benched areas.

A good example is the house at Aðalstræti, and the internal occupation of this house has been analysed in some detail by Milek (ibid, fig. 4.42). A similar pattern was identified by Myhre in his analysis of the distribution of artefacts at Oma in Norway (Myhre 1982, fig. 8), and by Einarsson in his report on the excavations at Granastaðir in Iceland (Einarsson 1995, fig. 43).

The evidence from Scotland currently does not fully conform to this model. Most Norse-period Scottish houses had an asymmetrically located entrance, placed at one end of a side wall, but very few houses appear to have a threefold internal division, with a central living space dominated by a hearth. The nearest approximation is Underhoull (Bond 2013), where a threefold division has been suggested. However, the proportions of these rooms at Underhoull were not similar to the Icelandic model, the living room was not central and evidence for a major hearth was missing, arguably because the living room had a wooden floor.

Most of the Scottish houses have been interpreted as having a twofold division of the internal space, with a benched living area located at one end of the house, e.g. Jarlshof and Skaill. These living areas seldom have evidence for a substantial hearth, which instead was located in a separate kitchen area. It is unclear why the Scottish evidence is not similar to that from Iceland. It could be a significant cultural difference possibly indicating the influence of native tradition, though this would raise some problems over what that tradition was. It is not impossible, however, that the difference simply reflects the incomplete archaeological record for the Scottish settlements.

Late Norse houses

The later Norse settlement of the twelfth to fourteenth centuries has been much more thoroughly explored; a wide variety of structures has been investigated and a large quantity of material recovered. However, until recently these later structures were not central to the discussion of the Norse settlement, scholarly interest declining as the temporal distance from the conquest increased. Important excavations of later settlements include Quoygrew, Westray (Barrett 2012) and Sandwick, Unst (Bigelow 1985), and the reports on these sites focus on the intrinsic significance of the architecture and economy of these periods.

Quoygrew in Orkney is one of the most informative recent excavations and has been fully published (Barrett 2012; Figures 17 E, 18 K). The site originated in the tenth century and an important eleventh-century house was discovered. The principal excavated structure was built on top of this around AD 1200, and comprised a well-defined rectangular house with narrow stone-built walls. The internal space was split into two large rooms by a wall, which appears to have been substantial though it was completely removed by later robbing. The main entrance lay in the south wall, slightly west of the central axis of the house and directly opposite the access to a small subsidiary room that projected from the north wall. The two main rooms comprised a living room to the east of the main entrance and a smaller byre to the west; the gable wall in the byre had a subsidiary entrance. The living room had an ill-defined central hearth and a raised benched area around three sides defined by slots for upright stones. The western room was identified as a byre by the presence of edge-set stones defining stalls and a drain running down the centre of the room that exited through the gable door.

Figure 18. Simplified plans of later Scandinavian houses in Atlantic Scotland. J Sandwick, Unst, Shetland; K Quoygrew, Westray, Orkney; L Pool, Sanday, Orkney

This structure was maintained for an estimated 200 years before a major architectural change, the addition of a large additional room to the east that was accessed by an entrance leading from the interior of the main house into the additional room through the house’s eastern gable wall. This room might have served as a bedroom and there is evidence for a major reconfiguration of the main living room, with the removal of the side benches and the construction of a large central drain. There was much subsequent modification of the internal space before the house eventually went out of use in the sixteenth century (Barrett 2012, 94).

Sandwick in Unst, Shetland is a single small farm (Bigelow 1987, 27; Figure 18 J) that started life in the twelfth century AD and was occupied through to the end of the fourteenth century. It comprised two spaces, the principal large longhouse and a subsidiary room attached to the side of the building. The longhouse was 17.5 m by 4 m internally and was divided into four separate spaces by changes in the floor height. The southern half the building was the living room. This was bounded by a paved crosspassage that ran between the east-facing external door and an internal door to the subsidiary room on the house’s west side; the subsidiary room also had a west-facing external door. To the north of the cross-passage was a kitchen and to the north of this was a byre area with a drain that led through a door in the gable end to a separate external room.

There is no clearly defined function for the subsidiary building on the west side but its floor was paved and a drain was present along its north wall. The presence of a byre in the house is identified as the major new feature of these Late Norse houses by Bigelow (1987, 31). Important changes made during the life of the Sandwick house include the transformation in the late twelfth or early thirteenth century of the long hearth at the centre of the living room into a box hearth placed in the corner of the room, and the reduction in size of the side benches in favour of a larger gable-end bench. Corner hearths are a feature of Scandinavian houses and may indicate continued contact with the homeland (Christie 2002).

The spatial arrangement and relationship of these defined spaces within a house appear to be distinctive to the different island groups of Scotland. The excavation of the structures at Quoygrew and Pool (Figure 18 K and L) demonstrated the importance of the linear arrangement of space in Orkney, and the evidence from Freswick suggests Caithness has a similar development. The excavation of Sandwick and Jarlshof in Shetland in contrast, demonstrated the increasing importance of buildings constructed parallel to each other. In both areas the fragmentation of the large domestic space of the Early Norse period and the appearance of well-defined and separated spaces seems to be a major motivating force (Crawford and Ballin Smith 1999).

Similar trends can be observed in the development of domestic housing in Iceland and Greenland. In Iceland the trend is for the separate rooms to be arranged along a single passage, with the main living room placed at the very end of the passage, and for an arrangement of ancillary structures creating a facade (Ágústsson 1982; Stoklund 1980, 127). Similar trends occur in Greenland (Roussell 1941) though these houses have been argued to have a less formal plan and to be better described as clustered (Skaaning Høegsberg 2009, 96). From the inception of our research, understanding the evolution of comparable structures on the Western Isles was a significant goal of the Bornais excavations.

Cille Pheadair – N Sharples

The settlement at Cille Pheadair on South Uist was excavated in three summers between 1996 and 1998 by Mike Parker Pearson, and the following synopsis is based on a draft text of the final report kindly provided by the director (Parker Pearson et al. 2018). The settlement comprised only a single house in each phase of occupation. The house underwent major structural modifications (Figure 19), or a complete rebuild, regularly over a period of between 160 and 220 years, from cal AD 980–1015 (68% probability) to cal AD 1175–1220 (68% probability) (Marshall et al. in Parker Pearson et al. 2018). The sequence was divided into nine