sign in sign up
<<
Home , All Cannings Cross, Banbury Hill, Barbury Castle, Blewburton Hill, Bratton Castle, Bredon Hill

3 The Regional Pattern by Mark Corney and Andrew Payne

The of : their P Williams Freeman. Both worked primarily morphology and environs in , although Sumner also extended his survey work into neighbouring by Mark Corney southern and . His pioneer­ ing survey of Chase (Sumner Introduction 1913) resulted in the presentation of plans produced to a very high standard of draughts­ The methodical study of Wessex hillforts has manship that had a profound influence on the its origins in the late 19th century and open­ graphic style of the early RCHM surveys in ing two decades of the 20th century and, . (These were largely produced with the notable exception of Pitt Rivers’ during the 1930s but, owing to the outbreak excavations at Winkelbury, Wiltshire (Pitt of the Second World War, were not published Rivers 1888), initially developed as a non- until 1952.) Williams Freeman published sur­ intrusive survey tradition. Given that this veys of many Hampshire hillforts (1915) and volume is presenting the results of non-intru­ although his graphical style did not match sive methodologies it is worth pausing to that of Sumner, he provided an important review the development of this tradition in record of many sites as they appeared 100 Wessex. Although the late 19th century years ago (see for example the section on marks the main starting point for investiga­ Norsebury, pp 66–71). tions, any review must acknowledge the con­ This auspicious start to non-intrusive tribution of the superb surveys produced by investigation was to prove something of a false Philip Crocker on behalf of Sir Richard Colt dawn as the emphasis began to move rapidly Hoare. These plans (Fig 3.1) are a remark­ to excavation. In Wiltshire Maud able and accurate record of many monu­ Cunnington, assisted by her husband B H ments, including hillforts (Colt Hoare 1812, Cunnington, investigated a number of hill- 1819). They frequently depicted major hill- forts (Cunnington 1908, 1925, 1932a, forts and their environs thus presenting the 1932b, 1933; Cunnington and Cunnington first ‘landscape’ plans specifically executed 1913, 1917) as well as the important Early to record the extant . Iron Age settlement at Cross Pitt Rivers was certainly aware of the (Cunnington 1923). In Hampshire it was C F importance of recording surface features C Hawkes who took the lead in hillfort exca­ and the volumes contain vation, investigating St Catherine’s Hill many plans of his sites prior to excavation (Hawkes et al 1930; Hawkes 1976), Buckland (eg Pitt Rivers 1888). In addition he also Rings (Hawkes 1936), Hill (Hawkes had scale models produced of many sites 1939) and (Hawkes 1940). that depict the condition of the monument Hawkes’ excavations in Hampshire were to be prior to excavation (Bowden 1991). Pitt central to his ‘ABC’ scheme for the British Rivers’ assistants, most notably Herbert Iron Age (Hawkes 1931, 1956, 1959). South­ Toms, were to develop this analytical survey west of our study area, in Dorset, Sir Mor- skill further (Bradley 1989). timer Wheeler and his team examined a Earthwork depictions of most of the hill- number of major hillforts at Maiden forts in the Wessex region to a common spec­ (Wheeler 1943), Poundbury (Richardson ification were first produced by the Ordnance 1940) and Chalbury (Whitley 1943). Survey for the first edition 6-inch and 25­ The growth of aerial photography in inch maps (Crawford 1955; Phillips 1980). In archaeology in the decades following the opening years of the 20th century a small the Second World War had a profound number of fieldworkers began to produce impact in the region. The potential of this larger-scale, divorced surveys of many Wessex method had already been demonstrated by hillforts. In 1908 Allcroft published Earth­ Crawford and Keiller (1928) but it was not work of , and of particular interest are until the formation of the Uni­ the investigations of Heywood Sumner and J versity Committee for Air Photography and,

131 THE WESSEX HILLFORTS PROJECT

Fig 3.1 The surveys of Oldbury and published in 1812 by Sir Richard Colt Hoare in his Ancient , Volume 2 (from NMRC Library – The Ancient History of Wiltshire, Volume 2, North Wiltshire, Plate VIII, pages 40–41, originally published 1812, re-published 1975).

132 THE REGIONAL PATTERN somewhat later, the Air Photography Unit lation. Whatever the undoubtedly complex of RCHME that the intensity and complex­ factors behind the choice of location might ity of the later prehistoric landscape in Wes­ have been, there were clear preferred loca­ sex could be fully appreciated. Collin tions within the landscape and strong Bowen (1975, 1978) began the elucidation regional trends can be discerned over much of this landscape by using the results of air of the area covered by this study. There is a photography and ground survey. Building tendency to view Wessex hillforts as part of upon this came a landmark study with the the classic chalkland prehistoric landscape. publication of a major air photographic However, even a cursory glance at the map of study of the environs (Palmer Wessex hillforts immediately shows that the 1984). Using only air photographic sources, great majority are located on the limits of the an area of 450 sq km was mapped at a scale chalk, either on the escarpment edge or over­ of 1:10,000 with select windows at 1:5000 looking the major valleys such as the Avon, and 1:2500. This study set a new standard Test, Kennet and Wylye. Only a relatively in air photographic analysis of extensive small number are within the main chalk mas­ relict landscapes that has been repeated sif and even here it is possible to discern pre­ since over many areas of the country (eg ferred locations in certain regions. Bowen 1990; Stoertz 1997). It must be stressed that many of those In addition to pure air survey, multidisci­ sites grouped together below have had little plinary extensive projects within Wessex if any modern excavation and the detailed using air photography, earthwork survey chronology is far from clear. It will be seen, and geophysics were undertaken by the for­ though, that there are certain common links mer RCHME (now part of English Her­ in the morphology of these groupings. Obvi­ itage). Studies of southern Wiltshire (NMR ously none of these sites will have existed in archive) and Plain (McOmish et al isolation; the other components of the envi­ 2002) demonstrate the level of detailed rons of those hillforts examined by this pro­ analysis attainable by these means and were ject, and noted in the preceding section, instrumental in stimulating further projects must be borne in mind (see also pp 139–41). (eg Bradley et al 1994). Escarpment locations The morphology of Wessex hillforts There are two major groups occupying north-facing escarpments: The main morphological characteristics of the hillforts of the British Isles have been /Marlborough Downs examined in detail by numerous authors in Group recent years (eg Cunliffe 1991, 312–70; The largest group within this category is Forde-Johnston 1976; Hogg 1975). These located on the north-facing escarpment of studies looked at the phenomenon of hill- the Downs and Marlborough forts from both national and regional per­ Downs – along the route of the ‘Ridgeway’ – spectives. In Wessex work has continued at a comprising ten hillforts: number of levels; ranging from major pro­ (Harding 1976), Segsbury, Rams Hill, Uffin­ jects involving large scale excavation, such gton Castle, Hardwell Camp, Liddington as at Danebury, to intensive non-intrusive Castle, (a ploughed out univallate survey utilising multi-disciplinary method­ site known only from air photographs (see ologies. Among the latter there have been a Fig 2.53) and Barbury Castle. The latter large number of earthwork and air photo­ three also overlook the junction between the graphic surveys that have added both a con­ Og Valley and the northern chalk escarp­ siderable amount of detail and important ment, allowing easy access from the chalk observations that, until now, have remained massif to the upper Thames Valley. Beyond largely unpublished. The following section Barbury Castle there is a gap in the is largely based upon this work. area and then two western fringe outliers of the group: Oldbury and Oliver’s Camp. Location North Hampshire Escarpment The choice of hillfort location cannot be con­ This group comprises six hillforts and over­ sidered as a random decision. The correlation looks the middle reaches of the Kennet Val­ between hillforts and earlier monuments is ley with extensive views to the north. The well known – even though there is still consid­ westernmost outlier of the group, erable debate on the significance of this corre­ Hill near Marlborough, is possibly a Late

133 THE WESSEX HILLFORTS PROJECT

Iron Age construct (Cunliffe 1991, 153; are (Clay 1935), Corney 1997) while the remainder; Wal­ Wick Ball Camp, (Tisbury) bury, Beacon Hill, , Bowry Walls and Castle Rings (Donhead). North-east of and Winklebury occupy the main escarp­ the confluence with the Avon, the high ment edge set back some distance from the ground overlooking the valley of the Valley. Bourne has (Cunnington Smaller scarp edge clusters can be dis­ 1925; Guido and Smith 1982), the major cerned in a number of areas. The north­ complex on Boscombe Down West western escarpment of in the (Richardson 1951) and Sidbury. area has four hillforts – Scratch- bury, Battlesbury, (McOmish The River Stour, Dorset et al 2002) and – forming a discrete The Stour and its tributaries host a number cluster. At the eastern end of the Vale of of major hillforts and, by way of the Black- Martinsell Hill and Giant’s Grave moor Vale, give access through to the dominate the south-facing escarpment. southern edge of the Somerset Levels and the two large and impressive Wessex fringe River Valley Foci hillforts of South Cadbury (Barrett et al In Wessex a number of the major river 2000) and Ham Hill (Dunn 1997). The valleys act as notable foci for hillfort loca­ main concentration of larger hillforts along tions. This applies to four principal river sys­ the Stour Valley is between Hengistbury tems and their tributaries: the in Head and the area, effec­ Hampshire, the River Avon in Hampshire tively defining the southern and western and Wiltshire, the River Stour in Dorset and limits of Cranborne Chase with its distinc­ the and Piddle in Dorset. Of these the tive Iron Age Settlement pattern (Barrett Stour and the Avon have the greatest concen­ et al 1991). This group comprises Duds- tration of hillforts and associated landscapes. bury, Spettisbury (Gresham 1940), Badbury (Crawford and Keiller 1928), The River Avon in Hampshire and Buzbury, (Richmond 1968; Wiltshire RCHM 1970c) and The River Avon and its tributaries, in partic­ (RCHM 1970b). Beyond Hambledon Hill, ular the , has the largest and where the valley broadens out into the most coherent group comprising 25 hillforts Blackmoor Vale, are smaller hillforts at and the major emporium at Hengistbury Rawlsbury and (RCHM Head (Cunliffe 1987) where the Stour and 1970c). Close to the source of the Stour and the Avon flow into the English Channel. also situated on the south-western extremity Along the lower stretches of the Avon, of the Wiltshire chalk lies , east of the river and on the fringe of the a multi-period prehistoric focus of compara­ are four small univallate enclo­ ble complexity to Hambledon Hill. sures: Castle Hill, Castle Piece, Gorley and Frankenbury (Smith N 1999). Upon enter­ The Frome and Piddle, Dorset ing the chalk the hillforts along the Avon This is the smallest of the river foci in Wes­ become more frequent and are often of sex, having five hillforts within the catch­ larger proportions: Castle Ditches, Whits- ment area: Bulbury (Cunliffe 1972), bury (Bowen 1990; Ellison and Rahtz Woodbury (RCHM 1970a), Weatherby 1987), Clearbury, Woodbury (Bersu 1940; Castle (RCHM 1970c), Poundbury Brailsford 1948, 1949), , Ogbury (Richardson 1940) and Maiden Castle (Crawford and Keiller 1928), Heale Hill, (Sharples 1991; Wheeler 1943). The latter ’s Camp (RCHME 1979, 20–1), is the only hillfort in Dorset to have had an Casterley Camp (McOmish et al 2002; intensive study of its immediate environs Cunnington and Cunnington 1913) and (Sharples ibid). Chisenbury Trendle (Cunnington 1932b). Along the River Wylye, north-west of the The River Test, Hampshire confluence with the Avon are Grovely Cas­ The valley of the River Test is the eastern­ tle, Bilbury ring, Stockton (Corney 1994), most of the major valley foci in Wessex. Yarnbury (Cunnington 1933), Codford Cir­ Beyond here the Itchen and Meon cle, and then Scratchbury, Battlesbury and valleys have but one major fort apiece, Cley Hill (the last three also being on scarp St Catherine’s Hill (Hawkes et al 1930; edge locations – see above). West of the Salis­ Hawkes 1976) and Old Hill bury confluence, along the valley of the (Chapter 1, this volume). The Test and its

134 THE REGIONAL PATTERN tributaries flow from the heart of the Hamp­ Hillforts Survey was to test this theory fur­ shire chalk into Water and ther by attempting, through non-destructive have 13 hillforts within the catchment. The means, to determine if the sites do in fact largest in terms of area enclosed and com­ contain recurring patterns of spatial organi­ plexity are north of Stockbridge: Ashley’s sation. Like the Danebury Environs Project Copse, , Danebury, Bury Hill, before it, the Hillforts of the Ridgeway Balksbury, and Norsebury. Project (Gosden and Lock 2003; Miles South of Stockbridge and beyond the south­ et al 2003; Lock et al 2005) is now beginning ern limit of the chalk are other smaller and to provide a more detailed chronological poorly understood sites: The Walls, Tatch­ framework for hillfort development in the bury, Toothill Camp, Dunwood Camp, Ridgeway area, which will help to resolve Camp and Holbury. some of the fundamental archaeological questions concerning the group. It is encour­ Further observations on location aging to observe that all of the Ridgeway hill- One curious grouping observed in parts of forts are currently under stable grassland the region is the occasional pairing of large management regimes, some formerly having hillforts in close proximity to each other. In been under arable cultivation. The present some cases the benefit of excavation has sympathetic management of the sites is likely indicated support for the model put forward to stay in place for the foreseeable future, by Cunliffe that sees one monument aban­ with beneficial effects for the preservation of doned while another continues to develop archaeological features contained within and become a multivallate or developed hill- them (many of which have been revealed for fort. This is clearly the most likely case in the first time by the geophysical surveys). In the Dorchester area with Poundbury and some cases the increased knowledge of the Maiden Castle. In other areas the evidence hillforts derived from the geophysical pro­ is not so clear cut and the possibility of an gramme has acted as a catalyst for improving earlier manifestation of the pairing of sites the management of the sites. The cultural seen in parts of Wessex (Barrett et al 1991; resource value of many of the sites had previ­ Corney 1989) in the Late Iron Age should ously been largely ignored owing to the not be discounted. Of especial note are the paucity of knowledge of their internal charac­ close proximity of Hambledon Hill and Hod ter. This had led to the misconception that Hill in Dorset, Battlesbury and Scratchbury there was little of archaeological interest sur­ in Wiltshire, and Martinsell Hill and Giant’s viving or worth preserving within the contin­ Grave also in Wiltshire. It is of passing inter­ uously ploughed sites. est to note in the cases of Hambledon Hill The hillforts of the Ridgeway exhibit and Scratchbury that both hillforts enclose considerable differences in size, ranging causewayed enclosures, as does from the largest at Segsbury Camp (Let­ Maiden Castle. combe Castle) with an internal area of some 12ha, to the smallest enclosure of 1.2ha at Observations on the Ridgeway and Alfred’s Castle. As well as the varying size of Avebury Environs grouping of hillforts the areas enclosed by the Ridgeway forts, The hillforts of the Ridgeway and Avebury there are also obvious differences in the lay­ Environs grouping were the main focus of out of the defensive circuits. More often investigation during the first season of the than not the ground plans reflect the partic­ survey programme in 1996. The sites are ular topographical position of the site, but arranged approximately equidistantly in a some sites also exhibit more elaborate linear fashion along the escarpment edge of defensive architecture than others in the the Berkshire and Marlborough Downs, form of the entrances, the presence of addi­ coincident with the route followed by the tional screening an entrance and Ridgeway giving rise to the frequently used multivallation of the ramparts. Univallate term ‘Ridgeway hillforts’. Based largely on sites with an internal bank fronted by a their even distribution but without the back­ and an outer are the most ing of reliable dating evidence, it has been common defensive arrangement in the contended that the Ridgeway forts represent Ridgeway group (illustrated by Uffington largely contemporary centres of adjacent ter­ and Liddington for example). Other sites ritorial blocks (Cotton 1962), in which case are multivallate for part of their circuit, such they might be expected to exhibit similar as Segsbury and Oldbury (generally to rein­ densities and character of internal occupa­ force sections of the defences with less of tion. One of the specific aims of the Wessex a natural terrain advantage that can be

135 THE WESSEX HILLFORTS PROJECT

approached over level ground or to provide Morphology a more impressive symbol of strength or sta­ tus visible from the main avenue of There has been a long tradition of categorising approach to the hillfort). In some cases (for hillfort ramparts according to the nature of example at Barbury and Segsbury) there is the circuit, construction method and the limited evidence for earlier pre-hillfort materials employed (cf Hawkes 1971; Cunliffe enclosures removed or built over by the later 1991, 313–29). This level of analysis can be defences (see for example Bowden 1998) or based partly on surface observation, but a full remodelled and extended enclosures (for elucidation of the often complex sequence of example Oldbury and Alfred’s Castle). Bar- construction requires excavation. The litera­ bury shows the greatest elaboration of the ture on this aspect of hillfort circuits is well group having completely bi-vallate defences known and will not be repeated here. There (the product of successive phases of con­ are, however, a number of observations on the struction) and a defensive screen­ nature of hillfort ramparts that seem to have ing the approach to the eastern entrance. escaped attention and are especially relevant Segsbury has an outward projecting horn- to a number of the monuments examined as work shielding the eastern entrance and yet part of the Wessex Hillfort project. another variation on entrance reinforcement Detailed examination of a number of Wes­ is present at Oldbury, where a northerly sex hillforts, especially although not exclu­ extension of the second outer rampart sively confined to the univallate examples, screens the eastern entrance preventing a reveals a geographically widespread common direct approach and creating an extended feature; the construction of the bank and corridor to the entrance, which itself is ditch in a series of short, straight sections of deeply inturned. relatively uniform length. Feachem (1971) The varying approaches adopted for noted this feature of hillfort construction in entrance augmentation at the hillforts of the connection with unfinished hillforts but it is Ridgeway and the Avebury Environs have also visible on many complete examples. Also clear parallels with other hillfort sites else­ visible on most surviving hillforts is another where in Wessex. The outer rampart screen­ characteristic feature: ‘peaks’ and ‘troughs’ ing the eastern entrance at Oldbury uses along the length of the rampart tops that can exactly the same technique employed at the be seen to correlate with similar features eastern entrance of Hod Hill in Dorset along the base of the ditch. Both of these (Cunliffe 1991, fig14.13, 336). The possible traits may prove to be related to the construc­ out-curving of the main ramparts at the tion of the circuits and could indicate some­ eastern entrance of Segsbury to create an thing of the organisation of labour in the extended corridor approach mirrors the construction and maintenance of hillforts. construction of the eastern entrance at Ralston (1996) has noted a similar trend in Danebury. The eastern outwork at Barbury some of the oblong-shaped hillforts in eastern Castle has some similarity with the one pro­ Scotland (such as Tap o’Noth, Grampian tecting the eastern entrance of Yarnbury Region), where the form of the enclosure was Castle also in Wiltshire. A similar feature is dictated partly by the materials employed, also present at Chiselbury, Witshire. Lid­ involving the maximum use of straight dington, Segsbury and Uffington all show lengths of timber for ease of construction. evidence of originally having possessed two The straight length construction form in entrances – east and west – one of which was southern Britain is most readily seen on the subsequently blocked. The same practice surviving univallate hillforts although bi-val­ can be seen at Danebury and Beacon Hill late and multivallate examples also display in Hampshire and at Camp, the trait. Analysis of the best surviving (Thomas 2005). It is inter­ examples appears to indicate two main esting to note that the examples of hillforts groups, each distinguished by the length of with multivallate defences in the wider rampart unit: Group 1 with the rampart Ridgeway grouping at Barbury and Oldbury constructed in 30m to 40m lengths with the retain two opposing entrances, as is also the average being 32m; Group 2 featuring case with other multivallate sites farther lengths averaging 50m. afield such as Maiden Castle (Dorset) and Particularly good examples of Group 1 Castle Ditches (Tisbury, Wilts). The wide­ can be seen at Chiselbury, Wiltshire (Clay spread occurrence of blocked entrances at 1935), Figsbury, Wiltshire (Guido and Smith the hillforts investigated by the project is 1982), Ogbury, Wiltshire (Crawford and discussed below and in Chapter 4. Keiller 1928; Hampton and Palmer 1977)

136 THE REGIONAL PATTERN and , (Miles That this phenomenon is so readily et al 2003). All of the cited examples are apparent on the univallate examples should univallate enclosures of proven or probable not come as a surprise. When certain early Iron Age date (see Table 2). One multi- hillforts develop into multivallate enclosures vallate example within this group is Yarn- the sequence of re-modelling so clearly bury, Wiltshire (Crawford and Keiller 1928; demonstrated by excavation can lead to a Cunnington 1933) where the inner and outer ‘blurring’ of the original configuration. ramparts display this feature with remarkable Even so it is still possible to see a hint of uniformity around the entire circuit. this construction method on many multi- vallate or developed hillforts, such as Table 2 Group 1 hillforts Danebury, Hambledon Hill, Badbury and Maiden Castle. site type average Group 2 hillforts, where the unit length unit length averages 55m, appear to be less frequent than Group 1, but still form a significant Chiselbury Univallate 30m number of those examined as part of this Codford Circle Univallate 30m analysis. Included in this group is the Figsbury Univallate 32m unfinished fort on Ladle Hill, Hampshire Grovely Castle Univallate 32m (see Fig 2.21; Piggott 1931). At Ladle Fig 3.2 Univallate 35m Hill this pattern is remarkably clear with Ground view of the hillfort Ogbury Univallate 35m each incomplete unit still being separate defences of Segsbury Camp, Rybury Univallate 32m from its adjacent components. The angular­ Oxfordshire illustrating the Uffington Castle Univallate 35m ity of the changes in alignment is especially remarkably long and straight Walbury Univallate 30m clear at Segsbury and is still a striking sections of rampart out of Woolbury Univallate 35m feature of this monument when viewed which the hillfort is Yarnbury Multivallate 30m from ground level (Fig 3.2). constructed (James Davies).

137 THE WESSEX HILLFORTS PROJECT

Table 3 Group 2 hillforts known feature of a number of Wessex hillforts. Within the study area blocked site type average entrances have been examined by excavation unit length at Danebury (Cunliffe and 1991, 23–32) and Uffington Castle (Miles et al Alfred’s Castle Univallate 50m 2003) and other examples can be suggested Barbury Castle Bivallate 50m on the basis of the earthwork evidence. At a Bury Hill 2 Univallate 80m number of sites there are common indica­ Casterley Camp Univallate 45m tions of such an event. This will usually take Chisenbury Trendle Univallate 55m the form of a characteristic indentation in Fosbury Univallate 60m the rampart, marking where the rampart Ladle Hill Univallate 45m terminals of the former entrance have been Martinsell Hill Univallate 50m infilled, and, occasionally, there will be the Univallate 50m remains of outworks associated with the for­ Segsbury Univallate 70m mer entrance. The latter feature is especially Stockton Univallate 50m pronounced at Danebury (Cunliffe and Poole 1991, 23–32). Eagles (1991), in a It was suggested above that this unit form of paper examining the surface evidence from construction may indicate the way in which Beacon Hill, Hampshire, has drawn atten­ the building of the circuit was organised. tion to another probable example marked by Whether this indicates discrete groups from a subtle change in the external ditch, and a the hillforts’ hinterland contributing to the marked increase in the height of the coun­ communal monument or, perhaps, a reflec­ terscarp where the former gap had been tion of seasonal construction is of some con­ infilled. Other examples can be postulated siderable interest, but beyond the scope of on the basis of field observation. this discussion. However, at two of the sites One relatively common characteristic of where this phenomenon is especially clear, at univallate hillforts in Wessex is pairs of Ogbury in Wiltshire (Crawford and Keiller opposed entrances. Where only one entrance 1928, plate xxiv; Hampton and Palmer is now visible the observer will have a rea­ 1977, fig 7) and Perborough Castle in Berk­ sonably good idea where to seek evidence for shire it is possible that the unit lengths have a blocked counterpart. This very simple been influenced by the presence of an exist­ maxim has been used to identify blocked ing field system. This is especially clear at entrances at four Wessex hillfort sites and Perborough Castle (see Fig 2.2; Wood and others doubtless exist. At Liddington Castle Hardy 1962) where, as at Ogbury, there are a single entrance survives on the east side of indications of settlement within the field sys­ the monument (Figs 2.54–6). This is an tem beyond the hillfort circuit. At Ogbury unelaborate affair consisting of a simple gap air photographs and antiquarian plans also through the defences with the bank display­ record two smaller ditched enclosures abut­ ing slightly expanded terminals. On the ting the east side of the circuit (Colt Hoare western side of the circuit, directly opposite 1812; Hampton and Palmer 1977). Records the eastern entrance, the inner rampart has a in the Wiltshire SMR note Middle Iron Age slight inward kink and is slightly wider for a pottery from these and they may well post­ distance of 10m. The ditch narrows at this date the construction and use of Ogbury. point and a gap through the counterscarp is Beyond Wessex this phenomenon has still very evident on the ground. This feature also been noted on Bathampton Down, has all the characteristics of a blocked Somerset. Here a large univallate hilltop entrance and is close to the probable junc­ enclosure was laid out over an existing field tion between the hillfort circuit and a linear system (Crawford and Keiller 1928, plate ditch that approached the site from the west xxiii; Wainwright 1967). This association (the precise relationship having been trun­ between rampart form and earlier field sys­ cated by later quarrying). tems is not common and appears to be the At Segsbury, excavation of the eastern exception rather than the rule. entrance has shown that it was protected by a projecting and that the existing Blocked Entrances southern entrance might have been a later opening, possibly Iron Age or Roman (Lock The blocking of entrances, especially on and Gosden 1998, Lock et al 2005). The univallate sites at the period when elabora­ form of the eastern entrance is clearly of tion of the circuit commences, is a well- earlier Iron Age type and it would be tempt­

138 THE REGIONAL PATTERN ing to postulate that there was once a west­ projects such as the Danebury Environs ern equivalent. Study of the earthworks on (Palmer 1984; Cunliffe 2000) and the large- the western arc of the circuit has produced scale mapping of extensive areas of the Wes­ two candidates. To the north-west there is sex chalk (eg Bowen 1990; Bewley 2001; an opening through the rampart and the McOmish et al 2002) have begun to redress counterscarp is noticeably narrower at this this imbalance and have graphically demon­ point. However, the fact that the inner ram­ strated the complexity of settlement forms part is broken at this point would tend to and land division that coexisted through argue against this being an earlier entrance. much of the 1st millennium BC. In this gen­ A stronger candidate can be seen on the eral discussion attention will focus upon the south-west where the inner rampart, ditch immediate environs of the hillforts and pay and counterscarp all make a characteristic special attention to the growing body of evi­ kink over a length of 20m. This is also dence for enclosed and unenclosed extra­ directly opposite the eastern entrance and mural settlement. A more detailed the magnetometer survey shows a broad discussion of the environs of those sites band with significantly fewer features run­ investigated by the project will be found in ning east–west between these points that the gazetteer (pp 39–130). could be interpreted as a former road. Addi­ The positioning of hillforts appears to tionally, immediately beyond this postulated be based on many complex factors that blocked entrance, air photographs and mag­ can include proximity to earlier monuments, netometer survey have located an area of significant points of junction between possible occupation (Chapter 2). landscape divisions and geomorphological Perborough Castle has suffered serious factors. The project under discussion in degradation to the earthwork circuit from this volume has also added valuable corrob­ modern ploughing in recent decades. How­ oration to observations made from the study ever, the northern arc displays the charac­ of air photography regarding potential teristic inturned kink suggestive of a blocked settlements in close proximity to the entrance. Close examination of the earth­ hillforts. Excavation of such sites has been work also shows that the inner rampart at all too rare, although the recent examination this point is slightly disjointed and may indi­ of an extensive settlement located on a cate an original entrance form with slightly spur north of Battlesbury hillfort in offset terminals. The feature is clearly visible Wiltshire has demonstrated a very early Iron on air photographs (for example NMR Age date that probably precedes the first 4229/17, SU 5278/9, 1988) and beyond the phase of the hillfort (M. Rawlings, Wessex fort there are traces of a slight hollow-way Archaeology, pers comm). beside one of the field lynchets that predate Air photographs show potential unen­ the monument. closed settlements marked by pit clusters In addition to these sites, at least seven and maculae immediately outside and more Wessex hillforts display convincing south-west of Perborough Castle (see for earthwork evidence for the blocking of example Ashmolean Museum 7093/929 entrances: Castle Ditches, Tisbury (see held in NMRC ) and Segsbury above); Chiselbury, Wiltshire; Eggardon (NMR 1703/264, SU 3884/17, October Hill, Dorset; Grovely Castle, Wiltshire; St 1979). In the case of Perborough Castle Catherine’s Hill, Hampshire; Weatherby these features can be seen to extend right up Castle, Dorset and Yarnbury, Wiltshire. In to the outer edge of the ploughed-out ditch every case these are located directly opposite suggesting that they predate the construc­ the principal surviving entrance. tion of the hillfort defences and its associated counterscarp. At both Perbor­ Beyond the ramparts: ough Castle and Segsbury the cropmark evi­ hillforts in their landscape dence was confirmed by the detection of significant anomalies during the magne­ The brief history of Wessex hillfort studies tometer survey (see Figs 2.3 and 2.40, outlined above illustrates how, until Chapter 2). In neither case has the settle­ recently, there had been a strong tendency ment been verified or dated. to view hillforts in isolation. This myopia Martinsell Hill, a large univallate hill-top had created many problems with the way enclosure overlooking the eastern end of the hillforts and indeed the Wessex Iron Age , has been shown by the mag­ had been studied and interpreted. The netometer survey (pp 118–23) to be largely growth in ‘Landscape Archaeology’ and devoid of significant archaeological features.

139 THE WESSEX HILLFORTS PROJECT

On the plateau to the west of the enclosure At Old Sarum in Wiltshire casual finds and extending as far as the small promon­ and limited observation and excavation have tory fort of Giant’s Grave, some 1km dis­ recorded Iron Age material of the 4th cen­ tant, fieldwalking by Owen Meyrick tury BC to early 1st century AD over an area recovered spreads of Late and of at least 10ha beyond the eastern entrance Iron Age material (Swanton 1987). (conveniently summarised in Borthwick and At Bury Hill air photographs indicate a Chandler 1984). Owing to the circum­ mix of enclosed and unenclosed settlement stances of discovery it is impossible to ascer­ over an area of at least 4ha located 150m tain the exact nature and full extent of the south-east of the entrance to the hillfort. occupation, but both enclosed and open set­ Again magnetometry has confirmed this loca­ tlement seems probable. There are similar tion and added clarity to the marks observed records of extensive spreads of later Iron on the air photographs (Figs 2.14–15). Age material outside the principal entrance There is now growing evidence that signifi­ to Badbury in Dorset. This spans the 3rd cant extramural settlement is a common fea­ century BC to early 1st century AD and ture of many hillforts in Wessex and in the case includes an area that developed into a small of developed or multivallate forts this often shrine in the Romano-British period (M takes the form of enclosed activity in relatively Papworth pers comm). close proximity to the principal entrance. The presence of these clusters of extra­ At Yarnbury in Wiltshire a large (approx­ mural activity appears to have been largely imately 3ha) oval enclosure is sited 400m ignored and yet they must surely represent south-east of the eastern entrance of the another potentially important component of developed hillfort. Although unexcavated, a hillfort landscape. To date these patterns pottery of 3rd–1st century BC date was appear to have relatively discrete distribu­ recorded when the site was damaged by tions, with a notable concentration in close road widening in the 1970s (Wilts SMR). proximity to the hillforts of the Wylye Valley This material is contemporary with the in southern Wiltshire. Farther east, on the ceramics recovered from the interior of the Hampshire chalk, this pattern has, with the developed hillfort during the 1932 excava­ exception of Bury Hill, so far failed to mani­ tions (Cunnington 1933). fest itself convincingly. The enclosure at Still in Wiltshire, air photographs held by Houghton Down (Cunliffe and Poole, at the National Monu­ 2000e) is, at just over 2km from Danebury, ments Record Centre (NMRC) in Swindon too far to be considered as an example of show oval ditched enclosures of approxi­ this phenomenon. The pattern seen in mately 1.5ha outside the hillforts of Battles- Hampshire is also similar to that observed bury and Scratchbury. In the case of the so far on the and the former this is situated approximately 300m Marlborough Downs where, with the possi­ beyond the eastern entrance of the hillfort ble exceptions of Segsbury and Perborough on the low spur giving access to the monu­ Castle, evidence of potential settlements in ment. Although undated, the form of the very close proximity to the hillforts appears enclosure is typical of other later prehistoric to be lacking. examples in this part of the county. At There is clearly an urgent need for a Scratchbury, only 1.5km south-east of Bat­ carefully planned sampling strategy to tlesbury, another ditched enclosure occupies obtain more information on those settle­ a similar spur-end position some 200m ments and other features hard by hillfort beyond the north-east entrance of the hill- entrances. Such a strategy will need to fort. Aerial reconnaissance is playing an address some very fundamental questions important role in the identification of these starting with: ‘are these settlements and extramural enclosures and unenclosed set­ other features contemporary with the use of tlements. At Grovely Castle, another hillfort the adjacent hillfort? If so, is there any dis­ along the Wylye Valley (and in the same cernible difference in the character of the locational group as Yarnbury, Battlesbury material assemblage that may indicate a dif­ and Scratchbury), an enclosure of approxi­ ferent economic/social pattern to that of the mately 2ha has been located in close prox­ hillfort? Do these sites remain in occupation imity to the entrance. Farther west, along after the decline of the hillfort and if they do the Nadder Valley in the , is there any major change in their charac­ another enclosure of approximately 3ha has ter?’ It is tempting to postulate that in the been discovered some 500m from the west­ absence of any major concentrations of ern entrance of Castle Ditches, Tisbury. obvious ‘high status’ material from many

140 THE REGIONAL PATTERN excavated hillforts in central Wessex that acres) that continues into the early 5th cen­ such a focus, should it exist, is not within tury AD (Corney 1989). A similar complex the hillfort but immediately adjacent, on the might also have developed adjacent to a approach to the monument. In the areas nearby complex centred on Bilbury Ring where the pattern is concentrated, the recur­ hillfort and Hanging Langford Camp (ibid). ring location, generally within 200m and Two hillforts in the project area have 500m of an entrance, does strongly suggest remarkable structures within their circuits. a close relationship. Tidbury Ring, Bullington in Hampshire has two substantial Roman buildings, set at 90° Hillforts of Wessex after the to each other, placed centrally within the Iron Age enclosure. Known only from air pho­ tographs (for example NMRC SU 4642/6, The use of hillforts in Wessex in the Late 1948) this complex appears to be a small Iron Age and beyond is an aspect that villa complex with an aisled building and a has yet to be given the study it deserves. simple corridor house. Such a siting is The patterns that are discernible appear highly unusual and poses questions as to again to be both regional and chronological. why this particular location was chosen. A The Danebury excavations show that here substantial Roman building is also known there is very little major activity after c within the small enclosure of Alfred’s Cas­ 100–50 BC (Cunliffe 1984a) and no evi­ tle. This again appears to be a domestic dence of Roman military activity in the mid­ structure constructed in the 1st or 2nd cen­ 1st century AD. Unlike Dorset (Hod Hill, tury and demolished in the late 3rd century Maiden Castle), South Somerset (Ham AD (Gosden and Lock 1999, 2001, 2003, Hill, South Cadbury) and East Lock and Gosden 2000). To seek a possible (Hembury), none of the hillforts in the core parallel it is necessary to look into the area of Wessex have produced convincing Cotswold region to The Ditches at North evidence of Roman military intervention. Cerney, Gloucestershire. Excavation here Only at Forest Hill near Marlborough, has recovered details of a simple corridor probably part of a Late Iron Age regional house of 1st century AD date set within a centre (Corney 1997), and Bilbury Ring in plough-levelled enclosure of hillfort propor­ the Wylye Valley is there a possibility of a tions and dated to the 1st century BC (Trow short-lived Roman military presence. This 1988; Trow and James 1989). It is possible lack of evidence can be accepted and in that Tidbury Ring may be a further example probability reflects the very different politi­ of a Romano-British villa developing within cal and social attitudes in the region towards a hillfort but only fieldwork can answer this the Roman invasion in AD 43. question. The Roman building at Alfred’s Evidence of non-military activity within Castle was recently excavated by the Hill- hillforts throughout the Roman period in forts of the Ridgeway Project during central Wessex is, however, plentiful even if, 1998–2000 (Gosden and Lock 1999, 2001 in many cases, the exact nature of this is still and 2003; Lock and Gosden 2000) and a obscure. In some cases the activity is clearly detailed summary of the results is included domestic and the relationship to the hillfort in Chapter 2. Tidbury also has other fea­ may be little more than convenience in tures suggesting post-Iron Age activity. defining an area of settlement activity. This South of the hillfort air photographs show a is surely the case at Balksbury, a Late substantial linear ditch mirroring the south­ Bronze Age–earliest Iron Age enclosure near ern arc of the hillfort and presumably of Andover. Here an aisled building of later prehistoric date (Fig 3.3). Close examina­ Roman date appears to be the focus of a tion of the photographs shows a series of small farming settlement (Wainwright and cropmarks that may represent an inhuma­ Davies 1995). At Yarnbury in Wiltshire tion . These are clustered around a excavation (Cunnington 1933) and surface small ring ditch of approximately 5–7m collection (unpublished, National Monu­ diameter. There are two possible contexts ments Record [NMR] archives) suggests the for this apparent cemetery. It could be very presence of a large settlement spanning the late Iron Age and compared with Mill Hill, entire Roman period. At Stockton Earth­ Deal (Parfitt 1995) or, and perhaps more works, overlooking the Wylye Valley in Wilt­ plausibly, be an early pagan Anglo-Saxon shire, an early univallate enclosure develops cemetery. Tidbury Ring is a site that into a major nucleated Late Iron Age and requires a great deal of further investigation Romano-British settlement of 32ha (79 and it is to be very much regretted that

141 THE WESSEX HILLFORTS PROJECT

Leach 1996). Within our study area Romano-British religious activity has been postulated at a number of examples includ­ ing Uffington Castle (Lock and Gosden 1997a), Old Sarum (Corney 2001), Lid­ dington Castle (this volume) and Oldbury (pp 123–7). Others, such as Ashley’s Copse on the Wiltshire-Hampshire border, are also likely candidates. The phenomenon of post-Roman reoc­ cupation and refortification of hillforts is, like reuse as a religious focus, best known in western Britain. Here, hillforts such as Cad- bury Congresbury (Rahtz et al 1992), South Cadbury (Alcock 1995, Barrett et al 2000) and Ham Hill (Burrow 1981) have all pro­ duced good evidence of reoccupation. The nature and character of this activity is still far from understood but clearly involved long distance contacts with the Byzantine world as evidenced by ceramic imports. This focus on Somerset is probably more a reflection of the work of individual archaeol­ ogists such as Philip Rahtz, rather than a true geographical pattern. In Dorset there is good evidence for post-Roman activity at Maiden Castle in proximity to the Romano- British temple (Woodward 1992) and Hod Hill has produced some items of late Roman style metalwork, weapons and two 5th cen­ tury AD Germanic brooches, the latter com­ ing from the site of a Roman building just below the hillfort defences (Eagles and Mor- timer 1994). At Oldbury, Wiltshire, close to the west­ Fig 3.3 access for geophysical survey as part of the ern terminal of the East Wansdyke, a penan­ Aerial photograph of Wessex Hillfort Project was denied. nular brooch of probable 5th-century date is Tidbury Ring, Bullington, In the Vale of Wardour in south-west known with another example from nearby Hampshire showing linear Wiltshire both surface finds and geophysical Calne (Youngs 1995). In this context the earthwork south of the fort survey suggest an extensive Roman period proximity of the hillfort to a major Roman with possible adjacent settlement within Castle Ditches, Tisbury villa below Cherhill village church, only 1km inhumation cemetery (pp 103–7). None of the features located by to the north-west (Johnson and Walters (NMRC, SU 4642/19/16, the geophysical survey resembles a temple of 1988) and the possibility of an extramural 1976). Romano-Celtic form and the settlement Romano-British temple (p 127) makes the may be a largely secular one. The area is geophysical evidence for a possible reduction intriguing as it is one where there is good of the hillfort circuit especially interesting. survival of pre-English place names indicat­ The proximity of major Roman structures to ing possible continuity from the Roman to hillforts with evidence for post-Roman reoc­ post-Roman period (Eagles 1994). cupation is impressive and includes Cadbury In western Britain the most common Congresbury, South Cadbury, Ham Hill, occurrence of substantial Roman buildings Crickley Hill and Old Sarum. on or in close proximity to hillforts is usually In Hampshire, small-scale excavations by associated with a religious focus. There are Philip Rahtz recovered post-Roman ceram­ numerous examples ranging from ‘intra­ ics and evidence for refurbishment of the mural’ cases such as Maiden Castle defences at Castle Ditches, (Elli­ (Wheeler 1943), and Lydney (Wheeler and son and Rahtz 1987). This site is in some Wheeler 1932) to those in close proximity to ways comparable to Oldbury in that it is the hillfort such as Uley (Woodward and close to another probable 5th century AD Leach 1993) and Henley Wood (Watts and boundary, Bokerley Dyke (Bowen 1990).

142 THE REGIONAL PATTERN

Such hints do suggest that reoccupation The third category consists of sites of hillforts in southern Britain may be far with evidence for scatters of pit-type anom­ more widespread than hitherto thought, and alies such as St Catherine’s Hill (only sam­ to this author it would appear that it may be pled because of tree cover), Woolbury, related to the area once covered by the for­ Perborough Castle and Uffington Castle. mer late Roman province of Britannia In many cases distinct clustering of pits Prima. The region has a growing body of can be observed in specific areas of the evidence for very late Roman activity in hillfort – either around the perimeter of the both coins and other artefacts (ibid) and it is enclosure or at the centre, often on the here that we may expect to see evidence of a highest ground – but the overall quantity social evolution develop before the final and density of pits is low. assertion of Anglo-Saxon hegemony. In the fourth category are sites such as Bury Hill II and Barbury Castle that contain An overview of the very dense and even pit distributions. This geophysical survey results response is consistent with the stronger, more developed, multivallate earthworks by Andrew Payne defending these forts, usually indicative of continued and prolonged occupation into The results of the programme of geophysical the Middle Iron Age and beyond or re-occu­ surveys span a wide range and do not divide pation at a late period in the Iron Age. simply into clear groups. The classification The fifth category includes a range of of sites based on the geophysical results is to hillforts that all contain similar patterns of a degree a matter of personal interpretation occupation, although the density of the and a range of quite different classifications anomalous activity varies. It is quite clear are clearly possible based upon using a that all these sites functioned as settlements range of different attributes for grouping the or at least foci of activity at one time or sites. The system adopted below is based on another because they contain zones of pits similarities in the density, form and pattern associated with small numbers of round of magnetic anomalies within the hillforts structures defined by ring-gullies. This and the presence of recurrent features such group makes up about a third of all the sites as circular gully structures. surveyed and therefore seems to be the most At one end of the spectrum, there are a representative of hillforts in general in our number of hillforts that exhibit a low level of sample region. It includes Segsbury Camp, internal activity. These could be termed Beacon Hill Camp, Liddington Castle and ‘empty hillforts’. In the case of Ladle Hill Oldbury Castle. this is entirely compatible with the unfin­ Finally we are left with two very distinc­ ished status of the hillfort, suggested by the tive sites that exhibit rather more elaborate irregular form of the earthwork. In other patterns of internal layout suggesting an ele­ cases, such results could reflect early aban­ ment of settlement planning and division of donment of the site (as happened at many the internal area into functional zones for hillforts in the early Iron Age) or sporadic, different activities. One site is more coher­ perhaps seasonal, usage. The small hillfort ent as a single phased layout; the other is of Oliver’s Camp appears to represent more suggestive of two separate distinct another example of this type of site. phases of internal arrangements. A second category of sites that appear to The first site – Norsebury – contains lin­ show features in common are the group ear sub-divisions and there is a particular known as hill-top enclosures – vast enclo­ concentration of occupation features adja­ sures following the contours of a plateau cent to the ramparts along the western side area defined by relatively slight earthwork of the hillfort, while the central area appears defences and datable to the very beginning to have been reserved for a large circular of the Iron Age. The examples of these sites enclosure of unknown date and purpose but that were surveyed at Walbury and Martin- possibly a shrine. Complex entrance fea­ sell appear to contain mainly geological dis­ tures are indicated by the magnetometer in turbances or areas of quarrying with little the ploughed-out section of the hillfort evidence for a settlement function. Total defences now clearly defined by the survey. coverage of these sites was thought to be At Castle Ditches the site is occupied by unnecessary after this disappointing large numbers of circular structures defined response. The internal areas were neverthe­ by ring-gullies, with enclosures and roadways less extensively sampled. aligned on the four entrances into the fort.

143 THE WESSEX HILLFORTS PROJECT

The enclosures are clearly of a different another remarkably well, but there is less of phase to the ring-gullies which they appear a clear match at Barbury Castle where sub­ to intersect in several places. The round surface features are much more prolific. It is structures appear to be aligned in rows sug­ likely that the two forms of survey at Bar- gesting an element of planning in their lay­ bury are picking up separate phases of occu­ out. Pits appear to be less plentiful at Castle pation and therefore providing a more Ditches than the round structures and complete picture of the sequence of activity enclosure features. in the enclosure than would be gained by using the techniques in isolation. The sub­ Small hillforts surface features detected by the magnetom­ etry are most likely earlier than the features More work needs to be done on under­ visible as surface indications. The land-use standing the function of smaller hillforts as history of the site and variation in past land- the results from those included in the study use across the site again plays a part in the were uninformative (Oliver’s Camp) or visibility of both surface and sub-surface complicated by later occupation (Roman in features – one set of features often being the case of Alfred’s Castle and relatively detectable at the expense of the other. recent activity at Oliver’s Camp). One ques­ tion that is frequently asked of such sites is, Notable discoveries at specific sites ‘do they represent a different level of social organisation to the larger hillfort enclo­ Important information on specific aspects of sures?’ The ranking of such sites in a settle­ a number of sites has also been recorded. At ment hierarchy depends on them being Oldbury a previously unknown boundary permanent settlements. The evidence from ditch partitioning the hillfort (no longer Oliver’s Camp suggests it was never inten­ clearly visible on the ground) suggests two sively occupied, implying that there may be distinct phases of hillfort development, some functional distinction between some involving expansion or retraction of the small hillforts and larger hillforts. In con­ enclosed/defended area. This may reflect trast, the magnetometer data from Alfred’s several stages of fortification of the site dur­ Castle shows signs of considerable activity ing the Iron Age involving phased expansion within the enclosure indicated by a high of the hillfort across high ground, as is density of pits. Some of these have now been already known, for example, at Maiden Cas­ excavated producing a rich assemblage of tle and Torberry (West Sussex) (Sharples Early Iron Age material suggestive of a high 1991, Cunliffe 1976). Alternatively, it might status site (Gosden and Lock 1999, 2001, represent a second line of defence added as 2003; Lock and Gosden 2000). a later partition of the enclosed area to pro­ vide greater protection to the core area of Other aspects of the results settlement. Such a feature has been recog­ nised through excavation at Conderton The results from ploughed or previously Camp (Worcestershire), where a secondary cultivated sites (such as Norsebury Ring and rampart was inserted across an earlier hill- Castle Ditches) were generally much clearer fort enclosure, and the smaller area so than those from uncultivated sites under formed occupied by a settlement, leaving an permanent grassland. Surveying sites that outer annex that was unoccupied (Thomas have been ploughed for many years is there­ 2005). Cadbury Hill, Congresbury, in Som­ fore a clear advantage for magnetometer erset provides another example of later par­ survey despite the likelihood of loss or trun­ tition of a pre-existing hillfort (Rahtz et al cation of archaeological deposits from agri­ 1992). The internal ditch at Oldbury might cultural erosion. have functioned simply to out animals The grassland sites often preserve earth­ or to divide agricultural or other activities work evidence for archaeological features in from habitation areas. Yet another alterna­ their interiors that can more easily be inter­ tive explanation for the ditch is that it repre­ preted from analytical earthwork surveys of sents an earlier prehistoric linear boundary the type carried out by the former Royal running through the area later occupied by Commission on the Historical Monuments the hillfort, although it does not appear to of England (RCHME; now part of English line up with any of the known ‘linears’ in the Heritage). At Beacon Hill the earthwork evi­ area. Further magnetometer survey could dence (Eagles 1991) and the evidence from be used to determine if the ditch does con­ magnetometer survey tie in with one tinue outside the hillfort.

144 THE REGIONAL PATTERN

The magnetometer data from Lidding- after the new rampart build. Elsewhere in ton Castle raises interesting questions about the region now covered by Hampshire the the nature of the activity within this hillfort. hillforts at and Woolbury also The singular nature and impressive diame­ seem to have declined after the end of the ter of the large round structure revealed by 4th century in common with St Catherine’s the geophysics inside the fort is suggestive of Hill. Only Danebury continued as a major a specialised function, such as a shrine or centre in the region after this decline (Cun­ temple. The large oval enclosure set apart liffe 2000, Cunliffe 1995). from the rest of the activity in the hillfort of Zones of dense pitting and occasional Norsebury may represent a similar sacred small round/oval structures also occur at enclosure, shrine or temple site. An enclo­ Norsebury although these concentrate sure mapped by magnetometry within the towards the edges of the enclosure in areas defences of Maiden Castle in 1985 (Baalam bordering the ramparts rather than the cen­ et al 1991) may represent another example tral area which seems to have been reserved of this type of feature. for a large, circular ditched feature. The A group of unusual features revealed deeply in-turned, slanted entrance on the inside Oliver’s Camp are thought to relate to south side of Norsebury is a possible parallel relatively modern (possibly Second World with remodeled strengthened approaches War) activity. through hillfort defences dating to the Mid­ dle Iron Age at sites such as Torberry, Parallels with Danebury and other Danebury and St Catherine’s Hill (Cunliffe excavated Wessex hillforts 1991, 330–4). The eastern entrance at Norsebury also shows signs of elaboration in Based on the magnetometer survey evi­ the form of additional projecting outworks dence, the hillforts of St Catherine’s Hill similar in design to the south-east entrance (with a central zone of pits), Segsbury, Lid­ at Beacon Hill and the blocked west dington (containing discrete zones of pits entrance at Danebury. with round structures), Oldbury (a moder­ The majority of the other sites do not ately high pit density, but more evenly scat­ seem to compare well with Danebury in its tered, plus round structures) and at the earlier phases. Barbury could correspond lower end of the scale Beacon Hill (a thin with Danebury nearer the end of its occupa­ scatter of pits plus round structures) all tion history – along with Maiden Castle the show elements of the early Danebury layout product of cumulative phases and a long in the 6th–5th century BC. sequence of activity. The resemblance St Catherine’s Hill also shares other fea­ between the magnetic results from Barbury tures in common with Danebury in the early and Maiden Castle (which is well under­ period, such as entrance/gate structures stood from excavation) is quite striking. known as a result of excavation in the 1930s. Sites such as Perborough and Ladle Hill In Period 2 of the St Catherine’s Hill have most in common with emptier sites sequence there is evidence of major recon­ such as Bury Hill I and Woolbury (plus struction and heightening of the original Figsbury and Quarley) examined during the dump rampart in parallel with narrowing Danebury Environs Project (Cunliffe 2000) and lengthening of the entrance passageway. and the smaller promontory-type fort of At St Catherine’s Hill these modifications Oliver’s Camp would fit in here too. Martin- are linked to a major change of pottery style sell and Walbury probably belong in the ear­ to saucepan pots of the St Catherine’s Hill liest, sparsely occupied, class of hillfort in group. Similar developments took place at their region similar to Balksbury in the Danebury about 270 BC (Danebury Period Danebury Environs – but this is difficult to 4) when the original box rampart built in state conclusively because of geological the middle of the 6th century BC was complications – and there is a question mark replaced by a more substantial dump ram­ over whether archaeological features are part fronted by a large V-shaped ditch and really absent. Features of an ephemeral the entrance passage was also narrowed and nature such as post-holes may not be ade­ lengthened. The first hillfort entrance at St quately resolved by the fluxgate type magne­ Catherine’s Hill also closely resembles tometers and ‘standard’ recording intervals Danebury Gate 2a-b (a wide dual carriage­ employed by the project. way entrance closed by double gates) in the There are other anomalous hillforts early period of Danebury. St Catherine’s that do not easily fit in with our current Hill shows evidence of destruction not long understanding. These include Fosbury,

145 THE WESSEX HILLFORTS PROJECT

Alfred’s Castle and Castle Ditches. On The character of internal activity the basis of the paucity of evidence for revealed by the magnetometer surveys can activity inside it, Fosbury is similar to Wool- not always be correctly anticipated from the bury or Bury Hill I – but it has elaborate layout and sophistication of the hillfort defensive architecture and a suggestion of defences, showing the value of magnetome­ internal quarry ditches more in keeping with ter survey for rapidly revealing the character a developed hillfort. This may indicate of occupation within a hillfort. This in turn that the enclosure circuit was redefined at can shed light on the likely duration of occu­ regular intervals involving heightening of pation and the character and intensity of the ramparts but never actually brought into past activity on the site. The case of Norse- use as a fortified static community. Alfred’s bury is a good example where the design of Castle is complicated by Romano-British the earthworks at first sight would suggest a occupation but appears to be a densely relatively simple form of hillfort, belying the used, primarily early, small hillfort akin to complex internal activity now revealed by the previously excavated site of Lidbury the magnetometer. The reverse seems to be Camp in the east of Salisbury Plain that the case at Fosbury. produced eleven storage pits in a limited Many sites that are superficially similar area of excavation (Cunnington and in terms of size, siting and rampart con­ Cunnington 1917; Cunliffe 1991, 348). struction contain very different and some­ Castle Ditches stands out on its own as an times unusual or unexpected patterns of untypical hillfort in the sample of sites activity. Two sites that appear very similar included in the Wessex Hillforts Survey, but on the ground based on the size of the areas is peripheral to the main area sampled and enclosed and the form and layout of the possibly belongs to a geographically distinct enclosing earthworks are Perborough Castle group with more in common with hillforts and Norsebury Ring but they exhibit very in Dorset and Somerset. If Castle Ditches different patterns of occupation. Differential does belong in this group it might have preservation may also have some part to been occupied until a much later date than play in these results but this is difficult to the hillforts farther east in what became the quantify without excavation. territory of the in the Late The size of a site and the complexity of Iron Age. Although defensively a hillfort the defences visible on the surface are not, Castle Ditches has, in one phase, the therefore, necessarily related to function or internal characteristics of an oppida-type socio-economic complexity. The large settlement or ‘valley-fort’ such as Salmons­ enclosed area of Segsbury (12ha) contains a bury in Gloucestershire or Dyke Hills in similar pattern of occupation to that Oxfordshire. Non-hillfort Iron Age settle­ observed inside the 3ha enclosed at Lid­ ment in the Danebury Environs shows con­ dington Castle. Norsebury, enclosed by a siderable variety to the extent that it is simple bank ditch and counterscarp, dis­ difficult to discern any regular pattern. plays a density of internal occupation on a There is no reason why this variety should par with larger sites with massive multival­ not extend to hillfort settlement. late defences such as Yarnbury, Oldbury, Bury Hill II and Castle Ditches. Univallate Some overall conclusions sites can contain a similar density and com­ plexity of internal activity as multivallate The project has revealed a wealth of new sites, but multivallate sites generally, but evidence for the nature of the internal with the notable exception of Fosbury, con­ utilisation of Wessex hillforts. While sup­ tain dense internal activity. porting some of the existing models of The overall impression given by the hillfort development, the surveys also show results is that far from all hillforts were that the pattern is considerably more inhabited or functioned primarily as settle­ complex and varied than previously realised ments. Although in some cases forts may (see Chapter 4). Some hillforts exhibit a very have been constructed to house settlements low density of archaeological features, while perhaps when the need arose or for socio­ others contain evidence for prolonged and political reasons, in many cases the sites may intensive usage (indicated by a very high not have been inhabited for very long density of magnetic anomalies mapped). or served other purposes, leaving few In some cases several discrete phases of detectable traces in the archaeological settlement activity are suggested by the record. Fosbury is one possible example of magnetic results. this. Some hillforts were obviously centres

146 THE REGIONAL PATTERN

Fig 3.4 Aerial photograph of Conderton (or Dane’s) Camp, Worcestershire, a small 1.5 hectare hillfort located on a narrow ridge between two dry valleys on the side of the upland massif of Hill (NMRC; NMR 18035/11, SO 9738/18, 1998, Crown Copyright).

of large permanent settled communities (as Some sites contain large numbers of illustrated by the houses, streets and enclo­ pits apparently with few house sites sures mapped at Castle Ditches). Others (Liddington, Barbury and Segsbury). At were probably only temporarily or sporadi­ others, house sites are fairly plentiful, but cally occupied while some may have had have few pits (Beacon Hill, Castle Ditches). more specialised functions possibly as reli­ Sites such as Perborough Castle appear gious or ceremonial centres or seasonal to have only ever been sparsely occupied Fig 3.5 (page 148) gathering places. The overall results of the leaving evidence only of limited scatters Magnetometer and survey allow for a considerable range of of pits. Barbury Castle appears to have earthwork surveys of functional variability between hillforts. been the most intensively used or longest Conderton Camp showing The internal planning and layout of occupied of all the sites surveyed. The the bi-lateral division of the structures in hillfort interiors is highly sheer profusion of anomalies at this site hillfort into storage and varied. Some sites appear more organised suggests numbers of pits running into occupation areas represented than others. At some sites the pattern of the thousands. by distinct zones of densely features appears to be quite random and A significant sample of the hillforts packed pits and round disorganised although nearly all sites display in central- has now been structures. The remains of some clustering of activity. In other cases surveyed, considerably broadening our a field system survive as a there is more evidence of zoned activities. knowledge and understanding of the sites. series of lynchets to the east One example is Norsebury, where there are The more detailed information that has of the fort and were partially zones containing a very high concentration emerged from the project is already begin­ subsumed by it (Mark of archaeological features in two discrete ning to show the diversity of patterns Corney and Andrew areas of the hillfort including pits, quarries of activity within Iron Age hillforts. The Payne). and circular structures, while the remaining evidence suggests that hillforts were con­ third of the hillfort appears much emptier. structed for a number of purposes and that Segsbury and St Catherine’s Hill both these purposes will have changed over time. Fig 3.6 (page 149) have concentrations of occupation near the The results prove that it is not possible The results of the magne­ centre of the site, on the highest ground, easily to predict the character of hillforts tometer survey carried out dominating the whole of the enclosed from surface evidence alone and therefore inside the hillfort of Castle area. At Oldbury occupation is concentrated there is clearly justification for the contin­ Hill, Little Wittenham, in the northernmost third of the hillfort ued and expanded use of geophysical meth­ Oxfordshire. The newly on a steep natural promontory separated ods for hillfort investigation. Preliminary identified inner enclosure at some time by a cross boundary ditch results of magnetic survey from a limited circuit revealed by the from the remainder of the area enclosed number of hillforts in the neighbouring survey is clearly visible in by the hillfort. Severn-Cotswold Region (Figs 3.4 and 3.5; the plot.

147 150m VEY OMETER SUR 0 TE MAGNET N FLUXGA VEY THWORK SUR N EAR

148 All rights 572 fice, (c) Crown copyright. , June 2002 571 The Controller of Her Majesty's Stationary Of This map is based upon Ordnance Survey material with the permission of on behalf of reserved. Unauthorised reproduction infringes Crown Copyright and may lead to prosecution or civil proceedings. Licence Number: 100019088, English Heritage, 2002. Magnetometer Survey 570 569 , LITTLE WITTENHAM, OXFORDSHIRE 568 CAMP 90m 567 OR SINODUN HILL N CASTLE HILL 149 0 925 924 923 922 THE WESSEX HILLFORTS PROJECT

Conderton Camp, Worcestershire) suggests practical value of thematic geophysical that here there is considerable future survey aimed at a single type of archaeologi­ promise for further expansion of our knowl­ cal site, for which there is a recognised range edge of hillforts. Recent survey within the of management challenges and a clear hillfort at Castle Hill, Little Wittenham, research agenda. Oxfordshire – a site overlooking the Thames To end with a final note of caution – we Valley – has continued to demonstrate need to ask the question: does magnetome­ the potential, revealing the presence of a try really represent the reality beneath the previously unknown inner enclosure circuit ground in a hillfort? The answer, it has to be provisionally dated to the late Bronze Age acknowledged, is probably ‘no – not totally’ (Fig 3.6; Payne 2002). The work of the based on the retrospective survey of Wessex Hillforts project has also demon­ Danebury, but if the results are interpreted strated the complementary academic and with care they can still tell us much.

150