By M A Cole and N T Linford
Ancient Monuments Laboratory
23 Savile Row
London W1X 1AB.
The results of detailed geophysical survey at Woodbury Farm, Axminster indicate the presence of buried foundations of buildings and walls within the Roman fort as well as evidence of an associated settlement. This information, allied to that from previous excavation and aerial photographs, permits a fuller interpretation of the archaeological significance of this Roman site.
The importance of Woodbury Farm, Axminster as a former Roman military and later Roman site has been recognised since a limited excavation was undertaken there in 1984, prior to the construction of a swimming pool by the current land owner. The subsequent discovery of an associated Roman settlement during an evaluation by the Exeter Museums Archaeological Field Unit (EMAFU), preceding the insertion of a South West Water Plc (SWW) water main, prompted the re-examination of the area surrounding this scheduled ancient monument (Devon 1031). Furthermore, concern about the extent of teh area protected by scheduling led to a request from Devon County Council's Archaeology service for English Heritage to conduct a geophysical survey to assist with the archaeological evaluation of this area and thus aid the accurate protection of archaeological remains in the locality. This paper reports the results of two weeks of field work conducted by the Archaeometry Branch, Ancient Monuments Laboratory, English Heritage, in October 1993 and March 1994. During the preliminary phase an extensive fluxgate gradiometer survey was conducted to the W of the scheduled monument and the success of this work instigated the extension of the coverage over the monument and into the field immediately S of Woodbury Farm. In addition, an earth resistance survey was conducted in the interior of the monument and values of topsoil magnetic susceptibility were recorded over the areas of magnetometer coverage.
Aims and Objectives
The aim of this survey was to investigate the ability of geophysical techniques to identify archaeological activity on both the scheduled monument and within its immediate environs. It was envisaged from previous excavation data (Silvester and Bidwell 1984 and Simpson1993) that a range of archaeological features, from buried ditches and rubbish pits to the stone wall footings of former buildings would be encountered and the techniques applied would have to be capable of detecting a distinctive geophysical anomaly from such features. The specific aims of the survey were therefore deemed to be:
This paper provides a brief introduction to the site and previous investigations, a summary of the techniques applied, a detailed report of the geophysical results and a discussion of their archaeological significance. It is further hoped that this study will provide an informative adjunct to the companion article presented by Peter Weddell within this same volume.
The area scheduled as a monument in 1988 takes the form of an eroded rectangular earthwork enclosing an area of approximately 2ha with modern farm buildings encroaching into the NE corner (Fig 1 - location plan). Excavation evidence prior to the construction of a swimming pool adjacent to the modern farm house within the enclosure (Silvester and Bidwell 1984) and the watching brief during the insertion of the SWW water main (Weddell 1991 and Simpson 1993) revealed the presence of a first century Roman fort, a well preserved Roman road and suggested extensive extra mural activity. The location of the site at the apparent confluence of two Roman roads (the Fosse Way and the Dorchester to Exeter Road - Silvester and Bidwell 1984 - Fig 1, Margary 1973) has led to the suggestion that the Roman site at Woodbury may well have developed into a villa or posting station (mansio) and indeed it is this latter interpretation that is adopted by Silvester and Bidwell (1984). The discovery of a variety of Roman features during the insertion of both the SWW water main (Simpson 1993) and the 1992 slurry disposal pipe reinforced this interpretation and provided additional evidence to suggest the location of a more substantial civilian settlement (vicus) immediately W of the scheduled monument, possibly the site of the lost Roman town of "Moridunum" (Weddell 1991, Griffith pers comm).
The site (NGR SY 297 974) lies over white Lias limestones, shales and marls of the Rhaetic beds (British Geological Survey Sheet 326 - Sidmouth). However, the drift geology mapping does not appear to have taken account of the clay-with-flints cover soils to be found at the site (Canti 1993). The soils at the site are developed from both the clay-with-flints and from additional, typically lighter substrates, perhaps including occupational material from the fort.
The first phase of the geophysical work consisted of a fluxgate gradiometer survey of the land parcel immediately W of the scheduled monument and a 60m x 120m control area within the fort itself (Linford 1993). A topsoil magnetic susceptibility survey was also conducted in conjunction with the magnetometer survey and was extended to include the entire area of the fort. The speed of ground coverage of magnetometer survey combined with the ability of the technique, under suitable conditions, to detect a wide range of buried archaeological features dictated its use as the most effective initial survey method. Examples of the ability of this technique to reveal anomalies associated with both Roman military and civilian occupation are plentiful (eg Clatterford Roman settlement, Isle of Wight - Payne 1993, the Roman Fort and extra mural activity at Burgh-by-Sands, Cumbria - Linford 1992 and similarly at Lake Farm, Dorset - Clark 1990) and it was envisaged that despite interference from ferrous pipelines a variety of archaeological features would be revealed. However, site conditions, particularly the local geology and geomorphology can severely influence the success of magnetic surveys (Tite and Mullins 1971).
Fortunately, the results from the initial magnetometer survey were extremely encouraging and it was decided to complete the coverage of the fort and to further examine its environs to the S and E. Regrettably access to the fields immediately E of the fort was denied by the land owner and thus the significance of the Roman pottery and tile scatter reported by the Chard History Group (Silvester and Bidwell 1984) could not be investigated. However, aerial photographs of parchmarks over the site (Silvester and Griffith 1984 - Plate 1) strongly suggested the presence of high resistance, possibly stone building the remains of which in view of the universal use of timber structures in the other known first century forts of the area, are best interpreted as evidence of a later phase of use. Whilst magnetometer surveys can detect the presence of stone buildings as negative anomalies, the earth resistance technique - although considerably slower - provides the most suitable technique for the location of such remains (cf Payne 1993). Thus a detailed earth resistance survey was conducted within the fort interior with the aim of further examining the function and possible later reuse of the site as a villa or mansio.
A survey grid divided into 30m squares was first established over the site ( Fig 1 ) utilising a series of non-magnetic marker canes to facilitate the detailed recording of both the gradiometer and topsoil susceptibility surveys. The same grid was reinstated during the second phase of field work and an independent grid was established in the parcel of land immediately S of the scheduled monument.
These areas were then surveyed with Geoscan FM36 fluxgate gradiometers along successive N-S traverses separated by 1m intervals. Readings of the local gradient of the Earth'smagnetic field were logged every 0.25m and the data was down-loaded to a microcomputer in the field.
The presence of subsurface features with contrasting magnetic properties to the surrounding subsoil will result in a local perturbation in both the total field intensity and the gradient of the Earth's magnetic field. These anomalies can result from the silting of cut features with a magnetically enhanced sediment, the enhancement of organic debris in-situ or by the accumulation of material containing a thermoremanent magnetic field (eg burnt clay). LeBorgne (1955, 1960) demonstrated that under normal pedogenic conditions a successive climatic cycle of chemical reduction and subsequent oxidisation alters the mineral state and magnetic properties of iron oxides within the soil. Many other authors have also confirmed that anthropogenic activity, particularly intense burning or industrial processes, can further accelerate the magnetic enhancement of archaeological sediments (Tite and Mullins 1971) and that the fermentation of organic matter (eg rubbish pits, wooden structural remains) will also produce features with contrasting magnetic properties (Fabinder 1994, Linford 1994). Thermoremanent features occur when an accumulation of an iron rich mineral is heated beyond its Curie point and the individual magnetic domains are then able to align themselves with the prevailing magnetic field at that time. As the temperature is reduced the magnetic domains become less mobile and are "frozen" into a common magnetic alignment producing a substantial permanent magnetic field. Archaeological features such as pottery kilns, ovens or hearths often produce distinctive magnetic anomalies through a combination of this process and the rich concentration of iron minerals found in the clay used in their construction.
Topsoil Magnetic Susceptibility
Topsoil magnetic susceptibility (MS) measurements were taken at a 15m sample interval using a Bartington MS2 meter and field search loop. In this study, the topsoil MS study was intended to act as a reference for the gradiometer results by evaluating the concentration of iron minerals encountered in the modern plough soil over the site. The combined effect of bioturbation and ploughing has been shown to produce detectable anomalies in topsoil MS due to the accumulation of anthropogenically enhanced sediments from buried archaeological features in the modern plough soil (Cole et al 1994). However, the efficacy of the technique for locating areas of archaeological activity is often hampered by current land-use (particularly differing agricultural practice between adjoining land parcels) and the need to cover a suitably large area to detect what may, in many cases, be a rather dispersed distribution of high magnetic susceptibility readings.
Earth Resistance survey
A Geoscan RM15 resistivity meter was used to collect earth resistance measurements in the interior of the fort ( Fig 1 - shaded squares) using the twin electrode configuration with a mobile probe separation of 0.5m (Clark 1990). Readings were logged at 1m intervals along successive N-S traverses separated by 1m and the data was down-loaded to a microcomputer in the field. The twin electrode array is particularly well suited to archaeological targets and measures the earth resistance of the volume of ground immediately below the mobile current-potential probes with the addition of a constant, and thus negligible contribution from the remote current-potential electrodes.
Electrical conduction in the soil is proportional to the concentration of dissolved ions it contains which is itself largely proportional to the overall water content present at the time of the survey. Local variations in the subsurface resistivity are thus determined by contrasts in moisture retention between buried archaeological features and the surrounding soil. For example, non-porous stone wall footings will retain less moisture than the surrounding soil and will result in a high resistance anomaly; conversely a ditch silted with a combination of topsoil sediment and organic humus will produce a water-retaining low resistance anomaly.
Data Processing and Display
The individual survey grids of data were initially combined into two separate composite mosaics of magnetic and earth resistance data; the topsoil magnetic susceptibility data was recorded in the field as a single data grid. The unprocessed ("raw") data from the three survey techniques is presented in Figures 2a [92Kb JPEG], 3a [62Kb GIF], and 4 in the form of greyscale images. This format prescribes a shade of grey between black and white to each data point which in turn forms an individual picture element of the final image. The densely sampled magnetometer and resistivity data has been interpolated between individual data points to simulate a continuously sampled survey. However, this treatment is not applicable to the coarse sample interval used to collect the topsoil susceptibility data and this data is therefore depicted as a series of shaded squares superimposed upon the Ordnance Survey map of the site. All plots follow the convention of positive values (ie increased magnetic gradient, magnetic susceptibility or resistance) being displayed as lighter shades and decreased values as denser shades of grey towards black. Each plot is presented with a histogram greyscale key depicting the distribution of data values within the data-set and the mapping of these to the appropriate shades in the greyscale range.
The numerals within the following text refer to the identification of geophysical anomalies in the two graphical summary guides where the prefix "m" denotes magnetometer ( Fig 2b [20Kb GIF] ) anomalies and the prefix "r" resistance anomalies ( Fig 3b [17Kb GIF] ). These numerals have been used to refer to both individual anomalies and to groups of similar anomalies. The guides are, by their nature, somewhat subjective and aim to elucidate the more significant anomalies discussed below.
A number of anomalies detected, both by the magnetometer and resistance surveys, can be excluded from the archaeological discussion by their identification as modern interference. This is particularly true of a number of anomalies occurring within the magnetic data which, due to the gradiometers' high sensitivity, is extremely susceptible to concentrations of ferrous metal. The most notable examples are the course of the SWW water main (m1), the Gas main (m2) and a third unidentified ferrous pipeline running from Woodbury Cottage to the modern farm buildings (m3) believed to be a foul drain. It is of interest to note the distinctive contrasting positive/negative response of anomaly m3 associated with the thermoremanent magnetic field present in the individual sections of pipe and the continuation of this anomaly into the N edge of the earthwork enclosure. The trench dug for this pipe appears in theresistivity data as a low resistance anomaly (r1). Unfortunately the intense response from these anomalies has severely marred the magnetometer results from the majority of this area by masking all but the most intense anomalies. Further magnetic interference is associated with the modern farm buildings in the N of the earthwork enclosure (m4) and from the electric fence posts (m5) to the W of the survey.
Both the resistivity and magnetometer data sets contain linear anomalies on a E-W alignment within the earthwork enclosure associated with the ridges (positive magnetometer anomaly) and furrows (negative magnetometer and low resistance anomaly) of a former orchard (m6 and r2) and indeed this pattern is also replicated in the aerial photograph evidence (Silvester and Griffith 1984 - photo DAP/AY7). bad switch yylook 3 Of further interest is the E-W field division within the earthwork recorded by the 1984 aerial photograph and the second edition Ordnance Survey map of the area. This appears to be paralleled by a trackway leading from the farm buildings to the gateway through the enclosure to the W. The relationship between this trackway and the high resistance anomaly (r3 - see below) is potentially significant as the local compaction of soil caused by the frequent use of the track could well have produced this high resistance anomaly. Further evidence for the course of the former field boundaries recorded in the OS second edition map are the ditch-type anomalies (m7) and the linear concentration of ferrous detritus (m8).
The course of a World War II tank trap is visible in the NW of the magnetometer survey (m9) appearing as a curved anomaly running between Wyke Lane and Woodbury Lane. Simpson (1991) reports that the tank trap originally consisted of two parallel ditches that were infilled shortly after the January 1948 RAF aerial photograph. The magnetic anomaly appears to be related to this infilling and is seen to contain a quantity of ferrous detritus.
The significance of the three broad positive anomalies (m10, m11, m12) is also questionable and the possibility remains that they may not be of archaeological importance. However, m12 coincides with the original orchard boundary (Weddell pers comm) and it seems feasible that this anomaly, although rather broad and somewhat intense, may be related to this more recent phase of agricultural activity on the site.
Roman road system
Much of the evidence for the interpretation of the earthwork as representing the remains of a Roman fort is based upon the Roman road system surrounding the site (Silvester and Bidwell 1984; Margary 1973) supported by the discovery of substantially built sections of Roman road in the 1991 SWW pipeline route (Simpson 1993). As a geophysical target, a paved Roman road should provide an easily distinguishable anomaly and both resistivity and magnetometer surveys should detect either the absence of moisture over the paving itself (eg Bartlett 1976, Cottrell and Payne 1993) or the presence of parallel drainage ditches to either side (eg David 1986). In this case only tentative evidence for the presence of metalled roadways exists within the earthwork and consists of the branched high resistance anomaly (r3) suggesting the junction of two roads or trackways from the N and the E. However, the occurrence of a modern farm track in the 1984 aerial photograph (see above) could also lead to the production of a high resistance anomaly caused by the compaction of topsoil due to frequent use. This would also explain the limited correlation with the magnetometer datawhich is restricted to the two negative linear anomalies (m13) which do not extend along the entire course of the resistivity anomaly, or beyond the confines of the earthwork enclosure.
Any interpretation of r3 as a Roman road is based largely upon the intensity of the anomaly being consistent with the presence of a metalled surface and the absence of a breach in the field boundary in either the current hedgerow or that depicted by the aerial photograph. The failure of r3 to be reproduced as a magnetic anomaly is perhaps explained by the poor contrast in magnetic susceptibility between the Lias blocks of the paving of the roadway (Silvester and Bidwell 1984 and Simpson 1993) and the surrounding subsoil developed over the same geology. In addition, the alignment of the buildings within the earthwork (see below) appears to respect r3 and only the modern services pipeline (m3) is seen to intersect it.
The three ditch anomalies (m14) immediately S of the SWW pipeline appear to be orientated parallel to the course of the Exeter-Dorchester Roman Road and may represent a series of Roman road gullies. However, results from the pipeline watching brief (Simpson 1993 - Fig 5) strongly suggest that the course of Exeter-Dorchester road ran to the N of the 1990 water main.
The earthwork enclosure
The course of the earthwork ditch is apparent within the enclosure as a low resistance anomaly (r4) which is particularly evident in the S and W of the survey and is complemented by a discontinuous high resistance anomaly (r5) which appears at the extreme S and E edges of the survey. The inner scarp of the S edge also exhibits a parallel high/low resistance linear anomaly (r6) that continues as a single linear low resistance anomaly along the E edge of the enclosure. These anomalies appear to be, in part, concurrent with the negative magnetic anomaly (m15) which turns through a right-angle following the SE corner of the enclosure. These anomalies, whilst defining the course of the fort ditch where it does not form part of the modern field boundary, also suggest the presence of an internal high resistance feature, possibly a road or track-way, within the perimeter defences.
Anomalies r7, r8 and r9 are distinguished by a similar linear low resistance form, indicative of a pair of parallel buried ditches. Whilst these may suggest the gullies of a former road or trackway it should be borne in mind that r7, r8 and possibly r10, show a close alignment with the boundary of the recent orchard (OS second edition, Weddell pers comm).
Both the magnetometer and resistivity surveys provide good evidence for the presence of internal buildings (m16, r11 and r12) within the earthwork enclosure amplifying the parchmark evidence reported by Silvester and Bidwell (1984). The strongest evidence for stone building foundations is revealed by the resistivity data which demonstrates the presence of an approx. 30m range of separate rooms orientated approximately N-S in the centre of the Roman fort. Additional rectilinear high resistance anomalies are seen to project at right angles from the ends of this structure and the entire complex is apparently enclosed by an approx. 50m square perimeter wall (m17, r13), extending along the S and W extremities.However, this pattern does not entirely respect the orientation of the fort disclosed by the extant ditches and internal perimeter features revealed by the geophysical survey.
To the N of this complex the magnetometer survey has revealed a series of negative rectilinear anomalies which appear to form a complex of quite substantial (approx. 15m x 10m) buildings (m16). These same anomalies are partially replicated as high resistance anomalies (r12) suggesting that they represent the response to non-magnetic wall footings within a more magnetic topsoil, thus producing negative "voids" within the magnetometer data. It is perhaps significant that the former range of buildings and the perimeter wall, which appear as strong high resistance anomalies, and parchmarks fail to produce a consistent response within the magnetometer data.
Further anomalies appear within the fort including a rectilinear low resistance anomaly (r14) in association with an adjacent positive magnetometer anomaly (m18). Whilst these anomalies fail to correlate with each other they tentatively provide evidence for either the location of possible timber buildings or the remains of stone-robbed foundation trenches.
Additional evidence of internal occupation features is provided by the plethora of discrete single anomalies distributed over the whole of the site (m19). Whilst many of these may be pits, a number demonstrate an intensity in excess of 50nT suggesting they represent strongly magnetised thermoremanent features such as kilns or hearths as opposed to other pit-type anomalies.
A most peculiar circular, weak low resistance anomaly (r15) exists in the SW corner of the fort. The archaeological significance of this feature is uncertain; it may well be of fairly modern origin, perhaps associated with agricultural activity at the site (the trampling of the ground by cows surrounding a cattle feeder, for example).
Despite the interference from the concentration of ferrous pipes in the NW of the survey area the magnetic survey has unequivocally corroborated the excavation evidence to support the presence of extensive extra-mural activity. The survey has revealed a number of linear ditch-type anomalies (m20) to the W of the fort and several groups of linear negative anomalies, possibly indicative of stone wall footings (m21 - m24). Unfortunately the results of the survey in the vicinity of the earlier excavation (Simpson 1993) have been severely hampered by the response to the ferrous water main. However, the series of linear magnetic anomalies confirms the suggestion of a number of enclosure or field boundaries extending W of the fort for approximately 200m to the boundary with Wyke Lane. A number of pit-type occupation anomalies (m19) are also evident within this area and one of the stronger of these (m25) was augered during the initial magnetometer survey for magnetic analysis of the recovered soil samples (Linford 1993). The latter demonstrated an enhanced low field magnetic susceptibility increasing with depth and the inclusion of bright orange burnt material and charcoal from 40-120cm confirming an archaeological origin.
Activity immediately S of the fort is apparently more restricted, the most outstanding anomaly probably being of geological/geomorphological origin (m26), and evidence of occupation features is limited to a scatter of three pit-type and two partial linear anomalies.
Topsoil Magnetic Susceptibility
The results from the topsoil magnetic susceptibility survey (Fig 4) provide a useful reference for magnetometer results discussed above. The low values recorded by the MS survey partially reflect attenuation caused by the pasture vegetation covering the site, but the data still broadly reflects the distribution of archaeological activity as revealed by the magnetometer survey. It is of particular interest to note the concentration of enhanced readings within the SE quadrant of the fort itself and a similar concentration immediately W of the monument. Both these areas correspond with concentrations of occupation-type anomalies identified within the magnetic and earth resistance surveys. This suggests that the consistently low magnetic susceptibility recorded N of the SWW 1990 pipeline may genuinely represent a fall-off in the concentration of occupation activity there. However, this interpretation should be tested by either trial excavation or the extension of the earth resistance survey into the areas of the magnetometer survey affected by the ferrous pipelines.
Discussion and Conclusion
The geophysical survey supports the interpretation of the earthwork monument as a Roman fort. In particular it provides evidence of anomalies possibly indicative of an earlier circuit of the fort defences. These findings go some way, perhaps, to explaining the disagreement that arises between the interpretation made by Silvester and Bidwell (1984) and that of Simpson (1993) in the positioning of the N limit of the defences. Of particular importance, however, is the revelation of a substantial number of stone buildings within the monument amplifying the evidence from the 1984 aerial photograph. The high resistance anomalies indicative of stone building foundations are at odds with the timber constructions expected of a first century fort in this locale (Griffith pers comm) and instead it may be suggested that both the dimensions and morphology of the anomalies may indicate the presence of a later villa complex.
Evidence for the location of Roman roads is restricted to the tentative interpretation of the high resistance linear anomalies within the fort itself. Sections of the two Roman roads encountered by the SWW 1990 pipeline (Simpson 1993 - Fig 5) suggest that extension of the resistivity survey to specifically target these anomalies should prove successful and the dimensions of these features correlate well with the geophysical anomalies within the fort. It would be of particular interest to determine the relationship between the road section observed in the SWW 1990 pipeline and the possible continuation of the road anomaly both within the fort and into the adjacent farmland to the E. Future survey should encompass this area to amplify the observations of Roman tile and pottery scatter made by the Chard History Group (Silvester and Bidwell 1984) as well as the trackway recorded during the 1984 excavation.
Once again the geophysical data allows for the expansion of the evidence proposed by Simpson (1993) for the presence of a substantial Roman settlement W of the fort. Whilst the identification of specific civilian activity is hampered by modern interference the distribution of pit-type anomalies in this area and the confirmation of magnetic enhancement to the sediments recovered from anomaly m25 confirms the archaeological sensitivity of the area W of the currently scheduled ancient monument. The survey S of the fort has failed to reveal a similar intensity of archaeologically significant activity, although this interpretation can only be tested against the trial excavation of anomalies identified within this area.
Bartlett, A.D.H. (1976). Report on Geophysical Survey at Corbridge, Northumberland, April 1976. Ancient Monuments Laboratory Report Series No. 2025.
Canti, M.G. (1993). A brief note on the soils and geology found at the Axminster pipeline (June 1990). In Simpson, S.J. (1993). Excavations on the Roman Fort and Settlement Site in Woodbury Great Close, Axminster 1990. Exeter Museums Archaeological Field Unit Report 93.11 .
Clark, A.J. (1990). Seeing Beneath the Soil. London: Batsford.
Cole, M.A., Linford, N.T., Payne, A.W., and Linford, P.K. (1994). Soil magnetic susceptibility measurements and their application to archaeological site investigation. In Beavis, J. (editor), Science and Site: Archaeological Sciences Conference 1993 (London: Archetype Books): (in press).
Cottrell, P. and Payne, A.W. (1993). Kingscote, Gloucestershire: Interim Report on Geophysical Survey, September 1993. Ancient Monuments Laboratory Report Series 103/93.
David, A.E.U. (1986). Report on the Magnetometer survey at Wharram Grange. In Rahtz, P. and Hayfield, C. Two Roman Villas at Wharram Le Street. York University Archaeological Publications 2: 24.
Fabinder, J.W.E. (1994) Die magnetischen Eigenschaften und die Genese Ferrimagnetischer Minerale in Bden. Phd Thesis, Verlag Marie L. Leidorf, Buch am Erlbach.
Le Borgne, E. (1955). Susceptibilit magntique anormale du sol superficiel. Annales de Gophysique 11: 399-419.
Le Borgne, E. (1960). Influence du feu sur les proprits magntiques du sol et du granite. Annales de Gophysique 16: 149-195.
Linford, N.T. (1992). Burgh-by-Sands, Cumbria, Report on Geophysical Survey Ancient Monuments Laboratory Report Series 88/92.
Linford, N.T. (1993). Woodbury Farm, Axminster, Devon. interim report on geophysical survey 1993. Ancient Monuments Laboratory Report Series 88/93.
Linford, N.T. (1994). Mineral Magnetic Profiling of Archaeological Sediments. Archaeological Prospection vol 1 (in press).
Margary, I.D. (1973). Roman Roads in Britain. John Baker, London.
Payne, A.W. (1993). Clatterford Roman Site, near Carisbrooke, Isle of Wight Report on Geophysical Survey, July 1993. Ancient Monuments Laboratory Report Series87/93.
Silvester, R.J. and Bidwell, P.T. (1984). A Roman Site at Woodbury, Axminster. Proceedings of the Devon Archaeological Society 42 .
Simpson, S.J. (1993). Excavations on the Roman Fort and Settlement Site in Woodbury Great Close, Axminster 1990. Exeter Museums Archaeological Field Unit Report 93.11 .
Tite, M.S. and Mullins, C.E. (1971). Enhancement of the magnetic susceptibility of soils on archaeological sites. Archaeometry 13: 209-219.
Weddell, P.J. (1991). Archaeological Appraisal of Potential Development Areas in Axminster and its Environs. Exeter Museums Archaeological Field Unit Report 91.14 .
We would like to thank Mr and Mrs P. J. Powell of Woodbury Farm for allowing us access to the site and for their interest in the project. We would also like to thank Peter Cottrell (Ancient Monuments Laboratory) and Tom Williams (Bradford University) for their assistance in the field and Andrew David for his comments on the draft text of this article. Thanks are also due to Frances Griffith and Peter Weddell for their most informative comments.