Preliminary report
In September 2008, the Roman Gask Project conducted a large scale geophysical survey of the northernmost of the Roman Gask frontier forts: Bertha, which lies on a promontory above the confluence of the rivers Tay and Almond, just north of modern Perth (NO 097 268). Both resistance and magnetic surveys were undertaken. The former employed two Geoscan RM15 resistance meters and used 1m traverses, with readings taken at 1m intervals. The latter used a Bartington Grad601 single fluxgate gradiometer, with traverses again at 1m, and readings at 0.25m intervals.
Little previous archaeological work had taken place on the site and, although it has been under aerial surveillance since at least 1941, it is a disappointing cropmark site. Some trenching was undertaken towards the fort's NW corner in 1973 (Adamson 1979 + Adamson & Gallagher 1986) prior to the addition of a second carriageway to the modern A9 (fig 1) and revealed the rampart, intervallum road and a single defensive ditch. The Gask Project has analysed a small collection of pottery picked up in the fort interior during fieldwalking by the Cumbernauld Historical Society (Woolliscroft 2002, 40ff) and these operations, along with a scattering of earlier finds (Hartley 1972, 5), were enough to confirm that the fort was held during both the Flavian and Antonine occupations. Otherwise, however, remarkably little was known about the site, except for its basic outline, and the survey seemed likely to reveal a great deal of new information without threatening surviving deposits.
Resistivity
Conditions on the site were far from ideal for resistance work: so much so that it stands as a tribute to the often much underrated robustness of the technique that this part of the survey produced data at all. The ground is not free draining and was saturated after a very wet summer, to the point that standing surface water was present in places. There was frequent further rain during the survey. The field was covered with lines of wet unbaled straw, and water managed to get into one of the two meters used, along with its PA5 electrode frame, resulting in such noisy data that part of the fort had to be re-surveyed. Nevertheless, a reasonable image was obtained after the meters were set to their slowest data acquisition speed for greater accuracy. This, however, along with the re-surveying mentioned above and the constant need to move heavy, water saturated straw to allow clean electrode contact with the ground, slowed proceedings to the extent that, although it was possible to cover the entire surviving area of the fort in the time available, we were not able to survey an area to its north that had been planned.
Despite these difficulties, the 
survey was still able to show a clear outline of most of the fort's defences and some of its internal road system (fig 2), all overlain by a pattern of later rig and furrow cultivation. The site measured 257m from east to west and at least 135m from north to south, giving a minimum area of 3.47 ha (8.57 acres) over the rampart. The road to the north gate showed particularly strongly, whilst the road between the east and west gates was rather fainter. Short sections of ditch and rampart have long been known as visible surface features in and just outside the southern boundary of the main fort field (fig 1: ba & bb). These run more or less parallel to the northern defences and have thus, not surprisingly, been assumed to represent the southern defences. It was not possible to conduct a geophysical scan of what little remains of these earthworks as they lie partly in a hedgerow and partly in a dense thicket above the Almond. Nevertheless, the survey does show virtually the whole of the north, east and west sides (minus the small area in the NW corner that was clipped by the A9 dual carriageway), and adds considerable detail to the rather poor indications built up through air photography. The site is bisected by the twin track railway to Perth, which has obscured a c. 19m wide swath, but the survey shows that there is good survival of remains right up to the railway boundary fence. Moreover, as the line here runs on a low embankment, it is possible that archaeological remains might even survive beneath it. Perhaps the least expected discovery was the trace of what seem to be two parallel ditches just east of the railway, which appear to cut the site in two. These raise two possibilities. Firstly, as the fort was occupied in both the Flavian and Antonine periods, one might have expected to find signs of a complex series of defences from different phases, similar to those known from the other two Gask forts: Ardoch and Strageath, which had a similar occupation history. In fact , however, the situation seems to be simpler than anticipated, but it is possible that these ditches might be part of a different phase. If so, the fort may have been cut down in area during its later occupations in the same manner as Ardoch. Alternatively, it is possible that the ditches represent the true western side of the fort, whilst the area to the west of the railway is an annexe. Certainly, in the past, the fort has seemed unusually large and yet remarkably narrow in proportion to its e-w length. The possibility of an annexe would reduce the fort proper to a more reasonable 2.32 ha (5.73 acres) (plus a 1.15 ha (2.84 acre) annexe) and would also explain why neither the survey, nor past air photographic work has detected a gate through the western defences, since such an annexe might only have been accessible from the fort itself and not from the outside world. The lack of a gate might, however, hint that these defences were a later addition, because the 1973 excavation detected cobbling and packed gravel to their west (Adamson 1979, 34 & fig 1, 2) which was interpreted as an incoming road and the resistance survey confirmed the existence of a band of high readings here, which head towards the putative annexe from the SW. The road cannot be traced inside the enclosure, but it would seem illogical to run a road up to a blank rampart if the two were contemporary. Moreover, parts of the western ditch show as a band of low readings sandwiched between narrow lines of high resistance. This may reflect the excavation's finding that the ditch here was clay lined, for the clay might well produce higher resistance as it reaches the ploughsoil base on either side of the main fills. It is noteworthy, however, that the fort ditch to the north of the railway does not show the same banding pattern, and thus might again belong do a different phase.
Finally, a number of faint areas of high resistance were found in the fort's interior. One runs from the southern edge of the survey area, c. 230m to the east of its base line, and may be the road to the former farm of Broxie, shown of Roy's (1793, PL 12) plan of the site. Other areas are more diffuse and may represent artificial gravel standings, geological features or some mixture of the two.
Magnetometry
The magnetic data more than made up for the resistivity problems, being the clearest yet produced during the Roman Gask Project's long series of whole 
Roman fort surveys (fig 3). It shows an almost complete plan of the internal buildings, along with other features, such as roads and probable rampart ovens. In the process, it seems to reveal a somewhat unusual fort plan. There is a line of six probable barrack blocks (or barrack pairs) in the fort proper, each c. 55m in length, three on each side of the road to the north gate, set with their long axes at right angles to the northern defences. To their south, a road runs e-w along the entire length of the fort, beyond which are rather fainter traces of further buildings. From east to west, these include at least two more rectangular buildings, this time set with their long axes parallel to the northern defences, followed by two much larger, courtyard structures that probably represent, respectively, the headquarters building and commanding officer's house. The fort would thus face north and the road running north would be the via praetoria. If so, the east-west road is the via principalis and might be slightly non-standard, although far from unique (e.g. Burgh-by- Sands I in Cumbria), in running along the fort's long axis. At first sight, however, the most unusual aspect is that there is no room for the additional barracks to the south of the main range that we would expect of a normal fort configuration, at least if the known stretches of ditch and rampart to the south belong to the same phase. This would result in a fort which had all of its barracks in the praetentura, and nothing but the main range behind the via principalis and so, in effect, no retentura.
To the west of the railway, at least three more long rectangular structures can be made out (with varying degrees of clarity) in the possible annexe. The most visible lies in the NW corner with its long axis parallel to the western defences, and seems to consist of a series of rounded features inside a rectangular structure, which even produced an area of low readings in the resistance survey. This might be a set of ovens protected by a timber shed, or a workshop containing furnaces, both of which might aid the conductivity of the soil as well as producing magnetic signatures. Given its straight line configuration, however, it seems less likely to be a bath block, although baths do occur in fort annexes elsewhere in Scotland. Indeed the survey showed no likely bath building anywhere in the area covered but, assuming that such a structure was built on the site, it might have lain on lower ground closer to one of the rivers and so may well have been destroyed by erosion.
To the north of the fort, air photography has shown a series of roundhouses, but these hardly show in the magnetic data and were not covered by the resistance survey. However, there are clear signs of more rig and furrow in an area indicated as being under such cultivation in Roy's 18th century plan. The survey also detected one much larger ring feature (never seen from the air), that might be a palisaded enclosure, and it also picked up part of a ditch that runs at a distinct angle to the rig, but parallel to the fort's northern defences, at a range of c. 85m. This is probably the same ditch as that shown on just one air photograph of the site (RCAHMS neg: PT14699, taken 1983), which has a very Roman looking curved corner close to the Tay, which again parallels the fort's own NE corner. This may be part of a temporary camp, or another fort or annexe phase but, whatever the case, no internal features from it can be seen in the survey data.
Interpretation
The effective absence of a retentura would not be unique (e.g. Zwammerdam fort in the Netherlands) and so the possibility cannot be dismissed out of hand, but it would be extremely rare. The (perhaps more likely) alternative would be that the ditch and rampart to the south of the fort field do not represent the fort's southern defences, at least in the same occupation phase as the buildings detected by the survey. Instead they might belong to a different fort phase or even to a non Roman enclosure, such as the Medieval Rathinveralmond, the precursor of Perth, which is reputed to have lain on or near this spot. If so, the fort might originally have been significantly larger. If its retentura was the same size as the known praetentura, i.e. c. 72m wide, including the rampart, the fort's n-s length would increase to c. 207m and its area would grow to 3.56 ha (8.79 acres) over the ramparts, excluding the possible annexe. Indeed, if we include the ditch system, the fort could have extended up to 90m beyond where we currently have it. At present this might seem impossible for lack of room, as the ground slopes away steeply almost immediately beyond the fort field boundary to form the deep, steep sided Almond valley. In practice, the fort could have been more square in shape, so that the retentura might have been rather narrower that the praetentura (as at i.a. Künzing fort in Germany), but even if its ditches had originally sat very close to the valley edge, we would still have to assume the destruction of a significant amount of ground by the river: possibly up to 100m. This would be an impressive erosion rate for c. 1,800 years and it is noteworthy that the much larger River Tay, to the east of the site, shows little sign of having caused similar damage to the site, despite the fact that this part of the fort sits almost right at the edge of a near vertical river cliff, which is clearly eroding towards the west. Nevertheless, the Almond is notorious for its spates and air photographic indications of paleochannels, along with geological analyses have shown that its confluence with the Tay has moved a good way to the north since the end of the last Ice Age. Moreover, its channel can be seen to have moved north here even in the limited time since the advent of accurate mapping, for Roy's mid 18th century plan of the site shows it noticeably further south, with enough room on the plateau to hold the now vanished farm of Bertha. The river's movement is now constrained by the design of the Victorian railway bridge and the 1827 road bridge that takes the original line of the A9 across the river a little further upstream, so that the difference between Roy's river and today's may represent less than a century's erosion. Certainly other Roman forts in Scotland have suffered almost as greatly from river erosion, with examples in the same general area including Cardean, Dalginross, Inverquharity and even the Inchtuthil legionary fortress. Furthermore, there are 18th century antiquarian accounts of erosion (OSA, 15, 527f) and of deep features, containing Roman material being revealed in section by the Almond (Adamson 1774, 52 & Stuart 1852, 206), which suggests that it was indeed cutting towards the north at this time, and a Discipulinae Augusti inscription has been recovered from the river (Keppie 1983, 402), which also presumably got there through erosion. Under these circumstances, the loss of a significant area of ground in and around the southern part of the fort does not seem implausible and, if true, it might mean that the via principalis originally ran across the fort's short axis, in the more usual manner. It is even possible that the visible southern defences might represent a later, possibly Medieval, restoration of a by then much reduced defended area after the fort's partial destruction. Indeed it is not difficult to question the traditional Roman interpretation of these remains, since the two surviving sections (fig 1: ba & bb) seem to lie at a slight, but distinct, angle to one another, whereas the opposite, northern defences run perfectly straight and, as the magnetic survey would suggest that the southern section (fig 1: bb) might also overlie part of the principia, they would not sit well at least as part of the same phase as the detected internal buildings.
The well preserved archaeological remains revealed by these results were somewhat unexpected, for the site has been under the plough since time immemorial and, as already stated, it has never shown well from the air, despite decades of observation. As a result, it had been widely expected that the interior would be largely ploughed out, and that little more than the defences would survive. This is obviously not the case and the site must now be seen as having considerable potential to yield valuable further information.
D.J.Woolliscroft & B.Hoffmann.
University of Liverpool.
Acknowledgements
The work was funded by the Roman Research Trust and took place with the kind permission of the landowner, Mr A. Richie. The survey was directed by the writers and crewed by our long standing members, Rachel Hunt, Paul Murdoch and Dan Boddice, assisted by local volunteers.
Bibliography
Adamson, H. (1774) The Muses Threnodie, Perth.
Adamson, H. C. (1979) 'The Roman fort at Bertha', in Breeze D. J. (ed) Roman Scotland some recent excavations, Edinburgh, 33ff.
Adamson, H. C. and Gallagher, D. B. (1986). 'The Roman fort at Bertha: the 1973 excavation', Proc Soc Antiq Scot, 116, 195ff.
Hartley, B. R. (1972) 'The Roman occupations of Scotland: the evidence of Samian Ware', Britannia, 3, 1-55.
Keppie, L. J. F. (1983) 'Roman inscriptions from Scotland: some additions and corrections to RIB I', Proc Soc Antiq Scot, 113, 391ff.
Roy, W. (1793) Military antiquities of the Romans in Britain.
Stuart, R. (1852) Caledonia Romana, Edinburgh.
Woolliscroft, D. J. (2002) The Roman frontier on the Gask Ridge, Perth & Kinross, BAR (British Series), 335.
