Background
For many years, the southernmost known installation on the late first century AD Roman Gask frontier has been the fortlet of Glenbank (fig 1 & 2:1), which lies at NN 812057, just inside the Perthshire border, on the northern side of Strathallan.  It has always seemed highly unlikely that this was the frontier's real terminus, however, for Roman frontiers almost always end with full size forts, rather than more minor installations.  They also usually rest on significant topographic features, such as coast- lines or major rivers, whereas Glenbank appears to be a more or less arbitrary position, of no obvious strategic importance.  Certainly the Gask road is known to run much further to the south, reaching at least as far as the fort of Camelon in northern Falkirk.  One more intermediate fort is known at Doune.  Another has long been anticipated at Stirling, and either of these positions would make a more typical and, indeed, logical end to the system, lying as they do on rivers and the former Forth mosses.

In fact, the current status of Glenbank might result at least as much from a change in the modern agricultural regime as from any ancient reality, thanks to the way in which most additional Gask sites now come to light.  Ten new Gask installations have been discovered since the 1930's: Glenbank itself, and the towers of Greenloaning, Shielhill North and South, Westerton, Ardunie, Roundlaw, Westmuir, Peel and Huntingtower (fig 1), and all except Ardunie (Crawford, 1949,52) were found from the air as cropmarks.  As a result, the area to the south-west of Glenbank has long been monitored during air photographic flights, but there are problems which have rendered the search less effective.  Firstly, although much of the Gask line between Glenbank and Bertha runs through arable land, large parts of the Forth valley, along with lower Stathallan, are in pasture and so far less effective at producing visible cropmarks.  To make matters worse, Gask installations are always found close to the Roman road, which on the known frontier makes it easy to decide where to look for additional sites.  But, partly because of the cropmark deficiency, there is a long (at least 13km) stretch between Kinbuck and the southern suburbs of Stirling where even the approximate course of the road is unknown.  There are a number of possible routes (Woolliscroft & Hoffmann 2006, 78ff) of which the most likely pass either through Doune and the Fords of Few, or through Dunblane and Stirling.  It is even possible that both of these lines may have been followed, but until the matter can be settled more conclusively, any aerial search for more installations is inevitably hamstrung by the difficulty of knowing where to look.  There is, though, one short exception to this gloomy position: the 1.3 km sector between Glenbank and Kinbuck.  Here the road line is known with certainty, indeed parts of it are still visible on the surface, and much of its surroundings are in at least intermittent arable cultivation.  As a result, Roman Gask Project flights have paid particularly close attention to the area and a number of new features have emerged as a result.  In particular, three ring features have come to light beside the road: at Glassingall, Kinbuck Muir and Lower Whiteston (fig 2: 2, 3 & 4).  The latter (NN 800051) seems too small to be a candidate watchtower when the air photographs are properly rectified and mapped.  Moreover, its single entrance faces east, rather than south towards the Roman road (the pattern of all the known Gask sites), and faint signs of a possible central macula may show it to be a ring cist.  The first two sites seemed more worthy of further consideration, however, and they will be dealt with in greater detail below.

Glassingall

The Site
During a Gask Project flight in July 2006, the cropmark of a ring ditch was photographed at NN 796048 (fig 3).  The site lay in a field belonging to Faulds farm, immediately to the north of Glassingall and 72m north of the Roman road, which is visible here as a woodland track.  Rectifications of the discovery air photographs (fig 4) suggested that it was c. 19m in external diameter, and so lies within the size range of the known Gask watchtower ditches.

From first sight, there were reasons for doubting an identification as a Roman tower, but none of them were conclusive.  Firstly, only a single ditch could be seen on the air photographs, whereas all of the known Gask installations to the south of Kaims Castle, including the Glenbank fortlet, have twin ditches.  This was only a minor issue, however, for the frontier seems to consist of a series of building sectors, each of which had sites of slightly different design, perhaps constructed by different military units (Woolliscroft 2002, 18ff and Woolliscroft & Hoffmann 2006, 235ff).  Most of these stretches have single-ditched towers and it is perfectly possible that Glenbank marks the end of the double ditched sector and that any sites to the south might only have one.  Moreover, even in the known double ditched area, the outer ditches are sometimes very slight and can show poorly from the air, so that it remained possible that the Glassingall site might also, in fact, prove to have two.  Secondly, all of the known Gask towers have entrance breaks facing the Roman road, but no such gap could be made out in the Glassingall ditch.  But, the cropmark was at its least clear on its southern side and so, again, this may not be crucial.  Likewise, the fact that the site lies to the north of the road need have no significance whatever.  For, unlike many Roman frontiers, for example Hadrian's Wall, the Gask installations were not protected by a running barrier, such as a wall, bank or palisade.  There was thus no need for them all to lie on what might be looked on as the inside of the system and, in practice, they are positioned apparently at random on either side of the road.  As a result, although 11 of the 18 known towers do sit inside the road line, the remaining 7 are outside it, whilst two of the three known fortlets lie outside the line and only one (Glenbank) is inside it.  More significantly, however, a surface examination showed that, although the site lay above the level of the Roman road, there was higher ground to both its immediate west and north-east in the form of two marked knolls.  The site would thus have been ill chosen for a watch/signal tower because, although its views would have been reasonable from the full likely height of a Roman tower (perhaps c. 10m), especially to the west, they would have been dramatically improved had one of the two hillocks been used.  In particular, use of the north-eastern example would have brought the site into visual (and so signalling) contact with the nearest major garrison fort, Ardoch, without taking the site more than 150m from the road.  This is admittedly quite a distance, for the average Gask tower is only 35m from the line, and some are as little as 10m (fig 5).  But it is still less than the known tower of Raith, which was built 170m from the road, to exploit a similarly advantageous viewpoint.

The Survey
To glean further information without damaging surviving archaeological deposits, resistance and magnetic surveys were conducted in September 2008.  The former used a Geoscan RM15 resistance meter with 1m traverses, a 0.5m electrode spacing and readings taken at 1m intervals.  The latter used a Bartington Grad601 fluxgate gradiometer, with traverses again at 1m, and readings at 0.25m intervals.  As time was short, the area covered consisted only of concentric 40m and 30m squares, these being the largest that the magnetic and resistance meters respectively were able to scan in a single data grid.  The work was done immediately after the field was harvested and with lines of unbaled straw still on the ground.  This was fairly dry, however, and does not appear to have interfered with the results.

Neither data set was particularly clear, but the ring ditch is visible as a 1-2m wide band of low readings towards the northern corner of the resistance plot (fig 6).  The feature proved to be rather smaller and more ovoid in plan than had been anticipated from the air, being c. 14m in external diameter from SW to NE, and 17m from NW to SE.  The air photographs had also shown slight signs of a second, much smaller ring a few meters further to the SSW and this too was visible on the resistance plot, c. 7m away.  It showed as a c. 10m diameter ring of high readings, despite being clearly a ditch feature from the air, but this is a common phenomenon in this area (e.g. Woolliscroft 2002, 62ff) and usually shows that a ditch has later been backfilled with stones, presumably during field clearance.

The main ring feature did not show at all on the magnetic survey, despite the fact that the range between the top and bottom cut off points in the grey-scale plot shown in fig 6 was brought down to just 12 nT.  There was some slight trace of the smaller ring to the SSW, however, with a particularly low magnetic and resistance spot in its centre, along with signs of further magnetic disturbance to the NW.  Perhaps the most interesting feature of the magnetic plot, though, was a 1-2m wide curving line of slightly elevated readings, which lay towards the grid's northern corner and did not show at all in the air photographs.  On re-examination, it also proved to be very faintly visible on the resistance plot, this time as low readings, although the smaller grid used covered a shorter length of its track.  The smooth curve suggests that the survey may have revealed just part of a much larger full circle, in which case the section of circumference revealed would extrapolate to a total diameter of c. 46m.

Interpretation
The survey evidence seems to make it less probable that the main ring feature seen from the air is a Roman watchtower.  It is true that its size is still not completely incompatible with the known Gask towers, because the northernmost examples: Westmuir, Peel and Huntingtower (fig 1), are all between 15 and 16m in diameter (Woolliscroft 2002, 20).  Nevertheless, the markedly oval shape does not sit well with any of the other known sites.  Moreover, there are indications that the ring might have two ditch entrances, something that is also unparalleled amongst the Gask towers.  One does face SE, towards the Roman road, but there are much stronger traces of another facing SW, i.e. more or less parallel to the road.  The fact that the magnetic survey failed to detect the feature might suggest that there was no burning present.  If so, this too would be unusual for a Gask tower, since almost all of the excavated examples have produced evidence that they were demolished at the end of their service lives and the timber-work burned on site (e.g. Woolliscroft & Hoffmann  1998, 450ff).  An obvious alternative interpretation of the feature, as a roundhouse, might also be cast into doubt by the lack of burning, although it is not impossible that any trace could have been ploughed away.  Under these circumstances, and especially given the evidence for more than one ditch break, the ring might best be seen as a barrow.  This would also explain why no band of high resistance readings was detected outside the ditch to mark the upcast thrown out from it.  Such features have been seen in almost all of the Gask Project's resistivity work on Roman towers, even in intensively ploughed areas (e.g. Woolliscroft & Hoffmann  1997, 566), and they can still be seen on the surface at well preserved sites such as Kirkhill.  The ditch or gully around a barrow, on the other hand, was not a deliberate construction in its own right.  It was simply a ring quarry for the central mound, so the upcast would be thrown inwards and the feature itself might be distinctly intermittent, thus giving the impression of multiple entrances in a resistance plot.  Even so, it remains possible that any external upcast has been completely ploughed away and the site's identity must still retain a considerable degree of uncertainty.

The smaller circle would be compatible with a small ring cist, similar to one excavated by the Gask Project some miles to the SW at East Coldoch (Woolliscroft & Hoffmann, 2002, Feature C).  If so, the central anomaly would presumably represent the burial cist itself.  On the other hand, the large curving feature found by the magnetic survey would be more compatible with a palisaded enclosure, especially as stones that may be pot boilers and an anvil stone were seen on the surface nearby.  If so, it lies on rather uneven ground and it will be interesting to see if future aerial observation will cast further light on this site.

Kinbuck Muir

The Site
An examination of Gask Project air photographs and the aerial coverage available on the "Live Search Maps" Web site (http://maps.live.com/), revealed a feature just south of the Gask road (fig 2.3 & fig 7) on Kinbuck Muir (NN 803051), 830m to the ENE of the Glassingall survey.  As the name suggests, the site lies on partly drained mossland, belonging to Lower Whiteston farm and took the form of a ring, estimated at c. 21m in diameter, defined by a waterlogged, reed filled hollow.  This was truncated in the north by a large drainage ditch several meters across, that runs just inside the field boundary, which itself marks the southern side of the Roman road.  Both aerial and surface inspection showed that the feature was at least partly made up of apparently natural drainage channels of a dendritic pattern, which were crossed by artificial drains that once led water to a railway reservoir.  Nevertheless, the fairly neatly circular overall shape looked suspiciously artificial, as if the natural gullies might have intersected a ring ditch of a size that would fit well with a Roman tower.  The site enjoys a superb field of view in almost all directions, especially to the north over Strathallan and would have had all of the known Gask installations between Kaims Castle and Glenbank in sight from the height of a Roman watchtower, including the fort of Ardoch and the Glassingall ring ditch.  Moreover, although the towers on most of the Gask system show a more of less random spacing pattern, those in the southernmost sector are set at regular intervals (Woolliscroft 2002, 18ff) of 3/5 of a Roman mile (851m), and scaling off such spacings from the nearest known tower, Greenloaning, would put a tower very close to this spot.

The Survey
To gain further information, a 35m (SE-NW) x 50m (SW-NE) resistance survey was conducted in May 2008 using a Geoscan RM4 meter, with 1m traverses, a 0.5m electrode spacing and readings taken at 1m intervals.  The work was done in a dry spell, but the reed beds in the damp hollow made clean electrode contact with the underlying surface difficult. The resulting plot (fig 8) shows the network of drainage features that had been apparent from the air as low resistance bands and some did run perfectly, or near perfectly straight, confirming their artificial origin.  The ring feature was detected, as a truncated single circuit of low readings (suggestive of a ditch), c. 2m wide.  This surrounded an area of significantly higher resistance than the external background, and which appeared to swell towards the west at the grid's northern edge.  The feature was slightly larger than had been estimated from the air, at c. 23 in diameter.  It was also less regular and more subrectangular in form, with the southern part of the supposed ditch, in particular, being almost straight.  As the aerial evidence had suggested, the ring showed considerable interference by drainage channels, but it does still seem to show signs of an underlying artificial construction.  It is, however, rather difficult to determine exactly what does and does not belong to this, and only excavation could clarify the matter completely.

Interpretation
The size of the ring feature is fully compatible with a Gask tower ditch, for the known examples range in diameter from 15m (Westmuir) to 25.55m (Shielhill South), averaging 21m, and a ditch width of around 2m is also perfectly normal (Woolliscroft 2002, 18ff).  Nor is it unusual for a Gask tower ditch to be somewhat irregular and subrectangular in plan, whilst the westerly swelling in the central high resistance area could be seen on the surface to be caused by upcast from the modern drainage ditch and so was nothing to do with the possible ring ditch.  As a result, it is perfectly possible that the site does represent a tower.  Nevertheless, there are reasons to doubt such an identification and the survey is certainly not enough to prove the case without excavation.  Firstly, the site plan deviates more from the circular than any other Gask tower.  Secondly, although the site is truncated by the modern drainage ditch, enough survives to make a reasonable extrapolation of the rest, and this would suggest that it originally extended approximately 5m further to the NW.  This would have brought it right up to the edge of the Roman road (if not slightly further) and although most of the Gask towers do lie close to the road, there is usually a reasonable separation of at least 10-15m (fig 5).  Finally, the high resistance seen inside the ring seems rather too homogenous for a tower, albeit the readings grow higher towards the east.  The Gask towers have consistently proved to have a turf or turf revetted earth rampart inside their ditches, with a single break in line with the ditch entrance to give access to the central timber tower (e.g. Robertson 1974).  One would not expect to see either the ditch or rampart entrance on this site, because it would inevitably have faced towards the road and so would have lain at some point in the truncated area, facing somewhere between the north and north-west.  One would, though, expect the pattern of internal readings to reflect the shape of the rampart, with a high band surrounding an area of lower readings towards the centre.  It is also worth noting that the ring ditches of other Gask towers have shown broad external bands of high readings, which mark the upcast thrown out from the ditch when it was dug.  This is true even on sites which have long been under the plough and where this feature must thus have been largely dispersed.  No such evidence can be detected on the present site, however, even though the damp ground and the surface remains of the possible ring ditch would both suggest that little ploughing has taken place.  On the other hand, the consistent high readings inside the ditch might suggest that the upcast was actually thrown into the interior, in which case the most likely interpretation of the site would be as a barrow, later passed (or possibly even impinged upon) by the Roman road, and then truncated by the drainage ditch.  This identification must remain unproven without excavation, and the interior of the surface feature does not stand significantly higher than its surroundings as one might expect of a barrow site.  But further evidence might be drawn from the fact that the survey showed signs of two slight breaks in the ring feature: one very faintly visible in the east and another, rather stronger, in the SW.  These do not seem large enough to be deliberate entrances, being no wider than 2m but, as stated in relation to Glassingall, barrow ditches are merely ring quarries from which the material for the central mound was dug, and it is not unusual for them to be intermittent.

Conclusion
Of the three ring features so far found from the air to the south-west of Glenbank, none can yet be claimed with any confidence as Roman Gask frontier installations.  Indeed, whilst not conclusively disproving the possibility, the geophysical surveys at Glassingall and Kinbuck Muir have made it seem less, rather than more, likely that any of the sites are Roman in date.  This is not to say that we should abandon the idea that the Gask system continued beyond Glenbank.  Quite the reverse: the hypothesis remains as likely as ever.  But, for the moment, it must remain just that: an unproven hypothesis which might seem to reflect the balance of military probability, but for which there is currently no positive evidence.

D.J. Woolliscroft & B. Hoffmann.
University of Liverpool

Background papers

Aerial photographs

Acknowledgements
The writers are grateful to Mrs Henderson of Faulds Farm and Mr A.P. Anderson of Lower Whiteston Farm for allowing us access to the sites.  We also thank Mr W. Fuller for piloting our air photographic flights and Mr D. Boddice for his help on the Glassingall survey.

Bibliography
Crawford, O. G. S.  (1949)   Topography of Roman Scotland north of the Antonine Wall, Cambridge.

Robertson, A. S.  (1974)  'Roman "signal stations" on the Gask Ridge',   Trans Perthshire Society of Natural Science, (Special Issue), 14-29.

Woolliscroft, D. J.  (2002)  The Roman frontier on the Gask Ridge, Perth & Kinross,  BAR (British Series), 335.

Woolliscroft, D. J. & Hoffmann, B.  (1997)  'The Roman Gask system tower at Greenloaning, Perth & Kinross', Proc Soc Antiq Scot, 127, 563-576.

Woolliscroft, D. J. & Hoffmann, B.  (1998)   'The Roman Gask series tower at Shielhill South, Perth & Kinross.  Excavations in 1973 and 1996',  Proc Soc Antiq Scot, 128, 441-460.

Woolliscroft, D. J. & Hoffmann, B.  (2002)   'Iron Age and Roman period native sites at East Coldoch, Stirling. an interim report for work up to 2002', www.theromangaskproject.org.uk/Pages/Papers/ Coldoch2.html.

Woolliscroft, D.J. & Hoffmann, B.  (2006)  Rome's first frontier, Stroud.