The Roman Gask Project was founded in 1995 and is funded largely by the Perth and Kinross Heritage Trust. For the last five years it has been engaged in an intensive campaign of surveys, excavation, and archival work which has included, where necessary, the preparation of past research for publication. The Project is still underway and the present work will attempt to provide an overview of its principal results to date.
First, let us define the field of study. The Project’s remit has been kept deliberately broad to cover all of Roman Scotland north of the Antonine Wall, but as the name suggests, the principle focus has been on the system of military works on and around the Gask Ridge in Perthshire (fig 1). This consists of a chain of forts, fortlets and towers strung out along the Roman road to the Tay. It appears to be a frontier, perhaps protecting Fife and, in particular, the strategically sensitive, potential invasion corridor down Strathearn and, at first glance, there is nothing particularly remarkable about it. The line belongs to a familiar class of early Roman linear defences which also includes the Wetterau Limes in Germany. It is, however, a particularly early example, dating to the reign of the Flavian Emperor Domitian in the 80’s AD. Indeed, as the German frontier, which was once thought to be slightly earlier, has recently been re-dated to 105-115 AD, i.e. the Trajanic period (Körtüm 1998), the Gask may be the very earliest of these systems, in effect the prototype fortified Roman land frontier. As such, it acquires a special place in Roman frontier studies, because the state of development of the prototype has obvious implications for any study of the later evolution of Roman frontier design. Yet until recently, although much good work has been done, the line has been relatively neglected, especially when compared to the two British mural frontiers and, in particular, there has been less interest than one might like in studying the line as a complete system, rather than simply as a collection of individual sites.
As for the Project’s results, let us start with the much debated question of the system’s date and length of occupation. Of the three forts on the line: Ardoch, Strageath and Bertha, the first two have long been known to have both Flavian and mid second century, Antonine occupation and we have recently been able to prove a similar picture for Bertha (Woolliscroft forthcoming (c)) thanks to a collection of field walking finds 1, but it has long been assumed that the whole of the rest of the system must be purely Flavian. Yet the evidence to support this is negligible thanks to a near total lack of datable material from the towers and fortlets. Indeed at the time the Gask Project was founded, the entire 37km system of (to date) 21 minor installations had yielded only two published datable finds, and even these were less helpful than they might have been. For, the first, a mortarium fragment from Gask House tower (Robertson 1974,20f), derived from ditch backfill and was thus essentially unstratified, whilst the second, from Westerton tower (Hanson and Friell 1995, 506), came from a similar context and was only tentatively datable as Flavian.
The Project’s own work cannot claim to have greatly improved this position because only four more datable finds have emerged (in 16 excavations), only two of which were more closely datable than that they are Roman. The remaining two did, though, prove to be Flavian which does at least add some small additional support for the traditional date, but again they came from less than helpful contexts. The first was a sherd of late first century glass from the top of the ditch fill at Shielhill South tower whilst the other, a piece of Flavian Samian pottery, found with a less closely datable fragment of Roman course ware, derive from a still rather mysterious site at Cuiltburn, just to the south of the fort of Strageath (figs 1 & 6).
Even worse, there is still no dating evidence whatever for the fortlets, despite our own extensive work at the southernmost example, Glenbank and a smaller scale excavation at Midgate, which means that there can be no guarantee that these installations belong with the Flavian system at all. They might still prove to be Antonine. There is certainly a strong counter argument in that Glenbank in the south, like the towers immediately around it, has a double ditch, whereas the northern fortlets, like their towers, had single ditches, which does suggest integration between the two site types and thus a Flavian date for the fortlets. But the forts were reused in the Antonine Period, presumably as outposts for the Antonine Wall and so, we have recently proved, was the former Flavian ‘Glen Blocker’ fort at Dalginross (to the system’s north-west), so the possibility cannot yet be discounted that we have a Flavian system of forts and towers and an Antonine one of forts and fortlets, which does seem a somewhat more balanced picture than the usual assumption that the Antonine forts stood alone. Indeed, at one point, the writer came to believe this quite strongly, partly because this small (c. 24m square) fortlet type does tend to be more a second than a first century phenomenon, but mostly because of an intriguing piece of evidence from the best known of the Gask fortlets: Kaims Castle. For its excavator (Christison 1901, 18ff) was puzzled to find that he could locate no internal buildings, only a layer of metalling, whereas one would normally expect to find a pair of barracks. It did not occur to him to look under this surfacing and, at the time, there was perhaps no reason why he should have done. But we now know something he didn’t which is that a number of the Antonine Wall milefortlets had their interiors cleared of buildings and cobbled over at some point during their operational lives (Wilkes 1974, 57) and the Gask fortlets, if they were Antonine, might well have had a similar history. Our own work at Glenbank failed to find any buildings even under the internal surfacing, however, except for a massive timber gate tower at the entrance, and so it would seem that the Gask fortlets either never had buildings or, perhaps more probably, that their buildings are for some reason no longer archaeologically detectable. They may have been of cob, mud brick or turf walled construction or they may have been founded on sleeper beams which rested on, rather than cutting into, the surface. Even tents may have been used for seasonal occupation and if any of these scenarios are true the remains will have been ploughed away long ago. Whatever the case, the site appeared to have been in use for some considerable time as the ditches had repeatedly silted to the point where they had to be re-cut, and only further excavation and a firm date for the fortlets will allow us to see with certainty where they fit.
Where our work has been of more value is in shedding light on the system’s likely life span. In the past this has usually been assumed to have been exceedingly brief, perhaps no more than a season or two (e.g. Breeze 1982, 65). For, if the historian Tacitus is to be believed, the Romans did not even get to this area until Agricola’s third campaign in 79 or 80, whilst Hobley 1989 has argued, from coin evidence, that they must have left again by 86 or 87. But many would doubt whether the system could even have lasted this long, partly because there has been reluctance to believe that it could have coexisted with the more northerly line of glen blocking forts around the highland fringe, and partly because, although the forts had shown signs of two Flavian structural periods, which suggested a need for refurbishment during their life times, the minor installations had been seen as single period, which has lead to a number of suggestions for suitable stages within the Flavian conquest and occupation where a short chronology Gask system might be fitted (i.a. Breeze 1982, 65, Maxwell 1989,121-128, Pitts and St.Joseph 1985, 263-281). But here, the Project has been able to modified the picture considerably, because excavations at two of the southernmost towers: Greenloaning and Shielhill South (Woolliscroft and Hoffmann 1997 and 1998) found signs of two structural periods in both, whilst more recent work by the CFA (Glendinning and Dunwell forthcoming) has produced similar evidence from the intervening tower of Blackhill Wood (fig 1).
At all three sites, the evidence took the form of a replacement of the tower’s timber corner posts at some point during its service life, whilst Shielhill South also showed obvious signs of ditch re-cutting. This means that all of these sites had to be completely rebuilt at some point and this should suggest more than just a brief occupation. Of course it could be argued that this might indicate only that the towers, like the forts, were reused 60 years later in the Antonine Period but this does seem very unlikely for, although Shielhill South produced one of the new Flavian dates, there have been no Antonine finds of any kind from the minor installations. Moreover, at Greenloaning, where the interior surfacing survived well, there were no indications of a prolonged abandonment between the two tower phases and the fact that at all three towers the replacement posts were set in almost exactly the same positions as their predecessors, whose locations were thus presumably still known, would suggest that there was no significant break between the two phases. We would, therefore, appear to have evidence for a more prolonged Flavian occupation, although just how long still remains open to question. Much probably depends on how long tower uprights of up to 40cm in diameter could be expected to last before needing replacement which in turn depends, at least partly, on what they were made of. Hanson (1978) has pointed out that the Romans sometimes used far from ideal timber in military structures, alder for example, which tends to rot fairly quickly once set into the ground. Environmental analyses from a number of Gask sites have suggested that virtually the only trees in this landscape in Roman times were water loving species, such as alder, around the rivers (Ramsay in both Woolliscroft forthcoming (a) andWoolliscroft and Davies forthcoming), but there are a number of wood fragments from the post holes of Raith and Roundlaw towers (Christison 1900, 28f and Ramsay in Woolliscroft forthcoming (a)) which suggest that the structural timber actually used was good solid oak which, even without preservative treatment, should have been reasonably durable. This sits well with evidence from the ditches of Shielhill South and Glenbank fortlet (whose gate tower had only a single discernible phase) which had been re-dug (in the latter case repeatedly) only after enough time had elapsed for a considerable depth of silt to form in the primary cuts. We might conclude, therefore, that we have either a single extended Flavian occupation, perhaps lasting for the full six to seven years of the known Flavian period, if not longer or, at the very least, that we have two distinct Flavian phases separated by such a short period of disuse that it is now archaeologically undetectable. Other scenarios are possible, for example a wide-ranging destruction of the tower line by a severe storm, but the post hole evidence would not appear to support this.
There was, however, a problem in extrapolating these results to the rest of the system because these three towers are part of a clear group of four, from Greenloaning to Shielhill North (fig 1) at the southern end of the line which have one marked difference to the rest of the series in that, as mentioned, they are surrounded by two ring ditches rather than the usual one. Given a profound lack of dating evidence, and to date we have only a single shard of late first century glass between all four sites, there could be no guarantee that these towers shared the same history as the rest of the line. It remained conceivable, for example, that the southern end of the system was built first and then extended using a slightly different tower design. Moreover, if that was the case, it might also be possible that the southern towers were rebuilt when this extension was added, even though this might not yet have been strictly necessary on mechanical grounds, so that everything was new together. In other words, it remained possible that we might still be dealing with a relatively short chronology. In fact, there had long been evidence from earlier excavations that the towers on the longer, northern, single ditched, part of the line had also been rebuilt (Woolliscroft 1999, 294), notably from Midgate, Raith, Moss Side and Westerton, but it had never been enough to be conclusive.
The only solution was to excavate another of the northern towers with this specific question in mind and in 1998 the Project excavated the northernmost tower currently known on the system, West Mains of Huntingtower (fig 1). The evidence here was less clear cut than further south as the tower post holes had been badly damaged by land drains. Nevertheless, the site was still able to provide relatively clear evidence for rebuilding (Woolliscroft forthcoming (a)) thus producing a consistent pattern which means that we should now have a reasonably unified history for the entire frontier and probably a longer one than had previously been though.
Whatever the duration of the occupation, however, the system was eventually abandoned and there has long been evidence to suggest that this was a matter of deliberate policy rather than the result of at least direct native coercion. For we find a general picture of careful demolition and the elimination or burial of useful material. Our own work has added further support to this picture, for Glenbank had its gate tower posts dug out, The fortlet of Midgate had its ramparts partly shovelled into its ditch and at both Greenloaning and Shielhill South the tower posts had been extracted and burned, along with wattle and daub panelling, at the end of the second period. But we can introduce one subtle extra nuance, for although the end when it came may have been orderly, it might also have been rather sudden and unexpected by the men on the ground. Because Midgate seems to have been abandoned part way through a ditch re-cut; not, one would have thought, a particularly long job on a site of this size and something that would surely not have been started had the garrison known that they were about to leave.
Another area being studied by the Project is the building of the system. Unlike those working on later frontiers, however, we do not have the luxury of being able to determine relative dates by means of stratigraphic relationships between individual sites and a running barrier of some sort, such as a wall or palisade. For no such structure has ever been traced on the Gask, where the only constant element is the road Moreover, it is worth saying at this point that there is no guarantee that even this is Flavian, at least in its currently known well engineered form, for despite the fact that the Project is currently publishing the records of two former road sections, one of which we have actually reopened, there is still no dating evidence and certainly no other part of Flavian north Scotland had such a well constructed road. In any case no stratigraphic link has ever been shown be-tween the road and any of the frontier installations and it has thus proved impossible, so far, to use it to determine the order in which the different elements of the system were built.
There is, though, one clue that might be relevant to the system’s construction. We have already noted the distinct group of double ditched installations at the southern end of the system but this sector also has another distinction. There have been a number of attempts over the years to identify a regular spacing system on the Gask, based on the Roman mile (e.g. Rivet 1964). These have been far from convincing and it has to be said that this is something of a British obsession. We have grown used to the idea that systems like Hadrian’s Wall have their installations set to a fairly rigid spacing and the same pattern may now be emerging with the milefortlets of the Antonine Wall (Woolliscroft 1996, 158-167). Elsewhere in the Empire, however, we find frontier sites positioned far more flexibly and so archaeologists there feel less obliged to, shall we say, try to impose order. In short, there is no need to expect the Gask to show regular spacings and much of it is indeed more or less random, but this southern area does shows regular tower spacings of 3/5 of a Roman mile (Woolliscroft 1993, 293). This might be another hint that we have two construction periods on the system, but there is an alternative explanation as Hanson and Friell (1995, 513) have recently raised an interesting point which we had also been considering. For, on other Roman frontiers, such as Hadrian’s Wall, differences in design often represent not different building periods, but different construction teams working simultaneously, and the same may be true on the Gask. If so, can we find evidence for any other such building sectors ?
Unlike the Wall, the Gask does not provide the luxury of building inscriptions, but we might be able to tell something from looking at structural differences and there are a number of criteria that could be chosen. For example the size of the towers themselves, or their shape, for whilst most are rectangular, a few are near perfect squares. But these differences appear to be more or less randomly distributed.
The sites’ ditches are a different story, however and have characteristics which may point to there being no fewer than four building sectors on the system. The first is the southern group of four towers which, with their double ditches, are not surprisingly the largest sites in overall ground plan (fig 2). Next comes a group of sites from Kirkhill to Midgate towards the eastern end of the Gask Ridge which show no regularity of spacing but which, although single ditched, have ditch diameters averaging c.22m, only slightly smaller than the southern group’s outer ditches (which average c. 24.5m). The ditches of this group are also much more substantial than those of the southern towers. They range up to 1.8m deep and 3-4m wide, whereas those of the southern group are often less than 50cm deep and only just over a meter wide and can have had little if any defensive value.
Next, lying between the southern and eastern Ridge groups is an intermediate group from Westerton to Roundlaw, whose ditches are noticeably smaller in diameter (average c. 19m) and which have intermediate ditch volumes at around 1m deep and 2m wide. This group also shows no regularity of spacing, but it does have another distinguishing feature in that, unlike the rest of the line, its rectangular towers have their long axes facing, rather than at right angles to, their ditch entrances. Moreover, at least one site, Westerton, had a pair of projecting slots running out towards its entrance which might be the foundation for a flight of steps to the tower top (Hanson and Friell 1995, 502ff).
Finally, Huntingtower yielded a ditch diameter of around 15.5m which is smaller even than some of the southern group’s inner ditches (which average c.16m) and a geophysical survey of its unexcavated western neighbour Peel shows this site to be almost identical. Huntingtower’s ditch sections were similar in volume to those of the intermediate group, but it had its tower set right at the back of the internal area, with its rear posts set into the turf rampart (fig 3), which may or may not prove to be a further diagnostic feature. There are also indications that this northern group might again have a regular spacing interval, this time of 2/3 of a Roman mile.
All this seems quite clear cut and logical, for the groups are not too dissimilar in length and they coincide reasonably well with what one might expect to be the spheres of influence of the forts. If we do have four groups, one can hardly resist noticing that there were also four legions in Britain at the time, which are the usual builder units and, as none of these designs repeat on further sectors, we may even have an indication that the entire frontier was built in a single season. There is, however, just one anomaly. For Parkneuk, at the western end of the Gask Ridge, which one would expect to belong to the intermediate group, has the widest ditch diameter of any tower known on the Ridge (22.5m) and a ditch volume in the eastern group range. Its excavator (Robertson 1974, 21ff) thought that its tower was oriented in the intermediate group fashion, with its long axis facing the entrance, but not even this much is certain and so for the moment these groupings must continue to be treated with a degree of caution. There are in fact gaps in the spacing pattern which suggest that there are still quite a number of towers to be discovered on the Gask, especially in the intermediate group sector, and only further study will tell whether Parkneuk is really an isolated aberration or whether the model presented above is a figment of the current, incomplete data.
In a number of other areas the Project has, as yet, had fewer results. We have not, for example, been able to trace any extension of the system to the north or south of its previously known extent from Glenbank (to the north of Dunblane) to Bertha on the Tay, although it does seem likely that it may extend further, at least to the south. Certainly the road is known further south, although so far without the accompanying towers but, despite a thorough Air search, we have yet to see convincing evidence that it ever ran north of the Tay, despite antiquarian reports and the presence further north of additional forts, at least one tower, and the legionary fortress of Inchtuthil. Likewise, an intensive search has been made, both on the ground and from the air, for additional sites on the known parts of the line. This time there can be little doubt that these are there to be found, since the spacing pattern still has obvious gaps, but for the moment we again have little to report except that we have been able to confirm the existence of two unusually small (c. 0.4ha) Roman temporary camps at Easter Powside and East Mid Lamberkin (fig 4) near the tower of Peel.
Work on the military aspects of the frontier continues, but the Project now intends to give more attention to the native/civilian populations of the area. One large geophysical survey has already been conducted outside the fort of Strageath in the hope of locating vicus activity (as yet with inconclusive results) and more work is planned.
For the moment the Project’s principle tool for studying the indigenous population has been environmental analysis, but we have been hampered here by the free draining and acidic nature of many of the local soils which tend to destroy both pollen and bone. The picture to date is thus far from complete, but what evidence we do have points to a virtually treeless landscape given over almost exclusively to grazing with, as yet, little sign of cereal cultivation. This pattern contrasts with a more arable regime seen in the area in still earlier times (e.g. Romans & Robertson 1983) but, as the same picture can be seen in both immediately pre-Roman contexts and in Roman ditch silts, the Flavian incursion, at least, may have had relatively little impact on agriculture.
As for the natives themselves and their interactions with the invading power, we have as yet only tantalising glimpses. Three of the Gask installations, including Huntingtower and Glenbank at opposite ends of the line, overlay native groove houses and it is tempting to conjure up a picture of local people being thrown out of their homes to make way for the army but, in the absence of adequate dating evidence, it remains equally possible that these features may have been abandoned for centuries before the Romans arrived. Air photographs of the area have revealed numerous Iron Age sites but, as always, it is difficult to date such features with any precision from the air to select promising targets for further study. As a compromise, the Project has been looking for multi-phased sites. These should be an indication of prolonged occupation which should, at least, provide a statistically higher chance of continuity into the Roman period, and an excavation is currently underway at East Coldoch (fig 5), near the Roman fort of Doune. This site shows at least five separate phases and C14 dates and environmental analyses from the first season are currently being awaited.
It is also worth mentioning an unusual site at Cuiltburn (Woolliscroft forthcoming (b)), just to the south of Strageath which may have a civilian if not native origin (fig 6). The feature was tentatively identified from the air as a new fortlet but large scale excavation by the Project has disproved this. For, although the site did produce a few sherds of Roman pottery (and nothing from any other period), along with Roman looking rectangular beam founded buildings, grouped around an internal courtyard and fronted by some sort of post hole based facade, the defences are not those of a fortlet. The ditch goes around only three of the four sides and, although the site sits right beside and parallel to the Roman road, the open side faces away from the road not towards it. The ditch itself has a steep sided, flat bottomed profile, quite unlike the Roman military “V” shape and there were no signs of an internal rampart or palisade. Nevertheless the site is quite probably Roman and it is certainly fascinating in its own right. Indeed we may be dealing with an entirely new site type, for we have yet to find a single excavated parallel.
Finally, a study has been made of the Gask’s signalling arrangements, to allow a further comparison with later Limes systems elsewhere. If the Gask really was the prototype Roman frontier, it would be useful to know how developed it was if we are to judge the level of innovation which look place later, and a technical aspect like signalling might well be a good indication.
On most of the later Roman frontiers of Europe we see signs of what have become known as Direct Signalling systems, in which almost all of the smaller sites, the fortlets and towers, are carefully sited so as to have a view to a fort (Woolliscroft 1989,1994 & 1996, Woolliscroft and Hoffmann 1991 & 1999). This would have allowed direct signalling via the visual techniques of the day, whilst any remaining sites could have their signals passed on to a fort via (usually single) relays. And, despite its early date, this is also the pattern we find on the Gask (Woolliscroft 1993). Almost all of the lower lying sites have a direct view to a fort, whilst the fortlet of Kaims Castle has been skilfully positioned to be able to see both Ardoch and Strageath forts simultaneously (fig 7). Moreover, although the topography of the Gask Ridge itself restricted the views of some of the towers, there is not an installation on the system that would have needed more than one relay to transmit its signals to a fort. This means that apart from the lack of a running barrier we already have on the Gask all of the basic elements of a classic Limes system present in the initial design, sprung, as it were, virtually fully formed from the head of, if not Zeus, then at least the Emperor Domitian and his governor of Britain at the time, either Agricola or his unnamed successor.
1. The writer would like to thank Mr J.J.Walker and the Cumberland Historical Society for donating this material.
Christison, D 1901 ‘Excavations Undertaken by the Society of Antiquaries of Scotland of Earthworks Adjoining the “Roman Road” Between Ardoch and Dupplin Perthshire’, Proc Soc Antiq Scot, 35, (1901), 16-43.
Hobley, A S 1989 ‘The Numismatic Evidence for the Post-Agricolan Abandonment of the Roman Frontier in Northern Scotland’, Britannia, 20, (1989), 69-74.
Woolliscroft, D J forthcoming (a) “The Roman Gask Series Tower at West Mains of Huntingtower, Perth & Kinross”, Proc Soc Antiq Scot, 130.
Woolliscroft, D J & Hoffmann, B 1997 “The Roman Gask System Tower at Greenloaning, Perth & Kinross”, Proc Soc Antiq Scot, 127, 563-576.