Proc Soc Antiq Scot, 123 (1993), 291-313

Signalling and the design of the Gask Ridge system D J Woolliscroft*

ABSTRACT This paper attempts describeto signallingthe arrangements along Gaskthe Ridge, Strathin Earn, and the influence that these may have had on the general layout of the system of Roman installations, in an effort to understand the purpose and context of the Roman line. It also re-excavationreportsthe on possiblea of Roman fortlet within metresGaskthe of Ridge tower of Midgate (NO 02112047). This juxtaposition suggests that the Gask Ridge fortlets may not be exactly contemporary with the towers. Although, for the moment, there is insufficient evidence proveto thatfortletsthe belong anotherto Flavian a phase, to or different period altogether, former the still may appear more likely.

INTRODUCTION Scientific study of the Gask Ridge began at the turn of the century with the excavation by D J Christison (1901e fortleth f f Kaimo o )t s Castl e eigh somd th an f et o ewatc h towers then known on the Ridge itself. Although a great deal was learned through this programme, none e siteoth f s yielded dating evidenc ee shal w and s l a ,see e excavatioth , e fortleth y f ma o nt have been flawed. Since then, another fortle coms ha tligho t e t Glenbana t k (Maxwell 1990), to the south of Ardoch, along with evidence for 10 new towers, bringing the number known or suspected1 to 18 (illus 1). The towers can also now be seen to fall into two distinct groups, although the significance of the change is uncertain. For, whilst those to the north of Kaims Castle, includin gGase thosth kn o eRidg e itself, hav singlea e surrounding ring ditch, those further south have two. Furthermore a singl, e stratified Flavian sherd from Gask House (Robertson 1974, 20) and a possible second from Westerton (private communication), both in e northerth n group, have finally provide probabla d e date for t leasta , e singlth , e ditched towers, whilst the discovery of the Flavian fort at has suggested a possible destination for the southern end of the system (Maxwell 1984). picturA e has, thus, bee integraten a n f builo p u t d Flavian system, stretchin t leastg(a ) fro Teite mth t Doun ha Bertho Taye t fortlete th th ;n a o f Glenban o s Kaimd kan s Castle have usually been seen as part of this system, despite their current total lack of dating evidence.2 With this much apparently settled debate th ,s move consideo ha t e n d o exace th r t contexn i t which the system was built. For example, some scholars would argue for an early date, with the line forming part of the provisions made during Agricola's fourth season (Hobley 1989, 73-4lated an r) replace glen-blockine th y db g forts. Others suppor latea t r date, wite th h

Holwoo4 2 * d Drive, Manchester TOWER-SINGLE DITCH

TOWER-TWIN DITCH

ILLUS 1 The Cask Ridge system WOOLLISCROFT CASE TH : K RIDGE SYSTEM 293

system representing an attempt to hold a frontier line defending Fife in the aftermath of the abandonmen f Inchtuthio t rese f northerth o t d an l n (Breeze 1982, 61-5). Case alsn made ca sob r regardinefo e Casa 'bacg th s ka k stop' frontier r whicfo , e hth glen-blocking forts served as outposts, or for seeing the glen-blockers as the true frontier wit e Gashth s merelka closela y y watched road supervisin e strategigth c invasion route through Strathearn (Pitts & St Joseph 1985, 278). Yet, despite the fact that three of the four systee th fort n mo - sArdoc h (Breeze 1983), Strageath (Frer Wilke& e s 1989 Berthd an ) a (Adamson & Gallagher 1986) - have Antonine as well as Flavian occupations, the assumptio more n th tha l e al tmino r installations belon singla o gt e Flavian lins persisteeha d throughout these discussions. If this was the case, we would appear to have a fairly logically laid out system. The presenc f Glenbaneo k between Doun Ardochd ean Kaimd an , s between Ardoc Strageathd han , suggest alternatinn sa g serie f fort t fortletsalono d ou an roa se intersperset d gth dse san d with a closely spaced serie f watco s h towers. Attempt fino t s d signregulaa f o s r tower spacing interval ,e Romabaseth n o dn mile (Rivet, 1964, 196-8), have never proved wholly convincing t thi(Tablbu s , neee1) d hardly surpris irregulae vien th i f s wo eu r tower spacings contemporare onth y Wetterau e nortth f Frankfurho o t , Germann i t y (Woolliscrof& t Hoffmann 1991).

TABLE 1 Gask Ridge site spacings

Glenban 230k- c Greenloanin 0 m g Greenloaning - Ardoch c 2750 m Ardoch - Blackhill Wood 900 m Blackhill Wood - Shielhill South 875 m Shielhill Sout h- Shielhil m 0 95 l North Shiehill North - Kaims Castle 875 m Kaims Castle - Westerton 2300 m Westerton - Strageath c 4200 m Strageat h- Parkneu k 1750m Parkneu k- Rait h 1520m Raith - Ardunie 1510m Arduni e- Roundla w 1110m Roundlaw - Kirkhill 960 m Kirkhill - Muir o' Fauld 1440 m Muir o'm Faul0 87 d- Gas k House Gask House - Witch Knowe 800 m Witch Know e- Mos s Side 1120m Moss Sid e- Midgat e 1400m Midgate - Westmuir c 915 m 397c 5m Westmui r- Pee l Peel - Huntingtower c 1940 m Huntingtowe 317c 5m r- Berth a

It is, however, apparent that the system as we have it today cannot be complete. Glenbank is not intervisible with Doune and, although Huntingtower would, from its likely original tower height, have been visible from Bertha, both spacing and signalling considerations suggest that further towers await discovery at both ends of the line. Other obvious gaps exist between Westmuir and Peel, Westerton and Strageath, and Greenloaning and Ardoch, whilst a number of smaller gaps (Table 1) might also repay attention. At one time it also seemed likely that the system continued south, beyond Doune, perhap Flaviae th o st n for f Camelono t well-preserveA . d hilring-ditcp to l t Wesha t Pleano t , | SOCIET 4 29 ANTIQUARIEF YO SCOTLANDF SO , 1993

the south of Stirling, was once interpreted as a tower of the Gask Ridge type (Crawford 1949, 18). This has been shown to be non-Roman (Steer 1956),3 however, and all that can now be sai thats di , whilst extension e soutth f Douno ho st , indeed norte or th f Berthho o t , a cannot rulee b d out, therfiro n ms ei evidence that they existed.

THE SIGNALLING SYSTEM Despite these gaps in our knowledge, it is possible to reconstruct at least a skeletal signalling systee Gasth r km fo Ridge s currentla , y understood, base a stud n f eacdo yo h installation's field of view, from its likely full original height (Woolliscroft 1989b). As all long-range Roman signalling relie n visuado l techniques (Leiner 1982), direct communications could take place only between intervisible sites readee Th . s cautionei r d fro outsete mth , however, that suc hstuda shon yca w only wha possibls t wa bestta , ,eor likely. Signallin procesa s gi s that leaves little physical trac attracted ean d scant detailed comment from ancient writerst I . must always be remembered, therefore, that the fact that a signalling system was apparently both possibl desirabld ean givea n eo n Roman deployment doe t necessarilsno y mean thay an t such system actually existed. That said, there are design oddities on a number of Roman frontiers that see mmako t e sense onl s movea y o accommodatt s e signalling (Woolliscroft 1989a; Woolliscroft & Hoffmann 1991) so that some signals provision does appear likely on manyt all f Rome'no o , f i , s major military systems. e knowth Almos f o n l Gasal t k Ridge installation e intervisiblar s e with bot f theiho r immediate neighbours casew fe se wherth n i ;t (Westmuir e no the e yar r instancefo , , cannoe b t seen from Peel), it already seems likely on spacing grounds that the sites are not true neighbours, as intermediate towers may still remain to be found. It should, therefore, have been possibl paso et s signals from towe toweo t r r along t leasta , ,e line mosth f . o Indeedt , mutual intervisibility between neighbours seems to have been regarded as important and, as Wetterae onth u Lime Germanyn si , ther signse ear thairregulae th t r inter-site spacings areo t , some extent, causeneee th ensurr o exampledt y db Fo . eit , ther tendencya s ei , especialln yo the Gask Ridge itself, for the towers to be sited on or near pronounced changes in the angle of grounde slopth f fiel e e o limie tha o f vieth s ,on dth o f numbe a tt f o t wa o t se f site ro e sar or othe f theio r r neighbour appeao s d thein o an rs r skyline. Where th e cas th n thi o es i s Ridge itself, it is usually a site's eastern neighbour that marks the limit of its vision, which suggests thasystee th thav y mma e been lait fro weste dou m th . Indeed pronounceo s , s thidi s tendency, tha t becomei t s first-tima eas r yfo e visito locato t re groun th e site eth n o sd with little help from a map. Nevertheless, as the inter-site spacings are generally quite short (if slightly longer than Wetterau)e thosth n ei , rangin gjusm0 o t fro 80 t movec r 1500 m (Table 1), and many of the installations have fairly extensive views, most can see sites well beyond their immediate neighbours directione t leasa on , n i t . There are, thus, considerably more sites than would have been require a simpl r dfo e lin f relao e y stations sucn I . ha syste m each installation migh r nea o limi e expectee t th rb f a tvie o t e f botwli o neighbour s o hdit t s a o s e longesth e us t possiblo t e smallese th link d an st numbe f stationso r . This apparent over- provision may require explanation and this, in turn, may have implications for signalling. So before attemptin interpreo gt intervisibilite th t y data fro sitese worts mth i t i , h takin gclosea r look at the likely role of the system. Donaldson (1988, 352-3 s argueha ) d thae facth t t thae Gasth t k Ridgeo s site e ar s closely spaced suggests that the signalling techniques employed on the system were limited in range to around a Roman mile. But experiments by the writer (Woolliscroft 1993, 26-61) WOOLLISCROFT: THE CASK RIDGE SYSTEM | 295

have shown that almosrecordee th l al t d Roman visual methods were capabl f muceo h greater ranges, even in less than perfect weather: especially the relatively primitive, beacon based, alarm signals thalikele ar t havo yt frontiere eth been o f mos no e s (Woolliscrofus t t 1989a. 9) , This means that althoug Gas e signallina havy hth d kma eha g system t cannoi , thoughe b t f o t as being only a signalling system. Furthermore, whilst it would certainly have been possible line fo havth o ret e operate strictla n di y linear manner, with signals being relayed from site o sitt e through ever ye towe th lin n ki r chain, this would have been both unnecessard yan inefficient (Woolliscrof Hoffman& t n 1991, 534) t woulI . d also have been slowa r fo , signalling system based on relays only a few hundred metres apart is likely to have been little faster tha mountena d messenger onls ,wa y eve t i intende f ni carro dt y simple alert beacons. If it was expected to compete with the messenger in terms of information carrying capacity it would have been even slower Donaldsod an , n would appea havo t r e ignore e numeroudth s instance f Romao s n forts sited very much further from outlying watch towers, with which the obviousle yar y associate Topping d(e g 1987; Woolliscroft 1988, 25-7). e GasTh k tower spacing e othes th are n r o ,hand , perfectly reasonabl lina f r o e fo e observation posts; the overlapping fields of view would have allowed tight surveillance and, between them, they enable almost ever e watched b ye lin th o inct e f o h . Despite th e misleading term 'signal tower', which is often applied to them, signalling would probably have formed onl ya mino r par f theio t r role thin I . s they would parallel certain modern deployments writee th r s informei rfo , d that observation post e Gree partn th o s f no s Linn ei Cyprus are set at very similar intervals and knows from experience that those on the former East German border were often even closer. Yet these are or were equipped with modern electronic communications for which no such range restrictions could be suggested. In practice, like its present-day equivalent, any small Roman frontier post faced with trouble would simply have wante contaco dt e nearesth t t garrison bas thin (i es contexe th t neares tquickese fortth n possibley i ) wa t rols It . e would have bee superviso nt providd ean e early warning, rather than direct defence. In studying any possible signalling system, therefore, we should probably bear this in mind and look for simplicity and efficiency, rather than complexity. We should also look to see how the forts themselves might have communicated with eac singla alloo hs t a systeothee s t ewa th co-ordinateac o s rm o t d whole under emergency conditions and to facilitate a general troop concentration should a threat emerge at any one point that was beyond the capacity of the local garrison unit to handle alone. intervisibilite turth w o nt no Wn eca y detailwhae se to st patter n emerges. Doune cannot be seen from any other known Roman installation, although it does have a view into the heart of Stirlin s stila , lg if seem s possible expectee b o t fora , s i td there.4 This meane s th that r fo , moment signallino n , g systetracee b n mdca beyond Glenban . Thik 2) (illu s& fortle1 s n ca t be seen from Ardoch, however, and, although not visible from either Doune or Stirling, it stands in a reasonable position to link the fort to points further south. Likewise, all the known towers between Glenban Kaimd kan s also have direct view f Ardocho s s shoula , e thredth e additional towers that might be expected, on spacing grounds, between Glenbank and Ardoch: around Balhaldie (NN 823064),^ Kingfisher Hotel (NN 834080) and Kierallan Farm (NN 837089). In other words, the area is set up to allow the sort of direct signalling system already found on Hadrian's Wall (Woolliscroft 1989a) and the northern part of the Wetterau Limes (Woolliscroft & Hoffmann 1991, 535-6) in which each minor installation has a direct linfora o kt rather than signalling laterally from towe toweo t r r alon linee gth . The position of Kaims Castle is slightly unexpected, although it may still have a certain BERTHA

Huntingtower

STRAGEATH

I ROMAN FORT

| FORTLET |? POSSIBLE FORTLET Shielhill.N

Shielhill.S TOWER

Blackhill Wood i PUTATIVE TOWER

—— LOCAL SIGNAL LINK

—H LOCAL LINK VIA RELAY

INTER-FORT SIGNAL LINK 8km >GLENBANK TRUE SITE POSITION e signallinILLUTh S2 g syste Gase th f kmo Ridge, schematic drawing WOOLLISCROFT: THE CASK RIDGE SYSTEM | 297

logic. The fortlet stands at the northern end of a flat hilltop which is the only point from whic e fort th hf Ardoc o s d Strageathan botn e hseeca hb n simultaneously thao s , a singlt e installation here could have serve relaa s da y point linkin e twogth . But, standing whert i e does, Kaim s onli s y just abl fulfio et l this clea a role e sits reha Th . vie Strageato wt h (illus 3,A). But, even assuming thae fortle th s equippet wa t d wit a gath e tower higaboum 0 h1 t (Woolliscroft 1993, 61-90), this would have been visible only from the very top of a similar tower at Ardoch, and even then only one at the southern end of the fort. The intervisibility is, thus, somewhat borderlin would an e d have been easily blocke treey db r busheo e s th t a s southern end of the hilltop. Yet Strageath would have retained its easy view to Kaims even if the fortlet had been sited up to 350 m further south, so it is somewhat surprising that Kaims was not built at the southern end of the summit, c 320 m to the south, from where the view to Ardoch would have been very much more secure. Nevertheless e linth ,k betweeo tw e th n forts could stil operatee b lfortlet e th y s sitedd b a ,Kaim d havy an , ma se been pushee th o dt absolute limit of Ardoch's field of vision in order to improve its own more immediate view north and/or to compensate, to some extent, for the fact that the hill is much closer to Ardoch tha Strageatho t n e rangesTh . , with Kaims Castl s siteda e , are: Kaims-Ardocd an m k 7 3. h Kaims-Strageath 6.6 km; the position chosen for Kaims, whilst at first sight illogical, shortened its distance from Strageath by c 5%. Westerton direc a alss oha t vie Strageathwo t intermediato tw e th ; e sites, which spacing considerations might suggest lie between the two, also seem likely to have done so, as would any intermediate tower between Westerto Kaimsd an n . Again, therefore, thie s th par f o t system shows signs of having been efficiently laid out, with all of the known sites certainly enjoying direc tforto view tw forta witd s e o themselvean t s,hth s connecte a simple a vi d , one-stage, relay. But we now come to look at the Gask Ridge itself. Both Parkneu e see b Raitd nn kan frohca m Strageath t Raithbu , , which e standth n o s western summit of the Ridge, is close to the eastern limit of view of the fort, even from a 10 m tower. This means that if any signals from the rest of the Ridge-top towers were to reach Strageath, they would have to have been relayed via Raith. Interestingly, like Kaims, Raith could have bee nfarthe m abouo site t 0 p du 35 rt fro stild fore mth an tl have remained visible but, again, there are good reasons for the site being built where it is. The tower, as site933185)N (N dquita s eha , spectacular fiel f visiono d , makin excellenn a t gi t lookout point as well as an important potential signal relay. Its view extends far to its west and north, allowing it to watch long stretches of the Highland fringe. To its south and south-east it looks right across Strathearn to Kaims Castle, as well as controlling an excellent view of the floor of the strath itself. But, perhaps most importantly, its view to the east takes in all the remaining ridge-top towers as far as Midgate. Only to its north-east is its view somewhat limited. Yet farthee eastm , e positioth onl0 th ,o t 35 yr n would have been rather less advantageous: after passing over this western summit e Romath , n roa a littld n starteru o t s way to the south of the ridge top. As all of the towers are built close to this road, their views e nortth o ht become seriously degrade mora d ean d easterly Raith would have beeo n exception. In s excellencvieit f wo lookoua s a esignallind an t g position s interestini t i , g that a t lease aeriaon t l photograp f Rait ho s show ha h ) (CUCAn 96 whaD te PAK appearth e b o t s ditcrathea f ho r larger rectangular site (illu , arros4 ) surroundinw2 modere gth n water tank, the construction of which destroyed the known tower (arrow 1). This ditch is quite distinct fro e well-definemth d crop marformea f ko r plantatiot n ye boundar t no s ha y t (arrobu , w3) been tested by excavation. It is, however, of just the right size and shape to be an additional 298 I SOCIETY OF ANTIQUARIES OF SCOTLAND, 1993

o J3

3~USVO SWIVX

\ O

o IX

HOOQHV- miva WOOLLISCROFT CASE TH : K RIDGE SYSTEM 299

'

ILLUS 4 Raith. 1 Water-tank on the site of the tower; 2 cropmark of ditch of rectangular site, possibly a fortlet; 3 cropmark of a former plantation boundary

fortlet of the Kaims Castle/Glenbank type and, although such an identification remains, at best, hypothetical possibls it , e implications wil consideree b l d below. e facTh t tha tremainine mosth f o t g ridge-top tower behine li s e apparendth t optimum lookout line, must, inevitably, e system'reflecth n o t s function e mighOn . t have expecteda Roman frontier garrison here to have been primarily interested in what was happening to its north, and yet, of the 11 known towers on the ridge itself, only Raith and Midgate have completely unimpeded views in this direction, even from tower-top height. The traditional explanation has been that the ridge would have been heavily forested in Roman time thad e toweran sth t s were designe monitoro dt , repor andn o t somo ,t e extent, contro movementy an l s acros clearea s d stri f grounpo d aroun roae dth d itself (Breeze 1982, 62) othen I . r words, they were designe looo dt k east/west alon systeme gth , rathere thath o nt north. It now appears that the level of forest cover in Roman Scotland may have been exaggerated in the past (Breeze 1992, 331-5). Whatever the local position may have been, it seems unlikely, in practice, that there would have been that much need for such a watch, because anyone crossing the road from the north would already have had to reach, and then climb, the Cask Ridge itself; although this is not especially difficult, the fact that almost no modern roads cros ridge reflectioa th s s ei face th t f thao n t ther e verear y much easier ways through the area. The main route which, to judge from the temporary camps, all Roman invasions see havo mt e followed (Hanson 1987, 121-7) passes through Strathearne th o t , south of the Gask, along what is now the main A9 road from Stirling to Perth. The only other viable route through the area is that followed by the modern A822 which passes quite close to Strageat thed han n runnorte du s h through Crief enteo t f Highlane th r d passe t Fendocha s t I . is interestin noteo gt , therefore towere Gase th th f ,k n o thaso Ridgel al t matteo n , poow ho rr their views north, have superb views over Strathear thatd an mann i , y cases, these views | SOCIET 0 30 ANTIQUARIEF YO SCOTLANDF SO , 1993

would have been seriously worsened had the sites been positioned so as to obtain better views to the north. It may also be significant that Raith is positioned so as to enjoy particularly good views alon e routfollowew gth e A822no eth y b d; although Christison's claim (1901, ) thabeesite n 28 th teca n seen from Fendoc mistakens hi s visibli t i , e from fort'e hillth n ssi vicinity. This raise numbea s f interestino r g possibilitie additionad an s l parallels with other frontiers. e towere contemporarManth th f n o yo s y Wetterau Limes also have limited viewt ou s from the line and, again, these sites almost all have superb views over their hinterland. This means that any invader entering the area, along the Fulda Gap (the most likely invasion route fro e north)mth , coul e monitoredb conspired an d d against fro e relativmth e e safetth f o y surrounding hills, so that the Limes could act, to some extent, as a defence in depth. This, in turn, would have mad e defenceth e s more flexibl mord an e e abl maintaio t e n co-ordinated action after being penetrated than a shallow defensive line (Woolliscroft, 1988, 23-5); the provision of special towers, apparently to facilitate cross-Wetterau signalling, may be a sign that these opportunities were exploited (O.R.L., BanA t d 2,1Ab , 1936, Streck , 67-725 & e4 ; Helmke, 1910; Kofler, 1898). s possiblIi t e thaGase th t k Ridg havy ema e been intende woro similaa dt n ko r basis, wit e towehth r crews watching routeway likeld an s y trouble spots fro ma saf e distances a , well as simply monitoring their own immediate environs. If so, the superb overlapping views norte th f Raitho o t Midgate d thorouge han th d an , h observation cove f Strathearo r n provided ridge-toe th rese b f yth o t p sites, woulexcellenn a n i d t i hav t t. positioepu so o d o nt Suc possibilita h y might also she de system' th ligh n o t s relationship wit e glenhth - blockin e thesse g eo fortst installatione ar . Fore w f ,i thein i s r traditional defensive light (Ogilvi Richmon& e d 1967, 76), rather tha s 'springboardsa n ' (Breeze 1982, 55-6r fo ) assaults intHighlande oth s that never materialized, their position e sucar s h that most would have received almost no prior warning of attack. Under normal conditions, one assumes that the Romans would have maintained intelligence cover ahea f theido r line givo st e advance warnin majoy an f r o g attack (Woolliscroft 1988, 23-7). But, in the fluid conditions that are likely to have existed for Flaviae mucth f ho n perio Scotlandn di , intelligence breakdowns could easily have placee dth glen-blocker serioun i s s jeopardy r exampleFo . ,e accept eveon f e supposeni th s d Fendoch towe s Roman,a r t couli 6 d have give e formoro nth n t eminute w thafe na s warnina f go surprise attack dow Smae nth positione ' Glenth d an ,f othe so r glen-blockers, wher suco en h tower e evear s n hinte , woulat d d have been very much worse. Some, suc s Bochastleha , command views of only a few hundred metres into their glens and so, even at full cohort strength, their garrisons could easily have become hostages to fortune. Midgate and Raith, on othee th r hand, could, between them, have watched over many e sammileth f e o sHighlan d fringe, but from a range of about 10 km, as well as guarding against flanking movements through Strathearn. There is obviously little point in an early warning system if the warnings it provides come lato effectivr to efo e counter-measure takene b Gase o st Th . k Ridgenougr fa s ei h back froe mth Highlands to give up to an hour's warning of any attack (even by horsemen), but is still close enough to provide a view of what was happening, at least during daylight and in reasonable weather thin I . s context, althoug s stili t hi l possibl envisago et scenariea whicn oi glene hth - blockers forme distincda t non-contemporary line replace proceeder do Gaske alss th i y t doi ,b possible that the two lines could have usefully existed together, with the Gask representing a 'back stop' frontie whico t r glen-blockere hth s were somo t , e extent, outposts. CULTMALUNDIEA WOOCU -"

FORTLET •TOWE PUTATIVA R E TOWER

topographe ILLUTh Gase th eastere S5 f k th o Ridgf d yo nen e | SOCIET 2 30 ANTIQUARIEF YO SCOTLANDF SO , 1993

returo e signallinT th o nt ge ridge-to th details f o l pal ; towers r eas fa s Midgat a ts a , e (illus 2 & 3,B), can be relayed to Strageath via Raith;7 Midgate itself occupies a knoll toward ridge th eastere f eth so wit d nhen extensive viewl directional n si s except, againe th , north-east e directioth , f Berthano e sit f Westmuireo Th . nexe th , e t towee th east th n o ,o t r other hand mucs i , h more difficul accouno t t t fortowee Th . m0 r liefro92 sc m Midgats i d ean situated beside the Roman road, which again, here, lies a little to the south of the ridge-top. A section drawn from the 1:10000 Ordnance Survey map would suggest that this position would also have been just about visible from writer'e Raithth t surven ,bu sow y contradicts thisd an , towee th f sigh o s als ri t f Bertha o to ou whico t , closes i t hi r tha Strageatho nt . (The half-way poin s closi t Moso et s Side.) Suc hpositioa n follow e e patter othemosy th th sb f t ro t nse towere ridgth n e o sgooa and s s everd a ,ha e south viet i th , wo t . But, fro ma signallin g viewpoint, one might have expected this site to be different, for only c 300 m to its north is highese th t poin whole th tn e(15o Gas) 4m k Ridg 029210O e(N : illu. s5) Westmuid Ha r been buil thin o t s summit t wouli , d hav considerabld eha e advantages. For example, without losing much of its view to the south, it would have been visible from both Raith and Bertha, almost from ground level (illus 3,B), allowing a fairly simple two- stage rela lino yt k Bertha with both Strageat Kaimd han s Castle t woulI . d alson a hav d eha even better view north than Midgate, which would have muce allowese o ht furthet de i th o t r north-east along the Highland line, and east towards Perth and the end of Strathearn. Yet the opportunity was ignored, producing a tower with no view north and with limited views in every other direction except south. Fortunately, on spacing grounds, the next tower to the east should lie somewhere in Cultmalundie 040213)WooO N c d( t cannoI . stressee b t highlo dto y momente thatth t a , e th , very existence of such a tower is no more than hypothesis. The writer can find no trace on the e densgrounth d e dan woodlan d will make aerial photograph betweep yga uselesse th n t Bu . Westmuir and Peel does seem far too long not to have held intermediate sites and a considerable area of the wood can be seen from both Bertha and Midgate, as well as from Westmuir itself. Such a site would still not have been visible from Raith, however, and so any inter-fort communications between Bertha and Strageath would have needed to pass through a rather more inefficient three stage relay via Cultmalundie, Midgate and Raith. To the east of Cultmalundie Wood the ground falls away rapidly towards the fort at Bertha and to the Tay but, although contour maps would suggest that Peel and the two or three towers that might be expected between it and Cultmalundie, may just have been visible from Bertha, this could not be confirmed on the ground; these links have not been included e signallinonth g plan (illu . Thess2) e sites could, however, have been relaye Bertho dt a avi Huntingtower, which does have a clear view of the fort and, as Huntingtower could also have relayed signals to Bertha from any sites between itself and Peel, the system, as a whole, is (dependin existence th n go towea f eo Cultmalundin ri e Wood) potentially complete.

MIDGATE (also called Thorny Hill) Unfortunately, however plausibl e aboveth e outlin e problee, ther systeon be s y ei mmma which makes it impossible to accept without reservation. reture e dawIw th f f archaeologo no nt e ridgeth n yo , Christison's original programme included excavations at Midgate (NO 02112047), which was then the most easterly Gask site known (Christison 1901, 32-5) e standar. th Here f o addition i ,e d timbeon o nt r towerse h , also reveale drectangulaa r structure, whic secona e tooe hb h o kdt fortleKaime th f o ts Castle J la

. v V * 1 i i i 1 I 1 M y , V v Upcas/ t Mound .

>"iS»

MODERN ROAD

ILLU S6 Midgate, showin excavatiotowe e o positione gfortleth d tw th e ran d th an tf so n trenches 304 I SOCIETY OF ANTIQUARIES OF SCOTLAND, 1993

ILLUS 7 Midgate fortlet seen from the approximate full height of Midgate tower (west). The summit of the distance seee th extreme ridgb n th ni n n eca eo e left

type, lying immediately e Romanortth f o hn metre w roaonld fe dan a y e towe s th eas f ro t (illu . Yet6) s , although Christison's repor s remaineha t e foundatioth d l subsequenal r fo n t studies of the area, this site seems largely to have been forgotten and, when mentioned at all, s usualli t i y dismisse s non-Romana d r evea virtuao , s a n l figmen s excavator'it f o t s imagination. Indeed, an observer of the calibre of O G S Crawford (1949, 54-5) was able to visi sitdene d th tean y that anythin visible s existence gwa th s fortleA a . f eo t thia t s particular site could cas tr mode entire doubth ou f n eo lo t system fels twa important i , t that Christison's identification shoul re-examinede db .

THE SITE The importance of the site lies in its very close proximity to the Midgate tower, for its western ditch comes to within 13 m of the tower ditch (illus 6) and it seems most unlikely thaRomao tw t n installation closo s s e together coul e exactldb y contemporary. Moreover,

where a group of small Roman military sites, such as the Gas8 k Ridge fortlets, form a distinct morphological class they might reasonabl expectee yb contemporarye b o dt thao s , data t r efo any one installation is highly suggestive as a date for them all.9 This means that if Midgate is e samfortle, a th f eo ttyp s Kaimea s Castl Glenbankd ean s positioit , n would cast doubtn (i s the absenc f evidenco e e contrary e fortletth th o whetheo e t t f e likel o s ar sa ) y havo yt an r e belonge towee th o drt system. The site lies at the eastern end of a small but distinct knoll, with its western half in the WOOLLISCROFT CASE TH : K RIDGE SYSTEM 305

uncultivated south-east s eastercorneit fiela d f an o ndr hal densn i f e coniferous woodland. Despite Crawford's denial, it is still clearly visible as an earthwork (illus 7), although it has been heavily damaged since Christison' e impositioth y b a Secon y f o nda s d Worlr Wa d building interiors (11.1it m8 n i )5. 3t consistx I . rectangulaa f so r enclosure, almost identical Kaimo t internallW E/ s Castlem 3 y2 (Kaim,x measurinS , N/ s m) measure m 2 0 2 g 2 x 0 2 s rather more square tha s showni Christison'y nb s original plan (Christison . Lik19018) eg ,Fi Kaim s surroundei t i s e remain th eartn a y b df h o s rampart which a surfac s a , e feature, measurethickm 5 ,4- excepe wesc s th tn i twhers beeha nt i e thickene sincm 1 1 e o t d Christison's plan was drawn, perhaps in connection with the Second World War activity. On lower averagen ground(o t m fro ramparte 7 ou ,)m th s (agai t Kaims)a s na singla , e ditcn hca be traced as a shallow hollow, c 3.5 m wide, round all but the south side, where it has been destroye modere th y db n road orbie f thiTh . o t s ditc rathes hi r less regular tha s normani r fo l a Roman installation, but this is inevitable, especially in the north, due to the shape and steep sides of the hill. As Christison (1901, 34-5) pointed out, the ditch is surrounded on its northern and eastern sides by a low mound. He interpreted this as a branch of the Roman road, to give s plahi acces nsitee d th show an ,o t s t continuini s g south pas e site'th t s south-east corneo t r the modern road, which largely follows its Roman predecessor, but is here probably on a slightly more northern line. The feature, however, begins to turn, parallel with the ditch, at this corner befor e soutth ehn f e sit wit t o beinconsisti th of sidet h i t s f (illu a o egcu ; s6) s mainly of rubble from the partly rock cut ditches, it is clearly an upcast mound. There is also a causeway acros e westerth s n ditch littla , e nort f centreho immediateld an , y norta f ho narrow channel through the rampart, which gives the impression of being an entrance. Neither feature appears on Christison's plan, however, despite being conspicuously visible, an boto s d h seem likele modernb o t ye channel Th . particularn i , s probabli , y meanr fo t drainag agai y interioe th connectee ma nb f d eo an r d wit Secone hth d Worl r activitydWa d an , the causeway is probably its upcast mound. No other signs of an entrance survive, but this need not surprise us since the entrance at Kaims Castle had been carefully blocked and was also invisible before excavation (Christison 1901 ,site 20) Romans th ei f .I woule on , d expect entrancs it havo et esoute beeth n hno side, lik towere thath f o t, facin Romae gth n road.

THE EXCAVATIONS Christison's excavations were confined largel interioe th produced o yt an r d similar resulto st Kaims Castle, wit buildingo hn s being uncovered, onl remaine yth f pavingso . Unfortunately, this area has now been so badly damaged by a combination of Christison, Second World War activity and exceptionally heavy infestation by rabbits, that it was thought unlikely that further excavation would produce useful results. Work, therefore, was confined to re- surveying the site and sectioning the ditch, an operation that Christison does not appear to have carried out. Two trenches were dug (illus 6) which, for reasons of accessibility, were grouped aroun ditch'e th d s north-west corner ) sectionem 1 . Trenc x westere 5 dth ( 1 h n ditch towards its northern end where it was just beginning to turn into the corner, whilst Trench 2 (4 x 1 m) cut the western end of the north ditch, again just as it was turning into the corner. aparm t thei6 a t 7. trenchery o closestla tw d e an Th s° .wer 70 angln a c e t f a ethuo t sse Both trenches (illus 8) sectioned a 'V'-shaped ditch with a bottom slot or 'ankle- breaker'. In both cases the ditch was 2.5 m wide and a little over 1.2 m deep (1.27 m in Trench 1 and 1.21 m in Trench 2). In Trench 1 the ditch had been dug into firm, but easily I SOCIET 6 30 Y OF ANTIQUARIES OF SCOTLAND, 1993

AB« Trench 1. XY< Trenc. h2

ILLUS 8 Midgate fortlet, the excavated ditch sections. It should be noted that points B and X are both on the inne ditce r sidth hf e o (se e topsoiltur d 1 illu . an f s6) gre2 , y gritty loam yello3 , w sand gre4 , y silt, 5 plum/grey sand, 6 grey sandy loam WOOLL1SCROFT CASE TH : K RIDGE SYSTEM 307

worked yellow sand, but in Trench 2 it was cut through a hard, if brittle, purple/grey sandstone thao bottos s , it t m slo unlikels i t havo yt e bee accidentae nth l resul f cleaningo t n I . both trenches the ditch had been deliberately backfilled with a slightly brown, grey loamy material (L2), probably degraded turf. Thiidenticas swa materiao t l l currently being produced (in disturbing quantity) from the remains of the ramparts by rabbit activity, and was probably partiae resule th th f o tl slightin f thosgo e ramparts botn I . h sections this laye s essentiallwa r y homogenous t botbu , h showe a faid r degre f gradationo e , wit preponderanca h f fineo e r particles toward bottome sth , consistent wit hlona g perio f watedo r action since backfillingn I . Trench 1 the backfill filled the entire ditch, with no detectable sign of primary silt, even in the bottom slot. This, coupled wit e verhth y sharply defined e san t sideditchth dcu f o s , suggest that the ditch was backfilled almost immediately after being dug or re-cut. Trench 2 showed a deep silt deposit layerso , thic somm tw lowere m 5 n k i mucTh d .e 0. 4 thickeran e 0. , hth a s wa , deep layer (L5 f silt)o y purple/grey sand, similae weatherinth o t r g produc e naturath f o t l bedrock, whicaree th an hi immediatel y outsid ditce weathered eth hha produco dt thicea c k( 0.2-0.25 m) layer of soft, but clean, purple/grey sand. Above this, the northern two-thirds of the ditch were covered by a layer (L4) of clean grey silt c 0.1 m thick. The border between these layers was somewhat vague, but that between them and the backfill (L2) was very clear cut; taken together, the two trenches may suggest that the ditch was open and subject to natural silting for some time and was then backfilled whilst in the process of being cleaned outdatino N . g evidenc s found t ewa thi bu s, activity must have taken place long enougo hag backfile foth r havo t l e been grade watey db r action.

INTERPRETATION In the absence of datable material, one cannot honestly claim that the excavations prove a great deal. Perhaps the most that can be said is that nothing was found in either trench or in generae th l layousite th e f thao t othes i t r than consistent wit Romaha n fortlet. Unfortunately, however a 'V'-shape, d ditch, eve e wit 'ankle-breaker'n on ha t necessarilno s i , y diagnostic s Romaa northern i n n Britain a numbe; f siteo r f t Iroe dateleasa o s f r Iroe nAg o o t, Ag n culture, have also produced such ditches r examplfo , e Cnoc a' Caisteil near Alness (Rideout Hartburnd an 1987 ) 68 , , Northumberland, whic s itselwa h f once thougha Roma e b o nt t fortlet (Jobey 1973, 17.). Taken together with the surface survey and Christison's results, however, the very close similarity of Midgate to Glenbank and to Kaims Castle appears to make it virtually certain that these nearly identical sites form a distinct class which, although presumably contemporary, do not fit comfortably with the towers. Indeed the writer would wonder whether there would ever have been any doubt over this if Midgate had not been so close to its tower. Furthermore, the aerial evidence from Raith (illus 4) may, as stated, present a fourth membe f thio r s class, occupyin e samgth e knowa sit s a e n tower and, therefore, actually over- (or under-) lying it. It is important to remember that none of these sites has yet produced dating evidence and, despite its traditional identification as such, Kaims Castle is moro n e certainl Romaya n fortlet than Midgate firA . m absolut relativd ean e dating (vis-a- vis the towers) must, thus, await further excavations, especially at Raith; in the meantime we onle us y n whaca t evidenc have ew possibilities weige o et th p hu . The site of Midgate itself is not as helpful as might have been hoped, because the tower and fortle juse r enougar tfa t h apar remaio t t n completely separate. Their ditchet no o d s intersect and, so far as can be seen, neither of their ditch upcast mounds overlies the other. It is, therefore, impossible to establish a relative dating between the two sites and we are left 308 SOCIETY OF ANTIQUARIES OF SCOTLAND, 1993

wit procesa h f eliminatioo s n between prehistoric, Roma post-Romad an n ne dateth r fo s 'fortlets'. A non-Roman date has been advocated to the writer in a number of private conversations but, although the possibility must continue to be borne in mind, it does seem somewhat improbable. This is partly a subjective conclusion, because the general form of the 'fortlets', coupled to their (albeit inconclusive) ditch-profiles appears so Roman, but it is also based on their relationship with the Roman road. For, whilst these are the only such sites withie knowmetre roadareaw e e ar havo fe th l th nf al a n Tw ,e.n i o s their entrances oriented toward t (Christisoi s n 1901; Maxwel2 g fi , l 1990, 353-9), and, whils positioe th t t Raitna s hi uncertain, even Midgate, whose entranc unknowns ei , lies parallet woulI . it do t lthu s appear most unlikely that these sites could have been constructed before the road was built, or at least planned, which tend prehistoria rulo t st eou c date. A post-Roman date might be more conceivable, although again the sites' morphologies argue agains . Althougit t h suc hdata e canno rulee b t d out, severe doub cass i e t th t ovey b t i r apparent presence of a unit of six Roman miles (8877 m) in the sites' spacings. Glenbank and Kaims Castl almose ear t exactl Romax ysi n miles apart Midgatd an ,Roma 2 1 s ei n miles from Kaims Romax si , n miles fro mfrox f thifore si Raito t m Bertha. th a sd te unihan us y e b t 1Th 0 non-Roman builders seems unlikely, whilst the very presence of such a systematic approach would make a Roman date more probable, especially since the degree of gradation in the Midgate ditch fill rule t relativelsou y modern dates. If, therefore assume w , e thae 'fortletsth t Romane ar ' choice th , betwees ei Flaviae nth n and Antonine periods timee th , t whica s system'e hth s forts were occupied momente th r Fo . , t musi e stresseb t d thae evidencth t r eithefo e r indee(o r d any) perio s inconclusivedi ; althoug e traditionath h l Flavian datstily e considereb ma el d more likely s worti t i , h reviewin evidencee gth f onli , remino yt d ourselves thacase th tunproves ei suggeso t d nan t area whicn si h further data coul soughte db . A Flavian date would certainly be more convenient because, although it would complicat r preseneou t pictur systeme th f eo t wouli , d leav s fundamentaleit s unchangede W . would simply hav envisago t e procesa e f developmeno s whicn i t h fortlets were addea o dt syste f towermo fortswhicd n i an sr o ,h towers wer f fortfortletsd o e e an sadde e on Th o .dt former, in particular, would leave the signalling arrangements discussed above substantially unchanged and there is a certain amount of evidence that might support such a scenario. For example have w , e already see recurrence nth Romax unisi a f f eo o t n systeme mileth n d si an , s noteworthi t i y that Glenban Romax si s ki n miles from Doune onle th e ,syste yth forn mo t that, so far, has shown no sign of re-occupation in the Antonine period (Maxwell 1984). More persuasive evidence might be seen in the fact that there is no sign, as yet, that either Kaims or Glenbank replaced towers thad an ,t Glenbank, like tower s vicinityth eit a doubln i ss ha , e ditch. It is always dangerous to argue from an absence of evidence and the Glenbank ditch could f courseo , coincidencee b , t thibu ,s does see linmo southert e kth n fortlets more closely to the towers and may suggest that a change was made to the system's overall plan during the course of its construction. If so, it would appear that the Gask Ridge sector itself was built first, for whilst at Midgate and Raith the fortlets may have replaced towers (or possibly vice- versa) that had already been completed, Kaims and Glenbank seem to have been built together with their surrounding towers, whic y allma h, therefore e slightlb , y later. Furthermore, G S Maxwell (1986, 354) has recently sought to establish a rational spacing towersystee e southerth th r n mi sfo n area , norte baseth hn d o gat f Ardoch eo f whic o d han , Glenban Kaimd kan s see foro mt integran ma l part l thiAl .s should have e th helpe e ti o dt WOOLLISCROFT CASE TH : K RIDGE SYSTE9 30 I M

system togethe l butal r ,r unfortunately fo onc d ean s stili t i l, unproven whether these double ditched southern towers are themselves Flavian. Only the single ditched towers have so far produced dating material and thess ,a e sho suco wn h regularit thein yi r spacing e (Tablth , e1) two types need not necessarily go together. An Antonin e 'fortletse th dat r efo ' would force rather mor changa r model f ou eo n ei . Current thinking envisage a Flavias n syste f fortsmo , fortlet towersd Antoninn a an s d an , e syste f fortmo s only, wit o signallinn h g links between them, becaus ee fort nonth s i sf o e intervisible. Antonine fortlets would change this to an arguably more balanced picture: a Flavian syste f fort mtowerAntonino d n a an sd san e syste f fort mfortletso d an s f o . Suce us ha small fortlets as road posts is hardly unknown in Antonine Scotland and three, admittedly tenuous, strand f evidencso e coul marshallee db favou n di Antoninn a f o r e date. First, there is a general tendency for small fortlets of this type to be more of a second- tha a first-centurn y phenomenon r exampleFo . , Maxwell (1990) compare e standarth s d pictur integraten a f eo d Flavian Gask Ridge system with fortlets Flaviae th o t , n systee th n mo Wetterau Lime Germanyn si noteworths i t i t Ye . y that, where they have been investigatede th , Kleinkastell e Germath f o e n Limes have often prove e lateb o rdt tha e originath n l Flavian design (Baatz & Herrmann 1989, 374-423) and usually date to the second century. Similar tiny fortlets proliferated in the mid second century, as milecastles or Kleinkastelle on the frontiers s roaa , d post thein i s r hinterlands components a r o , f coastao s l defences (Bellhouse 1989; Newall 1976). First- and early second-century fortlets, on the other hand, are rarer and usuall e seeyb t Haltwhistllargera n ca s a , e Burn (Gibso Simpso& n n 1909), Cargill (Richmond 1943, 47-8 Castleshad )an w (Walker 1989). Milecastle-sized fortlet occuo d s n i r the first century, for example at Old Burrow and Martinhoe in Devon, (Fox 1977, 15-20) but they are much less common. A second strand of evidence stems from possible parallels between the interiors of Kaims Castle and Midgate, and some of the milecastles of the . Christison (1901, 21, 33-4) expressed surprise that he could find no sign of internal buildings in these 'fortlets', and this continues to cause puzzlement. Christison had pioneered the excavation of timber feature Scotlann i s d (Christiso Cunningham& n , 1898 s thu, 443-6wa s d quitan ) e capabl f findineo g such buildings botn i t h Ye 'fortlets. foune h ' d onl paveya d interior t doeI . s not, however, see havmo t e looo occurret km undehi o dt r this pavin earlier n signa gfo f o s r period which may have involved buildings. This is a pity, because in recent years a number of Antonine Wall milecastles have been foun dtheid whicha r d interiorhha s cleare f buildingdo s and paved ove t soma r e time after their completion e milecastleTh . s r sucpickefo ht dou treatment (Wilderness Plantation, Kinnei possibld an l y Seabegs Wood: Wilkes 1974; ,57 Keppi Walke& e r 1981 n importani ,l 145-6al e ar t) potential signalling positions (Woolliscroft 1993, 208-307) and the Gask Ridge 'fortlets' may have had a similar history. Kaims Castle particularn i , , could have been vita r signallinfo l t wouli d interestine gdan b o gt whae se re-excavatioa t site th e f mighno t reveal. The third piece of evidence is the most tenuous of all. It has long been fashionable to use spacing criteria and the relationships between known Roman installations to help in searching for new ones. The value of such numbers games may sometimes be debatable, but we have already seen the recurrence of a unit of six Roman miles in the spacing of the Gask Ridge 'fortlets thein i d r an relationshi' forte th f Berth o so pt Doune d becauss aan wa t I f . eo this pattern that the writer sought out the aerial photograph of Raith (illus 4) discussed above, pattere th an f i ds projecte i n d farther sout t wouli h d also place installation e longth n -o s suspected sites of Stirling, and Torwood (c NS 834846). The principal interest, however, 310 SOCIETY OF ANTIQUARIES OF SCOTLAND, 1993

would lie in the fact that a further six-mile spacing to the south of Torwood would project, as th ee systeme Flavia basth th f t eo no n, for f Camelono t s migha , e expectedb t e th t bu , Antonine Wall fort of . e evidencTh e from signallin s ambiguousgi e righar thinkinn , i e teve w f ni g that there actuall s suc ywa hsystea placemn i . Certainly t migh i ,e Antonin th e though b f ti d eod t forts were left without relay site provido st e inter-fort communications Antonine th f i ; e systed mha used isolated relay stations, rather than a closely spaced line of watch posts, fortlets would have bee na safe r siz f installatioo e o use t nt ther problea . Bu s i e m because, althouge hth 'fortlets' come tantalizingly clos o beint e g abl o providt e e such links t quit, no the eo d y manage it and this must throw doubts on their relevance to an Antonine context. probleme Th , once againfailure th s occupi o ,t e eastere yth n summi Gase th f k o t Ridge, to the north of Westmuir. This appeared peculiar enough on a Flavian system, where it merely produce unnecessarn da y degre f inefficiencyeo Antoninn a n o t ebu ,fortle d foran t t systemt i would have been fatal. For, although Kaims can link Ardoch to Strageath, Raith (if it exists) n linca k Strageat o Midgatet h d Glenbanan , n linca kk Ardoc o pointt h s farther south, Midgate's view to Bertha is blocked by the ridge summit, so that one of the system's forts would have been left cut off. The puzzle is thus: if the 'fortlets' are Antonine, why was Midgate not built on the higher ground to its north-east so as to link Raith and Bertha and allow the whole system to work ? There are two possible solutions, but neither appears very satisfactory. For example, it could be that there actually was some sort of minor installation outstationed on the summit that we have simply not yet found. This seems unlikely, if only because it would have been suc ha needlessl y complex arrangement e summiTh . s onlhighei m t y7 r tha e 'fortlet'th n . Midgat s alreadi e y fully expose e prevailinth o dt g westerly windt wouli o s , d have gained little in the way of shelter by being positioned off the highest point and, as both Raith and Kaims are sited on hilltops, there seems to be little reason why Midgate should not also have been. e seconTh d possibilit ye weaknes th reste evidencn th o n sAntonin a f o sr fo e e occupatio f Bertho n a (Adamso Gallaghe& n r 1986). Excavation e site th faro s t , a s, have produced onl singlya estratifiee phasth f f defenceo e o l dal potterd an s Flavians yi e onlTh .y evidence for Antonine activity is a single sherd of probable Antonine pottery, picked up as a surfac ededicatioa findd an , Disciplinaeo nt Augusti rivee founth rn i d(Keppi e 1983, 402) which, although undated, is of a type which does not occur before Hadrian and is generally Severan (RIB 990, 1128, 1723, 1978, 2092). Obviously, if the fort did not have an Antonine phase, there would be no need for any signalling arrangements to include it. But even this is not quite enough because, if this were the case, Midgate would become the terminal installation of the system. This might be righn surprisinow t s (althougit n gi t impossiblehno e sit s th onlei s a fortleta y) e woulOn . d certainly expect that if there was the opportunity to place the terminal site of such a line on a magnificent position wit a superh b view toward e exacth s t direction from whic majoy an h r trouble might be expected to come, then that opportunity would be taken up. If this site could also have provided observation cover, fro msafa e distance greatea f Berthf o o , d r aan are a around it than the fort itself could command, one would have expected such a site to be doubly tempting. In other words, even if Bertha had not existed, one would expect that, if Midgat a systes par f wa eo t m consisting onl f fortfortletsd o y an s t wouli , d still have been ridge builth en o tsummit. 11 There is also another difficulty. Antonine fortlets at Midgate and Raith would leave our WOOLLISCROFT: THE CASK RIDGE SYSTEM | 311

modeFlaviae th f o l n system unaffectedpresence th o t f towerse eo du , . But f Kaim,i s Castle was Antonine there could be no general signalling system on the Flavian line since this site, which is the only possible link between the forts of Ardoch and Strageath, has no corresponding tower. This, however, is to begin to enter a circular argument since, despite mapping a possible Flavian signalling system we still cannot be sure that one ever existed, or was even aspired to. Under these circumstances, all that can be said is that, whilst the Cask Ridge's minor installation seeo sd mmako t e most traditionalle sensth n ei y envisaged, single- period, probably Flavian, context, there are peculiarities even here; the exact position will be known only whe gooa n d deal more excavatio aeriad nan l reconnaissanc s beeha e n carried out.

ACKNOWLEDGEMENTS writee Th r wishe thano st k Messrs lan Brow Sonsn& ownere th , f Midgateso r permissiofo , n to carry out excavations and Dr S A M Swain for her help in the field.

NOTES 1 Greenloaning, Westmuir, Peel and Huntingtower are known only from aerial photography, although there seems little doubt as to their identity. 2 G S Maxwell (1990, 354) mentions amphora fragments from Glenbank but has yet to publish their date and context. resulte excavatioe th Th f o s3 n were surprising becaus site eth e standexposen a n o s d hilltop witha superb view over Stirling to the Highland fringe, a marvellous signalling position but an unpleasant plac liveo et . 4 It is still not certain that Doune is part of the Gask System. Its discoverer (Maxwell 1984) has suggested that it might have done away with the need for a fort at Stirling, but the two are six Roman miles apart e samth ,s Ardocea Strageathd han . They would also hav complementard eha y roles because Stirling, the traditional gateway to the Highlands, watches the primary route over the Forth, whilst Doune guards an ancient secondary pathway through the extensive Forth mosses. 5 The National Monuments Record 6-inch map records a ring-ditch at this reference discovered from the air in 1984 which may be the missing tower, but the writer has been unable to obtain further information. 6 The site is unexcavated, but its prominent central mound and faint outer ditch do not look like any other Flavian tower. It stands in a cairnfield and until the 1770s had five standing stones on it. In 1783 large amounts of stone were removed to build a bridge over the Almond (Crawford 1949, 48-9). Yet all other Flavian towers in Scotland are built solely of timber. As for its position, Fendoc e newlth t s visiblhyi no discoveret ebu d putative fortlet (informatio JonesnB froD m)G which stands only a few hundred metres away. Its famous view up the Sma' Glen is actually extremely limited, even from the height of a Roman tower, perhaps no more than c 400 m. The site is also some distance from the fort; a vastly superior site can be found on the opposite (northern) glee sidth n f eo mout h whenc toweea r would dramaticalla hav d eha y improve dglene vieth p ,wu and a view to the fortlet, whilst being much closer to the fort. s bee e writer'ha nth t I 7 s usual citpolico intervisibility t et an no y t beeyno nthas personallha t y checked in the field but, as much of the ridge-top is heavily wooded, the intervisibilities between ease sitee f Kirkhilth o tth Rait o t sd han l have been checked usin contoure gth 1:1000n so 0 maps. The maps would suggest, however, that the views would have been very clear with no borderline cases. 8 Christison (1901 ) attempt 32 t ,aroun ge issuo e t s dth describiny eb towee outworn a gth e s a rth f ko fortlet. 312 | SOCIETY OF ANTIQUARIES OF SCOTLAND, 1993

9 The entire tower system has, after all, been dated on precisely this basis. e spacingTh exace 0 1 ar s withio t t n 1-3%. Within this margi sitee nth s appea havo t r e been siten di accordance with local tactical considerations s noteworthi t I . y thae fort th f tCamelon o s , Ardoch and t intStrageatfi ot thino so hd sequence . Ardoc Strageatd han e als Romax har o si n miles apart, but Kaims lies much closer to Ardoch than to Strageath, and Glenbank is much closer to Ardoch than Doune e measurementTh . e scalear s d from map thud an ss subjec o inaccurace t t th o t e ydu unevennes difference groundth e s th a f t so ebu , betwee statutRomax x si nsi d m5 enan , mile72 c s si almost hal Romaa f n mile s unlikeli t i , unito y tw thase coulth t mistakee d b anothere on r nfo . t mighI 11 t als interestine ob Secone th y speculato gt dwh Worlo t r installatio s ea dWa t buil no s t nwa on the ridge summit, since the writer is informed that it was an anti-aircraft position.

REFERENCES Adamson Gallagher& C H ,B 198D , 6 'The Roman for t Berthaa t e 197th , 3 excavation', Procc So AntiqScot, 116, (1986), 195-204. Baatz, D & Herrmann, F R 1989 Die Romer in Hessen (2nd edn). Stuttgart. Bellhouse, R L 1989 Roman Sites on the Cumberland Coast, a New Schedule of Coastal Sites. Kendal. Breeze J 198D , Northerne 2Th Frontiers of . London. Breeze J 198D , 3 'The Roman Fort t Ardoch'sa O'Connorn i , Clarke& FromA ,V Stonee D , th e Ag to the 'Forty-Five, Edinburgh, 224-36. Breeze J 199D , 2 'The Great Myt Caledon'f ho , Scott Forestry, (1992)6 4 , 331-5. Christison D ,190 1 'Excavations Undertake Societe th y nb f Antiquarie yo f Scotlanso f Earthworkdo s Adjoinin "Romae th g n Road" Between Ardoc Dupplid han n ', Proc Antiqc So Scot,5 3 (1900-1), 16-43. Christison, D & Cunningham, J H 1898 'Account of the Excavation of the Roman Station at Ardoch, Perthshire Undertaken by the Society of Antiquaries of Scotland in 1896-97', Proc Soc Antiq Scot, 32(1897-8), 399-476. Crawford, O G S 1949 Topography of Roman Scotland North of the Antonine Wall. Cambridge. Donaldson H 198G , 8 'Signalling Communication e Romath d nan s Imperial Army', Britannia,9 1 (1988), 349-56. Fox, A 1967 'Martinhoe and Old Burrow', in Studien zu den Militargrenzen Roms (Proc 6th Int Limes Congress), Koln, 15-20. WilkesFrere& S J S ,J 198, 9 Strageath, Excavations Within Romane th Fort 1973-86 (Britannia Monograph Series No.9). London. Gibson Simpson& P J , G 190 F , 9 'The Roma nStanegate th For n o t t Haltwhistlea e Burn', Archaeol Aeliana, 3 ser, 5 (1909), 213-85. Hanson S 198W , 7 Agricola Conqueste th d an North.e oth f London. Hartley, B R 1970 'The Evidence of Samian Ware for the Flavian-Trajanic Frontiers', Roman Northern Frontier Seminar, 2 (1970), 1-5. Helmke, P 1910 Romische specula iber einer Germanischen Anlage auf dem Johannisberg bei Bad Nauheim. Nauheimd Ba . Hobley S 198A , 9 'The Numismatic Evidenc e Post-Agricolath r efo n Abandonmen e Romath f o tn Frontie Northern ri n Scotland', Britannia, (1989)0 2 , 69-74. Jobey, G 1973 'A Native Settlement at Hartburn and the Devil's Causeway, Northumberland (1971)', Archaeol Aeliana, ser5 (1973),1 , 11-53. Keppie, L J F 1983 'Roman inscriptions from Scotland: some additions and corrections to RIB I', Proc Soc Antiq Scot, 113, (1983), 391-404. Walker& Keppie F J J J 198,L , 1 'Fortlet e Antoninth n o s e Wal t Seabega l s Wood, Kinneid an l Cleddans', Britannia, 12 (1981), 143-62. Kofler F , 1898 'StraBenturm im Wolfersheimer Walde', Westdeutsche Zeitschrift, , Limesblatt17 , (1898), 767-71. WOOLLISCROFT CASE :TH K RIDGE SYSTEM 313

Leiner, W 1982 Die Signaltechnik der Antike. Stuttgart. Maxwell, G S 1984 'New Frontiers: The Roman Fort at Doune and its possible significance', Britannia, 15 (1984), 217-23. Maxwell S 199G , 0 'Flavian Frontier Caledonian si Vettersn i ' Kandler& H , (edsM , . ) Akten14 s de Internationalen Limeskongresses 1986 in Carnuntum, Vienna, 353-65. Newall F , 1976 'The Roman Signal Fortle t Outerwardsa t , Ayrshire', Glasgow Archaeol (1976)4 , J , 111-23. Ogilvie, R M & Richmond, I A 1967 Cornelii Taciti de Vita Agricolae. Oxford. ORL = Fabricius, E Hettner, F & von Sarwey, O Der Obergermanisch-Raetische Limes Des Roemerreiches. Pitts, L F & St Joseph, J K 1985 : the Roman Legionary Fortress (Britannia Monograph Series, 6). London. RIB = Colling wood, R G & Wright, R P 1965 The Roman Inscriptions of Britain. Oxford. Richmond A 194I , 3 'Recent Discoverie Roman i s Field'e th nn i BritaiRomanJ , d an nr FroAi e mth Stud, (1943)3 3 , 45-54. Rideout, J S 1987 'Excavation of an Earthwork at Cnoc a' Caisteil, Alness, Easter Ross', Glasgow Archaeol (1987)4 1 , J , 63-9. Rivet, A L F 1964 'Kaims Castle' and 'The Gask Ridge Signal Stations', Archaeol J, 121 (1964), 196-8. Robertson, A S 1974 'Roman "Signal Stations" on the Gask Ridge', Trans Perthshire Soc Natur Sci, (Special Issue), (1974), 14-29. St Joseph, J K 1976 'Air Reconnaissance of Roman Scotland 1939-75', Glasgow Archaeol J, 4 (1976), 1-28. Steer A 195 K , 6 'An Early HomesteaIroe nAg t Wesda t Plean, Stirlingshire', Proc Antiqc So Scot,9 8 (1955-6), 227-51. Topping, P 1987 'A New Signal Station in Cumbria', Britannia, 18, (1987), 298-300. Walker, J (ed) 1989 Castleshaw, the Archaeology of a Roman Fortlet. Manchester. Wilkes J J 197, 4 'The Antonine Wall Fortle t Wildernesa t s Plantation, Lanarkshire', Glasgow Archaeol , 3(1974)J , 51-65. Woolliscroft J 198D , 8 'The Outpost Syste f Hadrian'mo s Wall', Trans Cumberland Westmorland Antiq Archaeol Soc, 2 ser, 88 (1988), 23-8. Woolliscroft J 1989D , a 'Signallin desige th d f Hadrian'no gan s Wall', Archaeol Aeliana, ser5 7 ,1 (1989), 5-19. Woolliscroft, D J 1989b 'Elevated Archaeological Photography', British Archaeol, 13 (June 1989), 18-21. Woolliscroft J 199D , 3 Signalling e Designth d f Romano an Frontier Systems. UnpublisheD dPh thesis, University of Manchester. Woolliscroft Hoffmann& J D , B , 1991 'Zum Signalsyste Aufbad mun s Wetterau-Limes'de u , Fundberichte Baden-Wurttemberg,s au (1991)6 1 , 531-43.